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
BMC Evol BiolBMC Evolutionary Biology1471-2148BioMed Central London 1471-2148-5-451610916110.1186/1471-2148-5-45Research ArticleAn ancient spliceosomal intron in the ribosomal protein L7a gene (Rpl7a) of Giardia lamblia Russell Anthony G [email protected] Timothy E [email protected] Russell F [email protected] Michael W [email protected] Program in Evolutionary Biology, Canadian Institute for Advanced Research, Department of Biochemistry and Molecular Biology, Dalhousie University, Halifax, Nova Scotia B3H 1X5, Canada2005 18 8 2005 5 45 45 22 4 2005 18 8 2005 Copyright © 2005 Russell et al; licensee BioMed Central Ltd.2005Russell 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
Only one spliceosomal-type intron has previously been identified in the unicellular eukaryotic parasite, Giardia lamblia (a diplomonad). This intron is only 35 nucleotides in length and is unusual in possessing a non-canonical 5' intron boundary sequence, CT, instead of GT.
Results
We have identified a second spliceosomal-type intron in G. lamblia, in the ribosomal protein L7a gene (Rpl7a), that possesses a canonical GT 5' intron boundary sequence. A comparison of the two known Giardia intron sequences revealed extensive nucleotide identity at both the 5' and 3' intron boundaries, similar to the conserved sequence motifs recently identified at the boundaries of spliceosomal-type introns in Trichomonas vaginalis (a parabasalid). Based on these observations, we searched the partial G. lamblia genome sequence for these conserved features and identified a third spliceosomal intron, in an unassigned open reading frame. Our comprehensive analysis of the Rpl7a intron in other eukaryotic taxa demonstrates that it is evolutionarily conserved and is an ancient eukaryotic intron.
Conclusion
An analysis of the phylogenetic distribution and properties of the Rpl7a intron suggests its utility as a phylogenetic marker to evaluate particular eukaryotic groupings. Additionally, analysis of the G. lamblia introns has provided further insight into some of the conserved and unique features possessed by the recently identified spliceosomal introns in related organisms such as T. vaginalis and Carpediemonas membranifera.
==== Body
Background
Spliceosomal introns have now been identified in all major eukaryotic lineages. Recently added to this list are several protists that are widely considered to represent deep divergences within Eucarya: the diplomonad Giardia lamblia [1], its close relative Carpediemonas membranifera [2], and the parabasalid Trichomonas vaginalis [3]. The distribution and conservation of proteins involved in the removal of spliceosomal introns [1,4,5] suggests that this intron type is a feature that was present in the ancestor of this domain of life. While the precise determination of intron frequency in the T. vaginalis and C. membranifera genomes awaits further analysis, these preliminary studies suggest that these two organisms likely contain many more spliceosomal introns than are evident in the compact genome of G. lamblia. To date, only one spliceosomal-type intron has been reported in G. lamblia despite the analysis of approximately 5800 predicted open reading frames (ORFs) [1].
One of the most interesting properties of the T. vaginalis introns is the presence of a functionally important motif and putative branch-point sequence, ACTAAC, invariantly located within a 'conserved' 12-nucleotide (nt) intron segment located directly adjacent to the 3' splice site. Surprisingly, the only known G. lamblia intron also contains the identical 12-nt motif at its 3' end. Additionally, the introns from these two organisms display significant sequence similarity at their 5' ends. This suggests that similar splicing mechanisms may be employed to remove introns in organisms from these two eukaryotic groups, members of the proposed eukaryotic supergroup Excavata [6]. These sequence motifs have been shown to be important for efficient in vivo splicing of introns in T. vaginalis [3]; however, since only one intron has been identified in G. lamblia so far, the extent of conservation of these sequence motifs in other G. lamblia introns is not known. It is conceivable that these intron features are indicative of ancestral eukaryotic spliceosomal introns. Conversely, these features may be derived and thus be unique among spliceosomal introns found in these long-branching eukaryotic taxa.
It has been proposed that most of the identified T. vaginalis introns are ancient because introns are also found at similar positions in homologous genes in some other eukaryotic taxa [3]. Since the taxonomic sampling addressing the prevalence of any of these introns is sparse and because many of these introns are not located in the same phase or even the same relative amino acid position, the actual conservation and ancient nature of these introns requires further verification. The possibility that a subset of spliceosomal introns could be ancestral to the eukaryotic radiation is an exciting one and, if true, suggests that these introns might be useful genomic markers to aid in the elucidation of deep phylogenetic relationships within the domain Eucarya.
Recently, hypotheses describing eukaryotic evolution have coalesced around a limited number of eukaryotic supergroups. One scheme [7] proposes six primary eukaryotic clades: Opisthokonta, Amoebozoa, Plantae, Chromalveolata [8], Rhizaria [9] and Excavata [6,10]. The strength of the evidence supporting each of these supergroups varies, as does the degree of organismal sampling within the proposed assemblages. The relationships among the supergroups themselves are currently undefined: i.e., their relative branching order in the eukaryotic tree is unknown. Even with concatenated data sets comprising hundreds of individual sequences, the approaches of molecular phylogenetics are increasingly challenged to provide robust and compelling answers to this question.
In this report, we identify an intron in Rpl7a, the gene encoding the G. lamblia ribosomal protein L7a and an additional spliceosomal intron in an unassigned ORF that encodes a non-conserved protein. We observe striking similarities among the G. lamblia, T. vaginalis and C. membranifera introns. At the same time, departures from the sequence constraints within these motifs may discriminate some of the splicing mechanisms employed within these eukaryotic groups. Our extensive examination of the phylogenetic distribution and properties of the Rpl7a intron indicate that it is an ancient spliceosomal intron, and our study also provides preliminary evidence uniting three eukaryotic supergroups (Opisthokonta, Amoebozoa, Excavata) to the exclusion of at least two others (Chromalveolata, Plantae). Our investigation of the distribution of the Rpl7a intron is the most extensive examination to date of a conserved intron position in eukaryotes. Taken together, these results argue that further examination of the patterns of intron conservation and distribution within the eukaryotic domain, like other shared derived characters such as gene fusions, insertions and gene replacements, can provide a valuable adjunct for evaluating proposed phylogenetic groupings and frameworks derived from sequence comparisons.
Results and discussion
A conventional spliceosomal intron in the Giardia lamblia Rpl7a gene
In comparing ribosomal protein L7a homologs from various organisms, we noted that the predicted sequence for the G. lamblia protein [GenBank:EAA41652] appeared abnormally truncated at the C-terminal end relative to other eukaryotic and archaeal sequences, terminating at position 171 of S. cerevisiae L7a (Fig. 1). In addition, this truncated sequence displays unusual divergence after yeast position 161, which is unexpected because this C-terminal region corresponds to a highly conserved portion of L7a. Further examination of the G. lamblia gene sequence revealed a 109-nt intron whose removal results in a predicted L7a protein sequence similar in length to that of other eukaryotic L7a homologs, and that aligns readily with them downstream of position 161 (Fig. 1). The assigned intron boundary sequences, GT...AG (Fig. 2), are those of conventional spliceosomal introns.
Figure 1 Clustal X alignment of ribosomal protein L7a amino acid sequences. The alignment comprises the portion corresponding to positions 135 to 200 in the Saccharomyces cerevisiae protein (full organism names are listed in Table 1). At highly conserved positions at which a single amino acid predominates, residues are indicated as white letters on a black background. The arrowhead denotes the location of a conserved spliceosomal intron. The predicted Giardia lamblia protein sequence either with (above arrowhead) or without (below arrowhead) removal of the intron is shown. Sources of the sequences are: S. cerevisiae [GenBank:AAB65045], Homo sapiens [GenBank:NP_000963], D. melanogaster [GenBank:AAN09172], A. castellanii [GenBank:AY925000], J. libera [GenBank:AY924997], T. pyriformis [GenBank:DQ118092], E. gracilis [GenBank:], H. marismortui [GenBank:YP_134885] and G. lamblia [GenBank:AACB01000019].
Figure 2 Conservation of intron boundary sequences in Giardia lamblia, Trichomonas vaginalis and Carpediemonas membranifera. Positions of nucleotide identity within the intron sequences in G. lamblia and C. membranifera are indicated with asterisks. Potential intron branch-point sequences are underlined. The distance between the intron boundary sequences is indicated for each G. lamblia intron as is the distance variation between these sequences seen in the T. vaginalis introns used to derive the consensus sequence shown. The ferredoxin intron sequence is from [1], the T. vaginalis introns are from [3], and the carbamate kinase intron sequences are from [2].
The Rpl7a intron is only the second reported spliceosomal-type intron in G. lamblia, a diplomonad widely considered to be a deep-branching eukaryote. The other example (35 nt in length) resides in a putative [2Fe-2S] ferredoxin gene and begins with CT rather than GT at the 5' intron boundary [1]. Comparison of the sequences of the two G. lamblia introns reveals striking sequence conservation at the intron termini (Fig. 2). Notably, this similarity includes the region predicted to contain the intron branch-point sequence [1] (underlined in Fig. 2). Likewise, both the 5' and 3' intron boundary sequences of the two G. lamblia introns display extensive sequence similarity to the consensus sequence of the known T. vaginalis introns (Fig. 2). However, there are some significant nucleotide differences between the G. lamblia Rpl7a intron and the T. vaginalis consensus motif that suggest somewhat different sequence requirements for splicing in these two organisms.
As mentioned previously, both CT and GT sequences have now been observed at the 5' intron boundary of G. lamblia introns. This observation is made even more significant by the demonstration of efficient splicing of the G. lamblia ferredoxin intron in a T. vaginalis in vivo splicing system, but only after changing the CT to a canonical GT at the intron 5' boundary [3]. Our identification of a second G. lamblia intron that contains a canonical GT indicates that the CT sequence is not an absolute requirement for splicing of these introns in G. lamblia and also suggests that the constraints on 5' intron boundary sequences are different in these two organisms.
The second important observation is derived from a comparison of the sequences of the putative branch-point motifs contained within the two G. lamblia introns. The G. lamblia ferredoxin intron has the putative branch-point sequence ACTAAC, which is identical to that observed in all the identified T. vaginalis introns. The G. lamblia Rpl7a intron instead contains the sequence ACTGAC. By creating site-directed mutants of a T. vaginalis splicing reporter construct using an intron from a poly(A) polymerase gene, Vaňáčová et al. [3] demonstrated the importance of this sequence element and its position relative to the intron 3' end for efficient in vivo splicing in T. vaginalis. However, single-nucleotide substitutions at the position corresponding to the G nucleotide in the branch-point sequence of the G. lamblia Rpl7a intron were not examined for their effects on splicing. Therefore, it is possible that ACTGAC could be a functional branch-point sequence for the splicing of T. vaginalis introns.
The majority of the introns identified in the T. vaginalis genome were found using a sequence-pattern search strategy based primarily on nucleotide conservation observed between the intron boundary sequences of the first identified T. vaginalis intron and the G. lamblia ferredoxin intron. The identified introns, 39 in total, vary in length between 59 and 196 nt. They contain a consensus 5' splice site sequence GTAYGT and, based on constraints imposed in the search parameters, the 3' splice-site sequence ACTAACACAYAG, where the nucleotides in bold italics indicate the potential branch-point sequence.
To investigate whether an additional group of T. vaginalis introns might instead contain variations at the fourth nucleotide position (bold italics) of the branch-point sequence, ACTAAC, the partial T. vaginalis genome sequence was searched for potential introns containing these properties. We also incorporated into our search parameters an allowance for other sequence differences within the 3' intron boundary sequence, such as those observed when comparing the two G. lamblia introns. Using the pattern-search algorithm PatScan [11], we searched the entire preliminary genomic data (3X coverage) for all sequences containing the sequence pattern [5'– GTAYGT...5–500 nt...ACTBACNCAYAG-3'], where B = C, G or T; Y = C or T; N = any nucleotide and the branch-point sequence is in bold italics. Remarkably, only two matches to this pattern were found in the entire T. vaginalis genome data set and neither of these sequence patterns appears to be an intron candidate. Furthermore, neither of these two sequences contains ACTGAC in the predicted position for the intron branch-point sequence. This result further emphasizes the importance of the ACTAAC sequence motif in the T. vaginalis introns and the differences in sequence constraints apparent between these introns and those of G. lamblia. If additional T. vaginalis spliceosomal introns are found (<519 nt in size) that contain alternative sequences at the 3' intron boundary, sequence differences must also be present in the 5' intron boundary that prevent the search algorithm from finding these intron candidates using the parameters we have employed.
Using search parameters similar to those above (but allowing G or C as the starting nucleotide at the intron 5' end), we searched the G. lamblia partial genomic data set for intron candidates. As expected, the Rpl7a and ferredoxin introns are detected by this method, as is an additional set of 15 matches to this sequence pattern. Of these matches, only one was a likely intron candidate that had the potential to disrupt a predicted ORF, among other criteria. Since the ORF in which this candidate intron resides (or that it disrupts) does not encode a conserved protein – unlike the Rpl7a intron – further experimentation was required to prove whether or not it is an intron. Removal of the intron candidate sequence, nt positions 1263 to 1482 of the G. lamblia contig [GenBank:AACB01000025], extends the predicted protein encoded by the unassigned ORF [GenBank:EAA41257] by an additional 118 amino acids at its amino terminus. Further, when the intron sequence is removed, the inferred initiation codon is now positioned directly adjacent to an AT-rich motif that Yee et al. [12] have previously identified as a promoter element for other G. lamblia genes. RT-PCR analysis and sequencing of cDNA clones obtained from the corresponding mRNA (see Additional File 1) confirm the existence of this third G. lamblia spliceosomal intron, which is 220 nt long and exhibits intron boundary sequences similar to those of the ferrodoxin and Rpl7a introns (Fig. 2).
Extensive genomic DNA information is not yet available for C. membranifera, and to date only two introns, both small, have been identified in this protist [2]. In these introns, potential branch-point sequences (underlined in Fig. 2) are found abutting the 3' intron boundary, as is observed in the G. lamblia and T. vaginalis introns. While the G. lamblia introns lack an obvious polypyrimidine tract in the vicinity of the 3' end of the intron, C. membranifera introns and the majority of the T. vaginalis introns do exhibit pyrimidine-rich sequences immediately upstream of the 3' intron/exon boundary. However, in the C. membranifera introns these pyrimidine-rich sequences include the potential branch-point sequences whereas in T. vaginalis these sequences (usually T-rich) are found upstream of the potential branch-point sequences. It is also apparent that the C. membranifera introns exhibit less sequence conservation at their 3' intron boundaries than do the G. lamblia and T. vaginalis introns. These important differences in the properties apparent in C. membranifera introns are particularly relevant to this comparison of intron structure because this organism appears to have a closer evolutionary relationship to G. lamblia than does T. vaginalis [13]. The variety of differences observed in intron structure from each of these three organisms seems to further differentiate the sequence requirements for the splicing mechanisms employed in these eukaryotic groups. The identification of homologs of snRNAs in these organisms may give insight and permit comparisons of novel splicing mechanisms employed in the removal of these introns, characterized by the above-mentioned conserved intronic elements and potential branch-point sequences unusually close to the intron 3' end.
Phylogenetic distribution and properties of the Rpl7a intron indicate that it is ancient
Given the phylogenetic conservation of the L7a protein in the eukaryotic domain, we searched available eukaryotic genomic databases to determine whether an intron at the same position as in G. lamblia is conserved in the Rpl7a gene of other organisms. Somewhat surprisingly, we identified a conventional spliceosomal intron at exactly the same position and in the same phase within Rpl7a in many animal (opisthokont) taxa and in the amoebozoon Dictyostelium discoideum (Fig. 3). This result, in concert with the proposed deep branching position of Giardia, suggested that the Rpl7a intron might be an ancient eukaryotic spliceosomal intron. Accordingly, we undertook a more detailed investigation of the occurrence of this particular intron within the eukaryotic domain. Using EST sequences available through the Protist EST Program (PEP), PCR primers were designed to amplify genomic Rpl7a sequences from a wider array of eukaryotes, particularly those placed in the controversial taxon Excavata [14]. We identified the Rpl7a intron in three additional Amoebozoa representatives spanning amoebozoan diversity [15]. We also found the intron in three additional groups from Excavata: jakobids, malawimonads and Trimastix pyriformis (Fig. 3). The intron is variable in size in these organisms but was only ever found in frame 0 of the coding region. The jakobid Rpl7a introns (541–988 nt) are the largest characterized to date in this taxon, notably longer than the 156-nt intron found in the J. libera β-tubulin gene [16]. The Rpl7a intron is also the first reported example of an intron from Trimastix, although other introns have been identified in this organism (A.J. Roger, pers. comm.). Malawimonas jakobiformis appears to have two copies of Rpl7a, each containing the intron. Table 1 summarizes the organisms surveyed for the presence of the Rpl7a intron.
Figure 3 An evolutionarily conserved spliceosomal intron within Rpl7a of representative organisms of the domain Eucarya. Nucleotide sequences at the exon-intron junctions are shown, with the exon sequences in uppercase letters and intron sequences in lowercase. The total length of each intron is shown above the dotted lines. The conventional spliceosomal intronic boundary sequences are bolded (gt...ag). One-letter amino acid abbreviations for the corresponding L7a protein sequences are indicated below the exon sequences. Amino acid position within the L7a protein is indicated at the bottom of the figure and corresponds to the S. cerevisiae sequence, as in Fig. [1]. Sequences from NCBI are: D. discoideum [GenBank:AC116100], D. melanogaster [GenBank:X82782]; H. sapiens [GenBank:X52138]; and G. lamblia [GenBank:AACB01000019]. Sequences determined in the present study are: A. castellanii [GenBank:AY925008]; J. libera [GenBank:AY925006]; T. pyriformis [GenBank:AY925011]; M. californianus [GenBank:AY925003].
Table 1 Organisms surveyed for the presence of the Rpl7a intron
Organism1 Number2 Intron Size (nt)
With intron:
Opisthokonta (Metazoa only)
Vertebrata (7)
Homo sapiens† 277
Chordata (2)
Ciona intestinalis† 586
Hexapoda (5)
Drosophila melanogaster† 401
Amoebozoa
Dictyostelium discoideum† 492
Physarum polycephalum*† 103
Acanthamoeba castellanii*† 82
Hartmannella vermiformis*† 58
Excavata
Jakobidae
Jakoba libera*† 541
Reclinomonas americana*† 877
Seculamonas ecuadoriensis*† 988
Trimastix (Trimastix pyriformis*†) 172
Malawimonadidae
Malawimonas jakobiformis*† 53, 553
Malawimonas californianus*† 54
Diplomonadida (Giardia lamblia) 109
Without intron:
Chromalveolata
Alveolata (14)
Heterokonta (stramenopiles) (3)
Plantae
Streptophyta (5)
Rhodophyta (2)
Chlorophyta (1)
Rhizaria
Cercozoa (Bigelowiella natans*) (1)
Opisthokonta
Fungi (28)
Nematoda/Trematoda (4)
Hexapoda (1)
Capsaspora owczarzaki*
Amoebozoa
Entamoebidae (4)
Excavata
Heterolobosea (Naegleria gruberi*)
Euglenozoa
Trypanosomatidae (7)
Euglenida (Euglena gracilis*)
Parabasalia (Trichomonas vaginalis)
Diplomonadida (Spironucleus barkhanus*)
1An asterisk (*) denotes organisms for which PCR amplification of the Rpl7a gene was performed; † indicates that expressed sequence tag (EST) data exist that confirm the absence of the intron in the corresponding mRNA.
2Numbers in brackets refer to the number of organisms examined within each group; a complete list is provided in Additional File 3.
3Two versions of the Rpl7a gene exist in this organism, containing introns of slightly different length.
Within Opisthokonta and Amoebozoa, several groups appear to lack the Rpl7a intron. Over time introns are randomly lost, so a punctate distribution within established eukaryotic groups is not unexpected. Independent cases of intron loss can be inferred when the relationship among groups is known and the intron is present in a common ancestor. This is the case for taxa from the Opisthokonta and the Entamoebidae (amoebozoons) that are missing the Rpl7a intron. The Rpl7a intron is present in several amoebozoons that branch outside [15,17] of those that lack the intron. A particularly interesting case of predicted intron loss is the Anopheles gambiae Rpl7a [GenBank:AAAB01008960], which lacks any introns. This situation is in stark contrast to other hexapod Rpl7a sequences, which contain introns in addition to the conserved intron discussed here. The A. gambiae case may be an example of intron loss mediated by a reverse transcription mechanism [18,19] and is also consistent with observed intron loss patterns in the hexapods [20].
None of the examined members of Plantae, Chromalveolata or Rhizaria contained the conserved Rpl7a intron. Only a single member of Rhizaria (B. natans) was investigated here so it is premature to conclude that the Rpl7a intron is absent altogether from this supergroup. Our survey failed to detect the Rpl7a intron in representatives of the Euglenozoa, Heterolobosea or Parabasalia, all eukaryotic groups for which inclusion in a larger Excavata assemblage is only weakly supported [6]. Whereas emerging data suggest that E. gracilis is well endowed with spliceosomal introns [21-23], intron distribution appears to be sparse in heteroloboseans [24]. The diplomonad Spironucleus barkhanus does not have the Rpl7a intron, likely a result of intron loss in this organism given the presence of the intron in other related Excavata.
We considered other (less probable) explanations that the phylogenetic distribution of the Rpl7a intron could be explained by independent events of intron gain or by lateral gene transfers. We note that the Rpl7a exon boundary sequences conform poorly to the conserved proto-splice site sequence (A,C)AG/G [25] (Fig. 1 and 3). The predicted amino acids flanking the intron insertion site are highly conserved, including in the archaeal homologs, resulting in a functional constraint on the DNA sequence abutting the intron. This observation argues against a distribution of the Rpl7a intron resulting from multiple intron gains at proto-splice sites. Supporting the argument against intron gain is the apparent very low density of introns in G. lamblia. This is also in agreement with recent data suggesting that few shared intron positions between distantly related taxa are due to parallel gain (i.e., independent insertion) at proto-splice sites [25].
The additional possibility exists that the phylogenetic distribution of the intron, particularly its presence in G. lamblia, could reflect eukaryote-to-eukaryote lateral gene transfer events. A phylogenetic analysis (see Additional File 2) of L7a protein sequences from 20 representative eukaryotes and two archaeons does not indicate an obvious unexpected affinity of G. lamblia with any other eukaryotic group. Cumulatively, the above observations argue that the Rpl7a intron is ancestral to many eukaryotic groupings and has been lost sporadically in various eukaryotic taxa, in which case the Rpl7a intron would be one of the oldest introns found to date.
Phylogenetic implications
Depending on one's views of the eukaryotic tree, there are two possibilities to explain the observed distribution of the Rpl7a intron with regard to intron loss. In the context of more recent proposals of a basally unresolved eukaryotic tree [7,26], the Rpl7a intron would have to have been lost multiple times at the base of Plantae, Chromalveolata and (tentatively) Rhizaria, but maintained in Opisthokonta, Amoebozoa and Excavata. Alternatively, if one assumes a specific relationship among Opisthokonta, Amoebozoa and Excavata, it is only necessary to invoke a single loss of the Rpl7a intron in a common ancestor of Plantae/Chromalveolata/Rhizaria, to explain the apparent absence of this intron in these three supergroups.
While multiple cases of intron gain seem unlikely (as discussed above), a single intron gain at the base of Opisthokonta, Amoebozoa and Excavata would result in the observed distribution, assuming the Plantae/Chromalveolata/Rhizaria had already diverged. While additional intermediate possibilities exist with regards to various intron loss and/or gain events, we propose that a single loss or gain is the most parsimonious explanation for the observed distribution of the Rpl7a intron, supporting a grouping of Opisthokonta, Amoebozoa and Excavata.
Although the distribution of the Rpl7a intron does not position the root of the eukaryotic tree, it may help to resolve some of the basal branches. It is important to note that Excavata is a tenuous grouping that may in fact be polyphyletic: thus, members of Excavata without the intron may fall on either side of a putative loss/gain event. In this instance, it is possible that one or more members of Excavata will be found to group with the Plantae/Chromalveolata/Rhizaria consortium. Conversely, if additional evidence robustly groups the organisms in question within Excavata, this result would imply that they lost the intron independently, as is evidently the case for Fungi within Opisthokonta.
New eukaryotic genome sequences may well reveal the Rpl7a intron in other representatives of Plantae, Chromalveolata and/or (particularly) Rhizaria than those listed in Table 1. Although such a finding would necessarily require reinterpretation of some of the conclusions reached here, discovery of the Rpl7a intron in these other eukaryotic supergroups would only strengthen the argument that this intron is indeed ancient. At present, the simplest explanation for the observed distribution of the Rpl7a intron is a specific relationship uniting Opisthokonta, Amoebozoa and Excavata to the exclusion of Plantae, Chromalveolata and possibly Rhizaria. However, additional characters will need to be found in order to strengthen this proposed assemblage.
Conclusion
The properties possessed by the first identified G. lamblia spliceosomal-type intron (35 nt in length) raised questions regarding the importance of its non-canonical 5' intron boundary and possible constraints on the size of Giardia introns. Furthermore, it was clear that the identification of additional spliceosomal introns would be required to assess the degree of conservation of predicted functional sequence elements within these introns [27]. In this study we have identified in G. lamblia two larger introns (109 and 220 nt) in genes encoding, respectively, L7a and a non-conserved, unassigned protein. Both of these newly identified introns exhibit a canonical GT 5' intron boundary. Evolutionary conservation of the exact position and phase of the Rpl7a intron (within a highly conserved region of the L7a protein sequence in members of diverse eukaryotic groups) indicates that this particular intron was likely present in the ancestor of these lineages. The Rpl7a intron exhibits all of the hallmark features of an ancient intron, such as widespread distribution, phase 0 positioning, location in an ancient eukaryotic gene, and lack of a proto-splice site sequence; thus, this intron can be considered to be a meaningful phylogenetic marker. Currently, no singular genomic marker definitively resolves the branching order of the eukaryotic supergroups. We propose that patterns of conservation of ancient introns, when larger data sets are examined, may provide such information.
Methods
Genomic DNA
Samples of genomic DNA from various protists (full organism names are listed in Table 1) were kindly provided by B.F. Lang (R. americana, ATCC 50394; S. ecuadoriensis, ATCC 50688; M. jakobiformis, ATCC 50310; M. californianus, ATCC 50740); A.J. Roger (S. barkhanus strain NOR-1A, ATCC 50380+; C. owczarzaki, ATCC 30864; T. pyriformis, ATCC 50562; N. gruberi, ATCC 30224); A.J. Lohan (A. castellanii, ATCC 30010; H. vermiformis, ATCC 50236); D.F. Spencer (E. gracilis strain Z); and J.M. Archibald (B. natans, CCMP 621). Genomic DNA from J. libera (ATCC 50422) and P. polycephalum (strain M3C) was obtained by lysing cells in 1% SDS followed by phenol extraction and ethanol precipitation.
Characterization of Rpl7a sequences
Polymerase chain reaction (PCR) was used to amplify Rpl7a from genomic DNA (100–200 ng), using Invitrogen Taq DNA polymerase. PCR primers used are listed in Additional file 3: Supplemental Tables. PCR cycling conditions were: 3 min at 95°C; 35 cycles of 30 sec at 95°C, 30 sec at 55°, 1 min at 72°C; and 10 min at 72°C. PCR product bands were isolated from gels using the Sephaglas™ BandPrep Kit (Amersham Pharmacia Biotech) and the recovered DNA was cloned into the pCR® 2.1-TOPO® vector using the TOPO TA Cloning® Kit (Invitrogen). DNA sequencing was performed using an automated ABI Prism 377 DNA sequencer.
Computer analyses
Ribosomal protein L7a gene and protein sequences were identified by searching relevant sequence databases using TBLASTN or BLASTP with the S. cerevisiae sequences as queries. Protein alignments were generated with ClustalX 1.82 [28] applying default alignment parameters. DNASIS V 2.5 was used for translation of DNA sequences and to assist in the identification of introns within Rpl7a sequences. Searches for spliceosomal-like introns within the preliminary genome sequence data of T. vaginalis and G. lamblia were performed using the sequence pattern search program PatScan [11], recompiled with the constant 'MAX_SEQ-LEN' redefined to '100000000'.
Authors' contributions
AGR analyzed the L7a protein sequences, discovered the G. lamblia Rpl7a intron, performed genomic searches for additional introns in G. lamblia and T. vaginalis and carried out the RT-PCR experiments to confirm the presence of the third G. lamblia spliceosomal intron. TES examined the phylogenetic conservation of the Rpl7a intron in currently available eukaryotic genome databases, performed all the PCR experiments to determine the prevalence of the intron, and participated in the genomic searches for additional G. lamblia introns. RFW performed the phylogenetic analysis of L7a sequences. MWG made substantial intellectual contributions to this work. All authors participated in the assembly and editing of the manuscript. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Evidence for a spliceosomal intron in an unassigned G. lamblia ORF. This file (PDF format) presents gel electrophoretic data documenting the results of PCR and RT-PCR experiments to confirm the existence of a putative spliceosomal intron in a non-conserved, unassigned ORF in G. lamblia, as described in the text.
Click here for file
Additional File 3
Supplemental Tables. This file (PDF format) contains two tables. Table 1 contains the sequences of the oligonucleotide primers used for amplifying Rpl7a sequences from various eukaryotic taxa. Table 2 is a complete list of all the organisms and sequence sources that were examined for the presence of the Rpl7a intron.
Click here for file
Additional File 2
L7aTree. This file (PDF format) presents a maximum likelihood tree generated using L7a sequences from representative eukaryotic and archaeal taxa. Also included are the methods used to generate the tree, the sources of the sequences and the alignment that was used.
Click here for file
Acknowledgements
We are grateful to B.F. Lang (Université de Montréal) and A.J. Roger, A.J. Lohan, D.F. Spencer and J.M. Archibald (Dalhousie University) for generous gifts of genomic DNA. We thank the following members (past and present) of the Roger laboratory for providing Rpl7a EST sequence data: J. Silberman and M. Baumgartner (T. pyriformis), Å. Sjögren and E. Gill (N. gruberi), J. Andersson (S. barkhaus), A.J. Roger (C. owczarzaki). Additional Rpl7a EST sequences were provided by P.J. Keeling, University of British Columbia (B. natans) and B.F. Lang (M. californianus, M. jakobiformis, J. libera, R. americana, S. ecuadoriensis). We also thank J. Yee (Trent University) for generously providing G. lamblia total RNA and for insightful discussions regarding promoter elements; J. Gott (Case Western Reserve University) for helpful discussions; D.F. Spencer for assistance with the PatScan program; and M. Dlutek and A. Fong for performing DNA sequencing. Preliminary sequence data for Trichomonas vaginalis were obtained from The Institute for Genomic Research website at Sequencing of the T. vaginalis genome was funded by the National Institute of Allergy and Infectious Diseases (NIAID). This work was supported by funding to MWG from the Canadian Institutes of Health Research (operating grant MOP-4124) and Genome Canada through Genome Atlantic and the Atlantic Innovation Fund (Protist EST Program). MWG is pleased to acknowledge salary support from the Canadian Research Chairs Program and the Canadian Institute for Advanced Research.
==== Refs
Nixon JE Wang A Morrison HG McArthur AG Sogin ML Loftus BJ Samuelson J A spliceosomal intron in Giardia lamblia Proc Natl Acad Sci USA 2002 99 3701 3705 11854456 10.1073/pnas.042700299
Simpson AG MacQuarrie EK Roger AJ Eukaryotic evolution: early origin of canonical introns Nature 2002 419 270 12239559 10.1038/419270a
Vaňáčová S Yan W Carlton JM Johnson PJ Spliceosomal introns in the deep-branching eukaryote Trichomonas vaginalis Proc Natl Acad Sci USA 2005 102 4430 4435 15764705 10.1073/pnas.0407500102
Fast NM Doolittle WF Trichomonas vaginalis possesses a gene encoding the essential spliceosomal component, PRP8 Mol Biochem Parasitol 1999 99 275 278 10340492 10.1016/S0166-6851(99)00017-1
Collins L Penny D Complex spliceosomal organization ancestral to extant eukaryotes Mol Biol Evol 2005 22 1053 1066 15659557 10.1093/molbev/msi091
Simpson AG Cytoskeletal organization, phylogenetic affinities and systematics in the contentious taxon Excavata (Eukaryota) Int J Syst Evol Microbiol 2003 53 1759 1777 14657103 10.1099/ijs.0.02578-0
Simpson AG Roger AJ The real 'kingdoms' of eukaryotes Curr Biol 2004 14 R693 R696 15341755 10.1016/j.cub.2004.08.038
Cavalier-Smith T A revised six-kingdom system of life Biol Rev Camb Philos Soc 1998 73 203 266 9809012 10.1017/S0006323198005167
Nikolaev SI Berney C Fahrni JF Bolivar I Polet S Mylnikov AP Aleshin VV Petrov NB Pawlowski J The twilight of Heliozoa and rise of Rhizaria, an emerging supergroup of amoeboid eukaryotes Proc Natl Acad Sci USA 2004 101 8066 8071 15148395 10.1073/pnas.0308602101
Simpson AGB Roger AJ Hirt RP, Horner DS Excavata and the origin of amitochondriate eukaryotes Organelles, Genomes, and Eukaryote Phylogeny: An Evolutionary Synthesis in the Age of Genomics 2004 Boca Raton, FL, CRC Press 27 54
Dsouza M Larsen N Overbeek R Searching for patterns in genomic data Trends Genet 1997 13 497 498 9433140 10.1016/S0168-9525(97)01347-4
Yee J Mowatt MR Dennis PP Nash TE Transcriptional analysis of the glutamate dehydrogenase gene in the primitive eukaryote, Giardia lamblia. Identification of a primordial gene promoter J Biol Chem 2000 275 11432 11439 10753960 10.1074/jbc.275.15.11432
Simpson AGB Roger AJ Silberman JD Leipe DD Edgcomb VP Jermiin LS Patterson DJ Sogin ML Evolutionary history of "early-diverging" eukaryotes: the excavate taxon Carpediemonas is a close relative of Giardia Mol Biol Evol 2002 19 1782 1791 12270904
Cavalier-Smith T The phagotrophic origin of eukaryotes and phylogenetic classification of Protozoa Int J Syst Evol Microbiol 2002 52 297 354 11931142
Fahrni JF Bolivar I Berney C Nassonova E Smirnov A Pawlowski J Phylogeny of lobose amoebae based on actin and small-subunit ribosomal RNA genes Mol Biol Evol 2003 20 1881 1886 12949141 10.1093/molbev/msg201
Archibald JM O'Kelly CJ Doolittle WF The chaperonin genes of jakobid and jakobid-like flagellates: implications for eukaryotic evolution Mol Biol Evol 2002 19 422 431 11919283
Baldauf SL Roger AJ Wenk-Siefert I Doolittle WF A kingdom-level phylogeny of eukaryotes based on combined protein data Science 2000 290 972 977 11062127 10.1126/science.290.5493.972
Roy SW Gilbert W The pattern of intron loss Proc Natl Acad Sci USA 2005 102 713 718 15642949 10.1073/pnas.0408274102
Sverdlov AV Babenko VN Rogozin IB Koonin EV Preferential loss and gain of introns in 3' portions of genes suggests a reverse-transcription mechanism of intron insertion Gene 2004 338 85 91 15302409 10.1016/j.gene.2004.05.027
Zdobnov EM von Mering C Letunic I Torrents D Suyama M Copley RR Christophides GK Thomasova D Holt RA Subramanian GM Mueller H-M Dimopoulos G Law JH Wells MA Birney E Charlab R Halpern AL Kokoza E Kraft CL Lai Z Lewis S Louis C Barillas-Mury C Nusskern D Rubin GM Salzberg SL Sutton GG Topalis P Wides R Wincker P Yandell M Collins FH Ribeiro J Gelbart WM Kafatos FC Bork P Comparative genome and proteome analysis of Anopheles gambiae and Drosophila melanogaster Science 2002 298 149 159 12364792 10.1126/science.1077061
Breckenridge DG Watanabe Y-I Greenwood SJ Gray MW Schnare MN U1 small nuclear RNA and spliceosomal introns in Euglena gracilis Proc Natl Acad Sci USA 1999 96 852 856 9927657 10.1073/pnas.96.3.852
Canaday J Tessier LH Imbault P Paulus F Analysis of Euglena gracilis alpha-, beta- and gamma-tubulin genes: introns and pre-mRNA maturation Mol Genet Genomics 2001 265 153 160 11370862 10.1007/s004380000403
Russell AG Watanabe Y-i Charette JM Gray MW Unusual features of fibrillarin cDNA and gene structure in Euglena gracilis: evolutionary conservation of core proteins and structural predictions for methylation-guide box C/D snoRNPs throughout the domain Eucarya Nucleic Acids Res 2005 33 2781 2791 15894796 10.1093/nar/gki574
Logsdon JM Jr The recent origins of spliceosomal introns revisited Curr Opin Genet Dev 1998 8 637 648 9914210 10.1016/S0959-437X(98)80031-2
Sverdlov AV Rogozin IB Babenko VN Koonin EV Conservation versus parallel gains in intron evolution Nucleic Acids Res 2005 33 1741 1748 15788746 10.1093/nar/gki316
Baldauf SL The deep roots of eukaryotes Science 2003 300 1703 1706 12805537 10.1126/science.1085544
Johnson PJ Spliceosomal introns in a deep-branching eukaryote: The splice of life Proc Natl Acad Sci USA 2002 99 3359 3361 11904397 10.1073/pnas.072084199
Chenna R Sugawara H Koike T Lopez R Gibson TJ Higgins DG Thompson JD Multiple sequence alignment with the Clustal series of programs Nucleic Acids Res 2003 31 3497 3500 12824352 10.1093/nar/gkg500
|
16109161
|
PMC1201135
|
CC BY
|
2021-01-04 16:29:16
|
no
|
BMC Evol Biol. 2005 Aug 18; 5:45
|
utf-8
|
BMC Evol Biol
| 2,005 |
10.1186/1471-2148-5-45
|
oa_comm
|
==== Front
BMC Emerg MedBMC Emergency Medicine1471-227XBioMed Central London 1471-227X-5-61612238010.1186/1471-227X-5-6Research ArticleWhich diagnostic tests are most useful in a chest pain unit protocol? Goodacre Steve [email protected] Thomas [email protected] Jane [email protected] Karen [email protected] Francis [email protected] Medical Care Research Unit, Regent Court, 30 Regent Street, Sheffield, S1 4DA, UK2 Emergency Department, Northern General Hospital, Herries Road, Sheffield, S5 7AU, UK2005 25 8 2005 5 6 6 27 1 2005 25 8 2005 Copyright © 2005 Goodacre et al; licensee BioMed Central Ltd.2005Goodacre 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 chest pain unit (CPU) provides rapid diagnostic assessment for patients with acute, undifferentiated chest pain, using a combination of electrocardiographic (ECG) recording, biochemical markers and provocative cardiac testing. We aimed to identify which elements of a CPU protocol were most diagnostically and prognostically useful.
Methods
The Northern General Hospital CPU uses 2–6 hours of serial ECG / ST segment monitoring, CK-MB(mass) on arrival and at least two hours later, troponin T at least six hours after worst pain and exercise treadmill testing. Data were prospectively collected over an eighteen-month period from patients managed on the CPU. Patients discharged after CPU assessment were invited to attend a follow-up appointment 72 hours later for ECG and troponin T measurement. Hospital records of all patients were reviewed to identify adverse cardiac events over the subsequent six months. Diagnostic accuracy of each test was estimated by calculating sensitivity and specificity for: 1) acute coronary syndrome (ACS) with clinical myocardial infarction and 2) ACS with myocyte necrosis. Prognostic value was estimated by calculating the relative risk of an adverse cardiac event following a positive result.
Results
Of the 706 patients, 30 (4.2%) were diagnosed as ACS with myocardial infarction, 30 (4.2%) as ACS with myocyte necrosis, and 32 (4.5%) suffered an adverse cardiac event. Sensitivities for ACS with myocardial infarction and myocyte necrosis respectively were: serial ECG / ST segment monitoring 33% and 23%; CK-MB(mass) 96% and 63%; troponin T (using 0.03 ng/ml threshold) 96% and 90%. The only test that added useful prognostic information was exercise treadmill testing (relative risk 6 for cardiac death, non-fatal myocardial infarction or arrhythmia over six months).
Conclusion
Serial ECG / ST monitoring, as used in our protocol, adds little diagnostic or prognostic value in patients with a normal or non-diagnostic initial ECG. CK-MB(mass) can rule out ACS with clinical myocardial infarction but not myocyte necrosis(defined as a troponin elevation without myocardial infarction). Using a low threshold for positivity for troponin T improves sensitivity of this test for myocardial infarction and myocardial necrosis. Exercise treadmill testing predicts subsequent adverse cardiac events.
==== Body
Background
The chest pain unit (CPU) has been developed to provide standardised care for patients presenting with acute chest pain, undiagnosed by initial clinical assessment, electrocardiogram (ECG) and chest radiograph. The CPU aims to rapidly diagnose acute coronary syndrome (ACS), providing early access to appropriate care for those with positive test results and discharge home for those who test negative. Although a variety of different tests have been used in CPU protocols, most units use a combination of ECG monitoring, biochemical diagnostic testing and provocative cardiac testing to diagnose ACS [1]. This has led to some debate regarding the most appropriate testing regime. In particular, the role of the exercise treadmill test has been questioned [2] and does not feature in some protocols [3].
Since 1999 a CPU has been operating at the Northern General Hospital in Sheffield. Patients are investigated with a combination of serial ECG recording, ST segment monitoring, biochemical cardiac markers (CK-MB(mass) and troponin T) and an exercise treadmill test [4]. We aimed to determine the value of each of the constituent elements of this protocol by measuring: a) the diagnostic accuracy (sensitivity, specificity and likelihood ratios) of each test for ACS at presentation, and b) prognostic value of each test for major adverse cardiac events over six months.
Methods
The Northern General Hospital is a 1100-bedded urban teaching hospital with the only adult emergency department for the 530,000 population of Sheffield, United Kingdom. The emergency department see approximately 90,000 new patients per year, of whom approximately 6% have chest pain. The hospital is a tertiary referral centre for cardiology, with cardiac catheterisation and cardiac surgery facilities. The CPU is a two-bedded, nurse-run unit based in the emergency department. Patients are assessed on the unit if they present with acute chest pain and have a normal or non-diagnostic ECG, no co-morbidity requiring admission, no serious alternative cause for chest pain (such as pulmonary embolus) and pain that is potentially compatible with cardiac ischaemia (i.e. not chest wall pain). Patients with known coronary heart disease can be assessed on the CPU provided they do not have prolonged (> one hour) or recurrent (more than one episode) pain that is characteristic of their angina.
The diagnostic testing regime consists of ST segment monitoring and serial ECG recording every hour for two to six hours; CK-MB(mass) assay on arrival, repeated either two hours later or six hours after symptom onset, whichever is later; troponin T assay at least six hours after symptom onset; and, if appropriate, exercise treadmill testing. Exercise testing is not performed if any preceding test is positive, if the patient has had recent diagnostic testing for coronary heart disease (e.g. coronary angiography), or if they are unable to exercise.
From 1/3/99 to 30/9/00 presenting details, CPU processes and diagnostic test results were prospectively recorded for all patients assessed on the CPU. Patients admitted after assessment received further testing and management at the discretion of the admitting physician and were followed up by case note review. Patients discharged after assessment were invited to attend a review appointment on the CPU 72 hours later for clinical assessment, ECG and troponin T measurement.
Six months after assessment the emergency department computer database was searched for details of any further hospital attendances or admissions. Case notes were reviewed for all cases identified. The general practitioners of all patients presenting between 1/3/99 and 29/2/00 were contacted by post to determine if they had suffered any adverse cardiac event or received diagnostic cardiac testing or procedures during the previous six months. This survey revealed no previously unidentified episodes so for patients attending from 1/3/00 to 30/9/00 the general practitioner was only contacted if the patient resided outside the Sheffield area. We did not attempt to contact patients at six months to confirm whether they were still alive.
ST segment monitoring and serial ECG recording was performed using Spacelabs monitors. These functions will be evaluated together, because: 1) serial ECG recording is facilitated and influenced by the availability of this monitoring equipment, and 2) ECG changes detected on ST segment monitoring may also be detected on serial recording and vice-versa. Each patient received between two and six hours of ST segment monitoring and at least two ECGs, one hour apart. Results were dichotomised into positive (>1 mm ST segment elevation or depression; T wave inversion > 2 mm or normalisation; ventricular arrhythmia; or second or third degree heart block) or negative (none of these changes). ECGs were interpreted by the specialist chest pain nurses who managed the unit. Positive changes were defined a priori and used for CPU decision-making.
The absolute value of CK-MB(mass) at baseline and on repeat sampling was defined as positive if it exceeded 5 ng/ml. The rise in CK-MB(mass) (delta CK-MB(mass)) was defined as positive if it exceeded 1.6 ng/ml. The diagnostic performance of baseline, repeat assay and change in CK-MB(mass) were analysed separately and then combined to give an overall result. This was defined as positive if either assay or the change were positive, and negative if all three results were negative. Two thresholds for positivity were analysed for troponin T: 1) the traditional threshold of 0.1 ng/ml or above, and 2) a more sensitive threshold of 0.03 ng/l. The CK-MB(mass) thresholds and the 0.1 ng/ml threshold for troponin were defined a priori and used for decision making on the CPU. The 0.03 ng/ml troponin threshold was only used for post hoc analysis.
The exercise treadmill test was performed according to the Bruce protocol. A positive test was defined as more than 1 mm horizontal or down-sloping ST segment depression; more than 1 mm ST elevation; or ventricular arrhythmia. For the purposes of this analysis, any test that did not produce these changes was considered to be negative, regardless of the duration of exercise or heart rate achieved.
The diagnostic performance of all tests except the exercise treadmill test was evaluated by comparison to the reference standard diagnosis of ACS at presentation, categorised as recently recommended [5] into: 1) ACS with clinical myocardial infarction (defined as a typical clinical syndrome associated with a troponin T elevation greater than 1.0 ng/ml), or 2) ACS with myocyte necrosis (defined as a typical clinical syndrome with any detectable troponin T less than 1.0 ng/ml). Sensitivity, specificity, positive and negative predictive values, and likelihood ratios were calculated for each test, along with 95% confidence intervals.
The prognostic value of each test was evaluated by measuring the relative risk of a major adverse cardiac event during the six months following a positive result, compared to a negative result. The following were defined as major adverse cardiac events: cardiac death, non-fatal myocardial infarction, arrhythmia, or revascularisation procedure. Because revascularisation procedures may be precipitated by the diagnostic tests under investigation, we repeated the analysis with these excluded from the definition.
Data were analysed using SPSS for windows version 11.5 (SPSS Inc., Chicago) and 95% confidence intervals for all proportions, likelihood ratios and risk ratios were calculated using CIA, Confidence Interval Analysis software. The North Sheffield Research Ethics Committee approved the evaluation of the chest pain unit.
Results
During the study period 706 patients were assessed on the CPU. Details of the study population characteristics are outlined in table 1. All patients received baseline blood sampling. This was performed at a mean time from symptom onset of 6.2 hours (median 3.0 hours). Repeat blood sampling was performed on 604 patients (86%) at a mean time from symptom onset of 8.6 hours (median 6.5 hours). After CPU assessment 596 (84%) were discharged home and 110 (16%) were admitted. Of those discharged, 515 (86%) returned for review at 72 hours. At review one patient had a previously undetected rise in troponin T (0.8 ng/ml) and was classified as having ACS with myocyte necrosis. All other cases of ACS were detected by the CPU protocol.
Table 1 Characteristics of the study population
Mean age 53.6 years (range 22 to 99)
Male 431 (61%)
Known coronary heart disease 150 (21%)
Diabetes 48 (7%)
Hypertension 181 (26%)
Hyperlipidaemia 114 (16%)
Smoker 215 (35%)
ACS with clinical myocardial infarction was ultimately in diagnosed in 30 patients (4.2%), while another 30 (4.2%) had ACS with myocyte necrosis. Over the following six months 32 patients (4.5%) had a major adverse cardiac event. These included eight cardiac deaths, five non-fatal myocardial infarctions, two arrhythmias, and 17 revascularisation procedures.
Serial ECG/ST segment monitoring was performed for 690 patients, baseline CK-MB(mass) was measured in 687, repeat CK-MB(mass) was measured in 601, troponin T was measured in 686 patients, and 422 patients received an exercise treadmill test. Table 2 shows the sensitivity, specificity, positive and negative predictive values and table 3 shows the likelihood ratios of the tests for ACS with clinical myocardial infarction at presentation. The same parameters are outlined for ACS with clinical myocardial infarction or myocyte necrosis in tables 4 and 5. Serial ECG / ST segment monitoring added relatively little to the assessment. None of the positive cases involved ST elevation myocardial infarction requiring reperfusion therapy. All the biochemical markers were useful for ruling in ACS with clinical myocardial infarction and ACS with myocyte necrosis. The second CK-MB(mass) sample and the delta CK-MB(mass) were useful for ruling out myocardial infarction, but did not reliably rule out myocyte necrosis. Using a lower threshold for troponin T positivity markedly improved sensitivity with only a modest loss of specificity. Only troponin T, using a lower threshold, was useful for ruling out ACS with myocyte necrosis. However, it should be recognised that this diagnosis is based upon detection of troponin T.
Table 2 Sensitivity, specificity, positive and negative predictive values for the diagnosis of ACS with clinical myocardial infarction at presentation
Sensitivity (95% CI) Specificity (95% CI) Positive predictive value (95% CI) Negative predictive value (95% CI)
Serial ECG & ST monitoring N = 690 33.3% (19.2 to 51.2) 95.3% (93.4 to 96.7) 24.4% (13.8 to 39.3) 96.9% (95.3 to 98.0)
Initial CK-MB(mass) N = 687 63.3% (45.5 to 78.1) 97.0% (95.3 to 98.0) 48.7% (33.9 to 63.8) 98.3% (97.0 to 99.0)
Delayed CK-MB(mass) N = 601 95.7% (79.0 to 99.2) 96.2% (94.3 to 97.5) 50.0% (35.8 to 64.2) 99.8% (99.0 to 100)
Delta CK-MB(mass) N = 601 95.7% (79.0 to 99.2) 98.4% (97.0 to 99.1) 70.0% (53.4 to 83.9) 99.8% (99.0 to 100)
Any positive CK-MB(mass) N = 601 100% (88.6 to 100) 95.6% (93.7 to 96.9) 50.8% (38.3 to 63.2) 100% (99.4 to 100)
Troponin T > = 0.1 ng/ml N = 686 83.3% (66.4 to 92.7) 98.8% (97.7 to 99.4) 75.8% (59.0 to 87.2) 99.3% (98.3 to 99.7)
Troponin T > = 0.03 ng/ml N = 686 96.4% (82.3 to 99.4) 95.9% (94.1 to 97.2) 50.0% (37.1 to 62.9) 99.8% (99.1 to 100)
Table 3 Likelihood ratios for the diagnosis of ACS with clinical myocardial infarction at presentation
Positive likelihood ratio (95% CI) Negative likelihood ratio (95% CI)
Serial ECG & ST monitoring N = 690 7.1 (3.8 to 13.1) 0.700 (0.543 to 0.901)
Initial CK-MB(mass) N = 687 20.8 (12.5 to 34.7) 0.378 (0.236 to 0.605)
Delayed CK-MB(mass) N = 601 25.1 (16.5 to 38.2) 0.045 (0.007 to 0.307)
Delta CK-MB(mass) N = 601 61.4 (39.1 to 118.1) 0.044 (0.006 to 0.300)
Any positive CK-MB(mass) N = 601 ~25* ~0*
Troponin T > = 0.1 ng/ml N = 686 70.4 (34.1 to 140.8) 0.169 (0.079 to 0.387)
Troponin T > = 0.03 ng/ml N = 686 23.5 (16.1 to 34.2) 0.037 (0.005 to 0.255)
*Unable to calculate precisely as no false negative were recorded
Table 4 Sensitivity, specificity, positive and negative predictive values for the diagnosis of ACS with myocyte necrosis or clinical myocardial infarction at presentation
Sensitivity (95% CI) Specificity (95% CI) Positive predictive value (95% CI) Negative predictive value (95% CI)
Serial ECG & ST monitoring N = 690 23.3% (14.4 to 35.4) 95.7% (93.8 to 97.0) 34.1% (21.6 to 49.5) 92.9% (90.7 to 94.6)
Initial CK-MB(mass) N = 687 43.3% (31.6 to 55.9) 97.9% (96.5 to 98.8) 66.7% (51.0 to 79.4) 94.8% (92.8 to 96.2)
Delayed CK-MB(mass) N = 601 63.5% (49.9 to 75.2) 98.0% (96.4 to 98.9) 75.0% (60.6 to 85.4) 96.6% (94.7 to 97.8)
Delta CK-MB(mass) N = 601 55.8% (42.3 to 68.4) 99.6% (98.7 to 99.9) 93.5% (79.3 to 98.2) 96.0% (94.0 to 97.3)
Any positive CK-MB(mass) N = 601 71.7% (59.2 to 81.5) 97.4% (95.9 to 98.4) 72.9% (60.4 to 82.6) 97.3% (95.7 to 98.3)
Troponin T > = 0.1 ng/ml N = 686 55.0% (42.5 to 66.9) 100% (99.4 to 100) 100% (89.6 to 100) 96.0% (94.2 to 97.2)
Troponin T > = 0.03 ng/ml N = 686 89.7% (79.2 to 95.2) 99.7% (98.8 to 99.9) 96.3% (87.5 to 99.0) 99.6% (97.9 to 99.9)
Table 5 Likelihood ratios for the diagnosis of ACS with myocyte necrosis or clinical myocardial infarction at presentation
Positive likelihood ratio (95% CI) Negative likelihood ratio (95% CI)
Serial ECG & ST monitoring N = 690 5.4 (3.0 to 9.8) 0.801 (0.696 to 0.922)
Initial CK-MB(mass) N = 687 20.9 (11.3 to 38.5) 0.579 (0.464 to 0.722)
Delayed CK-MB(mass) N = 601 31.7 (17.0 to 58.9) 0.373 (0.261 to 0.534)
Delta CK-MB(mass) N = 601 153.1 (37.6 to 623.5) 0.444 (0.327 to 0.602)
Any positive CK-MB(mass) N = 601 28.1 (16.9 to 46.7) 0.291 (0.194 to 0.435)
Troponin T > = 0.1 ng/ml N = 686 ~25* ~0*
Troponin T > = 0.03 ng/ml N = 686 281.5 (70.4 to 1126) 0.104 (0.049 to 0.221)
*Unable to calculate precisely as no false positives were recorded
Table 6 shows the relative risk of all cardiac events over the subsequent six months for each diagnostic test and table 7 shows the relative risk of cardiac death, non-fatal AMI or arrhythmia. Positive results for serial ECG / ST segment monitoring or biochemical cardiac markers showed a weak association with cardiac events over the following six months. This association disappeared when revascularisation procedures were excluded from the definition of cardiac events. Positive exercise stress test result showed a much stronger association with cardiac events, which was maintained even when revascularisation procedures were excluded.
Table 6 Relative risk of a cardiac event over the following six months
Test Event rate:
positive test Event rate:
negative test Relative risk (95% CI)
Serial ECG & ST monitoring N = 690 5/41 27/649 2.72 (1.10 to 6.74)
Initial CK-MB(mass) N = 687 5/39 27/648 2.84 (1.15 to 7.02)
Delayed CK-MB(mass) N = 601 5/44 23/557 2.57 (1.02 to 6.47)
Delta CK-MB(mass) N = 601 3/31 25/570 2.10 (0.67 to 6.61)
Any positive CK-MB(mass) N = 601 8/59 24/628 3.24 (1.52 to 6.93)
Troponin T > = 0.1 ng/ml N = 686 8/33 24/673 6.80 (3.31 to 13.96)
Troponin T > = 0.03 ng/ml N = 686 9/54 23/632 4.58 (2.23 to 9.40)
Exercise treadmill test N = 422 9/37 4/385 19.03 (6.10 to 59.34)
Table 7 Relative risk of cardiac death, non-fatal myocardial infarction or arrhythmia over the following six months
Test Event rate:
positive test Event rate:
negative test Relative risk
Serial ECG & ST monitoring N = 690 1/41 15/649 1.06 (0.14 to 7.79)
Initial CK-MB(mass) N = 687 0/39 16/648 -
Delayed CK-MB(mass) N = 601 1/44 13/557 0.97 (0.13 to 7.28)
Delta CK-MB(mass) N = 601 0/31 14/570 -
Any positive CK-MB(mass) N = 601 1/59 15/628 0.71 (0.09 to 5.28)
Troponin T > = 0.1 ng/ml N = 686 1/33 15/673 1.36 (0.18 to 9.98)
Troponin T > = 0.03 ng/ml N = 686 2/54 14/632 1.67 (0.39 to 7.17)
Exercise treadmill test N = 422 2/37 3/385 6.63 (1.14 to 38.50)
Discussion
These findings suggest that serial ECG recording and ST segment monitoring, as applied in our protocol, add little diagnostic value in patients with a normal or non-diagnostic initial ECG. Only a minority of cases of ACS were detected by ECG monitoring and most of the positive results from this test were false positives. Serial ECG recording and ST segment monitoring offers the potential advantage of allowing early detection of ST segment elevation myocardial infarction that may be eligible for thrombolysis [6], but no such cases occurred in our cohort. Our findings are consistent with those of Decker et al [7] who found that serial ECG was of limited value in their CPU protocol. It should be appreciated, however, that our protocol used a variable duration of ST segment monitoring and a variable number of serial ECGs, with some patients receiving only two hours of monitoring and two serial ECGs. Hence these findings may not apply to more prolonged regimes, or to populations with a higher prevalence of ACS.
Baseline CK-MB(mass) testing allows early detection of ACS and calculation of a CK-MB(mass) rise. However, the initial CK-MB(mass) alone has insufficient sensitivity to rule out ACS. The addition of a repeat sample at least two hours later and at least six hours after the onset of pain provides adequate sensitivity to rule out ACS with myocardial infarction, but not myocyte necrosis. Our estimate of sensitivity for myocardial infarction is consistent with previously published estimates [3,8,9]. If we simply wish to rule out myocardial infarction then this would appear to be a relatively cheap and effective way of achieving that aim.
Using a traditional threshold for positivity of 0.1 ng/ml for troponin T [10,11] in this protocol does not reliably rule out myocardial infarction or myocyte necrosis. However, if a lower threshold of 0.03 ng/ml is used then 96% of cases of myocardial infarction and almost 90% of cases of myocyte necrosis will be detected. This finding is consistent with previously published data showing that use of a lower threshold for positivity for troponin T is associated with improved early sensitivity [12]. However, since we tested this threshold as a post hoc analysis, further prospective validation of the performance of a lower threshold is required. Furthermore, lowering the threshold would be expected to reduce specificity and lead to more false positives being generated. Potential incorporation bias (see limitations section) means that it is difficult to determine the impact of altering the threshold upon specificity.
The question of whether to use CK-MB(mass) or troponin T or both is a subjective judgement in which the benefits of detecting and treating cases of ACS must be weighed against the costs of additional testing and management of false positives. ACS with myocardial infarction is associated with a markedly increased risk of adverse events[5,10] and thus an increased expectation of benefit from treatment [13]. The risks of ACS with myocyte necrosis are smaller, but still indicate potential benefit from detection and treatment [5]. In estimating the potential costs and benefits of detecting cases of ACS we also need to consider prevalence. CPU patients have a low prevalence of ACS, so a large number of additional patients will need to be tested to detect a small number of additional cases.
The value of exercise treadmill testing in a CPU protocol has been questioned [2]. It may be costly to implement and risks generating large number of false positives due to poor specificity. This study has shown that exercise treadmill testing offers useful prognostic information that is not provided by ECG or biochemical testing. Meanwhile, concerns about false positives are undermined by evidence from a recent randomised controlled trial [14] in which two-thirds of patients randomised to CPU care received an exercise treadmill test compared to only one-third of the routine care group. Despite this difference the rate of referral for angiography was identical. Whether the additional prognostic information provided by treadmill testing justifies the addition cost remains debatable.
This study has a number of limitations that should be appreciated. The reference standard for diagnosis was not independent of the tests being evaluated. Troponin elevation was the diagnostic criterion for ACS so estimates of sensitivity and particularly specificity will be subject to incorporation bias. Any patient who has an elevated troponin will, by definition, have a diagnosis of ACS, unless repeat sampling shortly afterwards is negative. Therefore specificity of troponin, using this definition of ACS, is expected to be high. The value of our analysis of troponin therefore lies in assessment of early sensitivity, particularly in comparing the early sensitivity of using different thresholds for positivity.
Caregivers were aware of the results of all the tests under evaluation and thus would be more likely to rigorously follow-up those with positive tests, raising the possibility of work-up bias. Patients with positive tests were admitted while those with negative tests were discharged. Although 86% of discharged patients attended review at 72 hours, it is possible that some patients with false negative initial tests may have failed to attend follow-up, and were thus misclassified as true negative. A similar bias may influence the association between test results and adverse events when revascularisation procedures are included in the definition, but not the association between treadmill testing and adverse events limited to cardiac death, non-fatal myocardial infarction or arrhythmia.
Statistical analysis inevitably requires some simplification of the data used. For ST segment monitoring / serial ECG and for exercise treadmill testing this involved categorising all results into positive or negative. For exercise treadmill testing inconclusive results were classified as negative. This approach may lead to under-estimation of the diagnostic value of the tests. Furthermore, since patients with new abnormalities on their ECG were excluded from CPU evaluation, it is perhaps not surprising that further ECG-based tests had limited value. This again reinforces the importance of not extrapolating these findings to different populations, such as those with a high prevalence of ACS.
Conclusion
ST segment monitoring and serial ECG recording, as applied in our protocol, appears to add little diagnostic value in patients with a normal or non-diagnostic initial ECG. CK-MB(mass) measurement allows reliable detection of ACS with myocardial infarction but not myocyte necrosis (defined as a troponin elevation without myocardial infarction). Using a low threshold for positivity for troponin T improves sensitivity of this test for myocardial infarction. Exercise treadmill testing offers additional prognostic information regarding the risk of subsequent adverse cardiac events.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SG conceived the idea, planned the study, analysed the data, and wrote the paper. TL, JA and KA collected the data and helped to write the paper. FM helped conceive the idea and write the paper. All authors contributed to the final draft.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Goodacre SW Should we establish chest pain observation units in theUnited Kingdom? A systematic review and critical appraisal of the literature J Accid Emerg Med 2000 17 1 6 10658981
Lindsay J Bonnet YD Pinnow EE Routine stress testing for triage of patients with chest pain: Is it worth the candle? Ann Emerg Med 1998 32 600 3 9795325
Herren KR Mackway-Jones K Richards CR Seneviratne CJ France MW Cotter L Is it possible to exclude a diagnosis of myocardial damage within six hours of admission to an emergency department? Diagnostic cohort study BMJ 2001 323 372 4 11509427 10.1136/bmj.323.7309.372
Goodacre S Morris FP Campbell S Angelini K Arnold J A prospective, observational study and cost analysis of a chest pain observation unit Emerg Med J 2002
Fox KA Birkhead J Wilcox R Knight C Barth J British Cardiac Society Working Group British Cardiac Society Working Group on the definition of myocardial infarction Heart 2004 90 603 609 15145852 10.1136/hrt.2004.038679
Fesmire FM Percy RF Bardoner JB Wharton DR Calhoun FB Usefulness of automated serial 12-lead ECG monitoring during the initial emergency department evaluation of patients with chest pain Ann Emerg Med 1998 31 3 11 9437335
Decker WW Prina LD Smars PA Boggust AJ Zinsmeister AR Kopecky SL Continuous 12-lead electrocardiographic monitoring in an emergency department chest pain unit: an assessment of potential clinical effect Ann Emerg Med 2003 41 342 51 12605201 10.1067/mem.2003.78
Fesmire FM Percy RF Bardoner JB Wharton DR Calhoun FB Serial creatinine kinase (CK) MB testing during the emergency department evaluation of chest pain: utility of a 2-hour deltaCK-MB of +1.6 ng/ml Am Heart J 1998 136 237 44 9704684 10.1053/hj.1998.v136.89571
Bholasingh R de Winter RJ Fischer JC Koster RW Peters RJ Sanders GT Safe discharge from the cardiac emergency room with a rapid rule-out myocardial infarction protocol using serial CK-MB(mass) Heart 2001 85 143 8 11156661 10.1136/heart.85.2.143
Ohman EM Armstrong PW Christenson RH Granger CB Katus HA Hamm CW O'Hanesian M Wagner GS Kleiman NS Harrell FE Califf RM Topol EJ Lee KL for the GUSTO Investigators Cardiac troponin T levels for risk stratification in acute myocardial ischaemia N Engl J Med 1996 335 1333 41 8857016 10.1056/NEJM199610313351801
Hamm CW Goldman BW Heeschen C Kreymnn G Berger J Meinertz T Emergency room triage of patients with acute chest pain by means of rapid testing for cardiac troponin T or troponin I N Engl J Med 1997 337 1648 53 9385123 10.1056/NEJM199712043372302
Collinson PO Stubbs PC Kessler A-C Multicentre evaluation of the diagnostic value of cardiac troponin T, CK-MB mass, and myoglobin for assessing patients with suspected acute coronary syndromes in routine clinical practice Heart 2003 89 280 286 12591831 10.1136/heart.89.3.280
Lindahl B Venge P Wallentin L The FRISC experience with troponin T. Use as a decision tool and comparison with other prognostic markers Eur Heart J 1998 19 N51 N58 9857941
Goodacre S Nicholl J Dixon S Cross E Angelini K Arnold J Revill S Locker T Capewell S Quinney D Campbell S Morris F Randomised controlled trial and economic evaluation of chest pain observation unit versus routine care BMJ 2004 328 254 7 14724129 10.1136/bmj.37956.664236.EE
|
16122380
|
PMC1201136
|
CC BY
|
2021-01-04 16:31:03
|
no
|
BMC Emerg Med. 2005 Aug 25; 5:6
|
utf-8
|
BMC Emerg Med
| 2,005 |
10.1186/1471-227X-5-6
|
oa_comm
|
==== Front
BMC GeriatrBMC Geriatrics1471-2318BioMed Central London 1471-2318-5-101612238910.1186/1471-2318-5-10Research ArticleHealth status in older hospitalized patients with cancer or non-neoplastic chronic diseases Corsonello Andrea [email protected] Claudio [email protected] Luciana [email protected] Francesco [email protected] Bruno [email protected] Incalzi Raffaele [email protected] Gruppo Italiano di Farmacovigilanza nell'Anziano (GIFA) investigators 1 Istituto Nazionale di Ricovero e Cura per Anziani (INRCA), Cosenza, Italy2 Cattedra di Geriatria, Università Campus-BioMedico, Rome, Italy3 Centro di Medicina dell'Invecchiamento, Policlinico "Agostino Gemelli", Università Cattolica del Sacro Cuore, Rome, Italy4 Dipartimento di Medicina Interna, Università degli Studi di Messina, Messina, Italy2005 25 8 2005 5 10 10 28 12 2004 25 8 2005 Copyright © 2005 Corsonello et al; licensee BioMed Central Ltd.2005Corsonello 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
Whether cancer is more disabling than other highly prevalent chronic diseases in the elderly is not well understood, and represents the objective of the present study.
Methods
We used data from the Gruppo Italiano di Farmacovigilanza nell'Anziano (GIFA) study, a large collaborative observational study based in community and university hospitals located throughout Italy. Our series consisted of three groups of patients with non-neoplastic chronic disease (congestive heart failure, CHF, N = 832; diabetes mellitus, N = 939; chronic obstructive pulmonary disease, COPD, N = 399), and three groups of patients with cancer (solid tumors without metastasis, N = 813; solid tumors with metastasis, N = 259; leukemia/lymphoma, N = 326). Functional capabilities were ascertained using the activities of daily living (ADL) scale, and categorical variables for dependency in at least 1 ADL or dependency in 3 or more ADLs were considered in the analysis. Cognitive status was evaluated by the 10-items Hodgkinson Abbreviated Mental Test (AMT).
Results
Cognitive impairment was more prevalent in patients with CHF (28.0%) or COPD (25.8%) than in those with cancer (solid tumors = 22.9%; leukemia/lymphoma = 19.6%; metastatic cancer = 22.8%). Dependency in at least 1 ADL was highly prevalent in patients with metastatic cancer (31.3% vs. 24% for patients with CHF and 22.4% for those with non-metastatic solid tumors, p < 0.001). In people aged 80 years or more, metastatic cancer was not associated with increased prevalence of physical disability. In multivariable analysis, metastatic cancer was associated with a greater prevalence of physical (OR 2.09, 95%CI 1.51–2.90) but not cognitive impairment (OR 1.34, 95%CI 0.94–1.91) with respect to CHF patients. Finally, diabetes was significantly associated with cognitive impairment (OR 1.40, 95%CI 1.11–1.78).
Conclusion
Cancer should not be considered as an ineluctable cause of severe cognitive and physical impairment, at least not more than other chronic conditions highly prevalent in older people, such as CHF and diabetes mellitus.
==== Body
Background
Acute disabling conditions such as stroke or hip fracture have obvious and dramatic effects on functional capabilities, whereas chronic conditions which do not cause a segmental motor deficit have a more complex and less predictable effect. Clinical observations suggest that chronic diseases may be associated with different patterns of physical decline. For example, a distinctive pattern of disability has been found in chronic obstructive pulmonary disease (COPD) compared with that characterizing patients with congestive heart failure (CHF) or diabetes mellitus [1].
Physical dependency can be seen as the end result of the complex interaction among physical, cognitive and affective factors. Cancer is commonly perceived as a highly disabling condition, whereas the impact of other chronic conditions such as diabetes mellitus or hypoxemic COPD on functional capabilities is underestimated [2,3].
Given et al reported that, at the time of the first diagnosis, older cancer patients have considerably better physical function than persons of the same age from the general population [4]. Thus, cancer, on average, might not be more disabling than other highly prevalent chronic diseases in the elderly. Clarifying this issue might be relevant to quantify the needs of care besides the expenditure directly related to the treatment of cancer as well as to select patients most likely to benefit from a comprehensive assessment program. [5]. Indeed, interventions guided by geriatric assessment have positive effects on a number of important health outcomes in frail older patients in different settings [6-9]. However, older cancer patients are underrepresented in geriatric assessment and intervention trials [10]. This makes desirable to clarify the impact of cancer on physical and mental capabilities in comparison with that of conditions such as CHF, COPD and diabetes mellitus which were highly prevalent in geriatric series proven to benefit form geriatric assessment [6-9]. This is the objective of the present study.
Methods
We used data from the Gruppo Italiano di Farmacovigilanza nell'Anziano (GIFA) study, a large collaborative observational study that periodically surveys drug consumption, occurrence of adverse drug reactions (ADR), and quality of hospital care. We used data on patients consecutively admitted to the participating centers during the 4 months surveys carried out in 1993, 1995, 1997 and 1998. Methods of the GIFA have been previously described [11]. Briefly, after obtaining a written informed consent, all patients admitted to the 81 participating wards of Geriatric or Internal Medicine in tertiary hospitals located throughout Italy were enrolled and followed until discharge. There were no inclusion or exclusion criteria. The majority of patients were admitted from the Emergency Room at each hospital, and the diagnosis made by the on-call physician in the Emergency Room was recorded. For each patient a questionnaire was completed at admission and updated daily by a study physician who received specific training for the study.
Data recorded included demographic characteristics, drugs taken prior to admission and during hospital stay, and those prescribed at discharge, ADR, routine blood examination tests, cognitive function, admission and discharge diagnoses. All data were recorded at the clinical center on a microcomputer by means of a dedicated software. Such a software controlled the suitability and the internal consistency of the data so that impossible values or contradictory information could not be entered. The software allowed automatic coding of diagnoses, of ADRs and of drugs by simple typing the description of the disease, of the ADR, or of the commercial name of the drug. Procedures conformed to guidelines provided by the Catholic University Ethical Committee.
Overall, 17,186 patients were enrolled in the study period. Patients who died during hospital stay were excluded from the analysis to avoid the bias due to the presence of terminal illness. We selected five groups of patients on the basis of their first-listed diagnosis using the International Classification of Diseases 9th revision Clinical Modification (ICD9-CM) codes [12]. Three groups consisted of patients with non-neoplastic chronic disease (congestive heart failure, N = 832; diabetes mellitus, N = 939; chronic obstructive pulmonary disease, N = 399), and were compared to three groups of patients with cancer: solid tumors (gastrointestinal, lung, breast, prostate, oro-pharyngeal, bone, and genito-urinary cancer) without metastasis (N = 813); solid tumors with metastasis (N = 259); leukemia/lymphoma (N = 326)
Variables specifically considered in this study were age, gender, length of hospital stay, number of diagnoses, use of drugs and prevalence of adverse drug reactions during hospital stay, and number of hospitalization in the last year. Functional capabilities were ascertained using the ADL scale [13], and categorical variables for dependency in at least 1 ADL or dependency in 3 or more ADLs were considered in the analysis. Cognitive status was evaluated by the Hodgkinson Abbreviated Mental Test (AMT), that is a 10-item version of the Blessed-Roth information-memory-concentration test [14,15], validated in an Italian population for screening for dementia [16]. Each question scores 1 point, and the total score ranges from 0 (no correct answer) to 10 (correct answers). On the weekday after admission, the study physician identified patients to be included in the study and interviewed them on the day before discharge to avoid any interference caused by an acute illness. The cut-off level of 7 (3 or more errors) has been reported to have 100% sensitivity and 71% specificity with respect to the DSM III diagnostic criteria of dementia [16].
We used contingency tables to compare the demographic and clinical characteristics of the groups studied. AMT and ADL scores of patients with lung or gastrointestinal cancer, i.e. of the most frequent cancers in the population studied, were separately analyzed to estimate the effects of metastases on the functional capabilities in homogeneous groups of cancer patients. Logistic regression analysis was used to obtain a deconfounded estimate of the association between the type of disease and physical or cognitive impairment. All analyses were performed using SPSS V10.0 (SPSS Inc., Chicago IL)[17].
Results
The prevalence of patients aged 80 or more was higher in CHF and COPD than in diabetes and cancer groups, while male gender was more frequent in patients with COPD and cancer. Comorbidity was greater in diabetic patients, while patients with leukemia/lymphoma or metastatic cancer had the longest average stay. Use of NSAIDs and analgesics was greater in patients with diabetes and cancer, particularly in those with metastases. The highest rate of hospitalization in the previous year was observed in patients with COPD and cancer (Table 1).
Table 1 Demographic and clinical characteristics of the groups studied.
CHF N = 842 Diabetes N = 939 COPD N = 399 Solid tumors N = 813 Leukemia/Lymphoma N = 326 Metastasis from solid tumors N = 259 P
Age, yrs 0.001
<65 8.4 29.9 16.3 20.9 27.3 30.1
65–79 45.2 45.8 44.6 46.4 46.0 41.7
80+ 46.3 24.3 39.1 32.7 26.7 28.2
Gender (males) 46.0 45.2 63.2 67.7 51.2 61.4 0.001
No of diagnoses>4 35.3 48.3 17.0 29.8 37.7 35.1 0.001
Length of stay>14 days 33.1 36.1 30.1 35.8 41.4 40.5 0.01
ADR during stay 12.1 10.4 6.3 7.4 7.7 10.4 0.002
Drugs during stay
NSAIDs 3.6 10.5 4.8 13.7 15.6 28.6 0.001
Analgesics 0.2 8.0 1.3 3.8 3.1 6.9 0.001
More than 2 hospitalization in the last year 9.4 8.7 12.0 12.2 19.0 15.1 0.001
CHF = congestive heart failure; COPD = chronic obstructive pulmonary disease; ADR = adverse drug reactions; NSAIDs = non-steroidal antinflammatory drugs.
Data are percentage. P values in the last column refer to the 6 groups for 2 levels chi-square test.
Patients with cancer had a lower prevalence of cognitive impairment (22.9% in patients with solid tumors; 19.6% in patients with leukemia/lymphoma; 22.8% in patients with metastatic cancer) compared to patients with CHF (28.0%) or COPD (25.8%). Physical dependency was highly prevalent in patients with metastatic cancer (dependent in at least 1 ADL, 31.3%; dependent in 3 or more ADLs, 27.4%). The corresponding figures for patients with CHF were 24.0% and 19.6%, respectively (Figure 1, panel A).
Figure 1 Panel A: prevalence of cognitive impairment and physical disability in patients divided according to their main diagnosis. Panel B: association between physical disability (dependency in at least 1 ADL) and comorbidity in the groups studied. Panel C and D: prevalence of cognitive impairment and physical disability in patients with gastrointestinal or lung cancer, with or without metastases.
Physical dependency in at least 1 ADL was significantly associated with higher comorbidity in patients with CHF (p < 0.05), diabetes (p < 0.01) or metastatic cancer (p < 0.05), but not in those with COPD, non-metastatic solid tumors or leukemia/lymphoma (figure 1, panel B). No significant association between cognitive performance and comorbidity was observed (data not shown).
For patients with gastrointestinal cancer (figure 1, panel C) and lung cancer (figure 1, panel D), the presence of metastases was associated to a slight increase in the prevalence of physical dysfunction, and to a slight decrease in the prevalence of cognitive impairment. However, these differences were not statistically significant.
When we repeated the analysis in people aged 80 years or more (figure 2), we found that the prevalence of cognitive dysfunction was similar in all the conditions considered, including metastatic cancer. Furthermore, the prevalence of physical disability did not distinguish metastatic cancer from the remaining conditions.
Figure 2 Prevalence of cognitive impairment and physical disability in patients aged 80 years or more (N = 1196) grouped according to their main diagnosis.
The gender-specific limitations of functional capabilities are reported in figure 3. The highest prevalence of cognitive impairment was observed in males with CHF (23.0%) or COPD (24.6%) and in females with CHF (32.3%) or metastatic cancer (32.0%). Metastatic cancer was associated with the highest prevalence of physical dependency in both genders (Figure 3, Panel A and B). When we divided patients according to the number of diseases, CHF and COPD were associated with the greatest prevalence of cognitive impairment (29.2% and 26.9%, respectively) in patients with less than 5 diagnoses, while among patients with more than 4 diagnoses, the greatest prevalence of cognitive impairment was observed in CHF (25.9%) and metastatic cancer (27.5%) groups. Metastatic cancer was associated with the highest prevalence of physical dependency regardless of comorbidity (Figure 3, panel C and D).
Figure 3 Prevalence of cognitive impairment and physical disability in: male patients divided according to their main diagnosis (Panel A); female patients divided according to their main diagnosis (Panel B); patients with less than 5 diagnoses divided according to their main diagnosis (Panel C); patients with more than 4 diagnoses divided according to their main diagnosis (Panel D).
Finally, after simultaneous adjustment for age, gender, number of drugs, number of diagnoses, and length of stay, only metastatic cancer was associated with a greater prevalence of physical but not cognitive impairment with respect to CHF patients. Diabetes was significantly associated with cognitive impairment (Table 2).
Table 2 Summary logistic regression models* of main diagnosis to cognitive impairment or physical dependency in at least 1 ADL.
Cognitive impairment OR (95%CI) Dependency in at least 1 ADL OR (95%CI)
CHF 1.0 (reference) 1.0 (reference)
Diabetes 1.40 (1.11–1.78) 0.93 (0.73–1.19)
COPD 1.13 (0.85–1.51) 0.73 (0.53–1.10)
Solid tumors 1.09 (0.85–1.39) 1.23 (0.95–1.57)
Leukemia/lymphoma 0.99 (0.70–1.38) 1.05 (0.75–1.47)
Metastatic cancer 1.34 (0.94–1.91) 2.09 (1.51–2.90)
*After adjusting for age, gender, number of drugs, number of diagnoses, and length of stay.
Discussion
Our study indicates that physical performance and cognitive status in patients with non-metastatic cancer did not significantly differ from those observed in older hospitalized patients with other non-neoplastic chronic diseases. In presence of metastases, however, physical dependency was more severe, whereas cognitive impairment was significantly more prevalent in CHF than in metastatic cancer patients. Furthermore, compared to patients with CHF, those with metastatic cancer had longer hospital stay, greater number of hospitalization in the last year, and used more anti-inflammatory and analgesic drugs. Thus, our data confirm the common perception of metastatic cancer as a disease dramatically impacting on the health status, but this view should take into account the differential effect of cancer on physical and mental domains. On the other hand, non-metastatic cancer does not outweigh non-neoplastic chronic diseases as a cause of physical and cognitive impairment. It is interesting to note, however, that the presence of metastases had a distinctive impact on health status only in people aged less than 80. This finding may be consistent either with selective survival up to older ages or with a real lack of difference in the effects on health status of non-malignant chronic diseases (especially diabetes and CHF) and metastatic cancer in the very old.
Assessing the health status is relevant to optimize the therapy of cancer in the elderly. On average, the elderly are as likely to benefit from standard cancer treatment as younger people do [4]. Only older patients with functional and cognitive impairment are at higher risk of developing complications in response to aggressive treatments [18]. Thus, there is no sound basis for the common practice of treating the elderly with substandard therapy because of the perceived minimal benefit of chemotherapy and great risk of toxicity [18]. Age bias may affect both physicians' attitudes toward the use of standard anticancer therapeutic regimens in elderly patients [19], and the recruitment of elderly cancer patients in clinical trials [20,21]. Our data show that only one out of three older patients with metastatic cancer has severely impaired physical capabilities, whereas cognitive impairment is less common. Thus, in an unselected elderly population most of patients with metastatic cancer seem to be amenable to standard oncologic therapy, at least on the basis of their physical and cognitive capabilities.
The effect of CHF on health status is the object of a growing number of reports [22,23]. Our study adds to current knowledge by showing that CHF approaches metastatic cancer as a disabling condition, but, compared with metastatic cancer, it impacts more on mental than on physical capabilities. Indeed, cognitive dysfunction is highly prevalent in CHF populations and represents an important health problem, for example by affecting the compliance with therapy [22]. These findings might help understand the positive effects of geriatric assessment and intervention trial in CHF [24,25]. Indeed, physical rehabilitation and strategies enhancing the compliance with drugs and life style measures were a primary component of such trials [24]. Similarly, the association between diabetes and cognitive impairment is well known [26]. Our findings confirm this association and further stresses that metastatic cancer does not primarily affect mental performance.
Limitations of our study deserve to be cited. First, a cross-sectional observation is exploratory in nature, and it should be prospectively replicated. Furthermore, by considering only patients admitted to the acute care hospital, our sample can not be considered fully representative of the general population of older people. Second, the general health status of patients admitted to Geriatric or Internal Medicine units may be different from those admitted to Oncology or other specialty units. We excluded people dying during the hospital stay to avoid the bias introduced in the analysis by people with terminal illness. The GIFA questionnaire, however, does not contain an item on explicit terminal prognosis, therefore we could have excluded people with more advanced, but not terminal, disease. This could have biased our results by inflating the proportion of people with less advanced cancer progression. Nonetheless, by excluding people who died regardless of the diagnosis, we also excluded people with more advanced CHF or COPD, and this is likely to have offset the potential bias. Third, the use of a single cognitive screening test did not allow us to investigate the impact of selected chronic conditions on specific cognitive domains. Finally, comorbidity variously affects functional capabilities and thus might be responsible for some of the differences among groups. However, also in patients with more than 4 diagnoses, only the presence of metastatic cancer was associated with physical but not cognitive impairment. Furthermore, comorbidity usually characterizes patients having these main diseases, which makes present findings representative of the clinical reality.
Conclusion
Cancer should not be considered as an ineluctable cause of severe cognitive and physical impairment, at least not more than other chronic conditions highly prevalent in older people, such as CHF. Further studies should be carried out to explain in which measure the impairment of mental and functional capabilities depends upon cancer per se or cancer related pain or comorbid conditions. Clarifying this issue in the individual patient would improve interventions aimed at reducing the burden of cognitive and physical dysfunction and improving health status in older patients with cancer.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AC, CP and RAI participated in the conception and design of the study, analysis of data, and drafting of the manuscript. LC, FC and BM participated in the collection and analysis of data, and helped to draft the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The Gruppo Italiano di Farmacovigilanza nell'Anziano (GIFA) is a research groupof the Italian Society of Gerontology and Geriatrics (SIGG) – Fondazione Italiana per la Ricerca sull'Invecchiamento (FIRI-ONLUS).
The GIFA is partially supported by a grant from the Italian National Research Council (No 94000402).
A complete list of GIFA investigators has been published previously (Pharmacol Res. 1999; 40: 287-95).
==== Refs
Antonelli Incalzi R Corsonello A Pedone C Corica F Carbonin P Bernabei R Construct validity of activities of daily living scale: a clue to distinguish the disabling effects of Chronic Obstructive Pulmonary Disease and Congestive Heart Failure Chest 2005 127 830 838 15764764 10.1378/chest.127.3.830
Wu JH Haan MN Liang J Ghosh D Gonzalez HM Herman WH Diabetes as a predictor of change in functional status among Mexican Americans: a population-based cohort study Diabetes Care 2003 26 314 319 12547855
Okubadejo AA O'Shea L Jones PW Wedzicha JA Home assessment of activities of daily living in patients with severe chronic obstructive pulmonary disease on long-term oxygen therapy Eur Respir J 1997 10 1572 1575 9230249 10.1183/09031936.97.10071572
Given B Given C Azzouz F Stommel M Physical functioning of elderly cancer patients prior to diagnosis and following initial treatment Nurs Res 2001 50 222 232 11480531 10.1097/00006199-200107000-00006
Balducci L Extermann M Management of cancer in the older person: a practical approach Oncologist 2000 5 224 237 10884501 10.1634/theoncologist.5-3-224
Stuck AE Aronow HU Steiner A Alessi CA Bula CJ Gold MN Yuhas KE Nisenbaum R Rubenstein LZ Beck JC A trial of annual in-home comprehensive geriatric assessments for elderly people living in the community N Engl J Med 1995 333 1184 1189 7565974 10.1056/NEJM199511023331805
Bernabei R Landi F Gambassi G Sgadari A Zuccala G Mor V Rubenstein LZ Carbonin P Randomised trial of impact of model of integrated care and case management for older people living in the community BMJ 1998 316 1348 1351 9563983
Rich MW Beckham V Wittenberg C Leven CL Freedland KE Carney RMA Multidisciplinary intervention to prevent the readmission of elderly patients with congestive heart failure N Engl J Med 1995 333 1190 1195 7565975 10.1056/NEJM199511023331806
Counsell SR Holder CM Liebenauer LL Palmer RM Fortinsky RH Kresevic DM Quinn LM Allen KR Covinsky KE Landefeld CS Effects of a multicomponent intervention on functional outcomes and process of care in hospitalized older patients: a randomized controlled trial of Acute Care for Elders (ACE) in a community hospital J Am Geriatr Soc 2000 48 1572 1581 11129745
Ferrucci L Guralnik JM Cavazzini C Bandinelli S Lauretani F Bartali B Repetto L Longo DL The frailty syndrome: a critical issue in geriatric oncology Crit Rev Oncol Hematol 2003 46 127 137 12711358
Carosella L Pahor M Pedone C Zuccala G Manto A Carbonin P Pharmacosurveillance in hospitalized patients in Italy. Study design of the 'Gruppo Italiano di Farmacovigilanza nell'Anziano' (GIFA) Pharmacol Res 1999 40 287 295 10479475 10.1006/phrs.1999.0508
PHS-HCF International classification of diseases, 9th revision. Clinical modifications Public Health Service – Health Care Financing Administration 1980 Washington, D.C
Katz S Ford AB Moskowitz RW Jackson BA Jaffe MW Studies of illness in aged: the index of ADL, a standardized measure of biological and psychosocial function JAMA 1963 185 94 106
Blessed G Tomlinson BE Roth M The association between quantitative measures of dementia and of senile change in the cerebral grey matter of elderly subjects Br J Psychiatry 1968 114 797 811 5662937
Hodkinson HM Evaluation of a mental test score for assesment of mental impairment in the elderly Age Ageing 1972 1 233 238 4669880
Rocca WA Bonaiuto S Lippi A Luciani P Pistarelli T Grandinetti A Cavarzeran F Amaducci L Validation of the Hodkinson Abbreviated Mental Test as a screening instrument for dementia in an Italian population Neuroepidemiology 1992 11 288 295 1291893
SPSS Release 10.0 software documentation 1999 Chicago: SPSS Inc
Dale DC Poor prognosis in elderly patients with cancer: the role of bias and undertreatment J Support Oncol 2003 1 11 17 15346995
DeMichele A Putt M Zhang Y Glick JH Norman S Older age predicts a decline in adjuvant chemotherapy recommendations for patients with breast carcinoma: evidence from a tertiary care cohort of chemotherapy-eligible patients Cancer 2003 97 2150 2159 12712466 10.1002/cncr.11338
Benson AB 3rdPregler JP Bean JA Rademaker AW Eshler B Anderson K Oncologists' reluctance to accrue patients onto clinical trials: an Illinois Cancer Center study J Clin Oncol 1991 9 2067 2075 1941065
Kemeny MM Peterson BL Kornblith AB Muss HB Wheeler J Levine E Bartlett N Fleming G Cohen HJ Barriers to clinical trial participation by older women with breast cancer J Clin Oncol 2003 21 2268 2275 12805325 10.1200/JCO.2003.09.124
Trojano L Antonelli Incalzi R Acanfora D Picone C Mecocci P Rengo F Cognitive impairment: a key feature of congestive heart failure in the elderly J Neurol 2003 250 1456 1463 14673579 10.1007/s00415-003-0249-3
Zuccala G Pedone C Cesari M Onder G Pahor M Marzetti E Lo Monaco MR Cocchi A Carbonin P Bernabei R The effects of cognitive impairment on mortality among hospitalized patients with heart failure Am J Med 2003 115 97 103 12893394 10.1016/S0002-9343(03)00264-X
Ahmed A Quality and outcomes of heart failure care in older adults: role of multidisciplinary disease-management programs J Am Geriatr Soc 2002 50 1590 1593 12383160 10.1046/j.1532-5415.2002.50418.x
Konstam V Salem D Pouler H Kostis J Gorkin L Shumaker S Mottard I Woods P Konstam MA Yusuf S for the SOLVD Investigators Baseline quality of life as a predictor of mortality and hospitalization in 5,025 patients with congestive heart failure Am J Cardiol 1996 78 890 895 8888661 10.1016/S0002-9149(96)00463-8
Allen KV Frier BM Strachan MW The relationship between type 2 diabetes and cognitive dysfunction: longitudinal studies and their methodological limitations Eur J Pharmacol 2004 490 169 175 15094083 10.1016/j.ejphar.2004.02.054
|
16122389
|
PMC1201137
|
CC BY
|
2021-01-04 16:30:32
|
no
|
BMC Geriatr. 2005 Aug 25; 5:10
|
utf-8
|
BMC Geriatr
| 2,005 |
10.1186/1471-2318-5-10
|
oa_comm
|
==== Front
BMC Health Serv ResBMC Health Services Research1472-6963BioMed Central London 1472-6963-5-531610217310.1186/1472-6963-5-53Research ArticleImplementation of a health care policy: An analysis of barriers and facilitators to practice change Watt Susan [email protected] Wendy [email protected] Paul [email protected] School of Social Work, McMaster University, Hamilton, Ontario, Canada2 School of Nursing, McMaster University, Hamilton, Ontario, Canada3 Department of Clinical Epidemiology & Biostatistics, McMaster University, Hamilton and Senior Research Associate, St. Joseph's Health System Research Network, Brantford, Ontario, Canada2005 15 8 2005 5 53 53 4 2 2005 15 8 2005 Copyright © 2005 Watt et al; licensee BioMed Central Ltd.2005Watt 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
Governments often create policies that rely on implementation by arms length organizations and require practice changes on the part of different segments of the health care system without understanding the differences in and complexities of these agencies. In 2000, in response to publicity about the shortening length of postpartum hospital stay, the Ontario government created a universal program offering up to a 60-hour postpartum stay and a public health follow-up to mothers and newborn infants. The purpose of this paper is to examine how a health policy initiative was implemented in two different parts of a health care system and to analyze the barriers and facilitators to achieving practice change.
Methods
The data reported came from two studies of postpartum health and service use in Ontario Canada. Data were collected from newly delivered mothers who had uncomplicated vaginal deliveries. The study samples were drawn from the same five purposefully selected hospitals for both studies. Questionnaires prior to discharge and structured telephone interviews at 4-weeks post discharge were used to collect data before and after policy implementation. Qualitative data were collected using focus groups with hospital and community-based health care practitioners and administrators at each site.
Results
In both studies, the respondents reflected a population of women who experienced an "average" or non-eventful hospital-based, singleton vaginal delivery. The findings of the second study demonstrated wide variance in implementation of the offer of a 60-hour stay among the sites and focus groups revealed that none of the hospitals acknowledged the 60-hour stay as an official policy. The uptake of the offer of a 60-hour stay was unrelated to the rate of offer. The percentage of women with a hospital stay of less than 25 hours and the number with the guideline that the call be within 48 hours of hospital discharge. Public health telephone contact was high although variable in relation to compliance the guideline that the call be within 48 hours of hospital discharge. Home visits were offered at consistently high rates.
Conclusion
Policy enactment is sometimes inadequate to stimulate practice changes in health care. Policy as a tool for practice change must thoughtfully address the organizational, professional, and social contexts within which the policy is to be implemented. These contexts can either facilitate or block implementation. Our examination of Ontario's universal postpartum program provides an example of differential implementation of a common policy intended to change post-natal care practices that reflects the differential influence of context on implementation.
==== Body
Background
A social policy expresses "ongoing strategies for structuring relationships and coordinating behaviour to achieve collective purposes...ways of exerting power, of getting people to do things that they might otherwise not do" [1]. The implementation of a policy requires that resources come from wherever necessary to enact the relevant program(s) and "that the economic structure, social institutions, and political processes will be shaped to protect and maintain that commitment" [2]. The purpose of this paper is to examine how a health policy initiative was implemented in two different parts of a single payer, government run health care system and to analyze the barriers and facilitators to achieving practice change.
The implementation of the 1999 policy-expansion of the Healthy Babies Healthy Children (HBHC) program, under the aegis of the government in Ontario, Canada, provided an instructive example of a policy intended to drive health care practice. The expansion of this universal program extended to women the option of staying in hospital for 60 hours following an uncomplicated vaginal birth for "assessment, support and follow-up purposes" [3]. The stated goal of the policy was to give "women some flexibility" in length of stay (LOS) in hospital after childbirth and to provide enhanced community-based postpartum services. This policy was in keeping with the 1996 recommendations of The Canadian Pediatric Society and the Society of Obstetricians and Gynecologists of Canada [4].
The policy prescribed a province-wide implementation mechanism for the community care portion of the policy that consisted of a public health initiated telephone call within 48 hours of postpartum discharge and the offer of a home visit to all mothers, regardless of their LOS, demographic profile or residential location in the province [3]. No specific implementation mechanism was prescribed for extending the LOS; hospitals were charged with developing their own implementation plans.
Grol and Grimshaw [5] categorized some of the environmental factors that may become barriers to the implementation of evidence in practice. The universal extension of the HBHC program through two distinctive components of the health care system, with different structural and functional accountabilities, provides an interesting opportunity to examine how contextual factors can act as barriers or facilitators, leading to very different implementation strategies and uptake levels.
These groupings provide a reasonable way of examining barriers to policy implementation at the practice level. The organizational, professional, and social contexts into which a policy is introduced may block or facilitate practice change. What happens at the level of implementation when a government imposes universal changes in practice standards not driven by, or even agreed to by, clinicians? What happens when the implementation of a policy is dependent upon two quite different segments of the health care system – hospitals and public health units – each sector controlled by different professions and operating from related, but different, mandates? What happens when women who are the objects of the policy are neither consulted nor required to be informed in any systematic way about the policy?
Far from being a new phenomenon, postpartum "early release" from hospitals for healthy mothers and newborn infants has been a contentious but familiar theme in both the practice and the politics of Canadian health care in the twentieth century [6]. By the mid twentieth century, in-hospital birth and postpartum care became the norm in North America [6]. This change in location from community to hospital was driven by a concern for the mother's health as much as a concern about infant mortality or morbidity. [7]
The shift from home deliveries to a hospital setting reached a peak in the 1960s [8,9]. Within the first 50 years of the twentieth century, the high and well-known numbers of maternal deaths and injuries pushed physicians to do what they did best – to medicalize pregnancy and childbirth [6]. As with most medical advice, this practice was couched in terms of safety for mothers and infants; [8,10]. Childbirth became firmly lodged within the mandate and practices of the 'sickness care system' [11].
Physicians prescribed 'lying in' for 14 days postpartum, in spite of evidence that suggested shorter hospital stays and getting out of bed sooner were better for the health of women [6,8]. However, by 1949, according to Mitchinson, postpartum stays had decreased to an average of 10 days post delivery. 'Early rising' began to be debated and occasionally practiced during this time, mostly popularized by the post WWII increase in births and concurrent shortage of hospital beds [6].
From the 1970s through the 1990s, postpartum hospital stays shortened. In Canada, Wen reported the mean length of hospital stay after delivery decreased from 5.3 days in 1984–85 to 3.0 days in 1994–95[12]. In response to an expressed wish on the part of women and their partners, the practice of family-centred obstetrics led hospitals to develop early discharge programs to accommodate strong patient desires [8]. By 1999, the average length of stay for a "delivery in a completely normal case" (Code 650) was 1.8 days [13]. These developments mirrored similar changes in practice in the U.S. [14]. Decter reports that, concurrent to this, hospital stays for a variety of procedures and illnesses were becoming shorter as well due in part to emerging technology and changing practice standards and in part to fiscal restraints [15].
Similarly, public health nursing practice changed over time. In the 20th century, the mandate of public health expanded to specifically include improving maternal child health [16]. Public health continued to place emphasis on the health and well being of mothers and infants in many ways (e.g., 1998 Ontario Healthy Babies, Healthy Children Program). This program is a prevention and early intervention initiative intended to provide support and services to families with children from before birth up to six years of age that could benefit from additional resources. At the time the universal postpartum program was introduced, the HBHC program was already in place, with public health nurses engaged in calling and providing home visits for mothers and infants identified through in-hospital screening [17].
The 1990s were troubled years for the Canadian healthcare system and, consequently, for the governments which managed provincial operations [18,19]. Increasingly, hospitals cared only for critically ill patients; convalescence occurred in the community with or without the support of community-based health care providers. At the same time, public health units in Ontario stopped routine visits to newborn infants and their mothers. By early 1998, an Ipsos-Reid poll found that 73% of Canadians believed that the healthcare system was worse than it had been 5 years previously [20]. Consumers and potential consumers of medical care services were expressing the fear that reductions in in-hospital care were compromising patient well-being.
These concerns extended to postpartum and newborn care. In Canada, in 1997, there occurred what has been defined as a "focussing event" – something that draws widespread attention and publicity to an existing problem [21,22]. The accidental death of a newborn infant from dehydration led to a Coroner's inquest that raised questions about a connection between the death of a seemingly healthy infant and maternity short stay hospital policies [23].
The political lessons from this event were not lost on a Canadian government with a health care system under criticism. By 1998, the Government of Canada had created Family-Centred Maternity and Newborn Care: National Guidelines [24]. The guidelines made clear that, while early discharge programs had been judged safe, satisfying, efficient and economical for their users, the success of these programs rested on the following factors: parental choice, post-natal screening, community support components, and appropriately trained professional staff [24]. They go on to say that where "administrative mandates may give rise to a non-voluntary, short hospital stay", this must be coupled with community support strategies and a program of community follow-up care [[24] 6.36]. It is very clearly stated that "the mother...should decide the length of hospital stay, based on her individual needs" [[24] 6.36].
Facing an election call the Conservative Government of Ontario announced in their 1999–2000 business plan [25] their intention to give additional funding to hospitals for extending the stay of postpartum women and to the HBHC program for a postpartum telephone call and home visit [3]. This universal policy had the advantage of being potentially popular, apparently caring, medically harmless, relatively inexpensive and appealing to an already concerned public, media and professional community, even though evidence of efficacy or system capacity to implement the policy was not clear.
The Ontario Mother and Infant Survey (TOMIS) of mother and infant health and service utilization was completed just prior to the introduction of this policy. A replication study, TOMIS II, provided an opportunity to examine the outcomes of this policy and to search for the factors that shaped the uptake of the policy by providers and consumers.
Methods
The context of our examination of the implementation of this policy was two research studies, one initiated before and one after the HBHC policy enhancement. The primary methodology used in both studies was a cross-sectional survey.
Data collection for TOMIS occurred between November 1998 and June 1999 just prior to the Hospital Stay and Postpartum Home Visiting Program extension to the HBHC program in November 1999. This provided us with baseline data to compare with findings from TOMIS II. These data were collected from September 2001 to June 2002.
The survey methods and instruments used for TOMIS II paralleled those used in TOMIS, which allowed for an appropriate comparison of data at two points in time. The same study sites, sample size, eligibility criteria, recruitment strategy, and instruments were used for the two surveys[13,14]. In both studies, women completed a questionnaire before discharge from hospital and participated in a structured telephone survey at 4-weeks post-discharge.
Five purposefully selected Ontario hospitals provided respondents who constituted a cross-section of mothers and newborn infants with diverse socio-economic characteristics and access to varying health and social services. The characteristics of the hospitals are presented in Table 5
Table 5 Site 1 Southern, suburban, teaching hospital, metropolitan catchment area, 3900 annual births
Site 2 Central east regional centre, urban & rural catchment areas, 1500 annual births
Site 3 Central south regional centre, urban & rural catchment areas, 4500 annual births
Site 4 Southern, urban, teaching, metropolitan catchment area, 2700 annual births
Site 5 Central north regional centre, urban & rural catchment areas, 2000 annual births.
Participants for both studies included the first 250 eligible, consenting women from each site, totaling 1,250 participants in each study. This sample size was determined to be large enough to allow for the examination of many variables together, and was in keeping with the generally accepted guideline of 30 subjects per variable[15]. Women were eligible if they (a) had given birth vaginally to a single live infant, (b) were being discharged from hospital at the same time as their infant, (c) were assuming care of their infant at the time of discharge, and (d) were competent to give consent to participate. Women were excluded if they (a) had an infant who required admission to a neonatal intensive care or special care nursery for more than 24 hours or (b) were unable to communicate in one of the study languages – English, French, Chinese, and Spanish. Each study hospital continued to utilize its own postpartum care protocols throughout the recruitment period. Participants received services from the public health units related to their residence, which by and large, were those located in the same geographic region as the hospital sites. A full description of the methodology has been previously published[14]. The ethics review committees of McMaster University and each of the hospitals involved in the study granted ethical approval.
Descriptive statistics were computed by site for all variables measured, including frequency counts and percentages. Chi-square tests were used to determine differences between sites or differences between TOMIS and TOMIS II data. A probability level of <0.05 was used to determine statistical significance. SPSS was used for all statistical computations.
In both of the studies, following preliminary analysis of the survey findings, focus groups were held at each of the sites. These groups were comprised of front-line clinical and administrative staff from each hospital and community agencies. TOMIS II focus group participants were asked to reflect not only on the findings in the context of local practices and policies, but also on local implementation of the universal Hospital Stay and Postpartum Home Visiting Program. They were asked to comment specifically on the extent to which the program had been implemented in their community and implementation challenges. It therefore is only the TOMIS II focus group findings that are relevant to this paper.
One TOMIS II site was unable to participate in the focus groups. At three of the four sites that did participate, we were able to hold a focus group for only community-based providers and managers, and a second focus group for only hospital-based participants. Due to planning issues, the two focus groups at the fourth site were a combination of community and hospital personnel. The size of the focus groups ranged from 8 to 12 individuals. Each group interview was about an hour long, and was audio taped and later transcribed verbatim.
In TOMIS II, focus group participants, in addition to commenting on the survey findings, were asked to describe local implementation of the universal Hospital Stay and Postpartum Home Visiting Program. They were asked to comment specifically on the extent to which the program had been implemented in their institution or community and on the implementation challenges that they had experienced.
Focus group data were analyzed using an inductive approach. Two research assistants independently coded the transcripts, with phrases and sentences that described specific aspects of program implementation being given a descriptive code. The research assistants then met and reached consensus on a coding scheme that resulted in the assignment of a common code to data that were similar. The emergent themes were reviewed and validated with one of the research team members.
Results
Information providers
In both TOMIS and TOMIS II there were no statistically significant differences in the sociodemographic characteristics of those recruited in hospital and those who completed the telephone interviews 4-weeks post discharge [26]. No anomalous results in relation to these characteristics were found when comparisons were made to the statistical profiles developed from Statistics Canada data about women ages 15 to 45 in each of the communities served by each hospital. Infants born to study participants were full-term and of normal birth weight [26]. Further, the focus groups endorsed the representativeness of the sample of the population of clients at each site. Therefore, the authors are confident that the respondent groups in both studies reflect a population of women who experienced an "average" or non-eventful hospital-based, vaginal, singleton delivery.
Focus groups were held at each site except Site 4. Participants in TOMIS II focus groups included nursing and physician administrators from obstetrical units in the site hospital, front-line nurses, lactation consultants, clinical educators, social workers, midwives, and public health nurses and nursing administrators. A total of approximately 80 people participated in these focus groups.
Length of stay
In TOMIS II, the findings shown in Table 1 demonstrate wide variance in the implementation of the offer of 60-hour stay among the sites. It is possible that characteristics of mothers and/or newborn infants could explain this variance. Although the offer was associated with a younger maternal age, first live birth, self-identified ethnicity as Canadian, English or French as 1st language, and maternal place of birth as Canada, these sociodemographic characteristics failed to explain the site variability [33]. Explanation was sought, therefore, in the sites' implementation of the policy.
Table 1 Offer and Acceptance of 60-hour Length of Stay
Site 1
No. (%) Site 2
No. (%) Site 3
No. (%) Site 4
No. (%) Site 5
No. (%)
Offered a 60-hr stay a b 20 (11.7) 78 (41.9) 168 (81.2) 69 (39.9) 80 (52.3)
Accepted a 60-hr stayc 4 (21.1) 28 (39.4) 51 (30.4) 21 (31.3) 17 (21.3)
a Chi-square test indicated a statistically significant difference (P < 0.001) across sites for offer of a 60-hr stay
b Offer is reported for those who took part in the scheduled telephone interview at 4 weeks post-discharge (n = 890)
c Acceptance is reported for those offered a 60-hr stay (n = 405)
It became quite clear in the focus groups that hospital participants did not view the policy of offering a 60-hour stay as prescriptive. They talked about the policy as difficult to implement given the lack of beds in their hospitals and some were convinced that, given an option, women would choose to stay for longer than medically necessary.
Representatives of the hospital frequently expressed surprise that mothers even knew about the policy and at the rates at which women reported being offered a 60-hour stay. They disavowed any responsibility on the part of the hospital for informing women about the LOS aspect of the policy. Some focus group participants were sure that, rather than being offered the 60-hour stay, women were "demanding 60 hours" because "we don't volunteer that information". Interestingly, focus group members sometimes attributed acknowledgement of an offer by a mother as "recall error" because "our policy is 2 days and out".
Focus group participants from the two sites with significantly higher offer rates (Sites 3 and 5) had a slightly different perspective when interpreting the results from their hospitals in comparison with other sites. They stated that they knew about the policy and while not necessarily agreeing with it, believed that they had an obligation to their patients to adhere to the policy. In neither site was information about the policy provided to women in a standard format; it was not part of the nursing admission protocol or available in writing to mothers. However, prenatal class teachers in both geographic areas were reported to have been telling expectant women in a consistent fashion that they were entitled to up to a 60-hour stay. Family physicians also were reported as a source of policy information for women. In both of these sites, nurses believed that women whom they viewed as being "at high risk" (e.g., teenage mothers, 1st time breastfeeding mothers, mothers living at a significant distance from the hospital) received information about the 60-hour option from postpartum hospital staff but were unable to explain the basis of their belief in any formal information provision. One representative who said that women "heard about it somehow – if not from us directly then from other women in the unit" expressed the view of focus group participants from several sites.
The uptake of the offer of up to a 60-hour stay was between 21 and 39% (Table 1) and was unrelated to the rate of offer. Sociodemographic characteristics did not differentiate between those who took up the offer and those who did not. However, uptake was associated with first live birth, infant health problems, maternal health problems, and the mother having two or more concerns related to herself or her infant (See Sword, Watt, Krueger, 2004, [26] for more details). In short, mothers who felt less sure of their own ability to care for themselves and their newborn infant were more likely to stay longer in hospital. It is interesting to note that one of these sites had decided, following the completion of data collection, to no longer universally offer a 60-hour stay.
If the intent of this policy was to produce fewer "drive through" deliveries, then the appropriate question may not be "Were you offered/did you accept a 60-hour stay?" Perhaps the more important issue is, did the policy result in increased LOS overall, and particularly in a decrease in stay of less than 25 and 48 hours?
Table 2 shows that the percentage of women with a stay of less than 25 hours declined in all sites following policy implementation. Similarly, it demonstrates that there were fewer women with a stay of 48 hours or less. The result is that there was a shift to marginally longer lengths of stays for the two groups staying in hospital for the shortest periods.
Table 2 Length of Stay a
LOS <25 hours LOS ≤48 hours LOS >48 hours
Site T1 % T2 % Change % p-value T1 % T2 % Change % p-value T1 % T2 % Change % p-value
Site 1 59.1 42.7 -16.4 0.005 98.7 91.8 -6.9 0.011 0.0 6.4 6.4 0.005
Site 2 11.0 9.7 -1.3 0.81 78.5 67.8 -10.7 0.028 15.5 15.1 -0.4 0.97
Site 3 32.5 12.6 -19.9 <0.001 91.8 62.8 -9.0 <0.001 4.3 23.2 18.9 <0.001
Site 4 45.3 25.9 -19.4 <0.001 94.2 80.4 -13.9 0.001 0.7 9.8 9.1 0.002
Site 5 23.6 13.1 -10.5 0.031 64.2 62.0 -2.2 0.79 14.5 10.5 -4.0 0.40
a Chi-square tests were used to determine whether statistically significant differences existed between TI and T2
It appears that a shift in LOS occurred following the policy initiation. This change marginally increased the most common LOS to the 25 to 48 hour period by decreasing the incidence of discharge in less than 25 hours. Practice changed to become more in line with professional recommendations.
Regardless of whether or not they were offered or accepted up to a 60-hour stay, and unrelated to their actual LOS, most women stated that at the time of discharge they were ready to leave hospital. Similarly, 4 weeks after discharge they continued to see their stay as appropriate for their needs (Table 3 . Despite different approaches to deciding on an appropriate LOS at each site, site was not statistically associated with maternal readiness for discharge or satisfaction with their LOS. On the other hand, having a choice about LOS was associated with maternal satisfaction (p < 0.01).
Public health initiated contact
The second aspect of the expansion of the HBHC program provided for contact with public health initially through a telephone call within 48 hours of discharge from hospital. During this phone call, a home visit to the mother and newborn infant by a public health nurse was to be offered. Women in TOMIS II reported very high rates of public health telephone contact (Table 4). There were significant drops in those rates when asked if the phone call had come within 48 hours of discharge.
Table 4 Public Health Initiated Contact a b
Site 1
No. (%) Site 2
No. (%) Site 3
No. (%) Site 4
No. (%) Site 5
No. (%) p-value
Telephone call anytime after discharge 150
(88.8) 178
(97.8) 180
(87.8) 136
(81.4) 143
(94.7) p < 0.001
Telephone call within 48 hrs of discharge 125
(74.0) 135
(75.0) 131
(64.2) 119
(71.7) 120
(80.0) p = 0.017
Home visit offered 143
(95.3) 161
(91.5) 169
(96.6) 129
(95.6) 135
(94.4) p = 0.276
Home visit accepted c 109
(76.2) 72
(44.7) 69
(40.8) 93
(72.1) 89
(65.9) p < 0.001
a Chi-square tests were used to determine whether statistically significant differences existed between sites.
b N = 890
c Acceptance is reported for those offered a home visit
Focus group participants explained that lack of staff on weekends and problems with the transfer of information from hospitals to health units accounted for delays in phoning women. Home visits were offered to virtually all mothers but were accepted at highly variable rates ranging from 40.8% to 76.2% [26]. Focus group participants unanimously endorsed the policy initiative seeing it as a positive move for both their organizations and for women and infants. They acknowledged receipt of additional resources to implement the policy and were clear about the factors that inhibited complete compliance with the published service standards. Most often participants stated that a lack of resources was the primary reason accounting for low visitation rates. (For further discussion, see Sword, Watt, Krueger, 2004 [26].)
Discussion
The policy addressed in this paper was intended by the government to be universally implemented. Its two parts – the offer of an option for mothers and infants to remain in hospital for up to 60 hours and public health contact through a telephone call within 48 hours of discharge and the offer of a home visit – were implemented at strikingly different rates in different locations by two sectors of a publicly funded, universal health care system.
In the hospital sector, the major barrier which appears to have influenced implementation is that of organizational context. Providers at two of the four hospital sites did not view extension of maternal LOS as a requirement of service delivery. Health care providers and administrators from these sites told us that there simply were not enough beds to allow for longer stays. In fact, since no extra operational funding came specifically to hospitals for this initiative, there was a financial disincentive to keep mothers longer. The dissenting sites reluctantly viewed the policy as an organizational requirement and consequently have high offer rates.
The downsizing of hospitals in general, and maternity wards in particular, were seen by focus group participants as having created a scarcity of beds to accommodate any increased LOS. They had anticipated high rates of acceptance of the offer and were convinced that resources were not available to implement the policy. Despite this reluctance hospitals lengthened their shortest stays suggesting that other factors were in play that modified organizational outcomes.
The general lengthening of stays is an important result because there is no evidence suggesting a longer LOS is universally beneficial. Although some studies have found a relationship between a shorter LOS and newborn infant readmission to hospital [27-29], others have not supported this relationship [9,30,31]. In sum, professional juries are still out on the optimal postpartum LOS for women who have uncomplicated deliveries.
The professional context provides some explanation for the implementation patterns. Some providers did not see themselves as responsible for informing women of the 60-hour stay option and, in fact, were surprised that women had knowledge of the policy and would exercise this option. They reflected a view of LOS as a clinical decision and one that is most appropriately made by mothers and practitioners rather than by policy makers. Providers were quick to point out that if they assessed a mother as being in need of more time in hospital, then she stayed longer. They saw part of their role as one of professional advocacy for patient services when they believed it necessary. Applying the policy would have meant giving up this discretionary professional activity. The association of longer stay with women who fall into traditional "high risk categories" suggests that discretionary offers were still being made and that the professional context modified organizational context in terms of implementation.
The dissonance created in hospital-based health care providers by this policy was evident when they talked about implementing the LOS policy. Their behaviours, including deliberately not telling women about the policy and actively discouraging extended stays, suggests the presence of barriers to implementation based on accepted standards of practice. For example, physicians remained in control of discharge orders and were reported by nursing focus group participants as unwilling to change their traditional practices. Lack of physician support for increasing LOS also accounted for the inability of nursing staff to see this policy as viable in their hospitals. Implementation of a change in-hospital health care policy relies on physician cooperation; they remain the final arbiters of most practice policy, including LOS.
What seemed to be at issue was who should make the decision and on what basis. Great concern was expressed by hospital providers that if women were given the option of staying longer in hospital, most would do so. That impression was not upheld by the responses of women in the study. Generally, women were eager to be discharged from hospital when they felt well and when they perceived their infants to be well. Even when offered an extended stay, most did not accept and whether or not they were offered a longer stay, they found their LOS to have met their needs.
Women's approaches to determining how long to stay in hospital provide one measure of the social context, consumer expectations. In this instance, perceptions of patient knowledge about the policy and expectation regarding LOS was influential in supporting implementation even in the face of organizational and professional contexts which erected strong implementation barriers.
These findings suggest that if implementation is to be successful, all the players need to be included. Consumers need to know about health policies and, if not involved in the formulation of the policy, at least systematically informed of their health care options. The provision of choice and active decision making appears to promote consumer satisfaction, at least in the instance of satisfaction with postpartum LOS.
On the other hand, the section of the policy extension that affected public health practice was implemented consistently and at a high rate at all sites. The strengthening of the community follow-up for mothers and newborn infants by public health nurses had a clearly identifiable proponent and implementation strategy. Direct funding flowed to health units specifically targeted for the provision of these additional services. Public health units received additional resources to provide a service that they had long wanted to offer in their communities, were trained to provide, and believed in professionally.
According to the Ministry responsible, this service expansion did not come at a cost to other public health services although some focus group participants did not share this view. We can only speculate about what might have happened if more women had accepted the offer of a home visit thereby placing more demands on the finite resources of public health units. However, the policy did not establish a specific goal of doing home visits, but rather of offering home visits. In this regard, public health units met the offer-target at all of the sites.
Direct accountability for implementation was assigned by the policy to public health care professionals who believed in the initiative and had been advocates for its adoption. In practice, the lack of weekend staff which public health units claimed occurred because of inadequate funding compromised the ability of providers to attain the specified timeframe for telephone contact, but did not lower the overall rate of contact or appear to interfere with the offer of home visits. Unlike hospitals that had been working with a goal of reducing LOS and fewer resources, health units found the policy to be consonant with their overall mission and feasible with the additional provincially provided resources. Also, unlike extending LOS, the approval of individual physician providers was not needed to implement this aspect of the policy. In addition to adequate funding, only the support of public health practitioners, who already advocated for this approach, was required for implementation. In short, the policy was consonant with the organizational, professional, and social contexts of public health practice and came with adequate resources to meet professionally endorsed implementation targets with strong historic roots.
In this instance women as service consumers again responded variably. Uptake of the offer of a home visit appear to have reflected women's perceptions of their own needs and a home visit by public health nurses as an appropriate way to meet those needs with a range of variables at play in determining these perceptions [26].
Conclusion
Policy implementation in any health care system relies upon provider commitment. Policies that do not address the organizational, professional and social contexts are unlikely to achieve successful implementation. Political objectives alone, however well intentioned, are inadequate to change practice. When barriers to policy implementation exist in any of these contexts, the policy may fail to meet its objectives.
The common goal of positive health outcomes is shared by providers, consumers, and policy makers. Policy makers in any system must respect the knowledge and experience of providers when developing policies that require practice change. Providers need to appreciate and endorse changes in practice, to be "on board" with at least the intent of the policy; they need to value, support, and act on any policy entitlement. Consumers need to be informed and prepared to hold both providers and policy makers accountable in the making and implementing of health policy. A consumer right, in the absence of provider responsibility and accountability, appears to lead to the implementation of only those aspects of a health policy with which providers agree and for which there are perceived to be adequate resources.
Often in health care, providers are cast in the role of service gatekeepers. When policies increase consumer entitlement, they also challenge the authority of the gatekeeper role. Therefore, any universal policy that challenges this role needs both policy makers and practitioners to view it not as a suggestion but rather as a requirement. There must be consequences for failure to comply with the policy. At the same time, providers must be convinced that the policy can be implemented and that the outcome will be positive. If implementation is to be successful, policy makers need to engage providers in the process of policy development by acknowledging and entering into the contexts of the providers.
In the case of the LOS policy examined in this paper, providers were not obligated to action. The policy was merely a statement in principle, leaving action largely on the shoulders of postpartum women themselves. The policy was permissive rather than prescriptive. It relied on hospitals and healthcare providers to act often against the pressures of organizational, professional, and social contexts such as shorter stays, prevailing practice trends, and patient expectations. In such circumstances it seems unlikely that the policy will be implemented.
It would appear that policy statements, no matter how convincing, cannot be assumed to change health care practice. Other facilitating and inhibiting factors must be addressed if policy is to be used as a tool to change practice. Policy makers need to carefully consider not only the intent and objectives of a policy, and the evidence for and against alternative approaches, but also the contextual barriers faced by policy implementers.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SW had a major role in designing the study and writing the proposal, co-supervised all aspects of study implementation, participated in data analysis, and was the lead writer of this manuscript.
WS had a major role in designing the study and writing the proposal, co-supervised all aspects of study implementation, participated in data analysis, contributed to the manuscript, and provided editorial comments.
PK contributed to the study design, implementation, analysis and interpretation, as well as the writing of the manuscript.
Table 3 Mother's Readiness for Discharge
In hospital 4 weeks post discharge
N (1237) % N (888) %
Yes (Definitely & probably) 1055 85.3 786 88.5
No (Not sure, Definitely & probably) 182 14.7 102 11.5
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The Canadian Health Services Research Foundation funded the Ontario Mother and Infant Survey (TOMIS). The Canadian Institutes of Health Research funded TOMIS II.
==== Refs
Stone D Policy Paradox: The Art of Political Decision Making 1997 New York: W.W. Norton & Company, Inc;
Lightman E Social Policy in Canada 2003 Toronto, Oxford Press;
Ontario Ministry of Health, OHIP Bulletin, 10004 September 2, 1999
Canadian Paediatric Society Facilitating discharge home following a normal term birth. A joint statement with the Society of Obstetricians and Gynaecologists of Canada J Paediatr Child Health 1996 1 162 168
Grol R Grimshaw J From best evidence to best practice: effective implementation of change in patients' care Lancet 2003 362 1225 30 14568747 10.1016/S0140-6736(03)14546-1
Mitchinson W Giving Birth in Canada – 1900–1950 2002 Toronto: University of Toronto Press
Fleming G Maternal Mortality and Postnatal Care CMAJ 1933 29 159 as cited in Mitchinson W: Giving Birth in Canada – 1900–1950. Toronto: University of Toronto Press; 2002
Brumfield CG Early Postpartum Discharge Clin Obstet Gynecol 1998 41 611 625 9742358 10.1097/00003081-199809000-00016
Liu S Wen SW McMillan D Trouton K Fowler D McCourt C Increased neonatal readmission rate associated with decreased length of hospital stay at birth in Canada Can J Public Health 2000 91 46 50 10765585
De Vries R Benoit C Teijlingen ER Wrede S Birth by Design Pregnancy, Maternity Care, and Midwifery in North America and Europe 2001 New York: Routledge
Hancock T The mandala of health: a model of the human ecosystem Fam Community Health 1985 8 1 10 10274086
Wen SW Liu S Maroux S Fowler D Trends and variations in length of hospital stay for childbirth in Canada CMAJ 1998 158 875 80 9559012
Canadian Institute for Health Information LOS data for newborns
US Fact Sheet – U.S. Department of Labor, Employee Benefits Security Administration, 2004 (accessed March 14, 2004)
Decter MB Healing Medicare 1994 Toronto: McGilligan Books
Duncan SM Leipert BD Mill JE "Nurses as health evangelists"?: The evolution of public health nursing in Canada, 1918–1939 ANS 1999 22 40 52
Ontario Healthy Babies Healthy Children Toronto, 2005
Rachlis M Kushner C Strong Medicine: How to Save Canada's Health Care System 1994 Toronto: Harper Collins
Editorial Doctors who go on strike Wall Str J June 17, 1998
Ipsos-Reid Healthcare and education are the number one issues in Ontario 1998
Declercq E Simms D The Politics of Drive-through Deliveries Milbank Q 1997 75 175 202 9184681 10.1111/1468-0009.00051
Kingdon JW Agendas, alternatives, and public policies 2003 New York: Longman
The College of Physicians and Surgeons of Ontario A Coroner's Case Members Dialogue 1997 5 8 14
Canada Family-Centered Maternity and Newborn Care-National Guidelines 1999 Health Canada, Division of Childhood and Adolescence
Harder J NOW, where were we ...? June 3, 1999
Sword W Watt S Krueger P Implementation, Uptake, and Impact of a Provincial Postpartum Program Can J Nurs Res 2004 36 60 82 15369165
Lee KS Perlman M Ballantyne M Elliot I To T Association between duration of neonatal hospital stay and readmission rate J Pediatr 1995 127 758 766 7472833
Lock M Ray JG Higher neonatal morbidity after routine early hospital discharge: Are we sending newborns home too early? CMAJ 1999 161 249 253 10463045
Malkin JD Border MS Keeler E Do longer postpartum stays reduce newborn readmissions? Analysis using instrumental variables Health Serv Res 2000 35 1071 1091 11130811
Sword W Watt S Krueger P Lee KS Sheehan D Roberts J Gafni A Understanding newborn infant readmission: Findings of The Ontario Mother and Infant Survey Can J Public Health 2001 92 196 200 11496629
Yanicki S Hasselback P Sandilands M Jensen-Ross C The safety of Canadian early discharge guidelines. Effects of discharge timing on readmission in the first year post-discharge and exclusive breastfeeding to four months Can J Public Health 2002 93 26 30 11925696
Ciliska D Hayward S Thomas H Mitchell A Dobbins M Underwood J Rafael A Martin E A systematic overview of the effectiveness of home visiting as a delivery strategy for public health nursing interventions Can J Public Health 1996 87 193 198 8771925
|
16102173
|
PMC1201138
|
CC BY
|
2021-01-04 16:31:52
|
no
|
BMC Health Serv Res. 2005 Aug 15; 5:53
|
utf-8
|
BMC Health Serv Res
| 2,005 |
10.1186/1472-6963-5-53
|
oa_comm
|
==== Front
BMC Health Serv ResBMC Health Services Research1472-6963BioMed Central London 1472-6963-5-551610916010.1186/1472-6963-5-55Research ArticleThe effect of a monetary incentive on return of a postal health and development questionnaire: a randomised trial [ISRCTN53994660] Kenyon Sara [email protected] Katie [email protected] David [email protected] David [email protected] Alison [email protected] Neil [email protected] Peter [email protected] Reproductive Sciences Section, Department of Cancer Studies and Molecular Medicine, University of Leicester, Leicester, UK2 Department of Epidemiology and Public Health, University of Leicester, Leicester, UK3 Paediatrics Department, Great Ormond Street Hospital for Sick Children, London, UK4 Department of Child Health, University of Nottingham, Nottingham, UK5 National Perinatal Epidemiology Unit, Oxford, UK2005 18 8 2005 5 55 55 16 3 2005 18 8 2005 Copyright © 2005 Kenyon et al; licensee BioMed Central Ltd.2005Kenyon 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
Postal questionnaires are widely used to collect data in healthcare research but a poor response rate may reduce the validity and reliability of results. There was a lack of evidence available relating to use of a monetary incentive to improve the response rate in the healthcare setting.
Methods
The MRC ORACLE Children Study is assessing the health and development of nearly 9000 seven year old children whose mothers' joined the MRC ORACLE Trial. We carried out a randomised controlled trial of inclusion of monetary incentive (five pound voucher redeemable at many high street stores) with the reminder questionnaire to parents. This trial took place between April 2002 and November 2003. When the parents were sent the reminder questionnaire about their child's health and development they were randomly assigned by concealed computer-generated allocation stratified by week of birthday to receive a five pound voucher or no incentive. The population were 722 non-responders to the initial mailing of a 12-page questionnaire. Main outcome measures: Difference in response rate between the two groups.
Results
Inclusion of the voucher with the reminder questionnaire resulted in a 11.7%(95% CI 4.7% to 18.6%) improvement in the response rate between the two groups.
Conclusion
This improvement in response rate and hence the validity and reliability of results obtained appears to be justified ethically and financially.
==== Body
Background
Postal questionnaires are widely used to collect data in health research, but a poor response rate may reduce the validity and reliability of results. In a systematic review of randomised controlled trials of strategies to improve the response rate to postal questionnaires[1], a monetary reward had a significant effect on response. However, caution was attached to the interpretation of the findings in this review [2]. On further examination of the updated review 20% of the participants included in the analysis of the effect of inclusion of a monetary incentive in the final response came from healthcare settings [3] and none of the studies evaluated the use of monetary incentives for a postal questionnaire to collect data from a follow-up of a clinical trial. To evaluate the impact of such an intervention on response rate in such a setting we undertook a randomised trial.
The MRC ORACLE Children Study (MOCS) is following up nearly 9000 seven-year-old children whose mothers joined the MRC ORACLE Trial [4,5] which evaluated the use of antibiotics to improve neonatal outcome after preterm labour or preterm rupture of the membranes. This trial of a monetary reward to enhance response to a postal questionnaire was undertaken between April 2002 – November 2003. Research Ethics Committee approval was obtained from the West Midlands Multicentre Research Ethics Committee
The questionnaire itself is 12 pages long and A4 in size. It contains questions relating to the child's health and development using a mixture of validated tools and questions pertinent to the study. In designing the study we implemented many of the strategies believed to influence response rates to postal questionnaires [1,3]. The questionnaire itself is set out in a user friendly way and we have used coloured ink in its design. The letters accompanying the questionnaire are individualised to the child concerned and the parents are warned the questionnaire will be sent to them. The University in which MOCS is housed franks the envelope, the return envelope is stamped, and reminder letters include a questionnaire.
Methods
When a child in MOCS is seven years old the parents receive an information leaflet about the follow-up Study, and two weeks later a questionnaire about their child's health and development. Contact with parents has already been established prior to this. If no response is obtained the child's General Practitioner is contacted to check the child's address and ensure that contact would be appropriate. Six weeks after the first questionnaire, a reminder one is sent to those who have not responded. At this point the parents were randomly assigned by computer-generated allocation to receive a five-pound voucher (redeemable at many high street shops) with their mailed questionnaire or not (see Figure 1).
Figure 1 Consort flow diagram.
The sample size was predefined by the numbers not responding at this stage of the Study (i.e. approx 700). This yielded 80% power to detect an increased response from 10% to 18% or from 15% to 24% (at the 5% significance level).
Results
Balance between voucher/not voucher groups on the main baseline covariates was good. 722 consecutive parents were randomly allocated to receive a voucher or not with the reminder questionnaire (see Table 1).
Table 1 Results of random allocation of voucher or not on response rate
Voucher No voucher
Questionnaire returned 156 (42.3%) 108 (30.6%)
Questionnaire not returned 213 245
Total 369 353
Inclusion of the voucher with the second questionnaire resulted in a 11.7% (95% CI 4.7% to 18.6%) improvement in the response rate between the two groups (χ2 = 10.6, P = 0.001).
Discussion
The inclusion of a five-pound voucher improved the proportion of questionnaires returned. MOCS will be completed in 2008 and a voucher is being sent to all parents with the reminder questionnaire. It is estimated that this will improve the response rate by 3% over the whole study, at a cost of £67 per additional questionnaire returned. This was calculated on the basis that vouchers will be sent to approximately 40% of parents (2842) and an additional 3% will return the questionnaire
This improvement in response rate, and hence of the validity and the reliability of results appears to be justified ethically and financially. This is particularly relevant in the follow up of children as there is some evidence [6] of raised levels of adverse outcomes in difficult to follow-up children.
Competing interests
The author(s) declare that they have no competing interests.
Contributors
SK and PB designed the trial. DJ and KP carried out the analysis. NM, AS, DT contributed to conception, design and interpretation. All authors read and approved the trial manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Ann Blackburn, Kate Taylor, Gill Grummitt.
Funding: UK Medical Research Council. Grant number 53994660
==== Refs
Edwards P Roberts I Clarke M Di Guiseppi C Pratap S Wentz R Increasing response rates to postal questionnaire: systemic review BMJ 2002 324 1183 1185 12016181 10.1136/bmj.324.7347.1183
Smeeth L Fletcher AE Editorial Improving response rates to questionnaires BMJ 2002 324 1168 1169 12016167 10.1136/bmj.324.7347.1168
Edwards P Roberts I Clarke M DiGuiseppi C Pratap S Wentz R Kwan I Cooper R Methods to increase response rates to postal questionnaires The Cochrane Database of Methodology Reviews 2003 Issue 4.Art No.:MR000008.pub2.DOI: 10.1002/14651858.MR000008.pub2
Kenyon SL Taylor DJ Tarnow-Mordi W ORACLE Collaborative Group Broad-spectrum antibiotics for preterm, prelabour rupture of fetal membranes: the ORACLE I randomised trial Lancet 2001 357 979 988 11293640 10.1016/S0140-6736(00)04233-1
Kenyon SL Taylor DJ Tarnow-Mordi W ORACLE Collaborative Group Broad-spectrum antibiotics for spontaneous preterm labour: the ORACLE II randomised trial Lancet 2001 357 989 994 11293641 10.1016/S0140-6736(00)04234-3
Tin W Fritz S Wariyar U Hey E Outcome of very preterm birth children reviewed with ease at 2 years differ from those followed up with difficulty Arch Dis Child Fetal Neonatal Ed 1998 79 F83 87 9828731
|
16109160
|
PMC1201139
|
CC BY
|
2021-01-04 16:31:52
|
no
|
BMC Health Serv Res. 2005 Aug 18; 5:55
|
utf-8
|
BMC Health Serv Res
| 2,005 |
10.1186/1472-6963-5-55
|
oa_comm
|
==== Front
BMC Health Serv ResBMC Health Services Research1472-6963BioMed Central London 1472-6963-5-551610916010.1186/1472-6963-5-55Research ArticleThe effect of a monetary incentive on return of a postal health and development questionnaire: a randomised trial [ISRCTN53994660] Kenyon Sara [email protected] Katie [email protected] David [email protected] David [email protected] Alison [email protected] Neil [email protected] Peter [email protected] Reproductive Sciences Section, Department of Cancer Studies and Molecular Medicine, University of Leicester, Leicester, UK2 Department of Epidemiology and Public Health, University of Leicester, Leicester, UK3 Paediatrics Department, Great Ormond Street Hospital for Sick Children, London, UK4 Department of Child Health, University of Nottingham, Nottingham, UK5 National Perinatal Epidemiology Unit, Oxford, UK2005 18 8 2005 5 55 55 16 3 2005 18 8 2005 Copyright © 2005 Kenyon et al; licensee BioMed Central Ltd.2005Kenyon 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
Postal questionnaires are widely used to collect data in healthcare research but a poor response rate may reduce the validity and reliability of results. There was a lack of evidence available relating to use of a monetary incentive to improve the response rate in the healthcare setting.
Methods
The MRC ORACLE Children Study is assessing the health and development of nearly 9000 seven year old children whose mothers' joined the MRC ORACLE Trial. We carried out a randomised controlled trial of inclusion of monetary incentive (five pound voucher redeemable at many high street stores) with the reminder questionnaire to parents. This trial took place between April 2002 and November 2003. When the parents were sent the reminder questionnaire about their child's health and development they were randomly assigned by concealed computer-generated allocation stratified by week of birthday to receive a five pound voucher or no incentive. The population were 722 non-responders to the initial mailing of a 12-page questionnaire. Main outcome measures: Difference in response rate between the two groups.
Results
Inclusion of the voucher with the reminder questionnaire resulted in a 11.7%(95% CI 4.7% to 18.6%) improvement in the response rate between the two groups.
Conclusion
This improvement in response rate and hence the validity and reliability of results obtained appears to be justified ethically and financially.
==== Body
Background
Postal questionnaires are widely used to collect data in health research, but a poor response rate may reduce the validity and reliability of results. In a systematic review of randomised controlled trials of strategies to improve the response rate to postal questionnaires[1], a monetary reward had a significant effect on response. However, caution was attached to the interpretation of the findings in this review [2]. On further examination of the updated review 20% of the participants included in the analysis of the effect of inclusion of a monetary incentive in the final response came from healthcare settings [3] and none of the studies evaluated the use of monetary incentives for a postal questionnaire to collect data from a follow-up of a clinical trial. To evaluate the impact of such an intervention on response rate in such a setting we undertook a randomised trial.
The MRC ORACLE Children Study (MOCS) is following up nearly 9000 seven-year-old children whose mothers joined the MRC ORACLE Trial [4,5] which evaluated the use of antibiotics to improve neonatal outcome after preterm labour or preterm rupture of the membranes. This trial of a monetary reward to enhance response to a postal questionnaire was undertaken between April 2002 – November 2003. Research Ethics Committee approval was obtained from the West Midlands Multicentre Research Ethics Committee
The questionnaire itself is 12 pages long and A4 in size. It contains questions relating to the child's health and development using a mixture of validated tools and questions pertinent to the study. In designing the study we implemented many of the strategies believed to influence response rates to postal questionnaires [1,3]. The questionnaire itself is set out in a user friendly way and we have used coloured ink in its design. The letters accompanying the questionnaire are individualised to the child concerned and the parents are warned the questionnaire will be sent to them. The University in which MOCS is housed franks the envelope, the return envelope is stamped, and reminder letters include a questionnaire.
Methods
When a child in MOCS is seven years old the parents receive an information leaflet about the follow-up Study, and two weeks later a questionnaire about their child's health and development. Contact with parents has already been established prior to this. If no response is obtained the child's General Practitioner is contacted to check the child's address and ensure that contact would be appropriate. Six weeks after the first questionnaire, a reminder one is sent to those who have not responded. At this point the parents were randomly assigned by computer-generated allocation to receive a five-pound voucher (redeemable at many high street shops) with their mailed questionnaire or not (see Figure 1).
Figure 1 Consort flow diagram.
The sample size was predefined by the numbers not responding at this stage of the Study (i.e. approx 700). This yielded 80% power to detect an increased response from 10% to 18% or from 15% to 24% (at the 5% significance level).
Results
Balance between voucher/not voucher groups on the main baseline covariates was good. 722 consecutive parents were randomly allocated to receive a voucher or not with the reminder questionnaire (see Table 1).
Table 1 Results of random allocation of voucher or not on response rate
Voucher No voucher
Questionnaire returned 156 (42.3%) 108 (30.6%)
Questionnaire not returned 213 245
Total 369 353
Inclusion of the voucher with the second questionnaire resulted in a 11.7% (95% CI 4.7% to 18.6%) improvement in the response rate between the two groups (χ2 = 10.6, P = 0.001).
Discussion
The inclusion of a five-pound voucher improved the proportion of questionnaires returned. MOCS will be completed in 2008 and a voucher is being sent to all parents with the reminder questionnaire. It is estimated that this will improve the response rate by 3% over the whole study, at a cost of £67 per additional questionnaire returned. This was calculated on the basis that vouchers will be sent to approximately 40% of parents (2842) and an additional 3% will return the questionnaire
This improvement in response rate, and hence of the validity and the reliability of results appears to be justified ethically and financially. This is particularly relevant in the follow up of children as there is some evidence [6] of raised levels of adverse outcomes in difficult to follow-up children.
Competing interests
The author(s) declare that they have no competing interests.
Contributors
SK and PB designed the trial. DJ and KP carried out the analysis. NM, AS, DT contributed to conception, design and interpretation. All authors read and approved the trial manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Ann Blackburn, Kate Taylor, Gill Grummitt.
Funding: UK Medical Research Council. Grant number 53994660
==== Refs
Edwards P Roberts I Clarke M Di Guiseppi C Pratap S Wentz R Increasing response rates to postal questionnaire: systemic review BMJ 2002 324 1183 1185 12016181 10.1136/bmj.324.7347.1183
Smeeth L Fletcher AE Editorial Improving response rates to questionnaires BMJ 2002 324 1168 1169 12016167 10.1136/bmj.324.7347.1168
Edwards P Roberts I Clarke M DiGuiseppi C Pratap S Wentz R Kwan I Cooper R Methods to increase response rates to postal questionnaires The Cochrane Database of Methodology Reviews 2003 Issue 4.Art No.:MR000008.pub2.DOI: 10.1002/14651858.MR000008.pub2
Kenyon SL Taylor DJ Tarnow-Mordi W ORACLE Collaborative Group Broad-spectrum antibiotics for preterm, prelabour rupture of fetal membranes: the ORACLE I randomised trial Lancet 2001 357 979 988 11293640 10.1016/S0140-6736(00)04233-1
Kenyon SL Taylor DJ Tarnow-Mordi W ORACLE Collaborative Group Broad-spectrum antibiotics for spontaneous preterm labour: the ORACLE II randomised trial Lancet 2001 357 989 994 11293641 10.1016/S0140-6736(00)04234-3
Tin W Fritz S Wariyar U Hey E Outcome of very preterm birth children reviewed with ease at 2 years differ from those followed up with difficulty Arch Dis Child Fetal Neonatal Ed 1998 79 F83 87 9828731
|
0
|
PMC1201140
|
CC BY
|
2021-01-04 16:28:15
|
no
|
BMC Infect Dis. 2005 Aug 19; 5:65
|
latin-1
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-65
|
oa_comm
|
==== Front
BMC ImmunolBMC Immunology1471-2172BioMed Central London 1471-2172-6-201612021510.1186/1471-2172-6-20Methodology ArticleQuantitative PCR for detection of the OT-1 transgene Wright Kate O [email protected] Debbie A [email protected] Nicholas I [email protected] Robert H [email protected] Department of Pathology, University of Rochester Medical Center, Rochester, NY 14642, USA2 David H. Smith Center for Vaccine Biology and Immunology, Aab Institute of Biomedical Sciences, University of Rochester, Rochester, NY 14642, USA2005 24 8 2005 6 20 20 8 3 2005 24 8 2005 Copyright © 2005 Wright 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
Transgenic TCR mice are often used experimentally as a source of T cells of a defined specificity. One of the most widely used transgenic TCR models is the OT-1 transgenic mouse in which the CD8+ T cells express a TCR specific for the SIINFEKL peptide of ovalbumin presented on kb. Although OT-1 CD8+ can be used in a variety of different experimental settings, we principally employ adoptive transfer and peptide-driven expansion of OT-1 cells in order to explore the distribution and fate of these antigen-specific OT-1 T cells. We set out to develop a quantitative PCR assay for OT-1 cells in order to assess the distribution of OT-1 CD8+ T cells in tissues that are either intrinsically difficult to dissociate for flow cytometric analysis or rendered incompatible with flow cytometric analysis through freezing or fixation.
Results
We show excellent correlation between flow cytometric assessment of OT-1 cells and OT-1 signal by qPCR assays in cell dilutions as well as in in vivo adoptive transfer experiments. We also demonstrate that qPCR can be performed from archival formalin-fixed paraffin-embedded tissue sections. In addition, the non-quantitative PCR using the OT-1-specific primers without the real-time probe is a valuable tool for OT-1 genotyping, obviating the need for peripheral blood collection and subsequent flow cytometric analysis.
Conclusion
An OT-1 specific qPCR assay has been developed to quantify adoptively transferred OT-1 cells. OT-1 qPCR to determine cell signal is a valuable adjunct to the standard flow cytometric analysis of OT-1 cell number, particularly in experimental settings where tissue disaggregation is not desirable or in tissues which are not readily disassociated
==== Body
Background
Adoptive transfer of CD8+ T cells from MHC Class I-restricted OVA specific T cell receptor (TCR) transgenic mice (OT-1) into host C57BL/6 mice allows for the activated CD8+ T cell population to be tracked and quantified after stimulation with the OVA peptide SIINFEKL [1] using standard flow cytometric analysis. Using this methodology, it has been demonstrated that activated CD8+ T cells undergo migration to the liver where they are trapped and undergo apoptosis [2]. While flow cytometry is sensitive and specific, it is best suited to tissues from which single cell suspensions can be obtained easily, such as the spleen and peripheral lymph nodes. Although flow cytometry can be performed on lymphocytes isolated from other tissues such as liver, muscle and intestine, it is technically more challenging and questions may be raised regarding whether the cell sample is representative of the in vivo state. Reports by Lang (1997), Lim (2002), and Gallard (2002), [3-5] also emphasize these points and describe PCR based approaches for tracking specific T-cell clonotypes. Additionally, sample preparation for histologic analysis and nucleic acid or protein-based assessment is incompatible with tissue dis-aggregation methods for flow cytometry. In order to augment standard flow cytometric analysis in these more challenging experimental settings, a real-time quantitative PCR assay was developed to quantify and track activated OT-1 cells with a high degree of sensitivity.
Because the OT-1 mouse was made by micronuclear injection of genomic DNA, which did not contain unique transgene non-coding sequence, we based a quantitative PCR strategy on the DNA sequence encoding the rearranged VDJ sequences corresponding to the CDR3 region of the TCR Beta chain. As genomic DNA is the starting point for qPCR, any tissue harvested from a host mouse can be assayed for the OT-1 transgene. Additionally, nucleic acid isolated from paraffin-embedded material is also a suitable template for real-time quantitative PCR, making this technique useful for analysis of archival paraffin-embedded material.
The utility and sensitivity of a qPCR assay designed to detect the OT-1 transgene was compared to the standard method of flow cytometry to track and quantify activated OT-1 cells in both in vitro dilution and in vivo adoptive transfer experiments.
Results
OT-1 Genotyping PCR is a specific and reliable method to determine positive OT-1 mice from wild-type littermates
A PCR strategy was developed to identify OT-1 positive transgenic mice, which express a T cell receptor (TCR) specific for the OVA peptide SIINFEKL, from wild-type littermates. Primers were designed in the VDJ region of the OT-1 TCRβ chain (Figure 1) to yield a 238 bp product. Figure 2 demonstrates that the OT-1 fragment is amplified from OT-1 positive transgenic mice (lanes 2, 3, and 5), but not from wild-type littermates (lanes 1,4, and 6). Flow cytometric phenotyping of PCR positive OT-1 mice confirmed the specificity of the reaction (data not shown).
A quantitative real-time PCR assay was developed using a 27 base pair fluorophore-coupled oligonucleotide probe, with sequence-dependent hybridization within the 238 base pair amplified VDJ region of the OT-1 TCR. This assay was successful at determining OT-1 transgenic mice from negative littermates. Figure 3 shows data from a real-time PCR experiment using the pCRTOPO2.1-OT1 plasmid at various concentrations and genomic DNA isolated from positive and negative OT-1 transgenic mice. Based on the plasmid standard curve, 8.408E+5 copies of the OT-1 transgene were present in 300 ng of genomic DNA from an OT-1 positive mouse. There was no signal amplification when genomic DNA from the OT-1 negative mouse was used as template, or in the no template control. To normalize all qPCR results for the OT-1 transgene, a fluorophore-coupled oligonucleotide probe and primer set for murine β-actin was also designed and used for each sample.
To improve upon the sensitivity of the quantitative PCR assay sequence-independent detection with SYBR Green I was also performed. Double-stranded DNA product was quantified by monitoring fluorescence of the DNA-binding SYBR Green I dye. Plasmid pCRTOPO2.1 containing the OT-1 fragment was used as template in a standard curve analysis at concentrations of 2 ng to 2 E-8 ng, correlating to 4.38E+8 to 4.38E+0 copies of transgene. This real-time PCR strategy was successful at detection down to 4.38 copies of the transgene and did not amplify signal in the absence of DNA template. Melting curve analysis also showed a single PCR product was generated. In subsequent experiments using genomic DNA isolated from animal tissue, however, high background signals and multiple fluorescent peaks were generated, indicating a quantitative PCR assay based on SYBR Green would not be sufficient for future experiments.
Sensitivity and specificity of real-time quantitative PCR on diluted OT-1 genomic DNA
The linear range of sensitivity of the qPCR assay was determined by diluting OT-1 DNA (isolated from the liver of an OT-1 transgenic mouse) with genomic DNA isolated from a non-transgenic C57BL/6 normal mouse. Figure 4 shows a regression analysis of the logarithmic values for normalized OT-1 copy number (defined as the number of OT-1 copies/10,000 β-actin copies) by real-time qPCR (done in triplicate) plotted against the logarithmic values of the actual DNA dilution values. The calculated R2 value for the regression analysis was 0.9684. The real-time assay was used to quantify a range of 5% to 0.1560% of OT-1 DNA diluted into background C57BL/6 DNA. The real-time qPCR could also distinguish 0.3125% OT-1 DNA from 0.1560% OT-1 DNA, demonstrating that the sensitivity and specificity of the assay would be sufficient for future applications.
Quantitative real-time PCR for detection of OT-1 CD8+T cells diluted into normal C57Bl/6 splenocytes
To test the sensitivity of the real-time assay, OT-1 CD8 +T cells were purified from an OT-1 transgenic mouse. OT-1 cells were diluted into splenocytes isolated from normal C57BL/6 mice so that the dilutions were comprised of 100%, 4%, 1,3%, 0.433%, 0.144%. 0.048%, and 0.016% OT-1 cells in a background of wild-type C57BL/6 splenocytes. Previous to the dilutions, flow cytometric analysis was performed to determine cell purity and yield of both cell suspensions, which were both resuspended at 8 million cells/mL. 200 μL was taken for flow cytometric analysis and genomic DNA was isolated from the remaining 3.2 million cells. 300 ng of genomic DNA from each dilution (done in triplicate) was used as template for quantitative PCR to determine OT-1 and β-actin copy number.
Figure 5 demonstrates the correlation between the results obtained from qPCR for the average normalized OT-1 transgene copy number and the average actual percentages of OT-1 cells determined by flow cytometry. To further analyze the data and to separate the cluster of data points, regression analysis for the logarithmic values of normalized OT-1 transgene copy (defined as the number of OT-1 copies/10,000 β-actin copies) and actual percentage of OT-1 cells by flow cytometry was done. The OT-1 dilutions of 100%, 4%, 1,3%, 0.433%, 0.144%. 0.048%, and 0.016% were used in the analysis. Figure 6 demonstrates the regression analysis with a calculated R2value of 0.9458 and also shows that linearity of the assay was maintained and the lowest dilution values of 0.048% versus 0.016% could be resolved. Again, the sensitivity and specificity of the real-time assay are evidence for the utility of qPCR for future applications.
Accumulation of activated OT-1 T-cells after adoptive transfer by real-time quantitative PCR
Accumulation of activated OT-1 CD8+ T cells in the liver after adoptive transfer and in vivo activation and expansion has been well documented [6-8]. The purpose of these experiments was to determine whether a quantitative PCR assay would be sensitive enough to also monitor and quantify the migration of T-cells to the liver and other organs after an adoptive transfer experiment compared to results obtained using standard flow cytometric analysis.
Figure 7 demonstrates the correlation between flow cytometric analysis and qPCR to quantify OT-1 cells in the liver of animals given i.p. injections of PBS or OVA stimulating peptide. Livers were harvested on days 3, 5, and 7 after the first i.p. injection of peptide. One lobe was reserved for DNA isolation and qPCR, the remainder of the tissue was prepared for flow cytometric analysis. The percent of OT-1 cells by flow cytometric analysis were compared to the OT-1 gene levels (shown as the number of OT-1 gene copies per 1*105 copies of β-actin). Although there is variation in the absolute numbers, both methods demonstrate an expansive OT-1 CD8+ T-cell population in the liver of OVA treated animals, which diminishes by the day 7 timepoint.
Spleens from the same animals, sacrificed on days 3, 5, and 7 after first i.p. injection with peptide, were analyzed in the same manner (Figure 8). Roughly half of the spleen was prepared for flow cytometric analysis and the remaining half was used for DNA isolation and qPCR. The percent of OT-1 cells by FACS analysis in the spleen samples were compared to the normalized OT-1 gene levels. Again, while there was variation in the absolute numbers, both methods show an expanded OT-1 CD8+ T-cell population in the spleens of OVA treated animals, which is most prominent on the day 3 timepoint, and had similar trends. When the normalized copies of the OT-1 transgene for all liver and spleen samples from each timepoint were plotted against the percent of OT-1 cells by flow cytometric analysis, the correlation coefficient was calculated at R2 = 0.7027.
The results from these experiments demonstrate the utility of a quantitative PCR assay on a tissue that is amenable (spleen) and one that is not so amenable (liver) to flow cytometry. While the trends of OT-1 cell accumulation in the various organs were very similar there were definite differences in the resulting numbers. Part of the discrepancy may be due to sampling artifact: Although the same tissue specimens were used for flow cytometry and qPCR, different sections of the tissue were used in each assay. A small portion of tissue (approximately 75 mg) was needed for DNA isolation, while one half to three-quarters of the tissue was used to make cell suspensions for flow cytometry, implying that variation inherent to each individual organ was present and was a possible contributor to the observed discrepancies. Also pertinent to this argument is to point out that flow cytometric analysis in the liver involves sampling the leukocyte fraction after a mononuclear cell isolation, where DNA isolation for qPCR analysis from whole liver includes the large parenchymal population. Quantitative PCR and flow cytometry also have different denominators to consider. Flow cytometry determines the percent of OT-1 cells out of the selected CD8+ CD45.1+ positive population in the leukocyte gate while qPCR determines the number of OT-1 copies in all of the sampled DNA.
To further test the utility of this assay, the lungs of each animal were also flash frozen and stored for later DNA isolation. Due to the difficulty in obtaining adequate materiel for single cell suspensions, flow cytometric analysis was not performed from lung tissue. Since qPCR requires isolated nucleic acid material and not single cell suspensions, quantitative OT-1 transgene numbers could be easily obtained from lung samples. After normalization to the β-actin control, results from quantitative PCR show a marked increase in OT-1 transgene copy number in the OVA treated animals compared to PBS controls (Figure 9). These results further demonstrate the utility of this assay for tissue samples that are difficult to disassociate for flow cytometry and also that a migratory OT-1 CD8+ T cell population is present in lung tissue of animals after adoptive transfer experiments.
Quantitative PCR analysis of OT-1 transgene on archival material
Paraffin embedded slides (5 uM) were made from liver sections of C57BL/6 mice used as hosts in OT-1 CD8+ T cell adoptive transfer experiments. The animals were sacrificed on day five after the first i.p injection of the stimulating OVA peptide or PBS control. Histological examination of the slides confirmed the presence of foci (trapped and apoptotic T cells) in the livers of animals given OVA peptide while livers of mice given PBS injections did not have significant foci formation. DNA isolated from the paraffin-embedded slides was used in real-time qPCR for detection of the OT-1 transgene. Although the paraffin slides had been stored at room temperature for approximately 18 months, genomic DNA was readily isolated from the samples. Figure 10 demonstrates average levels of the OT-1 transgene in animals that had received the OVA stimulating peptide were nearly five-fold higher than in PBS control animals, confirming the utility of the qPCR assay on archived experiments.
Conclusion
The aim of this work was to develop a qPCR method for the tracking and quantification of activated OT-1 CD8+ T cells as an alternative to flow cytometric analysis. While sensitivity and specificity of flow cytometry is sufficient for most applications, it is limited to tissue in which single cell suspensions can be obtained without a high degree of difficulty, such as the spleen and peripheral lymph nodes. Flow cytometry can also be performed on liver tissue, with more difficulty, by using a very time consuming protocol. The purification of mononuclear cells for flow cytometric analysis from brain or muscle tissue is also quite difficult and could limit the value of the transgenic model of T cell activation and clearance in an experimental setting where a complete tissue survey was required. Our real-time quantitative PCR assay is useful for detection of the OT-1 transgene using both in vitro and in vivo experiments, particularly where flow cytometry is difficult. On a more practical level, it should be noted that quantitative PCR assays are relatively easy to use and can be performed in a variety of settings. Also, samples from ongoing experiments with multiple time points can be harvested and stored until all specimens are obtained and can be prepared all at the same time.
We have validated the quantitative real-time PCR for the OT-1 transgene against known numbers of OT-1 cells in dilution experiments in vitro and correlated the PCR against flow cytometric assessment. In addition, we have validated the OT-1 qPCR against flow cytometric analysis of tissue samples. In all cases, we have observed excellent correlation. Therefore, we are confident that this quantitative PCR based approach to detect OT-1 cells will be a very useful tool in the detection and quantification of OT-1 cells in a variety of experimental models.
Methods
Animals
C57B/6J and B6.SJL-Ptprca Pep3b/BoyJ mice were obtained from the Jackson Laboratory (Bar Harbor, ME). OT-1 mice express a transgenic TCR that recognizes the 8-mer SIINFEKL peptide derived from residues 257–264 of ovalbumin. Transgenic animals were identified by staining PBL with antibodies against CD8 and the chains of the transgenic TCR, Vα 2 and Vβ 5. A colony of OT-1 mice were maintained on the B6.SJL (CD45.1 expressing) background. All animals were housed in a specific pathogen-free environment in accordance with institutional guidelines for animal care.
Adoptive transfer and in vivo activation
Donor CD8 + OT-1 T cells from mouse spleen and peripheral lymph nodes were purified through depletion of B cells, dendritic cells, NK cells and CD4 +T cells using primary antibodies (clone 212.Al specific for MHC class II molecules, clone 2.4.G2 specific for FcRs, clone GK1.5 specific for CD4 and clone HB.191 specific for NK1.1), followed by magnetic beads (Biomag goat anti-mouse IgM, goat anti-mouse IgG and goat anti-rat IgG, Qiagen). Purity was calculated to be greater than 90%. A suspension of 5 × 106 OT-1 cells was injected IV into each recipient host mouse. OT-1 T cells were activated after adoptive transfer by daily i.p. injections of 25 nM SIINFEKL peptide (New England Peptide Inc., Fitchburg, MA) in PBS for 3 days starting 24 h after OT-1 transfer as previously described by Mehal, et al [9]. Control mice received an equivalent volume of PBS.
Mononuclear cell isolation
Intrahepatic lymphocytes were isolated using a standard method (Huang, 1994). In brief, the liver was perfused through the portal vein with 5 ml ice-cold PBS, homogenized by forcing through a metal strainer, then digested with 10 ml of a digestion buffer (RPMI containing 0.02% (w/v) collagenase IV and 0.002% (w/v) DNase I (Sigma, St. Louis, MO)) at 37°C with shaking. The digested liver was centrifuged twice at 10 × g for 4 min at 4°C to remove hepatocytes. The non-parenchymal cells were resuspended to a final volume of 1.4 ml in RPMI before mixing with 2.6 ml of 40% metrizamide (ICN Biomedicals Inc. Aurora, OH) resulting in a final metrizamide concentration of 26%. The cell suspension/metrizamide mixture was then overlayed with 2 ml of RPMI and centrifuged at 1500 × g for 30 min at 4°C. The cells at the interface were collected, washed in HBSS + 5% FBS and counted. Spleens and lymph node cell suspensions were washed at 120 × g for 10 min at 4°C, filtered through 100 μm strainers to remove debris and counted.
Flow cytometric analysis
Cells were surface stained with Vα2-FITC, Vβ5-PE, CD45.2-FITC, CD45.1-PE, CD8-PerCP (BD Pharmingen) and Vα2-APC (Caltag, Burlingame, CA) for 30 min at 4°C, washed in PBS and fixed in a paraformaldehyde-based fixative. Cells were analyzed using a BD FACSCalibur. Data were analyzed using Cell Quest software (BD Biosciences). OT-1 T cells were identified using light scatter gates characteristic of lymphocytes, as well as positive staining for CD45.1 and CD8.
DNA isolation from mouse tissue
Tissue sections from donor mice were immediately flash frozen in liquid nitrogen after removal from animal and stored at -85 degrees C until ready for processing. The DNeasy kit (Qiagen) was used to isolated genomic DNA from approximately 25 to 75 mg of flash frozen mouse tissue. Alternatively, to make purified genomic DNA from OT-1 and wild-type C57B6 mice for DNA dilution experiments, whole livers were removed from animals and immediately flash frozen in liquid nitrogen. Tissues were homogenized and digested overnight, at 55 degrees C, in digestion buffer (100 mM NaCl, 10 mM Tric-Cl, pH 8.0, 25 mM EDTA, pH 8.0, 0.5% SDS, and 0.1 mg/mL Proteinase K). Genomic DNA was isolated by phenol:chloroform extraction followed by ethanol precipitation.
Cell dilutions
CD8 +T cells were isolated from the spleens of two MHC class I-restricted OVA-specific TCR transgenic mouse (OT-1), as described. Monocytes were isolated from the spleens of 5 wild-type C57B6 mice. FACS analysis was performed to determine cell purity and yield, and appropriate cell dilutions were made. 200 uL was taken for further FACS analysis and genomic DNA was isolated from the remaining 3.2 million cells (Qiagen DNeasy kit). 300 ng of genomic DNA from each dilution was used as template for quantitative PCR to determine OT-1 and β-actin copy number.
Hybridization probe based quantitative PCR
The LightCycler PCR and detection system (Roche Diagnostics) was used for amplification and quantification of the OT-1 transgene from each tissue sample. The LightCycler FastStart DNA Master Hybridization Probes kit was used as described by the manufacturer. Each PCR mixture contained 250 ng of genomic DNA template, 1X LightCycler FastStart reaction buffer, 5.0 mM magnesium chloride, 0.5 uM sense primer (ACGTGTATTCCCATCTCTGG), 0.5 uM antisense primer (CTGTTCATAATTGGCCCGA) and 0.3 uM OT-1 Hybridization probe (5' FAM/TGCCTTGGAACTGGAGGACTCTGCTAT 3'/TAMRA) to amplify 238 bp fragment of the OT-1 transgene. The OT-1 transgene fragment corresponds with nucleotides 150 to 388 of the transgenic T cell receptor. The PCR was performed in glass capillaries with 50 cycles of denaturation (5 s at 95°C) and annealing and extension (75 s at 61°C) after an initial 95°C melt for 10 minutes. Standard curve analysis was performed with each LightCycler run using the OT-1 transgene product cloned into pcCRTOPO2.1 (Invitrogen) for plasmid template. Each DNA sample was also analyzed for β-actin DNA following the same protocol using sense and antisense β-actin primers (agccatgtacgtagccatcc and ttcaccaccacagctgagag), respectively, and a β-actin hybridization probe (5' FAM/ACCTGACGGACTACCTCATGAAGATCC 3'/TAMRA). These primers correspond with mouse genomic DNA on chromosome 5 nucleotide positions 14162633 to 141626562 [10]. Standard curve analysis was performed with each LightCycler run using the mouse β-actin DNA product cloned into pCRBLUNT (Invitrogen) for plasmid template.
SYBR Green quantitative PCR
To determine sensitivity of a real time PCR assay for OT-1 detection, the LightCycler PCR and detection system (Roche Diagnostics) was used for amplification and quantification of OT-1 DNA. The LightCycler FastStart DNA Master SYBR Green kit was used as described by the manufacturer. Each PCR mixture contained 1X LightCycler reaction buffer, 5 mM magnesium chloride, and 0.5 uM sense (ACGTGTATTCCCATCTCTGG) and antisense (CTGTTCATAATTGGCCCGA) primers to amplify 238 bp fragment of OT-1 transgene. PCR was performed in glass capillaries with 55 cycles of denaturation (2 s at 95°C), annealing (2 s at 61°C), and chain extension (8 s at 72°C). Melting curve analysis from 65°C to 95°C, followed by cooling to 40°C, was also performed. Double-stranded DNA product was quantified by monitoring fluorescence of the DNA-binding SYBR Green I dye. Plasmid pCRTOPO2.1 containing the OT-1 fragment was used as template in standard curve analysis at concentrations of 2 ng to 28*10 -8 n, correlating to 4.38*108 to 4.38*100 copies of transgene.
DNA isolation from paraffin-embedded slides
C57BL/6 host mice were used in adoptive transfer experiments of OT-1 CD8+T cells, as described. Five days after the first i.p. injection with OVA peptide, animals were sacrificed and livers were fixed in paraffin. Five uM thick slides were cut from experimental livers and also from livers of normal C57BL/6 mice. Paraffin-embedded material was scraped from slides and DNA was isolated following the protocol outlined in the Qiagen QIAmp kit (Invitrogen).
Authors' contributions
KOW designed and carried out the genotyping and quantitative PCR assays, participated in final analysis and drafted the manuscript. DAM performed the adoptive transfer experiments and flow cytometric analysis. INC participated in design of the study and final data interpretation. RHP participated in design of the study, statistical analysis, and final data interpretation. All of the authors read and approved the final manuscript.
Acknowledgements
We would like to acknowledge Judith Cornejo for her technical assistance.
Figures and Tables
Figure 1 Schematic Representation of Primers and Probes. Schematic representation of primers and hybridization probe location within the VDJ region of the OT-1 T Cell Receptor Beta chain.
Figure 2 OT-1 Genotyping Results. OT-1 genotyping results demonstrate the 238 bp OT-1 fragment was amplified from OT-1 positive transgenic mice (lanes 2, 3, and 5) but not from negative littermates (lanes 1, 4, and 6). Plasmid pcRTOPO2.1 containing the cloned OT-1 product was used as a positive control (lane 7).
Figure 3 Quantitative Real-time PCR Data. Quantitative real-time PCR data for the OT-1 transgene using pCR TOPO2.-OT-1 plasmid standards and genomic DNA from OT-1 positive and OT-1 negative mice. 8.408 *105 copies of the OT-1 transgene were detected in 300 ng of genomic DNA from the OT-1 positive mouse while no signal was generated from the OT-1 negative mouse.
Figure 4 Regression Analysis of OT-1 Copy Number. Regression analysis of the logarithmic values for normalized OT-1 copy number (defined as the number of OT-1 copies/10,000 β-actin copies) by real-time qPCR (done in triplicate) plotted against the logarithmic values of the actual DNA dilution values. The R2 value was calculated to be 0.9684.
Figure 5 Correlation of qPCR and Flow Cytometry. Correlation of qPCR and flow cytometric analysis of for detection of OT-1 cells after dilution into wild-type C57BL/6 spleenocytes. The cell dilutions were calculated to represent 4%, 1,3%, 0.433%, 0.144%. 0.048%, and 0.016% total OT-1 CD8+T cells.
Figure 6 Regression Analysis of qPCR and Flow Cytometry. Regression analysis for the logarithmic values of normalized OT-1 transgene copy (defined as the number of OT-1 copies/10,000 β-actin copies) and actual percentage of OT-1 cells by flow cytometry. OT-1 dilutions of 100%, 4%, 1,3%, 0.433%, 0.144%. 0.048%, and 0.016% were used for the analysis and the R2 value was calculated to be 0.9458.
Figure 7 Correlation of qPCR and Flow Cytometry After Adoptive Transfer. Correlation between qPCR and flow cytometry for detection and quantification of OT-1 present in the livers of host animals after adoptive transfer. Mice were sacrificed on days 3, 5, or 7 after the first i.p. injection of stimulating OVA peptide, or PBS control.
Figure 8 Correlation of qPCR and Flow Cytometry After Adoptive Transfer. Correlation between qPCR and flow cytometry for detection and quantification of OT-1 present in the spleens of host animals after adoptive transfer. Mice were sacrificed on days 3, 5, or 7 after the first i.p. injection of stimulating OVA peptide, or PBS control.
Figure 9 Detection of OT-1 Cells in Lungs After Adoptive Transfer. Quantitative PCR for OT-1 transgene demonstrates migration of activated OT-1 cells into the lungs of host animals after adoptive transfer. The normalilzed OT-1 transgene copy numbers were averaged for each treatment (PBS or stimulating OVA peptide) on mice sacrificed on days 3, 5, or 7 after the first i.p. injection was administered.
Figure 10 Detection of OT-1 Cells from Archival Material. Detection of OT-1 cells from paraffin-embedded archived material is possible using the OT-1 qPCR assay. Paraffin embedded liver sections (5 uM) of C57BL/6 mice used as hosts in OT-1 CD8+ T cell adoptive transfer experiments were used to generate template DNA. The animals were sacrificed on day five after the first i.p injection of the stimulating OVA peptide or PBS control.
==== Refs
Sterry SJ Kelly JM Turner SJ Carbone FR T cell receptor V alpha bias can be determined by TCR-contact residues within an MHC-bound peptide Immunol Cell Biol 1995 73 89 94 7768549
Huang L Soldevila G Leeker M Flavell R Crispe IN The liver eliminates T cells undergoing antigen-triggered apoptosis in vivo Immunity 1994 1 741 9 7895163 10.1016/S1074-7613(94)80016-2
Lang R Pfeffer K Wagner H Heeg K A rapid method for semiquantitative analysis of the human VB-repertoire using Taqman PCR J Immunol Methods 1997 203 181 92 9149812 10.1016/S0022-1759(97)00028-8
Lim A Baron V Ferradini L Bonneville M Kourilsky P Pannetier C Combination of MHC-peptide multimer-based T cell sorting with the Immunoscope permits sensitive ex vivo quantitation and follow-up of human CD8+ T cell immune responses J Immunol Methods 2002 261 177 94 11861076 10.1016/S0022-1759(02)00004-2
Gallard A Foucras G Coureau C Guery JC Tracking T cell clonotypes in complex T lymphocyte populations by real-time quantitative PCR using fluorogenic complementarity-determining region-3-specific probes J Immunol Methods 2002 270 269 80 12379331
Mehal WZ Juedes AE Crispe IN Selective retention of activated CD8+ T cells by the normal liver J Immunol 1999 163 3202 10 10477588
John B Crispe IN Passive and active mechanisms trap activated CD8+ T cells in the liver J Immunol 2004 172 5222 9 15100260
Murray DA Crispe IN TNF-alpha controls intrahepatic T cell apoptosis and peripheral T cell numbers J Immunol 2004 173 2402 9 15294953
Mehal WZ Azzaroli F Crispe IN Antigen presentation by liver cells controls intrahepatic T cell trapping, whereas bone marrow-derived cells preferentially promote intrahepatic T cell apoptosis J Immunol 2001 167 667 73 11441069
UCSC Genome Bioinformatics
|
16120215
|
PMC1201141
|
CC BY
|
2021-01-04 16:03:32
|
no
|
BMC Immunol. 2005 Aug 24; 6:20
|
utf-8
|
BMC Immunol
| 2,005 |
10.1186/1471-2172-6-20
|
oa_comm
|
==== Front
BMC Med EducBMC Medical Education1472-6920BioMed Central London 1472-6920-5-301610517810.1186/1472-6920-5-30Research ArticleIdentifying inaccuracies on emergency medicine residency applications Katz Eric D [email protected] Lee [email protected] Lawrence [email protected] David [email protected] Janis P [email protected] Christopher [email protected] Osman R [email protected] Victoria [email protected] Jason [email protected] Diamond [email protected] Jackie [email protected] Timothy [email protected] Gene [email protected] Ralph [email protected] Antonio [email protected] Keith [email protected] Edward [email protected] Division of Emergency Medicine, Washington University in St. Louis, St. Louis, MO, USA2 Department of Emergency Medicine, Denver Metro Health Center, Denver, CO, USA3 Department of Emergency Medicine, Pennsylvania State University, Hershey, PA, USA4 Division of Emergency Medicine, University of Chicago, Chicago, IL, USA5 Department of Emergency Medicine, Indiana University School of Medicine, Indianapolis, IN, USA6 Department of Emergency Medicine, NewYork-Presbyterian Hospital, New York, NY, USA7 Department of Emergency Medicine, University of Alabama, Birmingham, AL, USA8 Department of Emergency Medicine, Medical College of Virginia, Richmond, VA, USA9 Department of Emergency Medicine, Highland General Hospital, Oakland, CA, USA10 Department of Emergency Medicine, Thomas Jefferson University, Philadelphia, PA, USA2005 16 8 2005 5 30 30 7 6 2005 16 8 2005 Copyright © 2005 Katz et al; licensee BioMed Central Ltd.2005Katz 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
Previous trials have showed a 10–30% rate of inaccuracies on applications to individual residency programs. No studies have attempted to corroborate this on a national level. Attempts by residency programs to diminish the frequency of inaccuracies on applications have not been reported. We seek to clarify the national incidence of inaccuracies on applications to emergency medicine residency programs.
Methods
This is a multi-center, single-blinded, randomized, cohort study of all applicants from LCME accredited schools to involved EM residency programs. Applications were randomly selected to investigate claims of AOA election, advanced degrees and publications. Errors were reported to applicants' deans and the NRMP.
Results
Nine residencies reviewed 493 applications (28.6% of all applicants who applied to any EM program). 56 applications (11.4%, 95%CI 8.6–14.2%) contained at least one error. Excluding "benign" errors, 9.8% (95% CI 7.2–12.4%), contained at least one error. 41% (95% CI 35.0–47.0%) of all publications contained an error. All AOA membership claims were verified, but 13.7% (95%CI 4.4–23.1%) of claimed advanced degrees were inaccurate. Inter-rater reliability of evaluations was good. Investigators were reluctant to notify applicants' dean's offices and the NRMP.
Conclusion
This is the largest study to date of accuracy on application for residency and the first such multi-centered trial. High rates of incorrect data were found on applications. This data will serve as a baseline for future years of the project, with emphasis on reporting inaccuracies and warning applicants of the project's goals.
==== Body
Background
The residency application process is predicated on the validity of the credentials submitted by an applicant. Previous studies, however, have found that 10–30% of applications contain errors. [1-6] In emergency medicine (EM), Roellig, et. al. showed that 13.3% of applicants to a single EM residency had at least one error on their application, and 4% had more than one.[7] Rates of erroneous claims were similar for claims of authorship (21.3% erroneous), Alpha-Omega-Alpha (AOA) claims (35.7%) and advanced degrees (26.7%). In a previous single center investigation, Gurudevan, et. al. reported that 20.4% of applicants who claimed authorship of a peer- reviewed paper had at least one error in their reference.[8]
Presently, the residency program discovering misrepresentations may only act on the information internally. Discussion of applicants among programs could be viewed as violating the ethics and rules of the National Residency Matching Program (NRMP) match, thus, programs are reluctant to share information. If errors are identified, corrective action can be implemented either by contacting the Dean's Office of the applicant's medical school or the NRMP. Further fact-checking and any disciplinary action is then the responsibility of the notified organization. In addition, ongoing anti-trust litigation (Jung vs. Association of American Medical Colleges) against the NRMP, residency review committees (RRC's), Accreditation Council for Graduate Medical Education (ACGME), etc. makes a coordinated fact-checking effort by any one agency unattractive.
To date, there have been no published reports of attempts to identify or impact the inaccuracy rate. Our study group was formed with the intent of documenting inaccuracy rates in publications, AOA claims and advanced degrees at multiple centers, and to seek ways of impacting the problem on a national level. This multi-phased study attempts to characterize the magnitude and characteristics of application inaccuracies while later stages will attempt to impact the error rate.
Methods
This is a prospective, multi-center, single-blinded, cohort study of applicants to EM residency at involved residency sites. Each study site obtained permission from their Human Studies Committee (or equivalent) prior to beginning the project. All inclusion criteria, exclusion criteria, endpoints and methods were prospectively defined. Applicants were not made aware of the study prior to investigation. Programs invited to participate were asked to keep the existence of the study confidential from their own students to avoid contamination of the applicant pool.
The applications of all residents applying to participating residency programs were eligible for analysis. Participating sites randomly reviewed 10% of their applicant pool or a minimum of 50 applications. Randomization was achieved by assigning each site a single-digit number. The site then reviewed all ERAS applications whose unique application identifiers ended with that number. If the number of applications at a given site did not meet the minimum of 50 applications, applications with unique identifiers ending with the next higher digit were reviewed. Additionally, each site had the opportunity to review other applications at their discretion without regard to randomization as long as they completed a minimum of 50 randomized reviews. Incomplete applications and those from medical schools not accredited by the United States Medical Licensing Examination were excluded from the analysis.
All study data was recorded on a secure, encrypted, internet database. The only personal identifier entered was the NRMP number. Once this was entered, the database irreversibly converted it to a unique study number by a fixed, but random, formula. This allowed for calculation of inter-rater reliability, while blinding investigators to the identity of the applicants.
Peer-reviewed publications were verified by searching at least two publication databases (Medline, PubMed, etc.) or review of the referenced journal. If either of these methods identified the publication in question, it was considered "verified". Applicants could also be asked to supply a copy of the publication or submission. Publications cited as being "submitted" or "in press" were excluded from analysis. Publication errors were classified as those which claimed improper order of authorship, journal citation, or those which could not be found with the above techniques. Each publication could have more than one error.
AOA status was verified by contact with the national headquarters of Alpha Omega Alpha (computer search of members) or by review of the applicant's Medical Student Performance Evaluation (MSPE) or dean's letter of recommendation.
Advanced degrees were confirmed by the awarding institution. If the degree was earned concurrently with medical training, comments in the MSPE or dean's letter were considered evidence of accuracy.
The assessed endpoints are listed in table 1.
Table 1 Initial results
Number % 95% CI (%)
Peer-reviewed publications claimed 256
Accurate 151 59.0 53.0–65.0
Inaccurate 101 41.0 35.0–47.0
Incorrect author placement 7 6.0 0.0 – 14.6
Unconfirmed 40 34.5 25.8 – 43.2
Errors of publication type 52 44.8 36.2 – 53.4
"Benign" errors 8 7.6 2.5 – 12.7
Other error in peer-reviewed journal 17 14.7 6.0 – 23.4
Claims of AOA membership 47 9.5 6.9 – 12.1
AOA Status verified 47 100 99.1–100
Advanced degree claimed 51 10.3 7.6 – 13.0
Accurate 22 43.1 29.5–56.7
Inaccurate 7 13.7 4.4 – 23.1
Unable to be confirmed 22 43.1 29.5 – 56.7
An error could be classified as "benign," if it was felt that the error could not possibly be due to malicious intent on the part of the applicant and could not benefit the application. Two examples of benign errors include typographical errors and incorrect page or journal numbers in a reference. Judgment about intent was neither made nor implied on any other misrepresentation.
For all non-benign errors, the programs were to inform the applicant's Dean's office and the NRMP of the error, to allow for corrective action. If an applicant was made aware of a concern about his or her credentials, the applicant could verify the claim with appropriate documentation. Notification of the applicant was not required by the study protocol, but was allowed.
Results
A total of 493 applications were screened (28.6% of all applicants submitting at least one ERAS application to an EM residency program). 56 applications (11.4%, 95% CI 8.6–14.2%) contained at least one inaccuracy. Eight of these were judged to be "benign" leaving 48 applications with a non-clerical error (9.8%: 95% CI 7.2–12.4%). The reviews were conducted at nine residencies. Please refer to table 1 for detailed results.
Thirty-three applications (6.7%) were screened by two study centers. Only 2 (6.1%, 95%CI 0.0–14.3) had disagreeing data. In one of these disagreements, one reviewer classifying an error as a "benign," while another reviewer did not record the error. A kappa could not be calculated as the total number of reviewers involved in these 33 cases was not tracked.
The 493 applications referenced 737 publications (mean 1.49 per applicant, range 0–12). Of these, 256 (34.7%, 95%CI 31.3–38.1%) were from peer-reviewed journals. Errors were identified on 105 of these (41.0%, 95%CI 35.0–47.0%).
Fifty-one applicants (10.3%, 95%CI 7.6–13.0%) claimed advanced degrees. Of these, seven (13.7%, 95%CI 4.4–23.1%) were inaccurate. However, 22 (43.1%, 95% CI 29.5–56.7%) claims could not be verified as either accurate or inaccurate. If these are excluded from analysis, 7/29 (24.1%, 95%CI 8.5–39.7%) of claims were inaccurate.
Due to a database flaw, it was not possible to accurately monitor how often the NRMP or Deans' Offices were notified. There was general agreement amongst sites that there had been reluctance to pursue inaccuracies. No site reported a corrected application being submitted by the NRMP. It was not possible to reliably track the actions taken by the deans' offices.
Discussion
This study represents the largest and only multi-centered study to date of inaccuracies on residency applications. Review of 28.6% of the total applicant pool revealed an error on 11.6% of all applications. This is similar to those described in prior reports in EM as well as other specialties [1-8]. If benign errors are excluded, the rate of major errors on applications is nearly 10%, which would place EM among the lowest of reviewed fields.
Our study design was mandated by features of the application system. For example, there is no centralized way to review applications, so a multi-centered trial was necessary, despite the duplication of effort this entails. In addition, the rules governing the behavior of programs during the NRMP match make discussion of applicants between residencies difficult, so programs are often reluctant to share negative information about applicants. This forced our study towards a design that would maximize protection of the applicant, while utilizing the available pathways for reporting of errors and for corrective action. In addition, ongoing legal efforts necessitated legal consultation with a lawyer assigned to the defense of Jung vs. Association of American Medical Colleges in order to ensure protection of the residency programs involved.
A prior retrospective study at one of the authors' institution (EDK) showed similar findings with a few notable differences. We found no erroneous AOA claims while Roellig[7] found an error rate of 35.7%. It is possible that this is due to the differences in included applications: the prior study included osteopathic students while ours did not. As three of the five erroneous AOA claims in that study were from osteopathic students, it is possible that this was due to students attributing this moniker to the American Osteopathic Association.
Secondly, our study found a higher rate of errors on publications (41.0% vs. 21.3% in Roellig[7] and 20.4% in Gurudevan[8]) than previously reported. There are many conceivable sources for this discrepancy. It is possible that the literature search methodology was insufficient in our project, as it relied on review by database, rather than journal review. Before further study is begun, time from publication to database entry must be assessed. In addition, the prior, retrospective, studies were able to assess journals listed as "in press," which was not possible with a prospective design, due to the time interval between acceptance, revision and publication. The inter-rater reliability implies that the methodology provided consistent results, but the disparity between our results and previous results from our field is concerning.
A third difference is that the prior study[7] found 73.3% of advanced degree claims to be verifiable while we were able to verify only 56.9%. Several institutions contacted by the authors had procedures that prevented verification. The previous, retrospective, trial could invest the time necessary to address these procedures, while a prospective, multi-centered trial could not.
Lastly, during the Roellig trial[7], we were able to confirm the AOA status of applicants up to 8 months after the application season. During the current project, the AOA offices were less able to accommodate our inquiries. Should this project become much larger, the AOA staff may become reluctant to verify applicant election to AOA. However, the MSPE's were uniformly helpful in confirming AOA status. It should be noted that we chose not to validate claims of AOA nomination as that process is more loosely defined, and other than each applicant's dean's office, there is not a reliable avenue for verification (the AOA database does not track nominations).
In our study, members of the research program did not reliably take corrective action when an inaccuracy was found. Because of this, we are unable to draw conclusions as to the efficacy of the reviews. If this persists, it may limit the effect of the project. This matter will be strongly emphasized before the next application season.
We did not attempt to assess whether the inaccuracies were intentional though we did attempt to differentiate "benign" and significant errors. For instance, typographical errors may be reasonably assumed to by unintentional whereas an erroneous publication or advanced degree claim is more likely assumed to be intentional. Others inaccuracies may be more difficult to assess. For instance, the misrepresentation of the order of authorship of an article could be assumed to be intentional or simply due to a lack of understanding regarding the significance of the order.
Since the interpretation of motivation is fraught with difficulty, the study did not make any recommendations for how each program used the findings on its applicants. In addition, we did not seek to recommend specific action by any dean's office or the NRMP. Decisions of this nature were outside our scope of research, but may prove to be fertile ground for future investigation.
During the next application season, we intend to expand our study to other sites, with a goal of eventually reviewing the majority of applicants to our specialty. In addition, we will publicize the existence of the trial to the applicants before their applications are submitted. In future studies, we aim to impact the number of erroneous claims. By notifying applicants' institutions and the NRMP regarding inaccuracies and publicizing these efforts, we anticipate a decrease in intentional erroneous claims on applications.
Limitations
Though multi-centered, this study was not large enough to review the entire NRMP EM applicant pool. Nevertheless, by randomly reviewing nearly 29% of all applications we feel the sample size is adequate to draw general conclusions regarding the error rate on these applications. The time-constraints of the match process limits review of some data (manuscripts referenced as "in press," advanced degrees, etc.) Future studies will assess the impact of specific interventions on the error rate.
Conclusion
We report the findings in the first year of a 3-year project aimed at assessing and impacting inaccuracies in residency applications in Emergency Medicine. 11.4% of applications had at least one error and 9.9% had at least one non-clerical error. Publication claims were found to contain errors in 41% of cases. Though at times difficult to verify, erroneous claims on advanced degrees were made in 14 – 24% of applications. All AOA claims were verified.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
All authors read the manuscript and approved the final manuscript.
EK was involved in study concept and design, data acquision, analysis and interpretation, drafting of the manuscript and statistical analysis.
LS was involved in study concept and design, data acquision, analysis and interpretation, and critical review of the manuscript.
LK was involved in study concept and design, data acquision, analysis and interpretation, and critical review of the manuscript.
DH was involved in study concept and design, data acquision, analysis and interpretation, and critical review of the manuscript.
JT was involved in data acquision, analysis and interpretation, and critical review of the manuscript.
CW was involved in study concept and design, data acquision, analysis and interpretation, and critical review of the manuscript.
OS was involved in study concept and design, data acquision, analysis and interpretation, and critical review of the manuscript.
JB was involved in data acquision, analysis and interpretation, and critical review of the manuscript.
DV was involved in study concept and design, data acquision, analysis and interpretation, and critical review of the manuscript.
JF was involved in data acquision, analysis and interpretation, and critical review of the manuscript.
TE was involved in study concept and design, data acquision, analysis and interpretation, and critical review of the manuscript.
GH was involved in study concept and design, data analysis and interpretation, and critical review of the manuscript.
RR was involved in study concept and design, data acquision, analysis and interpretation, and critical review of the manuscript.
AR was involved in data acquision, analysis and interpretation.
KK was involved in study concept and design, data acquision, analysis and interpretation.
EF was involved in study concept and design, statistical expertise and technical support.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors would like to thank Dr. Debra Perina for her advice and guidance with the project, and Jo Len Janes for her assistance with manuscript preparation.
==== Refs
Bilge A Shugerman RP Robertson WO Misrepresentation of authorship by applicants to pediatrics training programs Acad Med 1998 73 532 3 [PMID: 9609867] 9609867
Dale JA Schmitt CM Crosby LA Misrepresentation of research criteria by orthopaedic residency applicants J Bone Joint Surg Am 1999 81 1679 81 [PMID: 10608378] 10608378
Baker DR Jackson VP Misrepresentation of publications by radiology residency applicants Acad Radiol 2000 7 727 9 [PMID: 10987335] 10987335
Sekas G Hutson WR Misrepresentations of academic accomplishments by applicants for gastroenterology fellowships Ann Intern Med 1995 123 38 41 [PMID: 7762913] 7762913
Panicek DM Schwartz LH Dershaw DD Ercolani MC Castellino RA Misrepresentation of publications by applicants for radiology fellowships: is it a problem? AJR Am J Roentgenol 1998 170 577 81 [PMID:9490934] 9490934
Grover M Dharamshi F Goveia C Deception by applicants to family practice residencies Fam Med 2001 33 441 6 11411972
Roellig M Katz ED Inaccuracies on applications for emergency medicine residency training Acad Emerg Med 2004 11 922 924 10.1197/j.aem.2004.04.010
Gurudevan SV Mower WR Misrepresentation of research publications among emergency medicine residency applicants Ann Emerg Med 1996 27 327 330 [PMID: 8599492] 8599492
|
16105178
|
PMC1201142
|
CC BY
|
2021-01-04 16:30:55
|
no
|
BMC Med Educ. 2005 Aug 16; 5:30
|
utf-8
|
BMC Med Educ
| 2,005 |
10.1186/1472-6920-5-30
|
oa_comm
|
==== Front
BMC Med Inform Decis MakBMC Medical Informatics and Decision Making1472-6947BioMed Central London 1472-6947-5-261609553410.1186/1472-6947-5-26Research ArticlePretest probability assessment derived from attribute matching Kline Jeffrey A [email protected] Charles L [email protected] Charles V [email protected] Deborah B [email protected] Judd E [email protected] Craig D [email protected] J Lee [email protected] Department of Emergency Medicine, Carolinas Medical Center, Charlotte, NC, USA2 Computational Biology Program, BreathQuant Medical Systems Inc, Charlotte, NC, USA3 Department of Emergency Medicine, Pennsylvania Hospital, Philadelphia, PA, USA4 Department of Emergency Medicine, University of California at Davis, Sacramento, CA, USA5 Department of Emergency Medicine, Hospital of the University of Pennsylvania, Philadelphia, PA, USA6 Department of Emergency Medicine, Oregon Health & Science University Medical Center, Portland, OR, USA2005 11 8 2005 5 26 26 4 1 2005 11 8 2005 Copyright © 2005 Kline et al; licensee BioMed Central Ltd.2005Kline 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
Pretest probability (PTP) assessment plays a central role in diagnosis. This report compares a novel attribute-matching method to generate a PTP for acute coronary syndrome (ACS). We compare the new method with a validated logistic regression equation (LRE).
Methods
Eight clinical variables (attributes) were chosen by classification and regression tree analysis of a prospectively collected reference database of 14,796 emergency department (ED) patients evaluated for possible ACS. For attribute matching, a computer program identifies patients within the database who have the exact profile defined by clinician input of the eight attributes. The novel method was compared with the LRE for ability to produce PTP estimation <2% in a validation set of 8,120 patients evaluated for possible ACS and did not have ST segment elevation on ECG. 1,061 patients were excluded prior to validation analysis because of ST-segment elevation (713), missing data (77) or being lost to follow-up (271).
Results
In the validation set, attribute matching produced 267 unique PTP estimates [median PTP value 6%, 1st–3rd quartile 1–10%] compared with the LRE, which produced 96 unique PTP estimates [median 24%, 1st–3rd quartile 10–30%]. The areas under the receiver operating characteristic curves were 0.74 (95% CI 0.65 to 0.82) for the attribute matching curve and 0.68 (95% CI 0.62 to 0.77) for LRE.
The attribute matching system categorized 1,670 (24%, 95% CI = 23–25%) patients as having a PTP < 2.0%; 28 developed ACS (1.7% 95% CI = 1.1–2.4%). The LRE categorized 244 (4%, 95% CI = 3–4%) with PTP < 2.0%; four developed ACS (1.6%, 95% CI = 0.4–4.1%).
Conclusion
Attribute matching estimated a very low PTP for ACS in a significantly larger proportion of ED patients compared with a validated LRE.
==== Body
Background
Despite its importance, pretest probability assessment remains a relatively imprecise and inferential process, sometimes referred to as the "doctor's best guess" [1,2]. Previous authors have broadly defined pretest probability as the clinician's estimate of the probability of disease from the patient's words, physical findings, risk factors, and exposures, rendered prior to knowledge of objective test results [3,4]. Presently, the most widely recognized quantitative method of determining pretest probability employs the logistic regression equation [5-10]. The logistic regression equation can be used to estimate the probability of a target disorder in individual patients by assembling and weighing the importance of characteristics found to be predictive of that disorder in an equation. The predictive power of the characteristic is reflected in the magnitude of its coefficient in the equation. The equation can then be solved explicitly to provide a point estimate of probability based upon the sum of the coefficients and the intercept. From the perspective of probability estimation for acute disease, one of the main drawbacks to logistic regression equations is that they seldom output a pretest probability in the very low (0–5%) range. In particular, this very-low pretest probability range is extremely important in the context of evaluating a patient with chest pain in the emergency department, because this is the range where the clinician must decide whether or not to use the resources required for formal testing by a chest pain protocol [11].
We hypothesize that accurate pretest probability assessments can be obtained by matching an individual patient to a group of previously studied patients who shared the same clinical characteristic, and determining the percentage of these previously studied patients who had the outcome of interest. This hypothesis proposes a method of inference that differs substantially from the logistic regression equation. Instead of treating each clinical characteristic as an independent value and adding up the coefficients, we propose a system that forces the probability to be computed from a dependent set of clinical characteristics. In other words, all chosen characteristics of a patient of interest must be matched before a previously studied patient is eligible for pretest probability computation.
In this report, we derive and test a computerized method to estimate the pretest probability of acute disease by matching the clinical characteristics (or attributes) of a patient of interest to an identical profile of attributes shared by a group of patients with known outcomes contained in a large derivation database. The derivation database contains prospectively collected clinical attributes of emergency department patients who were evaluated for acute coronary syndrome and for whom the results of diagnostic testing plus 30-day follow-up are known. First, we used the classification and regression technique to select the variables for attribute matching in the derivation database. We then wrote a computer program to perform the attribute matching procedure. Next, we tested the computerized system in an independent validation population of emergency department patients. The focused question of this work was how often and how accurately the novel system would produce a pretest probability that is below the test threshold. The test threshold represents the point estimate of pretest probability derived from a formula that considers the risks of false positive testing versus the risk of untreated disease [12]. The formula predicts that a patient with a pretest probability below the test threshold treatment will not benefit from further testing.
Methods
Model derivation
The reference database used for model derivation and as the source for pretest probability assessment was drawn from the multicenter internet tracking of acute coronary syndrome (i*trACS) collaborative conducted in 1999–2001 at 7 hospitals in the United States and one in Indonesia [13]. Local Institutional Review Board approval was obtained to collect these data. An emergency department patient was eligible for enrollment into the registry (Access® Microsoft Corporation) when an emergency physician had sufficient suspicion of acute coronary syndrome to order a 12-lead electrocardiogram. Subsequent decision to perform objective testing for acute coronary syndrome was performed at the discretion of a board-certified emergency physician. While the patients were in the emergency department, a study associate (physician or registered nurse) recorded 70 clinical variables, including the standard historical, risk, physical examination, electrocardiographic and laboratory data using a structured clinical report form. We restricted the reference database to patients aged >15 years, with complete data and 30-day follow-up. With these restrictions, the reference database consisted of 14,796 patients with a mean age of 54 ± 16 years, with 35% identified as Caucasian, 51% female, and an overall 15.8% prevalence of acute coronary syndrome.
For the attribute-matching process, it was necessary to truncate the 70 recorded variables to a smaller subset with statistically significant predictive value. This truncation process was accomplished using classification and regression tree analysis (CART®, Salford Systems, San Diego, CA), a form of binary recursive partitioning. Classification and regression tree analysis is a nonparametric method of statistical analysis used to classify observations based on a large number of possible predictive variables, and is well-suited for identifying complex interactions among variables [14]. This methodology has been described previously [15]. A quantitative ranking of the variables (i.e. relative importance) was generated by CART®, based upon the frequency and importance of a given variable in the tree building process. Classification and regression tree analysis identified eight variables as having discriminatory value for the prediction of acute coronary syndrome: 1. Age (divided into four groups, <35, 35–38, 39–50, and >50 years), 2. Gender, 3. Race (white or asian, and nonwhite, non-asian), 4. Patient report or physician observation of sweating with symptoms, 5. Patient report of a prior history of coronary artery disease or myocardial infarction, 6. Chest pain worsened with manual palpation on physical examination, and 7. Electrocardiographic manifestation of ST segment depression > -0.5 mm (-50 μV) in any two leads, 8. T wave inversion > -0.5 mm in any two electrocardiographic leads. These eight variables were then denoted as "attributes" in the matching process. In the matching sequence, age produced a four-way split and the other 7 attributes were binary splitter nodes, allowing a maximum of 4*27 or 512 unique matching permutations (termed attribute profiles).
User interface
To facilitate the process of attribute matching, an author (CLJ) wrote computer source code in visual basic with standard query language subroutines to allow the user to mouse click the eight matching attributes into fields displayed on the screen of a personal computer or handheld personal digital assistant (Figure 1). After the eight attributes were populated and submitted for a new patient, the computer program then extracted from the reference database only those patients with the attribute profile exactly matching that of the patient of interest (Figure 1). Thus, the attribute-matching pretest probability estimate for any new patient was based upon assigning the patient to one of 512 exact, unique attribute profiles. Each profile returned a specific number of patients from the database, and no one patient could be assigned to more than one profile. Figure 2 illustrates the logic the attribute matching process for a hypothetical patient.
Figure 1 Screenshot of the user interface designed to perform the attribute matching process. The application is preloaded with the attributes of a hypothetical patient with chest pain, and shows the calculated pretest probability.
Figure 2 Illustration of the attribute matching algorithm for the hypothetical patient whose profile were input in Figure 1. The patient is a 41 year old African-American woman who presents without diaphoresis, has no history of coronary artery disease and chest pain that is not reproducible with palpation (Chest Tender), and has no ST depression or T wave inversion >0.5 mm (2 leads) on electrocardiography. This match process returned 439 patients from the derivation database. Four of 439 had acute coronary syndrome within 45 days, yielding a pretest probability of 0.9% for this hypothetical patient.
Validation testing
The validation population was a new set of emergency department patients who underwent evaluation for suspected acute coronary syndrome including 30 day outcomes based upon telephone contact and medical chart review. These patients were enrolled at two hospitals, the University of California, Davis and the Hospital of the University of Pennsylvania in Philadelphia from 2001–2002. In addition to other clinical data, all patients included in the validation population had the variables required for the attribute matching process as well as the logistic regression. The choice of diagnostic testing to rule out acute coronary syndrome was at the discretion of a board-certified emergency physician who had unrestricted (24-hour, 7 day) access to the resources of a tertiary academic hospital, including serial biochemical markers, exercise treadmill testing, nuclear imaging, and cardiac catheterization. At both centers, these variables were prospectively collected and recorded using an identical data collection form, which was completed in the emergency department, prior to knowledge of diagnostic testing outcome. Patients with an initial 12-lead electrocardiogram that demonstrated 1 mm (+100 μV) ST segment elevation in two or more leads were analyzed separately. Patients were excluded if data required for either method of pretest probability computation were absent, or if the patient were lost to follow-up.
Logistic regression equation
To provide a benchmark decision rule for comparison of diagnostic utility, we computed the pretest probability estimate from the logistic regression equation described by Selker and colleagues (the acute cardiac ischemia-time insensitive prediction instrument, ACI-TIPI). The variables and coefficients of this method have been published previously. This equation can output a maximum of 128 unique probability estimates [16-18].
Outcome definition
In both the source database and in the validation testing, the definition of acute coronary syndrome included any of the following events, occurring within 30 days of initial evaluation: 1. Acute myocardial infarction, defined by the European Society of Cardiology/American College of Cardiology consensus recommendations [19], 2. Performance of percutaneous coronary balloon angioplasty, with or without intracoronary stent placement, 3. Coronary artery bypass grafting, 4. Death within 30 days after initial evaluation.
Computation of test threshold
The focused question was how often and how accurately each method could produce a low enough pretest probability to potentially forestall a formal protocol to evaluate for acute coronary syndrome. This was done by defining the testing threshold as described by Pauker and Kassirer [see additional file 1] [12]. This estimate was 2.0%. This value is consistent with data in published literature indicating a 0.5% to 3.1% probability of acute coronary syndrome 30 days after negative evaluation in a chest pain evaluation protocol [11,20-25]. For the purpose of evaluating diagnostic accuracy of the attribute matching method, 2.0% is analogous to the positive cutoff in a diagnostic test. Patients with pretest probabilities below 2.0% are hereafter referred to as having "very low risk" of developing acute coronary syndrome. The boundaries defining low-, moderate-, and high-risk categories were based upon pretest probability intervals that are consistent with published practice patterns for admission status of emergency department patients [11,21,24].
Statistical analyses
Confidence intervals for proportions were computed from the exact binomial distribution using Wilson's method (StatsDirect, v 2.2.8) [26]. The median pretest probability estimates between the two methods were compared using the Mann-Whitney U test; Spearman's rank correlation coefficient statistic (rho) was used to test for concordance. The overall diagnostic performance of both methods of pretest probability assessment were determined by rounding the point estimate of pretest probability to the nearest whole percentage and construction of a receiver operating characteristic curve, where all patients with higher pretest probabilities were considered test positive. The overall diagnostic performance was assessed as area under the receiver operating characteristic curve computed by the trapezoidal rule with 95% confidence intervals computed using the Wilcoxon estimate. Gaussian curve fitting was performed in Sigmaplot v. 7.0, SPSS company, Chicago, IL.
Results
The validation population was drawn from 8,120 ED patients who were prospectively evaluated for possible acute coronary syndrome. One thousand sixty-one patients (13%) were not included in the primary analysis for the following reasons: 713 (8.9%) patients had 1 mm or more ST elevation in two or more leads on electrocardiogram; 77 (0.9%) did not have all required data fields for the logistic regression equation; 271 (3.3%) had no ACS-defining event and were lost to 30-day follow-up. The remaining 7,059 patients in the validation population had a mean age of 54 ± 15 years; 57% were female, 38% were Caucasian; 68% percent had a chief complaint of chest pain, but 8% had no chest pain. All patients had two blood cardiac markers drawn separated by 8 hours. Troponin I concentrations measured using the AxSYM174; system (Abbott Laboratories, Abbott Park, IL). 2780 (39%) patients underwent additional cardiac-specific diagnostic testing, including treadmill electrocardiography in 745 (10%), treadmill echocardiography in 105 (2%), or a nuclear cardiac imaging study in 1135 (16%) and cardiac catheterization in 795 (11%). All 7059 patients had 30 day follow-up performed. Six hundred-one of 7,059 patients, or 8.5%, were diagnosed with acute coronary syndrome within 30 days, including 348 acute myocardial infarctions (5%) and 92 deaths. The absence of acute coronary syndrome was established in the remaining 6,458 patients based upon scripted 30-day telephone follow-up with the patient.
The two methods generated contrasting sets of pretest probabilities in the validation population. The attribute matching method utilized 267 of the 512 possible unique pretest probability estimates to yield a median pretest probability value of 5.5% (1st–3rd quartiles: 1.0–10.5%). The median match size (denominator) used to compute the pretest probability from the attribute matching method was 192 patients (1st–3rd quartiles: 94 to 340), and 6,104 of the 7,059 (87%) pretest probability estimates had a match size ≥ 50 patients. The top 10 most frequent matching profiles are shown in Table 1. Four of the top ten profiles generated pretest probabilities less than 2.0%. In contrast, the logistic regression equation produced 96 unique pretest probability estimates, yielding a significantly higher median pretest probability value of 24.0% (1st–3rd quartile 10.0–30.0%) compared with the median from attribute matching (P < 0.001, Mann Whitney U). Figure 3 plots the frequency of each whole percentage pretest probability as a function of the rounded pretest probability estimate value for the validation population and plots an overlying best-fit Gaussian curve for both methods (Y = a*exp [-0.5((X-Xo)/b)2]. With the attribute matching method, the frequency of the pretest probability result tended to be inversely related to the magnitude of the pretest probability estimate with two-thirds of its pretest probability estimates less than 10%. In comparison, logistic regression categorized 54% of the pretest probability estimates between 10 and 30%.
Table 1 Top 10 most frequently matched attribute profiles in the validation population
Patient Count* %PTP Age Sex Race Sweating Coronary Artery Disease Reproducible Chest Pain ST Depression T-Wave Inversion
467 5 >50 F Non-white, non-Asian None No history of CAD None None None
354 1 39–50 F Non-white, non-Asian None No history of CAD None None None
353 6 >50 F White, Asian or other None No history of CAD None None None
272 12 >50 M White, Asian or other None No history of CAD None None None
237 6 >50 F Non-white, non-Asian None No history of CAD None None > -0.5 mm
225 1 39–50 M Non-white, non-Asian None No history of CAD None None None
217 10 >50 M Non-white, non-Asian None No history of CAD None None None
194 4 39–50 M White, Asian or other None No history of CAD None None None
160 1 39–50 F White, Asian or other None No history of CAD None None None
157 0 <35 F Non-white, non-Asian None No history of CAD None None None
Abbreviations: %PTP-Percentage pretest probability rounded to nearest whole number, CAD-Coronary artery disease, M-male, F-female
*Refers to the number of patients in the validation set (out of 7,059) matched to the profile
Figure 3 Plot of the frequency of each discrete pretest probability as a function of the actual pretest probability, estimated on the validation population using two methods. A Gaussian curve was fitted to the attribute matching points (solid line) and the logistic regression points (dashed line). The median pretest probability estimate from attribute matching was 5.5% (1st–3rd quartiles: 1.0–10.5%), versus 24.0% (1st–3rd quartile 10.0–30.0%) for the logistic regression method (P < 0.001, Mann Whitney U). Because of rounding, each point could represent more than one unique match profile.
To determine the overall diagnostic performance of the two methods, the pretest probability estimates were rounded to the nearest whole percentage, and receiver operating characteristic curves were constructed using each whole percentage as cutoff points on the curves (Figure 4). At each whole percentage, patients with higher pretest probability values were considered "test positive." Both methods demonstrated only fair overall discriminatory value; the area under the curve measurements were 0.74 (95% CI 0.65 to 0.82) for the attribute matching curve and 0.68 (95% CI 0.62 to 0.77) for the logistic regression method.
Figure 4 Receiver operating characteristic curves for two methods of pretest probability assessment. The area under the curve measurements were 0.74 (95% CI 0.65 to 0.82) for the attribute matching curve and 0.68 (95% CI 0.62 to 0.77) for the logistic regression method.
We then sought to aggregate the pretest probability estimates into four clinically relevant categories, (Table 2): 1. Very low-risk (<2.0% pretest probability of acute coronary syndrome, the testing threshold subgroup); 2. Low-risk category (2.0 -< 10.0%), 3. Moderate (10%–30%), and 4. High risk (>30%). To generate these categories using the attribute matching method, a stable pretest probability was required to have a match size ≥ 50 patients (maximum very low-risk 95% CI = 0–10%). Table two shows pretest probabilities computed in the validation population using both methods. The two methods demonstrated modest concordance (Spearman's rho = 0.69). The attribute matching method generated a match size >50 in 87% of subjects. Attribute matching yielded a very low pretest probability with an adequate match size in 1,670 or 23.7% of all patients, and among these patients, 28/1,670 or 1.7% (95% CI = 1.1 to2.4%) developed acute coronary syndrome, three of whom died during follow-up (3/1,670= 0.2%, 95% CI = 0 to 0.5%) and one of whom developed an acute myocardial infarction. The logistic regression method produced a very low pretest probability in 244 or 3.6% of all patients, and among these patients, 4/244 or 1.6% (95% CI = 0.4 to 4.1%) developed acute coronary syndrome, one of whom died (1/244 = 0.4%, 95% CI = 0 to 2.2%) and none had myocardial infarction. The sensitivity and specificity of the very low-risk designation for the detection of acute coronary syndrome at 30 days was 95.3% and 25.4% for attribute matching, versus 99.3% and 3.7% for the logistic regression method.
Table 2 Comparison of pretest probability estimates for acute coronary syndrome using two techniques in a validation population.
Category of Pretest Probability
Very Low (<2%) Low (2 to <10%) Moderate (10 to 30%) High (>30%) Indeterminate*
Attribute Matching
Number with ACS† 28 189 250 9 125
Total in category‡ 1670 2953 1453 28 955
Prevalence of ACS 1.7% 6.4% 17.2% 32.1% 13.1%
% of all patients assessed 23.7% 41.8% 20.6% 0.4% 13.5%
Logistic Regression
Number with ACS† 4 59 272 266 NA
Total in category‡ 244 1356 3813 1646 NA
Prevalence of ACS 1.6% 4.4% 7.1% 16.2% NA
% of all patients assessed 3.5% 19.2% 54.0% 23.3% NA
*Estimates from attribute matching with match size <50. Does not apply to logistic regression.
†Total number of patients with ACS = 601
‡Total number patients in validation study = 7059
Abbreviations: ACS acute coronary syndrome.
Figure 5 plots the actual, observed prevalence of acute coronary syndrome as a function of the decile of the predicted probability by each method in the 7,059 patients who did not have ST segment elevation. Neither method closely followed the diagonal line that represents the performance of the hypothetically perfect pretest probability system. However, the graph suggests that compared with attribute matching, ACI-TIPI produced a more reliably linear relationship between the observed versus predicted prevalence of ACS.
Figure 5 Plot of the actual prevalance of acute coronary syndrome as a function of the predicted probability of acute coronary syndrome in 7,059 patients without ST segment elevation. Data are aggregated within deciles. The diagonal line represents the ideal relationship between predicted probability versus observed probability.
Although the purpose of this report was to focus on the low-risk chest pain patients (i.e., the 7,059), we believe it is useful to compare the performance of the attribute matching system and the ACI-TIPI equation in the population that included the 713 patients who had pathological ST segment elevation on the initial 12 lead electrocardiogram. In this higher risk subgroup, 202/713 (28%) patients were ultimately diagnosed with an acute coronary syndrome. The attribute matching system categorized 78 of these 713 patients as having a pretest probability <2.0%, and among these patients, 4 (5.1%) were diagnosed with acute coronary syndrome. The ACI-TIPI system categorized 33 of these 713 patients as having a pretest probability <2.0% and none of these patients had an outcome of acute coronary syndrome. Thus, if ST segment elevation patients were included in the validation set (N = 7,772), attribute matching would have produced a sensitivity of 96% and specificity of 24.6%. The ACI-TIPI equation would have produced a sensitivity of 99.5% and a specificity of 3.9%. The false negative rates would have been 1.8% (95% CI: 1.0 to 3.0%), and 1.5% (0 to 4.0%). These results suggest that attribute matching tends to underestimate the probability of acute coronary syndrome in patients with ST segment elevation on electrocardiography.
The first troponin I concentration, measured in the emergency department, was elevated (>0.3 ng/mL) in 12 of 28 patients who developed acute coronary syndrome after being deemed very low-risk by attribute matching; the initial troponin I concentration was elevated in one of four patients who developed acute coronary syndrome after being deemed very low-risk by logistic regression. The prevalence of acute coronary syndrome in patients with a normal first troponin I concentration plus a very low-risk designation by attribute matching was 16/1,605 or 1.0% (95% CI = 0.5–1.6%).
Both methods produced a stepwise increase in pretest probability with each category. The attribute matching method categorized significantly fewer patients as high-risk (0.4%) compared with the logistic regression technique (23.3%), and the prevalence of acute coronary syndrome was significantly higher in the high-risk group predicted by attribute matching compared with the high-risk group predicted by logistic regression. The sensitivity of the high-risk designation by the logistic regression method was significantly higher (266/601 = 44%) compared with the attribute-matching method (9/601 = 1.5%; 95% CI for difference of 42.5% = 39 to 47%). However, the logistic regression equation categorized 1,380/6,458 or 21% of patients without acute coronary syndrome as high risk compared with 19/6,458 or 0.3% with the attribute-matching method. As a result, the specificity of the high-risk designation by the attribute matching method was significantly higher (99%, 95% CI = 99–100%) than the specificity of the high-risk designation by the logistic regression method (79%, 95% CI = 78–80%).
Discussion
This report introduces a novel method to estimate of the pretest probability of acute disease based upon computer-assisted, database-derived, attribute matching. The system operates by allowing the clinician to input a predefined set of clinical attributes for a subject for whom the pretest probability is desired. When executed, a computer program queries a large patient database, and returns only the patients who share the identical attribute profile as the new patient being evaluated. The proportion of these attribute-matched subjects who had a clinical outcome of interest comprises the point estimate of the pretest probability. We submit that this process conforms to the prose definition of pretest probability from a clinician's perspective: "the probability of the disease of concern in one patient, based upon prior experience with many patients who had similar clinical characteristics as the patient under consideration" [3,4].
In this report, we used an eight-node matching system capable of generating 512 unique pretest probabilities from a 14,796-patient database. We compared this system to an established logistic regression equation to estimate the probability of acute coronary syndrome. Both methods were tested in a validation study of 7,059 emergency department patients without ST segment elevation who were evaluated for possible acute coronary syndrome, and for whom the 30-day outcome was known. The attribute matching system produced significantly more pretest probability assessments that had a significantly lower median value, and categorized patients as very low-risk for acute coronary syndrome six times more often than the logistic regression equation. The rates of acute coronary syndrome-defining outcomes at 30-days in both very low probability groups were virtually identical at 1.7% (95% CI = 1.1–2.4%) with attribute matching, versus 1.6% (95% CI = 0.4–4.1%) with logistic regression. Among patients with an attribute matching-designated very low pretest probability and one normal initial troponin I concentration measured while the patient were in the emergency department, the rate of acute coronary syndrome at 30 days was 1.0% (95% CI = 0.5–1.6%).
A perfect pretest probability instrument would produce exactly two point estimates of pretest probability: 0% and 100%. In our validation population, this imaginary perfect instrument would yield 6,458 pretest probability estimates equal to zero and 601 estimates equal to 100% in the validation population. This would yield exactly two points on Figure 3, one point at 0% on the X-axis and 91.5% on the Y-axis and another at 100% on the X-axis and 8.5% on the Y-axis. Thus, assuming that its estimates were correct, a real instrument that produces a large cluster of very low pretest probabilities and a small cluster of very high pretest probabilities would begin to model the perfect results. Figure 3 shows that the Gaussian curve fitted to the attribute matching probabilities has an inverse function appearance, with the majority of the estimates occurring in the <5% range, whereas the Gaussian curve fitted to the logistic regression probabilities has a quasi-normal appearance, with the peak of the curve occurring at a pretest probability of 15%.
Physicians are well aware that they can be held negligent for failure to test patients for a clinical picture that provokes the slightest thought of acute coronary syndrome. In the present study, 91% of patients in the validation study had no acute coronary syndrome-defining event within 30 days of initial evaluation (including the "soft" endpoint of revascularization). In a 1997 multicenter study, Graff and colleagues found that only 2.5% of patients evaluated by a chest pain protocol were diagnosed with acute myocardial infarction [27]. In our experience, this "rule-in rate" is declining. We submit that the primary motivation for evaluating many low-risk patients in a chest pain evaluation unit are the clinicians' perceptions of reduced medicolegal risk and the patients' perceptions of increased safety. But approximately 25–30% of these patients will have a false positive or indeterminate provocative test – a result that usually mandates hospital admission, and can lead to performance of coronary angiography negative for coronary artery disease [11,20,22,24,27]. A potential application of the present system would be use of the combination of a pretest probability <2.0%, and one negative biomarker of cardiac ischemia or necrosis, in conjunction with the patient's risk tolerance to prevent unnecessary chest pain protocol evaluation [28]. Reports in the lay and medical literature underscore the desire of patients to become informed and active participants in medical decision-making [29].
The present study does not purport clinical utility of pretest probability derived from attribute matching at this stage. We tested attribute matching against a logistic regression equation that contains different variables, and was not designed to rule out ACS in very low risk patients. We did not test the output of attribute matching against a logistic regression equation that used the exact same variables so this report does not allow a head-to-head comparison of the two methodologies. The present work only tested the accuracy of the derived system by a secondary analysis of prospectively obtained data, albeit a large, two-center sample from opposite sides of the US. The next question to answer is whether the findings are valid in other populations, and if the computerized system adds any value to the implicit estimate of probability from clinicians with variable experience. Attribute matching should also be tested against other published research methods, including other computerized models, Bayesian and neural network systems [6,30]. The heuristic aspects of the attribute matching system warrant more research to define them quantitatively. We recognize that as the number and complexity of the input attributes increases, this will create a more specific and potentially more accurate clinical profile, but at a cost of reduced match size if the reference database remains the same size. In theory, the ideal attribute matching system would allow a very detailed clinical profile to be matched against a tremendously large reference database. In the present work, we used an eight-node attribute profile and a 14,796 patient database that produced a match size of 50 or more in about 87% of patients tested in the validation phase. This match size requirement appears to have resulted in relatively reliable pretest probability assessments in the validation population based upon the categorization data in Table 2. The optimal stoichiometry between the number of matching nodes that can be used to create an accurate pretest probability versus the database size remains uncertain. It could be hypothesized that a larger database and more complex attribute profile would produce a more accurate pretest probability assessment. Finally, the model only had good utility in patients with very low risk for acute coronary syndrome and cannot be used in patients with ST segment elevation. We believe that attribute-matching should be similarly tested for other clinical conundrums, including the evaluation of pulmonary embolism.
List of abbreviations
ACS: Acute Coronary Syndrome
CI: Confidence Interval
ED: Emergency Department
i*trACS: Internet Tracking of Acute Coronary Syndrome
LRE: Logistic Regression Equation
PTP: Pretest Probability
Competing interests
JAK and CLJ are both partially employed by BreathQuant Medical Systems, which now markets the ACS software described in this paper.
CDN serves as consultant for BreathQuant on a study that is separate from and unrelated to this paper's focus.
JAK (VP-Medical Director), CLJ (VP-Product Development), and CVP (Scientific Advisor) are all members of BreathQuant's board.
Authors' contributions
JAK conceived and designed the study, acquired data, assisted in the analysis and interpretation of this data and drafted the paper.
CLJ contributed to study design and made critical revisions to the paper.
CVP acquired data, assisted with the analysis and interpretation of this data and made critical revisions to the paper.
DBD acquired data, assisted with the analysis and interpretation of this data and made critical revisions to the paper.
JEH acquired data, assisted with the analysis and interpretation of this data and made critical revisions to the paper.
CDN acquired data, assisted with the analysis and interpretation of this data and made critical revisions to the paper.
JLG assisted with the analysis and interpretation of this data and made critical revisions to the paper.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Test threshold for ACS. Explanation of the formula used to compute the test threshold to rule out acute coronary syndrome.
Click here for file
==== Refs
Dawson NV Arkes HR Systematic errors in medical decision making: judgment limitations. [Review] [18 refs] Journal of General Internal Medicine 1987 2 183 187 3295150
Lilford RJ Pauker SG Braunholtz DA Chard J Decision analysis and the implementation of research findings. [Review] [22 refs] BMJ 1998 317 405 409 9694762
Richardson WS Wilson MC Guyatt GH Cook DJ Nishikawa J Users' guides to the medical literature: XV. How to use an article about disease probability for differential diagnosis. Evidence-Based Medicine Working Group JAMA 1999 281 1214 1219 10199432 10.1001/jama.281.13.1214
Diamond GA Forrester JS Analysis of probability as an aid in the clinical diagnosis of coronary-artery disease New England Journal of Medicine 1979 300 1350 1358 440357
Richardson WS Where do pretest probabilities come from? ACP Journal Club 1999 4 68 69
Goldman L Cook EF Brand DA Lee TH Rouan GW Weisberg MC Acampora D Stasiulewicz C Walshon J Terranova G A computer protocol to predict myocardial infarction in emergency department patients with chest pain New England Journal of Medicine 1988 318 797 803 3280998
Reilly BM Evans AT Schaider JJ Das K Calvin JE Moran LA Roberts RR Martinez E Impact of a clinical decision rule on hospital triage of patients with suspected acute cardiac ischemia in the emergency department JAMA 2002 288 342 350 12117399 10.1001/jama.288.3.342
Baxt WG Shofer FS Sites FD Hollander JE A neural network aid for the early diagnosis of cardiac ischemia in patients presenting to the emergency department with chest pain Ann Em Med 2002 40 575 583 10.1067/mem.2002.129171
Pozen MW D'Agostino RB Selker HP Sytkowski PA Hood WBJ A predictive instrument to improve coronary-care-unit admission practices in acute ischemic heart disease. A prospective multicenter clinical trial New England Journal of Medicine 1984 310 1273 1278 6371525
Antman EM Cohen M Bernink PJ McCabe CH Horacek T Papuchis G Mautner B Corbalan R Radley D Braunwald E The TIMI risk score for unstable angina/non-ST elevation MI: A method for prognostication and therapeutic decision making.[comment] JAMA 2000 284 835 842 10938172 10.1001/jama.284.7.835
Storrow AB Gibler WB Chest pain centers: diagnosis of acute coronary syndromes Ann Emerg Med 2000 35 449 461 10783407
Pauker SG Kassirer JP The threshold approach to clinical decision making New England Journal of Medicine 1980 302 1109 1117 7366635
Blomkalns AL Lindsell CJ Chandra A Osterlund ME Gibler WB Pollack CV Tiffany BR Hollander JE Hoekstra JW Can electrocardiographic criteria predict adverse cardiac events and positive cardiac markers? Academic Emergency Medicine 2003 10 205 210 12615583 10.1197/aemj.10.3.205
Breiman L Friedman JH Stone CJ Classification and regression trees 1984 New York, Chapman & Hall (Wadsworth, Inc.)
Kline JA Hernandez J Newgard CD Cowles DN Jackson RE Courtney DM Use of pulse oximetry to predict in-hospital complications in normotensive patients with pulmonary embolism American Journal of Medicine 2003 115 203 208 12935827 10.1016/S0002-9343(03)00328-0
Selker HP Griffith JL D'Agostino RB A tool for judging coronary care unit admission appropriateness, valid for both real-time and retrospective use. A time-Insensitive predictive instrument (TIPI) for acute cardiac ischemia: A multicenter study Medical Care 1991 29 610 626 2072767
Lau J Ioannidis JP Balk EM Milch C Terrin N Chew PW Salem D Diagnosing acute cardiac ischemia in the emergency department: a systematic review of the accuracy and clinical effect of current technologies Ann Emerg Med 2001 37 453 460 11326181 10.1067/mem.2001.114903
Selker HP Beshansky JR Griffith JL Aufderheide TP Ballin DS Bernard SA Crespo SG Feldman JA Fish SS Gibler WB Kiez DA McNutt RA Moulton AW Ornato JP Podrid PJ Pope JH Salem DN Sayre MR Woolard RH Use of the acute cardiac ischemia time-insensitive predictive instrument (ACI-TIPI) to assist with triage of patients with chest pain or other symptoms suggestive of acute cardiac ischemia. A multicenter, controlled clinical trial Annals of Internal Medicine 1998 129 845 855 9867725
Apple FS Wu AH Jaffe AS European Society of Cardiology and American College of Cardiology guidelines for redefinition of myocardial infarction: how to use existing assays clinically and for clinical trials.[comment] American Heart Journal 2002 144 981 986 12486421 10.1067/mhj.2002.124048
Manini AF Gisondi MA van V Schreiber DH Adverse cardiac events in emergency department patients with chest pain six months after a negative inpatient evaluation for acute coronary syndrome Academic Emergency Medicine 2002 9 896 902 12208678 10.1197/aemj.9.9.896
Tatum JL Jesse RL Kontos MC Nicholson CS Schmidt KL Roberts CS Ornato JP Comprehensive strategy for the evaluation and triage of the chest pain patient Ann Emerg Med 1997 29 116 125 8998090
Diercks DB Gibler WB Liu T Sayre MR Storrow AB Identification of patients at risk by graded exercise testing in an emergency department chest pain center American Journal of Cardiology 2000 86 289 292 10922435 10.1016/S0002-9149(00)00916-4
Newby LK Storrow AB Gibler WB Garvey JL Tucker JF Kaplan AL Schreiber DH Tuttle RH McNulty SE Ohman EM Bedside multimarker testing for risk stratification in chest pain units: The chest pain evaluation by creatine kinase-MB, myoglobin, and troponin I (CHECKMATE) study.[comment] Circulation 2001 103 1832 1837 11294799
Nichol G Walls R Goldman L Pearson S Hartley LH Antman E Stockman M Teich JM Cannon CP Johnson PA Kuntz KM Lee TH A critical pathway for management of patients with acute chest pain who are at low risk for myocardial ischemia: recommendations and potential impact.[comment] Annals of Internal Medicine 1997 127 996 1005 9412306
Farkouh ME Smars PA Reeder GS Zinsmeister AR Evans RW Meloy TD Kopecky SL Allen M Allison TG Gibbons RJ Gabriel SE A clinical trial of a chest-pain observation unit for patients with unstable angina. Chest Pain Evaluation in the Emergency Room (CHEER) Investigators.[comment] New England Journal of Medicine 1998 339 1882 1888 9862943 10.1056/NEJM199812243392603
Gardner MJ Altman PG Calculating confidence intervals for proportions and their differences Statistics With Confidence 1989 1 London, British Medical Journal 28 33
Graff LG Dallara J Ross MA Joseph AJ Itzcovitz J Andelman RP Emerman C Turbiner S Espinosa JA Severance H Impact on the care of the emergency department chest pain patient from the chest pain evaluation registry (CHEPER) study American Journal of Cardiology 1997 80 563 568 9294982 10.1016/S0002-9149(97)00422-0
Lange RA Hillis LD Cardiology patient pages. Diagnostic cardiac catheterization Circulation 2003 107 e111 e113 12732596 10.1161/01.CIR.0000070982.94049.A2
Braddock CH Edwards KA Hasenberg NM Laidley TL Levinson W Informed decision making in outpatient practice: time to get back to basics.[comment] JAMA 1999 282 2313 2320 10612318 10.1001/jama.282.24.2313
Baxt WG Shofer FS Sites FD Hollander JE A neural computational aid to the diagnosis of acute myocardial infarction Ann Emerg Med 2002 39 366 373 11919522 10.1067/mem.2002.122705
|
16095534
|
PMC1201143
|
CC BY
|
2021-01-04 23:52:11
|
no
|
BMC Med Inform Decis Mak. 2005 Aug 11; 5:26
|
utf-8
|
BMC Med Inform Decis Mak
| 2,005 |
10.1186/1472-6947-5-26
|
oa_comm
|
==== Front
BMC Med Inform Decis MakBMC Medical Informatics and Decision Making1472-6947BioMed Central London 1472-6947-5-271610517710.1186/1472-6947-5-27DatabasePHSkb: A knowledgebase to support notifiable disease surveillance Doyle Timothy J [email protected] Haobo [email protected] Samuel L [email protected] Richard S [email protected] Division of Public Health Surveillance and Informatics, Epidemiology Program Office, Centers for Disease Control and Prevention (CDC); Atlanta, Georgia, USA2 Bureau of Epidemiology, Florida Department of Health; Tallahassee, Florida, USA2005 16 8 2005 5 27 27 10 1 2005 16 8 2005 Copyright © 2005 Doyle et al; licensee BioMed Central Ltd.2005Doyle 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
Notifiable disease surveillance in the United States is predominantly a passive process that is often limited by poor timeliness and low sensitivity. Interoperable tools are needed that interact more seamlessly with existing clinical and laboratory data to improve notifiable disease surveillance.
Description
The Public Health Surveillance Knowledgebase (PHSkb™) is a computer database designed to provide quick, easy access to domain knowledge regarding notifiable diseases and conditions in the United States. The database was developed using Protégé ontology and knowledgebase editing software. Data regarding the notifiable disease domain were collected via a comprehensive review of state health department websites and integrated with other information used to support the National Notifiable Diseases Surveillance System (NNDSS). Domain concepts were harmonized, wherever possible, to existing vocabulary standards. The knowledgebase can be used: 1) as the basis for a controlled vocabulary of reportable conditions needed for data aggregation in public health surveillance systems; 2) to provide queriable domain knowledge for public health surveillance partners; 3) to facilitate more automated case detection and surveillance decision support as a reusable component in an architecture for intelligent clinical, laboratory, and public health surveillance information systems.
Conclusions
The PHSkb provides an extensible, interoperable system architecture component to support notifiable disease surveillance. Further development and testing of this resource is needed.
==== Body
Background
In the United States, notifiable disease reporting is mandated by state and local regulations. These regulations require medical providers and laboratories to notify state and local public health authorities of persons diagnosed with a reportable condition [1]. A reportable condition is one for which regular, frequent, and timely information regarding individual cases is considered necessary for the prevention and control of the disease. Each state determines which conditions are reportable within its jurisdiction. The Council of State and Territorial Epidemiologists (CSTE) determines which diseases all states will voluntarily report nationally to the federal Centers for Disease Control and Prevention (CDC) [2]. Each year, CDC publishes a summary of notifiable disease activity in the United States [3].
Surveillance case definitions provide uniform criteria for reporting notifiable diseases. Case definitions for nationally notifiable diseases have previously been published in printed copy [4] and are currently maintained on the CDC website [5]. The format of case definitions varies somewhat by condition but often contains information on clinical criteria, laboratory criteria, case classification categories, and criteria for classification. The case definitions are written in text to provide guidance to health providers and surveillance epidemiologists when determining whether or not an individual case meets the criteria for reporting. The case definitions are not derived from a formalized information model and are typically not developed for computational purposes. Their content, however, includes numerous terms found in standard medical vocabularies.
Routine notifiable disease surveillance often suffers from incomplete reporting [6] and poor timeliness [7,8]. New threats of bioterrorism have resulted in increased pressures to improve the sensitivity and timeliness of routine disease surveillance, particularly through the use of improved electronic data interchange. Pilot studies have demonstrated improved surveillance sensitivity and timeliness through electronic reporting of laboratory findings by laboratories to public health agencies [9,10]. Other electronic data (e.g., coded discharge diagnoses or pharmacy dispensing data) have also been used to improve the sensitivity and timeliness of routine notifiable disease surveillance [11,12]. Implementation of these methods often depends on the existence of tables that relate the coded laboratory or clinical findings to the notifiable conditions under surveillance. The systematized nomenclature of medicine (SNOMED) and logical observation identifier names and codes (LOINC) have been identified as important vocabulary standards for constructing these tables [13].
Previous efforts to separately maintain information regarding the diseases that are reportable, the content of case definitions, and mapping tables of coded observations to notifiable diseases, have resulted in a proliferation of disparate, unintegrated spread sheets and documents that are used to support notifiable disease surveillance activities. In 2003, we began to develop a database to integrate these information sources – the Public Health Surveillance knowledgebase (PHSkb™). The long-term goal of the PHSkb is to provide a resource to improve the sensitivity, timeliness, and quality of surveillance data through improved electronic data interchange. To achieve this goal, the notifiable disease domain is expressed by using methods of ontology development and knowledge representation, combined with integration of national vocabulary standards that cover the domain. Whereas these methods of ontology development and knowledge representation have been applied to health information retrieval, clinical information systems, and clinical decision support, their application to public health disease surveillance systems is less established. This paper describes the initial creation of the PHSkb. Further field testing will be needed, however, to determine the impact of such methods on surveillance sensitivity, timeliness, and data quality.
Construction and content
Data collection
The knowledgebase scope includes diseases, conditions, or other events that are reportable in one or more reporting jurisdictions in the United States. The reporting jurisdictions (n = 52) are those states or cities that report data weekly to CDC via the National Notifiable Diseases Surveillance System (NNDSS), which includes the 50 U.S. states, New York City, and the District of Columbia. Reportable conditions in each jurisdiction were ascertained from the health department website for each jurisdiction [see Additional file 1].
While all jurisdictions provided a list of reportable conditions on their website, in many instances this information was not prominently displayed and required substantial effort to identify. The median time interval needed to identify reporting requirements by navigating the website was approximately one minute, ranging from less than 15 seconds to more than eight minutes. Identifying requirements often required understanding of the organizational structure of the agency and knowledge of which bureau or division was responsible for posting such information to the website. During a 6-month interval between the time the data were first collected and later reconfirmed, 16 (31%) of 52 jurisdictions had updated or otherwise modified their reportable conditions list, suggesting a dynamic information domain.
Domain ontology
A knowledge representation model was created for the notifiable disease domain (Figure 1) that depicts the major root concepts in the ontology, their attributes, and relationships between concepts. By necessity, the figure is an oversimplification and does not account for the full hierarchical classification of concepts within the knowledgebase, or identify all the links between concepts. The reader is referred to the PHSkb for the full knowledge representation.
Figure 1 Knowledge representation model of notifiable condition domain.
Notifiable conditions include not only single instances of disease, but also infectious agents, substances, procedures, and findings. Reportable findings can include individual findings (e.g., a disease carrier state) or population findings (e.g., an outbreak or cluster of illness). Reportable procedures include concepts such as the administration of medications or prophylaxis specific for a particular disease. Reportable infectious agents include those microbial isolates which laboratories are required to report, independent of clinical illness. For reportable diseases, 18 sub-categories of disease were identified among jurisdiction reporting requirements. However, the majority of reportable diseases fall within the category of infectious disease. For this category, additional attribute knowledge was included, such as the causative agent, insect vector, and associated immunoprophylaxis.
Additional root classes were included for the jurisdictions where events are reportable, and the various terminology standards that cover the domain. Information specified in the surveillance case definition was included as attribute data for each disease and is described in more detail in the content section below.
Class hierarchy
Instances of the major classes were classified hierarchically, focusing initially on the disease class. The infectious agent, vector, and substance classes were constructed by including all organisms or substances with a causal or other obvious association to concepts in the reportable disease class. Semantic heterogeneity of notifiable events between different reporting jurisdictions was resolved by harmonizing term variants to a standard concept when possible. SNOMED CT (release 1/2003, with the Clue browser) was used extensively to construct the class hierarchy and to harmonize term variations between jurisdictions [14]. In some instances, individual jurisdiction or national data aggregation needs required modification or extension of the concept hierarchy in SNOMED CT. The full list of reportable diseases (n = 373), infectious agents (n = 174), substances (n = 48), findings (n = 58), and procedures (n = 3) is presented in this report along with the number of jurisdictions in which each event is reportable [see Additional files 2 thru 6]. The number of jurisdictions should be interpreted with caution, however, because a particular reportable event might be referred to as a broader or more narrowed concept in another jurisdiction. Furthermore, the reportable condition lists available on the jurisdiction websites, at the time of data collection, might not be the most current reporting requirements given to health providers. The tables are intended to provide a quick glimpse of the breadth of the notifiable condition domain; however, further work is needed to validate the information contained in the knowledgebase with state surveillance partners.
Additional concepts, not reportable in any jurisdiction, were included in the PHSkb to construct the hierarchy for each root class. In the disease class, for example, additional concepts were added either because the condition is a parent-level concept in the class hierarchy under which a reportable disease exists, or the condition is a clinical sub form with distinct clinical symptoms and findings and is already represented by a broader parent concept that is notifiable (i.e., pneumonic and bubonic plague are clinical forms of the notifiable disease plague). More than 570 additional concepts were added to the knowledgebase to construct the class hierarchies.
Content and relation to vocabulary standards
Concepts in the class hierarchies were mapped to SNOMED-CT and summary results are presented for events reportable in at least one jurisdiction (Table 1). In general, SNOMED-CT provides extensive domain coverage, particularly for the reportable diseases, infectious agents, and substances. Reportable population findings (e.g. disease outbreaks) are not covered as thoroughly in SNOMED. Concepts in the disease class were further mapped to other vocabulary standards. Approximately 76% of concepts in the disease class were in ICD-9 and 48% in ICD-10; 77% had a UMLS concept unique identifier (CUI). Therefore, SNOMED-CT provided better domain coverage than these other coding standards. For the jurisdiction class, Federal Information Processing Standards (FIPS) codes were used to identify each jurisdiction [15].
Table 1 Domain content coverage of reportable events with SNOMED CT.
Semantic class SNOMED CT Content coverage
Disease 323 of 373 (87%)
Infectious agent 168 of 174 (97%)
Substance 45 of 48 (94%)
Reportable findings 11 of 58 (19%)
Reportable procedures 1 of 3 (33%)
Additional attribute data for each class instance were then entered into the database. When a reportable disease, infectious agent, or substance has been identified as a possible bioterrorism agent or condition [16] the category of BT agent (A, B, or C) was specified in the knowledgebase. Knowledge pertaining to the disease class is the most fully developed in the PHSkb. It contains data on whether or not the disease is nationally notifiable, the year it first became notifiable, and the CDC-assigned code used by jurisdictions to report cases to CDC. For each infectious disease, the associated infectious agents, insect vectors, and incubation periods were identified (when not obvious) either from the surveillance case definition [5] or from the Control of Communicable Diseases Manual [17]. Associated immunoprophylaxes were coded by using the CVX-Vaccines Administered codes developed by CDC for use in immunization registries and adopted by HL7 as a standard code set [18]. Approximately 100 CVX-coded vaccines were associated with 68 reportable diseases in the knowledgebase. The surveillance case definition text was used to populate the clinical and laboratory criteria [5]. Terms from the case definition were parsed and mapped to SNOMED-CT concepts to populate the symptoms & findings, associated procedure, and associated medication slots. A total of 344 findings, 46 procedures, and 19 medications from SNOMED-CT were mapped to terms included in, or implied by, the case definition text.
Finally, the existing table created to support electronic laboratory reporting, that relates LOINC coded laboratory tests to notifiable diseases [19] was imported into the database and used to populate the associated laboratory test field for each disease. We did not attempt to precisely map the laboratory criteria from case definitions to LOINC. Experience has demonstrated that fully specified LOINC terms are often substantially more granular than criteria specified in surveillance case definitions. In addition, the lack of a hierarchical representation within LOINC causes mapping to the less granular case definition criteria to be difficult [20]. However, more that 3,500 LOINC codes have direct relevance as diagnostic tests for reportable diseases and are included in the knowledgebase.
Development environment
The model was instantiated by using the Protégé-2000 ontology and knowledgebase editing software (version 1.9)[21]. Protégé is an open-source, Java tool that provides an extensible architecture for creating customized knowledge-based applications. Multiple plug-ins have been developed to extend the functionality of Protégé, including various inference and reasoning tools. We used JESS (Java Expert System Shell) tab plug-in (version 1.1) to query the PHSkb[21].
Utility
The PHSkb has at least three potential uses: 1) providing a framework for the development and maintenance of a controlled vocabulary for reportable events of public health importance; 2) providing convenient, queriable domain knowledge to surveillance epidemiologists, data reporters, and others; and 3) providing a reusable domain knowledge component for intelligent surveillance information system architectures. Each of these broad areas of utility is discussed further in this report.
Development and maintenance of a controlled vocabulary for reportable conditions
To support the activities of the NNDSS, CDC maintains an authoritative code set for use when jurisdictions report notifiable diseases to CDC. This code set is maintained as a spreadsheet and distributed annually to states when changes occur [22]. This authoritative code set is not concept-based, does not express hierarchical relationships between terms, and focuses predominantly on those conditions that are nationally notifiable.
Our review of the jurisdiction specific websites [see Additional file 1] identified extensive semantic heterogeneity between jurisdictions when referring to reportable conditions, particularly for those conditions that are not nationally notifiable. Conditions reported locally within a jurisdiction that are not nationally notifiable are usually assigned a code by each jurisdiction for use within their system. Without a hierarchical representation, it is difficult to aggregate data across multiple jurisdictions for these diseases, because different jurisdictions use different levels of granularity when defining their own disease reporting requirements. Therefore, aggregation across jurisdictions requires extensive mapping and harmonization of jurisdiction-specific extensions to the code set for notifiable conditions. Having the notifiable disease domain organized across jurisdictions in a hierarchical classification will facilitate data aggregation and electronic data interchange across jurisdictions and between parties within jurisdictions.
CDC-assigned reportable disease codes exist for approximately 128 (34%) of the 373 diseases reportable in at least one jurisdiction. For the remaining diseases reportable in at least one jurisdiction but without a standard name or code (i.e. non-nationally notifiable), non-standard codes are assigned by each jurisdiction, making cross jurisdiction data aggregation difficult or impossible and resulting in a fragmented national surveillance approach. The PHSkb attempts to move from an authoritative code set characterized by incomplete domain coverage to managing the notifiable condition domain as a controlled vocabulary. When fully implemented, features of PHSkb would include harmonizing term variants across jurisdictions, assigning nationally standard codes for locally reportable events, expressing the hierarchical relation between notifiable conditions, and maintaining mappings between notifiable conditions and concept equivalents within other widely used coding standards.
Queriable domain knowledge
The PHSkb provides convenient, queriable domain knowledge for surveillance epidemiologists and other public health partners. The Protégé software has a built-in query development utility that allows users to construct standard queries and save them to a query library. Several standard queries of the PHSkb have been created and saved in the query library. Examples of such queries include the following.
• In what jurisdictions is a particular disease notifiable?
• What are the reportable conditions in a particular jurisdiction?
• What disease is caused by a particular microorganism?
• What diseases have a particular constellation of symptoms mentioned in their surveillance case definition?
• What diseases are transmitted by a particular insect vector?
• What diseases are associated with a particular laboratory test or finding?
In addition, custom queries can be developed using the inference tools in Protégé (e.g., the JESS tab).
A web interface to the PHSkb is needed to provide broad, web-based, public access to the query functions of the knowledgebase. In the interim, while the web interface is being developed, we have received numerous queries as part of our oversight responsibilities for the NNDSS and have used the PHSkb to respond directly to these domain-specific queries. If adequate, ongoing maintenance of the PHSkb exists, the effort needed to access jurisdiction specific requirements will be reduced by the availability of a central, queriable knowledgebase integrating domain information derived from > 50 different agencies. Such an integrated, centrally-accessed database could be particularly useful to data providers who report to multiple jurisdictions (e.g., large reference laboratories or regional and national provider networks serving communities in different states).
Reusable architecture component
The PHSkb can function as a reusable component in an architecture for intelligent public health surveillance and clinical information systems [23]. Private software developers have referenced the reporting requirements specified in the PHSkb when developing public health surveillance information systems, and it has been embedded in the architecture of at least one state-based system under development [24]. For users of this state web-based system, the class hierarchies could be navigated and related knowledge could be viewed within the system. When a particular disease was selected for reporting, the system was able to query the knowledgebase and dynamic data entry screens were generated, based on its content. In this way, domain knowledge can be maintained separately by subject matter experts without requiring extensive hard-coding changes to the surveillance software resulting from emerging public health threats or rapidly evolving domain knowledge for a particular disease.
The PHSkb could also be used as an inference engine to identify reportable events from one or more observations. Previous studies have demonstrated that knowledge-based patient screening methods can lead to earlier diagnosis of rare infections, thus improving both clinical patient management and disease surveillance [25]. Two-dimensional tables are currently used to infer cases of reportable disease from LOINC-coded laboratory test observations. In the future, the PHSkb might be able to provide more robust inference capability by integrating 1) laboratory observations with clinical findings and exposure criteria, 2) a hierarchical class structure, and 3) information on jurisdiction-specific requirements. Future development might also address the logic contained in case definitions. Before additional extensions are included, however, further testing is needed regarding the current inference capability of the PHSkb when interacting with laboratory and clinical data streams.
Discussion
The historical paradigm of notifiable disease surveillance is based on passive reporting of notifiable events from health-care providers to public health agencies. This paradigm relies on public health agencies informing providers of what should be reported (i.e., reportable disease list), a common understanding of the case reporting criteria (i.e., surveillance case definition) and a method for sending and receiving the reports (i.e. telephone hotline, fax, mailed morbidity reports, or web-based reports). Given that there is often minimal reward for reporting or punitive consequences for not reporting, it is not surprising that this passive, unautomated surveillance paradigm often results in poor surveillance sensitivity and timeliness. The information technology and internet revolution during the previous decade has created new opportunities to alter this paradigm and use pre-existing electronic health data to improve the sensitivity and timeliness of surveillance data while reducing the reporting burden on individual providers.
In the same way that clinical practice guidelines have been implemented into rule-based expert systems featuring clinical decision support, the surveillance case definitions provide a basis for developing a rules-based decision support capability for the public health surveillance function [26]. Reusable, extensible domain knowledge components such as the PHSkb are a necessary, but not sufficient, component for fulfilling the paradigm shift from passive disease reporting to efficient, comprehensive, automated electronic data interchange.
Multiple barriers remain, however, for achieving this paradigm shift. First, current reportable disease requirements are unnecessarily fragmented by jurisdiction. The current variability between states regarding reporting requirements makes it cumbersome to develop tools that are generalizable across jurisdictions to assist providers in meeting local reporting requirements. Second, current surveillance case definitions are not based on a uniform information model, are not written for automated interpretation, and often contain ambiguous or conflicting logic. Efforts to retrofit existing case definitions to a standard information model are necessarily awkward and difficult. However, the current definitions do represent an important starting point for standard public health surveillance guidelines, analogous to clinical practice guidelines that often require months or years of consensus-building to create. Third, the use of electronic medical records (EMR) is still not widespread and the cost of implementing EMRs is often a barrier. Finally, more robust clinical vocabulary standards such as SNOMED-CT are not yet widely used in health-care settings. In addition, much of the clinical information contained in patient charts is text that is not electronic and not coded to any vocabulary standard. Further advances are needed in the area of natural language processing and automated methods for converting text data to electronic vocabulary standards.
Despite these important barriers, reusable domain knowledge components such as the PHSkb hold promise for improved interoperability between surveillance information systems and their clinical and laboratory counterparts, through use of a set of integrated content standards for disease surveillance. Field testing of the PHSkb is needed, however, to determine its impact on surveillance metrics such as sensitivity, timeliness, and data quality. Future development of the PHSkb should focus on 1) validating its content with state surveillance partners and subject matter experts, 2) additional development of a queriable web interface to provide broad access to the knowledgebase content and query functions, 3) testing the inference capabilities of the knowledgebase when interacting with clinical and laboratory data streams, and 4) development of an organizational infrastructure and protocols for ongoing maintenance, versioning, and distribution of the knowledgebase. Revisions to the content and structure of the PHSkb should be guided by user feedback and the results of field testing.
Conclusions
The Public Health Surveillance knowledgebase (PHSkb) provides integrated, extensible domain knowledge regarding notifiable conditions in the United States. It can be used by public health professionals and information system developers to improve the quality of disease surveillance data.
Availability and requirements
Additional work is needed to validate the PHSkb contents with surveillance partners and subject matter experts. Therefore, the original source documents used to populate the knowledgebase should be regarded as the most definitive source of information regarding the notifiable disease domain. However, the PHSkb is provisionally available for access to demonstrate the methods used and to generate discussion among public health surveillance partners and system developers. The knowledgebase is available for download from the CDC ftp server . The database can be downloaded as three java files suitable for use with the Protégé software (version 1.9 or higher). The total file size is approximately 3 megabytes.
Abbreviations
CDC Centers for Disease Control and Prevention
CSTE Council of State and Territorial Epidemiologists
CUI concept unique identifier
Dxplain clinical decision support software product
EMR electronic medical record
FIPS Federal Information Processing Standards
FTP File transfer protocol
ICD International Classification of Diseases
JESS Java Expert System Shell
LOINC Logical Observation Identifier Names and Codes
NNDSS National Notifiable Diseases Surveillance System
PHSkb™ Public Health Surveillance Knowledgebase
SNOMED-CT Systematized Nomenclature of Medicine-Clinical Terms
UMLS unified medical language system
URL uniform resource locator
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
TD conceived of project idea, built knowledgebase, and drafted manuscript text and tables. HM assisted in knowledgebase development, imported data tables, developed custom data queries, contributed to manuscript text, and produced manuscript tables and figures. SG assisted with conceptualization of project, provided initial leadership to obtain project funding, and provided critical review of manuscript text. RH provided broad project management oversight and critical review of manuscript text. All authors read and approved the final manuscript. The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the funding agency.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Website addresses for notifiable disease reporting requirements, by jurisdiction
Click here for file
Additional File 2
Reportable diseases and number of jurisdictions where reportable
Click here for file
Additional File 3
Reportable infectious agents and number of jurisdictions where reportable
Click here for file
Additional File 4
Reportable substances and number of jurisdictions where reportable
Click here for file
Additional File 5
Reportable findings and number of jurisdictions where reportable
Click here for file
Additional File 6
Reportable procedures and number of jurisdictions where reportable
Click here for file
Acknowledgements
The authors wish to thank the following persons for their contributions to the development of the PHSkb: Sigrid Economou and Bill Parks, CDC; Cecil Lynch, University of California-Davis; Russ Cucina, Natasha Noy, David Buckeridge, Samson Tu, Larry Fagan, Mark Musen, and the Protégé Development team, Stanford Medical Informatics.
==== Refs
Roush S Birkhead G Koo D Cobb A Fleming D Mandatory reporting of diseases and conditions by health care professionals and laboratories JAMA 1999 282 164 170 10411198 10.1001/jama.282.2.164
Nationally Notifiable Infectious Diseases; United States 2005
Hopkins RS Jajosky RA Hall PA Adams DA Connor FJ Sharp P Anderson WJ Fagan RF Aponte JJ Nitsche DA Worsham CA Adekoya N Chang MH Centers for Disease Control and Prevention (CDC) Summary of notifiable diseases – United States, 2003 MMWR Morb Mortal Wkly Rep 2003 52 1 85 15889005
Case definitions for infectious conditions under public health surveillance. Centers for Disease Control and Prevention MMWR Recomm Rep 1997 46 1 55
Case definitions for infectious conditions under public health surveillance
Doyle T Glynn M Groseclose S Completeness of notifiable infectious disease reporting in the United States: an analytical literature review Am J Epidemiol 2002 155 866 874 11978592 10.1093/aje/155.9.866
Birkhead G Chorba T Root S Klaucke D Gibbs N Timeliness of national reporting of communicable diseases: the experience of the National Electronic Telecommunications System for Surveillance Am J Public Health 1991 81 1313 1315 1928531
Jajosky RA Groseclose SL Evaluation of reporting timeliness of public health surveillance systems for infectious diseases BMC Public Health 2004 4 29 15274746 10.1186/1471-2458-4-29
Effler P Ching-Lee M Bogard A Ieong M-C Nekomoto T Jernigan D Statewide system of electronic notifiable disease reporting from clinical laboratories JAMA 1999 282 1845 1850 10573276 10.1001/jama.282.19.1845
Panackal A M'ikanatha N Tsui F McMahon J Wagner M Dixon B Zubieta J Phelan M Mirza S Morgan J Jernigan D Pasculle A Rankin J JrHajjeh R Harrison L Automatic electronic laboratory-based reporting of notifiable infectious diseases at a large health system Emerg Infect Dis 2002 8 685 691 12095435
Yokoe D Subramanyan G Nardell E Sharnprapai S McCray E Platt R Supplementing tuberculosis surveillance with automated data from Health Maintenance Organizations Emerg Infect Dis 1999 5 779 787 10603211
Yokoe D Coon S Dokholyan R Iannuzzi M Jones T Meredith S Moore M Phillips L Ray W Schech S Shatin D Platt R Pharmacy data for tuberculosis surveillance and assessment of patient management Emerg Infect Dis 2004 10 1426 1431 15496244
National Electronic Disease Surveillance System (NEDSS) – LabReporting (ELR)
College of American Pathologists. SNOMED-CT
Federal Information Processing Standards (FIPS) Codes
Biological and chemical terrorism: strategic plan for preparedness and response – Recommendations of the CDC Strategic Planning Workgroup MMWR Recomm Rep 2000 49 1 14
Control of Communicable Diseases Manual 2000 17 American Public Health Association
HL7 Standard Code Set. CVX – Vaccines Administered
PHIN: Notifiable Condition Mapping Tables Accessed 7-24-2003.
Steindel S Loonsk J Sim A Doyle T Chapman R Groseclose S Introduction of a hierarchy to LOINC to facilitate public health reporting Proc AMIA Symp 2002 737 741 12463922
Stanford Medical Informatics Protege-2000 [1.9]
Nationally Notifiable Infectious Diseases – Event Code List Accessed 10-2004.
Musen M Schreiber A Architectures for intelligent systems based on reusable components Artif Intell Med 1995 7 189 199 7581622 10.1016/0933-3657(95)00003-O
ITI Public Health Sentinel Surveillance Solutions
Carter C Ronald N Steele J Young E Taylor J Russell L JrEugster A West J Knowledge-based patient screening for rare and emerging infectious/parasitic diseases: a case study of brucellosis and murine typhus Emerg Infect Dis 1997 3 73 76 9126449
Tu S Eriksson H Gennari J Shahar Y Musen M Ontology-based configuration of problem-solving methods and generation of knowledge-acquisition tools: application of PROTEGE-II to protocol-based decision support Artif Intell Med 1995 7 257 289 7581625 10.1016/0933-3657(95)00006-R
|
16105177
|
PMC1201144
|
CC BY
|
2021-01-04 23:52:11
|
no
|
BMC Med Inform Decis Mak. 2005 Aug 16; 5:27
|
utf-8
|
BMC Med Inform Decis Mak
| 2,005 |
10.1186/1472-6947-5-27
|
oa_comm
|
==== Front
BMC Med Res MethodolBMC Medical Research Methodology1471-2288BioMed Central London 1471-2288-5-261610917610.1186/1471-2288-5-26Research ArticleThe therapeutic effect of clinical trials: understanding placebo response rates in clinical trials – A secondary analysis Walach Harald [email protected] Catarina [email protected] Cornelia [email protected] Dick [email protected] University Hospital Freiburg, Brsg., Institute of Environmental Medicine and Hospital Epidemiology, Germany2 Samueli Institute, European Office, Universitatsklinikum Freiburg, Institut für Umweltmedizin und Krankenhaushygrene, Kugstether Strasse 55, 79106 Freiburg, Germany3 University of Freiburg, Brsg., Institute of Psychology, Germany4 University College Northampton, School of Social Sciences, UK5 University of Amsterdam, Department of Psychology, Netherlands2005 18 8 2005 5 26 26 9 6 2005 18 8 2005 Copyright © 2005 Walach et al; licensee BioMed Central Ltd.2005Walach 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 and purpose
Placebo response rates in clinical trials vary considerably and are observed frequently. For new drugs it can be difficult to prove effectiveness superior to placebo. It is unclear what contributes to improvement in the placebo groups. We wanted to clarify, what elements of clinical trials determine placebo variability.
Methods
We analysed a representative sample of 141 published long-term trials (randomized, double-blind, placebo-controlled; duration > 12 weeks) to find out what study characteristics predict placebo response rates in various diseases. Correlational and regression analyses with study characteristics and placebo response rates were carried out.
Results
We found a high and significant correlation between placebo and treatment response rate across diseases (r = .78; p < .001). A multiple regression model explained 79% of the variance in placebo variability (F = 59.7; p < 0.0001). Significant predictors are, among others, the duration of the study (beta = .31), the quality of the study (beta = .18), the fact whether a study is a prevention trial (beta = .44), whether dropouts have been documented (beta = -.20), or whether additional treatments have been documented (beta = -.17). Healing rates with placebo are lower in the following diagnoses; neoplasms (beta = -.21), nervous diseases (beta = -.10), substance abuse (beta = -.14). Without prevention trials the amount of variance explained is 42%.
Conclusion
Medication response rates and placebo response rates in clinical trials are highly correlated. Trial characteristics can explain some portion of the variance in placebo healing rates in RCTs. Placebo response in trials is only partially due to methodological artefacts and only partially dependent on the diagnoses treated.
==== Body
Background
Randomized controlled clinical trials (RCTs) usually employ a placebo control group to control for non-specific effects of therapy [1]. These effects comprise, among others, statistical regression to the mean, measurement artifacts, the natural course of diseases [2-4]. and true healing or improvement due to psychological or non-specific factors of therapy [5,6]. Thus, placebo response rates in clinical trials are always a mix of what has been termed true and false placebo effect [7]. True placebo effects have recently been defined as therapeutic effects due to the meaning of an intervention for a patient [8]. Some argue against the usage of the term "placebo effect" for such positive non-specific treatment effects, and observe that it is not possible to determine true placebo response rates in controlled clinical trials, since a natural history control group is lacking [9]. An analysis of three-armed trials with a natural-history control group in addition to treatment and placebo groups found no significant effect different from natural history for dichotomous outcomes, but a significant effect size of d = -.28 for continuos measures, which reflects mostly the placebo effect in pain [10]. This latter study, however, excluded a lot of high quality evidence from non-clinical, experimental studies, from non-pharmacological studies, from the psychological literature and older trials which were not randomized but are still very suggestive of clinical effects of placebos. If that material is also taken into account it seems difficult to dismiss a placebo effect in clinical trials outright [11]. A recent re-analysis of these data classifying studies into those that tried to maximize placebo effects and into trials that used placebo only as a control procedure found a large effect size of d = .95 for studies maximizing placebo effects, which was significantly different from studies using placebo as control (d = .15) [12]. It seems worthwhile, therefore, to try to understand what elements placebo response rates in randomized, placebo-controlled clinical trials (RCTs) are comprised of.
Although it would be ideal to launch a series of dismantling studies within the context of clinical trials to disentangle components of placebo response rates in RCTs, this is difficult. While in an experimental context many efforts have been presented recently, within a clinical context researchers, with a few exceptions [13,14]., seem reluctant to pursue the issue. A possible, albeit less reliable, route is the indirect study through secondary analysis of published trials. By using a non-selected sample of RCTs and correlating formal and informal study characteristics with the placebo response rate in these trials we tried to understand which characteristics of studies contribute to placebo response rates in RCTs.
In a previous secondary analysis of 26 RCTs with treatment duration of more than 12 weeks we found a significant correlation of r = .59 between placebo response rate and response rate in treatment groups [15]. Using a similar approach, Kirsch and Sapirstein [16] found an even higher correlation between treatment and placebo groups in trials of antidepressants (r = .90). When comparing these results with psychotherapy studies which used wait-list controls they estimated the specific treatment effect in those antidepressant trials to be 25% of the whole treatment effect. A more recent analysis of unpublished FDA licensing data of new antidepressant drugs, selective serotonin re-uptake inhibitors, corroborates these findings showing that 82% of the drug effect is duplicated by placebo, although the study arms are not significantly correlated [17]. Since the data of these study are from licensing studies which are proprietary and have not been subjected to peer review, it is difficult to know what to make of this information.
We decided to replicate and enlarge these findings by conducting another secondary analysis on a larger set of studies with a more heterogeneous set of diseases. Placebo response rates in RCTs might vary as a function of the disease studied, due to methodological artifacts because of low methodological study quality, as a function of time or due to other characteristics of a study. Thus, we retrieved a sample of unselected long-term trials. We coded several study characteristics and correlated them with the placebo response rate. We also operationalized study quality to find out whether placebo response rate is partially an artifact due to variation in quality. Our question was: Is the placebo response rate in RCTs dependent on time, study quality or other formal and informal study characteristics? In order to answer this question we conducted a systematic secondary analysis of published trial data.
Methods
Literature search and trial inclusion criteria
We searched the medical literature from 1992 to 1997 (MEDLINE, Cochrane Library) in order to find high quality long-term clinical trials. We also used existing meta-analyses in order to locate appropriate studies. Search terms were "placebo*", "double-blind*", "RCT*", "meta-analysis", "long*", "*studies", "treatment outcome". These search terms were combined with different strategies in order to optimize the results. We also handtracked the reference lists of existing studies and meta-analyses. Furthermore, we handsearched volume 1997 of main stream journals (Lancet, New England Journal of Medicine, British Medical Journal, Journal of the American Medical Association) as well as other leading medical journals. We also screened Journal Watch, Newsletter of the New England Journal of medicine, and AIDS Clinical Care.
Criteria for exclusion and inclusion of trials were formulated in a study protocol in advance.
All studies had to:
1 have a double blind placebo controlled randomized design
2 treat ill adults
3 provide a treatment of at least 12 weeks' duration
4 deal with a medical intervention with placebo control group
5 give sufficient information for to have outcome rates calculated for both groups
6 cover the period from 1992 to 1997.
Data extraction
All data were extracted from the report of each trial with the use of a pretested form and entered into a spreadsheet. Formal characteristics were coded: (1) disease treated (ICD-10 diagnosis according to the first and second level); (2) duration of study; (3) is it a multi-center study (4) attrition rate (number of drop-outs in each group); (5) is the statistical evaluation of the study results done according to intent-to-treat analysis; (6) rating of study quality; (7) 'improvement' = response rate with placebo and medical treatment in a manner similar to [18] (i.e. % patients improved according to main outcome parameter mentioned in the study), or alternatively, in prevention trials, the number of patients (%) without event (e.g., cardiac) or worsening of the condition being under investigation (e.g., dementia);
Quality rating
The rating of study quality is notoriously difficult [19]. Although the Jadad-score seems to be accepted since empirical validation studies exist [20-22], recent data also cast doubt on its usefulness [23]. Therefore, we followed recommendations to use a checklist adapted to the study question [19]. Our quality rating based on the one proposed by Detsky [24] and the Cochrane criteria [25] comprised the following items:
1. Description of inclusion and exclusion criteria?
2. Randomisation
3. Allocation concealment
4. Documentation of undesired events
5. Double-blinding of doctor, patient, and evaluator
6. Concomitant treatments
7. Description of statistical methods
8. Predefinition of outcome criteria
9. Documentation of patients lost to follow-up
Items 1, 4, 7-9 were answered in a yes-no format, items 2, 3, 5 and 6 were answered as "yes, and adequate", "yes, only mentioned", "no".
The coding of study quality was conducted by 2 of the authors with a subset of 20 studies rated jointly. A coding manual defined how to code for different aspects of study quality. The calculation of inter-rater reliability as intra-class correlation coefficient [26,27] showed sufficient consistency (r = .77); the widely used coefficient kappa was not applicable in our case due to the structure of the data matrix.
Additional questionnaire information
Research on placebo effects normally starts from the presumption that placebo effects are due to instructions and contexts of the study, which induce expectations and other cognitions in patients as well as in physicians. In turn, these expectations give rise to biological processes and physical changes.
To test this hypothesis we constructed a questionnaire asking for additional information (e.g., amount of time spent with patients, amount of effort invested in the trial, indications of unblinding, expectations during the study). Our question was: is there a correlation between surrogate measures for expectancies of the principal investigators and the response rates.
For a subset of studies with recent publication dates we contacted the original researchers to elicit more information about study characteristics which are not normally reported in publications. We faxed or e-mailed a standardized questionnaire to those investigators who had responded positively to initial telephone requests. This questionnaire asked for informal addditional information. Four questions referred to organisational aspects of the trial (who initiated and financed the study?). Seven questions referred to possible aspects of methodological quality like unblinding, four questions were about time and intensity of contacts between the nursing staff, researchers and patients, and eleven questions referred to the attitude of the principal investigators towards the study, the study result and their expectations during the study. The questionnaire was originally in German, translated into English by a native speaker, who was fluent in German, and retranslated by a German native speaker fluent in English.
Statistical evaluation
Diagnosis according to ICD 10 classification (first 2 descriptive levels) was converted into dummy variables (1,0-coding). All entered data were checked twice for plausibility and correctness by two independent persons. Descriptive information was taken from the study reports. The rating of study quality was added to a single unweighted quality score, after inspecting correlational patterns on single-item level. Improvement rate with treatment and placebo was defined as proportion of improved patients in relation to all patients treated within this group. Normally, this is reported in the original publications, or else it was calculated. Studies with more than two study arms were treated as two different studies, if separate placebo groups were employed. In dose finding studies with more than one treatment arm we always used the treatment group with the highest efficacy.
Data of the investigators' questionnaire was used on a single item level. Additionally, we formed different indices out of the items of the questionnaire which were meant to reflect "high involvement" (e.g. time, personal effort), high expectation or high importance of a study by grouping items according to their topic.
We calculated first order correlations of formal and informal study characteristics with placebo response rates. Based on these results and on our previous finding we formulated parsimonious regression models weighted by study size (n-3) as recommended in [28] to clarify which variables contribute to placebo response rates. In order to minimize capitalization of chance we only used theoretically interesting variables, those with significant first order correlations, and those that seemed promising after a first stepwise hierarchical model. In a second step we took out all non-significant predictors and entered all variables in the hypothesized sequence of importance in a forced regression model. Residuals were checked for possible non-linearity.
Results
Our search strategy produced after initial screening for sensitivity and specificity in its definite version 375 studies. The abstracts were screened for inclusion criteria. These were fulfilled by 141 studies. Figure 1 gives details about inclusion of trials.
Figure 1 Flowchart of study inclusion.
Table 1 gives the number of studies according to ICD categories.
Table 1 Number of studies according to ICD categories
ICD-10 Category Disease category No of studies
A, B Infectious and parasitical diseases 4
C Neoplasms (Tumours) 5
E Endocrine, nutritional and metabolic diseases 8
Psychological and behavioural disorders
F0 Organical, including symptomatical psychological disorders 4
F1 Psychol. and behav. disorders due to psychotropic substance abuse 22
F3 Affective disorders 13
F4 Neurotic, stress and somatoform disorders 12
F5 Behav. abnormalities with somatic disorders 5
F6 Personality and behavioural disorders 1
F8 Developmental disorders 1
G Nervous diseases 14
Diseases of circulatory system
I10/11 High blood pressure 4
I20-I52 Ischemic and other forms of heart disease 21
I6/7 Cerebrovascular and other peripheral vascular diseases 2
J4 Pulmonary diseases 2
K Digestive diseases 6
L Diseases of the skin 1
M Diseases of the muscular skeletal system and of connective tissue 9
N Diseases of the urogenital system 3
T8 Trauma, intoxications and other consequences of extraneous causes 2
Y4 Extraneous causes of morbidity and mortality 2
Number of all included studies N = 141
Since in some studies there were multiple arms of treatment and control groups the basis for the statistical calculations is N = 153 independent combinations of placebo and treatment effects. Table 2 gives the significant first order correlations between the relevant variables of study characteristics.
Table 2 First order correlations between study characteristics (Pearson), N = 153
%_T %_P DUR. DIA_E DIA_F0 DIA_F1 DIA_F4 DIA_G DIA_ I1 DIA_ I2 PREV INDEXQR
%_T 1,00 ,78*** ,29*** ,01 -,14 -,33*** ,01 -,31*** ,14 ,37*** ,49*** ,03
%_P 1,00 ,41*** ,07 -,16* -,19* -,11 -,30*** ,12 ,45*** ,59*** ,08
DUR. 1,00 ,26** -,08 -,18* -,18* -,16 ,41*** ,25** ,40*** ,12
DIA_E 1,00 -,04 -,11 -,07 -,07 -,04 -,10 ,25** -,15
DIA_F0 1,00 -,07 -,05 -,05 -,03 -,07 ,05 -,05
DIA_F1 1,00 -,13 -,14 -,08 -,19* -,27** -,03
DIA_F4 1,00 -,09 -,05 -,12 -,22** -,02
DIA_G 1,00 -,06 -,13 -,24** -,06
DIA_I1 1,00 -,08 ,01 ,05
DIA_I2 1,00 ,54*** ,14
PREV 1,00 ,11
INDEXQR 1,00
*) p < .05 **) p < .01 ***) p < .001
Legend: %_T Percent of patients improved with treatment
%_P Percent of patients improved with placebo
DUR. Duration of study in months
DIA_E: Diabetes, other diseases of secretory glands and metabolism
DIA_F0: Dementia
DIA_F1: Behavioural disorders due to psychotropic substances
DIA_F4: Panic disorders
DIA_G: Epilepsy
DIA_I1: High blood pressure
DIA_I2: Ischemic and other forms of heart disease
PREV. Studies with preventive targets
INDEXQR Quality index
Improvement rates in placebo groups were significantly correlated with improvement rates in the treatment groups (r = .78; Fig. 2), and with duration of study (r = .41). If only the subset of 97 therapeutic trials excluding prevention studies are analysed, the correlation remains significant (r = .61). Placebo improvement rates are lower in trials of behavioral disorders due to psychotropic substances (r = -.19; recall that disease categories were dummy coded as 1: belonging to category; 0: not belonging to category), in dementia (r = -.16) and anti-epileptic trials (r = -.30). They are higher in prevention studies (r = .59) and trials of ischemic and other forms of heart disease (r = .45). They are unrelated to other forms of diagnosis and most notably unrelated to the quality of the trials, on single item level as well as on the index level, when considered as zero-order correlations. Nearly the same patterns of correlations can be found for the improvement rates with treatment.
Figure 2 Scatterplot of response rate with drug and placebo for all studies
We checked in a regression model for the predictive power of the combined variables. A significant regression model emerged that was able to predict nearly 80% of the variability of response in the placebo group (adjusted R2 = .794; F 10/142 = 59,74; p < 0.0001). This is presented in table 3. Main predictors were from two categories: methodological and diagnostic. Placebo response rates were larger for studies with longer study duration (beta = 0.31), for prevention trials (beta = 0.44), for multicentre trials (beta = 0.13) and for studies with better methodological quality (beta = 0.18). The questions whether dropouts (beta = -0.17) and additional treatments (beta = -.21) were described, were the important methodological variables. Therapeutic effects in placebo groups were smaller for trials of antitumor agents (beta = -0.21), studies in dementia (beta = -0.12), in substance withdrawal studies (beta = -0.14) and in studies of nervous diseases, mainly anti-epileptic trials (beta = -0.10).
Table 3 Regression model predicting variability in response rates in placebo groups of clinical trials; n = 153 studies/comparisons; formal and diagnostic variables (ICD coding)
Beta t (DFs = 142) p-value
Constant 2.07 0.04
Duration in months 0.31 6.17 <0.000001
Multicenter trial 0.13 3.04 0.003
Quality Index 0.18 3.44 0.0008
Neoplasms (C) -0.21 -5.48 <0.000001
Organical psychological disorders (F0) -0.12 -3.24 0.001
Disorders due to substance abuse (F1) -0.14 -2.97 0.003
Nervous diseases (G) -0.10 -2.55 0.01
Prevention trial 0.44 7.94 <0.000001
Quality rating: Additional treatment described? -0.21 -4.34 0.00003
Quality rating: Dropouts described? -0.16 -3.33 0.001
Table 4 presents a subsidiary analysis for those 97 studies, which were not prevention trials. This regression model was also highly significant and explained 42% of the variance in placebo variability (adjusted R2 = .426; F 5/91 = 15.22; p < 0.0001). Main predictors in this model were five diagnostic categories, with studies on substance withdrawal (beta = -0.31) and on nervous diseases/dementia (beta = -.0.20) showing lower placebo response rates, and studies in affective disorders, i.e. antidepressant and anxiolytic trials, (beta = 0.37), studies in digestive diseases, mostly inflammatory bowel conditions (beta = 0.29) and studies in urogenital diseases (beta = 0.16) showing better placebo treatment rates. Most notably, methodological quality did not contribute to the model.
Table 4 Regression model predicting variability in response rates in placebo groups of clinical trials; n = 97 Studies, with prevention trials excluded; formal and diagnostic variables (ICD Coding); italics: different from full model
Beta t (DFs = 91) p-value
Constant 10.3 <0.000001
Disorders due to substance abuse (F1) -0.31 -3.57 0.0006
Affective disorders (F3) 0.37 4.49 0.00002
Nervous diseases (G) -0.20 -2.40 0.018
Digestive diseases (K) 0.29 3.65 0.0004
Diseases of the urigenital system (N) 0.16 2.09 0.039
Additional questionnaire information
Principal investigators of 57 recently published studies were asked for retrospective additional information. 50 of those answered our request (87%), and 44 supplied sufficient data. There is only one significant first order correlation out of all items. Apart from the correlation between placebo and treatment response rates there is no significant and substantial correlation except for chance fluctuations. But we found evidence for unblinding in 30% of the 44 responses.
Monte Carlo simulation
One of the possible sources of the correlation between placebo and verum effect might be publication bias, especially when there is no significant difference between the placebo and verum condition. In order to explore this potential source of the correlation we simulated the situation of publication bias by assuming that only trials were published where the differential effect had a chance probability of p < 0.01
The simulation parameters used were:
a) mean effect size of the placebo: ESp
b) mean effect size of the verum: ESv
c) Number of subject in a trial: Nss
d) Number of experiments per effect: Nexp
If no publication bias is assumed, there is no correlation between verum and placebo effect sizes. However, there were significant correlations when a publication bias was introduced. This was the case for each reasonable choice of simulation parameters. In Figure 3 a scatterplot for published placebo and verum effect sizes are given for a typical choice of simulation parameters. The correlation obtained is r = 0.63, close to the true correlation. In other words publication bias could, in principle, explain nearly all variance, assuming that all non-significant studies are suppressed. It should be noted, however, that the filedrawer for this particular simulation contains 86% of all experimental results. It is improbable that so many trials would remain unpublished.
Figure 3 Scatterplot of results of simulation of the effect of publication bias. Simulation parameters were: Esp = 0.5, ESv = 0.51 - 0.60 in steps of 0.01, Nexp = 50, Nss = 100. All non-significant trials removed
Discussion
This secondary analysis was motivated by the attempt to understand what influences placebo response rates in RCTs. We found a rather strong correlation between improvement rates with placebo and treatment of r = .78, which explains roughly 60% of the variance. The magnitude of the correlation drops somewhat to r = .61, explaining 37% of the variance, if only therapeutic trials are considered. This result is in concordance with our own and other previous findings [15,17,29]. It is incidentally backed by an early result reported by Evans [30]. Although Evans' finding has been explained as a result of the skewed distribution of the underlying binary data [31], this explanation does not apply to our data, which followed a Gaussian distribution.
This high correlation should not come as a surprise, as in clinical trials a certain basic effect, thought to be covered by the placebo control group, is compared with this same basic effect plus some specific element of pharmacologic intervention. That is, effects in clinical trials are bound to be correlated. Indeed, if there were no intervention effects at all we would have a perfect correlation of r = 1.0. The fact that the correlation is not perfect is a sign that treatment and control groups behave differently. The fact that the correlation is so high and in fact explains 60% of the variance is a sign that the commonalities of factors in groups within trials is greater than their difference. In other words, non-specific treatment effects are more important than the specific ones. Thus we have once more corroborated a now growing body of evidence about the importance of non-specific treatment effects [14,32].
Natural history or cohort effects might be an explanation for the high correlation between treatment and placebo response. The correlations found in our study are suggestive of an overall small to moderate treatment effect reflecting in an imperfect correlation. Although a cohort effect is a possible explanation, we would like to point out that durations of trials are normally chosen according to the disease studied and to cover a time period which captures a relevant portion of natural fluctuation. For instance, depression maintenance trials are normally conducted over a time span long enough to make a recurrence in untreated patients likely, or dementia trials cover a period, where a progression of the disease is to be expected. Thus, natural history is partly covered by the differing trial durations of the studies analyzed. If natural history were the only factor reflected in the improvement rates of placebo groups, on average we should expect deteriorations or little improvement in placebo groups, because the diseases studied here are mostly long-term chronic diseases with little self-limiting tendencies. Across all studies, diseases and trial durations of a heterogeneous set of studies we should expect that cohort effects average out and the correlation between treatment and placebo improvement should tend towards zero [31]. In fact, our simulation produces a zero-correlation for many different scenarios of study outcomes, confirming this intuitive reasoning. Thus, there seem to be non-specific elements of treatment at work here that reflect in improvement rates in placebo groups which are grossly comparable to those in the treatment groups.
What is important from a differential point of view is, whether these non-specific elements of treatments, or the variability of responses in the placebo groups, can be further elucidated. Our regression analysis shows that a series of formal characteristics is responsible for this variability of therapeutic responses in placebo groups. The fact that this analysis can explain nearly 80% of the variance is a support for the hypothesis that a large part of those non-specific effects in randomised placebo controlled trials is due to formal characteristics. The effect in placebo control groups is higher in studies with a longer duration and in prevention trials. Prevention trials are normally also longer in duration. In fact this effect of duration of trial vanishes when prevention trials are taken out of the analysis (compare Tables 3 and 4). This strong effect of prevention trials is worth considering. It means that in such trials the event rate in the control group is smaller, i.e. the placebo healing rate higher, than in comparable curative trials. It may be the psychological focus on preventing an event that triggers effects different from placebos in curative trials. It may simply be the fact that in prevention trials medications, and placebos, are given for a very long time, sometimes over years, and hence expectations for maintainance of a comparatively healthy state are continuously reinforced. Or it may be easier to harness expectations for maintaining a state of comparative well-being than to use them for getting well again. This effect of prevention trials warrants a deeper scrutiny than can be provided by a retrospective analysis.
This predictive power of prevention trials is by far the strongest effect in our data base, which also explains partially the other elements: prevention trials are normally not only longer, but also larger and require the effort of many centres. Therefore they are in tendency better planned and have a slightly better methodology. These methodological factors disappear, when the analysis is repeated with prevention trials excluded. It is obvious that placebo variability is different for different diseases. In the full data-set trials of anti-tumor agents, of anti-epileptics, of anti-dementia drugs, and of substance withdrawal contribute to variability by exhibiting smaller placebo effects. However, this should not blind us to the fact that the beta-weights, which are correlation coefficients adjusted for the effects of all other variables, of these diagnostic variables are rather small. In the reduced data-set, without the prevention trials, improvement rates in placebo groups are higher in studies of affective disorders, which comprise mainly antidepressant and anxiolytic studies, in studies of anti-inflammatory agents in bowel diseases and in studies of urogenital diseases. Improvement rates are lower for substance withdrawal studies and anti-epileptics. The fact that the regression model in the reduced data-set explains only 42% of the variance and thus has approximately only half the explanatory power of the full set shows that other unknown factors are operative, besides spontaneous improvement rates in different diseases. Otherwise we would expect a stronger effect of the diagnostic categories in our regression model.
Our simulation runs brought one other explanatory option to the fore: publication bias. Under the assumption that only trials are published that prove the significance of the treatment over placebo and the non-significant trials are filed away, simply by this publication bias alone, a correlation can be obtained which is close to the one we observed empirically. Thus it is tempting to assume that the true reason for the high correlation between treatment and placebo improvement rates is, at least partially, publication bias. This points, once more, to the importance of publishing all evidence, not only positive studies. However, bearing in mind that the simulation produced a correlation of r = .63 under the precondition that all non-significant evidence, i.e. 87% of all evidence remained unpublished, publication bias cannot account for the whole correlation, but perhaps for a substantial amount.
Although not quite as obvious, one of the most important messages of this study is hidden behind what we have not found:
There is no sizeable effect of methodological quality apart from the one discussed above which is explained by the prevention trials. This means that placebo variability in clinical trials is not only due to methodological artifacts, as is sometimes suggested [2,3], at least as assessed by our scale.
Frequently it is proposed that placebo responses are due to heightened expectations of investigators and subsequently patients in trials. Since trials are conducted with more frequent and more intense contacts between doctors and patients, the argument would run, patients could form stronger expectations of improvement, which in turn could lead to response expectancies, which again would result in clinical improvement [33]. We could not test this hypothesis directly. But we did test it indirectly by asking principal investigators to give information on study characteristics not normally reported in publications, like amount of time spent with patients, amount of effort invested in a trial, importance of the trial for further funding, expectation of investigator and the like. None of these indirect operationalizations of heightened expectation and increased effort on part of the investigators showed any notable correlation with placebo response rates.
We found substantial evidence for unblinding of investigators in 30% of 44 trials with additional questionnaire data. Surprisingly enough, placebo response rate was uncorrelated with unblinding. Thus, patients in studies with indications for investigator unblinding had a higher improvement rate only in the treatment group, but not in the placebo group. If investigator opinion was a decisive factor in the creation of patient expectancies, we would expect a substantial negative correlation between unblinding and placebo response rate. Since unblinding occurred in 30% of the cases, we should have been able to see such a correlation if it had been present. It is obviously not an issue. Thus, informal characteristics in studies, reflecting a higher awareness and a stronger engagement of investigators and patients in clinical trials, do not seem to influence the amount of improvement in placebo groups. We submit that this part of our data is rather weak, since it is retrospective evidence. Nevertheless, it is a first empirical data set which could be followed up by prospective studies.
What other explanations or mechanisms could be responsible for this effect? One idea is well known and posits that clinical trials are in fact healing rituals with a strong non-specific effect [33]. The other invokes the new concept of generalized entanglement in systems.
Trials as healing rituals
Treatment effects of mostly pharmacological interventions studied here do have some effect, but the effect is smaller than those of non-specific effects. It has been argued that therapeutic effects are meaning effects in general, with pharmacological or other medical interventions adding a plausibility factor that actually triggers this meaning effect [34]. A meaning effect would be an individual healing effect induced by the complex interaction between an individual, constructing meaning out of his or her medical situation, the medical system trying to intervene, physiological changes brought about by the interventions and psychological changes inferred by the individual and hence altered psychological states. These states have been documented as altered brain function [35-40] and hence it is plausible to assume that they can also affect complex medical conditions.
Such an interpretation could be supported by our data, since the non-specific elements are by far the more important ones, compared with the specific elements of treatment. Non-specific effects account for nearly 60% of the variance of all treatment effects. This is true for most disease categories and across a wide variety of interventions. This alone should be an intriguing result, stimulating more research effort into mechanisms and processes of unspecific effects of therapy in general and trials in particular.
Correlation effects as an instance of generalized entanglement:
According to a weaker and generalized version of quantum theory [41], applicable to different systems outside the realm of physics, entanglement, i.e. non-local correlations structurally similar to but factually different from quantum correlations, would be expected in any system with a global observable and a local observable that are complementary and do not commute; in addition they may have to share a common contextual history (D. Gernert, personal communication). A clinical, blinded trial is such a system by definition. The global observable is the blinding of the trial, the local observable is the actual allocation of patients to treatment groups. Global blinding and local definite allocation are complementary notions, and thus entanglement would be predicted, visible as a correlation across trials. The common context is given by the aim and the procedures of the trial. This hypothesis, which has received some support recently by experimental data in other areas [42], would have to be investigated further by means of direct experimentation. For the time being this is only a speculative possibility. However, should it bear out, it would have consequences both for study methodology and for clinical practice. On the one hand, it would be expected that medications show different, mostly stronger, effects in unblinded contexts, on the other hand it would follow that data from blinded trials are biased estimators of real effect sizes.
We would like to add some cautions:
It is difficult to draw firm conclusions about non-specific effects from two armed-trials [9]. This should be born in mind, when interpreting our data, since all our evidence and argument is indirect by default. Our retrospective questioning of principal investigators may be considered weak. While we agree that only a few investigators could be queried and retrospectively gathered evidence is far from compelling, there is no selection behind our sampling procedure but only temporary sequence. And among those selected we achieved a high return rate. Thus we are confident that we have at least given a small part of the picture reliably, and it goes without saying that this piece of our evidence can only be a hint for further research. It should be elucidated in prospective analyses, whether and how often unblinding is seen and how this affects outcome rates. It could be documented before knowledge of outcomes, how strong extra involvement and enthusiasm of doctors and investigators have been compared to conventional clinical practice. Until such prospective evidence is available, we have produced at least some estimators of trial effects and have not found convincing evidence for them.
Conclusion
We conclude that the placebo response rate in controlled clinical trials is not due to methodological artifacts, to disease history alone or to circumstantial characteristics of studies, but seems to reflect a genuine improvement, unless one invokes publication bias for all negative studies. This improvement accounts for roughly 60% of the variance of all therapeutic gains across trials.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Harald Walach devised the study, organised funding, supervised the study, participated in the final evaluation, wrote the manuscript and had part in the interpretation of the data.
Catarina Sadaghiani was a PhD-student on the project, conducted the study, collected, prepared and analysed the data, had part in the final evaluation and in the interpretation of the data.
Cornelia Dehm was a diploma student on the project, jointly supervised by HW and CS. She did the interview-part of the study and helped with interpretation.
Dick Bierman helped with data-analysis and interpretation and conducted the Monte-Carlo simulation.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This study was sponsored by the Institut für Grenzgebiete der Psychologie und Psychohygiene, Freiburg, Germany. HW and CS are supported by the Samueli Institute for Information Biology. The authors are grateful to Dr. David Reilly and Zelda di Blasi for helpful comments on an earlier version of this paper, and to Dan Moerman and Klaus Linde for suggestions for improvement.
==== Refs
Armitage P The role of randomization in clinical trials Stat Med 1982 1 345 352 7187102
Kienle GS Kiene H The powerful placebo effect: Fact or fiction? J Clin Epidemiol 1997 50 1311 1318 9449934
McDonald CJ McCabe GP How much of the placebo effect is really statistical regression? (letter) Stat Med 1989 8 1301 1302 2814076
McDonald CJ Mazzuca SA McCabe GP How much of the placebo 'effect' is really statistical regression? Stat Med 1983 2 417 427 6369471
Kirsch I Harrington A Specifying nonspecifics: Psychological mechanisms of placebo effects The Placebo Effect: Interdisciplinary Explorations 1997 Cambridge MA: Harvard University Press 166 186
Kirsch I Lynn SJ Automaticity in clinical psychology Am Psychol 1999 54 504 515 10424156
Ernst E Resch KL Concept of true and perceived placebo effects BMJ 1995 311 551 553 7663213
Moerman DE Jonas WB Deconstructing the placebo effect and finding the meaning response Ann Int Med 2002 136 471 476 11900500
Hróbjartsson A What are the main methodological problems in the estimation of placebo effects? J Clin Epidemiol 2002 55 430 435 12007544
Hróbjartsson A Gotzsche PC Is the placebo powerless? An analysis of clinical trials comparing placebo with no treatment N Engl Med 2001 344 1594 1602
Walach H Sadaghiani C Plazebo und Plazebo-Effekte: Eine Bestandsaufnahme Psychother Psychosom med Psychol 2002 52 332 342 12152127
Vase L Riley JL Price DD A comparison of placebo effects in clinical analgesic trials versus studies of placebo analgesia Pain 2002 99 443 452 12406519
Amanzio M Pollo A Maggi G Benedetti F Response variability to analgesics: a role for non-specific activation of endogenous opioids Pain 2001 90 205 215 11207392
Vase L Robinson ME Verne GN Price DD The contributions of suggestion, desire, and expectation to placebo effects in irritable bowel syndrome patients: An empirical investigation Pain 2003 105 17 25 14499416
Walach H Maidhof C Kirsch I Is the placebo effect dependent on time? Expectancy, Experience, and Behavior 1999 Washington, DC: American Psychological Association 321 332
Kirsch I Sapirstein G Listening to prozac but hearing placebo: A meta-analysis of antidepressant medication Prevention & Treatment 1998 1 2a
Kirsch I Moore TJ Scoboria A Nicholls SS The emperor's new drugs: An analysis of antidepressant medication data submitted to the U.S. Food and Drug Administration Prevention & Treatment 2002 5 Article 23
Walsh TB Seidman SN Sysko R Gould M Placebo response in studies of major depression: variable, substantial, and growing J Am Med Assoc 2002 287 1840 1847
Jüni P Witschi A Bloch R Egger M The hazards of scoring the quality of clinical trials for meta-analysis J Am Med Assoc 1999 282 1054 1060
Moher D Jadad AR Tugwell P Assessing the quality of randomized controlled trials. Current issues and future directions International Journal of Technology Assessment in Health Care 1996 12 195 208 8707495
Jadad AR Moore RA Carroll D Jenkinson C Reynolds DJM Gavaghan DJ Assessing the quality of reports of randomized clinical trials: is blinding necessary? Control Clin Trials 1996 17 1 12 8721797
Moher D Pham B Jones A Cook DJ Jadad AR Moher M Does quality of reports of randomised trials affect estimates of intervention efficacy reported in meta-analyses? Lancet 1998 352 609 613 9746022
Clark HD Wells GA Huet C McAlister FA Salmi LR Fergusson D Assessing the quality of randomized trials: reliability of the Jadad scale Control Clin Trials 1999 20 448 452 10503804
Detsky AS Naylor CD O'Rourke K McGeer AJ L'Abbé KA Incorporating variations in the quality of individual randomized trials into meta-analysis J Clin Epidemiol 1992 45 255 265 1569422
The Cochrane Library (database on disk and CDROM) Mulrow CD, Oxman AD The Cochrane Collaboration 1997 Oxford, Update Software. Cochrane Collaboration Handbook updated September 1997
Orwin RG Cooper H, Hedges LV Evaluating coding decisions The Handbook of Research Synthesis 1994 New York: Russell Sage Foundation 139 162
Fleiss JL Statistical Methods for Rates and Proportions 1981 2 New York: Wiley
Hedges LV Olkin I Statistical Methods for Meta-Analysis 1985 Orlando: Academic Press
Kirsch I Sapirstein G Kirsch I Listening to Prozac but hearing placebo: A meta-analysis of antidepressant medications Expectancy, Experience, and Behavior 1999 Washington, DC: American Psychological Association 303 320
Evans FJ Bonica JJ The placebo response in pain reduction International Symposium on Pain 1974 New York: Raven Press 289 296
McQuay H Carroll D Moore A Variation in the placebo effect in randomised controlled trials of analgesics: all is blind as it seems Pain 1996 64 331 335 8740611
Walach H Jonas WB Placebo research: The evidence b ase for harnessing self-healing capacities J Altern Complement Med 2004 10 S103 S112 15630827
Frank JD Shepherd M, Sartorius N Non-specific aspects of treatment: the view of a psychotherapist Non-Specific Aspects of Treatment 1989 Bern: Huber 95 114
Moerman DE Meaning, Medicine, and the "Placebo Effect" 2002 Cambridge: Cambridge University Press
de la Fuente-Fernández R Ruth TJ Sossi V Schulzer M Calne DB Stoessl AJ Expectation and dopamine release: Mechanism of the placebo effect in parkinson's disease Science 2001 293 1164 1166 11498597
Leuchter AF Cook IA Witte EA Morgan M Abrams M Changes in brain function of depresses subjects during treatment with placebo Am J Psychiatry 2002 159 122 129 11772700
Mayberg HS Silva JA Brannan SK Tekell JL Mahurin RK McGinnis S The functional neuroanatomy of the placebo effect Am J Psychiatry 2002 159 728 737 11986125
Pariente J White P Frackowiak RS Lewith G Expectancy and belief modulate the neuronal substrates of pain treated by acupuncture Neuroimage 2005 25 1161 1167 15850733
Petrovic P Kalso E Petersson KM Ingvar M Placebo and opioid analgesia – imaging a shared neuronal network Science 2002 295 1737 1740 11834781
Wager TD Rilling JK Smith EE Sokolik A Casey KL Davidson RJ Placebo-induced changes in fMRI in the anticipation and experience of pain Science 2004 303 1162 1167 14976306
Atmanspacher H Römer H Walach H Weak quantum theory: Complementarity and entanglement in physics and beyond Foundations of Physics 2002 32 379 406
Wackermann J Seiter C Keibel H Walach H Correlations between brain electrical activities of two spatially separated human subjects Neurosci Lett 2003 336 60 64 12493602
|
16109176
|
PMC1201145
|
CC BY
|
2021-01-04 16:32:51
|
no
|
BMC Med Res Methodol. 2005 Aug 18; 5:26
|
utf-8
|
BMC Med Res Methodol
| 2,005 |
10.1186/1471-2288-5-26
|
oa_comm
|
==== Front
BMC Musculoskelet DisordBMC Musculoskeletal Disorders1471-2474BioMed Central London 1471-2474-6-441608683910.1186/1471-2474-6-44Research ArticleTreatment of osteoarthritis of the knee with a topical diclofenac solution: a randomised controlled, 6-week trial [ISRCTN53366886] Baer Philip A [email protected] Lisa M [email protected] Zev [email protected] Malvern Medical Centre, Toronto, Canada2 Clinical Research, Dimethaid Research Inc., Markham, Canada2005 8 8 2005 6 44 44 14 2 2005 8 8 2005 Copyright © 2005 Baer et al; licensee BioMed Central Ltd.2005Baer 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
Topical NSAIDs have been proven to relieve the symptoms of osteoarthritis (OA) in short-term studies (2 weeks). To justify its chronic use, efficacy of a topical NSAID over a longer term of study should be demonstrated. The efficacy and safety of a topical diclofenac solution over a 6-week treatment course in symptomatic primary OA of the knee was investigated.
Methods
216 men and women, age 40–85 years, with radiologically confirmed primary OA of the knee and a flare of pain at baseline following discontinuation of prior therapy were enrolled into this double-blind study. Participants applied either a topical diclofenac solution (Pennsaid®) or vehicle control solution (carrier with no diclofenac); 40 drops 4 times daily directly to the painful knee(s), without massage, for 6 weeks. Pre-planned primary efficacy outcome measures included the core continuous variables pain relief and improved physical function measured by the Western Ontario and McMaster Universities (WOMAC) LK3.1 OA Index, and improved patient global assessment (PGA). Secondary efficacy measure was reduced stiffness. Safety assessments included adverse events and vital signs.
Results
The topical diclofenac group had a significantly greater mean change in score (final minus baseline) compared to the vehicle control group for pain (-5.2 vs. -3.3, p = 0.003), physical function (-13.4 vs. -6.9, p = 0.001), PGA (-1.3 vs. -0.7, p = 0.0001) and stiffness (-1.8 vs. -0.9, p = 0.002). The mean difference between treatment arms (95% confidence interval [CI]) was 1.9 (0.7 to 3.2), 6.5 (2.5 to 10.5), 0.6 (0.2 to 0.9), and 0.9 (0.3 to 1.4), respectively. Safety analyses showed that topical diclofenac caused skin irritation, mostly minor local skin dryness, in 42/107 (39%), leading to discontinuation of treatment in 5/107 (5%) participants.
Conclusion
This topical diclofenac solution demonstrated relief at 6 weeks of the symptoms of primary osteoarthritis of the knee.
==== Body
Background
Meta-analysis of previous trials of topical non-steroidal anti-inflammatory drugs (NSAIDs) concluded that they effectively treat the pain of acute soft tissue injuries [1] and chronic musculoskeletal conditions [2,3]. Current evidence-based recommendations for the management of osteoarthritis (OA) support the use of topical NSAIDs and rubefacients [4-6] as a therapeutic option potentially with fewer gastrointestinal risks than oral NSAIDs [7]. However, a recent critical meta-analysis concluded that claims of pain relief in OA by currently available topical NSAIDs are supported by only a limited number of randomised controlled trials of small size and brief duration, with no data demonstrating efficacy beyond 2 weeks [8].
In this report, we present the efficacy and safety results from a 6-week controlled trial using a newer topical diclofenac solution in knee OA. Effect size data and number-needed-to-treat (NNT) are presented, facilitating comparison with the previously reviewed data.
Methods
Participants and inclusion/exclusion criteria
This study was conducted from November 1999 to August 2000, at 17 medical centres across central Canada, following approval by a central ethics review board (Integrated Research Incorporated, Ethics Review Committee, Montreal, QC). Participants were recruited from the physician's private practice or the surrounding community. At the screening visit, after providing written, informed consent, each participant underwent a screening interview and was eligible to proceed to washout if all inclusion criteria and no exclusion criteria were met.
Inclusion criteria specified men and non-pregnant women, age 40–85 years, with primary OA of at least one knee, and a flare of pain after withdrawal of prior therapy with either an oral NSAID or acetaminophen (used at least 3 days per week during the previous month). Primary OA was defined by deterioration and abrasion of articular cartilage (joint space narrowing) or formation of new bone (osteophytes) at the joint surface of the knee (medial tibio-femoral, lateral tibio-femoral or patello-femoral), demonstrated on a radiological examination carried out within the previous 3 months [9]. Pain was measured by the Western Ontario and McMaster Universities LK3.1 OA Index (WOMAC) 5-item pain subscale, each item scored on a 5-point Likert scale (none = 0; mild = 1; moderate = 2; severe = 3; extreme = 4) [10]. Pain was scored at the screening visit, following which prior therapy was withdrawn. The patient scored the pain again at the baseline visit. A flare was defined as an increase in total pain subscale score of at least 2 and at least 25%, with a baseline total pain score of at least 6 (out of a possible 20), and a score of ≥2 (out of a possible 4) on at least one of the 5 items in the WOMAC pain subscale.
Participants were excluded if they had secondary arthritis related to systemic inflammatory arthritis (including rheumatoid arthritis, psoriatic arthritis, post-infectious arthritis and metabolic arthritis, traumatic arthritis or surgical joint replacement); corticosteroid use: (a) oral corticosteroid within the previous 14 days, or (b) intramuscular corticosteroid within 30 days, or (c) intra-articular corticosteroid into the study knee within 90 days, or (d) intra-articular corticosteroid into any other joint within 30 days, or (e) topical corticosteroid at the site of application within 14 days; intra-articular viscosupplementation (e.g., Synvisc®) into the study knee in the preceding 90 days; ongoing use of prohibited medication including NSAID, other oral analgesic, muscle relaxant, or low-dose antidepressant for any chronic pain management; ongoing use of glucosamine or chondroitin (unless used continuously for 90 days prior to study entry); sensitivity to diclofenac, acetylsalicylic acid (ASA) or any other NSAID, acetaminophen, dimethyl sulphoxide, propylene glycol, glycerine or ethanol; clinically-active renal, hepatic or peptic ulcer disease; history of alcohol or drug abuse; lactation; concomitant skin disease at the application site; current application for disability benefits on the basis of knee osteoarthritis; fibromyalgia; other painful or disabling condition affecting the knee; or participation in another investigational drug trial in the previous 30 days.
Interventions
At the baseline visit, all patients that met the final entry criterion of a flare of pain were randomly assigned to receive one of two treatments: (a) topical diclofenac solution (Pennsaid®; Dimethaid Research Inc.), consisting of 1.5% (w/w) diclofenac sodium in a patented carrier containing dimethyl sulphoxide (45.5%, w/w), propylene glycol, glycerine, ethanol and water, or (b) vehicle control solution, consisting of the complete carrier (including dimethyl sulphoxide, 45.5% w/w) but no diclofenac. Participants applied a dose of 40 drops of study solution (about 1.3 mL) to the affected knee 4 times daily for up to 6 weeks. The participant was instructed to apply 10 drops of solution to each side of the knee (front, back, medial and lateral) either dripped directly onto the knee or first into the hand, and then spread over the site without massage. Compliance was verified by weighing the solution bottles at each visit. If the other knee was painful at any time during the study, it was treated and evaluated for safety, but efficacy analysis was performed on only the study knee – the one with the greater baseline pain score (or the dominant knee if both had the same score). Consumption of acetaminophen (up to four 325-mg tablets per day) was permitted for residual knee or other body pain throughout the treatment period, but not during the washout period prior to baseline assessment or during the week prior to final assessment at week 6. ASA (≤ 325 mg/day) was permitted for cardiovascular prophylaxis.
Outcome measures
The primary outcome measures were defined as the change from baseline to final assessment of the study knee in the 3 core continuous variables [11] pain and physical function, assessed using the WOMAC subscales, and patient global assessment (PGA). There was no intermediate assessment of efficacy. The WOMAC is a validated questionnaire [12] consisting of 24 questions (5 on pain, 17 on physical function and 2 on stiffness), each scored on a 5-point Likert scale (see Participants). The PGA question asked: "How has the osteoarthritis in your study joint been over the last 48 hours?" and was scored on a Likert scale (very good = 0; good = 1; fair = 2; poor = 3; very poor = 4). This question focuses on the treated site, unlike a PGA in an oral NSAID trial that can probe the non-signal joints. Secondary measure was change in stiffness. Ancillary measures defined a posteriori were pain on walking – the first question of the WOMAC pain subscale (referred to as 'use-related pain' [13]) – and the following dichotomous variables: 50% improvement in pain [3]; final PGA score of "good" or "very good" [3]; and response based on OMERACT-OARSI responder criteria [14] (a responder is defined as a participant with ≥ 50% improvement in pain or function that was ≥ 20% of the scale, or ≥ 20% improvement in at least two of pain, function or PGA that was ≥ 10% of the scale).
Safety analyses
Safety was assessed during all clinic visits (weeks 3 and 6) and telephone 'visits' (weeks 1 and 5). Safety variables included adverse events, application-site dermatological reactions and vital signs. Adverse events were identified using open-ended questions and a checklist covering common oral NSAID side effects. Dermatological assessment of the knee was based on a standard scale [15] and any abnormality was recorded as an adverse event. All adverse events were categorised according to Coding Symbols for Thesaurus of Adverse Reaction Terms (COSTART) [16]. Laboratory assessment was not done.
Sample size
Based on a power of 80% and a Type I error rate of α = 0.052-tailed, a sample size of 80 participants per group was required to detect an estimated important difference of 2 between the treatment arms, in the change in WOMAC pain dimension score from baseline to final (with standard deviation of 4.5). A total sample size of 200 participants (100 per treatment group) was specified in the protocol, which allowed for a non-evaluable rate of up to 20%.
Randomisation and blinding
The study kits were prepared, labelled and numbered according to a computer-generated randomisation schedule created by an outside consultant using a randomly chosen block size of 4 or 6. They were shipped to the sites in multiples of complete blocks to ensure that a balanced number of participants was assigned to the two treatment arms within each site. As a participant qualified for study entry at the baseline visit, the investigator assigned him/her the next randomisation number in a sequential manner. The randomisation schedule was concealed from the investigators, their support staff, study participants and the sponsor's clinical research personnel, until final data lock. Except for the individual participant identification number on the label, the two study solutions were identical clear, colourless liquids packaged in opaque bottles.
Statistical analysis
Safety analyses were performed on all randomised participants who received at least one dose of study solution. There was no imputation of missing safety data. Efficacy analyses were performed on an intent-to-treat (ITT) group, defined as a subset of all randomised participants who met critical inclusion criteria (primary OA by history, an abnormal radiological study, and any degree of knee pain), as per ICH guidelines [17]. For any missing efficacy data in the ITT analysis, the last observation was carried forward. A per-protocol group was defined based on stricter adherence to study conduct, including requirement for a moderate flare of knee pain (see Participants) and treatment continuing for at least 40 days.
Baseline demographic and clinical variables were analysed by Chi-square or Student's t-Test. Adverse event incidence was analysed by Chi-square or Fisher's Exact Test. Continuous variables (WOMAC dimensions, PGA and pain on walking) were analysed by ANCOVA with baseline score as the covariate without adjustment for testing secondary/alternative objectives. The dichotomous variables were analysed by Chi-square test. All statistical tests were two-sided and were performed at the 0.05 level of significance.
Results
Participant flow
Two hundred and sixteen participants were randomised to treatment with either topical diclofenac (n = 107) or vehicle control (n = 109). All participants received their allocated intervention. More participants in the topical diclofenac group (86 [80%]) completed the entire 6-week treatment period compared to the vehicle control group (70 [64%]; p = 0.008). Discontinuation rate due to an adverse event was similar in both groups. Dropout due to lack of effect was lower for topical diclofenac (8 [7.5%]) compared to vehicle control (18 [16.5%]; p = 0.041). No participant was lost to follow-up (Fig. 1).
Figure 1 Flow of participants.
Baseline demographic and clinical characteristics
No significant difference was found between treatment groups in baseline demographic and clinical characteristics (Table 1). The mean (SD) screening and baseline pain scores were 8.2 (2.7) and 13.0 (3.2) in the topical diclofenac group versus 8.3 (3.0) and 12.8 (3.1) in the vehicle control group (12.9 [3.2] overall). Most participants treated both knees, either from baseline or by the end of the trial.
Table 1 Baseline demographic and clinical characteristics of treatment groups
Topical diclofenac (n = 107) Vehicle control (n = 109)
Age (years) 65.0 (11.0) 64.6 (10.9)
Women, number (%) 56 (52.3) 66 (60.6)
Race/ethnicity, number (%)
White 88 (82.2) 91 (83.5)
Black 8 (7.5) 3 (2.8)
Oriental 3 (2.8) 2 (1.8)
Other 8 (7.5) 13 (11.9)
Weight (kg) 89.9 (18.1) 86.5 (17.3)
Height (m) 1.65 (0.11) 1.65 (0.10)
Heart rate (bpm) 74.1 (10.0) 74.3 (9.1)
Systolic blood pressure (mm Hg) 137.6 (16.3) 133.6 (15.6)
Diastolic blood pressure (mm Hg) 81.4 (9.1) 79.7 (8.7)
Total x-ray score* 7.7 (5.4) 7.0 (5.0)
Screening pain score 8.2 (2.7) 8.3 (3.0)
Baseline† pain score 13.0 (3.2) 12.8 (3.1)
Baseline† physical function score 40.7 (11.9) 40.4 (11.2)
Baseline† stiffness score 5.2 (1.5) 5.2 (1.5)
Patient global assessment score‡ 3.1 (0.8) 3.2 (0.8)
Participants treating two knees at baseline, number (%) 64 (59.8) 70 (64.2)
Participants treating two knees at final, number (%) 84 (78.5) 89 (81.7)
Data are presented as mean (SD) unless otherwise indicated.
*Total score of joint space narrowing, marginal osteophytes formation and subchondrial sclerosis for each knee compartment (medial, lateral, patello-femoral); maximum score possible was 27.
†After washout of prior therapy; pain scale ranged from 0 (no pain) to 20 (extreme pain); physical function scale ranged from 0 (no difficulty) to 68 (extreme difficulty); stiffness scale ranged from 0 (no stiffness) to 8 (extreme stiffness).
‡Patient global assessment was measured using a Likert scale, ranging from 0 (very good) to 4 (very poor).
Mean (SD) duration of treatment in the topical diclofenac group was 38.2 (9.9) days versus 34.4 (12.5) days in the vehicle control group (p = 0.013). Compliance with the dosing regime was 83.1 % and 84.5% for the topical diclofenac and vehicle control groups, respectively. No significant difference was noted in the mean (SD) consumption of rescue acetaminophen tablets per day between the topical diclofenac (0.9 [0.9]) and vehicle control groups (1.1 [1.0]; p = 0.079).
Efficacy analyses
Four of 216 randomized participants were not included in the ITT analysis group because they violated major entry criteria: 2 participants lacked radiological confirmation of OA (no radiological examination for one participant and a normal examination for the other), and 2 participants had secondary OA (related to osteochondroma). Inclusion of these participants yielded the same results in a subsequent re-analysis (data not shown).
Planned analyses
There was a significantly greater improvement in score with topical diclofenac compared to vehicle control (Table 2) for pain (-5.2 vs. -3.3; p = 0.003,), physical function (-13.4 vs. -6.9; p = 0.001), PGA (-1.3 vs. -0.7; p = 0.0001) and stiffness (-1.8 vs. -0. 9; p = 0.002) Analysis of the per protocol group of 128 participants confirmed the statistical superiority of topical diclofenac over vehicle control for the primary and secondary outcome measures (p < 0.01; data not shown).
Table 2 Efficacy evaluation of the continuous variables
Efficacy variable Treatment group N Baseline score, mean (SD) Change in score mean (SD) Mean difference in change (95% CI) P-value Effect size (95% CI)
Pain Topical diclofenac 105 13.0 (3.1) -5.2 (5.0) 1.9 (0.7 to 3.2) 0.003 0.41 (0.14 to 0.68)
Vehicle control 107 12.7 (3.2) -3.3 (4.3)
Physical function Topical diclofenac 105 40.9 (11.9) -13.4 (16.3) 6.5 (2.5 to 10.5) 0.001 0.44 (0.16 to 0.71)
Vehicle control 107 40.3 (11.3) -6.9 (13.2)
Patient global assessment Topical diclofenac 105 3.1 (0.8) -1.3 (1.3) 0.6 (0.2 to 0.9) 0.0001 0.47 (0.19 to 0.74)
Vehicle control 107 3.2 (0.7) -0.7 (1.1)
Stiffness Topical diclofenac 105 5.3 (1.4) -1.8 (2.1) 0.9 (0.3 to 1.4) 0.002 0.43 (0.15 to 0.70)
Vehicle control 107 5.2 (1.5) -0.9 (2.0)
Pain on walking Topical diclofenac 105 2.7 (0.8) -1.2 (1.2) 0.4 (0.1 to 0.7) 0.014 0.34 (0.07 to 0.61)
Vehicle control 107 2.7 (0.8) -0.8 (1.1)
A posteriori analyses
There was a significantly greater improvement in score with topical diclofenac compared to vehicle control (Table 2) for pain on walking (-1.2 vs. -0.8; p = 0.014). The response rate for at least a 50% reduction in pain (Table 3) was significantly greater following topical diclofenac treatment compared to vehicle control (46/105 [43.8%] vs. 27/107 [25.2%]; p = 0.004). The topical diclofenac group had a significantly greater number of participants with good or very good PGA response (43.8% vs. 16.8%; p <0.0001) compared to the vehicle control group and of OMERACT-OARSI responders (65.7% vs. 49.5%; p = 0.017).
Table 3 Efficacy evaluation of the dichotomous variables
Efficacy variables Treatment group N Number (%) of participants p-value Number-needed-to-treat (95% CI)
50% reduction in pain Topical diclofenac 105 46 (43.8) 0.004 5 (3–17)
Vehicle control 107 27 (25.2)
OMERACT-OARSI responder* Topical diclofenac 105 69 (65.7) 0.017 6 (3–33)
Vehicle control 107 53 (49.5)
Good or very good PGA response Topical diclofenac 105 46 (43.8) <0.0001 4 (3–7)
Vehicle control 107 18 (16.8)
*A responder is defined as a participant with ≥ 50% improvement in pain or function that was ≥ 20% of the scale, or ≥ 20% improvement in at least two of pain, function or patient global assessment that was ≥ 10% of the scale.
Adverse events
The major adverse effect reported was dry skin at the application site, occurring in 42/107 (39.3%) and 23/109 (21.1%; p = 0.004) of topical diclofenac and vehicle control participants, respectively (Table 4). A skin-related adverse event led to discontinuation of only 5 participants in the topical diclofenac group. All skin reactions resolved promptly upon withdrawal of treatment. Abdominal pain and dyspepsia each were reported in 4 [3.7%] participants in the topical diclofenac group compared to 1 [0.9%] participant in the vehicle control group, but this difference was not significant (p = 0.21).
Table 4 Number (%) of adverse events
Adverse Event Topical diclofenac (n = 107) Vehicle control (n = 109)
Gastrointestinal reaction
Abdominal pain 4 (3.7) 1 (0.9)
Constipation 1 (0.9) 1 (0.9)
Diarrhea 1 (0.9) 0
Dyspepsia 4 (3.7) 1 (0.9)
Gastritis 1 (0.9) 0
Melena 0 1 (0.9)
Nausea 1 (0.9) 2 (1.8)
Application-site skin reaction
Dry skin 42 (39.3)* 23 (21.1)
Rash 2 (1.9) 4 (3.7)
Paresthesia 2 (1.9) 2 (1.8)
Pruritus 0 2 (1.8)
Other reaction
Headache 6 (5.6) 10 (9.2)
Halitosis 2 (1.9) 0
Taste Perversion 4 (3.7) 2 (1.8)
*p < 0.01 vs. vehicle control
Discussion
Published guidelines have incorporated topical NSAIDs as recommended treatment for OA of the knee [4-6]. However, there has been controversy surrounding the adequacy of data supporting their benefit beyond 2 weeks [2,3,8,18]. Moreover, the studies identified in these meta-analyses generally did not conform to current standards for OA trial design [11,19]. In contrast, the present trial utilized standardized radiological and clinical entry criteria and measured efficacy with validated outcome measures. Baseline pain score was substantial; mean (SD) score was 12.9 (3.2) out of a maximum of 20, indicating a flare of pain following withdrawal of prior therapy. Analysis of all of the primary and secondary measures demonstrated that treatment with this topical diclofenac solution relieved the symptoms of primary knee OA at 6 weeks in this study population. Two other recently published trials using this topical diclofenac solution showed it to be superior to vehicle control and/or placebo; a 4-week, non-flare trial of 248 participants [20] and a 12-week, flare trial of 326 participants [21]. As with most NSAID trials, the subject population in this study was selected by the inclusion criterion of a flare of pain, which demonstrates the potential to respond to NSAID/analgesic. In clinical practice, an individual not taking an analgesic may have considered previous NSAID therapy ineffective, in which case s/he would not be expected to respond to topical diclofenac. However, where an individual is intolerant to oral NSAID, one may consider topical diclofenac as a treatment option.
Comparison of efficacy results from independent trials with various treatments is facilitated by the introduction of benchmark determinants that are mathematically derived from the experimental raw data, such as effect size [22] for improvement of a continuous variable (e.g. "How much did the patient's pain improve, relative to placebo?"). We calculated an effect size (95% CI) of 0.41 (0.14 to 0.68) for pain relief, 0.44 (0.16 to 0.71) for improved physical function and 0.34–0.47 for improved measures of PGA, stiffness and pain on walking (Table 2). In contrast, Lin et al. [8] calculated a pooled effect size for pain relief of 0.04 (essentially no effect) in 3 placebo-controlled topical NSAID trials of 4 weeks duration. A meta-analysis of 23 oral NSAID trials for OA knee, lasting 2–13 weeks, reported a pooled effect size of 0.32 for pain reduction and 0.29 for improving physical function [23]. Another meta-analysis of 14 OA trials found a pooled effect size of 0.37 for pain reduction with oral NSAIDs and 0.44 for coxibs [24]. Zhang et al. [25], using data from 2 oral NSAID studies of 6–12 weeks duration, calculated a pooled effect size for OA pain reduction of 0.34.
Efficacy of a treatment is being expressed increasingly as a dichotomous result, e.g. "Did the patient's pain improve by 50%; yes or no?". We derived the response rate for each dichotomous variable from our raw data, and demonstrated the superiority of topical diclofenac over vehicle control for 50% reduction in pain, achieving a good or very good final PGA response, and 'response' by OMERACT-OARSI criteria (Table 3). The benchmark determinant for comparing dichotomous efficacy results of various treatments is the number-needed-to-treat (NNT) [26]. We calculated a NNT between 4 and 6, depending upon the variable (Table 3). In their meta-analysis of topical NSAIDs, Mason et al. [3] cited 5 placebo-controlled trials of short duration for OA knee pain – 8 days (1 trial), 14 days (3 trials), and 28 days (1 trial). Their definition of clinical success, representing approximately a 50% reduction of pain, was estimated using patient or physician global assessment as the outcome measure (4 trials and 1 trial, respectively). They calculated a NNT of 5.3.
Few oral NSAID studies have reported dichotomous data. Osiri et al. [26] reported a NNT for pain improvement of 4.4 with etodolac and 3.8 with tenoxicam. Defining improvement as an increase of at least 2 grades (on a 0–5 scale) in the patient's global rating of arthritis, Edwards et al. [27] reported a NNT of 11–13 for valdecoxib treatment of OA.
The OMERACT-OARSI initiative used a consensus approach to derive dichotomous 'responder' criteria [14]. Through their vast meta-analysis of suitable trials, the authors found that for trials of oral NSAIDs vs. placebo the responder rates were 65.4% and 45.9% respectively. Responder rates of 60–65% have been reported for 13-week treatment of OA with celecoxib and lumiracoxib, with placebo responder rates of 49–53% [28,29]. The OMERACT-OARSI initiative did not look at topicals but we applied its criteria to this study and found a responder rate for topical diclofenac of 65.7% with a placebo responder rate of 49.5%, similar to their oral NSAID data.
A caveat in the application of the mathematical benchmarks, effect size and NNT, is the influence of trial design, outcome measures and patient population on the apparent magnitude of response to a given treatment. Because the trials with topical diclofenac were designed according to the OARSI guidelines, like most recent NSAID and cyclooxygenase-2 (COX-2) inhibitor studies, such comparison of results is reasonable [19]. Although the data observed for topical diclofenac in this trial are comparable to other NSAID trials, a direct head-to-head comparison trial is required to prove equivalency of two treatments. A previously published 12-week comparative trial of 622 participants with OA knee confirmed the clinical equivalence between topical diclofenac solution and oral diclofenac [30].
Safety analysis revealed no serious clinical adverse effects and only minor application-site skin reactions, mostly skin dryness, following treatment with topical diclofenac. While dimethyl sulphoxide in the carrier acts as a penetrant [31], it also dissolves normal surface oils and leaves the skin dry. Common skin lubricants may prevent most application site reactions and any related discontinued therapy, but such products were not permitted in this trial in order to detect the maximum potential side effect profile of the study solutions. The low dropout rate due to skin reactions (5/107 [4.7%] for topical diclofenac) suggests patient acceptance of the overall topical treatment regime.
The use of a checklist to prompt the patient about possible adverse events likely yielded a high estimate of the true incidence of gastrointestinal adverse reactions caused by topical diclofenac. The report of abdominal pain and dyspepsia each in 3.7% of patients is consistent with what was seen in other published trials of this topical diclofenac [20,21] and much lower than commonly experienced with oral NSAIDs or COX-2s [30]. Those other trials included results of laboratory testing and found minor abnormality of liver enzymes in 2–5%, creatinine in 1% and haemoglobin in 2% of patients, significantly lower than with oral diclofenac [30]. This safety profile can be predicted from the low systemic availability of topically applied diclofenac. Although the patient applies a daily dose (40 drops, 4 times a day) of 86 mg of diclofenac to the knee, the blood level is only 12 ng/mL [31]. The level reported after oral administration of 50 mg Voltaren® is 1500 ng/mL [32]. Similar improved safety with topical NSAIDs has been reported previously [33].
Conclusion
Topical diclofenac solution provides 6-week relief of the symptoms of knee OA. The data in this and previous reports provide substantial evidence for the efficacy and safety of topical diclofenac solution in chronic OA.
Competing interests
LMT and ZS are employees of Dimethaid Research Inc.
PAB was a principal investigator in the trial, and was remunerated for his participation.
Authors' contributions
PAB was a major investigator in the trial and was involved in data interpretation. LMT was involved in analysis and interpretation of the data and writing of the manuscript. ZS was involved in trial design and conduct, data review and writing of the manuscript. All authors reviewed and approved the final draft of the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This study was supported by Dimethaid Health Care Ltd. We thank all trial site investigators for their dedication to this study.
==== Refs
Mason L Moore RA Edwards JE Derry S McQuay HJ Topical NSAIDs for acute pain: a meta-analysis BMC Fam Pract 2004 5 10 15147585 10.1186/1471-2296-5-10
Moore RA Tramer MR Carroll D Wiffen PJ McQuay HJ Quantitative systematic review of topically applied non-steroidal anti-inflammatory drugs BMJ 1998 316 333 338 9487165
Mason L Moore RA Edwards JE Derry S McQuay HJ Topical NSAIDs for chronic musculoskeletal pain: systematic review and meta-analysis BMC Musculoskelet Disord 2004 5 28 15317652 10.1186/1471-2474-5-28
Jordan KM Arden NK Doherty M Bannwarth B Bijlsma JW Dieppe P Gunther K Hauselmann H Herrero-Beaumont G Kaklamanis P Lohmander S Leeb B Lequesne M Mazieres B Martin-Mola E Pavelka K Pendleton A Punzi L Serni U Swoboda B Verbruggen G Zimmerman-Gorska I Dougados M EULAR Recommendations 2003: an evidence based approach to the management of knee osteoarthritis: Report of a Task Force of the Standing Committee for International Clinical Studies Including Therapeutic Trials (ESCISIT) Ann Rheum Dis 2003 62 1145 1155 14644851 10.1136/ard.2003.011742
Walker-Bone K Javaid K Arden N Cooper C Regular review: medical management of osteoarthritis BMJ 2000 321 936 940 11030685 10.1136/bmj.321.7266.936
Altman RD Hochberg MC Moskowitz RW Schnitzer TJ Recommendations for the medical management of osteoarthritis of the hip and knee: New OA practice management guidelines from the ACR Arthritis Rheum 2000 43 1905 1915 11014340 10.1002/1529-0131(200009)43:9<1905::AID-ANR1>3.0.CO;2-P
Evans JM McMahon AD McGilchrist MM White G Murray FE McDevitt DG MacDonald TM Topical non-steroidal anti-inflammatory drugs and admission to hospital for upper gastrointestinal bleeding and perforation: a record linkage case-control study BMJ 1995 311 22 26 7613317
Lin J Zhang W Jones A Doherty M Efficacy of topical non-steroidal anti-inflammatory drugs in the treatment of osteoarthritis: meta-analysis of randomised controlled trials BMJ 2004 329 324 15286056 10.1136/bmj.38159.639028.7C
Altman RD Hochberg M Murphy WAJ Wolfe F Lequesne M Atlas of individual radiographic features in osteoarthritis Osteoarthritis Cartilage 1995 3 Suppl A 3 70 8581752
Bellamy N WOMAC Osteoarthritis Index - A User's Guide 1995 London, Canada, London Health Sciences Centre
Bellamy N Kirwan J Boers M Brooks P Strand V Tugwell P Altman R Brandt K Dougados M Lequesne M Recommendations for a core set of outcome measures for future phase III clinical trials in knee, hip, and hand osteoarthritis. Consensus development at OMERACT III J Rheumatol 1997 24 799 802 9101522
Bellamy N Buchanan WW Goldsmith CH Campbell J Stitt LW Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to antirheumatic drug therapy in patients with osteoarthritis of the hip or knee J Rheumatol 1988 15 1833 1840 3068365
Recommendations for the registration of drugs used in the treatment of osteoarthritis. Group for the respect of ethics and excellence in science (GREES): osteoarthritis section Ann Rheum Dis 1996 55 552 557 8774185
Pham T van der Heijde D Altman RD Anderson JJ Bellamy N Hochberg M Simon L Strand V Woodworth T Dougados M OMERACT-OARSI initiative: Osteoarthritis Research Society International set of responder criteria for osteoarthritis clinical trials revisited Osteoarthritis Cartilage 2004 12 389 399 15094138 10.1016/j.joca.2004.02.001
Marzulli FN and Maibach HI Contact allergy: predictive testing in humans Dermatotoxicology and pharmacology Advances in modern toxicology 1977 Washington D.C., Hemisphere Publishing 353 372
Administration USFD COSTART: Coding Symbols for Thesaurus of Adverse Reaction Terms 1993 Springfield, VA, National Technical Information Service, US Department of Commerce
International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use E9: Statistical Principles for Clinical Trials 1998
Cooper C Jordan KM Topical NSAIDs in osteoarthritis BMJ 2004 329 304 305 doi:10.1136/bmj.329.7461.304 (published 30 July 2004) 15297323 10.1136/bmj.329.7461.304
Altman R Brandt K Hochberg M Moskowitz R Bellamy N Bloch DA Buckwalter J Dougados M Ehrlich G Lequesne M Lohmander S Murphy WAJ Rosario-Jansen T Schwartz B Trippel S Design and conduct of clinical trials in patients with osteoarthritis: recommendations from a task force of the Osteoarthritis Research Society. Results from a workshop Osteoarthritis Cartilage 1996 4 217 243 11048620
Bookman AA Williams KS Shainhouse JZ Effect of a topical diclofenac solution for relieving symptoms of primary osteoarthritis of the knee: a randomized controlled trial CMAJ 2004 171 333 338 15313991
Roth SH Shainhouse JZ Efficacy and safety of a topical diclofenac solution (Pennsaid) in the treatment of primary osteoarthritis of the knee: a randomized, double-blind, vehicle-controlled clinical trial Arch Intern Med 2004 164 2017 2023 15477437 10.1001/archinte.164.18.2017
Kazis LE Anderson JJ Meenan RF Effect sizes for interpreting changes in health status Med Care 1989 27 S178 S189 2646488
Bjordal JM Ljunggren AE Klovning A Slordal L Non-steroidal anti-inflammatory drugs, including cyclo-oxygenase-2 inhibitors, in osteoarthritic knee pain: meta-analysis of randomised placebo controlled trials BMJ 2004 329 1317 15561731 10.1136/bmj.38273.626655.63
Lee C Hunsche E Balshaw R Kong SX Schnitzer T Importance of evaluating common internal controls when assessing the relative efficacy of different pharmacologic agents through a metaanalytic approach: a case study of COX-2-selective inhibitors in the treatment of osteoarthritis Osteoarthritis Cartilage 2004 12 (Supplement B) S74 S75
Zhang W Jones A Doherty M Does paracetamol (acetaminophen) reduce the pain of osteoarthritis? A meta-analysis of randomised controlled trials Ann Rheum Dis 2004 63 901 907 doi 10.1136/ard.2003.018531 (published 5 March 2004) 15020311 10.1136/ard.2003.018531
Osiri M Suarez-Almazor ME Wells GA Robinson V Tugwell P Number needed to treat (NNT): implication in rheumatology clinical practice Ann Rheum Dis 2003 62 316 321 12634229 10.1136/ard.62.4.316
Edwards JE McQuay HJ Moore RA Efficacy and safety of valdecoxib for treatment of osteoarthritis and rheumatoid arthritis: systematic review of randomised controlled trials Pain 2004 111 286 296 15363872 10.1016/j.pain.2004.07.004
Sheldon E Beaulieu A Paster Z Dutta D Yu S Lumiracoxib 100 mg once daily is effective in treating osteoarthritis as assessed by OMERACT-OARSI Criteria Osteoarthritis Cartilage 2004 12 (Supplement B) S140 S141
Ouellet JP Zacher J Moore A Lee A V.S. S Long-term efficacy of lumiracoxib in osteoarthritis demonstrated using the OMERACT-OARSI criteria Osteoarthritis Cartilage 2004 12 (Supplement B) S84 S85
Tugwell PS Wells GA Shainhouse JZ Equivalence study of a topical diclofenac solution (Pennsaid) compared with oral diclofenac in symptomatic treatment of osteoarthritis of the knee: a randomized controlled trial J Rheumatol 2004 31 2002 2012 15468367
Hui X Hewitt PG Poblete N Maibach HI Shainhouse JZ Wester RC In vivo bioavailability and metabolism of topical diclofenac lotion in human volunteers Pharm Res 1998 15 1589 1595 9794502 10.1023/A:1011911302005
Association CP Voltaren Product Monograph Compendium of Pharmaceuticals and Specialties 2004 Ottawa, ON 2203 2204
Heyneman CA Lawless-Liday C Wall GC Oral versus topical NSAIDs in rheumatic diseases: a comparison Drugs 2000 60 555 574 11030467
|
16086839
|
PMC1201146
|
CC BY
|
2021-01-04 16:32:04
|
no
|
BMC Musculoskelet Disord. 2005 Aug 8; 6:44
|
utf-8
|
BMC Musculoskelet Disord
| 2,005 |
10.1186/1471-2474-6-44
|
oa_comm
|
==== Front
BMC PediatrBMC Pediatrics1471-2431BioMed Central London 1471-2431-5-251602949310.1186/1471-2431-5-25Study ProtocolA population-based nested case control study on recurrent pneumonias in children with severe generalized cerebral palsy: ethical considerations of the design and representativeness of the study sample
Veugelers Rebekka [email protected] Elsbeth AC [email protected] Corine [email protected] Arianne [email protected] Roos [email protected] Jan [email protected] Marc A [email protected] Peter JFM [email protected] Hubertus GM [email protected] Dick [email protected] Heleen M [email protected] Intellectual Disability Medicine, department of General Practice Erasmus MC, PO Box 1738, 3000 DR Rotterdam, The Netherlands2 Department of General Practice Erasmus MC, PO Box 1738, 3000 DR Rotterdam, The Netherlands3 Department of Paediatric Gastro-enterology Erasmus MC, PO Box 1738, 3000 DR Rotterdam, The Netherlands4 Department of Paediatric Gastro-enterology and Nutrition Academic Medical Centre / Emma's Children's Hospital, G8 217, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands5 Department of Paediatric Pulmonology Erasmus MC, PO Box 1738, 3000 DR Rotterdam, The Netherlands6 Department of Paediatric Pulmonology UMC, HP KH.01.419.0, PO Box 85590, 3508 AB Utrecht, The Netherlands7 Department of Paediatric Surgery Erasmus MC, Sophia Children's Hospital, PO Box 1738, 3000 DR Rotterdam, The Netherlands2005 19 7 2005 5 25 25 8 6 2005 19 7 2005 Copyright © 2005 Veugelers et al; licensee BioMed Central Ltd.2005Veugelers 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 children with severe generalized cerebral palsy, pneumonias are a major health issue. Malnutrition, dysphagia, gastro-oesophageal reflux, impaired respiratory function and constipation are hypothesized risk factors. Still, no data are available on the relative contribution of these possible risk factors in the described population. This paper describes the initiation of a study in 194 children with severe generalized cerebral palsy, on the prevalence and on the impact of these hypothesized risk factors of recurrent pneumonias.
Methods/Design
A nested case-control design with 18 months follow-up was chosen. Dysphagia, respiratory function and constipation will be assessed at baseline, malnutrition and gastro-oesophageal reflux at the end of the follow-up. The study population consists of a representative population sample of children with severe generalized cerebral palsy. Inclusion was done through care-centres in a predefined geographical area and not through hospitals. All measurements will be done on-site which sets high demands on all measurements. If these demands were not met in "gold standard" methods, other methods were chosen. Although the inclusion period was prolonged, the desired sample size of 300 children was not met. With a consent rate of 33%, nearly 10% of all eligible children in the Netherlands are included (n = 194). The study population is subtly different from the non-participants with regard to severity of dysphagia and prevalence rates of pneumonias and gastro-oesophageal reflux.
Discussion
Ethical issues complicated the study design. Assessment of malnutrition and gastro-oesophageal reflux at baseline was considered unethical, since these conditions can be easily treated. Therefore, we postponed these diagnostics until the end of the follow-up. In order to include a representative sample, all eligible children in a predefined geographical area had to be contacted. To increase the consent rate, on-site measurements are of first choice, but timely inclusion is jeopardised. The initiation of this first study among children with severe neurological impairment led to specific, unexpected problems. Despite small differences between participants and non-participating children, our sample is as representative as can be expected from any population-based study and will provide important, new information to bring us further towards effective interventions to prevent pneumonias in this population.
==== Body
Background
Children with severe generalized cerebral palsy often have a combination of motor and intellectual disabilities. They frequently experience co-morbidity and their life expectancy is low [1-11] with respiratory disease as a main cause of death [1-3,8,10,12]. Although it is common clinical knowledge that children with neurological impairment often have respiratory problems [13-17], get hospitalised for this [18] with a major impact on their quality of life and life expectancy [14], prevalence rates have not been studied prospectively. Retrospective prevalence estimates of pneumonias range from 31% per 6 months; 38% single episodes to 19% recurrent pneumonias per year [19,20]. Although several clinical specialists presume several conditions to be risk factors for pneumonias, population-based studies on this subject are lacking. Epidemiological identification of such risk factors will bring us further towards effective interventions to prevent pneumonias.
Hypothesized risk factors of respiratory disease in children / adolescents with neurological impairment / intellectual disabilities from the literature are listed in Table 1. These factors may co-exist and interact with each other. On top of this, normal childhood factors may exist, such as asthma or passive smoking. Pneumonias can be infectious or chemical of nature. To prevent pneumonias, adequate function of the protection mechanisms of the airways is essential. But in children with severe generalized cerebral palsy this protection system is often compromised or endangered due to several conditions [14,15,20-29]
Table 1 Hypothesized risk factors of pulmonary disease in children with neurological impairment / intellectual disabilities
recurrent aspiration (dysphagia, gastro-oesophageal reflux) [14-16, 20, 28, 53, 54]
inefficient cough / poor cough reflex [14, 15, 28]
poor airway clearance (immobility and retained secretions) [14, 15]
respiratory muscle weakness and in-coordination [14, 15, 28]
chest wall or spinal deformities (poor pulmonary reserve) [14, 15, 28]
inadequate nutritional status (feeding problems, gastro-oesophageal reflux) [14, 15]
miscellaneous factors [2, 8, 10, 14-17]
bronchopulmonary dysplasia in preterm survivors
immune problems (Down's syndrome)
lipid aspiration in mineral oil treatment of constipation
reduced lung growth in skeletal dysplasias
normal childhood factors (e.g. asthma, passive smoking) [14, 15]
immobility [3, 10, 27, 28, 55, 56]
We hypothesize that malnutrition, dysphagia, gastro-oesophageal reflux, decreased respiratory function and constipation are the most relevant risk factors for recurrent pneumonias. Since scientific evidence for a relationship between these disorders and the occurrence of pneumonias is lacking, we aim to evaluate this in a large-scale epidemiological study. Our research questions are the following: (1) What is the prevalence of pneumonias in children with severe generalized cerebral palsy? (2) Are malnutrition, dysphagia, gastro-oesophageal reflux, decreased respiratory function and constipation risk factors for pneumonias in this group of children? The design of the study also allows us to determine the prevalence and presentation of the studied hypothesized risk factors.
This article describes the study design, diagnostic methods and the study population. Attention is paid to adaptations in the study design arising from ethical considerations as well as from the diagnostic methods required to study medical conditions in children with severe generalized cerebral palsy.
Methods / Design
Study design
This study has a nested case-control design and will be conducted in a representative group of children with severe generalized cerebral palsy, recruited through care centres (specialized day-care centres and residential facilities) and through specialized schools. In our study population, the hypothesized risk factors dysphagia, respiratory function and constipation will be assessed at baseline. However, for ethical reasons explained in the discussion paragraph, malnutrition and gastro-oesophageal reflux will be assessed at the end of the follow-up period. Cases are defined as children with recurrent pneumonias, and controls as children without pneumonias during a follow-up of 18 months. Cases and controls are matched on age, gender and GMFCS level. A duration of the follow-up period of 18 months was considered sufficient, since we defined recurrent pneumonias as 2 or more episodes within a year. The study will not interfere with common medical practice and interventions in the study population during the follow-up period. Thus, children might be diagnosed and treated by their own physicians during the course of the study. The study design is depicted in Figure 1.
Figure 1 Study design. In this nested case-control study, a cohort of 194 children with severe generalized cerebral palsy is followed up for 18 months in order to record recurrent pneumonias (2 or more episodes per year). Possible risk factors are measured during the follow-up. Dysphagia, constipation and pulmonary function are diagnosed at baseline, while nutritional state and gastro-oesophageal reflux are diagnosed at the end of the study period.
Setting
All diagnostic assessments in this study will be carried out on-site at the different care centres and specialized schools. In order to obtain a complete inclusion and therewith a representative study population, we had to keep the burden for the participants as small as possible. Hospital visits were considered an obstacle for participation. Furthermore, performing measurements in a familiar setting might improve cooperation of the children.
Sample size
Calculating a required sample size for this study was hampered, since valid prevalence numbers of both pneumonias and most of the supposed risk factors in this population, were lacking in the literature. Prevalence numbers were estimated based on the available literature and on clinical experience. We calculated the required sample size for a univariate analysis, since the number of children required for a multivariate analysis including five separate variables will probably be quite large. In addition, we estimated that for logistical purposes a maximum number of 300 children could be included in this study. Required sample size was calculated for each possible risk factor separately, assuming a prevalence rate of recurrent pneumonias of 30% with a required power of 0.80 and an alpha of 0.05. The analysis for dysphagia, based on an estimated prevalence of dysphagia of 19% in the controls and 38% in the cases, resulted in the highest sample size (n = 260). Assuming a loss-to-follow-up rate of 13%, recruitment numbers were set to 300 participants.
Inclusion criteria
In this study we aimed to include children (2 to 18 years), who have a combination of moderate to profound intellectual disabilities and a severe motor disability. The intellectual disability was defined as an IQ below 55 (or estimated by dividing the developmental age by the calendar age times 100). The motor disability was defined by hypertonic or hypotonic generalized cerebral palsy or a motor developmental delay to such an extent that a child can at best crawl. This corresponds to a Gross Motor Function Classification Scale (GMFCS) score IV or V [30]. These broad criteria, resulting in a heterogeneous cohort with regard to aetiology and disabilities, was chosen deliberately, because in daily practice, it is this heterogeneous group that causes a lot of concern for parents and physicians regarding the studied illnesses. Furthermore, the inclusion criteria had to be clear to non-medical personnel, to ascertain they could identify the eligible children.
Consent procedure
We approached all children with severe generalized cerebral palsy in a certain geographical area, an important prerequisite when studying a prevalence rate, to obtain a representative sample of the total population. For pragmatic reasons, we chose an area of 50 kilometres around the cities of Rotterdam and Utrecht. We estimated that we could reach 500 children in this area. With an assumed consent rate of 0.60, this would provide the desired 300 participants. Within this area, we traced all facilities that might provide care to children and adolescents with severe generalized cerebral palsy, using the Dutch address guide for disability care. These centres were contacted and asked to participate in the study if they indeed provided care for such children. In the participating centres, parents or guardians of all children that met the inclusion criteria were informed, unless children were in a critical health status, when home situations were considered very unstable, or if parents were known to have a strong aversion to research. Information for parents was available in Dutch, English, and Turkish. For Moroccan families, a spoken introductory compact disc was available, since Berber is only a spoken language. Because gastro-oesophageal reflux can only be measured properly using an invasive method, parents had the opportunity to give consent with or without this measurement.
Inclusion period
Of the 93 care centres and specialized schools that had been contacted, 61 provided care for one or more children with severe generalized cerebral palsy. Fifty-six of these centres agreed to participate in our study. The other centres did not cooperate due to personnel shortage and besides this, one centre also considered the burden of the study for parents, children and personnel too large.
Participants
Within the participating care centres and specialized schools, 593 children were eligible for participation. Parents of 573 children were informed while the parents of 9 children were not contacted based on the previously mentioned reasons and 11 were not contacted because of ineffective internal procedures of care centres. Four children, for whom consent was given, appeared not to meet our inclusion criteria at first visit and were excluded. After a prolonged inclusion period of 20 months, this resulted in the informed consent for 194 children (consent rate of 33%). Although recruitment numbers were set to 300 participants, we stopped the inclusion for practical reasons. We had included nearly 10% of the Dutch population of children with severe generalized cerebral palsy [31]. Parents of 98 children gave consent including assessment of gastro-oesophageal reflux (Figure 2). Because of the broad inclusion criteria, not all children fulfilled the strict definition of cerebral palsy [32], but all children had comparable disabilities. The different aetiologies of the disabilities of the participants are depicted in Table 2. Basic characteristics of the participants are listed in Table 3. All participating parents that gave consent preferred the questionnaires in Dutch, even when their native language was Turkish.
Table 2 Aetiology of disabilities
n %
Congenital diseases
Miller Dieker Syndrome / lissencephaly 7
corpus callosum agenesis 5
Cornelia de Lange syndrome 2
Walker-Warburg syndrome 2
unspecified abnormal brain development 16
other non progressive syndromes 6
other chromosomal abnormalities 9
Rett syndrome 3
Alpers syndrome 4
Aicardi-Goutieres syndrome 2
other progressive syndromes 5
other congenital diseases 4
65 33.5
Pre and perinatal complications
perinatal asphyxia 18
cerebral palsy e.c.i. 13
cerebral haemorrhage 6
intra uterine CMV infection 5
other infections 4
other causes 7
53 27.3
Acquired
meningitis / encephalitis 5
Trauma 3
near drowning accident 2
Other 2
12 6.2
Combinations of causes
congenital and acquired disease 6
congenital disease and perinatal complications 5
perinatal and acquired 3
perinatal and hereditary progressive 1
15 7.7
Unknown cause
25 12.8
Missing
24 12.3
Total 194 children
Table 3 Characteristics of the participants
% valid*
GMFCS score V 82.7 0.95
Can communicate "yes" and "no" 20.6 0.87
Can verbally communicate "yes" and "no" 3.1 0.87
Living with parents at home 81.4 1
Intentional movements none 34.8
little 27.9
regularly 37.7 0.66
Involuntary movements most of the day 29.6
regularly 35.2
< 2 hours a week 35.2 0.64
Seated > 3 hours / day 84.5 0.68
Standing < 30 minutes / week 38.3 0.59
Activity < 30 min / day 51.3 0.58
GMFCS = Gross Motor Function Classification Scale, *valid = fraction of the population with known information
Figure 2 Flow chart of inclusion period. This figure depicts the inclusion of eligible children in the study from a predefined geographical area. 593 children met our inclusion criteria and parents or guardians of 573 children were informed. For several reasons, parents of 20 children were not informed. For 194 children informed consent was obtained and for 98 of those with additional consent for assessment of gastro-oesophageal reflux. For 379 children no consent was obtained. Carers of 298 of these children filled in a small questionnaire. Of 101 children no information was obtained.
Representativeness
Global written information on children that did not participate was obtained from parents, care centres or specialized schools, concerning reasons for no consent, frequency of pneumonias, gastro-oesophageal reflux, body mass index and diet. To our clinical experience, parental judgement of eating skills is unreliable. Therefore we asked which food types the child received and reformulated this into a rough scale of dysphagia. Children were categorised as severe dysphagic if they received daily tube feeding, with or without additional oral food. Children with dietary restrictions (liquid, solid, ground, pureed) were categorised as having moderate dysphagia. All other children were categorised as having "no or mild" dysphagia.
Brief written information on children's characteristics was acquired for 298 of the non-participants (for 169 children from parents and for 129 children from the care centre and school personnel). Information from 101 children that were asked to participate (17%) is lacking. The main reported reasons for not participating were that parents were reluctant to any additional "hassle" with their child, mostly because of the extended medical history. Parents also considered the burden too large for themselves. Table 4 shows that the children that participate are slightly younger of age, and therewith have shorter height and lower body weight than the eligible children not participating in the study (BMI is not different between the groups). Gender is equally distributed. According to the parents' reports, the participating children have more severe dysphagia, more lower respiratory infections, and more gastro-oesophageal reflux than the non-participants.
Table 4 Comparison of the parent-reported characteristics between the participants and non-participants
Non-participants Participants
Valid* valid*
Total number 379 194
Mean Age (years) 10.6 (4.3) 0.67 8.9 (4.4) 1
Gender (% of boys) 50.2 0.7 53.1 1
Mean Height (cm) 130.3 (21.9) 0.52 124.0 (20.1) 0.91
Median Weight (kg) 28.0 [17.0] 0.59 24.7 [16.1] 0.88
Median BMI (kg/m2) 16.4 [4.2] 0.51 15.9 [4.0] 0.85
Dysphagia severe (%) 27.3 37.8
moderate (%) 17.7 51.2
no / mild (%) 55.0 0.68 11.0 0.65
Lower respiratory tract infections (%) 16.9 0.68 27.3 0.45
recurrent** (%) 12.5 0.67 18.2 0.45
Reported gastro-oesophageal reflux (%) 25.1 44.3 0.72
Standard deviations are between brackets, inter quartile range is between square brackets, *valid = fraction of the population with known information, ** recurrent = two or more episodes per year, BMI = body mass index.
Diagnostic methods
Diagnostic methods had to be chosen with great care. Because all assessments are performed on-site, diagnostic methods should be ambulatory available. Moreover, standard methods are often not feasible, due to the severity of the handicaps of these children, and the required level of cooperation. The Dutch ethics committee also demanded methods to be non invasive, if possible.
Pneumonia
In clinical practice, pneumonia is diagnosed based on a chest X-ray together with symptoms and signs. In the present study however, we needed to use a definition that could be used without requiring extra diagnostic procedures. A previous study showed that retrospective examination of medical files was not accurate for detection of pneumonias [33]. Therefore, the research team agreed upon the following definition for an episode of pneumonia: fever (> 38.5°C, or 1,5°C above basal temperature) during more than 24 hours, likely due to a pneumonia, characterized by: (increase of) dyspnoea (tachypnoea, use of assistant respiratory muscles, wheezing) during the last 6 hours, and/or (increase of) hyper secretion of mucus, and/or, tachypnoea and regular coughing. In addition, no other explanation for fever (such as middle ear infection or a urinary tract infection) should be present. Because this is a population-based study, participating children all have their own treating physicians. To limit the number of people that are involved in gathering data on pneumonias, parents were asked to complete a questionnaire whenever their child has a fever and airway symptoms. If a physician is contacted, parents ask him or her to fill in a questionnaire for physicians. Every 4 months, parents will be reminded to complete the questionnaires if their child was ill.
Respiratory function
The gold standard technique, spirometry, is not feasible for this population due to the low developmental age and motor disabilities [34]. We will measure respiratory function using the interruption technique. A reversibility test will be done using Salbutamol. This is a well-studied technique that is commonly used in infants. Reliability is high and the ambulatory equipment is commercially available. [35-40] In addition, reference values are available for children. [34,41-44]
Dysphagia
In a hospital setting, aspiration can be assessed with videofluoroscopy. Since this technique is not ambulatory available, we will assess severity of dysphagia instead of aspiration. For this epidemiological study we have chosen a standardized observation method: the Dysphagia Disorder Survey (DDS) / Dysphagia Management Staging Scale (DMSS). This method has been developed especially for people with developmental disabilities [45]. We will combine this method with cervical auscultation and measurements of oxygen saturation, to increase accurateness of the observation.
Constipation
To assess constipation, we will use structured parental interviews, a two-week defecation diary and a one-week diary on food intake. This will be combined with a physical examination of the abdomen and the anal area [46]. In clinical practice, the physical examination also includes a digital rectal palpation to assess faecal impaction. However, this was considered too invasive by the ethics committee.
Nutritional state
To assess nutritional state, we will use classical anthropometry in accordance with Gerver & de Bruin [47] and single frequency Bioelectric Impedance Assessment (BIA) [48].
Gastro-oesophageal reflux
Gastro-oesophageal reflux will be assessed using the gold standard method, 24-hour pH-metry [49]. However, to make this test feasible for on-site measurements, catheter placement will not verified by X-ray, but the step-up method will be used [50,51].
Analysis and statistics
Incidence of pneumonia will be studied prospectively and the prevalence of the hypothesized risk factors will be studied cross-sectionally. The association between the hypothesized risk factors and recurrent pneumonias will be assessed using logistic regression. A Poisson regression will be used to analyse their influence on pneumonia incidence. In these analyses, only the cases and their controls will be used. The required number of controls will depend on the number of cases. P-values less than 0.05 will be considered significant.
Ethical approval
Ethical approval was obtained (P02.0188C) from the national ethics committee (The Central Committee on Research Involving Human Subjects). Care centres and specialized schools formally consented to participate. Parents or legal guardians gave informed consent, with or without consent for gastro-oesophageal reflux. Because gastro-oesophageal reflux can only be measured properly using an invasive method, parents had the opportunity to give consent with or without this measurement.
Discussion
Designing and conducting an epidemiological study in children with severe generalized cerebral palsy is associated with characteristic difficulties. Even though we have considerable experience with research through care organisations [52], the initiation of this first study in children lead to specific, not always anticipated, problems, which caused a substantial delay. In the present study several obstacles needed to be overcome, which will most likely be encountered in future studies as well. This started with the design of a realistic, ethically acceptable study, including the choice of feasible diagnostic assessment methods and was followed by the recruitment of a representative cohort. In addition, one should bear in mind that on-site measurements and therewith inclusion through care centres (specialized day-care centres and residential facilities) and specialized schools can jeopardise timely inclusion due to potential lengthy procedures.
Dealing with encountered obstacles
Designing the study was complicated by ethical issues, which were resolved by a limited concession in the study design. In standard (nested) case-control studies, hypothesized risk factors are determined at baseline. In the present study, indeed, we will determine respiratory function, constipation and dysphagia at the start of the study, as risk factors. However, gastro-oesophageal reflux and malnutrition are disorders that are likely to cause a considerable loss of quality of life, apart from their possible effects on pneumonias, and both can easily be treated. Therefore, it was considered ethically unacceptable to determine the presence of these conditions at the start of the follow-up and then postponing treatment until the study would be finished. For that reason, we decided to perform the diagnostic tests for these conditions at the end of the follow-up period. This theoretically reduces the power of the analysis, but this reduction is relative since both conditions have a chronic character. We consider this design ethically acceptable, even though we purposely will not assess gastro-oesophageal reflux and nutritional state at baseline, because we will not interfere with common medical practice. Therefore, medical diagnosing and treatment of these disorders will not be hampered.
To conduct this study, a group of children with recurrent pneumonias needed to be identified prospectively. It would make sense to do this retrospectively. However, a previously conducted pilot study indicated that medical records, even when combined with interviews of paediatricians and intellectual disability physicians, provided incomplete and therefore unreliable information on pneumonias in these children [33].
Getting informed consent of the carers of all eligible children in a geographical area within a reasonable time span was difficult. Firstly, there was no clear registration of the centres that provide care for this specific population in the Netherlands, which resulted in a search amongst a range of organisations. Secondly, centres all had their own procedure to decide on cooperation with a study, often including management, medical staff, other personnel, parent boards and ethics committees. In some centres no standard procedure existed, since they had never been asked to participate in a study before. Thirdly, the national ethics committee considered this study as a multi-centre study and required a consent-form from each centre in advance of their final approval. Although this procedure works well in studies with 2 or 3 participating hospitals, for the present study it meant that 56 centres needed to decide on participation in advance. The resulting delay was a new and unsatisfying experience for the national ethics committee as well. Fourthly, privacy regulations lead to great dependence on willingness and organizational skills of the participating centres. The selection of eligible children had to be done by care centre personnel, and information brochures were sent while researchers were blinded for names and addresses. Despite these encountered difficulties, we have approached a representative sample of children with severe generalized cerebral palsy.
All diagnostic measurements should be ambulatory available and require no active cooperation. Therefore, not all diagnostic methods in this study are "gold-standard" methods. To date, only few diagnostic tests are available, validated for this specific population. Some diagnostic tests used in the present study are applied for the first time in this population, resulting in valuable feasibility data for future validation studies. Since ethical regulations also required methods to be non-invasive when possible, assessment of constipation need to be done without the rectal digital examination, which will therefore provide less information in comparison to the normal diagnostic procedure.
To ensure that people of different nationalities participate in a prevalence study, information needs to be provided in several languages. However, our experience is that there is no need for translated written information brochures and questionnaires. A spoken introduction on compact disc can provide an introduction and interested parents will ask a family member for translation of the brochure and questionnaires.
Finally, the inclusion period was stopped before target sample size was reached, due to delay because of practical reasons discussed above. By the end of our inclusion period, almost a quarter of the children with severe generalized cerebral palsy in the Netherlands had been approached and nearly 10% of the Dutch population of these children participates. Even with less power than desired, this study will be able to put a subject on the map that got little attention up to now.
Representativeness
To stay close to clinical practice, we used inclusion criteria based on disabilities rather than on aetiology, resulting in a heterogeneous group of children. Obviously, this might also cause more heterogeneity of the results.
The participating children are slightly younger of age than the eligible children that did not participate. However, we do not regard an age difference of less then 2 years with a standard deviation of over 4 years, as a clinical relevant discrepancy. Height and weight differences can be explained by age, since BMI is not different between both groups. A relevant discrepancy does seem to be present between the groups with regard to the reported severity of dysphagia, the frequency of lower respiratory tract infections and the presence of gastro-oesophageal reflux. We assume that the parents of the children with more severe health problems were more likely to recognize the health issues of their child in the information brochure and therefore decided to participate more often. Since swallowing strongly depends on motor skills, it seems likely that participants have poorer motor skills in general then the non-participants. Another part of the discrepancy might be explained by the selection of non-eligible children by staff of the centres. On first visit, we had to exclude four children whose motor or intellectual skills were of a higher level than those defined by our inclusion criteria. This might also have been the case in the group that did not consent to participate. Because of the slight discrepancies in characteristics, the final results, especially prevalence rates, have to be interpreted with caution. Despite the discrepancies, our sample is as representative as can be expected in population-based research.
Implication for future studies
Preventive medicine needs to play a major role in the healthcare for children with severe neurological impairment. Consequently, intervention studies are needed in which effects can be measured in a valid and reproducible way, and reference values need to be established. As in any discipline, intervention studies should be based on epidemiological data. To avoid complex epidemiological studies, a health register seems to be a requisite. In such a registry, data on health status, diagnostic assessments and applied medical treatments of children with severe neurological impairment should be recorded. This would also enable specialists to combine knowledge and to monitor trends.
For every study question, one should contemplate on the choice between diagnostic assessments in hospital or on-site. When a representative cohort of children with severe generalized cerebral palsy is required, one should perform a community-based study to keep the burden low and therewith the consent rate as high as possible, but one can expect to encounter the discussed obstacles. The main disadvantage of a hospital-based study is that a selective population will be recruited, even when performed through an outpatient clinic. Furthermore, one should consider that feasibility of diagnostic assessments might be better on-site, due to the fact that the setting is familiar to the child. On the other hand, in hospital-based studies, logistics are less complicated and hospital assessments, such as X-rays, are easily applied.
In conclusion, this study will fill in some of the lacunas in the knowledge of the health status of these children such as prevalence numbers of several health conditions, associations with recurrent pneumonias. It will also provide new information on the diagnostic tools available for these children, and provide experience in performing scientific studies in this specific field.
List of abbreviations
BIA Bioelectric Impedance Assessment
BMI Body Mass Index
IQ Intelligence Quotient
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RV contributed to the design of the study, coordinated the inclusion and data acquisition, acquired and analysed data and wrote the article. EACC contributed to the acquisition of data and has been involved in revising the article critically for important intellectual content. CP was responsible for the conduct of the study and helped to draft the manuscript. HGMA has contributed to the design and had been involved in revising the article critically for important intellectual content. AV, RB, JB, MAB, PJFMM contributed to the conception of the study and design and have been involved in revising the article critically for important intellectual content. DT participated in the design and conception of the study and helped to draft the manuscript. HME was the initiator of the study, participated in its design and helped to draft the manuscript. All authors read and approved the final version of the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We kindly acknowledge and thank Prof. Dr. J.M. Bogaard, Drs. J.E.A.R. De Schryver, Dr. R.H.J. Houwen and Dr. CK van der Ent for their contribution in the development of this research project, research nurse Annelies A. Bos for her assistance with the inclusion and data acquisition, all medical students that have graduated on this project for their contribution to data acquisition, all participating children and their carers and finally all participating children's care centres and special schools for their hospitality and cooperation.
==== Refs
Katz RT Life expectancy for children with cerebral palsy and mental retardation: implications for life care planning NeuroRehabilitation 2003 18 261 270 14530591
Blair E Watson L Badawi N Stanley FJ Life expectancy among people with cerebral palsy in Western Australia Dev Med Child Neurol 2001 43 508 515 11508916
Reddihough DS Baikie G Walstab JE Cerebral palsy in Victoria, Australia: mortality and causes of death J Paediatr Child Health 2001 37 183 186 11328476 10.1046/j.1440-1754.2001.00644.x
Shavelle RM Straus DJ Day SM Comparison of survival in cerebral palsy between countries Dev Med Child Neurol 2001 43 574 11508926
Strauss D Cable W Shavelle R Causes of excess mortality in cerebral palsy Dev Med Child Neurol 1999 41 580 585 10503915 10.1017/S001216229900122X
Hutton JL Colver AF Mackie PC Effect of severity of disability on survival in north east England cerebral palsy cohort Arch Dis Child 2000 83 468 474 11087278 10.1136/adc.83.6.468
Singer RB Strauss D Shavelle R Comparative mortality in cerebral palsy patients in California, 1980-1996 J Insur Med 1998 30 240 246 10537930
Plioplys AV Kasnicka I Lewis S Moller D Survival rates among children with severe neurologic disabilities South Med J 1998 91 161 172 9496870
Strauss DJ Shavelle RM Anderson TW Life expectancy of children with cerebral palsy Pediatr Neurol 1998 18 143 149 9535300 10.1016/S0887-8994(97)00172-0
Hollins S Attard MT von Fraunhofer N McGuigan S Sedgwick P Mortality in people with learning disability: risks, causes, and death certification findings in London Dev Med Child Neurol 1998 40 50 56 9459217
Strauss D Shavelle R Life expectancy of adults with cerebral palsy Dev Med Child Neurol 1998 40 369 375 9652777
Williams K Alberman E Survival in cerebral palsy: the role of severity and diagnostic labels Dev Med Child Neurol 1998 40 376 379 9652778
Fischer-Brandies H Avalle C Limbrock GJ Therapy of orofacial dysfunctions in cerebral palsy according to Castillo-Morales: first results of a new treatment concept Eur J Orthod 1987 9 139 143 3472892
Seddon PC Khan Y Respiratory problems in children with neurological impairment Arch Dis Child 2003 88 75 78 12495971 10.1136/adc.88.1.75
Couriel JM Respiratory complications of neurological disease in children Current Medical Literature: Respiratory Medicine 1997 10 70 75
Morton RE Wheatley R Minford J Respiratory tract infections due to direct and reflux aspiration in children with severe neurodisability Dev Med Child Neurol 1999 41 329 334 10378759 10.1017/S0012162299000729
Liptak GS O'Donnell M Conaway M Chumlea WC Wolrey G Henderson RC Fung E Stallings VA Samson-Fang L Calvert R Rosenbaum P Stevenson RD Health status of children with moderate to severe cerebral palsy Dev Med Child Neurol 2001 43 364 370 11409824 10.1017/S001216220100069X
Mahon M Kibirige MS Patterns of admissions for children with special needs to the paediatric assessment unit Arch Dis Child 2004 89 165 169 14736636 10.1136/adc.2002.019158
Saito N Ebara S Ohotsuka K Kumeta H Takaoka K Natural history of scoliosis in spastic cerebral palsy Lancet 1998 351 1687 1692 9734885 10.1016/S0140-6736(98)01302-6
Sullivan PB Lambert B Rose M Ford-Adams M Johnson A Griffiths P Prevalence and severity of feeding and nutritional problems in children with neurological impairment: Oxford Feeding Study Dev Med Child Neurol 2000 42 674 680 11085295 10.1017/S0012162200001249
Couriel JM Bisset R Miller R Thomas A Clarke M Assessment of feeding problems in neurodevelopmental handicap: a team approach Arch Dis Child 1993 69 609 613 7504911
Fung CW Khong PL To R Goh W Wong V Video-fluoroscopic study of swallowing in children with neurodevelopmental disorders Pediatr Int 2004 46 26 30 15043660 10.1111/j.1442-200X.2004.t01-1-.x
Berquist WE Rachelefsky GS Kadden M Siegel SC Katz RM Fonkalsrud EW Ament ME Gastroesophageal reflux-associated recurrent pneumonia and chronic asthma in children Pediatrics 1981 68 29 35 7243506
Gangil A Patwari AK Bajaj P Kashyap R Anand VK Gastroesophageal reflux disease in children with cerebral palsy Indian Pediatr 2001 38 766 770 11463964
Booth IW Silent gastro-oesophageal reflux: how much do we miss? Arch Dis Child 1992 67 1325 1327 1471881
Gisel EG Patrick J Identification of children with cerebral palsy unable to maintain a normal nutritional state Lancet 1988 1 283 286 2893092 10.1016/S0140-6736(88)90360-1
Loughlin GM Respiratory consequences of dysfunctional swallowing and aspiration Dysphagia 1989 3 126 130 2700953
Toder DS Respiratory problems in the adolescent with developmental delay Adolesc Med 2000 11 617 631 11060558
Martin TR The relationship between malnutrition and lung infections Clin Chest Med 1987 8 359 372 3117482
Palisano RJ Hanna SE Rosenbaum PL Russell DJ Walter SD Wood EP Raina PS Galuppi BE Validation of a model of gross motor function for children with cerebral palsy Phys Ther 2000 80 974 985 11002433
IGZ Severe multiple handicaps, then what? A study on quality of heath care for people with multiple complex handicaps. [Ernstig Meervoudig gehandicapt en dán? Een onderzoek naar de kwaliteit van zorg voor mensen met meervoudige complexe handicaps.] Dutch Health Care Inspectorate in cooperation with the Dutch Ministry of Health, Welfare and Sport [Inspectie voor de Gezondheidszorg ism ministerie van VWS] 2000 Den Haag, The Netherlands,
Bax MC Terminology and Classification of Cerebral Palsy Dev Med Child Neurol 1964 11 295 297 14155190
Huisman S Een pilotonderzoek naar luchtweginfecties bij mensen met mcg problematiek TVAZ, tijdschrift van de vereniging van artsen in de zorg voor mensen met een verstandelijke handicap 2001 nr. 4 9 11
Merkus PJ Mijnsbergen JY Hop WC de Jongste JC Interrupter resistance in preschool children: measurement characteristics and reference values Am J Respir Crit Care Med 2001 163 1350 1355 11371400
Arets HG Brackel HJ van der Ent CK Applicability of interrupter resistance measurements using the MicroRint in daily practice Respir Med 2003 97 366 374 12693796 10.1053/rmed.2002.1452
Beelen RM Smit HA van Strien RT Koopman LP Brussee JE Brunekreef B Gerritsen J Merkus PJ Short and long term variability of the interrupter technique under field and standardised conditions in 3-6 year old children Thorax 2003 58 761 764 12947133 10.1136/thorax.58.9.761
Bridge PD Ranganathan S McKenzie SA Measurement of airway resistance using the interrupter technique in preschool children in the ambulatory setting Eur Respir J 1999 13 792 796 10362042 10.1034/j.1399-3003.1999.13d16.x
Child F Clayton S Davies S Fryer AA Jones PW Lenney W How should airways resistance be measured in young children: mask or mouthpiece? Eur Respir J 2001 17 1244 1249 11491172 10.1183/09031936.01.00089501
Hadjikoumi I Hassan A Milner AD Effects of respiratory timing and cheek support on resistance measurements, before and after bronchodilation in asthmatic children using the interrupter technique Pediatr Pulmonol 2003 36 495 501 14618641 10.1002/ppul.10384
Phagoo SB Wilson NM Silverman M Evaluation of the interrupter technique for measuring change in airway resistance in 5-year-old asthmatic children Pediatr Pulmonol 1995 20 387 395 8649919
Beydon N Amsallem F Bellet M Boule M Chaussain M Denjean A Matran R Wuyam B Alberti C Gaultier C Pre/postbronchodilator interrupter resistance values in healthy young children Am J Respir Crit Care Med 2002 165 1388 1394 12016101 10.1164/rccm.2011082
Merkus PJ Arets HG Joosten T Siero A Brouha M Mijnsbergen JY de Jongste JC van der Ent CK Measurements of interrupter resistance: reference values for children 3-13 yrs of age Eur Respir J 2002 20 907 911 12412682 10.1183/09031936.02.01262001
Lombardi E Sly PD Concutelli G Novembre E Veneruso G Frongia G Bernardini R Vierucci A Reference values of interrupter respiratory resistance in healthy preschool white children Thorax 2001 56 691 695 11514689 10.1136/thorax.56.9.691
McKenzie SA Chan E Dundas I Bridge PD Pao CS Mylonopoulou M Healy MJ Airway resistance measured by the interrupter technique: normative data for 2-10 year olds of three ethnicities Arch Dis Child 2002 87 248 251 12193444 10.1136/adc.87.3.248
Sheppard JJ Hochman R Dysphagic disorders in a large residential setting. In: Dysphagic disorders in a large residential setting. city: Washington DC; 1988
Benninga MA Voskuijl WP Taminiau JA Childhood constipation: is there new light in the tunnel? J Pediatr Gastroenterol Nutr 2004 39 448 464 15572881
Gerver WJ de Bruin R Body composition in children based on anthropometric data. A presentation of normal values Eur J Pediatr 1996 155 870 876 8891556 10.1007/s004310050505
NIH Consensus statement. Bioelectrical impedance analysis in body composition measurement. National Institutes of Health Technology Assessment Conference Statement. December 12-14, 1994 Nutrition 1996 12 749 762 8974099 10.1016/S0899-9007(97)85179-9
A standardized protocol for the methodology of esophageal pH monitoring and interpretation of the data for the diagnosis of gastroesophageal reflux. Working Group of the European Society of Pediatric Gastroenterology and Nutrition J Pediatr Gastroenterol Nutr 1992 14 467 471 1517953
Pehl C Boccali I Hennig M Schepp W pH probe positioning for 24-hour pH-metry by manometry or pH step-up Eur J Gastroenterol Hepatol 2004 16 375 382 15028969 10.1097/00042737-200404000-00002
Klauser AG Schindlbeck NE Muller-Lissner SA Esophageal 24-h pH monitoring: is prior manometry necessary for correct positioning of the electrode? Am J Gastroenterol 1990 85 1463 1467 2239874
Evenhuis H van Splunder J Vink M Weerdenburg C van Zanten B Stilma J Obstacles in large-scale epidemiological assessment of sensory impairments in a Dutch population with intellectual disabilities J Intellect Disabil Res 2004 48 708 718 15494060 10.1111/j.1365-2788.2003.00562.x
Loughlin GM Lefton-Greif MA Dysfunctional swallowing and respiratory disease in children Adv Pediatr 1994 41 135 162 7992682
Sheikh S Allen E Shell R Hruschak J Iram D Castile R McCoy K Chronic aspiration without gastroesophageal reflux as a cause of chronic respiratory symptoms in neurologically normal infants Chest 2001 120 1190 1195 11591559 10.1378/chest.120.4.1190
Evans PM Alberman E Certified cause of death in children and young adults with cerebral palsy Arch Dis Child 1991 66 325 329 2025010
Maudsley G Hutton JL Pharoah PO Cause of death in cerebral palsy: a descriptive study Arch Dis Child 1999 81 390 394 10519709
|
16029493
|
PMC1201147
|
CC BY
|
2021-01-04 16:31:05
|
no
|
BMC Pediatr. 2005 Jul 19; 5:25
|
utf-8
|
BMC Pediatr
| 2,005 |
10.1186/1471-2431-5-25
|
oa_comm
|
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-801605352710.1186/1471-2458-5-80Research ArticleThe influence of in-pregnancy smoking cessation programmes on partner quitting and women's social support mobilization: a randomized controlled trial [ISRCTN89131885] Aveyard Paul [email protected] Terry [email protected] Olga [email protected] KK [email protected] Department of Public Health and Epidemiology University of Birmingham Birmingham B15 2TT UK2005 29 7 2005 5 80 80 3 12 2004 29 7 2005 Copyright © 2005 Aveyard et al; licensee BioMed Central Ltd.2005Aveyard 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
Smoking cessation interventions in pregnancy could influence a woman's social behaviour and her partner's smoking behaviour, but this has not been examined in any published randomized trials.
Method
918 women smoking at booking for antenatal care were enrolled in a cluster-randomized trial of three interventions: standard care, self-help manual and enhanced stage-based counselling, or self-help manual, enhanced stage-based counselling and use of an interactive computer program. The outcomes were change in social support received by women between booking for maternity care and 30 weeks gestation and 10 days postpartum and reported cessation in the woman's partner at these times.
Results
Few pregnant women's partners stopped smoking (4.1% at 30 weeks of gestation and 5.8% at 10 days postpartum) and the probability of quitting did not differ significantly by trial arm. Women's scores on the Inventory of Socially Supportive Behaviors showed a slight decline from booking to 30 weeks gestation, and a slight increase to 10 days postpartum, but these changes did not differ significantly by trial arm.
Conclusion
The stage-based interventions tested in this trial aimed partly to influence women's mobilization of support and might have influenced partners' quitting, but there was no evidence that they did so. Given that women and their partners often stopped smoking together, future interventions to prevent smoking in pregnant women could encourage both partners to quit together.
==== Body
Background
There are 44 trials in the Cochrane review of interventions for smoking cessation in pregnancy[1], which show that advice and support to stop smoking doubles the quit rate. None of these trials have assessed outcomes outside women's smoking and indices of perinatal well being of the fetus and child. The authors of the Cochrane review suggest that future studies of smoking cessation in pregnancy record the effects on family functioning, meaning a basket of social and emotional outcomes.
In a comment on the Cochrane review of smoking cessation interventions in pregnancy[1], Oliver contrasts it with another Cochrane review[2], which evaluated the outcomes of social support for disadvantaged mothers on maternal and child well being[3]. Oliver makes the point that the determinants of continuing smoking throughout pregnancy include younger age[4], lower socio-economic status[5], continued psychological stress[6], low social participation, low instrumental support, and low support from a woman's partner[7], all of which is associated with nicotine dependence[8]. Given that nicotine dependence and smoking in pregnancy is embedded within socio-economic disadvantage to such a degree, we might expect that both sets of studies within the Cochrane reviews to have a similar range of broad outcomes. However, this is only true for those studies where the intervention was social support and not for those studies examining smoking cessation advice in pregnancy. Commenting particularly on the oft-cited role of cigarettes as stress-reduction agents, Oliver states "it seems irrational to try to take away a coping mechanism and not look for any social and emotional consequences such as strained family relationships" (p275)[3]. An alternative view, not discussed by Oliver, is that interventions to assist pregnant women stop smoking might actually improve family functioning if the intervention has at least some potential to do so. People who stop smoking are on average less stressed than when they were smoking[9]. This is the report of one trial of smoking cessation advice in pregnancy and its influence on two disparate secondary outcome measures that reflect family functioning. These are partners quitting and the social support given to the pregnant woman.
We found only one non-randomized intervention study that has examined the influence of smoking cessation advice for pregnant women on any of the outcomes that might constitute family functioning. Wakefield et al trained midwives to help women stop smoking[10]. Women were shown a model of a fetus in utero and played the recording of the change in fetus' heartbeat when a woman smoked, and women were given an explanation of this. In addition, women were given additional time with the midwife to discuss smoking cessation, and given a small booklet to take home to assist smoking cessation. Women's partners in the intervention hospital were more likely to quit compared to the partners of both a group of women in the same hospital whose smoking was monitored prior to the intervention period and to women contemporaneously observed in another hospital. The OR for partners trying to quit while the woman was pregnant in the intervention group relative to the control groups was 2.94 (1.10–7.88), representing 34.0% versus 14.9%. However, the rates of successful quitting by partners were low and not significantly different at 1.8% in the intervention group and 2.1% in the control group. Nor were there any significant differences in partners' daily cigarette consumption. At six months postpartum, there were no significant differences in attempts to quit, successful quitting, or daily cigarette consumption. Presumably, if this intervention affected the partners, it could have acted either through changing the personal interaction between the pregnant woman and her partner, or directly through sharing the self-help intervention. However, given the control group was non-randomized, any differences in partners' smoking habits could be due to inherent differences in the intervention and control groups initially, and not a result of the intervention. No currently available randomized trials have examined these issues.
We have previously reported the primary outcome at the end of pregnancy and 18 months postpartum and two other secondary outcomes of the trial[11,12]. Women who were still smoking at booking for antenatal care were enrolled in a randomized controlled trial to assist smoking cessation. The trial compared a standard care intervention with two different programs based on the Transtheoretical Model (TTM). We found that there were small differences between the two TTM arms. Combining the two TTM arms in the analysis (as pre-planned), the OR (95% confidence intervals (CI)) for stopping smoking at 30 of gestation weeks were 2.09 (0.90–4.85) for 10-week sustained abstinence and 2.92 (1.42–6.03) for point prevalence abstinence relative to controls. At 10 days after delivery, the OR (95%CI) for quitting were 2.81 (1.11–7.13) and 1.85 (1.00–3.41) for 10-week and point prevalence abstinence respectively. The absolute benefit of the intervention was low because only 3% or so of women managed sustained abstinence. Nevertheless, there was some evidence that the intervention benefited women's smoking, but what effect did the intervention have on women's social functioning and their partners' smoking habits? These were pre-planned secondary outcomes. The trial was not explicitly designed to influence these outcomes, but had elements of the intervention that made it plausible that it might do so and we were responding to the calls described above to report these outcomes.
Methods
We obtained ethical approval from the relevant NHS ethical committees. The methods and main outcome of this trial have previously been reported in detail elsewhere[11,12], as have an analysis on stages of change outcomes[13], and stress in the pregnant woman[14]. Briefly, we recruited 16 of the 19 midwifery services for the West Midlands to participate in the trial. Midwives deliver antenatal care mainly in community settings, meaning general practices, rather than hospitals. About half of the available general practices were selected to participate, with only one midwife declining. Midwives were asked to attempt to recruit all women aged 16 years and over who were still smoking at booking for maternity care (about 12 weeks of gestation on average). We estimate that they recruited approximately 42% of potentially eligible smokers, described fully in the previous report[11,12]. In brief, nearly all were white, almost two thirds of women had had a baby previously, were of mean (SD) age 26.5 (5.9) years, of average net household income of £100–£200 per week, and, on average, left education aged 16 years. Women smoked on average 6 cigarettes per day at booking, but this increased to 11 cigarettes from mid-pregnancy onward[15]. The median Fagerstrom Test for Nicotine Dependence (measured at booking for maternity care) was 3, with the 10th and 90th percentiles being 0 and 6[16]. Three points and below represents low dependency, which might reflect a reduction in cigarette consumption that occurred in women at around the time of booking for maternity care[15]. Six points and above represents high dependency. Two thirds of women lived with partners that smoked.
Interventions
In this pragmatic trial[17], we examined three interventions: Arm A, Controls; Arm B, Manuals; and Arm C, Computer. Midwives in each trial arm were aware that they were one of three trial arms.
Arm A, controls
The intervention in Arm A was intended to be standard smoking cessation advice given by midwives. Midwives in Arm A received half a day's training on the research protocol only. They were asked to deliver smoking cessation advice as they would normally do. Midwives gave women the Health Education Authority (of England) leaflet Thinking about Stopping. This single 21 × 30 cm sheet was folded into a 3-page leaflet and contains one section on why women should stop smoking, and five sections on how to do so.
Arm B, manuals
Midwives in Arms B received 2 1/2 days training. Two days covered the Transtheoretical Model, and a half a day covered the research protocol, as for Arm A midwives. Following this, midwives practiced recruiting women and using the materials and then had a half-day's reflective session on their experiences and for them to recheck details of the intervention.
At booking for maternity care, midwives gave participants in trial Arm B received a set of six 15 × 21 cm 30-page stage-based self-help manuals; "Pro-change programme for a healthy pregnancy". The set consisted of one manual for each stage of change and a further one for "recycling". These manuals explained the concepts of stage of change, helped participants to stage themselves, and contained quizzes and exercises to engage the stage-appropriate processes of change. Additionally, at each of three occasions during pregnancy booking, (generally about 12 weeks of gestation but up to 20 weeks) (named T1); 23–25 weeks (T2); and 28–30 weeks (T3), and 10 days postpartum (T4), the midwife assessed a participant's stage of change. Midwives were encouraged to discuss the use of manuals for no more than 15 minutes, such as by going through one of the self-help exercise.
Arm C, computer
The midwives in this arm received the same training as midwives in Arm B. The participants also received the same stage-based self-help manual intervention as Arm B and the midwife explained how to use the stage-based manuals in the same way. Additionally, these participants used a computer program installed on a laptop computer at times T1 to T4. Women worked alone without the midwife using the computer program. This consisted of questions to stage the woman, and this was followed by on-screen and audio feedback of what stage women were in and what that meant. This format was repeated for the other concepts: decisional balance, temptation, and processes of change, with strategies to use to move stage. On second and third use, women also received feedback on progress or lack of it since the last use. It took about 20 minutes to complete, and, consequently, midwives often needed an additional visit to allow women the time to complete the computer program. Following each use of the computer, the feedback was printed out and sent to the participant within one-week of the intervention.
How could the intervention influence social support mobilization?
Neither the control intervention nor the TTM-based interventions had as a primary goal the changing of social support mobilization. Nevertheless, an intervention in smoking is an intervention on a complex biopsychosocial phenomenon, and the TTM-based intervention in particular had important elements that encouraged women to make changes in their support mobilization.
The self-help leaflet given to women in Arm A advised women to get support, but gave no advice on how this was achieved. Given that midwives offer little detailed counselling in smoking cessation to pregnant women[18], women in Arm A received very little if any intervention that could have influenced social support mobilization except that which midwives would give to all women regardless of smoking status.
Each of the TTM-based manuals gave women advice on mobilizing social support. In Precontemplation, women were advised to recognize pressure to quit from others ('nagging'), and make a plan to address this more constructively. Women were offered three pieces of advice-to acknowledge the person's concern reflected in the nagging, to tell the person that stopping smoking is a personal decision, and to remind the person that stopping smoking takes a lot of energy and stopping the nagging would allow them to consider whether stopping smoking was the right thing to do or not. In Contemplation, women were asked to imagine themselves as a non-smoker. In this guided imagining, women were told to think of the praise that family and friends might give if they stopped smoking. In Preparation, women were advised to set up a support team to help them quit (and write this down). They were asked to get the support team to congratulate them for every day without cigarettes. The Action manual emphasized the support team in the same way. In Maintenance, women were advised to anticipate the stressors they might face and make a plan of how to deal with those. They were offered a menu of items as a prompt, including 'Don't be afraid to ask for help.' The value of thinking through new ways of doing familiar tasks was also emphasized. Most of this advice described above was followed by blank areas of the book where women could write down their personal plans. Thus women in Arm B received much more advice and support to increase social support and create more positive environment than did women in Arm A.
Women in Arm C would have therefore had at least the same content as women in Arm B on social support. Plus, they would have assessed themselves on the processes of change[19]. Several processes involve social interaction. These are social liberation, which includes the creation of new social opportunities, stimulus control, which includes the control over social stimuli to smoke, and helping relationships, which means creating therapeutic relationships and enhancing the rapport in existing relationships. Women were assessed on their use of these social processes and received on-screen and subsequent written feedback on these processes to increase their use if appropriate. Additionally, women would have been assessed on the Temptation (to smoke) Scale (the complement of self-efficacy)[20], and given advice on handling social temptations, one of the dimensions of this scale, receiving on-screen and written feedback. Thus women in Arm C would have received the most intense advice addressing social support mobilization of the three groups of women.
How could the intervention influence partner quitting?
Neither the leaflet given to the control group nor the stage-based manuals directly addressed partner quitting. The only way that these self-help interventions could therefore influence partners' smoking would be if the partners' motivation to stop was bolstered by the pregnant woman attempting to quit, or the woman shared the manual with her partner. We have no direct data on the former mechanism, which must be inferred from the quit rate data for partners presented below. However, given that the TTM-based arms increased the quit rate relative to the control group[11], this means of influencing partners is possible. Although a minority view, some men report being prepared to quit smoking to support their pregnant partners doing so[21].
The second possible means of influencing partners' quitting, lending their self-help materials, is supported by data from the follow up of these women 18 months after they had given birth. Women in Arms B and C valued their self-help materials more than did women in Arm A. For example, 10% of women in Arm A, 14% of women in B, and 22% of women in C found the self-help materials they were given either very helpful or extremely helpful. Similarly, women in the TTM arms were slightly more likely to lend their self-help materials to someone (unspecified), with 9% in A, 7% in B, and 17% in C doing so. Self-help materials are known to improve slightly the rate of quitting[22] so this may directly influence partners' quitting.
Allocation
This was a cluster-randomized trial. The midwifery teams in each family practice were allocated by computerized minimization algorithm designed to balance the family practices across arms of the trial. The characteristics balanced by minimization were a measure of socio-economic status of the population served by the family practice (4 groups), urban/rural location (2 groups), and birth rate (3 groups).
Outcome measures
Two outcomes were used. Women reported their partners' smoking status at booking for maternity care (approx 12 weeks of gestation) and whether their partner quit or not by 30 weeks gestation and 10 days postpartum. A previous study has shown that pregnant women's reporting of their partners' smoking habits were nearly in complete agreement with the partners' self-reports[23]. Adults in socially neutral settings, such as in response to surveys, report their own smoking accurately when checked against biochemical measures[24]. Given that women's reports of their partners' smoking agree with their partners' reports, and that partners' smoking habits are accurately reported, this implies that pregnant women report their partners' smoking habits accurately. We could not verify quitting in partners by biological measurement, but given these arguments and data, there is no reason to assume that partners' reports would be wrong, or, in particular, be more likely to be wrong in one arm rather than another. This was the first outcome.
The second outcome was the Inventory of Socially Supportive Behaviors at 30 weeks gestation and 10 days postpartum. The Inventory of Social Supportive Behaviors (ISSB) is a 40-item self-report measure that was designed to assess how often individuals received various forms of assistance during the preceding month[25]. The ISSB is an appropriate measure of support mobilization or aid provision. It measures a concept that differs from qualitative measures of support such as support satisfaction or perceived availability of social support. It asks respondents to rate how frequently certain events have happened to them in the past month. Caldwell and Reinhardt's factor structure is the most parsimonious for this scale; with clusters labelled Guidance, Emotional Support, and Tangible Support[26]. These three subscales were used as outcomes along with the global score. Guidance covered items such as 'suggested some action you should take, taught you how to do something, gave you feedback on how you were doing, gave you some information to help you understand a situation.' Emotional support covered items such as 'expressed interest and concern in you situation, was right there with you in a stressful situation, comforted you by showing you some physical affection, told you that he or she feels very close to you'. Tangible support covered items such as 'gave you under £20, provided you with a place to stay, gave you transportation, loaned you under £20.' For the component and global change in ISSB outcomes, the scores at baseline were taken from the scores at 30 weeks gestation or 10 days postpartum and this change score constituted the outcome.
In total, 918 women entered the study, of which 791 (86.1%) had a partner at booking for maternity care. Of these partners, 571 (72.2%) smoked when these women booked for maternity care. Of these 571 women, 106 (18.6%) were not followed up. The most common reasons for this were an early end to pregnancy, losing contact with the midwife, or moving house. Importantly, drop out did not differ according to arm. There was one statistically significant though fairly small difference between the women that dropped out and those that did not. Women who dropped out were less well educated, with 37% compared to 22% having no educational qualifications. However, there was no difference in drop out by baseline stage of change, cigarettes per day, Fagerstrom Test for Nicotine Dependence[16], household income, age, parity, gestation at booking, and ethnic group. Thus, drop out appeared random with respect to most characteristics.
Of the 918 women, 595 (64.8%) women had data on the change in ISSB between booking and 30 weeks gestation and 615 (70.0%) women had such data at 10 days postpartum. There was one small statistically significant difference between women with data and women in whom it was absent. Twenty percent of women with ISSB change data had no educational qualifications compared to 33% of women without ISSB change data. There were no differences in the other baseline characteristics; stage of change, cigarettes per day, Fagerstrom Test for Nicotine Dependence, household income, age, parity, gestation at booking, the proportion of women with a partner, the proportion of partners that smoked, and ethnic group. Importantly, the proportion of women with missing data was the same in each arm. Thus, drop out appeared random with respect to most characteristics.
Analysis was conducted using Multilevel Modelling for Windows (MLwiN) using random effects regression models. This accounted for the cluster randomization design. Logistic models were used for partner quitting, a binary outcome, and linear models for the change in ISSB. For both outcomes, we created both unadjusted models and adjusted models. The latter models adjusted for baseline cigarette consumption, Fagerstrom Test for Nicotine Dependence, weeks of gestation at booking, ethnic group, parity, education, income, and baseline stage of change. However, in no case did the adjusted models produce different results to the unadjusted ones and these results have been omitted.
Results
Few pregnant women stopped smoking and also few of their partners. Of the 465 women who had smoking partners at baseline, 30 (6.5%) women had stopped smoking at 30 weeks of gestation. At the same time point, 19 (4.1%) partners had stopped at 30 weeks gestation, of which 10 (52.6%) lived with women that had quit. There was no evidence that the probability of quitting by partners differed significantly by trial arm (Table 1).
Table 1 The probability of partners quitting smoking by trial arm
Arm A Arm B Arm C Difference between arms
% % OR (95%CI) % OR (95%CI) χ2, p*
Partner quitting at 30 weeks gestation 3.3% 4.1% 1.24 (0.35–4.41) 5.2% 1.59 (0.45–5.60) 0.52, 0.77
Partner quitting at 10 days postpartum 4.8% 4.7% 0.99 (0.35–2.79) 7.9% 1.71 (0.66–4.48) 1.86, 0.40
* 2 degrees of freedom
At 10 days postpartum, 46 (9.9%) women had stopped smoking. Twenty-seven (5.8%) partners had stopped, of whom 12 (44.4%) lived with women that had stopped smoking. At 10 days postpartum, there was no evidence that the probability of quitting by partners differed by trial arm (Table 1).
The mean (SD) ISSB score at T1 was 2.0 (0.6). The means were 2.0 in all three arms, with SDs of 0.6, 0.7, and 0.6 in Arms A to C respectively. At T3, it was lower at 1.9 (0.6), but had risen after delivery to 2.1 (0.6). Table 2 shows that this pattern of small decrease to 30 weeks gestation and small increase to 10 days postpartum was the same in all three arms of the trial. There were no significant differences in change in ISSB score by trial arm. The pattern of change scores was similar for all three subscales of the ISSB.
Table 2 The effects of trial arm on the difference in ISSB from baseline to outcome
Arm A Arm B Arm C Difference between arms
Mean Mean Difference B-A (95%CI) Mean Difference C-A (95%CI) χ2, p*
Outcome at 30 weeks gestation
ISSB combined score -0.13 -0.09 0.04 (-0.08–0.16) -0.08 0.05 (-0.07–0.17) 0.77, 0.68
Guidance subscale -0.12 -0.05 0.07 (-0.06–0.20) -0.05 0.07 (-0.06–0.20) 1.42, 0.49
Emotional support subscale -0.15 -0.15 0.00 (-0.16–0.15) -0.18 -0.03 (-0.18–0.12) 0.16, 0.92
Tangible support subscale -0.01 0.02 0.03 (-0.10–0.17) -0.06 -0.05 (-0.18–0.08) 1.71, 0.43
Outcome at 10 days postpartum
ISSB combined score 0.09 0.10 0.01 (-0.12–0.14) 0.16 0.07 (-0.06–0.19) 1.27, 0.53
Guidance subscale 0.04 0.10 0.06 (-0.08–0.20) 0.12 0.08 (-0.06–0.22) 1.44, 0.49
Emotional support subscale 0.12 0.10 -0.02 (-0.19–0.14) 0.12 0.00 (-0.17–0.17) 0.10, 0.95
Tangible support subscale 0.10 0.08 -0.01 (-0.16–0.14) 0.15 0.05 (-0.10–0.21) 0.85, 0.65
* 2 degrees of freedom
Discussion
Smoking is a complex bio-psychosocial phenomenon; so smoking cessation interventions usually involve quitters making changes to their social world. Hence all smoking cessation interventions have the potential to have effects on individuals other than the quitter. Smoking cessation interventions in pregnancy might be particularly able to do so because of the shared social change implied by the pregnancy for the partners involved[21]. The smoking cessation intervention in this trial had beneficial effects on women's smoking. Although the intervention did not primarily aim to influence the social world of the woman concerned and were not the major component of the intervention, the more intensive stage-based advice arms did contain considerably more advice on how to make changes than did the standard care arm. Additionally, women in the TTM-based intervention arms took home an attractive set of self-help books that they could have shared with their partner, although they were not given specific advice to do so, more than twice as many women did this in the most intensive advice arm. There was no evidence from this trial that such computerized advice and self-help literature resulted in more of the women's partners quitting smoking or women receiving more social aid provision from those around them in the intensive advice arms of the trial, however. Nevertheless, the data do emphasise that many of those women making sustained changes to their smoking behaviour were accompanied by their partners doing likewise.
There seems little scope for bias to explain these results. Approximately 42% of all potential smokers were recruited into the study. To explain these negative results by this potential bias, we would have to postulate that among the majority of non-recruits, the TTM-based intervention would have influenced social support mobilization and partner quitting favourably relative to the control intervention, but did not do so among those women recruited. Given the main reason that women were not recruited was due to midwives inactivity within the trial[11], rather than a characteristic of the women themselves, there seems to be no reason to suspect bias from this source, though it clearly cannot be excluded. Once women were recruited, the cluster randomization resulted in good balance of the characteristics of women between the arms[11], so bias from this source is unlikely also. Around 80% of women had data on change in partners' smoking habits and around 70% on change in ISSB. Clearly, if women with absent data in Arm A were different to women with absent data in Arm B or C, then bias could result. However, reassuringly, while women with fewer educational qualifications were more liable to have missing data, this effect was similar in all three arms, and bias from this source therefore also seems unlikely. It must be acknowledged that the trial had little power to exclude a worthwhile effect on partners' quitting, though it had ample power to detect small differences in social support provision.
The quit rate among pregnant women's partners was low, at 4–6%. The quit rate among partners of pregnant women was low at 4% in a nationally representative English sample[27]. Similarly, it was around 2–4% in the intervention study by Wakefield and colleagues discussed in the Introduction[10]. These data suggest that the partners of pregnant women are unlikely to stop, although this quit rate is slightly higher than the annual quit rate among all smokers (around 3%[28]). Observational data suggest that living with a partner that smokes is a major risk factor for pregnant women continuing to smoke through pregnancy[29]. Our trial data suggest that standard interventions, even those with potential to influence the partner, such as by self-help manuals used in our study, do not currently influence partners' smoking. Nevertheless, it is striking that one quarter to one third of pregnant women that stopped smoking lived with partners that also stopped. Perhaps future interventions need to test interventions that intervene on both partners in pregnancy, and not the pregnant woman alone. A qualitative study found evidence that the issue of men's smoking was not usually addressed in antenatal clinics even when the man and pregnant woman attended together. Furthermore, men reported that they would have generally welcomed support to stop smoking in that context[21].
The issue of the effect of in-pregnancy smoking cessation advice on the social functioning of women has not, to our knowledge, been addressed in any previous study. Wakschlag describe the past and current life experiences of women who continued to smoke through pregnancy[30]. Compared to women who either had never smoked or quit smoking in pregnancy, continuing smokers were more likely to exhibit problem behaviour by creating interpersonal difficulties, display problems in adaptive functioning, and engage in problematic health behaviours. Given that nicotine addiction and smoking throughout pregnancy is embedded in this constellation of problem behaviours and social disadvantage, it is unsurprising that even well-placed advice to change one's social world had negligible effect on the smokers enrolled in this trial.
Conclusion
This self-help and midwife intervention had a small beneficial effect on women's smoking, but no benefit on either partners' smoking or on women's social functioning. More comprehensive interventions to address factors outside of the pregnant woman's smoking may have effects on family functioning and might therefore influence women's smoking more effectively also. In particular, given that a large minority of women who stopped successfully also lived with a partner who stopped, interventions that target both partners in pregnancy might be the most effective means of protecting the fetus and the child into the future.
Abbreviations
TTM Transtheoretical Model
OR Odds ratio
CI Confidence interval
ISSB Inventory of Socially Supportive Behaviors
NHS National Health Service (of the UK)
SD Standard deviation
MLwiN Multilevel Modelling for Windows
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
The trial was designed by Terry Lawrence, KK Cheng, Olga Evans, and Paul Aveyard. The study was managed by Olga Evans and Terry Lawrence. This particular data analysis was planned by Olga Evans, Paul Aveyard, and Terry Lawrence. Paul Aveyard did the data analysis and produced the initial draft of the manuscript and all other authors contributed to the revision of the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This trial was funded by the health authorities of the West Midlands. The interventions described are copyright of Pro-Change . Helen Evans supported this work admirably.
==== Refs
Lumley J Oliver SS Chamberlain C Oakley L Interventions for promoting smoking cessation during pregnancy The Cochrane Database of Systematic Reviews 2004 CD001055 15495004 10.1002/14651858.CD001055.pub2
Hodnett ED Fredericks S Support during pregnancy for women at increased risk of low birthweight babies The Cochrane Database of Systematic Reviews 2003 CD000198 12917888 10.1002/14651858.CD000198
Oliver S Maynard A and Chalmers I Exploring lay perceptions on questions of effectiveness Non-random reflections on health services research On the 25th anniversary of Archie Cochrane's Effectiveness and Efficiency 1997 London, BMJ Publishing Group 272 291
Lu Y Shilu T Oldenburg B Determinants of smoking and cessation during and after pregnancy Health Promotion International 2001 16 355 365 11733454 10.1093/heapro/16.4.355
Olsen J Predictors of smoking cessation in pregnancy Scand J Soc Med 1993 21 197 202 8235506
Wergeland E Strand K Bjerkedal T Smoking in pregnancy: a way to cope with excessive workload? Scand J Prim Health Care 1996 14 21 28 8725090
Dejin-Karlsson E Hanson BS Ostergren PO Ranstam J Isacsson SO Sjoberg NO Psychological resources and persistent smoking in early pregnancy - a population study of women in their first pregnancy in Sweden J Epidemiol Community Health 1996 50 33 39 8762351
Ludman EJ McBride CM Nelson JC Curry SJ Grothaus LC Lando HA Pirie PL Stress, depressive symptoms, and smoking cessation among pregnant women Health Psychol 2000 19 21 27 10711584 10.1037//0278-6133.19.1.21
Chassin L Presson CC Sherman SJ Kim K Long-term psychological sequelae of smoking cessation and relapse Health Psychol 2002 21 438 443 12211510 10.1037//0278-6133.21.5.438
Wakefield M Jones W Effects of a smoking cessation program for pregnant women and their partners attending a public hospital antenatal clinic Aust N Z J Public Health 1998 22 313 320 9629815
Lawrence PT Aveyard P Evans O Cheng KK A cluster randomised controlled trial of smoking cessation in pregnant women comparing interventions based on the transtheoretical (stages of change) model to standard care Tobacco Control 2003 12 168 177 12773727 10.1136/tc.12.2.168
Lawrence T Aveyard P Cheng KK Griffin C Johnson C Croghan E Does stage-based smoking cessation advice in pregnancy result in long-term quitters? 18-month postpartum follow-up of a randomized controlled trial Addiction 2005 100 107 116 15598198 10.1111/j.1360-0443.2005.00936.x
Aveyard P Lawrence T Cheng KK Griffin C Croghan E Johnson C A randomized controlled trial of smoking cessation for pregnant women to test the effect of a Transtheoretical Model based intervention on change in stage and interaction with baseline stage. British Journal of Health Psychology 2005 in press
Aveyard P Lawrence T Croghan E Evans O Cheng KK Is advice to stop smoking from a midwife stressful for pregnant women who smoke? Data from a randomized controlled trial Prev Med 2005 40 575 582 15749141 10.1016/j.ypmed.2004.07.012
Lawrence T Aveyard P Croghan E What happens to women's self-reported cigarette consumption and urinary cotinine levels in pregnancy? Addiction 2003 98 1315 1320 12930219 10.1046/j.1360-0443.2003.00485.x
Heatherton TF Kozlowski LT Frecker RC Fagerstrom KO The Fagerstrom Test for Nicotine Dependence: a revision of the Fagerstrom Tolerance Questionnaire British Journal of Addiction 1991 86 1119 1127 1932883
Roland M DJ T Understanding controlled trials. What are pragmatic trials? BMJ 1998 316 285 285 9472515
Cooke M Mattick RP Barclay L Predictors of brief smoking intervention in a midwifery setting Addiction 1996 91 1715 1725 8972929 10.1046/j.1360-0443.1996.9111171512.x
Prochaska JO Velicer WF The transtheoretical model of health behavior change American Journal of Health Promotion 1997 12 38 48 10170434
Velicer WF Diclemente CC Rossi JS Prochaska JO Relapse situations and self-efficacy: an integrative model Addict Behav 1990 15 271 283 2378287 10.1016/0306-4603(90)90070-E
Wakefield M Reid Y Roberts L Mullins R Gillies P Smoking and smoking cessation among men whose partners are pregnant: a qualitative study Soc Sci Med 1998 47 657 664 9690848 10.1016/S0277-9536(98)00142-7
Lancaster T Stead LF Self-help interventions for smoking cessation The Cochrane Database of Systematic Reviews 2002 CD001118 12137618 10.1002/14651858.CD001118
Passaro KT Noss J Savitz DA Little RE Team TALSPACS Agreement between self and partner reports of paternal drinking and smoking Int J Epidemiol 1997 26 315 320 9169166 10.1093/ije/26.2.315
Bajekal M Boreham R Erens B Falaschetti E Hirani V Primatesta P Prior G Tait C Health Survey for England:cardiovascular disease 1998 London, The Stationery Office
Barrera M Sandler IM Ramsay TB Preliminary development of a scale of social support: Studies on college students. Am J Community Psychol 1981 9 435 447 10.1007/BF00918174
Caldwell RA Reinhart M An empirical investigation of the structure of social support: An effort at concept clarification. 2003 MI 48824 Michigan, University of Michigan
Bolling K Owen L Smoking and Pregancy: a survey of knowledge, attitudes and behaviour 1997 London, Health Education Authority 1 13
Parrott S Godfrey C Raw M West R McNeill A Guidance for Commissioners on the Cost Effectiveness of Smoking Cessation Interventions Thorax 1998 53 (Suppl 5) S1 S37 10226676
Ockene JK Ma Y Zapka JG Pbert LA Goins KV Stoddard AM Spontaneous cessation of smoking and alcohol use among low-income pregnant women Am J Prev Med 2002 23 150 159 12350446 10.1016/S0749-3797(02)00492-0
Wakschlag LS Pickett KE Middlecamp MK Walton LL Tenzer P Leventhal BL Pregnant smokers who quit, pregnant smokers who don't: does history of problem behavior make a difference? Soc Sci Med 2003 56 2449 2460 12742608 10.1016/S0277-9536(02)00248-4
|
16053527
|
PMC1201148
|
CC BY
|
2021-01-04 16:28:54
|
no
|
BMC Public Health. 2005 Jul 29; 5:80
|
utf-8
|
BMC Public Health
| 2,005 |
10.1186/1471-2458-5-80
|
oa_comm
|
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-831609297310.1186/1471-2458-5-83Research ArticleSuicidal tendencies and attitude towards freedom to choose suicide among Lithuanian schoolchildren: results from three cross-sectional studies in 1994, 1998, and 2002 Zemaitiene Nida [email protected] Apolinaras [email protected] Institute for Biomedical Research, Kaunas University of Medicine, 4, Eiveniu str., Kaunas, LT-50009, Lithuania2005 11 8 2005 5 83 83 25 3 2005 11 8 2005 Copyright © 2005 Zemaitiene and Zaborskis; licensee BioMed Central Ltd.2005Zemaitiene and Zaborskis; 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
Suicidal behaviour is increasingly becoming a phenomenon associated with young people and an important public health issue in Lithuania. However, there are very few studies evaluating impact of young peoples' attitudes towards suicide to their suicidal behaviour. A better understanding of the relations among the variables associated with suicidal ideation and threats in the normal population of adolescents may eventually result in a better understanding of the more serious forms of adolescent suicidal behaviour. The aim of the present study was to evaluate prevalence of suicidal tendencies among Lithuanian schoolchildren and to estimate its association with an attitude towards suicide in 1994 – 2002.
Methods
Three country representative samples of schoolchildren, aged 11, 13 and 15, were surveyed in 1994 (n = 5428), 1998 (n = 4513), and 2002 (n = 5645) anonymously in conformity with the methodology of the World Health Organization Cross – National study on Health Behaviour in School-aged Children (HBSC).
Results
About one third of respondents reported about suicidal ideation, plans or attempts to commit suicide. In the study period of eight years, the percentage of adolescents who reported sometime suicidal ideation decreased but the percentage of adolescents who declared serious suicidal behaviour remained on the same high level (8.1%, 9.8% and 8.4% correspondingly in 1994, 1998 and 2002). Moreover, the number of suicidal attempts changed from 1.0% in 1994 to 1.8% in the year 1998 and to 1,7% in the year 2002. The schoolchildren's attitude towards suicide became more agreeable: 36.6%, 41.9% and 62.5% of respondents, correspondingly in 1994, 1998 and 2002, answered that they agree with a person's freedom to make a choice between life and suicide. A multiple logistic regression analysis with low level of suicidality and high level of suicidality versus non suicidal behaviour as dependent variables for gender, age, year of the survey and attitude towards freedom to choose suicide as independent variables approved a significant association between studied covariates over the entire study period.
Conclusion
Suicidal tendencies are quite frequent among Lithuanian adolescents. An increasing number of schoolchildren are expressing an agreeable attitude towards suicide. The approving attitude towards suicide among adolescents correlates with suicidal ideation and behaviour.
==== Body
Background
Suicidal behaviour is becoming a phenomenon increasingly associated with young people. The rise in the overall suicide rates in many countries is, to a large extent, due to the increase in suicides in the younger age groups. Lithuania has been among the countries with the highest suicide rate for more than ten years. It's extremely disturbing that this problem is becoming more and more associated with the youngest inhabitants of the country. Over the recent period of ten years, the mortality due to suicide in the youngest age group from birth to 19 year old has increased more than 55.8%, therefore, suicide took the third place among external causes of death [1]. According to the statistical data of the year 2001, suicide mortality in this age group was 5.85 per 100 000 of population. Boys were committing suicides more frequently than girls (10.3 versus 1.29 per 100000 of population) [2]. In comparison to the data of other countries Lithuania is loosing a great amount of young lives. Even in the neighbouring countries the incidence rate of young people's suicides is less [3].
It is important to note, that our investigation was conducted in the period between 1990 and 1996 when suicide mortality in Lithuania rose 82.4%, with the rate peaking at more than 47 per 100,000 persons in 1996. After a slight decrease in 1997 (to 45.6) and in 1998 (to 43.8), suicide rates stabilized at a very high level (in 1998–2002 the average rate was 44.6) [4].
The existing data of the previously carried out surveys in Lithuania hardly show all the extent of the phenomenon, especially while talking about children and adolescents: a lack of data do not allow estimating a degree of spread, intensity and dynamics of suicidality among young people. Furthermore, suicide is a culturally sensitive phenomenon, that is why, in order to understand it, it is necessary to begin with attitudes. Some authors conceptualise suicide as a struggle among various conflicting attitudes towards life and death [5]. As it is stated in the literature, permissive attitudes are mediating processes regarding suicide acts, therefore this issue is important in assessing risk for suicide [6-9]. In Lithuania, though, this topic has not received much attention yet, and it is still little known about young people's attitudes towards suicide. Therefore, a better understanding of the relations among the variables associated with suicidal ideation and threats in the normal population of adolescents will eventually result in a better understanding of the more serious forms of adolescent suicidal behaviour.
In this article we aimed to assess the prevalence of suicidal ideation and behaviour among Lithuanian adolescents, and its changes in the period from 1994 till 2002. Likewise we set a task to evaluate variations in attitude towards suicide and its associations with suicidal tendencies among schoolchildren.
Methods
The study was based on the data of three surveys conducted in Lithuania in 1994, 1998 and 2002 by the methods of the WHO Cross – National study on Health Behaviour in School-aged Children (HBSC). The philosophy and methods of the HBSC project have been described in more details elsewhere [10,11].
Participants and study procedures
The samples were expected to represent the whole country from the point of view of age, sex, nationality and the place of living. The guidelines for the survey state that at least 1500 respondents in each of three age groups – 11, 13 and 15 years – should be targeted. As the survey was planned in spring months, the appropriate grade levels corresponding to the desired age ranges were 5, 7 and 9. A stratified cluster sampling design was used to draw a representative sample of schoolchildren from the whole Lithuania. There were five strata by regions of the country (including cities Vilnius, Kaunas, Klaipeda, Siauliai and Panevezys) and three strata by language (Lithuanian, Russian and Polish) used for education at school. On the first level of sampling the schools were randomly selected from each stratum. The number of the selected schools was proportional to the size of stratum. Then 5th, 7th and 9th grades were included into the sample. If two or more classes of the desired grade level occurred in the selected school only one class was randomly selected. Afterwards we surveyed all the pupils of the selected class.
The surveys were coordinated by the laboratory for Social Pediatrics of Institute for Biomedical Research, Kaunas University of Medicine. The ethical clearance for the study and the support with relevant information were obtained from the Ministry of Education and Science as well as from the regional departments of education and the administrations of the selected schools. The study was exempt from the need for Institutional Review Board approval.
The teachers were asked to administrate the questioning and to follow the agreed guidelines. The survey was conducted in the school classes with a teacher overseeing the process. The pupils responded anonymously. We ensured the self-dependent work of the pupils and confidentiality of their answers. The completed questionnaires were enclosed into envelopes and returned to the research institution for completion. Altogether 5688, 4655 and 5761 questionnaires were returned correspondingly in 1994, 1998 and 2002. Regarding the actual number of the pupils in the lists of the selected classes the response rate for all surveys was approximately 96 percent.
The national data files were prepared and exported to the HBSC international databank at the University of Bergen (Norway). The data were checked and cleaned according to strict criteria, e.g. 90 percent of the respondents should fall within one-half a year of the mean age and the remaining 10 percent no more than one-half a year beyond this point. The schoolchildren outside the targeted age ranges were removed.
The final population of the cleaned data consisted of 5428, 4513 and 5645 schoolchildren correspondingly for the surveys in 1994, 1998 and 2002. The studied population was representative to the population of school-aged children from the whole Lithuania in respect of demographic and social values (Table 1). The groups were also well balanced according to the living environment of the respondents (urban and rural areas) and the language used for education at school (Lithuanian, Russian and Polish).
Table 1 Studied population in the years 1994, 1998 and 2002, by gender and age group
Years of the survey Gender Age group Total
Boys Girls 11-year-olds 13-year-olds 15-year-olds
1994 N 2 429 2 999 1 783 1 886 1 759 5 428
% 44.8 55.2 32.9 34,7 32,4 100.0
1998 N 2 150 2 363 1 566 1 512 1 435 4 513
% 47.6 52.4 34.7 33.5 31.8 100.0
2002 N 2 887 2 758 1 867 1 873 1 905 5 645
% 51.1 48.9 33.1 33.2 33.7 100.0
Total N 7 466 8 120 5 216 5 271 5 099 15 586
% 47.9 52.1 33.5 33.8 32.7 100.0
Measures
The survey instrument was a standardized anonymous questionnaire that included structured questions followed by alternative answers. The questionnaire topics for each survey were devised through a cooperative research among the members of the HBSC research network and finally approved by the Protocols [12].
In addition to the international standard questionnaire each country was allowed to produce a set of 'national' items for their topic area to examine the research questions more thoroughly or to explore new aspects of the topic and, thus, increase the scope of the research. Due to this agreement a focus question group concerning suicides was included into the Lithuanian version of the HBSC questionnaire.
In 1994, 1998 and 2002 surveys two questions on suicide were maintained:
1. "What do you think, does a person have the right to choose: to live or to take away his own life?" Two alternative answers were given: 'yes' and 'no'.
2. "Have you ever thought about suicide?" There were five alternative answers (categories) to this question: 1 = 'have never thought about that'; 2 = 'sometimes have had such thoughts'; 3 = 'have frequently thought about suicide'; 4 = 'have thought about suicide seriously, even made plans how to carry it out'; 5 = 'tried to commit suicide'.
In order to minimize a possible impact of translation errors on the data all the versions of the questionnaires that were used in Lithuania were retranslated into English. They were approved by the Coordinating Committee of an international network of research teams.
In the analyses, the respondents were subdivided into the groups according to the chosen answer to the second question:
Group 1 – 'Non suicidal' (with no suicidal tendencies expressed). To this group were assigned those adolescents who had chosen the first variant of the possible answers 'I have never thought about suicide'.
Group 2 – 'Low level of suicidality'. The second group was formed of the respondents who had chosen the answer 'Sometimes have had thoughts about suicide'.
Group 3 – 'High level of suicidality'. The third group was made up of the respondents who had reported often thoughts, concrete plans and actions trying to commit suicide. Different studies prove that such manifestations show a high risk of suicide [13-15].
Statistical analysis
Prevalence (%) of the answers to the questions on suicide and its 95% confidence intervals (CI) were calculated, stratified by sex, age and year of the study. Chi-square (χ2) tests were performed in order to test the sex, age and study year differences in prevalence for each of the variables. Crude odds ratios and 95% intervals (OR, 95% CI) were calculated in order to analyse associations between suicidal tendencies and attitude towards suicide in the tables 2 × 2. A multivariate analysis was performed using a logistic regression model to investigate the potential importance of adolescents' attitude towards suicide and possible confounders on dependent variables low level of suicidality and high level of suicidality versus non suicidal. All possible interaction terms among adolescents' attitude based on gender, age and year of the study were included and tested. The statistical analysis was performed using the SPSS software package.
Results
Prevalence rate of suicidal ideation and behaviour
On the whole, the question "Have you ever thought about suicide?" was answered by 15 414 (98.9%) of the respondents. The distribution of answers to this question is presented in the Table 2. It seems that suicidal tendencies were quite frequent among adolescents in general. In the period from 1994 to 1998 the variation in prevalence rates of suicidal tendencies (suicidal ideation, serious plans how to commit suicide and reported attempts to commit suicide) among the Lithuanian schoolchildren was not significant. In the period of the last four years this rate slightly decreased from 40.7% (95% CI 38.2 – 42.1) in the year 1998 to 32.5% (95% CI 31.3 – 33.8) in the year 2002 (p < 0.001). The decrease was achieved only due to decrease in the prevalence rate of 'low level of suicidality'.
Table 2 Frequency of answers to the question "Have you ever thought about suicide?" in the years 1994, 1998 and 2002
Level of suicidality Category of the answer 1994 1998 2002 Total
N % N % N % N %
Non suicidal 'Have never thought about that' 3269 60.4 2660 59.3 3722 67.5 9651 62.6
Low level of suicidality 'Sometimes have had such thoughts' 1706 31.5 1387 30.9 1332 24.1 4425 28.7
High level of suicidality 'Have frequently thought about suicide' 224 4.2 203 4.5 218 4.0 645 4.2
'I have thought about suicide seriously, even made plans how to carry it out' 164 3.0 154 3.5 150 2.7 468 3.0
'Tried to commit suicide' 50 0.9 79 1.8 96 1.7 225 1.5
Total 5413 100.0 4483 100.0 5518 100.0 15414 100.0
Test for homogeneity of the prevalence of answers in 1994, 1998 and 2002: χ2 = 118.0; df = 8; p < 0.001.
During the study period of eight years, the prevalence rate of 'high level of suicidality' remained approximately on the same high level: 8.1%, 9.8% and 8.4% correspondingly in 1994, 1998 and 2002.
A comparison of indicators for suicide in the groups by age and gender was made. Suicidality, in general, was increasing with age. The girls were more likely to express the ideas of suicide than the boys. According to the data received, the problem of suicidality becomes evident already in a young age. The surveys conducted in 1994 and 1998 showed that the rate of high level of suicidality was more characteristic with younger boys. The survey of the year 2002 as compared to the data of the year 1998 demonstrated a significant decrease in the prevalence of a high level of suicidality among the 11-year-old boys while the same indicators among the 15-year-old boys remained at the same level (Figure 1).
Figure 1 The rate of high level of suicidality among boys. * – p < 0.05 in comparison with the previous year of the survey.
According to the data of the surveys of the years 1994, 1998 and 2002 among the girls there was a remarkable difference in prevalence rate of high level of suicidality among age groups. The lowest prevalence of a high level of suicidality was established among the 11-year-olds and the highest prevalence of a high level of suicidality was detected among the 15-year-old girls (Figure 2).
Figure 2 The rate of high level of suicidality among girls. * – p < 0.05 in comparison with the previous year of the survey.
Attitude towards suicide as a human right
An answer to the question "What do you think, does a person have the right to choose: to live or to take away his own life?" was considered to show an affirmative or negative attitude towards suicide. The data of the surveys demonstrated that an increasing number of the Lithuanian schoolchildren were advocating suicide as a basic human right. In 1994 and 1998, correspondingly 36.3% (95% CI 35.3 – 37.9) and 41.9% (95% CI 40.5 – 43.4) (p < 0.05) of the adolescents pointed that they agreed with the freedom to make a choice between life and suicide. In 2002, an approving attitude towards suicide was declared already by 62.5% (95% CI 61.2 – 63.8) of the respondents. Gender differences were relatively small, but boys were more prone to express acceptance of the right for such a choice (in 1994 38.6% of boys versus 34.5% of girls, p < 0.05; in 1998 – 44.1% versus 39.9, p < 0.05; in 2002 – 63.9% of boys versus 61.1% of girls, p < 0.05). The Figures 3 and 4 present the variations in prevalence rate of an affirmative attitude towards suicide among boys and girls by the age and the year of the survey.
Figure 3 The rate of approval attitude towards suicide among boys, by the age and year of the survey. * – p < 0.05 by the year of the survey.
Figure 4 The rate of approval attitude towards suicide among girls, by the age and year of the survey. * – p < 0.05 by the year of the survey.
It was observed that an approving attitude towards suicide among adolescents could be identified as a significant factor associated with suicidality. The Figures 5 and 6 show, correspondingly, the percentage of boys and girls in every level of suicidality who approved a person's freedom to choose suicide. This phenomenon is especially evident in 1994 and 1998 years of the survey, where the proportion of the adolescents expressing such an attitude among the individuals with high level of suicidality was approximately twice higher in comparison with those who did not express any suicidal ideation.
Figure 5 The percentage of boys who approved a person's freedom to choose suicide, by level of suicidality in 1994, 1998 and 2002. * – p < 0.05 by the risk for suicide.
Figure 6 The percentage of girls who approved a person's freedom to choose suicide, by level of suicidality in 1994, 1998 and 2002. * – p < 0.05 by the risk for suicide.
The suicidal risk impact of a positive attitude towards a person's freedom to choose suicide was measured by calculating the odds ratio (OR). The crude OR's of the low and high levels of suicidality in comparison to non suicidal, by gender, age and year of the survey, are given in the Table 3. Overall, a statistically significant suicidal risk impact of a positive attitude towards freedom to choose suicide was established in all the groups of the adolescents by gender, age and the study year. The observed associations were more evident among the girls.
Table 3 Crude odds ratios (OR) and 95% confidence intervals (CI) of low and high level of suicidality in relation to approving a person's freedom to choose suicide, by gender, age and year of the survey
Year of the survey Gender and level of suicidality 11-year-olds 13-year-olds 15-year-olds Total
OR (95% CI) OR (95% CI) OR (95%CI) OR (95%CI)
Boys:
1994 Non suicidal 1 1 1 1
Low 1.48 (1.03–2.12) 1.77 (1.31–2.41) 1.86 (1.36–2.53) 1.27 (1.03–1.55)
High 1.52 (0.91–2.55) 2.89 (1.70–4.89) 3.31 (1.78–6.15) 1.51 (1.09–2.08)
1998 Non suicidal 1 1 1 1
Low 1.34 (0.94–1.90) 1.75 (1.25–2.44) 2.08 (1.47–2.94) 1.45 (1.20–1.74)
High 2.34 (1.37–4.00) 3.70 (2.09–6.53) 3.52 (1.74–7.10) 2.86 (2.06–3.96)
2002 Non suicidal 1 1 1 1
Low 1.64 (1.09–2.45) 1.63 (1.13–2.36) 1.63 (1.17–2.27) 1.63 (1.36–1.96)
High 2.39 (1.02–5.57) 3.50 (1.75–7.00) 1.72 (0.95–3.12) 2.50 (1.74–3.59)
Total Non suicidal 1 1 1 1
Low 1.27 (1.03–1.55) 1.45 (1.20–1.74) 1.63 (1.36–1.96) 1.47 (1.32–1.64)
High 2.51 (1.09–2.08) 2.86 (2.06–3.96) 1.50 (1.34–3.59) 2.20 (1.82–2.67)
Girls:
1994 Non suicidal 1 1 1 1
Low 2.23 (1.62–3.07) 1.73 (1.32–2.28) 1.34 (1.02–1.77) 1.62 (1.34–1.95)
High 5.40 (3.00–9.73) 2.24 (1.43–3.51) 3.15 (2.05–4.83) 3.33 (2.37–4.68)
1998 Non suicidal 1 1 1 1
Low 1.32 (0.92–1.89) 1.33 (0.98–1.82) 1.56 (1.14–2.15) 1.40 (1.19–1.65)
High 2.79 (1.58–4.91) 3.63 (2.10–6.24) 4.54 (2.97–6.95) 2.81 (2.14–3.70)
2002 Non suicidal 1 1 1 1
Low 1.87 (1.31–2.70) 1.51 (1.12–2.04) 1.77 (1.32–2.38) 1.47 (1.24–1.73)
High 3.15 (1.59–6.22) 3.32 (1.98–5.58) 3.13 (2.03–4.82) 3.51 (2.75–4.47)
Total Non suicidal 1 1 1 1
Low 1.62 (1.34–1.95) 1.40 (1.19–1.65) 1.47 (1.24–1.73) 1.54 (1.39–1.69)
High 3.33 (2.37–4.68) 2.81 (2.14–3.70) 3.51 (2.75–4.47) 3.40 (2.90–3.97)
We performed a multiple logistic regression analysis with the low and high level of suicidality versus having non suicidal as the dependent variables for gender, age and the year of the survey, and an attitude towards freedom to choose suicide as the independent variables (Table 4). Gender (girls vs boys), age (13- and 15-year-olds vs 11-year-olds) and attitude (approved vs not approved freedom to choose suicide) were the factors that increased the level of suicidality statistically significantly. Suicidal risk of the adolescents surveyed in the year 2002 was significantly reduced in comparison with their peers surveyed previously. The interaction terms among an attitude and gender, age and the year of the study were not significant with the exception of interaction between approval attitude towards suicide and gender for a high level of suicidality. This indicates a significant modification of interaction between positive attitude towards suicide and a high level of suicidality by gender group.
Table 4 Odds ratios (OR), 95% confidence intervals (CI) and significance levels (p) for gender, age, year of the survey and attitude towards suicide of the multiple logistic regression analysis
Variables in the equation Dependent variable: low level of suicidality versus non suicidal Dependent variable: high level of suicidality versus non suicidal
OR (95% CI) p OR (95% CI) p
Girls vs boys (GIRLS) 1.71 (1.59–1.85) <0.001 1.87 (1.65–2.12) <0.001
13-year-olds vs 11-year-olds (AGE13) 1.82 (1.66–2.00) <0.001 1.63 (1.40–1.91) <0.001
15-year-olds vs 11-year-olds (AGE15) 2.50 (2.27–2.74) <0.001 2.35 2.01–2.74) <0.001
Survey in 1998 vs survey in 1994 (S1998) 1.00 (0.91–1.10) 0.98 1.16 (0.99–1.34) 0.058
Survey in 2002 vs survey in 1994 (S2002) 0.61 (0.56–0.67) <0.001 0.73 (0.63–0.86) <0.001
Approved vs not approved freedom to choose suicide (ATTITUDE) 1.63 (1.51–1.77) <0.001 2.83 (2.49–3.23) <0.001
ATTITUDE × GIRLS 1.03 (0.89–1.20) 0.69 1.51 (1.17–1.93) 0.001
ATTITUDE × AGE13 0.93 (0.77–1.12) 0.43 1.23 (0.90–1.68) 0.20
ATTITUDE × AGE15 0.98 (0.81–1.18) 0.81 1.28 (0.94–1.74) 0.12
ATTITUDE × 1998 0.90 (0.74–1.08) 0.24 1.24 (0.92–1.68) 0.15
ATTITUDE × 2002 0.95 (0.79–1.14) 0.59 1.00 (0.73–1.37) 0.98
Discussion
The surveys of the recent years show that thoughts about suicide as well as suicidal behaviour are more characteristic to the young generation of nowadays than to the one some years ago. Lithuania, unfortunately, is not an exception. In the period of the last three decades the suicide rate among 15–18-year-olds increased much more dramatically than it has among general population [16]. Accidents, suicides and homicides make more than half of reasons in the mortality of the youngest inhabitants (less than 14 year old) [17]. This has led to an increased interest in suicidal threats and behaviour among adolescents.
Findings of the anonymous questionnaires conducted in different countries suggest that approximately 20%-30% of adolescents have been having thoughts of suicide and from 2% to 12% of young people claim about attempts to commit suicide [13,18-21]. On the other hand, population – based surveys designed to estimate the prevalence of suicidal ideation and behaviour, are often difficult to compare because of a great variability in definitions of suicidality and research methods used. Continuum of suicidality includes wide spectrum of suicidal thoughts and behaviour that begins with ephemeral thoughts about suicide and, proceeds to attempted suicide without injury, to attempted suicide with serious injury, and finally, to completed suicide. The cognitive development of children and adolescents also affects their understanding of the concepts of suicidal ideation and behaviour [22]. However, the data received in our study claim that suicidal indicators of Lithuanian adolescents remain constantly at a rather high level. According to the results of the 1998 survey, almost 40% of schoolchildren at the age of 11, 13, and 15 reported about suicidal thoughts, threats or attempts to commit suicide [23]. Relatively a low rate of reported suicide attempts, in comparison to the results of similar designed studies conducted in other countries may be influenced by young age and developmental state of respondents. Numerous researchers are prone to emphasize the developmental variations in death cognition. The concept of death develops in late childhood while full awareness of the implications of death is not gained until early adolescence [24].
The data of the year 2002 indicate that the frequency of schoolchildren's suicidal tendencies has slightly decreased. It is important to note, however, that these changes occurred due to the decrease in the group of the schoolchildren having "sometimes" suicidal thoughts. Another factor which influenced this change was significant decrease in the frequency of suicidality among the youngest boys. The decrease in the prevalence of suicidal tendencies among the 11 year-old boys could be somewhat attributed to an improved situation in schools, because in the period of the recent years more attention has been paid to the 11 year-olds' adaptation in schools (as it is a transitional period: children become pupils of a secondary school). A previous research of psychosocial correlates indicated a strong association between suicidal tendencies of 11-year-old boys with school difficulties [23]. On the other hand, the number of the schoolchildren with high suicidal risk indicators (often thoughts, concrete plans or attempts to commit suicide) has remained on the same high level. Such schoolchildren make up almost ten percent.
Analysis by gender revealed some changes in the patterns of the dynamics of boys and girls' suicidal tendencies during the period of eight years. According to the data of the first and the second survey, the frequency of the high level of suicidality among the boys was changing according to the age and had a tendency to decrease. The same indicators among the girls have an opposite trend and were increasing with age. Subsequently, the prevalence of a high level of suicidality was more characteristic for younger boys and elder girls. The last investigation, however, demonstrated similar rates to the ones found in most European countries: the numbers of adolescents contemplating suicide or making suicidal attempts increase correspondingly with age, girls' suicidal manifestations are almost twice higher than boys' [19,25]. This difference was not so evident among pupils aged eleven, but with every year of adolescence this difference increases due to the fact that the indicators the girls' suicidality changed faster than those of the boys.
The explanation of these patterns could not be homologous. Some authors associate this with different manifestations of adolescents' depression, pointing out the fact that girls with a medium level of depression tend to speak about suicide more often than boys [25]. According to other authors considering suicide as one of the possible alternatives when confronted with problems is rather a normal and prevalent way of coping in adolescents [26]. Nonetheless, the vast majority of the investigators tend to point out that any suicidal display, despite age and sex, must be treated seriously as a sign of possible suicide [27,28]. Most studies in samples of normal adolescents assume that suicidal ideation, threats and behaviour could be considered as a part in an overall continuum of suicidality [19,26]. A high prevalence rate of suicidal risk indicators among schoolchildren as well as the prognosis that a stable decrease of suicides is unexpected in Lithuania in the near future because of the rate of suicides in the young generation, born after 1970, make the situation urgent and stimulate a research in this field [16,23].
Culture influences timing, development, and shape of children's concept of death in general, and suicide in particular [29]. Despite a long standing research interest in the social correlates of suicide, the associations between attitudes and suicidal behaviour have been a topic of discussions and a growing research interest in the recent years. It has been marked, that suicide as a concept is more and more becoming a life norm and a suicidogenics atmosphere has widely set in our society [30,31].
The data of our survey demonstrated that children have increasingly expressed an acceptance of the choice to commit suicide. During the period of the last eight years the number of the adolescents supposing that a person has the right to decide – to live or to commit suicide – has credibly grown from 36,3% in the year 1994 to 62,5% in the year 2002. Consequently, an increasing number of the Lithuanian schoolchildren have been advocating suicide as a basic human right.
The most important finding of the present study is that an approving attitude towards suicide among adolescents is a significant predictor of suicidal ideation. The data of the surveys conducted in 1994, 1998 and 2002 confirmed a stable association between an approval attitude towards freedom to choose suicide and suicidal potential of the schoolchildren, aged 11, 13 and 15.
The adolescents who thought that suicide was an acceptable way – out of a complicated situation made an attempt of suicide twice more frequently than those who showed a negative attitude towards this phenomenon. Though gender differences were relatively small, boys were more likely than girls to endorse the idea that one has the right to take away one's life. The similar gender differences were observed in some other studies among adolescents [32]. Our results are consistent with the findings of the study of Israeli adolescents' attitudes to suicide that have suggested the participants' attitudes towards suicide correlate with their suicidal ideation. The more approving attitudes are being associated with a greater personal preoccupation with suicide [9]. The received data indicates that a tolerant view on suicide might be considered as a suicide risk factor. A positive attitude is working as a kind of a conduct influencing the action [30]. An approval attitude towards freedom to choose suicide in a complicated life situation could serve as a positive motivational force for death. Nowadays the rising suicide rates among youth may in part be attributable to that fact, that they are more tolerant of suicide and less fearful of its consequences.
A limitation of the present study is that the decision about student's attitude towards suicide was made referring to one question. Consequently, it may be possible to get a clearer picture of interaction between attitudes and person's suicidality using special attitude research scales. More the less, regarding high suicide rates in the society, a widespread attitude towards suicide as an appropriate choice among children and adolescents could be considered as very dangerous. These findings should be considered in the prevention of suicides. Norms, values and attitudes are not static. Usually these changes have taken place slowly and rather smoothly, but in order to affect suicidal risk of young people, such attitudes must be tackled purposefully.
Conclusion
Based on the data provided by the current study it could be concluded that suicidal tendencies are quite frequent among schoolchildren; that consists a growing problem of public health in Lithuania. Moreover, the schoolchildren's attitude towards suicide became more agreeable: 36.6%, 41.9% and 62.5% of the respondents, correspondingly in 1994, 1998 and 2002, answered that they agree with a person's freedom to make a choice between life and suicide. Over the entire study period it was observed that an approving attitude towards suicide among the adolescents could be identified as a significant factor associated with suicidal ideation and suicidal behaviour. Therefore any attempt for development and implementation of a school program for prevention of suicides should consider these findings and set priorities accordingly.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
NZ outlined the questions on suicidal ideation and attitude towards suicide, substantially contributed to the conception and the design of the article and to the interpretation of data. AZ coordinated the surveys, performed the statistical analysis and helped to draft the manuscript. Both authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Zemaitiene N Jaruseviciene L Gailiene D Gailiene D Suicide in school (in Lithuanian) Ideas for suicide prevention 2001 Vilnius: Tyto Alba 138 170
Lithuanian Department of Statistics Lithuanian children 2003 2004 Vilnius
The MONEE Project CEE/CIS/Baltics A Decade of Transition 2001 Regional monitoring report No. 8. Florence, Italy: Innocenti Research Centre and UNICEF
Gailiene D Suicide in Lithuania during the years of 1990 to 2002 Arch suicide research 2004 8 389 395 16081405 10.1080/13811110490476806
Orbach I Carlson G Feshbach S Ellenberg L Attitudes towards life and death in suicidal, normal, and chronically ill children: an extended replication J Consult Clin Psychol 1984 52 1020 1027 6520272 10.1037//0022-006X.52.6.1020
Peltzer K Cherian VI Cherian L Cross-cultural attitudes towards suicide among South African secondary school pupils East Afr Med J 2000 77 165 167 12858894
Kocmur M Dernovsek MZ Attitudes towards suicide in Slovenia: a cross-sectional survey Int J Soc Psychiatry 2003 49 8 16 12793511 10.1177/0020764003049001537
Gutierrez P King CA Ghziuddin N Adolescent attitudes about death in relation to suicidality Suicide Life Threat Behav 1996 26 8 18 9173613
Stein D Brom D Elizur A Witztum E The association between attitudes toward suicide and suicidal ideation in adolescents Acta Psychiatr Scand 1998 97 195 201 9543307
Currie C Hurrelmann K Settertobulte W Smith R Todd J (Eds) Health and Health Behavior in Young People: International Report From the HBSC 1997/1998 Survey 2000 Copenhagen, Denmark: World Health Organization Regional Office for Europe (WHO Policy Series: Health Policy for Children and Adolescents, Issue 1).
Zaborskis A Makari J Health behaviour of Lithuanian schoolchildren: trends in 1994–1998 and cross-national comparison (in Lithuanian, summary in English) 2001 Panevezys, publisher E. Vaicekauskas
Currie C (Ed) Health Behaviour in School-Aged Children: Research Protocol for the 1997–98 Survey A Word Health Organization Cross-National Study 1998 Edinburgh, Scotland: University of Edinburgh
de Wilde EJ Kienhorst I Suicide attempts in adolescence – "self-report" and "other-report" Crisis 1995 16 59 62, 65 7587291
Moscicki EK O'Carroll P Rae DS Locke BZ Roy A Regier DA Suicide attempts in the Epidemiologic Catchment Area Study Yale J Biol Med 1988 61 259 268 3262956
Kalafat J Elias M Adolescents' experience with and response to suicidal peers Suicide Life Threat Behav 1992 Fall 22 315 321 1440745
Kalediene R Time trends in suicide mortality in Lithuania Acta Psychiatr Scand 1999 99 419 422 10408263
World Health Organization (WHO) World Health Report 2001 – Mental Health – New Understanding, New Hope 2002 Geneva, WHO
Goldney RD Suicide in the young J Paediatr Child Health 1993 29 S50 52 8268024
Beautrais AL Gender issues in youth suicidal behaviour Emerg Med (Fremantle) 2002 14 35 42 11993833
Hawton K Rodham K Evans E Weatherall R Deliberate self harm in adolescents: self report survey in schools in England BMJ 2002 325 1207 1211 12446536 10.1136/bmj.325.7374.1207
Toros F Bilgin NG Sasmaz T Bugdayci R Camdeviren H Suicide attempts and risk factors among children and adolescents Yonsei Med J 2004 45 367 374 15227721
Apter A Waserman D King RA, Apter A Adolescent attempted suicide Suicide in children and adolescents 2003 Cambridge University Press 63 86
Zemaitiene N Psychosocial characteristics of Lithuanian adolescents with suicidal tendencies and its dynamics in 1994–1998 (in Lithuanian) PhD thesis 2000 Kaunas: Kaunas University of Medicine, Department of Public Health
Gothelf D Apter A Brand-Gothelf A Offer N Ofek H Tyano S Pfeffer CR Death concepts in suicidal adolescents J Am Acad Child Adolesc Psychiatry 1998 37 1279 1286 9847500 10.1097/00004583-199812000-00011
Allison S Roeger L Martin G Keeves J Gender differences in the relationship between depression and suicidal ideation in young adolescents Aust N Z J Psychiatry 2001 35 498 503 11531732 10.1046/j.1440-1614.2001.00927.x
Kerkhof Ad JFM Arensman E van Heeringen K Pathways to suicide: The epidemiology of the suicidal process Understanding suicidal behaviour: the suicidal process approach to research, treatment and prevention 2001 John Willey & Sons LTD 15 40
Pelkonen M Marttunen M Child and adolescent suicide: epidemiology, risk factors, and approaches to prevention Paediatr Drugs 2003 5 243 265 12662120
King RA Schwab-Stone M Flisher AJ Greenwald S Kramer RA Goodman SH Psychosocial and risk behavior correlates of youth suicide attempts and suicidal ideation J Am Acad Child Adolesc Psychiatry 2001 40 837 846 11437023 10.1097/00004583-200107000-00019
Kelleher MJ Chambers D King RA, Apter A Cross – cultural variation in child and adolescent suicide Suicide in children and adolescents 2003 Cambridge University Press 170 192
Gailiene D They have not to die: suicides in Lithuania (in Lithuanian) 1998 Vilnius: Tyto Alba
Gailiene D Suicides in Lithuania: social and psychological aspects of the prevalence and the model of prevention PhD Thesis 2000 Kaunas: University of Vytautas Magnus, Department of Sociology
Bille-Brahe U van Heeringen K The suicidal process and society Understanding suicidal behaviour: the suicidal process approach to research, treatment and prevention 2001 John Willey & Sons LTD 182 209
|
16092973
|
PMC1201149
|
CC BY
|
2021-01-04 16:28:54
|
no
|
BMC Public Health. 2005 Aug 11; 5:83
|
utf-8
|
BMC Public Health
| 2,005 |
10.1186/1471-2458-5-83
|
oa_comm
|
==== Front
BMC PhysiolBMC Physiology1472-6793BioMed Central London 1472-6793-5-121610721610.1186/1472-6793-5-12Research ArticleFatigue resistance of rat extraocular muscles does not depend on creatine kinase activity McMullen Colleen A [email protected]ß Katrin [email protected] Francisco H [email protected] Department of Physiology, University of Kentucky, 800 Rose St., Lexington KY 40536-0298, USA2 Department of Cell Biology, University of Potsdam, D-144171 Potsdam, Germany2005 17 8 2005 5 12 12 25 6 2005 17 8 2005 Copyright © 2005 McMullen 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
Creatine kinase (CK) links phosphocreatine, an energy storage system, to cellular ATPases. CK activity serves as a temporal and spatial buffer for ATP content, particularly in fast-twitch skeletal muscles. The extraocular muscles are notoriously fast and active, suggesting the need for efficient ATP buffering. This study tested the hypotheses that (1) CK isoform expression and activity in rat extraocular muscles would be higher, and (2) the resistance of these muscles to fatigue would depend on CK activity.
Results
We found that mRNA and protein levels for cytosolic and mitochondrial CK isoforms were lower in the extraocular muscles than in extensor digitorum longus (EDL). Total CK activity was correspondingly decreased in the extraocular muscles. Moreover, cytoskeletal components of the sarcomeric M line, where a fraction of CK activity is found, were downregulated in the extraocular muscles as was shown by immunocytochemistry and western blotting. CK inhibition significantly accelerated the development of fatigue in EDL muscle bundles, but had no major effect on the extraocular muscles. Searching for alternative ATP buffers that could compensate for the relative lack of CK in extraocular muscles, we determined that mRNAs for two adenylate kinase (AK) isoforms were expressed at higher levels in these muscles. Total AK activity was similar in EDL and extraocular muscles.
Conclusion
These data indicate that the characteristic fatigue resistance of the extraocular muscles does not depend on CK activity.
==== Body
Background
In order to maintain a high effective free energy change of ATP hydrolysis (ΔGATP), metabolically active tissues need to control cellular ATP and ADP levels within a relatively narrow range. This is particularly true in skeletal muscle, as its energy requirements can fluctuate rapidly by more than two orders of magnitude. In mammalian skeletal muscles and other tissues with widely fluctuating metabolic needs, the creatine-phosphocreatine system buffers intracellular ATP concentration: creatine kinase catalyzes the reversible transfer of the phosphoryl group from phosphocreatine to ADP and sustains normal ATP levels [1]. Cellular creatine kinase (CK) activity is due to a family of oligomeric enzymes: two cytosolic, ubiquitous "brain-type" CK-B and "muscle-type" CK-M, and two mitochondrial isoforms, ubiquitous mitochondrial CK (uCK) and "sarcomeric" mitochondrial CK (sCK). The mitochondrial CK isoforms are found as homo-octamers in the inner mitochondrial membrane [2,3]. The cytosolic -M and -B subunits form homo- and heterodimers, CK-MM, -MB- and -BB isoenzymes [2]. In differentiated skeletal muscle, CK-MM and sCK are the predominant isoforms. In fast-twitch muscles, most CK activity is due to the CK-MM isoform, which is preferentially associated with the sarcomeric M line, the sarcoplasmic reticulum and T-tubules [1,4]. This arrangement couples the CK-dependent ATP buffering system to the cellular sites with the highest ATPase activity, and is important for normal contractile function [5,6]. In addition, targeting of CK to these cellular microenvironments localizes CK activity where it is needed, optimizing enzyme distribution [7].
The extraocular muscles, responsible for voluntary and reflexive movements of the eyes, are arguably the fastest and most active skeletal muscles [8-10]. These muscles are typically activated in a biphasic fashion: a high-intensity burst followed by a lower-frequency step (pulse-step) [11]. These functional properties depend on a reliable energy supply and suggest that the extraocular muscles may rely on cytosolic CK-M activity to a great extent. Furthermore, the extraocular muscles are characterized by abundant mitochondria and the presence of developmental and cardiac markers. Then, this muscle group may potentially express CK-B and mitochondrial CK isoforms at higher levels than limb skeletal muscles. Therefore, we tested the hypotheses that (1) the expression and content of CK isoforms and CK activity in rat extraocular muscles would be higher than in fast limb skeletal muscle, and (2) that the fatigue resistance of the extraocular muscles would be critically dependent on normal CK activity. The results showed that the expression and content of muscle-specific cytosolic and mitochondrial CK isoforms is actually lower in extraocular muscle than in extensor digitorum longus (EDL), a prototypical fast limb skeletal muscle. Hence, total CK activity in extraocular muscle is significantly less than in EDL. In addition, the fatigue resistance of the extraocular muscles is not altered when CK activity is inhibited. These data are further evidence that the extraocular muscles manage the metabolic load imposed by their constant activity in a manner not usually seen in typical fast skeletal muscles.
Results
Lower CK activity and isoform expression in extraocular muscle
We used quantitative PCR to compare the expression of all the CK isoforms in the extraocular muscles and EDL. Message for the main cytosolic CK isoform in skeletal muscle, CK-M, was decreased in the extraocular muscles, as was the other cytosolic isoform, CK-B (figure 1A). The lower expression of cytosolic CK isoforms in extraocular muscle does not result in a compensatory increase in the expression of the mitochondrial CKs: mRNA abundance for muscle-specific sCK was also significantly less in the extraocular muscles (figure 1A); uCK was found in EDL by quantitative PCR, but it was below detectable levels in the extraocular muscles. It can then be inferred that uCK expression in these muscles is also significantly lower than in EDL, although no relative comparison could be calculated. In turn, total CK activity in rat extraocular muscles was only ~20% of that in the fast-twitch muscle EDL (4.1 ± 0.6 vs. 20.7 ± 3.0 U/mg protein, respectively, figure 1B). Figure 1C shows western blots that demonstrate that CK-B and uCK isoforms were not detectable in EDL and extraocular muscles. The muscle-specific isoforms CK-M and sCK were found in the two muscles but not in brain. CK-M protein content was less in the extraocular muscles. Although the mitochondrial volume density of the extraocular muscles is ~3- to 4-fold higher than in EDL (McMullen and Andrade, not shown), sCK protein was also less abundant in the extraocular muscles.
Less CK binding sites in extraocular muscle
A fraction of CK-M in skeletal muscles associates with the sarcomeric M lines. Rat extraocular muscle fibers do not have M lines [12,13]. Myomesin and M-protein, the main cytoskeletal components of the M lines, were not detected in extraocular muscles by immunocytochemistry, but were clearly found in EDL fibers in a stereotypical banding pattern (figure 2A). Western blotting failed to detect M-protein in the extraocular muscles (figure 2B). On the other hand, the extraocular muscles contained an alternatively spliced isoform of myomesin, EH-myomesin. This isoform was also detected in embryonic heart together with myomesin (figure 2B). The level of M-protein mRNA measured with quantitative PCR was correspondingly lower in extraocular muscle (13.9-fold less than in EDL); however, myomesin mRNA content was higher (2.1-fold higher than in EDL).
Effect of CK inhibition on fatigue development
The baseline in vitro contractile properties of extraocular and EDL muscles are presented in Table 1. Twitch kinetics (time to peak twitch force, half-relaxation time, and twitch-to-tetanus ratio) were significantly lower in the extraocular muscles than in the EDL muscle bundles. Maximal tetanic forces (P0) of the extraocular muscles were only ~25% of the P0 generated by EDL bundles. Although the contractile properties of the extraocular muscles are very different from those of EDL, they are similar to those reported previously by us and others [14,15].
In our initial experiments, extraocular muscles and EDL bundles were incubated with the CK inhibitor 2,4-dinitro-1-fluorobenzene (DNFB) immediately following the determination of baseline contractile properties. To check for non-specific DNFB-induced toxicity, its effect on P0 was measured with maximal tetanic contractions at 2-min intervals for up to 30 min. DNFB at 50 μM resulted in a drastic loss of force in EDL bundles after only 6 min; P0 decreased by 22.4 ± 3.5% from pre-DNFB baseline (n = 4 muscles, P < 0.05). The magnitude of the force drop in extraocular muscles after 6 min in 50 μM DNFB was less but still significant: P0 decreased by 12.1 ± 2.1% (n = 4 muscles, P < 0.05). On the other hand, DNFB at 10 μM for = 20 min did not alter force significantly in either muscle group: EDL 98.6 ± 1.1% of baseline, and extraocular muscles 97.7 ± 2.0%, n = 4 muscles/each, P = not significant. Therefore, for the rest of the study we incubated muscles in 10 μM DNFB for 20 min, and then washed it out for 10 min before the start of the fatigue protocol. Five minutes before the start of fatigue, during the washout period, P0 was measured to ensure the functional integrity of the muscles. Fatigue was induced with 500-ms submaximal tetani (~50% of peak tetanic force) at 1.5 s intervals until force declined by 50% or for 10 min, whichever occurred first. Untreated control extraocular muscles sustained forces above 50% of initial for the full duration of the fatigue protocol. Control EDL bundles, on the other hand, sustained forces above 50% of the initial for only 4.8 ± 0.5 min. DNFB accelerated the development of fatigue in EDL bundles, such that 60 s into the fatigue protocol there was already a significant difference between the control and DNFB-treated muscles (figure 3A). At the end of the fatigue protocol (270 s), control EDL muscles generated 44 ± 3% of the initial force vs. 32 ± 3% for DNFB-treated EDL bundles (P < 0.05). By contrast, the extraocular muscles proved more resistant to DNFB; the changes in force during the fatigue protocol followed almost the same trajectory with and without DNFB, and after 10 min peak forces were not significantly different: control 60 ± 2 and DNFB 57 ± 3% of initial force (figure 3B).
Upregulation of AK isoforms in extraocular muscle
The "myokinase" reaction (2 ADP → ATP + AMP), catalyzed by AK, serves as an alternative ATP buffering system in skeletal muscle. Total AK activity in extraocular muscles was ~86% of the activity measured in EDL muscles, and this difference was not statistically significant (figure 4A). There are four known AK isoforms in rats, two of which (AK1 and AK2) are found at relatively high levels in skeletal muscle [16,17]. We used quantitative PCR to determine the relative expression of the different AK isoforms in EDL and extraocular muscles. AK1 and AK2 were detected at equivalent levels in the extraocular and EDL muscles; differences were less than the 2-fold threshold for significance (Figure 4B). However, mRNAs for AK3 and AK4 were over 13-fold more abundant in the extraocular muscles.
Discussion
The results from this study demonstrate that CK expression, content and activity are significantly lower in rat extraocular muscles and that CK activity does not explain the fatigue resistance of this muscle group, leading us to reject our initial hypotheses.
Decreased CK expression and activity in extraocular muscle
CK is particularly abundant in fast-twitch skeletal muscles; therefore, we measured total CK activity and CK isoform expression in EDL, a prototypical fast limb muscle, and in the extraocular muscles, which are also composed mostly of fast-twitch fibers. Surprisingly, CK activity was significantly less in the extraocular muscles, only 20% of the level measured in EDL. Message abundance and protein content of the muscle-specific CK-M were also lower in the extraocular muscles (figure 1). A fraction of CK-M is normally associated with the sarcomeric M lines, structures that support the thick filament lattice and are more prominent in fast-twitch muscle fibers [4,6]. In consequence, we explored whether low CK-M content correlated with the absence of M lines characteristic of most rat extraocular muscle fibers [12,13]. Our results confirmed the previously described altered expression pattern of the M line cytoskeletal components, myomesin and M-protein, in rodent extraocular muscles (figure 2) [13,18]. The lack of M-protein is particularly intriguing since this cytoskeletal protein is found almost exclusively in fast twitch fibers [19,20]. This finding also correlates with the description of fainter M lines in skeletal muscles from CK-M knockout mice [21]; apparently the association of the enzyme with the M lines contributes importantly to the electron density of these structures. Myomesin was also less abundant in extraocular muscle, as shown by immunocytochemistry and western blot, despite the fact that message levels were higher than in EDL. This finding is at variance from our recent study of mouse extraocular muscles [18]. It may indicate that post-transcriptional control of myomesin expression is relatively more important in the extraocular muscles of adult rats. Western blotting confirmed that rat extraocular muscles contain an alternatively spliced myomesin isoform, EH-myomesin, originally described in embryonic heart [22,23]. The fact that this isoform was not detected by immunocytochemistry in the present study may reflect low protein abundance.
Extraocular muscles express developmental and cardiac markers [24]. Since the CK-MB heterodimer is the prevalent CK isoform in developing skeletal muscles and in adult myocardium, we anticipated a relative upregulation of CK-B in extraocular muscle compared to EDL. Our results did not bear this out; CK-B mRNA and protein were also significantly less in the extraocular muscles (figure 1).
Interestingly, the fast-twitch skeletal muscle fibers from CK-M knockout mice have higher mitochondrial contents and increased sCK activity [21]. Extraocular muscle fibers have very high mitochondrial contents [12]. It has been suggested that a greater number of mitochondria reduces the diffusion distance between these organelles and myofibrills and other ATP sinks, compensating for the loss of cytosolic CK [21,25]. By analogy, the high mitochondrial content of extraocular muscles would achieve the same goal, minimizing the need for the temporal and spatial ATP buffering capacity normally provided by cytosolic CK activity. Then, the movement of ATP from mitochondria to cytosolic phosphocreatine may not be required in the extraocular muscles. Despite the high mitochondrial content of extraocular muscles, mRNA and protein levels of sCK and uCK were lower in these muscles; uCK was actually undetectable in the extraocular muscles by quantitative PCR. The lower expression and content of sCK are particularly puzzling; it has been proposed that sCK mediates the high-energy phosphoryl flux from mitochondria to cytosol and regulates oxidative phosphorylation [26].
CK inhibition and fatigue resistance
The most physiologically significant finding in this study was that CK inhibition with DNFB did not change the fatigue resistance of the extraocular muscles, even though it significantly accelerated the development of fatigue in the fast-twitch muscle EDL (figure 3). The lack of effect of DNFB on extraocular muscle fatigue was more impressive because these muscles were stimulated at a higher frequency than the EDL bundles in order to fulfill the requirement that all muscles start the fatigue protocol at 50% of maximal force. In addition, endurance (time to decrease force by 50%) was much longer in the extraocular muscles, since all EDL bundles reached this target force by 3 min, and all extraocular muscles sustained forces >50% of initial for 10 min. The use of phosphocreatine in skeletal muscle energetics allows for high power output and effective buffering of ATP concentration. While maintenance of constant ATP concentration during contractile activity is neither universal not absolutely required, the use of phosphocreatine as fuel increases the power output attainable by skeletal muscles. This and delaying fatigue after the onset of high-intensity stimulation are the most perturbed functions in CK knockout mice [21,25]. However, CK activity may actually contribute to fatigue development during prolonged stimulation by increasing the cytosolic concentration of inorganic phosphate [27]. CK would then be deleterious to a muscle group that is constantly active and whose energy requirements may not rely on traditional ATP buffers [28,29]. Because most skeletal muscle fibers are recruited during fairly short periods, peak energy demand is much greater than the average demand [30]. During these sporadic periods of peak demand, CK provides the "metabolic capacitance" needed to allow the system to have only enough mitochondria to provide for average energy demand. Such arrangement may not serve constantly active muscles such at the extraocular muscles. Interestingly, skeletal muscles with genetically impaired CK activity resemble the normal extraocular muscle phenotype: high mitochondrial content, low force and power [12,31]. Both phenotypes increase ATP generating capacity and decrease the size of the ATP sinks.
Upregulation of AK isoforms in extraocular muscle
Preferential targeting of CK to myofibrillar and mitochondrial compartments serves to localize activity to where it is needed, thus economizing enzyme distribution. This would be particularly important in typical skeletal muscle fibers where sparsely and non-uniformly distributed mitochondria impose large diffusion distances [7]. But it would be much less so in the mitochondria-rich and very small extraocular muscle fibers, explaining the lack of M lines and low CK activity. Instead, mRNAs for the mitochondria-associated AK3 and AK4 isoforms are present at higher levels in these muscles, confirming an earlier gene expression profile study that found AK4 at higher levels in mouse extraocular muscles [28]. Interestingly, AK activity is increased in CK-deficient mice, and inhibition of CK in intact skeletal muscle results in increased phosphoryl transfer via AK [32,33]. However, total AK activity was not significantly different between EDL and extraocular muscle. Whether the putative mitochondrial localization of AK3 and AK4 is designed to replace CK activity associated with mitochondria in the extraocular muscles remains to be determined.
Another possible role for AK in extraocular muscle is that it may mediate an alternative strategy for metabolic control. The lower expression of mitochondrial CK in extraocular muscles would diminish its proposed role for fine regulation and amplification of the energy state signal from the cytoplasm and control of mitochondrial oxidative phosphorylation [26,34]. Instead, AMP from the AK-catalyzed myokinase reaction would serve as a strong positive allosteric signal on 6-phosphofructo-1-kinase (PFK) and inhibit fructose 1,6-bisphosphatase, activating glycolysis. This control step could be particularly important in the extraocular muscles given their apparent reliance on fatty acid oxidation and glycolysis and not on glycogen breakdown [28,29].
Conclusion
Our data indicate that the fatigue resistance of the fast and constantly active extraocular muscles does not depend on high CK activity. In consequence, the expression of the known CK isoforms is downregulated in these muscles. These findings strengthen emerging evidence that the extraocular muscles follow a different strategy or design to cope with their peculiar mechanical and metabolic loads.
Methods
Animals
Ethical use of experimental animals was approved by the Institutional Animal Care and Use Committee. Three-month old male Sprague Dawley rats (Harlan, Indianapolis, IN) were anesthetized with ketamine hydrochloride/xylazine hydrochloride (100 mg/8 mg per kg body weight, i.p. injection) and killed by exsanguination following a medial thoracotomy. For in vitro function, extraocular muscles were dissected intact from bony origin to distal tendon. Small EDL bundles were taken to yield a representative fast limb skeletal muscle sample. For gene expression and protein studies, all extraocular muscles and mid-belly samples of EDL were quickly excised, frozen in liquid nitrogen and stored at -80°C. For immunocytochemistry, whole muscles were dissected, pinned to cork at resting length, covered with OCT embedding medium and frozen in 2-methylbutane cooled in liquid nitrogen.
Enzyme assays
EDL and extraocular muscle samples were homogenized (1:100 w/v) in 26 mM Tris, 30 mM dithiothreitol, 0.3 M sucrose and 1% Triton X-100 (pH 8.0), and extracted on ice for 1 hr. CK activity was determined in triplicate by a hexokinase/glucose-6-phosphate dehydrogenase coupled system, which yields NADH at a rate proportional to CK activity (Sigma Chemical Co., St. Louis, MO), and expressed as units/mg of protein. AK activity was determined in triplicate by a pyruvate kinase/lactate dehydrogenase coupled system which follows the consumption of NADH, and expressed as pmoles NADH/min/mg of protein. Protein content of muscle homogenates was determined by the Lowry method [35].
Real-time quantitative PCR
Muscles were pulverized under liquid nitrogen and total RNA was isolated using Trizol (GibcoBRL, Rockville, MD) according to the manufacturer's instructions. Muscles from four animals were combined into each total RNA sample to lessen the effect of inter-subject variability. Reverse transcription was carried out using Superscript II RNAse H- Reverse Transcriptase (Invitrogen, Carlsbad, CA) with random hexamers. Primers for the mRNAs of interest were designed with the software package Primer Express version 1.5 (Applied Biosystems, Foster City, CA) from GenBank nucleotide sequences and are shown in Table 2. cDNA samples (2 μg each) were analyzed in triplicate with the ABI Prism 7700 Sequence Detection System (Applied Biosystems) using ABI SYBR Green and β-actin as the calibrator housekeeping gene. The relative abundance of target mRNAs in the extraocular and EDL muscles was determined with the comparative cycle threshold method [36,37].
Immunocytochemistry
Five-μm thick longitudinal cryosections from extraocular and EDL muscles were fixed with 2% paraformaldehyde, blocked with 0.1% bovine serum albumin in phosphate-buffered saline, and incubated overnight at 4°C with monoclonal antibodies specific for myomesin or M-protein (BB78 and AA259, respectively) [38]. After washing with phosphate-buffered saline, slides were incubated for 1 h in Alexa Fluor 488-conjugated secondary antibody (1:50; Molecular Probes Inc., Eugene, OR), rinsed in phosphate-buffered saline, and mounted in Immu-Mount (Shandon, Pittsburgh, PA). Sections were examined and imaged with a Zeiss LSM 410 confocal microscopy system.
Immunoblotting
For the analysis of myomesin, EH-myomesin, M-line protein and CK isoform content, tissue samples were homogenized in a buffer containing 26 mM Tris-HCl, 0.3 M sucrose, 30 mM DTT and 1% Triton X-100; pH was adjusted to 8.0. Western blots for myomesin, EH-myomesin and M-protein were carried out using previously described monoclonal antibodies and chemiluminescence [38]. Polyclonal antisera to synthetic peptides identical to rat amino acid sequences of the cytosolic CK-M and CK-B (Rockland Immunochemicals), and sCK and uCK (Affinity Bioreagents) were raised in New Zeland white rabbits and collected by terminal bleeding. Specificity of antisera was determined by ELISA and dot blot. Western blots for CK isoforms were carried out as follows: 50 μg of total protein per sample were electrophoresed and transferred to PVDF membranes. Blocked membranes were probed with CK isoform-specific sera (1:1000), followed by a horseradish peroxidase-conjugated secondary antibody (Bio-Rad). Membranes were then developed with 4-chloro-1 naphthol, scanned and analyzed using ImageJ 1.30v [39].
Isolated muscle preparation
Whole extraocular muscles and small EDL muscle bundles (~10–15% of total muscle mass) were placed in a small muscle chamber. The distal tendon was attached to a micropositioner and the proximal bone fragment to a force transducer (AE801, SensoNor, Horten, Norway or ELG-H, Entran, Fairfield, NJ). The chamber was superfused with a physiological salt solution: (in mM) 137 NaCl, 5 KCl, 2.0 CaCl2, 1.0 MgSO4, 1.0 Na2HPO4, 24 NaHCO3, 11 glucose, and 0.026 d-tubocurarine, bubbled with a 95% O2-5% CO2 gas mixture to maintain pH at 7.4 at 25°C. To inhibit muscle CK activity, 2,4-dinitro-1-fluorobenzene (DNFB, Sigma Chemical Co.) was added to the bath from a 100 mM stock solution in dimethyl sulfoxide (DMSO) [33,40,41]. Muscles were stimulated with 0.5 ms pulses delivered by an S48 stimulator (Grass Instruments, Braintree, MA) to platinum electrodes in the muscle chamber and stretched to the length giving maximum tetanic force (optimal length, L0). Force signals were stored in a personal computer for analysis. At the end of the study, the length of muscle fibers at L0 was measured and bone and tendons removed. The muscles were blotted dry and weighed. Force (in Newtons) was normalized to muscle cross sectional area (cm2) [42].
Fatigue protocol
After measuring baseline contractile properties, muscles were divided into two groups: DNFB-treated and time-matched controls. Pilot studies demonstrated that 30-min exposures to DMSO at the concentrations used here (0.01 to 0.05%) followed by 10-min washout had no effect on the response to the fatigue protocol. Fatigue was induced with the following protocol: muscles were stimulated at a frequency giving approximately one half of maximal tetanic force (30–50 Hz for EDL bundles, 50–70 Hz for extraocular muscles) for 500 ms, followed by 1.5 s interval between contractions, until force in the control group declined to approximately 50% of the level at time = 0 or for 10 min, whichever occurred first.
Data analysis
All results are presented as means ± s.e.m. of n observations, unless otherwise noted. Statistical significance was determined at the 95% confidence level using Student's t test for unpaired or paired samples as indicated; the treatment effect in the fatigue runs was determined by analysis of variance.
Authors' contributions
CAM was responsible for real-time quantitative PCR, immunocytochemistry and confocal microscopy. KH contributed the immunoblotting of M line proteins. FHA performed the functional studies, enzyme assays and immunoblotting of CK isoforms. All authors participated in the experimental design and manuscript preparation.
Acknowledgements
The authors thank Dr. Peter van der Ven (University of Potsdam) for stimulating discussions and for the monoclonal antibodies for myomesin and M-protein. The advice and technical assistance of Denise Hatala and Carol Luckey (Vision Science Research Center), and Midori Hitomi (Department of Neurosciences) at Case Western Reserve University are gratefully acknowledged. This work was supported by grants from the National Eye Institute (EY12998 and EY13724 to FHA).
Figures and Tables
Figure 1 Lower CK isoform expression and activity in rat extraocular muscle. (A) Relative abundance of three CK isoform mRNAs in EDL and extraocular muscle by quantitative PCR. Each sample includes pooled muscles from 4 rats. Data are means from 3–4 samples, and are normalized to levels measured in EDL = 100%. Message levels for the cytosolic CK-M and CK-B, and the mitochondrial sCK were much lower in extraocular muscles compared to EDL. (B) Total CK activity in EDL and extraocular muscles. EDL muscles have ~5-fold greater CK activity than extraocular muscles (n = 8, P < 0.001). (C) CK protein isoforms found in brain, EDL and extraocular muscle by western blot.
Figure 2 Myomesin and M-protein are downregulated in rat extraocular muscles. (A) Representative confocal micrographs showing that myomesin and M-protein are found in EDL muscle fibers (EDL, top) in the same stereotypical banding pattern, but are not detectable in extraocular muscle (EOM, bottom). Scale bar = 20 μm. (B) Western blots demonstrating the presence of myomesin and M-protein in EDL and heart (H). Embryonic heart (He) did not have M-protein, but two bands reacted with the myomesin antibody: myomesin and EH-myomesin, an alternatively spliced isoform of higher molecular weight. Extraocular muscle (EOM) contained EH-myomesin, but did not have detectable levels of myomesin or M-protein.
Figure 3 CK inhibition does not alter fatigue resistance of rat extraocular muscle. (A) DNFB (black triangles) decreases the fatigue resistance of EDL muscle bundles (n = 6 muscles per treatment, * P < 0.05 control vs. DNFB at the corresponding time points). (B) DNFB (black circles) does not accelerate fatigue development in extraocular muscles (n = 6 muscles per treatment). Vertical dotted line is included for comparison to mark to the end of the fatigue protocol for EDL muscles.
Figure 4 Upregulation of AK isoforms in rat extraocular muscle. (A) Total AK activity in EDL and extraocular muscles (n = 9, P > 0.35). (B) Relative abundance of mRNAs for four AK isoforms in extraocular muscle by quantitative PCR. Each sample includes pooled muscles from 4 rats. Data are mean -fold differences (3–4 samples) in AK isoform expression and normalized to EDL muscle = 1.0, shown by the horizontal reference line. The abundance of AK1 and AK2 mRNAs was not significantly different in extraocular and EDL muscles (black bars). AK3 and AK4 are expressed at significantly higher levels in extraocular muscle (13.5- and 16.2-fold greater than EDL, respectively).
Table 1 Contractile properties of EDL and extraocular muscles
TPT (ms) HRT (ms) PT/P0 (%) P0 (N/cm2) n
EDL 25.9 ± 0.5 25.4 ± 0.5 24.6 ± 1.7 24.2 ± 0.4 16
EOM 13.2 ± 0.4* 13.4 ± 0.4* 8.0 ± 0.6* 6.4 ± 0.2* 16
TPT, time to peak twitch force; HRT, half-relaxation time; PT/P0, twitch-to-tetanus ratio; P0, maximal tetanic force. EOM = extraocular muscles. * P < 0.001 EDL vs. extraocular muscle.
Table 2 Primer sequences for quantitative PCR
Name Size (bp) Forward / Reverse primers (5' → 3')
CK-M 76 GCCATGGTGGCTTCAAACC / CAGATCGTCTCCACCCTTGAG
CK-B 67 AAGCCCTGTCCAGCCTAGATG / CGCCTCGGTCATGCTCTT
sCK 73 GAGCTGGTGTCCACGTTAGGA / CCGCAGGTTCTCCAAGATCTT
uCK 82 CTGACCTTGATGCCAGCAAA / ACTTCGGCCAGTTCTGACTCTT
AK1 78 CTTCCAACGGCTTCTTGATTG / TGTGCGATCTTCCGTTCAAA
AK2 71 TGGAGGCCTACCACATCAGA / CAATGGCGCAGTGAATGA
AK3 75 ACCAGTTGTAGCTGGCTGTTG / TCTGATAAACTCTCCAGGGCTTCT
AK4 71 CTTCCAGCGGGAGGGTCTTAT / CGCCGGTGATGTCATCAAC
Myomesin 75 GGCCCACATTTCGCTGAGTA / TTGCCACCTTGCATTTCAAC
M-protein 96 GAGAGGGCGAGACAGTCACACT / CTCGTGGCTGTACTCTCCTCAGA
==== Refs
Wallimann T Wyss M Brdiczka D Nicolay K Eppenberger HM Intracellular compartmentation, structure and function of creatine kinase isoenzymes in tissues with high and fluctuating energy demands: the "phosphocreatine circuit" for cellular energy homeostasis Biochem J 1992 281 21 40 1731757
Schlegel J Zurbriggen B Wegmann G Wyss M Eppenberger HM Wallimann T Native mitochondrial creatine kinase forms octamer structures. I. Isolation of two interconvertible mitochondrial creatine kinase forms, dimeric and octameric mitochondrial creatine kinase: characterization, localization, and structure-function relationships J Biol Chem 1988 263 16942 16953 3182823
Schlegel J Wyss M Eppenberger HM Wallimann T Functional studies with the octameric and dimeric form of mitochondrial creatine kinase. Differential pH-dependent association of the two oligomeric forms with the inner mitochondrial membrane J Biol Chem 1990 265 9221 9227 2345172
Squire JM Architecture and function in the muscle sarcomere Curr Opin Struct Biol 1997 7 247 257 9094325 10.1016/S0959-440X(97)80033-4
Ventura-Clapier R Kuznetsov AV d'Albis A van Deursen J Wieringa B Veksler VI Muscle creatine kinase-deficient mice. I. Alterations in myofibrillar function J Biol Chem 1995 270 19914 19920 7650006 10.1074/jbc.270.34.19914
Wallimann T Schlösser T Eppenberger HM Function of M-line-bound creatine kinase as intramyofibrillar ATP regenerator at the receiving end of the phosphorylcreatine shuttle in muscle J Biol Chem 1984 259 5238 5246 6143755
Meyer RA Sweeney HL Kushmerick MJ A simple analysis of the "phosphocreatine shuttle" Am J Physiol 1984 246 C365 C377 6372517
Porter JD Baker RS Ragusa RJ Brueckner JK Extraocular muscles: basic and clinical aspects of structure and function Surv Ophthalmol 1995 39 451 484 7660301
Asmussen G Beckers-Bleukx G Maréchal G The force-velocity relation of the rabbit inferior oblique muscle; influence of temperature Pflügers Arch 1994 426 542 547 8052524 10.1007/BF00378532
Fuchs AF Scudder CA Kaneko CRS Discharge patterns and recruitment order of identified motoneurons and internuclear neurons in the monkey abducens nucleus J Neurophysiol 1988 60 1874 1895 2466962
Robinson DA Oculomotor unit behavior in the monkey J Neurophysiol 1970 33 393 404 4985725
Mayr R Structure and distribution of fibre types in the external eye muscles of the rat Tissue Cell 1971 3 433 462
Porter JD Merriam AP Gong B Kasturi S Zhou X Hauser KF Andrade FH Cheng G Postnatal suppression of myomesin, muscle creatine kinase and the M-line in rat extraocular muscle J Exp Biol 2003 206 3101 3112 12878677 10.1242/jeb.00511
McMullen CA Andrade FH Stahl JS Functional and genomic changes in the mouse ocular motor system in response to light deprivation from birth J Neurosci 2004 24 161 169 14715949 10.1523/JNEUROSCI.3234-03.2004
Asmussen G Gaunitz U Mechanical properties of the isolated inferior oblique muscle of the rabbit Pflügers Arch 1981 392 183 190 7322845 10.1007/BF00581270
Noma T Song S Yong-Shik Y Tanaka S Nakazawa A cDNA cloning and tissue-specific expression of the gene encoding human adenylate kinase isozyme 2 Biochim Biophys Acta 1998 1395 34 39 9434148
Tanabe T Yamada M Noma T Kajii T Nakazawa A Tissue-specific and developmentally regulated expression of the genes encoding adenylate kinase isoenzymes J Biochem 1993 113 200 207 8468325
Andrade FH Merriam AP Guo W Cheng G McMullen CA Hayeß K van der Ven PFM Porter JD Paradoxical absence of M lines and downregulation of creatine kinase in mouse extraocular muscle J Appl Physiol 2003 95 692 699 12716871
Carlsson E Grove BK Wallimann T Eppenberger HM Thornell L-E Myofibrillar M-band proteins in rat skeletal muscles during development Histochemistry 1990 95 27 35 1704877 10.1007/BF00737225
Grove BK Cerny L Perriard J-C Eppenberger HM Thornell L-E Fiber type-specific distribution of M-band proteins in chicken muscle J Histochem Cytochem 1989 37 447 454 2926123
van Deursen J Heerschap A Oerlemans F Ruitenbeek W Jap P ter Laak H Wieringa B Skeletal muscles of mice deficient in muscle creatine kinase lack burst activity Cell 1993 74 621 631 8358791 10.1016/0092-8674(93)90510-W
Agarkova I Schoenauer R Ehler E Carlsson L Carlsson E Thornell L-E Perriard J-C The molecular composition of the sarcomeric M-band correlates with muscle fiber type Eur J Cell Biol 2004 83 193 204 15346809 10.1078/0171-9335-00383
Agarkova I Auerbach D Ehler E Perriard J-C A novel marker for vertebrate embryonic heart, the EH-myomesin isoform J Biol Chem 2000 275 10256 10264 10744711 10.1074/jbc.275.14.10256
Porter JD Baker RS Muscles of a different "color": the unusual properties or the extraocular muscles may predispose or protect them in neurogenic and myogenic disease Neurology 1996 46 30 37 8559415
Steeghs K Benders A Oerlemans F de Haan A Heerschap A Ruitenbeek W Jost C van Deursen J Perryman B Pette D Brückwilder M Koudijs J Jap P Veerkamp J Wieringa B Altered Ca2+ responses in muscles with combined mitochondrial and cytosolic creatine kinase deficiencies Cell 1997 89 93 103 9094718 10.1016/S0092-8674(00)80186-5
Kay L Nicolay K Wieringa B Saks V Wallimann T Direct evidence for the control of mitochondrial respiration by mitochondrial creatine kinase in oxidative muscle cells in situ J Biol Chem 2000 275 6937 6944 10702255 10.1074/jbc.275.10.6937
Dahlstedt AJ Katz A Wieringa B Westerblad H Is creatine kinase responsible for fatigue? Studies of isolated skeletal muscle deficient in creatine kinase FASEB J 2000 14 982 990 10783153
Porter JD Khanna S Kaminski HJ Rao JS Merriam AP Richmonds CR Leahy P Li J Andrade FH Extraocular muscle is defined by a fundamentally distinct gene expression profile Proc Natl Acad Sci USA 2001 98 12062 12067 11572940 10.1073/pnas.211257298
Fischer MD Gorospe JR Felder E Bogdanovich S Pedrosa-Domellöf F Ahima RS Rubinstein NA Hoffman EP Khurana TS Expression profiling reveals metabolic and structural components of extraocular muscles Physiol Genomics 2002 9 71 84 12006673
Sweeney HL The importance of the creatine kinase reaction: the concept of metabolic capacitance Med Sci Sports Exerc 1994 26 30 36 8133735
Close RI Luff AR Dynamic properties of inferior rectus muscle of the rat J Physiol 1974 236 259 270
LaBella JJ Daood MJ Koretsky AP Roman BB Sieck GC Wieringa B Watchko JF Absence of myofibrillar creatine kinase and diaphragm isometric function during repetitive activation J Appl Physiol 1998 84 1166 1173 9516180
Dzeja PP Zeleznikar RJ Goldberg ND Suppression of creatine kinase-catalyzed phosphotransfer results in increased phosphoryl transfer by adenylate kinase in intact skeletal muscle J Biol Chem 1996 271 12847 12851 8662747 10.1074/jbc.271.22.12847
Wallimann T Dolder M Schlattner U Eder M Hornemann T O'Gorman E Ruck A Brdiczka D Some new aspects of creatine kinase (CK): compartmentation, structure, function and regulation for cellular and mitochondrial bioenergetics and physiology Biofactors 1998 8 229 234 9914824
Lowry OH Rosebrough NJ Farr AL Randall RJ Protein measurement with the Folin phenol reagent J Biol Chem 1951 193 265 272 14907713
Giulietti A Overbergh L Valckx D Decallonne B Bouillon R Mathieu C An overview of real-time quantitative PCR: applications to quantify cytokine gene expression Methods 2001 25 386 401 11846608 10.1006/meth.2001.1261
Livak KJ Schmittgen TD Analysis of relative gene expression data using real-time quantitative PCR and the 2ΔΔCt method Methods 2001 25 402 408 11846609 10.1006/meth.2001.1262
Vinkemeier U Obermann W Weber K Fürst DO The globular head domain of titin extends into the center of the sarcomeric M band J Cell Sci 1993 106 319 330 7505783
ImageJ 1.30v
Dahlstedt AJ Westerblad H Inhibition of creatine kinase reduces the rate of fatigue-induced decrease in tetanic [Ca2+]i in mouse skeletal muscle J Physiol 2001 533 639 649 11410623 10.1111/j.1469-7793.2001.00639.x
Tombes RM Shapiro BM Metabolite channeling: A phosphorylcreatine shuttle to mediate high energy phosphate transport between sperm mitochondrion and tail Cell 1985 41 325 334 3995586 10.1016/0092-8674(85)90085-6
Andrade FH Anzueto A Napier W Levine S Lawrence RA Jenkinson SG Maxwell LC Effects of selenium deficiency on diaphragmatic function after resistive loading Acta Physiol Scand 1998 162 141 148 9550226 10.1046/j.1365-201X.1998.0265f.x
|
16107216
|
PMC1201150
|
CC BY
|
2021-01-04 16:03:51
|
no
|
BMC Physiol. 2005 Aug 17; 5:12
|
utf-8
|
BMC Physiol
| 2,005 |
10.1186/1472-6793-5-12
|
oa_comm
|
==== Front
BMC Plant BiolBMC Plant Biology1471-2229BioMed Central London 1471-2229-5-151610217010.1186/1471-2229-5-15Research ArticleHighly syntenic regions in the genomes of soybean, Medicago truncatula, and Arabidopsis thaliana Mudge Joann [email protected] Steven B [email protected] Peter [email protected] Giles ED [email protected] Bruce A [email protected] Christopher D [email protected] Nevin D [email protected] Dept of Plant Pathology, 495 Borlaug Hall, University of Minnesota, St. Paul, MN 55108 USA2 Dept. of Disease and Stress Biology, John Innes Centre, Norwich Research Park, Colney Norwich, NR4 7UH, UK3 The Advanced Center for Genome Technology (ACGT), Stephenson Research & Technology Center, University of Oklahoma, Norman OK 73019 USA4 The Institute for Genomic Research (TIGR), 9712 Medicago Center Drive, Rockville, MN 20850 USA2005 15 8 2005 5 15 15 31 3 2005 15 8 2005 Copyright © 2005 Mudge 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
Recent genome sequencing enables mega-base scale comparisons between related genomes. Comparisons between animals, plants, fungi, and bacteria demonstrate extensive synteny tempered by rearrangements. Within the legume plant family, glimpses of synteny have also been observed. Characterizing syntenic relationships in legumes is important in transferring knowledge from model legumes to crops that are important sources of protein, fixed nitrogen, and health-promoting compounds.
Results
We have uncovered two large soybean regions exhibiting synteny with M. truncatula and with a network of segmentally duplicated regions in Arabidopsis. In all, syntenic regions comprise over 500 predicted genes spanning 3 Mb. Up to 75% of soybean genes are colinear with M. truncatula, including one region in which 33 of 35 soybean predicted genes with database support are colinear to M. truncatula. In some regions, 60% of soybean genes share colinearity with a network of A. thaliana duplications. One region is especially interesting because this 500 kbp segment of soybean is syntenic to two paralogous regions in M. truncatula on different chromosomes. Phylogenetic analysis of individual genes within these regions demonstrates that one is orthologous to the soybean region, with which it also shows substantially denser synteny and significantly lower levels of synonymous nucleotide substitutions. The other M. truncatula region is inferred to be paralogous, presumably resulting from a duplication event preceding speciation.
Conclusion
The presence of well-defined M. truncatula segments showing orthologous and paralogous relationships with soybean allows us to explore the evolution of contiguous genomic regions in the context of ancient genome duplication and speciation events.
==== Body
Background
The rapid increase in eukaryotic genome sequence in recent years enables genome-wide alignments, megabase (Mb)-scale comparisons between species, and fine-scaled phylogenetic footprinting. Recent sequenced-based studies in a variety of organisms have described high levels of synteny (conservation of gene content and order between species) within kingdoms and between families, but have also highlighted frequent synteny loss and degradation due to gene duplication, deletion, and rearrangement. In some cases, observed synteny has been extensive. In vertebrates, over 90% of the mouse and human genomes (separated by 91 million years; My) lie in syntenic blocks [1,2], some exceeding 40 Mb [2,3]. At a greater evolutionary distance (310 My), the human and chicken genomes show large synteny blocks, including at least 70 Mb of highly conserved sequence [2,4]. Regions syntenic to 1.8 Mb of human DNA were identified in twelve different species including fish, which separated from humans 450 Mya [2,5].
High levels of synteny have also been found in plant families. Molecular marker analysis has allowed chromosome-by-chromosome alignments of several genera within the Solanaceae, Fabaceae, and Poaceae [6-8]. Generally, syntenic relationships are complicated by micro- and macro-rearrangements as well as duplications [9]. Complete genome sequences of rice and A. thaliana, models representing the two major clades of flowering plants, allows comparisons across a greater evolutionary distance. Separated by 200 My, rice and Arabidopsis thaliana nonetheless retain substantial conserved syntenic blocks, including one region spanning 119 A. thaliana genes [10].
Though genomic relationships within legumes are less well characterized, a growing number of studies have begun to reveal extensive synteny between the members of this important plant family. Based on restriction fragment length polymorphisms (RFLPs), substantial genome conservation was discovered among Phasoloid species, including mungbean (Vigna radiata) and cowpea (V. unguiculata), extending as long as entire chromosomes [11]. Comparable levels of synteny were later demonstrated between Vigna and the common bean, Phaseolus vulgaris [12]. Synteny with the more distant soybean, Glycine max, was more limited, typically on the order of 10 – 20 cM. Later, Lee et al. [13] observed higher levels of conservation between bean, mungbean, and soybean, where A. thaliana also showed conservation to some conserved legume regions and even helped to elucidate duplicated regions in soybean. Choi et al. [6] described genome-wide macrosynteny among legumes using a large set of cross-species genetic markers. Though genomic correspondence was reduced by chromosomal rearrangements increasing with phylogenetic distance, they could align chromosomes from a variety of Papilionoid species, including Medicago truncatula and soybean.
M. truncatula and Lotus japonicus are two model legumes that are now targets of large-scale genome sequencing. With more than 100 Mb of genome sequence publicly available in both, genome-scale comparisons at both the macro- and micro-syntenic level are possible. Young et al [14] compared all finished and anchored sequence between these two genomes (111 Mb) and concluded that more than 75% of both genomes reside in conserved, syntenic segments. At a microsyntenic scale, Choi et al. [6] analyzed ten BAC/TAC clone pairs and found 80% of genes were conserved and colinear. Soybean has also been compared to M. truncatula because of its economic importance. With few sequences 100 kbp or more in length available, however, comparisons of soybean with reference legumes have been limited to low resolution surveys and short contiguous segments. Nevertheless, conserved synteny is widespread between M. truncatula and soybean. Yan et al. [15] analyzed three homologous BAC contig groups in detail by comparative physical mapping and cross-hybridization and found six of eight genome regions exhibited conserved synteny, including three that were extensively conserved. In genome-wide survey of synteny, slightly more than half of 50 RFLP-based soybean BAC-contigs, each approximately 200 kbp in size, exhibited conserved synteny with M. truncatula [16] and nearly 75% of these cases were extensive.
In the course of our genome sequencing work in M. truncatula [14], two regions were observed to be significantly conserved with previously sequenced regions of soybean. These soybean regions contain two important soybean cyst nematode (SCN) (Heterodera glycines) resistance loci, rhg1 and Rhg4, which have been studied extensively reviewed in Concibido, et al. [17]. In previous work, our lab and others localized the genetic positions of these genes and characterized their role in resistance [18-24]. We saturated the regions with genetic markers, developed high throughput molecular markers, created physical maps, and characterized homoeologous and surrounding genome regions [23,25-30]. As a result of the extensive information available and the importance of SCN resistance, these genome regions were eventually sequenced [31,32], including the tentative cloning of rhg1 and Rhg4 (gene 29 in Figure 1 and gene 21 in Figure 2, respectively).
A preliminary examination of the soybean rhg1 region described in the present study concluded that nearly 70% of genes were conserved and colinear between soybean and M. truncatula [6]. Previously, Foster-Hartnett et al. [29] had used survey sequences along a 1 Mb stretch that included and extended beyond the region described here to examine syntenic relationships with A. thaliana. Based on survey (primarily BAC-end) sequence that included both genic and non-genic regions, 35% of soybean sequences were conserved in one or more syntenic A. thaliana regions. The soybean region formed a network of synteny with six different A. thaliana regions. The longest syntenic segment in A. thaliana was more than 2 Mb in length, highlighting the existence of long stretches of conserved sequence between distantly related genomes [29].
In the present study, we describe the gene content in two soybean regions totaling approximately 1 Mb in size with more than 150 genes [32] that exhibit extensive synteny to M. truncatula and A. thaliana. The two soybean regions reside on different chromosomes but are functionally linked – each contains a receptor-like kinase gene (rhg1 or Rhg4) tentatively identified as a resistance gene to SCN (Heterodera glycines). Up to 75% of soybean genes in this region are colinear with M. truncatula, including one 300 kbp segment with 33 of 35 soybean genes colinear to M. truncatula. Nearly 60% of the genes in this same soybean region exhibit colinearity with one or more A. thaliana regions. These highly syntenic blocks are discussed in the context of phylogenetic data revealing patterns of evolution and orthologous and paralogous relationships.
Results
Identifying and mapping homologous contigs
We used the Genbank soybean sequences surrounding SCN resistance loci as a basis for searching all available M. truncatula BAC sequences and the A. thaliana proteome for syntenic regions (Table 1). Homologous regions were then used to create corresponding sequence assemblies for all three genera (Figures 1, 2). Where possible, we also merged sequence M. truncatula assemblies by identifying end-sequenced BACs that spanned gaps between non-overlapping BACs. Most of the M. truncatula BACs could be anchored to the M. truncatula genetic map (Table 1) [33].
There is a gap in all three species between synteny blocks 1a and 1b which we were unable to span. In soybean, these two sequence assemblies are genetically linked and located 2 cM apart on LG-G. In addition, synteny block 1b contained two gaps in M. truncatula (Figure 1), one a 90 kbp gap toward the bottom M. truncatula sequence assembly. This gap and its surrounding M. truncatula sequence (totaling about 175 kbp) correspond to an insertion/deletion in soybean just 25 kbp in size and containing two gene models without hits to Genbank's nonredundant (nr) database along with one repetitive sequence. There is also a gap of unknown size between the top and bottom sequence assemblies in synteny block 1b that could not be spanned with end sequenced BACs. In M. truncatula, the top sequence assembly maps to M. truncatula chromosome 4, while the bottom maps to chromosome 3 (Table 1). These two M. truncatula assemblies therefore appear to be unlinked, even though they show substantial synteny and are apparently both orthologous (see below) to a contiguous region in soybean.
M. truncatula sequences in synteny block 2 also map to chromosomes 3 and 4. One of the M. truncatula homoeologs in synteny block 2, Mt_2ii, maps 5–8 cM below the M. truncatula bottom sequence assembly in synteny block 1b (Figure 1, 2, Table 1). The other M. truncatula duplicate in synteny block 2, Mt_2i, maps to chromosome 4, more than 25 cM from the top sequence assembly in synteny block 1b (Figure 1, 2, Table 1).
Synteny
Soybean/Medicago truncatula synteny within synteny block 1
The soybean and M. truncatula regions in synteny block 1b are highly syntenic, with nearly complete conservation of orientation and order of conserved genes (Table 2, Figure 1). Seventy-five percent of soybean genes in this region (33 of 44) have M. truncatula homologs and 59% of M. truncatula genes (33 of 56) have soybean homologs. When this comparison is revised to include only genes with significant matches to Genbank's nr database (thereby eliminating potentially poor gene calls) even more extensive levels of synteny emerge: 33 of 35 soybean genes with nr hits (94%) have M. truncatula homologs conserved in order and orientation. The two soybean genes without M. truncatula homologs include the rhg1 resistance locus itself, as well as one hypothetical protein. Both of these genes along with a soybean gene that extends beyond available M. truncatula sequence data all show synteny to A. thaliana. Therefore, all 36 confirmed soybean genes in this 336 kbp region have homologs in syntenic regions either of M. truncatula or A. thaliana.
Synteny block 1a shows lower, yet still impressive synteny. Nearly half of the genes in synteny block 1a are conserved in order and orientation. Indeed, 44% of M. truncatula genes are conserved in order and orientation in soybean in block 1a, increasing to 50% when only genes with database support are considered. Of soybean genes, 37% are conserved in M. truncatula, increasing to 43% of genes with database support.
Soybean/Medicago truncatula synteny in synteny block 2
Like block 1, extensive synteny is also evident throughout synteny block 2 (Figure 2, Table 2). In synteny block 2, there are two duplicated regions of M. truncatula syntenic to soybean, Mt_2i and Mt_2ii, which flank the soybean segment in Figure 2. The Mt_2i homoeolog and soybean share 60% (28 of 47) of their genes. With two exceptions, orientation is conserved between Mt_2i and soybean, and remarkably, a run of 13 out of 13 confirmed soybean genes are perfectly conserved in Mt_2i in the bottom portion of block 2. The corresponding soybean region extends nearly 110 kbp (Figure 2).
The other M. truncatula homoeolog, Mt_2ii, shows synteny with soybean extending more than 300 kbp (Figure 2, Table 2). In this region, soybean shares 32% (12 of 38) of genes and M. truncatula homoeolog Mt_2ii, 24% (12 of 50) in this syntenic region. One gene, with similarity to a rapid alkalinization factor in Solanum chacoense, shows synteny between soybean and Mt_2ii but appears to have been lost from Mt_2i (Figure 2, Gm gene 34, Mt_2ii gene 18). The middle portion of synteny block 2 exhibits multiple rearrangements and duplications between soybean and Mt_2ii (Figure 2). While much of the corresponding Mt_2i region has not yet been sequenced, it is less than half the size of the rearranged region in Mt_2ii/soybean on the basis of BAC-end sequenced clones that span the Mt_2i region.
The Mt_2i and Mt_2ii homoeologs themselves share nine genes, only one of which is absent from soybean (Figure 2). The gene absent from soybean encodes a putative AMP-binding protein (Figure 2, Mt_2i gene 14, Mt_2ii genes 9–10) present in one copy in Mt_2i and two adjacent copies in Mt_2ii. By contrast, Mt_2i and soybean share three times as many homologous pairs as the two M. truncatula duplicates themselves, including 19 homologous pairs that are absent from Mt_2ii. These observations help to illuminate the orthologous and paralogous relationships of these genome regions, which are described in further detail below.
Comparisons with A. thaliana
High levels of synteny are also maintained between the two legume species and networks of duplicated A. thaliana regions, each with a unique pattern of gene loss (Table 2). For example, nearly 62% (28 of 45) soybean genes and half of M. truncatula genes in synteny block 1b have a homolog within a syntenic network of four A. thaliana duplicated regions (Figure 1). With any one A. thaliana region, much lower levels of conserved synteny are observed (between 23% and 34% of soybean genes; 17% and 27% of M. truncatula genes). These results are consistent with the model of large-scale genome duplication followed by gene loss in Arabidopsis [34] and mirror the results of Foster-Hartnett et al. [29] in their low resolution synteny analysis of the soybean rhg1 region. By contrast, we found only one region in A. thaliana syntenic to block 1a (Figure 1, Table 2) with 20% (4 of 20) of soybean genes and 29% (5 of 17) of M. truncatula genes. In synteny block 2, levels of synteny between the legume species and individual A. thaliana regions were comparable to those in synteny blocks 1a and 1b, but composite syntenies were much lower (Table 2). For instance, just 29% of soybean genes were conserved in At4_2i and 17% in At3_2ii, with only 31% conserved in the network of both A. thaliana regions.
Perspectives on cyst nematode resistance genes
Despite the absence of an rhg1 homolog in the syntenic M. truncatula region examined here (synteny block 1b, Mt, top sequence assembly in Figure 1), it is clear that there is an rhg1 homolog elsewhere in M. truncatula. Though not present in any of the full-length BAC sequences that currently comprise >40% of M. truncatula's genespace [33], a homolog does exist on a M. truncatula BAC-end sequence (mth2-60m7), showing 78% amino acid identity to rhg1 over 279 amino acids (e-value = e-118). This percent identity is higher than that of the syntenic M. truncatula homologs compared to soybean's Rhg4 locus in synteny block 2 (~70%). BAC mth2-60m7 and the next three BAC-end hits all belong to the same region of M. truncatula FPC contig 949 [33], which tentatively maps to chromosome 5. This indicates that either the M. truncatula rhg1 homolog was translocated out of the remainder of the syntenic region on chromosome 4 or that contig 949 represents a paralogous region with the rhg1 homolog lost from the orthologous region that we examined here. Given the extensive synteny that exists in the region surrounding rhg1, it is surprising that the M. truncatula homolog of this gene, in particular, has undergone such rearrangement and/or loss. Notably, a homolog to rhg1 does exist among the network of A. thaliana duplicates (Figure 1, Synteny block 1b, At3_1iii, gene 5).
On the other hand, homologs of Rhg4 were found in the corresponding regions of M. truncatula and A. thaliana, including both M. truncatula homoeologs (Figure 2, Mt_2i gene 8, Mt_2ii gene 7, and At3_2ii gene 11). The M. truncatula Rhg4 homologs both show approximately 70% identity with Rhg4 and the surrounding region is greatly conserved as well. Mt_2i shares 75% and Mt_2ii 47%, of confirmed genes with soybean in the region immediately surrounding the Rhg4 homologs. Mt_2i even shows conservation of 100% of confirmed genes in a region over 300 kbp away. The high conservation of genome context surrounding Rhg4 indicates that, had the M. truncatula sequence been available before Rhg4 was cloned, it would have greatly facilitated cross genomic chromosome walking and cloning of the Rhg4 gene.
Tandem duplications
Genes have undergone tandem duplication in all species (soybean: 7 genes, M. truncatula: 9 genes, A. thaliana: 6 genes). In four cases, homologous soybean/M. truncatula genes are both duplicated. In no cases are homologous A. thaliana/legume genes both duplicated.
Tandemly duplicated genes with the highest copy numbers occur in a highly rearranged region in the middle of synteny block 2 (Figure 2). The rearranged region in soybean contains 11 copies of chalcone synthase genes in three separate groups of four, four, and three genes (genes 39–42, 50–51, 53–54, 57–59 in Figure 2). The latter group appears to have originated from a 25 kbp segmental duplication of the top CHS group and surrounding genes. While soybean has 11 copies of the CHS gene in this region, including CHS1, CHS2, CHS3, CHS4, and CHS5, the Mt_2ii region has only one CHS cluster with two genes, CHS1A and CHS1B (genes 52–53 in Figure 2). In addition, Mt_2ii contains a group of 10 genes with similarity to A. thaliana auxin-induced proteins 6B and X10A that are absent in soybean (genes 24–33 in Figure 2). It was not possible to analyze the corresponding region in Mt_2i, as this genome segment has not yet been sequenced.
Tandem duplications occur in other regions as well (Figures 1, 2). There are examples of tandemly duplicated genes whose homolog(s) are not duplicated, as well as cases in which two or more homologs have duplicated. For example, soybean and both M. truncatula duplicates in synteny block 2 have three copies of a glucosyltransferase (Mt_2i genes 2–4, Gm genes 18–20, and Mt_2ii genes 4–6 in Figure 2). Cases of differential tandem duplication may have resulted from duplication in only one species or loss of duplicates from one species.
Phylogenetic analysis
Phylogenetic trees were successfully generated for 21 gene families with members in synteny block 1 and for 23 in synteny block 2, many of which included homologs from both M. truncatula duplicates (Figures 1, 2) [see Additional Files 1, 2]. These phylogenies were examined to determine whether soybean and M. truncatula homologs within synteny blocks were more closely related to each other than to homologs elsewhere in the genomes, as represented by expressed sequences. For all 16 phylogenies in synteny block 1b, M. truncatula/soybean homologs were more closely related to each other than to homologs in other genomic regions, strongly suggesting orthology (Figure 1) [see Additional File 1]. Synteny block 1a contained a mix of tentative orthologs (two comparisons) and paralogs (three comparisons) (Figure 1) [see Additional File 1]. In synteny block 2, soybean genes showed orthologous relationships with their homologs in M. truncatula block Mt_2i every time (20 of 20 comparisons) and paralogous relationships in M. truncatula homoeolog Mt_2ii (11 of 11 comparisons) (Figure 2) [see Additional File 2].
Nucleotide substitution levels were determined to measure the evolutionary distance between soybean and M. truncatula (synonymous substitution levels) and to identify differences, if any, in selection pressure (nonsynonymous substitution levels). In synteny block 1, estimates of synonymous and nonsynonymous substitution levels were obtained for 34 sets of M. truncatula and soybean homologs (Figure 3a), six in synteny block 1a and 28 in synteny block 1b. In comparing these two blocks, we observed no difference in the number of synonymous substitutions per site (Table 3; 1a: 0.71, 1b: 0.71, p = 0.96), suggesting similar times of divergence between soybean and M. truncatula in both regions. This result is somewhat surprising given the fact that block 1a is composed of a mixture of apparent orthologous and paralogous relationships, while block 1b exhibits exclusively orthologous relationships.
In synteny block 2, the two M. truncatula homoeologs shared the same eight genes with soybean. We therefore focused on paired comparisons using these eight genes. The extent of synonymous substitutions between soybean and Mt_2i (0.87), an orthologous relationship based on phylogenetic tree analysis, was significantly lower than the extent between soybean and Mt_2ii (1.21), a paralogous relationship (p = 0.008) (Table 3; Figure 3b). All eight paired comparisons show higher levels of synonymous substitutions in the paralogous comparison. Not surprisingly, therefore, the paralogous region has evolved farther from soybean in evolutionary time than the orthologous region, implying that a duplication spanning the entire synteny block 2 preceded speciation between M. truncatula and soybean. The number of nonsynonymous substitution levels per site were comparable between the orthologous and paralogous M. truncatula/soybean relationships, with no significant difference (p = 0.69) (Table 3).
Comparisons of the distance between the two M. truncatula homoeologs in synteny block 2 revealed levels of synonymous substitutions (0.82) comparable to those of the orthologous Mt_2i/soybean comparison in the same block (0.87) (Table 3; Figure 3b; p = 0.85), suggesting that the duplication may have occurred close in time with speciation. It is surprising that the synonymous distance between M. truncatula homoeologs (0.82) is not closer to that of the paralogous M. truncatula/soybean comparison (1.21), which should be comparable given a duplication event followed by speciation. However, the difference between them was not significant (p = 0.22). There were no significant differences when comparing homoeologs between synteny blocks (data not shown) for either synonymous or nonsynonymous substitution levels. No differences were observed between tandemly duplicated and single copy genes for nonsynonymous or synonymous substitution levels (data not shown).
Estimates of synonymous substitution distance between orthologous soybean and M. truncatula regions allowed us to estimate the time of the Medicago/Glycine speciation event, as did the synonymous distance between the two M. truncatula duplicates in synteny block 2 in timing the underlying genome duplication. Orthologous regions between soybean and M. truncatula in both synteny blocks 1b (Mt/Gm) and 2 (Mt_2i/Gm) (Figures 1, 2) give similar estimates of divergence since speciation. Synteny block 1b has a median synonymous substitution level of 0.61 per site (Table 3), suggesting 50 My since the divergence through speciation, using an estimate of 6.1 × 10-9 substitutions per synonymous site per year [35]. Synteny block 2 has a median synonymous substitution level of 0.59 per site (Table 3) when comparing all orthologs, suggesting 48 Mya since speciation. By contrast, the duplication event in M. truncatula evident in synteny block 2 (Figure 2) appears to have predated speciation, with a median of 0.79 synonymous substitutions per site and an inferred divergence between duplicates of 64 Mya.
Discussion
Synteny
Soybean and M. truncatula synteny
We examined two regions that are highly syntenic between soybean, M. truncatula, and A. thaliana. These two regions comprise approximately 0.5 Mb each surrounding the rhg1 and Rhg4 SCN resistance loci of soybean and their corresponding regions in M. truncatula and A. thaliana. In the process, we discovered remarkably high levels of colinearity between soybean and M. truncatula, including cases of near perfect conservation of gene order and orientation. For example, we observed one case where 33 of 35 genes and a second where 13 out of 13 genes were perfectly conserved and colinear. Because M. truncatula and soybean are estimated to have diverged approximately 50 Mya, these examples of conserved synteny are truly remarkable. Overall, soybean synteny to M. truncatula in orthologous relationships averages 79% and reaches 94% when only genes with Genbank's nonredundant database support are considered. Moreover, synteny between the legume species and a network of A. thaliana segmental duplications exceeds 60%.
Levels of synteny are clearly not this high genome-wide, though macrosynteny exists over much of the genome [6]. Yan et al. [16] estimated that synteny between soybean and M. truncatula exists in only about half of soybean genomic regions anchored by RFLPs, and of these cases, just 75% exhibit extensive synteny. In other genome regions not described in here, we have found much lower levels of synteny. In a targeted search for syntenic relationships between soybean and M. truncatula in genome regions surrounding the soybean disease resistance gene, rpg1, Cannon et al. [36] discovered much lower levels of synteny. The best syntenic candidate regions that could be identified showed significant differential gene expansion, multiple rearrangements, indels, and translocations. In this case, the rpg1 soybean region on linkage group F of approximately 300 kbp corresponded to an M. truncatula region of more than 500 kbp. Of 30 soybean genes and 45 M. truncatula genes confirmed by hits to the Uniref database [37] at BLASTP, e ≤ -4, just nine (20%) were syntenic.
Why the regions that we examined in the current study should have remained so highly conserved is unknown. Several explanations for differential conservation of synteny have been proposed. Regions with disease resistance genes often evolve rapidly and show frequent rearrangements [38-40]. Nevertheless, the soybean regions in this study were characterized because they contain important disease resistance genes – though not members of the more widespread NBS-LRR gene family most frequently associated with rearrangements [38]. Often, regions near centromeres tend to be more conserved than telomeric regions [41]. But the soybean region in synteny block 1 is known to be located at the very end of the chromosome, while still retaining high levels of synteny with M. truncatula and A. thaliana. Of the mapped M. truncatula sequence assemblies, only the top sequence assembly in synteny block 1b is close to the centromere (R. Geurts et al., personal communication). In some organisms, regions of housekeeping genes are clustered and thought to be more conserved [42-44]. Though housekeeping genes are certainly present in the regions of this study, the regions do not represent clusters of housekeeping genes [see Additional File 3]. Finally, the presence of transposons and other repetitive sequence may decrease stability in a region [45-47]. In the regions described here, just 12 of 356 predicted gene models in soybean and M. truncatula are transposons or retrotransposons. On a base pair level, this translates to 2.2% of M. truncatula and 1.7% of soybean genomic sequence. These levels of transposons and retrotransposons are not unusual. Cannon et al. (personal communication) estimated that transposons of this type occupy ~3% of the M. truncatula genome when examining over 100 Mb of sequence. In soybean, Graham et al. [48] found that more than 10% of a soybean BAC containing multiple NBS-LRR sequences was composed of retrotransposons. Although we cannot identify hallmarks of the sequences examined that would cause them to be more conserved than usual, they do appear to be highly conserved and care should be taken in drawing general conclusions from this comparison.
Legume and A. thaliana synteny
The relationship between legumes and A. thaliana in synteny block 1b, described in detail here, seem to follow a pattern of post-speciation duplication followed by gene loss [34]. Several syntenic regions exist that vary slightly in overall degree of synteny, while gene loss/insertion has occurred in every case. A composite of all the partially syntenic regions come together to form a network that together recapitulates substantial genome conservation. The retention of soybean and M. truncatula genes in A. thaliana is impressive, given the roughly 90 My thought to separate A. thaliana from legumes [49]. For example, synteny block 1b has 60% of its genes occurring in at least one of four syntenic A. thaliana regions, though individually, the most conserved of these regions contains just half that level.
Though synteny between legumes and A. thaliana in this region is impressive, previous results suggest it extends beyond the region analyzed here. Foster-Hartnett et al [29] described conserved synteny involving the genome region around rhg1 twice the size examined in the present study, though at low sequence resolution primarily using BAC-end sequences [29]. Simillion, et al. [50] also found conservation among the A. thaliana regions syntenic to synteny block 1b extending up to 182 kbp in length.
Exceptions to synteny
There were also some consistent exceptions to synteny. A total of 13 transposons, including retrotransposons, were found among the three species. Just one was found in syntenic positions, where an M. truncatula retroelement in synteny block 1b was located in a comparable position to an element in soybean. However, the soybean (a copia-type polyprotein) and M. truncatula (an RNA-directed DNA polymerase) genes do not share sequence homology.
Additionally, genes predicted by FGENESH but without database support were much less likely to have homologs than those with database confirmation (9% of soybean predicted genes without database support versus 62% of those with database support have homologs). Still, a small number – roughly 10% – of nearly 60 unconfirmed genes (no hits in nr) in either soybean or M. truncatula did show synteny. Conserved legume genes without homologs in the database may be the most interesting genes of all, since they are likely to be novel or highly diverged from known proteins and may play a role in important plant or legume-specific processes.
Gene density
Although soybean's genome size is more than double that of M. truncatula [51], gene density is comparable in these regions (Synteny block 1: M. truncatula = 1 gene/7.2 kb, soybean: 1 gene/6.7 kb; Synteny block 2: M. truncatula = 1 gene/6.2 kb, soybean: 1 gene/5.8 kb). These values are not far from those of A. thaliana overall (1 gene/5 kb) [52]. In this region, A. thaliana gene density is approximately half to two-thirds that of the legumes (Synteny block 1: A. thaliana = 1 gene/3.4 kb; Synteny block 2: A. thaliana = 1 gene/4.1 kb), although soybean and M. truncatula genome sizes are 7X and 3X that of A. thaliana, respectively [51]. Higher than expected gene densities in soybean and M. truncatula suggest the possibility of gene clustering. Indeed, gene clustering has identified in M. truncatula [53] and forms the basis of targeting M. truncatula sequencing to the gene-rich regions of the genome [14].
Genome duplication and speciation
The networks of synteny we have identified reflect duplication and gene loss between species, including M. truncatula regions with both orthologous and paralogous relationships to soybean. In synteny block 2, for example, there are clear-cut examples of regions with orthologous and paralogous relationships. Mt_2i shows only orthologous relationships with soybean, while Mt_2ii shows only paralogous relationships (Figures 1, 2). M. truncatula regions in synteny block 1b also unambiguously display orthology to soybean.
M. truncatula duplications in synteny block 2 allowed us to systematically examine corresponding orthologous and paralogous relationships to soybean. The percentage of conserved genes between soybean and Mt_2i (orthologous region) was twice as high as with Mt_2ii (paralogous region) (Figure 2, Table 2). Given that orthology indicates the most closely related regions evolutionarily (reflected in the phylogenetic trees) [see Additional File 2], it is not surprising that fewer genes have been deleted/inserted or experienced substitutions in orthologous comparisons. The fact that all the M. truncatula genes in the orthologous region (Mt_2i) are more closely related to soybean as evidenced by phylogenetic trees, synonymous substitution levels, percent identity, and extent of synteny, than either is to the M. truncatula genes in the paralogous region (Mt_2ii) suggests that the duplication seen in M. truncatula occurred before the speciation event splitting Medicago and Glycine lineages. Presumably, soybean also has (or had) a duplicate region as well, a possibility with some phylogenetic support [see Additional File 2].
We date the duplication event, possibly as a part of a genome duplication, in M. truncatula at 64 Mya, preceding a speciation event approximately 48–50 Mya. Indeed, the possibility of a genome duplication event predating the split between M. truncatula and soybean has been suggested previously [16,54,55]. Median synonymous substitution levels between the two M. truncatula duplicates in synteny block 2 (0.79 synonymous substitutions per site) fall within [55] or near [54] synonymous distance peaks, which were interpreted by the authors as a genome duplication event in M. truncatula. Schleuter et al. [55] estimates that this event occurred 58 million years ago, while Blanc and Wolfe [54] inferred a more recent event based on a substantially different molecular clock [56]. Likewise, we estimate the speciation event between Medicago and Glycine at 48 – 50 Mya, while Blanc and Wolfe [54] inferred a much more recent date of 13.3–15 million years ago, though again, the differences are primarily due to the use of differing molecular clocks [56].
Comparatively long (~500 kbp) and contiguous sets of homologous segments from different species with known phylogenetic relationships and nucleotide substitution levels bring power to the study of molecular evolution. Though median synonymous substitution levels of duplication and speciation events correspond well to published values [54,55] (see above), the extent of synonymous substitutions varies significantly between neighboring genes despite a common genomic context (Figure 3). Estimates comparing the two M. truncatula segments created by a duplication event range from 0.62–1.12 while those comparing soybean and M. truncatula orthologs (speciation event) range from 0.42–2.68. Since the duplicates in each one of these cases presumably diverged at the same moment, one must postulate different evolutionary trajectories for the different gene lineages. Knowing that all the genes on a contiguous genomic block duplicated (and later speciated) together removes an important unknown from evolution analyses in contrast to comparable EST-based studies [54,55].
Conclusion
We analyzed genome regions of soybean, M. truncatula, and A. thaliana with remarkable levels of conservation of gene content and order. Such high levels of colinearity within the legumes and with the model plant A. thaliana bode well for leveraging information from model genomes to crop plants like soybean. Further, we described substantial blocks of genes with the same evolutionary (duplication) history, allowing us to study and compare the individual evolution of genes within a common genomic context. These blocks include two duplicates in M. truncatula, one orthologous and the other paralogous with soybean. This duplication may be part of a larger genome duplication event in the common ancestor of soybean and M. truncatula. If so, the analysis described here is just the first step in understanding the evolution of legume genomes and a useful addition to our knowledge about genomic reorganization that occurs at a the scale of megabase or less.
Methods
Sequences
Glycine max sequences [GenBank:AX196294.1, GenBank:AX196295.1, GenBank:AX196297.1, and GenBank:AX197417.1] [32] were obtained from Genbank. All soybean sequences were derived from cultivar 'A3244'. [GenBank:AX196295.1] (Figure 1, gene 29) includes the susceptible allele of soybean rhg1 gene on molecular linkage group G. [GenBank:AX197297] (Figure 2, gene 21) includes the susceptible allele of soybean Rhg4 gene on molecular linkage group A2.
M. truncatula BACs were sequenced as part of an international effort to sequence the genespace of this model legume [14]. Two additional M. truncatula BACs were sequenced and examined before the international genome sequencing had begun [57]. Putative homologs of soybean sequences in M. truncatula and A. thaliana were identified by searching the soybean sequences against all sequenced M. truncatula BACs and the A. thaliana proteome using BLAST [58] (The Institute for Genomic Research, Arabidopsis Proteome version 5). After identifying genes (see below), protein/protein comparisons (BLASTP) were performed in order to confirm that BACs were syntenic and to identify syntenic genes (see below). Genbank accessions for soybean and M. truncatula sequences, A. thaliana gene numbers, and mapping information are shown in Table 1.
Sequence assemblies
Sequences were aligned and merged in regions of sequence overlap on the basis of 99% identity or better. End-sequenced BAC clones that tentatively spanned gaps in the sequence were identified based on strong hits (e-value = 0, ≥99% identity) to sequenced BACs on either side. Gap sizes were estimated by removing overlap from the estimated size of end-sequenced BAC(s).
Nomenclature
Throughout the manuscript, the following nomenclature is used. Regions surrounding and syntenic to the SCN resistance rhg1 locus are collectively referred to as synteny block 1 (Figure 1). Regions surrounding and syntenic to SCN Rhg4 gene are collectively referred to as synteny block 2 (Figure 2). Synteny block 1 is divided into blocks 1a and 1b, which are separated by gaps in all three species (Figure 1). Within each synteny block, species are labeled as Gm (soybean), Mt (M. truncatula), or At (A. thaliana). The chromosome number follows the "At" abbreviation for A. thaliana. If more than one homoeolog is present, the species abbreviation is appended with an underscore followed by the synteny block and lower case roman numerals (i.e. Mt_2i, Mt_2ii, At4_2i, and At3_2ii in Figure 2) (Figures 1, 2). Sequence assemblies separated by physical gaps are labeled as sequence assemblies "top" and "bottom" in arbitrary order (Figure 1).
Gene prediction and identification of synteny
Genes were predicted in G. max and M. truncatula genomic sequences using the dicot (Arabidopsis) matrix of FGENESH [59,60]. BLASTP was used to compare predicted proteins between databases containing these G. max or M. truncatula predicted genes and all A. thaliana proteins with an e-value cutoff of e-8 and percent identity cutoff of 40% for the top high scoring segment pair for soybean and M. truncatula comparisons and an e-value cutoff of e-8 for comparisons to A. thaliana [58]. These cutoff values generally identified homologs in syntenic positions while rejecting related genes in nonsyntenic positions.
In this study, we defined synteny to include both conservation of gene content and order between species. In estimating syntenic density (the percentage of genes conserved between two species), repetitive sequences (genes with similarity to transposable elements, including retroelements) were not included and tandemly duplicated genes were counted as one. Synteny between two species was estimated from the first to the last pair of conserved genes in the available sequence for both species.
Phylogenetic analysis
To distinguish between orthologous and paralogous regions, we constructed phylogenetic trees as follows. BLASTP or TBLASTN, as appropriate, were used to compare all G. max genes with the following sequences: all G. max and M. truncatula proteins in the corresponding genomic regions of this analysis; the nonredundant A. thaliana proteome; soybean and M. truncatula EST unigene sets [61] (GMGI v.11 and MTGI v.7; The Institute for Genomic Research. Rockville, MD). The top 25 hits ≥100 amino acids with e-values ≤ e-10 were included in the analysis. Tandem duplications and highly related genes in the same gene family were grouped for analysis.
Initial alignments were calculated using T-COFFEE [62] with manual evaluations and edits in Jalview [63] for poorly aligning sequences. For subsequent phylogenetic analysis, an HMM calculated for each alignment using hmmer [64] was used to realign sequences and to identify and remove indel regions and sequences with fewer than 60% matches to the model. Parameters for hmmbuild were: archpri = 0.7, gapmax = 0.3.
Parsimony trees were calculated using the protpars of Phylip [65], with maximum likelihood branch lengths calculated using TREE-PUZZLE [66]. Parameters for protpars were: randomize input order; use ordinary parsimony; search for best tree; select one best tree for further analysis in TreePuzzle. Parameters for TreePuzzle were: user defined tree (from parsimony search); approximate parameter estimates; Whelan-Goldman substitution model [67] estimate amino acid frequencies from data set; allow rate heterogeneity with eight gamma-distributed rates.
Nucleotide substitutions
Codon-aligned nucleic acid sequences were created with TranslateAlign.pl (courtesy Dan Kortschak, University of Adelaide, Adelaide, Australia). Nucleotide substitutions levels were calculated using these alignments with SNAP (Synonymous/Non-synonymous Analysis Program) [68,69]. In this program, the levels of synonymous and nonsynonymous substitutions per site are approximated using methods developed by Nei and Gojobori [70], incorporating Ota and Nei's statistic [71]. Median synonymous substitution levels were converted into estimates of time since divergence using an estimate of 6.1 × 10-9 substitutions per synonymous site per year [35].
Abbreviations
Megabases (Mb); kilobase pairs (kbp); million years (My); million years ago (Mya); soybean cyst nematode (SCN); restriction fragment length polymorphism (RFLP); Medicago truncatula (Mt); Glycine max (Gm); Arabidopsis thaliana (At)
Authors' contributions
JM carried out the synteny, phylogenetic, and nucleotide substitution analyses and drafted the manuscript. SBC participated in the design of the study, the phylogenetic analysis, and the interpretation of the data. PK and GEDO participated in the sequencing, including identification of BACs in the area of interest. BAR participated in the sequencing. CDT participated in the interpretation of data and writing of the manuscript. NDY participated in the design, analysis, and interpretation of the data as well as in the writing of the manuscript. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Phylogenetic trees of genes in synteny block 1. For soybean and M. truncatula, genes are labeled with their homoeolog name and gene number as in Figure 1. A. thaliana genes are labeled by their standard gene numbers. Other genes represent expressed sequences. Scales are in PAM units.
Click here for file
Additional File 2
Phylogenetic trees of genes in synteny block 2. For soybean and M. truncatula, genes are labeled with their homoeolog name and gene number as in Figure 2. A. thaliana genes are labeled by their standard gene numbers. Other genes represent expressed sequences. Scales are in PAM units.
Click here for file
Additional File 3
Gene Annotation Table
Click here for file
Acknowledgements
We thank A. Abell for assistance with computational analysis. We thank D. Foster-Hartnett for helpful discussions. This research was supported by National Science Foundation grant DBI 0321460 and the Samuel Roberts Noble Foundation. This paper is published as part of the series of the Minnesota Agricultural Experiment Station.
Figures and Tables
Figure 1 Synteny block 1. A syntenic block of soybean, M. truncatula and A. thaliana genes surrounding soybean's rhg1 gene (Gm gene 21). Solid black lines connect homologs. Dotted black lines indicate that the absence of a homolog in the syntenic position. Blue lines connect orthologs. Pink lines connect paralogs. M. truncatula genes are shown in red, soybean in brown, and A. thaliana in blue. Lighter colored genes represent those that had no significant similarity to Genbank's nonredundant protein database. Gray genes are repetitive elements. A thick gray vertical line connecting sequence assemblies indicate regions in which sequence is not yet available but in which linkage and approximate distance were determined. Genbank accessions are shown in gray.
Figure 2 Synteny block 2. A syntenic block of soybean, M. truncatula and A. thaliana sequence assemblies surrounding soybean's Rhg4 gene (Gm gene 29). Solid black lines connect homologs. Blue lines connect orthologs. Pink lines connect paralogs. Dotted black lines indicate that the absence of a homolog in the syntenic position. M. truncatula genes are shown in red, soybean in brown, and A. thaliana in blue. Lighter colored genes represent those that had no significant similarity to Genbank's nonredundant protein database. Gray genes are repetitive elements. A thick gray line connecting sequence assemblies indicate regions in which sequence is not yet available but in which physical linkage and approximate physical distance were determined. Numbers along the sequence assemblies indicate gene numbers. Genbank accessions are shown in gray. Green boxes identify a 25 kbp duplication.
Figure 3 Histograms of synonymous distance between soybean and M. truncatula homologs or M. truncatula homoeologs in a) synteny block 1 and b) synteny block 2. For synteny block 2, only the 8 genes with homologs in soybean and both M. truncatula duplicates are shown.
Table 1 Sequence accession, contig, and map positions
Synteny Block Species Homoeolog Sequence assembly Genbank accession or Arabidopsis gene numbers1 Orientation Contig2 length (kbp) chromosome Position3 Anchoring marker
1a M. truncatula Mt AC141115.22 + 246 126
1a Soybean Gm AX196294.1 - 127 G 5.8 B053_14
1a A. thaliana At4 + 30 4 18.4
1b M. truncatula Mt top AC149303.10 - 945 141 4 32.5 DK379L5,7
1b M. truncatula Mt bottom CR378662.1a + 1072 119 3 59
1b M. truncatula Mt bottom CR937029.1a - 1072 112 3 59 h2_105b15c, h2_6m1c6
1b M. truncatula Mt bottom AC142498.20 - 1072 114 3 59
1b Soybean Gm AX196295.1 + 336 G 3.9–4.5 Sat_168, Satt3094
1b A. thaliana At2_1i At2g40130-At2g40400 - 107 2 16.8
1b A. thaliana At5_1ii At5g05350-At5g05600 + 89 5 1.6
1b A. thaliana At3_1iii At3g55970-At3g56140 - 66 3 20.8
1b A. thaliana At3_1iv At3g10980-At3g11180 + 69 3 3.4
2 M. truncatula Mt_2i AC146585.18b - 273 125 4 58.7
2 M. truncatula Mt_2i AY224188.1b + 99 4 58.7
2 M. truncatula Mt_2i AC145021.13c - 114 120 4 58.7 EST9485
2 M. truncatula Mt_2i AC130798.14c + 114 92 4 58.7
2 Soybean AX196297.1d + 350 A2 48.8 I4
2 Soybean AX197417.1d + 214 A2 48.8 I4
2 M. truncatula Mt_2ii AC146706.8e + 1132 107 3 64.2 h2_108g5a6
2 M. truncatula Mt_2ii AY224189.1e - 68 3 67.6 AY_224189_a6
2 M. truncatula Mt_2ii AC146705.11e + 1132 120 3 63.5 h2_101f3d6
2 M. truncatula Mt_2ii AC146683.9e - 1132 131 3 63.5
2 A. thaliana At4_2i At4g13600-At4g14200 - 278 4 7.9
2 A. thaliana At3_2ii At3g23670-At3g24000 + 156 3 8.5
1 Letters denote groups of sequence accessions that overlap
2 Physical contig number [33]
3 Measured in cM for soybean and M. truncatula and in Mb for A. thaliana
4 [72]
5 [33]
6 Denny et al., unpublished
7 Marker is from BAC Mth2-31m16, which is adjacent in the contig but not overlapping
Table 2 Synteny
Synteny Block Reference species Homoeolog Syntenic Species Homoeolog Synteny (% genes) Synteny (# of genes) Total Genes1 Synteny (% confirmed genes) Synteny (# of confirmed genes) Total Confirmed Genes1
1a Soybean M. truncatula 37% 7 19 43% 6 14
1a M. truncatula Soybean 44% 7 16 50% 6 12
1a Soybean A. thaliana 20% 4 20
1a M. truncatula A. thaliana 29% 5 17
1a A. thaliana Soybean 50% 4 8
1a A. thaliana M. truncatula 50% 5 10
1b Soybean M. truncatula 75% 33 44 94% 33 35
1b M. truncatula Soybean 59% 33 56 79% 33 42
1b Soybean A. thaliana Composite At2_1i-At3_1iv 62% 28 45
At2_1i 34% 14 41
At5_1ii 25% 11 44
At3_1iii 25% 10 40
At3_1iv 23% 7 30
1b M. truncatula A. thaliana Composite At2_1i-At3_1iv 50% 28 56
At2_1i 27% 13 48
At5_1ii 21% 12 56
At3_1iii 17% 9 52
At3_1iv 24% 8 33
2 Soybean M. truncatula Mt_2i 60% 28 47 66% 27 41
2 M. truncatula Mt_2i Soybean 60% 28 47 72% 26 36
2 Soybean M. truncatula Mt_2ii 32% 12 38 33% 12 36
2 M. truncatula Mt_2ii Soybean 24% 12 50 26% 12 46
2 Soybean M. truncatula Composite Mt_2i, Mt_2ii 66% 31 47 73% 30 41
2 Soybean A. thaliana Composite 31% 20 64
2 Soybean A. thaliana At4_2i 29% 16 56
2 Soybean A. thaliana At3_2ii2 17% 11 64
2 M. truncatula Mt_2i A. thaliana Composite At4_2i, At3_2ii 26% 12 47
2 M. truncatula Mt_2i A. thaliana At4_2i 23% 11 47
2 M. truncatula Mt_2i A. thaliana At3_2ii 27% 6 22
2 M. truncatula Mt_2ii A. thaliana Composite At4_2i, At3_2ii 37% 15 41
2 M. truncatula Mt_2ii A. thaliana At4_2i 32% 13 41
2 M. truncatula Mt_2ii A. thaliana At3_2ii 22% 8 36
1 Calculated from first to last syntenic gene in region
2 If the last syntenic pair is removed, 10 of 35 genes are syntenic (29%)
Table 3 Extent of Nucleotide Substitution between Soybean and M. truncatula (Mean/Median)
Synteny Block Homoeologs Nonsynonymous/Synonymous1,2 Synonymous substitutions1,3 Nonsynonymous substitutions1,4
1 Mt and Gm a 0.42/0.27 a 0.71/0.73 a 0.19/0.23
1 Mt and Gm a 0.21/0.18 a 0.71/0.61 a 0.13/0.12
25 Mt_2i and Gm a,b 0.26/0.25 a 0.87/0.63 a 0.16/0.12
25 Mt_2ii and Gm a 0.17/0.16 b 1.21/0.98 a 0.17/0.15
25 Mt_2i and Mt_2ii b 0.24/0.24 a,b 0.82/0.79 b 0.21/0.21
1 Letters denote significant difference at p ≤ 0.05
2 Ratio of synonymous to nonsynonymous substitutions
3 Synonymous substitutions per site
4 Nonsynonymous substitutions per site
5 Synteny block 2 includes only the eight genes in soybean and both M. truncatula duplicates
==== Refs
Mouse Genome Sequencing Consortium Initial sequencing and comparative analysis of the mouse genome Nature 2002 420 520 562 12466850 10.1038/nature01262
Hedges SB The origin and evolution of model organisms Nat Rev Genet 2002 3 838 849 12415314 10.1038/nrg929
Zhao S Shetty J Hou L Delcher A Zhu B Osoegawa K de Jong P Nierman WC Strausberg RL Fraser CM Human, mouse, and rat genome large-scale rearrangements: Stability versus speciation Genome Res 2004 14 1851 1860 15364903 10.1101/gr.2663304
International Chicken Genome Sequencing Consortium Sequence and comparative analysis of the chicken genome provide unique perspectives on vertebrate evolution Nature 2004 432 695 716 15592404 10.1038/nature03154
Thomas JW Touchman JW Blakesley RW Bouffard GG Beckstrom-Sternberg SM Margulies EH Blanchette M Siepel AC Thomas PJ McDowell JC Maskeri B Hansen NF Schwartz MS Weber RJ Kent WJ Karolchik D Bruen TC Bevan R Cutler DJ Schwartz S Elnitski L Idol JR Prasad AB Lee-Lin SQ Maduro VV Summers TJ Portnoy ME Dietrich NL Akhter N Ayele K Benjamin B Cariaga K Brinkley CP Brooks SY Granite S Guan X Gupta J Haghighi P Ho SL Huang MC Karlins E Laric PL Legaspi R Lim MJ Maduro QL Masiello CA Mastrian SD McCloskey JC Pearson R Stantripop S Tiongson EE Tran JT Tsurgeon C Vogt JL Walker MA Wetherby KD Wiggins LS Young AC Zhang LH Osoegawa K Zhu B Zhao B Shu CL De Jong PJ Lawrence CE Smit AF Chakravarti A Haussler D Green P Miller W Green ED Comparative analyses of multi-species sequences from targeted genomic regions Nature 2003 424 788 793 12917688 10.1038/nature01858
Choi HK Mun JH Kim DJ Zhu H Baek JM Mudge J Roe B Ellis N Doyle J Kiss GB Young ND Cook DR Estimating genome conservation between crop and model legume species Proc Natl Acad Sci U S A 2004 101 15289 15294 15489274 10.1073/pnas.0402251101
Moore G Devos KM Wang Z Gale MD Cereal genome evolution. grasses, line up and form a circle Curr Biol 1995 5 737 739 7583118 10.1016/S0960-9822(95)00148-5
Tanksley SD Ganal MW Prince JP de Vicente MC Bonierbale MW Broun P Fulton TM Giovannoni JJ Grandillo S Martin GB High density molecular linkage maps of the tomato and potato genomes Genetics 1992 132 1141 1160 1360934
Salse J Piegu B Cooke R Delseny M New in silico insight into the synteny between rice (oryza sativa L.) and maize (zea mays L.) highlights reshuffling and identifies new duplications in the rice genome Plant J 2004 38 396 409 15086801 10.1111/j.1365-313X.2004.02058.x
Goff SA Ricke D Lan TH Presting G Wang R Dunn M Glazebrook J Sessions A Oeller P Varma H Hadley D Hutchison D Martin C Katagiri F Lange BM Moughamer T Xia Y Budworth P Zhong J Miguel T Paszkowski U Zhang S Colbert M Sun WL Chen L Cooper B Park S Wood TC Mao L Quail P Wing R Dean R Yu Y Zharkikh A Shen R Sahasrabudhe S Thomas A Cannings R Gutin A Pruss D Reid J Tavtigian S Mitchell J Eldredge G Scholl T Miller RM Bhatnagar S Adey N Rubano T Tusneem N Robinson R Feldhaus J Macalma T Oliphant A Briggs S A draft sequence of the rice genome (Oryza sativa L. ssp. japonica) Science 2002 296 92 100 11935018 10.1126/science.1068275
Menanciohautea D Fatokun CA Kumar L Danesh D Young ND Comparative genome analysis of mungbean (Vigna-radiata l-wilczek) and cowpea (V-unguiculatal walpers) using RFLP mapping data Theoretical & Applied Genetics 1993 86 797 810 10.1007/BF00212605
Boutin SR Young ND Olson TC Yu ZH Shoemaker RC Vallejos CE Genome conservation among three legume genera detected with DNA markers Genome 1995 38 928 937
Lee JM Grant D Vallejos CE Shoemaker RC Genome organization in dicots. II. Arabidopsis as a 'bridging species' to resolve genome evolution events among legumes Theoretical & Applied Genetics 2001 103 765 773 10.1007/s001220100658
Young NevinD Cannon StevenB Shusei Sato Dongjin Kim Cook DouglasR Town ChrisD Roe BruceA Satoshi Tabata Sequencing the genespaces of Medicago truncatula and Lotus japonicus Plant Physiology 2005 137 1174 1181 15824279 10.1104/pp.104.057034
Yan HH Mudge J Kim DJ Shoemaker RC Cook DR Young ND Comparative physical mapping reveals features of microsynteny between Glycine max, Medicago truncatula, and Arabidopsis thaliana Genome 2004 47 141 155 15060611 10.1139/g03-106
Yan HH Mudge J Kim DJ Larsen D Shoemaker RC Cook DR Young ND Estimates of conserved microsynteny among the genomes of Glycine max, Medicago truncatula and Arabidopsis thaliana Theor Appl Genet 2003 106 1256 1265 12748777
Concibido VC Diers BW Arelli PR A decade of QTL mapping for cyst nematode resistance in soybean Crop Science 2004 44 1121 1131
Concibido VC Denny RL Boutin SR Hautea R Orf JH Young ND DNA marker analysis of loci underlying resistance to soybean cyst nematode (Heterodera glycines ichinohe) Crop Science 1994 34 240 246
Concibido VC Denny RL Lange DA Orf JH Young ND RFLP mapping and marker-assisted selection of soybean cyst nematode resistance in PI 209332 Crop Science 1996 36 1643 1650
Concibido VC Young ND Lange DA Denny RL Danesh D Orf JH Targeted comparative genome analysis and qualitative mapping of a major partial-resistance gene to the soybean cyst nematode Theoretical & Applied Genetics 1996 93 234 241 10.1007/s001220050271
Concibido VC Lange DA Denny RL Orf JH Young ND Genome mapping of soybean cyst nematode resistance genes in Peking, PI 90763, and PI 88788 using DNA markers Crop Science 1997 37 258 264
Meksem K Doubler TW Chancharoenchai K Njiti VN Chang SJC Arelli APR Cregan PE Gray LE Gibson PT Lightfoot DA Clustering among loci underlying soybean resistance to Fusarium solani, SDS and SCN in near-isogenic lines Theoretical & Applied Genetics 1999 99 1131 1142 10.1007/s001220051317
Meksem K Pantazopoulos P Njiti VN Hyten LD Arelli PR Lightfoot DA 'Forrest' resistance to the soybean cyst nematode is bigenic: Saturation mapping of the Rhg1 and Rhg4 loci Theoretical & Applied Genetics 2001 103 710 717 10.1007/s001220100597
Webb DM Baltazar BM Raoarelli AP Schupp J Clayton K Keim P Beavis WD Genetic mapping of soybean cyst nematode race-3 resistance loci in the soybean PI 437.654 Theoretical & Applied Genetics 1995 91 574 581
Cregan PB Mudge J Fickus EW Danesh D Denny R Young ND Two simple sequence repeat markers to select for soybean cyst nematode resistance conditioned by the rhg1 locus Theoretical & Applied Genetics 1999 99 811 818 10.1007/s001220051300
Meksem K Zobrist K Ruben E Hyten D Quanzhou T Zhang HB Lightfoot DA Two large-insert soybean genomic libraries constructed in a binary vector: Applications in chromosome walking and genome wide physical mapping Theoretical & Applied Genetics 2000 101 747 755 10.1007/s001220051540
Meksem K Ruben E Hyten DL Schmidt ME Lightfoot DA High-throughput genotyping for a polymorphism linked to soybean cyst nematode resistance gene Rhg4 by using TaqMan (TM) probes Molecular Breeding 2001 7 63 71 10.1023/A:1009610009663
Meksem K Ruben E Hyten D Triwitayakorn K Lightfoot DA Conversion of AFLP bands into high-throughput DNA markers Mol Genet Genomics 2001 265 207 214 11361330 10.1007/s004380000418
Foster-Hartnett D Mudge J Larsen D Danesh D Yan H Denny R Penuela S Young ND Comparative genomic analysis of sequences sampled from a small region on soybean (Glycine max) molecular linkage group G Genome 2002 45 634 645 12175066 10.1139/g02-027
Danesh D Penuela S Mudge J Denny RL Nordstrom H Martinez JP Young ND A bacterial artificial chromosome library for soybean and identification of clones near a major cyst nematode resistance gene Theoretical & Applied Genetics 1998 96 196 202 10.1007/s001220050727
Clough SJ Tuteja JH Li M Marek LF Shoemaker RC Vodkin LO Features of a 103-kb gene-rich region in soybean include an inverted perfect repeat cluster of CHS genes comprising the I locus Genome 2004 47 819 831 15499396 10.1139/g04-049
Hauge BM Wang ML Parsons JD Parnell LD Nucleic acid molecules and other molecules associated with soybean cyst nematode resistance Patent: WO 0151627-A 2 19-JUL-2001
Medicago truncatula sequencing resources
Ku HM Vision T Liu J Tanksley SD Comparing sequenced segments of the tomato and arabidopsis genomes: Large-scale duplication followed by selective gene loss creates a network of synteny Proc Natl Acad Sci U S A 2000 97 9121 9126 10908680 10.1073/pnas.160271297
Lynch M Conery JS The evolutionary fate and consequences of duplicate genes Science 2000 290 1151 1155 11073452 10.1126/science.290.5494.1151
Cannon S Makarevich G Savage E Denny R Mudge J Seigfried M Lai H Ashfield T Roe BA Young ND Innes R A phylogenetic and structural comparison of homologous Rpg1 R-gene-containing regions in soybean and Medicago truncatula Plant and Animal Genomes XIII: San Diego, CA 2005 P437
Apweiler R Bairoch A Wu CH Barker WC Boeckmann B Ferro S Gasteiger E Huang H Lopez R Magrane M Martin MJ Natale DA O'Donovan C Redaschi N Yeh LS UniProt: The universal protein knowledgebase Nucleic Acids Res 2004 D115 9 14681372 10.1093/nar/gkh131
Baumgarten A Cannon S Spangler R May G Genome-level evolution of resistance genes in Arabidopsis thaliana Genetics 2003 165 309 319 14504238
Leister D Kurth J Laurie DA Yano M Sasaki T Devos K Graner A Schulze-Lefert P Rapid reorganization of resistance gene homologues in cereal genomes Proc Natl Acad Sci U S A 1998 95 370 375 9419382 10.1073/pnas.95.1.370
Michelmore RW Meyers BC Clusters of resistance genes in plants evolve by divergent selection and a birth-and-death process Genome Res 1998 8 1113 1130 9847076
Akhunov ED Akhunova AR Linkiewicz AM Dubcovsky J Hummel D Lazo G Chao S Anderson OD David J Qi L Echalier B Gill BS Miftahudin Gustafson JP La Rota M Sorrells ME Zhang D Nguyen HT Kalavacharla V Hossain K Kianian SF Peng J Lapitan NL Wennerlind EJ Nduati V Anderson JA Sidhu D Gill KS McGuire PE Qualset CO Dvorak J Synteny perturbations between wheat homoeologous chromosomes caused by locus duplications and deletions correlate with recombination rates Proc Natl Acad Sci U S A 2003 100 10836 10841 12960374 10.1073/pnas.1934431100
Chuang JH Li H Functional bias and spatial organization of genes in mutational hot and cold regions in the human genome PLoS Biol 2004 2 E29 14966531 10.1371/journal.pbio.0020029
Zhang L Li W Mammalian housekeeping genes evolve more slowly than tissue-specific genes Mol Biol Evol 2004 21 236 239 14595094 10.1093/molbev/msh010
Lercher MJ Urrutia AO Hurst LD Clustering of housekeeping genes provides a unified model of gene order in the human genome Nat Genet 2002 31 180 183 11992122 10.1038/ng887
Hua-Van A Daviere JM Kaper F Langin T Daboussi MJ Genome organization in Fusarium oxysporum: Clusters of class II transposons Curr Genet 2000 37 339 347 10853772 10.1007/s002940050537
Zhang J Peterson T Genome rearrangements by nonlinear transposons in maize Genetics 1999 153 1403 1410 10545468
Zhang J Peterson T Transposition of reversed Ac element ends generates chromosome rearrangements in maize Genetics 2004 167 1929 1937 15342530 10.1534/genetics.103.026229
Graham MA Marek LF Shoemaker RC Organization, expression and evolution of a disease resistance gene cluster in soybean Genetics 2002 162 1961 1977 12524363
Gandolfo MA Nixon KC Crepet WL A new fossil flower from the turonian of new jersey – Dressiantha bicarpellata gen. et sp. nov. (capparales) American Journal of Botany 1998 85 964 974
Simillion C Vandepoele K Saeys Y Van de Peer Y Building genomic profiles for uncovering segmental homology in the twilight zone Genome Res 2004 14 1095 1106 15173115 10.1101/gr.2179004
Royal Botanic Gardens, Kew: Plant DNA C-values Database (release 3.0, December 2004)
The Arabidopsis Genome Initiative Analysis of the genome sequence of the flowering plant arabidopsis thaliana Nature 2000 408 796 815 11130711 10.1038/35048692
Kulikova O Gualtieri G Geurts R Kim DJ Cook D Huguet T de Jong JH Fransz PF Bisseling T Integration of the FISH pachytene and genetic maps of medicago truncatula Plant J 2001 27 49 58 11489182 10.1046/j.1365-313x.2001.01057.x
Blanc G Wolfe KH Widespread paleopolyploidy in model plant species inferred from age distributions of duplicate genes Plant Cell 2004 16 1667 1678 15208399 10.1105/tpc.021345
Schlueter JA Dixon P Granger C Grant D Clark L Doyle JJ Shoemaker RC Mining EST databases to resolve evolutionary events in major crop species Genome 2004 47 868 876 15499401 10.1139/g04-047
Koch MA Haubold B Mitchell-Olds T Comparative evolutionary analysis of chalcone synthase and alcohol dehydrogenase loci in Arabidopsis, Arabis, and related genera (Brassicaceae) Mol Biol Evol 2000 17 1483 1498 11018155
Zhu H Kim DJ Baek JM Choi HK Ellis LC Kuester H McCombie WR Peng HM Cook DR Syntenic relationships between Medicago truncatula and Arabidopsis reveal extensive divergence of genome organization Plant Physiol 2003 131 1018 1026 12644654 10.1104/pp.102.016436
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
Salamov A Solovyev V Ab initio gene finding in drosophila genomic DNA Genome Research 2000 10 516 522 10779491 10.1101/gr.10.4.516
SoftBerry – fgenesh
Quackenbush J Cho J Lee D Liang F Holt I Karamycheva S Parvizi B Pertea G Sultana R White J The TIGR gene indices: Analysis of gene transcript sequences in highly sampled eukaryotic species Nucleic Acids Res 2001 29 159 164 11125077 10.1093/nar/29.1.159
Notredame C Higgins DG Heringa J T-coffee: A novel method for fast and accurate multiple sequence alignment J Mol Biol 2000 302 205 217 10964570 10.1006/jmbi.2000.4042
Clamp M Cuff J Searle SM Barton GJ The Jalview java alignment editor Bioinformatics 2004 20 426 427 14960472 10.1093/bioinformatics/btg430
Eddy SR Profile hidden markov models Bioinformatics 1998 14 755 763 9918945 10.1093/bioinformatics/14.9.755
Felsenstein J PHYLIP (Phylogeny Inference Package) Version 3.6. Distributed by the Author 2000 Seattle: Department of Genetics, University of Washington
Schmidt HA Strimmer K Vingron M von Haeseler A TREE-PUZZLE: Maximum likelihood phylogenetic analysis using quartets and parallel computing Bioinformatics 2002 18 502 504 11934758 10.1093/bioinformatics/18.3.502
Goldman N Whelan S Statistical tests of gamma-distributed rate heterogeneity in models of sequence evolution in phylogenetics Mol Biol Evol 2000 17 975 978 10833204
Korber B Rodrigo AG, Learn GH Signature and sequence variation analysis Computational Analysis of HIV Molecular Sequences 2000 Dordrecht, Netherlands: Kluwer Academic Publishers 55 72
HIV Sequence Database: SNAP Submission Form
Nei M Gojobori T Simple methods for estimating the numbers of synonymous and nonsynonymous nucleotide substitutions Mol Biol Evol 1986 3 418 426 3444411
Ota T Nei M Variance and covariances of the numbers of synonymous and nonsynonymous substitutions per site Mol Biol Evol 1994 11 613 619 8078400
Song QJ Marek LF Shoemaker RC Lark KG Concibido VC Delannay X Specht JE Cregan PB A new integrated genetic linkage map of the soybean Theor Appl Genet 2004 109 122 128 14991109 10.1007/s00122-004-1602-3
|
16102170
|
PMC1201151
|
CC BY
|
2021-01-04 16:03:52
|
no
|
BMC Plant Biol. 2005 Aug 15; 5:15
|
utf-8
|
BMC Plant Biol
| 2,005 |
10.1186/1471-2229-5-15
|
oa_comm
|
==== Front
BMC Struct BiolBMC Struct. BiolBMC Structural Biology1472-6807BioMed Central 1472-6807-5-131609295310.1186/1472-6807-5-13Research ArticleCrystal structure of nitrogen regulatory protein IIANtr from Neisseria meningitidis Ren Jingshan [email protected] Sarah [email protected] Nick S [email protected] David [email protected] Joanne E [email protected] David K [email protected] Nigel J [email protected] Raymond J [email protected] The Oxford Protein Production Facility, Henry Wellcome Building for Genomic Medicine, University of Oxford, Roosevelt Drive, Oxford, OX3 7BN, UK2 Division of Structural Biology, Henry Wellcome Building for Genomic Medicine, University of Oxford, Roosevelt Drive, Oxford, OX3 7BN, UK3 The Bacterial Pathogenesis and Functional Genomics Group, The Sir William Dunn School of Pathology, University of Oxford, South Parks Road, Oxford, OX1 3RE, UK2005 10 8 2005 5 13 13 14 4 2005 10 8 2005 Copyright ©2005 Ren et al; licensee BioMed Central Ltd.2005Ren 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 properly cited.Background
The NMB0736 gene of Neisseria meningitidis serogroup B strain MC58 encodes the putative nitrogen regulatory protein, IIANtr (abbreviated to NM-IIANtr). The homologous protein present in Escherichia coli is implicated in the control of nitrogen assimilation. As part of a structural proteomics approach to the study of pathogenic Neisseria spp., we have selected this protein for structure determination by X-ray crystallography.
Results
The NM-IIANtr was over-expressed in E. coli and was shown to be partially mono-phosphorylated, as assessed by mass spectrometry of the purified protein.
Crystals of un-phosphorylated protein were obtained and diffraction data collected to 2.5 Å resolution. The structure of NM-IIANtr was solved by molecular replacement using the coordinates of the E. coli nitrogen regulatory protein IIAntr [PDB: 1A6J] as the starting model. The overall fold of the Neisseria enzyme shows a high degree of similarity to the IIANtr from E. coli, and the position of the phosphoryl acceptor histidine residue (H67) is conserved. The orientation of an adjacent arginine residue (R69) suggests that it may also be involved in coordinating the phosphate group. Comparison of the structure with that of E. coli IIAmtl complexed with HPr [PDB: 1J6T] indicates that NM-IIANtr binds in a similar way to the HPr-like enzyme in Neisseria.
Conclusion
The structure of NM-IIANtr confirms its assignment as a homologue of the IIANtr proteins found in a range of other Gram-negative bacteria. We conclude that the NM- IIANtr protein functions as part of a phosphorylation cascade which, in contrast to E. coli, shares the upstream phosphotransfer protein with the sugar uptake phosphoenolpyruvate:sugar phosphotransferase system (PTS), but in common with E. coli has a distinct downstream effector mechanism.
==== Body
Background
Neisseria spp. are Gram-negative Beta-Protobacteria which include many species found only in humans. Two Neisseria spp. are pathogenic to human: N. meningitidis and N. gonorrhoeae, responsible for bacterial meningitis and septicaemia, and gonorrhoea, respectively. In the last few years, the genomes of N. meningitidis serotypes A (strain Z2491) [1] and B (strain MC58) [2] and N. gonorrhoeae (strain FA1090) (currently unpublished) have been sequenced and annotated. As part of a structural proteomics approach to the study of pathogenic Neisseria, we have solved the structure of the putative nitrogen regulatory protein IIANtr of N. meningitidis (abbreviated to NM-IIANtr) (Gene NMB0736), the sequence of which is highly conserved amongst the Neisseria spp.
In Escherichia coli, the IIANtr gene (ptsN) is located within the sigma-54 factor coding operon, rpoN, with which it is co-transcribed [3]. Insertional mutagenesis of ptsN has been shown to suppress the conditional lethality of temperature sensitive era (erats) mutants [4]. The era gene of E. coli encodes a GTPase which appears to be essential for cell growth [5]. IIANtr is a member of the mannitol-fructose family of IIA protein/domains which forms part of the phosphoenolpyruvate: sugar phosphotransferase system (PTS) [4]. PTS controls sugar uptake by bacteria through a series of phosphoryl transfer reactions in which the sugar-specific IIA enzyme is phosphorylated on a highly conserved histidine residue by the histidine-containing phosphocarrier protein, HPr [6]. In E. coli, the rpoN operon also includes a gene encoding a protein related to HPr, designated NPr; both HPr and NPr have been shown to phosphorylate E. coli IIANtr [4]. The sequence similarity between HPr and the molybdenum-iron protein of the nitrogenase complex of Rhizobium trifolii [6] and the frequent association of PTS proteins with rpoN have been taken to suggest that IIANtr may provide a regulatory link between carbon and nitrogen assimilation in bacteria [7].
The role of the NM-IIANtr in Neisseria has not been characterized, and the potential functions of these proteins in cellular behaviour are diverse [8]. The gene encoding this protein is not found in proximity to the equivalent of the E. coli rpoN gene (NMB0217, which may not be functional) nor is it close to the HPr-like protein (NMB2045). The structures of the E. coli IIANtr [9] and related IIAmannitol (IIAmtl) domain of PTS [10] have been solved by X-ray crystallography. More recently, a solution structure of a phosphoryl transfer complex between IIAmtl domain and HPr has been described [11]. In this report, we compare the structure of the neisserial IIANtr to the homologous E. coli enzymes and complex and discuss the functional implications of our findings.
Results and discussion
Expression and purification
The NM-IIANtr was over-expressed in E. coli with an N-terminal His-tag which was removed prior to crystallization. Mass spectrometric analysis of the purified protein showed that it comprised a 60:40 mixture of mono-phosphorylated and un-phosphorylated protein. The conserved H67 in the protein was thus capable of phosphorylation by endogenous HPr/NPr from the expression system. Analyses of re-dissolved crystals of the protein by mass spectrometry showed that only the un-phosphorylated form of the protein was crystallized presumably due to labile nature of the phospho-histidine bond in the acidic crystallisation conditions, 0.1 M sodium citrate, pH 4 (data not shown).
Overall structure
The model of NM-IIANtr contains 146 (out of 149) protein residues and 72 water molecules. Residues 1 and 148–149 are not modelled in the structure due to lack of clearly defined electron density. The structure of NM-IIANtr consists of a 5-stranded mixed β-sheet which is sandwiched by six α-helices, two on one side and four on the other (Fig. 1a &1b). The overall fold, as expected from the sequence homology, is similar to that of E. coli IIANtr (34 % sequence identity) and IIAmtl (23% sequence identity). 122 Cαs out of 150 of E. coli IIANtr [PDB:1A6J] can be overlapped onto NM-IIANtr with a root mean square deviation (rmsd) of 0.78 Å. The major differences between the neisserial and E. coli structures appear at surface loops linking the secondary structure elements, for examples α2–α3, β3–β4 and β5-α4 loops (residues 34–40, 70–75 and 107–112 respectively in NM-IIANtr), and the termini (Fig. 1c). The α1 is shorter and there is a one residue insertion at the α2–α3 loop in NM-IIANtr. Whilst comparison of E. coli IIAmtl [PDB: 1A3A] with NM-IIANtr gave 120 equivalent Cαs with an rmsd of 1.48 Å. E. coli IIAmtl does not have the short α-helix at the N-terminus, but has a extra helix between β3 and β4. Residue insertions and deletions at α2–α3, β3–β4 and β5-α4 loops have resulted in large structural differences at these places between the two proteins (Fig. 1d). Similar differences between E. coli IIANt and IIAmtl have been noted [12].
Figure 1 Structure of NM-IIAntr. (a) Stereo figure of a Cα-trace of NM-IIANtr with every twentieth residue numbered (b) Stereo figure of a ribbon diagram of NM-IIANtr with secondary structure elements labelled (α-helices 1–4; β-strands 1–5). The active site residue H67 is displayed as ball-and-stick. (c) Stereo figure of an overlay of NM-IIANtr (green) and E. coliIIANtr (red) (d) Stereo figure of an overlay of NM-IIANtr(green) and E. coli IIAmtl(red).
Active site and interface with H-Pr
The active site of NM-IIANtr is located in a concave area of protein surface consisting of α3–α4 and β2–β3. The active site residue, H67, the equivalent of H73 in E. coli IIANtr or H65 in E. coli IIAmtl, is the fifth residue of β3 and is surrounded by hydrophobic residues L69, I65, L114, L117 and A121 on one side, and by hydrophilic residues R51 and R69 on the other side (Fig. 2a). H67 and R51 are conserved amongst all IIANtr and IIAmtl proteins. H67 is stabilized by a hydrogen bond to the carbonyl oxygen of L65 from its ND1 atom, while the NH1 atom of the R51 side-chain hydrogen bonds to the carbonyl groups of both G55 and itself; a similar pattern is also observed in E. coli IIANtr. The side-chain of R69 is folded toward the active site H67 unlike the corresponding K75 of E. coli IIANtr, which is folded away from the active site H73. In the E. coli IIANtr crystal structure there is a sulphate ion forming a salt bridge/hydrogen bonding interactions to the side-chains of both R57 and H73 in one molecule of the crystal asymmetric unit. Interestingly, in our structure there is also a strong peak of electron density located on the 2-fold crystallographic axis, which could be modelled as a sulphate ion interacting with the side-chains of R51, H67 and R69 via salt bridge/hydrogen bonding, as well as to the equivalent residues of the symmetry related molecule (Fig. 2a). The distance from the ND2 atom of H67 to the two-fold symmetry axis is 3.2 Å, providing insufficient space to accommodate a phosphoryl group, consistent with the finding that only un-phosphorylated protein is found in the crystals by mass spectrometry.
Figure 2 Active site of NM-IIANtr. (a) Stereo figure of an overlay of the active sites of NM-IIANtr (green line for main chain and orange for side-chains displayed as ball-and stick) and E. coli IIAmtl (grey line for main chain and individual side-chains) showing the positions of residues around the active site H67 and the sulphate ions observed in the structures of both NM-IIANtr and E. coli IIANtr [9]. Residues are numbered according to the sequence of NM-IIANtr. Hydrogen bonds are shown by broken yellow lines and nitrogen, oxygen and sulphur atoms are displayed as blue, red and yellow respectively. (b) Stereo figure of an overlay of NM-IIANtr (red for the main chain ribbon and green for interface residues displayed as ball-and-stick) and E. coli IIAmtl (grey for the main chain and orange for interface residues) showing the relative positions of the residues which form the interface between E. coli IIAmtl HPr and the corresponding residues in NM-IIANtr. Residues are numbered according to the sequence of NM-IIANtr
In the active site of the E. coli IIAmtl there is a second histidine residue (H111) which is conserved amongst all IIAmtl sequences [13] and is proposed by analogy to the glucose PTS to be essential for phosphoryl transfer to the next component in the cascade, IIBmtl[7]. H111 shows two side-chain configurations in both crystal and NMR structures [10,14] either being parallel to and pointing away from, the active site H65 (Fig. 1a). These differing H111 conformations might be related to the two active site geometries required for the phosphoryl group accepting and donating functions of the protein [10]. The equivalent histidine, H113, of NM-IIANtr and H120 of E. coli IIANtr has similar conformations and are located at the back of helix α4 pointing away from the active site. It is unlikely that this histidine can be repositioned at the active site by unwinding the helix and therefore it appears improbable that it play a role in phosphoryl transfer from IIANtr to other proteins. The question arises as to whether another residue in the active site of NM-IIANtr could play a role in phosphotransfer. Intriguingly, R69 in NM-IIANtr is positioned close to the active site and at a similar position to H111 of IIAmtl (Figure 2a). The structure therefore suggests mutagenesis experiments which could be carried out to investigate further the active site of the protein. More generally, it remains an open question as to how NM-IIANtr is de-phosphorylated and whether this results in the phosphorylation of another protein(s). Certainly in E. coli, IIANtr cannot substitute for IIAmtl in the PTS mediated phosphorylation of mannitol [4] and therefore cannot be de-phosphorylated as a consequence of transfer to IIBmtl. Since no IIBNtr component has been identified, it remains unclear what effectors IIANtr proteins bind to and what role phosphorylation plays in this interaction.
By contrast, more is known about the protein(s) that interact with IIANtr to phosphorylate the protein. Again, results from E. coli can be used to inform the likely situation in Neisseria. Thus phosphorylation of E. coli IIANtr involves association with the phospho-transfer proteins HPr, as well as the related NPr, and transfer of a phosphoryl group between histidine residues. HPr and NPr are in turn phosphorylated by the upstream histidine kinases Enzyme I (EI) and EINtr respectively, leading to the proposal that there are two parallel phosphoryl transfer chains in E. coli, with NPr the preferential donor to IIANtr [15]. Interestingly in Neisseria meningitidis, there is only one HPr-like protein, encoded by the gene NMB2045. The N. meningitidis HPr-like protein (abbreviated to NM-HPr) shares 32% and 38% sequence identity with E. coli HPr and NPr respectively. Insight into how IIA proteins bind to the upstream effectors has been obtained by nmr studies. A solution structure has been reported for the complex formed between E. coli HPr and IIAmtl [PDB:1J6T] showing that binding is achieved through a central core of hydrophobic contacts strengthened by a few hydrophilic interactions [11]. It is expected that the interactions between the two Neisseria proteins should resemble the E. coli HPr and IIAmtl complex. Superimposing NM-IIANtr onto E. coli IIAmtl and aligning the HPr sequence with NM-HPr have indeed revealed common features of protein-protein interactions among the two systems. The key residues of the hydrophobic core of E. coli IIAmtl involved in contacts with HPr are L57, I61, I112, I115, T119 and L122, which correspond to residues L59, V63, L114, L117, A121 and F124 in NM-IIANtr. The aromatic residue F48 of HPr makes most extensive interactions with the hydrophobic core, especially to I61 and L122 of E. coli IIAmtl, whereas in Neisseria the corresponding residues are M48 in NM-HPr, V63 and F124 in IIANtr, an example of change in shape complementarity. However, since helix α4 of NM-IIANtr is 3 residues longer and about 25° different in orientation compared to the same helix of E. coli IIAmtl, one would anticipate that the NM-HPr could bind in a different orientation. More recently, nmr has been used to characterise the interaction between E. coli IIANtr and N-Pr [16]. Chemical shift mapping identified the surface on IIANtr for NPr binding, which generally corresponds to the HPr -binding region of IIAmtl but specifically implicates G61, D115, S125, T156 and nearby residues in the interaction. The corresponding residues in NM-IIANtr are G55, N108, S118 and E149, with the region around G55 being the most highly conserved.
Conclusion
The structure of NM-IIANtr confirms its assignment as a homologue of the IIANtr proteins found in a range of other Gram-negative bacteria. In fact the overall fold of the Neisseria enzyme shows a high degree of similarity to both the IIANtr and IIAmtl proteins from E. coli. Further, the orientations of the two histidine residues in the active site region is conserved between the Neisseria and E. coli IIANtr proteins and is distinct from the IIAmtl.
The availability of a second IIANtr structure enables certain generalizations to be made. The effector mechanism of this sub-group of regulatory proteins is distinct from the IIA components of the PTS controlling sugar uptake, which involve a transfer of phosphate via histidine residues to a IIB acceptor protein. The nature of the downstream effectors of IIANtr proteins and the role of IIANtrphosphorylation in the process remain to be established. In contrast, the mechanism of phosphorylation of all IIA components appears to be broadly similar and involves inter-molecular transfer between histidine residues in a complex formed between HPr and IIA proteins. In Neisseria a single HPr-like protein appears to be responsible for phosphorylation of both IIANtr and the IIA components of the sugar PTS whereas in E. coli, there is a parallel pathway involving the HPr-related protein NPr. In E. coli, NPr is expressed, with IIA Ntr and σ54, from the rpoN operon implying common regulation of gene expression. This is not the case in Neisseria where these genes are found on different transcriptional units. Therefore, although the structure of Neisseria IIANtr indicates that it is part of a similar phosphotransfer cascade to E. coli, details of the regulation of the gene are likely to be distinct.
Methods
Protein production
The NM-IIANtr expression construct was generated by means of ligation-independent cloning using Gateway™ technology (InVitrogen). The NMB0736 gene [Genbank: AE002098 for complete genome sequence] was amplified from genomic DNA (Neisseria meningitidis strain MC58) with KOD HiFi™ polymerase (Novagen) using the forward primer:- 5'ggggacaagtttgtacaaaaaagcaggcttcctggaagttctgttccagggcccgATGAG CCTTATCGGCGAAATTTTG 3' and the following reverse primer:- 5'ggggaccactttgtacaagaaagctgggtctcaTTATTCTTCAGTCAGGATGGCACG 3'
The PCR product was purified using QIAquick 96 plates (Qiagen) and inserted into the vector pDONR221 by recombination between attB (PCR product) and attP (vector) sequences (BP reaction). The insert from this vector was then transferred in to the expression vector pDEST17 by recombination between attL (pDONOR vector) and attR (pDEST vector) sequences (LR reaction). The expression construct contained the following N-terminal His tag and 3C protease cleavage site (underlined) MAHHHHHHAGFLEVLFQGP. BP and LR reactions were carried out according to the manufacturer's instructions. Recombinant LR clones were identified by PCR using a gene specific forward primer and a T7 reverse primer and verified by DNA sequencing. Protein was produced in the E. coli strain, B834(DE3). The cells were grown at 37°C in a 1L of GS96 media (QBiogene) to an A600 of 0.6, induced with isopropyl β-D-thiogalactopyranoside (IPTG) to 0.5 mM and then incubated for a further 20 h. at 20°C reaching an A600 of approximately 5. The cells were harvested by centrifugation at 6000 g for 15 min. and lysed using a Basic-Z Cell Disruptor (Constant Systems Ltd) at 30 Kpsi in 30 ml of 50 mM Tris pH 7.5, containing 500 mM NaCl, 0.2% Tween-20. The NM-IIANtr was purified by Nickel affinity chromatography followed by size exclusion chromatography using the standard His Affinity-Gel filtration program on the Akta 3D™ (GE Healthcare). After centrifugation at 30000 g for 30 min., the lysate was loaded onto a 1 ml pre-charged HiTrap ™ Chelating Sepharose™ FF column (GE Healthcare). The column was washed with 50 mM Tris pH7.5, 500 mM NaCl, 20 mM imidazole. The protein was then eluted in 50 mM Tris pH 7.5, 500 mM NaCl, 500 mM imidazole and injected on to a 16/60 HiLoad™ Superdex 200 column (GE Healthcare) equilibrated in 20 mM Tris pH 7.5, 200 mM NaCl. Protein-containing fractions were analysed on SDS-Page gels (Criterion -Biorad). The N-terminal tag was removed by over-night incubation at 4°C with His-tagged 3C protease (prepared from a pET24a/His3C expression vector kindly provided by A. Geerlof, EMBL, Heidelberg). The 3C protease and any uncleaved protein were removed by Nickel affinity chromatography and the protein concentrated to 9.7 mg/ml using a 5 K MWCO Vivaspin 15 concentrator (Vivascience) in 20 mM Tris pH7.5, 200 mM NaCl, 1 mM TCEP. Mass spectrometry of protein and crystals was carried as described [17].
Crystallization and data collection
The protein was crystallized using the nanodrop crystallization procedure with standard OPPF protocols [18]. Crystals were grown by the sitting drop vapour diffusion method at room temperature from 3.2 M ammonium sulphate, 0.1 M citrate pH 4.0 over a period of 14 days. Diffraction data were collected at station PX14.2 of SRS (Daresbury, UK). Data images were recorded using an ADSC Quantum 4 CCD detector. A crystal mounted in a fibre loop was placed 250 mm from the detector and exposed to the X-ray beam with a wavelength of 0.945 Å. A total of 69 oscillation images of 2.0 degree per exposure were collected from a single crystal frozen under a stream of nitrogen at 100K. The diffraction data were indexed and integrated with DENZO [19] and merged with SCALEPACK. The crystal belongs to the trigonal system, with space group of either P3221 or P3121 and unit cell dimensions of a = b = 61.02 Å and c = 63.31 Å. There is one molecule in an asymmetric unit, the crystal has a solvent content of 41.5% in the crystal. The data set is 100% complete to 2.5 Å resolution (Table 1).
Table 1 X-ray data and refinement statistics
X-ray data
Space group P3121
Unit cell dimensions (a,b,c in Å) 61.02, 61.02, 63.31
Resolution range (Å) 30.0 - 2.50 (2.59-2.50)‡
Unique reflections 4988 (486)
Redundancy 7.6 (5.7)
Completeness (%) 100 (100)
Average I/σ(I) 10.2 (2.9)
Rmerge* 0.183 (0.552)
Refinement statistics:
No. atoms (protein/water) 1105/73
R-factor†(Rwork/Rfree) 0.201/0.276
Rms bond length deviation (Å) 0.006
Rms bond angle deviation (°) 1.21
Ramachandran plot statistics
Residues in most favoured region (%) 87.1
Residues in additional allowed region (%) 12.1
Residues in generously allowed region (%) 0
Residues in disallowed region (%) 0.8 (Arg 35)#
‡ data in brackets are for the high resolution shell.
* Rmerge = Σ |I-<I> |/Σ <I> † R-factor = Σ|FO - FC|/Σ FO;
# Arg 35 has well defined main-chain electron density.
Structure solution and refinement
The structure was solved using molecular replacement with CNS [20] and the coordinates of E. coli nitrogen regulatory protein IIANtr [PDB:1A6J] as the starting model. The real space cross rotation search was carried out using data from 15–4 Å and Patterson vectors of 5–24 Å. The cross rotation peaks were then subjected to
PC-refinement with e2e2 target followed by translation search with fastf2f2 target. At this stage the space group of the crystal was confirmed to be P3121. The highest peak of the translation search (θ1 = 150.1, θ2 = 68.4, θ3 = 82.6, x = 16.5, y = 24.1, z = -35.0), which is 4.2 σ above the mean and 3.2 σ above the 2nd highest one (noise peak), is corresponding to 11th of the 33 cross rotation peaks. Rigid-body refinement of the rotated and translated model at 30-4.0 Å gave an R-factor of 0.462. Rounds of simulated annealing, conjugate gradient minimization and B-factor refinement followed by model rebuilding and solvent molecule addition with O have resulted in the current structure which has a Rwork/Rfree of 0.201/0.276 for all data from 30-2.5 Å resolution. The rms deviation of the model from the ideal is 0.006 Å for bond lengths and 1.21° for bond angles (Table 1).
The atomic coordinates of NM-IIANtr and structure factors have been deposited in the Protein Data Bank under the accession code 2A0J.
Authors' contributions
JR collected and processed the diffraction data, modelled, refined and analyzed the structure, SS purified and crystallized the protein, JN carried out mass spectrometry of the protein and crystals, NB and DA cloned and expressed the protein, NJS initiated the study DKS and RJO coordinated the study. RJO and JR prepared the manuscript with additional input from DKS and NJS.
Acknowledgements
The Oxford Protein Production Facility is funded by the Medical Research Council UK and is part of the Structural Proteomics IN Europe (SPINE) consortium (European Commission Grant No. QLG2-CT-2002-00988).
==== Refs
Parkhill J Achtman M James KD Bentley SD Churcher C Klee SR Morelli G Basham D Brown D Chillingworth T Davies RM Davis P Devlin K Feltwell T Hamlin N Holroyd S Jagels K Leather S Moule S Mungall K Quail MA Rajandream MA Rutherford KM Simmonds M Skelton J Whitehead S Spratt BG Barrell BG Complete DNA sequence of a serogroup A strain of Neisseria meningitidis Z2491 Nature 2000 5 502 506 10.1038/35006655 10761919
Tettelin H Saunders NJ Heidelberg J Jeffries AC Nelson KE Eisen JA Ketchum KA Hood DW Peden JF Dodson RJ Nelson WC Gwinn ML DeBoy R Peterson JD Hickey EK Haft DH Salzberg SL White O Fleischmann RD Dougherty BA Mason T Ciecko A Parksey DS Blair E Cittone H Clark EB Cotton MD Utterback TR Khouri H Qin H Vamathevan J Gill J Scarlato V Masignani V Pizza M Grandi G Sun L Smith HO Fraser CM Moxon ER Rappuoli R Venter JC Complete Genome Sequence of Neisseria meningitidis Serogroup B Strain MC58 Science 2000 5 1809 1815 10.1126/science.287.5459.1809 10710307
Merrick MJ Coppard JR Mutations in genes downstream of the rpoN gene (encoding sigma 54) of Klebsiella pneumoniae affect expression from sigma 54-dependent promoters Mol Microbiol 1989 5 1765 1775 2695747
Powell BS Court DL Inada T Nakamura Y Michotey V Cui X Reizer A Saier MH Reizer J Novel proteins of the Phosphotransferease System encoded within the rpoN Operon of Escherichia coli J Biol Chem 1995 5 4822 4839 10.1074/jbc.270.9.4822 7876255
Inada T Kawakami K Chen SM Takiff HE Court DL Nakamura Y Temperature-sensitive lethal mutant of era, a G protein in Escherichia coli J Bacteriol 1989 5 5017 5024 2527846
Reizer J Regulation of sugar uptake and efflux in gram-positive bacteria FEMS Microbiol Rev 1989 5 149 156 10.1016/0168-6445(89)90019-3 2699246
Reizer J Reizer A Saier MH Jacobson GR A proposed link between nitrogen and carbon metabolism involving protein phosphorylation in bacteria Protein Science 1992 5 722 726 1304914
Saier MH Reizer J The bacterial phosphotransferase system: new frontiers 30 years later Mol Microbiol 1994 5 755 764 7815935
Bordo D van Montfort RLM Pijning T Kalk K Reizer J Saier MH Dijkstra BW The Three-dimensional Structure of the Nitrogen Regulatory Protein IIANtr from Escherichia coli J Mol Biol 1998 5 245 255 10.1006/jmbi.1998.1753 9636714
Van Montfort RLM Pijning T Kalk K Hangyi I Kouwijzer MLCE Robillard GT Dijkstra BW The structure of the Escherichia coli phosphotransferase IIAmannitol reveals a novel fold with two conformations of the active site Structure 1998 5 377 388 10.1016/S0969-2126(98)00039-2 9551558
Cornilescu G Lee BR Cornilescu CC Wang A Clore M Solution structure of the phosphoryl transfer complex between the cytoplasmic A domain of the mannitol transporter IIMannitol and HPr of the Escherichia coli phosphotransferase system J Biol Chem 2002 5 42289 42298 10.1074/jbc.M207314200 12202490
Van Montfort RL Dijkstra BW The functional importance of structural differences between the mannitol-specific IIAmannitol and the regulatory IIAnitrogen Protein Sci 1998 5 2210 2216 9792109
Reizer J Reizer A A voyage along the bases; novel phosphotransferase genes revealed by in silico analyses of the Escherichia coli genome Res Microbiol 1996 5 458 471 10.1016/0923-2508(96)84000-9 9092011
Van Dijk AA Scheek RM Dijkstra K Wolters GK Robillard GT Characterization of the protonation and hydrogen bonding state of the histidine residues in IIAmtl, a domain of the phosphoenolpyruvate-dependent mannitol-specific transport protein Biochemistry 1992 5 9063 9072 10.1021/bi00152a050 1390693
Rabus R Reizer J Paulsen I Saier MH Enzyme INtr from Escherichiai coli J Biol Chem 1999 5 26185 26191 10.1074/jbc.274.37.26185 10473571
Wang GS Peterkofsky A Keifer PA Li X NMR characterization of the Escherichia coli nitrogen regulatory protein IIANtr in solution and interaction with its partner protein, NPr Protein Science 2005 5 1082 1090 10.1110/ps.041232805 15741344
Nettleship JE Walter TS Aplin R Stammers DK Owens RJ Sample preparation and mass-spectrometric characterization of crystal-derived protein samples Acta Crystallogr 2005 5 643 645
Walter TS Diprose JD Brown J Pickford M Owens RJ Stuart DI Harlos K A procedure for setting up high-throughput nanolitre crystallization experiments. 1. Protocol design and validation J Appl Crystallogr 2003 5 308 314 10.1107/S0021889803001997
Otwinowski Z Minor W Processing of X-ray diffraction data collected in oscillation mode Methods Enzymol 1996 5 307 326
Brunger AT Adams PD Clore GM Delano WL Gros P Grosse-Kunstleve RW Jiang JS Kuszewski J Nilges M Pannu NS Read RJ Rice LM Simonson T Warren GL Crystallography & NMR system: A new software suite for macromolecular structure determination Acta Crystallogr 1998 5 905 921
|
16092953
|
PMC1201152
|
CC BY
|
2021-01-04 22:45:16
|
no
|
BMC Struct Biol. 2005 Aug 10; 5:13
|
utf-8
|
BMC Struct Biol
| 2,005 |
10.1186/1472-6807-5-13
|
oa_comm
|
==== Front
BMC Struct BiolBMC Structural Biology1472-6807BioMed Central London 1472-6807-5-141610517210.1186/1472-6807-5-14Research ArticleThe Ramachandran plots of glycine and pre-proline Ho Bosco K [email protected] Robert [email protected] Department of Pharmaceutical Chemistry, University of California San Francisco, 600 16th St, San Francisco, CA 94107, USA2 Centre de Biophysique Moléculaire Numérique, 2 Passage des déportés, B-5030 Gembloux, Belgium2005 16 8 2005 5 14 14 15 3 2005 16 8 2005 Copyright © 2005 Ho and Brasseur; 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 Ramachandran plot is a fundamental tool in the analysis of protein structures. Of the 4 basic types of Ramachandran plots, the interactions that determine the generic and proline Ramachandran plots are well understood. The interactions of the glycine and pre-proline Ramachandran plots are not.
Results
In glycine, the ψ angle is typically clustered at ψ = 180° and ψ = 0°. We show that these clusters correspond to conformations where either the Ni+1 or O atom is sandwiched between the two Hα atoms of glycine. We show that the shape of the 5 distinct regions of density (the α, αL, βS, βP and βPR regions) can be reproduced with electrostatic dipole-dipole interactions. In pre-proline, we analyse the origin of the ζ region of the Ramachandran plot, a region unique to pre-proline. We show that it is stabilized by a COi-1···CδHδi+1 weak hydrogen bond. This is analogous to the COi-1···NHi+1 hydrogen bond that stabilizes the γ region in the generic Ramachandran plot.
Conclusion
We have identified the specific interactions that affect the backbone of glycine and pre-proline. Knowledge of these interactions will improve current force-fields, and help understand structural motifs containing these residues.
==== Body
Background
The Ramachandran plot [1] is the 2d plot of the φ-ψ torsion angles of the protein backbone. It provides a simple view of the conformation of a protein. The φ-ψ angles cluster into distinct regions in the Ramachandran plot where each region corresponds to a particular secondary structure. There are four basic types of Ramachandran plots, depending on the stereo-chemistry of the amino acid: generic (which refers to the 18 non-glycine non-proline amino acids), glycine, proline, and pre-proline (which refers to residues preceding a proline [2]). The generic and proline Ramachandran plots are now well understood [3] but the glycine and pre-proline Ramachandran plots are not.
The generic Ramachandran plot was first explained by Ramachandran and co-workers in terms of steric clashes [1]. This has become the standard explanation for the observed regions in the Ramachandran plot [4,5]. However, recent studies found significant discrepancies between the classic steric map and the Ramachandran plot of high-resolution protein structures [6-9]. These discrepancies have now been resolved. The first discrepancy is that the N···Hi+1 and Oi-1···C steric clashes in the classic steric map have no effect in the observed Ramachandran plot [3]. By removing these steric clashes, a better steric map can be constructed. The second discrepancy is that the Ramachandran plot cluster into distinct regions within the sterically-allowed regions of the Ramachandran plot [8,10]. These clusters have now been explained in terms of backbone dipole-dipole interactions [3,11,12].
The proline Ramachandran plot has been reproduced in a calculation [13]. The proline Ramachandran plot is severely restricted by the pyrrolidine ring, where the flexibility in the pyrrolidine ring couples to the backbone [14].
The observed glycine Ramachandran plot has a distinctive distribution (Figure 1A) quite different to the generic Ramachandran plot. An early attempt to explain the observed Ramachandran plot in terms of a steric map of glycine [15] (Figure 2A) fails to account for the observed distribution. It does not explain the observed clustering at ψ = 180° and ψ = 0°, nor the clustering into 5 distinct regions [8]. Using a molecular-dynamics simulation of Ace-Gly-Nme [16], Hu and co-workers found that the glycine Ramachandran plot generated by standard force-fields reproduced the original steric map but not the observed Ramachandran plot. They calculated a somewhat better result with a quantum-mechanics/molecular-mechanics model, which reproduced the observed clustering along ψ, but not the partitioning into the 5 clusters. In this study, we identify the specific interactions that define the observed glycine Ramachandran plot by studying the conformations of glycine in the structural database. We test these interactions with a simple model based on electrostatics and Lennard-Jones potentials.
Although the overall shape of the pre-proline Ramachandran plot (Figure 1B) is well understood, there exists a region unique to pre-proline that remains unexplained. The basic shape of pre-proline was predicted by Flory using steric interactions [17]. This was later confirmed in a statistical analysis of the protein database [2]. However, the statistical analysis also revealed the existence of a little leg of density poking out below the β-region (Figure 1B; purple in Figure 2C), which Karplus called the ζ region [10]. More recent calculations using standard molecular mechanics force-fields reproduced the energy surface of the original Flory calculation [13,18] but not the ζ region. In this study, we focus on the physical origin of the ζ region.
Results
A non-redundant PDB data-set
To extract the statistical distributions of the glycine and pre-proline Ramachandran plots, we chose a high-resolution subset of the PDB [19] provided by the Richardson lab [9] of 500 non-homologous proteins. These proteins have a resolution of better than 1.8 Å where all hydrogen atoms have been projected from the backbone and optimized in terms of packing. Following the Richardsons, we only consider atoms that have a B-factor of less than 30.
Regions in the glycine Ramachandran plot
Glycine is fundamentally different to the other amino acids in that it lacks a sidechain. In particular, glycine does not have the Cβ atom, which induces many steric clashes in the generic Ramachandran plot. We call the hydrogen atom that is shared with the other amino acids, the Hα1 atom. We call the hydrogen atom that replaces the Cβ atom, the Hα2 atom. The absence of the Cβ atom allows the glycine Ramachandran plot to run over the borders at -180° and 180° (Figure 1A).
The observed glycine map has 5 regions of density [8]. In order to display the observed density in one continuous region, we shift the coordinates from φ-ψ to φ'-ψ' where φ': 0° < φ' < 360°, and ψ': -90° < ψ' < 270°. With the shifted glycine Ramachandran plot (Figure 3A), we can clearly identify the different regions. Along the horizontal strip ψ' ~ 180°, there are three separate regions. One of these is an elongated version of the βP region of the generic Ramachandran plot. The βP region corresponds to the polyproline II structure, which forms an extended left-handed helix along the protein chain [20]. The βPR region is a reflection of the βP region where a sequence of glycine residues in the βPR conformation will form a right-handed helix. Finally, there is a region that corresponds to the βS region of the generic Ramachandran plot. This region corresponds to the extended conformation of residues in β-sheets. However, the glycine βS region, centered on (φ', ψ') = (180°, 180°), is slightly displaced from the βS region of the generic Ramachandran plot. There is also the diagonal α and αL regions (Figure 3A), which are associated with helices and turns [3]. Unlike the generic Ramachandran plot, the glycine α region is symmetric to the αL region [8,21]. In the generic Ramachandran plot, there is also a γ region corresponding to the hydrogen bonded γ-turn [12]. The glycine Ramachandran plot does not have any density in the γ region.
Steric interactions in glycine
The original steric map of glycine (Figure 2A) [15] fails to explain large parts of the observed glycine Ramachandran plot (Figure 1A). In the observed glycine Ramachandran (Figure 3A), there are two large excluded horizontal strips at 50° < ψ' < 120° and -120° < ψ' < -50°, which are not excluded in the glycine steric map (Figure 2A). Conversely, the glycine steric map excludes a horizontal strip at -30° < ψ' < 30° (Figure 2A), but this region is populated in the observed plot (Figure 1A). There are also diagonal steric boundaries in the observed glycine Ramachandran plot (Figure 1A), whereas the steric map predicts vertical boundaries (Figure 2A).
We carried out a re-evaluation of the steric map of glycine (Figure 2B) by following the methodology of Ho and co-workers [3]. For each interaction in the glycine backbone, we consider the variation of the inter-atomic distance with respect to the φ'-ψ' angles. We compare the observed variation to the variation generated from a model that uses canonical backbone geometry. We divide these interactions into 3 categories: the φ' dependent, ψ' dependent and φ'-ψ' co-dependent distances.
For some of the interactions, the results for glycine are identical to that of the generic Ramachandran plot [3]. For brevity, we omit the analysis of these interactions and summarize the results. The excluded horizontal strip -30° < ψ' < 30°, due to the N···Hi+1 steric interaction in the glycine steric map (Figure 2A), does not exist in the observed distribution (Figure 1A). Similarly, the Oi-1···C steric clash in the original glycine steric map, which excludes a vertical strip centered on φ' = 0° (Figure 2A), does not exist in the observed distribution (Figure 1A). We ignore the effect of the N···Hi+1 and Oi-1···C steric clashes. The diagonal boundaries of the observed distribution are defined by the φ'-ψ' co-dependent steric interactions Oi-1···O and Oi-1···Ni+1. In Figure 3A, we show the fit of these steric interactions to the data.
Here, we analyze the most distinctive feature of the glycine Ramachandran plot – the tendency for ψ' to cluster near 180° and 0°. We focus on the ψ'-dependent interactions. For each interaction, we first calculate the model curve of the corresponding inter-atomic distance as a function of ψ' (see Methods). We then compare the observed ψ' distribution (bottom of Figure 3B) to the curve. If a hard-sphere repulsion restricts ψ', then, in regions of ψ' where the model curve is below the van der Waals (VDW) diameter (horizontal dashed line in Figure 3B), the ψ' frequency distribution should drop correspondingly.
In the region (60° < ψ' < 100°), we find that the drop-off in the ψ frequency distribution (bottom of Figure 3B) corresponds to values of Hα1···Ni+1 (bottom of Figure 3B) and Hα2···O (top of Figure 3B) that are smaller than their VDW diameters. In the region (-90° < ψ' < -60; 210° < ψ' < 270°), the drop-off in the ψ frequency distribution corresponds to regions where Hα2···Ni+1 and Hα1···O are found below their VDW radii. In contrast, the values of Hα1···Hi+1 and Hα2···Hi+1 are never found significantly below their VDW diameter (middle of Figure 3B).
The observed ψ' dependence in glycine is due to the Hα1···O, Hα2···O, Hα1···Ni+1 and Hα2···Ni+1 steric clashes. A simple interpretation is that the ψ' dependence in glycine arise from conformations that place either the Ni+1 or O atom between the two Hα atoms (Figure 4A). The observed limits in the distributions have been drawn in Figure 3A as horizontal lines.
We thus obtain a revised steric map of glycine, consisting of the steric clashes Oi-1···O, Oi-1···Ni+1, Hα1···O, Hα2···O, Hα1···Ni+1 and Hα2···Ni+1. Using parameters from CHARMM22 [22], we calculate the Lennard-Jones 12-6 potential due to the revised steric clashes (Figure 5A). The minimum-energy region accounts for much of the shape of the observed distribution (Figure 3A).
Dipole-dipole interactions in glycine
The revised glycine steric map does not explain the diagonal shape of the α, αL, βP, βPR and βS regions. In the generic Ramachandran plot, it was found that the diagonal shape of regions could be reproduced using electrostatic dipole-dipole interactions [3] but only when the dipole-dipole interactions were considered individually. The overall electrostatic interaction does not reproduce the observed Ramachandran plot [23]. Here, we use the same approach of treating individual electrostatic dipole-dipole interactions along the glycine backbone.
We calculate the energy map of φ-ψ for the 4 dipole-dipole interactions in the glycine backbone interaction: COi-1···CO, NH···NHi+1, CO···NH and COi-1···NHi+1 (Figure 5C-F). The electrostatic interactions are calculated with the Lennard-Jones potentials of the steric clashes identified in the section above. We find that the shapes of the different regions of the glycine Ramachandran plot (Figure 3A) are reproduced (Figure 5). The CO···NH interaction produces the diagonal αL, α and βS region (Figure 5E). The NH···NHi+1 interaction also produces a diagonal αL and α region (Figure 5D). The α region is symmetric to the αL region. The COi-1···CO interaction produces minima corresponding to the βP and βPR regions (Figure 5C).
In the original glycine steric map (Figure 2A), the region near (φ, ψ) = (-180°, 180°) is forbidden due to a steric clash between O and H. Yet glycine has density in this region in the observed Ramachandran plot (Figure 3A). This can also be seen in the frequency distribution of d(O···H) (Figure 3C), where there is a peak at d(O···H) ~ 2.4 Å. At this peak, the O and H atoms are in contact, as the VDW diameter is 2.5 Å. Thus, in the βS region of glycine, the favorable CO···HN dipole-dipole interaction overcomes the steric repulsion of the O and H atoms (Figure 5E).
The pre-proline Ramachandran plot
Schimmel and Flory argued in 1968 that pre-proline – amino acids preceding proline – has a particularly restricted Ramchandran plot, compared to the generic Ramachandran plot [17]. This was finally observed in the protein database by MacArthur and Thornton (Figure 1B) [2].
There are three main differences between the pre-proline Ramachandran plot and the generic Ramachandran plot. In the pre-proline Ramachandran plot, there is a large excluded horizontal strip at -40° < ψ < 50°, which restricts αL and α regions. The αL region is shifted up higher. These two features were reproduced in the Schimmel-Flory calculation [17] and subsequent calculations [13,18]. The third feature is a little leg of density poking out below the β-region (Figure 1B; purple in Figure 2C). Karplus called this the ζ region [10], which is unique to pre-proline.
Previous calculations [2,17,18] did not focus on the individual interactions, and did not account for the ζ region. Here, we identify the exact steric clashes that determine the pre-proline Ramachandran plot. We will then analyse the interactions responsible for the ζ region.
Steric interactions in the pre-proline backbone
In pre-proline, instead of an interaction with the NH atom in the succeeding generic amino acid, the pre-proline interacts with a CH2 group of the succeeding proline (Figure 1B). The CH2 group exerts a much larger steric effect on the pre-proline Ramachandran plot. MacArthur and Thornton [2] suggested that the dominant effect is due to the N···Cδi+1 and Cβ···Cδi+1 steric clashes. Here we can analyse the efficacy of each clash by analysing the statistical distributions directly.
We consider the φ-ψ co-dependent interactions that involve the Cδ, Hδ1 and Hδ2 atoms of the succeeding proline (Figure 1B). For each interaction, we generate the contour plot in φ-ψ of the VDW diameter distance. By comparing the contour plot to the observed density in the pre-proline Ramachandran plot, we identify the interactions that induce the best match in the boundaries (Figure 6A, the interactions are identified in Figure 2C). We found that the chunk taken out of the bottom-left β-region of the observed density is due to the Oi-1···Cδi+1 steric clash. Another restriction on the αL and α regions is due to the H···Cδi+1 steric clash.
We next consider the ψ dependent interactions. In the pre-proline ψ frequency distribution, we found three distinct peaks (bottom Figure 6B). The left-most peak at ψ ~ -50° corresponds to the α region of pre-proline. We focus on the two peaks in the β-region 50° < ψ < 180° The larger peak centred on ψ ~ 150° corresponds to the βS region of the generic Ramachandran plot. In the generic Ramachandran plot, this βS region is bounded by the Cβ···O and Cβ···Ni+1 steric clashes. In pre-proline, the smaller peak centred on ψ ~ 70° corresponds to the ζ region and occurs in a region that would be excluded by the Cβ···O steric clash. Instead the smaller peak is bounded from below by the N···Cδi+1 steric clash. This can be seen by comparing the ψ distribution to the model curve of N···Cδi+1 vs. ψ (middle of Figure 6B).
Using parameters from CHARMM22, we calculate the Lennard-Jones 12-6 potential due to the revised steric clashes (Figure 7A). Lennard-Jones potentials cannot account for the ζ region.
Interactions that stabilize the pre-proline ζ region
As the ζ region (purple in Figure 2B) brings the Cβ···O interaction into steric conflict, there must be a compensating interaction that stabilizes the ζ region. What is this interaction? To understand this interaction, we consider an analogy with the γ region in the generic Ramachandran plot. In the γ region, a distorted COi-1···HNi+1 hydrogen bond is formed, which brings the Hi+1 atom into contact with the Oi-1 atom. Similarly, in the ζ region of pre-proline, the Oi-1 atom of pre-proline is in contact with the Hδ1 and Hδ2 atoms (see Figure 4B; Table 1), suggesting that the COi-1 group interacts with the CδHδi+1 group of the succeeding proline.
Can the Cδ Hδi+1 group interact with COi-1? Such an interaction would fall under the class of the CH···O weak hydrogen bond, a well-documented interaction in proteins [24]. Studies of the CH···O weak hydrogen bond use a distance criteria of d(H···O) < 2.8 Å [25-27]. There is little angular dependence found in the CH···O bond around the H atom where an angle criteria of ∠OHX > 90° is often used. This is much more permissive than the geometry of the canonical hydrogen bond. In Table 1, we list the hydrogen bond parameters of the COi-1···CδHδi+1 interaction in the ζ region. As proline can take on two different major conformations, the UP and DOWN pucker, measurements of the geometry of the COi-1···CδHδi+1 interaction must also be divided in terms of the UP and DOWN pucker. The observed geometry of the COi-1···CδHδi+1 geometry satisfies the geometric criteria of the weak hydrogen bond (Table 1).
As the COi-1···CδHδi+1 weak hydrogen bond is a close contact, we need to model the interaction in order to understand its dependence on the φ-ψ angles. For the modelling, we consider strategies that have been used for the analogous COi-1···HNi+1 hydrogen bond. The COi-1···HNi+1 hydrogen bond has been modelled in quantum-mechanical studies where the γ region was found to be the minimum energy conformation in vacuum [12]. A simpler approach, which modelled the hydrogen bond with electrostatic dipole-dipole interactions, also find a minimum in the γ region [23].
Here, we model the COi-1···CδHδi+1 weak hydrogen bond as an electrostatic dipole-dipole interaction (see Methods). How do we model the CδHδi+1 group as an electrostatic dipole? Bhattacharyya and Chakrabarti [28] found that, of the CH groups in proline, the CδHδ group forms the most CH···O hydrogen bonds. The Cδ atom sits next to the electron-withdrawing N atom and thus, is more acidic than the other C atoms. Consequently, we place a small negative partial charge on the Cδ atom. In our model, we find an energy minimum in the ζ region for both the UP pucker (Figure 7B) and the DOWN pucker (Figure 7C). We conclude that the COi-1···Cδi+1Hδ1i+1 weak hydrogen bond stabilizes the ζ region in pre-proline.
Conclusion
We have identified the interactions that determine the high-resolution Ramachandran plots of glycine and pre-proline.
For glycine, the Ramachandran plot of the glycine backbone modeled by standard force-fields fails to reproduce the observed Ramachandran plot [16]. Instead the modeled Ramachandran plot resembles the original steric map of glycine [1]. The failure of these calculations arises from the inadequate treatment of the Hα atoms. We have identified a revised set of steric interactions that can reproduce the observed glycine Ramachandran plot. These are Oi-1···O, Oi-1···Ni+1, Hα1···O, Hα2···O, Hα1···Ni+1 and Hα2···Ni+1 (Figure 2B). These steric interactions constrain either the Ni+1 or O atom to be sandwiched between the two Hα atoms, which clusters glycine to ψ = 180° and ψ = 0°. The five clustered regions can be traced to electrostatic dipole-dipole interactions: the CO···NH interaction induces diagonal αL, α and βS regions; and the COi-1···CO interaction induces the diagonal βP and βPR regions.
Previous calculations of the pre-proline Ramachandran reproduced most of the observed pre-proline Ramachandran plot with the notable exception of the ζ region. Previous studies did not identify the specific steric interactions involved in defining the pre-proline Ramachandran plot. Here, we have identified them: N···Cδi+1, Oi-1···Cδi+1 and H···Cδi+1 (Figure 2C). We have also identified the physical mechanism that stabilizes the ζ region (purple in Figure 2C). It is the COi-1···CδHδi+1 weak hydrogen bond, which is directly analogous to the COi-1···NHi+1 hydrogen bond that stabilizes γ-turns in the generic amino acid.
Combined with the analysis of the generic Ramachandran plot [3] and the proline Ramachandran plot [13,14], we have identified the interactions that define the high-resolution Ramachandran plots of all 20 amino acids. Although our analysis uses simple modeling techniques, the interactions identified here suggest concrete ways to resolve the inadequacies in current force-fields.
Methods
VDW radii
In the steric clash analysis, we used the VDW radii given by the Richardson lab [29]: Hα = 1.17Å, H = 1.00Å, C = 1.65Å, Cα = Cβ = 1.75Å, O = 1.40Å and N = 1.55Å. From the database, we extracted 7277 glycine and 4336 pre-proline residues.
Local conformations of the φ-ψ map
To calculate the model curves of the inter-atomic distances as a function of the φ-ψ angles, we modeled the glycine and pre-proline protein fragments shown in Figure 1. Covalent bond lengths and angles were fixed to CHARMM22 values [22]. Only the φ-ψ angles vary. The φ-ψ angles of the central residue were incremented in 5° steps and the corresponding distance parameters and energies of the inter-atomic interactions were calculated. We used 2 types of interactions, partial charge electrostatics, Eelec = 331·(q1·q2) kcal·mol-1, and Lennard-Jones 12-6 potentials, ELJ = ε (σ/d)12 – 2 (σ/d)6) kcal·mol-1, where the parameters were taken from CHARMM22 [22]. There are no parameters in CHARMM22 for the Hδ and Cδ atoms. As such, we have assigned a partial charge of -0.20 to Cδ and 0.10 to Hδ1 and Hδ2. These are not based on any detailed arguments but are merely used to estimate the effect that such charges would have.
Authors' contributions
Both authors conceived the study. BKH carried out the data analysis and modelling, and drafted the manuscript. RB provided guidance and mentorship.
Acknowledgements
BKH was supported by a post-doctoral grant from the Fonds National de la Recherche Scientifique (FNRS), Belgium. RB is director of research of FNRS.
Figures and Tables
Figure 1 Backbone conformations of glycine and pre-proline. Backbone schematic (left) and observed Ramachandran plot (right) of (A) glycine and (B) pre-proline. Taken from the data-set of Lovell et al. (2003). The clustered regions are labeled on the Ramachandran plots.
Figure 2 Schematic of the Ramachandran plot. (A) original steric map of glycine, in standard (left) and shifted (right) coordinates; (B) revised schematic of glycine, in standard (left) and shifted (right) coordinates; (C) pre-proline. The clustered regions are: grey – sterically allowed; red – α and αL; yellow – βS; blue – βP and βPR; purple – ζ. See text for explanation of the regions.
Figure 3 Glycine parameters. (A) The Ramachandran plot in shifted coordinates φ'-ψ'. The dashed lines show the steric clashes that define the boundaries of the observed densities (Figure 2B describes the specific interactions). (B) The distributions of various inter-atomic interactions as a function of ψ'. The dashed line show the limit of the VDW diameters. The grey line gives the model curve calculated with ideal geometry. At the bottom is the frequency distribution of the ψ' angle. (C) Frequency distribution of the inter-atomic distance d(O···H). There are 3 peaks, of which, the smallest at d(O···H) = 2.4 Å, which corresponds to the βS region.
Figure 4 Stick figure representation of glycine and pre-proline. (A) glycine in the ψ ~ 180° conformation where the Ni+1 atom is sandwiched between the two Hα atoms, and (B) pre-proline in the ζ conformation where the Oi-1 atom interacts with the Hδ atoms of the succeeding proline.
Figure 5 Dipole-dipole interactions in glycine. Axes are shown in the shifted φ'-ψ' angles [°]. Energy plots [kcal/mol] of (a) the Lennard-Jones 12-6 potentials of the revised set of steric clashes; (b) all electrostatic interactions; (c)-(f) the individual dipole-dipole interactions of the glycine backbone (see Figure 1A for backbone schematic of the dipoles). Energy parameters were taken from CHARMM22. The light areas show regions of minimum energy.
Figure 6 Pre-proline parameters. (A) The Ramachandran plot. The dashed lines show the steric clashes that define some of the boundaries of the observed densities (see Figure 2C). (B) The distributions of various inter-atomic interactions as a function of ψ. The dashed lines show the limit of the VDW diameters. The solid grey line gives the model curve calculated with ideal geometry. At the bottom is the frequency distribution of the ψ angle.
Figure 7 Energy plots in pre-proline as a function of φ-ψ. Energy plots [kcal/mol] of (a) the Lennard-Jones 12-6 potentials of the revised set of steric clashes; the COi-1···CδHδi+1 dipole-dipole interactions when the succeeding proline ring is in (b) the UP pucker and (c) the DOWN pucker. The light areas show regions of low energy.
Table 1 Parameters of the CO···HX hydrogen bond
n φ ψ O···H ∠COH ∠OHX ∠dihCOHX
γ region of the generic amino acid
CO···HN 518 -85(3)° 81(11)° 2.39(0.24) Å 107(5)° 123(10)° 178(11)°
ζ region of pre-proline: UP PUCKER
CO···Hδ1Cδ 105 -129(6)° 80(6)° 3.64(0.27) Å 79(8)° 75(10)° 25(9)°
CO···Hδ2Cδ 2.59(0.22) Å 99(12)° 144(14)° -108(18)°
ζ region of pre-proline: DOWN PUCKER
CO···Hδ1Cδ 406 -129(6)° 74(9)° 3.16(0.38) Å 95(32)° 101(20)° 22(11)°
CO···Hδ2Cδ 2.98(0.36) Å 71(10)° 113(22)° -118(14)°
==== Refs
Ramachandran GN Ramakrishnan C Sasisekharan V Stereochemistry of polypeptide chain configurations J Mol Biol 1963 7 95 99 13990617
MacArthur MW Thornton JM Influence of proline residues on protein conformation. Journal of Molecular Biology 1991 218 397 412 2010917 10.1016/0022-2836(91)90721-H
Ho BK Thomas A Brasseur R Revisiting the Ramachandran plot: hard-sphere repulsion, electrostatics, and H-bonding in the alpha-helix Protein Sci 2003 12 2508 2522 14573863 10.1110/ps.03235203
Mandel N Mandel G Trus BL Rosenberg J Carlson G Dickerson RE Tuna cytochrome c at 2.0 A resolution. III. Coordinate optimization and comparison of structures J Biol Chem 1977 252 4619 4636 194885
Richardson JS The anatomy and taxonomy of protein structure Adv Protein Chem 1981 34 167 339 7020376
Kleywegt GJ Jones TA Phi/psi-chology: Ramachandran revisited Structure 1996 4 1395 1400 8994966 10.1016/S0969-2126(96)00147-5
Chakrabarti P Pal D The interrelationships of side-chain and main-chain conformations in proteins Prog Biophys Mol Biol 2001 76 1 102 11389934 10.1016/S0079-6107(01)00005-0
Hovmöller S Zhou T Ohlson T Conformations of amino acids in proteins Acta Crystallogr D Biol Crystallogr 2002 58 768 776 11976487 10.1107/S0907444902003359
Lovell SC Davis IW Arendall WB de Bakker PI Word JM Prisant MG Richardson JS Richardson DC Structure validation by Calpha geometry: phi,psi and Cbeta deviation Proteins 2003 50 437 450 12557186 10.1002/prot.10286
Karplus PA Experimentally observed conformation-dependent geometry and hidden strain in proteins Protein Sci 1996 5 1406 1420 8819173
Maccallum PH Poet R Milner-White EJ Coulombic interactions between partially charged main-chain atoms not hydrogen-bonded to each other influence the conformations of alpha-helices and antiparallel beta-sheet. A new method for analysing the forces between hydrogen bonding groups in proteins includes all the Coulombic interactions J Mol Biol 1995 248 361 373 7739046
Milner-White EJ Situations of gamma-turns in proteins. Their relation to alpha-helices, beta-sheets and ligand binding sites Journal of Molecular Biology 1990 216 386 397 2254936
Summers LN M. K Modeling of globular proteins. A distance-based data search procedure for the construction of insertion/deletion regions and Pro-non-Pro mutations Journal of Molecular Biology 1990 216 991 1016 2266566
Ho BK Coutsias EA Seok C Dill KA The flexibility in the proline ring couples to the protein backbone Protein Sci 2005 14 1011 1018 15772308 10.1110/ps.041156905
Ramakrishnan C Ramachandran GN Stereochemical criteria for polypeptide and protein chain conformations. II. Allowed conformations for a pair of peptide units Biophys J 1965 5 909 933 5884016
Hu H Elstner M Hermans J Comparison of a QM/MM force field and molecular mechanics force fields in simulations of alanine and glycine "dipeptides" (Ace-Ala-Nme and Ace-Gly-Nme) in water in relation to the problem of modeling the unfolded peptide backbone in solution Proteins 2003 50 451 463 12557187 10.1002/prot.10279
Schimmel PR Flory PJ Conformational energies and configurational statistics of copolypeptides containing L-proline J Mol Biol 1968 34 105 120 5760450 10.1016/0022-2836(68)90237-4
Hurley JH Mason DA Matthews BW Flexible-geometry conformational energy maps for the amino acid residue preceding a proline Biopolymers 1992 32 1443 1446 1457725 10.1002/bip.360321104
Berman HM Westbrook J Feng Z Gilliland G Bhat TN Weissig H Shindyalov IN Bourne PE The Protein Data Bank Nucleic Acids Research 2000 28 235 242 10592235 10.1093/nar/28.1.235
Stapley BJ Creamer TP A survey of left-handed polyproline II helices Protein Sci 1999 8 587 595 10091661
Eswar N Nagarajaram HA Ramakrishnan C Srinivasan N Influence of solvent molecules on the stereochemical code of glycyl residues in proteins Proteins 2002 49 326 334 12360522 10.1002/prot.10226
Mackerell ADJ Bashford D Bellott M Dunbrack Jr. RL Evanseck JD Field MJ Fischer S Gao J Guo H Ha S Joseph-McCarthy D Kuchnir LKK Lau FTK Mattos C Michnick S Ngo T Nguyen DT Prodhom B Reiher WEIII Roux B Schlenkrich M Smith JC Stote R Straub J Watanabe M Wiorkiewicz-Kuczera J Yin D Karplus M All-atom empirical potential for molecular modeling and dynamics Studies of proteins Journal of Physical Chemistry 1998 B102 3586 3616
Roterman IK Lambert MH Gibson KD Scheraga HA A comparison of the CHARMM, AMBER and ECEPP potentials for peptides. II. Phi-psi maps for N-acetyl alanine N'-methyl amide: comparisons, contrasts and simple experimental tests J Biomol Struct Dyn 1989 7 421 453 2627294
Derewenda ZS Lee L Derewenda U The occurrence of C-H...O hydrogen bonds in proteins J Mol Biol 1995 252 248 262 7674305 10.1006/jmbi.1995.0492
Taylor R Kennard O Crystallographic evidence for the existence of C-H...O, C-H...N, and C-H...Cl hydrogen bonds J Am Chem Soc 1982 104 5063 5070 10.1021/ja00383a012
Steiner T Saenger W Role of C-H...O hydrogen bonds in the coordination of water molecules. Analysis of Neutron Diffraction Data J Am Chem Soc 1993 115 4540 4547 10.1021/ja00064a016
Vargas R Garza J Dixon DA Hay BP How strong is the C{alpha}-H...O=C hydrogen bond? J Am Chem Soc 2000 122 4750 4755 10.1021/ja993600a
Bhattacharyya R Chakrabarti P Stereospecific interactions of proline residues in protein structures and complexes J Mole Biol 2003 331 925 940 10.1016/S0022-2836(03)00759-9
Word JM Lovell SC LaBean TH Taylor HC Zalis ME Presley BK Richardson JS Richardson DC Visualizing and quantifying molecular goodness-of-fit: small-probe contact dots with explicit hydrogen atoms J Mol Biol 1999 285 1711 1733 9917407 10.1006/jmbi.1998.2400
|
16105172
|
PMC1201153
|
CC BY
|
2021-01-04 16:03:52
|
no
|
BMC Struct Biol. 2005 Aug 16; 5:14
|
utf-8
|
BMC Struct Biol
| 2,005 |
10.1186/1472-6807-5-14
|
oa_comm
|
==== Front
BMC Struct BiolBMC Structural Biology1472-6807BioMed Central London 1472-6807-5-151610517610.1186/1472-6807-5-15Research ArticleInteraction preferences across protein-protein interfaces of obligatory and non-obligatory components are different De Subhajyoti [email protected] O [email protected] N [email protected] N [email protected] Department of Biotechnology, Indian Institute of Technology, Kharagpur, 721 302, India2 Molecular Biophysics Unit, Indian Institute of Science, Bangalore 560 012, India2005 16 8 2005 5 15 15 30 11 2004 16 8 2005 Copyright © 2005 De 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 polypeptide chain of a protein-protein complex is said to be obligatory if it is bound to another chain throughout its functional lifetime. Such a chain might not adopt the native fold in the unbound form. A non-obligatory polypeptide chain associates with another chain and dissociates upon molecular stimulus. Although conformational changes at the interaction interface are expected, the overall 3-D structure of the non-obligatory chain is unaltered. The present study focuses on protein-protein complexes to understand further the differences between obligatory and non-obligatory interfaces.
Results
A non-obligatory chain in a complex of known 3-D structure is recognized by its stable existence with same fold in the bound and unbound forms. On the contrary, an obligatory chain is detected by its existence only in the bound form with no evidence for the native-like fold of the chain in the unbound form. Various interfacial properties of a large number of complexes of known 3-D structures thus classified are comparatively analyzed with an aim to identify structural descriptors that distinguish these two types of interfaces. We report that the interaction patterns across the interfaces of obligatory and non-obligatory components are different and contacts made by obligatory chains are predominantly non-polar. The obligatory chains have a higher number of contacts per interface (20 ± 14 contacts per interface) than non-obligatory chains (13 ± 6 contacts per interface). The involvement of main chain atoms is higher in the case of obligatory chains (16.9 %) compared to non-obligatory chains (11.2 %). The β-sheet formation across the subunits is observed only among obligatory protein chains in the dataset. Apart from these, other features like residue preferences and interface area produce marginal differences and they may be considered collectively while distinguishing the two types of interfaces.
Conclusion
These results can be useful in distinguishing the two types of interfaces observed in structures determined in large-scale in the structural genomics initiatives, especially for those multi-component protein assemblies for which the biochemical characterization is incomplete.
==== Body
Background
Proteins interact with other proteins and bring about myriad of molecular activities in the cell. Interacting proteins are known to play key roles in almost all cellular and biological processes such as metabolism, endocrine, exocrine and paracrine signaling, protein synthesis and trafficking [1]. With the availability of genomic data in abundance, it is important to conceive protein-protein interactions structurally and be able to predict those proteins that might potentially bind to each other.
Many protein-protein interfaces are permanent and the polypeptide chains remain bound to each other throughout their functional lifetime. The complex between β and γ-subunits of hetero-trimeric G-proteins forms a classical example. There are several examples of dimeric enzymes, such as triose phosphate isomerase, in which the interface formed between the subunits can be considered permanent. We refer such interfaces and subunits as obligatory.
On the contrary, there also exist protein-protein complexes that are transiently formed and the proteins detach from each other in specific biological situations. The overall structures of these proteins are stable in the unbound form and as they bind to each other. Conformational changes are possible in one or both the proteins as they switch between bound and unbound forms. There are several known examples of this kind. One of the examples is the complex formed between cyclin and cyclin-dependent protein kinase. Such complexes can be deemed as non-obligatory interactions and they act as switches and bring about regulation of a number of proteins in the pathway in which they occur.
Physical interaction between proteins is viewed best as interactions between structural domains as a domain is often the minimal module corresponding to a biochemical function. However the interacting domains could arise from the same polypeptide chain or different polypeptide chain. Inter domain interfacial properties, with both the domains arising from the same polypeptide chain, are observed to be intermediate to homodimeric (inter-chain obligatory) and non-obligatory complexes [2]. Inter chain protein-protein interaction may be formed between two identical chains (Homodimers) or between two different chains (hetero complexes) or it could be a special category of association such as antigen-antibody complexes. Each of the above types of interfaces is ranked based on the chemical and geometrical parameters and it was detected that a single parameter could not be used to definitively distinguish one type of interface from rest of the tertiary surface [3].
The interactions of non-obligatory components can be transient or weaker (compared to interactions between obligatory subunits), although specific. Weak interactions may exist also in the interfaces of those protein homooligomers that are known to exist as both monomer and oligomer at physiological conditions. Transient interactions are those that are more stable and the association or dissociation process requires a molecular trigger. The interfacial property analysis on such a dataset revealed that there exist distinct physicochemical and geometrical properties between these two types of transient complexes [4].
Ofran and Rost [5] classified protein-protein interaction complexes from the protein data bank (PDB) [6] in conjunction with Swissprot [7] into various categories such as the interfaces formed between domains within a polypeptide chain, homo and hetero types of obligatory and non-obligatory complexes. They detected characteristic amino acid compositional preference for each type of interfaces so obtained. Thus, it is generally expected that obligatory and non-obligatory subunits may be characterized by distinct physico-chemical properties.
Bahadur and coworkers [8] analyzed a set of homodimers and reported that the interfacial properties such as the interface area and the hydrophobicity of the interface of these homodimers are distinctly different from that of the protein complexes formed after the subunits fold into tertiary structure.
Attempts have been made to distinguish protein-protein interaction contacts from crystal contacts. It has been noticed that the protein-protein interaction sites are generally larger in surface area than non-specific crystal contacts [9]. This distinction alone however, is not accurate in clearly demarcating the specific and non-specific contacts. In addition to interface area, the conservation of interfacial residues seems to be more robust descriptor to distinguish the crystal contact and specific protein-protein interaction contact [10]. In a recent analysis on a selection of 122 homodimeric protein-protein interaction complexes and 70 protein-protein complexes (representing nonspecific interactions) it has been noticed that the residue propensity and hydropathy along with interfacial area at the contact surface helps in distinguishing the two types of interfaces [11].
Taking lead from the earlier work, the present analysis aims at addressing more complex problem of recognition of the obligatory and non-obligatory complexes from the PDB. Much of the earlier reported analyses relied on datasets derived from hand picked cases from the PDB or are concentrated on homodimeric protein-protein complexes for representation of permanent (obligatory) complexes. We try and overcome both these drawbacks by devising a homology-based method to identify non-obligatory complexes. The current analysis takes advantage of the fact that PDB [6] is a storehouse of all known protein structures and the present multimeric population of proteins in the PDB reveals that there exists a large repertoire of monomeric and oligomeric structures [12]. We have formed a non-redundant data set of obligatory and non-obligatory chains identified by analyzing various protein-protein interaction complexes. The steps involved in the formation of our dataset help us in the identification of homo and hetero protein-protein interaction complexes clearly distinguished as obligatory or non-obligatory. We then analyze various interface properties with an objective to interpret the differences at the interface level that would help in identifying and distinguishing the obligatory and non-obligatory interactions starting from the 3-D structures of protein-protein complexes.
Glaser and co-workers [13] have analyzed residue contact preferences in a dataset of 621 protein-protein complexes. They observe that all types of contacts are observed at the interfaces, especially the hydrophobic contacts. They also report that large interfaces are more abundant in non-polar contacts and small interfaces are abundant in polar contacts. In present study we observe that obligatory and non-obligatory residue contacts across interface together makes up for the general trend. The highest number of contacts between residues across interface that is reported for obligatory homo oligomer is between identical residues [5]. The hetero obligatory complexes and both types of transient oligomers (homo and hetero type) have predominance of polar interactions. Frequency of non-polar contacts reported here are generally lower, when compared to polar contacts [5]. Complementary to earlier studies we not only study the residue-residue interactions across interfaces, we also analyze the local secondary structures involved in interface formation and contribution of main -chain atoms to the interactions.
The rules thus derived can be applied to protein engineering projects aimed at stabilizing otherwise weak interactions as reported in the case of monomeric protein L [14] or destabilize the interactions. This notion regarding the nature of interaction can guide in designing small molecules that might disrupt associations between two proteins and probably have the potential to act as drug molecules. Our results could also help in predicting obligatory or non-obligatory nature of a polypeptide chain that is seen to form protein-protein interactions in the experimentally determined 3-D structures of protein assemblies.
Results and discussion
The basis of our classification of protein chains as obligatory and non-obligatory relies on the fact that an obligatory chain depends on binding of another chain for maintenance of overall structure, stability and function. Thus one would expect that in the PDB, the three dimensional structures of obligatory chains are deposited with its partner chain. Further, availability of crystal or NMR structure of an obligatory chain not bound to another polypeptide chain is not expected. A simple-minded PSI-Blast [15] based homology search is implemented here to distinguish between the different types of chains.
Present dataset is derived from a recent release of PDB that contained tertiary structure for 47,550 chains. This set was subjected to a number of filters to obtain a dataset of obligatory and non-obligatory protein complexes.
A chain and its close homologues with no independent (unbound) tertiary structures is deemed obligatory chain and the region on its tertiary surface that makes the contact with the other chain in the complex is said to be obligatory interface. Those complexed chains that have clear sequence similarity to a monomeric structure are deemed non-obligatory chains and the interface such a chain makes with its partner chain in the complexed state is considered as non-obligatory. We have manually scrutinized initial lists obtained, by these considerations, by referring to the literature and removed erroneous and doubtful cases from the final dataset for analysis. The final dataset contains 82 obligatory chain entries and 30 non-obligatory chain entries. These examples are listed in Table 1. While this list is unlikely to be comprehensive, since we have removed doubtful cases, the entries in the final dataset are often either clearly obligatory or non-obligatory. All analysis presented here are performed on the observed protein-protein interfaces of each of the obligatory or non-obligatory chain and its partner chain. The interfaces are defined by identifying those residues that show significant change in the solvent accessibility upon complex formation.
The stringent criterion identifies only those interfacial residues that show large changes in accessibility on complex formation and are located at the center of the interaction patch (which is confirmed by manual inspection), and are shielded from the solvent molecules in the bound state.
The generous criterion, on the other hand, is a lenient measure and identifies all the residues that show variation in accessibility, however small it is, between complexed and free forms. This criterion identifies all the residues in the interface. The residues that are located in the central region of the interface as well as the residues in the surrounding region (periphery of the interface patch) are also picked up by this scheme of defining interface. The interfacial residues at the periphery of the patch are not completely buried and remain solvated even in the bound form.
Residue propensity
Propensities of individual residues to exist in the protein-protein interface are calculated for both obligatory and nonobligatory interfaces for the entire interface as well as for core of the interfaces and the results are shown in figure 1. The propensity values aid in identifying the residues commonly occurring at the interface and also reveal the chemical nature of the interfaces. A propensity value greater than 1 indicates that the frequency of occurrence of that amino acid at the interface is higher than rest of the domain surface.
The comparative graph of residue propensities for obligatory and nonobligatory interfaces plotted for core and entire interface shows that center of the obligatory interfaces have high propensity for mainly nonpolar residues such as Ile and Met although other hydrophobic residues are also preferred. Other than a slight preference for Trp, Ser and His, no other polar residue is found to occur preferentially at the center of obligatory interface patch. This observation can be explained by considering the residue-residue contacts and are analyzed as described in a later section. These residues appear to strengthen binding at the interface by aromatic and polar interactions.
The core of non-obligatory interface has a large frequency of occurrence of short non-polar residues and aromatic residues such as Tyr. Interestingly, polar residues such as Arg and Gln show a slight, preference for occurrence at the core of the non-obligatory interface. High propensity of short nonpolar residues such as Leu, Val can provide the necessary flexibility in transient interaction in case of non-obligatory interfaces and polar residues can bring about the necessary strength and specificity. A slight preference for Pro to occur at non-obligatory center is noted. Pro is a structurally constrained residue, and shows feeble participation in regular secondary structures, and is chemically non-polar, contributing to hydrophobic interaction in protein structures. Thus Pro is expected to favor non-obligatory interfaces, providing irregular regions and turns, which is a hallmark of non-obligatory interfaces.
Trp is seen to be present in both types of interfaces, both at the center and at the periphery. The relative occurrence of Trp in proteins is small [16], and it is also known to be well-conserved [17]. Evidence suggests that Trp is the most favored residue as the interaction hot spots [18]. Hot spot residues are those residues that contribute maximally to the binding energy. It can thus be postulated that it plays a role in domain-domain recognition in transient interactions apart from balancing the positively charged Arg by cation-π interaction [13]. This could explain the larger propensity of Trp in non-obligatory interfaces, where recognition of the interface patches becomes essential for association and dissociation steps during the course of the functionality of the protein. In obligatory interfaces, Trp could be assisting in interface formation by virtue of burial of its large surface area upon complex formation.
It is surprising to note the high propensity of occurrence of Cys in interfaces. Cys may be considered a weakly polar residue when it is not involved in disulfide formation. Compared to Lys, Arg has higher probability of occurrence at interface. The acidic residues are not preferred at the interface in comparison to domain surface. The detailed picture of residue contacts observed in the dataset is discussed in a later section.
Polar residues are primarily picked up for both obligatory and nonobligatory interfaces, when both the center and periphery of the interface (entire interface) is considered. From the figure 1 it becomes clear that polar residues such as Thr and Tyr show roughly equal tendency to occur in interfaces of both types- obligatory and non-obligatory. It can be suggested that the polar residues are more frequent at the periphery of the interfaces than at the center. Also, since polar interactions are directional they may play a role in maintaining specificity of the interaction.
The analysis of overall hydrophobicity of the interface for the core and entire interface (as plotted in figure 2) reveals that the centers of obligatory and non-obligatory interfaces are predominantly apolar. Additionally, the periphery of both the types of interfaces is more polar in nature as compared to center. Similar observation was made, in an earlier analysis on a general dataset of protein complexes [19]. This could result in favorable interactions of the residues at the periphery of the interface with the solvent. Interestingly, the non-obligatory interfaces are more polar in nature as compared to obligatory interfaces when the center and periphery of the interfaces is compared, probably because the interfacial residues interact with solvent when the non-obligatory pairs exist as tertiary structures not bound to each other.
Residue contacts at the interfaces
The residue contact analysis is aimed at identifying the pairing pattern of the local regions and interacting residues across the interface. Only the contacts made by the obligatory or non-obligatory chain to the interface formation is considered here. The contribution of partner chain is not considered here unless, it is also present in the dataset of either obligatory or non-obligatory class. The contacts made by each chemical group of residue from the obligatory or non-obligatory chain with another chemical group from the partner chain are considered for the contact matrix generation. Both main-chain and side-chain atoms are considered for the analysis. The interactions were broadly classified into polar and nonpolar. Inter-subunit disulphide links are rare and we observe only three disulphide bridges across the subunits of obligatory chains in the present dataset.
The results are summarized in figure 3. All the interactions observed are normalized and are color coded, with lighter shades indicating fewer contacts observed and darker shades indicating larger number of contacts observed. The values of the observed contacts between different residues that are colour coded vary between 0 and 14 and colour intensity increases in discrete steps (white color indicates no interaction). These values represented in the matrix are the normalised values of the observed number contacts for the obligatory and non-obligatory chain. (Materials and Methods section covers the details of calculation of the values represented in the matrices). The extent of non-polar contacts observed for non-obligatory examples is shown in figure 3a and that observed for obligatory cases in depicted in figure 3b. Similarly, the extent of polar contacts observed for non-obligatory examples is shown in figure 3c and that observed for obligatory cases in depicted in figure 3d. The residues that contribute to the contacts made by obligatory or non-obligatory chain is shown along the rows, and the partner chain residues are shown along the columns.
Comparison of apolar contacts (as shown in figure 3a and 3b) between obligatory and non-obligatory complexes reveal that the contact frequency is marginally higher in the case of obligatory complexes. The cumulative values of normalised contacts are 1229 and 1215 respectively for obligatory and non-obligatory chains. This observation is suggestive of the fact that the obligatory interfaces are dominated by apolar contacts. The cumulative contribution of non-polar residues to the contacts in obligatory is 49.3% of the total apolar contacts, while the contribution of non-polar residues to in contacts made across interface in non-obligatory complexes is 42% of the total apolar contacts. The normalised average contacts made by non-polar residues like Leu, and Phe is higher in case of obligatory chains (8.73 and 5.3 contacts respectively) as compared to non-obligatory chains (6.29 and 4.4 contacts respectively). In case of non-obligatory chains, weakly polar residue such as Cys and Thr contribute to non-polar contacts. The average contact values for Cys and Thr are 1.87 and 3.59 contacts respectively for non-obligatory chains and 0.68 and 2.37 contacts respectively for obligatory chains. Though an isolated van der Waal's contact is weak, large numbers of such contacts can have a collective effect and could contribute to large binding energies. Similar effect could also contribute to stable bound states of the obligatory interfaces.
The residue contact matrix for polar interactions across interface for non-obligatory and obligatory interfaces (as shown in figure 3c and 3d) reveals that the polar contacts are formed mainly between side chains of polar or charged residues for both obligatory and nonobligatory interfaces. However, the main chain amide and carbonyl groups are also seen to contribute to some of the polar interactions shown in the matrix.
The polar interactions that the non-obligatory chains make with their partner chain are represented in figure 3c. We observe that larger number of polar contacts are made by the non-obligatory chain when compared to obligatory chain. The cumulative numbers of normalised contacts are 378 and 391 for obligatory and non-obligatory chains respectively. The polar groups of polar side chains primarily contribute to the contacts in the non-obligatory chain. For example, on an average, 2.62 and 2.39 contacts are made by acidic residues, Glu and Asp present in interfaces of non-obligatory chains. However, Glu and Asp makes only 1.69 and 1.63 average contacts respectively in interaction mediated by obligatory chains. Polar contacts observed in obligatory chains are more distributed, in terms of residue involvement, when compared to contribution from non-obligatory chains. Interestingly, polar atoms in the main chain of the residues mediate a large section of the obligatory polar contacts.
Comparing all the contact matrices in figure 3, we observe that obligatory interactions show extensive apolar contacts and the polar contacts in the obligatory interfaces are largely mediated by main chain atoms. Polar atoms of the polar residues on the other hand mediate non-obligatory polar contacts.
Interfacial residue propensity of Thr is slightly higher for non-obligatory chains, and the interfacial propensity of Cys is higher for non-obligatory chains when compared to obligatory chains (as can be visualised from figure 1). Both these residues can be considered as weakly polar and on dissociation of the non-obligatory chains, it would be favorable for them to interact with solvent. Thus nature has carefully designed the non-obligatory interface, with precise balance of polar, non-polar and weakly polar residues.
Involvement of Arg, Tyr and Cys in contacts at interface
Interestingly, the involvement of Arg in polar and nonpolar interaction in both non-obligatory and obligatory interfaces is significant. The long nonpolar part of the side-chain of Arg is observed to interact with large nonpolar residues. The interaction of Arg with aromatic side chains indicates the involvement of cation-π interaction. Propensity analysis indicates that aromatic resides are found to be abundant in interfaces, and specifically Tyr is frequent in non-obligatory interfaces.
Tyr is a special case, as it can contribute both to aromatic and polar interactions. On the other hand, the center of nonobligatory interface consists of both polar and nonpolar residues. An interesting observation is the high propensity of Arg at the center of non-obligatory interface. Probably, the ability of Arg to take part in polar as well as in nonpolar interaction using its long nonpolar side chain or by cation-π interaction with phenyl ring of aromatic residues assist in formation of nonobligatory interface significantly. It favors Arg to interact with the solvent (water) in unbound state, and on the other hand, in the complex form, Arg can potentially interact with all types of polar, non-polar or aromatic residues by virtue of the carbon atoms in the side chain, and the positively charged guanidino group.
Similary, it is surprising to find high prevalence of Cys at the interfaces. Cys may be considered as weakly polar if it is not involved in the formation of disulfide. From the figure 3a, we infer that Cys does participate in apolar contacts. Interaction of sulphur with aromatic groups in proteins has been reported. [20,21]. Such a possibility of Cys interacting with aromatic ring systems is raised (P. Chakrabarthy, personal communication).
Secondary structure analysis
The secondary structures at the interfaces are classified as helix (H), β-strand (E) and others such as turns and loops both collectively represented (T). The conformation of the interfacial residues contributed by both obligatory and non-obligatory chains falls into all the three above-mentioned classes.
In obligatory interfaces 45.8% of total interface residues were involved in helix-helix interaction while only 31.3% of total interface residues are involved in helix-helix interactions in case of non-obligatory interfaces. Thus, interactions between two helices were noticed in both obligatory and non-obligatory types of complexes.
Non-obligatory interfaces have higher involvement of irregular secondary structural region (either defined as turns 'T' or as unassignable). 12 and 37.4% of the total interface residues in case of non-obligatory and 9.1% and 16.9% of the total interface residues in case of obligatory complexes are observed to form turns or irregular secondary structures. This probably provides the necessary flexibility to the interface to favor the interacting subunits to dissociate under appropriate conditions.
While examining the examples of non-obligatory interactions we found no instance of β-sheet formation across the two subunits at the interface. On the other hand in the case of obligatory interactions, out of the 28.3% of total interfacial residues participating in formation of strands 3.4% of it were detected to form inter-subunit β-sheet. Only 19.3% of total interfacial residues from non-obligatory class contribute to strands at interface. Such β-sheet formation across interface makes the complex formed very stable, and in such examples, polar contacts are the driving force in interface formation, and non-polar contacts are less prominent.
Hence it can be inferred that the involvement of secondary structures elements for interface formation is more characteristic of obligatory surfaces (P value < 0.05 for the involvement of helix as well as β-sheet at the interface).
The interaction between the secondary structures especially the β-sheet formation is mediated by the interaction between the main chain atoms. We quantified the main chain-main chain (MC-MC), main chain-side chain (MC-SC), side chain-side chain interactions (SC-SC) in both cases of obligatory and nonobligatory interactions. The extent of MC-MC (16.9% of total contacts in case of obligatory and 11.2% of total contacts in case of non-obligatory) is the most distinguishing between two types of interfaces when contacts are considered at atomic level. The values obtained for MC-SC (about 42.6% of total contacts in case of obligatory and 49.3% of total contacts in case of non-obligatory) and SC-SC (40.5% of total contacts in case of obligatory and 39.6% of non-obligatory) are mostly comparable. However, we note that the main chain involvement is clearly higher for obligatory examples (P value < 0.15 using t-test).
Interface area distribution
Interface areas are calculated for both obligatory and non-obligatory protein complexes and the results are summarized in figure 4. The plot presented in figure 4a is the frequency of absolute interface areas for both types of interfaces. The average interface area in case of obligatory interfaces is 492.74 Å2 and in the case of non-obligatory complexes it is 279.55 Å2. From the plot given in Figure 4a, we observe that the obligatory interface has a higher mean value and a broader distribution in raw interfacial area. This implies that the nonobligatory interfacial areas are generally smaller (P value < 0.05 using t-test) and this translates to less strong interaction that might help in making the interaction transient. This point is further validated by considering the average number of contacts per interface in the two cases. The obligatory complexes make 20 contacts per chain on an average whereas non-obligatory complexes make 13 contacts per chain. The number of contacts per chain can be taken as a rough measure of the strength of interaction. In the dataset derived in this work, the number of contacts in obligatory interfaces is shown to be significantly different than the average number of contacts made by non-obligatory interfaces (P value <0.05 using modified t-test). It must be pointed out that, even though the number of contacts seen is different in the two cases, the contact density is similar (0.08 contacts / Å2 in case of obligatory complexes and 0.06 contacts / Å2 in the case of non-obligatory complexes). This means that the interfacial packing density is not different in the two cases.
Viewing interfacial areas as a fraction of total domain surface area, we observe 41% of the non-obligatory interfaces occupies ≤ 2% area of the domain surfaces. However, in the cases of obligatory interfaces, there is an even distribution of examples between 0–6% area of domain surface, with 80% of examples in this range (figure 4b). The surface areas of the domains considered here are large, hence, we observe that the interface occupation on tertiary domain surface is small. However, their absolute areas in these are mostly comparable with the other obligatory interfaces in (Å2).
Among the cases of obligatory complexes, there are instances of huge multi-subunit protein machinery like proteosome (1ryp_1), where single interface (formed between two subunits of multi subunit complex) occupancy is very low on the total domain surface. Most of these examples correspond to those proteins that have small domain surface area and part of a large multi-domain complex.
Analysis of the topology of the interfaces
Shape complementarity of the interface using SC-program of Lawrence and Colman [22] for the pair of interacting proteins. Overall both type of interfaces showed a robust clustering of shape complementarity value within a range of 0.6 to 0.8. This implies that the geometrical complementarity at the interfaces of both types of complexes is similar. The average shape complementarity value for non-obligatory interfaces was 0.649 while it is 0.686 for obligatory interfaces. Thus it appears that overall the obligatory interfaces have slightly better shape complementarity though the difference between the obligatory and non-obligatory types of interfaces is very small (P value > 0.4 using t-test).
Conclusion
We have arrived at a set of protein complexes from the PDB, classified in broad terms as belonging to obligatory or non-obligatory categories using a simple sequence analysis based procedure. The assignment of obligatory or non-obligatory nature is restricted to the chain level and the interaction interface of this chain.
Present analysis is attempted to find the distinguishing features of the two types of interfaces. While nonpolar contacts dominate the interaction interfaces, especially the obligatory interactions, the polar interactions are also observed in interfaces. The polar interactions are mediated by hydrophilic sidechains in the cases of non-obligatory interactions probably provides favorable binding energies, and also helps to stabilize the tertiary structure when the complex dissociates. On the other hand main chain polar groups have a substantial representation in the obligatory interfaces.
Non-obligatory or transient interactions are likely to be characterized by optimal binding energy, so that the complex can be disassembled into its constituent elements upon a molecular stimulus. Additionally, the interacting subunits of the non-obligatory complex interacts with polar solvent in the uncomplexed form and hence the non-obligatory interfaces are less hydrophobic. While we have not made an explicit analysis of role of water in protein-protein interfaces, it is expected that water can interact favorably with polar amino acid residues at the interface. Indeed a water molecule is observed to contribute to polar contacts between two macromolecules [23].
The β-sheet formation across the interface is a feature seen in the case of obligatory interfaces. However, none of the non-obligatory cases analyzed here has this feature. The main chain contribution to the interface is clearly more prominent in obligatory interfaces.
Covalent association in the form of disulphide bridges between the subunits is a feature of obligatory complexes. The covalent associations make the protein-protein interaction permanent. There are exceptions to this rule, like the case of the type II ribosome inactivating plant toxins where the toxic chain is covalently linked via a disulphide bond to a carrier lectin moiety. The disulphide bond reduction and release of toxin is an essential step for biological activity of the protein. Here although the plant toxin- lectin association is non-obligatory, a covalent association of the subunits is observed making the complex formed between the two proteins extremely stable and are dissociated only under specialized conditions.
The key feature of obligatory type of interfaces is its stability of association. This stability is achieved in a number of systems in diverse manners. For example, if extensive β-sheet formation is the strategy adopted to form obligatory interface in a certain protein, for example the lectins, then it is observed that the interfaces can be more polar than the generally observed trend, and the protein-protein interaction is sensitive to pH variations [24]. In such cases, the residue propensities do not obey the general rules. For this reason, the test data set considered here has marked deviations in the residue propensities compared to the original dataset.
Thus from the carefully chosen set of obligatory and non-obligatory complexes, the analysis shows distinction between obligatory and non-obligatory interfaces in terms of some of the features such as patterns of interaction across the interfaces. There is a clear trend for the obligatory interfaces to be larger in area, the center of obligatory interface to be non-polar, and to involve stable secondary structural elements across the interface. Since the variations between different types of interfaces are subtle, a single feature cannot be reliably used to predict different types of complexes. However a cumulative effect of all these features can aid recognizing obligatory and non-obligatory interfaces. The results of statistical tests on various features suggest that differences in only some of the features are statistically significant. However our analysis provides an indication of the trend which may be strengthened by the accumulation of more 3-D structures of protein-protein complexes.
Thus, a combination of above said features, when considered concurrently and appropriately weighed can add value to the prediction of obligatory and non-obligatory interaction sites on the tertiary surface. Such an approach is shown to be successful using a test dataset not used in the original analysis.
Association of wide variety of proteins mediates many vital cellular processes. To be able to model the tertiary and quaternary structure from the primary structure is the goal of comparative modelling approaches [12]. Such problems are best addressed by considering the structural information of a homologous protein, since it is observed that protein-protein interaction sites are evolutionarily conserved among close homologues [10,11]. However, in cases where the information on association cannot be directly derived based on homology, the present analysis can aid in determining the nature of the interface. This information about the nature of interface formed gives an indication of the stability of the complex.
The results presented in this paper can also be useful in distinguishing the obligatory and non-obligatory types of interfaces observed in structures determined in large-scale in the structural genomics initiatives, especially for those multi-component protein assemblies for which the biochemical characterization is incomplete.
Methods
Dataset generation
The co-ordinate sets of protein structures used in the analysis were extracted from the April 2003 version of PDB . All nucleic acid, hetero-atoms, small peptides (<30aa) and extremely large chains (>1000aa) were excluded from this raw set. All other protein chains were retained and taken for further analysis.
The polypeptide chains in many of the protein-protein complexes could be classified into one of obligatory or non-obligatory subunit. Each entry in the PDB is classified as monomer or multimer depending on number of chains in the structure and by consulting the Protein Quaternary Structure server [9]. Each chain from the monomer data set was searched for homologues in the multimer chain set using single round of PSI-blast run using 0.01 as E-value and 0.001 as inclusion value (h-value) [15]. The protein chains in monomeric set that have homologues having 95% sequence identity over 90% of the monomer chain length is assumed to exist in monomeric form as well. This means that being in the oligomeric state is not mandatory for their structure and functionality. Subsequently their interactions with physically adjacent protein chains within the same protein were considered as nonobligatory interaction. On the contrary, in case of obligatory entries, it was assumed that the state of oligomerisation is essential for the structure and functionality of the subunit. So the protein chains that are not nonobligatory were considered to be probably obligatory and their corresponding interactions with neighboring subunits were also considered obligatory. e.g. in G-protein the β and γ subunits are always bound to each other whereas α-subunit alternate between bound and unbound forms with β,γ subunits depending upon if GTP or GDP is bound to the α-subunit. So we consider α-β interactions are nonobligatory while β-γ binding is an obligatory interaction. The interactions between polypeptide chains within a protein that have no physical contact between them were kept out of consideration.
The polypeptide chain that is considered for search against the monomer dataset is called the representative chain and the chain with which it physically interacts is called the partner chain. The obligatory or non-obligatory nature of the interaction is defined with respect to the interaction of representative chain with its partner in the crystal structure. The obligatory or non-obligatory nature of interaction is restricted to the interactions contributed by the representative chain.
Crystal structures having resolution better than 2.5Å and the best model of NMR structures were considered for the analysis. To avoid redundancy in the dataset, Only those representative structures that had lesser than 25% sequence identity to other structures in the dataset were selected for the analysis. We consulted the PDB-Select [25] that gives a listing of non-redundant collection of PDB structures.
Due to low occurrence of monomeric homologues in the PDB, the entries classified as obligatory interactors purely based on an automatic procedure as described above have high scope for contamination. Hence, we consulted other sources, especially the literature to retain only those entries for which the biologically relevant oligomeric state was clearly and explicitly mentioned. Those cases which are either unclear or lack sufficient information or ambiguous are excluded from the dataset for analysis. We have consulted the published works of Lo Conte and co-workers [26] and Nooren and Thornton [4] and have incorporated entries from their reported datasets in appropriate classes, in case we failed to identify them by our automated procedure.
Antigen-antibody complexes behave as non-obligatory complex in the unbound state; but the binding of the antigen with the antibody is associated with large energy of binding and thus, the interaction can be considered as obligatory. Thus, antigen-antibody complexes are deviant from our definition of obligatory and nonobligatory complexes and hence we did not include antigen-antibody complexes in the present analysis. Similarly, integral membrane proteins have intrinsic amino acid preferences so that they could be accommodated in a hydrophobic environment – the cell membranes. These entries were weeded out from the dataset.
Defining interface
Inter-chain interfaces were defined following the accessibility changes. A residue is said to be at the core of protein-protein interaction interface if its accessibility values show large variation between exposed (>10%) and buried state (<7%) upon oligomerisation (dimerisation) with the corresponding interacting subunit. This method identifies those residues that are almost fully buried in the complex state and well exposed in the uncomplexed state. Mainly the residues that are at the center of the interface are picked up by this method.
On the other side, the residues at the surface (ASA >7%) which lose solvent-accessible surface area by >1Å2 on oligomerisation, are also considered to be at the interface. This encompasses a larger number of interfacial residues than the number of residues at the core of the interface. Thus using this criterion, the residues are picked up over a broader area or in other words, the residues at the center as well as the periphery of the interface are picked up. The notion of the location of interacting residues in the center and periphery of the interacting region on the surface of protein complex was confirmed by visual inspection on a number of cases.
Residue propensity
We analyzed residue propensity at the interface to study the preference of the amino acid residue to occur at the interface with respect to the preference of the residue to occur at the surface of protein at the domain level. We referred to the SCOP [27] for the definitions of the composite structural domains of the polypeptide chains.
The residue propensity was defined as -
P(int)i = N(int)i / N(surf)i
Where, P(int)i = Propensity of ith amino acid at the interface
N(int)i = Normalised number of ith amino acid at the interface
N(surf)i = Normalised number of ith amino acid at the domain surface
Propensities for each of the 20 residue types were calculated for both obligatory and nonobligatory interfaces following both core and entire interface. The propensity analysis reflects the residue preferences of the interfaces and also reveals the chemical nature of the interfaces and kind of interactions present.
The hydrophobic nature of the interfaces was studied by a hydropathy analysis using standard Kyte and Doolittle scale [28]. Hydrophobicity value is calculated as-
Hydrophobicity value = hydrophobicity index * residue propensity.
The analyses were done both for the core and entire interfaces for both obligatory and non-obligatory types of interfaces.
Residue contacts
One of the objectives of the present analysis is to discern the residue level interactions across the interacting and representative chains. This information is very crucial since it can reveal the nature of interactions and residue pairing preferences across interface.
The residue interactions were classified broadly in 3 groups -
i. Covalent bond forming: Disulphide linkages
ii. Electrostatic and H- bond forming: Polar (salt bridge and H-bond) interactions,
iii. Van der Waal interactions: Nonpolar interaction, interactions involving aromatic ring systems
The polar interactions were considered between uncharged polar as well as charged groups. Hydrogen bonds considered here are formed when the hydrogen associated with nitrogen is shared with acceptor oxygen of carbonyl or carboxyl group. Other hydrogen bonds occur under special geometrical and chemical constraints and are weaker than the above said class. Hence their involvement is not considered here. All types of polar interactions are significant when the N and O are at a distance between 2.4–3.4Å. The apolar interactions were considered to be significant only when the deviation of the sum of the Van der Waal radii of the two atoms is within 1Å distance. Covalent interactions like the disulphide bonds can also be formed at the interface although they are not common at the protein-protein interfaces. The disulphide bonds are considered to exist if the sulphur atoms of the two Cys residues from the interacting chains are at a distance = 2.1 Å. Disulphide linkage provides rigidity and stability in the interaction as compared to electrostatic and Van der Waal interaction.
The inter-chain interacting residue-pairs were picked up on the basis of the kind of interactions they are involved-in. We have classified all the pairwise residue interactions into polar or nonpolar. For a single pair of residues, polar and disulphide interactions were given more priority than Van dar Waal interaction i.e. if a pair of residues present all the types of contacts – polar and non-polar contact, the residue-interaction was considered to be primarily polar and this pair is not considered for its contribution to apolar contacts. The interaction data was classified in polar and nonpolar interactions and presented in the form of 20 × 20 matrices with matrix elements represent the normalized frequency of occurrence of interaction between the residue-pairs. The normalisation was done to account for the disparity in the dataset sizes for obligatory and non-obligatory complexes and also to account for the higher interfacial areas observed in obligatory chains. This normalization ensures that the values obtained for the obligatory and non-obligatory contacts are comparable.
Interface area
The interface areas for obligatory and nonobligatory interfaces were calculated following core of interface and were represented both in absolute (Å2) and as percentage of the total domain surface occupied by the interface. Surface residues were identified if the percent accessibility is >10%. The results were classified according as the fraction of the dataset that have an interface area within a specified range.
Secondary structures
Secondary structures of the interacting and representative chain were identified using SSTRUC software developed by David Keith Smith (1989, unpublished data) based on the DSSP algorithm [29]. The secondary structures were considered mainly in the broad grouping of Helix (H), B-strand (E) and others (T). A secondary structural element was deemed to be present at the interface of the representative chain, if a secondary structural element contributes at least two residues for interface formation. Similar contributions of helix, strand or loops were calculated for the partner chain also. We then determine the percent of interfacial residues that participate in interactions between the secondary structures across the interface. In the case of extended strands, the possibility of formation of β-sheets across the interface was analyzed by considering potential main chain main chain interactions.
Shape complementarity
Another geometrical measure, the shape complementarity for interacting chain-pairs was determined by SC program developed by Lawrence and Colman [22]. This program calculates the geometrical packing at the interface between two chains and determines how well the interacting surfaces of the protein complex complement one another. Higher value indicates good geometric complementation, while small values generally indicate bad complementarity.
Statistical analysis of results
Test for the statistical significance of the results was done using the students t-test. The mean and the variance in the parameter under study of the complexes in obligatory and non-obligatory complexes were calculated. If the variances were found to be significantly similar using the F-test, the normal t-test was used with pooled variances. If the variances were not equal, then a modified t-test with adjusted variances was used. The test statistic in each case was tested at the 0.05 level of significance.
The t-test statistic used to compare the means of absolute interface areas and shape complementarity analysis is given in equation (1) below.
Eq (1) where, Sp is given by
The t-test statistic used for the statistical analysis of main-chain main-chain contact analysis, secondary structure composition analysis and residue contacts per interface is given in equation (2) given below.
Eq(2)
Authors' contributions
NS conceived of this study, SD generated the general dataset and carried out the analysis. NR and OK refined the general dataset and carried out the analysis. All four authors have read and approved the manuscript.
Acknowledgements
We would like to thank Ms. Sujatha S for her valuable inputs on deriving the data set from the PDB. We thank Mr. Kiran Kulkarni who helped in setting up the shape complementarity program of CCP4. This research is supported by the Wellcome Trust, UK in the form of International Senior Fellowship in Biomedical Sciences to NS. NR is a recipient of fellowship from the CSIR, India. SD received summer training fellowship from JNCASR, India.
Figures and Tables
Figure 1 Plot showing the residue propensity in obligatory and non-obligatory interfaces. Propensities of the residues in the entire interface as well as at the core of the interface are shown.
Figure 2 Hydropathy plot for the interface patch. The figure shows hydropathy plot for both obligatory and non-obligatory interfaces and for the residues in the core of the interface and in entire interface.
Figure 3 Residue contact matrix showing the frequency of the contact between two residues at the interface. a: Non-obligatory non-polar interactions, b: obligatory non-polar interactions c: non-obligatory polar interactions d: obligatory polar interactions. The residues contributed by the obligatory or non-obligatory chain of the complex is represented in rows. The residues from the partner chain (for which the assignment of obligatory or non-obligatory is ambiguous) is shown in columns. Colour gradation: white- No interaction; Cyan to black- increasing gradation of interaction with normalised frequency varying between 0–14 in discrete steps.
Figure 4 a: Distribution of the absolute interface area for obligatory and non-obligatory protein complexes in the general dataset. b: Distribution of percentage occupancy of interface for obligatory and non obligatory complexes in the general dataset.
Table 1 The dataset of obligatory and non-obligatory polypeptide chains identified from the PDB
LIST OF NON-OBLIGATORY PROTEIN-CHAINS
PDB Representative-chain Partner-chain Resolution Names of proteins
1dx5 I M 2.30 Thrombomodulin, Thrombin Heavy Chain
1gua A B 2.00 Rap1A, C-Raf1
1efu A B 2.50 Elongation Factor Tu, Elongation Factor Ts
1avw B A 1.75 Trypsin, Trypsin Inhibitor
1emv A B 1.70 Immunity Protein Im9, Colicin E9
1ay7 B A 1.70 Barstar, Guanyl-Specific Ribonuclease Sa
1stf E I 2.37 Papain (Cys 25 Carboxymethylated), Papain (Cys 25 Carboxymethylated)
3eza B A NMR Histidine-Containing Phosphocarrier Protein H, Phosphotransferase System, Enzyme I
1ggr B A NMR Phosphocarrier Protein Hpr, Pts System, Glucose-Specific Iia Component
1pyt B A 2.35 Procarboxypeptidase A, Procarboxypeptidase A
1pyt B D 2.35 Procarboxypeptidase A, Chymotrypsinogen C
1fle I E 1.90 Elafin, Elastase
1sgp I E 1.40 Turkey Ovomucoid Inhibitor, Streptomyces Griseus Proteinase B
1ycs A B 2.20 P53, P53
1ycs B A 2.20 P53, P53
1efn A B 2.50 Fyn Tyrosine Kinase, Hiv-1 Nef Protein
1efn B A 2.50 HIV-1 Nef Protein, Fyn Tyrosine Kinase
1tx4 A B 1.65 P50-Rho GAP, Transforming protein rhoa
1tx4 B A 1.65 Transforming protein Rhoa, P50-Rho GAP
1a2k C A 2.50 Nuclear Transport Factor 2, Nuclear Transport Factor 2
1fin A B 2.30 Cyclin-Dependent Kinase 2, Cyclin A
1fin B A 2.30 Cyclin A, Cyclin-Dependent Kinase 2
1ak4 A C 2.36 Cyclophilin A, Hiv-1Capsid
1ak4 C A 2.36 Hiv-1Capsid, Cyclophilin A
1dhk A B 1.85 Porcine Pancreatic Alpha-Amylase, Bean Lectin-Like Inhibitor
1gla F G 2.60 Glycerol Kinase, Factor III Glc
1ydr E I 2.20 C-Amp-Dependent Protein Kinase, Protein Kinase Inhibitor Peptide
2pcc A B 2.30 Cytochrome C Peroxidase, Cytochrome C
2pcc B A 2.30 Cytochrome C, Cytochrome C Peroxidase
1d2z A B 2.00 Death Domain Of Pelle, Death Domain Of Tube
LIST OF OBLIGATORY PROTEIN-CHAINS
PDB Representative chain Partner chain Resolution Names of proteins
1bmq B A 2.50 Interleukin-1 Beta Convertase, Interleukin-1 Beta Convertase
1qdl B A 2.50 Anthranilate Synthase (Trpg-Subunit), Anthranilate Synthase (Trpe-Subunit)
1poi B A 2.50 Glutaconate Coenzyme A-Transferase, Glutaconate Coenzyme A-Transferase
1ihf B A 2.50 Integration Host Factor, Integration Host Factor
1poi B C 2.50 Glutaconate Coenzyme A-Transferase, Glutaconate Coenzyme A-Transferase
1bcr B A 2.50 Serine Carboxypeptidase II, Serine Carboxypeptidase II
1wdc A B 2.00 Scallop Myosin, Scallop Myosin
1wdc A C 2.00 Scallop Myosin, Scallop Myosin
1ryp 1 I 1.90 20S Proteasome, 20S Proteasome
1ryp 1 J 1.90 20S Proteasome, 20S Proteasome
1ryp 1 S 1.90 20S Proteasome, 20S Proteasome
1ryp 1 T 1.90 20S Proteasome, 20S Proteasome
1ryp 1 Z 1.90 20S Proteasome, 20S Proteasome
1ryp 1 2 1.90 20S Proteasome, 20S Proteasome
1kve A B 1.80 Smk Toxin, Smk Toxin
1luc A B 1.50 Bacterial Luciferase, Bacterial Luciferase
1fm0 E D 1.45 Molybdopterin Convertin Factor, Subunit 2, Molybdopterin Convertin Factor, Subunit 1
1fm0 D E 1.45 Molybdopterin Convertin Factor, Subunit 1, Molybdopterin Convertin Factor, Subunit 2
1h2r S L 1.40 Periplasmic [Nife] Hydrogenase Small Subunit, [Nife] Hydrogenase Large Subunit
1h2r L S 1.40 Periplasmic [Nife] Hydrogenase Large Subunit, Periplasmic [Nife] Hydrogenase Small Subunit
1svf C D 1.40 Fusion Glycoprotein, Fusion Glycoprotein
1svf B A 1.40 Fusion Glycoprotein, Fusion Glycoprotein
1hbn B C 1.16 Methyl-Coenzyme M Reductase I Beta Subunit, Methyl-Coenzyme M Reductase I Gamma Subunit
1ycp K J 2.50 Epsilon Thrombin, Epsilon Thrombin
1hfe S L 1.60 Fe-Only Hydrogenase (Smaller Subunit), Fe-Only Hydrogenase (Larger Subunit)
1f95 A B NMR Dynein, Dynein
1a2k A B 2.50 Nuclear Transport Factor 2, Nuclear Transport Factor 2
2thi A B 2.50 Thiaminase I, Thiaminase I
1e5d A B 2.50 Rubredoxin: Oxygen Oxidoreductase, Rubredoxin: Oxygen Oxidoreductase
1b55 B A 2.40 Tyrosine-Protein Kinase Btk, Tyrosine-Protein Kinase Btk
3nos A B 2.40 Endothelial Nitric-Oxide Synthase, Endothelial Nitric-Oxide Synthase
1qfx A B 2.40 pH 2.5 Acid Phosphatase, pH 2.5 Acid Phosphatase
2tmk A B 2.40 Thymidylate Kinase, Thymidylate Kinase
1f37 A B 2.30 Ferredoxin [2Fe-2S], Ferredoxin [2Fe-2S]
1qdn A B 2.30 N-Ethylmaleimide Sensitive Fusion Protein (N), N- Ethylmaleimide Sensitive Fusion Protein (N)
1dcp A B 2.30 Dcoh, Dcoh
1dcp A C 2.30 Dcoh, Dcoh
1dcp A D 2.30 Dcoh, Dcoh
1dcp A E 2.30 Dcoh, Dcoh
1del B A 2.20 Deoxynucleoside Monophosphate Kinase, Deoxynucleoside Monophosphate Kinase
1otg A B 2.10 5-Carboxymethyl-2-Hydroxymuconate Isomerase, 5-Carboxymethyl-2-Hydroxymuconate Isomerase
1bft A B 2.00 Nuclear Factor Nf-Kappa-B P65, Nuclear Factor Nf-Kappa-B P65
1coz A B 2.00 Glycerol-3-Phosphate Cytidylyltransferase, Glycerol-3-Phosphate Cytidylyltransferase
1mka A B 2.00 Beta-Hydroxydecanoyl Thiol Ester Dehydrase, Beta-Hydroxydecanoyl Thiol Ester Dehydrase
1b66 A B 1.90 6-Pyruvoyl Tetrahydropterin Synthase, 6-Pyruvoyl Tetrahydropterin Synthase
1otf A C 1.90 4-Oxalocrotonate Tautomerase, 4-Oxalocrotonate Tautomerase
1otf A D 1.90 4-Oxalocrotonate Tautomerase, 4-Oxalocrotonate Tautomerase
1otf A E 1.90 4-Oxalocrotonate Tautomerase, 4-Oxalocrotonate Tautomerase
1j9l A B 1.90 Stationary Phase Survival Protein, Stationary Phase Survival Protein
1dfn A B 1.90 Defensin HNP-3 – Chain A, Defensin HNP-3 – Chain B
1b93 A B 1.90 Methylglyoxal Synthase, Methylglyoxal Synthase
1b93 A C 1.90 Methylglyoxal Synthase, Methylglyoxal Synthase
1ext B A 1.85 Tumor Necrosis Factor Receptor, Tumor Necrosis Factor Receptor
1kve A C 1.80 Smk Toxin, Smk Toxin
1atl A B 1.80 Atrolysin C, Atrolysin C
1d2v A B 1.75 Myeloperoxidase, Myeloperoxidase
1tvx B A 1.75 Neutrophil Activating Peptide 2 Variant, Neutrophil Activating Peptide 2 Variant
1tvx B C 1.75 Neutrophil Activating Peptide 2 Variant, Neutrophil Activating Peptide 2 Variant
6gsv A B 1.75 Mu Class Glutathione S-Transferase Of Isoenz, Mu Class Glutathione S-Transferase Of Isoenz
1qsg G E 1.75 Enoyl-Reductase, Enoyl-Reductase
1qsg G F 1.75 Enoyl-Reductase, Enoyl-Reductase
1qsg G H 1.75 Enoyl-Reductase, Enoyl-Reductase
1a2z A B 1.73 Pyrrolidone Carboxyl Peptidase, Pyrrolidone Carboxyl Peptidase
1a2z A D 1.73 Pyrrolidone Carboxyl Peptidase, Pyrrolidone Carboxyl Peptidase
1mjh A B 1.70 ATP-Binding Domain Of Protein Mj0577, ATP-Binding Domain Of Protein Mj0577
1dqi A B 1.70 Superoxide Reductase, Superoxide Reductase
1dqi A D 1.70 Superoxide Reductase, Superoxide Reductase
1jqc B A 1.61 Protein R2 Of Ribonucleotide Reductase, Protein R2 Of Ribonucleotide Reductase
1f74 A C 1.60 N-Acetyl-Neuraminate Lyase, N-Acetyl-Neuraminate Lyase
3pvi A B 1.59 Pvuii Endonuclease, Pvuii Endonuclease
1f9z A B 1.50 Glyoxalase I, Glyoxalase I
1jr8 A B 1.50 Erv2 Protein, Mitochondrial, Erv2 Protein, Mitochondrial
1a4i B A 1.50 Methylenetetrahydrofolate Dehydrogenase / Me, Methylenetetrahydrofolate Dehydrogenase / Me
1dvj A B 1.50 Orotidine 5'-Phosphate Decarboxylase, Orotidine 5'-Phosphate Decarboxylase
1qh4 A B 1.41 Creatine Kinase, B Chain, Creatine Kinase, B Chain
2tnf A B 1.40 Tumor Necrosis Factor Alpha, Tumor Necrosis Factor Alpha
3sdh A B 1.40 Hemoglobin I (Homodimer) (Carbon-Monoxy) – C, Hemoglobin I (Homodimer) (Carbon-Monoxy) – C
1i0h A B 1.35 Manganese Superoxide Dismutase Y174F Mutant, Manganese Superoxide Dismutase Y174F Mutant
1dbf A B 1.30 Chorismate Mutase, Chorismate Mutase
1qks A B 1.28 Cytochrome Cd1 Nitrite Reductase, Cytochrome Cd1 Nitrite Reductase
1hbn B E 1.16 Methyl-Coenzyme M Reductase I Beta Subunit, Methyl-Coenzyme M Reductase I Beta Subunit
==== Refs
Stryer L ISBN, USA Signal transduction cascades Biochemistry 1995 4th WH. Freeman and Company, New York 325 360
Jones S Marin A Thornton JM Protein domain interfaces: characterization and comparison with oligomeric protein interfaces Protein Eng 2000 13 77 82 10708645 10.1093/protein/13.2.77
Jones S Thornton JM Prediction of protein-protein interaction sites using patch analysis J Mol Biol 1997 272 133 143 9299343 10.1006/jmbi.1997.1233
Nooren IM Thornton JM Structural characterisation and functional significance of transient protein-protein interactions J Mol Biol 2003 325 991 1018 12527304 10.1016/S0022-2836(02)01281-0
Ofran Y Rost B Analysing six types of protein-protein interfaces J Mol Biol 2003 325 377 387 12488102 10.1016/S0022-2836(02)01223-8
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
Boeckmann B Bairoch A Apweiler R Blatter MC Estreicher A Gasteiger E Martin MJ Michoud K O'Donovan C Phan I Pilbout S Schneider M The SWISS-PROT protein knowledgebase and its supplement TrEMBL in 2003 Nucleic Acids Res 2003 31 365 370 12520024 10.1093/nar/gkg095
Bahadur RP Chakrabarti P Rodier F Janin J Dissecting subunit interfaces in homodimeric proteins Proteins 2003 53 708 719 14579361 10.1002/prot.10461
Henrick K Thornton JM PQS: a protein quaternary structure file server Trends Biochem Sci 1998 23 358 361 9787643 10.1016/S0968-0004(98)01253-5
Valdar WS Thornton JM Conservation helps to identify biologically relevant crystal contacts J Mol Biol 2001 313 399 416 11800565 10.1006/jmbi.2001.5034
Bahadur RP Chakrabarti P Rodier F Janin J A dissection of specific and non-specific protein-protein interfaces J Mol Biol 2004 336 943 955 15095871 10.1016/j.jmb.2003.12.073
Sali A Glaeser R Earnest T Baumeister W From words to literature in structural proteomics Nature 2003 422 216 225 12634795 10.1038/nature01513
Glaser F Steinberg DM Vakser IA Ben-Tal N Residue frequencies and pairing preferences at protein-protein interfaces Proteins 2001 43 89 102 11276079 10.1002/1097-0134(20010501)43:2<89::AID-PROT1021>3.0.CO;2-H
Kuhlman B O'Neill JW Kim DE Zhang KY Baker D Conversion of monomeric protein L to an obligate dimer by computational protein design Proc Natl Acad Sci U S A 2001 98 10687 10691 11526208 10.1073/pnas.181354398
Altschul SF Madden TL Schaffer 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
Creighton TE Chemical properties of polypeptides Proteins structure and molecular properties 1993 2 WH. Freeman and Company 16 17
Gonnet GH Cohen MA Benner SA Analysis of amino acid substitution during divergent evolution: the 400 by 400 dipeptide substitution matrix Biochem Biophys Res Commun 1994 199 489 496 8135790 10.1006/bbrc.1994.1255
Samanta U Chakrabarti P Assessing the role of tryptophan residues in the binding site Protein Eng 2001 14 7 15 11287674 10.1093/protein/14.1.7
Chakrabarti P Janin J Dissecting protein-protein recognition sites Proteins 2002 47 334 343 11948787 10.1002/prot.10085
Duan G Smith VH JrWeaver D Characterization of aromatic- thiol π-type hydrogen bonding and phenylalanine- cysteine side chain interactions through ab initio calculations and protein database analyses Molecular Physics 2001 99 1689 1699 10.1080/00268970110063917
Pal D Chakrabarti P Non-hydrogen bond interactions involving the methionine sulfur atom J Biomol Struct Dyn 2001 19 115 28 11565843
Lawrence MC Colman PM Shape complementarity at protein/protein interfaces J Mol Biol 1993 234 946 950 8263940 10.1006/jmbi.1993.1648
Janin J Wet and dry interfaces: the role of solvent in protein-protein and protein-DNA recognition Structure Fold Des 1999 7 R277 R279 10647173 10.1016/S0969-2126(00)88333-1
Svensson C Teneberg S Nilsson CL Kjellberg A Schwarz FP Sharon N Krengel U. High-resolution crystal structures of Erythrina cristagalli lectin in complex with lactose and 2'-alpha-L-fucosyllactose and correlation with thermodynamic binding data J Mol Biol 2002 321 69 83 12139934 10.1016/S0022-2836(02)00554-5
Hobohm U Scharf M Schneider R Sander C Selection of representative protein data sets Protein Science 1992 1 409 417 1304348
Lo Conte L Chothia C Janin J The atomic structure of protein-protein recognition sites J Mol Biol 1999 285 2177 98 9925793 10.1006/jmbi.1998.2439
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
Kyte J Doolittle RF A simple method for displaying the hydropathic character of a protein J Mol Biol 1982 157 105 132 7108955 10.1016/0022-2836(82)90515-0
Kabsch W Sander C Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features Biopolymers 1983 22 2577 2637 6667333 10.1002/bip.360221211
|
16105176
|
PMC1201154
|
CC BY
|
2021-01-04 16:03:53
|
no
|
BMC Struct Biol. 2005 Aug 16; 5:15
|
utf-8
|
BMC Struct Biol
| 2,005 |
10.1186/1472-6807-5-15
|
oa_comm
|
==== Front
Cardiovasc UltrasoundCardiovascular Ultrasound1476-7120BioMed Central London 1476-7120-3-181608079210.1186/1476-7120-3-18ReviewThe non-invasive documentation of coronary microcirculation impairment: role of transthoracic echocardiography Dimitrow Pawel Petkow [email protected] Maurizio [email protected] Fausto [email protected] 2nd Department of Cardiology, Collegium Medicum, Jagiellonian University, Cracow, Poland2 Division of Cardioangiology with CCU, Department of Clinical and Experimental Medicine, Federico II University of Naples, Italy3 Department of Cardiology Umberto I° Hospital Mestre-Venice, Italy2005 4 8 2005 3 18 18 8 6 2005 4 8 2005 Copyright © 2005 Dimitrow et al; licensee BioMed Central Ltd.2005Dimitrow 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.
Transthoracic Doppler echocardiographic-derived coronary flow reserve is an useful hemodynamic index to assess dysfunction of coronary microcirculation. Isolated coronary microvascular abnormalities are overt by reduced coronary flow reserve despite normal epicardial coronary arteries. These abnormalities may occur in several diseases (arterial hypertension, diabetes mellitus, hypercholesterolemia, syndrome X, aortic valve disease, hypertrophic cardiomyopathy and idiopathic dilated cardiomyopathy). The prognostic role of impaired microvascular coronary flow reserve has been shown unfavourable especially in hypertrophic or idiopathic dilated cardiomyopathies. Coronary flow reserve reduction may be reversible, for instance after regression of left ventricular hypertrophy subsequent to valve replacement in patients with aortic stenosis, after anti-hypertensive treatment or using cholesterol lowering drugs. Coronary flow reserve may increase by 30% or more after pharmacological therapy and achieve normal level >3.0. In contrast to other non invasive tools as positron emission tomography, very expensive and associated with radiation exposure, transthoracic Doppler-derived coronary flow reserve is equally non invasive but cheaper, very accessible and prone to a reliable exploration of coronary microvascular territories, otherwise not detectable by invasive coronary angiography, able to visualize only large epicardial arteries.
==== Body
The concept of coronary flow reserve (CFR) and transmural reduction of CFR
The coronary arterial tree consists of four basic segments. Epicardial coronary arteries give off small transmural penetrating arteries, which have branches in the myocardial layers. These branches are defined arterioles, terminating in capillary vessels, directly supplying myocardial cells. Each of these coronary segments produces different level and degree of resistance to coronary blood flow. Normal (non-stenosed) large epicardial coronary arteries play a minor role in the regulation of coronary vascular resistance and act mainly as conductance (conduit) vessels. Most of the resistance, which opposes coronary blood flow, arises from resistance arterioles. The resistance is manifest by decreased coronary perfusion pressure. The percent distribution of the length and resistance of individual segments of the coronary vascular tree is summarised in Table 1 and depicted in Figure 1.
Table 1 Distribution of coronary resistance in normal coronary arterial bed.
Large epicardial arteries Medium-sized and small arteries Arterioles Capillaries
Diameter >1000 μm 1000–100 μm. 100–10 μm <10 μm
% of total resistance 5% 15–25% 400–100 μm 50–60% 20%
% length of coronary bed 5–10% 15–25% 60–75%
Figure 1 Sequential decrease of coronary perfusion pressure in consecutive segments of coronary vasculature. The largest fall in perfusion pressure occurs in coronary resistance arterioles.
The difference between coronary blood flow corresponding to flow autoregulation plateau at rest and coronary blood flow after maximal vasodilatation is traditionally defined as coronary flow reserve (CFR) [1-6]. CFR is usually calculated as the ratio of maximal (hyperemic) to resting coronary blood flow (Figure 2). CFR is an important functional parameter to understand the pathophysiology of coronary circulation and can be used to examine the integrity of microvascular circulation.
Figure 2 Transmural distribution of coronary flow reserve (CFR); CFR subepicardial > CFR subendocardial.
The assessment of regional coronary flow reveals marked spatial heterogeneity of CFR across the myocardial wall. According to the model of Hoffman [2], the highest CFR is that measurable in the subepicardial layer of myocardium. In relation to a transmural flow reduction, CFR is significantly lower in the subendocardial layer, also due to elevated left ventricular (LV) diastolic pressure increasing extravascular compressive forces. As a result of this transmural coronary flow distribution, CFR is exhausted firstly in the subendocardial layer [2] (Figure 2). The lower limit of coronary flow autoregulation is unfavourably shifted to the higher value of coronary perfusion pressure in the subendocardial layer as compared with the subepicardial layer (i.e. 55–65 mm Hg versus 30–40 mmHg, respectively) [5] (Figure 2). According to Hoffman [2] the reduction of global CFR from 4.0 to 2.0 could be associated with the loss of flow reserve in a part or all of the subendocardial layer of the myocardium.
In human studies, due to limited spatial resolution of positron emission tomography (PET), the demonstration of subendocardial hypoperfusion has been made possible only in the hypertrophied myocardium [7]. In patients with aortic valve stenosis and normal epicardial coronary arteries [7], subendocardial CFR (1.43 ± 0.33) was lower than subepicardial CFR (1.78 ± 0.35; P = 0.01). In the subgroup of severe aortic stenosis (aortic valve area <0.8 cm2) CFR was <2.0 both at subendocardial and subepicardial levels and in two of these patients subendocardial CFR was <1.0 (lack of CFR).
An alternative approach to document transmural steal phenomenon is the calculation of subendocardial/subepicardial flow ratio [8,9]. A hyperemic value of subendocardial / subepicardial flow ratio <0.8 (i.e., subendocardial flow is at least 20% lower than subepicardial) has been proposed as marker of subendocardial hypoperfusion. In some patients with hypertrophic cardiomyopathy (HCM), subendocardial flow was 40% lower than subepicardial flow (ratio = 0.6) after infusion of a vasodilator agent [8,9].
Magnetic resonance imaging (MRI) with gadolinium contrast agent, has higher resolution and appears more sensitive providing the opportunity to assess even patients without left ventricular hypertrophy, i.e. with normal wall thickness [10]. Accordingly, cardiac MRI demonstrates subendocardial hypoperfusion in patients with syndrome X during the intravenous administration of adenosine when compared with healthy control subjects who showed increase of subendocardial perfusion after induced hyperemia [10].
The lower limit of normal CFR and criteria for normal reference values
An important practical issue is to confirm in clinical studies that the lower limit of normal CFR is >3.0. This aspect is summarised in Table 2.
Table 2 Comparison of CFR using different methods in reference control groups.
Number of patients Method CFR Reference
17 (HTX) DI 5.0 ± 0.3* [11]
26 (HTX) DI 5.2 ± 1.3* [12]
18 (young subjects) PET 4.1 ± 0.9 [13]
22 (elderly subjects) PET 3.0 ± 0.7 [13]
10 D.TTE 4.5 ± 0.9 [14]
10 PET 4.1 ± 1.0 [15]
10 D.TTE 5.2 ± 1.6 [16]
19 D.TTE 3.7 ± 0.7 [17]
26 (athletes) D.TTE 5.9 ± 1.0 [17]
* intracoronary papaverine
CFR = coronary flow reserve; DI. = invasive intracoronary Doppler, D.TTE = noninvasive transthoracic Doppler echocardiography, HTX = routine assessment of coronary flow reserve early after heart transplantation
Before defining the normal value of CFR, we should identify appropriately reference groups of healthy subjects. In order to achieve this aim, non invasive studies present a particularly useful approach because we may recruit any subject who gives informed consent to examination. The perfect candidate is a healthy volunteer without cardiovascular signs/symptoms and/or risk factors for both vascular dysfunction and coronary artery disease. In contrast, invasive studies are performed in subjects who complain about chest pain or other cardiac symptoms, implying that coronary microvascular vasodilator dysfunction may limit coronary blood flow and determines angina despite normal coronary epicardial arteries. Two studies [18,19] have shown that such a coincidence is quite frequent. Bearing in mind this limitation it has been proposed by Baumgart et al [20] that in invasive measurements normal limits of CFR may be derived only from highly selected subjects according to the following restricted criteria:
- truly normal epicardial arteries confirmed by intracoronary ultrasound examination
- age <50 years
- absence of symptoms (in addition, we propose an obligatory absence of risk factors for vascular/endothelial dysfunction).
These highly-selected subjects (table 2) exhibit CFR markedly exceeding the cut-off point of 3.0, this value is near to the highest value obtained in an athletes' populations (CFR > 5.0).
The recruitment of reference controls subjects with normal epicardial coronary arteries verified by intracoronary ultrasound is crucial as indicated by the following example. Positive exercise myocardial scintigraphy, primarily considered as false positive in relation to angiographically normal coronary vessels, may frequently turn out to be true positive when control intracoronary ultrasound reveals vascular lesions [21]. In this clinical setting, CFR is a fairly good predictor of "soft lesions", non-visualizable even by coronary angiography [21].
Baumgart [20] proposed a range of age <50 years since PET revealed a significantly higher CFR in younger than in the elderly subjects (Table 2). Other investigators [13] demonstrated that aging-induced reduction of CFR is a result of increased coronary blood flow at rest, whereas maximal blood flow remained relatively unchanged during the years. Only over 70 years, maximal coronary blood flow is reduced. According to these findings [20], the lower limit of normal CFR should be reasonably fixed to 3.0 for subjects up to the age of 50.
Factors limiting coronary flow reserve
In general, the reduction of CFR may be associated with three main kinds of abnormalities [2,3,22,23] (Table 3, Figure 3) and it is even possible for two factors to co-operate simultaneously. However, it has to be taken into account that epicardial coronary artery stenosis, the most visible factor of patients with angina pectoris (examined alone by coronary angiography in daily practice), is only one possible determinant in contrast to several other multi-factorial mechanisms involving coronary microvascular dysfunction. Structural changes (remodelling) in coronary microcirculation can themselves be responsible of CFR reduction. By using myocardial biopsy, some studies have provided the opportunity to compare pathomorphological changes of coronary microcirculation and CFR reduction, documenting the relationship between morphological and hemodynamic abnormalities. In patients with hypertrophic cardiomyopathy, a reduced arteriolar lumen was associated to a reduced CFR [24,25]. Also in hypertensive patients [26] reduced CFR correlated with increased coronary arteriolar wall thickness, i.e, arteriolar remodelling. A decreased arteriolar wall/lumen ratio correlates with reduction of CFR as well as with abnormalities of derived parameters as coronary resistance reserve. All together, these studies strongly support the hypothesis that the microvascular factor is a further important contributor (as extravascular, myocardial factor, i.e. LV hypertrophy, excessive intramyocardial pressure) of CFR reduction. From a pathophysiological point of view, coronary microvascular disease is associated with alternative ischemic cascade where stress test induces chest pain, ECG ST-segment depression and myocardial scintigraphic perfusion defect despite the lack of changes in echocardiographic-derived regional myocardial wall motion [27].
Table 3 Three groups of factors limiting CFR:
1. Increase of resting coronary blood flow due to increased myocardial oxygen demand as a result of:
• tachycardia
• increased myocardial contractility
• myocardial hypertrophy
2. Decrease of maximal (hyperemic) coronary blood flow:
• epicardial coronary artery stenosis
• decrease mean aortic pressure = coronary perfusion pressure e.g. aortic insufficiency, exaggerate response to vasodilator agent
• wall thickening (remodeling) of resistance arterioles
• reduced density of arterioles
• cardiomyocyte hypertrophy
• perivascular fibrosis
• interstitial fibrosis
• endothelial dysfunction
• increased blood viscosity: policythemia, macroglobulinemia
• elevated LV diastolic pressure increasing extravascular compressive forces and resistance (particularly in subendocardial layer).
3. Shift to the right in the pressure-flow relation through maximally dilated vessels due to an increase in zero flow pressure line:
• increased left ventricular diastolic pressure
• tachycardia
• myocardial hypertrophy
Figure 3 Complexity of CFR concept. Percent values on the curves represent the severity of coronary epicardial stenosis.
How much coronary microvascular disease may reduce CFR?
In a recent study Voci et al. [28] minimised the role of reduced microvascular CFR, probably underestimating the unfavourable influence of impaired coronary microcirculation on prognosis. These authors stated that patients die of epicardial coronary artery disease, not of microvascular disease. In other studies [29-35], however, patients with no or minimal coronary stenosis of epicardial coronary arteries exhibited a significantly blunted CFR in relation to microvascular abnormalities induced by various cardiovascular diseases (Table 4). Several patients had markedly reduced CFR < 2.0 and in some pediatric patients with hypertrophic cardiomyopathy CFR was <1.0 [29]. On these grounds, the authors [29] postulated that non-hypertrophic free wall steals the blood flow from the hypertrophied septum after a vasodilator infusion. A gradual decrease of CFR, parallel to more advanced stages of microvascular disease, was observed in diabetic patients, patients with syndrome X and also hypertensive patients without overt coronary artery stenosis (Table 4). In this view it is notable the experience of Rigo [36], who collected CFR values measured by transthoracic Doppler echocardiography in large population samples affected by various cardiac diseases (Table 5). It is also of interest that the reduction of CFR is reversible in some cases, for instance after regression of LV hypertrophy subsequent to aortic valve replacement in patients with aortic valve stenosis, after anti-hypertensive treatment or even after cholesterol treatment [37-42]. (Table 6). Importantly, even drugs of the same group (angiotensin converting enzyme inhibitors [perindopril versus enalapril] and statins [simvastatin versus pravastatin]) exhibited different influence on CFR (Table 6). After simvastatin treatment [39], the percent increase of CFR correlated with percent decrease of cholesterol and triglycerides levels.
Table 4 CFR in microvascular disease with normal coronary angiogram.
Clinical setting CFR
HCM pediatric pts (septum) [29] 0.84 ± 0.33
Control 2.94 ± 0.35
Aortic insufficiency [30] 1,67 ± 0,4
Control 4,03 ± 0,52
Dilated cardiomyopathy [31] 2.2 ± 0.8
Control 3.3 ± 0.8
Dilated cardiomyopathy [32] 2,0 ± 0,6
NYHA class I 2,43 ± 0,4
NYHA class II 2,21 ± 0,2
NYHA class III 1,98 ± 0,3
NYHA class IV 1,78 ± 0,3
Control 3,2 ± 0,5
Diabetes [33]
Without retinopathy 2.8 ± 0.3
with early diabetic retinopathy 2.3 ± 0.3
with slightly advanced retinopathy 1.6 ± 0,2
Control 3.3 ± 0,4
Patients with chest pain and [34]
Without ST depression in ECG exercise test 3.0 ± 0.6
With up-slope ST depression in ECG exercise test 3.1 ± 0.6
With flat ST depression in ECG exercise test 2.1 ± 0.6
With down-slope ST depression ECG exercise test 2.0 ± 0,4
Hypertension [35]
Concentric remodeling 2.0 ± 0.7
Concentric hypertrophy 2.3 ± 0.8
Eccentric hypertrophy 2.9 ± 0.6
Normal geometry 2.7 ± 0.4
Control 4.2 ± 0.5
Table 5 Findings of CFR in some diseases associated to coronary microvascular dysfunction and in healthy controls [36]
Clinical setting CFR
Hypertrophic cardiomyopathy 2.21 ± 0.2
Dilated cardiomyopathy 1.9 ± 0.2
Syndrome X 2.27 ± 0,3
Control group 3.3 ± 0.3
Table 6 Increase in CFR in microvascular disease after treatment.
Disease CFR Effect of treatment p
Aortic stenosis [37] 1.8 ± 0.5 2.6 ± 0.7 (valve replacement) <0.05
Familial hypercholesterolemia [38] 2.3 ± 0.6 3.3 ± 1.2 (simvastatin) <0.05
Hypercholesterolemia [39] 2.4 ± 0.7 3.2 +1.2 (simvastatin) <0.05
2.2 ± 0.7 2.3 ± 0.6 (pravastatin) >0.05
Arterial hypertension [40] 1.9 ± 0.31 2.1 ± 0.3 (nebivolol) <0.05
Arterial hypertension [41] 2.1 ± 0.6 3.5 ± 1.9 (perindopril) <0.05
Arterial hypertension [42] 2.4 ± 0.7 2.4 ± 0.6 (enalapril) >0.05
2.7 ± 0.8 3.7 ± 1.8 (verapamil) <0.05
Interestingly, in patients with epicardial coronary artery stenosis and severe hypercholesterolemia, single LDL-apheresis improved microcirculation by increasing CFR from 1.91 ± 0,68 to 2.48 ± 0.68 [43]. This finding demonstrates that, even after single LDL-lowering intervention, some patients can move quickly from a group where PTCA appears to be required (corresponding to CFR < 2.0) to a group with CFR >2.0, where PTCA may not be needed anymore.
Vasodilators inducing hyperemia
Two main pharmacological vasodilators, adenosine and dipyridamole, are used in humans to recruit CFR. The characteristics of these agents are compared in Table 7. These agents have an advantage over exercise and dobutamine, which represent submaximal stimuli for coronary flow reserve and are much more technically demanding for ultrasound imaging of CFR [44]. Either adenosine or dipyridamole were used in referred studies of Tables 4, 5, where only studies with control group are reported. The control group provides an opportunity to compare how much CFR is reduced (sometimes more than 50%) in patients with different involvement of coronary microvascular disease.
Table 7 Comparison between adenosine and dipyridamle characteristics
Adenosine Dipyridamole
Duration of action 30 sec 30 min
Time to max. Effect 30–55 sec 6–16 min
Advantage Short action, short-lasting adverse effects prolonged action allow to assess CFR and wall motion abnormalities during the same examination
Disadvantage Frequent- hyperventilation
Rare – bradycardia, AV block, hypotension, flushing, headache, possibility of antidote-resistance prolonged ischemia, hypotension, flushing, headache, hyperventilation,
Prognostic value of impaired microvascular CFR
The prognostic role of impaired microvascular CFR has been found unfavourable. In a study by Marks at al. [45], reduced CFR due to unspecified microvascular disease predicted increased mortality – 20% vs. 7% in a group with normal CFR (p < 0.016). The relationship between unfavourable prognosis and reduced CFR in patients with hypertrophic or idiopathic dilated cardiomyopathies was recently reported [46,47]. In other two studies where patients were divided according to CFR tercentyle [48] or maximally stimulated coronary blood flow tercentyle [49], the subgroup defined as the lowest tercentyle had the worst outcome. The markedly blunted maximal blood flow was related with poor prognosis not only in HCM patients [49] but also in patients with idiopathic dilated cardiomyopathy (DCM) [50]. In this clinical setting, strongly depressed dipyridamole-stimulated maximal coronary blood flow was associated, with a 3.5 relative risk of death or development or progression of heart failure. These results support the hypothesis that chronic myocardial hypoperfusion or repetitive myocardial ischemia attributable to abnormal coronary microcirculatory flow could exert a detrimental role in the evolution of idiopathic LV dysfunction toward overt DCM. Cecchi et al. [49] hypothesised that coronary microvascular dysfunction may represent a common pathway leading to a disease progression in different cardiomyopathies, including conditions as aortic valve stenosis and hypertensive heart disease.
Limitations and hypothesis
In several of the referred reports, dipyridamole was used to produce vasodilating hyperemia. Adenosine (140 ug/kg/min) has been shown to be either similar [51] or more potent vasodilator agent [52] than high-dose dipyridamole (0.84 mg/Kg). In relation to the possibility that dipyridamole-recruited CFR could be submaximal, we can not be absolutely sure about the appropriateness of the lower limit of normal CFR = 3.0. To resolve this problem, apart from vasodilator selection, the choice of appropriate control groups is also due, by excluding smokers and patients with arterial hypertension, hyperlipidemia, obesity, diabetes mellitus, and, possibly, those affected by hyperhomocysteinemia. These highly selecting criteria probably was not fulfilled in previous studies, in particular the oldest, where the newest recognized factors limiting CFR were not yet known. Recently, new evidence has been given that in healthy subjects even single exposition to passive smoking [53], fat meal inducing hypertriglyceridemia [54], hyperhomocysteinemia [55], and estrogen decrease in menstrual phase of cycle [56] can reduce CFR by approximately 30%. Thus, we can not be certain that normal CFR starts from 3.0 because CFR values higher than 5.0 were recorded in humans. Interestingly, in a transplanted heart which can not be equivalent of intact heart in healthy volunteers, CFR may also achieved more 5.0 (see Table 2). This high value is probably the result of intracoronary papaverine, which is not used currently. However, these findings strongly supports the hypothesis that CFR may achieve much higher level than 3.0.
As regard validation of a non-invasive methods, transthoracic Doppler echocardiography closely agrees with intracoronary Doppler flow wire results in assessing CFR. Good correlation of non-invasive and invasive Doppler assessment of CFR has been shown both in LAD [57] and RCA [58]. The feasibility of transthoracic Doppler echocardiography to detect coronary flow is 80–98% in LAD, 50–87% in RCA and 43–72% in Cx [6,36,59].
Conclusion
Isolated coronary microvascular abnormalities are overt by reduced CFR despite normal epicardial coronary arteries. These abnormalities may occur in several diseases (arterial hypertension, diabetes mellitus, hypercholesterolemia, syndrome X, aortic valve disease, hypertrophic cardiomyopathy and idiopathic dilated cardiomyopathy). The prognostic role of impaired microvascular CFR has been shown unfavourable, in particular in patients with hypertrophic or idiopathic dilated cardiomyopathies. CFR reduction may be reversible, for instance after regression of left ventricular hypertrophy subsequent to valve replacement in patients with aortic stenosis, after anti-hypertensive treatment or using cholesterol lowering drugs. CFR may increase by 30% or more after pharmacological therapy and achieve level >3.0. In contrast to other non invasive tools as PET, very expensive and associated with radiation exposure, transthoracic Doppler-derived CFR is equally non invasive but cheaper, very accessible [60] and prone to a reliable exploration of coronary microvascular territories, not detectable by invasive coronary angiography, able to visualize only large epicardial arteries.
==== Refs
Ganz P Ganz W Braunwald E, Zipes DP, Libby P Coronary blood flow and myocardial ischemia Heart Disease 2001 New York: W.B.Sauders 1087 1113
Hoffman JI Problems of coronary flow reserve Ann Biomed Eng 2000 28 884 896 11144672 10.1114/1.1308503
Klocke FJ Measurements of coronary flow reserve: defining pathophysiology versus making decisions about patient care Circulation 1987 76 1183 1189 2960470
Mosher PJ Ross J JrMcfate PA Shaw RF Control of coronary blood flow by an autoregulatory mechanism Circ Res 1964 14 250 259 14133952
Dimitrow PP Coronary flow reserve- measurement and application: focus on transthoracic Doppler echocardiography 2002 Boston/Dordrecht/London: Kluwer Academic Publishers
Dimitrow PP Transthoracic Doppler echocardiography- noninvasive diagnostic window for coronary flow reserve assessment Cardiovasc Ultrasound 2003 1 4 12740038 10.1186/1476-7120-1-4
Rajappan K Rimoldi O Dutka DP Ariff B Pennell DJ Sheridan DJ Camici PG Mechanisms of coronary microcirculatory dysfunction in patients with aortic stenosis and angiographically normal coronary arteries Circulation 2002 105 470 476 11815430 10.1161/hc0402.102931
Choudhury L Elliott P Rimoldi O Ryan M Lammertsma AA Boyd H McKenna WJ Camici PG Transmural myocardial blood flow distribution in hypertrophic cardiomyopathy and effect of treatment Basic Res Cardiol 1999 94 49 59 10097830 10.1007/s003950050126
Lorenzoni R Gistri R Cecchi F Olivotto I Chiriatti G Elliott P McKenna WJ Camici PG Coronary vasodilator reserve is impaired in patients with hypertrophic cardiomyopathy and left ventricular dysfunction Am Heart J 1998 136 972 981 9842009
Panting JR Gatehouse PD Yang GZ Grothues F Firmin DN Collins P Pennell DJ Abnormal subendocardial perfusion in cardiac syndrome X detected by cardiovascular magnetic resonance imaging N Engl J Med 2002 346 1948 1953 12075055 10.1056/NEJMoa012369
McGinn AL White CW Wilson RF Interstudy variability of coronary flow reserve. Influence of heart rate, arterial pressure, and ventricular preload Circulation 1990 81 1319 1330 2317912
Chauhan A Mullins PA Petch MC Schofield PM Is coronary flow reserve in response to papaverine really normal in syndrome X? Circulation 1994 89 1998 2004 8181123
Czernin J Muller P Chan S Brunken RC Porenta G Krivokapich J Chen K Chan A Phelps ME Schelbert HR Influence of age and hemodynamics on myocardial blood flow and flow reserve Circulation 1993 88 62 69 8319357
Hirata K Shimada K Watanabe H Otsuka R Tokai K Yoshiyama M Homma S Yoshikawa J Black tea increases coronary flow velocity reserve in healthy male subjects Am J Cardiol 2004 93 1384 1388 15165919 10.1016/j.amjcard.2004.02.035
Uren NG Marraccini P Gistri R de Silva R Camici PG Altered coronary vasodilator reserve and metabolism in myocardium subtended by normal arteries in patients with coronary artery disease J Am Coll Cardiol 1993 22 650 658 8354794
Kubo S Tadamura E Toyoda H Mamede M Yamamuro A Magata Y Mukai T Kitano H Tamaki N Konishi J Effect of caffeine intake on myocardial hyperemic flow induced by adenosine triphosphate and dipyridamole J Nucl Med 2004 45 730 738 15136619
Hildick-Smith DJ Johnson PJ Wisbey CR Winter EM Shapiro LM Coronary flow reserve is supranormal in endurance athletes: an adenosine transthoracic echocardiographic study Heart 2000 84 383 389 10995406 10.1136/heart.84.4.383
Hasdai D Holmes DR JrHigano ST Burnett JC JrLerman A Prevalence of coronary blood flow reserve abnormalities among patients with nonobstructive coronary artery disease and chest pain Mayo Clin Proc 1998 73 1133 1140 9868410
Reis SE Holubkov R Lee JS Sharaf B Reichek N Rogers WJ Walsh EG Fuisz AR Kerensky R Detre KM Sopko G Pepine CJ Coronary flow velocity response to adenosine characterizes coronary microvascular function in women with chest pain and no obstructive coronary disease. Results from the pilot phase of the Women's Ischemia Syndrome Evaluation (WISE) study J Am Coll Cardiol 1999 33 1469 1475 10334410 10.1016/S0735-1097(99)00072-8
Baumgart D Haude M Liu F Ge J Goerge G Erbel R Current concepts of coronary flow reserve for clinical decision making during cardiac catheterization Am Heart J 1998 136 136 149 9665231
Verna E Ceriani L Giovanella L Binaghi G Garancini S "False-positive" myocardial perfusion scintigraphy findings in patients with angiographically normal coronary arteries: insights from intravascular sonography studies J Nucl Med 2000 41 1935 1940 11138675
Gould KL Lipscomb K Hamilto GW Physiology basis for assessing critical coronary stenosis: instantaneous flow response and regional distribution during coronary hyperemia as measures of coronary flow reserve Am J Cardiol 1974 33 87 94 4808557 10.1016/0002-9149(74)90743-7
Gould KL Kirkeeide RL Buchi M Coronary flow reserve as a physiologic measure of stenosis severity J Am Coll Cardiol 1990 15 459 474 2137151
Krams R Kofflard MJ Duncker DJ Von Birgelen C Carlier S Kliffen M ten Cate FJ Serruys PW Decreased coronary flow reserve in hypertrophic cardiomyopathy is related to remodeling of the coronary microcirculation Circulation 1998 97 230 233 9462521
Schwartzkopff B Mundhenke M Strauer BE Alterations of the architecture of subendocardial arterioles in patients with hypertrophic cardiomyopathy and impaired coronary vasodilator reserve: a possible cause for myocardial ischemia J Am Coll Cardiol 1998 31 1089 1096 9562012 10.1016/S0735-1097(98)00036-9
Schwartzkopff B Motz W Frenzel H Vogt M Knauer S Strauer BE Structural and functional alterations of the intramyocardial coronary arterioles in patients with arterial hypertension Circulation 1993 88 993 1003 8353927
Picano E Palinkas A Amyot R Diagnosis of myocardial ischemia in hypertensive patients J Hypertens 2001 19 1177 1183 11446706 10.1097/00004872-200107000-00001
Voci P Pizzuto F Romeo F Coronary flow: a new asset for the echo lab? Eur Heart J 2004 25 1867 1879 15522465 10.1016/j.ehj.2004.07.029
Tadamura E Yoshbayashi M Yonemura T Kudoh T Kubo S Motooka M Significant regional heterogeneity of coronary flow reserve in paediatric hypertrophic cardiomyopathy Eur J Nucl Med 2000 27 1340 1348 11007516 10.1007/s002590000300
Nitenberg A Foult JM Antony I Blanchet F Rahali M Coronary flow and resistance reserve in patients with chronic aortic regurgitation, angina pectoris and normal coronary arteries J Am Coll Cardiol 1988 11 478 486 3278034
Canetti M Akhter MW Lerman A Karaalp IS Zell JA Singh H Mehra A Elkayam U Evaluation of myocardial blood flow reserve in patients with chronic congestive heart failure due to idiopathic dilated cardiomyopathy Am J Cardiol 2003 92 1245 1249 10.1016/j.amjcard.2003.08.002
Santagata P Rigo F Gherardi S Pratali L Drozdz J Varga A Picano E Clinical and functional determinants of coronary flow reserve in non-ischemic dilated cardiomyopathy An echocardiographic study Int J Cardiol 2005
Akasaka T Yoshida K Hozumi T Takagi T Kaji S Kawamoto T Morioka S Yoshikawa J Retinopathy identifies marked restriction of coronary flow reserve in patients with diabetes mellitus J Am Coll Cardiol 1997 30 935 941 9316521 10.1016/S0735-1097(97)00242-8
Youn HJ Park SJ Cho EJ Jung HO Jeon HK Lee JM Oh YS Chung WS Kim JH Choi KB Hong SJ Choi K Pattern of exercise-induced ST change is related to coronary flow reserve in patients with chest pain and normal coronary angiogram Int J Cardiol 2005 101 299 304 15882679 10.1016/j.ijcard.2004.03.037
Schafer S Kelm M Mingers S Strauer BE Left ventricular remodeling impairs coronary flow reserve in hypertensive patients J Hypertens 2002 20 31 37 10.1097/00004872-200207000-00031
Rigo F Coronary flow reserve in stress-echo lab. From pathophysiologic toy to diagnostic tool Cardiovasc Ultrasound 2005 3 8 15792499 10.1186/1476-7120-3-8
Hildick-Smith DJ Shapiro LM Coronary flow reserve improves after aortic valve replacement for aortic stenosis: an adenosine transthoracic echocardiography study J Am Coll Cardiol 2000 36 1889 1896 11092661 10.1016/S0735-1097(00)00947-5
Yokoyama I Yonekura K Inoue T Ohtomo K Nagai R Long-term effect of simvastatin on the improvement of impaired myocardial flow reserve in patients with familial hypercholesterolemia without gender variance J Nucl Cardiol 2001 8 445 451 11481566 10.1067/mnc.2001.115517
Yokoyama I Inoue Y Moritan T Ohtomo K Nagai R Impaired myocardial vasodilatation during hyperaemic stress is improved by simvastatin but not by pravastatin in patients with hypercholesterolaemia Eur Heart J 2004 25 671 679 15084372 10.1016/j.ehj.2004.02.017
Galderisi M Cicala S D'Errico A de Divitiis O de Simeone G Nebivolol improves coronary flow reserve in hypertensive patients without coronary heart disease J Hypertens 2004 22 2201 2208 15480106 10.1097/00004872-200411000-00024
Schwartzkopff B Brehm M Mundhenke M Strauer BE Repair of coronary arterioles after treatment with perindopril in hypertensive heart disease Hypertension 2000 36 220 225 10948081
Parodi O Neglia D Palombo C Sambuceti G Giorgetti A Marabotti C Gallopin M Simonetti I L'Abbate A Comparative effects of enalapril and verapamil on myocardial blood flow in systemic hypertension Circulation 1997 96 864 873 9264494
Mellwig KP Baller D Gleichmann U Moll D Betker S Weise R Notohamiprodjo G Improvement of coronary vasodilatation capacity through single LDL apheresis Atherosclerosis 1998 139 173 178 9699905 10.1016/S0021-9150(98)00055-0
Takeuchi M Miyazaki C Yoshitani H Sakamoto K Yoshikawa J Assessment of coronary flow velocity with transthoracic Doppler echocardiography during dobutamine stress echocardiography J Am Coll Cardiol 2001 38 117 123 11451260 10.1016/S0735-1097(01)01322-5
Marks DS Gudapati S Prisant LM Leir B Donato-Gonzalez C Baller JL Houghton JL Mortality in patients with microvascular disease J Clin Hypertens 2004 6 304 309
Rigo F Cortigiani L Gherardi S Zanella C Di Pede F Pasanisi E Raviele A Picano E The prognostic meaning of coronary flow reserve assessed by Doppler echocardiography in nonischemic dilated cardiomyopathy. [Abstract] Circulation 2004 110 III 511
Rigo F Pasanisi E Gherardi S Zanella C Cutaia V Cortigiani L Raviele A Picano E Coronary microvascular dysfunction and prognosis in hypertrophic cardiomyopathy: a echocardiographic study. [Abstract] Eur Heart J 2004 25 16
Britten MB Zeiher AM Schachinger V Microvascular dysfunction in angiographically normal or mildly diseased coronary arteries predicts adverse cardiovascular long-term outcome Coron Artery Dis 2004 15 259 264 15238822 10.1097/01.mca.0000134590.99841.81
Cecchi F Olivotto I Gistri R Lorenzoni R Chiriatti G Camici PG Coronary microvascular dysfunction and prognosis in hypertrophic cardiomyopathy N Engl J Med 2003 349 1027 1035 12968086 10.1056/NEJMoa025050
Neglia D Michelassi C Trivieri MG Sambuceti G Giorgetti A Pratali L Gallopin M Salvadori PA Sorace O Carpeggiani C Poddighe R L'Abbate A Parodi O Prognostic role of myocardial blood flow impairment in idiopathic left ventricular dysfunction Circulation 2002 105 186 193 11790699 10.1161/hc0202.102119
Lim HE Shim WJ Rhee H Kim SM Hwang GS Kim YH Seo HS Oh DJ Ro YM Assessment of coronary flow reserve with transthoracic Doppler echocardiography: comparison among adenosine, standard-dose dipyridamole, and high-dose dipyridamole J Am Soc Echocardiogr 2000 13 264 270 10756243 10.1067/mje.2000.103508
Holdright DR Lindsay DC Clarke D Fox K Poole-Wilson PA Collins P Coronary flow reserve in patients with chest pain and normal coronary arteries Br Heart J 1993 70 513 519 8280515
Otsuka R Watanabe H Hirata K Tokai K Muro T Yoshiyama M Takeuchi K Acute effects of passive smoking on the coronary circulation in healthy young adults JAMA 2001 286 436 441 11466122 10.1001/jama.286.4.436
Hozumi T Eisenberg M Sugioka K Kokkirala A Change in coronary flow reserve on transthoracic Doppler echocardiography after a single high-fat meal in young healthy men Ann Inter Med 2002 136 523 528
Ascione L De Michele M Accadia M Rumolo S Sacra C Alberta Ortali V Inserviente L Petti M Russo G Tuccillo B Effect of acute hyperhomocysteinemia on coronary flow reserve in healthy adults J Am Soc Echocardiogr 2004 17 1281 1285 15562267 10.1016/j.echo.2004.07.011
Hirata K Shimada K Watanabe H Muro A Modulation of coronary flow velocity reserve by gender, menstrual cycle and hormone replacement therapy J Am Coll Cardiol 2001 38 1879 1884 11738288 10.1016/S0735-1097(01)01658-8
Caiati C Montaldo C Zedda N Montisci R Ruscazio M Lai G Cadeddu M Meloni L Iliceto S Validation of a new noninvasive method (contrast-enhanced transthoracic second harmonic echo Doppler) for the evaluation of coronary flow reserve: comparison with intracoronary Doppler flow wire J Am Coll Cardiol 1999 34 1193 1200 10520812 10.1016/S0735-1097(99)00342-3
Lethen H Tries HP Kersting S Lambertz H Validation of noninvasive assessment of coronary flow velocity reserve in the right coronary artery. A comparison of transthoracic echocardiographic results with intracoronary Doppler flow wire measurements Eur Heart J 2003 24 1567 1575 12927192 10.1016/S0195-668X(03)00284-7
Fujimoto K Watanabe H Hozumi T Otsuka R Hirata K Yoshiyama M Yoshikawa J New noninvasive diagnosis of myocardial ischemia of the left circumflex coronary artery using coronary flow reserve measurement by transthoracic Doppler echocardiography: comparison with thallium-201 single photon emission computed tomography J Cardiol 2004 43 109 16 15067798
Krzanowski M Bodzon W Dimitrow PP Imaging of all three coronary arteries by transthoracic echocardiography: an illustrated guide Cardiovasc Ultrasound 2003 1 16 14622441 10.1186/1476-7120-1-16
|
16080792
|
PMC1201155
|
CC BY
|
2021-01-04 16:38:32
|
no
|
Cardiovasc Ultrasound. 2005 Aug 4; 3:18
|
utf-8
|
Cardiovasc Ultrasound
| 2,005 |
10.1186/1476-7120-3-18
|
oa_comm
|
==== Front
Cardiovasc UltrasoundCardiovascular Ultrasound1476-7120BioMed Central London 1476-7120-3-221610722110.1186/1476-7120-3-22ResearchQuantitative assessment of harmonic power doppler myocardial perfusion imaging with intravenous levovist™ in patients with myocardial infarction: comparison with myocardial viability evaluated by coronary flow reserve and coronary flow pattern of infarct-related artery Tani Tomoko [email protected] Kazuaki [email protected] Minako [email protected] Fumie [email protected] Minako [email protected] Koichi [email protected] Shuichiro [email protected] Atsushi [email protected] Kunihiko [email protected] Kenichi [email protected] Shigefumi [email protected] Yasuki [email protected] Division of Cardiology, Kobe General Hospital, 4–6 Minatojima-Nakamachi, Chuo-ku, Kobe, Japan2005 18 8 2005 3 22 22 16 7 2005 18 8 2005 Copyright © 2005 Tani et al; licensee BioMed Central Ltd.2005Tani 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
Myocardial contrast echocardiography and coronary flow velocity pattern with a rapid diastolic deceleration time after percutaneous coronary intervention has been reported to be useful in assessing microvascular damage in patients with acute myocardial infarction.
Aim
To evaluate myocardial contrast echocardiography with harmonic power Doppler imaging, coronary flow velocity reserve and coronary artery flow pattern in predicting functional recovery by using transthoracic echocardiography.
Methods
Thirty patients with anterior acute myocardial infarction underwent myocardial contrast echocardiography at rest and during hyperemia and were quantitatively analyzed by the peak color pixel intensity ratio of the risk area to the control area (PIR). Coronary flow pattern was measured using transthoracic echocardiography in the distal portion of left anterior descending artery within 24 hours after recanalization and we assessed deceleration time of diastolic flow velocity. Coronary flow velocity reserve was calculated two weeks after acute myocardial infarction. Left ventricular end-diastolic volumes and ejection fraction by angiography were computed.
Results
Pts were divided into 2 groups according to the deceleration time of coronary artery flow pattern (Group A; 20 pts with deceleration time ≧ 600 msec, Group B; 10 pts with deceleration time < 600 msec). In acute phase, there were no significant differences in left ventricular end-diastolic volume and ejection fraction (Left ventricular end-diastolic volume 112 ± 33 vs. 146 ± 38 ml, ejection fraction 50 ± 7 vs. 45 ± 9 %; group A vs. B). However, left ventricular end-diastolic volume in Group B was significantly larger than that in Group A (192 ± 39 vs. 114 ± 30 ml, p < 0.01), and ejection fraction in Group B was significantly lower than that in Group A (39 ± 9 vs. 52 ± 7%, p < 0.01) at 6 months. PIR and coronary flow velocity reserve of Group A were higher than Group B (PIR, at rest: 0.668 ± 0.178 vs. 0.248 ± 0.015, p < 0.0001: during hyperemia 0.725 ± 0.194 vs. 0.295 ± 0.107, p < 0.0001; coronary flow velocity reserve, 2.60 ± 0.80 vs. 1.31 ± 0.29, p = 0.0002, respectively).
Conclusion
The preserved microvasculature detecting by myocardial contrast echocardiography and coronary flow velocity reserve is related to functional recovery after acute myocardial infarction.
==== Body
Introduction
In patients with myocardial infarction (MI), the distinction between irreversible fibrotic scar and akinetic but viable myocardium has important clinical implications. Several imaging techniques have been used to detect myocardial viability. Coronary blood flow reserve (CFR) has been established as a useful method for assessing microvascular function [1-3]. Previous studies revealed that the measurement of CFR in the infarct-related coronary artery might help to assess myocardial viability [4-6].
Kawamoto T et al. showed that coronary blood flow spectrum immediately after primary percutaneous transluminal coronary angioplasty (PTCA) by use of a Doppler guidewire reflected a greater degree of microvascular damage in the risk area, and was useful in predicting recovery of regional left ventricular (LV) function [7].
The diastolic deceleration slope of coronary flow velocity is steeper in patients with substantial 'no reflow' phenomenon assessed by myocardial contrast echocardiography (MCE). Recently, Shintani et al revealed that patients with a shorter deceleration half time (DHT) of diastolic coronary flow velocity have a poorer functional outcome among patients with anterior AMI and that the transthoracic Doppler echocardiography (TTDE)-determined DHT is a useful predictor of myocardial viability after anterior AMI [8].
The assessment of MCE with intracoronary or venous injection of ultrasound contrast agents composed of microbubbles has been shown to provide information on perfusion territories, collateral flow, infarct size, myocardial viability, and success of reperfusion therapy [9-12]. Recent advances in both of the contrast agents and ultrasound technology have enabled the improvement of detection of myocardial perfusion using intravenous contrast application. The evaluation of myocardial perfusion is an important component of the risk stratification of patients with ischemic heart diseases. Harmonic power Doppler imaging (HPDI) has emerged as a promising tool to detect myocardial perfusion after intravenous injection of contrast agents. MCE may be more versatile than perfusion scintigraphy for identifying the presence and extent of perfusion defects after MI. Recent studies indicated that HPDI can reliably detect myocardial perfusion at rest and during pharmacological stress [13,14]. Previous works, however, used qualitative analysis of the HPDI. Digital acquisition of HPDI should lend itself to quantitative analysis, which may be more accurate in distinguishing normal perfusion from mild defects. In addition, quantitative analysis offers the potential to measure flow reserve ratio noninvasively.
The aim of this study was to compare the MCE results with myocardial viability evaluated by CFR two weeks after AMI and CF of LAD at acute phase measured with high-frequency TTDE echocardiography and evaluated MCE with HPDI, CFR and CF in predicting functional recovery and LV remodeling.
Methods
Patient Population
This study included 30 patients (age 62 ± 10 years, 5 women and 25 men) with the first anterior AMI. All patients underwent successful revascularization by PTCA and stent placement in the acute phase (within 6 hours after the onset of chest pain). Inclusion criteria were typical anginal chest pain lasting > 30 min and ST-segment elevation of > 0.2 mV in at least two contiguous electrocardiographic leads. And inclusion criteria were also echocardiographic images of sufficient quality to allow adequate visualization of all myocardial segments from the apical views and angiographically documented patency of the infarct-related artery (TIMI grade 3 flow) on coronary angiograms taken in the repeated study. Six patients were excluded from analysis because of inadequate image quality of echocardiograpic images and coronary flow by TTDE.
Exclusion criteria were pregnancy, lactation, unstable angina and second or third degree atrioventricular block, bronchial asthma and systolic blood pressure < 90 mmHg. All patients gave written informed consent.
Myocardial Contrast Echocardiography
Transthoracic 2-dimensional (2D) echocardiographic images were obtained with an ultrasound imaging system (Sonos 5500, Philips Medical systems, Andover, MA, USA). MCE was performed two weeks after AMI. The contrast agent used for this study was Levovist™, which is composed of galactose (mean diameter 1.2 μm). MCE was performed by administration of Levovist at a dose of 300 mg/dl by injecting 3 ml/3 sec using by Medrad Pulser™ (Ultrasound injection system, MEDRAD Service Department, Indianola, PA), followed by a 10 ml saline flushed. Contrast enhanced images using HPDI mode (scanhead transmit to receive frequency = 1.8:3.6 MHz) were acquired on intermittent mode at each pulse intervals of 4 cardiac cycles with the ultrasound transmission gated to the T wave of the electrocardiogram (i.e. end-systole). The dynamic range of this system is 40 dB. The mechanical index was set as high as possible to increase microbubble destruction. Ultrasound system gains were optimized at the beginning of the study and held constant for subsequent image acquisitions. The filter threshold was optimized to reduce the appearance of any color over the myocardium before contrast injection. The apical 4-chamber view was used to assess myocardial perfusion in all patients. We set the focus point at the mid-portion of the LV and kept constant for the quantitative analysis at baseline and during ATP stress. Images were stored on a 2.3 Gbytes MO disk.
We examined MCE under conditions at initial baseline and during an infusion of ATP. The dosage of ATP was selected of 0.15 mg/kg/min. Levovist was pushed into the intravenous infusion at baseline imaging and during ATP infusion for stress imaging. Since peak hyperemia begins 2 minutes after the ATP infusion, image acquisition was initiated 2 minutes after the start of the infusion. Since only 2–2.5 minutes of maximal hyperemia was available for imaging, HPDI was performed only in the apical views by obtaining 3 to 5 beats gated to every fourth cardiac cycle.
Image Analysis of Echocardiogram
The analysis was performed off-line using QuantiCon software (Echo Tech 3D Imaging Systems, Germany) for the quantification of contrast based ultrasound images. Regions-of-interest (ROIs) were placed in the basal septum, mid septum, apex and apical portion of the lateral segment. The ROIs can be individually repositioned in each frame to match the anatomy of the unaligned images. The information is displayed as the number of scatters that have moved in the power mode. Thus, intensity curves of the mean Doppler information in the ROIs can be plotted overtime using this software. Peak intensities of color pixels (dB) were then measured in 4 different segments of the LV. The peak intensity ratios of the risk area (apical segment) to the control area (PIR) at rest and during hyperemia were calculated. The control region was set in the basal septum in this study (Fig. 1).
Figure 1 Quantitative analysis of harmonic power Doppler imaging. A. Four different regions of interest are drawn on the digital clip. The ROI can be individually repositioned in each frame to match the anatomy. B. The graph shows time-intensity curves of the mean Doppler information (dB) from the ROIs. y1 = basal septum, y2 = mid septum, y3 = apex, y4 = apical portion of lateral segment.
Doppler Echocardiographic Studies
We performed Doppler measurements of LAD flow velocity by TTDE within 24 hours after successful recanalization. Doppler echocardiographic examinations were performed with an SONOS 5500 digital ultrasound system with a broadband high frequency (5-12MHz) transducer (S12) and a Logic 500 digital ultrasound system with a frequency of 8 MHz (GE medical system). The color gain was adjusted to provide optimal images. The acoustic window was around the midclavicular line in the fourth and fifth intercostal. The ultrasound beam was transmitted toward the heart to visualize coronary blood flow in the LAD by color Doppler echocardiography. With a sample volume positioned on the color signal in the LAD, Doppler spectral tracings of flow velocity in the LAD were recorded by fast Fourier transformation analysis. All studies were continuously recorded on 1/2-inch super-VHS videotape for off-line analysis.
The digitized coronary blood flow velocity spectrum provided the following parameters: peak systolic velocity (PSV: cm/sec), mean diastolic velocity (MDV: cm/sec), peak diastolic velocity (PDV: cm/sec), and deceleration time of diastolic flow velocity (DDT: msec). Retrograde flow was calculated as a negative value (Fig. 2).
Figure 2 Measurement of parameters from systolic and diastolic flow velocity patterns. By tracing the contour the coronary flow velocity wave form, we measured the peak diastolic velocity (PDV: cm/sec), mean diastolic velocity (MDV: cm/sec), peak systolic velocity (PSV: cm/sec) and deceleration time of diastolic flow velocity (DDT: msec).
CFR Measurements by TTDE
We measured CFR of LAD two weeks after AMI at the same time of MCE. At first, we recorded baseline spectral Doppler signals in the distal portion of the LAD. Then, ATP was administered for two minutes and spectral Doppler signals were recorded during hyperemic conditions. All patients had continuous heart rate and ECG monitoring. Blood pressure was recorded at baseline, every minute during ATP infusion and at recovery. One experienced investigator who was unaware of the other patient data analyzed each study. Measurements were performed off-line by tracing the contour of the spectral Doppler signal using the computer incorporated in the ultrasound system. Mean diastolic velocity (MDV) and peak diastolic velocity (PDV) were measured at baseline and peak hyperemic conditions. CFR was defined as the ratio of hyperemic to basal peak diastolic coronary flow velocity.
Two-dimensional Echocardiographic Measurements of LV Function at Baseline
Before CFR measurements by TTDE, we measured LV wall (septal and posterior wall thickness) at end diastole by two-dimensional echocardiography. LV volume measurements were performed according to the recommendation of the American Society of Echocardiography. Apical two- and four-chamber views were obtained at baseline. End-diastolic and end-systolic LV volumes were computed by use of modified Simpson's method (method of disks). Furthermore, we assessed regional wall motion at rest on the basis of 16 segments of the LV as recommended by the American Society of Echocardiography.
Coronary Angiography
Left heart catheterization was performed by the femoral approach after local anesthesia induced with 0.5% lidocaine. Biplane left ventriculography was performed after injection of 4000 IU IV heparin to assess LV wall motion and to measure LV volume by the area-length method in the acute phase and 6 months after AMI. Selective coronary angiography was carried out by the Judkins technique after an intravenous injection of 3 mg of isosorbide dinitrate. Coronary angiography was analyzed quantitatively with the use of videodensitometric analysis performed with a commercially available system (CAMAC-300, Goodman, Inc.) in the manner previously reported [15].
Statistical Analysis
Data were expressed as a mean ± standard deviation. Continuous data between groups were compared by student t-test, correlation between Doppler parameters and PIR was performed using linear regression analysis. Statistical significance was defined as p < 0.05.
Receiver operating characteristics (ROC) curve analysis was performed. Analyses were done with SPSS for Windows software.
Results
Patient Characteristics
The clinical characteristics of the patients are summarized in Table 1. The mean time interval between the MCE studies and coronary angiography was 16 ± 2 days. There were no clinical events among these examinations in any of the patients. In accordance with previous findings[7], patients were divided into two groups based on CF pattern: group A (n = 20) with DDT ≧ 600 msec and group B (n = 10) with DDT ≧ 600 msec. Representative coronary flow pattern is shown in Fig. 3. All patients received conventional drug therapy based on individual needs, as determined by the attending physician.
Table 1 Patient Characteristics
Group A Group B
Number 20 10
Age 63 ± 12 64 ± 10
Men (%) 17 (85%) 8 (80%)
Cardiovascular risk factors
Diabetes Mellitus 7 4
Hypertension 11 6
Smoking 12 7
Total Cholesterol (mg/dl) 200 ± 38 189 ± 13
HDL-cholesterol (mg/dl) 43 ± 8 48 ± 12
Triglyceride (mg/dl) 155 ± 94 105 ± 44
Visible Collaterals to Infarct-related artery 5/20 3/10
Referece Diameter (mm) 3.4 ± 0.7 3.1 ± 0.5
TIMI 3 flow after PCI 18/20 5/10*
*p < 0.01
PCI: Percutaneous Coronary Interventional Therapy
TIMI: Thrombolysis in Myocardial Infarction
Figure 3 Representative coronary flow pattern in Group A (A) and Group B (B).
Myocardial Contrast Echocardiography
There were no significant changes in the average heart rate (65 ± 4 vs. 77 ± 9, p = 0.08), systolic blood pressure (118 ± 11 vs. 94 ± 26, p = 0.21), and diastolic blood pressure (59 ± 8 vs. 58 ± 15, p = 0.96) after intravenous ATP administration in all patients. ATP did not cause significant side effects in the study patients. Figure 4 shows a representative perfusion image in a patient with anterior AMI of Group A. MCE revealed mild perfusion defect in the risk area and improved perfusion defect during hyperemia. Figure 5 shows a representative perfusion image in a patient with anterior AMI of Group B. There was a contrast perfusion defect in the risk area by MCE. In patients of Group A, PIR both at rest and during hyperemia of HPDI were significantly higher compared with those in patients of Group B (at rest: 0.668 ± 0.178 vs. 0.248 ± 0.015, p < 0.0001: during hyperemia 0.725 ± 0.194 vs. 0.295 ± 0.107, p < 0.0001).
Figure 4 The results of MCE and CFR in a patient with Group A. 1) HPDI showing increase in contrast signals in the anteroseptal and apical lesion in 4-chamber view during ATP stress image (B) when compared with baseline image (A). 2) PDV in the infarct-related artery increased during ATP stress (D) compared with baseline (C) by TTDE.
Figure 5 The results of MCE and CFR in a patient with Group B. 1) HPDI showing resting apical perfusion defect (A) and without improvement of perfusion during ATP stress (B) in the anteroseptal and apical lesion. 2) PDV in the infarct-related artery slightly increased during ATP stress (D) compared with baseline (C) by TTDE
According to receiver operating characteristics curve analysis, the optimal cutoff value to identify myocardial viability by MCE was 861 msec for DDT (sensitivity 87%, specificity 80%) and 1.8 for CFR (sensitivity 87%, specificity 73%).
Coronary Blood Flow Pattern and Coronary Flow Reserve
FR in patients of Group A was significantly higher compared with CFR of Group B (2.60 ± 0.80 vs. 1.31 ± 0.29, p < 0.0001). In patients without a perfusion defect in the risk area by MCE, CFR was > 2.0 (Table 2). Representative imagings of MCE and CFR in two groups were shown in Fig. 4 and Fig. 5.
Table 2 Differences of Parameters between Two Groups
Group A Group B
PIR at rest 0.668 ± 0.178 0.248 ± 0.015#
PIR at ATP stress 0.725 ± 0.194 0.295 ± 0.107#
EDV in the acute phase (ml) 112 ± 33 146 ± 38
EDV at follow-up (ml) 114 ± 30 192 ± 39*
EF in the acute phase (%) 50 ± 7 45 ± 9
EF at follow-up (%) 52 ± 7 39 ± 9*
LVWMI in the acute phase 2.1 ± 0.7 2.4 ± 0.3
LVWMI at follow-up 1.7 ± 0.4 2.4 ± 0.3*
Peak creatine kinase (U/L) 2734 ± 868 6198 ± 2265*
CFR 2.60 ± 0.8 1.31 ± 0.29*
#p < 0.05, *p < 0.01
PIR, peak intensity ratio; EDV, end-diastolic volume; EF, ejection fraction; CFR, coronary blood flow reserve; LVWMI, left ventricular wall motion index
DDT of Group A was significantly longer than that of Group B (985 ± 136 vs. 222 ± 115, p < 0.0001). CFR was correlated with DDT (p < 0.0001, r2Q = 0.566).
PSV of Group A was significantly larger than that of Group B (12.2 ± 2.81 vs. -22.0 ± 23.5, p = 0.0006). PDV of group A was significantly smaller than that of Group B (23.5 ± 6.19 vs. 45.1 ± 21.8, p = 0.012).
According to present/absent post-infarct LV remodelling at follow-up, I divided two groups (Table 3).
Table 3 Differences of Parameters between Two Groups
Group 1 Group 2 p
PIR at rest 0.700 ± 0.161 0.334 ± 0.162 0.0001
PIR at ATP stress 0.736 ± 0.151 0.358 ± 0.057 0.0007
Peak creatine kinase (U/L) 1526 ± 1021 3724 ± 684 0.02
DDT 997 ± 161 427 ± 350 0.0003
CFR 2.35 ± 0.50 1.71 ± 0.72 0.03
Group 1: LV remodeling(-)
Group 2: LV remodeling(+)
Left Ventricular Volume and Function
In the acute phase, there were no significant differences of LVEDV and EF between the two groups.
LVEDV of Group B was, however, significantly larger than that of Group A, and EF of Group B was significantly decreased as compared with EF of Group A 6 months after AMI (Table 2).
LV Wall Motion Recovery and CFR, CF Pattern
According to receiver operating characteristics curve analysis, the optimal cutoff value to predict wall motion recovery was 809 ms for DDT (sensitivity 87%, specificity 73%) and 2.0 for CFR (sensitivity 87%, specificity 87%) (Fig. 6).
Figure 6 The result of receiver operating characteristics curve about DDT and CFR to predict wall motion recovery.
Discussion
In patients with reperfused anterior AMI, a shorter DDT of the LAD flow detected within 24 hrs after coronary reperfusion and CFR less than 1.75 were feasible predictors of poor functional outcome. These parameters correlated with the quantitative results by HPDI with intravenous MCE. TTDE provides a simple and promising means of noninvasive assessment of myocardial viability in patients with reperfused anterior AMI.
Coronary Flow Pattern
Recent studies with intracoronary MCE have shown that about one fourth to one third of patients with AMI treated with primary PTCA have an inadequate tissue perfusion (no-reflow phenomenon) despite angiographically successful coronary recanalization [16]. Characteristic coronary flow velocity pattern (systolic retrograde flow and rapid deceleration of diastolic flow) by using a Doppler guidewire is associated with the "no-reflow" phenomenon [17]. This "no-reflow" phenomenon is thought to be the result of microvascular dysfunction.
In patients with MCE no reflow, the coronary microvasculature was profoundly damaged and it seemed that microvascular impedance increased and the intramyocardial blood pool decreased. The coronary microvasculature would be diffusely obstructed in patients with MCE no reflow. Early systolic retrograde flow (ESRF) has been reported to be observed frequently in patients with no reflow in case of AMI after successful recanalization and was explained by an occluded coronary microvasculature.
Kawamoto et al. showed that the degree of reduced systolic antegrade flow or the deceleration time of diastolic flow reflected the degree of microvasculature damage and was predictive of residual myocardial viability [7]. They revealed that low average peak velocity and rapid DDT of coronary blood flow spectrum immediately after primary PTCA reflected a greater degree of microvascular damage in the risk area and analysis of coronary blood flow spectrum immediately after primary PTCA by use of a Doppler guidewire was useful in predicting recovery of regional LV function.
Recently, Wakatsuki et al. revealed that the coronary flow velocity pattern measured immediately after successful primary stenting is predictive of the recovery of regional and global LV function in patients with AMI [18]. The changes of regional wall motion score (RWM) and EF were significantly greater in the non-ESRF group than it was in the ESRF group. These authors reported that ESRF is a parameter predicting poor functional recovery of LV wall motion. They explained that decreased extravascular pressure of the infarcted myocardium during systole might increase the apparent systolic flow, and increased vascular resistance might decrease the diastolic antegrade flow in mildly to moderately damaged myocardium.
In our study, there were 6 patients with early systolic retrograde flow. They had no myocardial perfusion in infarcted area by HPDI and revealed poor CFR. LV wall motion in all these patients did not recover at follow-up. The obstruction of the microvasculature with subsequent high impedance results in the inability to squeeze blood forward into the venous circulation during systole. It is consequently pushed back to the epicardial coronary artery to produce early systolic retrograde flow. The reduced intramyocardial blood pool, which fills rapidly during diastole, has been used to explain the rapid decline of diastolic velocity.
Recently, Lepper et al. reported that the coronary flow reserve immediately and 24 hr after PTCA for AMI relates to myocardial perfusion determined by MCE and LV function in four weeks [19]. From our results, the patients with significantly altered coronary flow pattern, which showed rapid DDT, were found to have subsequent depression of LV function at follow up, confirming the prognostic importance of altered coronary blood flow patterns reported in previous studies [6,7]. This study confirms recent reports on the difference in coronary blood flow pattern between patients with and without reperfusion determined by MCE.
There were no reports that were compared coronary blood flow velocity pattern and CFR by TTDE with intravenous MCE by using HPDI. Our study revealed that DDT at acute phase correlated with microvascular integrity by HPDI and with CFR after two weeks. The quantitative analysis by HPDI revealed the degree of microvascular dysfunction. HPDI is a feasible technique for the detection of a myocardial perfusion defect in patients with coronary artery disease after a venous injection of contrast agent. Previous works, however, used qualitative analysis of the HPDI [14]. In the present study, we performed quantitative analysis of HPDI at rest and during hyperemia in patients with AMI.
We used Levovist, which is sensitive to microbubble destruction at diagnostic ultrasound frequencies [19]. Several recent studies have demonstrated that the feasibility of applying this contrast agent to the assessment of myocardial perfusion by using HPDI [21].
In patients who had myocardial viability, the peak intensity of the risk area was 0.668 ± 0.178 dB at rest. In patients who had no myocardial viability in the risk area, the curves of intensity showed no increase and peak intensity was 0.248 ± 0.015 dB at rest. During ATP induced hyperemia, peak intensities were increased in the segments with preserved myocardial integrity. In the segments without preserved myocardial integrity, however, there was no change in peak intensities. We diagnosed quantitatively myocardial viability using by HPDI. HPDI is strictly dependent on microvascular integrity and PIR shows quantitatively the degree of microvascular damage. There were significant differences of LV function at follow-up between two groups. Our data may provide additional information from the point of coronary flow pattern.
Coronary Artery Flow Reserve
Our main findings were that the myocardial perfusion status assessed by HPDI at rest and during ATP stress corresponds closely to CFR and DDT of CF. CF at acute phase and CFR two weeks after AMI onset correspond to left ventricular remodelling at chronic phase.
A previous study showed that, in patients with myocardial infarction, CFR is inversely correlated with the extent of myocardial infarction and directly correlated with the improvement in wall motion contractility during the recovery period [4,6]. A value above 1.75 is associated with an improved wall motion index. Reduction of MCE perfusion defects were associated with improvement of CFR, whereas persistent MCE perfusion defects were associated with unchanged depression of CFR, indicating a relation between microvascular integrity assessed by CFR and by intravenous MCE [22].
There were no reports that had been investigated the relationship between CFR and DDT of LAD. From our study, CFR correlated with DDT. CFR and DDT revealed the degree of microvascular damage in myocardium.
The patients with MCE no reflow, which had no viability in the infarcted area by HPDI, showed poor functional outcomes and left ventricular remodeling. The detection of viability with HPDI and CFR were useful for predicting of LV remodeling.
Study Limitations
We compared the myocardial opacification obtained by HPDI with CFR and DDT of LAD in patients after PTCA for AMI. The machine settings used in this study relate to the best knowledge at the time of study initiation. Optimal echocardiographic machine settings for MCE are rapidly evolving and are dependent on the applied contrast agent. Thus, it is very challenging to set up and adhere to a study protocol in a field in which knowledge of how to use an evolving technology is improving very quickly.
The differences in the shell structure as well as gas compositions of different microbubbles are likely to influence their efficacy for myocardial perfusion assessment by HPDI. Unfortunately, real-time perfusion imaging is not available on the instrument used in this study. Using real-time assessment of perfusion defects, we can diagnose both wall motion and myocardial perfusion at the same time [23].
We could not perform MCE study immediately after the primary angioplasty. This analysis included only a limited number of patients.
Coronary flow reserve is related to the microcirculatory status, which will be either indirectly affected by epicardial coronary stenosis or directly affected by a previous myocardial infarction and other related factors.
The coronary flow pattern may be influenced by left ventricular pressure. But we didn't compare CFR and coronary flow pattern with the left ventricular pressure. Further study is needed to clarify this issue. Systolic reversal flow is a specific indication of the no reflow phenomenon, but it was sometimes difficult to record the complete Doppler spectral envelope throughout the entire cardiac cycle by TTDE because of the systolic heart motion.
Conclusion
Coronary flow velocity patterns relate to myocardial perfusion determined by intravenous MCE and CFR by TTDE.
Intravenous MCE and intermittent harmonic power Doppler imaging has the potential to noninvasively identify myocardial viability in patients with a previous anterior MI. Quantitative assessment of microvascular integrity corresponds to the evaluation of the microcirculation by CFR using transthoracic echocardiography.
The assessment of microvascular damage by HPDI, CFR and CF corresponds to left ventricular remodeling.
DDT and PIR were more useful in detecting functional recovery.
List of Abbreviations
AMI-acute myocardial infarction
MCE-myocardial contrast echocardiography
CFR-coronary blood flow reserve
CF-coronary artery flow pattern
LV-left ventricular
TTDE-transthoracic Doppler echocardiography
HPDI-harmonic power Doppler imaging
PIR-peak intensity ratios of the risk area to the control area
LAD-left anterior descending artery
DDT-deceleration time of diastolic flow velocity
Competing interests
The author(s) declare that they have no competing interests.
==== Refs
Doucette JW Corl PD Payne HM Validation of a Doppler guide wire for intravascular measurement of coronary artery flow velocity Circulation 1992 85 1899 1911 1572046
Akasaka T Yoshida K Hozumi T Takagi T Kaji S Kawamoto T Morioka S Yoshikawa J Retinopathy identifies marked restriction of coronary flow reserve in patients with diabetes mellitus J Am Coll Cardiol 1997 30 935 941 9316521 10.1016/S0735-1097(97)00242-8
Dimitrow PP Transthoracic Doppler echocardiography-noninvasive diagnostic window for coronary flow reserve assessment Cardiovascular Ultrasound 2003 11 4 12740038 10.1186/1476-7120-1-4
Teiger E Garot J Aptecar E Bosio P Woscoboinik J Pernes JM Gueret P Kern M Dubois-Rande JL Dupouy P Coronary blood flow reserve and wall motion recovery in patients undergoing angioplasty for myocardial infarction Eur Heart Journal 1999 20 285 292 10.1053/euhj.1998.1195
Lim DS Kim YH Lee HS Park CG Seo HS Shim WJ Oh DJ Ro YM Coronary flow reserve is reflective of myocardial perfusion status in acute anterior myocardial infarction Cathet Cardiovasc Intervent 2000 51 281 286 10.1002/1522-726X(200011)51:3<281::AID-CCD7>3.0.CO;2-O
Rigo F Varga Z Di Pede F Grassi G Turiano G Zuin G Coli U Raviele A Picano E Early assessment of coronary flow reserve by transthoracic Doppler echocardiography predicts late remodeling in reperfused anterior myocardial infarction J Am Soc Echocardiogr 2004 17 750 755 15220900
Kawamto T Yoshida K Akasaka T Hozumi T Takagi T Kaji S Ueda Y Can coronary blood flow velocity pattern after primary percutaneous transluminal coronary angiography predict recovery of regional left ventricular function in patients with acute myocardial infarction? Circulation 1999 100 339 345 10421592
Shintani Y Ito H Iwakura K Sugimoto K Yamamoto K Masuyama T Kuzuya T Hori M Higashino Y Fujii K Prediction of wall motion recovery from the left anterior descending coronary artery velocity pattern recorded by transthoracic Doppler echocardiography in patients with anterior wall myocardial infarction. Retrospective and prospective studies Jpn Circ J 2001 65 717 722 11502048 10.1253/jcj.65.717
Cheirif J Desir RM Bolli R Mahmarian JJ Zoghbi WA Verani MS Quinones MA Relation of perfusion defects observed with myocardial contrast echocardiography to the severity of coronary stenosis: correlation with thallium-201 single-photon emission tomography J Am Coll Cardiol 1992 19 1343 1349 1564236
Ito H Iwakura K Oh H Masuyama T Hori M Higashino Y Fujii K Minamimoto T Temporal changes in myocardial perfusion patterns in patients with reperfused anterior wall myocardial infarction. Their relation to myocardial viability Circulation 1995 91 656 662 7828290
Ragosta M Camarano G Kaul S Powers ER Sarembock IJ Gimple LW Microvascular integrity indicates myocellular viability in patients with recent myocardial infarction. New insights using myocardial contrast echocardiography Circulation 1994 89 2562 2569 8205665
Marwick TH Brunken R Meland N Brochet E Baer FM Binder T Flachskampf F Kamp O Nienaber C Nihoyannopoulos P Pierard L Vanoverschelde JL van der Woun P Lindvall K Accuracy and feasibility of contrast echocardiography for detection of perfusion defects in routine practice: comparison with wall motion and technetium-99m sestamibi single-photon emission computed tomography J Am Coll Cardiol 1998 32 1260 1269 9809934 10.1016/S0735-1097(98)00373-8
Senior R Kaul S Soman P Lahiri A Power Doppler harmonic imaging: A feasibility study of a new technique for the assessment of myocardial perfusion Am Heart J 2000 139 245 251 10650297
Heinle SK Noblin J Goree-Best P Mello A Ravad G Mull S Mammen P Grayburn PA Assessment of myocardial perfusion by harmonic power Doppler at rest and during adenosine stress. Comparison with 99mTc-Sestamibi SPECT imaging Circulation 2000 102 55 60 10880415
Hoshio A Shirota K Sawada Y Fukuki M Doi T Kotake H Mashiba H Kasahara T Endo S Comparison of coronary responses to intracoronary and intraaortic ergonovine and isosorbide dinitrate in patients with atypical chest pain Am J Cardiol 1992 70 1004 1009 1414896 10.1016/0002-9149(92)90351-X
Ito H Tomooka T Sakai N Yu H Higashino Y Fujii K Masuyama T Kitabatake A Minamino T Lack of myocardial perfusion immediately after successful thrombolysis: a predictor of poor recovery of left ventricular function in anterior myocardial infarction Circulation 1992 85 1699 1705 1572028
Iwakura K Ito H Takiuchi S Taniyama Y Nakatsuchi Y Negoro S Higashino Y Okamura A Masuyama T Hori M Fujii K Minamimoto T Alternation in the coronary blood flow velocity pattern in patients with no reflow and reperfused acute myocardial infarction Circulation 1996 94 1269 1275 8822979
Wakatsuki T Nakamura M Tsunoda T Toma H Degawa T Oki T Yamaguchi T Coronary flow velocity immediately after primary coronary stenting as a predictor of ventricular wall motion recovery in acute myocardial infarction J Am Coll Cardiol 2000 35 1835 1841 10841232 10.1016/S0735-1097(00)00632-X
Lepper W Sieswerda GT Franke A Henssen N Kamp O de Cock CC Schwarz ER Voci P Visser CA Hanrath P Hoffmann R Repeated assessment of coronary flow velocity pattern in patients with first acute myocardial infarction J Am Coll Cardio 2002 39 1283 1289 10.1016/S0735-1097(02)01753-9
Wei K Skyba DM Firschke C Jawayeera AR Lindner JR Kaul S Interactions between microbubbles and ultrasound: In vitro and in vivo observations J Am Coll Cardiol 1997 29 1081 1088 9120163 10.1016/S0735-1097(97)00029-6
Agati L Funaro S Bilotta F Assessment of no-reflow phenomenon after acute myocardial infarction with harmonic angiography and intravenous pump infusion with Levovist: comparison with intracoronary contrast injection J Am Soc Echocardiogr 2001 14 773 781 11490325
Lepper W Hoffmann R Kamp O Franke A de Cock CC Kuhl HP Sieswerda GT Dahl J Janssens U Voci P Visser CA Hanrath P Assessment of myocardial reperfusion by intravenous myocardial contrast echocardiography and coronary flow reserve after primary percutaneous transluminal coronary angiography in patients with acute myocardial infarction Circulation 2000 101 2368 2374 10821812
Tiemann K Lohmeier S Kuntz S Koster Pohl C Burns PRT Nand NC Luderitz B Becher H Real-time contrast echo assessment of myocardial perfusion at low emission power. First experimental and clinical results using power pulse inversion Echocardiography 1999 16 799 809 11175224
|
16107221
|
PMC1201156
|
CC BY
|
2021-01-04 16:38:32
|
no
|
Cardiovasc Ultrasound. 2005 Aug 18; 3:22
|
utf-8
|
Cardiovasc Ultrasound
| 2,005 |
10.1186/1476-7120-3-22
|
oa_comm
|
==== Front
Cell Commun SignalCell communication and signaling : CCS1478-811XBioMed Central London 1478-811X-3-91607638810.1186/1478-811X-3-9ResearchAnti-lipid phosphate phosphohydrolase-3 (LPP3) antibody inhibits bFGF- and VEGF-induced capillary morphogenesis of endothelial cells Wary Kishore K [email protected] Joseph O [email protected] Institute of Biosciences and Technology, Texas A&M University System Health Science Center, Texas Medical Center, 2121 W. Holcombe Blvd., Houston TX-77030, USA2 Department of Cell and Tissue Biology, University of California San Francisco, 521 Parnassus Ave., CA-94143, USA2005 2 8 2005 3 9 9 3 6 2005 2 8 2005 Copyright © 2005 Wary and Humtsoe; licensee BioMed Central Ltd.2005Wary and Humtsoe; 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
Angiogenesis, or the remodeling of existing vasculature serves as a lifeline to nourish developing embryos and starved tissues, and to accelerate wound healing, diabetic retinopathy, and tumor progression. Recent studies indicate that angiogenesis requires growth factor activity as well as cell adhesion events mediated by α5β1 and αvβ3 integrins. We previously demonstrated that human lipid phosphate phosphohydrolase-3 (LPP3) acts as a cell-associated ligand for α5β1 and αvβ3 integrins. Here, we test the hypothesis that an anti-LPP3 antibody can inhibit basic fibroblast growth factor (bFGF)-and vascular endothelial growth factor (VEGF)-induced capillary morphogenesis of endothelial cells (ECs).
Results
We report that bFGF and VEGF up-regulate LPP3 protein expression in ECs. Immunoprecipitation analyses show that LPP3 is a cell surface protein and undergoes N-glycosylation. Fluorescent activated cell sorting (FACS) data suggest that anti-LPP3-RGD detects native neoepitope on the surface of activated ECs. Moreover, we demonstrate LPP3 protein expression in tumor endothelium alongside VEGF. The embedding of ECs into three-dimensional type I collagen in the presence of bFGF and VEGF induce capillary formation. Importantly, we show that the addition of an anti-LPP3 antibody specifically and significantly blocks bFGF- and VEGF-induced capillary morphogenesis of ECs.
Conclusion
These data suggest that activated ECs as well as tumor endothelium express LPP3 protein. In an in vitro assay, the anti-LPP3-RGD specifically blocks bFGF and VEGF induced capillary morphogenesis of ECs. Our results, therefore, suggest a role for LPP3 in angiogenesis.
bFGFcapillary morphogenesiscollagen matricesendothelial cellsVCIPVEGF
==== Body
Background
Angiogenesis, the sprouting or remodeling of preexisting quiescent blood vessels, is critical for embryonic development, wound healing, and various pathological conditions such as tumor progression, complications associated with acquired immune deficiency syndrome (AIDS), rheumatoid arthritis, and diabetic retinopathy [1-4]. Angiogenesis can be initiated by hypoxic tumors, inflammation or an increased accumulation of pro-angiogenic factors. These factors, in turn, trigger secretion of matrix metalloproteinases (MMPs) that dissolve the basement membrane. This MMP-mediated membrane dissolution is an essential event for subsequent EC activation, migration, and capillary formation [1-6]. Angiogenesis is regulated through a dynamic balance between pro- and anti-angiogenic factors [1-4]. Angiogenic mediators include growth factors such as basic fibroblast growth factor (bFGF), vascular endothelial growth factor (VEGF), collagen and fibronectin, and proteases such as MMPs [2,4,6-8]. VEGF signaling activates ECs through VEGF receptor-1 (VEGFR-1, also known as Flt) and VEGFR2 (KDR/Flk-1) tyrosine kinase receptors, and promotes cell migration, survival, proliferation and differentiation [5,6,9]. The microenvironment surrounding a tumor is generally rich in VEGF, which is upregulated in response to hypoxia and can directly activate ECs to initiate tumor angiogenesis, growth and metastatic deposits [1-4,9]. Both bFGF and VEGF are able to induce tumor angiogenesis and wound healing, as well as contribute to unwanted angiogenesis [2,4-6,9]. Addition of bFGF and VEGF can increase the expression of EC integrins, a family of cell surface receptors that regulate cell adhesion events [2,10-13]. In particular, α5β1 and αvβ3 integrins mediate adhesion, migration, and proliferation of endothelial cells by interacting with extracellular matrix (ECM) proteins such as fibronectin, fibrin, and vitronectin [13-15]. In addition, integrins also mediate cell-cell interactions by associating with counter-receptors or cell associated integrin ligands [16,17]; such interactions generate both chemical and mechanical signals that influence cellular behavior [18-22].
Our ability to target neo-epitopes expressed by tumor-endothelium could potentially minimize the toxicity and drug-resistance associated with conventional chemotherapy treatment of solid tumors [9]. Recently, we identified lipid phosphate phosphohydrolase-3 (LPP3), also called phosphatidic acid phosphatase-2b (PAP2b), VEGF and type I collagen inducible protein (VCIP) in a functional assay of angiogenesis [23,24]. Lipid phosphate phosphohydrolases (LPPs) dephosphorylate polar lipid signaling molecules, both within and outside cells [25-27]. Structurally, all LPPs display a 6-transmembrane channel-like organization [29-32]. Both the N-and C-terminal segments are located in the cytoplasm [32,33]. There are three extracellular loops, and the proposed 2nd extracellular loop of LPP3 contains a lipid phosphatase, one cell-adhesion motif, and a N-glycosylation site [23,29,32,33]. LPP3 protein has been identified within intracellular organelles as well as on the cell surface, and in both locations it exhibits ectoenzyme activity [29-32]. Previously we have shown that LPP3-RGD (RGE in mice) can act as a cell-associated integrin ligand and mediate cell-cell interactions [23,28]. Consistent with our findings, confocal image analyses demonstrated that green fluorescent protein-LPPl remains apically sorted, whereas green fluorescent protein-LPP3 co-localized with E-cadherin in cell-cell junctions and the basolateral domains of polarized MDCK cells [23,33]. Transfection of mutants as well as swapping experiments have established that LPP1 protein contains an apical targeting signal sequence (FDKTRL) in its N-terminal segment; in contrast, LPP3 protein contains dityrosine (109Y/110Y) cell-cell and basolateral sorting motifs [33]. Unlike Lpp2, whose function is dispensable for embryonic development, Lpp3 is required for extra-embryonic vasculogenesis and axis patterning [34,35], raising the possibility that the function of the LPP3 protein may also be to mediate adult, as well as pathological, angiogenesis.
We previously showed that anti-LPP3-RGD blocks cell aggregation (cell-cell interactions) that is mediated by α5β1 and αvβ3 integrins [23]. In the current study, we examine whether an anti-LPP3-RGD antibody can inhibit bFGF- and VEGF-mediated capillary morphogenesis of ECs. In this study, we demonstrate that the addition of bFGF and VEGF angiogenic cytokines stimulate the expression of LPP3 protein of ECs. We further show that tumor endothelium express LPP3 protein. By embedding ECs in a three-dimensional type I collagen matrix followed by treatment with bFGF and VEGF to induce formation of capillaries, we demonstrate the ability of anti-LPP3 antibodies to inhibit bFGF- and VEGF-induced capillary morphogenesis. These findings are the first to our knowledge to suggest a mechanism by which anti-LPP3-RGD antibodies may inhibit capillary morphogenesis of ECs.
Results
Basic FGF and VEGF induce expression of LPP3 in HUVECs
Hypoxic tumors in vivo and many cell lines in vitro secrete bFGF and VEGF. Both bFGF and VEGF are components of the tumor microenvironment capable of activating ECs. To evaluate the potential role of LPP3 in angiogenesis, we investigated the effects of treatment of HUVECs with VEGF and bFGF. We stimulated monolayer HUVECs with either VEGF165 or bFGF for various time periods between 0 and 18 h, and subjected lysates to Western blot analyses using an affinity purified anti-LPP3-cyto antibody (Fig. 1A). The expression of LPP3 protein levels was increased by >3-fold in response to VEGF165 treatment (100 ng/ml) for 6 or 12 h (relative to control levels), whereas bFGF (20 ng/ml) had a significantly less robust effect on LPP3 levels during the same treatment duration (Fig. 1A). The concentrations of VEGF165 and bFGF used in this experiment were optimal as evidenced by the observation that both factors activated extracellular-signal-regulated kinase (Erkl/2) in these cells in an independent experiment (data not shown). We observed that anti-LPP3-c-cyto antibody detects three major polypeptides, two of which (~52 and ~46 kDa) are slow and one (~36 kDa) that exhibits high mobility (Fig. 1A). Since, LPP3 contains a single consensus N-glycosylation site (170N) on the proposed 2nd extracellular loop of the LPP3 protein, the high mobility anti-LPP3-c-cyto immunoreactive species is likely to be a post-translationally modified form. This idea is actually supported by our data with N-glycanase described elsewhere. As a positive control for the cytokines used, the membrane was stripped and re-blotted with an anti-proliferating cell nuclear antigen (PCNA) antibody (Fig. 1B). As expected, both cytokines increased PCNA expression to optimal levels. The level of Focal adhesion kinase (Fak) protein was measured to confirm equivalent protein loading (Fig. 1C).
Figure 1 Expression of LPP3 protein in monolayer ECs. Confluent ECs (passage 4) were starved for 6 h in conditioned medium (M199 supplemented with 0.2% BSA + 1× ITS), and then stimulated with VEGF165 (100 ng/ml) or bFGF (20 ng/ml) for various durations, as indicated. Cells were solubilized, clarified by centrifugation, and the protein concentrations were determined. Samples were subjected to SDS-PAGE and analyzed by immunoblotting with: (A) Rabbit anti-LPP3-c-cyto polyclonal antibody (2.0 μg/ml). Anti-LPP3-c-cyto antibody detects unprocessed LPP3 protein that appears as ~36 kDa, and two major polypeptides migrate below the ~52 kDa molecular weight marker, LPP3 polypeptides are indicated by arrowheads; (B) Anti-PCNA monoclonal antibody (0.5 μg/ml). (C) Anti-Fak monoclonal antibody (1.0 μg/ml). (D) Cell surface biotinylation of intact cells and immunoprecipitation analysis. K562 (lane 1, unstimulated), HUVECs (lane 2, stimulated with VEGF, 100 ng/ml 6 hours), and HUVECs (lane 3, stimulated with bFGF, 20 ng/ml for 6 h) subjected to cell surface biotinylation, lysed in RIPA buffer, clarified by centrifugation and immunoprecipitated using anti-LPP3-c-cyto (5 μg) antibody. Immunocomplexes were resolved by 10% SDS-PAGE under reducing condition and analyzed by ligand blotting with streptavidin-HRP (1:10000). Data shown are representative of those obtained in at least three separate experiments, with similar results. (E) De-N-glycosylation of LPP3 protein. Monolayer HUVECs (passage 4) were stimulated with VEGF165 for 6 h and subjected to cell surface biotinylation, lysed in RIPA buffer, clarified by centrifugation, and immunoprecipitated by either rabbit IgG (control) or anti-LPP3-c-cyto antibodies, as indicated. Following immunoprecipitation with anti-LPP3-c-cyto antibodies, the contents were equally divided into two tubes. One tube was left untreated, and the second tube treated with N-Glyganase (50 units of PNGaseF) enzyme at 37°C for 3 h. Immunocomplexes were analyzed by ligand blotting with streptavidin-HRP (1:10000). Arrowheads indicate LPP3 polypeptides. Arrows indicate unknown polypeptides. Molecular weights are given in kiloDaltons (kDa). Data shown are representative of those obtained in at least three separate experiments, with similar results.
Next, to determine the ability of anti-LPP3-c-cyto antibody to immunoprecipitate LLP3 antigen, we used ECs and human erythroleukemia (K562) cells (Fig. 1D). K562 cells that do not express LPP3 protein was included as a negative control. K562 cells were left unstimulated (Fig. 1D, lane 1), while ECs were stimulated with VEGF (Fig. 1D, lane 2) and bFGF (Fig. 1D, lane 3), and subsequently subjected to cell surface biotinylation and immunoprecipitation with an anti-LPP3-c-cyto antibody and analyzed by ligand blotting with Streptavidin-HRP (Fig. 1D). We found that the anti-LPP3-c-cyto did not immunoprecipitate 36–52 kDa polypeptides from K562 cells (negative control cell line); in contrast, ligand blotting with streptavidin-HRP detected three major polypeptides (36, 42 and 52 kDa, indicated by arrows) (Fig. 1D). These data suggest that the LPP3 antigen is exposed on the extracellular surface of ECs and can be immunoprecipitated by an anti-LPP3-c-cyto antibody.
LPP3 contains a single N-glycosylation site (170N, accession number O14495) [23]. Several pilot experiments suggested that, depending upon how cells are cultured, the variation in the extent of LPP3 glycosylation can be complex and dramatic. To examine this possibility, we prepared cell extracts from ECs and subjected lysates to immunoprecipitation as indicated (Fig. 1E). After several washes with cell lysis buffer, immunoprecipitates were incubated with N-glycanase F (PNGaseF), an enzyme that cleaves the carbohydrate moiety. In doing so, we observed that most of the slow mobility (smeared) polypeptides disappeared, leaving behind a ~36 kDa unprocessed polypeptide (Fig. 1E). Consistent with previous reports, we observed that LPP3 is N-glycosylated [30,31], and the extent of N-glycosylation appears to be cell type- and culture condition-dependent.
The LPP3 is a cell surface antigen
Anti-LPP3-RGD antibody was raised by injecting rabbit with a synthetic peptide modeled after the proposed 2nd extracellular loop of LPP3 protein (EGYIQNYRCRGDDSKVQEAR) [23]. Previously we relied on the specificity of the anti-LPP3-RGD antibody to analyze tumor sections and inhibit LPP3-mediated cell-cell interactions [23]. In the current study, we determined the ability of anti-LPP3-RGD to detect native LPP3 antigen of intact ECs. Towards this end, ECs were either left unstimulated or were stimulated with VEGF, incubated with anti-LPP3-RGD and subjected to fluorescence activated cell sorting (FACS). In contrast to unstimulated ECs that do not express LPP3 protein significantly, the addition of VEGF induced cell surface expression of LPP3 protein of monolayer ECs (Fig. 2A). This result indicates that VEGF stimulates expression of LPP3 antigen on the surface of ECs, which can be detected by anti-LPP3-RGD antibody. This data also suggests that anti-LPP3-RGD antibody detects intact and native LPP3 antigen neoepitope expressed by activated ECs. A schematic diagram showing topological and structural organization of LPP3 is shown (Fig. 2B).
Figure 2 LPP3 is a cell surface antigen. A) HUVECs (passage 4) were serum-starved for 6 h, thereafter, left untreated (-) or treated with VEGF165 (100 ng/ml) for 6 h and subjected to fluorescent activated cell sorting (FACS) using indicated antibodies. B) Schematic representation of human LPP3 protein showing 6-transmembrane organization. Proposed 3 extracellular loops (L-l, -2, and -3) are as shown. One lipid phosphatase motif, a cell adhesion sequence, a putative proton donor sequence and a dityrosine basolateral targeting sequence are as shown. Both N-and C-terminals of LPP3 protein are located inside the cytoplasm.
Tumor endothelium expresses LPP3 protein
It is apparently clear that the tumor microenvironment contains bFGF and VEGF. Because bFGF and VEGF induce LPP3 expression in cultured ECs, we hypothesized that LPP3 protein might be similarly expressed by tumor-endothelium. To examine this hypothesis, serial angioma and hemangioma sections were subjected to immunostaining with anti-Flk-1, anti-PECAM-1 (CD31), and von Willebrand factor (vWF) to establish the presence of ECs and blood vessels as previously described [23]. We have previously shown that quiescent skin blood vessels are negative for anti-LPP3-RGD immunoreactivity [please see reference 23 for online supplemental data]. Hypoxic tissues as well as inflamed tissues are known to express VEGF [1,4]. Consistent with these reports, our data indicate that VEGF (green) is diffusedly distributed throughout angioma and hemangioma tissues, with significantly higher expression in the blood vessels (Fig. 3B,3E,3H). As shown in Fig. 3A,3D and 3G, LPP3 (red) protein appears to co-localize with VEGF in the angioma, including in the endothelium where the yellow ring like structure indicates colocalization; however, their molecular proximity remains unknown. Similarly, the expression of LPP3 protein was coincident with VEGF expression in the hemangioma section examined (Fig. 3G,3H,3I). These data demonstrate that both LPP3 and VEGF are diffusedly distributed within tumor vasculature, and LPP3 expression may not be exclusively restricted to ECs. Incubation of the antibodies with peptides that had been used to generate the primary antibody blocked immunoreactivity, confirming the specificity of antibodies used, as previously described [23]. This data is consistent with our earlier observation that the LPP3 protein is highly expressed in tumor endothelium [23].
Figure 3 Tumor endothelium express LPP3 protein alongside VEGF. Paraffin-embedded angioma (A-C) and hemangioma (D-I) tumor tissue sections (4 μm) were subjected to antigen retrieval, and sequentially incubated with the indicated antibodies. After washing with PBS, sections were incubated with donkey anti-goat/rabbit IgG conjugated to Texas-red (red) and goat anti-mouse IgG conjugated to FITC (green). C, F, and I images represent overlays of A, B; D, E; and G, H respectively. Images were captured below saturation level. Merged yellow represents co-expression. Data shown are representative of those obtained in at least three separate experiments, with similar results. (L, lumen; Magnification, 100×; Bars, 50 μM).
Inhibition of bFGF and VEGF induced capillaries by anti-LPP3-RGD antibodies
A considerable number of studies indicate that bFGF- and VEGF-mediated signaling, as well as cell adhesion events mediated by α5β1 and αvβ3 integrins, critically determine the outcome of angiogenesis. Because the LPP3-RGD protein acts as a cell-associated integrin ligand for α5β1 and αvβ3 integrins, we hypothesized that an anti-LPP3-RGD antibody could inhibit cell-cell interactions that may impede the ability of ECs to undergo capillary morphogenesis (also called tubulogenesis). To evaluate the capacity of anti-LPP3-RGD to block capillary morphogenesis of ECs, we employed early passage (between 3–4 total) ECs. Typically, we embed ECs between two layers of type I collagen matrices, and treat them with bFGF and VEGF in the presence of 20% serum. As previously described, the process of capillary formation by ECs in a three-dimensional type I collagen take place over a period of 24 to 72 h, and requires the addition of bFGF and VEGF [23,24].
For the purpose of this study, capillaries formed after 24 h of culture in 3D collagen were considered "pre-formed vessels", whereas capillaries formed after more than 24 h of culture were considered "new capillaries". We used affinity purified rabbit anti-IgG (control) polyclonal antibodies (pAbs) and mouse anti-MHC class II (W6/32) monoclonal Abs (mAbs), and mouse anti-β1 (4B4) and anti-αvβ3 (LM609) mAbs, as negative and positive controls, respectively.
In the absence of antibodies, we observed an increased number of capillaries formed from 36 to 72 h in culture (Fig. 4, filled black bar). The addition of an anti-MHC mAb (empty bar) and rabbit IgG (filled blue bar) resulted in minimal inhibition, and for the purpose of our study we consider these minimal inhibitions as baseline (Fig. 4). In contrast, the addition of the anti-β1 (filled red bar) and anti-αvβ3 (filled green) mAbs caused regression of the pre-formed capillaries (Fig. 4). Anti-αvβ3 mAbs reduced the number of pre-formed capillaries by more than 50%, suggesting that other cell surface proteins, such as fibronectin-binding integrin α5β1 and collagen/laminin binding integrins may also mediate capillary morphogenesis. Indeed, anti-β1 integrin subunit mAbs inhibited pre-formed interconnections and reduced the number of capillaries by ~60–70% (Fig. 4). No two mAbs were added simultaneously since it has been reported that α5β1 and αvβ3 integrin antibodies together induce complete collapse and regression of tubules in vitro [37,38]. Unexpectedly, the addition of the anti-LPP3-RGD antibody (filled yellow bar) had a dramatic effect on bFGF and VEGF-induced capillary formation (Fig. 4). The effect of anti-LPP3-RGD was comparable to anti-αvβ3 mAbs (Fig. 4, solid yellow bar). Representative cross sections of 3D gel are shown (Fig. 5). Arrows indicate lumen like structures. These data suggest that antibodies affecting cell adhesion, migration and cell-cell interaction events inhibit capillary morphogenesis by ECs in vitro. These data demonstrate that an affinity-purified rabbit anti-LPP3-RGD polyclonal antibody can inhibit bFGF- and VEGF-induced capillary formation in vitro.
Figure 4 Effect of specific antibodies on pre-formed capillaries. ECs were cultured in 3D collagen matrices in the presence of bFGF and VEGF165. Cultures were treated with the indicated antibodies at 24 h and processed at various time points (indicated). The 3D cultures were fixed, serial sections prepared and stained with eosin. The capillaries were counted as described in methods section. Values represent the mean ± SEM obtained from three independent experiments that used five wells in each case. * P < 0.02.
Figure 5 Inhibition of bFGF and VEGF induced capillary morphogenesis of ECs in 3D type I collagen matrix. The samples shown in the upper panels (A-E) were treated with anti-MHC class II mAbs, whereas those in the lower panels (F-J) with anti-αvβ3 integrin antibodies. Cross sections were stained with acidified eosin as described in methods. Magnification, 100×. Bar, 200 μM.
Discussion
We previously identified LPP3 in a functional assay of angiogenesis and reported that human LPP3 protein mediates cell-cell interactions [23,24]. Although the proposed cell adhesion sequences of human (RGD) and mouse (RGE) LPP3 are not identical, we observed that, in response to long-term cell adhesion, both sequences efficiently ligate α5β1 and αvβ3 integrins, and these adhesion events do not require protein synthesis [28]. Many previous studies have described a correlation between over-expression of LPP3 as an enzyme and down-regulation of cell signaling. However, the critical importance of Lpp3 gene function during developmental angiogenesis is underscored by the observation that mice embryos lacking Lpp3 die at day E7.5 due to a dearth of functional vasculature [35]. This report suggested that LPP3 protein regulates cellular interactions [35]. It is clear that further investigation will be necessary to elucidate the function of LPP3. Regardless of the mechanism, LPP3 is likely to be required during both adult and pathological angiogenesis. Therefore, we hypothesize that inhibition of LPP3 protein function blocks angiogenesis. Here, we asked the question whether an anti-LPP3-RGD antibody inhibits bFGF and VEGF induced capillary morphogenesis of ECs. Indeed, we found evidence that an anti-LPP3-RGD antibody can inhibit capillary morphogenesis of ECs.
It has become increasingly clear that bFGF- and VEGF-induced angiogenesis requires integrin-mediated adhesion events, a process by which ECs maintain cell-cell contact, survive, migrate, and proliferate [11-14,17]. ECs are known to express α2β1, α3β1, α5β1, α6β1, αvβ3 and αvβ5, integrins [13,17,18]. Several integrins have been suggested to play a role in each step of these processes; of these, α5β1 and αvβ3 integrins appear to be key players [11-14]. Previously, we provided evidence that a subset of integrins, including α1β1, α5β1 and αvβ3 are able to promote the EC cell cycle progression through the Shc pathway [17,18,21]. In vitro and in vivo assays have provided evidence that interference with the function of α5β1 and αvβ3 integrins block bFGF- and VEGF-induced angiogenesis, suggesting that α5β1- and αvβ3-mediated signaling networks cooperate by regulating a similar angiogenic signaling cascade [39-42]. Because signaling from α5β1 and αvβ3 integrins are critical for EC functioning, perhaps depriving LPP3-mediated cell-cell interactions interrupts several important signaling pathways, thereby inducing regression of capillary morphogenesis by ECs.
Our data indicate that pathophysiologically relevant agonists, e.g. bFGF and VEGF cytokines, can stimulate the expression of LPP3 protein in ECs. These data imply that LPP3 protein expression is likely to be associated with inflamed tissues and organs that require angiogenesis. We also demonstrate that LPP3 protein is up-regulated in tumor endothelium. We also show the ability of anti-LPP3-RGD to detect native LPP3 antigen on the cell surface of ECs. Considering the ability of Lpp3 to regulate embryonic vasculogenesis, and to ligate α5β1 and αvβ3 integrins, we propose that the function of LPP3 protein is pro-angiogenic. Accordingly, we show that specific inhibition of LPP3 with an anti-LPP3-RGD polyclonal antibody inhibit bFGF- and VEGF-induced capillary morphogenesis. Currently, a function blocking anti-LPP3 monoclonal antibody is not available. It is clear that further studies would be necessary to test the ability and efficiency of various anti-LPP3 antibodies and peptides to inhibit angiogenesis in vivo.
Methods
Cells and Reagents
Human umbilical vein endothelial cells (HUVECs) were purchased from Cambrex Bio Science Inc. (Walkersville, MD). Media 199, antibiotic solution, L-glutamine and all other cell culture reagents were purchased from InVitrogen Corp. (Carlsbad, CA). Recombinant human vascular endothelial growth factor (VEGF165), basic fibroblast growth factor (bFGF), and anti-VEGF (MAB293) were purchased from R&D Systems Inc. (Minneapolis, MN); adult human serum-AB from Gemini Bioproducts (Woodland, CA); bovine skin-derived type I collagen from Cohesion Technologies, Inc. (Palo Alto, CA). Affinity-purified anti-αvβ3 integrin (LM609), anti-PCNA and anti-VE-cadherin (MAB1989) monoclonal antibodies (mAbs) were purchased from Chemicon International Inc. (Temecula, CA) and anti-KDR/Flk-1 (sc-6251) mAb from Santa Cruz Biotechnology Inc. (Santa Cruz, CA). Mouse anti-Fak monoclonal antibody was purchased from Upstate Biotechnology Inc. (Lake Placid, NY); anti-human β1 integrin subunit (clone 4B4) mAbs from Beckman Coulter Inc. (Fullerton, CA); and anti-MHC class II (W6/32) from Sigma Chemical Com., (St. Louis, MO). The preparation of rabbit anti-LPP3-RGD (previously called anti-VCIP-RGD or anti-PAP2b-RGD) and anti-LPP3-c-cyto (previously called anti-VCIP-c-cyto or anti-PAP2b-c-cyto) polyclonal antibodies (pAbs) has been previously described [23,24].
De-glycosylation of LPP3 protein
Monolayer ECs (2 × 107) at passage 4 were stimulated with VEGF165 and subjected to cell surface biotinylation as previously described [23,24]. Biotinylated ECs were solubilized in RIPA cell extraction buffer [50 mM HEPES, pH 7.5, 150 mM NaCl, 1.0% Triton X-100 (non-ionic), 0.25% SDS (anionic), 0.25% sodium deoxycholate (anionic), and 2 mM EDTA, to which appropriate concentrations of proteases were added prior to use]. Extracts were clarified by centrifugation at 21,000 × g for 45 min at 4°C. Lysates were pre-absorbed at 4°C for 2 h by incubating with sepharose beads conjugated to rabbit IgGs. For each immunoprecipitation, approximately 1.5 mg total protein was used. Immunoprecipitation was carried out for 3 hr at 4°C. Immunocomplexes were washed five times with RIPA cell extraction buffer. For deglycosylation, immunoprecipitates were denatured in 1.0% SDS for 20 min at 90°C and washed twice with deglycosylation reaction buffer (New England Biolab., Beverly, MA). The deglycosylation (PNGaseF) reaction was initiated in a reaction volume of 100 μl containing 0.5% NP40 detergent and 50 units of PNGaseF enzyme at 37°C for 3 h. Samples were boiled in Lammeli reducing sample buffer and resolved by 10% SDS-PAGE and transferred to a nitrocellulose (NC) membrane. The NC membrane was blocked with 5% milk and 1% BSA in 1× TBS, 0.1% Tween and analyzed by incubating with streptavidin conjugated to horse-radish peroxidase (HRP) at a 1:10000 dilution. For FACS, monolayer ECs were starved for 6 h, thereafter, either left unstimulated or stimulated with VEGF (100 ng/ml) for 6 h. Cells were then non-enzymatically detached and subjected to FACS analyses as previously described [36].
Monolayer and three-dimensional (3D) cell culture
Monolayer EC culture was performed as previously described [17,18,23,24]. The preparation of 3D collagen matrix has also been previously described [23,24]. Briefly, a viscous gel-like solution was prepared by mixing 7 ml of 3.0 mg/ml type I collagen solution with 1 ml of 10× Ml99 medium at 4°C, adjusting the pH to 7.5 with 0.1 N sodium hydroxide, adding 0.1 ml of 100× ITS, and adjusting to a final volume of 10 ml with sterile distilled water. The matrix was allowed to polymerize (solidify) for 30 min at 37°C. Next, unstarved proliferating ECs were gently resuspended (at 6 × 105 cells/ml in complete media), seeded onto solidified gels, and the dishes (24 well Coster cell culture dishes) were returned to a CO2 incubator for 2–3 h. At the end of 3 h, unattached cells were removed by gentle aspiration. Onto monolayer cells, a second layer of collagen gel was added and returned to a 37°C humidified CO2 incubator. Following solidification (~3 h), the matrix was layered with Ml99 medium containing 20% adult human serum-AB, 4 mM L-glutamine, 1× ITS, bFGF (20 ng/ml) and 100 ng/ml VEGF165. The old growth medium was removed and fresh medium was added every 24 h. Capillary formation was examined under a phase contrast microscope every 12 h.
Inhibition Capillary Morphogenesis in 3D culture by anti-LPP3-RGD antibodies
ECs were embedded in 3D gels as described above, using media M199. At least 10 random 100× fields were examined to assess capillary formation (also called tubulogenesis). The formation of lumen-like structures was visible as early as 24 h. Antibodies were dialyzed in sterile dialysis buffer (25 mM Tris, 175 mM sodium chloride, 2 mM potassium chloride pH 7.4) overnight and passed through a 0.22 μM filter prior to use. The integrity of antibodies was determined by SDS-PAGE. To determine the effects of specific antibodies on pre-formed capillaries, the method of Bayless et al. was used [37,38]. Briefly, at the end of 24 h, mAbs (50 μg/ml) and pAbs (75 μg/ml) were added. Fresh aliquots of antibodies were added every 12 h for a total of 60 h. To quantify the degree of capillary formation, 3D matrices were fixed at 24, 36, 48, 60, and 72 h by aspirating the medium, washing with PBS, and then fixing with 4% glutaraldehyde in PBS, pH 7.4, overnight at 4°C. Matrices were then washed with distilled water and embedded in paraffin according to the manufacturer's instructions (Richard Allen Scientific, Kalamazoo, MI). Serial sections (4 μm) were prepared, dehydrated, stained with acidified eosin, and destained with distilled water. Capillaries were counted and photographed using a Zeiss Axiovert 25C microscope at 100× magnification. Each capillary tubule was surrounded by least 2 to 5 ECs. Capillary formation was defined as the induction of a minimum of 3 separate capillary events within a single field. At least 10 random fields were counted for each sample. Experiments were performed in duplicate, using triplicate wells in each case. Results were expressed as mean ± SEM.
Statistics
Student's t tests and ANOVA were used to detect significant comparisons as previously described [23].
Abbreviations
3D, three dimensional; ECs, endothelial cells; bFGF, basic fibroblast growth factor; hr, hour; kDa, kiloDalton; LPP1, lipid phosphate phosphohydrolase-1; LPP3, lipid phosphate phosphohydrolase-3; mAb, monoclonal antibodies; mg, milligram; μg, microgram; pAb, polyclonal antibodies; SDS-PAGE, sodium dodecyl sulphate - Polyacrylamide gel electrophoresis; VEGF, vascular endothelial growth factor;
Authors' contributions
JOH was responsible for maintaining the monolayer ECs, FACS and protein expression analyses. KKW was responsible for 3D gel preparation, capillary assays, sectioning, staining, data analysis, interpretation, and preparation of manuscript. KKW and JOH were involved in study design, and read and approved the manuscript.
Acknowledgements
This study was supported by an award from the American Heart Association (AHA) to KKW. KKW is a member of Mission Connect (TIRR) and Cardiovascular Research Institute (CVRI) of Texas A & M University. The authors thank Shu Feng for immunostaining of tumor sections.
==== Refs
Folkman J Angiogenesis in cancer, vascular, rheumatoid and other disease Nat Med 1995 1 27 31 7584949 10.1038/nm0195-27
Hanahan D Folkman J Patterns and emerging mechanisms of the angiogenic switch during tumorigenesis Cell 1996 86 353 364 8756718 10.1016/S0092-8674(00)80108-7
Ferrara N Davis-Smyth T The biology of vascular endothelial growth factor Endocr Rev 1997 8 4 25 9034784 10.1210/er.18.1.4
Carmeliet P Jain RK Angiogenesis in cancer and other diseases Nature 2000 407 249 257 11001068 10.1038/35025220
Sato Y Kanno S Oda N Abe M Ito M Shitara K Shibuya M Properties of two VEGF receptors, Flt-1 and KDR, in signal transduction Ann N Y Acad Sci 2000 902 201 205 10865839
Ferrara N Role of vascular endothelial growth factor in regulation of physiological angiogenesis Am J Physiol Cell Physiol 2001 280 C1358 1366 11350730
Werb Z ECM and cell surface proteolysis: regulating cellular ecology Cell 1997 91 439 442 9390552 10.1016/S0092-8674(00)80429-8
Shapiro SD Matrix metalloproteinase degradation of extracellular matrix: biological consequences Curr Opin Cell Biol 1998 10 602 608 9818170 10.1016/S0955-0674(98)80035-5
Wary KK Molecular targets for anti-angiogenic therapy Curr Opin Mol Ther 2004 6 54 70 15011782
Varner JA Cheresh DA Tumor angiogenesis and the role of vascular cell integrin alphavbeta3 Important Adv Oncol 1996 69 87 8791129
Varner JA The role of vascular cell integrins alpha v beta 3 and alpha v beta 5 in angiogenesis EXS 1997 79 361 390 9002227
Eliceiri BP Cheresh DA The role of alpha v integrins during angiogenesis Mol Med 1998 4 741 750 9990860
Ruoslahti E Engvall E Integrins and vascular extracellular matrix assembly J Clin Invest 1997 100 S53 56 9413402
Elicieri BP Cheresh DA Emile E Voest, Patricia A Damore Alphav beta3 integrin and its antagonists in the control of angiogenesis Tumor angiogenesis and Microcirculation 2001 Mercel Dekker, Inc, New York City, NY 387 398
Hynes RO Integrins: bidirectional, allosteric signaling machines Cell 2000 110 673 687 12297042 10.1016/S0092-8674(02)00971-6
Hynes RO Integrins: versatility, modulation, and signaling in cell adhesion Cell 1992 69 11 25 1555235 10.1016/0092-8674(92)90115-S
Wary KK Mainiero F Isakoff SJ Marcantonio EE Giancotti FG The adaptor protein Shc couples a class of integrins to the control of cell cycle progression Cell 1996 87 733 743 8929541 10.1016/S0092-8674(00)81392-6
Wary KK Mariotti A Zurzolo C Giancotti FG A requirement for caveolin-1 and associated kinase Fyn in integrin signaling and anchorage-dependent cell growth Cell 1998 94 625 634 9741627 10.1016/S0092-8674(00)81604-9
Guan JL Focal adhesion kinase in integrin signaling Matrix Biol 1997 16 195 200 9402009 10.1016/S0945-053X(97)90008-1
Assoian RK Marcantonio EE The extracellular matrix as a cell cycle control element in atherosclerosis and restenosis J Clin Invest 1996 98 2436 2439 8958203
Giancotti FG Ruoslahti E Integrin signaling Science 1999 285 1028 1032 10446041 10.1126/science.285.5430.1028
Stupack DG Cheresh DA Integrins and angiogenesis Curr Top Dev Biol 2004 64 207 238 15563949
Humtsoe JO Feng S Thakker GD Yang J Hong J Wary KK Regulation of cell-cell interactions by phosphatidic acid phosphatase 2b/VCIP EMBO J 2003 22 1539 1554 12660161 10.1093/emboj/cdg165
Wary KK Thakker GD Humtsoe JO Yang J Analysis of VEGF-responsive Genes Involved in the activation of endothelial cells Mol Cancer 2003 2 1 12 12537587 10.1186/1476-4598-2-1
Brindley DN Lipid phosphate phosphatases and related proteins: signaling functions in development, cell division, and cancer J Cell Biochem 2004 92 900 912 15258914 10.1002/jcb.20126
Saba JD Lysophospholipids in development: Miles apart and edging in J Cell Biochem 2004 92 967 992 15258919 10.1002/jcb.20128
Siess W Tigyi G Thrombogenic and atherogenic activities of lysophosphatidic acid J Cell Biochem 2004 92 1086 1094 15258894 10.1002/jcb.20108
Humtsoe JO Bowling RA Feng S Wary KK Murine Lipid phosphate phosphohydrolase acts as a cell associated integrin ligand Biochem Biophys Res Commun
Kai M Wada I Imai S Sakane F Kanoh H Identification and cDNA cloning of 35-kDa phosphatidic acid phosphatase (type 2) bound to plasma membranes J Biol Chem 1996 271 18931 18938 8702556 10.1074/jbc.271.31.18931
Roberts R Sciorra VA Morris AJ Human type 2 phosphatidic acid phosphohydrolases: substrate specificity of the type 2a, 2b and 2c enzymes and cell surface activity of the 2a isoform J Biol Chem 1998 273 22059 22067 9705349 10.1074/jbc.273.34.22059
Sciorra VA Morris AJ Sequential actions of phospholipase D and phosphatidic acid phosphohydrolase 2b generate diglyceride in mammalian cells Mol Biol Cell 1999 10 3863 3876 10564277
Ishikawa T Kai M Wada I Kanoh H Cell surface activities of the human type 2b phosphatidic acid phosphatase J Biochem (Tokyo) 2000 127 645 651 10739957
Jia YJ Kai M Wada I Sakane F Kanoh H Differential localization of lipid phosphate phosphatases 1 and 3 to cell surface subdomains in polarized MDCK cells FEBS Lett 2003 552 240 246 14527693 10.1016/S0014-5793(03)00931-1
Zhang N Sundberg JP Gridley T Mice mutant for Ppap2c, a homolog of the germ cell migration regulator wunen, are viable and fertile Genesis 2000 27 137 140 10992322 10.1002/1526-968X(200008)27:4<137::AID-GENE10>3.0.CO;2-4
Escalante-Alcalde D Hernandez L Le Stunff H Maeda R Lee HS Cheng Gang JrSciorra VA Daar I Spiegel S Morris AJ Stewart CL The lipid phosphatase LPP3 regulates extra-embryonic vasculogenesis and axis patterning Development 2003 130 4623 4637 12925589 10.1242/dev.00635
Humtsoe JO Kim JK Xu Y Keene DR Höök M Lukomski S Wary KK A streptococcal collagen-like protein interacts with the alpha2betal integrin and induces intracellular signaling J Biol Chem 2005 280 13848 13857 15647274 10.1074/jbc.M410605200
Bayless KJ Salazar R Davis GE RGD-dependent vacuolation and lumen formation observed during endothelial cell morphogenesis in three-dimensional fibrin matrices involves the alpha(v)beta(3) and alpha(5)beta(l) integrins Am J Pathol 2000 156 1673 1683 10793078
Bell SE Mavila A Salazar R Bayless KJ Kanagala S Maxwell SA Davis GE Differential gene expression during capillary morphogenesis in 3D collagen matrices: regulated expression of genes involved in basement membrane matrix assembly, cell cycle progression, cellular differentiation and G-protein signaling J Cell Sci 2001 114 2755 7273 11683410
Brooks PC Clark RA Cheresh DA Requirement of vascular integrin alpha v beta 3 for angiogenesis Science 1994 264 569 571 7512751
Brooks PC Montgomery AM Rosenfeld M Reisfeld RA Hu T Klier G Cheresh DA Integrin alpha v beta 3 antagonists promote tumor regression by inducing apoptosis of angiogenic blood vessels Cell 1994 79 1157 1164 7528107 10.1016/0092-8674(94)90007-8
Friedlander M Brooks PC Shaffer RW Kincaid CM Varner JA Cheresh DA Definition of two angiogenic pathways by distinct alpha v integrins Science 1995 270 1500 1502 7491498
Matter ML Ruoslahti E A signaling pathway from the alpha5beta1 and alpha(v)beta3 integrins that elevates bcl-2 transcription J Biol Chem 2001 276 27757 27763 11333270 10.1074/jbc.M102014200
|
16076388
|
PMC1201157
|
CC BY
|
2021-01-04 16:39:15
|
no
|
Cell Commun Signal. 2005 Aug 2; 3:9
|
utf-8
|
Cell Commun Signal
| 2,005 |
10.1186/1478-811X-3-9
|
oa_comm
|
==== Front
Environ HealthEnvironmental Health1476-069XBioMed Central London 1476-069X-4-151609113510.1186/1476-069X-4-15ResearchMenstrual function among women exposed to polybrominated biphenyls: A follow-up prevalence study Davis Stephanie I [email protected] Heidi Michels [email protected] Vicki S [email protected] Paige E [email protected] Carol [email protected] Lorraine L [email protected] Alden K [email protected] Michele [email protected] Department of Epidemiology, Rollins School of Public Health, Emory University, 1518 Clifton Road, Alanta, GA, 30322, USA2 Graduate Division of Biological and Biomedical Sciences, Emory University, 1462 Clifton Road, NE, Atlanta, Georgia, 30322, USA3 Department of Biostatistics, Rollins School of Public Health, Emory University, 1518 Clifton Road, NE, Atlanta, Georgia, 30322, USA4 Department of Environmental and Occupational Health, Rollins School of Public Health, Emory University, 1518 Clifton Road, NE, Atlanta, Georgia, 30322, USA5 Division of Environmental Hazards and Health Effects, National Center for Environmental Health, Centers for Disease Control and Prevention, 4770 Buford Highway, MS F-46, Atlanta, Georgia, 30341, USA6 Division of Environmental and Occupational Epidemiology, Michigan Department of Community Health, P.O. Box 30195, Lansing, Michigan, 48909, USA2005 9 8 2005 4 15 15 29 3 2005 9 8 2005 Copyright © 2005 Davis et al; licensee BioMed Central Ltd.2005Davis 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
Alteration in menstrual cycle function is suggested among rhesus monkeys and humans exposed to polybrominated biphenyls (PBBs) and structurally similar polychlorinated biphenyls (PCBs). The feedback system for menstrual cycle function potentially allows multiple pathways for disruption directly through the hypothalamic-pituitary-ovarian axis and indirectly through alternative neuroendocrine axes.
Methods
The Michigan Female Health Study was conducted during 1997–1998 among women in a cohort exposed to PBBs in 1973. This study included 337 women with self-reported menstrual cycles of 20–35 days (age range: 24–56 years). Current PBB levels were estimated by exponential decay modeling of serum PBB levels collected from 1976–1987 during enrollment in the Michigan PBB cohort. Linear regression models for menstrual cycle length and the logarithm of bleed length used estimated current PBB exposure or enrollment PBB exposure categorized in tertiles, and for the upper decile. All models were adjusted for serum PCB levels, age, body mass index, history of at least 10% weight loss in the past year, physical activity, smoking, education, and household income.
Results
Higher levels of physical activity were associated with shorter bleed length, and increasing age was associated with shorter cycle length. Although no overall association was found between PBB exposure and menstrual cycle characteristics, a significant interaction between PBB exposures with past year weight loss was found. Longer bleed length and shorter cycle length were associated with higher PBB exposure among women with past year weight loss.
Conclusion
This study suggests that PBB exposure may impact ovarian function as indicated by menstrual cycle length and bleed length. However, these associations were found among the small number of women with recent weight loss suggesting either a chance finding or that mobilization of PBBs from lipid stores may be important. These results should be replicated with larger numbers of women exposed to similar lipophilic compounds.
==== Body
Background
In 1973, the fire retardant chemical, FireMaster®, was mistaken for NutriMaster®, a magnesium oxide-based cattle feed supplement, and was inadvertently introduced into cattle feed (Michigan Chemical Corporation, St. Louis, MI). Thousands of farm families and farm product consumers were exposed to this commercial mixture of polybrominated biphenyl (PBB) congeners throughout Michigan. Concern about possible adverse health effects led to the establishment of the Michigan PBB Long-Term Study in 1976 by the Michigan Department of Community Health (MDCH). The incident is described in detail elsewhere [1,2].
From 1997–1998, the Michigan Female Health Study (MFHS) was conducted to assess whether PBB exposure disrupted endocrine function among cohort women. Only one rhesus monkey study was found, suggesting that PBBs may be associated with longer menstrual cycles [3]. Animal studies of structurally similar polychlorinated biphenyls (PCBs) have found that menstrual cycle length was either not different [4-7] or longer [8] with Aroclor 1248 [4,8] or Aroclor 1254 [5-7] dosing. Bleed length was marginally longer in the higher Aroclor 1254 dose groups [5-7]. Human PCB studies are also inconsistent. Shorter cycle length was associated with indices of PCB-contaminated fish consumption [9]. When summed over measures of selected PCB congeners, cycle length was longer with higher total PCB levels in a large multicenter cohort [10], but not among Southeast Asian immigrants [11]. Cooper et al reported that bleed length was not associated with PCBs [10], but Yucheng women, with cooking oil exposure to PCBs, reported abnormal bleeding more often than controls [12].
Endocrine regulation is a complex process and may be disrupted at many points. The immune and neuroendocrine systems are integrated in a network of the ovarian, thyroid, thymus, and adrenal axes. This network is regulated or disrupted through feedforward and feedback loops of the hypothalamus, the pituitary, and associated end-organs of the neuroendocrine axes [13-15].
Cycle length and bleed length, although nonspecific, are markers for the reproductive status of women undergoing cyclical ovulation and endometrial angiogenesis under hormonal feedback and control [16-20]. This study assessed whether serum PBB levels were associated with these markers of menstrual cycle function among women still menstruating in the year before interview.
Methods
The Michigan Female Health Study design and data collection
The release of PBBs into the Michigan food chain was limited in geographic distribution and in duration due to the environmental remediation of quarantined farms and the food chain. Michigan PBB cohort enrollment data, along with blood samples for exposure biomonitoring that included PBB and PCB, were collected from 1976–1987 by the MDCH.
The study design of the MFHS is a follow-up prevalence study [21]. It is a hybrid study of an exposure cohort followed prospectively to assess the prevalence of current menstrual cycle outcomes. Data were collected for the MFHS during August 1997 through April 1998 by computer-assisted telephone interview. Participants were asked about reproductive history, menstrual cycle characteristics, physical activity, smoking habits, medication, physician-diagnosed medical conditions, and other health and demographic information. The MFHS was approved by the institutional review boards of Emory University, MDCH, and the Centers for Disease Control and Prevention.
Eligibility for menstrual cycle analyses
A total of 1020 women in the MFHS had serum levels of PBB available from cohort enrollment. This study was restricted to women who were born before the exposure incident, which is estimated to have occurred as early as July 1, 1973 [22]. In this manner, we restricted the analyses to women solely exposed by food ingestion and not in utero.
Women were asked to report their usual non-pregnancy related cycle length during the past year, defined as the number of days from the first day of menstrual bleed to the first day of the next menstrual bleed. The distribution of menstrual cycle length across the reproductive lifespan represents a mixture between normal ovulatory cycles and short or long anovulatory bleeding episodes at both age extremes [23-25]. Long cycles in the right tail are influenced by older women in perimenopausal status [24,25] with more frequent ovulatory failure [23]. Among women aged 25–44 years, Metcalf observed that 93.6% of all cycles were between 20–35 days in length with a pattern of consistently ovulatory cycles [23]. We lacked laboratory confirmation and referred to the women in this study as having normal, not ovulatory, cycles. We decomposed the distribution of cycle length into an approximately symmetric part centered about 28 days (n = 338) and a long right tail (n = 12). We restricted our analysis to women with cycle lengths ranging from 20–35 days (n = 337), excluding a subject with a 14-day cycle length (Figure 1). Other studies define normal cycles from 21–35 days [26], in general agreement with our definition, while others recommend cycle lengths of 18–40 days [25].
Figure 1 Eligibility for menstrual cycle analysis, Michigan Female Health Study, 1997–1998.
We excluded women from menstrual cycle analyses if they indicated any of the following during the past year: no menstrual cycles, menstrual cycles shorter than 20 days or longer than 35 days, breastfeeding, or use of hormone medications. We also excluded ten women with missing data (Figure 1).
We restricted bleed length analyses to the same women who were eligible for cycle length analyses. Women reported their usual length of menstrual bleeding in the past year as a single estimate or a range in days. Bleed length for women reporting a range was taken as the midpoint between the two values.
Measurement of enrollment serum PBB and PCB concentrations
PCBs and PBBs have 209 theoretically possible congeners. By weight, FireMaster® was 54%–68% PBB 153, a non-coplanar congener. Several coplanar congeners were present in the mixture in smaller amounts [22,27]. Serum concentrations of PBB 153, in parts per billion (ppb), were quantified by gas chromatography with electron capture detection using established protocols. The limit of detection (LOD) was 1 ppb. These PBB 153 measures are indicators of total PBB exposure [28,29].
Up to 1982, total PCBs were quantified as Aroclor 1254 (Montsano Company, St. Louis, MO) levels with the LOD at 5 ppb. Between 1982–1993, Aroclor 1260 (Montsano Company, St. Louis, MO) levels were assessed with the LOD at 3 ppb. We combined initial total PCB levels from the two assay methods, assuming any total below 5 ppb was below the LOD. Only three women (0.9%) had initial Aroclor 1260 levels, all below 3 ppb. Changing PCB analytical methods did not result in misclassification of exposure rank in this sample.
Estimation of current serum PBB concentration
We hypothesized that current PBB levels were associated with current menstrual function. We used serum PBB enrollment measurements to estimate individual current PBB levels, using an exponential decay model described previously [30]. The original decay model included women who were at least 16 years old at enrollment with initial PBB levels of 2 ppb or higher. They also had at least two nonpregnancy serum samples drawn. The authors found that overall, median half-life was 13.5 years, but decay was slower among women with high enrollment BMI. Parity, age at enrollment, smoking history, and breastfeeding duration were not significant predictors of decay [30]. The estimated levels from the decay model were highly correlated with measured levels (r = 0.92) [31].
We used the same equation specified in the decay model for our estimates. We assigned enrollment PBB levels less than or equal to 1 ppb a value of 0.5 ppb, a value half way between zero and the LOD. This enabled log-transformations [32], and has been shown to be an appropriate imputation technique [33]. We included data from the MDCH archives: linear and quadratic terms for the logarithm of enrollment PBB level, categories of body mass index (BMI) at enrollment (less than/greater than or equal to 23.0 kg/m2), and age at enrollment. We also included data from the MFHS: number of full-term pregnancies between enrollment and interview, breastfeeding duration (months), and smoking status (ever/never). Because these data were modeled, we did not impose an LOD on the estimated levels.
Classification of estimated current PBB exposure
PBB and PCB exposures were ranked in low, middle, and high exposure categories. Tertiles were our primary classification for estimated current PBBs: with low, from zero to 0.06 ppb (n = 112) as referent; middle, from greater than 0.06–0.32 ppb (n = 113); and high, at greater than 0.32 ppb (n = 112). We also performed analyses with estimated current PBBs categorized for highest exposure, retaining the lowest tertile as the referent and classifying the upper decile at greater than 3.08 ppb (n = 35).
Classification of PBB exposure at cohort enrollment
We categorized laboratory measures of PBB at enrollment into approximate tertiles: at or below the LOD of 1 ppb (n = 111) as referent; from greater than 1 to 3 ppb (n = 120); and greater than 3 ppb (n = 106). We classified enrollment PBB levels greater than 12 ppb (n = 37) as the upper decile, retaining the same referent group.
Classification of PCB exposure
We used initial serum PCB measurements. We did not attempt to model the decay and bioaccumulation of PCBs due to changes in the laboratory assessment from Aroclor 1254 to Aroclor 1260. We also did not have information on important factors of PCB bioaccumulation, such as fish consumption. We divided PCB exposure into four categories. The lowest referent category included measurements at or below the LOD or 5 ppb (n = 178); the middle category, from greater than 5 to 7 ppb (n = 53); the high category, at greater than 7 ppb (n = 62); and missing (n = 44).
Statistical analyses
We performed multiple linear regressions with cycle length and with the natural logarithm of bleed length as continuous outcomes. We examined the distributions of the continuous variables to better meet the assumptions of normality in linear regression models. We did not apply data transformations for age and BMI at interview, or for cycle length restricted to 20–35 days (median 28.0; mean 27.7 days; standard deviation (SD) 2.9; skewness -0.42; kurtosis 0.70) (Figure 2). Bleed length was right-skewed (median 5.0; mean 5.3; SD 1.3; skewness 0.87; kurtosis 2.5) (Figure 3). We log-transformed bleed length to center the distribution (median 1.61; mean 1.63; SD 0.25; skewness -0.37; kurtosis 1.91). We verified that the quantile-quantile (QQ) plot for log-transformed bleed length linearized the curved point pattern we observed in the QQ plot for bleed length. We performed collinearity diagnostics, and centered age and BMI about their respective means (37.8 years and 26.6 kg/m2) to reduce inflated variances in the models.
Figure 2 Distribution of menstrual cycle length, Michigan Female Health Study, 1997–1998
Figure 3 Distribution of bleed length, Michigan Female Health Study, 1997–1998
We retained the same covariates in all regression models across both cycle length and bleed length analyses. Categorical covariates included PCBs, usual weekly physical activity, smoking status, educational attainment, income level, and loss of 10% of body weight or more in the past year unrelated to pregnancy. We defined never exercising as low, one or two times per week as moderate, and exercising three or more times per week as high levels of recreational physical activity.
We evaluated interactions of PBB categories with age and BMI at interview, weight loss, and physical activity for all models. Interaction assessment for age or BMI with PBBs included testing linear and quadratic terms. When significant interactions between age and PBB categories were detected, diagnostic tests were repeated with age quintiles. The age range for the first quintile was 24–31 years (n = 59); the second, 32–35 years (n = 70); the third, 36–38 years (n = 64) as the referent group; the fourth, 39–43 years (n = 74); and the fifth, 44–56 years (n = 70). We also evaluated possible additive or synergistic effects between exposures to PBBs and PCBs or cigarette smoking by interaction assessment. We exponentiated point estimates and confidence intervals for the difference in the logarithm of bleed length and reported the converted differences in bleed length as a ratio on the original scale, in days.
The decay model we used to estimate current PBB levels was originally developed using different criteria than the inclusion criteria for this study. We performed sensitivity analyses by repeating our models excluding females less than 16 years at enrollment (n = 140).
We performed secondary analyses for the full distribution of menstrual cycle length, including women in the long right tail. Logistic regression models would not converge due to small cell sizes with long cycles classified as greater than 35 days (n = 12; upper 3%). We redefined long cycles as greater than 30 days (n = 43; upper 10%) and examined the relationship between current estimated PBB exposure and the prevalence of long cycles.
We retained women with self-reported thyroid conditions (n = 39, 11.6%) in analyses to avoid adjusting for possible PBB-related thyroid-mediated menstrual cycle effects [34-41]. We did not include in utero diethylstilbestrol exposure as a covariate or exclusion due to the high number of women reporting unknown exposure status (n = 101, 30.0%). Five women (1.5%) reported having in utero diethylstilbestrol exposure. We performed sensitivity analysis removing these five women from the final models. Most women lacked historical pesticide information (n = 298; 88.4 percent); therefore, it was not included in this analysis.
We performed Pearson partial correlation analyses between cycle length and the logarithm of bleed length, assessing overall associations and correlations stratified by history of weight loss. We also performed correlation analyses between the logarithm of both PBB and PCB, and age in years. All analyses were performed using SAS Version 9.1 (SAS Institute Inc., Cary, NC).
Results
Descriptive statistics of the study population
After restricting our sample to women with cycle lengths between 20–35 days, average cycle and bleed lengths were 27.7 days and 5.3 days, respectively. The women, all white, ranged in age from 24 to 56 years. Most never smoked and reported some form of weekly exercise. Educational attainment was evenly distributed across PBB tertiles. Forty-four women (13.0%) reported losing 10% or more of their body weight in the past year (Table 1).
Table 1 Characteristics of 337 women,* Michigan Female Health Study, 1997–1998.
Population Characteristics Distribution
Continuous Variables Point Estimate (Range)
Median PBB level (ppb) measured at cohort enrollment 2.0 (0.5–1490)
Median PBB level (ppb) estimated at study interview 0.1 (0–1005)
Median initial PCB level (ppb) measured in cohort follow-up 5.0 (1.5–78) †
Mean age (years) at study interview 37.8 (24–56)
Mean BMI (kg/m2) at study interview 26.6 (16.9–53.1)
Categorical Variables N (%)
Past Year History of Weight Loss = 10% No 293(86.9%)
Yes 44 (13.0%)
Amount of weekly exercise Low (Never) 48 (14.2%)
Moderate (1 or 2 times) 137 (40.6%)
High (3 or more times) 152 (45.1%)
Lifetime smoking status Never Smoker 222 (65.9%)
Ever Smoker 115 (34.1%)
Educational attainment High School Grad or Less 132 (39.2%)
Some College or Technical 110 (32.6%)
College or Postgrad 95 (28.2%)
Annual Household Income Less Than $35,000 113 (33.5%)
At Least $35,000 207 (61.4%)
Unknown 17 (5.0%)
* Cycle lengths ranging from 20–35 days. † 44 missing PCB levels.
In our sample, 318 women (94.4%) had enrolled in the cohort by the end of 1977. The remaining 19 (5.6%) enrolled over the next 9 years. Using the exponential decay model [30], we extrapolated enrollment levels of PBB over an average of 20.5 years (median: 20.6; 25th %: 20.3; 75th %: 20.9; range: 10.3–22.4; in years) to estimate PBB levels at the time of the MFHS interview. Most women categorized as having low, middle, or high PBB exposure at enrollment were classified the same way by estimated current PBB tertiles (75.7%, 57.5%, and 79.2%, respectively in Table 2). If laboratory LODs were applied to the decay estimates, 277 women (82.2%) would currently have PBB levels below 1 ppb (median: 0.13; mean: 7.73; 25th %: 0.03; 75th %: 0.51; range: 0–1005; in ppb). A total of 86 women were common to our study (25.5% of 337) and the decay study (22.6% of 380).
Table 2 Cross tabulation of PBB tertiles from two timeframes, Michigan Female Health Study, 1997–1998.
No. in Tertiles of PBB Exposure Measured at Cohort Enrollment, (Range, ppb) No. in Tertiles of PBB Exposure Estimated at Time of Study Interview (Range, ppb) Total No.
Low (0.00–0.06) Middle (>0.06–0.32) High (>0.32)
Low (≤ 1.0) 84 27 0 111
Middle (1.0–3.0) 23 69 28 120
High (>3.0) 5 17 84 106
Total No. 112 113 112 337
We observed digit preferences in reporting menstrual cycle length for 30 days and 7-day multiples, noting peaks at 21 (n = 18), 28 (n = 112), and 30 (n = 64) days (Figure 2). The distribution of bleed length did not exhibit digit preferences. Because we allowed women to report their usual bleed length either directly or in a range, 84 women (24.9%) had half-day estimates (Figure 3).
Crude associations between PBB, PCB, and population characteristics
Average cycle length did not differ among women when stratified by PBB exposure at enrollment or by PBB exposure estimated at the time of the interview. There was a suggested increase in bleed length with increasing PBB tertiles at enrollment and time of interview (Table 3).
Table 3 Menstrual cycle outcomes by PBB tertiles from two time frames, Michigan Female Health Study, 1997–1998.
Outcomes Tertiles of PBB Estimated at Study Interview (No.) P * (df) Tertiles of PBB Measured at Cohort Enrollment (No.) P * (df)
Low (112) Middle(113) High (112) Low (111) Middle (120) High (106)
Mean Cycle Length (SD), days 27.4 (3.1) 28.0 (2.9) 27.7 (2.8) 0.33 (2) 27.5 (3.0) 27.6 (2.8) 28.0 (3.0) 0.34 (2)
Mean Bleed Length (SD), days† 5.0 (1.3) 5.1 (1.3) 5.2 (1.3) 0.36 (2) 4.9 (1.3) 5.1 (1.3) 5.3 (1.3) 0.14 (2)
* ANOVA F test for continuous variables. † Exponentiated from log(bleed length).
Age and BMI were positively correlated (r = 0.13, p = 0.02). The data suggested that women in the lowest tertile of enrollment PBBs were older and had higher BMIs; conversely, women in the highest tertile of estimated current PBBs were older and had higher BMIs (Table 4). Previously, Blanck et al also noted slower PBB decay among women with higher BMIs [30]. We found little difference in the frequency of weight loss in the past year and in the usual levels of weekly physical activity by PBB tertiles (Table 4).
Table 4 Menstrual cycle covariates by PBB tertiles from two time frames, Michigan Female Health Study, 1997–1998.
Covariates Tertiles of PBB Estimated at Study Interview (No.) P * (df) Tertiles of PBB Measured at Cohort Enrollment (No.) P * (df)
Low (112) Middle (113) High (112) Low (111) Middle (120) High (106)
Continuous Variables
Mean Age (SD), years 37.7 (6.9) 36.8 (6.1) 38.8 (6.9) 0.07 (2) 38.9 (6.4) 37.1 (6.4) 37.3 (7.2) 0.07 (2)
Mean BMI (SD), kg/m2 25.8 (5.4) 26.9 (6.5) 27.0 (6.8) 0.28 (2) 27.5 (6.4) 26.3 (6.4) 25.8 (5.9) 0.12 (2)
Amount of Weekly Exercise
Low No. (%) 14 (12.5) 17 (15.0) 17 (15.2) 0.27 (4) 13 (11.7) 15 (12.5) 20 (18.9) 0.27 (4)
Moderate No. (%) 38 (33.9) 48 (42.5) 51 (45.5) 41 (36.9) 50 (41.7) 46 (43.4)
High No. (%) 60 (53.6) 48 (42.5) 44 (39.3) 57 (51.4) 55 (45.8) 40 (37.7)
Past Year History of Weight Loss = 10%
Without No. (%) 95 (84.8) 97 (85.8) 101 (90.2) 0.45 (2) 97 (87.4) 103 (85.8) 93 (87.7) 0.90 (2)
With No. (%) 17 (15.2) 16 (14.2) 11 (9.8) 14 (12.6) 17 (14.2) 13 (12.3)
* ANOVA F test for continuous variables; Chi square test for categorical variables.
The frequency of diagnosed thyroid conditions did not differ across either PBB tertiles estimated at interview or measured at enrollment (p = 0.76 and 0.46, respectively; 2 df). Mean menstrual cycle length and bleed length did not differ when stratified by history of thyroid disorders (p = 0.09 and 0.58, respectively; 1 df). Enrollment PBBs and enrollment PCBs were positively correlated among the 293 women with both measures available (r = 0.12, p = 0.04); however, estimated current PBBs and enrollment PCBs, were not (r = 0.06, p = 0.27). Age was positively correlated with PCBs (r = 0.18, p = 0.002), but not PBBs (enrollment: r = -0.09, p = 0.10; adjusted current: r = 0.02, p = 0.66).
Multiple regression models for cycle length and bleed length
No grossly influential observations were noted in model diagnostics. The adjusted R2 values for linear regression models were 5% for cycle length and 7% for the logarithm of bleed length.
We observed an association between physical activity and bleed length, but not cycle length, regardless of PBB categories used in the models. Women with high levels of physical activity had bleed lengths 0.92 times shorter than those with moderate levels (95% confidence limits: 0.87, 0.97).
We found no overall association between current estimated PBBs and either menstrual cycle length or bleed length; however, the associations between PBB exposure and menstrual cycle length or bleed length differed for women based on their history of weight loss in the past year. The interaction terms for past year weight loss and PBB tertiles were not statistically significant for cycle length (Table 5). When we considered women with weight loss in the highest decile of estimated current PBB exposure, the interaction term was significant. This small group of women (n = 4) had cycle lengths 3.55 days shorter (95% confidence limits: -6.45, -0.65) than the referent group. We observed a similar association using upper decile enrollment PBB levels with even fewer women with weight loss (n = 3). Their mean cycle lengths were 5.54 days shorter (95% confidence limits: -8.83, -2.26) than their respective referent group. These three were among the four women with weight loss in the upper decile of estimated current PBBs.
Table 5 Cycle length by weight loss and PBB tertiles, Michigan Female Health Study, 1997–1998. Adjusted for PCB, BMI and age at interview, physical activity, smoking history, education, income.
Difference in Cycle Length, (days)
Without Weight Loss With Weight Loss
No. Diff (95% CI) No. Diff (95% CI)
Tertiles of PBB Estimated at Study Interview §
Low 95 0.00 Ref 17 -0.28 (-1.79 to 1.22)
Middle 97 0.35 (-0.48 to 1.18) 16 0.21 (-1.34 to 1.76)
High 101 0.30 (-0.53 to 1.13) 11 -1.04 (-2.86 to 0.77)
Test for Interaction: p = 0.57, 2 df
Tertiles of PBB Measured at Cohort Enrollment
Low 97 0.00 Ref 14 0.35 (-1.27 to 1.98)
Middle 103 0.01 (-0.80 to 0.82) 17 -1.02 (-2.53 to 0.48)
High 93 0.54 (-0.30 to 1.39) 13 -0.22 (-1.88 to 1.46)
Test for Interaction: p = 0.45, 2 df
* Cycle lengths ranging from 20–35 days.
Women with weight loss also exhibited a monotonic increase in bleed length with increasing PBB exposure relative to referent women in the low PBB tertile without weight loss. Among women with weight loss in estimated current PBB tertiles, those in the low tertile had shorter bleed lengths than the referent group with a ratio of 0.82; those in the middle tertile had a bleed length ratio of 0.92; and those in the high tertile had longer bleed lengths with a ratio of 1.27 (test for interaction: p < 0.0001) (Table 6). We observed consistent results in the model examining the upper decile (test for interaction: p = 0.002, data not shown). We also noted PBB-weight loss interactions with enrollment PBB tertiles that were consistent with the models for estimated current PBB tertiles and bleed length (test for interaction: p = 0.08) (Table 6).
Table 6 Bleed length by weight loss and PBB tertiles, Michigan Female Health Study, 1997–1998. Adjusted for PCB, BMI and age at interview, physical activity, smoking history, education, income.
Ratio of Bleed Length *
Without Weight Loss With Weight Loss
No. Ratio (95% CI) No. Ratio (95% CI)
Tertiles of PBB Estimated at Study Interview §
Low 95 1.00 Ref 17 0.82 (0.72 to 0.93)
Middle 97 0.92 (0.85 to 1.01) 16 0.92 (0.80 to 1.06)
High 101 1.00 (0.91 to 1.09) 11 1.27 (1.08 to 1.50)
Test for Interaction: p < 0.0001, 2 df
Tertiles of PBB Measured at Cohort Enrollment
Low 97 1.00 Ref 14 0.89 (0.78 to 1.03)
Middle 103 1.03 (0.96 to 1.10) 17 0.98 (0.86 to 1.12)
High 93 1.03 (0.96 to 1.11) 13 1.16 (1.00 to 1.34)
Test for Interaction: p = 0.08, 2 df
* Exponentiated from difference in log(bleed length).
We also examined the cumulative distribution functions (CDF) for log-transformed current estimated PBBs. There was no difference in the distribution of PBBs between women who did and did not report past year weight loss (exact two-sided Wilcoxon p = 0.61). We further examined the CDFs for PBB distributions among the 44 women with past year weight loss by menstrual cycle length and bleed length categories (Figure 4). We dichotomized PBB strata for shorter cycle length in the lower quartile (≤ 25 days) and also for longer bleed length in the upper quartile (≥ 6 days). CDFs for PBBs did not differ for women with shorter cycle lengths (Wilcoxon p = 0.63); however, the CDFs for PBBs stratified for longer bleed length were different (Wilcoxon p = 0.006). We observed that women who had longer bleed lengths with past year weight loss were more likely to have higher PBB levels (Figure 4).
Figure 4 Cumulative distribution functions for estimated current log(PBB) among women with weight loss. Shorter cycle length defined as lower 25th%. Longer bleed length defined as upper 75th%. T1 = Low Tertile; T2 = Middle Tertile; T3 = High Tertile.
Aging one year shortened cycle length by 0.10 day (95 percent confidence interval: -0.14 to -0.05) regardless of the PBB categories included in multiple regression models. Bleed length had a J-shaped relationship with age that was modified by estimated current PBB categories (tertile and decile models, tests for interaction: p = 0.04). Due to the curvilinear effect with age modeled as a continuous variable, a clear dose-response in these models was difficult to discern. Therefore, we repeated tests for these interactions with age quintiles cross-classified with estimated current PBB categories (tertile model: p = 0.03; decile model: p = 0.01). The women who were 24–31 years old in the upper tertile of estimated PBB exposure (n = 17) had bleed lengths 0.79 times shorter than the 36 to 38 year olds in the lowest tertile (n = 20) (95 percent confidence interval: 0.67, 0.93). The 24–31 year olds in the upper decile of exposure (n = 8) had bleed lengths 0.70 times shorter (95 percent confidence interval: 0.57, 0.87). We did not observe the same effect for bleed length with enrollment level PBBs. The MFHS women in the youngest age quintile (n = 59) were 6–13 years old at the time of the Michigan incident.
In our sensitivity analysis of women who were the same age as those in the exponential decay model, we excluded 140 women less than 16 years at enrollment (41.5%). We found that, despite this restriction to 197 women, all significant PBB-weight loss interactions persisted for both cycle length (adjusted PBB for upper decile, p = 0.04; initial PBB for upper decile, p= 0.02) and bleed length (adjusted PBB tertiles and upper decile, p < 0.0001; initial PBB tertile, p = 0.01; initial PBB for upper decile, 0.04). For the PBB-age interaction for bleed length with estimated current PBBs, we no longer noted a significant age interaction (tertile model: p = 0.38; decile model: p = 0.32). This is reasonable due to the age restriction imposed on this even smaller sample of women. In addition, among the 197 women, age was no longer an independent predictor of cycle length.
We tested the interaction between PBBs and PCBs for the full sample of 337 women including 44 women missing PCBs, and for the subset of 293 women with both PBB and PCB measures available. No significant interactions were detected for either enrollment or estimated current PBBs with enrollment PCBs, for either cycle length or bleed length. In our secondary analysis of long cycle length, we found no overall association between PBB exposure and menstrual cycle length greater than 30 days (n = 43; upper 10%).
In addition, the sensitivity analyses removing the five women with in utero DES exposure did not affect our results.
Discussion
Normal menstrual cyclicity is a marker of successful follicular recruitment and ovulation. Factors such as aging, weight loss, body composition, and physical activity are thought to affect menstrual cycle function through the hypothalamic-pituitary-adrenal axis [42-52]. Physical activity and age were associated with menstrual cycle function in this sample of women. We observed bleed length shortening with increased physical activity that was consistent with the literature [45,46]. We did not observe menstrual cycle lengthening with physical activity as was reported in another sample of Michigan women [47]. Our observation of cycle length shortening with increasing age is supported by the literature as well. Older women who are still menstruating exhibit accelerated development of dominant follicles and shorter mean cycle lengths due to shorter follicular phases [53], and follicle density has been clinically measured and mathematically modeled to decrease with age [54-56].
Weight loss, lactation, and pregnancy mobilize lipophilic chemicals like PBBs, normally resistant to elimination, from adipose tissue into systemic circulation, and potentially induce menstrual cycle and other endocrine changes [57-61]. We found no overall association between PBB exposure and menstrual cycle length or bleed length, but significant interactions between PBB exposure and past year weight loss not related to pregnancy were found for both cycle length and bleed length. Our results were consistent for both estimated current and measured enrollment PBB concentrations. Among women with weight loss, those in the highest tertile of PBB exposure experienced longer bleed length than women in the lowest tertile. Among the women in the lowest PBB tertile with weight loss, we observed shorter bleed length, as might be expected among women with weight loss in the general population [48-51]. We did not observe any differences in cycle length with PBB exposure or weight loss when PBB was categorized in tertiles. However, among the small number of women in the upper decile of PBB exposure who experienced weight loss, cycle length was shorter by 3.5 days. The direction of this effect is consistent with observations of women exposed to PCBs through fish consumption [9], but not with the longer cycles observed by Cooper et al in their multicenter cohort [10]. On the other hand, we observed no association between PCB exposure and menstrual cycle length or bleed length. Multiple testing and small numbers were a major concern. It is possible that our results were due to chance or artifactual findings. We did, however, observe the same cycle length and bleed length differences among women with weight loss when we restricted our analysis to an even smaller subsample of 197 women at least 16 years old at cohort enrollment.
We also observed significant age-related interactions with PBB exposure for bleed length. We found no evidence in the literature that bleed length is a correlate of age among regularly cycling women in the general population. We observed age-related bleed length differences only in association with current estimated PBB exposure, not with measured enrollment PBB levels. The observed age interaction with PBB exposure could have been spuriously introduced by the exponential decay model we used to estimate current PBB exposure or from small numbers. Nevertheless, we found that women who were 6–13 years old at the time of the Michigan incident (now 24–31 years) in the highest tertile and decile of current estimated PBBs had shorter bleed length relative to 36–38 year old women in the low tertile (n = 20). We restricted women in this study to those who were not transplacentally exposed to PBBs. Therefore, our sample was restricted to women who were exposed to PBBs only by the oral route. The age interaction we observed may indicate that women who were orally exposed before they reached menarche may have been affected during developmental maturation toward menses. Eskenazi et al reported longer cycle length in relation to higher levels of postnatal premenarcheal dioxin exposure, but no differences in bleed length [62]. Blanck et al found earlier age at menarche among a different sample of Michigan girls exposed both in utero to high maternal PBB levels and through breastfeeding [31]. These studies suggest that reproductive events may be affected by the timing and route of developmental and pre-pubertal exposure to halogenated hydrocarbons.
Laboratory assays of urine hormone levels have been useful to document concomitant ovulatory status with self-reported cycle length and bleed length [11,22]. Our reliance on self-reported menstrual cycle information alone is a limitation reported by others [10,63] and may be a source of misclassification. The resulting bias from selective over-reporting of certain days would most likely weight the resulting differences in cycle length toward the median (or toward the null hypothesis) (Figure 2). We did not observe digit preferences in the distribution of bleed length (Figure 3). Bleed length, normally shorter in duration compared to cycle length, may be less prone to recall bias. A larger sample and prospectively collected menstrual data would improve the reliability of these self-reported outcome measures.
We were also limited by our reliance on decay estimates of current PBB exposure. Uncertainties in the original decay model were previously described: lack of information on weight gain or loss during cohort follow-up; lack of fasting requirements for blood draws; lack of serum lipid standardization of the PBB measures; detection methods evolving over time; and most women having only two PBB measurements [30]. Strengths of the decay model over previous PBB half-life studies included: inclusion of PBB determinations over longer follow-up; inclusion of women at much lower initial PBB levels; inclusion of a much larger sample of women; and in addition to age and BMI, the inclusion of information on breastfeeding duration, smoking, and parity [30].
Eligibility requirements for our study differed from the decay model study. The decay analysis restricted inclusion to females who were at least 16 years of age. Our lower age bound only required women to have been born before the contamination episode. Our sensitivity analysis excluding females below 16 years found similar results to our full model. We also included women with enrollment PBB levels below 2 ppb (n = 112), essentially those women in our low enrollment tertile below the LOD (Table 2). It is possible that different kinetics apply at levels below the lower enrollment PBB inclusion bound of the decay model. The decay model found that half-life was shorter among women with low compared to high enrollment PBB levels [30]. Assuming the decay model holds for levels below 2 ppb, and assuming BMI constant, the current PBB estimates would not be biased. If the decay model fails below 2 ppb, third-order kinetics may apply. If the true half-life is shorter for enrollment PBBs below 2 ppb than above, then the decay model would overestimate current PBB levels, given a woman's BMI. In this case, women with low PBB levels at interview would be misclassified in higher adjusted PBB tertiles, and differences in cycle length or bleed length may be underestimated. If the true half-life is longer for enrollment PBBs below 2 ppb, then we would observe an overestimation of effect. We extrapolated our results further out to the time of the MFHS interview (22.4 years maximum) than the original decay model (11.1 years maximum). We acknowledge that this longer extrapolation is a limitation; however, data on long-term decay are not available. We believe that by using this decay model to estimate current PBB exposure that either no bias or an underestimation of effect is more likely, given the complex and nonlinear relationship between initial PBB level, BMI, and half-life described by Blanck et al [30].
Laboratory detection methods for PBBs changed over time to include PCB and pesticide determinations, and are described elsewhere [30,64], while early MDCH biologic sampling protocols did not require fasting blood draws nor standardization for serum lipids [2,60,64-68]. Nonfasting samples have higher mean concentrations of lipophilic compounds and higher total serum lipids than fasting samples. Postprandial increases in serum lipids have been shown to fluctuate with serum chlorinated hydrocarbon levels over a 24-hour period [69]. The lack of fasting requirements may add some measurement error in this present study; however; we believe that it would result in nondifferential misclassification. A woman's exposure status, even if known, is unlikely to prompt her decision to eat in relation to the time of a blood draw. Therefore, bias would most likely be toward the null hypothesis.
More recently, Schisterman et al proposed that lipid standardization is highly prone to bias, and advocate careful definitions of a causal framework for exposure, lipids, and health outcomes. There are study design frameworks when models not standardized for total lipids, equivalent to wet weight analyses, would be preferred [70]. Estrogens and other exogenous hormones are believed to alter plasma lipid and lipoprotein levels [71], and we hypothesize that potential endocrine disruptors, like PBBs or PCBs, may do the same. Since adiposity, energy homeostasis, and ovulation are believed to be intrinsically related [72], adjusting PBB or PCB levels for serum lipids could inappropriately adjust for the exposures themselves [70]. If this framework for the interrelationships holds true, then our lack of lipid standardization may not necessarily be a limitation. Statistical differences have been previously noted comparing lipid adjusted and wet weight analyses. When Cooper et al adjusted their menstrual cycle analyses for lipids, their serum-lipid analyses between PCBs with cycle length were somewhat attenuated compared to wet-weight analyses, but not with bleed length [10].
Our PBB exposure levels quantify only of the main congener, PBB 153. Different PBB congeners may have different endocrine-related effects related to menstrual function as suggested in some studies of PCBs [10,11] and by other proposed disruptors [62,73-78]. In congener-specific analyses, Cooper et al observed shorter cycle length with increasing exposure to PCB 52. This is in contrast to cycle lengthening when total PCBs are considered [10]. Windham et al noted that cycle length may decrease with increasing levels of PCB 187 but not with other congeners [11]. Rat and hamster models for follicular atresia and ovulatory delay with phenobarbital administration have been demonstrated [75-78], and PBB 153 is a phenobarbital-type inducer [27]. Among smokers, folliculotoxic effects of polycyclic aromatic hydrocarbons on primordial oocytes have been implicated in earlier menopause [79-82], while no relation between PBB exposure and time to menopause was found among women in the MFHS [83]. Other chemicals believed to alter endocrine function include dichlorodiphenyltrichloroethane (DDT), 1,1-dichloro-2,2-bis(p-chlorophenyl)ethylene (DDE), and polybrominated diphenyl ethers (PBDEs) [10,11,84,85]. The MFHS is limited by the lack of extensive information on these additional exposures.
Conclusion
This study suggests that PBB exposure may impact ovarian function as indicated by menstrual cycle length and bleed length. However, these associations were found among the small number of women with recent weight loss suggesting either a chance finding or that mobilization of PBBs from lipid stores may be important. These results should be replicated with larger numbers of women exposed to lipophilic compounds.
List of Abbreviations
BMI body mass index
CDF cumulative distribution function
DDE 1,1-dichloro-2,2-bis(p-chlorophenyl)ethylene
DDT dichlorodiphenyltrichloroethane
LOD limit of detection
MDCH Michigan Department of Community Health
MFHS Michigan Female Health Study
PBB polybrominated biphenyl
PBDE polybrominated diphenyl ether
PCB polychlorinated biphenyl
Ppb parts per billion
QQ quantile-quantile
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SID conceived, carried out, and interpreted the analytical models and drafted the manuscript. MM was Principal Investigator and PET, co-PI of the MFHS funded by the U.S. Environmental Protection Agency and the National Institute of Environmental Health Sciences. In that capacity they were responsible for scientific decisions regarding the study. LLC is PI of the Michigan PBB Long-Term Study at the MDCH. VSH oversaw the statistical analyses. MM, PET, HMB, CR and AKH, developed and implemented the protocol for the MFHS, assisted by LLC. HMB, MM, PET, AKH, and VSH conceived and developed the exponential decay mathematical model for the MFHS. LLC and the MDCH study staff provided the investigators access to the cohort, laboratory, and interview data archives on study participants and provided cohort outreach. All authors read and approved the final manuscript.
Acknowledgements
Research was funded by the U.S. Environmental Protection Agency (R 825300-01-1), the National Institute of Environmental Health Sciences and the National Institutes of Health Office of Women's Health (RO1 ES08341-01), and the Centers for Disease Control and Prevention cooperative agreement (U37/CCU500392).
The assistance of Peter McGuire and the staff of the Michigan PBB Long-Term Study, MDCH, and the generosity of cohort members are gratefully acknowledged. Dana Flanders provided valuable advice on statistical analyses and interpretation.
==== Refs
Meester WD The effect of polybrominated biphenyls on man: the Michigan PBB disaster Vet Hum Toxicol 1979 21 131 135 228470
Landrigan PJ Wilcox KR Silva J Humphrey HEB Kauffman C Heath C Cohort study of Michigan residents exposed to polybrominated biphenyls: epidemiologic and immunologic findings Ann NY Acad Sci 1979 320 284 294 222186
Lambrecht LK Barsotti DA Allen JR Responses of nonhuman primates to a polybrominated biphenyl mixture Environ Health Perspect 1978 23 139 145 209967
Allen JR Carstens LA Barsotti DA Residual effects of short-term, low-level exposure of nonhuman primates to polychlorinated biphenyls Toxicol Appl Pharmacol 1974 30 440 451 4141140 10.1016/0041-008X(74)90265-8
Allen JR Barsotti DA Carstens LA Residual effects of polychlorinated biphenyls on adult nonhuman primates and their offspring J Toxicol Environ Health 1980 6 55 66 6770099
Arnold DL Bryce F Stapley R McGuire PF Burns D Tanner JR Karpinski K Toxicological consequences of Aroclor 1254 ingestion by female rhesus (Macaca mulatta) monkeys. Part 1A. Prebreeding phase: clinical health findings Food Chem Toxicol 1993 31 799 810 8258409 10.1016/0278-6915(93)90218-N
Truelove JF Tanner JR Langlois A Stapley RA Arnold DL Mes JC Effect of polychlorinated biphenyls on several endocrine reproductive parameters in the female rhesus monkey Arch Environ Contam Toxicol 1990 19 939 943 2124092 10.1007/BF01055065
Bryce F Hayward S Stapley R Arnold DL Consequences of Aroclor 1254 ingestion on the menstrual cycle of rhesus (Macaca mulatta) monkeys Food Chem Toxicol 2000 38 1053 1064 11038242 10.1016/S0278-6915(00)00094-6
Mendola P Buck GM Sever LE Maria Zielezny Vena JE Consumption of PCB-contaminated freshwater fish and shortened menstrual cycle length Am J Epidemiol 1997 146 955 960 9400337
Cooper GS Klebanoff MA Promislow J Brock JW Longnecker MP Polychlorinated biphenyls and menstrual cycle characteristics Epidemiology 2005 16 191 200 15703533 10.1097/01.ede.0000152913.12393.86
Windham GC Lee D Mitchell P Anderson M Petreas M Lasley B Exposure to organochlorine compounds and effects on ovarian function Epidemiology 2005 16 182 190 15703532 10.1097/01.ede.0000152527.24339.17
Yu ML Guo YL Hsu CC Rogan WJ Menstruation and reproduction in women with polychlorinated biphenyl (PCB) poisoning: long-term follow-up interviews of the women from the Taiwan Yucheng cohort Int J Epidemiol 2000 29 672 677 10922344 10.1093/ije/29.4.672
Marchetti B Morale MC Guarcello V Cutuli N Raiti F Batticane N Palumbo G Farinella Z Scapagnini U Cross-talk communication in the neuroendocrine-reproductive-immune axis. Age-dependent alterations in the common communication networks Ann NY Acad Sci 1990 594 309 325 2165760
Mosley RL Aging, immunity and neuroendocrine hormones Adv Neuroimmunol 1996 6 419 432 9183521
Genazzani AR Petraglia F Gamba O Sgarbi L Greco M Genazzi AD Neuroendocrinology of the menstrual cycle Ann NY Acad Sci 1997 816 143 150 9238263
Brenner PF Differential diagnosis of abnormal uterine bleeding Am J Obstet Gynecol 1996 175 766 769 8828559
Sillem M Rabe T Runnebaum B Effect of central and ovarian endocrine disturbances on the female genital tract – clinical signs and symptoms Verh Dtsch Ges Pathol 1997 81 191 196 9474869
Wathen PI Henderson MC Witz CA Abnormal uterine bleeding Med Clin North Am 1995 79 329 344 7877394
Munro MG Dysfunctional uterine bleeding: advances in diagnosis and treatment Curr Opin Obstet Gynecol 2001 13 475 489 11547028 10.1097/00001703-200110000-00006
Reynolds LP Grazul-Bilska AT Redmer DA Angiogenesis in the female reproductive organs: pathological implications Int J Exp Pathol 2002 83 151 164 12485460 10.1046/j.1365-2613.2002.00277.x
Kleinbaum DG Kupper LL Morgenstern H Epidemiologic Research: Principles and Quantitative Methods 1982 New York, New York: Van Nostrand Reinhold
Fries GF The PBB episode in Michigan: an overall appraisal CRC Crit Rev Toxicol 1985 16 105 156
Metcalf MG Incidence of ovulation from the menarche to the menopause: observations of 622 New Zealand women NZ Med J 1983 96 645 648
Harlow SD Zeger SL An application of longitudinal methods to the analysis of menstrual diary data J Clin Epidemiol 1991 44 1015 1025 1940994 10.1016/0895-4356(91)90003-R
Harlow SD Lin X Ho MJ Analysis of menstrual diary data across the reproductive life span applicability of the bipartite model approach and the importance of within-woman variance J Clin Epidemiol 2000 53 722 733 10941950 10.1016/S0895-4356(99)00202-4
Creinin MD Keverline S Meyn LA How regular is regular? An analysis of menstrual cycle regularity Contraception 2004 70 289 292 15451332 10.1016/j.contraception.2004.04.012
Dent JG Characteristics of cytochrome P-450 and mixed function oxidase enzymes following treatment with PBBs Environ Health Perspect 1978 23 301 307 209991
Needham LL Burse VW Price HA Temperature-programmed gas chromatographic determination of polychlorinated and polybrominated biphenyls in serum J Assoc Off Anal Chem 1981 64 1131 1137 6270054
Burse VW Needham LL Liddle JA Bayse DD Price HA Interlaboratory comparison for results of analyses for polybrominated biphenyls in human serum J Anal Toxicol 1980 4 22 26 6098783
Blanck HM Marcus M Hertzberg V Tolbert PE Rubin C Henderson AK Zhang RH Determinants of polybrominated biphenyl serum decay among women in the Michigan PBB cohort Environ Health Perspect 2000 108 147 152 10656855
Blanck HM Marcus M Tolbert PE Rubin C Henderson AK Hertzberg VS Zhang RH Cameron L Age at menarche and tanner stage in girls exposed in utero and postnatally to polybrominated biphenyl Epidemiology 2000 11 641 647 11055623 10.1097/00001648-200011000-00005
Schantz SL Jacobson JL Humphrey HB Jacobson SW Welch R Gasior D Determinants of polychlorinated biphenyls in the sera of mothers and children from Michigan farms with PCB-contaminated silos Arch Environ Health 1994 49 452 458 7818287
Hornung RW Reed LW Estimation of the average concentrations in the presence of non-detectable values Appl Occup Environ Hyg 1990 5 46 51
Bahn AK Mills JL Snyder PJ Gann PH Houten L Bialik O Hollmann L Utiger RD Hypothyroidism in workers exposed to polybrominated biphenyls N Engl J Med 1980 302 31 33 6243165
Hood A Hashmi R Klaassen CD Effects of microsomal enzyme inducers of thyroid-follicular cell proliferation, hyperplasia, and hypertrophy Toxicol Appl Pharmacol 1999 160 163 170 10527915 10.1006/taap.1999.8752
Hood A Klaassen CD Differential effects of microsomal enzyme inducers on in vitro thyroxine (T4) and triiodothyronine (T3) glucuronidation Toxicol Sci 2000 55 78 84 10788562 10.1093/toxsci/55.1.78
Barter RA Klaassen CD Reduction of thyroid hormone levels and alteration of thyroid function by four representative UDP-glucuronosyltransferase inducers in rats Toxicol Appl Pharmacol 1994 128 9 17 8079359 10.1006/taap.1994.1174
Byrne JJ Carone JP Hanson EA Hypothyroidism and abnormalities in the kinetics of thyroid hormone metabolism in rats treated chronically with polychlorinated biphenyl and polybrominated biphenyl Endocrinology 1987 121 520 527 3036477
Zhu YS Yen PM Chin WW Pfaff DW Estrogen and thyroid hormone interaction on regulation of gene expression Proc Natl Acad Sci 1996 93 12587 12592 8901626 10.1073/pnas.93.22.12587
Wakim AN Paljug WR Jasnosz KM Alkahim N Brown AB Burholt DR Thyroid hormone receptor messenger ribonucleic acid in human granulosa and ovarian stromal cells Fertil Steril 1994 62 531 534 8062948
Doufas AG Mastorakos G The hypothalamic-pituitary-thyroid axis and the female reproductive system Ann NY Acad Sci 2000 900 65 76 10818393
Loucks AB Mortola JF Girton L Yen SSC Alterations in the hypothalamic-pituitary-ovarian and the hypothalamic-pituitary-adrenal axes in athletic women J Clin Endocrinol Metab 1989 68 402 411 2537332
Hu RM Han ZG Song HD Peng YD Huang QH Ren SX Gu YJ Huang CH Li YB Jiang CL Fu G Zhang QH Gu BW Dai M Mao YF Gao GF Rong R Ye M Zhou J Xu SH Gu J Shi JX Jin WR Zhang CK Wu TM Huang GY Chen Z Chen MD Chen JL Gene expression profiling in the human hypothalamus-pituitary-adrenal axis and full-length cDNA cloning Proc Natl Acad Sci 2000 97 9543 9548 10931946 10.1073/pnas.160270997
Kondoh Y Uemura T Murase M Yokoi N Ishikawa M Hirahara F A longitudinal study of disturbances of the hypothalamic-pituitary-adrenal axis in women with progestin-negative functional hypothalamic amenorrhea Fertil Steril 2001 76 748 752 11591409 10.1016/S0015-0282(01)02000-3
Cooper G Sandler DP Whelan EA Smith KR Association of physical and behavioral characteristics with menstrual cycle patterns in women aged 29–31 years Epidemiology 1996 7 624 628 8899389
Harlow SD Campbell BC Host factors that influence the duration of menstrual bleeding Epidemiology 1994 5 352 355 8038253
Sternfeld B Jacobs MK Quesenberry CP Gold EB Sowers M Physical activity and menstrual cycle characteristics in two prospective cohorts Am J Epidemiol 2002 156 402 409 12196309 10.1093/aje/kwf060
Falsetti L Pasinetti E Mazzani MD Gastaldi A Weight loss and menstrual cycle: clinical and endocrinological evaluation Gynecol Endocrinol 1992 6 49 56 1580168
Schweiger U Laessle R Pfister H Hoehl C Schwingenschloegel M Schweiger M Pirke KM Diet-induced menstrual irregularities: effects of age and weight loss Fertil Steril 1987 48 746 751 3117591
Schweiger U Tuschl RJ Platte P Broocks A Laessle RG Pirke KM Everyday eating behavior and menstrual function in young women Fertil Steril 1992 57 771 775 1555687
Pirke KM Schweiger U Strowitzki T Tuschl RJ Laessle RG Broocks A Huber B Middendorf R Dieting causes menstrual irregularities in normal weight young women through impairment of episodic luteinizing hormone secretion Fertil Steril 1989 51 263 268 2912772
De Souza MJ Miller BE Loucks AB Luciano AA Pescatello LS Campbell CG Lashley BL High incidence of luteal phase deficiency and anovulation in recreational female runners: blunted elevation in follicle-stimulating hormone observed during luteal-follicular transition J Clin Endocrinol Metab 1998 83 4220 4232 9851755 10.1210/jc.83.12.4220
Klein NA Battaglia DE Fujimoto VY Davis GS Bremner WJ Soules MR Reproductive aging: accelerated ovarian follicular development associated with a monotropic follicle-stimulating hormone rise in normal older women J Clin Endocrinol Metab 1996 81 1038 1045 8772573 10.1210/jc.81.3.1038
Richardson SJ Senikas V Nelson JF Follicular depletion during the menopausal transition: evidence for accelerated loss and ultimate exhaustion J Clin Endocrinol Metab 1987 65 1231 1237 3119654
Faddy MJ Gosden RG A mathematical model of follicle dynamics in the human ovary Hum Reprod 1995 10 770 775 7650119
Faddy MJ Gosden RG A model conforming the decline in follicle numbers to the age of menopause in women Hum Reprod 1996 11 1484 1486 8671489
Chevrier J Dewailly E Ayotte P Mauriege P Despres JP Tremblay A Body weight loss increases plasma and adipose tissue concentrations of potentially toxic pollutants in obese individuals Int J Obes Relat Metab Disord 2000 24 1272 1278 11093288 10.1038/sj.ijo.0801380
Walford RL Mock D MacCallum T Laseter JL Physiologic changes in humans subjected to severe, selective calorie restriction for two years in Biosphere 2: health, aging, and toxicological perspectives Toxicol Sci 1999 52 61 65 10630592
Longnecker MP Klebanoff MA Gladen BC Berendes HW Serial levels of serum organochlorines during pregnancy and postpartum Arch Environ Health 1999 54 110 114 10094288
Eyster JT Humphrey HEB Kimbrough RD Partitioning of polybrominated biphenyls (PBBS) in serum, adipose tissue, breast milk, placenta, cord blood, biliary fluid, and feces Arch Environ Health 1983 38 47 53 6299210
Stoker TE Goldman JM Cooper RL Delayed ovulation and pregnancy outcome: effect of environmental toxicants on the neuroendocrine control of the ovary Environ Toxicol Pharmacol 2001 9 117 129 11167156 10.1016/S1382-6689(00)00066-1
Eskenazi B Warner M Mocarelli P Samuels S Needham LL Patterson DG Lippman S Vercellini P Gerthoux PM Brambilla P Olive D Serum dioxin concentrations and menstrual cycle characteristics Am J Epidemiol 2002 156 383 392 12181109 10.1093/aje/kwf046
Taffe J Dennerstein L Retrospective self-report compared with menstrual diary data prospectively kept during the menopausal transition Climacteric 2000 3 183 191 11910620
Kreiss K Roberts C Humphrey HE Serial PBB levels, PCB levels, and clinical chemistries in Michigan's PBB cohort Arch Environ Health 1982 37 141 147 6284069
Wolff MS Haymes N Anderson HA Selikoff IJ Family clustering of PBB and DDE values among Michigan dairy farmers Environ Health Perspect 1978 23 315 319 209993
Wolff MS Aubrey B PBB homologs in sera of Michigan dairy farmers and Michigan chemical workers Environ Health Perspect 1978 23 211 215 209978
Wolff MS Aubrey B Camper F Haymes N Relation of DDE and PBB serum levels in farm residents, consumers, and Michigan Chemical Corporation employees Environ Health Perspect 1978 23 177 181 209972
Wolff MS Anderson HA Rosenman KD Selikoff IJ Equilibrium of polybrominated biphenyl (PBB) residues in serum and fat of Michigan residents Bull Environ Contam Toxicol 1979 21 775 781 223695 10.1007/BF01685504
Phillips DL Pirkle JL Burse VW Bernert JT JrHenderson LO Needham LL Chlorinated hydrocarbon levels in human serum: effects of fasting and feeding Arch Environ Contam Toxicol 1989 18 495 500 2505694 10.1007/BF01055015
Schisterman EF Whitcomb BW Louis GM Louis TA Lipid Adjustment in the Analysis of Environmental Contaminants and Human Health Risks Environ Health Perspect 2005 113 853 857 16002372
Barnett JB Woods MN Lamon-Fava S Schaefer EJ McNamara JR Spiegelman D Hertzmark E Goldin B Longcope C Gorbach SL Plasma lipid and lipoprotein levels during the follicular and luteal phases of the menstrual cycle J Clin Endocrinol Metab 2004 89 776 782 14764796 10.1210/jc.2003-030506
Steel JH White R Parker MG Role of the RIP140 corepressor in ovulation and adipose biology J Endocrinol 2005 185 1 9 15817822 10.1677/joe.1.05896
Matikainen T Perez GI Juriscova A Pru JK Schlezinger JJ Ryu HY Laine J Sakai T Korsmeyer SJ Casper RF Sherr DH Tilly JL Aromatic hydrocarbon receptor-driven Bax gene expression is required for premature ovarian failure caused by biohazardous environmental chemicals Nature Genetics 2001 28 355 360 11455387 10.1038/ng575
Gao X Son DS Terranova PF Rozman KK Toxic equivalency factors of polychlorinated dibenzo-p-dioxins in an ovulation model: validation of the toxic equivalency concept for one aspect of endocrine disruption Toxicol Appl Pharmacol 1999 157 107 116 10366543 10.1006/taap.1999.8649
Braw RH Tsafriri A Follicles explanted from pentobarbitone-treated rats provide a model for atresia J Reprod Fertil 1980 59 259 265 7431284
Terranova PF Effects of phenobarbital-induced ovulatory delay on the follicular population and serum levels of steroids and gonadotropins in the hamster: a model for atresia Biol Reprod 1980 23 92 99 6774784
Na JY Garza F Terranova PF Alterations in follicular fluid steroids and follicular hCG and FSH binding during atresia in hamster Proc Soc Exp Biol Med 1985 179 123 127 3921972
Terranova PF Uilenbroek JT Saville L Horst D Nakamura Y Serotonin enhances oestradiol production by hamster preovulatory follicles in vitro: effects of experimentally induced atresia J Endocrinol 1990 125 433 438 2373985
Westhoff C Murphy P Heller D Predictors of ovarian follicle number Fertil Steril 2000 74 624 628 11020495 10.1016/S0015-0282(00)01527-2
Midgette AS Baron JA Cigarette smoking and the risk of natural menopause Epidemiology 1990 1 474 480 2090286
Harlow BL Signorello LB Factors associated with early menopause Maturitas 2000 35 3 9 10802394 10.1016/S0378-5122(00)00092-X
Kato I Toniolo P Akhmedkhanov A Koenig KL Shore R Zeleniuch-Jacquotte A Prospective study of factors influencing the onset of natural menopause J Clin Epidemiol 1998 51 1271 1276 10086819 10.1016/S0895-4356(98)00119-X
Blanck HM Marcus M Tolbert PE Schuch C Rubin C Henderson AK Zhang RH Hertzberg VS Age at menopause in relation to PBBs, PCBs, and smoking Maturitas 2004 49 97 106 15474753 10.1016/j.maturitas.2003.10.011
Schecter A Papke O Tung KC Joseph J Harris TR Dahlgren J Polybrominated diphenyl ether flame retardants in the U.S. population: current levels, temporal trends, and comparison with dioxins, dibenzofurans, and polychlorinated biphenyls J Occup Environ Med 2005 47 199 211 15761315 10.1097/01.jom.0000158704.27536.d2
Petreas M She J Brown FR Winkler J Windham G Rogers E Zhao G Bhatia R Charles MJ High body burdens of 2,2',4,4'-tetrabromodiphenyl ether (BDE-47) in California women Environ Health Perspect 2003 111 1175 1179 12842770
|
16091135
|
PMC1201158
|
CC BY
|
2021-01-04 16:36:32
|
no
|
Environ Health. 2005 Aug 9; 4:15
|
utf-8
|
Environ Health
| 2,005 |
10.1186/1476-069X-4-15
|
oa_comm
|
==== Front
Filaria JFilaria Journal1475-2883BioMed Central London 1475-2883-4-51601416910.1186/1475-2883-4-5ResearchImpact of two rounds of mass drug administration using diethylcarbamazine combined with albendazole on the prevalence of Brugia timori and of intestinal helminths on Alor Island, Indonesia Oqueka Tim [email protected] Taniawati [email protected] Is Suhariah [email protected] [email protected]ückert Paul [email protected] Mark [email protected] Peter [email protected] Bernhard Nocht Institute for Tropical Medicine, Bernhard-Nocht-Strasse 74, D-20359 Hamburg, Germany2 Department of Parasitology, Faculty of Medicine, University of Indonesia, Salemba 6, Jakarta 10430, Indonesia3 U.S. Naval Medical Research Unit No. 2, Jakarta, Indonesia4 German Technical Co-operation (GTZ), P.O. box 1217, Kupang 85000, Indonesia5 Global Community Partnerships, GlaxoSmithKline, 980 Great West Road, Brentfort Middlesex TW8 9GS, U.K6 Department of Internal Medicine, Infectious Diseases Division, Washington University School of Medicine, 660 S. Euclid, Campus box 8051, St. Louis, MO 63110, USA2005 13 7 2005 4 5 5 18 10 2004 13 7 2005 Copyright © 2005 Oqueka et al; licensee BioMed Central Ltd.2005Oqueka 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
Annual mass drug administration (MDA) using diethylcarbamizine (DEC, 6 mg/kg) combined with albendazole (alb, 400 mg) is recommended by the Global Programme to Eliminate Lymphatic Filariasis (GPELF). This strategy has been shown to be efficient in the of control bancroftian filariasis, but data on brugian filariasis as well as on the positive side effects on intestinal helminths are lacking.
Methods
The effect of one selective treatment and two rounds of MDA using DEC and alb on the prevalence and intensity of Brugia timori infection were studied on Alor island using a cross-sectional and a cohort approach. Before the campaign and ten months after each treatment cycle microfilariae (mf) were assessed by filtration of night blood. Before and ten months after MDA, stool samples were collected and the prevalence of intestinal helminths were determined.
Results
In all, the mf-rate dropped from 26.8% before any treatment to 3.8% following the second MDA. Almost all mf-positive, treated individuals showed very low mf densities. The crude prevalence of hookworm dropped from 25.3% to 5.9%. The reduction of prevalence of Ascaris lumbricoides (32.3% to 27.6%) and Trichuris trichiura (9.4% to 8.9%) was less pronounced. Within a cohort of 226 individuals, which was examined annually, the prevalence of A. lumbricoides dropped from 43.8% to 26.5% and of T. trichiura from 12.8% to 6.6%. The results indicate that this MDA approach reduces not only the mf prevalence of B. timori but also the prevalence of hookworm and to a lesser extent also of A. lumbricoides and T. trichiura.
Conclusion
The MDA using DEC and alb as recommended by GPELF is extremely effective for areas with brugian filariasis. The beneficial effect of MDA on intestinal helminths may strengthen the national programme to eliminate lymphatic filariasis in Indonesia and may set resources free which are otherwise used for deworming campaigns of schoolchildren.
==== Body
Background
Lymphatic filariasis (LF) has been targeted by the World Health Organization for elimination as a public health problem by the year 2020 [1,2]. It is caused by three species of filarial parasites: Wuchereria bancrofti infects about 115 million people in Africa, India and other tropical and subtropical areas, whereas Brugia malayi infects about 13 million people in south India and south-east Asia and it is replaced by its sibling species Brugia timori in eastern Indonesia and Timor-Leste [3]. In Asia, the key strategy of the Global Programme to Eliminate LF (GPELF) is an annual mass drug adminstration (MDA) of all individuals at risk of infection with a single annual dose of diethylcarbamazine (DEC) combined with albendazole (alb) for at least four to five subsequent years [2,4]. This approach has been shown to reduce microfilaraemia of W. bancrofti and B. malayi efficiently [5-7]. Studies in areas endemic for W. bancrofti also indicate that the treatment with DEC combined with alb has the additional long-term beneficial effect of reducing prevalence and intensity of infection with intestinal helminths such as Ascaris lumbricoides, hookworms and Trichuris trichiura [8-11].
In 2001, The Department of Health of the Indonesian Government decided to participate in the GPELF. Although Indonesia has a long history in filariasis control programmes [12], filariasis is still in many areas a large public health problem and the new strategy recommended by GPELF was never evaluated. Therefore, studies were initiated to investigate the efficacy of a single dose DEC combined with alb to control W. bancrofti and B. timori infections on Alor island [13]. The treatment was judged to be efficient and safe enough to be employed in an MDA approach and an area endemic for B. timori was selected for annual follow-up studies [13-15].
In the present study we examined the prevalence of B. timori microfilaraemia following one selective treatment and two annual rounds of MDA using DEC combined with alb in a highland village of Alor island. We assessed the prevalence of intestinal helminths before MDA and its impact on the prevalence of the most common intestinal helminths, A. lumbricoides, hookworms and T. trichiura. We provide data on the effectiveness of two annual single doses of DEC and alb treatment to control B. timori and intestinal helminths.
Methods
Study area
The study was performed in Mainang village on Alor island (East Nusa Tenggara Timor, Indonesia). This village was first characterized in April 2001, and it was found to have a prevalence of B. timori microfilaraemia of about 25%. [16]. About 40% of the inhabitants showed signs of infection and 80% presented IgG4 antibodies reactive with a recombinant B. malayi antigen, BmR1 [16,17]. Anopheles barbirostris mosquitoes were identified as vectors for B. timori [18]. An animal reservoir for this Brugia species is not known and wild-life as well as domestic animals were rare in the study village. Before treatment intestinal helminths were common, especially A. lumbricoides (prevalence 2002, 32.3%) and hookworms (prevalence 2002, 25.3%) as well as T. trichuria (prevalence 2002, 9.4%). In the study area a few individuals received diethylcarbamazine (DEC) before 1990 and no anthelminthics were commercially available on the island. However, a deworming campaign was performed in schoolchildren at irregular intervals using generic benzimidazole derivates before the year 2000. In the village about 60% of eligible children attended school.
Sample collection
Originally, the study was planned as a cohort-study with annual re-examinations. Since the most appropriate travel period to Alor is also harvesting season, many farmers were absent from their home during the re-examinations. Therefore, in 2001, 2002, 2003 and 2004 between 37 and 51% of the total resident population (about 1,500 individuals) were examined and data were analysed using a cross-sectional approach. The percentage of newly registered individuals, decreased from 45% during the first annual re-examination in 2002 to 25% in 2003 and to 20% in 2004. By summarizing all surveys, about 90% of the resident population living in the area in 2001 attended at least one survey. Only 145 (10%, 70 males, 75 females, median age in the year 2001, 26 years) individuals participated in all four filariasis surveys and 226 (15%, 117 males, 109 females, median age in the year 2002, 22 years) in the three surveys which included stool collection. Data of these individuals were also analysed separately as a cohort.
In May 2001 about 200 selected individuals with microfilariae (mf) and/or lymphoedema received a single dose of DEC (6 mg/kg) combined with alb (400 mg) [13]. This number represented about 40% of the total number of individuals with mf and/or lymphoedema in the study area. A second survey was performed in April/May 2002 to assess the long-term efficacy of the filariasis treatment in selected individuals. The treatment results of these mf-positive individuals were reported previously [13,14]. For comparison their data were also included in this report. In addition, during the 2002 survey, baseline data on the prevalence of intestinal helminths were collected. Re-examination following MDA was carried out from April to May 2003 and from March to April 2004. Inhabitants of the endemic villages were called by their local health workers to a central place, usually the Puskesmas (primary health center). For each volunteer, sex, age and name were noted and after a brief clinical examination venous blood was collected between 7.00 p.m. and 11 p.m. One labeled stool container was provided to each individual participating in the study and collected one day later by local health workers.
The participants came from three different residential quarters of Mainang, Welai Selatan, Malaipea and Tominuku, but no significant differences with regard to the prevalence of B. timori or intestinal helminths were found [16]. All individuals over the age of two years were asked to participate in the study. Informed consent was obtained from all adults or, in the case of children, from their parents. The study was approved by the ethical board of the University of Indonesia, Jakarta. Following the registration, the individuals were examined by experienced physicians for clinical signs of lymphatic filariasis. During the survey patients with lymphoedma were introduced to hygiene of affected legs and other procedures which may help to stop the progression or alleviate symptoms of their disease.
Mass drug administration
In May and June 2002 and 2003, after each survey, a community based MDA using a single dose of DEC and alb for the entire eligible population was performed. Pregnant or breastfeeding women, children younger than two years or persons suffering from acute illness were not treated. A medication regimen based on the classification of age rather than on weight was used (Table 1), since this was found to facilitate treatment.
Table 1 Simplified dosing regimen used for mass drug administration of DEC (100 mg tablets) and albendazole (400 mg tablets) to control lymphatic filariasis on Alor island, Indonesia, compared to the average body weight (2002, standard deviation, SD).
Age DEC Albendazole Mean body weight in kg (SD)
2–6 years, pre-school 1 tablet 1 tablet 14.4 (6.4)
7–12 years, primary school 2 tablets 1 tablet 23.6 (5.2)
13 years and older, high school and adults 3 tablets 1 tablet 45.7 (9.1)
The MDA of the rest of the island was conducted by the District Health Authority, which claims a coverage rate (number of distributed doses per number of residents) of 78%. Staff from the GTZ performed MDA in 6 randomly selected B. timori and/or W. bancrofti endemic villages with a total population of about 6,000 inhabitants and introduced a developed educational programme [19]. For MDA a mixed approach was used, involving both, the primary health care system and the community members directly. Health staff and volunteers were trained for correct drug distribution and documentation as well as side effect recognition and its treatment. The impact of this campaign was tested by using KAP (knowledge, attitudes, practice) surveys before and after the intervention [19]. In this project a MDA coverage rate of 75% was achieved, whereas 15% of the population, according to WHO guidelines, were not eligible for treatment. Besides the effect of treatment the educational programme has increased the knowledge about filariasis. In our study village Mainang in April 2003 a compliance rate (number of persons who reported to have taken the drug per number of residents) for the 2002 MDA of 67% was determined (n = 772).
Assessment of microfilariae
For the identification of mf-positive individuals, 1 ml of anticoagulated blood (EDTA) was filtered through a polycarbonate membrane with a 5 μm pore size (Millipore, Eschborn, Germany). Subsequently mineral water was passed through the membrane for blood cell lysis. The membrane was placed on a slide, air-dried and fixed with methanol. Following Giemsa staining, slides were air-dried, examined microscopically using 100-fold magnification and mf were counted.
Assessment of helminth eggs
In the field, the Harada-Mori hatching test was used to detect living hookworm larvae. Briefly about 0.5 g fresh stool was spread on a wet filter paper and placed with 1 ml water in a specially designed plastic bag. The bags were placed upright in a window, covered by paper and incubated for 6 to 10 days. The samples were examined microscopically for the presence of hookworm larvae at a 63-fold magnification. The rest of the stool samples were preserved in the field using 4% formaldehyde. In the laboratory in Jakarta, 1–2 g of stool was examined by the formalin/ether enrichment method for the presence of helminth eggs. The most prevalent geohelminths, A. lumbricoides, hookworm and T. trichiura were analysed in this study, since other species such as Hymenolepis spp. and Strongyloides stercoralis were only found in a few cases.
In the baseline survey in 2002 the Kato-Katz smear was used to assess the helminth eggs quantitatively. Compared to the enrichment method and the Harada-Mori test the Kato Katz technique had a poor sensitivity and this method was not applied in the following years. In 2002 the number of eggs per gram (epg) in infected individuals as determined by the Kato Katz smear was usually relatively low with medians for A. lumbricoides, hookworms and T. trichiura of 3,500 epg, 50 epg and 75 epg, respectively. For the enrichment method and the Harada Mori test, the numbers of helminth eggs or hookworm larvae were scored as follows: low density (1–10 eggs per slide or 1–50 hookworm larvae per plastic bag), moderate density (11–100 eggs or 51–500 larvae) or high density (more than 100 eggs or 500 larvae). Usually, scoring data are in good agreement with results obtained by the Kato Katz Smear [20].
Statistical analysis
EpiInfo 2002 Revision2 was used for documentation and analysis of the data. As index for the mf density within a study group the geometric mean was used. For the estimation of the community mf load (CMFL) a log (x+1) transformation was used. Data on distribution and density of mf were compared using the chi-square test or the Mann-Whitney U test.
Results
Brugia timori
In the years 2001 to 2004 a total of 586, 769, 768, and 704 individuals, respectively, were examined for the presence of mf (Table 2, Fig. 1). In 2001, a B. timori mf prevalence of 26.8% was observed, which dropped one year after selective treatment in 2002 to 17.6 %. In 2003, about one year following the first round of MDA, a mf prevalence of 6.1% was detected, while in 2004, about one year following the second round of MDA, a mf prevalence of 3.8% was recorded (Fig. 1A). There was no difference in prevalence reduction between male and female individuals (P > 0.05). In total, from 2001 to 2004 a reduction of mf prevalence of 85% was observed.
Table 2 Number of individuals examined for B. timori from 2001 to 2004 grouped by mf density and the numbers of individuals who claimed to have been treated with DEC/albendazole in the previous year.
Mf/ml 2001 No. (%) 2002 No. (%) 2003 No. (%) 2004 No. (%)
Examined* Treated** Examined* Treated** Examined* Treated** Examined* Treated**
0 429 (73.2) 0 634 (82.4) 126 (19.9) 721 (93.9) 482 (66.9) 677 (96.2) 548 (80.9)
1–100 66 (11.3) 0 90 (11.7) 22 (24.4) 25 (3.3) 17 (68.0) 20 (2.8) 14 (70.0)
101–500 42 (7.2) 0 29 (3.8) 6 (20.7) 14 (1.8) 6 (42.9) 4 (0.6) 2 (50.0)
>500 49 (8.3) 0 16 (2.1) 1 (6.3) 8 (1.0) 4 (50.0) 3 (0.4) 0 (0)
Total 586 (100) 0 769 (100) 155 (20.5) 768 (100) 509 (66.3) 704 (100) 564 (80.1)
*percentage of total examined, **percentage of negative or positive examined
Figure 1 (A) Prevalence of B. timori mf positive individuals in Mainang village, Alor, Indonesia, from 2001 to 2004 by age. The number of individuals examined is noted on top of each column. Selective DEC/albendazole treatment was performed after the survey in 2001. MDA was performed after the surveys in 2002 and 2003. (B) Communtiy microfilarial load (CMFL) of B. timori in Mainang village, Alor, Indonesia, from 2001 to 2004 by age. Due to definition the minimum of CMFL is 1 microfilaria per ml night blood (mf/ml).
The CMFL, the geometric mean number of mf per examined person including the mf-negative individuals, dropped from 3.8 mf/ml in 2001 to 1.1 mf/ml in 2004 (Fig. 1B). However, due to the used definition of the CMFL its minimum is 1 mf/ml and in areas with low prevalence and low mf densities the CMFL becomes inaccurate. While in 2001, 31.2% of 157 mf-positive individuals had high mf densities of more than 500 mf/ml, this percentage decreased in 2002 to 12.2% of 135, in 2003 to 17.0% of 47 and in 2004 to 11.1% of 27 microfilaraemics (Table 2).
The mf-positive individuals, who received their first treatment in 2001, were re-examined after six, twelve, 24 and 34 months. In 2004, 73 of these individuals could be re-examined. Most of them had received three treatments and the prevalence dropped from 100% to 5.5%. The geometric mean mf density of the mf-positive individuals dropped from 142 mf/ml before treatment (2001) to 1.1 mf/ml after 34 months (2004).
In order to determine the dynamics of the mf-status during the investigation period, the mf prevalence in a cohort of 145 individuals, who participated in all four annual surveys, was determined (Fig. 2). In this group 42 (29.0%) individuals, with a geometric mean mf density of 148.6 mf/ml, were mf-positive in 2001. In 2002, 7 formerly mf-negative individuals became microfilaraemic, while 30 formerly mf-positive individuals became amicrofilaraemic after the selective DEC and alb treatment. In 2003, following the first round of MDA, no new microfilaraemics were observed, while 5 individuals remained mf-positive. Three of these individuals had very low mf densities. Two individuals claimed not to have participated in MDA a year before. In both individuals the mf density increased from 2002 to 2003 from 5 to 27 mf/ml and from 726 to 2,723 mf/ml, respectively. In 2004, following the second round of MDA two individuals became microfilaraemic who were mf-negative the year before and one person remained microfilaraemic. The mf density for the 3 mf-positives, who were all positive at the beginning of the observation period, was 1.3 mf/ml, which is a reduction from 2001 to 2004 of >99% for the mf-positives (Fig. 2).
Figure 2 Number of people examined each year from 2001 to 2004 for B. timori and geometric mean of the mf-positives. The white boxes are mf-negative individuals, the grey boxes the mf-positive individuals. The percentages on the arrow indicate the treatment compliance. In the table the boxes marked with the stars differentiate the mf density.
Ascaris lumbricoides
To asses the effects of the MDA on the prevalence of intestinal helminths stool samples were collected in 2002 before MDA and in 2003 and 2004 about 10 months after MDA. In the years 2002, 2003 and 2004 stool samples were collected from 651, 565 and 576 individuals, respectively (Table 3, Fig. 3).
Table 3 Number of individuals examined for intestinal helminths from 2002 to 2004 grouped by infection status with A. lumbricoides, hookworms and Trichuris trichiura and the numbers of individuals who claimed to have been treated with DEC/albendazole in the previous year.
2002 No. (%) 2003 No. (%) 2004 No. (%)
Examined* Treated** Examined* Treated** Examined* Treated**
Ascaris
Negative 450 (67.8) 84 (18.7) 440 (77.9) 309 (70.2) 417 (72.4) 357 (85.6)
Positive 201 (32.2) 26 (12.9) 125 (22.1) 98 (78.4) 159 (27.6) 121 (76.1)
Hookworm
Negative 486 (74.7) 90 (18.5) 515 (91.8) 376 (73.0) 542 (94.1) 458 (84.5)
Positive 165 (25.3) 20 (12.1) 50 (8.2) 31 (62.0) 34 (5.9) 20 (58.8)
Trichuris
Negative 590 (90.6) 104 (17.6) 516 (91.3) 369 (71.5) 525 (91.1) 435 (82.9)
Positive 61 (9.4) 6 (6.6) 49 (8.7) 38 (77.6) 51 (8.9) 43 (84.3)
Total 651 (100) 110 (16.9) 565 (100) 407 (72.0) 576 (100) 478 (83.0)
*percentage of total examined, **percentage of negative or positive examined
Figure 3 (A) Prevalence of A. lumbricoides positive individuals in Mainang village, Alor, Indonesia, from 2002 to 2004 by age. The number of persons examined is noted on top of each column. DEC/albendazole MDA was performed after the surveys in 2002 and 2003. (B) Prevalence of hookworm positive individuals in Mainang village, Alor, Indonesia, from 2002 to 2004 by age. (C) Prevalence of T. trichuria positive individuals in Mainang village, Alor, Indonesia, from 2002 to 2004 by age.
The crude prevalence of A. lumbricoides dropped from 32.3% in 2002 to 22.1% in 2003. However, in 2004 a crude prevalence of 27.6% was observed (Table 3). No difference in prevalence reduction was observed in male and female individuals. Children under the age of ten years had the highest infection rate of 28.7% (2002), 29.6% (2003) and 43.7% (2004), whereas in the other age groups no increase of prevalence was observed (Fig. 3A). In contrast to these crude data, a significant decrease of A. lumbricoides prevalence was observed in a cohort of 226 individuals, who were examined in 2002, 2003 and 2004 (Fig. 4A). In these three years the prevalence dropped in this cohort from 43.8% to 28.3% and to 26.5%. This is a reduction of 39.7% after two rounds of MDA with a compliance rate of 86% and 95%. In 2003, 33 new infections occurred, while 68 individuals who were infected with A. lumbricoides became negative. In 2004, 38 new infections occurred, while 41 persons turned negative (Fig. 4A). In 2003 and 2004 about 90% of the infected individuals had light infections with an estimate of less than 2,000 eggs /g stool, whereas before treatment 50% of the examined individuals had 3,500 eggs /g or more.
Figure 4 (A) Number of people examined each year from 2002 to 2004 for Ascaris and the percentage (in brackets) of the infected (egg-positive) and non-infected (egg-negative) individuals for each group. The percentages on the arrow indicate the treatment compliance. (B) Number of people examined each year from 2002 to 2004 for hookworm and the percentage of the infected and non-infected individuals for each group. The percentages on the arrow indicate the treatment compliance. (C) Number of people examined each year from 2002 to 2004 for Trichuris and the percentage of the infected and non-infected individuals for each group. The percentages on the arrow indicate the treatment compliance.
Hookworms
The crude prevalence of hookworm infection decreased from 25.3% in 2002 to 8.2% in 2003 and the 5.9% in 2004 (Table 3). This is an average reduction of prevalence of 76.7%. In children the rate of infection dropped from 19.3 to 7.2% in 2003 and remained low with 8.2% in 2004, while in the other age groups the decrease of prevalence continued also in 2004 (Fig. 3B). No difference in hookworm reduction was observed in male and female individuals. In the cohort of 226 individuals with full annual data sets, the prevalence dropped from 2002 to 2004 from 34.5% to 17.3% and 14.2%, respectively. This equals following two rounds of MDA within this time course to a reduction of 59% (Fig. 4B). In 2003, 20 new infections were observed and 59 individuals became hookworm negative. In 2004, 25 new infections were detected, while 32 persons turned negative. This corresponds to an average rate of new infections of 13.5% and an average clearance rate of about 80%. The mean egg count before MDA was already relatively low, with about 65% of individuals with light infections of less than 2,000 eggs /g. In 2003 and 2004 the intensity of infection was even lower with more than 90% having light infections. The results from the formol/ether enrichment methods were in principal confirmed by the data obtained from Harada Mori culture, although in a few cases the hatching test was more sensitive.
Trichuris trichiura
The crude prevalence of T. trichiura was almost stable from 2002 to 2004 with only a non-significant decrease from 9.4% to 8.7% and to 8.9%, respectively (Table 3). The prevalence of infected children under the age ten years increased only slightly from 9.6% in 2002 to 11.8% in 2003, but dropped to 7.6% in 2004 (Fig. 3C). A more pronounced reduction of prevalence was observed in the cohort of 226 individuals, who where examined in all three years. In this group the prevalence was in 2002, 12.8% and a year later it was 10.2% and in 2004 it was only 6.6%. This is a significant reduction over two years of 48.4%. In 2003, 18 individuals became negative for T. trichiura eggs, while 12 new infections occurred. A year later 22 individuals turned negative, while 2004, 14 new infections occurred, of which 13 individuals were negative at the surveys in 2002 and 2003 (Fig. 4C). The rate of new infections in 2003 and 2004 was 6.1% and 7.0% respectively. In this cohort 80% of the individuals who were positive in 2002 became negative for T. trichiura in 2004, but new infections occurred. The mean egg count before MDA was already very low, with about 80% of individuals having light infections of less than 2,000 eggs /g. In 2003 and 2004 the intensity of infection continued to be low and almost all individuals had light infections.
Discussion
This study shows the effect of one selective treatment and two rounds of MDA using DEC combined with alb on the most prevalent helminths, B. timori, A. lumbricoides, hookworms and T. trichiura in a community on Alor island. The results confirm and show for the first time at the community level, that the MDA approach as recommended by the GPELF is highly effective at reducing the prevalence and the intensity of infection of B. timori. Furthermore, a positive impact on the prevalence, especially of hookworm infection, but also of A. lumbricoides and of T. trichiura, was observed. The data support the hypothesis that B. timori is an excellent candidate for elimination, while the campaign will also have a large impact on the reduction of intestinal helminths [15]. B. timori occurs only east of the Wallace line in Indonesia and Timor-Leste. This restricted distribution makes this species not only a good candidate for local elimination, but B. timori may also be a prime candidate for eradication of lymphatic filariae [15,21].
B. timori is closely related to B. malayi and tools and strategies developed to support the elimination of B. malayi infection may apply for both species [13-15,17,18]. It has been shown that the combination of a single annual dose of DEC combined with alb is very efficient in the control of brugian filariasis and a higher efficacy of this regimen has been suggested as compared to bancroftian filariasis [6,13,14]. The successful control of B. timori on parts of Flores island using multiple doses of DEC has been reported already some decades ago [12]. However, this strategy has caused many logistical problems and was never extended to the remaining parts of Flores and other islands. The present study proved the principle that a single annual dose of DEC combined with alb is highly suitable to control and most probably to eliminate B. timori infection. Before treatment, among a cohort of 145 individuals, 45 were mf-positive with a geometric mean mf density of almost 150 mf/ml. Three years later, following one selective treatment and two rounds of MDA in this group only 3 individuals were found to be mf-positive, who had mf densities of 1–2 mf/ml.
Our data show that a larger percentage of mf-positive individuals claimed not to have participated in MDA, compared to the average compliance rate. It can be concluded that a high compliance rate is necessary for reducing the mf prevalence to levels under which transmission cannot be sustained. In addition, a few individuals who received no treatment still had high mf densities. If these individuals participate in the next rounds of MDA, side effects may occur and affected individuals may spread their problems in the community which eventually reduce compliance. However, extensive health information campaigns can help to ensure high compliance rates [19].
On the community level not much data exists about the control of brugian filariasis by annual MDA using DEC combined with alb. For the control of W. bancrofti infection in Asia a number of extensive field studies were published using annual MDA with DEC alone or in combination with alb. Results from Papua New Guinea indicate promising prospects for elimination [22,23]. In India, after six rounds of a single dose treatment with DEC the prevalence of microfilaraemics was reduced by 86% and the mf density by 91% [7]. Computer models predict that in this area a further decline of mf prevalence will occur even after the cessation of MDA [24]. In this study (B.timori), similar reductions of prevalence and mf density were observed after only one selective treatment and two rounds of MDA. It is possible that models would also predict a further decline of mf prevalence for our study, but epidemiological parameters differ largely between W. bancrofti infections in Pondicherry (India) and B. timori infections on Alor island and further studies on the dynamics of B. timori control are needed. In addition, it has been discussed for W. bancrofti in the pacific area that MDA should be accompanied by local vector control [25].
Our results showed that MDA using a combination of DEC and alb also has an impact on the reduction of intestinal nematode infections. In the highland village examined the original prevalence and intensity of infections with A. lumbricoides, hookworms and T. trichiura was relatively low. This is in agreement with previous surveys on Alor and on other islands of volcanic origin in east Nusa Tenggara Timur [26,27]. Although, both hookworm species occur in eastern Indonesia, data from Flores indicate that Necator americanus, may be the prominent species [27].
The strongest reduction of prevalence following two rounds of MDA among the intestinal helminths was observed in hookworm infections, in both, the cross-sectional group of an average of 600 individuals and a cohort of 226 individuals. Although a large number of re-infections occurred, the crude prevalence dropped from 25.3% to 5.9% and in the cohort from 34.5% to 14.2%, ten months after the second round of MDA. This equals a reduction of 76.7% and 58.8%, respectively. The hookworm prevalence is usually reduced by about 80% shortly after treatment with alb [28,29]. Although DEC alone may reduce the output of hookworm eggs, it is assumed that it has no influence on its prevalence [30]. In our study we observed a large number of hookworm re- or new infections. In 13.5% (2003) and 14.0% (2004) of the cohort new infections were observed. From Java, an even higher re-infection rate with N. americanus of about 50% one year after anthelminthic treatment was reported [31]. Despite of the occurrence of re-and new infections, the drop in hookworm prevalence can be explained by the relatively short survival time of hookworm larvae in the environment as compared to the mean survival time of eggs of A. lumbricoides and of T. trichiura.
Following the first round of MDA the crude prevalence of A. lumbricoides dropped from 32.2% to 22.1%, but after the second round it was 27.6%. More consistently were the results in the cohort. Before MDA the prevalence was 43.3%, following the first round it was 28.3% and following the second round it was 26.5%. Albendazole is very effective against A. lumbricoides and median cure rates are over 95% [28,29,32]. DEC alone has a minor therapeutic effect on Ascaris [10,33]. Although some treated individuals may expel adult worms after DEC, the overall prevalence of infection may be not affected [30]. These observations are confirmed by other studies, which show that DEC alone has no significant impact on A. lumbricoides, but the combination of DEC with alb has relevant cure and egg reduction rates [34]. As in other intestinal helminths, re- and new infections occur regularly, and the time point of re-examination is critical. In our cohort we observed an annual rate of new infections of 25%. Children have an especially high risk for re- or new-infections and show a lower decrease in worm burden compared to adults [35]. In another study it was observed that eight months after treatment 55% of children were re-infected [36]. Re-infection with A. lumbricoides may return six months after treatment to almost 90% of the pre-treatment prevalence and worm density may drop to about 75% [35].
The crude prevalence of T. trichiura in the community before and after MDA was almost identical, ranging between 9.4% and 8.7%. However, following two rounds of MDA the prevalence dropped in the cohort from 12.8% to 6.6%. Although re- and new infections occurred, it is important to note that in 2004 most new infections were observed in those individuals which were negative for T. trichiura for the previous two years. Trichuris trichiura is known to be only poorly sensitive to albendazole and the reported reduction rates for alb range between 38% and 47.7% [28,29,32]. From Sri Lanka a cure rate of T. trichiura of 43.6% and an egg reduction rate of 70.3% was reported [37]. The combination of DEC with alb showed different results, ranging from no significant impact on the prevalence but with significant egg reduction of 79.4% one week after treatment, to a cure rate of 81.6% and an egg reduction of 84% [10,34]. The study from Sri Lanka reported that this drug combination has a cure rate of 30% and an egg reduction rate of 70% [36].
For areas endemic for W. bancrofti there are an increasing number of studies which show the positive effect of filariasis control using MDA with DEC combined with alb on the reduction of intestinal helminths [8-11,38]. The results of the present study can extend this observation to areas endemic for Brugia infections. Although it is unlikely that MDA as used for filariasis elimination will eliminate intestinal helminths from most areas, a reduction may be achieved to levels which may cause no significant morbidity. Other intervention strategies, such as for example the development of a hookworm vaccine [39], may take advantage of reduced prevalences in order to achieve a long-lasting elimination of intestinal helminths, as it has been accomplished in most industrialised countries. In areas with filariasis control by the MDA using DEC combined with alb, separate de-worming campaigns for school-age children may become superfluous. This could set available resources free which can then be used to support MDA. Co-ordination is needed within the local health administration to use the limited funds more efficiently. The present study showed that MDA using DEC combined with alb is effective to control B. timori and that this has also impact on the reduction of geohelminths.
Conclusion
Annual MDA using DEC in combination with alb is highly effective in the control of B. timori infection and has a positive impact on the reduction of intestinal helminths. Given a high compliance rate the strategy can lead to elimination of B. timori on Alor and on other islands in Indonesia.
List of abbreviations
Alb, albendazole
CMFL, community microfilarial load
DEC, diethylcarbamazine
GPELF, Global Programme to Eliminate Lymphatic Filariasis
MDA, mass drug admistratrion
Mf, microfilariae
Competing interests
Mark Bradley is employee of GlaxoSmithKline, which donate albendazole for filariasis elimination.
Authors' contributions
Tim Oqueka cand. MD, participated in field work, performed data analysis and wrote the first draft of the manuscript
Taniawati Supali, PhD, conceived the study, participated in field work, performed stool examinations and edited the manuscript
Is Suhariah Ismid MD, participated in the base-line survey and made comments on the manuscript
Purnomo PhD, participated in the base-line survey, performed stool examinations and made comments on the manuscript
Paul Rückert MD, MPH, PhD conceived the study, provided logistic support and edited the manuscript
Mark Bradley PhD, conceived the study, helped with data analysis and with writing of the manuscript
Peter Fischer PhD, conceived the study, participated in all field surveys, helped with data analysis and drafted the manuscript
Acknowledgements
The authors appreciate the help of Dr. Paul Manoempil and his team of the Alor District Health Administration. This study would have been not possible without the help and understanding of the local health workers and the entire population of Mainang village. Furthermore, we like to thank Yenny Djuardi for technical assistance and Alison Krentel for sharing unpublished data. This study formed part of a doctoral study of T.O. at the Faculty of Medicine of the University of Hamburg, Germany. GlaxoSmithKline, London, U.K. and the "Deutsche Gesellschaft für Technische Zusammenarbeit (GTZ/SISKES)" provided financial support for the field work. P.F. received a scholarship of the "Vereinigung der Freunde des Tropeninstituts Hamburg".
==== Refs
Bebehani K Candidate parasitic diseases Bull World Health Organ 1998 76 64 67 10063677
Ottesen EA Duke BO Karam M Behbehani K Strategies and tools for the control/elimination of lymphatic filariasis Bull World Health Organ 1997 75 491 503 9509621
Michael E Bundy DA Global mapping of lymphatic filariasis Parasitol Today 1997 13 472 476 15275135 10.1016/S0169-4758(97)01151-4
Molyneux D Lymphatic filariasis (elephantiasis) elimination: a public health success and development opportunity Filaria J 2003 2 13 13129436 10.1186/1475-2883-2-13
Horton J Witt C Ottesen EA Lazdins JK Addiss DG Awadzi K Beach MJ Belizario VY Dunyo SK Espinel M Gyapong JO Hossain M Ismail MM Jayakody RL Lammie PJ Makunde W Richard-Lenoble D Selve B Shenoy RK Simonsen PE Wamae CN Weerasooriya MV An analysis of the safety of the single dose, two drug regimens used in programmes to eliminate lymphatic filariasis Parasitology 2000 147 160 10.1017/S0031182000007423
Shenoy RK John A Babu BS Suma TK Kumaraswami V Two-year follow-up of the microfilaraemia of asymptomatic brugian filariasis, after treatment with two, annual, single doses of ivermectin, diethylcarbamazine and albendazole, in various combinations Ann Trop Med Parasitol 2000 94 607 614 11064762
Ramaiah KD Vanamail P Pani SP Yuvaraj J Das PK The effect of six rounds of single dose mass treatment with diethylcarbamazine or ivermectin on Wuchereria bancrofti infection and its implications for lymphatic filariasis elimination Trop Med Int Health 2002 9 767 774 12225508 10.1046/j.1365-3156.2002.00935.x
De Silva NR Pathmeswaran A Fernando SD Weerasinghe CR Padmasiri EA Montressor A Impact of mass chemotherapy for the control of filariasis on geohelminth infections in Sri Lanka Ann Trop Med Parasitol 2003 97 421 425 12831528 10.1179/000349803235002407
Rajendran R Mani TR Munirathinam A Sunish IP Abdullah SM Augustin DJ Satyanarayana K Sustainability of soil-transmitted helminth control following a single-dose co-administration of albendazole and diethylcarbamazine Trans R Soc Trop Med Hyg 2003 97 355 359 15228259 10.1016/S0035-9203(03)90168-4
Mani TR Rajendran R Munirathinam A Sunish IP Abdullah S Augustin DJ Satyanarayana K Efficacy of co-administration of albendazole and diethylcarbamazine against geohelminthiases: a study from South India Trop Med Int Health 2002 7 541 548 12031078 10.1046/j.1365-3156.2002.00894.x
Mani TR Rajendran R Sunish IP Munirathinam A Arunachalam N Satyanarayana K Dash AP Effectiveness of two annual, single-dose mass drug administrations of diethylcarbamazine alone or in combination with albendazole on soil-transmitted helminthiasis in filariasis elimination programme Trop Med Int Health 2004 9 1030 1035 15361118 10.1111/j.1365-3156.2004.01298.x
Partono F Maizels RM Purnomo Towards a filariasis-free community: evaluation of filariasis control over an eleven year period in Flores, Indonesia Trans R Soc Trop Med Hyg 1989 83 821 826 2617653 10.1016/0035-9203(89)90343-X
Supali T Ismid I Rückert P Fischer P Treatment of Brugia timori and Wuchereria bancrofti infections in Indonesia using DEC or a combination of DEC and albendazole: adverse reactions and short-term effects on microfilariae Trop Med Int Health 2002 7 894 901 12358626 10.1046/j.1365-3156.2002.00921.x
Fischer P Djuardi Y Ismid I Rückert P Bradley B Supali T Long-lasting reduction of Brugia timori microfilariae following a single dose of diethylcarbamazine combined with albendazole Trans R Soc Trop Med Hyg 2003 97 446 448 15259479 10.1016/S0035-9203(03)90086-1
Fischer P Supali T Maizels RM Lymphatic filariasis and Brugia timori: prospects for elimination Trends Parasitol 2004 20 351 355 15246315 10.1016/j.pt.2004.06.001
Supali T Wibowo H Rückert P Fischer K Ismid I Purnomo Djuardi Y Fischer P High prevalence of Brugia timori infection in the highland of Alor island, Indonesia Am J Trop Med Hyg 2002 66 560 565 12201590
Supali T Rahmah N Djuardi Y Sartono E Ruckert P Fischer P Detection of filaria specific IgG4 antibodies using Brugia Rapid test in individuals from an area highly endemic for Brugia timori Acta Trop 2004 90 255 261 15099812 10.1016/j.actatropica.2004.02.001
Fischer P Wibowo H Pischke S Ruckert P Liebau E Ismid IS Supali T PCR-based detection and identification of the filarial parasite Brugia timori from Alor Island, Indonesia Ann Trop Med Parasitol 2002 96 809 821 12625936 10.1179/000349802125002239
Krentel A Rückert P Servais G Manoempil P Fischer P A knowledge, attitudes and practice (KAP) survey of lymphatic filariasis to prepare and evaluate mass drug administration in Alor District, Indonesia
Muller R Worms and Human Disease 2002 2 CABI Publishing, Wallingford, UK
Molyneux DH Hopkins DR Zagaria N Disease eradication, elimination and control: the need for accurate and consistent usage Trends Parasitol 2004 20 347 351 15246314 10.1016/j.pt.2004.06.004
Bockarie MJ Tisch DJ Kastens W Alexander ND Dimber Z Bockarie F Ibam E Alpers MP Kazura JW Mass treatment to eliminate filariasis in Papua New Guinea N Engl J Med 2002 347 1841 1848 12466508 10.1056/NEJMoa021309
Bockarie MJ Kazura KW Lymphatic filariasis in Papua New Guinea: prospects for elimination Med Microbiol Immunol 2003 192 9 14 12592558
Subramanian S Stolk WA Ramaiah KD Plaisier AP Krishnamoorthy K Van Oortmarssen GJ Dominic Amalraj D Habbema JD Das PK The dynamics of Wuchereria bancrofti infection: a model-based analysis of longitudinal data from Pondicherry, India Parasitology 2004 128 467 482 15180315 10.1017/S0031182004004822
Burkot T Ichimori K The PacELF programme: will mass drug administration be enough? Trends Parasitol 2002 18 109 115 11854087 10.1016/S1471-4922(01)02221-8
Joesoef A Dennis DT Intestinal and blood parasites of man on Alor Island Southeast Indonesia Southeast Asian J Trop Med Public Health 1980 11 43 47 7403953
Higgins DA Jenkins DJ Kurniawan L Purnomo Harun S Juwono SS Human intestinal parasitism in three areas of Indonesia: a survey Ann Trop Med Parasitol 1984 78 637 648 6543421
Horton J Albendazole: a review of anthelminthic efficacy and safety in humans Parasitlogy 2000 113 132 10.1017/S0031182000007290
Bennett A Guyatt H Reducing intestinal nematode infection: efficacy of albendazole and mebendazole Parasitol Today 2000 16 71 74 10652492 10.1016/S0169-4758(99)01544-6
Meyrowitsch DW Simonsen PE Short communication: efficacy of DEC against Ascaris and hookworm infections in schoolchildren Trop Med Int Health 2001 6 739 742 11555442 10.1046/j.1365-3156.2001.00766.x
Soeripto N Reinfection and infection rates of soil-transmitted-helminths in Kemiri Sewu, Yogyakarta, Indonesia Southeast Asian J Trop Med Public Health 1991 22 216 221 1948282
Ottesen EA Ismail MM Horton J The role of albendazole in programmes to eliminate lymphatic filariasis Parasitol Today 1999 15 382 386 10461168 10.1016/S0169-4758(99)01486-6
Turner P Michael E Recent advances in the control of lymphatic filariasis Parasitol Today 1997 13 410 411 10.1016/S0169-4758(97)90021-1
Belizario VY Amarillo ME de Leon WU de los Reyes AE Bugayong MG Macatangay BJC A comparison of the efficacy of single doses of albendazole, ivermectin and diethylcarbamazine alone or in combinations against Ascaris and Trichuris spp Bull World Health Organ 2003 81 35 42 12640474
Hall A Anwar KS Tomkins AM Intensity of reinfection with Ascaris lumbricoides and its implications for parasite control Lancet 1992 339 1253 1257 1349668 10.1016/0140-6736(92)91593-W
Hagel I Lynch NR Di Prisco MC Perez M Sanchez JE Pereyra BN Soto de Sanabria I Helminth infection and anthropometric indicators in children from a tropical slum: Ascaris reinfection after anthelminthic treatment J Trop Pediatr 1999 45 215 220 10467832 10.1093/tropej/45.4.215
Ismail MM Jayakody RL Efficacy of albendazole and its combinations with ivermectin or diethylcarbamazine (DEC) in the treatment of Trichuris trichiura infections in Sri Lanka Ann Trop Med Parasitol 1999 93 501 504 10690245 10.1080/00034989958230
Pani SP Subramanyam Reddy G Das LK Vanamail P Hoti SL Ramesh J Das PK Tolerability and efficacy of single dose albendazole, diethylcarbamazine citrate(DEC) or co-administration of albendazole with DEC in the clearance of Wuchereria bancrofti in asymptomatic microfilaraemic volunteers in Pondicherry, South India: a hospital-based study Filaria J 2002 1 1 12537598 10.1186/1475-2883-1-1
Hotez PJ Zhan B Bethony JM Loukas A Williamson A Goud GN Hawdon JM Dobardzic A Dobardzic R Ghosh K Bottazzi ME Mendez S Zook B Wang Y Liu S Essiet-Gibson I Chung-Debose S Xiao S Knox D Meagher M Inan M Correa-Oliveira R Vilk P Shepherd HR Brandt W Russell PK 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 10.1016/S0020-7519(03)00158-9
|
16014169
|
PMC1201159
|
CC BY
|
2021-01-04 16:38:25
|
no
|
Filaria J. 2005 Jul 13; 4:5
|
utf-8
|
Filaria J
| 2,005 |
10.1186/1475-2883-4-5
|
oa_comm
|
==== Front
Front ZoolFrontiers in Zoology1742-9994BioMed Central London 1742-9994-2-141613139410.1186/1742-9994-2-14ResearchThe unsuitability of html-based colour charts for estimating animal colours – a comment on Berggren and Merilä (2004) Stevens Martin [email protected] Innes C [email protected] School of Biological Sciences, University of Bristol, Woodland Road, Bristol, BS8 1UG, UK2005 30 8 2005 2 14 14 28 4 2005 30 8 2005 Copyright © 2005 Stevens and Cuthill; licensee BioMed Central Ltd.2005Stevens and Cuthill; 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 variety of techniques are used to study the colours of animal signals, including the use of visual matching to colour charts. This paper aims to highlight why they are generally an unsatisfactory tool for the measurement and classification of animal colours and why colour codes based on HTML (really RGB) standards, as advocated in a recent paper, are particularly inappropriate. There are many theoretical arguments against the use of colour charts, not least that human colour vision differs markedly from that of most other animals. However, the focus of this paper is the concern that, even when applied to humans, there is no simple 1:1 mapping from an RGB colour space to the perceived colours in a chart (the results are both printer- and illumination-dependent). We support our criticisms with data from colour matching experiments with humans, involving self-made, printed colour charts.
Results
Colour matching experiments with printed charts involving 11 subjects showed that the choices made by individuals were significantly different between charts that had exactly the same RGB values, but were produced from different printers. Furthermore, individual matches tended to vary under different lighting conditions. Spectrophotometry of the colour charts showed that the reflectance spectra of the charts varied greatly between printers and that equal steps in RGB space were often far from equal in terms of reflectance on the printed charts.
Conclusion
In addition to outlining theoretical criticisms of the use of colour charts, our empirical results show that: individuals vary in their perception of colours, that different printers produce strikingly different results when reproducing what should be the same chart, and that the characteristics of the light irradiating the surface do affect colour perception. Therefore, we urge great caution in the use of colour charts to study animal colour signals. They should be used only as a last resort and in full knowledge of their limitations, with specially produced charts made to high industry standards.
==== Body
Background
The use of colour charts to estimate or categorise the colours of animal signals is a technique utilised in numerous studies [e.g. [1-5]]. In particular, a recent proposal [1] argues that researchers could produce custom-made charts, designed from the HTML colour code (the standard for colour representation on the World Wide Web). In this paper we present theory and data showing why the use of such charts to estimate colour is seriously flawed and only should be undertaken as a last resort.
'Colour is in the eye of the beholder'
A major problem with using colour charts is one frequently stressed: that human vision differs markedly from most animals other than Old World primates [6-11]. Signals are often aimed at specific animals, and it has long been realised that there is an association between the evolution of a particular signal and the receivers' visual system [12], and so signals should be considered from the perspective of the signal receivers' sensory experience [6,13]. The description of a certain colour is something specific to a particular visual system, and this perception may differ greatly between animals [6-8,14].
Colour perception is the product of reflectance, the irradiant light characteristics, the transmission characteristics of the medium, and the characteristics of the animal's visual system [6]. Most of the hues an animal can perceive can be produced by mixing wavelengths of light (called primaries) in different proportions, and so: (a) different light spectra can produce a sensation of the same hue if the output of the animal's photoreceptor types is the same (metamerism), (b) the same spectra will produce different hues to animals that differ in the absorption spectra of their photoreceptors, and (c) the dimensionality of colour space is determined by the number of interacting receptor types [8,15,16]. For instance, birds typically have four single cone types, compared to three in humans, and unlike humans, most birds are capable of perceiving light into the ultraviolet spectrum [[17-24], reviewed by [25-28]]. This means that birds should be capable of perceiving a wider range of hues, and will differ from humans in the magnitude of perceived colour differences, even for those spectra visible to humans.
Whilst avian vision has been described to illustrate why human colour-matching can never quantify the colours perceived by other animals, it is equally important to realise that colour matching to charts of the type proposed in [1] is not even adequate for human perception.
The inadequacy of colour charts to classify colour
To understand why using certain colour charts to study visual signals is often inadequate, it is helpful to briefly consider some of the main aspects of the colour spaces from which charts are created.
One way to represent colour is to agree on a set of primaries and describe a colour by the values of the weights of those primaries used by subjects to match a test light (additive colour mixing) [29,30]. The CIE (Commission International d'Éclairage) XYZ colour space is one such specification of colour stimuli, produced by additive mixing of three imaginary primaries.
It is usually important to know if a colour difference is perceivable, determined by experimentally modifying colours in small degrees to determine threshold perceptible differences. When plotted in colour space, these differences form the boundary of a region of colours that are indistinguishable from other colours, with ellipses fitted to the boundaries [29]. In colour spaces such as CIE XYZ, the shape and size of the ellipses depends strongly on the location of the difference in the colour space, meaning that the magnitude of the difference in CIE XYZ space is a poor indicator of the real perceived difference between colours [29-31]. Therefore, what is often preferable is a uniform colour space where the distance in coordinate space is comparable to the perceived difference in colour by an observer.
Currently, the most popular uniform colour space is the CIELAB space, obtained by a non-linear transformation of the XYZ space. This colour space is uniform, meaning that equations allow the Euclidian distance between two points in the CIELAB space to predict more accurately the observed difference in colour, although comparisons of 'colour constancy' in CIELAB to empirically measured colour constancy are still often quite poor [32].
It is a misconception that RGB (red, green, blue) colour space is an accurate method to classify colours as seen by humans; it is not (and certainly is not for non-human animals). Indeed, it is not generally associated with studies aiming to match colours to charts. RGB space is most readily associated with colour reproduction on computers, and with its associated CMY colour space for printing. RGB values are those used to represent digital photographs on a colour monitor, with values of R, G, and B usually ranging from 0 to 255 (an 8-bit scale). HTML colour-coding (as advocated by [1]) is simply a concise encoding of the RGB colour format.
There are several criticisms of using RGB colour codes to specify colours from charts. Firstly, RGB space is non-uniform, and therefore differences in RGB values do not equate to equal differences in colour perception. Secondly, unlike the CIELAB space, RGB ratios are not capable of producing all the possible perceptual combinations of colours to humans (let alone to other species). For instance, values of L* = 100, a* = -80 and b* = -2 changing continuously to values of L* = 100, a* = -80 and b* = -59 in CIELAB space, are all represented by the same RGB values (R = 0, G = 255, B = 255), and therefore, differences in the colours produced by CIELAB space over this range simply cannot be reproduced on a computer screen or on a printed chart. Thirdly, and perhaps most importantly, unlike the various CIE colour spaces, the colours generated from RGB colour co-ordinates are device-dependent [31]. That is, a photographic image of a given colour may be represented by different RGB values in different cameras, and a given RGB coordinate in a camera or computer may translate to different colours on different printers.
The experiments in this paper were designed to demonstrate that the faults with Berggren & Merilä's [1] approach are not simply theoretical abstractions. The experiments illustrate the contentions that individual human observers vary in their assessment and ranking of colours, that the surface irradiance characteristics may affect perceptions of colour and, finally, that different printers will produce colour sheets with significantly different reflectance values, even if the RGB/HTML values of the colour charts on a computer are identical (discounting variation arising from different toner levels which are, nonetheless, an important consideration in practice). These are not new arguments, and more thorough experiments are routine in colour science, but rather are presented here to illustrate the pitfalls for those studying animal colouration.
Results
Spectrophotometry Results
Spectrophotometric data supported the results obtained from the colour matching experiments (below). Reflectance spectra of the colour charts showed that there is significant variation in reflectance (shape and intensity) between the different printer types (Fig. 1). This variation leads to the large diversity in the perceived colours and brightness of the different chart components between printers. No two printers were the same in their reflectance for each of the colour blocks, and this was the case for both the red and the blue-green charts. Also, and perhaps most worryingly, whilst some printers showed a fairly constant increase in reflectance as the R or G/B value increased, many printers produced charts where there were sudden large jumps between what should have been equal steps between the colour blocks, or had several colour blocks at the top end of the RGB values having very similar reflectance (i.e. even parts of the colour chart that were separated by RGB values of 25, 50, or more, would sometimes produce similar reflectance spectra) (Fig. 2), showing that some printers are constrained to smaller variations in colour spacing (lower colour resolution).
Figure 1 Spectrophotometry Results Showing the Variation that Exists Between the Different Printers. Two plots of the average reflectance spectra from the 8 different printer types, for the blue-green colour blocks with B & G values of 200 and R value of 0 (top chart), and the red colour blocks with an R value of 175 and B & G values of 0 (bottom chart). This shows the amount of variation that exists between the different printers, in producing what should have been colour blocks with the same spectra. The spectra of the coloured paint cards used in the matching experiments are included for reference.
Figure 2 Spectrophotometry Results Showing the Uneven Scaling of Spectra with Equally Spaced RGB Values. The two plots show the average reflectance spectra for all the colour blocks included in the study on the red charts, for two of the printers used in the experiments. The top chart shows a 'good' printer, which produced a printed chart with approximately even spacing between the blocks with increasing R-values. This contrasts strongly with the bottom chart, produced from a different printer, which shows an uneven spacing between the spectra, and a 'bunching' of spectra at low and high R-values, meaning that the even spaces in RGB space did not correspond to an even spacing in spectral intensity.
Colour Matching Experiments
Results from the colour matching experiments showed that for the red charts, there were significant and large effects on subjects' colour matching of the printer and a marginal, non-significant effect of lighting conditions (repeated measures GLM; Printer: F(7,70) = 33.00, P < 10-19, partial eta2 = 0.767; Light source: F(3,30) = 2.74, P = 0.061, partial eta2 = 0.215; Printer*Light source: F(21,210) = 1.178, P = 0.273, partial eta2 = 0.105) (Figs 3 &4).
Figure 3 Colour Matching Results with the Red Charts for the Different Printer Types. Variation in colour matching choices that subjects made (mean red value on RGB scale, with standard deviation bars) for the eight different red colour charts produced from different printers. Results from all colour matches made, averaged across lighting conditions and subjects.
Figure 4 Colour Matching Results with the Red Charts Under Different Light Conditions. Variation in colour matching choices that subjects made (mean red value on RGB scale, with standard deviation bars) under the Xenon, incandescent, laboratory and skylight light conditions. Results from all colour matches made, averaged across printers and subjects.
For the blue-green charts, there was also a significant effect on colour matching of printer (F(7,70) = 145.54, P < 10-38), with the effect size even greater (partial eta2 = 0.936). Light type had no detectable main effect (F(3,30) = 1.40, P = 0.263, partial eta2 = 0.122) but there was a significant light*printer interaction (F(21,210) = 2.09, P = 0.005, partial eta2 = 0.173) (Figs 5 &6). The results contained a single large outlier in the data, with a large standardised residual value. However, there was little change to the results of the GLM when this outlier was removed (to determine its potential influence on the data).
Figure 5 Colour Matching Results with the Blue-Green Charts for the Different Printer Types. Variation in colour matching choices that subjects made (mean blue/green value on RGB scale, with standard deviation bars) for the eight different blue-green colour charts produced from different printers. Results from all colour matches made, averaged across lighting conditions and subjects.
Figure 6 Colour Matching Results with the Blue-Green Charts Under Different Light Conditions. Variation in colour matching choices that subjects made (mean blue/green value on RGB scale, with standard deviation bars) under the Xenon, incandescent, laboratory and skylight light conditions. Results from all colour matches made, averaged across printers and subjects.
These results show that, for both the blue-green and red chart experiments, the colour matching choices that subjects made were very different for charts produced from different printers. There was also a suggestion of a smaller effect of light conditions on colour matching. The variation in colour-matching judgements was large: for the red experiment, the same target colour was matched to chart elements with R-pixel values ranging from 125 to 250 (C.V. = 16%); for the blue experiment, the best match ranged from B/G-pixel values of 75 to 250 (C.V. = 25%). Although not usually interpretable in repeated-measures ANOVA, one could argue that the between-subject variation is of direct interest in this application, because in many applications of colour-matching, only one or a few researchers would be responsible. When treated as a fixed effect, 'subject' was significant for both colour-matching tasks (red: F(10,210) = 7.70, P < 10-9; blue: F(10,210) = 2.55, P = 0.006), although the effect sizes were not as substantial as that of printer type (partial eta2 = 0.268 and 0.108, respectively).
Discussion
The experiments detailed in this study show three important results with respect to the use of printed colour charts to identify the colours of animal signals. Firstly, people vary with respect to the choices they make when asked to match one colour to the perceived closest match from a set of colours on a chart. This means that there may be differences in colour matches made by different individuals. These differences may, to some extent, be reduced by using high-quality charts with a greater range of colour matching options. Secondly, there were very large and significant differences in the matching choices made by individuals between charts produced from different printers. Different inks have significantly different spectral properties [30]. This is a critical problem of 'self-made' charts, such as those made from RGB colour space. The aim of matching a colour signal to a specific section of a colour chart is that the chromatic content of the signal can be recorded, such as in terms of an RGB value. However, when charts with identical RGB values are produced from different printers these do not have the same properties, making comparisons to an RGB value irrelevant. Even if charts are printed from the same printer, with the same cartridge model and paper type, the exact values of the printed charts will still vary depending upon the toner levels (Cuthill & Stevens unpublished data). This means that complex printer calibrations are needed to ensure that the reproduction of more than one colour chart is accurate and invariable with respect to the chart properties. Furthermore, there still remains the problem that a linear increase in a colour value on a chart may not be linear in terms of the measured reflectance spectra and the perceived difference in colour, for many, if not all colour spaces. Thirdly, and perhaps least expected, there was the suggestion that colour matching results were significantly affected by the irradiant light conditions (non-significant for red charts, at P = 0.06, but in a significant interaction with printer type for the blue chart). For the red charts, the largest mean difference for judgements measured under different light sources was a difference of 14 pixel values on an 8-bit scale. For the blue-green charts, the effect of light varied between charts from different printers: for one printer the average difference in pixel values of best matches was 29, for another it was only 5. That the effects of lighting were modest was expected because humans possess colour constancy, where the visual system is capable of maintaining a constant appearance of colour quasi-independently of changes in the irradiant light. Presumably many other animals also show colour constancy [33-37]. However, the adapting mechanisms are not perfect [38-40] and changes in the irradiant light do have some effect on colour constancy. Different environments can vary significantly in their irradiant light characteristics [41], and thus influence colour appearance.
Whilst in a natural situation, the perception of colours under different conditions will be about the same (i.e. a light red signal will always look light red), in terms of quantitative, or even qualitative scientific experiments, changes in perception may significantly impact upon results, sometimes by large degrees. Results in the field will be affected by the time of day, the weather, and the natural environment [6,41], and will also differ under various laboratory lighting conditions. Finally, charts based on RGB colour space, even if used in studies of human vision, are incapable of reproducing the full range of colours perceptible to humans. The use of charts based on CIE data to estimate colours are better than the use of charts based on RGB colour space, but still unfortunately are based upon human subjective assessment.
The spectrophotometry analysis further supports the argument that printed charts could produce seriously inaccurate colour matching results. Firstly, different printers vary significantly in the reflectance properties of the colour charts that they produce. This means that comparisons between different printers will be unreliable, even discounting the effects of toner level. More seriously however, is the result that some printers do not show even gradations in reflectance between colour blocks with an even spacing in RGB colour space. Therefore, comparing the colours of animal signals between individuals (for example) via charts to obtain an R, G or B values could be seriously flawed – made worse when considering the non-linearity of RGB space in terms of visual perception.
Finally, as stated above, there are crucial differences between the visual perceptions of humans and non-human animals. The perception of a given colour signal to a human may be markedly different from that of the animal towards which the signal is directed. The fact that colour charts have numerous errors associated with them, especially self-made charts, in terms of human judgement, only emphasises the inadequacy of the method when used with respect to non-human animals. The fact remains that other animals' perceptions of a signal may drastically differ from our own.
Conclusion
The inadequacy of colour charts as a means to estimate the colour of animal signals is not a new topic, yet too often researchers outside of the technical colour sciences have adopted this procedure, despite the serious implications of doing so. Theoretically, the use of colour charts is poor practice when considering signals aimed at non-human animals, since these will often have significantly different visual perceptions. Also, some colour spaces on which colour charts are based are not linear in the perceptual differences between one point in space and another, even for humans. This is the case for the RGB/HTML colour space used to create charts by Berggren & Merilä [1]. Even charts that are uniform in colour space are far from perfect and an active area of research in the human vision sciences. Our results cast further doubt on the use of charts to estimate colour, in that individual people vary in their colour matching choices, and that the light environment also can affect colour perception, albeit to a far smaller degree than printer variation. Matching between different people will be variable and error prone and, even if the experiments are all performed by the same individual, their perceptions can also change based on the environmental conditions. In the case of 'self-produced' printed colour charts, different charts vary in their properties, and the same printers may not produce equal steps between colour blocks on a chart, even if that is the case on a computer.
In some instances, access to expensive equipment such as spectrophotometers and calibrated digital cameras may be difficult. In this case, the use of a colour chart may be the only option, and is certainly better than an abstract description of an observed colour. However, we urge caution with the use of colour charts, and advocate that they are used only as a last resort. We would also not recommend that anyone produces self-made charts based on empirically or perceptually non-uniform colour spaces, and are extremely dubious of results obtained in this way. If colour charts are to be used, we recommend the use of a well studied, perceptually uniform colour space, such as CIELAB, with colour matching experiments undertaken by the same individual in as carefully controlled conditions as is possible.
Methods
Whilst theoretical arguments indicate why the use of colour charts, in particular those based on RGB/HTML colour space, are a poor method to estimate the chromatic components of animal signals, we wished also to provide quantitative evidence. Our experiments aimed to show that there may be at least three problems with using humans to match the colour of an object to a set of charts, even discounting differences between human vision and that of other species. Firstly, perception of colour may vary between individuals. Secondly, the exact reproduction of a colour chart will vary depending on the printer from which the charts are produced (not to mention the toner levels and paper type). Thirdly, the light environment may also affect colour matching results.
Colour Matching Experiments
We designed colour charts in Jasc Paint Shop Pro® based on RGB values, consisting of coloured rectangles 68 mm by 20 mm in size, with 12 different rectangles per sheet, inserted into a Microsoft Word® file for easy printing. Colour charts were of two types, red or blue-green, with RGB values ranging from 0,0,0 to 250,0,0 for the red charts, and 0,0,0 to 0,250,250 for the blue-green charts. Copies of each chart type were printed from eight different printer types. The quality of the printers ranged from relatively inexpensive office ink jet types, to high quality laser printers used by the University of Bristol Print Services (HP Colour LaserJet 2500, HP InkJet Combi, HP DeskJet 1220c, HP DeskJet 6127, Epson Stylus Photo 915, Epson Stylus Colour 760, Cannon LaserJet 2100, Cannon LaserJet 5100).
The experiment was a repeated-measures design, with each of the 11 subjects (normally sighted according to self-report) asked to match a colour sample to what they perceived to be the most similar colour on each of the printed colour charts, under each of four different lighting conditions. The order of testing was randomised across subjects, with the authors blind to which colour match was optimal, and the subjects blind to the experimental aims. For each chart and lighting condition, a different colour sample was selected at random from an envelope. In fact, to simplify the subsequent analysis, within the two categories of colour stimuli (red and blue-green), all the samples to be matched were nominally identical, but subjects were unaware of this. The samples were obtained from paint charts (Dulux, Slough, UK, 'Spring 04 colour card') and their similarity verified using spectrophotometry (see below). The pretence of random selection from an apparently large set of samples was introduced to discount the possibility that subjects would recognise the same card, and bias their choices to the same match with each choice made.
Colour matching experiments were performed under four different light conditions: a 150 Watt Xenon arc lamp (Light Support, Berkshire, UK), a 20 Watt desktop incandescent lamp (Philips, PL-Electronic-T), general laboratory fluorescent lighting (Sylvania T5 FHE, Raunheim, Germany), and outside skylight. Incandescent lights contain filaments heated to high temperatures, and typically emit light richer in longer wavelengths (giving a reddish tinge) [30]. The xenon arc lamp tends to have an output richer in short wavelengths. The outside conditions used to test people were under sunshine and cloud, but avoiding direct sunlight, and would tend to be white-ish [41]. The order that each colour chart (from the different printers) was presented, and the order of light conditions under which the charts were viewed, was randomised for each subject so that the possibility of any biases developed towards specific colour patches were controlled for.
Spectrophotometry of Colour Charts
We aimed to quantify the properties of each colour rectangle on each of the colour charts from different printers via spectrophotometry. These results would also show how much variation exists between the different printed charts, which, in theory, should all show the same reflectance spectra. Reflectance measurements of each colour block on each chart was undertaken with a Zeiss MCS 230 diode array photometer, with illumination by a Zeiss CLX 111 Xenon lamp (Carl Zeiss Group, Jena) held at 45° to normal to reduce specular reflection. Measurements were taken normal to the surface, from a 2 mm area, recorded in 1-nm intervals from 300 to 700 nm, and expressed relative to a Spectralon 99% white reflectance standard (Labsphere, Congleton). White standard measurements were taken between measurements of each colour chart, to avoid error associated with drift in the light source and sensor. In all, 10 measurements, in different locations, were taken of each of the 12 colour blocks, on the 8 red and the 8 blue-green charts. Plus, 10 measurements were taken from a random sample of 8 red and 8 blue-green paint cards, giving a total of 2080 reflectance spectra measurements. For each colour block measured, or for the paint cards, the repeated samples were used to produce average spectra.
Authors' contributions
ICC devised the idea of the paper and the general experiments. MS designed the colour charts, and undertook the colour matching experiments. ICC and MS undertook the spectrophotometry measurements. MS and ICC analysed the colour matching, and MS the spectophotometry results. MS & ICC wrote the manuscript.
Acknowledgements
MS was supported by a Biotechnology & Biological Sciences (UK) studentship; further support was provided by BBSRC grants to ICC, Tom Troscianko and Julian Partridge. The comments of three anonymous referees greatly improved the clarity of our paper.
==== Refs
Berggren Å Merilä J WWW design code – a new tool for colour estimation in animal studies Frontiers in Zool 2004 1 2 10.1186/1742-9994-1-2
Dawkins MS Guilford T Design of an intention signal in the bluehead wrasse (Thalassoma bifasciatum) Proc R Soc Lond B 1994 257 123 128
Heinen JT The significance of color-change in newly metamorphosed American toads (Bufo americanus americanus) J Herpetol 1994 28 87 93
Ellis T Howell BR Hughes RN The cryptic responses of hatchery-reared sole to a natural sand substratum J Fish Biol 1997 51 389 401 10.1111/j.1095-8649.1997.tb01674.x
Bortolotti GR Fernie KJ Smits JE Carotenoid concentration and coloration of American kestrels (Falco sparverius) disrupted by experimental exposure to PCBs Funct Ecol 2003 17 651 657 10.1046/j.1365-2435.2003.00778.x
Endler JA On the measurement and classification of colour in studies of animal colour patterns Biol J Linn Soc 1990 41 315 352
Goldsmith TH Optimisation, constraint, and history in the evolution of eyes Q Rev Biol 1990 65 281 322 2146698 10.1086/416840
Bennett ATD Cuthill IC Norris KJ Sexual selection and the mismeasure of color Am Nat 1994 144 848 860 10.1086/285711
Jacobs GH Ultraviolet vision in vertebrates Am Zool 1992 32 544 554
Jacobs GH The distribution and nature of colour vision among the mammals Biol Rev Camb Philos Soc 1993 68 413 471 8347768
Tovée MJ Ultra-violet photoreceptors in the animal kingdom: their distribution and function TREE 1995 10 455 459
Cott HB Adaptive Colouration in Animals 1940 London: Methuen & Co. Ltd
Guilford T Dawkins MS Receiver psychology and the evolution of animal signals Anim Behav 1991 42 1 14
Cuthill IC Bennett ATD Partridge JC Maier EJ Plumage reflectance and the objective assessment of avian sexual dichromatism Am Nat 1999 160 183 200 10.1086/303160
Kelber A Vorobyev M Osorio D Animal colour vision – behavioural tests and physiological concepts Biol Rev Camb Philos Soc 2003 78 81 118 12620062 10.1017/S1464793102005985
Thompson E Palacios A Varela FJ Ways of coloring: comparative color vision as a case study for cognitive science Behav Brain Sci 1992 15 1 74
Huth HH Burkhardt D Der Spektrale Sehbereich eines Violetta Kolibris Naturwissenschaften 1972 59 650 4657435 10.1007/BF00609559
Wright AA The influence of ultraviolet radiation on the pigeon's color discrimination J Exp Anal Behav 1972 17 325 337 5033890
Goldsmith TH Hummingbirds see near ultraviolet light Science 1980 207 786 788 7352290
Chen D Collins JS Goldsmith TH The ultraviolet receptor of bird retinas Science 1984 225 337 340 6740315
Chen D Collins JS Goldsmith TH The ultraviolet receptor of bird retinas Science 1984 225 337 340 6740315
Maier EJ Bowmaker JK Colour vision in the passeriform bird, Leiothrix lutea : correlation of visual pigment absorbency and oil droplet transmission with spectral sensitivity J Comparative Physiol A 1993 172 295 301 10.1007/BF00216611
Hart NS The visual ecology of avian photoreceptors Prog Retin Eye Res 2001 20 675 703 11470455 10.1016/S1350-9462(01)00009-X
Odeen A Hastad O Complex distribution of avian color vision systems revealed by sequencing the SWS1 opsin from total DNA Mol Biol Evol 2003 20 855 861 12716987 10.1093/molbev/msg108
Cuthill IC Partridge JC Bennett ATD Church SC Hart NS Hunt S Ultraviolet vision in birds Adv Stud Behav 2000 29 159 214
Honkavaara J Koivula M Korpimaki E Siitari H Viitala J Ultraviolet vision and foraging in terrestrial vertebrates Oikos 2002 98 505 511 10.1034/j.1600-0706.2002.980315.x
Shi YS Yokoyama S Molecular analysis of the evolutionary significance of ultraviolet vision in vertebrates PNAS 2003 100 8308 8313 12824471 10.1073/pnas.1532535100
Church SC Bennett ATD Cuthill IC Partridge JC van Emden H, Rothschild M Avian ultraviolet vision and its implications for insect protective colouration Insect and Bird Interactions 2004 Andover (Hants, UK): Intercept Ltd 165 184
Wyszecki IG Stiles WS Color science: Concepts and Methods, Quantitative Data and Formulae 1982 2 New York: John Wiley & Sons
Forsyth DA Ponce J Computer Vision: A Modern Approach Pearson Educational International 2003
Westland S Ripamonti C Computational Colour Science Using MATLAB 2004 John Wiley & Sons. Ltd
Munsell Color Company Munsell Book of Color Glossy Finish Collection 1976 2 Munsell/Macbeth/Kollmorgen Corporation, Baltimore
Dyer AG Broad spectral sensitivities in the honeybee's photoreceptors limit colour constancy J Comp Physiol A 1999 185 445 453 10.1007/s003590050405
Dorr S Neumeyer C Color constancy in goldfish: the limits J Comp Physiol A 2000 186 885 896 11085641 10.1007/s003590000141
Vorobyev M Marshall J Osorio D de Ibarra NH Menzel R Colourful objects through animal eyes Color Res Appl 2001 26 S214 S217 10.1002/1520-6378(2001)26:1+<::AID-COL45>3.0.CO;2-A
Neumeyer C Dorr S Fritsch J Kardelky C Colour constancy in goldfish and man: influence of surround side and lightness Perception 2002 31 171 187 11922131 10.1068/p05sp
Lotto RB Chittka L Seeing the light: Illumination as a contextual cue to color choice behavior in bumblebees PNAS 2005 102 3852 3856 15731346 10.1073/pnas.0500681102
Arend LE Reeves A Simultaneous color constancy J Opt Soc Am 1986 3 1743 1751
Troost J de Weert C Naming versus matching in color constancy Percept and Psychophys 1991 50 591 602
Lucassen M Walraven J Quantifying color constancy – evidence for non-linear processing of cone-specific contrast Vis Res 1993 33 739 757 8351846 10.1016/0042-6989(93)90194-2
Endler JA The color of light in forests and its implications Ecol Monogr 1993 63 1 27
|
16131394
|
PMC1201160
|
CC BY
|
2021-01-04 16:38:34
|
no
|
Front Zool. 2005 Aug 30; 2:14
|
utf-8
|
Front Zool
| 2,005 |
10.1186/1742-9994-2-14
|
oa_comm
|
==== Front
Health Qual Life OutcomesHealth and Quality of Life Outcomes1477-7525BioMed Central London 1477-7525-3-481608351210.1186/1477-7525-3-48ResearchUsefulness of five-item and three-item Mental Health Inventories to screen for depressive symptoms in the general population of Japan Yamazaki Shin [email protected] Shunichi [email protected] Joseph [email protected] Epidemiology and Exposure Assessment Section, National Institute for Environmental Studies, Tsukuba, Japan2 Department of Epidemiology and Healthcare Research, Graduate School of Medicine and Public Health, Kyoto University, Kyoto, Japan3 Office of International Academic Affairs, Graduate School of Medicine, University of Tokyo, Japan2005 8 8 2005 3 48 48 2 6 2005 8 8 2005 Copyright © 2005 Yamazaki et al; licensee BioMed Central Ltd.2005Yamazaki 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 five-question Mental Health Inventory (MHI-5) is a brief questionnaire that can be used to screen for depressive symptoms. Removing the 2 anxiety-related items from the MHI-5 yields the MHI-3. We assessed the performance of the Japanese versions of the MHI-5 and MHI-3 in detecting depressive symptoms in the general population of Japan.
Methods
From the population of Japan, 4500 people 16 years old or older were selected by stratified-random sampling. The Medical Outcomes Study 36-Item Short Form Health Survey (SF-36, which includes the MHI-5) and the Zung Self-rating Depression Scale (ZSDS) were included in a self-administered questionnaire. ZSDS scores of 48 and above were taken to indicate the presence of moderate or severe depressive symptoms, and scores of 56 and above were taken to indicate the presence of severe depressive symptoms. We computed the correlation coefficient between the ZSDS score and the scores on the MHI-5 and MHI-3. We also computed the sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve.
Results
Of the 3107 subjects (69% of the 4500 initially selected), 14.0% had moderate or severe depressive symptoms, and 2.0% had severe depressive symptoms as measured with the ZSDS. The correlations of ZSDS scores with MHI-5 scores and with MHI-3 scores were similar: -0.63 and -0.61, respectively. These correlation coefficients were almost the same whether or not the data were stratified by age and sex. For detecting severe depressive symptoms with the MHI-5, the area under the ROC curve was 0.942 (95%CI: 0.919 – 0.965); for the MHI-3, it was 0.933 (95%CI: 0.904 – 0.962).
Conclusion
The MHI-5 and MHI-3 scores were correlated with the ZSDS score, and can be used to identify people with depressive symptoms in the general population of Japan.
==== Body
Background
Depression disorders are a major health problem in Japan. Depressive mood is associated with suicide in middle-aged workers [1], and the number of suicides has increased as economic conditions have worsened since 1998 [2]. Nonetheless, there are few studies of the prevalence of depression or of depressive symptoms in communities in Japan [3,4].
To assist in detecting depression or depressive symptoms, many screening questionnaires have been developed. Some of these have 20 to 30 items, take only a few minutes to complete, use the number of symptoms as the score, and have good performance to detect depressive state. Instruments that are even shorter but nonetheless have good performance to detect depressive state have also been developed [5-7]. One such questionnaire is the five-item version of the Mental Health Inventory (MHI-5) [6,7]. The MHI-5 is used as the "Mental Health" domain of the Medical Outcomes Study 36-Item Short Form Health Survey (SF-36). The SF-36 has been translated into Japanese [8], and the Japanese version has been validated for use in the general population of Japan [9], but the performance of the MHI-5 has not been evaluated in detail. In addition, two of the items in the MHI-5 are almost identical to two items in a scale developed to measure anxiety [10]. We hypothesized that removing those two anxiety-related items would result in a scale (the MHI-3) that performs as well as the MHI-5 in detecting symptoms of depression.
In this study, we compared the Japanese version of the MHI-5 and MHI-3 to the 20-item Zung Self-rating Depression Scale (ZSDS) [11], and assessed the performance of the Japanese versions of the MHI-5 and MHI-3 in detecting depressive symptoms among the general population.
Methods
Setting and participants
We used data that had been collected previously for a study of the validity of the Japanese version of the SF-36, and calculated national norm scores of all subscales of the SF-36 [8,9]. Details of the nationwide survey have been described previously [9]. Briefly, a total of 4500 people 16 years old or older were selected from the entire population of Japan by stratified-random sampling in 1995. A self-administered questionnaire was mailed, and the subjects were visited to collect the questionnaires. The SF-36, the ZSDS [11] (described below), and questions about demographic characteristics were included in the questionnaire.
The ZSDS consists of 10 positively worded items and 10 negatively worded items asking about symptoms of depression. Several studies have established the ZSDS as a reliable and valid instrument for measuring depressive symptoms [12-14]. The ZSDS scores were used to define four categories of the severity of depression: within normal range or no significant psychopathology (below 40 points); presence of minimal to mild depression (40–47 points); moderate to marked depression (48–55 points); presence of severe to extreme depression (56 points and above). These score ranges result from the studies of Zung [15] and Barrett et al [16]. The ZSDS has been translated into Japanese and studies of the validity of the Japanese version have been published [17]. Because the ZSDS is not a clinical diagnostic tool, subjects with high scores are said to have depressive symptoms rather than "depression."
Like the rest of the SF-36, the MHI-5 was administered as a paper-and-pencil questionnaire. The instrument contains the following questions: 'How much of the time during the last month have you: (i) been a very nervous person?; (ii) felt downhearted and blue?; (iii) felt calm and peaceful?; (iv) felt so down in the dumps that nothing could cheer you up?; and (v) been a happy person?' For each question the subjects were asked to choose one of the following responses: all of the time (1 point), most of the time (2 points), a good bit of the time (3 points), some of the time (4 points), a little of the time (5 points), or none of the time (6 points). Because items (iii) and (v) ask about positive feelings, their scoring was reversed. The score for the MHI-5 was computed by summing the scores of each question item and then transforming the raw scores to a 0–100-point scale [18].
Items (i) and (iii) are almost identical to 2 items in the Zung Self-rating Anxiety Scale [10]. To make a scale that is even shorter than the MHI-5 and is focused on depression we removed those two anxiety-related items. Thus, the MHI-3 comprised only (ii), (iv), and (v) above. Possible scores on the MHI-3 ranged from 3 to 18 points.
Statistical methods
First, we computed the correlation coefficient (Pearson's) between the ZSDS scores and the scores on the MHI-5 and the MHI-3. We computed the sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve. Analysis of ROC curves has been described in detail and ROC analysis is used extensively in health-related diagnostics [19,20]. ROC analysis can be used to study the performance of diagnostic or screening tests across a wide range of sensitivities and specificities. For example, it can be used to compute the sensitivity (the true-positive rate) and specificity (the true-negative rate) for any specified test score. The area under the ROC curve (AUC) is an index of the amount of information the test provides over its entire scoring range [21,22]. In general, an AUC can range from 0.5, which indicates a test with no information, to 1.0, which indicates a perfect test. The "gold standard" criteria for diagnosing depression are considered to be those of the Diagnostic and Statistical Manual of Mental Disorders (DSM) [7]. In this study, because we could not interview all subjects, we used, instead, scores on the ZSDS. For each of the three categories of the severity of depressive states (ZSDS scores of 40 or higher), we computed the AUC of each of the five items, the MHI-5, and the MHI-3. To define the cut-off points, we first considered each of the actually measured MHI-5 scores as a possible cut-off point. For each score, we took the sum of the sensitivity and the specificity. The score with the highest sum was used as the cut-off point. One cut-off point was determined for each of the three levels of severity defined by ZSDS scores (mild, moderate, and severe).
Results
The nationwide survey targeted 4500 people, and 3395 (male: 1704; female: 1691) responded to the questionnaire (75% response rate). Of these 3395 individuals, 3107 (male: 1573; female: 1534) completed all of the items on the ZSDS. The mean score on the MHI-5 was 72.8 (SD = 19.1). The mean scores on the MHI-5 for respondents of different demographic categories are shown in Table 1. These mean scores ranged from 68.5 to 76.6. Almost 23% of the respondents had ZSDS scores indicating mild depressive symptoms, 12% had scores indicating moderate depressive symptoms, and 2% had scores indicating severe depressive symptoms.
Table 1 MHI-5 scores by demographic categories
N (%) Score of the MHI-5
3107 (100) Mean (SD)
Sex
Male 1573 (51) 73.31 (18.63)
Female 1534 (49) 72.32 (19.55)
Age (years)
<30 619 (20) 70.17 (18.47)
30 – 39 506 (16) 72.50 (17.47)
40 – 49 665 (21) 72.38 (20.28)
50 – 59 617 (20) 74.22 (18.60)
60 – 69 479 (15) 75.21 (19.11)
≥70 221 (7) 73.23 (21.09)
Annual household income (million yen)
<3 385 (12) 69.37 (20.83)
3 – 4.9 670 (22) 71.87 (19.08)
5 – 6.9 685 (22) 72.62 (19.38)
7 – 9.9 648 (21) 73.72 (18.27)
10 – 11.9 228 (7) 74.57 (18.78)
≥12 266 (9) 76.63 (16.72)
Missing values 225 (7) 73.30 (19.4)
Schooling
Junior high school 613 (20) 72.64 (19.88)
High school 1426 (46) 72.97 (18.88)
Junior college, college, or higher 1028 (33) 72.84 (18.85)
Missing values 40 (1) 69.95 (20.88)
Marital status
Single 622 (20) 70.04 (18.89)
Married 2227 (72) 73.74 (18.8)
Separated 28 (1) 75.43 (16.86)
Divorced 65 (2) 68.66 (20.91)
Widowed 152 (5) 72.18 (22.27)
Missing values 13 (0) 70.00 (19.71)
Occupational status
Full time worker 1610 (52) 73.05 (18.25)
Part time worker 299 (10) 74.27 (17.96)
Retired 164 (5) 72.51 (22.74)
Unemployed 171 (6) 69.34 (21.95)
Homemaker 533 (17) 73.06 (19.71)
Student 226 (7) 73.36 (18.77)
Other 83 (3) 68.48 (21.49)
Missing values 21 (1) 70.86 (16.64)
The correlations of ZSDS scores with MHI-5 scores and with MHI-3 scores were similar: -0.63 and -0.61, respectively. These correlation coefficients were almost the same whether or not the data were stratified by age and sex (Table 2).
Table 2 Correlations of ZSDS scores with MHI-5 and MHI-3 scores, by demographic category
MHI-5 MHI-3
All -0.634 -0.614
Sex
Male -0.634 -0.610
Female -0.635 -0.618
Age (years)
<30 -0.653 -0.643
30 – 39 -0.686 -0.685
40 – 49 -0.619 -0.591
50 – 59 -0.576 -0.549
60 – 69 -0.635 -0.608
≥70 -0.698 -0.671
Annual household income (million yen)
<3 -0.666 -0.638
3 – 4.9 -0.612 -0.596
5 – 6.9 -0.642 -0.642
7 – 9.9 -0.637 -0.602
10 – 11.9 -0.654 -0.642
≥12 -0.562 -0.554
Missing values -0.613 -0.551
Schooling
Junior high school -0.612 -0.579
High school -0.637 -0.617
Junior college, college, or higher -0.651 -0.636
Missing values . .
Marital status
Single -0.661 -0.638
Married -0.624 -0.602
Separated . .
Divorced . .
Widowed -0.658 -0.642
Missing values . .
Occupational status
Full time worker -0.618 -0.601
Part time worker -0.533 -0.509
Retired -0.741 -0.711
Unemployed -0.714 -0.692
Homemaker -0.646 -0.636
Student -0.680 -0.646
Other . .
Missing values . .
With ZSDS scores as the basis for classifying depressive symptoms, ROC analysis allowed us to evaluate the performance of the MHI-5 and the MHI-3. The AUC values are shown in Table 3, and other performance characteristics are shown in Table 4. We also evaluated the performance of each of the MHI-5 question items individually (Table 3). For the individual items, the range of "cut-off scores" was determined by the range of each question's response options: from "none of the time" to "all of the time." The best-performing item for detecting severe depressive symptoms was the one asking about the frequency of "feeling downhearted and blue". That item had a sensitivity of 0.88 and a specificity of 0.77 (based on a score of 4 points or less). The AUC of the MHI-3 was only slightly lower than that of the MHI-5 (Figure 1).
Table 3 ROC analysis of individual MHI-5 items, the whole MHI-5, and the MHI-3, by severity of depressive symptoms
Items and scales Severity of depressive symptom (range of ZSDS scores)
Mild, moderate, or severe (40 through 80) Either moderate or severe (48 through 80) Severe (56 through 80)
AUC (95% CI) AUC (95% CI) AUC (95% CI)
(i) Nervous person 0.696 (0.677–0.716) 0.707 (0.680–0.734) 0.826 (0.774–0.879)
(ii) Down in the dumps 0.713 (0.694–0.733) 0.741 (0.714–0.769) 0.862 (0.813–0.910)
(iii) Calm and peaceful 0.745 (0.726–0.764) 0.755 (0.728–0.782) 0.845 (0.797–0.892)
(iv) Downhearted and blue 0.739 (0.720–0.758) 0.748 (0.721–0.776) 0.898 (0.855–0.941)
(v) Happy person 0.747 (0.729–0.765) 0.738 (0.711–0.765) 0.858 (0.811–0.905)
MHI-5* 0.810 (0.793–0.826) 0.819 (0.795–0.843) 0.942 (0.919–0.965)
MHI-3† 0.800 (0.783–0.817) 0.803 (0.779–0.828) 0.933 (0.904–0.962)
Values shown are areas under the ROC curves (AUC), and their 95% CIs, for three levels of depressive symptoms as measured by ZSDS scores. *The MHI-5 includes all 5 items. †The MHI-3 includes only items ii, iv, and v.
Table 4 Performance of the MHI-5 and MHI-3 for detecting depressive symptoms
Mild, moderate, or severe depressive symptoms (ZSDS scores of 40 or higher) Moderate or severe depressive symptoms (ZSDS scores of 48 or higher) Severe depressive symptoms (ZSDS scores of 56 or higher)
Prevalence 37% 14% 2%
Instrument MHI-5 MHI-3 MHI-5 MHI-3 MHI-5 MHI-3
(cut-off score) (68) (14) (60) (13) (52) (11)
Sensitivity 71.5% 76.4% 74.7% 77.1% 91.8% 90.0%
Specificity 79.1% 71.1% 80.0% 71.8% 84.6% 84.2%
Positive predictive value 66.7% 60.8% 37.1% 30.8% 10.8% 10.4%
Negative predictive value 82.5% 83.7% 95.1% 95.1% 99.8% 99.8%
Figure 1 ROC curves of the MHI-5 and MHI-3 for detecting severe depressive symptoms (ZSDS above 55).
Using the MHI-5, the prevalence of severe depressive symptoms (cut-off: 52 points) was 17%, that of moderate or severe depressive symptoms (cut-off: 60 points) was 28%, and that of mild, moderate, or severe depressive symptoms (cut-off: 68 points) was 40%.
Discussion
These data show that the MHI-5 and MHI-3 scores were each correlated with the ZSDS score and had good screening accordance with the ZSDS in the general population of Japan. We also found that the MHI-3 performs almost as well as the MHI-5. The best-performing single item was the one asking about "feeling downhearted and blue," which was also the case in the US [6]. The usefulness of the MHI-5 is consistent with results of a study done in the US [6]. Each scale and each item performed best as a detector of severe depressive symptoms, but each also contributed some information even for detecting moderate and mild depressive symptoms (Table 3). Both scales performed better than did any item alone.
Because prevalence affects positive predictive value, the latter was lowest for severe depressive symptoms and was highest for mild, moderate, and severe depressive symptoms (Table 4). For all levels of symptom severity, the positive predictive values of the MHI-3 were similar to those of the MHI-5, and for severe depressive symptoms they were nearly identical (10.8% and 10.4%) (Table 4).
A previous study showed that the prevalence of mood disorders (major depression, bipolar disorders, and dysthymia) as measured using the DSM criteria in Japanese people 20 years old and older was 3.1% [4]. On the other hand, 37% of the sample in the present study had mild, moderate, or severe depressive symptoms as measured using the ZSDS. People in whom depression is diagnosed using the DSM criteria are probably only a small number of those who report at least some depressive symptoms. In a previous study that also used the ZSDS, the prevalence of mild depressive symptoms among Japanese male workers was 45% [23], which is similar to that in our study.
In addition to its performance as shown in the present ROC analysis, an advantage of the MHI-5 may be the fact that it is part of the SF-36. The reason is that the possibility of a Hawthorne-type effect (i.e. an effect on study participants that results from their knowing that they are being studied) can be an obstacle to screening for depressive state. Specifically, the subjects' responses on a mental-health screening instrument may be affected by their knowledge that they are subjects in a study of mental health. Embedding the mental-health screening instrument in a more general survey, as the MHI-5 is embedded in the SF-36, could help minimize any such effect.
While the results of this study may be useful for public-health purposes, surveys done in primary-care settings could provide information that is more directly applicable to clinical work. Also, it should be kept in mind that ZSDS scores alone cannot be used to diagnose clinical depression. Studies using psychiatrist-diagnosed depression in addition to ZSDS scores would provide further information about the utility of the Japanese version of the MHI-5.
Another limitation is that the data set was obtained from a 1995 survey. Further studies are needed to confirm the performance of the MHI-5 and MHI-3 using data obtained in recent years.
In conclusion, the MHI-5 and MHI-3 scores were correlated with the ZSDS score, and can be used to identify people with depressive symptoms in the general population of Japan.
List of abbreviations
AUC: area under the ROC curve; MHI-5: the five-item version of the Mental Health Inventory; MHI-3: those 3 of the MHI-5 questions that were thought to be most directly related to depression; ROC: receiver operating characteristic; SF-36: the Medical Outcomes Study 36-Item Short Form Health Survey; ZSDS: the Zung Self-rating Depression Scale.
Authors' contributions
SY: analysis of the data, interpretation of results, manuscript writing; SF: initiation and study design, supervision, collection of data; JG: supervision, interpretation of results, manuscript writing.
==== Refs
Tamakoshi A Ohno Y Aoki K Hamajima N Wada M Kawamura T Wakai K Lin YS Depressive mood and suicide among middle-aged workers: findings from a prospective cohort study in Nagoya, Japan J Epidemiol 2000 10 173 178 10860301
Lamar J Suicides in Japan reach a record high BMJ 2000 321 528 10968804 10.1136/bmj.321.7260.528
Kawakami N Shimizu H Haratani T Iwata N Kitamura T Lifetime and 6-month prevalence of DSM-III-R psychiatric disorders in an urban community in Japan Psychiatry Res 2004 121 293 301 14675748 10.1016/S0165-1781(03)00239-7
The WHO World Mental Health Survey Consortium Prevalence, severity, and unmet need for treatment of mental disorders in the World Health Organization World Mental Health Surveys JAMA 2004 291 2581 2590 15173149 10.1001/jama.291.21.2581
Whooley MA Avins AL Miranda J Browner WS Case-finding instruments for depression. Two questions are as good as many J Gen Intern Med 1997 12 439 445 9229283 10.1046/j.1525-1497.1997.00076.x
Berwick DM Murphy JM Goldman PA Ware JE JrBarsky AJ Weinstein MC Performance of a five-item mental health screening test Med Care 1991 29 169 176 1994148
Rumpf HJ Meyer C Hapke U John U Screening for mental health: validity of the MHI-5 using DSM-IV Axis I psychiatric disorders as gold standard Psychiatry Res 2001 105 243 253 11814543 10.1016/S0165-1781(01)00329-8
Fukuhara S Bito S Green J Hsiao A Kurokawa K Translation, adaptation, and validation of the SF-36 Health Survey for use in Japan J Clin Epidemiol 1998 51 1037 1044 9817121 10.1016/S0895-4356(98)00095-X
Fukuhara S Ware JE JrKosinski M Wada S Gandek B Psychometric and clinical tests of validity of the Japanese SF-36 Health Survey J Clin Epidemiol 1998 51 1045 1053 9817122 10.1016/S0895-4356(98)00096-1
Zung WW A rating instrument for anxiety disorders Psychosomatics 1971 12 371 379 5172928
Zung WWK A Self-Rating Depression Scale Arch Gen Psychiatry 1965 12 63 70 14221692
Biggs JT Wylie LT Ziegler VE Validity of the Zung Self-rating Depression Scale Br J Psychiatry 1978 32 381 385 638392
Gabrys JB Peters K Reliability, discriminant and predictive validity of the Zung Self-rating Depression Scale Psycholog Reports 1985 57 1091 1096 4095223
Agrell B Dehlin O Comparison of six depressive rating scales in geriatric stroke patients Stroke 1989 20 1990 1994 2772980
Zung WWK From art to science Arch Gen Psychiatry 1973 29 328 337 4724142
Barrett J Hurst MW DiScala C Rose RM Prevalence of depression over a 12-month period in a nonpatient population Arch Gen Psychiatry 1978 35 741 744 306804
Fukuda K Kobayashi S A study on a self-rating depression scale Seishin Shinkeigaku Zasshi 1973 75 673 679 (in Japanese) 4798819
Ware JE Snow KK Kosinski M Gandek B SF-36 health survey manual & interpretation guide 1993 Boston, New England Medical Center
Metz CE Basic principles of ROC analysis Semin Nucl Med 1978 8 283 298 112681
Swets JA Pickett RM Whitehead SF Getty DJ Schnur JA Swets JB Freeman BA Assessment of diagnostic technologies Science 1979 205 753 759 462188
Hanley JA McNeil BJ The meaning and use of the area under a receiver operating characteristic (ROC) curve Radiology 1982 143 29 36 7063747
Hanley JA McNeil BJ A method of comparing the areas under receiver operating characteristic curves derived from the same cases Radiology 1983 148 839 843 6878708
Kawakami N Takatsuka N Shimizu H Ishibashi H Depressive symptoms and occurrence of type 2 diabetes among Japanese men Diabetes Care 1999 22 1071 6 10388970
|
16083512
|
PMC1201161
|
CC BY
|
2021-01-04 16:38:13
|
no
|
Health Qual Life Outcomes. 2005 Aug 8; 3:48
|
utf-8
|
Health Qual Life Outcomes
| 2,005 |
10.1186/1477-7525-3-48
|
oa_comm
|
==== Front
Hum Resour HealthHuman Resources for Health1478-4491BioMed Central London 1478-4491-3-81610917410.1186/1478-4491-3-8ReviewContemporary specificities of labour in the health care sector: introductory notes for discussion Campos Francisco Eduardo [email protected] Eduardo da Motta e [email protected] Center of Study of Collective Health, School of Medicine, Federal University of Minas Gerais (NESCON, UFMG), Minas Gerais, Brazil2 Center of Regional Development and Planning, Department of Economics, Federal University of Minas Gerais (CEDEPLAR-UFMG), Minas Gerais, Brazil2005 18 8 2005 3 8 8 20 10 2004 18 8 2005 Copyright © 2005 Campos and Albuquerque; licensee BioMed Central Ltd.2005Campos and Albuquerque; 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
This paper combines the literature on public health, on economics of health and on economics of technological innovation to discuss the peculiarities of labour in the health care sector.
Method and framework
The starting point is the investigation of the economic peculiarities of medical care.
Results and discussions
This investigation leads to the identification of the prevalence of non-market forms of medical care in the countries of the Organisation for Economic Co-operation and Development (OECD). Furthermore, the health care system has a distinctive characteristic from other economic sectors: it is the intersection between social welfare and innovation systems. The relationship between technological innovation and cost in the health care sector is surveyed. Finally, the Brazilian case is discussed as an example of a developing country.
Conclusion
The peculiarities of labour in the health care sector suggest the need to recognize the worth of sectoral labour and to cease to treat it separately. This process should take into account the rapid development of the health innovation system and one important consequence: the obsolescence of the acquired knowledge. One way to dignify labour is to implement continued education and training of health professions personnel.
==== Body
Background
Labour in the health care sector has specific characteristics that are evidenced by its institutional organization. The special economic properties of medical care determine the generalized emergence of what is called "market failures" in the economic literature. That is, the operation of market forces alone is not sufficient for the working of this sector, as is recognized in a recent report of the World Bank [1]. Society constructs varied institutional forms in order to offset market flaws by assigning an essential role to non-market institutions to render adequate services. The welfare state institutions may be viewed as an expression of social attempts to offset generalized market failures in the health care sector.
The next section of this paper presents the method of investigating the economic peculiarities of medical care as the starting point of this paper. The economic literature extensively discusses this subject, and the theoretical basis for social welfare institutions is well-grounded. This literature shows how superficial and theoretically poor are those approaches that wholly or predominantly insist on the market role for the operation of the health care sector. Kenneth Arrow's (Nobel economics prize, 1972) seminal contribution is a powerful remedy for such superficiality.
Arrow's analysis [2] is a conducting line for this paper. Arrow highlights that "the special economic problems of medical care can be explained as adaptations to the existence of uncertainty in the incidence of disease and in the efficacy of treatment". The weight of uncertainty and expressive information asymmetries determine the emergence of market failures, and hence the need of institutions to deal with such activities.
Based on these theoretical topics, the results of this approach are presented in the third section, which points to the prevalence of non-market forms of medical care organizations – an empirical confirmation of Arrow's analysis according to Barr [3]. Another specificity of labour in this sector is derived from this topic: the way these services are articulated (existing institutions, regulation and competition pressure) may affect both the quality of medical care and the pace of scientific research (and future scientific-technological progress as well).
The fourth section discusses a very distinctive characteristic of health care: its location in the intersection of two constituting institutional arrangements in advanced capitalist societies – the social welfare system is connected to the national innovation system. In other words, the professionals' performance in the sector affects – and is strongly affected by – the pace of scientific advancement and technological innovations. To simplify, one could say that a hospital is part of both systems [4].
In the fifth section, an assessment of this articulation between both systems introduces a discussion on cost peculiarities and those of technological progress in the health care sector.
The sixth section discusses the specific characteristics of the Brazilian situation. The precarious and rudimentary character of the social welfare institutions, the serious problems of access to health services and the general determinants of health conditions in Brazil, in addition to the incipient Brazilian innovation system (including that in health matters), simply add new problems to Arrow's list. Problems of resource allocation are crucial for defining the profile of the systems being settled, which amounts to one more labour peculiarity in the sector: that is, the involvement of professionals in the area in defining such a profile is not a trivial issue.
By articulating all questions so far developed, the seventh section concludes the paper by assessing the crucial role of recognizing the worth of labour in the health care sector and of ending its isolation.
Method and framework
The starting point of this paper is the analysis of the economic peculiarities of medical care. Medical practice is full of lessons on the special character of medical care. Any physician or health care manager is able to describe a set of properties distinguishing the activity of a health professional from those of other workers in more conventional activities.
Perhaps one of the most often perceived differences in the sector is the lack of consumers' ability to choose their own basket of goods and services due to their lack of information for decision-making. It is useless to ask a patient whether he/she would prefer chemotherapy instead of radiotherapy in case he/she is able to afford only one of the alternatives. By the same token, it would also be useless to ask the patient whether he/she prefers an immunological examination or a magnetic resonance scan.
Such a situation is aggravated because the decision must be made at a time of personal or family distress – an illness threatens to take the patient's life or that of a loved one. For this reason, in contrast to other items whose consumption can be postponed, the consumer will make a heroic effort and will certainly not hesitate to try all available alternatives. This breaks one of the basic rules for adequate market resource allocation, as there is no symmetry of information. One side – the service provider – supposedly holds the information by accumulating esoteric knowledge [5] that is inaccessible to the other side.
Another important difference resides in the existence of limits to "rationalizing the production" as in other economic sectors. Every emergency service must always have a neurosurgeon available, even if traumas requiring nocturnal procedures rarely occur. It would be inadmissible to deny treatment to a person with multiple traumas on the basis of statistical evidence that his or her condition accounts for a small incidence of traumas and that it would not be economically justified to keep staff for this purpose. As a further example, even though snakebite is becoming increasingly rare, every health care unit must keep antivenin in stock – duly cooled and periodically checked – most of which is discarded.
Possible functions of standardized production are one more example of major differences. In industry, such possibilities stem from a relative standardization and monotonous industrial processes – the inputs are constant, the processes are repetitive and the outcome is always expected and predicted. In the health sector things are not as simple, as inputs and processes cannot be fully standardized.
An extensive literature shows that distinct health agents or even distinct medical "cultures" come to entirely different diagnoses regarding similar groups of patients submitted to them and prescribe utterly different therapies. According to the patient's social, economic and cultural level, the same disease may be given a totally different treatment. A verminosis, for example, may be treated with broad-spectrum drugs instead of a simple faeces examination or it could be treated using a series of examinations that could lead to a similar therapy (broad-spectrum drugs).
All this happens because there is much subjectivity in the health labour process, which remains basically artisanal, grounded on sight, touch and smell instead of being readily captured in a defined algorithm. Although a great deal of information may be objective – blood pressure, coronary permeability, electrocardiographic waves – other data or even the interpretation of those considered objective are quite subjective. There is always a feeling, a clinical look, based on subjective hints, that may trigger a dozen supplementary examinations. In addition, there is a sort of association between the physician and the patient that has been ontologically constructed during human history, ever since someone's sufferings were first alleviated by someone else.
Hindrances to standardization, however, should be qualified. In some health areas, there are – within certain limits – standard procedures, such as: laboratory procedures; highly standardized procedures in hospital sectors, resulting in serial surgeries; in public health, a number of advancements were achieved by means of standard protocols (e.g. for the treatment of diarrhoea and acute respiratory infections); medical procedures can be partially standardized through detailed classification (as for example, the Diagnosis Related Groups – DRGs).
Such elements, however, encounter a significant limit – the clinical contact controls all other procedures. And the clinical contact is based on much more shifting variables in which there are imponderables and from which it is difficult to construct closed algorithms.
Such observations constitute a source of empirical elements for a substantive theoretical elaboration of the peculiar characteristics of medical care as an economic category. Such peculiarities thus present a series of limitations to the market's ability to provide such services in a quantitatively and qualitatively adequate manner. Arrow had the merit to present this discussion based on an elaborate economic conception.
Arrow's paper [2] is pedagogically structured. Firstly, the market's working is described in accordance with the neoclassical economic theory that it should lead to the occurrence of a competitive equilibrium and optimum status. Then, the author poses hindrances to medical care marketability. The first basic difference from common commodities is related to the risk-bearing associated with medical care – "to a great extent a disease is an unpredictable phenomenon". A subtle outcome arises: "When there is uncertainty, information or knowledge becomes a commodity ..." But information, in the form of skilled care, is precisely what is being bought from most physicians and, indeed, from most professionals.
The elusive character of information as a commodity suggests that it departs considerably from the usual marketability assumptions about commodities." (p. 183). Thus, he sustains that "all the special features of this industry, in fact, stem from the prevalence of uncertainty". Finally, Arrow considers that "when the market fails to achieve an optimum state, society will, to some extent at least, recognize the gap, and non-market social institutions will arise attempting to bridge it" (p. 184). Therefore, such unique characteristics call for "a special place for medical care in economic analysis" (p. 186).
First, the demand for it is irregular and unpredictable (contrary to the demand for food and clothing, for example). Another important aspect is that the demand for health care is usually associated with an assault on personal integrity. A disease is not just a risk, but a risk associated with a cost per se (reduction or loss of labour capacity, even temporary, obviously affecting one's earning capacity), which is distinct from the specific cost of medical care (p. 187). Furthermore, there is an "opportunity cost": the time lost for earning in the labour market while undergoing treatment.
Second, the physician's behavior cannot be fully known in advance – medical care is one of the activities of which "the product and the activity are identical". In these cases, the commodity purchased cannot be tested before consuming and "there is an element of trust in the relation". The physician's behavior "is assumed to be ruled with concern to the patient's welfare, which is not expected from a salesperson". "In Talcott Parsons's terms, there is a 'collectivity-orientation', which distinguishes medicine and other professions from business, where self-interest on the part of participants is the accepted norm" (p. 187).
Other typical differences from business people would be that: advertisement and price competition are virtually absent among physicians; counseling by doctors concerning other treatment is supposedly given without self-interest; and treatment should be oriented by the needs of the specific case and not restrained by financial considerations (p. 187). Finally, resource allocation in this area is significantly affected by "ethic compulsion" (p. 188).
Third, there is uncertainty concerning the output – recovery from a disease is as unpredictable as its incidence: "Because medical knowledge is so complicated, the information possessed by the physician as to the consequences and possibilities of treatment is necessarily very much greater than that of the patient, or at least so it is believed by both parts. Further, both parts are aware of this informational inequality, and their relation is colored by this knowledge" (p. 190). Information asymmetry proves to have a crucial weight in the physician-patient relationship.
A relevant aspect should be added here: although the physician knows better than the patient, his/her knowledge is still extremely limited, given the extensive ignorance areas of scientific knowledge concerning the working of a human body, aetiology of a number of diseases, etc. Thus, there is a great difference between purchasing a chair from a cabinetmaker and a medical appointment – the cabinetmaker knows how to make the chair that the client desires, but the doctor has every chance of knowing very little about how to treat the patient or even being unable to do so.
Fourth, supply conditions are uncertain – entry into the market is not free, which restrains the assumption of full mobility of the production factors. Doctors must be licensed to provide medical services. Furthermore, medical education costs are high and apparently are only partially incurred by the student (p. 191), which means another dissociation from the requirements for the working of competitive markets, i.e. private benefits granted to students after graduation exceed private costs. Arrow associates the high costs of medical education with the quality requirements imposed by the American Medical Association (AMA) since the Flexner Report (p. 191–192) became known.
Fifth, pricing is uncertain – this topic is not usual in economic texts. There is an extensive price discrimination according to the patient's income, at one extreme reaching zero cost for indigent patients. Price competition is strongly disapproved.
Sixth, there exist indivisibilities – specialists and some sort of equipment constitute significant indivisibilities (p. 194).
As risk (of the disease and its treatment outcome) determines the medical care "market", Arrow calls for the possibility of an insurance market that might be able to organize and distribute such risks. If such a market were possible, the problems identified so far could be solved. However, the analysis of a hypothetical ideal insurance market (pp. 199–207) points to a set of problems:
• uncovered population segments (unemployed population, aged people, chronic disease sufferers, low-income population);
• differentiated risk pooling (if the market is competitive, high-risk individuals will tend to have to pay higher premiums);
• existence of moral hazard, to the extent that individuals covered by the insurance plans would tend to overuse them;
• adverse selection – an aspect pointed out by Akerlof [6], since in case premiums increased so as to cover elderly people, individuals bearing higher risks would be precisely those who would tend to agree to pay for the insurance (therefore, the individuals selected by the insurer would be exactly those with greater health problems: the costlier individuals for the system);
• uninsurable diseases (e.g. AIDS at the epidemic outset);
• existence of interdependent probabilities (when a problem affecting a person reaches other people, as in epidemics) [3];
• high administrative costs (which would serve as an argument in favor of quite generalized plans, particularly the compulsory ones).
Such problems determine the market's incapacity to provide a comprehensive insurance for medical care (p. 210). In a recent interview, Arrow [7] maintained the diagnosis of the 1963 paper, suggesting that funding the system through contributions to a centralized system "can be accomplished in a cheaper way than having many competitive insurance plans". An interesting aspect of this interview is that it confirms the basic elements of a diagnosis accomplished more than 20 years ago.
In the postscript to his original paper [2], Arrow highlights two aspects as follows: the failure of the market in developing policies of insurance against uncertainty has stimulated the emergence of many social institutions; in these institutions, the usual market premises are "contradicted to a certain extent". He notes that this is not an exclusive problem of the medical profession. All through the text, he emphasizes the role of nonprofit institutions in the sector (e.g. p. 191).
Results: the prevalence of non-market forms of medical care in the OECD countries
Barr [3], in his evaluation of the social welfare states, points out that the structures of medical care organizations are internationally more divergent than those providing benefits, the other major function of the welfare state. According to Barr, there are various arrangements that can be grouped into three categories:
• a quasi-actuarial approach (purchase of private insurance by individuals and employees as well as private ownership of "medical factors of production" – as found in the United States);
• social security related to earnings (compulsory and funded by employees and/or employers' contributions sometimes supplemented by taxes, services rendered by a big private sector – Canada – or by a small private sector – Germany); universal medical services (funded by taxes and publicly owned and/or controlled factors of production – New Zealand, Sweden and United Kingdom);
• social service (most countries adopt schemes of this kind).
Another way of evaluating the distinct characteristics of medical care systems is accomplished by the OECD [8]. Its characterization does not contradict that explained by Barr (1992). In a different approach, Esping-Anderson [9] points to three categories of welfare systems: the Nordic, that of continental Europe and the Anglo-Saxon. This agenda contributes to a comparison between Arrow's diagnosis and the reality.
A general scenario can be drawn based on data from the World Bank Report [1], which show the relative weight of government expenditure in capitalist developed countries: 60% of total expenditure, according to the 1990 data. Even in the United States, where the private sector shows the highest participation among advanced countries, government health expenditure reached 6.17% of GDP in 2001, which amounted to 44.6% of total health expenditure (Table 1).
Table 1 Total health care expenditure, relative participation of private sector and public sector health expenditure
Country Total expenditure (% of GDP) Private expenditure (% of total) Public expenditure (% of GDP)
United States 13.9 55.6 6.17
Canada 9.5 29.2 6.73
Sweden 8.7 14.8 7.41
United Kingdom 7.6 17.8 6.25
Germany 10.8 25.1 8.09
France 9.6 24.0 7.29
The Netherlands 8.9 36.7 5.63
Average: countries with high HDI 9.8 29.0 6.79
Brazil 7.6 58.4 3.16
Source: WHO (2001)
According to Barr's scheme, the American health system is one of those that is closer to the private market model: this system "shows those problems predicted in the theory". In terms of resource allocation, public expenditure covers exactly those areas in which policies designed for health insurance are not capable of paying for the risks: Medicare for aged people; Medicaid for the poor; military veterans (partially due to chronic health problems); maternity and children's welfare. Furthermore, an increasingly high cost and unequal access to services can also be found: at the end of the 1980s, 17.5% of the population above 65 years of age could not count on adequate insurance coverage in the United States [3]. Finally, based on an assessment of the most pro-market health care system, Arrow's view – outlined in his classical paper – is confirmed.
It is worth noting that the pattern of the American public expenditure on health is comparable to those in countries found at the other extreme of a description of welfare systems proposed by Barr: the Swedish government invests 6.8% of the country's total GDP in health (Table 1).
The issue concerning the effectiveness of the several institutional arrangements is very controversial. Hurst [10] compares the systems in the United States, Canada and the United Kingdom, concluding that the British system would be the most efficient, as it was the cheapest, with similar results.
An international comparison raises relevant questions on the relative effectiveness of health care systems. The World Bank report, for example, compares countries in terms of health expenditure and results [1]. According to this evaluation, the United States is at one extreme, i.e. the countries with the worst performance and highest expenditure. China is placed at the opposite extreme – better performance with lower expenditure (Figure 3.1, p. 54).
Such evaluation is not a trivial task. Measuring productivity is (usually) an old problem in economics and it is becoming even more serious with the emergence of information and communication technologies [11]. Measuring productivity in the service sector (where the health care sector is placed) is even more troublesome. Gordon [12] presents a general survey of the discussion for the United States. In assessing the growth rates of sectoral products by employee, he found that the health care services presented a positive variation rate only during the period 1960–1972. In the remaining periods (1972–1979, 1979–1987, 1987–1992), it was negative. According to Griliches [11], these services would be ranked among those hard-to-measure sectors (as opposed to measurable economic activities).
Although measuring productivity in the sector is troublesome and controversial, measuring cost increases is not; the difficulty here is to determine the reasons for the rise in medical care costs. What is consensual in the literature is the role that the structure of incentives plays in the cost dynamics and even in the direction technological progress will take in the sector [13]. In other words, the way medical care is organized (in the various existing structures) contributes to the definition of the activity performance and the expenditure policy.
In order to understand the effect of medical care organization on its performance, a study developed in Brazil is quite instructive. Campos [14] found that the physician's decision – considering all the degrees of liberty that professional autonomy grants him/her – strongly affects the consumption pattern and the impact of medical action on the epidemiological indicators. By studying the resolvability of health care services in homogeneous cities – whose differences are only those relative to the health professionals' labour ties – the author found a significant difference between a system hiring its professionals under a regime of labor exclusiveness as compared to a traditional system of multiple labour ties. An explanation for this would be that the first model forces an on-the-spot resolution of problems, since the lack of resolution of a problem implies the patient's return, sometimes with some inconvenience regarding time and circumstances to the professional.
The second model, by its segmentation, is limited to the traditional treatment prescription without taking account of the outcome of such a behavior, as a responsibility tie is not established between the professional and the patient. Campos [14] concludes that "working in an exclusiveness regime is the major determinant of such a differential behavior". In all dichotomies found in a tree of decisions studied – e.g. concerning the commitment to a drug prescription, laboratory examinations prescribed and hospitalizations, among others – significant differences in behaviour were found among medical personnel.
The outcome of this investigation can be generalized in terms of determining the way labour is organized considering the quality of its results. With regard to the United States, it is believed that the structure of medical insurance through third-party payments and fee-for-service encourages the overuse of services (leading to increasing prices). Barr [3] considers this structure to be one of the causes of high costs in the American system. The emergence of health maintenance organizations – HMOs – has been an alternative of shared responsibilities by the users and service providers, a way by which the agents share the consequences of increased expenses (cutting harmful incentives granted to third-party payments). The growth of HMOs is also related to encouragement of competition among service providers, a policy suggested by the World Bank to high-income countries [1].
Such changes affect academic research, however. Studies show that "in regions where managed care plans are dominant and where there is stiff competition for dollars and patients among hospitals, physicians at academic medical centers report more pressure to take care of patients – and thus conduct fewer human studies, do less clinical research, and publish fewer papers" (NSF) [15]. A problematic result of higher competition at the service level is that nowadays medical research is directly related to the future quality of medical care.
Discussion: the health care sector's articulation of two institutional arrangements – social welfare systems and national innovation systems
The health care system possesses a distinctive characteristic relative to other economic sectors, i.e. it is the intersection between social welfare and innovation systems. These two systems (two institutional constructions) endeavour to surmount market restrictions. Arrow [16] points to a market economy trend to under-invest in research and development activities, which, as in the medical sector, would lead to the emergence of non-profit institutions so as to reach more desirable levels of R&D investment. These two institutional arrangements may be justified by Arrow's analysis [17], which considers that the market poses restrictions to efficiency (a task for the innovation systems) and equity (a task for social welfare systems).
The scientific and technological progress of nations – a decisive source of economic growth and development – is an outcome of complex institutional articulations involving firms, their R&D laboratories, universities and research institutions, financial systems, teaching institutions in general and the interaction of all such institutions, especially among firms [18,19].
The development of innovation systems is derived from a trend predicted by Marx [20], i.e. the systematic application of science to production. National systems of innovation may be studied as an institutionalization of this phenomenon discussed in the Grundrisse.
A national innovation system can be disaggregated into different sectors as the characteristics of technological progress vary significantly among the several sectors [21,22]. It is appropriate to say that innovation in the textile sector is quite different from innovation in the computer industry, i.e. the latter, for instance, depends much more on scientific knowledge and has a closer relation to universities and research outcomes [23]. Scholars of innovation economics have been surprised at the existence of a close relation between science and technology in the health care sector [24]. Following this rationale, the health sector can be outlined by the innovative dynamics differently from other economic sectors. Cautiously, however, the existence of an innovative subsector in the health care sector could be suggested.
A starting point already determined in the specific literature of the sector [25] is the notion of medical-industrial complex, which is an articulation involving medical care, education networks (schools, universities), the pharmaceutical industry and the medical equipment and diagnostic instrument industry. Back to that suggested formulation, the existence of an innovative subsystem within the health sector stemming from the literature on economics of technology and innovation adds to it a major aspect, i.e. it is necessary to study the information flows of technology and the mechanisms generating innovation in the medical-industrial complex. Gelijns & Rosenberg [26] presented a review of the literature on complex interactions between universities, industries and medical care systems, which pushed forward the advancement of medical technology. As in other sectors, interactions between demand for and supply of innovation are complex and varied.
On the one side, the study of a sectoral system necessarily contributes to the understanding of the medical care system. That is, the quantity and quality of treatment supplied, diagnostic methods and available equipment constitute a direct result of investment accomplished in scientific and technological research. By the way – and this is an important aspect for the specific objectives of this paper – a good deal of the "guilt" over the increased cost of the health care sector has been laid on technological innovation [3].
Lessons from the literature on economics of technology broke the traditional vision of technological progress, known as a "linear model". According to this model, there would be a process "from the top down", starting with basic research towards the laboratories of firms where applied research is accomplished and finally reaching the production phase. It could roughly be depicted in a scheme as follows: SCIENCE → TECHNOLOGY → PRODUCTION
This linear scheme is considered a distant representation of reality. Sources of technical progress are much more complex. For example, solving problems and bottlenecks in production is a major source of innovation (this is the way new methods of production appear). In many instances, the arrows' sense is the opposite – radio astronomy has been developed as a new scientific discipline based on the work of two physicists (Penzias and Wilson), employees at the laboratories of the Bell Company, in an attempt to solve a noise problem in transcontinental telephone calls. Therefore, science can be viewed as both leading and following technological advances [19]. For this reason, the existence of a dynamic nucleus of firms is crucial for the maturing of national innovation systems.
Hospitals play a major role as an authentic reservoir of innovation "coming from the top". Hospitals contribute to scientific development, i.e. the arrows of the scheme above point to both sides in the health sector, too. As a matter of fact, service-providing would usually play a similar role to that of firms in other sectoral systems: solving problems and escaping from bottlenecks is a significant source of innovation [4].
What is peculiar in the interaction between health innovation systems and health care systems is their closer tie to each other and the more immediate impact between technological progress and social welfare, the latter a decisive component of economic growth sources. Therefore, innovation in the health sector would have a double effect on the economic dynamics in general – the "usual" effects of every innovation and also those on health and welfare.
The relevance of such an articulation simply adds a new peculiarity and a new source of labour heterogeneity to the sector. It would not be possible to grasp the health care system integrally by neglecting academic research in the life sciences. The life sciences spent 54.4% (USD 10.83 billion) of total resources for R&D in academic institutions in 1993. In that year, the total United States expenditure on R&D amounted to USD 134.4 billion [27]. In 1994, the health sector absorbed 16.5% of federal expenditure on R&D, which reached USD 68.33 billion. The weight of R&D investment in health can also be assessed in the industrial sector. Drugs and medicines were the industrial sector with the most intense R&D (R&D expenditure in relation to revenues), and the sector of optical and surgical instruments and others was placed fifth [27]. Total expenditure on biomedical R&D funding reached USD 30 billion in 1993 [28]. Industry accounted for 50% of the total. American industry increasingly depends on the funding of science with public resources; the leading position of the biomedical sector is highlighted in this regard [29].
Therefore, it is possible to trace a labour repositioning, in relation to which the pole constituted by activities connected with intellectual labour can be found [30]. The peculiarity of labour repositioning in the health care sector would not reside in the shift of functions of manual labour (as in the industrial sector), but in the increased participation of more skilled professionals (including scientists and researchers), in addition to the demand for better-trained professionals in the area to deal with diagnostic methods and electronic equipment, etc. Such a dynamic stresses the need for training and retraining the whole set of professionals in the sector: the speed of technological progress enhances the role of continued learning.
Discussion: technological innovation and cost in the health care sector
The technological innovation dynamics in health care has been considered one of the reasons for the increased expenses in the sector. This would be partly explained by another peculiarity in the sector, i.e. the cumulative introduction of technology, as opposed to other production sectors, where new technology replaces the old. For example, the introduction of cardiac imaging (electrocardiogram, echocardiography, Doppler) did not replace traditional auscultation; the old technology and the new are used interchangeably. The obstetrician works with the time-tested Pinnard stethoscope together with modern sonar devices in order to listen to the fetal heartbeat.
In the case of the United States, there is another question derived from the pressure the demand for medical care exerts on R&D activities: the health insurance mode of organization based on retrospective payments (the mode of medical care and the incentive structure hence derived) exerts pressure on R&D activities in terms of producing costly and effort-consuming innovation [13]. Recent changes in the system would lead to a reversal of this pressure so as not to encourage the use of expensive technologies. As an example, General Electric had frozen the development of a diagnostic technology called positron emission tomography (PET) which "produces tridimensional images reflecting chemical and metabolic activities in the tissues". The reason for such a freeze, according to the company, was that government was too stringent in approving reimbursement to patients for PET scans. Previously, the company had invested heavily in developing computerized tomography scanners (CT) and magnetic resonance imaging (MCI), and had taken them to the market [13].
Such sensitivity in terms of technological progress in relation to the incentive structure is quite important. Two examples of how the structure of service provision affects both its quality [14] and the degree of involvement of hospital units with research [15] were presented in the third section. Surveying this subject, Halm & Gelijns [31] say: "it is clear that the critical point here is not medical technology per se, but a combination of economic, professional, and social incentives in the health care system which tend to diminish apprehensions as to the decision-making in medical care".
There is evidence that technological innovations are not exclusively price-raising in medical care [13]. This is an open question in the OECD document, which is uncertain "whether new technologies are part of the problem, part of the solution or both" [32].
In order to analyse this aspect, Weisbrod compares vaccines and transplants, their costs, effects and respective demands for innovation. He used the biologist Lewis Thomas' elaboration, which distinguishes three technological development stages in medicine so as to make his position explicit:
• at the lowest level, the non-technology level, where the relation between the patient and the disease is entirely understood. Little can be done by the patient, hospitalization and infirmary services, and there is little hope of recovery (untreatable cancer, severe rheumatoid arthritis, multiple sclerosis, advanced cirrhosis);
• at a somewhat higher level, halfway technologies, which would include dealing with the disease and its disabling effects. These are technologies that adjust the patient to his/her disease and postpone death (artificial organ implants, cancer treatment by means of surgery, radiation and chemotherapy);
• high technologies, as for example immunization, antibiotics and prevention of nutritional disorders, are designed for diseases whose mechanisms are known and whose treatment/prevention is feasible.
Weisbrod suggests the use of the following dynamic scheme: historically, knowledge goes from the first kind of technology to the second and then to the third. From this scheme, it arises that the cost function associated with such a dynamic process is an inverted U-shaped one. In the case of non-technology, not much can be done and the costs are low. The most expensive would be that of intermediate technologies, then decreasing again in the third stage – high technology. Weisbrod uses the evolution of poliomyelitis as an example: in the beginning (two generations ago), its victims died rapidly as a result of paralysis; then the development of intermediate technology came about with the emergence of the iron lung, which prolonged the patient's life at high cost; finally, the vaccines (Sabin and Salk) in the high technology phase reduced dramatically the costs associated with polio [13].
Based on this scheme, Weisbrod suggests, for the case of the United States, that "the development of halfway technologies was implicitly encouraged by the cost-reimbursement insurance system that has dominated hospital and medical care until recently, because there was little or no incentive for medical care providers to avoid costly technologies that were even marginally effective". (p. 534). That is, it is not the technology that accounts for increased costs, but the scheme of incentives that guides its evolution.
Further on, Weisbrod lists the technologies with a demand for health insurance: "the demand for health insurance tends to increase most rapidly when changes in technology are of the expenditure-increasing, halfway type" However, high technologies (vaccines) would tend to diminish the demand for insurance."
In order to confirm his suppositions, he describes some impacts of the emergence of HMOs being more concerned about costs; they would have broadened their R&D profitability by directing their actions to: drugs that could avoid costly treatments; drugs that replace surgeries (e.g. cimetidine, a substitute for ulcer surgery) [13]. Lichtenberg studied the relation between new medicine and the demand for hospitalization and found that hospital bed-days declined rapidly "for those diagnoses with a larger number of drugs prescribed and a greater change in distribution of medicines". He estimated that an increment of 100 prescriptions is associated with a reduction of 16.3 days of hospitalization [33].
However, perhaps Weisbrod has added another problem to Arrow's paper. As it is basically bought to pay for hospital treatments, health insurance provides incentives for R&D to search for ways to treat patients rather than prevent diseases. And this is not "optimum" in social terms.
Weisbrod's analysis is interesting as it contributes to assessing the demand for innovation in the health sector and mainly to show how the sector's organization affects technological progress. However, in the debate on health care changes, only the side of the demand for technological innovation has been focused upon; the conditions governing the supply of innovation have been neglected [26]. Undoubtedly, further developments in cancer prevention are restrained in part by the state of science.
A good example of such a restraint is provided by biotechnology: "a revolution in health care", announced OCDE [32]. The development of genetic therapies might mean treating cancers, genetic diseases and others (such as rheumatoid arthritis). Some research efforts are in the phase of clinical trials. However, the development of such therapies is complex and difficult, and so far "clinical efficacy has not been demonstrated in any of the genetic therapies". Perhaps "the genetic therapy will take a long time before reaching the patients" [32].
However, it is possible to conjecture that the biotechnology revolution will make high technology-type innovation available as soon as it is developed and employed in health care systems, in accordance with Weisbrod's scheme – efficacy derived from the understanding of the processes of a number of diseases treated with (cost-reducing) vaccine-type therapies.
Discussion: issues concerning the situation in Brazil
So far the discussion has centred on advanced countries. A brief introduction to the situation in Brazil requires caution so that conspicuous differences may not be neglected.
The first major difference is the country's development stage. According to the World Bank, Brazil is a high-medium income country, with a GDP per capita of USD 3640 in 1995, and was placed 46th in the world ranking [34]. In relation to the Human Development Index (HDI), it was ranked 58th in 1993 and 62nd in 1995. The technological gap and the social gap come together. If expressed by the terms used all through this paper, such a reality can be translated into precariousness of social welfare systems in the country (with severe influence on its health care structure) and the rudimentary and incipient character of the national innovation system [35].
This standpoint contributes to determining the scenario of health and morbidity in the country. Brazil has been undergoing, according to the jargon used in the health milieu, an epidemiological transition, which combines features found in a low-income country, such as sanitation shortages, malnutrition and infectious-parasitic diseases, with features of high-income countries, such as a growing incidence of degenerative diseases. Such differences pose complex tasks to be tackled by the whole health apparatus in the country.
As for the labour peculiarity, the health care sector must count on broad competences, ranging from the treatment of simple verminous diseases to modern techniques of emergency treatment. A neurosurgeon and a plastic surgeon are not luxuries – these two kinds of specialists are needed in a country with a high incidence of labour accidents such as Brazil (one need consider only the kind of accidents found in construction and those accidents requiring hand-repairing surgery, recovery of persons with burns, etc.).
To an extent, the capacities the advanced countries have been building through time – which have been substituting for each other since their introduction – must coexist in the country. The result is a health care system more complex and differentiated than those of countries at either extreme (high-income countries: no verminous diseases and better labour conditions; low-income countries: lower incidence of degenerative diseases).
Another effect of Brazil's economic stage is the existing budgetary restriction: basic necessities (education, sanitation, health, infrastructure investment) compete with each other in budgets with relatively scarce resources.
The World Bank proposal for governmental action is based on an intervention divided into three basic levels, corresponding to three different foundations [1]: alleviation of poverty and access of the poor to health care services; public health; extension of medical care to the population through insurance as well as its regulation. The World Bank proposal concentrated the discussion on the latter (insurance and regulation). The Brazilian peculiarity would be the relevance of an action based on a combination of these three levels [1].
That the Brazilian constitutional text has adopted the proposed organization of a single health care system (SUS) that is universal and equitable, with an integral and socially controlled approach, is equally a significant conceptual advancement and a great operational complication, when the health care scenario in the country is considered.
Boelen [36] proposes a comparative analysis of the social accountability of health care services that are oriented by four polar concepts: equity versus quality and relevance versus cost-effectiveness. According to the author, it would be relatively easy to design an equitable system by following a formula containing only the basic health care actions for the most vulnerable population groups. In the same way, it would be theoretically simple to construct systems following the quality criterion alone, unconcerned with the coverage of the actions developed. In this case, the concept of relevance of the health care actions would oppose the analysis of their cost effectiveness.
When writing its constitution chapter on health, and hence including the SUS proposal, Brazil posed a great challenge to itself, i.e. to reach the four cardinal points (equity, quality, relevance and cost-effectiveness) at the same time. The very fact that the health chapter of the Brazilian constitution is within the social security area, joining health care actions with social security and social work activities, accounts for the size of such a challenge.
The difficulty resides in correctly developing the matching of such elements. As compared with more advanced countries (Table 1), it is worth assessing the need for and possibility of a general increase in health care expenditure (public and private). Public expenditure, by strengthening basic programmes, public health and investment in regulatory activities (the recent scandal of falsified drugs is tragic proof of the price to be paid for weakness in such areas) cannot be replaced.
Based on the ideas discussed throughout the present paper – that the health care system is placed in the intersection of the welfare system and the innovation system – the relevance of social and economic investment in research should be considered. Investment should be made in the country to enhance scientific and technological capacity in biotechnology as well as to improve and extend sewage systems. The range of activities to be achieved by the construction of these two indispensable systems is sizable.
From the viewpoint of technological innovation in health care and that of the rest of the innovation system as well, the technological gap of the country, in relation to the international technological frontier, reveals some advantages and requires some efforts [37]. The advantages are as follows: investment in initial phases of development is not necessary, as the country is in its absorption phase of technologies generated in the technological frontier; it is quite possible for the country to adopt a technology after defining its development "path" (i.e. expenditure on technologies that could later be replaced by new developments can be avoided).
However, such advantages cannot be exploited without making important domestic investments – constructing "absorption capacity" is a must, since: the absorption and necessary adaptation of such technologies are not passive processes; they require knowledge, critical mass, financial and entrepreneurial capacity; they presuppose a follow-up and monitoring capacity of current scientific and technological progress, to the extent that basic research frequently means an entrance ticket to a circuit of scientific and technological information, as pointed out by Mowery & Rosenberg [38]; previous knowledge is necessary even for a simple purchase of equipment, machines and processes.
Conclusions: recognizing the worth of sectoral labour and integrating it
The incapacity of the market to allocate resources in health care production, the asymmetry of knowledge and the relation of trust between the physician and the patient are among the reasons why attempts at external regulation and control of health care labour have failed or can be bypassed.
The traditional compendiums of health administration have already recognized that the three basic modes of health care labour remuneration – fixed-salary payment based on time spent in the procedure, fee for service and the different modes of capitation – present advantages together with remarkable failures as to its performance controllability.
Those receiving fixed salaries tend to evade providing services, unlike those paid for procedures accomplished, who tend to overestimate service provision, while mechanisms for capitation may be bypassed through selecting less vulnerable groups. Perhaps for these very reasons, these mechanisms are rarely used separately, and there is a trend towards combining them, as for example, incentives for increased productivity combined with wages.
No matter how creative health service managers may be, there will always be circumstances standing in their way, aggravated by a disguised protection of some attitudes by corporatism, professional complicity or a strictly organized hierarchical structure. There are differential requirements that are rarely unequivocally expressed beforehand. A clear example of this is the strict control of the hours worked by subordinates, on the one hand, and a relative leniency with physicians as to the same requirement, on the other hand. Another striking example of such a situation is the failure of management to attempt to limit the number of examinations and hospitalizations provoked by a given number of visits to the doctor.
First, standardization is impossible if the input of this system is not known, i.e. the seriousness and complexity of the pathologies to be treated. Furthermore, the control of the denominator of this equation is very difficult, as it consists of visits and not of assisted patients. As it would hardly be reasonable to forbid the patient's return visits, which would be desirable as a demonstration by the service provider of concern with solving the problem, the number of visits for a similar group could be unnecessarily multiplied, which would allow a striking inflation of procedures, in this way bypassing the external control.
For all these reasons, a tight wage policy and precarious health care labour conditions, instead of saving resources, may result in their waste by provoking an increment of unnecessary examinations and hospitalizations that could be avoided if a specific agreement to solve the problems were reached. In this case, a positive labour incentive – including improved wages and working conditions and encouragement for training, i.e. positive organizational "climate" and "culture" – would certainly represent a positive impact on health conditions without a burst of final costs.
In the Brazilian experience it can be said that, on the negative side, doubling the salaries of social security physicians after months of strike in the early 1990s did not have measurable positive impacts on productivity and outcome, despite the economic burden for the public accounts. In contrast, some managerial micro-decisions with low added costs – such as personal recognition, prestige, kindness in interpersonal relations and a smooth work environment, encouragement to participate in scientific events, flexibility in order to attain personal expectations and demands – can have positive impacts in the outcomes. Therefore, it seems clear that it would not be possible to rationalize health care labour without the workers' adherence and collaboration, in an agreement that could simultaneously benefit users and providers.
Additionally, viewing the worth of labour exclusively as a wage issue would be simplistic, although it is still crucial. Another way to dignify labour is by implementing continued education and training of professional staff. This process should take into account the rapid development of the innovation system whose consequence is the obsolescence of acquired knowledge. More than half of medical techniques are estimated to become useless in 16 to 18 years. The contribution of the academic apparatus – which was so relevant during the medical education "boom" period about 20 years ago and surely made this professional activity available to large population contingents, previously unassisted – will be lessened if such institutions continue giving only initial education background to physicians and other professionals in the sector, which was a very important mission when innovation was a relatively slow-paced process.
Competing interests
The author(s) declared that they have no competing interest.
Authors' contributions
FEC and EMA shared the research and co-authored this paper.
Acknowledgements
This article is based on a paper prepared for the Pan American Health Organization (PAHO) in 1998. The authors thank Ana Luíza Lara, Elaine Rodrigues and Thais Henriques for research assistance. The authors also thank Dr Binod Khadria for corrections and suggestions. Any remaining errors are ours.
==== Refs
World Bank World Development Report 1993 1993 Oxford: Oxford University
Arrow K Arrow K Uncertainty and the welfare economics of medical care Essays in theory of risk-bearing 1971 Amsterdam/London: North Holland 177 219
Barr N Economic theory and the Welfare State: a survey and interpretation Journal of Economic Literature 1992 30 741 803
Hicks D Katz J Hospitals: the hidden research system Science and Public Policy 1996 23 297 304
Machado MH Os médicos e sua prática profissional: as metamorfoses de uma profissão Tese de Doutoramento Rio de Janeiro: Escola Nacional de Saúde pública da FIOCRUZ 1996
Akerlof G The market for 'lemons': quality uncertainty and the market mechanism Quarterly Journal of Economics 1970 84 488 500
Arrow K Interview with Kenneth Arrow 1995
Kalisch DW Aman T Buchele L Social and health policies in OECD countries: a survey of current programmes and recent developments OECD Labour Market and Social Policy – Occasional Papers 33 1998
Esping-Anderson G The Three Worlds of Welfare Capitalism 1990 Cambridge: Polity
Hurst J Financing Health Services in the United States, Canada and Britain 1985 London: King Edward's Hospital Fund
Griliches Z Productivity, R&D, and the data constraint American Economic Review 1994 84 1 23
Gordon R Problems in the Measurement and Performance of Service-Sector Productivity in the United States 1996 Cambridge, Massachusetts: National Bureau of Economic Research Working Paper 5519
Weisbrod B The health care quadrilemma: an essay on technological change, insurance, quality of care, and cost containment Journal of Economic Literature 1991 29 523 552
Campos FE Resolutividade: uma aproximação à avaliação qualitativa dos serviços de saúde Tese de Doutoramento Rio de Janeiro: Escola Nacional de Saúde pública da FIOCRUZ 1988
National Science Foundation Science and Engineering Indicators 1998 Washington: National Science Foundation
Arrow K Lamberton D Economic welfare and the allocation of resources for invention Economics of Information and Knowledge 1971 Harmondsworth: Penguin Books
Arrow K The Limits of Organization 1974 New York: WW Norton & Co
Freeman C The "National System of Innovation" in historical perspective Cambridge Journal of Economics 1995 19 5 24
Nelson R Ed National Innovation Systems: A Comparative Analysis 1993 New York/Oxford: Oxford University
Marx K Grundrisse 1992 London: Penguin Books/New Left Review
Freeman C Soete L The Economics of Industrial Innovation 1997 London: Pinter
Pavitt K Sectoral patterns of technical change Research Policy 1984 13 343 373 10.1016/0048-7333(84)90018-0
Klevorick A Levin R Nelson R Winter S On the sources and significance of inter-industry differences in technological opportunities Research Policy 1995 24 185 205 10.1016/0048-7333(93)00762-I
Nelson R Rosenberg N, Gelijns A, Dawkins H The intertwining of public and proprietary in medical technology Sources of Medical Technology: Universities and Industry Medical Innovation at the Crossroads 1995 5 Washington: National Academy 219 220
Cordeiro H A Indústria da Saúde no Brasil 1980 Rio de Janeiro: Graal
Gelijns A Rosenberg N Rosenberg N, Gelijns A, Dawkins H The changing nature of medical technology development Sources of Medical Technology: Universities and Industry Medical Innovation at the Crossroads 1995 5 Washington: National Academy 3 14
National Science Foundation Science and Engineering Indicators 1996 Washington: National Science Foundation
Bond E Glynn S Rosenberg N, Gelijns A, Dawkins H Recent trends in support for biomedical research and development Sources of Medical Technology: Universities and Industry Medical Innovation at the Crossroads 1995 5 Washington: National Academy 35 57
Narin F Hamilton KS Olivastro D The increasing linkage between U.S. technology and public science Research Policy 1997 26 317 330 10.1016/S0048-7333(97)00013-9
Albuquerque E Invenções, Mutações: o Progresso Científico Tecnológico em Habermas, Offe e Arrow Belo Horizonte: UNA 1996
Halm E Gelijns A Gelijns A, Halm E An introduction to the changing economics of technological innovation in medicine The Changing Economics of Medical Technology Medical Innovation at the Crossroads 1991 2 Washington: National Academy 1 20
Organisation for Economic Co-operation and Development Economic Aspects of Biotechnologies Related to Human Health Part I 1998
Lichtenberg F Do (more and better) drugs keep people out of hospitals? American Economic Review 1996 86 384 388 10172775
World Bank World Development Report 1997 1997 Oxford: Oxford University Press
Albuquerque E Sistema nacional de inovação no Brasil: uma análise introdutória a partir de dados disponíveis sobre a ciência e a tecnologia Revista de Economia Política 1996 16 56 72
Boelen C Towards Unity for Health 1997 Geneva: World Health Organization
Albuquerque E Notas sobre os determinantes tecnológicos do catching up: uma introdução à discussão sobre o papel dos sistemas nacionais de inovação na periferia Estudos Econômicos 1997 27 69 96
Mowery D Rosenberg N Technology and the Pursuit of Economic Growth 1989 Cambridge: Cambridge University Press
|
16109174
|
PMC1201162
|
CC BY
|
2021-01-04 16:37:43
|
no
|
Hum Resour Health. 2005 Aug 18; 3:8
|
utf-8
|
Hum Resour Health
| 2,005 |
10.1186/1478-4491-3-8
|
oa_comm
|
==== Front
Int J Health GeogrInternational Journal of Health Geographics1476-072XBioMed Central London 1476-072X-4-201609296910.1186/1476-072X-4-20ResearchInequalities in neighbourhood socioeconomic characteristics: potential evidence-base for neighbourhood health planning Odoi Agricola [email protected] Ron [email protected] Marion [email protected] Stephen [email protected] Brian [email protected] John [email protected] Tom [email protected] Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario, Canada2 Formerly with Hamilton District Health Council, Hamilton, Ontario, Canada3 Canadian Institute for Health Information, Toronto, Ontario, Canada4 Centre for Health Economics and Policy Analysis, McMaster University, Hamilton, Ontario, Canada5 School of Geography and Geology, McMaster University & McMaster Institute of Environment & Health, Hamilton, Ontario, Canada2005 10 8 2005 4 20 20 15 6 2005 10 8 2005 Copyright © 2005 Odoi et al; licensee BioMed Central Ltd.2005Odoi 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
Population health planning aims to improve the health of the entire population and to reduce health inequities among population groups. Socioeconomic factors are increasingly being recognized as major determinants of many aspects of health and causes of health inequities. Knowledge of socioeconomic characteristics of neighbourhoods is necessary to identify their unique health needs and enhance identification of socioeconomically disadvantaged populations. Careful integration of this knowledge into health planning activities is necessary to ensure that health planning and service provision are tailored to unique neighbourhood population health needs. In this study, we identify unique neighbourhood socioeconomic characteristics and classify the neighbourhoods based on these characteristics. Principal components analysis (PCA) of 18 socioeconomic variables was used to identify the principal components explaining most of the variation in socioeconomic characteristics across the neighbourhoods. Cluster analysis was used to classify neighbourhoods based on their socioeconomic characteristics.
Results
Results of the PCA and cluster analysis were similar but the latter were more objective and easier to interpret. Five neighbourhood types with distinguishing socioeconomic and demographic characteristics were identified. The methodology provides a more complete picture of the neighbourhood socioeconomic characteristics than when a single variable (e.g. income) is used to classify neighbourhoods.
Conclusion
Cluster analysis is useful for generating neighbourhood population socioeconomic and demographic characteristics that can be useful in guiding neighbourhood health planning and service provision. This study is the first of a series of studies designed to investigate health inequalities at the neighbourhood level with a view to providing evidence-base for health planners, service providers and policy makers to help address health inequity issues at the neighbourhood level. Subsequent studies will investigate inequalities in health outcomes both within and across the neighbourhood types identified in the current study.
==== Body
Background
Traditional health planning has typically focused on the practice and delivery of health care services. Population health planning, on the other hand, aims to improve the health of the entire population and to reduce health inequities among population groups [1]. The health of a population is influenced by several factors including but not limited to socioeconomic status [2], social support networks [3] education [4], ethnicity [5], employment [6], working conditions [7-10], physical environment [11], personal health behaviours [12,13], health care services [14] and individual coping skills [15,16]. Therefore, health planning, policy and interventions need to take into consideration not only the health care services, but also these broad determinants of health.
Since socioeconomic and demographic characteristics are important determinants of population health, adopting a population health approach to health planning at the neighbourhood level requires improved knowledge of the distribution of population socioeconomic and demographic characteristics at this level. Globally, there is an increasing interest in understanding the relationship between neighbourhood of residence and health of the population [17-19]. To this end, some researchers have suggested that improving the health of those living in the worst areas calls for systematically exploring area differences to inform social and health policy [20].
Currently, the lowest geographical level at which most health-planning data in Canada are analyzed is the municipal (city) level. Obviously, the use of such a large unit of analysis limits the ability to identify specific population characteristics as well as health variations and needs at the lower levels. The implication is that disparities in health outcomes and access to health care services across population sub-groups at these lower levels are unclear. Moreover, most large cities have diverse populations [21-24]; therefore the neighbourhoods within them have diverse socioeconomic and demographic characteristics that may influence neighbourhood population health outcomes and therefore health needs [25-27]. A number of studies have shown the extent and causes of neighbourhood socioeconomic inequalities [21-23,28-31]. In Canada, there is evidence that neighbourhood socioeconomic inequality has been rising since 1970 [32,33]. Moreover, numerous studies have reported associations between neighbourhood socioeconomic characteristics and various health outcomes [34-42]. Therefore, taking into consideration the diverse socioeconomic and demographic characteristics of the different neighbourhoods during health planning would ensure that planning and health services are tailored to the unique needs of the local residents of each neighbourhood.
Studies of geographical distribution of determinants of health have mainly used one of three approaches. The first involves either production of a single map showing the spatial distribution of a single variable (determinant of health) or production of a series of maps each showing the distribution of a single determinant of health [43-45]. The limitations of this method are that only one determinant can be assessed at a time(if a single variable approach is used) and assumes each determinant is independent of other determinants (if a series of maps is used). Moreover, when a series of maps is used, interpretation may be difficult. In the second approach, a composite index is created from combining two or more variables [46-48]. Although this approach mitigates the limitations of approach 1 and is effective in highlighting areas considered to be "high-risk", its drawback is that specific population characteristics (e.g. education, ethnicity, income, etc) are rolled into an index so that one cannot identify distinct characteristics (with respect to these variables) attributable to specific geographical areas. A third analytical approach uses factor analysis (or principal components analysis) to investigate several determinants [49,50]. This approach is a data reduction technique used to reduce the dimensionality of the data from several variables to a few factors (or principal components) that explain most of the variability in the original data. The current study uses principal components analysis and cluster analysis and Geographical Information Systems (GIS) to mitigate the limitations of the above approaches.
The objectives of this study were to use multivariate statistical techniques to identify the socioeconomic and demographic characteristics of neighbourhoods in the city of Hamilton, Ontario, Canada; and to classify the neighbourhoods based on similarities of these characteristics. Potential applications of the methodology in needs-based neighbourhood population health planning, service delivery, and policy development are proposed.
Methodology
Study area and geographical scale of analysis
The study was carried out in the city of Hamilton, Ontario, Canada. The city has a population of over 490,000 people and spans over 1,117 square kilometres [51]. A number of factors were considered in selecting the appropriate level of geography for the study. These included homogeneity of socioeconomic variables within the geographical unit; large enough population size to minimize the "small number problem"; data availability; acceptability by health planners and health service providers; and stability of the boundaries over time for maximum temporal data comparability for future analyses.
Based on the above criteria, census tracts were chosen as the most appropriate level of geography for the analysis. A census tract (CT) is a small, relatively stable geographic unit usually having a population of 2,500 to 8,000 persons with an average of approximately 4,000 [52]. There are 132 CTs in the City of Hamilton. Census tracts are used in this study to represent neighbourhoods because CTs have 'neighbourhood-like' characteristics due to their homogeneity with respect to socioeconomic and demographic characteristics [53]. Therefore, throughout this paper, CTs and neighbourhoods are used interchangeably. There were several advantages of adopting this as the level of geography for analysis and future health planning: (i) Most census data are reported at this level of geography; (ii) Administrative health data can easily be aggregated to this level, if the postal codes of the health care recipients are known (this is because, in Canada, the postal code areas are smaller than CTs are so postal code data can easily be aggregated to the CT level); (iii) Census tracts are homogeneous with respect to socioeconomic and demographic characteristics; (iv) Their population sizes are large enough to allow calculation of relatively stable rates of most health events; (v) The boundaries of the CTs follow permanent and easily recognizable physical features and changes to their boundaries are discouraged to maintain maximum data comparability over time [52].
Data source and variable selection
Socioeconomic and demographic data for 833 dissemination areas (DA) in the city of Hamilton, Ontario, Canada, were extracted from the 2001 Canadian census data [52]. A DA is a small area composed of one or more neighbouring blocks and is the smallest standard geographic area for which all census data are disseminated in Canada [52]. The data were then aggregated to the CT level at which all analyses were performed. The variables used in the analyses were chosen based on their usefulness as determinants of health [54], reliability, and availability at the DA level [55]. Care was taken to include as many socioeconomic and demographic variables as possible in order to enhance the highest statistical differentiation between the CTs [55]. A total of 18 variables measuring the following characteristics were included in the analyses: demographic structure, social status, economic status, ethnicity, aboriginal status, and housing (see Table 1 for a complete list of variables).
Table 1 Variables included in the Hamilton neighbourhood analysis. Most definitions were adopted from Statistics Canada health region peer groups Study [55]. Those not adopted from the Statistics Canada study are: median income, married, live alone, population under 20 and non-official language population.
Variable name Definition
Persons with less than grade 9 education Percentage of the pop 20 years and over with less than grade 9 education
New immigrants The percentage of immigrants who came to Canada from 1996 to 2001
Visible minority Percentage of the population belonging to a visible minority group. As defined by the employment equity act (1986), visible minorities are persons (other than Aboriginal people) who are non-Caucasian in race or non-white in colour.
Aboriginal persons Percentage of population reporting at least one Aboriginal origin (North American Indian, Métis or Inuit)
Median income Median personal income for persons aged 15 and over, from all sources.
Government transfer income Percentage of all income that came from government transfers (e.g., Canadian pension plan (CPP), guaranteed income supplement (GIS), old age security, etc.) for the population 15 years of age and older.
Incidence of low income (LICO) Percentage of persons in economic families and unattached individuals with 2000 incomes below the Statistics Canada low-income cut-off (LICO). The cut-offs represent levels of income where people spend disproportionate amounts of money for food, shelter, and clothing. LICOs are based on family size and degree of urbanization; cut-offs are updated to account for changes in the consumer price index.
Non-official language pop Percentage of the population not speaking any of the two official languages.
Unemployment rate Total number of unemployed individuals 15 and older divided by the total number of individuals 15 and older participating in the labour force.
Average dwelling value Average expected value of an owner-occupied, non-farm, non-reserve dwelling (including the value of the land the dwelling is on) at the time of the census
Owner-occupied dwellings Percentage of dwellings in which the owner also lives. Band housing and collective dwellings (i.e. rooming houses, nursing homes, military camps etc.) Are excluded from both numerator and denominator.
Population under 20 years old Percentage of the population under the age of 20 years
Population 65 years or older Percentage of the population aged 65 years or older
Single-parent Families Percentage of single-parent families among all census families living in private households. A census family refers to a married or common-law couple or lone parent with at least one never-married son or daughter living in the same household.
Married Percent of legally married persons 15 & over
Live alone Percent of persons living alone in private households
Internal migrant mobility Percent of the population that lived in a different Canadian municipality at the time of the previous census. Excludes Canadians in households outside Canada (military & government personnel)
Population density Total population of a census tract divided by its area in Km2
Statistical analyses
Variable standardization and correlation analysis
All statistical analyses were performed in STATA [56]. To overcome the impact of differing variances and different scales of measurements (e.g. dollars vs percentages), all variables used in the analyses were standardized to mean 0 and unit variances [57]. Had standardization not been performed, variables with high variances would unduly dominate the results of the analyses. A correlation matrix was constructed to explore relationships among the variables.
Principal components analysis
Principal component analysis (PCA) was used to reduce the dimensionality of data and to investigate the nature of the relationships among the CTs with the main objective of isolating the general features that best describe the variations in the data. Using this method, 18 inter-correlated socioeconomic and demographic factors were reduced to 5 principal components each of which represented different aspects of the original data. Kaiser criterion (eigenvalue one test) was used to guide the decision on the number of principal components to retain; all components with eigenvalues equal to or less than 1 were not retained since they explained variations equal to or less than any one of the original variables [58,59]. To maximize the variance of factor loadings and therefore aid the separation of CTs (or neighbourhoods) into homogeneous groups, varimax rotation was used [60].
Cluster analysis
Cluster analysis is a multivariate statistical technique used to organize observations into groups (or clusters) such that observations within a cluster have a high degree of similarity (or natural associations) among themselves while the clusters are relatively distinct from each other [57]. There are many different definitions of a cluster [57]. For the purpose of this study, we define a cluster as a set of entities that are alike.
There are two major classifications of cluster analysis techniques: hierarchical and non-hierarchical (or partition) techniques. This study adopted the partition cluster analysis, using k-means clustering methodology, to group CTs (or neighbourhoods) based on socioeconomic and demographic characteristics into clusters or neighbourhoods types. Using this methodology, the user specifies the number of clusters, say x, to create. These x clusters are formed through an iterative process. The algorithm begins with x seed values which act as the initial x group means. Observations are then assigned to the nearest group seed. After all observations have been assigned, cluster means are computed for each group. The initial cluster seeds are then replaced by their respective cluster means. The observations are then re-assigned to the nearest cluster mean. These steps continue until no observations change groups (clusters).
The optimum number of groups or clusters or "neighbourhood types" to be identified was decided upon using Calinski-Harabasz pseudo-F test [61] as well as the distribution of the CTs within the cluster. Five clusters (groups) were found to provide the most optimal separation of the CTs within the clusters. This would allow a reasonable number of relatively homogeneous CTs (or neighbourhoods) per group (or neighbourhood type). Formation of more groups (neighbourhood types) resulted in some with too few CTs whereas formation of fewer groups resulted in loss of homogeneity within the groups.
The similarity (or dissimilarity) measure (also known as distance measure) used for the classification of the CTs was Minkowski distance metrics with argument 2 (L2) [57]. This measure, commonly known as euclidean distance, was calculated as follows:
where: p (in this case 18) is the number of variables included in the cluster analysis; xik and xjk are the values of variable k for CTs i and j respectively. The summations are over the p (or 18) variables involved in the cluster analysis. The 5 initial cluster centres were obtained randomly from among the CTs or neighbourhoods in the study area. For reproducibility a random number seed was applied before the 5 CTs were randomly chosen.
Since neighbourhoods belonging to the same group have certain socioeconomic and demographics characteristics in common, the resultant grouping provides useful insights into understanding the socioeconomic and demographic characteristics of each group (or neighbourhood type).
Cartographic manipulations
All cartographic manipulations were performed in ArcView GIS [62]. The principal components extracted during the PCA and the identified cluster resulting from the cluster analysis were exported to ArcView GIS. The geographical distribution of principal components 1–4 across the CTs was cartographically displayed in four different maps, one map per principal component. The spatial distribution of the five identified clusters were also displayed in one map. All CTs belonging to one cluster were represented with the same colour resulting in a map with five different colours each representing CTs belonging to the same cluster.
Results
Correlations
All the observed correlations were in the expected directions (Table 2). For instance, neighbourhoods with high proportions of low-income earners were more likely to have high percentages of non-official language population (r = 0.62), single-parent families (r = 0.8), population living alone (r = 0.6), and high population density (r = 0.6). Moreover, neighbourhoods with high percentage of visible minority population also tended to have high proportions of new immigrants (r = 0.69), low-income persons (r = 0.61), and low percentage of owner-occupied dwellings (r = -0.61). Additionally, neighbourhoods with high percentage of population with less than Grade 9 education, tended to have low median income (r = -0.65), and a high percentage of population receiving government transfer income (r = 0.7).
Table 2 Correlation matrix of variables used in multivariate analyses of socioeconomic and demographic variables in Hamilton neighbourhoods, 2004. Numbers not in brackets are pair-wise correlation coefficients whereas those in brackets are p-values. A: Persons with less than grade 9 education; B: New immigrants; C: Visible minority; D: Aboriginal persons; E: Median income; F: Government transfer income; G: Low-income persons; H: Non-official language population; I: Unemployment rate; J: Dwelling value; K: Owner-occupied dwellings; L: Population aged under 20 years; M: Population aged 65 years or older; N: Single-parent families; O: Married population; P: Population living alone; Q: Internal migrants R: Population Density
A B C D E F G H I J K L M N O P Q R
A 1.00
B 0.12 (0.166) 1.00
C 0.28 (0.001) 0.69 (<.001) 1.00
D 0.24 (0.006) -0.04 (0.623) 0.12 (0.176) 1.00
E -0.65 (<.001) -0.35 (<.001) -0.48 (<.001) -0.59 (<.001) 1.00
F 0.70 (<.001) 0.31 (<.001) 0.37 (<.001) 0.11 (0.229) -0.68 (<.001) 1.00
G 0.52 (<.001) 0.50 (<.001) 0.61 (<.001) 0.22 (0.013) -0.74 (<.001) 0.82 (<.001) 1.00
H 0.66 (<.001) 0.59 (<.001) 0.62 (<.001) 0.04 (0.660) -0.51 (<.001) 0.56 (<.001) 0.60 (<.001) 1.00
I 0.49 (<.001) 0.34 (<.001) 0.44 (<.001) 0.11 (0.201) -0.54 (<.001) 0.68 (<.001) 0.66 (<.001) 0.56 (<.001) 1.00
J -0.49 (<.001) -0.45 (<.001) -0.48 (<.001) -0.19 (0.027) 0.68 (<.001) -0.77 (<.001) -0.82 (<.001) -0.50 (<.001) -0.54 (<.001) 1.00
K -0.25 (0.004) -0.67 (<.001) -0.61 (<.001) -0.17 (0.054) 0.59 (<.001) -0.62 (<.001) -0.79 (<.001) -0.50 (<.001) -0.55 (<.001) 0.68 (<.001) 1.00
L -0.07 (0.426) -0.12 (0.190) 0.10 (0.246) 0.09 (0.300) 0.14 (0.103) -0.33 (<.001) -0.16 (0.077) -0.07 (0.413) -0.12 (0.180) 0.28 (0.001) 0.38 (<.001) 1.00
M 0.13 (0.132) -0.06 (0.485) -0.26 (0.003) -0.22 (0.010) -0.02 (0.859) 0.38 (<.001) 0.01 (0.948) -0.01 (0.906) 0.06 (0.484) -0.16 (0.072) -0.11 (0.195) -0.76 (<.001) 1.00
N 0.46 (<.001) 0.31 (<.001) 0.43 (<.001) 0.18 (0.039) -0.57 (<.001) 0.67 (<.001) 0.80 (<.001) 0.41 (<.001) 0.57 (<.001) -0.75 (<.001) -0.58 (<.001) 0.08 (0.394) -0.12 (0.184) 1.00
O -0.32 (<.001) -0.36 (<.001) -0.44 (<.001) -0.23 (0.007) 0.64 (<.001) -0.67 (<.001) -0.86 (<.001) -0.39 (<.001) -0.56 (<.001) 0.79 (<.001) 0.81 (<.001) 0.40 (<.001) -0.10 (0.240) -0.73 (<.001) 1.00
P 0.10 (0.252) 0.34 (<.001) 0.29 (0.001) 0.08 (0.344) -0.42 (<.001) 0.58 (<.001) 0.62 (<.001) 0.27 (0.002) 0.41 (<.001) -0.59 (<.001) -0.77 (<.001) -0.72 (<.001) 0.44 (<.001) 0.33 (<.001) -0.80 (<.001) 1.00
Q -0.33 (<.001) 0.16 (0.071) 0.14 (0.103) 0.35 (<.001) -0.03 (0.698) -0.19 (0.032) 0.02 (0.791) -0.04 (0.671) 0.00 (0.962) 0.10 (0.251) -0.25 (0.004) -0.03 (0.756) -0.24 (0.005) -0.11 (0.231) -0.16 (0.076) 0.24 (0.006) 1.00
R 0.18 (0.038) 0.42 (<.001) 0.43 (<.001) 0.11 (0.219) -0.41 (<.001) 0.44 (<.001) 0.60 (<.001) 0.32 (<.001) 0.36 (<.001) -0.62 (<.001) -0.67 (<.001) -0.28 (0.002) 0.05 (0.553) 0.48 (<.001) -0.64 (<.001) 0.59 (<.001) -0.02 (0.821) 1.00
Principal components analysis
Only the first five principal components, that had eigenvalues greater than 1, were extracted (Table 3). The first PC is associated with the largest eigenvalue. This PC is a linear combination of the variables that account for the highest variability (46.1%) in the data. The second PC explains the highest variability not accounted for by the first PC while the third PC explains the largest variability not accounted for by the first two PC and so on. The first five PCs together accounted for a total of 84.23% of the total variation in the data (Table 3).
Table 3 Component loadings of socio-economic and demographic factors in the Hamilton, Ontario, Canada (2004)
Variable Principal component Uniqueness
1 2 3 4 5
Persons with <grade 9 education 0.575 0.176 0.659 -0.128 0.292 0.103
New immigrants 0.583 0.164 -0.337 0.539 0.243 0.171
Visible minority 0.643 0.419 -0.236 0.361 0.159 0.200
Aboriginal percentage 0.246 0.333 -0.165 -0.788 0.238 0.123
Median income -0.790 -0.200 -0.125 0.378 -0.203 0.136
Government transfer income 0.847 -0.153 0.372 -0.053 0.017 0.118
Low income 0.941 0.110 0.018 -0.026 -0.144 0.081
Non-official language pop 0.675 0.255 0.197 0.342 0.424 0.143
Unemployment rate 0.721 0.103 0.165 0.036 0.064 0.437
Average dwelling value -0.874 0.044 -0.103 0.042 0.203 0.181
Owner-occupied dwellings -0.867 0.099 0.354 -0.096 0.000 0.104
Population under 20 years old -0.324 0.867 0.089 0.045 -0.228 0.082
Population 65 years or older 0.139 -0.861 0.282 0.014 0.217 0.113
Single-parent families 0.759 0.252 0.160 -0.073 -0.437 0.138
Married percent -0.875 0.148 0.190 0.172 0.251 0.084
Percent living alone 0.711 -0.552 -0.337 -0.075 -0.010 0.071
Percent of internal migrants 0.036 0.112 -0.790 -0.280 0.292 0.198
Population density 0.681 -0.113 -0.225 0.101 -0.325 0.357
Eigenvalue 8.30 2.42 1.92 1.47 1.06 -
Percentage of variation explained 46.10 13.42 10.69 8.14 5.88 -
Cumulative % of variation explained 46.10 59.52 70.35 78.35 84.23 -
The first Principal Component is the most important since most (12) of the variables loaded heavily on it (Table 3). The component loadings measure the relationships of the socioeconomic and demographic variables with each of the PCs. The values of the loadings range from -1 to 1. The uniqueness values of almost all the variables were relatively low with the highest being 0.437 and the lowest 0.071. Uniqueness is the percentage of variance for a variable that is not explained by the PCs. For instance, almost all the variations for low-income, population under 20, percent married and percent living alone are explained by the five PCs (Table 3).
Principal component 1 is mainly an economic status component but also had a social component. Neighbourhoods (CTs) with high values of this PC had high proportions of aboriginal people, low median income, high percentage of low-income earners, relatively high percentage of individuals receiving government transfer income, high unemployment rates, low dwelling values, a high percentage of single-parent families, few owner-occupied dwellings, few married people, high percentage of persons living alone and high population density (Figure 1 and Table 3). These neighbourhoods could be described as high risk neighbourhoods because they had high values of most of the undesirable socioeconomic determinants of health.
Figure 1 Principal component 1. Spatial distribution of first principal component extracted in the principal components analysis of socio-economic factors in Hamilton, Ontario, Canada (2004)
Principal component 2 was generally a demographic component. Neighbourhoods with high values of PC 2 had high percentages of children less than 20 years of age, but low percentages of seniors (Figure 2 and Table 3). Neighbourhoods with high values of PC 3 had high percentages of population with less than grade 9 education but low percentages of internal migrants (Table 3 and Figure 3). Finally, neighbourhoods that had high values of the fourth PC, an immigration and aboriginal status component, had high percentage of new immigrants but low population of people of aboriginal origin (Figure 4 and Table 3). The last PC (PC 5) did not load highly on any of the variables but was extracted because its eigenvalue was slightly higher than 1 (Table 3).
Figure 2 Principal component 2. Spatial distribution of second principal component extracted in the principal components analysis of socio-economic factors in Hamilton, Ontario, Canada (2004)
Figure 3 Principal component 3. Spatial distribution of third principal component extracted in the principal components analysis of socio-economic factors in Hamilton, Ontario, Canada (2004)
Figure 4 Principal component 4. Spatial distribution of fourth principal component extracted in the principal components analysis of socio-economic factors in Hamilton, Ontario, Canada (2004)
Cluster analysis
Figure 5 shows the geographical distribution of the identified neighbourhood types (clusters). The detailed descriptive statistics of the socioeconomic and demographic factors for each of the neighbourhood types compared to the entire city of Hamilton is presented in Table 4. Statistical significance tests were performed to compare the characteristics of each of the neighbourhood types with the Hamilton average. Measures that were significantly different from the Hamilton average are described as high or low otherwise they are described as medium.
Figure 5 Neighbourhood types. Spatial distribution of identified neighbourhood types in Hamilton Ontario, Canada (2004)
Neighbourhood Type A is primarily located within an inner ring surrounding the downtown core. These neighbourhoods can be described as "mature" areas (i.e., many seniors) with some indications of transition (i.e., neighbourhood turnover with arrival of new immigrants). Neighbourhood type A consists of 46 census tracts. A high population density, percentage of seniors and low-income earners, and a medium percentage of new immigrants characterize this neighbourhood type. It also has a high percentage of single-parent families, low dwelling values and a medium percentage of persons not able to speak English or French. Approximately 32.7% of Hamilton residents live in this neighbourhood type.
Neighbourhood Type B includes high economic status neighbourhoods in low-density rural or suburban environments. Since it covers the largest geographical area, its low population density has a great impact on the overall population density of Hamilton. These areas constitute the geographic periphery of the city, forming an outer ring from east to west. Neighbourhood type B is comprised of 24 census tracts and has approximately 23% of Hamilton residents. It has high median income, dwelling values, and percentage of owner-occupied dwellings and a medium percentage of seniors. This neighbourhood type also has a low percentage of new immigrants, single-parent families, and individuals with less than grade 9 education.
Neighbourhood Type C represents a relatively high economic status neighbourhoods within a more urban environment. Unlike Neighbourhood Type B, an area similar in income and dwelling value levels, this NT has a relatively high percentage of visible minority groups. It consists of 17 census tracts and is characterized by high median income, and dwelling values. In addition, this NT has few seniors and persons living alone, but a medium percentage of individuals who cannot speak either English or French. It also has high dwelling values and population density. Approximately 16.9% of Hamilton residents live in this type of neighbourhood.
Neighbourhood Type D depicts a "mature" urban area with a high percentage of seniors and owned dwellings. It has a relatively low percentage of low-income earners and a high percentage of individuals with less than grade 9 education. In addition, this neighbourhood type has few recent immigrants and internal migrants, low unemployment rate, high percentage of owner-occupied dwellings and medium percentage of persons not able to speak English or French. It is composed of 20 census tracts and 9.6% of Hamilton residents live in this type of neighbourhood.
Neighbourhood Type E constitutes the inner city core and a few areas scattered in the inner ring around the core, and is comprised of 24 census tracts. It has a high prevalence of low-income earners, new immigrants, visible minority groups, and persons with less than grade 9 education. It also has many single-parent families, those receiving government transfer income and high unemployment rate. Approximately 17.7% of Hamiltonians live in these neighbourhoods. Note that the sum of the population percentages of the groups is not 100% because one census tract was not included in the analysis because of missing data.
Discussion
This study has used multivariate techniques to characterize neighbourhoods based on differences and/or similarities of their socioeconomic and demographic characteristics. The positive correlation between single-parenthood and low-income is consistent with observations from other studies that single-parents generally tend to spend more time in low-income neighbourhoods compared to childless couples and unattached individuals [63]. Moreover, it has also been reported that single-parenthood is common among socially disadvantaged groups and compounds social disadvantage [64]. In low socioeconomic neighbourhoods, people experience barriers in creating and benefiting from social capital, leading to social exclusion. The societal costs of social exclusion are lack of cohesion, higher crime rates, increased pressure on societal services and the stigma associated with particular neighbourhoods. Social exclusion is especially a problem in neighbourhoods with high unemployment rates, low-incomes, poor housing, etc., all of which combine to create a vicious cycle of poverty, low social capital and increased health risks [65].
The negative correlation between housing ownership and visible minority has been reported in other studies [66]. In addition, the observed positive correlation between visible minority and low-income has also been reported in other Canadian studies which reported that visible minority Canadians (people of colour) experience persistent income gap, above average levels of living on low-income and higher levels of unemployment [67]. The high negative correlation of low-income and housing ownership is not surprising and is in agreement with observations by Anderson and co-workers [68] who noted inadequate supply of affordable housing for low-income families and the increasing spatial segregation of some households by income, race, ethnicity, or social class into "unsafe neighbourhoods". Moreover, when affordable housing is not available to low-income households, family resources needed for food, medical or dental care, and other necessities are diverted to housing costs leading to the concept of "concentrated poverty" in certain neighbourhoods [68].
The low uniqueness values of the PCA imply that the five PCs appropriately represent the socioeconomic and demographic variables included in the analysis. Uniqueness values higher than 0.6 are considered high [69]. The advantage of using either PCA or cluster analysis in this kind of study is that they allow incorporation of many variables in the characterization of neighbourhoods. Therefore, from a population health planning perspective, they provide a better understanding of neighbourhood characteristics compared to representations based on only one variable. This is because the health of a population is determined by several socioeconomic, demographic and health care service factors and therefore analyses that incorporate only one variable would provide insufficient information for population health planning purposes.
Choice of the unit of analysis is critical in these kinds of analyses due to the modifiable areal unit problem (MAUP) since choice of a different and/or inappropriate unit could lead to quite different results [70,71]. As has been pointed out by Ross and coworkers (2004) [72], it is more meaningful to use 'naturally' defined neighbourhoods, rather than arbitrary geostatistical or political units since the distribution of population characteristics or health outcomes may not necessarily follow these arbitrary/political boundaries. Ross and co-workers compared the performance of census tracts to more 'natural' neighbourhoods and found very similar results and concluded that census tracts, used as proxies of neighbourhoods in our study, are good proxies for natural neighbourhood boundaries [72].
In this study, the results of the PCA were generally similar to those of cluster analysis since the distribution of areas identified as high risk by PC1 tended to follow similar spatial patterns as the high risk areas identified by cluster analysis. Both methods are therefore useful in identifying neighbourhood socioeconomic characteristics that would enhance health planning. However, as has been pointed out by Luginaah and co-workers [49], interpretation of the results of PCA is difficult due to its subjective nature and the fact that as many maps as number of principal components have to be produced. This makes cluster analysis methodology better for these purposes. Moreover, cluster analysis allows computation of statistics for each of the clusters (neighbourhood types) making the methodology much more objective than PCA.
Similar to the pattern seen in other industrial North American cities [28,73-75] most of the high risk neighbourhoods in this study (i.e. with high percentage of low-income earners, low educational attainment, etc), were located in the downtown core with the risk decreasing towards the suburban environments. The observed diverse neighbourhood socioeconomic characteristics may imply great variability in the health needs of the different population subgroups living in the different neighbourhoods since the conditions in which people live strongly influence their health. Health inequalities are produced by the clustering of several of these socioeconomic risk factors [76]. Therefore, populations living in different neighbourhood types differ in the type and number of socioeconomic risk factors to which they are exposed [77,78]. Although it is obvious that neighbourhood type E has the lowest socioeconomic status and highest risk while neighbourhood type B has the highest status and lowest risk, the intent of this study was not merely to classify the neighbourhoods based on economic status. Rather, this study was intended to generate neighbourhood socioeconomic information on which needs-based health planning and service delivery can be based. There is benefit in targeting improvement strategies to materially and socially deprived groups [79].
Future directions and potential applications
The current study is the first of a series of projects designed to investigate neighbourhood health inequalities and provide information to foster health planning with a view to reducing health inequities. The identified neighbourhood clusters will be used, in subsequent studies, as units of analyses in investigating equity in health status, access and utilization of health services. Additionally, the identified of neighbourhood characteristics are expected to provide useful information on which health planning decisions will be based in order to:
1) Identify population health needs at the neighbourhood level
2) Assess health service utilization patterns across neighbourhoods and compare these with neighbourhood population characteristics and needs
3) Create geographic boundaries for the integrated delivery of social and community health care services
4) Allow for the development of strategies tailored and responsive to the unique characteristics and needs of each neighbourhood.
5) Enhance the use of empirical data for local advocacy for marginalized and under-served neighbourhoods and other populations in need.
Incorporation of the differences in neighbourhood socioeconomic characteristics in population health planning decisions such as decisions on funding allocation to community health agencies will help ensure that health planning strategies are best tailored to address the unique needs of each population. This is because a "one-size-fits-all" planning approach is neither efficient nor practical due to the different socioeconomic and demographic characteristics of the different neighbourhoods. It is expected that inclusion of neighbourhood socioeconomic and demographic characteristics in population health planning will provide health planners with more evidence to guide needs-based decisions that would be more appropriate for the socioeconomically diverse neighbourhoods. Therefore, it is hoped that the results of these analyses will be useful in ensuring that planning is tailored to the unique needs of the different neighbourhood population groups. For instance, neighbourhood types A and D have very similar median incomes and therefore if income was the only variable used to characterize the neighbourhoods, they would be treated similarly. However, the rest of the characteristics of these neighbourhood types are different. For example, neighbourhood type D has a much lower percentage of new immigrants, visible minority population and single-parent families than neighbourhood type A. Moreover, there are significantly more owner occupied dwellings in neighbourhood type D than A. The implication is that these neighbourhoods have potentially different challenges and health needs. If only median income was used (as is most often done) to classify the neighbourhoods, the two NTs would inevitably erroneously be treated as similar. Planning strategies based on such single variable analysis may not be appropriate since the strategies would not be tailored to the unique characteristics and therefore needs of the NTs.
Conclusion
In this study, we have used multivariate techniques to identify unique neighbourhood characteristics and classify the neighbourhoods into groups with similar characteristics. Since the identified neighbourhood types are homogeneous with respect to the broad determinants of health, they offer potentially excellent opportunities for health planners and service providers to understand the characteristics and potential health needs of the different neighbourhoods and therefore better plan for them. Through continuous monitoring of health information across these neighbourhoods, health planners, service providers and policy makers could better make decisions based on knowledge of the local communities.
Authors' contributions
AO was involved in the study design, execution and writing up of the draft and final copies of the manuscript. RW conceived the need for the study, was involved in the study design and preparation of the manuscript. ME, SB, BH, JE, and TA participated in the study design, guiding implementation and preparation of the draft of the manuscript.
Table 4 Summary statistics of socio-economic and demographic features of identified neighbourhood types in Hamilton, 2004. Data source: Statistics Canada, 2001 census. NT = Neighbourhood type. The sum of the populations from the 5 neighbourhood types is not 490,268 due to missing data in one census tract.
Variable Means
Hamilton NT A NT B NT C NT D NT E
Persons with <grade 9 education (%) 10.3 11.3 4.7 9.2 13.7 16.9
New immigrants (%) 3.2 3.4 0.9 2.7 1.2 6.6
Visible minority (%) 10.9 9.4 4.2 14.4 5.6 20.7
Aboriginal persons (%) 2.2 2.6 1.0 1.6 3.0 4.1
Median income ($) 22927 22262 30351 25958 22837 16250
Government transfer income (%) 12.1 15.9 7.3 8.3 13.4 23.2
Low income (%) 19.8 21.7 5.6 12.7 10.7 42.4
Persons not speaking English or French (%) 1.8 1.6 0.33 1.6 1.6 4.4
Unemployment rate (%) 6.4 5.7 2.2 4.2 3.5 10.1
Average dwelling value ($) 166783 125271 224342 173801 162157 89174
Owner-occupied dwellings (%) 65.2 61.9 87.4 81.8 86.7 34.7
Population under 20 years old (%) 26.1 23.5 27.5 33.6 21.2 23.6
Population 65 years or older (%) 14.3 17.2 13.2 7.2 21.0 14.3
Single-parent families (%) 16.6 19.9 9.5 15.4 10.7 25.7
Married (%) 51.7 47.3 60.9 58.9 58.5 35.7
Live alone (%) 10.3 13.3 5.4 2.9 8.7 21.1
Internal migrants (%) 10.6 9.2 14.5 7.7 6.9 12.2
Population Density (No. of persons per Km2) 438.9 3627.4 357.1 2543.2 2352.1 5706.0
Population (count) 490268 160438 112875 82976 47137 86683
Acknowledgements
We thank all stakeholders who participated in the consultative meetings to discuss the study findings and appropriateness of findings for health planning.
==== Refs
Kindig D Stoddart G What is population health? Am J Public Health 2003 93 380 383 12604476
Yu IT Wong TW Wong SL Small area variations of cancer mortality in Hong Kong – the roles of health care and socio-economic status Neoplasma 2004 51 144 149 15190424
Singh GK Miller BA Health, life expectancy, and mortality patterns among immigrant populations in the United States Can J Public Health 2004 95 I14 I21 15191127
Paulander J Axelsson P Lindhe J Association between level of education and oral health status in 35-, 50-, 65- and 75-year-olds J Clin Periodontol 2003 30 697 704 12887338 10.1034/j.1600-051X.2003.00357.x
Epstein AM Health care in America – still too separate, not yet equal N Engl J Med 2004 351 603 605 15295055 10.1056/NEJMe048181
Hnizdo E Sullivan PA Bang KM Wagner G Association between chronic obstructive pulmonary disease and employment by industry and occupation in the US population: a study of data from the Third National Health and Nutrition Examination Survey Am J Epidemiol 2002 156 738 746 12370162 10.1093/aje/kwf105
Bartosinska M Ejsmont J [Health condition of employees exposed to noise – extra auditory health effects] Stan zdrowia pracownikow narazonych na halas – skutki zdrowotne pozasluchowe Wiad Lek 2002 55 20 25 15002214
Blau G Tatum DS McCoy K Dobria L Ward-Cook K Job loss, human capital job feature, and work condition job feature as distinct job insecurity constructs J Allied Health 2004 33 31 41 15053218
Garbarino S De Carli F Mascialino B Beelke M Nobili L Squarcia S Penco MA Ferrillo F Sleepiness in a population of Italian shiftwork policemen J Hum Ergol (Tokyo) 2001 30 211 216 14564884
Shields M Shift work and health Health Rep 2002 13 11 33 15069802
Dearry A Impacts of our built environment on public health Environ Health Perspect 2004 112 A600 A601 15289172
Kasmel A Helasoja V Lipand A Prattala R Klumbiene J Pudule I Association between health behaviour and self-reported health in Estonia, Finland, Latvia and Lithuania Eur J Public Health 2004 14 32 36 15080388 10.1093/eurpub/14.1.32
Ndubani P Bond V Liljestrom R Hojer B Understanding young men's sexual health and prospects for sexual behaviour change in rural Zambia Scand J Public Health 2003 31 291 296 15099035 10.1080/14034940210164975
Cree MW Juby AG Carriere KC Effect of home care service levels on health outcomes in hip fracture patients in Alberta Health 2004 7 49 53
Gerrard N Kulig J Nowatzki N What doesn't kill you makes you stronger: determinants of stress resiliency in rural people of Saskatchewan, Canada J Rural Health 2004 20 59 66 14964928
Gil KM Anthony KK Carson JW Redding-Lallinger R Daeschner CW Ware RE Daily coping practice predicts treatment effects in children with sickle cell disease J Pediatr Psychol 2001 26 163 173 11259518 10.1093/jpepsy/26.3.163
Ellaway A Macintyre S Does where you live predict health related behaviours?: a case study in Glasgow Health Bull 1996 54 443 446
Macintyre S Ellaway A Neighbourhood cohesion and health in socially contrasting neighbourhoods: implications for the social exclusion and public health agendas Health Bull 2000 58 450 456
Krueger PM Bond Huie SA Rogers RG Hummer RA Neighbourhoods and homicide mortality: an analysis of race/ethnic differences J Epidemiol Community Health 2004 58 223 230 14966236 10.1136/jech.2003.011874
Macintyre S Maciver S Sooman A Area, class and health: should we be focussing on places or people? Journal of Social Policy 1993 22 213 234
Massey DS American Apartheid: Segregation and the Making of the Underclass The American Journal of Sociology 1990 96 329 357 10.1086/229532
Massey DS Eggers ML The Ecology of Inequality: Minorities and the Concentration of Poverty, 1970–1980 The American Journal of Sociology 1990 95 1153 1188 10.1086/229425
Massey DS Denton NA Suburbanization and Segregation in U.S. Metropolitan Areas The American Journal of Sociology 1988 94 592 626 10.1086/229031
Myles J Picot G Pyper W Neighbourhood inequality in Canadian cities 2000 Ottawa, Ontario, Canada: Statistics Canada, Analytical Studies Branch
Kirby JB Kaneda T Neighborhood socioeconomic disadvantage and access to health care J Health Soc Behav 2005 46 15 31 15869118
Marmot M Social determinants of health inequalities Lancet 2005 365 1099 1104 15781105
Arntzen A Nybo Andersen AM Social determinants for infant mortality in the Nordic countries, 1980–2001 Scand J Public Health 2004 32 381 389 15513672 10.1080/14034940410029450
Jargowsky PA Poverty and Place: Ghettos, Barrios and the American city 1997 New York, USA: Rusell Sage Foundation
Denton NA Massey DS Patterns of neighborhood transition in a multiethnic world: U.S. metropolitan areas, 1970–1980 Demography 1991 28 41 63 2015944
Massey DS Denton NA Hypersegregation in U.S. metropolitan areas: black and Hispanic segregation along five dimensions Demography 1989 26 373 391 2792476
Massey DS The age of extremes: concentrated affluence and poverty in the twenty-first century Demography 1996 33 395 412 8939412
MacLachlan I Sawada R Measures of inequality and social polarization in Canadian metropolitan areas The Canadian Geographer 1997 41 377 397
Lee KK Urban Poverty in Canada 2000 Ottawa, Ontario, Canada: Canadian Council on Social Development
Henderson C Diez Roux AV Jacobs DRJ Kiefe CI West D Williams DR Neighbourhood characteristics, individual level socioeconomic factors, and depressive symptoms in young adults: the CARDIA study J Epidemiol Community Health 2005 59 322 328 15767387 10.1136/jech.2003.018846
Stafford M Cummins S Macintyre S Ellaway A Marmot M Gender differences in the associations between health and neighbourhood environment Soc Sci Med 2005 60 1681 1692 15686801 10.1016/j.socscimed.2004.08.028
Drukker M Kaplan C van Os J Residential instability in socioeconomically deprived neighbourhoods, good or bad? Health Place 2005 11 121 129 15629680 10.1016/j.healthplace.2004.02.002
Diez-Roux AV Borrell LN Haan M Jackson SA Schultz R Neighbourhood environments and mortality in an elderly cohort: results from the cardiovascular health study J Epidemiol Community Health 2004 58 917 923 15483307 10.1136/jech.2003.019596
Stafford M Martikainen P Lahelma E Marmot M Neighbourhoods and self rated health: a comparison of public sector employees in London and Helsinki J Epidemiol Community Health 2004 58 772 778 15310804 10.1136/jech.2003.015941
Schneiders J Drukker M van der Ende J Verhulst FC van Os J Nicolson NA Neighbourhood socioeconomic disadvantage and behavioural problems from late childhood into early adolescence J Epidemiol Community Health 2003 57 699 703 12933776 10.1136/jech.57.9.699
Stafford M Marmot M Neighbourhood deprivation and health: does it affect us all equally? Int J Epidemiol 2003 32 357 366 12777420 10.1093/ije/dyg084
Winkleby MA Cubbin C Influence of individual and neighbourhood socioeconomic status on mortality among black, Mexican-American, and white women and men in the United States J Epidemiol Community Health 2003 57 444 452 12775792 10.1136/jech.57.6.444
Sundquist J Malmstrom M Johansson SE Cardiovascular risk factors and the neighbourhood environment: a multilevel analysis Int J Epidemiol 1999 28 841 845 10597980 10.1093/ije/28.5.841
Blake BJ Bentov L Geographical mapping of unmarried teen births and selected sociodemographic variables Public Health Nurs 2001 18 33 39 11251871 10.1046/j.1525-1446.2001.00033.x
Pettit KLS Kingsley GT Coulton CJ Neighborhoods and Health: Building Evidence for Local Policy 2003 Washington DC, USA: The Urban Institute submitted to The Office of the Assistant Secretary for Planning and Evaluation U.S. Department of Health and Human Services
Odoi A Martin SW Michel P Middleton D Holt J Wilson J Investigation of clusters of giardiasis using GIS and a spatial scan statistic Int J Health Geogr 2004 3 11 15176979 10.1186/1476-072X-3-11
Lorant V Thomas I Deliege D Tonglet R Deprivation and mortality: the implications of spatial autocorrelation for health resources allocation Soc Sci Med 2001 53 1711 9 11762895 10.1016/S0277-9536(00)00456-1
Malmstrom M Sundquist J Johansson SE Neighborhood environment and selfreported health status: a multilevel analysis Am J Public Health 1999 89 1181 6 10432903
Pampalon R Raymond G A deprivation index for health and welfare planning in Quebec Chronic Dis Can 2000 21 104 13 11082346
Luginaah I Jerrett M Elliott S Eyles J Parizeau K Birch S Abernathy T Veenstra G Hutchinson B Giovis C Health profiles of Hamilton: Spatial characterisation of neighbourhoods for health investigations GeoJournal 2001 53 135 147 10.1023/A:1015724619845
Diez-Roux AV Kiefe CI Jacobs DR JrHaan M Jackson SA Nieto FJ Paton CC Schulz R Area characteristics and individual-level socioeconomic position indicators in three population-based epidemiologic studies Ann Epidemiol 2001 11 395 405 11454499 10.1016/S1047-2797(01)00221-6
Hamilton District Health Council Hamilton Health System Fact Book 2004 Hamilton, Ontario, Canada: Hamilton District Health Council
Statistics Canada 2001 Census Dictionary 2003 Ottawa, Ontario, Canada. Ministry of Industry
Frenette M Picot G Sceviour R How long do people live in low-income neighbourhoods? Evidence for Toronto, Montreal and Vancouver 2004 Ottawa, Canada: Statistics Canada
Population Health Approach: What Determines Health?
Statistics Canada Health Region Peer Groups 2002 Ottawa, Ontario, Canada. Statistics Canada
Stata Corporation STATA Version 8.2 2004 4905 Lakeway Drive, College Station, Texas, USA
Everitt BS Landau S Leese M Cluster Analysis 2001 Oxford, United Kingdom: Oxford University Press
Norman GR Streiner DL Biostatistics: The Bare Essentials 2000 B C Decker Inc
Dunteman GH Principal components analysis Quantitative Applications in the Social Sciences Series, No 69 1989 Thousand Oaks, California, USA: Sage Publications
Kaiser HF The varimax criterion for analytic rotation in factor analysis Psychometrika 1958 23 187 200
Caliński T Harabasz J A dendrite method for cluster analysis Communications in Statistics 1974 3 1 27
ESRI ArcView GIS. Version 3.2 1999 Redlands, California, USA: Environmental Systems Research Institute Inc
Frenette M Picot G Sceviour R How long do people live in low-income neighbourhoods? Evidence for Toronto, Montreal and Vancouver 2004 Ottawa, Canada: Statistics Canada
Martin SP Delayed marriage and childbearing: implications and measurement of diverging trends in family timing 2002 Maryland, USA: Department of Sociology and Maryland Population Research Center, University of Maryland, College Park
What is Social Exclusion?
Krivo LJ Kaufman RL Housing and wealth inequality: racial-ethnic differences in home equity in the United States Demography 2004 41 585 605 15461016 10.1353/dem.2004.0023
Galabuzi GE Canada's creeping economic apartheid 2001 Toronto, Ontario, Canada: CSJ Foundation for Research and Education
Anderson LM Charles JS Fullilove MT Scrimshaw SC Fielding JE Normand J Providing affordable family housing and reducing residential segregation by income. A systematic review Am J Prev Med 2003 24 47 67 12668198 10.1016/S0749-3797(02)00656-6
Stata Corporation Cluster Analysis 2003 College Station, Texas, USA: Stata Press
Openshaw The modifiable areal unit problem. CATMOG No. 38 1984 Norwich: Geo Books
Steel GG Holt D Rules for random aggregation Environ Planning A 1996 28 957 978
Ross NA Tremblay SS Graham K Neighbourhood influences on health in Montréal, Canada Soc Sci Med 2004 59 1485 1484 15246176 10.1016/j.socscimed.2004.01.016
Allard SW Access to social services: The Changing Urban Geography of Poverty and Service Provision 2004 Washington DC, USA: The Brookings Institute
Wilson WJ The Truly Disadvantaged 1987 Chicago, USA: University of Chicago Press
Wilson WJ When Work Disappears: The World of the New Urban Poor 1996 New York, USA: Alfred A Knopf
Shaw M Dorling D Gordon D Davey SG The widening gap: health inequalities and policy in Britain 1999 Bristol, United Kingdom: The Policy Press
Shaw M Dorling D Davey SG Marmot MG, Wilkinson RG Poverty, social exclusion and minorities Social Determinants of Health edition 1999 Oxford: Oxford University Press
Kunst AE Groenhof F Mackenbach JP Health EW Occupational class and cause specific mortality in middle aged men in 11 European countries: comparison of population based studies. EU Working Group on Socioeconomic Inequalities in Health BMJ 1998 316 1636 1642 9603745
Martinez J Pampalon R Hamel D Deprivation and stroke mortality in Quebec. Chronic Diseases in Canada 2003 Ottawa, Ontario, Canada: Population and Public Health Branch, Health Canada
|
16092969
|
PMC1201163
|
CC BY
|
2021-01-04 16:39:06
|
no
|
Int J Health Geogr. 2005 Aug 10; 4:20
|
utf-8
|
Int J Health Geogr
| 2,005 |
10.1186/1476-072X-4-20
|
oa_comm
|
==== Front
J Immune Based Ther VaccinesJournal of Immune Based Therapies and Vaccines1476-8518BioMed Central London 1476-8518-3-51611148510.1186/1476-8518-3-5ReviewThe significance of glucose, insulin and potassium for immunology and oncology: a new model of immunity Hill Albert F [email protected] William J [email protected] Darcy B [email protected] Hill Medical, LLC, 1755 Monaco Parkway, Denver, CO. 80220-1644, USA2 Rejuvenon Corporation, 621 Shrewsbury Ave., Shrewsbury NJ, 07702, USA3 Torrey Pines Institute for Molecular Studies, 3550 General Atomics Court, San Diego, CA, 92121-1122. USA2005 19 8 2005 3 5 5 16 6 2005 19 8 2005 Copyright © 2005 Hill et al; licensee BioMed Central Ltd.2005Hill 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 recent development in critical care medicine makes it urgent that research into the effect of hormones on immunity be pursued aggressively. Studies have demonstrated a large reduction in mortality as a result of infusion with glucose, insulin and potassium. Our work in the oncology setting has led us to propose that the principal reason for such an effect is that GIK stimulates lymphocytes to proliferate and attack pathogens, sparing the patient the stress of infection. That suggestion is based on a new model of immunity that describes the effect of hormones on lymphocytes. We hypothesized that the application of glucose, insulin, thyroid and potassium would awaken inert tumor infiltrating lymphocytes to destroy the tumor.
Methods
The antitumor effect of a thyroxine, glucose, insulin, and potassium (TGIK) combination was studied in a series of controlled experiments in murine models of tumor progression to assess the biologic activity of the formulation, the effect of route of administration, the effect on tumor type, and the requirement for insulin in the TGIK formulation.
Results
Melanoma and colon tumors inoculated with TGIK were significantly reduced in size or retarded in growth compared to controls injected with saline. I.P. and I.M. injections showed that the formulation had no effect systemically at the doses administered.
Conclusion
We conclude that TGIK has anti-tumor activity when administered intratumorally, probably by stimulating lymphocytes to attack tumors. This is similar to the effect of GIK on reducing sepsis in critical care patients. We suggest that when GIK is administered exogenously, it restores immune competence to the critically ill or cancer patient and causes destruction of pathogens or tumors, while endogenous resources are devoted to repair. This implies that hormonal therapy may be useful in treating various other pathologies involving immune suppression, as well as malignancies. We also propose research that could bring resolution of the controversy over mechanism and point the way to new therapeutic strategies for numerous diseases including chronic infections and auto-immune diseases.
==== Body
Background
In a turnaround from the usual laboratory research-to-clinical usage sequence, critical care has become the focus for one of the most interesting developments in medicine: the use of glucose, insulin and potassium (GIK) in treating the critically ill. Van den Berghe et al., in a landmark study, demonstrated a 46% reduction in mortality [1]. Krinsley, with a less aggressive protocol, produced similar results [2]. Since the greatest reduction was in deaths due to multiple-organ failure with a septic focus, the implications for immunology could be significant. Steinman and Mellman recently made a strong case that only research in human beings can advance our understanding of the human immune system [3]. The discoveries involved in the use of GIK supports that. It has been known for years that lymphocytes have receptors for numerous hormones and neurotransmitters, but that fact is seldom incorporated into models of the immune system [4]. Impressive progress has been made in many areas of immunology, particularly in the ways cells communicate with and affect each other. Now the success of GIK suggests that a hormone, insulin, strongly enhances the immune response. The time has come to examine more closely the role endocrine hormones play in regulating immunity. Deciphering the mechanism of GIK is crucial, not only for critical care, but also for a better understanding of immune response mechanisms.
Van den Berghe first speculated that strict glycemic control provided the beneficial effect of GIK; more recently she has suggested that the most important benefit may be from the "powerful anti-inflammatory effect" of insulin. Hyperglycemia can contribute to inflammation, and insulin has anti-inflammatory properties (e. g. inhibiting production of tumor necrosis factor-alpha and super-oxide radicals, macrophage migration inhibitory factor) [5-7], and TNFα and IL-1 have been shown to depress myocardial function in a dose-dependent fashion [8]. Still, it is unlikely that inflammation is producing the deleterious effects in the critically ill. IL-1, which is so central in inflammation, is known to suppress the expression of insulin-like growth factor-1 [9]. Yet Van den Berghe found levels of IGF-1 to be high in her patients, particularly those near death. Also, inflammation is an early, indispensable part of a robust immune response. Without phagocytes ingesting pathogens, presenting antigen and releasing cytokines, lymphocytes would not become activated effector cells. Infection would rage unabated. To be maximally effective, the immune sequence must move from the inflammatory to the acquired, lymphocytic phase. A remarkable aspect of immunity is the way the body selects and produces the right response to a given challenge. If an infection is contained, inflammation will be chosen as the appropriate defense, and the cytokines released will actually restrain the expansion of lymphocyte clones. If the response must proceed from inflammation to the adaptive phase, cytokines from damaged tissue, macrophages and dendritic cells instruct CD4 cells to become Th1 or Th2 cells, according to which kind of lymphocyte, CTL or B cell, is needed. Cytokines released by those cells then restrain inflammation but advance the lymphocyte response. For example, Interleukin 6, which is both pro- and anti-inflammatory at times, promotes proliferation of CD8 cells, and suppresses inflammation by down-regulating TNF-α. IL-1 and chemokine expression [10]. Interleukin 4, produced by Th2 cells also suppresses the production of Il-1, TNF-α, and chemokines [11]. Interleukin 10, another anti-inflammatory Th2 cytokine, down-regulates synthesis of IL-1, IFN-γ, IL-2, TNF-α [12].
Cytokines also have a powerful effect on metabolism. Il-6 and TNF-α cause loss of skeletal muscle protein and lean tissue wasting, insulin resistance, increased glucogenesis, increased lipolysis in adipose tissue, and development of cachexia [13]. These changes provide a rich substrate for use by dividing immune cells. The body will also increase the secretion of endocrine hormones that will further enhance the expansion of the cells needed for the particular challenge. For example, insulin will suppress inflammation but, as we shall see, it will also stimulate a rapid expansion of lymphocyte clones. It has been known for decades that following trauma, hyperglycemia without increased insulin secretion occurs [14-16], and that the degree of hyperglycemia is correlated with the severity of the injury [17,18]. We therefore suggest that hyperglycemia is the normal response of the body as it tries to make nutrients available for the repair of damaged tissues. If, after a trauma or inflammation, systemic infection occurs, insulin will rise as the body supports the expansion of lymphocyte clones. (see below)
Years ago it was discovered and confirmed that insulin powerfully enhances the capacity of cytotoxic T lymphocytes in vitro to kill targets bearing the sensitizing antigen [19] and to do so in a dose-dependent manner within the physiological range [20,21]. While circulating quiescent lymphocytes have no detectable insulin receptors, once they have received antigenic challenge, they acquire approximately 6,000 per cell [22-26]. Since acquisition of these receptors is an early event in cellular transformation, it seems probable that the emergent insulin receptors are a prerequisite for, rather than a consequence of cell enlargement and subsequent cell division [27-29]. Insulin is, therefore, an immuno-regulatory hormone [30].
The effect of insulin on lymphocytes becomes significant when seen as part of the profile of events when a body is challenged by infection. More than twenty years ago Beisel mapped the response of the body to an infectious challenge [31]. He showed that the first detectable response was phagocytic activity, followed by increased secretion of glucocorticoids and growth hormone, deiodination of thyroxine, secretion of acute phase proteins, carbohydrate intolerance, increased secretion of aldosterone and ADH and eventually an increased secretion of thyroxine. One of his many contributions included the discovery that IL-1 (then called Leukocyte Endogenous Mediator) also acts as a hormone, stimulating uptake of amino acids and increasing synthesis of acute phase reactants [32]. Beutler et al., pointed out that the inflammatory cytokine, Tumor Necrosis Factor (TNF), once called cachectin, suppresses lipoprotein lipase, and causes peripheral tissues to lose nutrients [33]. The net effect of this is to mobilize energy reserves and make them available to dividing inflammatory and immune cells [34].
Rayfield and associates studied the effect of acute endotoxemia on volunteers and showed that during the febrile phase of an infection insulin increases to three times basal levels (35 ± 5 μU/ml) and, paradoxically, glucagon increases to five times normal [35]. Other investigators have confirmed this threefold rise in insulin during an infection [36,37]. In this "Infectious Mode," lymphocytes produce insulin receptors at the very time the hormone is rising in the blood, and are able to bind it and acquire glucose. But if insulin is low in the blood, even lymphocytes displaying insulin receptors cannot activate. The rise in glucagon assures a supply of glucose for the expanding clone of lymphocytes. They are then able to pump ions, which, we shall see, is the sine qua non of full lymphocyte activation. Insulin and thyroid increase the activity of the sodium potassium pump [38].
The endocrine mix produced after an infection or trauma, when the body is repairing damaged tissues, is quite different. In this "Healing Mode," insulin levels drop to normal or lower levels, counter-regulatory hormones such as growth hormone and cortisol continue to be high [39], and the liver increases production of insulin-like-growth-factor-1 (IGF-1). IGF-1 and autocrine growth factors enable the dividing reparative tissues to acquire nutrients from the blood even as peripheral tissues are starved. Thus, the body cannibalizes peripheral tissues for the sake of repairing the wound [40]. This endocrine mix is powerfully immuno-suppressive, as all the body's resources are devoted to repair. The degree of hyperglycemia and IGF-1 are indices of the degree of injury. Van den Berghe found that rising IGF-1 levels predict mortality accurately [41].
When a patient is critically ill, the body responds quickly with "...a highly coordinated and powerful acute phase reaction, whereby the immune system is switched from the adaptive mode of response to the amplification of natural immune mechanisms." "The increased serum level of cytokines and the array of neuroendocrine changes lead to fever, catabolism and to the suppression of the T lymphocyte-dependent adaptive immune system. At the same time natural immune mechanisms are amplified" [42]. If pathogens are present, lymphocytes will later enter the battle. However, if the injury itself is life-threatening, we propose the body will not proceed to the next phase of supporting the expansion of lymphocyte clones but instead will move into the Healing Mode, described above, so that all bodily resources can be devoted to repair of damaged tissues. In this environment, inflammation can continue, sometimes with destructive force, but there can be no significant involvement by lymphocytes because insulin is too low. Immune competence in the seriously wounded patient is severely reduced.
Therefore we propose that it is not inflammation per se that harms the critically ill patient; it is the incapacity of the body to complete the immune sequence and protect itself against infection. Exogenous GIK enables inert lymphocytes to proliferate and perform cytotoxic tasks, even as endogenous resources are devoted to repair of tissues.
As evidence of how GIK stimulates immunity in vivo, we offer this. A few years ago, we developed a new model of immunity that incorporates the effects of endocrine hormones and neurotransmitters on lymphocytes. Lymphocytes are chemotactically attracted to a tumor and actually invade it (TILs), but they do little damage. Some of that failure is due to the immunosuppressive effect of autocrine growth factors produced by the tumor (e.g. Transforming Growth Factor beta (TGFβ) [43]. But there is more to the problem: in a tumor-bearing animal, the suppression is systemic [44].
We proposed that the brain of a tumor-bearing animal is "deceived" by growth factors released by the tumor. The brain treats the malignancy as if it were a healing wound and commands an endocrine mix to support growth and suppress immunity. The mix features decreased levels of insulin and increased amounts of counter-regulatory hormones. Peripheral tissues become insulin resistant and lose nutrients into the blood, sometimes producing hyperglycemia and eventually the familiar cachexia of the cancer patient. The dividing tumor cells (like those involved in repair of damaged tissue) can utilize the materials lost by peripheral tissues, because they produce autocrine growth factors [45]. And, again, the liver increases production of IGF-1. As does a healing wound, the tumor cannibalizes the body for the materials it needs to grow [46].
As mentioned above, when the lymphocyte is deprived of high levels of insulin, it cannot acquire glucose and the sodium/potassium pump cannot restore ionic integrity. With its stores of potassium reduced, the lymphocyte cannot complete its enzymatic actions and transform or proliferate. This effect on the sodium/potassium pump is crucial; at every point in a lymphocyte's activation and proliferation, and in the performance of its function, the cell loses its surface charge, ion channels open, potassium escapes and sodium rushes in, down the electro-chemical gradient [47-49]. Before the lymphocyte can proceed in its cycle, it must replenish stores of potassium [50-52]. If it is 20% deficient in that ion, it cannot continue its cycle of mitosis or perform its function [53]. Yet cancer patients are as much as 40% deficient in total body potassium [54]. It is also significant that when insulin is administered i.v. and blood levels rise to three times normal, potassium moves into the cells [55,56].
We hypothesized that if a cancer patient were to be administered thyroid and insulin (to stimulate the sodium/potassium pump), glucose and potassium (TGIK), all in quantities to mimic those reached during an infectious challenge, inert lymphocytes would activate and destroy a tumor.
Presented here are partial results from controlled studies with mice. At the request of investors, Hill Medical has not heretofore published any results.
Methods
Melanoma cells were injected into mice, and when the tumors became palpable they were inoculated with TGIK or saline solution. In another study mice were injected with only part of the combination to determine if insulin were necessary, or if irritation by potassium were producing the results. In further experiments the formula was tested by injecting I.M. and I.P. Still another tested the effect of the formulation on colon cancer.
Experiment 1
Five groups of C57BL/6 mice (ten mice per group) were injected subcutaneously on Day 1 with murine melanoma B16-F10 cells (1.8 × 106 cells) in the ventral aspect of the right hind limb. Injections with saline control and the TGIK formulation were begun on Day 6. Each milliliter of the TGIK formulation contained: insulin 3U, sodium thyroxine 50 μg, KCl 8 μEq, and glucose 50 mg. Tumor dimension (average length × average width) was determined on Days 10, 11, 13, 15, 17, and 19 and the results are indicated in Figure 1.
Figure 1 Antitumor activity against murine melanoma B16-F10 in C57BL/6 mice following TGIK administration via different routes of administration.
The results shown in Figure 1 demonstrate the antitumor efficacy of TGIK when administered by twice-daily intratumoral injection. Systemic administration (IP or SC) at these doses did not appear to offer any therapeutic benefit. The experimental design however, did not fully assess the possibility of a dose response relationship and consequently a potential benefit from larger doses administered systemically cannot be ruled out.
Experiment 2
In order to determine whether the combination of all four ingredients of the TGIK formulation was required, and specifically to rule out the possibility that the antitumor effects observed in Experiment 1 were due only to an irritant effect of potassium, an experiment was conducted using the B16-F10 melanoma line in C57BL/6 mice in which the complete TGIK formulation was compared against GK and TGK as well as a saline control.
The results shown in Figure 2 demonstrate the activity of intratumoral TGIK and the finding that the formulation is rendered ineffective by removal of insulin. Consequently, this experiment demonstrates that the antitumor activity of TGIK is not due to an irritant effect from KCl alone.
Figure 2 Antitumor activity against murine melanoma B16-F10 in C57BL/6 mice following administration of TGIK in comparison to incomplete formulations.
Figure 3 shows an incidental finding of this study. There was a reduction in mortality in the TGIK group relative to the other treatments.
Figure 3 Mortality resulting from murine melanoma B16-F10 in C57BL/6 mice following administration of TGIK in comparison to incomplete formulations.
Experiment 3
Two additional groups of mice were injected with tumor cells in both hind limbs with only one hindlimb receiving subsequent TGIK injections to assess whether there was any effect on the contralateral tumor. The results are indicated in Figure 4.
Figure 4 Antitumor activity against murine melanoma B16-F10 in C57BL/6 mice following administration of TGIK into the tumor site in comparison to growth in the contralateral tumor site.
Figure 4. In contrast to the potent antitumor activity of the formulation injected directly into the tumor site, there was no evidence of effect on the contralateral tumor site.
Experiment 4
These experiments were conducted in an analogous fashion to Experiment 1 except that the tumor line studied was the CT26 colon carcinoma line, the mouse model was the BALB/c mouse, and the tumor injection was of 50–100,000 cells per injection. Only the IT route of TGIK administration was evaluated. Because the tumors formed were more indurated, the mice were shaved to improve measurement determinations. The results of this experiment are presented in Figure 5.
Figure 5 Antitumor activity against murine colon carcinoma CT26 in Balb/C mice following administration of TGIK.
As can be seen from Figure 5, TGIK is active against murine colon carcinoma cells, although the effect is somewhat more modest than its demonstrated activity against murine melanoma cells, perhaps a consequence of the slower growth rate of the colon carcinoma cell line. The colon carcinoma tumors tended to be more nodular and grow into deeper tissues making the tumor size more difficult to assess.
Conclusions from the Preclinical Pharmacology Controlled Experiments
• The purpose for creating this model was to develop a more effective treatment for cancer. The aim of this series of controlled experiments was to prove that the cocktail would have anti-cancer activity. We realize these experiments do not prove the mechanism was immunological. However, the data produced in these experiments and in the low-dose human trials described below strongly suggest that immunity is the mechanism. An in vitro study in which tumor cells are exposed to the hormone cocktail without lymphocytes present would help to settle the issue. Also, a trial with nude mice would give more credence to immunity as the effective agent if the tumor's growth in that animal is not retarded, but those studies are not feasible for us at this time.
However, we believe the following conclusions are justified
• TGIK demonstrates potent antitumor activity against murine cancer cell lines transplanted into murine models
• Insulin is a required component of the TGIK formulation
• At the doses and regimens studied, antitumor activity is mediated by a direct response within the tumor without evidence of a systemic response affecting distant sites
Preliminary human trials
Early low-dose Phase I trials for Hill Medical, using one injection of long lasting insulin per day with other materials administered orally, produced large rises in the CD4/CD8 ratio, with one patient reaching 71:1. Levels for normal patients are 3:1, for cancer patients ca. 2:1 or lower, and for AIDS patients much lower. More trials, better controlled, with higher doses of all materials administered intravenously, and with frequent measurements of blood insulin are in the planning stage. It is of interest that a psychiatrist in the 1950s administered a modified insulin shock treatment to two depressed cancer patients and the patients' tumors disappeared [57].
Discussion
Great progress has been made in understanding the factors that regulate immunity. Immunologists have identified cytokines that up- or down-regulate immune functions. Others have created effective vaccines. Yet vaccines cannot be created for many diseases. Attempts to stimulate the immune system with cytokines to attack tumors have been disappointing. The doses most effective are unacceptably toxic [58]. But just as dreams of stimulating the immune system to attack tumors or more effectively deal with pathogens seem to be fading, there comes news of the surprisingly beneficial effect of GIK in treating the critically ill. Already both the American College of Cardiology and the American Heart Association have recommended that intravenous GIK be given to patients with acute myocardial infarction, even though the mechanism is still controversial. Since GIK apparently provides no benefit for patients with heart failure [59], we think it unlikely that the major benefit comes from a direct action on the heart.
We have proposed that GIK provides benefit to the critically ill patient because it stimulates lymphocytes. As the adaptive phase intensifies, activated lymphocytes release cytokines (IL-4, Il-10) [60] that down-regulate inflammation. Because septic shock is still the most common cause of death in the Intensive Care Unit, is the 10th leading cause of death overall, has increased 86% between 1979 and 1997, and costs $5–10 billion for treatment, an effective prophylactic or treatment is urgently needed. We propose that GIK (and TGIK) are capable of protecting the patient against what are probably hospital-acquired infectious agents.
Van den Berghe also reported a reduction in critical illness polyneuropathy among her patients receiving GIK [61]. That syndrome is more likely due to a pre-existing, smoldering infection by an unidentified pathogen. Flare-ups of chronic, often unperceived, infections when a patient is immune-compromised as from the stress of surgery or serious injury are common. Inflammation is being implicated in more and more diseases, from Alzheimer's [62] to cancer, [63] and to autoimmune diseases such as lupus and diabetes [64]. But we propose that if patients threatened with polyneuropathy benefit from GIK, it is not because GIK reduces inflammation per se. It is due to GIK stimulating lymphocytes to efficiently remove the offending pathogen and to down-regulate inflammation with appropriate cytokines. In a recent discussion of the ideal treatment for Chlamydia, Ojcius, Darville and Bavoil have proposed that any intervention should evoke just enough inflammation to help the body's other immune defenses eliminate the bacteria [65]. In our model that is what happens when high doses of GIK are administered intravenously for a period of several hours. Reactivated lymphocytes attack pathogens and release cytokines to reduce harmful inflammation. If GIK prevented or ameliorated polyneuropathy, it might do the same for other chronic infections or auto-immune diseases.
We propose that chronic diseases like AIDS and atherosclerosis and amyotrophic lateral sclerosis (ALS) are caused by an inadequate immune response with little involvement by lymphocytes. We also suggest that auto-immune diseases are not due to an overly zealous attack by lymphocytes but to a continual, ineffective and destructive defense by inflammatory cells.
It is known that the development of many auto-immune diseases (e.g. insulin dependent diabetes mellitus (IDDM) [66], rheumatoid arthritis [67], Reiter's syndrome [68], Guillam-Barre Syndrome (GBS) [69], multiple sclerosis (MS) [70]) is preceded by a viral or bacterial infection or a vaccination. The course of these diseases is more like that of a chronic inflammation. Rheumatoid arthritis is an unrelenting disease that can continue for decades, and while "T cells are a prominent component of the inflammatory infiltrate in the rheumatoid synovium,... the more striking observation is the general paucity of T-cell-derived cytokines in the synovial tissue. In contrast, there is a wide range of readily detectable macrophage-derived products, including proinflammatory cytokines such as tumor necrosis factor-α and interleukin-1, that can activate synovial fibroblasts and other cells to produce matrix metalloproteinases involved in the degradation of cartilage" [71]. As Dinarello and Moldawer have said "...there is now growing recognition that persistent activation of the innate immune system occurs in a variety of autoimmune diseases, including rheumatoid arthritis. This prolonged activation leads to the constitutional complaints, metabolic abnormalities, and the destruction and remodeling of tissues experienced by patients with chronic and uncontrolled progressive diseases" [72].
We further propose that both chronic infections and many autoimmune diseases occur because of Antigenic Competition. It has long been known that if a patient is fighting one pathogen, infection by a second meets little resistance. To pathogen #2, there most likely will be an automatic, inflammatory response with phagocytosis of pathogen #2 by dendritic cells and tissue macrophages followed by presentation of antigen to lymphocytes. In our model there even may be minimal proliferation of lymphocyte clones, but those cells will be unable to mount an effective attack on the second pathogen. The inflammatory attack will cause some destruction of pathogens but also damage surrounding tissues. Fibroblasts may attempt to contain the infection by erecting fibrin barriers. But if the pathogen is multiplying more rapidly than the inflammatory attack, the infection will become chronic. Such an inflammation can go on for months, even years if lymphocytes are not activated to destroy pathogens.
In short, because of Antigenic Competition, the body can mount only one adaptive response at a time. Besedovsky and colleagues proposed that the phenomenon is caused by the increased level of corticosteroids induced by the first antigen [73]. If cortisol increases after the lymphocyte has already been stimulated by antigen, it will have no effect on the lymphocyte at physiological levels. But if cortisol rises before the lymphocyte is presented with antigen, the cell will be unable to respond. Also, it has been shown that "...CD8 lymphocytes after 4 hours of hyperinsulinemia in the normal subjects... had a sharp reduction in insulin-supported lymphocyte mediated cytotoxicity" [74]. A lymphocyte cannot respond if levels of insulin are high before it is challenged by an antigen.
So we proposed that the effect of high levels of cortisol and of insulin in the blood at the time of the second challenge is that the clone of lymphocytes that would ordinarily attack pathogen #2 are rendered helpless. We propose that even after infection #1 is resolved, the paralysis of clone #2 will often continue. It cannot activate without high levels of insulin for a prolonged period. Insulin will ordinarily rise only in response to another infection. But that is preceded by another surge of cortisol, which will continue the suppression of clone #2. However, in all cases of local inflammation (e.g. Pancreas, joints, myelin), there will be some activity by lymphocytes, both cellular and humoral. For acetylcholine, released from endings of cholinergic nerves, has much the same effect of enhancing the ability of cytotoxic lymphocytes to injure target cells [75]. The teleological benefit is that the body can send lymphocytes into a lesion to finish the killing of pathogens without having to mount a full scale systemic attack involving insulin. It seems unlikely, however, that the few infiltrating lymphocytes could fully meet the challenge presented to it by a disease such as rheumatoid arthritis.
We also suggest that if pathogen #2 is not contained in a local site but becomes systemic, it is likely that one of two things will happen. If the pathogen is virulent, sepsis will develop. The infection will rage uncontained, defended against only by the innate limb of the immune system, which, under such circumstances may itself be destructive. If the pathogen is a bacterium susceptible to antibiotics, the patient may be saved. Or, if the pathogen is less virulent, it may lodge in various tissues, only emerging at times of reduced immunity. It will produce shingles or attack skin or even organs, as in SLE or scleroderma.
Thus, in our model there are two circumstances in which the body cannot mount an effective adaptive immune response. The first is when the body abandons all effort to rid itself of pathogens and turns its energies to healing, as in the critical care setting. The second is Antigenic Competition.
We suggest that the only cure for lingering infections such as atherosclerosis, HIV or tuberculosis or for some auto-immune diseases, is infusion by GIK or TGIK to achieve levels of insulin that mimic those produced during an infection and for a long enough time for lymphocyte clones to fully proliferate and destroy the pathogen.
Unfortunately, it is likely that only studies with humans would conclusively prove or disprove this hypothesis. Animal models are of limited value in many of these diseases. Yet human experiments would be unacceptably dangerous. If conventional thought concerning autoimmune diseases is correct, the patient's condition would worsen, perhaps catastrophically.
However, it is possible that such studies have already, inadvertently, been conducted. Surely, some of the hundreds of patients who have been treated with high dose, long duration GIK in the critical care setting must have had Parkinson's or MS or ALS or Alzheimer's or Chlamydia or SLE or rheumatoid arthritis or GBS or scleroderma or atherosclerosis or tuberculosis or AIDS in addition to the acute condition that caused their hospitalization. What were the results for such patients? Was the condition ameliorated or exacerbated or did it remain unchanged? Follow-up studies of these patients could be helpful.
Before the possible full benefits of GIK can be assessed, questions of correct dosage, method of administration and duration of treatment must be settled. Treating a patient for 20 minutes [76], or even for a few hours, especially with low doses, would have little effect on immunity. More time is needed for full proliferation of activated lymphocyte clones. As Das has observed "Studies in which higher concentrations of insulin were used showed better results than did those studies that employed a lesser dose" [77]. We propose that GIK should be administered continuously and intravenously in whatever doses will maintain blood insulin levels at 35 ± 5 μU/ml for 48 to 96 hours to produce maximal benefit. In order to reach that level it may be necessary to adjust the dosage of insulin to each patient, but it is likely that insulin in the range of .1 to .15 U/kg/hr for non-diabetic patients should achieve this target level [78]. The patient must also receive enough glucose and potassium to avoid hypoglycemia and hypokalemia. Low doses of thyroid may be added to achieve maximum effect. Future researchers can contribute to the data base if they will perform pre-prandial testing of serum insulin and CD4/CD8 levels before, during, after treatment. Only studies with human patients can establish correct doses, duration of treatment and method of administration, but one of the advantages of GIK is that it is not a new drug. Clinicians are familiar with the signs of toxicity and counter-measures. The work of Van den Berghe and Krinsley show that can be done safely if patients are carefully monitored.
While van der Horst, et al. are correct that conclusive evidence GIK has a positive effect on sepsis is lacking [79], our work and that of others in a different setting are indicative of the importance of more research. For example, in 1985 Kowli, et al. reported that when they gave insulin in significant amounts to surgical patients, the infection rate was significantly lower than in controls and infection-related mortality was also reduced [80]. Also, if our experience with the increase in CD4 cells after treatment with low-dose TGIK could be reproduced, GIK may prove helpful in the treatment of AIDS.
The significance of the mounting evidence from GIK studies and the oncology studies cited above is obvious. For the first time physicians may be able not only to reduce inappropriate inflammatory and immune reactions, as with glucocorticoids, but also to enhance lymphocytic action to destroy pathogens and tumors without the use of toxic cytokines. It is, therefore, important that more research be devoted to establishing the mechanism and optimum dose and duration of treatment of GIK. Clinicians are already engaged in seeking that mechanism and the parameters for treatment. But immunologists have special knowledge that would be helpful in exploiting this important discovery.
Competing interests
AFH holds multiple domestic and foreign patents on the use of TGIK and GIK for stimulating immunity and treating cancer. DBW and WJP have no competing interests.
Authors' contributions
AFH conceived the model of immunity and the use of TGIK and GIK for treating cancer.
DBW designed and conducted the studies with mice and provided helpful advice on human trials.
WJP wrote the report on mice studies and is designing the protocol for a new trial of TGIK in humans.
==== Refs
Van den Berghe G Wouters P Weekers F Verwaest C Bruyninckx F Schetz M Vlasselaers D Ferdinance P Lauwers P Bouillon R Intensive Insulin Therapy in Critically Ill Patients New Eng Jnl Med 345 1359 Nov 8 2001 10.1056/NEJMoa011300
Krinsley JS Effect of an intensive glucose management protocol on the mortality of critically ill adult patients Mayo Clin Pros 2004 79 992 1000
Steinman RM Mellman I Immunotherapy: bewitched, bothered and bewildered no more Science 305 197 200 Jul 9 2004 15247468 10.1126/science.1099688
Roitt I Essential Immunology 1991 7 Oxford: Blackwell Scientific Publications 168
Groeneveld AB Beishuizen A Visser FC Insulin: a wonder drug in the critically ill? Crit Care 2002 6
Dandona P Aljada A Chaudhuri A Bandyopadhyay A The potential influence of inflammation and insulin resistance on the pathogenesis and treatmentof atherosclerosis-related complications in type-2 diabetes Jnl Clin Endo Metab 2003 88 2422 2429 10.1210/jc.2003-030178
Das UN Insulin: an endogenous cardioprotector Cur Op In Crit Care 2003 9 375 383 10.1097/00075198-200310000-00007
Cain BS Meldrum DR Dinarello CA Meng X Joo KS Banerjee A Harken AH Tumor necrosis factor-alpha and interleukin-1 synergistically depress human myocardial function Crit Care Med 1999 7 1309 10446825 10.1097/00003246-199907000-00018
Dinarello C Moldawer LL Proinflammatory and Anti-Inflammatory Cytokines in Rheumatoid Arthritis 2001 3 Amgen, Inc. One Amgen Center Drive, Thousand Oaks, CA,91320-1799 17
Dinarello C Moldawer LL Proinflammatory And Anti-Inflammatory Cytokines In Rheumatoid Arthritis 1999 1 Amgen, Inc. One Amgen Center Drive, Thousand Oaks, CA,91320-1799
Ibid 3 329
Ibid
Op cit 1 13
Aarimaa M Syvalanti E Ovaska J Does adrenergic activity suppress insulin secretion during surgery? Ann Surg 1978 187 68 202205
Wright PD Glucose homeostasis following injury Ann Roy Coll Surg Eng 1979 61 430
Carey LC Lowery BD Cloutier CT Blood sugar and insulin response of humans in shock Ann Surg 1970 172 343
Wright op cit 429
Carey op cit 344
Strom TB Bear RA Carpenter CB Insulin-induced augmentation of lymphocyte-mediated cytotoxicity Science 187 1206 Mar 28 1975 163492
Koffler M Raskin P Womble D Helderman JH Immunobiological consequence of regulation of insulin receptor on alloactivated lymphocytes in normal and obese subjects Diabetes 1991 40 368
Strom Science op cit 1206
Ibid
Helderman JH Strom TB Role of protein and RNA synthesis in the development of insulin binding sites on activated thymus-derived lymphocytes Jnl Bio Chem 15 7203 Aug 10 1979
Buffington CK El-Shiekh T Kitabchi AE Matteri R Phytohemagglutinin (PHA) activated human T-lymphocytes: concomitant appearance of insulin binding, degradation and insulin-mediated activation of pyruvate dehydrogenase (PDH Biochem Biophys Res Comm 1986 1
Cunningham VL Wolken KW Ackerman GA Insulin complex binding to human peripheral and mitogen-stimulated lymphocytes Jnl Histochem Cytochem 1984 5 517 525
Krug U Krug F Cuatrecasas P Emergence of insulin receptors on human lymphocytes during in vitro transformation Pro Natl Acad Sci 1972 9 USA 2604
Helderman JH Strom TB Role of protein and RNA synthesis in the development of insulin binding sites on activated thymus-derived lymphocytes Jnl Bio Chem 1979 15 7206 Aug 10 1979
Murphy RF Bisaccia E Cantor CR Berger C Edelson RL Internalization and acidification of insulin by activated human lymphocytes Jnl Cell Phys 1984 121 351 356 10.1002/jcp.1041210212
Krug op cit 2607
Helderman JH Role of insulin in the intermediary metabolism of the activated thymic-derived lymphocyte Jnl ClinInvest 1981 67 1636
Beisel WR Magnitude of the host nutritional responses to infection Amer Jnl Clin Nutr 1977 30 1236 1247
Beisel WR Metabolic response to infection Ann Med Rev 1975 26 14
Beutler B Cerami A Cachectin: more than a tumor necrosis factor New Eng Jnl Med 1987 7 79 386
Beutler BA Milsark IW Cerami A Cachectin/tumor necrosis factor: production, distribution and metabolic fate in vivo Jnl Imm 1985 6 3975
Rayfield EJ Curnow RT Reinhard D Kochicheril NM Effects of Acute Endotoxemia on Glucoregulation in Normal and Diabetic Subjects Jnl Clin End Metab 1977 3
Ryan NT Blackburn GL Clowes GHA Jr Differential tissue sensitivity to elevated endogenous insulin levels during experimental peritonitis in rats Metab 1974 11
Rocha MD Santeusanio F Faloona GR Unger RH Abnormal pancreatic alpha-cell function in bacterial infections New Eng Jnl Med 14 701 Apr 5 1973
Jungas RL Brownlee M Action of insulin on enzymes Handbook Of Diabetes Mellitus 1981 2 New York: Garland STPM Press 162
Levenson S Seifter E Van Winkle W Jr Hunt TK, Dunphy E Nutrition Fundamentals of Wound Management 1979 Philadelphia:Appleton-Century-Crofts 289
Moore FD Metababolic Care Of The Surgical Patient 1959 Philadelphia: WD Saunders 128
Van den Berghe G The neuroendocrine response to the severe stress of critical illness
Berczi I Matera L, Corso AM, Rappaport R, Matera L, Rappaport R The neuroimmune biology of growth and lactogenic hormones. Foreword Growth and Lactogenic Hormones Neuroimmune Biology 2002 2 New York: Elsevier
Mizel SB Todaro GJ Moore MAS In vitro studies on interleukin-1 and tumor cell derived immunosuppressive peptides Maturation Factors And Cancer 1982 New York: Raven Press 205
Szentivani A Szentivani J Sodeman WA Jr, Sodeman TM The pathophysiology of immunologic and related diseases Sodeman's Pathologic Physiololgy: Mechanisms Of Disease 1985 7 Philadelphia: WB Saunders Comppany 168
Benz CC Lewis BJ Greenspan FS, Forsham PH, Los Altos Hormones and Cancer Basic and Clinical Endocrinology 1986 2 California: Lange Medical Publications 624
Homburger F The Biological Basis Of Cancer Management 1957 New York: Hoeber-Harper 140 1
Harold FM The Vital Force: A Study Of Bioenergetics 1986 New York: WH Freeman and Company 332
Handwerger BS Douglas S Gerald Weissman Cell biology of blastogenesis The Cell Biology Of Inflammation 1980 New York: Elsevier/North-Holland Biomedical Press, Amsterdam 654
Segal GB Lichtman MA Potassium transport in human blood lymphocytes treated with phytohemagglutinin Jnl Clin Invest 1976 58 1358
Handwerger Douglas loc cit
Quastel MR Kaplan JB Early stimulation of potassium uptake in lymphocytes treated with PHA Exp Cell Res 1970 63 230 5496279 10.1016/0014-4827(70)90360-5
Segal Lichtman op cit
Handwerger Douglas op cit 655
Eden E Edstrom S Bennegard K Schersten T Ludholm K Glucose flux in relation to energy expenditure in malnourished patients with and without cancer during periods of fasting and feeding Can Res 1984 44 1719
Williams ME Epstein FH Massry SG, Glassock RJ Extrarenal control of serum potassium Textbook Of Nephrology 1989 1 2 London: Williams and Wilkins 267
Bia MJ DeFronzo RA Extrarenal potassium homeostasis Amer Jnl Phys, 240 (Renal Fluid Electrolyte Physiology) 1981 9 F259
Koroljow S Two cases of malignant tumors with metastases apparently treated successfully with hypoglycemic coma Psych Quar 1962 1 261 271 10.1007/BF01586115
Oppenheim JJ Ruscetti FW Faltynek CR Stites D, Stobo JD, Wells VJ Interleukins and interferons Basic and Clinical Immunology 1986 Los Altos. California: Appleton and Lange 90
Van der Horst IC Zijlstra F van't Hof AW Doggen CJ de Boer MJ Suryapranata H Hoorntje JC Dambrink JH Gans RAO Bilo HJ Glucose-insulin-potassium infusion in patients treated with primary angioplasty for acute myocardial infarction: the glucose-insulin-potassium study: a randomized trial Jnl Amer Coll Card 5 792 795 Sep 3 2003
Das UN Insulin in sepsis and septic shock Jnl Assoc Phyns India 2003 51 695 700
New Eng Jnl Med op cit 1359
Brod SA Unregulated inflammation shortens human functional longevity Inflam Res 2000 11 561 570 10.1007/s000110050632
Marx J Inflammation and cancer: the link grows stronger Science 306 966 Nov 5 2004 15528423 10.1126/science.306.5698.966
Das UN Hypothesis: can glucose-insulin-potassium regimen in combination with polyunsaturated fatty acids suppress lupus and other inflammatory conditions? Prostaglandins Leukot Essent Fatty Acids 2001 109 113 11545628 10.1054/plef.2001.0297
Ojcius DM Darville T Bavoil PM Can Chlamydia be stopped? Sci Amer 2005 74 15859216
Baker JR Parslow TG, Stites DP, Terr AI, Imboden JB Endocrine diseases Medical Immunology 2001 10 New York:Lange Medical Books/McGraw Hill 429
Sack KAE Fye KH Rheumatic diseases Medical Immunology, op cit 406
Ibid 418
Stove O Zamvil SS Neurological diseases Medical Immunology, op cit 518
Ibid 511
Sack KE Fye KH op cit 406
Dinarello C Moldawer LL Proinflammatory and Anti-Inflammatory Cytokines in Rheumatoid Arthritis, Op cit 3 16
Besodovsky HO Del Rey A Sorkin E Antigenic Competition between horse and sheep red blood cells as a hormone-dependent phenomenon Clin Exp Imnmunol 1979 37 111
Koffler M Raskin P Womble D Helderman JH Immunological consequences of the regulation of insulin receptor on alloactivated lymphocytes in normal and obese subjects." Diabetes 1991 40 368
Illiano G Tell GPE Siegel MI Cuatrecasas P Guanosine 3':5'-cyclic monophosphate and the action of insulin and acetylcholine Proc Nat Acad Sci USA 1973 70 2443 4365701
Bronsveld W van den Bos GC Thijs LG Use of glucose-insulin-potassium (GIK) in human septic shock Crit Care Med 1985 70 566 70 3891230
Das UN Is insulin an endogenous cardioprotector? Crit Care 2002 6 389 393 12398773 10.1186/cc1535
Bia MJ DeFronzo RA Extrarenal potassium homeostasis Am J Physiol 1981 9 F257 268 6111930
van de Horst ICC Lightenberg JJM Bilo HJG Zijlstra F Gans RO Glucose-insulin-potassium infusion in sepsis and septic shock: no hard evidence yet Crit Care 2003 7 13 15 12617733 10.1186/cc1832
Kowli SS Parikh SK Shirahatti RG Relekar RR Bhalerao RA Jnl Postgrad Med 1985 31 11 5
|
16111485
|
PMC1201164
|
CC BY
|
2021-01-04 16:37:45
|
no
|
J Immune Based Ther Vaccines. 2005 Aug 19; 3:5
|
utf-8
|
J Immune Based Ther Vaccines
| 2,005 |
10.1186/1476-8518-3-5
|
oa_comm
|
==== Front
J Neuroengineering RehabilJournal of NeuroEngineering and Rehabilitation1743-0003BioMed Central London 1743-0003-2-201604280510.1186/1743-0003-2-20ResearchGait dynamics in mouse models of Parkinson's disease and Huntington's disease Amende Ivo [email protected] Ajit [email protected] Scott [email protected] Scott [email protected] James P [email protected] Thomas G [email protected] Division of Medicine, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA 02215 USA2 The CuraVita Corporation, Boston, MA 02109 USA2005 25 7 2005 2 20 20 2 4 2005 25 7 2005 Copyright © 2005 Amende et al; licensee BioMed Central Ltd.2005Amende 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
Gait is impaired in patients with Parkinson's disease (PD) and Huntington's disease (HD), but gait dynamics in mouse models of PD and HD have not been described. Here we quantified temporal and spatial indices of gait dynamics in a mouse model of PD and a mouse model of HD.
Methods
Gait indices were obtained in C57BL/6J mice treated with the dopaminergic neurotoxin 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP, 30 mg/kg/day for 3 days) for PD, the mitochondrial toxin 3-nitropropionic acid (3NP, 75 mg/kg cumulative dose) for HD, or saline. We applied ventral plane videography to generate digital paw prints from which indices of gait and gait variability were determined. Mice walked on a transparent treadmill belt at a speed of 34 cm/s after treatments.
Results
Stride length was significantly shorter in MPTP-treated mice (6.6 ± 0.1 cm vs. 7.1 ± 0.1 cm, P < 0.05) and stride frequency was significantly increased (5.4 ± 0.1 Hz vs. 5.0 ± 0.1 Hz, P < 0.05) after 3 administrations of MPTP, compared to saline-treated mice. The inability of some mice treated with 3NP to exhibit coordinated gait was due to hind limb failure while forelimb gait dynamics remained intact. Stride-to-stride variability was significantly increased in MPTP-treated and 3NP-treated mice compared to saline-treated mice. To determine if gait disturbances due to MPTP and 3NP, drugs affecting the basal ganglia, were comparable to gait disturbances associated with motor neuron diseases, we also studied gait dynamics in a mouse model of amyotrophic lateral sclerosis (ALS). Gait variability was not increased in the SOD1 G93A transgenic model of ALS compared to wild-type control mice.
Conclusion
The distinct characteristics of gait and gait variability in the MPTP model of Parkinson's disease and the 3NP model of Huntington's disease may reflect impairment of specific neural pathways involved.
Gait variabilityGaitMouse modelsNeurodegenerationMovement disordersAmyotrophic Lateral SclerosisSOD1
==== Body
Background
Disturbances in gait are symptomatic of Parkinson's disease (PD) and Huntington's disease (HD). Gait abnormalities in PD include shortened stride length [1,2], a dyscontrol of stride frequency [3], and postural instability [4]. Gait abnormalities in HD include reduced walking speed [5], widened stance width [6], reduced stride length [6,7], and sway [8]. Gait variability has also been shown to be significantly higher in patients with PD [9-11] and HD [7,9] compared to control subjects. Early detection of gait disturbances may result in earlier treatment. Therapies for PD and HD patients are often developed to ameliorate gait abnormalities [12,13]. Mouse models of PD and HD are used to understand the pathologies of the diseases and to accelerate the testing of new therapies to correct motor defects. Although spatial gait indices have been reported [14,15], gait dynamics in mouse models of PD and HD have not yet been described.
One common mouse model of PD is obtained by repeatedly administering the neurotoxin 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) [16-18]. MPTP causes damage of the nigrostriatal dopaminergic system [19], resulting in PD symptoms, including reduced stride length [14] and posture disturbances in mice [20]. One common mouse model of HD is obtained by repeatedly administering the mitochondrial toxin 3-nitropropionic acid (3NP) [21,22]. 3NP causes striatal neurodegeneration resulting in mild dystonia and bradykinesia comparable to HD in people [23,24].
Motor defects in MPTP-treated mice or 3NP-treated mice are often quantified using the rotarod test that measures the time a subject can balance on a rotating rod [25,26]. MPTP has been shown to reduce performance on the rotarod [27] or to have no effect on rotarod performance [17,28]. 3NP has been shown to reduce rotarod performance [29], or to have no effect on rotarod performance [30]. The swim test [31], balance beam test [32], and the pole test [33] have also been used to investigate the effects of MPTP and 3NP on motor function in mice. Results regarding motor dysfunction in the MPTP model of PD and the 3NP model of HD may vary due to the heterogeneity in protocols followed. Disparities in the degree of motor dysfunction have suggested that large doses of MPTP or 3NP may be required to detect motor defects after nigrostriatal damage [18,29,34].
Several studies in mouse models of PD and HD have described "gait" by estimating stride length [14], and stance width [15] determined by painting the animals' paws. Fernagut et al. reported that stride length is a reliable index of motor disorders due to basal ganglia dysfunction in mice [15]. Gait dynamics in humans, however, extend beyond the measure of stride length. Gait dynamics in humans include spatial indices such as stance width and foot placement angle. Gait dynamics in humans also include temporal indices, such as stride frequency, stride duration, swing duration, and stance duration.
Step-to-step gait variability in humans has also provided important information about possible mechanisms involved in neurodegenerative diseases, including PD and HD [7,9-11]. In patients with PD, higher step-to-step variability has been reported [9-11,35]. The stride length variability increased with the progression of PD suggesting that this index is useful in assessing the course of PD [10]. Hausdorff et al. demonstrated significantly higher variability in several gait indices, including stride duration and swing duration, in patients with PD and HD [9], and in subjects with amyotrophic lateral sclerosis (ALS) [36]. It has been proposed that a matrix of gait dynamic markers could be useful in characterizing different diseases of motor control [36]. Comparable analyses of gait and stride variability in mouse models of PD and HD have not yet been reported.
We recently described ventral plane videography using a high-speed digital camera to image the underside of mice walking on a transparent treadmill belt [37,38]. The technology generates "digital paw prints", providing spatial and temporal indices of gait. Here we applied ventral plane videography to study gait dynamics in the MPTP model of PD and the 3NP model of HD. We studied the C57BL/6 strain, which has been shown to be sensitive to both toxins [14,18,21,29]. Since PD, HD, and ALS share aspects of pathogenesis and pathology of motor dysfunction, we also studied gait dynamics in the SOD1 G93A transgenic mouse model of ALS [39] to compare gait variability in mouse models of basal ganglia disease to a mouse model of motor neuron disease.
Methods
Mice
Male C57BL/6J mice (7–8 weeks; ~22 gm) were purchased from The Jackson Laboratory (Bar Harbor, ME). Mice transgenic for the mutated human SOD1 G93A (TgN [SOD1-G93A]1Gur) (SOD1 G93A) and wild-type human SOD1 (TgN [SOD1]2Gur) wild-type controls) were purchased from The Jackson Laboratory (Bar Harbor, ME) when the mice were ~7.5 weeks old. Animals were maintained on a 12-hour light: 12-hour dark schedule with ad libitum access to food and water. Handling and care of mice were consistent with federal guidelines and approved institutional protocols.
Experimental groups
MPTP
1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) (Sigma-Aldrich, St. Louis, MO) dissolved in saline was administered 30 mg/kg i.p. to 7 mice every 24 hours for 3 days (MPTP-treated mice), based on previously published studies [40,41]. Equivolume (0.2 ml) of saline was administered i.p. to 7 control mice every 24 hours for 3 days (saline-treated mice).
3NP
3-nitropropionic acid (3NP) (Sigma-Aldrich, St. Louis, MO) dissolved in saline was administered 3 times to 6 mice: 25 mg/kg i.p. twice, separated by 12 hours (cumulative dose of 50 mg/kg), then 25 mg/kg 24 hours later (cumulative dose of 75 mg/kg) (3NP-treated mice). Equivolume (0.2 ml) of saline was administered i.p. according to the same schedule to 6 control mice. The intoxication protocol was based on published studies [29,42], and our own pilot observations that higher doses resulted in high mortality rates or the inability of the mice to walk at all on the treadmill belt.
SOD1 G93A transgenic mice
To compare gait variability in the MPTP and 3NP mouse models of basal ganglia disease to a mouse model of motor neuron disease, we also examined gait in a mouse model of amyotrophic lateral sclerosis (ALS). Gait dynamics in SOD1 G93A mice were measured at ages ~8 weeks (n = 3), ~10 weeks (n = 3), ~12 weeks (n = 5), and ~13 weeks (n = 5), time points this model has been shown to exhibit motor dysfunction [43-45], and compared to wild-type control mice studied at ages ~8 weeks (n = 3), ~10 weeks (n = 3), ~12 weeks (n = 6), and ~13 weeks (n = 6).
Gait dynamics
Gait dynamics were recorded using ventral plane videography, as previously described [37,38]. Briefly, we devised a motor-driven treadmill with a transparent treadmill belt. A high-speed digital video camera was mounted below the transparent treadmill belt. An acrylic compartment, ~5 cm wide by ~25 cm long, the length of which was adjustable, was mounted on top of the treadmill to maintain the mouse that was walking on the treadmill belt within the view of the camera. Digital video images of the underside of mice were collected at 80 frames per second. Each image represents 12.5 ms; the paw area indicates the temporal placement of the paw relative to the treadmill belt. The color images were converted to their binary matrix equivalents, and the areas (in pixels) of the approaching or retreating paws relative to the belt and camera were calculated throughout each stride. Plotting the area of each digital paw print (paw contact area) imaged sequentially in time provides a dynamic gait signal, representing the temporal record of paw placement relative to the treadmill belt (Figure 1). Each gait signal for each limb comprises a stride duration (stride time), which includes the stance duration when the paw of a limb is in contact with the walking surface, plus the swing duration when the paw of the same limb is not in contact with the walking surface. Stance duration was further subdivided into braking duration (increasing paw contact area over time) and propulsion duration (decreasing paw contact area over time) (Figure 1B).
Figure 1 Ventral view of walking saline-treated mouse. A. Two images depicting the ventral view of a saline-treated C57BL/6J mouse on a transparent treadmill belt walking at a speed of 34 cm/s. The example on the left depicts full stance for the right hind limb, and the example on the right depicts sequential full stance for the left hind limb. Cartesian coordinates are used to determine stance width and paw placement angles for the forelimbs and hind limbs. B. Representative gait signals of the left forelimb and right hind limb of a saline-treated C57BL/6J mouse walking at a speed of 34 cm/s. Duration of stride, stance, and swing are indicated for the right hind limb. Duration of braking and propulsion are indicated for the left fore limb.
Stride frequency was calculated by counting the number of gait signals over time. Stride length was calculated from the equation: speed = stride frequency × stride length. To obtain stance widths and paw placement angles at full stance, ellipses were fitted to the paws, and the centers, vertices, and major axes of the ellipses were determined. Forelimb and hind limb stance widths were calculated as the perpendicular distance between the major axes of the left and right paw images during peak stance. Gait data were collected and pooled from both the left and right forelimbs, and the left and right hind limbs.
Measures of stride-to-stride variability (gait variability) for stride length, stride time, and stance width were determined as the standard deviation and the coefficient of variation (CV). The standard deviation reflects the dispersion about the average value for a parameter. CV was calculated from the equation: 100 × standard deviation/mean value.
Gait was recorded ~24 hours after each administration of saline or MPTP. Gait was recorded ~12 hours after the 1st administration, and ~24 hours after the 2nd and 3rd administration of 3NP. Each mouse was allowed to explore the treadmill compartment for ~1 minute with the motor speed set to zero since our previous experience with C57BL/6J mice [37] indicated they do not require extended acclimatization to the treadmill. The motor speed was then set to 34 cm/s and images were collected. Approximately 3 seconds of videography were collected for each walking mouse to provide more than 7 sequential strides. Only video segments in which the mice walked with a regularity index of 100% [46] were used for image analyses. The treadmill belt was wiped clean between studies if necessary.
Statistics
Data are presented as means ± SE. ANOVA was used to test for statistical differences among saline-treated, MPTP-treated, and 3NP-treated mice. When the F-score exceeded Fcritical for α = 0.05, we used post hoc unpaired Student's two-tailed t-tests to compare group means. Gait indices between forelimbs and hind limbs within the saline-treated mice were compared using Student's two-tailed t-test for paired observations. Gait indices between SOD1 G93A and wild-type control mice were compared using unpaired Student's two-tailed t-test. Differences were considered significant with P < 0.05.
Results
Gait in saline-treated mice
The ventral view of a C57BL/6J mouse walking on a transparent treadmill belt is shown in the upper panel of Figure 1 (and Additional file 1). Representative gait dynamics signals for the left forelimb and right hind limb of a saline-treated mouse walking at a speed of 34 cm/s are shown in the lower panel of Figure 1. Walking at a speed of 34 cm/s, C57BL/6J mice achieved ~5 steps every second, completed one stride within ~200 ms, and traversed ~7 cm with each step. The contributions of stance and swing durations to stride duration were ~55% (stance/stride) and ~45% (swing/stride) respectively. Forelimb stance width was significantly narrower than hind limb stance width (1.7 ± 0.1 cm vs. 2.4 ± 0.2 cm, P < 0.05). The paw placement angle of the hind limbs was significantly more open than the paw placement angle of the forelimbs (13.9 ± 1.6 vs. 2.6 ± 0.6, P < 0.05). Stride length variability of hind limbs was lower than of forelimbs (0.63 ± 0.08 cm vs. 0.78 ± 0.03 cm, P < 0.05). Likewise, stance width variability of hind limbs was lower than of forelimbs (0.14 ± 0.01 cm vs. 0.21 ± 0.02 cm, P < 0.05) in saline-treated mice walking on a treadmill belt at 34 cm/s.
Gait in MPTP-treated mice
Gait dynamics in MPTP-treated mice after 3 administrations of 30 mg/kg MPTP were significantly different than gait dynamics in saline-treated mice (Table 1 and Figure 2). Stride length was decreased in MPTP-treated mice compared to saline-treated mice (6.6 ± 0.1 cm vs. 7.1 ± 0.1 cm, P < 0.05) at a walking speed of 34 cm/s. Stride frequency was increased in MPTP-treated mice. Stride duration was significantly shorter in MPTP-treated mice (194 ± 1 ms vs. 207 ± 2 ms, P < 0.05). This was attributable to a shorter swing duration of the hind limbs (92 ± 3 vs. 104 ± 2 ms, P < 0.05), and a shorter stance duration of the forelimbs (116 ± 2 ms vs. 126 ± 2 ms, P < 0.05). The contributions of stance and swing to stride duration in MPTP-treated mice were not different than in saline-treated mice, despite the shorter stride duration. Forelimb stance width and hind limb stance width were comparable in MPTP-treated mice and saline-treated mice. The paw placement angles of the forelimbs and hind limbs of MPTP-treated mice were not different than in saline-treated mice. Figure 2 illustrates the gait signal from the right hind limb of a MPTP-treated mouse superimposed over the gait signal from the right hind limb of a saline-treated mouse.
Table 1 Gait dynamics in saline-treated, MPTP-treated (90 mg/kg cumulative dose), and 3NP-treated (75 mg/kg cumulative dose) mice walking on a treadmill belt at a speed of 34 cm/s.
Saline (n = 7) MPTP (n = 7) 3NP (n = 3)
Stride Length (cm) 7.1 ± 0.1 6.6 ± 0.1* 7.3 ± 0.1
Stride Frequency (Hz) 5.0 ± 0.1 5.4 ± 0.1* 4.9 ± 0.1
Stride Duration (ms) 207 ± 2 194 ± 1* 217 ± 5
% Stance Duration 54.3 ± 0.9 55.9 ± 1.1 59.4 ± 2.3*
% Swing Duration 45.7 ± 0.9 44.1 ± 1.1 40.6 ± 2.3*
Forelimb Stance Width (cm) 1.7 ± 0.1 1.6 ± 0.1 1.7 ± 0.1
Forelimb Paw Placement Angle (°) 2.6 ± 0.6 2.6 ± 0.4 3.5 ± 1.1
Hind limb Stance Width (cm) 2.4 ± 0.2 2.2 ± 0.1 2.8 ± 0.2
Hind limb Paw Placement Angle (°) 13.9 ± 1.6 10.8 ± 1.3 15.2 ± 1.0
Means ± SE. *P < 0.05, compared to saline-treated mice.
Figure 2 Gait signals in a MPTP-treated mouse. Gait signal of the right hind limb of a MPTP-treated mouse superimposed over the gait signal of the right hind limb of a saline-treated mouse. Stride frequency was higher in MPTP-treated mice compared to saline treated mice. Stance duration and swing duration were shorter in MPTP-treated mice compared to saline-treated mice.
Stride time dynamics for 14 sequential strides in a MPTP-treated mouse are shown in the top panel of Figure 3. For comparison, stride time dynamics in a 3NP-treated mouse are illustrated in the middle panel, and in saline-treated mouse in the bottom panel of Figure 3. Gait variability was significantly higher in MPTP-treated mice after 3 treatments compared to saline-treated mice. Stride length variability of the forelimbs was higher in MPTP-treated than in saline-treated mice (0.91 ± 0.04 cm vs. 0.78 ± 0.03 cm, P < 0.05). Stride length variability of the hind limbs, however, was not different in MPTP-treated mice. The coefficient of variation (CV) of forelimb stride length was significantly higher in MPTP-treated than in saline-treated mice (13.6 ± 0.8 % vs. 11.1 ± 0.8 %, P < 0.05). The CV of hind limb stride length was somewhat higher in MPTP-treated than in saline-treated mice (10.0 ± 1.5 % vs. 8.0 ± 0.7 %, NS).
Figure 3 Stride time dynamics. Examples of stride time (gait cycle duration) in MPTP-treated, 3NP-treated, and saline-treated mice of forelimbs (left panels) and hind limbs (right panels). In saline-treated animals, forelimb stride variability was higher than hind limb stride variability. MPTP-treated and 3NP-treated mice exhibited significantly higher stride variability. The coefficient of variation (CV), a measure of stride-to-stride variability, was highest in the forelimbs of 3NP-treated mice.
Stance width variability of the forelimbs was significantly higher in MPTP-treated than in saline-treated mice (0.26 ± 0.01 cm vs. 0.21 ± 0.02 cm, P < 0.05). Stance width variability of the hind limbs was higher in MPTP-treated than in saline-treated mice (0.20 ± 0.02 cm vs. 0.14 ± 0.01 cm, P < 0.05). The CV of forelimb stance width was higher in MPTP-treated than in saline-treated mice (16.7 ± 1.3 % vs. 12.3 ± 1.2 %, P < 0.05). The CV of hind limb stance width was higher in MPTP-treated than in saline-treated mice (9.1 ± 1.1 % vs. 5.9 ± 0.5 %, P < 0.05).
Gait in 3NP-treated mice
Stride length, stride frequency, stance duration, and swing duration were not affected by 3NP after the 1st and 2nd administrations of 25 mg/kg. The paw placement angle of the hind limbs, however, was significantly more open in 3NP-treated mice (n = 6) compared to saline-treated mice (16.6 ± 1.2° vs. 12.4 ± 1.5°, P < 0.05) after the 2nd administration of 3NP (cumulative dose of 50 mg/kg). Stance width variability of the forelimbs, moreover, was higher in 3NP-treated than in saline-treated mice (0.28 ± 0.01 cm vs. 0.22 ± 0.02 cm, P < 0.05) after the 2nd administration of 3NP. The CV of forelimb stance width was higher in 3NP-treated than in saline-treated mice (15.0 ± 1.2 % vs. 11.7 ± 0.6 %, P < 0.05) after the 2nd administration of 3NP. Neither stride length variability nor stance width variability of the hind limbs was affected after the 2nd administration of 3NP (cumulative dose of 50 mg/kg).
After the 3rd administration of 3NP (cumulative dose of 75 mg/kg), half of the 3NP-treated mice could not walk on the treadmill belt at a speed of 34 cm/s. Forelimb gait indices in the three 3NP-treated mice that could walk on the treadmill belt were similar to saline-treated mice. Hind limb gait indices, however, were affected in the three 3NP-treated mice that could walk on the treadmill belt. The hind limb stance width (2.8 ± 0.2 cm) and paw placement angle (15.2 ± 1.0°) in the 3NP-treated mice that could walk on the treadmill belt (n = 3) tended to be greater than in saline-treated mice. The percentage of stride spent in stance was significantly greater in 3NP-treated mice than in saline-treated mice (59.4 ± 2.3% vs. 54.3 ± 0.9 %, P < 0.05). The percentage of stance duration spent in propulsion (propulsion/stance) was greater of the hind limbs in 3NP-treated mice than in saline-treated mice (45.2 ± 2.5 % vs. 40.2 ± 0.9 %, P < 0.05). This was at the expense of a smaller contribution of swing to stride duration (40.6 ± 2.3 % vs. 45.7 ± 0.9 %, P < 0.05).
Stride length variability of the forelimbs, moreover, was significantly higher in the three 3NP-treated mice that could walk than in saline-treated mice (1.31 ± 0.09 cm vs. 0.87 ± 0.07 cm, P < 0.05). Stance width variability of the forelimbs was also higher in 3NP-treated than in saline-treated mice (0.31 ± 0.04 cm vs. 0.22 ± 0.01 cm, P < 0.05). The CV of forelimb stride length was higher in 3NP-treated than in saline-treated mice (17.9 ± 1.6 % vs. 11.8 ± 0.8 %, P < 0.05) (Figure 3). The CV of forelimb stance width was higher in 3NP-treated than in saline-treated mice (17.3 ± 2.4 % vs. 11.7 ± 0.6 %, P < 0.05). Hind limb stride length variability and hind limb stance width variability were not different in the 3NP-treated mice that could walk on the treadmill belt compared to saline-treated mice.
Hind limb gait failure in 3NP-treated mice
Two 3NP-treated mice that could not walk on the moving treadmill belt at a speed of 34 cm/s, however, attempted to walk, but failed to engage the hind limbs in coordinated stepping. Rather, these mice braced their hind paws onto the base of the sidewalls of the walking compartment (Figure 4, upper panel; Additional file 2), avoiding the moving treadmill belt. The forelimbs of these 3NP-treated mice, however, executed coordinated stepping on the moving treadmill belt. Forelimb stride dynamics in these 3NP-treated mice did not differ significantly from saline-treated mice and the three 3NP-treated mice that were able to walk on the treadmill belt at 34 cm/s (Figure 4, lower panel). Despite the limitation of these 3NP-treated mice to only execute forelimb stepping, stride length of forelimbs was 7.1 ± 0.1 cm, stride frequency was 5.0 ± 0.1 Hz, and stance duration was 133 ± 5 ms, all values similar to forelimb gait indices in saline-treated mice.
Figure 4 Ventral view of a 3NP-treated mouse attempting to walk. A. The ventral view of a 3NP-treated mouse attempting to walk on the treadmill belt moving at a speed of 34 cm/s but failing to engage the hind limbs in coordinated stepping. This animal braced its hind paws onto the base of the sidewalls of the walking compartment avoiding the moving treadmill belt. Only the forelimbs execute coordinated stepping sequences. B. Gait signals of the left and right forelimbs of a 3NP-treated mouse demonstrating coordinated stepping, despite hind limb failure of stepping. The signals of left and right hind limbs are not coordinated and reflect artefacts associated with the belt contacting the braced paws.
Gait in SOD1 G93A transgenic mice
Stride length was significantly greater in SOD1 G93A mice (n = 5) than in wild-type mice (n = 6) at ~12 weeks and ~13 weeks of age. At ~12 weeks of age, stride length was significantly increased in SOD1 G93A mice compared to wild-type control mice (7.1 ± 0.1 cm vs. 6.7 ± 0.1 cm, P < 0.05). Stride frequency was lower in SOD1 G93A mice (5.0 ± 0.1 vs. 5.4 ± 0.1 Hz, P < 0.05), and stride duration was longer compared to wild-type control mice (210 ± 2 vs. 197 ± 3 ms, P < 0.05) at ~12 weeks of age. At ~13 weeks of age, stride length remained significantly increased in SOD1 G93A mice compared to wild-type control mice (7.1 ± 0.1 cm vs. 6.8 ± 0.1 cm, P < 0.05). Stride frequency remained lower in SOD1 G93A mice (5.0 ± 0.1 vs. 5.3 ± 0.1 Hz, P < 0.05), and stride duration remained longer compared to wild-type control mice (209 ± 2 vs. 198 ± 3 ms, P < 0.05) at ~13 weeks of age.
Gait variability was monitored in SOD1 G93A mice at ~8 weeks, ~10 weeks, ~12 weeks, and ~13 weeks of age, coinciding with the appearance of motor dysfunction reported in this model [43-45]. Gait variability was not different in SOD1 G93A mice compared to wild-type control mice at age ~8 weeks, ~10 weeks, ~12 weeks, and ~13 weeks. Stride length variability of the forelimbs and hind limbs were comparable between SOD1 G93A mice and wild-type control mice at all ages studied. Stance width variability of the forelimbs and hind limbs were also comparable between SOD1 G93A and wild-type control mice at age ~8 weeks, ~10 weeks, ~12 weeks, and ~13 weeks.
Discussion
Gait disturbances are characteristic of Parkinson's disease, Huntington's disease, and amyotrophic lateral sclerosis. Gait reflects several variables, including balance, proprioception, and coordination. There are several mouse models of PD [20,47] and HD [22,48-50], and one widely studied model of ALS [39,43-45]. Mouse models that replicate PD, HD, and ALS symptoms could improve understanding of their pathogenesis and treatment. Gait variability indices are increasingly being recognized as important markers of neurological diseases [4,9-11,36]. We found gait disturbances, including increased gait variability, in the MPTP-treated mouse model of PD and the 3NP-treated mouse model of HD, which may be the consequence of the affected neural pathways. Gait variability was not increased, however, in the SOD1 G93A transgenic mouse model of ALS.
Gait in MPTP-treated mice
The MPTP-treated mouse model of PD has been extensively studied for its ability to injure the nigrostriatal dopaminergic system, damage neurons, and deplete the brain of dopamine [16-18]. Several studies have described motor function disturbances in MPTP-treated mice to relate the deficits to symptoms in humans with PD. Motor function tests in MPTP-treated mice have included grip strength [40], the ability of the animals to balance on a rotating rod [27,40], and swimming performance [51]. MPTP significantly affects locomotor activity [17,40,52] and motor performance [17,20,28,51], thus providing functional readouts to test potential therapies. Shortened stride length is one of the cardinal features of PD [1,4,11], yet reports of reduced stride length in MPTP-treated animals are sparse. Fernagut et al., using the paw-inking method, measured stride length in mice one week after acute MPTP intoxication [14] and concluded that stride length was a reliable indicator of basal ganglia dysfunction. Smaller doses of MPTP (3 mg/kg) were also found to significantly reduce stride length in rats [53]. The difficulties associated with the paw-inking method and the variability in overground walking speeds in mice [54] have possibly limited reports of stride length in MPTP-treated mice. Using digital paw prints obtained by ventral plane videography, we found that stride length was significantly decreased in MPTP-treated mice after 3 days of administration (i.p. 30 mg/kg/day).
Gait indices, including stride duration, stance duration, swing duration, and stride length, change with changes in walking speed. We eliminated the confounding effects of differences in walking speed on gait dynamics by setting the motorized treadmill belt to 34 cm/s for all mice. Accordingly, since stride length was decreased in MPTP-treated mice, stride frequency was increased and stride duration was decreased in forelimbs and hind limbs of MPTP-treated mice. A decrease in stride duration can be attained by decreases in stance duration and swing duration. We found that the decrease in stride duration in MPTP-treated mice was attained by significantly shorter hind limb swing duration and forelimb stance duration. A reduction of the stance duration may result in a shorter time for limb muscles to be activated for stabilization [55]. This may account for the significant increase in stride-to-stride variability observed in MPTP-treated mice. Fleming et al. studied mice overexpressing wild-type human α-synuclein (ASO mice), a model of early onset familial PD [47]. The authors found that although stride length was comparable to control mice, stride frequency and stride length variability were increased in ASO mice [47]. ASO mice did not exhibit a loss of dopaminergic neurons, but developed accumulation of α-synuclein in the nigrostriatal system and show enhanced sensitivity of nigrostriatal neurons to MPTP administration [47].
Gait in 3NP-treated mice
Gait dynamics in 3NP-treated mice were difficult to study. Aggressive doses of 3NP resulted in high mortality or the inability of the mice to walk at all on the treadmill belt (data not shown). The earliest effect of 3NP (12 hours after 1st dose of 25 mg/kg) on gait was an increase in forelimb stride length variability. Subsequent gait disturbances included increased gait variability of the forelimbs and eventual failure of hind limb stepping. Our findings of different effects of 3NP on gait dynamics of forelimbs and hind limbs are in accordance with previous motor behavioral assessments in 3NP-treated animals [29,56]. Fernagut et al. found no differences in stride length of forelimbs and hind limbs after a cumulative dose of 3NP (340 mg/kg) [29]. With a cumulative dose of 560 mg/kg of 3NP, forelimb stride length was comparable to saline-treated mice, but hind limb stride length was shortened [29]. Administration of 3NP may affect hind limb gait dynamics differently than forelimb gait dynamics via different effects on the neostriatum and the nucleus accumbens [14,57]. Shimano et al. showed that hind limb muscles in 3NP-treated rats became hypotonic with low voltage electromyogram activity and impaired movement [58]. Activation of the motor program required for the two 3NP-treated mice that braced their hind limbs against the inside walls of the walking compartment while simultaneously maintaining coordinated gait of the forelimbs [59] may suggest that 3NP-induced cognitive defects [60] did not contribute to the gait disturbances in 3NP-treated animals.
Lin et al. reported that stride length and stance width in a knock-in mouse model of HD did not differ from wild-type mice [48]. Stride length variability and stance width variability were higher, however, in the mutants [48]. In a transgenic mouse model for HD, R6/2 mice exhibited unevenly spaced shorter strides, staggering movements, and an abnormal step sequence pattern [49]. No significant abnormalities in stride length were observed in the R6/1 HD transgenic mouse [50]. The significantly higher gait variability of the forelimbs we observed in 3NP-treated mice may reflect the jerky and highly variable arm movements in HD gene carriers and patients with HD [61]. Taken together, increases in forelimb stride variability appear to be more characteristic of motor control deficits in early HD than decreases in stride length.
Gait in SOD1 G93A mice
Impaired performance in SOD1 G93A mice in some motor function tests have been observed at ~8 weeks of age [45]. Others have reported motor impairments in SOD1 G93A mice at ~11–16 weeks of age [43,44]. It was of interest, therefore, to find that stride length was significantly longer in SOD1 G93A mice compared to wild-type mice at ~12 weeks and ~13 weeks of age. Increased stride length is often associated with increased amplitude of electromyogram activity and enhanced motor performance. Gurney et al. first described significantly shorter stride length in SOD1 G93A mice with severe pathological changes in the late stage of disease [39]. Puttaparthi et al. also reported significantly shorter stride length in SOD1 G93A mice at ~24 weeks of age [44]. The reported decrease in stride length at later stages could be due to muscle weakness, fatigue, and motor neuron loss. The data of Puttaparthi et al. also indicate, however, that stride length in SOD1 G93A mice may tend to be longer at ~16 weeks of age [44]. Wooley et al., moreover, recently reported significantly longer stride duration in SOD1 transgenic mice compared to wild-type mice walking on a treadmill at 23 cm/s at 8 and 10 weeks of age [62], which would mean that SOD1 transgenic mice had significantly longer stride lengths at 8 and 10 weeks of age. It is notable that patients with ALS who walked overground at speeds comparable to healthy subjects also had longer stride duration [36]. One explanation for the increased stride length in the presymptomatic SOD1 G93A mice we observed walking 34 cm/s could be aberrant electrical activity of the muscles involved in treadmill walking. Kuo et al., in fact, identified significantly elevated intrinsic electrical excitability in cultured embryonic and neonatal mutant SOD1 G93A spinal motor neurons [63]. Dengler et al. surmised that new motor unit sprouting and resulting increases of twitch force could compensate for the loss of motor neurons in patients with early stages of ALS [64]. To our knowledge, there are no reports regarding stride length in patients with ALS walking on a treadmill. An early indication of ALS could be an increase in stride length.
Gait variability indices
The CVs of stride length and stance width in healthy humans are ~3–6% and ~14–17%, respectively [65,66]. The CV of stride time in humans with intact neural control is <3%, and is significantly higher in patients with PD, HD, and ALS [36]. Stride time variability was highest in patients with HD [36]. The CV for stride length in saline-treated C57BL/6 mice is higher than in healthy humans, but the CV for stance width is comparable. Stride length may be determined predominantly by gait-patterning mechanisms, whereas stance width may be determined by balance-control mechanisms [67]. Stride length may be more variable in mice because of a greater number of gait patterns [37]. Gait variability may also be high in mice walking on a treadmill belt at a speed of 34 cm/s compared to mice walking overground at preferred speeds.
We found that gait variability of the forelimbs in mice was significantly higher than gait variability of the hind limbs. This may be attributable to the role of the forelimbs in balance and navigation [68,69]. We further found that the MPTP mouse model recapitulated the higher gait variability in patients with PD, as evidenced by a significant increase in stride length variability of the forelimbs and a significant increase in stance width variability of the forelimbs and hind limbs. We also found that the 3NP mouse model may reflect the higher gait variability in patients with HD, as evidenced by a significant increase in forelimb stride length variability and stance width variability. We found that gait variability of the forelimbs was highest in 3NP-treated mice, in parallel with the higher gait variability in patients with HD as compared to patients with PD [35]. The higher forelimb stride length variability in 3NP-treated mice may reflect the jerky movements of arms in HD patients [61]. Although pathology of PD and HD involve different portions of the basal ganglia, postural instability is common to both diseases. Postural instability was characteristic of MPTP-treated and 3NP-treated mice. Increased stride length and step width variability of the hind limbs was more characteristic in the MPTP model of PD than in the 3NP-model of HD. The more open paw placement angle of the hind limbs in 3NP-treated mice was not accompanied by higher stance width variability and stride length variability. Moreover, the eventual failure of the hind limbs in 3NP-treated mice (75 mg/kg cumulative dose) to engage in coordinated stepping was not preceded by changes in hind limb gait variability (50 mg/kg cumulative dose). We did not find an increase in gait variability in transgenic SOD1 G93A mice. Neither forelimb nor hind limb stride length variability or stance width variability in SOD1 G93A mice were different than in wild-type controls at ~12 weeks or ~13 weeks, ages when motor function deficiencies have been observed. In patients, gait variability was shown to be higher with well-established ALS [36]. We do not yet know if gait variability increases in SOD1 G93A mice as the disease progresses. Our findings suggest, however, that gait variability is not increased in the early stages of motor neuron disease. Differences in gait variability among MPTP-treated, 3NP-treated, and SOD1 G93A mice may reflect differences in neuropathology.
Limitations
We do not know the long-term effects of extended administrations of MPTP or 3NP on gait dynamics. Different schedules of neurotoxin administration result in differences in the mechanisms of neuronal death [34,70], which could affect gait. We did not observe morbidity and mortality in the MPTP-treated mice. Results in 3NP-treated mice, however, were variable, consistent with reports of significant inter-animal variation in response to 3NP toxicity [71]. MPTP- and 3NP-induced neuronal damage in mice are age-dependent [72,73], and both toxins have systemic effects, including the heart [42,74]. Since no postmortem analyses were performed demonstrating neurodegeneration, the pathogenesis of the gait disturbances is unclear. We did not measure striatal dopamine; previous reports indicate, however, that 30 mg/kg/day MPTP for 3 days reduce striatal dopamine by >50% [18,20]. Neither the MPTP nor the 3NP toxin models exactly replicate the pathological phenomena of PD and HD. Future studies could compare gait dynamics in different chemically-induced models and genetic models of PD and HD. We did not consider effects of habituation to treadmill walking [61] on gait indices. Gait dynamics are strain-dependent [75], making it difficult to compare gait dynamics in the SOD1 G93A transgenic mouse model of ALS, which is a mix of C57BL/6 and SJL mice, to gait in the MPTP-treated and 3NP-treated C57BL/6 mice.
Conclusion
MPTP-treated mice demonstrated significant gait disturbances, including shortened stride length, increased stride frequency, and increased stride-to-stride variability, symptoms characteristic of patients with Parkinson's disease. 3NP-treated mice demonstrated an increased forelimb stride-to-stride variability and a more open paw placement angle of the hind limbs. Gait failure in 3NP-treated mice resulted from an inability of the hind limbs to engage in stepping while forelimb gait remained intact. Gait variability was not significantly higher in SOD1 G93A mice, a model of motor neuron disease, compared to wild-type control mice. The present study provides a basis for additional studies of gait and gait variability in mouse models of PD, HD, and ALS.
Competing interests
Thomas G. Hampton is owner of Mouse Specifics, Inc., a company organized to commercialize the gait imaging technology described in the methods.
Authors' contributions
IA participated in data collection, analyses, interpretation, and manuscript preparation.
AK assisted in the design and development of the gait imaging system and developed the software for analyses of gait data via ventral plane videography. AK also participated in the collection and analyses of data. SM participated in the design of the walking compartment for mice on the moving treadmill belt, and participated in the collection of data and in manuscript preparation. SG participated in the design of the treadmill system, automation of image acquisition and modulation of treadmill belt speed. SG also participated in manuscript preparation. JPM participated in study design, pharmacology and physiology, data interpretation, and manuscript review. TGH designed the study, and participated in the collection and analyses of data, data interpretation, and manuscript preparation and submission.
Supplementary Material
Additional File 1
Movie of the ventral view of a C57BL/6J saline-treated mouse walking at a speed of 34 cm/s. File is playable using Windows Media Player.
Click here for file
Additional File 2
Movie of the ventral view of a 3NP-treated (cumulative dose 75 mg/kg) C57BL/6J mouse attempting to walk at a speed of 34 cm/s, demonstrating coordinated gait of the forelimbs but gait failure of the hind limbs. Compare this to the coordinated gait of the forelimbs and hind limbs in a saline-treated C57BL/6J mouse (Additional file 1). Files areplayable using Windows Media Player.
Click here for file
Acknowledgements
I. Amende was generously supported by Förderkreis zur Verbesserung des Gesundheitswesens e.V. We thank Walter R. Hampton and Mary K. Hampton for their valuable clinical insights. We gratefully acknowledge the excellent engineering design and craftsmanship of MK Automation (Bloomfield, CT) in the development and construction of the mouse treadmill, and Advanced Digital Vision (Natick, MA) for expertise in image capture and processing.
==== Refs
Salarian A Russmann H Vingerhoets FJ Dehollain C Blanc Y Burkhard PR Aminian K Gait assessment in Parkinson's disease: toward an ambulatory system for long-term monitoring IEEE Trans Biomed Eng 2004 51 156 159 10.1109/TBME.2004.827933
Weller C O'Neill CJ Charlett A Bowes SG Purkiss A Nicholson PW Dobbs RJ Dobbs SM Defining small differences in efficacy between anti-parkinsonian agents using gait analysis: a comparison of two controlled release formulations of levodopa/decarboxylase inhibitor Br J Clin Pharmacol 1993 35 379 385 8485018
Bartolic A Pirtosek Z Rozman J Ribaric S Postural stability of Parkinson's disease patients is improved by decreasing rigidity Eur J Neurol 2005 12 156 159 15679705 10.1111/j.1468-1331.2004.00942.x
Nieuwboer A Dom R De Weerdt W Desloovere K Fieuws S Broens-Kaucsik E Abnormalities of the spatiotemporal characteristics of gait at the onset of freezing in Parkinson's disease Mov Disord 2001 16 1066 1075 11748737 10.1002/mds.1206
Thaut MH Miltner R Lange HW Hurt CP Hoemberg V Velocity modulation and rhythmic synchronization of gait in Huntington's disease Mov Disord 1999 14 808 819 10495043 10.1002/1531-8257(199909)14:5<808::AID-MDS1014>3.0.CO;2-J
Koller WC Trimble J The gait abnormality of Huntington's disease Neurology 1985 35 1450 1454 3162109
Bilney B Morris ME Churchyard A Chiu E Georgiou-Karistianis N Evidence for a disorder of locomotor timing in Huntington's disease Mov Disord 2005 20 51 57 15390128 10.1002/mds.20294
Tian J Herdman SJ Zee DS Folstein SE Postural stability in patients with Huntington's disease Neurology 1992 42 1232 1238 1534875
Hausdorff JM Cudkowicz ME Firtion R Wei JY Goldberger AL Gait variability and basal ganglia disorders: stride-to-stride variations of gait cycle timing in Parkinson's disease and Huntington's disease Mov Disord 1998 13 428 437 9613733 10.1002/mds.870130310
Blin O Ferrandez AM Serratrice G Quantitative analysis of gait in Parkinson patients: increased variability of stride length J Neurol Sci 1990 98 91 97 2230833 10.1016/0022-510X(90)90184-O
Schaafsma JD Giladi N Balash Y Bartels AL Gurevich T Hausdorff JM Gait dynamics in Parkinson's disease: relationship to Parkinsonian features, falls and response to levodopa J Neurol Sci 2003 212 47 53 12809998 10.1016/S0022-510X(03)00104-7
Djaldetti R Melamed E New drugs in the future treatment of Parkinson's disease J Neurol 2002 249 II30 35 12375061
Bonelli RM Wenning GK Kapfhammer HP Huntington's disease: present treatments and future therapeutic modalities Int Clin Psychopharmacol 2004 19 51 62 15076012 10.1097/00004850-200403000-00001
Fernagut PO Diguet E Labattu B Tison F A simple method to measure stride length as an index of nigrostriatal dysfunction in mice J Neurosci Methods 2002 113 123 130 11772434 10.1016/S0165-0270(01)00485-X
Carter RJ Lione LA Humby T Mangiarini L Mahal A Bates GP Dunnett SB Morton AJ Characterization of progressive motor deficits in mice transgenic for the human Huntington's disease mutation J Neurosci 1999 19 3248 3257 10191337
Kopin IJ MPTP: an industrial chemical and contaminant of illicit narcotics stimulates a new era in research on Parkinson's disease Environ Health Perspect 1987 75 45 51 3319563
Sedelis M Hofele K Auburger GW Morgan S Huston JP Schwarting RK MPTP susceptibility in the mouse: behavioral, neurochemical, and histological analysis of gender and strain differences Behav Genet 2000 30 171 182 11105391 10.1023/A:1001958023096
Jakowec MW Petzinger GM.1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine-lesioned model of Parkinson's disease, with emphasis on mice and nonhuman primates Comp Med 2004 54 497 513 15575363
Gupta M Felten DL Gash DM MPTP alters central catecholamine neurons in addition to the nigrostriatal system Brain Res Bull 1984 13 737 742 6335836 10.1016/0361-9230(84)90234-X
Sedelis M Schwarting RK Huston JP Behavioral phenotyping of the MPTP mouse model of Parkinson's disease Behav Brain Res 2001 125 109 125 11682102 10.1016/S0166-4328(01)00309-6
Schulz JB Matthews RT Klockgether T Dichgans J Beal MF The role of mitochondrial dysfunction and neuronal nitric oxide in animal models of neurodegenerative diseases Mol Cell Biochem 1997 174 193 197 9309687 10.1023/A:1006852306789
Santamaria A Perez-Severiano F Rodriguez-Martinez E Maldonado PD Pedraza-Chaverri J Rios C Segovia J Comparative analysis of superoxide dismutase activity between acute pharmacological models and a transgenic mouse model of Huntington's disease Neurochem Res 2001 26 419 424 11495354 10.1023/A:1010911417383
Guyot MC Hantraye P Dolan R Palfi S Maziere M Brouillet E Quantifiable bradykinesia, gait abnormalities and Huntington's disease-like striatal lesions in rats chronically treated with 3-nitropropionic acid Neuroscience 1997 7 45 56 9178864 10.1016/S0306-4522(96)00602-1
Brouillet E Hantraye P Ferrante RJ Dolan R Leroy-Willig A Kowall NW Beal MF Chronic mitochondrial energy impairment produces selective striatal degeneration and abnormal choreiform movements in primates Proc Natl Acad Sci U S A 1995 92 7105 7109 7624378
Diguet E Fernagut PO Wei X Du Y Rouland R Gross C Bezard E Tison F Deleterious effects of minocycline in animal models of Parkinson's disease and Huntington's disease Eur J Neurosci 2004 19 3266 3276 15217383 10.1111/j.0953-816X.2004.03372.x
Dunham NW Miya TS A note on a simple apparatus for detecting neurological deficit in rat and mice J Am Pharm Ass 1957 46 208 209
Rozas G Lopez-Martin E Guerra MJ Labandeira-Garcia JL The overall rod performance test in the MPTP-treated-mouse model of Parkinsonism J Neurosci Methods 1998 83 165 175 9765130 10.1016/S0165-0270(98)00078-8
Willis GL Donnan GA Histochemical, biochemical and behavioural consequences of MPTP treatment in C-57 black mice Brain Res 1987 402 269 274 3493826 10.1016/0006-8993(87)90033-3
Fernagut PO Diguet E Stefanova N Biran M Wenning GK Canioni P Bioulac B Tison F Subacute systemic 3-nitropropionic acid intoxication induces a distinct motor disorder in adult C57Bl/6 mice: behavioural and histopathological characterisation Neuroscience 2002 114 1005 1017 12379255
Fernagut PO Diguet E Jaber M Bioulac B Tison F Dopamine transporter knock-out mice are hypersensitive to 3-nitropropionic acid-induced striatal damage Eur J Neurosci 2002 15 2053 2056 12099912 10.1046/j.1460-9568.2002.02047.x
Weihmuller FB Hadjiconstantinou M Bruno JP Acute stress or neuroleptics elicit sensorimotor deficits in MPTP-treated mice Neurosci Lett 1988 85 137 42 3258974 10.1016/0304-3940(88)90443-0
Ryu JK Kim J Cho SJ Hatori K Nagai A Choi HB Lee MC McLarnon JG Kim SU Proactive transplantation of human neural stem cells prevents degeneration of striatal neurons in a rat model of Huntington disease Neurobiol Dis 2004 16 68 77 15207263 10.1016/j.nbd.2004.01.016
Ogawa N Hirose Y Ohara S Ono T Watanabe Y A simple quantitative bradykinesia test in MPTP-treated mice Res Commun Chem Pathol Pharmacol 1988 50 435 441 3878557
Fornai F Schluter OM Lenzi P Gesi M Ruffoli R Ferrucci M Lazzeri G Busceti CL Pontarelli F Battaglia G Pellegrini A Nicoletti F Ruggieri S Paparelli A Sudhof TC Parkinson-like syndrome induced by continuous MPTP infusion: convergent roles of the ubiquitin-proteasome system and alpha-synuclein Proc Natl Acad Sci U S A 2005 2 3413 3418 15716361 10.1073/pnas.0409713102
Vieregge P Stolze H Klein C Heberlein I Gait quantitation in Parkinson's disease – locomotor disability and correlation to clinical rating scales J Neural Transm 1997 104 237 248 9203085 10.1007/BF01273184
Hausdorff JM Lertratanakul A Cudkowicz ME Peterson AL Kaliton D Goldberger AL Dynamic markers of altered gait rhythm in amyotrophic lateral sclerosis J Appl Physiol 2000 88 2045 2053 10846017
Kale A Amende I Meyer GP Crabbe JC Hampton TG Ethanol's effects on gait dynamics in mice investigated by ventral plane videography Alcohol Clin Exp Res 2004 28 1839 1848 15608600 10.1097/01.ALC.0000148103.09378.81
Hampton TG Stasko MR Kale A Amende I Costa AC Gait dynamics in trisomic mice: quantitative neurological traits of Down syndrome Physiol Behav 2004 82 381 389 15276802 10.1016/j.physbeh.2004.04.006
Gurney ME Pu H Chiu AY Dal Canto MC Polchow CY Alexander DD Caliendo J Hentati A Kwon YW Deng HX Chen W Zhai P Sufit RL Siddique T Motor neuron degeneration in mice that express a human Cu, Zn superoxide dismutase mutation Science 1994 264 1772 1775 8209258
Colotla VA Flores E Oscos A Meneses A Tapia R Effects of MPTP on locomotor activity in mice Neurotoxicol Teratol 1990 12 405 407 2392101 10.1016/0892-0362(90)90061-G
Shimoji M Zhang L Mandir AS Dawson VL Dawson TM Absence of inclusion body formation in the MPTP mouse model of Parkinson's disease Brain Res Mol Brain Res 2005 134 103 108 15790534 10.1016/j.molbrainres.2005.01.012
Gabrielson KL Hogue BA Bohr VA Cardounel AJ Nakajima W Kofler J Zweier JL Rodriguez ER Martin LJ de Souza-Pinto NC Bressler J Mitochondrial toxin 3-nitropropionic acid induces cardiac and neurotoxicity differentially in mice Am J Pathol 2001 159 1507 1520 11583977
Fischer LR Culver DG Tennant P Davis AA Wang M Castellano-Sanchez A Khan J Polak MA Glass JD Amyotrophic lateral sclerosis is a distal axonopathy: evidence in mice and man Exp Neurol 2004 185 232 240 14736504 10.1016/j.expneurol.2003.10.004
Puttaparthi K Gitomer WL Krishnan U Son M Rajendran B Elliott JL Disease progression in a transgenic model of familial amyotrophic lateral sclerosis is dependent on both neuronal and non-neuronal zinc binding proteins J Neurosci 2002 22 8790 8796 12388585
Barneoud P Lolivier J Sanger DJ Scatton B Moser P Quantitative motor assessment in FALS mice: a longitudinal study Neuroreport 1997 8 2861 2865 9376520
Hamers FP Lankhorst AJ van Laar TJ Veldhuis WB Gispen WH Automated quantitative gait analysis during overground locomotion in the rat: Its application to spinal cord contusion and transection injuries J Neurotrauma 2001 18 187 201 11229711 10.1089/08977150150502613
Fleming SM Salcedo J Fernagut PO Rockenstein E Masliah E Levine MS Chesselet MF Early and progressive sensorimotor anomalies in mice overexpressing wild-type human alpha-synuclein J Neurosci 2004 24 9434 9440 15496679 10.1523/JNEUROSCI.3080-04.2004
Lin CH Tallaksen-Greene S Chien WM Cearley JA Jackson WS Crouse AB Ren S Li XJ Albin RL Detloff PJ Neurological abnormalities in a knock-in mouse model of Huntington's disease Hum Mol Genet 2001 10 137 144 11152661 10.1093/hmg/10.2.137
Carter RJ Lione LA Humby T Mangiarini L Mahal A Bates GP Dunnett SB Morton AJ Characterization of progressive motor deficits in mice transgenic for the human Huntington's disease mutation J Neurosci 1999 19 3248 3257 10191337
Naver B Stub C Moller M Fenger K Hansen AK Hasholt L Sorensen SA Molecular and behavioral analysis of the R6/1 Huntington's disease transgenic mouse Neuroscience 2003 122 1049 1057 14643771 10.1016/j.neuroscience.2003.08.053
Muralikrishnan D Mohanakumar KP Neuroprotection by bromocriptine against 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine-induced neurotoxicity in mice FASEB J 1998 12 905 912 9657530
Rousselet E Joubert C Callebert J Parain K Tremblay L Orieux G Launay JM Cohen-Salmon C Hirsch EC Behavioral changes are not directly related to striatal monoamine levels, number of nigral neurons, or dose of parkinsonian toxin MPTP in mice Neurobiol Dis 2003 14 218 228 14572444 10.1016/S0969-9961(03)00108-6
Tsai YF Tsai HW Tai MY Lu KS Age-related changes in locomotor behavior induced by MPTP in rats Neurosci Lett 1991 129 153 155 1922966 10.1016/0304-3940(91)90743-D
Clarke KA Still J Gait analysis in the mouse Physiol Behav 1999 66 723 729 10405098 10.1016/S0031-9384(98)00343-6
Prochazka A Gillard D Bennett DJ Positive force feedback control of muscles J Neurophysiol 1997 77 3226 3236 9212270
Koutouzis TK Borlongan CV Scorcia T Creese I Cahill DW Freeman TB Sanberg PR Systemic 3-nitropropionic acid: long-term effects on locomotor behavior Brain Res 1994 646 242 246 8069670 10.1016/0006-8993(94)90085-X
Cools AR Jongen-Relo AL Role of neostriatum and nucleus accumbens in stepping induced by apomorphine and dexamphetamine Brain Res Bull 1991 26 909 917 1933411 10.1016/0361-9230(91)90256-J
Shimano Y Kumazaki M Sakurai T Hida H Fujimoto I Fukuda A Nishino H Chronically administered 3-nitropropionic acid produces selective lesions in the striatum and reduces muscle tonus Obes Res 1995 3 779S 784S 8653562
Abernethy B Hanna A Plooy A The attentional demands of preferred and non-preferred gait patterns Gait Posture 2002 15 256 265 11983500 10.1016/S0966-6362(01)00195-3
Shear DA Haik KL Dunbar GL Creatine reduces 3-nitropropionic-acid-induced cognitive and motor abnormalities in rats Neuroreport 2000 11 1833 1837 10884028
Smith MA Brandt J Shadmehr R Motor disorder in Huntington's disease begins as a dysfunction in error feedback control Nature 2000 403 544 549 10676962 10.1038/35000576
Wooley CM Sher RB Kale A Frankel WN Cox GA Seburn KL Gait analysis detects early changes in transgenic SOD1(G93A) mice Muscle Nerve 2005 32 43 50 15880561 10.1002/mus.20228
Kuo JJ Schonewille M Siddique T Schults AN Fu R Bar PR Anelli R Heckman CJ Kroese AB Hyperexcitability of cultured spinal motoneurons from presymptomatic ALS mice J Neurophysiol 2004 91 571 575 14523070 10.1152/jn.00665.2003
Dengler R Konstanzer A Kuther G Hesse S Wolf W Struppler A Amyotrophic lateral sclerosis: macro-EMG and twitch forces of single motor units Muscle Nerve 1990 13 545 550 2366827 10.1002/mus.880130612
Brach JS Berlin JE VanSwearingen JM Newman AB Studenski SA Too much or too little step width variability is associated with a fall history in older persons who walk at or near normal gait speed Journal of NeuroEngineering and Rehabilitation 2005
Menz HB Latt MD Tiedemann A Mun San Kwan M Lord SR Reliability of the GAITRite walkway system for the quantification of temporo-spatial parameters of gait in young and older people Gait Posture 2004 20 20 25 15196515 10.1016/S0966-6362(03)00068-7
Gabell A Nayak US The effect of age on variability in gait J Gerontol 1984 39 662 666 6491179
Budsberg SC Verstraete MC Soutas-Little RW Force plate analysis of the walking gait in healthy dogs Am J Vet Res 1987 48 915 918 3605807
Cohen AH Gans C Muscle activity in rat locomotion: movement analysis and electromyography of the flexors and extensors of the elbow J Morphol 1975 146 177 196 1152066 10.1002/jmor.1051460202
Bezard E Dovero S Bioulac B Gross C Effects of different schedules of MPTP administration on dopaminergic neurodegeneration in mice Exp Neurol 1997 148 288 292 9398471 10.1006/exnr.1997.6648
Brownell AL Chen YI Yu M Wang X Dedeoglu A Cicchetti F Jenkins BG Beal MF 3-Nitropropionic acid-induced neurotoxicity – assessed by ultra high resolution positron emission tomography with comparison to magnetic resonance spectroscopy J Neurochem 2004 89 1206 1214 15147513 10.1111/j.1471-4159.2004.02408.x
Gupta M Gupta BK Thomas R Bruemmer V Sladek JR JrFelten DL Aged mice are more sensitive to 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine treatment than young adults Neurosci Lett 1986 70 326 331 3490639 10.1016/0304-3940(86)90573-2
Brouillet E Jenkins BG Hyman BT Ferrante RJ Kowall NW Srivastava R Roy DS Rosen BR Beal M Age-dependent vulnerability of the striatum to the mitochondrial toxin 3-nitropropionic acid J Neurochem 1993 60 356 359 8417157
Fuller RW Hahn RA Snoddy HD Wikel JH Depletion of cardiac norepinephrine in rats and mice by 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) Biochem Pharmacol 1984 33 2957 2960 6333240 10.1016/0006-2952(84)90593-8
Leblond H L'Esperance M Orsal D Rossignol S Treadmill locomotion in the intact and spinal mouse J Neurosci 2003 23 11411 11419 14673005
|
16042805
|
PMC1201165
|
CC BY
|
2021-01-04 16:37:40
|
no
|
J Neuroengineering Rehabil. 2005 Jul 25; 2:20
|
utf-8
|
J Neuroeng Rehabil
| 2,005 |
10.1186/1743-0003-2-20
|
oa_comm
|
==== Front
J Neuroengineering RehabilJournal of NeuroEngineering and Rehabilitation1743-0003BioMed Central London 1743-0003-2-251608684310.1186/1743-0003-2-25ResearchAttention demanding tasks during treadmill walking reduce step width variability in young adults Grabiner Mark D [email protected] Karen L [email protected] Department of Movement Sciences, University of Illinois at Chicago, USA2005 8 8 2005 2 25 25 27 4 2005 8 8 2005 Copyright © 2005 Grabiner and Troy; licensee BioMed Central Ltd.2005Grabiner and Troy; 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 variability of step time and step width is associated with falls by older adults. Further, step time is significantly influenced when performing attention demanding tasks while walking. Without exception, step time variability has been reported to increase in normal and pathologically aging older adults. Because of the role of step width in managing frontal plane dynamic stability, documenting the influence of attention-demanding tasks on step width variability may provide insight to events that can disturb dynamic stability during locomotion and increase fall risk. Preliminary evidence suggests performance of an attention demanding task significantly decreases step width variability of young adults walking on a treadmill. The purpose of the present study was to confirm or refute this finding by characterizing the extent and direction of the effects of a widely used attention demanding task (Stroop test) on the step width variability of young adults walking on a motorized treadmill.
Methods
Fifteen healthy young adults walked on a motorized treadmill at a self-selected velocity for 10 minutes under two conditions; without performing an attention demanding task and while performing the Stroop test. Step width of continuous and consecutive steps during the collection was derived from the data recorded using a motion capture system. Step width variability was computed as the standard deviation of all recorded steps.
Results
Step width decreased four percent during performance of the Stroop test but the effect was not significant (p = 0.10). In contrast, the 16 percent decrease in step width variability during the Stroop test condition was significant (p = 0.029).
Conclusion
The results support those of our previous work in which a different attention demanding task also decreased step width variability of young subjects while walking on a treadmill. The decreased step width variability observed while performing an attention demanding task during treadmill walking may reflect a voluntary gait adaptation toward a more conservative gait pattern emphasizing frontal plane control of the trunk. Extension of the experimental paradigm to older adults and mechanistic approaches to link step width variability to dynamic stability, and falls, in a cause-effect manner are necessary.
==== Body
Introduction
Dynamic stability during locomotion can be negatively affected by concomitant information processing and the effect appears to increase with age [1]. These effects have generally been studied using a dual-task paradigm that introduces performance of an attention demanding secondary task during performance of the primary task, locomotion. The basis of the dual-task paradigm is the assumption that humans possess limited information processing capacity. When simultaneously performing the primary and secondary tasks, each of which require some level of attention, a negative influence on the performance of either task is reflective of task interference [1]. The interference may indicate structural interference or capacity interference. The former is associated with tasks that share common input and output resources whereas the latter is associated with the total information processing capacity having been exceeded.
The use of dual-task paradigms to investigate locomotion is, in part, based on the frequency with which locomotion, generally considered a highly automated motor task, is performed concurrently with cognitive tasks. The changes in reaction time and gait-related variables [e.g., [2-5]] reported for older adults during dual-task paradigms have been associated with increased fall-risk. For example, performing a verbal reaction time task during an obstacle avoidance task significantly increased the risk of obstacle contact by young adults [6]. In another study, a verbal reaction time task increased the risk of obstacle contact by both younger and older adults although the increase was larger in the older adults [7]. These results broadly suggest that performing cognitive tasks during locomotion may increase the risk of tripping. It is notable that secondary task interference on a primary task may be eliminated by practice [8] and, not surprisingly, priority can be directed at the primary task at the expense of decreasing performance on the secondary task [9].
Aging is associated with an increasingly conservative walking pattern marked by changes in basic step kinematics. The changes, which include decreased step length, increased step width and increased double support time, may reduce fall-risk. For example, longer double support time translates to a longer period of time during which the vertical projection of the total body center of mass is within the base of support. A larger step width essentially extends the lateral margins of the base of support and perhaps improves laterally directed control of whole body center of mass position and velocity. The variability of step kinematics has been strongly linked with falls by older adults. In particular, cross-sectional and prospective studies have consistently linked increased step time variability to falls by normally aging [10-12] and pathologically aging older adults [13,14]. Although older adults without a history of falls appear to have increased step width and step width variability compared to young adults [15], a prospective study reported that increased step width and decreased step width variability discriminated older adults who fell from those who did not fall [16]. If viewed as reflecting the presence of noise in a physiological system, variability is deleterious. In this view the decreased step width variability of the older adults who fell is counterintuitive. However, if viewed as arising from multiple interacting control systems, variability has been suggested as reflecting a desirable trait of an adaptive system [17]. From this viewpoint, the diminished step width variability of the older adults who fell is a consistent observation. The biomechanical and physiological significance of altered step width variability (and step time variability), however, has not yet been defined.
The apparent relationship between increased fall-risk when performing an attention demanding task while walking, and the relationship between step kinematic variability and fall risk raises the question of whether attention demanding tasks influence step kinematic variability. There is evidence that the answer is affirmative For example, the step time variability of patients with Parkinson's Disease and Alzheimer's disease, which is significantly larger than that of healthy controls, demonstrates additional and significant increases when walking and performing an attention demanding task [13,14]. Regarding step width variability, in one study performance of an attention demanding task (walking with a cup of water placed in a saucer) did not influence step width variability of either young or older adults [18]. However, methodologic issues may have influenced that result. These issues, specifically related to the technology used to measure step width and number of consecutive steps used to compute step width variability, were subsequently resolved using different instrumentation [19]. Using this technology an attention demanding task (maintaining the beam of an activated laser pointer within a target area) was found to significantly decrease step width variability in young adults [20]. This was a surprising finding in light of the body of literature indicating that increased step time variability is associated with both normal and pathological aging.
Our current interest in step width variability is driven by its role in maintaining laterally directed dynamic stability during gait and its potential utility as a clinical index of laterally directed dynamic stability. Our long-term goal is to determine if subtle changes in step width variability can alter the sensitivity of laterally directed dynamic stability to unexpected postural disturbances and thereby increase fall risk in older adults. The purpose of the present study was to confirm or refute the findings of Walters et al. [20] by characterizing the extent and direction of the effects of an attention demanding task (Stroop test) on the step width variability of young adults during treadmill walking.
Methods
Fifteen young healthy individuals (8 males and 7 females, age: 24.5 ± 3.4 years, height: 1.66 ± 0.12 m, and mass: 68.5 ± 8.0 kg) volunteered to participate in the study. The protocol was reviewed and approved institutionally and all subjects provided written informed consent prior to participation in the study.
The experiment consisted of three protocols, the order of which was randomly assigned and performed in a single laboratory session. In one protocol, subjects walked on a motorized treadmill at a self-selected speed for 10 minutes. This served as the control walking condition. During the second 10 minute protocol, the subjects walked on the treadmill at the same self-selected speed while performing an attention demanding task [[21], described in the next section]. The third protocol, conducted with the subjects in an upright standing position, provided a baseline measure of performance of the attention demanding task.
During the control walking condition, subjects were asked to walk while looking straight ahead at a wall that was approximately five meters away. The attention demanding task was the Stroop test. During the Stroop test images consisting of the name of one of four colors, printed in text of a different color, were projected onto the wall. The height of the letters, when projected on the wall was 15 cm. The images changed at a frequency of one Hz. The subjects were instructed to verbally identify the color of the text and to ignore the word itself. Incorrect answers were recorded by an investigator. The metric of Stroop test performance was the percentage of wrong answers.
Step width was quantified using motion analysis. The longitudinal axes of the right and left feet were marked using passively reflecting markers placed on the shoe over the heel and the over the third metatarsal. The height of the heel marker was matched to the height of the metatarsal marker. The motion of the reflecting markers was recorded using an eight-camera motion analysis system (Motion Analysis, Santa Rosa, CA, USA) operating at 60 Hz.
For each pair of sequential left-right foot placements, step width was calculated using the midpoints of the foot segments using a custom MATLAB algorithm [22]. Global coordinates were aligned so that the direction of the walk was along the Y-axis and the step width was measured in the X direction. Stance phase was characterized by the period during which the vertical position of the heel markers was approximately zero. Step width was computed as the distance, in the X direction, between the positions of the midpoint of the feet during two sequential stance phases. The midpoint of each foot was calculated as the midpoint of the segment marked by the reflective markers placed over the heel and metatarsal. Step width variability was calculated as the standard deviation of step width from all of the collected steps [19].
The effect of performing the Stroop test on step width and step width variability was determined by comparing the values to those of the control walking condition using paired t-tests. Paired t-tests were used to compare the error rate on the Stroop test during the upright standing condition to that during the control walking condition. A Pearson correlation was calculated to describe the relationship between the error rates on the Stroop test during the control walking and standing conditions. All analyses were performed using SPSS (Version 12.0).
Results
Performance of the Stroop test while walking had a significant influence on step width variability. Compared to the control walking condition, step width variability decreased 16 percent while performing the Stroop test (p = 0.029). The step width variability measured during the control condition and the Stroop test condition was 22.4 ± 5.5 mm and 18.9 ± 4.5 mm, respectively (Figure 1).
Figure 1 Composite means and standard deviations of step width variability for the control walking condition and the Stroop test condition are illustrated with individual subject data point pairs (n = 15).
In contrast to step width variability, performing the Stroop test did not affect step width. Compared to the control walking condition, step width decreased four percent while performing the Stroop test (p = 0.10). The step width measured during the control condition and the Stroop test condition was 152.0 ± 28.3 mm and 146.1 ± 27.1 mm, respectively.
Compared to the standing condition, walking on the treadmill appeared to diminish performance on the Stroop test. During the upright standing condition the error rate was 2.4 ± 3.5 percent whereas during the treadmill test the error rate more than doubled to 5.2 ± 4.7 percent (p = 0.052). Both values were significantly different than zero as indicated by a one-sample t-test (p = 0.02 and 0.001 for the standing and walking conditions, respectively). Notably, there was virtually no relationship between the error rates on the Stroop test during the control walking and standing conditions (r = 0.01, p = 0.73).
Discussion
The purpose of the present study was to confirm or refute a previous finding of decreased step width variability while walking on a treadmill and performing an attention demanding task [20]. The results confirm those previous results by demonstrating a significant decrease in the step width variability of young adults performing the Stroop test while walking on a treadmill. In the previous study [20], maintaining the beam of a handheld laser pointer within the boundary of a target placed about two meters in front of the subject was associated with a 12.2 percent decrease in step width variability; from 20.5 ± 4.1 mm to 18.0 ± 3.8 mm (p < 0.001). These values bear notable similarity to those observed in the present work.
The findings appear consistent in context with a number of related published studies. Performance of attention demanding tasks increases fall-risk by older adults (2–5). In addition, decreased step width variability distinguished older adults who, in a prospective study, fell from those who did not fall (10). Thus, the present results may implicate performance of attention demanding tasks with changes to a characteristic of gait previously associated with falls by older adults. However, in the absence of any published data related to the influence of performing attention demanding tasks on the step width variability of older adults the implication is indirect.
Despite the consistency of the findings during treadmill walking, however, the present data differ from those of Bauby and Kuo [23] who observed a 53 percent increase in step width variability of young subjects during overground walking with their eyes closed. The amplitude and direction differences between our results and those of Bauby and Kuo may reflect differences in the availability of vision. Humans veer, or deviate from straight line walking, after just a couple of meters without vision [24]. It is possible in the experiment of Bauby and Kuo, during which subjects received verbal stimuli to help them maintain a straight line gait, that the increased step width variability resulted from an interaction between the veering due to the absence of vision and the corrections in response to the verbal stimuli. In the present study, visual information was not absent although the extent to which it was available for guidance may have been reduced due to need to direct vision at the projection of the Stroop test words. Thus, the between-study differences in protocols render meaningful comparison of the results difficult. However, the biomechanical and physiological significance of these disparate findings may have considerable clinical importance. It is possible that directional changes (increase vs. decrease) are contextual and must be considered relative to the specific experimental conditions. In addition, it may be that changes in step width variability can not be considered in isolation from other relevant variables. For example, in the present study and that of Walter et al. [20] there was no effect of the attention demanding task on step width. In contrast, in the study of Bauby and Kuo step width increased by 11 percent. In the work of Maki [16], the older adults who fell demonstrated increased step width (compared to young adults) and decreased step width variability compared to older adults who did not fall.
A question raised by the present results is whether decreased step width variability, an outcome of performing an attention demanding task during treadmill walking, is causally linked to falls in the same manner as is apparent in the results of Maki [16]. From an empirical standpoint, step width variability may represent a manifestation of a mechanism underlying frontal plane control of the trunk. For example, external pelvic stabilization significantly reduced step width variability of young subjects walking on a treadmill by 33 percent [25]. This accompanied a 60 percent decrease in the peak lateral displacement of the center of mass, which implies a reduction in the amplitude of the trunk motion. In other studies, step width variability was significantly reduced while walking with a cane and maintaining contact between a hand and a wall [26] and grasping handles while walking on a motorized treadmill [15]; all of which would be expected to decrease the amplitude of frontal plane trunk motion. If so, then step width variability would be expected to parallel changes in the variability of frontal plane trunk motion. It merits mention that although recent data argue to the contrary [27] pilot data in our laboratory suggest a strong and statistically significant relationship between step width variability and the variability of frontal plane trunk kinematics.
Subtle age-related changes in control of step width may be associated with similarly subtle changes in frontal plane trunk control. The mass of the trunk and its location relative to the base of support during gait underscores the need for active dynamic stabilization in the lateral direction [23]. Disturbances that influence the position, velocity and acceleration of the trunk relative to the base of support, could lead to potentially deleterious biomechanical events. Further investigation of the relationship between step kinematic variability, biomechanics of the trunk, and motor control of the trunk seem warranted. This is particularly relevant for older adults, for whom control of the trunk appears to be decreased. For example, when subjected to a 7.5 degree laterally directed tilt of the platform on which they stood, the trunk of young subjects moved in the direction opposite to that of the tilt within 30 milliseconds [28]. In contrast, the response latency of the older adults was greater than 150 milliseconds and the subsequent trunk motion was in the direction of the impending fall.
If decreased step width variability is associated with decreased frontal plane trunk motion, it may be reasonable to expect that decreased step width variability reflects increased dynamic stability. If so, performance of the attention demanding task in the present study may have caused subjects to adopt a more conservative gait pattern, implying an increase in the voluntary control of gait. This makes sense given the visual resources invested in performing the Stroop test. Reduced availability of visually-derived information of the limbs and treadmill may increase uncertainty about foot placement. Given that a step causing a foot to be placed to some extent off the treadmill belt would be a destabilizing event that subjects could be expected to want to avoid. Indeed, one might speculate that the potential for a considerable destabilizing event might be associated with increased trunk stiffness, a condition that can significantly increase the risk for laterally directed falls [25,29]. Thus, it is proposed that in this manner, decreased step width variability reflects decreased dynamic stability.
The published work related to the influence of an attention demanding task on step width variability is quite limited. However, there is a considerable body of literature related to step time variability as it relates to normal and pathological aging. This literature consistently reports increased step time variability in older adults with a history of falls [10], patients with Huntington's disease [30], Parkinson's disease [14,30] and cardiovascular disease [31]. Notably, healthy older adults have been reported to have step time variability that is not different from that of young adults [10] although step width variability of healthy older adults is significantly larger than that of young adults [15]. The functional meaning of the directional differences in the effect of performing an attention demanding task on step time variability (increased variability) and step width variability (decreased variability) have not been resolved at this time. However, the opposite directions in which the changes occur provide an impetus to more fully investigate the relationship between changes in spatial and temporal step kinematic variability as well as the extent to which these variables provide dependent or independent information related to the neuromuscular control gait. Further work that characterizes the mechanisms by which subtle changes in step time variability and step width variability can be causally related to falls by older adults seems warranted.
Two methodological issues, which limit the extent to which results may be generalized, appear to warrant further study. The first relates to the uncertainty of the extent to which step kinematic variability measured during treadmill walking reflects that measured during unrestricted overground walking. Previous work has suggested that with respect to the variability of spatial step kinematics treadmill walking may be an acceptable representation of overground walking [32]. In light of the need to acquire hundreds of continuous steps for the accurate calculation of step kinematic variability [19,23] the methodological solutions to the question, although available, have yet to be applied. From the standpoint of clinical utility, it is not necessary for treadmill walking to perfectly represent overground walking. It may be sufficient for treadmill walking to be a reliable and valid surrogate for overground walking.
The second issue, perhaps the more easily addressed of the two, relates to the present study having been limited to young subjects. Clearly, the danger of falls and fall-related injuries is an issue that is of greatest interest as it relates to older adults. Because the magnitude of the effect of performing attention demanding tasks, and thus fall-risk, increases with age [7] the present results provide the impetus to extend the hypotheses and method to older adults. Our previous experience provides a basis for the expectation that healthy older subjects will demonstrate decreased step width variability under the described experimental conditions.
In conclusion, step width variability of young adults has been shown to be significantly decreased by the concurrent performance of an attention demanding task. This finding is consistent with our previous pilot work and may have important clinical ramifications. Because step width variability reflects frontal plane dynamic stability, disturbances to step width variability could reflect increased fall-risk. If this is so, measures of step width variability could potentially provide the means to clinically track age-related changes to and the effects of interventions on dynamic stability. To that end, mechanistic studies linking step width variability changes to altered dynamic stability, and falls, in a cause-effect manner are necessary.
Authors' contributions
MDG conceived the study, evaluated the data and results and was responsible for the initial drafting of the manuscript.
KLT wrote/modified software necessary for the analysis and was involved in drafting and revising the manuscript.
Both authors read and approved the final manuscript.
Acknowledgements
This work was partially funded by NIA R01AG10557. The author(s) declare that they have no competing interests. The authors wish to acknowledge the assistance of Rijuta Dhere who was instrumental in the collection of the data.
==== Refs
Woollacott M Shumway-Cook A Attention and the control of posture and gait: a review of an emerging area of research Gait Posture 2002 16 1 14 12127181 10.1016/S0966-6362(01)00156-4
Lajoie Y Teasdale N Bard C Fleury MM Attentional demands for static and dynamic equilibrium Exp Brain Res 1993 97 139 144 8131825 10.1007/BF00228824
Ebersbach G Dimitrijevic MR Poewe W Influence of concurrent tasks on gait: a dual-task approach Percept Mot Skills 1995 81 107 113 8532444
Sparrow WA Bradshaw EJ Lamoureux E Tirosh O Ageing effects on the attention demands of walking Hum Mov Sci 2002 21 961 972 12620728 10.1016/S0167-9457(02)00154-9
Beauchet O Dubost V Gonthier R Kressig RW Dual-task-related gait changes in transitionally frail older adults: the type of the walking-associated cognitive task matters Gerontology 2005 51 48 52 15591756 10.1159/000081435
Weerdesteyn V Schillings AM van Galen GP Duysens J Distraction affects the performance of obstacle avoidance during walking J Mot Behav 2003 35 53 63 12724099
Chen HC Schultz AB Ashton-Miller JA Giordani B Alexander NB Guire KE Stepping over obstacles: dividing attention impairs performance of old more than young adults J Gerontol A Biol Sci Med Sci 1996 51 M116 122 8630704
Ruthruff E Jonston JC van Selst M Whitsell S Remington R Vanishing dual-task interference after practice: Has the bottleneck been eliminated or is it merely latent? J Exp Psychol 2003 29 280 289
Schrodt LA Mercer VS Giuliani CA Hartman M Characteristics of stepping over an obstacle in community dwelling older adults under dual-task conditions Gait Posture 2004 19 279 287 15125917 10.1016/S0966-6362(03)00067-5
Hausdorff JM Edelberg HK Mitchell SL Goldberger AL Wei JY Increased gait unsteadiness in community-dwelling elderly fallers Arch Phys Med Rehabil 1997 78 278 283 9084350 10.1016/S0003-9993(97)90034-4
Hausdorff JM Nelson ME Kaliton D Layne JE Bernstein MJ Nuernberger A Singh MA Etiology and modification of gait instability in older adults: a randomized controlled trial of exercise J Appl Physiol 2001 90 2117 2129 11356774
Hausdorff JM Rios DA Edelberg HK Gait variability and fall risk in community-living older adults: a 1-year prospective study Arch Phys Med Rehabil 2001 82 1050 1056 11494184 10.1053/apmr.2001.24893
Sheridan PL Solomont J Kowall N Hausdorff JM Influence of executive function on locomotor function: divided attention increases gait variability in Alzheimer's disease J Am Geriatr Soc 2003 51 1653 1637 10.1046/j.1532-5415.2003.51516.x
Hausdorff JM Balash J Giladi N Effects of cognitive challenge on gait variability in patients with Parkinson's Disease J Geriatr Psych Neurol 2003 16 53 58
Owings TM Grabiner MD Variability of step kinematics in young and older adults Gait Posture 2004 20 26 29 15196516 10.1016/S0966-6362(03)00088-2
Maki BE Gait changes in older adults: predictors of falls or indicators of fear J Am Geriatr Soc 1997 45 313 320 9063277
Lipsitz LA Physiological complexity, aging, and the path to frailty
Grabiner PC Biswas ST Grabiner MD Age-related changes of spatial and temporal stride kinematic variability Arch Phys Med Rehabil 2001 82 31 35 11239283 10.1053/apmr.2001.18219
Owings TM Grabiner MD Measuring step kinematic variability on an instrumented treadmill: how many steps are enough? J Biomech 2003 36 1215 1218 12831749 10.1016/S0021-9290(03)00108-8
Walters A Owings TM Grabiner MD Performance of attention-splitting tasks has different effects on static and dynamic stability Proceedings of the 27th annual meeting of the American Society of Biomechanics, Toledo, Ohio 2003
Stroop JR Studies of interference in serial verbal reactions J Exp Psychol 1935 18 643 661
Dhere R Troy KL Grabiner MD Step width variability is insensitive to measurement using either kinetic and kinematic methods Gait Posture in review
Bauby CE Kuo A Active control of lateral balance in human walking J Biomech 2000 33 1433 1440 10940402 10.1016/S0021-9290(00)00101-9
Vuillerme N Nougier V Camicioli M Veering in human locomotion Neurosci Lett 2002 331 175 178 12383925 10.1016/S0304-3940(02)00876-5
Donnelan JM Shipman DW Kram R Kuo AD Mechanical and metabolic requirements for active lateral stabilization in human walking J Biomech 2004 37 827 835 15111070 10.1016/j.jbiomech.2003.06.002
Richardson JK Thies SB DeMott TK Ashton-Miller JA Interventions improve gait regularity in patients with peripheral neurpathy while walking on an irregular surface under low light J Am Geriatr Soc 2004 52 510 515 15066064 10.1111/j.1532-5415.2004.52155.x
Moe-Nilssen R Helbostad JL Interstride trunk acceleration variability but not step width variability can differentiate between fit and frail older adults Gait Posture 2005 21 164 170 15639395 10.1016/j.gaitpost.2004.01.013
Allum JHJ Carpenter MG Honegger F Adkin AL Bloem BR Age-dependent variations in the directional sensitivity of balance corrections and compensatory arm movements in man J Physiol 2002 542 643 663 12122159 10.1113/jphysiol.2001.015644
Gruneberg C Bloem BR Honegger F Allum JH The influence of artificially increased hip and trunk stiffness on balance control in man Exp Brain Res 2004 157 472 485 15138751
Hausdorff JM Cudkowicz ME Firtion R Wei JY Goldberger AL Gait variability and basal ganglia disorders stride to stride variations of gait cycle timing in Parkinson's and Huntington's Disease Movement Dis 1998 13 428 437 9613733 10.1002/mds.870130310
Hausdorff JM Forman DE Ladin Z Goldberger AL Rigney DR Wei JY Decreased walking variability in elderly persons with congestive heart failure J Am Geriatr Soc 1994 42 1056 1061 7930329
Owings TM Grabiner MD Step width variability, but not step length variability or step time variability, discriminates gait of healthy young and older adults during treadmill locomotion J Biomech 2004 37 935 938 15111081 10.1016/j.jbiomech.2003.11.012
|
16086843
|
PMC1201166
|
CC BY
|
2021-01-04 16:37:40
|
no
|
J Neuroengineering Rehabil. 2005 Aug 8; 2:25
|
utf-8
|
J Neuroeng Rehabil
| 2,005 |
10.1186/1743-0003-2-25
|
oa_comm
|
==== Front
Lipids Health DisLipids in Health and Disease1476-511XBioMed Central London 1476-511X-4-151609296810.1186/1476-511X-4-15ResearchRelationship between Sialic acid and metabolic variables in Indian type 2 diabetic patients Nayak B Shivananda [email protected] Geetha [email protected] Department of Biochemistry, Kasturba Medical College, Manipal- 576104, India2 Department of Preclinical Sciences, Biochemistry unit, University of West Indies, Trinidad2005 10 8 2005 4 15 15 15 6 2005 10 8 2005 Copyright © 2005 Nayak B and Bhaktha; licensee BioMed Central Ltd.2005Nayak B and Bhaktha; 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
Plasma sialic acid is a marker of the acute phase response. Objective is to study the relationship between sialic acid relationship with metabolic variables in Indian type 2 diabetes with and without microvascular complications.
Research design and Methods
Fasting Venous blood samples were taken from 200 subjects of which 50 were of diabetes mellitus (DM) and nephropathy patients, 50 patients with type 2 diabetes and retinopathy, 50 patients with type 2 diabetes without any complications and 50 healthy individuals without diabetes. The Indian subject's aged 15–60 years with type 2 diabetes were recruited for the study. Simultaneously urine samples were also collected from each of the subjects. All the blood samples were analyzed for total cholesterol, triglyceride (TG), low-density lipoprotein (LDL), high-density lipoprotein (HDL), fasting and postprandial glucose on fully automated analyzer. Serum and urine sialic acid along with microalbumin levels were also estimated.
Results
There was a significantly increasing trend of plasma and urine sialic acid with severity of nephropathy (P < 0.001) and with degree of urinary albumin excretion (P < 0.001). Serum sialic acid correlated with increasing serum creatinine concentration (P < 0.001). Elevated serum sialic acid concentrations were also associated with several risk factors for diabetic vascular disease: diabetes duration, HbA1c, serum triglyceride and cholesterol concentrations, waist-to-hip ratio and hypertension. Significant correlations were found between sialic acid concentration and cardiovascular risk factors like LDL and TG in the diabetic subjects.
Conclusion
The main finding of this study is that elevated serum and urinary sialic acid and microalbumin concentrations were strongly related to the presence of microvascular complications like diabetic nephropathy and retinopathy and cardiovascular risk factors in Indian type 2 diabetic subjects. Further study of acute-phase response markers and mediators as indicators or predictors of diabetic microvascular complications is therefore justified.
Sialic acidMicroalbuminuriamicrovascularand type-2 diabetes mellitus
==== Body
Introduction
Diabetes mellitus is a group of metabolic disorders characterized by elevation of blood glucose concentration and is associated with increased prevalence of microvascular complications. Type 1 diabetes mellitus results from cellular mediated autoimmune destruction of pancreatic β-cells of islets of langerhans and results in loss of insulin production. Type 2 diabetes mellitus is the most common form of diabetes accounting for 90% of cases. An estimated 16 million Americans have type-2 diabetes, and half are unaware they have it. Type 2 diabetes is characterized by insulin resistance or abnormal insulin secretion.
One of the more debilitating aspects of diabetes is the numerous complications that can arise from the disease. These complications include diabetic retinopathy, kidney nephropathy and peripheral neuropathy. The development and severity of these complications are dependent on the duration of the disease and how well it is managed. Prospective studies have reported associations among various markers of inflammation and incidence of diabetes [1], and it has been proposed that inflammation has a causal role in the development of diabetes [2]. Diabetes is another risk factor for myocardial infarction and stroke [3,4]. The relationship between diabetes and other traditional cardiovascular risk factors, e.g., an adverse lipid profile, obesity, hypertension and physical inactivity explain the increased risk in diabetic individuals. Even though it has been suggested that inflammation contributes to the increased incidence of cardiovascular diseases among diabetic subjects and only few prospective studies have addressed this question [1]. Plasma sialic acid is one of the markers for acute phase response [5]. Sialic acid is a terminal component of the non-reducing end of carbohydrate chains of glycoproteins and glycolipids [6]. Elevated total serum sialic acid (SA) concentration is a risk factor for cardiovascular mortality in humans [7]. Increased total serum sialic acid leads to increased excretion of sialic acid in urine of the patient presented with high urinary microalbumin.
It has been reported earlier that total serum sialic acid concentration increase in type 2 diabetes mellitus associated with microvascular complications [8,9]. The aim of this study was to measure serum and urine sialic acid and their relation with urinary microalbumin, serum cholesterol, TG, LDL cholesterol in diabetic subjects with and without microvascular complications.
Microalbumin is a risk factor for cardiovascular disease; it may be associated with chronic inflammation and investigated the relationship of urinary albumin excretion and urinary sialic acid.
Materials and methods
We investigated the relationship of sialic acid concentrations with serum lipids, and urinary albumin excretion in Indian type 2 diabetic subjects of Kasturba Medical College Hospital, Manipal. The study includes 200 subjects (male and female). There were 50 healthy controls. The diabetic subjects were divided into three groups according to the level of complications. Group A-50 patients with diabetes mellitus (DM) and nephropathy, group B-50 patients with type 2 diabetes and retinopathy and group-C 50 patients with type 2 diabetes without any complications. The Indian subject's aged 15–60 years with type 2 diabetes were recruited for the study. The estimation of serum and urine sialic acid may prove to predictive and preventive of microvascular diseases and their complications in people with type 2 diabetes.
All the subjects were reported fasting in the morning after 10–12 hr overnight fast. Standing height and weight were measured. Body mass Index (BMI) defined as weight in kg/height (in meters) squared was calculated, and used as an index for obesity. Measurements of the waist circumference were taken at the mid point between umbilicus and xiphoid, and for hip circumference at the widest point around the hips. The waist hip ratio was thereafter calculated [10]. Blood pressure was measured according to the standard procedure [11].
Venous blood samples collected without the use of tourniquet from each of the patients were analyzed for total serum cholesterol, TG, LDL, HDL, fasting and postprandial glucose on fully automated analyzer (Hitachi 912 analyzer, Roche, Switzerland) with the reagents supplied by Roche. The HbA1c is estimated with the principle based on affinity chromatography technique.
Serum and urinary sialic acid was measured by a colorimetric assay using standard chemicals and reagents. In this method a protein precipitate of serum containing sialic acid will react with diphenylamine producing a purple color, which is quantitatively measured on a spectrophotometer at 540 nm.
The fresh urine samples collected from the test and control group subjects were used for microalbumin estimation in an electrochemiluminiscence analyzer (Roche, Switzerland).
Statistical Method
Results were expressed as mean ± S.D. except where otherwise stated. Data were analyzed using the statistical package for social science, SPSS and p value ≤ 0.05 was taken as the cut off level for significance. Because the distribution of most variables was not symmetric. We used non-parametric statistical methods. Chi square tests was be used to examine, type 2 diabetes mellitus, the various clinical and biochemical markers.
Results
The table 1 shows the relationship between serum sialic acid, urine sialic acid and microalbumin concentrations with metabolic variables in diabetic subjects with and without microvascular complications. The table depicts significant increase of serum sialic acid (< 0.001) among the Indian diabetic subjects compared to the control subjects. Furthermore, in the diabetic subjects urine sialic acid and microalbumin were significantly higher (< 0.001). The Table also shows the association of sialic acid and several risk factors for diabetic vascular disease; diabetes duration, serum TG and cholesterol concentration. It is observed that the sialic acid values were statistically significantly higher with increasing urinary albumin excretion (p < 0.001). Similarly HbA1c, FBS, PPBS, TG and cholesterol showed marked increase in patients with elevated level of microalbumin and urine sialic acid when compared to normal subjects without any complications.
Table 1 Serum and urinary sialic acid and microalbumin levels in Type 2 diabetes with nephropathy and retinopathy
Parameters Diabetes without any complications Diabetic nephropathy Diabetic Retinopathy Non-diabetic subjects p value
Serum Sialic acid (mg%) 55.05 ± 2.9* 85.05 ± 2.7*** 75.05 ± 2.5*** 46.6 ± 2.08 < 0.001
Urine sialic acid (mg%) 6.02 ± 2.58** 13.06 ± 1.58*** 11.03 ± 1.78*** 3.2 ± 0.65 < 0.001
Microalbumin (mg %) 8.2 ± 3.24 132.2 ± 35.24*** 102.2 ± 29.24*** 7.67 ± 3.28 < 0.001
FBS (mg%) 140.02 ± 70.08 155.6 ± 50.7 150.6 ± 49.9 90.02 ± 80.08 < 0.01
PPBS (mg%) 150.02 ± 102.10 207.3 ± 57.6 200.1 ± 57.6 120.02 ± 102.10 < 0.01
HbA1C (%) 9.10 ± 5.20 11.1 ± 2.3 10.1 ± 2.5 6.10 ± 5.20 < 0.05
Triglyceride (mg%) 122.04 ± 75.01 178.02 ± 78.01 180.04 ± 75.01 120.04 ± 76.01 < 0.05
Cholesterol(mg%) 148.04 ± 120.01 256.03 ± 134.01 246.03 ± 130.01 140.04 ± 119.01 NS
HDL (mg %) 35.01 ± 20.04 38.01 ± 26.02 35.01 ± 20.04 36.01 ± 19.04 NS
LDL (mg %) 90.00 ± 76.06 165.00 ± 97.01 160.00 ± 95.01 87.00 ± 76.05 < 0.05
Creatinine (mg %) 2.00 ± 1.6.06 10.05 ± 2.03 10.03 ± 2.01 1.40 ± 1.20 < 0.001
Urine creatinine (mg%) 145.00 ± 102.6 155.03 ± 65.02 150.03 ± 66.01 146.00 ± 113.06 < 0.05
Mean ± SD *** p < 0.001, NS = not significant, n = 50
Discussion
In recent years, much attention has been given to the relationships among adiposity, inflammation, and diabetes. High inflammation sensitive plasma protein levels increased the cardiovascular risk slightly more in diabetic. Studies of diabetic subjects have reported increased incidences of cardiovascular diseases or increased diabetes complications among subjects with high fibrinogen [12] and other markers of inflammation [13,14]. Measurement of inflammation sensitive markers may be useful for assessment of the cardiovascular risk in diabetic patients. Results from prospective studies suggest that inflammation involved in the pathogenesis of diabetes [15] and atherosclerosis [16]. Inflammation could be a common antecedent for both diabetes and cardiovascular disease. Hyperglycemia and insulin resistance could also promote inflammation, and may be factor linking diabetes to the development of atherosclerosis. Elevated glucose levels could promote inflammation by increased oxidative stress [17]. Yet another possibility is that the inflammatory response is a result of vascular complications following diabetes. In type-2 diabetes, the circulating sialic acid concentration is elevated in comparison with non-diabetic subjects [18]. The results of our study showed serum and urine SA concentration increased in diabetic patients as compared to the general population, especially in type-2 diabetic patients with either microalbuminuria or albuminuria. Furthermore, the serum and urine sialic acid levels were independent of the duration of diabetes mellitus and degree of metabolic control (as estimated by HbA1c). Also, a good correlation was observed between sialic acid and important cardiovascular risk factors such as cholesterol, LDL and TG.
It has been reported that serum sialic acid levels are increased in type1 DM patients with albuminuria [19]. Several authors found the increased urinary concentration of sialic acid in type 2 diabetes with microangiopathy. The vascular permeability is regulated by sialic acid moieties, with increased vascular permeability resulting from the shedding of vascular endothelial sialic acid into the circulation. It is well established that vascular endothelium carries a high level of sialic acid [20], and the vascular damage leads to its release into the circulation. A relationship between serum sialic acid levels and microvascular complications has been observed before for microalbuminuria and clinical proteinuria in type 1 [21] and type 2 diabetes [22].
Our findings clearly indicated that serum and urinary sialic acid concentrations were elevated in type-2 diabetes. Crook M et.al found that serum sialic acid was significantly higher in men with diabetic complications than in those without any of the complications [23]. There may be an association between sialic acid and complications through the acute phase response.
Acknowledgements
The authors are thankful to Dr. Shivashankar K N, associate professor department of Medicine for sending the blood and urine samples with the consent of the patient. The authors also thankful to the staff of Kasturba Medical College hospital, helped to measure the BMI, height and weight of the subjects. We are grateful to Kasturba Medical College, Manipal Academy of Higher Education (MAHE- A Deemed University), Manipal, India for the constant support and encouragement throughout this study.
==== Refs
Pradhan AD Manson JE Rifai N Buring JE Ridker PM C-reactive protein, interleukin 6, and risk of developing type 2 diabetes mellitus JAMA 2001 286 327 334 11466099 10.1001/jama.286.3.327
Pickup JC Crook MA Is type II diabetes mellitus a disease of the innate immune system? Diabetologia 1998 41 1241 1248 9794114 10.1007/s001250051058
Vilbergsson S Sigurdsson G Sigvaldason H Sigfusson N Coronary heart disease mortality amongst non-insulin-dependent diabetic subjects in Iceland: the independent effect of diabetes: the Reykjavik Study 17-year follow up J Intern Med 1998 244 309 316 9797494 10.1046/j.1365-2796.1998.00368.x
Kannel WB D'Agostino RB Wilson PW Belanger AJ Gagnon DR Diabetes, fibrinogen, and risk of cardiovascular disease: the Framingham experience Am Heart J 1990 120 672 676 2389702 10.1016/0002-8703(90)90026-T
Pickup JC Mattock MB Chusney GD Burt D NIDDM as a disease of the innate immune system: association of acute phase reactants and interleukin-6 with metabolic syndrome X Diabetologia 1997 40 1286 1292 9389420 10.1007/s001250050822
Ng S Dain JA Rosenbur, Schengrund S The natural occurrence of sialic acid Biological role of sialic acid 1976 New York, Plenum 59 102
Taniuchi K Chifu K Hayashi N Nakamachi Y Yamaguchi N Miyamato Y Doi K Buba S Uchida Y Tsukada Y Sugimori T A new enzymatic method for the determination of sialic acid in serum and its application for a marker of acute phase reactants Kobe J Med Sci 1981 27 91 102 7021999
Crook M Tutt P Pickup JC Elevated serum sialic acid determination in non-Insulin dependent diabetes and its relationship to blood pressure and retinopathy Diabetes care 1993 16 57 60 8422833
Crook MA Earle A Morocutti A Yip J Vibereti GC Pickup JC Serum sialic acid, a risk factor for cardiovascular disease, is increased in IDDM patients with microalbuminuria and clinical proteinuria Diabetes Care 1994 17 305 310 8026286
Ashwell M Chinn S Stalley S Garrow JS Female fat distribution, a photographic and cellularity study Int J Obes 1978 2 289 302 721368
Cohen JD Grimm RH Smith WM The multiple risk factor intervention trial (MRFIT) VI: intervention on blood pressure Prev Med 1981 10 501 518 7027240 10.1016/0091-7435(81)90062-1
Ganda OP Arkin CF Hyperfibrinogenemia: an important risk factor for vascular complications in diabetes Diabetes Care 1992 15 1245 1250 1425083
Stehouwer CD Gall MA Twisk JW Knudsen E Emeis JJ Parving HH Increased urinary albumin excretion, endothelial dysfunction, and chronic low-grade inflammation in type 2 diabetes: progressive, interrelated, and independently associated with risk of death Diabetes 2002 51 1157 1165 11916939
Jager A van Hinsbergh VW Kostense PJ Emeis JJ Nijpels G Dekker JM Heine RJ Bouter LM Stehouwer CD C-reactive protein and soluble vascular cell adhesion molecule-1 are associated with elevated urinary albumin excretion but do not explain its link with cardiovascular risk Arterioscler Thromb Vasc Biol 2002 22 593 598 11950696 10.1161/01.ATV.0000013786.80104.D4
Ford ES Body mass index, diabetes, and C-reactive protein among U.S. adults Diabetes Care 1999 22 1971 1977 10587828
Ross R Atherosclerosis: an inflammatory disease N Engl J Med 1999 340 115 126 9887164 10.1056/NEJM199901143400207
Baynes JW Thorpe SR Role of oxidative stress in diabetic complications: a new perspective on an old paradigm Diabetes 1999 48 1 9 9892215
Crook MA Tutt P Simpson H Pickup JC Kuroda T Nago N Matsua H Shimada K Serum sialic acid and acute phase proteins in type 1 and 2 diabetes Clin Chim Acta 1993 219 131 138 7508342 10.1016/0009-8981(93)90204-H
Powie JK Watts GF Crook MA Ingham JN Shaw KM Serum sialic acid and the long term complications of insulin dependent diabetes mellitus Diabet Med 1996 13 238 242 8689844 10.1002/(SICI)1096-9136(199603)13:3<238::AID-DIA29>3.0.CO;2-W
Born GV Palinski W Unusually high concentration of sialic acid on the surface of vascular endothelium Br J Exp Pathol 1985 66 543 549 4063159
Yokoyama H Jensen JS Jensen T Deckert T Serum sialic acid concentration is elevated in IDDM especially in early diabetic nephropathy J Intern Med 1995 237 519 523 7738493
Chen J Gall MA Yokoyama H Jensen JS Deckert M Parving HH Raised serum sialic acid concentration in NIDDM patients with and without diabetic nephropathy Diabetes Care 1996 19 130 134 8718432
Crook MA Pickup JC Lumb PJ Georgino F Webb DJ Fuller JH Relationship between Plasma Sialic Acid Concentration and Microvascular and Macrovascular Complications in Type1 Diabetes Diabetes care 2001
|
16092968
|
PMC1201167
|
CC BY
|
2021-01-04 16:39:18
|
no
|
Lipids Health Dis. 2005 Aug 10; 4:15
|
utf-8
|
Lipids Health Dis
| 2,005 |
10.1186/1476-511X-4-15
|
oa_comm
|
==== Front
Malar JMalaria Journal1475-2875BioMed Central London 1475-2875-4-371606096710.1186/1475-2875-4-37ResearchIn vitro antimalarial drug susceptibility in Thai border areas from 1998–2003 Chaijaroenkul Wanna [email protected] Bangchang Kesara [email protected] Mathirut [email protected] Stephen A [email protected] Faculty of Allied Health Sciences, Thammasat University, Rangsit, Patumthani 12121, Thailand2 Department of Parasitology, Phramongkutklao College of Medicine, Ratchathewi, Bangkok 10400, Thailand3 Division of Molecular and Biochemical Parasitology, Liverpool School of Tropical Medicine, Pembroke Place, Liverpool L35QA, UK2005 2 8 2005 4 37 37 30 6 2005 2 8 2005 Copyright © 2005 Chaijaroenkul et al; licensee BioMed Central Ltd.2005Chaijaroenkul 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 Thai-Myanmar and Thai-Cambodia borders have been historically linked with the emergence and spread of Plasmodium falciparum parasites resistant to antimalarial drugs. Indeed, the areas are often described as harbouring multi-drug resistant parasites. These areas of Thailand have experienced significant changes in antimalarial drug exposure patterns over the past decade. This study describes the in vitro antimalarial susceptibility patterns of 95 laboratory-adapted P. falciparum isolates, collected between 1998 and 2003,.
Methods
Ninety five P. falciparum isolates were collected from five sites in Thailand between 1998 and 2003. After laboratory adaptation to in vitro culture, the susceptibility of these parasites to a range of established antimalarial drugs (chloroquine [CQ], mefloquine [MQ], quinine [QN] and dihydroartemisinin [DHA]) was determined by the isotopic microtest.
Results
Mefloquine (MQ) sensitivity remained poorest in areas previously described as MQ-resistant areas. Sensitivity to MQ of parasites from this area was significantly lower than those from areas reported to harbour moderate (p = 0.002) of low level MQ resistance (p = 000001). Importantly for all drugs tested, there was a considerable range in absolute parasite sensitivities. There was a weak, but statistically positive correlation between parasite sensitivity to CQ and sensitivity to both QN and MQ and a positive correlation between MQ and QN. In terms of geographical distribution, parasites from the Thai-Cambodia were tended to be less sensitive to all drugs tested compared to the Thai-Myanmar border. Parasite sensitivity to all drugs was stable over the 6-year collection period with the exception of QN.
Conclusion
This study highlights the high degree of variability in parasite drug sensitivity in Thailand. There were geographical differences in the pattern of resistance which might reflect differences in drug usage in each area. In contrast to many other studies there were weak, but statistically significant positive, correlations between sensitivity to CQ and sensitivity to MQ and QN. Over the six years of sample collection, parasite sensitivity appears to have stabilized to these drugs in these sites.
==== Body
Background
Malaria is still a major health problem in Thailand, especially along the Thai-Myanmar and Thai-Cambodia borders where multi-drug resistant malaria is highly prevalent [1]. Chloroquine-resistant (CQR) parasites were first reported in the late 1950s from Southeast-Asia and South-America. Since those early reports CQR has spread throughout the malaria endemic countries of the world. The Malaria Control Programme of the Ministry of Public Health of Thailand was established in 1963. The objective of the control programme is to monitor the susceptibility of Plasmodium falciparum to currently used antimalarial drugs using both in vitro and in vivo tests with the ultimate goal of providing effective malaria control strategies and delaying the emergence of drug resistance [2]. Both approaches to the sensitivity assessment have their strengths and weaknesses [3]. The in vitro sensitivity monitoring system is considered a suitable system for assessing absolute sensitivity without the confounding influences of host-related factors, such as host immunity and drug pharmacokinetics. The results from in vitro tests, therefore, provide a more objective insight into inherent drug sensitivity than do in vivo tests. However, compared to the in vivo test, there are technical requirements which make this type of analysis operationally more difficult.
Several in vitro sensitivity test systems have been developed and applied to sensitivity monitoring of P. falciparum in endemic areas. These include traditional in vitro tests based on the measurement of the effect of drugs on the growth and development of malaria parasites, i.e., schizont maturation or growth inhibition [4,5], incorporation of radiolabeled precursors [6], enzymatic activity of parasite lactate dehydrogenase (pLDH) [7] or histidine-rich protein II (HRP II) [8]. The in vitro sensitivity test based on the standard micro-technique recommended by the World Health Organization [5] using the schizont maturation inhibition test has been applied successfully in certain highly multi-drug resistant areas of Thailand. However, the technique requires immediate laboratory work at the sites, where availability of qualified personnel is not readily available. One approach to solve this problem is to collect patient's blood samples and perform the test at the central laboratories where qualified personnel exist. This approach requires preservation of parasite samples at low temperature (-196°C) in a liquid nitrogen tank and adaptation of the parasite isolates to short-term culture prior to in vitro sensitivity testing. An additional advantage of this approach is that it allows multiple assessment of drug susceptibility, increasing confidence in the data collected, compared to the one-off drug sensitivity analysis afforded by the use of freshly collected isolates. The main purpose of this study was to explore the in vitro susceptibility of recently collected and adapted P. falciparum isolates to chloroquine, quinine, mefloquine and dihydroartemisinin in order to establish the degree of variability in parasite sensitivity, any geographical patterns and the stability of parasite sensitivity over a six-year collection period.
Methods
Parasite isolates
This study was carried out between 1998–2003, P. falciparum isolates were collected from malaria endemic areas of Thailand (Figure 1), including the Thai-Myanmar border regions (Kanchanaburi, Tak, Ratchaburi and Ranong) and the Thai-Cambodia border (Chantaburi). Approval of the study protocol was obtained from the Ethics Committee of the Ministry of Public Health, Thailand. Fresh isolates of P. falciparum were collected from patients with acute uncomplicated falciparum malaria who presented at Mae Sot General Hospital and Malaria Clinics located in the various provinces. Inclusion criteria included those who had no previous history of antimalarial treatment within the preceding one month and with an asexual parasitaemia between 1,000 and 80,000/μl. All gave informed consent for study participation. Three to five millilitres of blood were collected into EDTA tubes from patients with a confirmed diagnosis of uncomplicated P. falciparum malaria. The fresh blood samples were then centrifuged to remove the buffy coat and cryopreserved in liquid nitrogen following the method of Rowe [9] before being transported to the laboratory.
Figure 1 Map of Thailand showing the sampling sites: Tak, Kanchanaburi, Ratchaburi, Ranong and Chantaburi provinces.
Parasites were maintained in in vitro culture using the method of Trager and Jensen [10]. Briefly, infected blood samples were removed from liquid nitrogen tank storage and thawed in a 37°C water bath. An equal volume of sterile 3.5% sodium chloride was added, and the mixture was centrifuged at 1000 × g for 5 min. The pellet was washed three times and re-suspended with RPMI-1640 medium supplemented with 20% human serum and placed in a 25 ml tissue culture flask in a total volume of 10 ml containing a 5% RBC suspension. The flask was flushed with a gas mixture of 5%CO2, 5% O2 and 90% N2 and incubated at 37°C. The culture medium was changed once a day; group O red blood cells were added to maintain the 5% cell suspension. Parasite growth was monitored by Giemsa-stained (2% v/v, pH 6.8, 30 min) thin smear examination.
In vitro sensitivity tests of P. falciparum isolates
Drug testing was performed using the cultured parasites once the parasitaemia of the culture reached an optimum density (0.5%–1% parasitaemia and 1.5% haematocrit). In vitro susceptibility was tested by monitoring [3H] hypoxanthine uptake [6]. Each drug was prepared as 10 mM stock solution and further diluted in RPMI-1640 medium to the desired concentrations. The plates (96-wells microtitre plates) were dosed with antimalarial drugs at a total of eight final concentrations as follows: mefloquine (1, 5, 10, 25, 50, 100, 150, 200 nM), chloroquine (5, 10, 25, 50, 100, 150, 250, 500 nM), quinine (10, 25, 50, 100, 150, 250, 500, 1,000 nM), and dihydroartemisinin (0.1, 0.2, 0.4, 0.8, 1.0, 5.0, 7.5, 10.0 nM). Ten microlitres of blood (1% parasitaemia, 20% haematocrit) was dispensed into each well of the sterile plates, followed by 100 μl of drug solution in RPMI-1640 without [3H] hypoxanthine. The plates were incubated at 37°C in a candle jar for 24 hours, pulsed with 5 μl of [3H] hypoxanthine solution (0.1 mCi/ml), and reincubated for an additional 24 hours. The plates were harvested using a Tomtec March III M semi-automatic harvester, using Wallac A Printed Filtermats (Finland), dried, and mixed with 5 ml of scintillation fluid. Each filter mat was then placed in a cassette and the radioactivity were measured using a 1450 MicroBeta Trilux liquid scintillation and luminescence counter (Wallac, Finland). The drug concentration that inhibited 50% parasite growth (IC50) were determined. Reported data represent the mean results from at least three independent drug sensitivity tests.
Statistical analysis
The in vitro activity of antimalarials is presented as the geometric mean of the IC50s for all isolates. The Mann-Whitney U-test was used to determine whether the observed differences in the in vitro response of geographically distinct parasites to antimalarial drugs were significantly different. The potential for in vitro cross-resistance was evaluated by standard linear regression analysis. For all statistical tests the significance level (p) was set at 0.05.
Results
In vitro susceptibility of P. falciparum to antimalarials
A total of 95 P. falciparum isolates were collected from five malaria endemic areas of Thailand: Mae-sot District, Tak (27), Ratchaburi (1), Kanchanaburi (37), Ranong (16) and Chantaburi Province (14). The in vitro susceptibility data are shown in Figure 2A to 2C. The geometric mean 50% inhibitory concentration (IC50) and 95%CI were 74.3 (45.2–122.2), 162.9 (91.9–288.8), 22.2 (10.8–45.6) and 1.35 (0.73–2.47) nM for CQ, QN, MQ and DHA respectively. There were no major differences in drug susceptibility over the six years the samples were collected with the exception of quinine sensitivity, which was lower in the isolates collected in 2003 (Table 1).
Figure 2 The scatter plot between the IC50 (nM) and the mefloquine resistance areas and the border regions. A represents the chloroquine data, B the quinine data and C the mefloquine data. (a) is the geometric mean 50% inhibitory concentration in vitro with the 95% CI and (b) is the median.
Table 1 In vitro susceptibility of Thai P. falciparum isolates according the years
Year GMIC50 CQ (nM) GMIC50 QN (nM) GMIC50 MQ (nM) GMIC50 DHA (nM)
Mean 95%CI Mean 95%CI Mean 95%CI Mean 95%CI
1998 72.5 56.7–89.7 178.3 121.1–262.4 20.1 11.7–34.5 1.07 0.82–1.39
2000 65.1 44.8–94.5 243.4 163.7–361.8 28.1 15.8–50.1 1.64 0.92–2.92
2002 67.8 53.2–86.5 184.7 152.1–224.5 30.4 25.5–36.1 1.46 1.16–1.84
2003 79.1 68.8–90.1 136b 115.2–160.4 19.1 15.3–23.7 1.35 1.12–1.63
a Geometric mean 50% inhibitory concentration in vitro
b Significant difference with 2000 (p = 0.012) and 2002 (p = 0.024)
In vitro susceptibility and geographical distribution
The malarious areas of Thailand have been categorized by the clinical level of mefloquine sensitivity i.e. high level MQ resistance (cure rate of MQ 750 mg is less than 50%; Tak and Chantaburi provinces), moderate level MQ resistance (cure rate between 50% and 70%; Kanchanaburi and Ratchaburi provinces) and low level MQ resistance (cure rate more than 70%; Ranong province). These groupings were maintained in the isolates studied here in vitro. Interestingly, parasites from the least MQ resistant area tended to demonstrate highest susceptibility to the other drugs tested.
In vitro cross-resistance
In contrast to previous studies conducted on African and Thai isolates [11,12], there was a significant positive correlation between the in vitro susceptibility to CQ and susceptibility to MQ (n = 91, Pearson r = 0.075, p = 0.010) and QN (n = 93, Pearson r = 0.134, p < 0.0001) (Figure 3). Similarly there was a significant positive correlations between IC50 values of QN and the IC50 values of MQ (n = 91, Pearson r = 0.230, p < 0.0001). The strong correlation between the activities of MQ and QN was not surprising given the structural similarities between these two drugs and has been noticed by several authors [13,14]. In contrast, susceptibility to DHA did not correlate with susceptibility to any of the other drugs tested (data not shown).
Figure 3 The scatter diagram and the regression line representing the relationship between IC50 (nM) values of CQ, QN and MQ.
Discussion
In response to concerns about the malaria situation in Thailand, the government established the Malaria Control Programme since 1963. This programme has monitored and modified Thai policy on malaria control which was last revised in 1995. The major aim of the programme is to monitor the evolution of drug resistant malaria parasites [2]. The ability to implement a quick and relatively cheap assay to predict drug susceptibility of P. falciparum infections is vitally important, so that the spread of drug resistant parasites can be monitored and controlled. The first report of CQ resistance came from Thailand's borders in the late 1950s. More recently, MQ resistance was shown to develop rapidly after its introduction into clinical practice. This problem has been remedied by the introduction and deployment of the artemisinins in combination with MQ. Experience with CQ has highlighted the utility of drug sensitivity assays in plotting the emergence and spread of resistance. More recently molecular tools have found utility. For example the use of the K76T mutation in pfcrt gene as a marker for CQ susceptibility is the most useful molecular marker for CQ resistance, however, some parasite isolates that are sensitive to CQ also carried this mutation [11,15] and the use of this strategy is based on a full knowledge of all potential resistance mutations. Furthermore, the large variability in parasite sensitivity to drugs such as CQ suggests that although one molecular event may be the dominant control of susceptibility, other molecular events must contribute to this variability [16].
There have been many studies looking at parasite chemosensitivity in S.E.Asia. In general, the pattern of sensitivity shows consistency across the region, although parasite responses in parasites from Vietnam tend to show better responses to drugs [17,18]. Decreased susceptibility to quinine was first reported at the beginning of the 1980s in patients living near the Thai-Cambodia border [19,20]. The data reported in this current study indicate that the P. falciparum susceptibility to QN is stable or may have slightly improved recently. With respect to the other drugs studied against parasite sensitivity apperas to have stabilized.
Several studies have reported evidence of cross-resistance among certain group of quinoline-containing antimalarial drugs, i.e., QN, MQ and HF, while showing an inverse relationship between this group of drugs and sensitivity to CQ [14,21-24]. Similarly, in other studies, the sensitivity of P. falciparum field isolates to the two structurally related drugs QN and MQ was strongly correlated [13,14,25]. This pattern of cross-resistance was also confirmed in this study. In contrast, a weak, but statistically significant, positive correlation was observed between sensitivity to CQ and sensitivity to either MQ or QN. The mechanisms behind this observation remains to be determined, but it does suggest that additional factors must be operational in these parasites compared to those displaying a clear negative correlation in their susceptibility patterns. Although a number of reports showed a positive correlation between in vitro susceptibilities of DHA and quinolines such as mefloquine and quinine, this cross-resistance was not apparent in this study [11,12,25]. In terms of distribution within Thailand, parasites from the Thai-Cambodia tended to be more resistant to all drugs evaluated compared to parasites from the Thai-Myanmar border.
The in vitro sensitivity data from adapted parasites showed the clearer view of drug profiles, which allows multiple assessment of drug testing. The drug profile in this study showed that the pattern of MQ resistance was spread to most of endemic areas, especially along the borders. The old drug policy was based on the level of MQ resistance areas, which treated the falciparum malaria with single dose of MQ 750 mg in low MQ resistance area, whereas the moderate and high MQ resistance areas were treated with multiple dose of MQ 750 mg plus 300 mg and 500 mg of artesunate at 6-hour interval, respectively [2]. From the fact of increasing the MQ resistance, the malaria control policy has been revised, which aim to reduce the MQ resistance. Thus, the combination of artemisinin derivatives with MQ has been introduced as first line drug for treating non-severe falciparum malaria throughout the country. However, QN still remains the first line drug for severe malaria treatment [26].
Conclusion
This study indicates a relative stability in parasite susceptibility to four key drugs in the border areas of Thailand. Notably, resistance to CQ and MQ remains at a high level although there is substantial variation. There are indications that QN sensitivity may have increased slightly and, importantly for combination chemotherapy, parasite susceptibility to the artemisinin derivative, DHA remains high. This supports the decision of the Thai malaria control programme to introduce artemisinin containing MQ combinations as first line malaria treatment for the country.
Acknowledgements
This work was supported by Thailand Research Fund for the scholarship of The Royal Golden Jubilee (RGJ) Ph.D. programme to W.C. M.M. and K.N. were supported by Thailand-Tropical Disease Research Programme (ID 02-2-MAL-24-047). We would like to thank Malaria Research Unit, Chulalongkorn University, Thailand for parasites isolates.
==== Refs
Wongsrichanalai C Pickard AL Wernsdorfer WH Meshnick SR Epidemiology of drug-resistant malaria Lancet Infect Dis 2002 2 209 218 11937421 10.1016/S1473-3099(02)00239-6
Malaria-Division Malaria Control Programme in Thailand 1998 Malaria Division, Department of Communicable Disease Control (CDC) of the Ministry of Public Health, Thailand
World Health Organization Chemotherapy of Malaria and Resistance in Plasmodium falciparum: What next? Trends Parasitol 2001 17 582 588 11756042 10.1016/S1471-4922(01)02085-2
Rieckmann KH Campbell GH Sax LJ Mrema JE Drug sensitivity of Plasmodium falciparum. An in-vitro microtechnique Lancet 1978 i 22 23 74500 10.1016/S0140-6736(78)90365-3
World Health Organization In vitro microtest (MARK II) for the assessment of the response of Plasmodium falciparum to chloroquine, mefloquine, quinine, sulfadoxine/pyrimentamine and amodiaquine 1990 World Health Organization: Geneva, Switzerland
Desjardins RE Canfield CJ Haynes JD Chulay JD Quantitative assessment of antimalarial activity in vitro by a semiautomated microdilution technique Antimicrob Agents Chemother 1979 16 710 718 394674
Makler MT Hinrichs DJ Measurement of the lactate dehydrogenase activity of Plasmodium falciparum as an assessment of parasitemia Am J Trop Med Hyg 1993 48 205 210 8447524
Noedl H Wernsdorfer WH Miller RS Wongsrichanalai C Histidine-rich protein II: a novel approach to malaria drug sensitivity testing Antimicrob Agents Chemother 2002 46 1658 1664 12019072 10.1128/AAC.46.6.1658-1664.2002
Rowe AW Eyster E Kellner A Liquid nitrogen preservation of red blood cells for transfusion; a low glycerol-rapid freeze procedure Cryobiology 1968 5 119 128 5717951
Trager W Jensen JB Human malaria parasites in continuous culture Science 1976 193 673 675 781840
Ringwald P Bickii J Basco LK In vitro activity of dihydroartemisinin against clinical isolates of Plasmodium falciparum in Yaounde, Cameroon Am J Trop Med Hyg 1999 61 187 192 10463665
Brockman A Price RN Van Vugt M Heppner DG Walsh D Sookto P Wimonwattrawatee T Looareesuwan S White NJ Nosten F Plasmodium falciparum antimalarial drug susceptibility on the northwestern border of Thailand during five years of extensive use of artesunate-mefloquine Trans R Soc Trop Med Hyg 2000 94 537 544 11132385 10.1016/S0035-9203(00)90080-4
Wongsrichanalai C Wimonwattratee T Sookto P Laoboonchai A heppner DG Kyle DE Wernsdorfer WH In vitro sensitivity of Plasmodium falciparum to artesunate in Thailand Bull World Health Organ 1999 77 392 398 10361756
Suebsaeng L Wernsdorfer WH Rooney M Sensitivity to quinine and mefloquine of Plasmodium falciparum in Thailand Bull World Health Organ 1986 64 759 765 3542267
Labbe AC Patel S Crandall I Kain KC Molecular surveillance system for global patterns of drug resistance in imported malaria Emerg Infec Dis 2003 9 33 36 12533279
Mu J Ferdig MT Feng X Joy DA Duan J Furuya T Subramanian G Aravind L Cooper RA Wootton JC Xiong M Su XZ Multiple transporters associated with malaria parasite responses to chloroquine and quinine Mol Microbiol 2003 49 977 989 12890022 10.1046/j.1365-2958.2003.03627.x
Singhasivanon P Mekong Malaria. Malaria, multi-drug resistance and economic development in the greater Mekong subregion of Southeast Asia Southeast Asian J Trop Med Public Health 1999 30 1 101
Huong NM Hewitt S Davis TME Dao Le D Toan TQ Kim TB Hanh NT Phuong VN Nahn DH Cong Le D Resistance of Plasmodium falciparum to antimalarial drugs in a highly endemic area of southern Viet Nam: a study in vivo and in vitro Trans R Soc Trop Med Hyg 2001 95 325 329 11491008 10.1016/S0035-9203(01)90254-8
Giboda M Desnis MB Response of Kampucheam strains of Plasmodium falciparum to antimalarials: in vivo assessment of quinine and quinine plus tetracycline; multiple drug resistance in vitro J Trop Med Hyg 1988 91 205 211 3043009
Denis MB Kouznetsov RL Gibboda M In vivo response of multi-resistant Plasmodium falciparum infections to mefloquine and its combination with sulfadoxine/pyrimethamine in Cambodia Folia Parasitol 1991 38 187 188 1937278
Cowman AF Galatis d Thompson JK Selection for mefloquine resistance in Plasmodium falciparum is linked to amplification of the pfmdr1 gene and cross-resistance to halofantrine and quinine Proc Natl Acad Sci U S A 1994 91 1143 1147 8302844
Peel SA Bright P Yount B Handy J Baric RS A strong association between mefloquine and halofantrine resistance and amplification, overexpression, and mutation in the P-glycoprotein gene homolog (pfmdr) of Plasmodium falciparum in vitro Am J Trop Med Hyg 1994 51 648 658 7985758
Webster HK Boudreau EF Payanand K Yongvanitchit K Pang LW Antimalarial drug susceptibility testing of Plasmodium falciparum in Thailand using a microdilution radioisotope method Am J Trop Med Hyg 1985 34 228 235 3885770
Wilson CM Volkman SK Thaithong S Martin SK Kyle DE Milhous WK Wirth DF Amplification of pfmdr1 associated with mefloquine and halofantrine resistance in Plasmodium falciparum from Thailand Mol Biochem Parasitol 1993 57 151 160 8426608 10.1016/0166-6851(93)90252-S
Lim P Chim P Sem R Nemh S Poravuth Y Lim C Seila S Tsuyuoka R Denis MB Socheat D Fandeur T In vitro monitoring of Plasmodium falciparum susceptibility to artesunate, mefloquine, quinine and chloroquine in Cambodia: 2001–2002 Acta Trop 2005 93 31 40 15589795 10.1016/j.actatropica.2004.09.002
Malaria Division Guideline for the treatment of uncomplicated falciparum malaria 2004 Malaria Division, Department of Communicable Disease Control (CDC) of the Ministry of Public Health, Thailand
|
16060967
|
PMC1201168
|
CC BY
|
2021-01-04 16:24:13
|
no
|
Malar J. 2005 Aug 2; 4:37
|
utf-8
|
Malar J
| 2,005 |
10.1186/1475-2875-4-37
|
oa_comm
|
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-301604277010.1186/1477-7827-3-30ResearchEvidence that 17alpha-estradiol is biologically active in the uterine tissue: Antiuterotonic and antiuterotrophic action Perusquía Mercedes [email protected] Erika [email protected] Department of Cell Biology and Physiology, Institute for Biomedical Research, National Autonomous University of Mexico (UNAM), Apartado Postal 70228, Mexico City 04510, Mexico2005 21 7 2005 3 30 30 19 4 2005 21 7 2005 Copyright © 2005 Perusquía and Navarrete; licensee BioMed Central Ltd.2005Perusquía and Navarrete; 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
17alpha-Estradiol has been considered as the hormonally inactive isomer of 17beta-estradiol. Recently, nongenomic (smooth muscle relaxation) and genomic (light estrogenic activity) effects of 17alpha-estradiol have been reported, but no reports have yet determined its possible antiestrogenic activity. Therefore, this study investigated: the nongenomic action of 17alpha-estradiol on uterine contractile activity and its potential agonist-antagonist activity on uterine growth.
Methods
Uterine rings from rats were isometrically recorded. Different concentrations (0.2–200 microM) of 17alpha-estradiol were tested on spontaneous contraction and equimolarly compared with 17beta-estradiol. To examine the mechanism of 17alpha-estradiol action, its effect was studied in presence of beta2-antagonist (propranolol), antiestrogens (tamoxifen and ICI 182,780) or inhibitors of protein synthesis (cycloheximide) and transcription (actinomycin D). Moreover, contractions induced by high potassium (KCl) solution or calcium in depolarized tissues by KCl-calcium free solution were exposed to 17alpha-estradiol. Collaterally, we performed an uterotrophic assay in adult ovariectomized rats measuring the uterine wet weight. The administration for three days of 0.3 microM/day/Kg 17beta-estradiol was equimolarly compared with the response produced by 17alpha-estradiol. Antiuterotrophic activity was assayed by administration of 0.3 microM/day/Kg 17beta-estradiol and various doses ratios (1:1, 1:3, 1:5, and 1:100) of 17alpha-estradiol.
Results
The estradiol isomers elicited an immediate relaxation, concentration-dependent and reversible on spontaneous contraction. 17alpha-Estradiol presented lower potency than 17beta-estradiol although it did not antagonize 17beta-estradiol-induced relaxation. Relaxation to 17alpha-estradiol was not inhibited by propranolol, tamoxifen, ICI 182,780, cycloheximide or actinomycin D. The KCl contractions were also sensitive to 17alpha-estradiol-induced relaxation and calcium contractions in depolarized tissues were markedly prevented by 17alpha-estradiol, implying a reduction of extracellular calcium influx through voltage-operated calcium channels (VOCCs). Uterotrophic assay detected significant increase in uterine weight using 17alpha-estradiol, which was significantly minor as compared with 17beta-estradiol. 17alpha-Estradiol, at all doses ratios, significantly antagonized the hypertrophic response of 17beta-estradiol.
Conclusion
17alpha-Estradiol induces a relaxing effect, which may be independent of the classical estrogen receptor, nongenomic action, apparently mediated by inactivation of VOCCs. 17alpha-Estradiol is also a weak estrogen agonist (uterotrophic response); likewise, 17alpha-estradiol may act as an antiestrogen (antiuterotrophic response). The overall data document a nongenomic relaxing action and a novel antiestrogenic action of 17alpha-estradiol, which are relevant in estrogen-mediated uterine physiology.
==== Body
Background
17α-Estradiol (17α-E2) has long been considered as the hormonally inactive isomer of 17β-estradiol (17β-E2) useful in determining the hormonal specificity of response to 17β-E2 [1,2]. Consequently, it has been generally accepted that 17α-E2 is devoid of genomic estrogenic effects [3-6]. Nevertheless, in the past few years it has been documented that 17α-E2 may induce genomic effects such as partial estrogenic activity [7-11]. In addition, this estrogen possesses important nongenomic (membrane) actions by inducing neuroprotective [12,13] and mitochondrial protective [14] effects, as well as relaxing effects in isolated vascular [15-17], uterine [18] and urinary [19] smooth muscle. In this respect it is reasonable to assume that, 17α-E2 may play a relevant physiological role, but little attention has been paid to examine its potential regulatory function.
On the other hand, the available data have shown that 17α-E2 is the predominant estrogen in some mammals, whereas only few studies exist concerning the detection of 17α-E2 in humans which has only been found in the urine and serum at low concentrations [reviewed in [20,21]]. However, is important to highlight that 17α-E2 is used as an ingredient of estrogen replacement therapy and hormone replacement therapy applied in the treatment of peri- and post-menopausal women [22].
Therefore, the present study was designed to explore the feasible actions of 17α-E2 in the uterine tissue. Specifically, we have examined the possible effects of this hormone on both nongenomic and genomic actions in the rat uterus: (1) some studies were performed on uterine contractile activity by using a well established isometric system for isolated tissue. The effects were observed by application of 17α-E2 on the spontaneous and KCl-induced myometrial contraction. The mechanism of action of 17α-E2 was delineated to determine if its potential relaxing effect on uterine contractility is genomically mediated or if this estrogen is interacting with membrane proteins (calcium channels and/or adrenoceptors); and (2) on the basis that some natural stereoisomers, as in the case of testosterone and epitestosterone which elicit nongenomic uterine relaxing action [23] and only epitestosterone has antiandrogenic activity [24-26], the estradiol isomers, 17α- and 17β-E2, should also induce agonist-antagonist activities. Thus, we have quantified estrogenicity and antiestrogenicity in a classical sense, determining these actions on uterine wet weigh. Accordingly, this study set out to investigate the potential antagonist (antiestrogenic) activity of 17α-E2 on the uterotrophic response induced by 17β-E2.
Methods
Animals
Female Wistar rats weighing 180–220 g were obtained from Charles River Breeding Laboratories (Wilmington, MA), housed in our animal facility under controlled lighting (lights-on from 0700–1900 h) and temperature (21°C) conditions, and given ad libitum water and food. The project was approved by our Animal Care Committee, and experiments were conducted in accordance with the published Guiding Principles in the Care and Use of Animals approved by the American Physiological Society. The vaginal smears of these animals were inspected daily for 2 weeks, and animals showing regular 4-day estrous cycle were selected on the day of diestrus.
Myometrium contractile activity
The rats were killed and the uterine tissues were immediately removed and transferred to warmed (37°C), oxygenated (O2/CO2 95:5) Krebs-bicarbonate solution of the following composition (mM): NaHCO3 (25), NaCl (119), KCl (4.6), KH2PO4 (1.2), MgSO4 (1.2), CaCl2 (1.5) and glucose (12), with the pH adjusted to 7.4. The uterine horns were isolated, cleaned of surrounding fat and loose connective tissue, and transversally bisected into two rings, approximately 1 cm in length.
The uterine rings were placed vertically in a 10 ml tissue chamber and bathed in Krebs-bicarbonate solution, under optimum resting force of 10 mN (1 g tension), and allowed for 30 min before starting the experiment. The contractile response of each tissue was recorded isometrically using transducers (FTO3C; Grass Instruments, Quincy, MA) connected to a polygraph (79; Grass Instruments).
After a stabilization period (1 h) of the tissues in Krebs-bicarbonate solution, the spontaneous uterine contraction was recorded for 10 min, and this was taken as the control value (100%). Immediately, 17α-E2, dissolved in absolute ethanol and added to the bath tissue in a final volume of 0.1%, was tested by adding increasing concentrations in a non-accumulative manner (in concentrations ranging from 0.2 to 200 μM; each concentration never exceeded 0.1% v/v of vehicle). Only one concentration was used for each uterine ring from different animals. The estrogen effects were also recorded for 10 min, and the response was compared with the control. In a separate group of experiments, uterine tissue was exposed to vehicle alone (0.1 % ethanol). The concentration-response curves to estrogen on spontaneous contraction were plotted, and the medium inhibitory concentration (IC50; value for estrogen concentration at which 50% of the maximum inhibition of uterine contraction was achieved) was calculated as described by Litchfield and Wilcoxon [27]. In order to evaluate the inhibitory response of 17α-E2, its potency was compared with that of 17β-E2, which was used as a positive control under the same experimental conditions.
In a series of experiments the tissues were incubated with: antiestrogens, 1 μM tamoxifen (estrogen receptor antagonist) or 1 μM ICI 182,780 (the pure estrogen receptor antagonist), as well as 10 μM actinomycin D (inhibitor of gene transcription) or 100 μM cycloheximide (inhibitor of protein synthesis), 30 min before of 17α-E2 addition at 89.39 μM (IC50). The effect of 17α-E2 was evaluated for 10 min in presence of each drug, and compared with the effect that this estrogen produces alone. In addition, the final volume of vehicle (0.1% absolute ethanol) by each substance added was observed in independent experiments.
To examine the potential blocking effect of 17α-E2 on 17β-E2-induced inhibition, the effect of both 17β- and 17α-E2 was also studied on the spontaneous contraction. The tissue was pretreated with 17β-E2 at 8.42 μM (IC50), 10 min before of 17α-E2 addition (89.39 μM; IC50), and then the response was recorded for 10 min. The relaxing response induced by both estrogens was evaluated and compared when 17α-E2 was not present. In the same way, the opposite treatment (89.39 μM 17α-E2 before 8.42 μM 17β-E2 addition) was also determined.
The effect of 17β- and 17α-E2 was also studied on the contraction induced by high potassium (KCl 40 mM) solution, after replacing normal Krebs-bicarbonate solution with an equimolar substitution of 40 mM KCl and 84 mM NaCl. Thus, KCl solution induces a tonic contraction and after a stable contractile tension was attained (~20 min) each estradiol, 17β- or 17α-E2, at 200 μM (highest concentration tested on spontaneous contraction) was separately added and the effect was recorded for 10 min. This effect was compared with their inhibitory responses at 200 μM on spontaneous contraction. Finally, the tissues were washed and a next contraction induced by KCl was observed for 60 min to check the tissue recovery. Collaterally, the contraction induced by KCl was exposed to vehicle (0.1% ethanol) alone.
Some additional experiments were carried out to analyze the possible interaction of 17α-E2 with the β-adrenoceptors. For this purpose, 5 μM of noradrenaline (β2-adrenoceptor agonist) was applied 10 min after the KCl stimulus and its relaxing effect was evaluated for 10 min. Previously, we observed that the relaxation induced by 5 μM of noradrenaline on KCl contraction was completely blocked when the tissues were preincubated 5 min before with 20 μM propranolol (β2-adrenoceptor antagonist). Following the same protocol, the effect of 17α-E2 at 89.39 μM was observed with 20 μM propranolol preincubation.
Other uterine rings were depolarized with a high potassium-calcium free solution; depolarizing solution (KCl) modified by addition of 2 mM EGTA and without CaCl2. A transient contraction was obtained by high potassium-calcium free solution and when the baseline was reached, 1 mM CaCl2 was added to evoke a tonic contraction, which was recorded for 20 min. This process was repeated until a reproducible response was obtained (control); then, the tissues were preincubated with 17α-E2 at 89.39 μM (IC50 on spontaneous contraction) 5 min before the addition of CaCl2 at 1 mM. Under this conditions, the contraction induced by CaCl2 was recorded for 20 min in the presence of 17α-E2, which was compared with the control. Subsequently, the tissues were washed out and a CaCl2-induced contraction was elicited again. The washout was done after all calcium contractions, three times, with depolarizing free-calcium solution. This protocol was used to test the potential voltage-operated calcium channel blocking properties of 17α-E2.
Uterotrophic and antiuterotrophic activity
Other rats in diestrus were ovariectomized under ether anesthesia. Fifteen days later, the animals were divided into seven groups (n ≥ 6 each), and they were injected subcutaneously, once daily for three days, with 0.4 ml/Kg body weight of vehicle (corn oil; group I), 0.3 μmol/Kg body weight of 17β-E2 or 17α-E2 (group II and III, respectively), and in combination with varying concomitant doses of 17β-E2/17α-E2 (μmol/day/Kg body weight): 1:1 (0.3:0.3; group IV), 1:3 (0.3:0.9; group V), 1:5 (0.3:1.5; group VI) and 1:100 (0.3:30; group VII), all dissolved and administered in 0.4 ml/day/Kg of the vehicle. The rats were weighed 24 h after the last dose, and vaginal smears were taken and examined under the microscope. Autopsy was performed and the uteri were carefully dissected out, blotted and the organ wet weights were recorded.
Data presentation and statistical analysis
The total contractile activity (the area under the curve inscribed by the frequency and amplitude of contraction) was measured during each 10-min interval by using PolyView system 2.1 (Grass Instruments Division/Astro-Med. Inc, West Warwick, RI) data acquisition and playback software. The data for compound action on the uterine contractility were calculated as mean value of more than 6 independent determinations, each from different experiments and expressed as percentages ± SEM. In order to evaluate the inhibitory response of 17α-E2, its potency was compared with the inhibitory effect of induced by 17β-E2. The gain in uterine weight of each group was calculated as mg uterine weight/100 g body weight. 17α-E2 (group III) was compared with the vehicle (group I) and 17β-E2 (group II) effect. The treated groups at different doses range of 17β-/17α-E2 were compared with the uterotrophic effect of 17β-E2 alone (uterine growth induced at 0.3 μmol/day/Kg body weight = 100%). Non-paired Student's t-test was utilized to compare the responses between two groups. We used two-way-ANOVA to compare the concentration-response curves in isolated tissues. For multiple comparisons, one-way-ANOVA with Bonferroni correction was used for antiuterotrophic assay. A value of P < 0.05 was accepted as statistical significance.
Chemicals
The following compounds were used: 1,3,5(10)-estratriene-3,17β-diol (17β-estradiol; 17β-E2), 1,3,5(10)-estratriene-3,17α-diol (17α-estradiol; 17α-E2), tamoxifen (estrogen receptor antagonist), ICI 182,780 (the pure estrogen receptor antagonist), actinomycin D (transcription inhibitor), cycloheximide (protein synthesis inhibitor), propranolol hydrochloride (β2-adrenoceptor antagonist; P) and noradrenaline hydrochloride (noradrenaline; NA). With the exception of ICI 182,780 (obtained from Tocris Cookson, Ellisville, MO, USA), the remaining compounds used in the present study were all purchased from Sigma Chemical Co., St. Louis MO, USA. In isolated tissue preparations, all compounds were prepared as stock solution (for each concentration) in absolute ethanol and added to the bath chamber in a final volume of 0.1% (absolute ethanol), except for NA and P which were dissolved in distilled water. Actinomycin D and NA were kept in the dark until use in order to avoid light-induced degradation. With respect to the in vivo experiments, the estrogens were dissolved and administered in the same volume of corn oil (0.4 ml/Kg).
Results
Inhibitory effect of 17α-E2 on spontaneous contractility
As shown in Fig. 1A, the vehicle of estrogens, ethanol (0.1%; a final volume identical to those added as solvent for estrogens), did not significantly modify spontaneous uterine contractility (2.95 ± 0.25% of inhibition, n = 6, P > 0.05). 17α- and 17β-E2 caused a concentration-dependent inhibition of spontaneous uterine contractility (Fig. 1B), with an IC50 value of 89.39 and 8.42 μM, respectively. Therefore, the effect of 17β-E2 was 10.6 fold more potent than 17α-E2 to inhibit the spontaneous uterine contractility. As shown in Fig. 1B, the concentration-response curves to 17α- and 17β-E2 were significantly different between them (P < 0.0005). The inhibitory effect of 17α-E2 and its 17β isomer was observed within 1 min after the uterine tissue was exposed to the estrogen (Fig. 1A) and the spontaneous contractility was reversed after estrogen was removed (washed out) from the tissue. We also observed that addition of 89. 39 μM 17α-E2 did not stop the previous inhibitory effect induced by 17β-E2 at 8.42 μM. On the contrary, 17β-E2-induced inhibition (30.6 ± 1.08%) was significantly enhanced after addition of 17α-E2 (70.53 ± 1.77% of inhibition; P < 0.0005). Furthermore, the opposite treatment (17α-E2 before 17β-E2 addition) revealed that the inhibitory effect of 17β-E2 was not antagonized by 17α-E2 pretreatment, but was also significantly enhanced (73.97 ± 1.97% of inhibition; n = 6, P < 0.0005). Thus, this observation implies that both inhibitory effects were synergized.
Figure 1 Inhibitory effect of 17α- and 17β-E2 on spontaneous uterine contractile activity of rats in diestrus. A) The vehicle utilized to dissolve each estradiol, ethanol (ETOH 0.1%), did not significantly modify (P > 0.05) the spontaneous contractility. 17α- and 17β-E2 induce inhibition of contractile activity. Note their different efficacy when they are added at the same concentration (20 μM). B) Concentration-response curves to 17α- and 17β-E2 on spontaneous contractility, which were significantly different (P < 0.0005) between them. Each point represents the mean ± SEM of six independent experiments. Statistical significance between concentrations: *P < 0.005, **P < 0.0005. The effect induced by both estradiol isomers at all concentrations tested were significantly different (P < 0.0005) from the vehicle control.
Effect of estrogens on KCl-induced uterine contraction
The tonic contraction induced by KCl was also inhibited by each estradiol at the highest concentration tested on spontaneous contraction (200 μM), the development of the relaxing effect in precontracted tissues also started within a few seconds (~30 sec) after addition of each estrogen (Fig. 2A). Likewise, after washout the amplitude and tone of the next KCl-induced contraction was totally recovered. As shown in Fig. 2B, the relaxing efficacy of 17α- and 17β-E2 was not significantly different in both spontaneous and KCl-induced contraction; however, the relaxing effect induced by 17β-E2 was higher than that induced by 17α-E2 in both contractile responses. The vehicle of estrogens (ethanol 0.1%) did not significantly affect (1.39 ± 0.17% of relaxation; n = 6, P > 0.05) the tone of KCl contraction, but the effect induced by each estradiol was significantly different (P < 0.0005) from the vehicle control (Fig 2A).
Figure 2 Effect induced by 17α- and 17β-E2 on uterine contraction achieved by high potassium (KCl 40 mM). A) The vehicle of estrogens, ethanol (ETOH 0.1%), did not significantly modify KCl-induced contraction (P > 0.05) whereas this tonic contraction was significantly inhibited by 17α- or 17β-E2 and this response was also significantly different from the vehicle control (P < 0.0005). The black circles represent the time of washout. B) Comparison of inhibitory effect at equimolar concentration (200 μM) of 17α- and 17β-E2 on spontaneous and KCl(40 mM)-induced contraction. The student's t test demonstrated that the relaxing efficacy of 17α- and 17β-E2 was not significantly different (P > 0.05) on spontaneous or KCl-induced contraction. The plotted values represent the mean ± SEM of six independent experiments.
Inhibitory effect of 17α-E2 in presence of different drugs
These results are illustrated in Fig. 3. The inhibitory effect elicited by 17α-E2 (89.39 μM) was not blocked by gene transcription (Fig. 3A) or protein synthesis (Fig. 3B) inhibitor. Likewise, we observed that the estrogen receptor antagonists (tamoxifen or ICI 182,780) failed to affect 17α-E2-induced uterine inhibition (Fig. 3C and 3D, respectively). Moreover, the response to 17α-E2 has rapid time-courses, and the spontaneous uterine contractility was totally recovered after estrogen was removed from the tissue (washout). As shown in Fig. 3E, the control vehicle (absolute ethanol; final volume 0.1%) by each drug (inhibitor and estrogen) did not significantly modify the spontaneous contractility (3.47 ± 1.14% of inhibition; P > 0.05). Propranolol at 20 μM blocked the inhibitory effect of 5 μM noradrenaline, but this β2-adrenoceptor antagonist (at the same concentration) did not block the relaxation induced by 17α-E2 on KCl-induced contraction (Fig. 3F).
Figure 3 Original recordings demonstrating typical effect of 17α-E2 on uterine contractions. The inhibitory effect of 17α-E2 on spontaneous contractility was not blocked by preincubation with inhibitors of transcription, 10 μM actinomycin D (A) and protein synthesis, 100 μM cycloheximide (B) or antiestrogens, 1 μM tamoxifen or 1 μM ICI 182,780 (C and D, respectively). The vehicle utilized to add each compound, absolute ethanol (ETOH 0.1%), did not affect (P > 0.05) the spontaneous contractility (E). KCl-induced contraction was inhibited by addition of 5 μM noradrenaline (NA), which was antagonized by preincubation with β2-adrenoceptor antagonist, 20 μM propranolol (P), while 17α-E2-induced relaxation was not antagonized by P (F). The relaxing efficacy of estrogen was not modified in tissues pretreated with these blocking agents. The calcium-induced contraction at 1 mM (Ca2+) in tissues previously depolarized by high potassium-calcium free solution (KCl-Ca2+φ) was notably prevented by 17α-E2 (G). The solid black line indicates the incubation time with 17α-E2 at 89.39 μM. Note the contraction recovery after washout (represented by the black circles), showing that the estrogen effect was reversible.
Effect of 17α-E2 on calcium-induced contraction
The uterine tissues were depolarized by high potassium-calcium free solution. Under these experimental conditions, a tonic contraction was induced by CaCl2 (1 mM), which was antagonized when tissues were preincubated with 89.39 μM 17α-E2 (71.03 ± 1.93% of inhibition, n = 6), observing that the amplitude was significantly decreased (Fig. 3G). The calcium antagonistic effect induced by 17α-E2 was reversible upon washing out the tissue and removing the estrogen (see Fig. 3G; third CaCl2 addition). Importantly, the prevention of this calcium contraction by 17α-E2 turned out significantly higher (P < 0.0005) than its inhibitory effect on spontaneous contraction (52.12 ± 2.05 %). The final volume of estrogen vehicle (0.1% absolute ethanol) did not significantly prevent calcium-induced contraction (3.55 ± 0.45%; n = 6, P > 0.05).
Agonistic and antagonistic activity on uterine growth
17α-E2 action on the gain uterine weight was significantly different (P < 0.05) to vehicle group (corn oil), thus this hormone presented a light uterotrophic activity in contrast with the potent uterotrophic activity induced by its 17β isomer at the same dose (Fig. 4). However, the antiuterotrophic activity of 17α-E2 was assayed and we observed that this hormone had a significant antagonistic effect on the uterotrophic action of 17β-E2 at all doses ratios (Table 1 and Fig. 4). Additionally, the vaginal smears from rats treated with 17β-E2 showed that the vaginal cornification was decreased as 17α-E2 dose was increased (data not shown).
Figure 4 Action of 17α- and 17β-E2 on uterine mass in adult ovariectomized rats. 17α-E2 significantly increased the uterine weight as compared with the vehicle (corn oil; *P < 0.05), although this increment was significantly less than those of 17β-E2 (+P < 0.00005). In rats treated with 17β-E2 plus different doses of 17α-E2, less uterotrophic effect of 17β-E2 was evident. Statistical significance: **P < 0.005, ***P < 0.0005, ****P < 0.0001. The analysis shows that the treatment 1:1 (a) was different from 1:3, 1:5 and 1:100 (b; P < 0.05), which were not different between them (b; P > 0.05). Each bar represents the mean ± SEM, n = 6 rats per group.
Table 1 Antiuterotrophic activity of 17α-E2 in adult ovariectomized rats.
Dosage of 17α-E2 (μmol/day/Kg body weight) Ratioa of 17β-E2:17α-E2 Uterine wet weight/100 gb % inhibitionc
0.3 1:1 266.1 ± 8.2 18.55
0.9 1:3 229.9 ± 10.7 29.63
1.5 1:5 228.3 ± 10.4 30.11
30.0 1:100 213.1 ± 7.8 34.77
aFixed dose of 17β-E2 used = 0.3 μmol/day/Kg body weight.
bData from six independent experiments ± SEM (six rats per group).
cAs compared with uterotrophic effect of 17β-E2.
Discussion
The results of the present study demonstrate that 17α-E2 is a natural estrogen hormonally active in the rat uterine tissue. Our findings show that this estrogen is capable of inducing both nongenomic (antiuterotonic effect) and genomic (estrogenic/antiestrogenic effect) action. This evidence may, in fact, account for its potential regulatory biological function.
In particular, 17α-E2 elicits inhibition of uterine contractile activity by inducing antiuterotonic responses on spontaneous and KCl-induced contraction. These findings are in line with a previous study in isolated rat uterus, which reported that 17α-E2 has the ability to produce relaxation on contractions induced by KCl, calcium and vanadate [18]. We observed that 17α-E2-induced inhibition was significantly different as compared to its 17β isomer, with about 10-fold lower potency than 17β-E2, implying a partial agonist effect of 17α-E2 on uterine contractile activity. Nevertheless, this inhibitory effect could be relevant to promote uterine quiescence during pregnancy. Admittedly, the concentrations of the inhibitory responses to 17α- and 17β-E2 may be in pharmacological ranges; however, these are close to the therapeutic doses used.
Regarding the mode of action of 17α-E2-induced uterine relaxation, it is important to emphasize that the instantaneous relaxing effect of 17α-E2, plus the evidence that the effect disappears after the estrogen is removed from the tissue, are uncharacteristic of classical genomic activities. Thus, this effect is presumably through nongenomic (membrane) actions. This idea is supported by the fact that the observed effect of 17α-E2 on uterine contractility occurred within 1 min of its addition, and 1 min is not enough time for genomic effects to occur [28]. Another approach to discriminate between genomic and nongenomic action is to use antihormones that bind to the intracellular steroid receptor, but that do not block the rapid membrane effects. Our results show that 17α-E2 on rat uterus does not seem to be mediated by genomic events since the antiestrogens, tamoxifen or ICI 182,780, and the transcriptional and translational inhibitors did not antagonize the uterine inhibitory effect of 17α-E2. This observation is in partial agreement with the study of Gutiérrez and coworkers [18], where it was also found that the relaxation induced by 17α-E2 on KCl-induced contraction was not blocked by tamoxifen, but differed by the fact that 17α-E2-induced relaxation was blocked by cycloheximide and actinomycin D, where an interaction with transcriptional ways was suggested. This apparent discrepancy may be explained because their evaluation was done under special experimental conditions; rats were estrogen-primed for 24 h. In this way, the administration of 17β-E2 results in changes of density and distribution of several receptors (proteins), called estrogen-dependent receptors, which are enhanced under conditions of estrogen dominance and, consequently, this treatment may modify the physiological response. In contrast, our data were obtained on spontaneous contractility of rats in diestrus, a model close to the physiological condition. Nevertheless, it should be kept in mind that the relaxing effect of 17α-E2 is too fast and reversible, a point of discussion to distinguish if the transcriptional process is or not involved.
Indeed, our findings indicate that 17α-E2 has a nongenomic action to induce myometrial relaxation, as previously reported to progestins and androgens in rat [29] and humans [30,31]. Likewise, our study has shown different sensitivity of uterine tissue to 17α- and 17β-E2-induced relaxation, suggesting a specific relaxing efficacy for each isomer. In this sense, we have reported before the presence of large differences in uterine relaxing potency in a series of closely related steroids, such as androgens and progestins, some of them without effect [29-32], pointing to a defined structure-activity relationship. Nevertheless, it is important to consider that not only is the uterine muscle the target of 17α-E2 to induce relaxation but also other types of smooth muscles such as those in the blood vessels [15-17] and bladder [19] are relaxed by this estrogen.
In view that the uterine relaxing effect of 17α-E2 can be explained by a nongenomic action, this effect may be associated with different sites of action at the cell surface (e.g. membrane proteins). In this respect, the possibility of an interaction of 17α-E2 with inhibitory β2-adrenoceptors can be dismissed since its specific antagonist (propranolol) did not block the uterine relaxing effect of 17α-E2. This evidence indicates that 17α-E2 induced a nonadrenergic inhibitory response on uterine contractility.
Since the involvement of β2-adrenoceptors seems improbable, the possibility has to be discussed finally that the nongenomic-relaxing effect of 17α-E2 involves a diminution on the intracellular concentration of calcium in the smooth muscle cells. In this connection, our preliminary results show that the efficacy induced by 17α-E2 was more prominent to antagonize contractions induced by calcium in depolarized tissues than to inhibit the spontaneous contractility. In this context, it is known that KCl depolarizes the membrane and opens voltage-operated calcium channels, resulting in calcium entry. Consequently, the inhibition induced by 17α-E2 in uterine rings precontracted by KCl and the marked antagonism of calcium entry in depolarized tissues are implying a direct reduction of extracellular calcium influx by producing inactivation of voltage-operated calcium channels. This hypothesis has also been proposed to the nongenomic relaxing effect of progestins and androgens in uterine [29-32] and vascular [33] smooth muscle. In support of this view, it has been reported that 17α-E2 might behave as a calcium channel antagonist in vascular cells [17,34] and this estrogen may also induce a reduction on voltage-dependent calcium currents in vascular smooth muscle cells [35,36]. Obviously, further patch-clamp experiments on uterine smooth muscle cells will be required to evaluate this issue.
It is also tempting to suggest that the nongenomic relaxing effect of 17α-E2 could be mediated through a subpopulation of the classical estrogen receptor (ER), ERα and ERβ, that is located at the plasma membrane [37,38] and specifically identified in caveolae and cell membranes from endothelial cells [39]. However, this possibility appears not to be supported by the fact that endogenous membrane and nuclear ER was found to be the same protein [40] and a much weaker affinity of 17α-E2 to the human ER compared to 17β-E2 has been shown [41]. In this context, our experimental evidence that two ER antagonists (tamoxifen or ICI 182,780) did not block 17α-E2-induced uterine relaxation implies that the estrogen binding site at the myometrial cell membrane could be a protein unrelated to the ERα or ERβ. Thus, the present findings indicate that 17α-E2 also acts at the cell surface of myometrial cells to initiate rapid, nongenomic, responses and this action may be mediated by interaction of estrogen with calcium channel proteins, which produces inactivation of voltage-operated calcium channels. Taken together these findings, it is important to consider that a nonspecific effect may also occur for the uterine relaxing effect of 17α-E2, such as changes in membrane fluidity.
Although 17α-E2 had been considered without estrogenic activity [3-6], we confirmed that 17α-E2 presents a significant weak estrogenic activity by inducing increase on uterine weight; this activity is much less potent than that induced by 17β-E2 and did not achieve full estrogenicity at higher doses. Consistent with this observation, it has also been reported that 17α-E2 induces light estrogenic activity [7-11]. However, to our knowledge, no evidence had been shown on the action of 17α-E2 as antagonist.
Remarkably, we have now demonstrated that 17α-E2 possesses antiestrogenic properties by antagonizing the uterotrophic response of 17β-E2, and we believe it plays an important role in estrogen-mediated uterine physiology. Furthermore, there is also the possibility that 17α-E2 should induce antiestrogenic activity in several tissues as well as in other species, including human; however, additional experiments are needed in order to address this question. The uterotrophic response of 17β-E2 was significantly antagonized by 17α-E2 at all doses ratios, although its antagonism was not dose-dependent at a dosage of 0.9 to 30 μM/Kg, with a maximal effect of ~30% of inhibition. Indeed, this may indicate that 17α-E2 acts as a partial antagonist of ER. Interestingly, the antiuterotrophic efficacy of 17α-E2 turned out similar to the antiuterotrophic action of tamoxifen, in a range of doses ratios from 1:2 to 1:10, as previously reported [42].
It is of note that the reduced efficacy of 17α-E2 to induce uterine relaxation and estrogenic activity, as compared with 17β-E2, is similar to that induced by tamoxifen, which deserves further consideration. In the first instance, tamoxifen is also capable of eliciting uterine relaxation [43-45]; consequently, this antiestrogen has been reported to have some nonreceptor-mediated effects [46] by unclear mechanisms. Moreover, since the ER is essential for 17β-E2-induced uterine proliferative responses [47], the present data suggest that 17α-E2 appears to regulate its slight agonist (estrogenic) and antagonist (antiestrogenic) activity through the same mechanism as tamoxifen. Recently, it has been documented that tamoxifen displays partial agonist-antagonist activities in different tissues and cells, and these differences may be related to the milieu of ER coactivators and corepressors in these tissues. The ER has two transcriptional activation domains, AF-1 and AF-2; thus, AF-1 activity is stimulated by tamoxifen binding to induce its partial agonist activity but AF-2 is inhibited when tamoxifen acts as antagonist [48-51]. With this line of evidence, could be reasonable to speculate that the mode of action of tamoxifen could be analogous for the agonist and antagonist properties of 17α-E2. Obviously, further studies that fall beyond the scope of the present investigation will be required to explore the agonistic-antagonistic mechanism of 17α-E2.
The present study revealed that both estradiol isomers, 17α- and 17β-E2, possess the same nongenomic action by inducing uterine relaxation but 17α-E2 is incapable of antagonizing the 17β-E2-induced uterine relaxation. Collaterally, these two isomers may also induce the same genomic action by eliciting an estrogenic uterotrophic response; however, the estrogenic activity of 17β-E2 is antagonized by 17α-E2. In this context, this evidence is correlated with the biological action emerged to other natural isomers, such as testosterone and epitestosterone which present the same nongenomic relaxing action [23], but only epitestosterone possesses a genomic antiandrogenic activity [24-26].
Finally, we also noted that the different structural conformation of both estradiol isomers could be important to induce diverse responses. The differences between both molecules is the α/trans or β/cis configuration at C-17. Thus, estrogen antagonistic activity of 17α-E2 could be presumed by its α/trans configuration, in contrast to the β/cis which is responsible for the marked antiuterotonic and estrogenic activity induced by 17β-E2.
Conclusion
The preliminary investigation presented here reveals interesting features of the uterine function regulated by 17α-E2. The data indicate that this estrogen is an agonist inducing nongenomically mediated uterine relaxation, but also appears to have mixed agonist-antagonist activity on uterine growth, presumably through genomic processes. This study shows that a rapid nongenomic action (antiuterotonic response) of 17α-E2 takes place before its genomic action (uterotrophic-antiuterotrophic response). Additionally, this evidence could account for our knowledge of effects produced by 17α-E2 and some sulfate derivatives, which are components applied in the treatment of peri- and post-menopausal women.
Authors' contributions
MP conceived and designed the study and wrote the article, and was also involved in acquisition, analysis and interpretation of data. EN carried out the in vitro studies, performed the statistical analysis and helped to draft the manuscript. The two authors read and approved the final manuscript.
Acknowledgements
This work was supported by a grant from PAPIIT/DGAPA (project #IN221102-2).
==== Refs
Girasole G Jilka RL Passeri G Boswell S Boder G Williams DC Manolagas SC 17 beta-estradiol inhibits interleukin-6 production by bone marrow-derived stromal cells and osteoblasts in vitro: a potential mechanism for the antiosteoporotic effect of estrogens J Clin Invest 1992 89 883 891 1541679
Vargas R Tomas G Wroblewska B Ramwell PW Differential effects of 17α and 17β-estradiol on PGF2α mediated contraction of the porcine coronary artery Adv Prostaglandin Thromboxane Leukot Res 1989 19 277 280 2526484
Pietras RJ Szego CM Estrogen receptors in uterine plasma membrane J Steroid Biochem 1979 11 1471 1483 513766 10.1016/0022-4731(79)90124-9
Merriam GR MacLusky NJ Johnson LA Naftolin F 2-Hydroxyestradiol-17α and 4-hydroxyestradiol-17α, catechol estrogen analogs with reduced estrogen receptor affinity Steroids 1980 36 13 20 7414655 10.1016/0039-128X(80)90063-X
Schwartz Z Soskolne WA Neubauer T Goldstein M Adi S Ornoy A Direct and sex-specific enhancement of bone formation and calcification by sex-steroids in fetal mice long bone in vitro (biochemical and morphometric study) Endocrinology 1991 129 1167 1174 1874163
Nasatzky E Schwartz Z Boyan BD Soskolne WA Ornoy A Sex dependent effects of 17-beta-estradiol on chondrocyte differentiation in culture J Cell Physiol 1993 154 359 367 8425917 10.1002/jcp.1041540220
Lundeen SG Carver JM McKean ML Winneker RC Characterization of ovariectomized rat model for the evaluation of estrogen effects on plasma cholesterol levels Endocrinology 1997 138 1552 1558 9075715 10.1210/en.138.4.1552
Hajek RA Robertson AD Johnston DA Van NT Tcholakian RK Wagner LA Conti CJ Meistrich ML Contreras N Edwards CL Jones LA During development, 17alpha-estradiol is a potent estrogen and carcinogen Environ Health Perspect 1997 3 577 581 9167998
Yamasaki K Takeyoshi M Yakabe Y Sawaki M Imatanaka N Takatsuki M Comparison of reporter gene assay and immature rat uterotrophic assay of twenty-three chemicals Toxicology 2002 170 21 30 11750080 10.1016/S0300-483X(01)00505-4
Yamasaki K Takeyoshi M Sawaki M Imatanaka N Shinoda K Takatsuki M Immature rat uterotrophic assay of 18 chemicals and Hershberger assay of 30 chemicals Toxicology 2003 183 93 115 12504345 10.1016/S0300-483X(02)00445-6
Fujimoto N Honda H Kitamura S Effects of environmental estrogenic chemicals on AP1 mediated transcription with estrogen receptors α and β J Steroid Biochem Mol Biol 2004 88 53 59 15026083 10.1016/j.jsbmb.2003.10.006
Green PS Bishop J Simpkins JW 17α-estradiol exerts neuroprotective effects on SK-N-SH cells J Neorosci 1997 17 511 515
Dicko A Morissette M Ben Ameur S Pezolet M Di Paolo T Effect of estradiol and tamoxifen on brain membranes: investigation by infrared and fluorescence spectroscopy Brain Res Bull 1999 49 401 405 10483916 10.1016/S0361-9230(99)00066-0
Morin C Zini R Simon N Tillement JP Dehydroepiandrosterone and α-estradiol limit the functional alterations of rat brain mitochondria submitted to different experimental stress Neuroscience 2002 115 415 424 12421607 10.1016/S0306-4522(02)00416-5
Salas E López MG Villarroya M Sánchez-García P De Pascua R Dixon WR García AG Endothelium-independent relaxation by 17-α-estradiol of pig coronary arteries Eur J Pharmacol 1994 258 47 55 7925599 10.1016/0014-2999(94)90056-6
Rodríguez J García de Boto MJ Hidalgo A Mechanisms involved in the relaxant effect of estrogens on rat aorta strips Life Sci 1996 58 607 615 8632714 10.1016/0024-3205(95)02330-5
Seeger H Mueck AO Ottel M Schwarz S Lippert TH Calcium antagonistic effect of 17 alpha-estradiol derivatives: in vitro examinations Gynecol Endocrinol 1999 13 246 248 10533159
Gutiérrez M Fernández AI Revuelta MP Cantabrana B Hidalgo A Partial contribution to the relaxant effect of 17α-estradiol in rat uterine smooth muscle Gen Pharmacol 1998 30 71 77 9457484
Ratz PH McCammon KA Altstatt D Blackmore PF Shenfeld OZ Schlossberg SM Differential effects of sex hormones and phytoestrogens on peack and steady state contractions in isolated rabbit detrusor J Urol 1999 162 1821 1828 10524942 10.1097/00005392-199911000-00077
Sievernich A Wildt L Lichtenberg-Frate H In vitro bioactivity of 17α-estradiol J Steroid Biochem Mol Biol 2004 92 455 463 15698550 10.1016/j.jsbmb.2004.09.004
Hobe G Schon R Goncharov N Katsiya G Koryakin M Gesson-Cholat I Oettel M Zimmermann H Some new aspects of 17α-estradiol metabolism in man Steroids 2002 67 883 893 12234624 10.1016/S0039-128X(02)00058-2
Bhavnani BR Estrogens and menopause: pharmacology of conjugated equine estrogens and their potential role in the prevention of neurodegenerative diseases such as Alzheimer's J Steroid Biochem Mol Biol 2003 85 473 482 12943738 10.1016/S0960-0760(03)00220-6
Perusquía M Hernández R Kubli-Garfias C Epitestosterone induces testosterone-like uterine relaxation [abstract] 10th International Congress of Endocrinology 1996 549 578
Nuck BA Lucky AW Epitestosterone: a potential new antiandrogen J Invest Dermatol 1987 89 209 211 3598212 10.1111/1523-1747.ep12470564
Starká I Bicíková M Hampl R Epitestosterone -an endogenous antiandrogen? J Steroid Biochem 1989 33 1019 1021 2532272 10.1016/0022-4731(89)90255-0
Bicíková M Hampl R Starká L Epitestosterone a potent competitive inhibitor of C21-steroid side chain cleavage in testis J Steroid Biochem Mol Biol 1992 43 721 724 1472464 10.1016/0960-0760(92)90298-W
Lictchfield JT Wilcoxon FA A simplified method of evaluating dose-effect experiments J Pharmacol Exp Ther 1949 96 99 108
Shang Y Hu X DiRenzo J Lazar MA Brown M Cofactor dynamics and sufficiency in estrogen receptor-regulated transcription Cell 2000 103 843 852 11136970 10.1016/S0092-8674(00)00188-4
Perusquía M Nongenomic action of steroids in myometrial contractility Endocrine 2001 15 63 72 11572328 10.1385/ENDO:15:1:063
Perusquía M Jasso-Kamel J Influence of 5α- and 5β-reduced progestins on the contractility of isolated human myometrium at term Life Sci 2001 68 2933 2944 11411793 10.1016/S0024-3205(01)01089-X
Perusquía M Navarrete E Jasso-Kamel J Montaño LM Androgens induce relaxation of contractile activity in pregnant human myometrium at term: a nongenomic action on L-type calcium channels Biol Reprod 2005 10.1095/biolreprod.104.036954
Perusquía M García-Yañez E Ibáñez R Kubli-Garfias C Non-genomic mechanism of action of Δ4 and 5-reduced androgens and progestins on the contractility of isolated rat myometrium Life Sci 1990 47 1547 1553 2174488 10.1016/0024-3205(90)90183-R
Perusquía M Villalón CM Posible role of Ca2+ channels in the vasodilating effect of 5β-dihydrotestosterone in rat aorta Eur J Pharmacol 1999 371 169 178 10357254 10.1016/S0014-2999(99)00161-2
Collins P Rosano GM Jiang C Lindsay D Sarrel PM Poole-Wilson PA Cardiovascular protection by oestrogen -a calcium antagonist effect? Lancet 1993 341 1264 1265 8098404 10.1016/0140-6736(93)91158-I
Jiang C Poole-Wilson PA Sarrel PM Mochizuki S Collins P MacLeod KT Effect of 17β-oestradiol on contraction, Ca2+ current and intracellular free Ca2+ in guinea-pig isolated cardiac myocytes Br J Pharmacol 1992 106 739 745 1504758
Zhang F Ram JL Standley PR Sower JR 17β-estradiol attenuates voltage-dependent Ca2+ currents in A7r5 vascular smooth muscle cell line Am J Physiol 1994 266 C975 C980 8178970
Pappas TC Gametchu B Watson CS Membrane estrogen receptors identified by multiple antibody labeling and impeded-ligand binding FASEB J 1995 9 404 410 7896011
Razandi M Pedram A Greene GL Levin ER Cell membrane and nuclear estrogen receptors (ERs) originate from a single transcript: studies of ERα and ERβ expressed in Chinese hamster ovary cells Mol Endocrinol 1999 13 307 319 9973260 10.1210/me.13.2.307
Chambliss KL Yuhanna IS Anderson RG Mendelsohn ME Shaul PW ERβ has nongenomic action in caveolae Mol Endocrinol 2002 16 938 946 11981029 10.1210/me.16.5.938
Razandi M Pedram A Merchenthaler I Greene GL Levin ER Plasma membrane estrogen receptors exist and function as dimmers Mol Endocrinol 2004 18 2854 2865 15231873 10.1210/me.2004-0115
Fang H Tong W Shi LM Blair R Perkins R Branham W Hass BS Xie Q Dial SL Moland CL Sheehan DM Structure-activity relationships for a large diverse set of natural, synthetic, and environmental estrogens Chem Res Toxicol 2001 14 280 294 11258977 10.1021/tx000208y
Agarwal AK Bindal RD Estrogen receptor-binding affinity of tamoxifen analogs with various side chains and their biologic profile in immature rat uterus Steroids 1991 56 486 489 1805462 10.1016/0039-128X(91)90007-I
Lipton A Vinijsanum A Martin L Acute inhibition of rat myometrial responses to oxytocin by tamoxifen stereoisomers and oestradiol J Endocrinol 1984 103 383 388 6502064
Fernández AI Cantabrana B Hidalgo A Estrogen and antiestrogen non-genomic effect in rat uterus contraction in calcium-free solution Gen Pharmacol 1993 24 391 395 8482524
Kostrzewska A Laudanski T Batra S Potent inhibition by tamoxifen of spontaneous and agonist-induced contractions of the human myometrium and intramyometrial arteries Am J Obstet Gynecol 1997 176 381 386 9065186
Altan N Chen Y Schindler M Simon SM Tamoxifen inhibits acidification in cells independent of estrogen receptor Proc Natl Acad Sci 1999 96 4432 4437 10200279 10.1073/pnas.96.8.4432
Korach KS Insights from the study of animals lacking functional estrogen receptor Science 1994 266 1524 1527 7985022
Smith CL Nawaz Z O'Malley BW Coactivator and corepressor regulation of the agonist/antagonist activity of the mixed antiestrogen, 4-hydroxytamoxifen Mol Endocrinol 1997 11 657 666 9171229 10.1210/me.11.6.657
McDonnell DP The molecular pharmacology of SERMs Trends Endocrinol Metab 1999 10 301 311 10481160 10.1016/S1043-2760(99)00177-0
Levin ER Integration of the extra-nuclear and nuclear actions of estrogen Mol Endocrinol 2005 10.12107me.2004-0390
Zhang H McElrath T Tong W Pollard JW The molecular basis of tamoxifen induction of mouse uterine epithelial cell proliferation J Endocrinol 2005 184 129 140 15642790 10.1677/joe.1.05987
|
16042770
|
PMC1201169
|
CC BY
|
2021-01-04 16:37:11
|
no
|
Reprod Biol Endocrinol. 2005 Jul 21; 3:30
|
utf-8
|
Reprod Biol Endocrinol
| 2,005 |
10.1186/1477-7827-3-30
|
oa_comm
|
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-321609296310.1186/1477-7827-3-32ResearchSpam1-associated transmission ratio distortion in mice: Elucidating the mechanism Martin-DeLeon Patricia A [email protected] Hong [email protected] Carlos R [email protected] Yutong [email protected] Michelle [email protected] Barry L [email protected] Hong [email protected] Deni S [email protected] Department of Biological Sciences, University of Delaware, Newark, DE 19716, USA2 Department of Anatomy and Cell Biology, McGill University, Montreal, Quebec, Canada3 Department of Genetics, Thomas Jefferson University, Philadelphia, PA, USA2005 10 8 2005 3 32 32 2 5 2005 10 8 2005 Copyright © 2005 Martin-DeLeon et al; licensee BioMed Central Ltd.2005Martin-DeLeon 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
While transmission ratio distortion, TRD, (a deviation from Mendelian ratio) is extensive in humans and well-documented in mice, the underlying mechanisms are unknown. Our earlier studies on carriers of spontaneous mutations of mouse Sperm Adhesion Molecule 1 (Spam1) suggested that TRD results from biochemically different sperm, due to a lack of transcript sharing through the intercellular cytoplasmic bridges of spermatids. These bridges usually allow transcript sharing among genetically different spermatids which develop into biochemically and functionally equivalent sperm.
Objectives
The goals of the study were to provide support for the lack of sharing (LOS) hypothesis, using transgene and null carriers of Spam1, and to determine the mechanism of Spam1-associated TRD.
Methods
Carriers of Spam1-Hyal5 BAC transgenes were mated with wild-type female mice and the progeny analyzed for TRD by PCR genotyping. Sperm from transgene and Spam1 null carriers were analyzed using flow cytometry and immunocytochemistry to detect quantities of Spam1 and/or Hyal5. Transgene-bearing sperm with Spam1 overexpression were detected by fluorescence in situ hybridization. In wild-type animals, EM studies of in situ transcript hybridization of testis sections and Northern analysis of biochemically fractionated testicular RNA were performed to localize Spam1 transcript. Finally, AU-rich motifs identified in the 3' UTR of Spam1 RNA were assayed by UV cross-linking to determine their ability to interact with testicular RNA binding proteins.
Results
The Tg8 line of transgene carriers had a significant (P < 0.001) TRD, due to reduced fertilizing ability of transgene-bearing sperm. These sperm retained large cytoplasmic droplets engorged with overexpressed Spam1 or Hyal5 protein. Caudal sperm from transgene carriers and caput sperm of null carriers showed a bimodal distribution of Spam1, indicating that the sperm in a male were biochemically different with respect to Spam1 quantities. Spam1 RNA was absent from the bridges, associated exclusively with the ER, and was shown to be anchored to the cytoskeleton. This compartmentalization of the transcript, mediated by cytoskeletal binding, occurs via protein interactions with 3' UTR AU-rich sequences that are likely involved in its stabilization.
Conclusion
We provide strong support for the LOS hypothesis, and have elucidated the mechanism of Spam1-associated TRD.
==== Body
Introduction
A remarkable feature of mammalian testicular maturation of sperm is the syncytial organization that results from the presence of intercellular cytoplasmic bridges among the germ cells. These bridges allow transcript sharing among genetically different spermatids and provide a mechanism by which these cells develop synchronously into biochemically and functionally equivalent sperm [1]. Studies of spermatid-expressed genes for Protamine [2] and several X-linked sperm-specific proteins [3] provide strong evidence for transcript sharing. However sharing may not be a global phenomenon for all spermatid-expressed genes, particularly those encoding membrane proteins [4]. Moreover there is compelling evidence for functionally different sperm in a male leading to TRD, as best exemplified by mice carrying different alleles at the t-complex [5,6]. The TRD seen for the t-haplotypes has been explained by unequal sharing of post-meiotic products [1], but there is no evidence for this mechanism.
Earlier our laboratory provided evidence for a Lack-of-Sharing hypothesis (LOS) for TRDs that were discovered in the progeny of Robertsonian (Rb) translocation-bearing mice [7-9], and shown to be associated with carriers of spontaneous mutations of the murine Sperm adhesion molecule1 (Spam1) gene [10,11]. SPAM1 encodes a widely conserved sperm membrane protein [12] which has multiple essential roles in mammalian fertilization [13]. The murine gene which maps to proximal chromosome 6 [14] in a cluster of hyaluronidase genes containing Hyalp1, Hyal4, and Hyal5 [15], is spermatid-expressed and the RNA is transcriptionally regulated since it first appears together with the protein in the testis of postnatal Day 21.5 mice [10]. The TRDs were seen in heterozygotes of either of two Rb translocations, Rb(6.15) and Rb(6.16), in which multiple Spam1 point mutations were shown to be present [11], leading to reduced expression of both the RNA and the protein [10]. We have since observed that in these mice Hyalp1 and Hyal5 which have overlapping functions with Spam1 also have point mutations that would have contributed to the TRDs [16]. Furthermore, the fact that Spam1 null mice are fertile suggests that other hyaluronidases are able to compensate for this gene [17].
Our LOS hypothesis for the Spam1-related TRDs is based on our finding of compartmentalization of the RNA, as assessed by RNA-FISH [10]. This compartmentalization precludes transcript sharing in normal as well as mutant mice, and leads to biochemically different sperm with respect to the protein. Importantly, the protein is inserted in the acrosomal membrane soon after formation [10]. Our study showed that in males carrying different alleles of Spam1, compartmentalization leads to biochemically and functionally different sperm and the resulting TRD [10]. The objectives of this study were to use transgene and null carriers of Spam1 to garner support for the LOS hypothesis and to study the transcript localization in normal mice to gain insights into the underlying mechanism leading to the TRD.
Materials and methods
Breeding and Transmission Study
The studies were approved by the Animal Care Committee at the University of Delaware and conform to the guide for the Care and Use of Laboratory animals published by the National Institutes of Health (publication 85-23, revised 1985). A 150 kb mouse BAC (bacterial artificial chromosome) clone, described earlier [14] and after sequencing shown to contain Spam1 and Hyal5 and their regulatory regions, was used to generate transgenic mice, as reported elsewhere [Zhang et al., submitted]. Male trangenic founders, Tg8, Tg9, and Tg11, and their F1 progeny were mated to C57BL females, and transgenic and non-transgenic offspring identified by PCR genotyping at weaning. Tail DNA samples from progeny were screened using BAC vector-specific primers
(F-5'AACATACGAGCCGGAAGCAT 3' and R-5'GATTCAGGTTCATCATGCCG 3') for PCR. Additionally, four females which were mated with Tg8 carriers or wild-type C57BL males (the background for the transgenic mice) were examined for resorption sites on the 14th day of pregnancy.
Hemizygotes or carriers for Spam1 null were generated by mating C57BL males with Spam1 null females. Null mice were obtained from the laboratory of Tadashi Baba where they were generated [17]. Backcross matings were then set up by pairing sexually mature hemizygous null males with Spam1 null females.
Flow Cytometry and Immunoflourescence
Flow Cytometric Analysis
Flow cytometry was performed to quantify the amount of Spam1 on the sperm surface. Caput and/or caudal sperm from adult wild-type C57BL mice and carriers of Tg8, Tg11, and the null allele were collected in PBS and fixed in 1.5% paraformaldehyde for 1 hr at RT. After washing and blocking in 2% BSA in PBS they were stained using the rabbit antipeptide mouse Spam1 antiserum generated from a C-terminal 15-mer (#381 – 395) oligopeptide (custom made by Zymed, San Francisco, CA) (diluted 1:400) specific for Spam1 [18,19]. The secondary antibody was FITC-conjugated goat anti-rabbit IgG (diluted 1:320). After several washes of the cells, fluorescence was measured (up to 30,000 sperm for each sample) using a FACScan (Becton-Dickinson, San Jose, CA) flow cytometer with a Lysis II software package. The FACScan instrument uses an argon laser at 488 nm with detectors for FITC. Prior to preparing for cytometric analysis, an aliquot of the sperm suspension was used for indirect immunofluorescence of Spam1.
Indirect Immunofluorescence
Caput and caudal sperm recovered from adult wild-type C57BL mice, and Tg8 and Tg11 hemizygotes and homozygotes were fixed and processed as above. They were then stained with rabbit antipeptide mouse Spam1 or Hyal5 antiserum generated from a 20-mer at C-terminal #474–492 (custom made by Zymed, San Francisco, CA) (diluted 1:400). Peptide blocking of the antiserum showed that the signal was specific. The secondary antibody was FITC-conjugated goat anti-rabbit IgG (diluted 1:320). Controls were incubated with preimmune rabbit/FITC-conjugated secondary antibody. The sperm were then mounted on slides in ρ-phenylenediamine antifade with 1.5 μg/ml of 4' 6-diamidino-2-phenylindole (DAPI) for standard fluorescence microscopy. The specimens were examined using a Zeiss Axiophot (Carl Zeiss, Oberkochen, Germany) with the appropriate FITC filter set, and imaged using a CCD-cooled camera and IPLab software. A total of 200 sperm were examined from each group to identify abnormal expression of Spam1. The analysis of Hyal5 was qualitative only.
Flow sorting of Sperm and Fluorescence In Situ Hybridization
Sorting
After flow cytometric analysis, sperm were sorted in preparation for fluorescence in situ hybridization. A sorting gate was set on the histogram to collect sperm from the half of the bimodal distribution with the higher fluorescence intensity. Sperm were sorted on the FACSCalibur in exclusion mode at an approximate sort rate of 200 events/second and collected into 50 ml tubes in PBS. They were then pelleted and fixed with methanol-acetic acid.
Fluorescence In Situ Hybridization
BAC DNA was labeled by nick-translation using SpectrumRed™ direct-labeled dUTP (Vysis, Downers Grove, IL). Methanol-acetic acid fixed sperm were treated with 10 mM dithiothreitol in 0.1 M Tris-HCl for 45 min on ice to decondense the chromatin. They were then dipped in ddH2O before air-drying and hybridization was as described [14].
In Situ Transcript Hybridization
CD-1 mice (n = 3 per group) were anesthetized with sodium pentobarbital, and the testes were perfused through the left ventricle with 4% paraformaldehyde, 0.1 % glutaraldehyde, and 3% dextran sulfate in 0.05 M phosphate buffer (pH 7.4) for 15 min. Following perfusion, the testes were removed and immersed in the same fixative for 5 hr at 4°C. The tissues were then cut into small blocks of approximately 8 mm3, embedded in 2.5% melted agarose (at 60°C), and cut into 60 μm-thick frontal sections with a vibrotome. Groups of 10 sections were collected in autoclaved vials and washed three times in RNAse-free 0.05 M phosphate buffer (pH 7.4) at room temperature. Glycine (1 M) was added to the buffer to neutralize aldehyde groups.
Prehybridization and hybridization procedures were performed as described previously [20]. Briefly, testicular sections were transferred from the phosphate buffer to the prehybridization buffer containing 4 × SSC (1 × SSC is 0.15 M NaCl plus 0.015 M sodium citrate) and 1 × Denhardt's solution for 1 hr at room temperature with gentle agitation. Sections were then immersed in hybridization buffer containing 1 ml of 8 × SSC, 1 ml of deionized formamide, 100 μl of Sarkosyl (2.3 mg/ml), 200 μl of 1.2 M phosphate, and 1.50 μg per vial of 3H-labeled Spam-1 antisense probe (specific activity 1.47 × 107 cpm/μg) or 1.50 μg per vial of a 3H-labeled control sense probe (specific activity 1.57 × 107 cpm/μg). The antisense probe was generated from a unique PCR fragment from the 3' UTR of Spam1 and thus would not cross-hybridize with other hyaluronidases. After hybridization overnight at 40°C, the sections were rinsed sequentially at the same temperature in 4 × SSC and 0.1 × SSC for 1.5 hr. Following the washes, the sections were quickly dehydrated in 50%, 70%, 90%, and 100% ethanol and embedded in Epon.
Radioautography
Ultrathin sections (65 nm thick) were cut from selected areas of seminiferous tubules within the epon blocks for electron microscope radioautography. The sections were placed on celloidin coated glass slides, coated and dipped in Ilford L4 emulsion according to the method of Kopriwa [21]. After 3 months exposure, the sections were developed in a solution physical development, which produce round silver grains [22]. The sections were then transferred to electron microscopy nickel grids, immersed for 45 sec in glacial acetic acid to remove the celloidin and carbon films.
Quantitative Analysis
For selected steps of spermiogenesis 10 EM micrographs, corresponding to 10 different cells per testis/animal, were selected for analysis according to the method of Nadler [23]. In most of the cases 50 silver grains were scored over each step spermatid cytoplasm. Since 85% of the silver grains were associated to the ER, a circle with a radius of 20 mm (equivalent to 0.23 μm resolution at 60,000x) was centered over silver grains that did not overlay any organelle. When an organelle was found within the circle, the radioautographic silver grain was attributed to such an organelle and considered as "exclusive". If the circle included more than one organelle the silver grains was considered "not exclusive". According to Haddad et al. [24] and Nadler [23] this procedure permits identification of the source of radioactivity, with a probability of 95%. For quantitative analysis steps spermatids were grouped as follow: steps 1–5 and 6–8 (round spermatids), steps 9–11 (early elongated spermatids), steps 12–16 (late elongated spermatids).
Biochemical fractionation of Spam1 RNA
Testicular RNA was extracted from sexually mature CD-1 males and free cytosolic-, cytoskeleton-bound, and membrane-bound fractions were separated by subcellular fractionation techniques as described [25]. Northern blotting was then performed with the fractions and hybridization carried out sequentially using Spam1 and β-actin 32P-labeled probes.
RNA Probe labeling and in vitro label transfer Assay by UV cross-linking
A 77 bp 3' UTR fragment (nts 1909–1985) of Spam1 cDNA containing AU-rich elements (AREs) was obtained by PCR and cloned into pSTBlue-1 vector according to the manufacturer's instructions (Novagen, Madison, WI). Sense RNA probe was generated by T7 RNA polymerase transcription of the HindIII-linearized plasmid in the presence of digoxigenin-11-UTP (DIG) using an in vitro transcription system [Riboprobe® System (Promega, Madison, WI)] in accordance with the manufacturer's protocol.
AU-rich sequence binding protein (AUBP) assays were performed using UV cross-linking (UVXL) label transfer. Testes of wild-type C57BL 4–5 month-old mice were used for cytoplasmic protein extraction as described [26]. To test for specificity of binding, unlabeled antisense DNA oligos (100-fold molar excess) were used in competition assays. The oligos were mixed with the labeled probe at 70°C for 10 min and renatured for 1 hr at 22°C before adding the protein extract. The label transfer was performed as described [25,26] with slight modifications. Briefly, 40 μg of cytoplasmic protein extract was incubated with 5 ng of digoxigenin-labeled RNA for 30 min at 22°C in a reaction volume of 20 μl with cytoplasmic extraction buffer. Subsequently, RNase T1 (0.3 U) was added to the mixture for 10 min at 22°C, followed by heparin (final 5 μg/μl) for 10 min at 22°C. The mixture was transferred to a microplate and UV-cross-linked in a GS Gene Linker™ UV Chamber (3 × 105 μJ, 254-nm bulbs) (BIO-RAD, Hercules, CA) by placing it 1.0 cm from the source for 15 min on ice. The mixture was then incubated with RNase A (final concentration 100 μg/ml) at 37°C for 15 min. Sodium dodecyl sulfate (SDS) sample loading dye was added, samples were boiled for 3 min, subjected to 12.5% SDS-PAGE and then transferred to a nitrocellulose membrane according to standard protocols. Proteins on blots were visualized using the WesternBreeze Chemiluminescent Immunodetection kit (Invitrogen, Carlsbad, CA), following the manufacturer's protocol.
Results
We analyzed the progeny of several Spam1-Hyal5 BAC transgene carriers (Tg9, Tg11, and Tg8 with transgene-copy numbers of 2, 8, and 10; respectively) for the rate of transmission of the transgenes. While the transgenes were transmitted in Mendelian proportions for Tg9 and Tg11 males (P > 0.05), Tg8 demonstrated a highly significant (P < 0.001) TRD arising from a deficiency of transgene-bearing progeny analyzed at postnatal Day 21 (Fig. 1a). The most severe TRD, 2.8:1 was seen for the progeny of the founder and three transgene-bearing F1 males (Tg8A), while males from subsequent generations showed a ratio of 1.8:1 (Tg8-B). The combined population of 339 progeny (Tg8-T) had an overall ratio of 2:1 (P < 0.001) (Fig. 1a), and reveals that the TRD is heritable.
Figure 1 Transmission frequencies of Spam1-Hyal5 BAC transgenes in hemizygotes of three lines, Tg8, Tg9, and Tg11, reveal a distortion only for Tg8/+. a) Histograms showing the rate of production of transgenic and non-transgenic progeny analyzed at weaning on Day 21. Tg8-A represents the progeny of the founder and 3 F1 hemizygous males, while progeny from F2-F4 hemizygous males are seen in Tg8-B. Tg8-T represents the total progeny analyzed for the Tg8 line. There is a significant deviation from 1:1 for Tg8-A (χ2 = 27.30; P < 0.001), Tg8-B (χ2 = 17.47; P < 0.001), and Tg8-T (χ2 = 41.8; P < 0.001) all represented by an asterisk; while there was no difference in 1:1 ratios (P > 0.05) for Tg9 and Tg11, represented by the diamond. b) The TRD for Tg8/+ mice does not result from post-zygotic selection against transgene-bearing zygotes, as revealed by the average litter sizes of transgenic lines. In addition to the BAC transgenic lines we included a Spam1 cDNA transgenic line, Cinn (179), which like Tg9 and Tg11 also had a 1:1 ratio in the progeny. The means for the litters ranged from 6.68 to 8.50 and are shown at the top of the histograms with their SDs at the sides. The highest mean, 8.50 ± 1.51, was seen for Tg8A which had the highest TRD, 2.8:1.
There was no evidence that post-zygotic selection could explain the TRD in Tg8 mice, as the average litter size in Tg8A progeny which had the most severe TRD was the highest, 8.50, among the transgenic lines (Fig. 1b). Further, examination of 14-day fetuses retrieved from matings of four Tg8 hemizygous males with wild-type females showed 0/40 resorptions compared to 3/36 from congenic wild-type males. This indicated that progeny of Tg8 carriers had no greater tendency for post-implantation loss than those for wild-type males. Thus the progeny of Tg8 hemizygotes have a TRD that is likely not due to in utero selection, but rather to meiotic drive.
To determine if the sperm population in the Tg8 hemizygotes was heterogeneous with respect to Spam1 expression, caudal sperm from mature males were subjected to flow cytometric analysis. Unimodal distributions of sperm, with similar peaks, were seen for the congenic C57BL/6J wild-type and for Tg11 transgene carriers (Fig. 2a) which showed no TRD. On the other hand, Tg8 carriers showed a bimodal distribution with a shift to the right, indicating that there was a subpopulation of sperm with Spam1 overexpression (Fig. 2a). To corroborate the flow cytometric finding of two phenotypic classes of sperm in Tg8 carriers we performed immunocytochemistry on aliquots of caudal (mature) sperm analyzed using flow cytometry in Fig. 2a. Surprisingly, a highly significant (P < 0.01) proportion of the caudal sperm, (16.5%, Fig. 3a) showed retention of enlarged cytoplasmic droplets (CDs) which were immunopositive for large deposits of Spam1 (Fig. 3b). Sperm with Spam1-containing CDs were lacking the protein on the heads, the normal location, and had far more Spam1 than was found on the heads of normal sperm (Fig. 3b). Thus sperm with the retention of the CDs are consistent with transgenic overexpression of Spam1, as represented by the subpopulation in Fig. 2a with a shift to the right.
Figure 2 Flow cytometric analyses of sperm from hemizygotes for an overexpressed or a Spam1 null allele show bimodal distributions. a) Caudal sperm from the congenic wild-type, Tg11/+ and Tg8/+ show a bimodal distribution only for Tg8/+. The second peak (on the right) with the greater intensity in this bimodal distribution indicates the presence of a subpopulation of sperm with overexpression. The distributions for the wild-type and Tg11/+ are unimodal with lower mean intensities, indicating a lack of Spam1 overexpression. b) Caudal sperm from Tg8 carriers showing the analysis and gating of the sperm for sorting. c) Caput sperm from hemizygous null mice show a bimodal distribution in sperm in A and B. The first peak (on the left) in each shows background levels of fluorescence, likely representing sperm with the null allele.
Figure 3 Immunocytochemistry demonstrates that Tg8/+ males have significantly increased numbers of sperm retaining enlarged cytoplasmic droplets (CDs) with overexpressed Spam1 or Hyal5, and a concomitant absence of the proteins on the heads. a) Histograms showing the proportion of sperm, from populations of 200, with CDs in wild-type and Tg8/+ males. The greater than 10-fold increase in caudal sperm compared to wild-type is highly significant (χ2 = 10.8; P < 0.01), as is the >30-fold increase in caput sperm (χ2 = 28.5; P < 0.001). b) Retention of CDs with overexpressed Spam1 (green staining) in sperm taken from an aliquot used for flow cytometry in Fig. 2a. A sperm with the normal amount and normal location of Spam1 is shown for comparison with the overexpressed protein in the CD. Note that there is non-specific background staining on the tails. c) Sperm with overexpressed Hyal5 (green staining) in enlarged CDs near the neck and the absence of the protein on the heads are seen in B and C, while A shows a normal CD without Hyal5. FISH signals on flow sorted sperm showing double signals in transgenic cells d) and a single signal in wild-type cells e).
We thus studied mice that were homozygous for the Tg8 transgene and observed that CDs were present on 25–30% of caudal sperm, suggesting that in hemizygotes they are associated with the transgene-bearing sperm. To confirm the significant increase of phenotypically abnormal caudal sperm with CDs in Tg8 carriers, we examined immature sperm from the caput of these mice. As might be expected, there were higher numbers of CDs containing large deposits of Spam1 (Fig. 3a) and the difference between the Tg8 and wild-type mice was highly significant (P < 0.001) (Fig. 3a). It should be noted that CDs in wild-type sperm were not only rare, but they were not enlarged (Fig. 3c-A).
To unequivocally demonstrate that the Tg8/+ sperm with large amounts of Spam1 were a result of transgenic overexpression, sperm were sorted after flow cytometric analysis and the subpopulation with the more intense fluorescence (Fig. 2B) recovered for analysis by fluorescence in situ hybridization (FISH). The Spam1-Hyal5 BAC used to generate the transgenic lines was used as the FISH probe, and 137/217 or 63.1% of the sperm had double hybridization signals representing the transgene and the endogenous locus (Fig. 3d) while 80/217 or 36.9% had a single signal (Fig. 3e). Since double signal due to chromosome 6 disomy has a spontaneous frequency of only 0.9% (our unpublished data), these proportions of double and single signal sperm are highly significantly different from 1:1 (χ2 = 14.96, P < 0.001). Thus they reveal that high Spam1 expression is a result of enrichment of transgene-bearing sperm or transgenic overexpression. The fact that a third of the sperm analyzed had a single signal can be explained by the overlapping peaks in the bimodal distribution (Fig. 2b), resulting in collection of some of the subpopulation without Spam1 overexpression.
We also analyzed the CDs in Tg8 hemizygotes for the presence of the closely related Hyal5 protein encoded by Hyal5 which is present on the transgene. In Fig. 3c-B and 3c-C we show large amounts of Hyal5 in CDs at the neck of sperm and its absence on the head, similarly to what was seen for Spam1. These CDs are distinctly different from normal CDs which were devoid of the protein (Fig. 3c-A), and were not seen in sperm from normal males. The overexpression of Hyal5 and its presence in retained CDs are consistent with the findings for Spam1 in Tg8 mice.
To test the LOS hypothesis using the Spam1 null carriers, caput sperm from sexually mature carriers and wild-type males were analyzed by flow cytometry for Spam1 quantities. There was a bimodal distribution including a subpopulation of sperm with only the baseline fluorescence, as can be seen for two animals in Fig. 2C, which was not seen for caput sperm in wild-type animals (data not shown). Note that in Fig. 2C the position of the two peaks in A and B, are consistent with the presence of sperm carrying the null allele (background or baseline fluorescence) and those with the normal allele.
We addressed the underlying mechanism for the TRD by performing ultrastructural studies to examine the precise localization of Spam1 transcript. Using in situ transcript hybridization with a 3H-labeled Spam1 antisense probe and electron microscopy to reveal the precise subcellular location of the RNA, we show that silver grains were compartmentalized. They were predominantly associated to the ER (Table 1). Conversely, the silver grains were not associated to structures such as the nucleus, chromatoid bodies or radial bodies (Table 1). Silver grains were sometimes located near the vicinity of intercellular bridges. However, they were associated to the ER, suggesting that the transcripts were not in transit but anchored to/near this organelle (Fig. 4a-D).
Table 1 Distribution of radioautographic silver grains in spermatids expressed as percentages (±SD)
Steps 3–5 Steps 6–8 Steps 9–11
ER 85 ± 5 85 ± 2 87 ± 1
Assigned to ER 13 ± 3 14 ± 6 13 ± 1
Nucleus 0 ± 0 0 ± 0 0 ± 0
Chromatoid Ba 0 ± 0 0 ± 0 0 ± 0
Radial Bodies 0 ± 0 0 ± 0 0 ± 0
Cytoplasmic Bb 0 ± 0 0 ± 0 0 ± 0
Unknown origin 2 ± 0 1 ± 0 0 ± 0
a Chromatoid bodies, b Cytoplasmic bridges. Some of the data in this table are taken from Molecular Reproduction & Development © copyright 2004 Wiley-Liss, Inc, A Wiley Company.
Figure 4 Spam1 transcripts which are compartmentalized are absent from the bridges and are associated with the cytoskeleton. a) EM autoradiography of seminiferous tubules after in situ hybridization with a tritiated (3H-labeled) Spam-1 antisense RNA probe. Note that silver grains are associated with the ER (arrowheads) but not with any other major spermatid structures such as the chromatoid bodies (A), the radial bodies (B) or the microtubules of the manchette (C). D is an intercellular bridge where the curvatures at the top and bottom represent the outer limits of the bridge. While some grains are seen in association with the ER in the vicinity, they are absent from the bridge. The circles centered over the silver grains include profiles of ER (arrowheads). (E) Late spermatids (S) and Sertoli cells (Se) are unreactive. (F) Shows the cytoplasm of round spermatid (RS) step 8 of a control section hybridized to a sense probe. Co, chromatoid body; Rb, radial body; M, manchette. X19,000 b. Northern hybridization of Spam1 and β-actin mRNAs in free cytosolic-, cytoskeletal-, and membrane-bound testicular RNA fractions. The fractions were separated by subcellular fractionation techniques. A) shows Northern blotting, while B) shows total RNA as a loading control with ethidium bromide staining. The presence of cytoskeletal-bound β-actin RNA in the free cytosolic fraction suggests that Spam1 in the latter could be present as a contaminant due to the preparation procedure.
To further probe the subcellular location of the transcript and determine the nature of the anchoring, we performed biochemical fractionation of testicular RNA to identify cytoskeletal-associated, membrane-associated, and cytosolic-related fractions, as previously described [25]. These fractions were probed in Northern analysis with Spam1 cDNA, using β-actin (which is known to be cytoskeletal-bound) as an internal control. Fig. 4b shows that Spam1 mRNA is not membrane-associated, but was found sequestered with the cytoskeleton. The similar pattern for β-actin and Spam1, allows us to conclude that Spam1 is cytoskeletal-bound. Transcripts in the cytosol are either newly formed ones that have just exited the nucleus and have not yet been bound to the cytoskeleton, or those that are a contaminant of the preparation process.
RNA-cytoskeletal binding has been shown to occur via AU-rich motifs in the 3' UTR [28]. Thus we searched Spam1 RNA sequence and identified four AREs in a 77 nucleotide (nt) sequence in the 3' UTR (nt 1909–1985), as shown in Fig. 5A. To investigate whether the sequence was a target for RNA-binding proteins that mediate cytoskeletal binding we generated a full-length 77 nt riboprobe for use in in vitro label transfer by UV-cross-linking. Testicular cytoplasmic proteins that bind specifically to the riboprobe were identified and are seen in Lanes T1, C3 and T3 in Fig. 5B. Addition of ~100-fold molar excess of unlabeled competitor antisense DNA oligomers for all four AREs (C1) virtually abolished RNA-protein complex formation as seen in Lane C1 (Fig. 5B), indicating the binding specificity of the ARE(s).
Figure 5 Spam1 RNA contains four AREs in the 3' UTR and UV-cross-linking reveals that one or two specifically bind testicular proteins. A) Four AU-rich motifs, 5' AUUUG...AUUUUA...AUUUUUA...AUUUUUG...3', are found in a 77 nucleotide sequence (nt 1909–1985) in Spam1 3' UTR. B) In the upper panel the thin line represents the 77 nt riboprobe sequence containing the AU motifs while the thick lines show the antisense DNA oligomers used in competition assays. C1 contains antisense oligos for all four AREs, C2 for the two at the 5' end of the sequence, C3 for the two at the 3' end, C5for the third from the 5' end, and C6 for the 3' end motif. The lower panel shows the results of in vitro label transfer by UV cross-linking and the RNA-protein complexes formed with the riboprobe and testicular protein extract, without competition, in Lanes T1 and T3. Lanes C1 and C2 show the disappearance of the complexes after competition with antisense oligos for all four motifs and the two at the 5' end, respectively. Lane C3 shows that the complexes are not diminished with competition with antisense oligos for the two 3' end AREs.
Similarly, pre-incubation of the riboprobe with unlabeled antisense DNA oligos for the two AREs in the 5' end of the probe (C2,) (prior to mixing with the protein extract) also abolished the binding as seen in Lane C2 (Fig. 5B), confirming the binding specificity. However, the addition of ~100-fold excess unlabeled antisense DNA oligos for the two AREs in the 3' end of the RNA (C3) did not diminish the formation of the RNA-protein complexes as seen in Lane C3 (Fig. 5B), indicating that these ARE's do not participate in the binding. When antisense oligos for these two 3' AREs were individually used in competition (C5 and C6), the formation of the RNA-protein complexes were also not diminished (data not shown). These results of in vitro binding are consistent with binding activity of Spam1 RNA, via one or both of the 5' AREs, to proteins that mediate cytoskeleton binding.
Discussion
The results show a highly significant (P < 0.001) heritable TRD in Tg8 carriers, in favor of normal sperm and against transgene-bearing ones with 10 copies of Spam1/Hyal5. [The Mendelian 1:1 transmission ratio that was detected for carriers of the other transgenic lines can be attributed to the difficulty in obtaining overexpression of these genes due to naturally occurring antisense RNA [Zhang et al., submitted]. Importantly the caudal sperm population of Tg8 carriers showed a bimodal distribution, reflecting the presence of sperm with different quantities of Spam1. FISH analysis showed that the high Spam1-expressing sperm were enriched for the transgene, indicating transgenic overexpression of Spam1. That transgene-bearing sperm produced significantly less progeny than normal sperm shows that optimal levels of Spam1 and Hyal5 are required for fertility, similar to β1,4-galactosyl transferase where overexpression is associated with acrosome instability [29].
The failure of transgene-bearing sperm to effect fertilization in the expected ratio is a direct result of the retention of cytoplasmic droplets (CDs). A CD is an organelle with residual cytoplasm on the neck or tail of sperm. It results from a defect in the final stages of spermiogenesis and its presence in mature sperm renders them infertile [30,31]. It is interesting that sperm with overexpressed Spam1 and Hyal5 in CDs were lacking the proteins on the surface of the heads where they are normally found. Thus in addition to the sperm motility defects associated with CDs [32] there would be a decreased ability of penetration of the cumulus cells, leading to their infertility. It should be noted that in domestic animals CD-associated infertility has been shown to be due to poor passage through hyaluronate swim-up medium and failure to bind to the zona pellucida [33], both of which are functions of Spam1 [13]. However this is the first report of the presence of hyaluronidases in CDs, although a number of other enzymes have been identified in them [30].
The fact that the number of sperm with the enlarged CD phenotype is less than 50% and 100% in Tg8 hemizygotes and homozygotes, respectively, may be due to technical factors such as their loss during preparation [30]. It is also likely that the highly regulated Spam1 silencing which is mediated by antisense transcription [Zhang et al., submitted] could be responsible. This silencing would be adaptive since overexpression of Spam1 and Hyal5 leads to their mis-expression in CDs, which are abnormally retained and which lead to infertility. Thus in Tg8 carriers there are functionally different sperm within a male, due to their different quantities of Spam1 and Hyal5. It is not known if Spam1 and Hyal5 are overexpressed in the same CDs, consequently their co-localization will be investigated in future studies.
The finding of structurally and functionally different sperm in Tg8 transgene carriers is reminiscent of the findings for heterozygotes of spontaneous mutant alleles of Spam1 and is also consistent with compartmentalization of the RNA and protein [10]. The accumulation of the overexpressed Spam1 and Hyal5 protein in the CDs of trangene-bearing sperm is a result of a lack of transcript sharing between these sperm and those with the normal alleles. These observations on the transgenic model therefore provide support for the LOS Hypothesis in the etiology of TRDs.
Similarly, the LOS hypothesis is also supported by the findings from carriers of Spam1 null allele. Generated by insertion of a neo cassette in exon II (which contains the hyaluronidase domain) of Spam1 in the laboratory Tadashi Baba [17], null mice were shown to be fertile (despite a delay in cumulus penetration) due to the compensating effect of the redundant Hyal5 [17]. Carriers of Spam1 null showed a bimodal distribution of Spam1 in caput sperm, with one subpopulation having background levels of fluorescence, consistent with the presence of sperm with a null allele. Since Spam1 is expressed in the epididymis where it may be acquired by sperm during transit [18,19] caput, but not cauda, sperm would reflect the spermatid phenotype, and the finding from these sperm is supportive of a lack of transcript sharing.
To determine the mechanism for the lack of transcript sharing we focused on RNA compartmentalization. It has been proposed that mRNA localization facilitates protein sorting and that nascent polypeptide chain targeting of membrane proteins is a major mechanism that accounts for mRNA localization [34]. We show that 85% of the grains from in situ transcript hybridization localized to the ER and the remaining 15% could be assigned to this region. This indicates that the transcripts are not dispersed in the cytoplasm where they would gain ready access to the bridges. Table 1 shows that they are not associated with any major spermatid structure, such as the chromatoid bodies (which indicates that the RNA is not stored and is not translationally regulated), the radial bodies, or the microtubules of the manchette (Fig. 4a). Importantly, they are absent from the bridges although they may be in the vicinity. In this connection it should be pointed out that the absence of the transcript from the chromatoid bodies which have recently been seen to cross the bridges (Parvinen and Sassone-Corsi, personal communication), bolsters the evidence for the RNA compartmentalization.
The compartmentalization of Spam1 transcripts and their absence from the bridges suggest that they are anchored, and this would preclude sharing and support our LOS hypothesis. Based on the restricted location of the transcript at the ER, it was expected that Northern analysis would reveal its association with the membrane-bound fraction. However, this was not the case as Spam1 RNA was shown to be associated with the cytoskeletal fraction and therefore anchored. More importantly, we show that the transcript has AU-rich elements (AREs) in the 3' UTR that are known to bind cytoplasmic proteins (AUBPs) that mediate binding to the cytoskeleton. Interestingly, AREs are also present in the 3' UTR of the rodent-specific Hyal5 (Genbank Accession# ABO85680) which (at the nucleotide level) is 71% homologous to Spam1 with which it shares all functional domains [16]. Thus Hyal5 transcripts are likely to be cytoskeletal-bound and compartmentalized.
While cytoskeletal-binding may assist with anchoring and maintaining a pool of Spam1 and Hyal5 transcripts which can be recruited to the ER for co-translational assembly which occurs for membrane proteins [34], the data also suggest its involvement in posttranscriptional regulation. AREs are well-known to mediate RNA (in)stability by interacting with trans-acting proteins [35-37]. Further, their protein interaction which mediates cytoskeletal binding is known to be involved in mRNA turnover and posttranscriptional regulation of RNA [28,37]. We have recently observed that testicular Spam1 RNA may be stabilized by interactions with RNA-binding proteins [Zhang et al., submitted]. Therefore the presence of AREs in Spam1 RNA and the demonstration of their ability to specifically bind testicular proteins, potentially relates TRD with posttranscriptional regulation of the RNA. Simply put, the cytoarchitecture that facilitates the co-translational assembly of the transcript in the ER is involved in regulating mRNA decay and ultimately precludes RNA sharing via the bridges Taken together, our findings provide strong support for the LOS hypothesis. They also mechanistically relate RNA compartmentalization, mediated by cytoskeletal binding, and the regulation of the mRNA turnover to TRD.
TRD has been seen for a) the transmission of disease alleles such as delta F508 of the CFTR gene for cystic fibrosis [38] and b) the inheritance of the most common Robertsonian translocation [39], and is a phenomenon for which there is extensive evidence in the human genome [40]. Based on the diversity of genes involved, it is likely that there may be many different underlying mechanisms. However the findings in this study have uncovered, to our knowledge, the first molecular mechanism for a mammalian TRD.
Acknowledgements
We are grateful to Dr. Tadashi Baba for providing the Spam1 null mice. The work was supported by NIH grant RO1 HD38273 and NSF 9974808 to P.A.M-D.
==== Refs
Willison K Ashworth A Mammalian spermatogenic gene expression Trends Genet 1987 3 351 355 10.1016/0168-9525(87)90296-4
Caldwell KA Handel MA Protamine transcript sharing among postmeiotic spermatids Proc Natl Acad Sci USA 1991 88 2407 2411 2006178
Moss SB VanScoy H Gerton GL Mapping of a haploid transcribed and translated sperm-specific gene to the mouse X chromosome Mamm Genome 1997 8 37 38 9021145 10.1007/s003359900342
Hendriksen P Do X and Y spermatozoa differ in proteins? Theriogenology 1999 52 295 1307 10.1016/S0093-691X(99)00218-6
Olds-Clarke P Genetic analysis of sperm function in fertilization Gamete Res 1988 20 241 264 3069685 10.1002/mrd.1120200214
Silver LM Mouse t haplotypes Ann Rev Genet 1985 19 179 208 3909941
Aranha IP Martin-DeLeon PA The murine Rb(6.16) translocation: evidence for sperm selection and a modulating effect of aging Hum Genet 1991 87 278 284 1864601 10.1007/BF00200904
Chayko CA Martin-DeLeon PA The murine Rb(6.16) translocation: alterations in the proportion of alternate sperm segregants effecting fertilization in vitro and in vivo Hum Genet 1992 90 79 85 1427792 10.1007/BF00210748
Aranha IP Martin-DeLeon PA Mouse chromosome 6 in Rb translocations: consequences in singly and doubly heterozygous males Cytogenet Cell Genet 1995 69 253 259 7698024
Zheng Y Deng X Martin-DeLeon PA Lack of sharing of Spam1 (Ph-20) among mouse spermatids and transmission ratio distortion Biol Reprod 2001 64 1730 1738 11369602
Zheng Y Deng X Zhao Y Zhang H Martin-DeLeon PA Spam1 (PH-20) mutations and sperm dysfunction in mice with the Rb(6.16) or Rb(6.15) translocation Mamm Genome 2001 12 822 829 11845284 10.1007/s00335-001-1008-3
Lathrop WF Carmichael EP Myles DG Primakof P cDNA cloning reveals the molecular structure of a sperm surface protein, PH-20, involved in sperm-egg adhesion and the wide distribution of its gene among mammals J Cell Biol 1990 111 2939 2949 2269661 10.1083/jcb.111.6.2939
Cherr GN Yudin AI Overstreet JW The dual functions of GPI-anchored PH-20: hyaluronidase and intracellular signaling Matrix Biol 2001 20 515 525 11731269 10.1016/S0945-053X(01)00171-8
Deng X Moran J Copeland NG Gilbert DJ Jenkins NA Primakoff P Martin-DeLeon PA The mouse Spam1 maps to proximal chromosome 6 and is a candidate for the sperm dysfunction in Rb(6.16)24Lub and Rb(6.15)1Ald heterozygotes Mamm Genome 1997 8 94 97 9060406 10.1007/s003359900365
Csoka AB Frost GI Stern R The six hyaluronidase-like genes in the human and mouse genomes Matrix Biol 2001 20 499 508 11731267 10.1016/S0945-053X(01)00172-X
Zhang H Shertok S Miller K Martin-DeLeon PA Sperm Dysfunction in the Rb(6.16)- and Rb(6.15)-bearing Mice Revisited: Involvement of Hyalp1 and Hyal5 Mol Reprod Dev 2005 epublished Aug 4, DOI 10. 1002/mrd.20360
Baba D Kashiwabara S Honda A Yamagata K Wu Q Ikawa M Okabe M Baba T Mouse sperm lacking cell surface hyaluronidase PH-20 can pass through the layer of cumulus cells and fertilize the egg J Biol Chem 2002 277 30310 30314 12065596 10.1074/jbc.M204596200
Zhang H Martin-DeLeon PA Mouse epididymal Spam1 (pH-20) is released in the luminal fluid with its lipid anchor J Androl 2003 24 51 58 12514083
Zhang H Morales CR Badran H El-Alfy M Martin-DeLeon PA Spam1 (PH-20) Expression in the Extratesticular Duct and Accessory Organs of the mouse: A Possible Role in Sperm Fluid Reabsorption Biol Reprod 2004 71 1101 1107 15175239 10.1095/biolreprod.104.030403
Morales CR Badran H El-Alfy M Men H Zhang H Martin-DeLeon PA Cytoplasmic localization during testicular biogenesis of the murine mRNA for Spam1 (PH-20), a protein involved in acrosomal exocytosis Mol Reprod Dev 2004 69 475 482 15457544 10.1002/mrd.20177
Kopriwa B A reliable standardized method for ultrastructural electron microscope radioautography Histochemistry 1973 37 1 17 4129434 10.1007/BF00306855
Kopriwa B A comparison of various procedures for fine grain development in electron radioautographiy Histochemistry 1975 44 201 224 1184416 10.1007/BF00491492
Nadler NJ The interpretation of the grain counts in electron microscope radioautography J Cell Biol 1971 49 877 882 5092210
Haddad A Smith MD Herscovics A Nadler NJ Leblond CP Radioautographic study of in vivo and in vitro incorporation of fucose-3H into thyroglobulin by rat thyroid follicular cells J Cell Biol 1971 49 856 882 4326459 10.1083/jcb.49.3.856
Hesketh JE Campbell GP Whitelaw PF c-myc mRNA in cytoskeletal-bound polysomes in fibroblasts Biochem J 1991 274 607 609 2006923
Thomson AM Rogers JT Walker CE Staton JM Leedman PJ Optimized RNA gel-shift and UV cross-linking assays for characterization of cytoplasmic RNA-protein interactions Biotechniques 1999 25 1032 1042 10572651
Balmer LA Beveridge DJ Jazayeri JA Thoson AM Walker CE Leedman PJ Identification of a novel AU-Rich element in the 3' Untranslated region of epidermal growth factor receptor mRNA that is the target for regulated RNA-binding proteins Mol Cellular Biol 2001 21 2070 2084 11238942 10.1128/MCB.21.6.2070-2084.2001
Henics T Nagy E Szekeres-Bartho J Interaction of AU-rich sequence binding proteins with actin: possible involvement of the actin cytoskeleton in lymphokine mRNA turnover J Cell Physiol 1997 173 19 27 9326445 10.1002/(SICI)1097-4652(199710)173:1<19::AID-JCP3>3.0.CO;2-M
Shur BD Hall NG A role for mouse sperm surface galactosyltransferase in sperm binding to the egg zona pellucida J Cell Biol 1982 95 574 579 6815212 10.1083/jcb.95.2.574
Oko R Hermo L Chan PT Fazel A Bergeron JJ The cytoplasmic droplet of rat epididymal spermatozoa contains saccular elements with Golgi characteristics J Cell Biol 1993 123 809 821 8227142 10.1083/jcb.123.4.809
Cooper TG Cytoplasmic droplets: the good, the bad or just confusing? Hum Reprod 2005 20 9 11 15471925 10.1093/humrep/deh555
Cooper TG Yeung C-H Fetic S Sobhani A Nieschlag E Cytoplasmic droplets are normal structures of human sperm but are not well preserved by routine procedures for assessing sperm morphology Hum Reprod 2004 19 2283 2288 15242996 10.1093/humrep/deh410
Thundathil J Palasz AT Barth AD Mapletoft RJ The use of in vitro fertilization techniques to investigate the fertilizing ability of bovine sperm with proximal cytoplasmic droplets Anim Reprod Sci 2001 65 181 192 11267798 10.1016/S0378-4320(00)00231-1
Lawrence JB Singer RH Intracellular localization of messenger RNAs for cytoskeletal proteins Cell 1986 45 407 415 3698103 10.1016/0092-8674(86)90326-0
Tourriere H Chebil K Tazi J mRNA degradation machines in eukaryotic cells Biochimie 2002 84 821 837 12457569 10.1016/S0300-9084(02)01445-1
Chen C-YA Xu N Shyu A-B mRNA decay mediated by two distinct AU-rich elements from c-fos and granulocyte-macrophage colony-stimulating factor transcripts: different deadenylation kinetics and uncoupling from translation Mol Cell Biol 1995 15 5777 5788 7565731
Bevilacqua Bevilacqua A Ceriani M Capaccioli S Nicolin A Post-Transcriptional Regulation of Gene Expression by Degradation of Messenger RNAs J Cell Physiol 2003 195 356 372 12704645 10.1002/jcp.10272
Kitzis A Chomel JC Haliassos A Tesson L Kaplan JC Feingold J Giraud G Lable A Dastugue B Dumur V Unusual segregation of cystic fibrosis alleles Nature 1988 366 316 3194017 10.1038/336316a0
Pardo Manuel de Villena F Sapienza C Transmission ratio distortion in offspring of heterozygous female carriers of Robertsonian translocations Hum Genet 2001 108 31 36 11214904 10.1007/s004390000437
Zollner S Wen X Hanchard NA Herbert MA Ober C Pritchard JK Evidence for extensive transmission distortion in the human genome Am J Hum Gen 2004 74 62 72 10.1086/381131
|
16092963
|
PMC1201170
|
CC BY
|
2021-01-04 16:37:12
|
no
|
Reprod Biol Endocrinol. 2005 Aug 10; 3:32
|
utf-8
|
Reprod Biol Endocrinol
| 2,005 |
10.1186/1477-7827-3-32
|
oa_comm
|
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-361610217210.1186/1477-7827-3-36DebateCan ovarian infertility be treated with bone marrow- or ovary-derived germ cells? Bukovsky Antonin [email protected] Laboratory of Development, Differentiation and Cancer, Department of Obstetrics and Gynecology, the University of Tennessee Graduate School of Medicine, Tennessee, USA2005 15 8 2005 3 36 36 29 7 2005 15 8 2005 Copyright © 2005 Bukovsky; licensee BioMed Central Ltd.2005Bukovsky; 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.
A year ago, reproductive biologists and general public were astonished with evidence reported by Johnson et al. in Nature 428:145 that mammalian ovaries possess persisting large germline stem cells, which allegedly enable follicular renewal in adult females. Recently, the same research group declared such view obscure, and reported that mammalian oocytes originate from putative germ cells in bone marrow and are distributed by peripheral blood to the ovaries (Cell 122:303). While neglecting available data on the germ cell origin from the ovarian surface epithelium (OSE) in adult mouse and human females and complexity of follicular renewal in humans, the authors widely extrapolated their observations on formation of allogeneic oocytes after bone marrow (or blood) transplantation in ovaries of adult mice treated with cytostatics to clinical implications in the public media. Yet, the resulting outcome that such allogeneic oocytes may enable the propagation of ovarian cycles is a poor alleviation for the women with ovarian infertility. Women lacking primary follicles, or carrying follicles with low quality eggs persisting in aging ovaries, are not concerned about the lack of menstrual cycles or ovarian steroids, but about virtually no chance of having genetically related children. Johnson et al. also reported that the germ cell formation in bone marrow disappears in ovariectomized mice. Such observation, however, raises solid doubts on the bone marrow origin of oocytes. Since germ cells developing from the OSE cells of adult human ovaries during periodical follicular renewal are known to enter blood vessels in order to enable formation of primary follicles at distant ovarian sites, they also contaminate peripheral blood and hence bone marrow. Better knowledge on the complexity of follicular renewal in humans and exploration of a potential of human OSE cells to produce new oocytes in vitro are essential for novel approaches to the autologous treatment of premature ovarian failure and age induced ovarian infertility.
==== Body
Comment
The claim that female germ cells in adult mammals (mice and humans) originate from bone marrow and are delivered to the ovaries via the blood stream [1] is another controversial postulate of Johnson et al. Last year this research group, which is lead by Dr. Jonathan L. Tilly of the Massachusetts General Hospital/Harvard Medical School, claimed that large germline stem cells persist in postnatal mammalian ovaries [2]. They allegedly documented that these cells divide and differentiate into new oocytes required for the follicular renewal and concluded: "Therefore, in addition to providing new directions to explore with respect to elucidation the biology of mammalian female germline stem cells, this work has significant clinical implications related to therapeutic expansion of the follicle reserve as a means to postpone normal or premature ovarian failure" [2].
After facing immediate critique, which indicated that such large cells rather resemble earlier described superfluous oocytes leaving mouse and human ovaries [3,4], the Tilly's group now declared that the permanent persistence of female germline stem cells contributing to the follicular renewal in adult mammalian ovaries is obscure [1]. When searching for another original (now extragonadal) source of germ cells, the Tilly's group stated a new surprise for the scientists and general public: Female germ cells cyclically develop in bone marrow of the adult mammalian females [1]. Yet, the experienced reproductive biologists stay calm. Why? There is no need to oppose this new "discovery" for another year or so, since the authors killed their claim instantly, by themselves, and in the same article.
To explore an idea on the extragonadal origin of germ cells in adult mammalian females, one will compare observations in animals with and without ovaries. Johnson et al. did and reported: In the absence of ovaries the evidence of germ cell formation in bone marrow completely disappears. However, instead of realizing that this observation is a perfect evidence on ovarian origin of germ cells, which enter the ovarian blood stream and, therefore, some of them bone marrow too, the authors surprised the reader with an additional uncovering: "The results from the ovariectomy experiments lend further support to the existence of a novel communication loop between the ovaries and bone marrow that may regulate the extent of de novo oocyte production each cycle" [1].
If Johnson et al. are more respectful for the available literature, they will consider that the germ cells derived from ovarian surface epithelium (OSE) utilize blood vessels for colonization of distant targets during follicular renewal [3], and hence contaminate peripheral blood and bone marrow. In other words, instead of the periodical flow of putative bone marrow-derived germ cells to the ovaries, the ovarian germ cells may periodically enter bone marrow via the blood stream.
Curiously, although Johnson et al. are apparently aware of relevant literature, the studies on the germ cell and oocyte origin from OSE and follicular renewal in adult mouse [5,6] and human ovaries [3,7] are unprofessionally ignored. They were neither referenced in the introduction nor discussed as a possible alternative on the origin of germ cells and oocytes in articles of Johnson et al. on persisting germline stem cells [2] or on bone marrow origin of germ cells in adult mammalian ovaries [1]. Although article of Allen from 1923 entitled "Ovogenesis during sexual maturity" [5] was recently mentioned by Johnson et al. [1], the idea on the oocyte origin from OSE in adult mammals has not been acknowledged.
Two highly distinguished American pioneers of modern reproductive physiology in 1920s and 1930s, Edgar Allen and Herbert M. Evans, concluded their studies of oogenesis and follicular renewal in adult mouse ovaries as follows: "A cyclical proliferation of the germinal epithelium (OSE) gives rise to a new addition of young ova to the cortex of the adult ovary at each normal oestrous period" [5] and "New oocytes are formed throughout life, and in phase with the reproductive cycle, from germinal epithelium of the adult mammal, at the same time as vast numbers of already-formed oocytes become eliminated through atresia" [6]. Why these articles are not appropriately referenced by Johnson et al.? The last year "discovery" on follicular renewal in postnatal mouse ovaries, which compensates concomitant atresia [2], has been, in fact, described more than 70 years ago.
What Johnson et al. have recently shown is that the mouse ovaries lacking primary follicles after treatment with cytostatics exhibit follicles with allogeneic oocytes following bone marrow or peripheral blood transplantation from distinct mouse donors. The concomitant immunosuppression may enable allogeneic oocytes to develop, either from the donor-derived circulating ovarian germ cells, or due to the transplantation of immune-system related cells (blood monocytes and lymphocytes), which accompany as ovarian macrophages and T lymphocytes the development of germ cells from some OSE cells in adult [7] and fetal human ovaries [8]. Yet, the resulting outcome that such follicles with allogeneic oocytes "play a critical role in the propagation of each ovarian cycle" [1] represents a poor promise for the women with ovarian infertility. Women lacking primary follicles with own oocytes, or carrying follicles with low quality eggs persisting in aging ovaries, are not concerned about the lack of menstrual cycles or ovarian steroids, but about virtually no chance of conceiving and having genetically related children.
Recently, we suggested animal experiments comparing the effect of autologous vs. allogeneic white (nucleated) blood cell concentrate (buffy coat) transfusions for induction of follicular renewal after chemotherapy in order to determine possible advantage of the former against the latter procedure (submitted May, 2005). Yet, compared to the small laboratory rodents with available clusters of primitive granulosa cells resembling human fetal ovaries [8], a re-colonization of adult human ovaries with new primary follicles requires their readiness by the means of the presence of nests of primitive granulosa cells. Such nests are required for follicular renewal, since superfluous oocytes are not preserved and degenerate [3]. In other words, even transplantation (transfusion) of autologous germ cells may not be sufficient for follicular renewal in aging women, which lack nests of primitive granulosa cells in their ovaries (unpublished observations). On the other hand, one may imagine a collection of autologous blood sample during ovarian oogenesis in younger females prior to anti-cancer chemotherapy or autologous white blood cells, and subsequent (after chemotherapy) rejuvenation of younger ovaries, which may contain nests of primitive granulosa cells ready to form primary follicles.
However, the age limitless in vitro production of new autologous eggs from the OSE cells of human ovaries [8-10], and their in vitro fertilization and utilization of embryos for intrauterine implantation, may represent more suitable variant for providing genetically related children to women with ovarian infertility, worth of consideration and further exploration.
==== Refs
Johnson J Bagley J Skaznik-Wikiel M Adams GB Niikura Y Tschudy KS Tilly JC Cortes ML Forket R Iacomini J Scadden DT Tilly JL Oocyte generation in adult mammalian ovaries by putative germ cells in bone marrow and peripheral blood Cell 2005 122 303 315 16051153 10.1016/j.cell.2005.06.031
Johnson J Canning J Kaneko T Pru JK Tilly JL Germline stem cells and follicular renewal in the postnatal mammalian ovary Nature 2004 428 145 150 15014492 10.1038/nature02316
Bukovsky A Caudle MR Svetlikova M Upadhyaya NB Origin of germ cells and formation of new primary follicles in adult human ovaries Reprod Biol Endocrinol 2004 2 20 15115550 10.1186/1477-7827-2-20
Gosden RG Germline stem cells in the postnatal ovary: is the ovary more like a testis? Hum Reprod Update 2004 10 193 195 15140866 10.1093/humupd/dmh023
Allen E Ovogenesis during sexual maturity Am J Anat 1923 31 439 481 10.1002/aja.1000310502
Evans HM Swezy O Ovogenesis and the normal follicular cycle in adult mammalia Mem Univ Calif 1931 9 119 224
Bukovsky A Keenan JA Caudle MR Wimalasena J Upadhyaya NB Van Meter SE Immunohistochemical studies of the adult human ovary: possible contribution of immune and epithelial factors to folliculogenesis Am J Reprod Immunol 1995 33 323 340 7546251
Bukovsky A Caudle MR Svetlikova M Wimalasena J Ayala ME Dominguez R Oogenesis in adult mammals, including humans: a review Endocrine 2005 26 301 316 16034186 10.1385/ENDO:26:3:301
Bukovsky A Svetlikova M Caudle MR Oogenesis in cultures derived from adult human ovaries Reprod Biol Endocrinol 2005 3 17 15871747 10.1186/1477-7827-3-17
Bukovsky A Origin of germ cells and follicular renewal in adult human ovaries Microscopy & Microanalysis Conference Honolulu, Hawaii 2005 2005 – July 31 – August 4, Invited presentation
|
16102172
|
PMC1201171
|
CC BY
|
2021-01-04 16:37:12
|
no
|
Reprod Biol Endocrinol. 2005 Aug 15; 3:36
|
utf-8
|
Reprod Biol Endocrinol
| 2,005 |
10.1186/1477-7827-3-36
|
oa_comm
|
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-391612238410.1186/1477-7827-3-39ResearchNon-infected preterm parturition is related to increased concentrations of IL-6, IL-8 and MCP-1 in human cervix Törnblom Susanne Abelin [email protected] Aurelija [email protected]öm Birgitta [email protected] Milan [email protected] Annelie [email protected] Gunvor [email protected] Dept of Women and Child Health, Division for Obstetrics and Gynecology, Karolinska Institute, Karolinska University Hospital Solna, 171 76 Stockholm, Sweden2 Microbiology and Tumor Biology Center, Division for Clinical Microbiology, Karolinska Institute, Karolinska University Hospital Solna, 171 76 Stockholm, Sweden2005 25 8 2005 3 39 39 20 6 2005 25 8 2005 Copyright © 2005 Törnblom et al; licensee BioMed Central Ltd.2005Törnblom 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
Human cervical ripening is an inflammatory process. In labour at term the mRNA-levels and protein concentrations for interleukin-6 (IL-6) and IL-8 in cervix significantly increase. The aim of this study was to investigate if there are differences in the inflammatory process of preterm and term cervical ripening.
Methods
Cervical biopsies from 50 singleton pregnant women without clinical signs of infection were allocated to four groups: preterm labour, term labour, preterm not in labour and term not in labour. The protein levels of IL-8, IL-6, monocyte chemotactic protein-1 (MCP-1), regulated upon activation normal t cells expressed and secreted (RANTES) and tumour necrosis factor-alpha (TNF-alpha) were quantified in tissue homogenates by ELISA or Immulite. The mRNA expression of IL-8, MCP-1 and RANTES was studied using RT-PCR. White blood cell count (WBC) and C-reactive protein (CRP) in the blood were determined. For determination of statistically significant differences between study groups Mann-Whitney U test or Kruskal-Wallis test were applied.
Results
Protein concentrations of IL-8, IL-6, and MCP-1 were significantly increased during labour compared to non-labouring groups, whereas no changes were observed for RANTES and TNF-alpha. The mRNA levels of representative cytokines such as IL-8 and MCP-1 increased significantly during labour whereas RANTES mRNA expression remained unchanged. WBC and CRP were significantly higher in the labouring groups as compared to groups not in labour. For neither of the analysed cytokines, WBC or CRP levels were there any changes between preterm and term respective groups.
Conclusion
Our findings indicate that non-infected preterm cervical ripening is an inflammatory process, just as cervical ripening at term, with cytokines as important mediators.
==== Body
Background
Preterm delivery occurs in approximately 6–15% of all pregnancies [1]. Furthermore, it accounts for 70–75% of neonatal mortality and morbidity [2]. The association between lower and upper genital tract infections and preterm delivery is well established [3]. However, in many cases preterm labour seems to be idiopathic [1]. The basic mechanisms underlying the initiation of both preterm and term cervical ripening and labour remain largely unknown. Both term and preterm labour requires cervical softening and regular myometrial contractions. Cervical softening is the result of an intensive remodelling process of the extracellular matrix (ECM). During cervical ripening the concentration of collagen decreases [4] and its physical state changes [5]. Additionally, the concentration of the main cervical proteoglycan decorin declines [6] in contrast to an increase of the mRNA expression and the protein concentration of the large proteoglycan versican [7]. Cytokines and several other mediators such as oestrogen, progesterone, nitric oxide and prostaglandins are involved in the human cervical ripening and the remodelling of extracellular matrix [8-11]. This remodelling process can therefore be regarded as an inflammatory reaction [12,13]. The density of immunoactive cells such as leucocytes and macrophages increases 6 to 10-fold at labour compared to first trimester pregnancy [14]. They are known producers of a variety of proinflammatory cytokines and matrix metalloproteinase's, and promote the cervical extracellular matrix degradation [13,15-17]. Furthermore, chemokines and cytokines are capable of attracting these immunoactive cells to the site of inflammation and are involved in human pregnancy and parturition [18,19].
Earlier studies have shown an increase of IL-8, IL-6 and TNF-α in human gestational membranes during spontaneous term labour [18,20,21]. Furthermore, the amniotic fluid levels of IL-8, RANTES, IL-6 and TNF-α [20] and MCP-1 [22] increase markedly at the onset of spontaneous labour at term. Cytokine levels in preterm parturition are widely analysed in relation to infection. However, it is estimated that less than 50% of the cases of preterm labour with elevated cytokines are due to an infection [23]. In placental cell cultures from non-infected preterm labour significantly larger amounts of IL-1β, IL-6, and TNF-α were registered compared to those shown by non-labouring women at term [24]. RANTES in non-infected amniotic fluid in preterm delivery is significantly higher compared to women with term delivery [25].
Only a few studies have investigated the role of the cytokines in human cervical tissue at term. The gene expression of IL-6 and IL-8 is significantly up regulated with a corresponding increase in the protein concentrations in patients in labour compared to not in labour [17,26,27]. Furthermore, animal studies have shown that IL-8 when applied intracervically induces cervical ripening in guinea pigs [28]. Studying the conditions at preterm parturition, Winkler et al. has shown that in the lower uterine segment, the IL-1β, IL-6 and IL-8 concentrations increased in relation to progressing cervical dilatation, whereas TNF-α remained unchanged [29]. Cytokines in the preterm cervix, to our knowledge, have not been investigated before.
We hypothesized that non-infected preterm cervical ripening and labour are associated with increased cytokine concentrations in the cervical tissue and increased inflammatory markers in the peripheral blood. The aim of this study was to investigate if there are any differences between inflammatory process during preterm and term cervical ripening. Therefore, protein concentrations of IL-6, IL-8, MCP-1, RANTES and TNF-α as well as mRNA expression of IL-8, MCP-1 and RANTES were determined in preterm and term groups. Furthermore, an analysis of inflammatory markers in peripheral circulation – White Blood Cell count (WBC) and C-reactive protein (CRP) – was undertaken.
Methods
Study Participants
A total of fifty women undergoing singleton pregnancies were included in the present investigation. The two study groups included 17 women in spontaneous preterm labour (PTL) and 8 not in labour at preterm (PTnotL). Premature delivery was defined as delivery before the 37th week of gestation (Table 1). 14 women in term labour (TL) and 11 not in labour at term (TnotL) served as controls. Characterization of all study groups is summarized in Table 1. There were no significant differences regarding maternal age, parity and previous preterm births between the four groups.
Table 1 Characterization of the study groups
Parameter Preterm labour (PTL) Term labor (TL) Preterm not in labour (PTnotL) Term not in labor (TnotL)
N 17 14 8 11
Age (median, range) 28 (18–37) 28 (20–39) 32.5 (27–44) 29 (25–39)
Parity (median, range) 1 (0–3) 1 (0–3) 1 (0–7) 1 (0–2)
Previous preterm births 2 0 1 1
Gestational age in days (median, range) 242 (184–255) 281 (259–294) 219 (183–256) 270 (263–284)
Gestational age in full gestational weeks (range) 26+2 – 36+3 37+0 – 42+0 26+1 – 36+4 37+4 – 40+4
Treated with corticosteroids 2 0 5 0
The labour groups were in active labour and demonstrated a ripe cervix dilated >4 cm. In all patients delivered by caesarean section the assessment of cervical dilatation was established immediately before surgery, by the same obstetrician (SAT) through vaginal digital examination. The women in preterm labour were either delivered vaginally, or by emergency caesarean section due to malpresentation. The women in term labour were also either delivered vaginally or by emergency caesarean due to threatening foetal asphyxia. Women not in labour had unripe cervices (Bishop score <5p) and were delivered by caesarean section before the onset of labour. The preterm indications were suspected ablatio or intra-uterine growth retardation and the term indications were breech presentation, humanitarian or disproportion.
In all women clinical signs of infection were absent, during parturition as well as during the postpartal period.
Specimen Collection
Immediately following parturition, a biopsy from the anterior cervical lip was taken transvaginally at the 12 o'clock position with scissors and tweezers. Our group has since 25 years applied this technique to get samples including squamous and cylindrical epithelium, vessels, glands and ECM. The samples were immediately frozen in liquid nitrogen and stored thereafter at -70°C until further investigation. A venous blood sample was taken for analysis of CRP and WBC.
Due to the limited amount of tissue from each woman, all analyses could not be performed on every sample.
The Local Ethics Committee of the Karolinska Institute approved the study (Ref. No. 97-089) and all women gave their informed consent.
Tissue homogenisation
Frozen tissue was cut into small slices on a block of dry ice and transferred to a pre-chilled (liquid nitrogen) capsule containing Teflon coated tungsten ball. The capsule was kept in liquid nitrogen for two minutes and thereafter shaken in a dismembranation apparatus (Retsch KG, Haan, Germany) at full speed for two minutes. The procedure was repeated after intermediated freezing in liquid nitrogen until the tissue became powder. Hereafter followed either RNA extraction or measurement of cytokine concentrations.
Measurement of cytokine concentrations
Tissue preparation
Following the tissue homogenisation, 1 ml of phosphate-buffered saline (PBS) was added. After centrifugation at 400 g for 5 min, supernatant was retrieved. Protein levels of IL-6, IL-8, TNF-α, MCP-1 and RANTES were expressed as picograms of cytokine per mg of total protein (pg/mg protein). Total protein concentration was determined by Bio-Rad's Protein Asssay, based on Bradford dye-binding procedure (Bio-Rad Laboratories Inc., Hercules, CA, USA), according to the manufacturer's instructions.
Determination of cytokine levels
Cytokine IL-6, IL-8 and TNF-α protein analysis were performed employing IMMULITE Automated Analyser (Diagnostic Products Corp., Los Angeles, CA, USA), using the commercially available immulite chemiluminescent enzyme immunometric assays (Immulite®, DPC, Los Angeles, CA, USA) according to the manufacturer's instructions. Analytical sensitivity and intra-assay and between assay coefficients of variation were respectively 2 pg/ml, 6.2% and 7.5% for IL-6; 2 pg/ml, 3.8% and 7.4% for IL-8; 1.7 pg/ml, 3.6% and 6.5% for TNF-α.
RANTES and MCP-1 concentrations in the supernatants of homogenized cervical samples were determined in duplicates using the quantitative sandwich enzyme-linked immunoassays (ELISA) by commercially available kits (Quantikine, R&D Systems, Minneapolis, MN, USA). The detailed procedures are described in the instruction booklets supplied by the manufacturers. The results were interpolated from the standard reference curve provided with each kit. The sensitivity of kits was 5 pg/ml for MCP-1 and 8 pg/ml for RANTES. The intra-assay and inter-assay coefficients of variation were respectively 7.8% and 6.7% for MCP-1; 3.6% and 10.3% for RANTES.
Detection of mRNA by RT-PCR
RT-PCR was performed on the four groups of cervical biopsies; women in PTL (n = 17), TL (n = 14), PTnotL (n = 8) and TnotL (n = 10).
RNA extraction
Following the tissue homogenization (see above), total RNA was extracted with the help of Trizol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions. The RNA concentration was measured at 260/280 nm by the help of Eppendorf Bio Photometer (Eppendorf AG, Hamburg, Germany). The quality of total RNA was controlled by running on 1.5% agarose gels and visualised under ultraviolet light after ethidium bromide staining. Total RNA was subsequently stored at -70°C until further investigation.
Reverse transcription (RT)
From each sample 1 μg RNA was taken, to which 1 μl (250 ng) of pd(N)6 Random Hexamer 5'-Phosphate primers (Amersham Biosciences, Pistacaway, NJ, USA), 1 μl of 10 mM dNTP (Amersham Biosciences) and sterile water was added to 12 μl. The mixture was incubated for 5 min at 65°C, cooled down and centrifuged. The reaction mixture consisting of 4 μl of 5 × First-Strand Buffer, 2 μl of 0.1 M DTT (Invitrogen, Carlsbad, California) and 1 μl (40 U/μl) Protector RNase Inhibitor (Roche, Mannheim, Germany) was added and incubated 2 min at 42°C. 1 μl (200 U/μl) of SuperScript™ Rnase H- Reverse Transcriptase (Invitrogen, Carlsbad, California, USA) was added to each tube and mixed up. The RT step was carried out at 42°C for 50 min, followed by heating at 70°C for 15 min to inactivate the enzyme. The cDNA was stored at -70°C until used.
RT-PCR
All gene-specific primers (Table 2) were obtained from Invitrogen (Carlsbad, California, USA). Ribosomic 28S was used as housekeeping gene. PCR was performed on 2 μl cDNA using Master Taq kit (Eppendorf, Hamburg, Germany) in a 25 μl reaction mixture containing 2.5 μl 10 × PCR buffer, 0.625 8 mM dNTP, 0.15 μl Taq polymerase (5 U/μl), 1 μl of specific primers (5 μM) and water. The PCR reaction was run in Eppendorf Mastercycler® gradient (Eppendorf AG, Hamburg, Germany). Repeated experiments were performed to ensure that the PCR reaction was within the linear phase. 1 min at 95°C was followed by cycles of 1 min denaturation at 94°C, 1 min of annealing and 1 min of extension at 72°C. The final extension lasted for 5 min at 72°C and thereafter cooling to 4°C. The number of cycles and annealing temperatures for each primer are presented in table 2.
Table 2 Description of the primers used for RT-PCR. The sequences, the number of cycles, annealing temperature (Tm) and product size of primers used for RT-PCR analysis.
Genes 5' Primer 3' Primer Cycles Tm °C Size bp
IL-8 TCTCTTGGCAGCCTTCCT AATTCTCAGCCTCTTCAAAAACTT 34 61 276
MCP-1 CTCTGCCGCCCTTCTGTGCC GTCTTCGGAGTTTGGGTTTGC 28 61 288
RANTES CGGCACGCCTCGCTGTCATC TGTACTCCCGAACCCATTT 34 61 240
28S GTGCAGATCTTGGTGGTAGTAGC AGAGCCAATCCTTATCCCGAAGTT 19 58 552
Semi-quantification of IL-8, MCP-1 and RANTES mRNA
The PCR products were separated by electrophoresis on a 1.5% agarose gel (Amersham Biosciences AB, Uppsala, Sweden). Following staining with ethidium bromide, the gels were photographed and band intensity measured under UV light using Gel Doc 2000 (BioRad, Hercules, CA, USA). The specific mRNA level of every sample was expressed as the product's intensity, divided by the housekeeping gene 28S intensity (the product/28S intensity ratio).
The identity of PCR products was confirmed by sequencing them at KISeq, Center for Genomics and Bioinformatics, Karolinska Institutet.
Determination of WBC and CRP
WBC and CRP were determined in blood in the routine laboratory of Karolinska University Hospital Solna (Stockholm, Sweden).
Statistical Analysis
Comparison between two groups was performed using the Mann-Whitney U test. When all four groups were compared, the Kruskal-Wallis test was applied, followed by multiple comparison with Bonferroni correction. The level of significance was set at p < 0.05. Calculations were performed employing STATISTICA 6.0 software (StatSoft Inc, Tulsa, OK, USA).
Results
Protein concentrations of cytokines
Every analysed sample revealed detectable levels of proteins.
There were no significant differences between preterm and term respective groups, but differences reached significance when comparing groups in labour with not in labour.
The concentration of IL-6 was significantly higher in the PTL versus the PTnotL (p = 0.02) and TnotL (p = 0.002). Similarly, concentration was significantly higher in the TL group compared to groups not in labour (Figure 1).
Figure 1 IL-6 concentration in the cervical tissue in all study groups. Protein concentration of IL-6 is expressed in picograms/mg of total protein. The groups are: preterm labour (PTL), term labour (TL), preterm not in labour (PTnotL), term not in labour (TnotL). The number of patients analysed in each group is marked in each bar in the bar chart. The box represents median value with 25%–75% of all data falling within the box. The whiskers extend to the non-outlier range. Outliers are marked as circles. Significant differences between the groups are shown above the box plots.
Concentrations of IL-8 were significantly higher in the TL group compared to the PTnotL and the TnotL (p < 0.01). Although the same tendency was noted in the preterm groups, it didn't reach statistical significance (Figure 2A).
Figure 2 Protein concentration and mRNA expression of IL-8 in the cervical tissue. Box and whisker plots represent (A) the protein concentration of IL-8 (picograms/mg of total protein) and (B) expression of the IL-8 mRNA (expressed as the product/28S intensity ratio) in the cervical tissue of the study groups. The groups are: preterm labour (PTL), term labour (TL), preterm not in labour (PTnotL), term not in labour (TnotL). The number of patients analysed in each group is marked in each bar in the bar chart. The box represents median value with 25%–75% of all data falling within the box. The whiskers extend to the non-outlier range. Outliers are marked as circles. Significant differences between the groups are shown above the box plots.
MCP-1 protein levels showed a tendency to be higher in the PTL compared to PTnotL, but did not reach statistical significance, whereas in the PTL was significantly higher compared to TnotL (p = 0.02). The concentration of MCP-1 was significantly higher in the TL group than in groups not in labour (Figure 3A).
Figure 3 Protein concentration and mRNA expression of MCP-1 in the cervical tissue. Box and whisker plots represent (A) the protein concentration of MCP-1 (picograms/mg of total protein) and (B) expression of the MCP-1 mRNA (expressed as the product/28S intensity ratio) in the cervical tissue of the study groups. The groups are: preterm labour (PTL), term labour (TL), preterm not in labour (PTnotL), term not in labour (TnotL). The number of patients analysed in each group is marked in each bar in the bar chart. The box represents median value with 25%–75% of all data falling within the box. The whiskers extend to the non-outlier range. Outliers are marked as circles. Significant differences between the groups are shown above the box plots.
Even more significant differences (p < 0.0001) between labouring and not labouring groups were achieved when it was looked upon the data irrespective gestational age (Table 3).
Table 3 In labour group compared with not in labour group. In labour group includes preterm labour (PTL) and term labour (TL) groups, not in labour group includes preterm not in labour (PTnotL) and term not in labour (TnotL) groups. Protein concentrations are expressed in pg/mg of total protein. Expression of mRNA is the product/28S intensity ratio. Significant differences between the groups were determined using Mann-Whitney U Test. NS-statistically not significant difference.
Measured factor In Labour Group (PTL and TL) Not in Labour Group (PTnotL and TnotL) p value
Median Min Max N Median Min Max N
IL-6 protein 396.7 7.7 1286.4 28 5.9 2.1 215.0 18 <0.0001
IL-8 protein 818.9 13.9 1288.1 28 71.0 4.8 1278.9 19 <0.0001
MCP-1 protein 420.8 21.2 1293.5 28 39.5 15.4 598.7 19 <0.0001
TNF-α protein 4.6 0.8 31.4 18 3.5 1.7 7.8 10 NS
RANTES protein 576.5 165.25 1075.3 18 439.5 141.0 869.6 10 NS
IL-8 mRNA 0.4 0 8.9 31 0 0 0.1 18 <0.0001
MCP-1 mRNA 1.9 0 67.2 31 0.4 0 1.4 18 <0.0001
RANTES mRNA 1.3 0 54.0 31 1.1 0.03 13.8 18 NS
WBC 15.9 7.3 28.4 21 9.6 6.9 14.1 14 <0.001
CRP 13.0 6.9 123.0 18 7.5 6.9 17.0 8 0.0221
Interestingly, there were no significant changes between preterm or term groups as regards the protein concentrations of TNF-α or RANTES (data not shown, combined data – table 3).
IL-8, MCP-1 and RANTES mRNA expression
In line with protein data, there were no differences registered in the representative cytokine mRNA expression between preterm and term respective groups. On the other hand, mRNA expression of IL-8 and MCP-1 was significantly higher in labour compared to not in labour groups.
IL-8 mRNA levels were significantly higher in the PTL group compared to the PTnotL (p = 0.01) and to the TnotL group (p = 0.03). Furthermore, there were significant differences comparing TL with groups not in labour (Figure 2B).
In line with the IL-8 expression, MCP-1 mRNA levels were significantly higher in labouring groups compared to not labouring groups (Figure 3B).
In line with protein data, differences were even more significant (p < 0.0001) comparing labouring (including PTL and TL) and non-labouring groups (including PTnotL and TnotL) (Table 3).
Similarly to the protein levels, no significant changes were found in RANTES mRNA expression (data not shown, combined data – table 3).
WBC and CRP levels
In line with cytokine levels, there were no differences in WBC (109/l) and CRP (mg/l) levels between preterm and term respective groups. However, higher levels of inflammatory markers were found in the labouring groups compared to the nonlabouring groups. WBC was significantly (p = 0.017) higher in the preterm labour group with a median value of 15.9 (range 9.2–24.3) compared to the preterm group not in labour, where the median value was 8.6 (6.9–13.6). The same tendency was seen in the term groups, where median and range in the TL group was 15.65 (7.3–28.4) and in the not labour group 10.0 (7.4–14.0) respectively, however without statistical significance.
CRP levels also revealed the same tendency, although differences didn't reach statistical significance. Median and range in the groups were respectively 13.0 (6.9–32.0) in PTL, 13.0 (6.9–123.0) in TL, 12.5 (8.0–17.0) in PTnotL and 7.0 (6.9–10.0) in TnotL.
WBC and CRP were significantly higher when compared the labouring groups (PTL and TL) to the groups not in labour (PTnotL and TnotL) (p < 0.001 and p = 0.02 respectively) irrespective to gestational age (Figure 4, Table 3).
Figure 4 WBC and CRP levels in labour compared to not in labour group. Box and whisker plots represent (A) white blood cell count (109/l) (B) C-reactive protein (mg/l) in the blood of women in the study groups. The two groups are: In labour (including preterm labour (PTL) and term labour (TL) and not in labour (including preterm not in labour (PTnotL) and term not in labour (TnotLabour). The number of patients analysed in each group is marked in each bar in the bar chart. The box represents median value with 25%–75% of all data falling within the box. The whiskers extend to the non-outlier range. Outliers are marked as circles. Significant differences between the groups are shown above the box plots.
Discussion
To our knowledge, this is the first investigation on cytokines in preterm cervical tissues in non-infected subjects.
The hypothesis that clinically non-infectious preterm cervical ripening and labour is associated with increased cytokine levels was tested and verified for IL-8, IL-6 and MCP-1 protein concentrations. This was not valid for RANTES or TNF-α protein concentrations which all remained unchanged.
Interestingly, no significant differences were revealed between the preterm and term labour samples (ripe cervices) or between preterm and term not in labour (unripe cervices) samples. In other words, this study could not identify any differences related to the gestational age, while differences were found related to the cervical state.
Our findings concerning term cervical ripening agree with those in previous studies, where increase in IL-8 and IL-6 concentrations and mRNA expression was found in cervical tissue in labouring compared to not labouring subjects [17,26,27,30]. Analogous changes were registered in the lower uterine segment [31-33].
Looking at the changes in preterm labour, Winkler et al has shown similarities in IL-1β, IL-6 and IL-8 increase in the lower uterine segment during preterm and term parturition [29]. The only observed difference was that this increase at preterm starts at earlier stages of cervical dilatation [29,34]. Even though the lower uterine segment contains lower amounts of extracellular matrix, undergoes less intense changes and has another function than the cervix during pregnancy and parturition, the remodelling process there may still somewhat reflect events in the cervix [35,36]. Our findings in the preterm cervical tissue agree with the study on the lower uterine segment, but we are not able to judge at which stage of cervical dilatation the increase in cytokine levels begins.
In our earlier study, we have shown a decrease in 15-hydroxyprostaglandin dehydrogenase expression related to the cervical state irrespective of gestational age [37]. All these findings suggest that cervical ripening at preterm is a similar process as at term. This similarity can be further confirmed, looking upon our results allocated into two groups: in labour (containing preterm and term labour) and not in labour (containing preterm and term not in labour), where even more significant differences between labouring and not labouring groups are achieved. Analogous results are seen in WBC count and CRP levels, where these markers seem to rise significantly in the labouring groups compared to the groups not in labour. This suggests that local inflammatory process in the cervix can be reflected in the peripheral blood. IL-6 is a prominent stimulator of the acute phase response in inflammatory reactions and stimulates CRP. Taking that into account, it is logical that when we identify a significant rise in IL-6 in the labour groups we can expect a concomitant rise in the CRP level. The WBC count and the CRP level are used routinely as markers of infection in daily clinical use, but our findings may suggest that WBC and CRP could probably be markers of active labour without any infection. Further studies with larger number of patients are needed to confirm this.
All these findings cannot answer the question which signals cause the cytokine increase in the labouring cervix. Preterm labour in correlation with infection is extensively studied. A significant cytokine increase is seen in gestational membranes, amniotic and cervicovaginal fluid in infected preterm labour [3,21,38-40]. In our study, none of the women showed clinical or laboratory signs of infection before or after labour. This could also account for the unchanged levels of TNF-α in our and Winkler et al study [29], as TNF-α is shown to be related to infection in the amniotic fluid [41]. However, no correlation was found between chorioamnionitis and elevated IL-6 and IL-8 levels of amniotic fluid in term pregnant women [42]. Our findings are in line with earlier studies, where cytokine increase was seen in non infected preterm parturition [23,24,29] suggest that some other signals than infection could be responsible for starting preterm cervical ripening and labour process. The down-regulation of progesterone receptors and oestrogen receptor α with significant up-regulation of oestrogen receptor β (ERβ) at term pregnant cervix [8,43] may be involved. Furthermore, there is a possibility of direct effect of estrogens via ERβ on cervical leukocytes [44], which are a major source of proinflammatory cytokines in the cervix during labour [45]. Foetal fibronectine could be also involved in foetal-maternal signalling, as its level in the cervicovaginal fluid rises in line with the ripening of cervix and is elevated at preterm birth [46-48]. It can also be localised in the cervical epithelium as well as IL-8 [26,47]. Corticotropine releasing hormone (CRH) is elevated in maternal serum in preterm labour [49] and there is possible cross-regulation between CRH and cytokines, but earlier studies show controversial results [50-52]. Further investigations are required to clarify the involvement of these factors in the preterm cervical ripening and labour and their relation with the cytokine increase.
Conclusion
In conclusion, preterm cervical ripening can be likened to an inflammatory process with cytokines as important mediators, corresponding to the process at term cervical ripening. The local changes in the cervix may be reflected in the peripheral blood as an increase in WBC count and CRP.
Authors' contributions
SAT have selected and recruited the patients, collected all the biopsies, participated in the design of the study, did a part of laboratory analyses, drafting of the manuscript. AK participated in the analysis and discussion of the results, performed a part of statistical analysis and drafted the manuscript. BB participated in design of the study, did RT-PCR, ELISA, Immulite analyses, participated in the discussion of results and revising the manuscript. MC participated in the design of the study, RT-PCR analysis, discussion of the results and statistical analysis. AB participated in the design of the study, discussion of the results, revising the manuscript. GEO participated in the design of the study, analysis and discussion of the results, drafting and critical revising of the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors of the article would like to thank Yvonne Pierre for the help with ELISA and Immulite analyses.
The present study was made possible by financial support from The Swedish Research Council Grant (Grant GEO 349-2002-7189) and the Karolinska Institute Funds.
==== Refs
Slattery MM Morrison JJ Preterm delivery Lancet 2002 360 1489 1497 12433531 10.1016/S0140-6736(02)11476-0
Challis JRG Mechanism of parturition and preterm labor Obstet Gynecol Surv 2000 55 650 660 11023206 10.1097/00006254-200010000-00025
Romero R Espinoza J Chaiworapongsa T Kalache K Infection and prematurity and the role of preventive strategies Semin Neonatol 2002 7 259 274 12401296
Uldbjerg N Ekman G Malmstrom A Olsson K Ulmsten U Ripening of the human uterine cervix related to changes in collagen, glycosaminoglycans, and collagenolytic activity Am J Obstet Gynecol 1983 147 662 666 6638110
Ekman G Malmstrom A Uldbjerg N Ulmsten U Cervical collagen: an important regulator of cervical function in term labor Obstet Gynecol 1986 67 633 636 3457329
Norman M Ekman G Malmstrom A Changed proteoglycan metabolism in human cervix immediately after spontaneous vaginal delivery Obstet Gynecol 1993 81 217 223 8423954
Westergren-Thorsson G Norman M Bjornsson S Endresen U Stjernholm Y Ekman G Malmstrom A Differential expressions of mRNA for proteoglycans, collagens and transforming growth factor-beta in the human cervix during pregnancy and involution Biochim Biophys Acta 1998 1406 203 213 9573366
Wang H Stjernholm Y Ekman G Eriksson H Sahlin L Different regulation of oestrogen receptors alpha and beta in the human cervix at term pregnancy Mol Hum Reprod 2001 7 293 300 11228250 10.1093/molehr/7.3.293
Stjernholm Y Sahlin L Malmstrom A Barchan K Eriksson HA Ekman G Potential roles for gonadal steroids and insulin-like growth factor I during final cervical ripening Obstet Gynecol 1997 90 375 380 9277647 10.1016/S0029-7844(97)00245-7
Abelin Tornblom S Ostlund E Granstrom L Ekman G Pre-term cervical ripening and labor induction Eur J Obstet Gynecol Reprod Biol 2002 104 120 123 12206923 10.1016/S0301-2115(02)00101-X
Hertelendy F Zakar T Prostaglandins and the myometrium and cervix Prostaglandins Leukot Essent Fatty Acids 2004 70 207 222 14683694 10.1016/j.plefa.2003.04.009
Bowen JM Chamley L Keelan JA Mitchell MD Cytokines of the placenta and extra-placental membranes: roles and regulation during human pregnancy and parturition Placenta 2002 23 257 273 11969336 10.1053/plac.2001.0782
Liggins GC Ripening of the cervix Semin Perinatol 1978 2 261 271 366764
Bokstrom H Brannstrom M Alexandersson M Norstrom A Leukocyte subpopulations in the human uterine cervical stroma at early and term pregnancy Hum Reprod 1997 12 586 590 9130764 10.1093/humrep/12.3.586
Thomson AJ Telfer JF Young A Campbell S Stewart CJ Cameron IT Greer IA Norman JE Leukocytes infiltrate the myometrium during human parturition: further evidence that labour is an inflammatory process Hum Reprod 1999 14 229 236 10374126 10.1093/humrep/14.1.229
Ledingham MA Thomson AJ Jordan F Young A Crawford M Norman JE Cell adhesion molecule expression in the cervix and myometrium during pregnancy and parturition Obstet Gynecol 2001 97 235 242 11165588 10.1016/S0029-7844(00)01126-1
Osman I Young A Ledingham MA Thomson AJ Jordan F Greer IA Norman JE Leukocyte density and pro-inflammatory cytokine expression in human fetal membranes, decidua, cervix and myometrium before and during labour at term Mol Hum Reprod 2003 9 41 45 12529419 10.1093/molehr/gag001
Garcia-Velasco JA Arici A Chemokines and human reproduction Fertil Steril 1999 71 983 993 10360897 10.1016/S0015-0282(99)00120-X
Kayisli UA Mahutte NG Arici A Uterine chemokines in reproductive physiology and pathology Am J Reprod Immunol 2002 47 213 221 12069388 10.1034/j.1600-0897.2002.01075.x
Keelan JA Blumenstein M Helliwell RJ Sato TA Marvin KW Mitchell MD Cytokines, prostaglandins and parturition--a review Placenta 2003 24 Suppl A S33 46 12842412 10.1053/plac.2002.0948
Keelan JA Marvin KW Sato TA Coleman M McCowan LM Mitchell MD Cytokine abundance in placental tissues: evidence of inflammatory activation in gestational membranes with term and preterm parturition Am J Obstet Gynecol 1999 181 1530 1536 10601939
Esplin MS Romero R Chaiworapongsa T Kim YM Edwin S Gomez R Gonzalez R Adashi EY Amniotic fluid levels of immunoreactive monocyte chemotactic protein-1 increase during term parturition J Matern Fetal Neonatal Med 2003 14 51 56 14563093
Farina L Winkelman C A review of the role of proinflammatory cytokines in labor and noninfectious preterm labor Biol Res Nurs 2005 6 230 238 15583363 10.1177/1099800404271900
Steinborn A Gunes H Roddiger S Halberstadt E Elevated placental cytokine release, a process associated with preterm labor in the absence of intrauterine infection Obstet Gynecol 1996 88 534 539 8841213 10.1016/0029-7844(96)00224-4
Athayde N Romero R Maymon E Gomez R Pacora P Araneda H Yoon BH A role for the novel cytokine RANTES in pregnancy and parturition Am J Obstet Gynecol 1999 181 989 994 10521766
Sennstrom MB Ekman G Westergren-Thorsson G Malmstrom A Bystrom B Endresen U Mlambo N Norman M Stabi B Brauner A Human cervical ripening, an inflammatory process mediated by cytokines Mol Hum Reprod 2000 6 375 381 10729321 10.1093/molehr/6.4.375
Sennstrom MK Brauner A Lu Y Granstrom LM Malmstrom AL Ekman GE Interleukin-8 is a mediator of the final cervical ripening in humans Eur J Obstet Gynecol Reprod Biol 1997 74 89 92 9243210 10.1016/S0301-2115(97)02757-7
Chwalisz K Benson M Scholz P Daum J Beier HM Hegele-Hartung C Cervical ripening with the cytokines interleukin 8, interleukin 1 beta and tumour necrosis factor alpha in guinea-pigs Hum Reprod 1994 9 2173 2181 7868694
Winkler M Kemp B Fischer DC Maul H Hlubek M Rath W Tissue concentrations of cytokines in the lower uterine segment during preterm parturition J Perinat Med 2001 29 519 527 11776683 10.1515/JPM.2001.072
Sakamoto Y Moran P Searle RF Bulmer JN Robson SC Interleukin-8 is involved in cervical dilatation but not in prelabour cervical ripening Clin Exp Immunol 2004 138 151 157 15373918 10.1111/j.1365-2249.2004.02584.x
Osmers RG Adelmann-Grill BC Rath W Stuhlsatz HW Tschesche H Kuhn W Biochemical events in cervical ripening dilatation during pregnancy and parturition J Obstet Gynaecol 1995 21 185 194 8556580
Kemp B Menon R Fortunato SJ Winkler M Maul H Rath W Quantitation and localization of inflammatory cytokines interleukin-6 and interleukin-8 in the lower uterine segment during cervical dilatation J Assist Reprod Genet 2002 19 215 219 12099551 10.1023/A:1015354701668
Maul H Nagel S Welsch G Schafer A Winkler M Rath W Messenger ribonucleic acid levels of interleukin-1 beta, interleukin-6 and interleukin-8 in the lower uterine segment increased significantly at final cervical dilatation during term parturition, while those of tumor necrosis factor alpha remained unchanged Eur J Obstet Gynecol Reprod Biol 2002 102 143 147 11950481 10.1016/S0301-2115(01)00606-6
Winkler M Role of cytokines and other inflammatory mediators BJOG 2003 110 Suppl 20 118 123 12763128
Granstrom L Ekman G Ulmsten U Malmstrom A Changes in the connective tissue of corpus and cervix uteri during ripening and labour in term pregnancy Br J Obstet Gynaecol 1989 96 1198 1202 2590655
Hjelm AM Barchan K Malmstrom A Ekman-Ordeberg GE Changes of the uterine proteoglycan distribution at term pregnancy and during labour Eur J Obstet Gynecol Reprod Biol 2002 100 146 151 11750954 10.1016/S0301-2115(01)00476-6
Tornblom SA Patel FA Bystrom B Giannoulias D Malmstrom A Sennstrom M Lye SJ Challis JR Ekman G 15-hydroxyprostaglandin dehydrogenase and cyclooxygenase 2 messenger ribonucleic acid expression and immunohistochemical localization in human cervical tissue during term and preterm labor J Clin Endocrinol Metab 2004 89 2909 2915 15181076 10.1210/jc.2003-031149
Jacobsson B Mattsby-Baltzer I Andersch B Bokstrom H Holst RM Wennerholm UB Hagberg H Microbial invasion and cytokine response in amniotic fluid in a Swedish population of women in preterm labor Acta Obstet Gynecol Scand 2003 82 120 128 12648172 10.1034/j.1600-0412.2003.00047.x
Jacobsson B Holst RM Wennerholm UB Andersson B Lilja H Hagberg H Monocyte chemotactic protein-1 in cervical and amniotic fluid: relationship to microbial invasion of the amniotic cavity, intra-amniotic inflammation, and preterm delivery Am J Obstet Gynecol 2003 189 1161 1167 14586371 10.1067/S0002-9378(03)00594-5
El-Bastawissi AY Williams MA Riley DE Hitti J Krieger JN Amniotic fluid interleukin-6 and preterm delivery: a review Obstet Gynecol 2000 95 1056 1064 10808034 10.1016/S0029-7844(00)00875-9
Romero R Mazor M Sepulveda W Avila C Copeland D Williams J Tumor necrosis factor in preterm and term labor Am J Obstet Gynecol 1992 166 1576 1587 1595815
Axemo P Brauner A Pettersson M Eriksson L Rwamushaija E Bergstrom S Amniotic fluid interleukins in Swedish and Mozambican pregnant women Gynecol Obstet Invest 1996 41 113 117 8838971
Stjernholm Y Sahlin L Akerberg S Elinder A Eriksson HA Malmstrom A Ekman G Cervical ripening in humans: potential roles of estrogen, progesterone, and insulin-like growth factor-I Am J Obstet Gynecol 1996 174 1065 1071 8633638
Stygar D Wang H Vladic YS Ekman G Eriksson H Sahlin L Co-localization of oestrogen receptor beta and leukocyte markers in the human cervix Mol Hum Reprod 2001 7 881 886 11517296 10.1093/molehr/7.9.881
Young A Thomson AJ Ledingham M Jordan F Greer IA Norman JE Immunolocalization of proinflammatory cytokines in myometrium, cervix, and fetal membranes during human parturition at term Biol Reprod 2002 66 445 449 11804961
Ekman G Granstrom L Malmstrom A Sennstrom M Svensson J Cervical fetal fibronectin correlates to cervical ripening Acta Obstet Gynecol Scand 1995 74 698 701 7572103
Sennstrom MB Granstrom LM Lockwood CJ Omazic B Johansson O Malmstrom A Ekman GE Cervical fetal fibronectin correlates to prostaglandin E2-induced cervical ripening and can be identified in cervical tissue Am J Obstet Gynecol 1998 178 540 545 9539523
Leitich H Kaider A Fetal fibronectin--how useful is it in the prediction of preterm birth? Bjog 2003 110 Suppl 20 66 70 12763115
Challis JR Smith SK Fetal endocrine signals and preterm labor Biol Neonate 2001 79 163 167 11275645 10.1159/000047085
Dudley DJ Immunoendocrinology of preterm labor: the link between corticotropin-releasing hormone and inflammation Am J Obstet Gynecol 1999 180 S251 6 9914628
Angioni S Petraglia F Gallinelli A Cossarizza A Franceschi C Muscettola M Genazzani AD Surico N Genazzani AR Corticotropin-releasing hormone modulates cytokines release in cultured human peripheral blood mononuclear cells Life Sci 1993 53 1735 1742 8246669 10.1016/0024-3205(93)90160-5
Sehringer B Schafer WR Wetzka B Deppert WR Brunner-Spahr R Benedek E Zahradnik HP Formation of proinflammatory cytokines in human term myometrium is stimulated by lipopolysaccharide but not by corticotropin-releasing hormone J Clin Endocrinol Metab 2000 85 4859 4865 11134154 10.1210/jc.85.12.4859
|
16122384
|
PMC1201172
|
CC BY
|
2021-01-04 16:37:12
|
no
|
Reprod Biol Endocrinol. 2005 Aug 25; 3:39
|
utf-8
|
Reprod Biol Endocrinol
| 2,005 |
10.1186/1477-7827-3-39
|
oa_comm
|
==== Front
RetrovirologyRetrovirology1742-4690BioMed Central London 1742-4690-2-491609114310.1186/1742-4690-2-49ResearchPhosphatidylserine treatment relieves the block to retrovirus infection of cells expressing glycosylated virus receptors Coil David A [email protected] A Dusty [email protected] Division of Human Biology, Fred Hutchinson Cancer Research Center, Seattle, Washington 98109-1024 USA2 Molecular and Cellular Biology Program, Fred Hutchinson Cancer Research Center, Seattle, Washington 98109-1024 USA2005 9 8 2005 2 49 49 19 5 2005 9 8 2005 Copyright © 2005 Coil and Miller; licensee BioMed Central Ltd.2005Coil and Miller; 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 determinant of retrovirus host range is the presence or absence of appropriate cell-surface receptors required for virus entry. Often orthologs of functional receptors are present in a wide range of species, but amino acid differences can render these receptors non-functional. In some cases amino acid differences result in additional N-linked glycosylation that blocks virus infection. The latter block to retrovirus infection can be overcome by treatment of cells with compounds such as tunicamycin, which prevent the addition of N-linked oligosaccharides.
Results
We have discovered that treatment of cells with liposomes composed of phosphatidylserine (PS) can also overcome the block to infection mediated by N-linked glycosylation. Importantly, this effect occurs without apparent change in the glycosylation state of the receptors for these viruses. This effect occurs with delayed kinetics compared to previous results showing enhancement of virus infection by PS treatment of cells expressing functional virus receptors.
Conclusion
We have demonstrated that PS treatment can relieve the block to retrovirus infection of cells expressing retroviral receptors that have been rendered non-functional by glycosylation. These findings have important implications for the current model describing inhibition of virus entry by receptor glycosylation.
==== Body
Background
Many of the cellular receptors for retroviruses have been well characterized (for review see [1]). These receptors perform a wide variety of cellular functions and can be single-transmembrane, GPI-anchored, or multiple-membrane-spanning proteins. The presence or absence of functional receptors on the cell surface is a major determinant of virus tropism. In some cases, otherwise functional receptors are glycosylated and therefore unusable by particular retroviruses [2-6]. Since these sites of glycosylation are often near the binding sites used by viruses, glycosylation is thought to be an important defense mechanism evolved by cells in their battle against virus infection (for example see [7]).
One particularly well-studied example of glycosylation-blocked receptors involves those for the cat endogenous retrovirus RD114, which is unable to enter NIH 3T3 mouse cells unless these cells have been treated with agents, including tunicamycin, that prevent the addition of N-linked oligosaccharides to proteins in the endoplasmic reticulum. The receptor for RD114 in tunicamycin-treated NIH 3T3 cells is a multiple-membrane spanning protein called ASCT1 (standard name SLC1A4), which is a neutral amino acid transporter [5]. RD114 also uses a closely related human protein, ASCT2 (standard name SLC1A5, also called RDR) as a receptor [8,9]. Sequence differences in the mouse ortholog of human ASCT2 prevent it from serving as a receptor, even after tunicamycin treatment [5].
Other examples of glycosylation-blocked receptors are the hamster and rat orthologs of the receptor for Moloney murine leukemia virus (MoMLV), CAT1 (standard name SLC7A1). Prevention of receptor glycosylation by treatment of rat or hamster cells with tunicamycin relieves the block to infection by MoMLV [3,10]. Like ASCT1 and ASCT2, CAT1 is an amino acid transporter, in this case for lysine, arginine, and ornithine [11-13]. If the N-linked glycosylation sites of mouse ASCT1, hamster CAT1, or rat CAT1 are removed through mutagenesis, these proteins are fully functional as virus receptors [3,10,14]. To date, removal of N-linked glycosylation through either mutagenesis of the oligosaccharide attachment sites or by treatment with inhibitors of glycosylation are the only ways known to relive the block to infection by RD114 and MoMLV viruses in the respective rodent cell lines.
We recently have shown that treatment of target cells with phosphatidylserine (PS) enhances enveloped virus infection by up to 20-fold [15]. This effect is not observed with other phospholipids, and is thought to occur through an enhancement of virus fusion [15]. Importantly, in all cases tested where a functional receptor was present, PS treatment enhanced virus infection. Conversely, when a functional receptor was not present, PS treatment did not allow infection of target cells. Here we show that phosphatidylserine treatment can relieve the block to infection mediated by glycosylation-blocked receptors and further investigate this phenomenon.
Results
PS treatment allows infection of cell types expressing glycosylation-blocked receptors
Our previous work demonstrated that PS-dependent enhancement of infection requires functional receptors [15], and we will refer to this effect as "non-specific enhancement" of virus infection by PS. We wanted to extend these observations by examining the effects of PS on virus entry in the case where the receptor was present but was inactive due to receptor glycosylation. We used the LAPSN retroviral vector [16] that encodes human placental alkaline phosphatase (AP) as a marker for infection. Viruses carrying this vector contained Gag-Pol proteins from MoMLV and Env proteins from either MoMLV or RD114. For simplicity we will call these viruses MoMLV or RD114 vectors, respectively. MoMLV vectors are unable to enter CHO cells and RD114 vectors are unable to enter NIH 3T3 cells unless these cells are first treated with tunicamycin to prevent receptor glycosylation [3-5]. Table 1 shows that pretreatment of CHO and NIH 3T3 cells with 400 μM PS for 24 h allowed efficient entry of MoMLV and RD114 vectors, respectively. Hereafter we will refer to this effect as "glycosylation-specific enhancement" by PS, in contrast to the "non-specific enhancement" described in our previous work.
Table 1 PS treatment allows infection of cells expressing glycosylation-blocked retrovirus receptorsa
Target cells Vector PS treatment Vector titer (AP+ FFU/ml)
NIH 3T3 RD114 - <1
RD114 + 2.3 × 104
CHO-K1 MoMLV - <5
MoMLV + 2.3 × 103
a Virus infections and PS preparation were performed as described in Materials and Methods. Where indicated, cells were treated with 400 μM PS for 24 h. Data shown are the averages of three independent experiments, each done in duplicate. Values from different experiments varied no more than three-fold.
PS treatment does not affect receptor glycosylation
A simple explanation for these results might be that PS inhibits receptor glycosylation, as does tunicamycin treatment. As described above, murine ASCT1 functions as a receptor for RD114 in NIH 3T3 cells treated with tunicamycin [5]. Treatment of cell lysates with peptide N-glycosidase F (PNGase F) causes an increase in the electrophoretic mobility of ASCT1 as a result of removal of the N-linked glycosylation [14]. We attempted to examine the glycosylation status of a myc-tagged ASCT1 protein in NIH 3T3 cells but were unable to clearly visualize the protein due to technical problems including high background antibody binding. However, we were able to examine the glycosylation state of a hemagglutinin (HA)-tagged human ASCT2 protein in NIH 3T3/ASCT2 cells (Figure 1A). In the non-PS treated cells there was a clear increase in mobility of ASCT2 when incubated with PNGase F, demonstrating that this protein is normally glycosylated. Furthermore, none of the protein is found in the unglycosylated state prior to PNGase F treatment. The same mobility shifts were observed in cells treated with PS, indicating that treatment with PS does not affect the glycosylation state of this protein in NIH 3T3 cells.
Figure 1 Analysis of N-linked oligosaccharide modification of ASCT1 and ASCT2 with or without PS treatment. (A) NIH 3T3/ASCT2 cells that express HA-tagged human ASCT2 were treated with 400 μM PS for 24 h. Cell lysates were treated with or without PNGase F as described in Materials in Methods, and lysates were analyzed by Western immunoblotting with anti HA-tag monoclonal antibody. (B) 293T cells were transiently transfected with a myc-tagged murine ASCT1 expression plasmid. 400 μM PS was added 24 h post-transfection. Cell lysates were made 48 h post-transfection, were treated with or without PNGase F as described in Materials in Methods, and were analyzed by Western immunoblotting with anti Myc-tag monoclonal antibody.
To examine ASCT1 glycosylation directly, we transiently expressed a myc-tagged mouse ASCT1 in 293T cells and examined the effects of PS treatment on glycosylation (Figure 1B). These cells were treated with either 35 μM PS or were left untreated. This concentration of PS was chosen because it induced the highest vector infection rate in 293T cells and a high concentration of PS (400 μM) was toxic to 293T cells (data not shown). As for the HA-tagged ASCT2 protein, there was no detectable unglycosylated receptor present in the PS treated cells, indicating that ASCT1 glycosylation is unaffected by PS treatment.
The non-specific enhancement of infection by PS treatment occurs rapidly
We have previously postulated that the non-specific enhancement of virus infection by PS occurs through an effect on virus fusion [15]. If this were true, the effect should happen relatively quickly since all that is required is for the PS liposomes to fuse with the plasma membrane of the cell and change the physical characteristics of the membrane. We undertook infections using RD114 vector on normally infectable NIH 3T3/ASCT2 cells given only a short exposure to PS, in contrast to the 24 h treatment used in previous experiments. Cells were treated with PS for 1 h, virus was added for 2 h, and the cells were trypsinized and replated. With only 1 h of PS treatment, virus infection was increased almost 4-fold. This experiment was repeated twice with the same results. While not as much as the full 10 to 20-fold increase in infection when treated for 24 h, this demonstrates that the effect of PS on virus infection is indeed rapid. However when the parental NIH 3T3 cells, containing the glycosylation-blocked receptor, were treated in the same manner, no infection by the RD114 vector was observed (data not shown).
The non-specific and glycosylation-specific enhancements of infection have different time courses
The preceding results suggest that the glycosylation-specific enhancement of PS treatment is delayed when compared to the non-specific enhancement of virus infection. To compare these two effects we examined RD114 vector infection of both NIH 3T3 cells and NIH 3T3/ASCT2 cells over a longer time course. Cells were treated with PS at time points from 4–24 h and were then infected with the RD114 vector. The cell surface PS levels were also measured at each timepoint by annexin-V staining. We found a linear relationship between the time after PS addition and the amount of PS present in the outer leaflet of the membrane (Figure 2, top panel). Furthermore, there was a direct relationship between the amount of PS present in the membrane and infection of normally-infectable NIH 3T3/ASCT2 cells by the RD114 vector (Figure 2, middle panel). In contrast, there was a long delay in the increase in RD114 vector infection of NIH 3T3 cells following PS addition, with the major enhancement of virus infection occurring after 12 h Figure 2, bottom panel).
Figure 2 Time course of cell-surface PS levels and cell susceptibility to RD114 vector infection of NIH 3T3/ASCT2 and NIH 3T3 cells during treatment with PS. Cells were plated on day 0. 400 μM PS was added on day 1 at 24, 20, 16, 12, 8, and 4 h pre-infection. At the time of infection, cells were either infected with the RD114 vector [LAPSN(RD114)] or were assayed for cell-surface PS levels by using annexin-V. Top panel: annexin-V staining of NIH 3T3 cells was undertaken as described in Materials and Methods. Middle panel: LAPSN(RD114) infection of NIH 3T3/ASCT2 cells. Bottom Panel: LAPSN(RD114) infection of NIH 3T3 cells. Data points shown are means of duplicates, and each series represents an independent experiment. Data is represented as a percentage of the highest value observed.
Effects of PS at reduced concentrations on RD114 vector infection of NIH 3T3 cells
The long delay between addition of PS and the glycosylation-specific enhancement of virus infection suggests that a threshold amount of PS in the cell membrane may be required for the observed enhancement. To address this possibility we undertook a 24-h time course as described above, using half the amount of PS (200 μM) (Figure 3). The total amount of PS incorporated into the plasma membrane was reduced at each timepoint, and saturation did not appear to be reached. The reduced incorporation of PS had the result of increasing the delay of RD114 vector infection of NIH 3T3 cells from 12 to more than 16 h, supporting the hypothesis that a threshold amount of PS is required for the glycosylation-specific enhancement of virus infection.
Figure 3 Effects of PS at a reduced concentration on RD114 vector infection of NIH 3T3 cells. PS liposomes were generated and added to NIH 3T3 cells at either 400 μM or 200 μM concentration. Cells were analyzed for cell-surface PS levels by using annexin-V or were infected with the RD114 vector [LAPSN(RD114)] as described in Materials and Methods. Top panel: Annexin-V staining of NIH 3T3 cells. Bottom panel: LAPSN(RD114) infection of NIH 3T3 cells. Data shown are the average of duplicates. The entire experiment was repeated with very similar results.
The dose-response of non-specific and glycosylation-specific enhancement of virus infection by PS differs
It appears from the results shown in Figure 3 that there is a simple relationship between amount of PS present in the membrane and the non-specific enhancement of virus infection. We next examined the effect of 24 h treatment with various concentrations of PS on RD114 vector infection of both NIH 3T3/ASCT2 cells and NIH 3T3 cells (Figure 4). Infection and annexin-V measurements were undertaken as previously described. At very low levels of PS, which are not detectable by annexin-V, no infection on either cell type was observed. As soon as an increase in PS levels was observed, there was a corresponding increase in RD114 infection of the NIH 3T3/ASCT2 cells. However, infection of NIH 3T3 cells was not detectable until a higher concentration of PS was reached, further supporting the hypothesis of a required threshold concentration for infection through the glycosylation-specific pathway.
Figure 4 Effects of PS concentration on cell-surface PS levels and RD114 vector infection of NIH 3T3 or NIH 3T3/ASCT2 cells. PS liposomes were generated and added to cells at concentrations of 0, 6.4, 32, 80, 240, 320, and 400 μM. Annexin-V staining and infections were undertaken as described in Materials and Methods. Top panel: Annexin-V staining of NIH 3T3 cells. Middle panel: RD114 vector [LAPSN(RD114)] infection of NIH 3T3/ASCT2 cells. Bottom panel: LAPSN(RD114) infection of NIH 3T3 cells. Data shown are the average of duplicates. The entire experiment was repeated twice with very similar results.
Discussion
Here we report that PS treatment of target cells containing glycosylation-blocked viral receptors allows virus infection. Importantly, this occurs without removal of the oligosaccharide itself, unlike the case with tunicamycin treatment. Furthermore, this glycosylation-specific effect takes place in NIH 3T3 cells on a different timescale than the non-specific enhancement of virus infection by PS, and appears to require a threshold concentration of cell-surface PS. When NIH 3T3 cells are treated with 200 μM PS, they reach the same level of infectivity after 24 h as when treated with 400 μM PS, but take longer before infection is observable, suggesting that the observed enhancement of infection is not merely a signaling cascade initiated by the addition of PS to the cell. One explanation for such a long delay is that de novo protein synthesis is required for the glycosylation-specific effect of PS treatment. Additional experiments will be needed to address this question. Unfortunately, preliminary experiments have demonstrated that PS treatment combined with inhibition of protein synthesis by cycloheximide is lethal to cells (data not shown), further complicating this analysis.
Additionally we have shown that the non-specific enhancement by PS occurs rapidly, and there is a direct correlation between amount of cell-surface PS and the amount of non-specific enhancement of virus infection. This result supports our previous hypothesis that the non-specific enhancement occurs through an influence of virus fusion.
Our results suggest that the block to infection of glycosylated receptors may occur at a different stage of virus entry than previously assumed. It has been proposed that glycosylation prevents MoMLV or RD114 from binding to their cognate receptors, thereby terminating virus entry at a very early step [7]. However, our results demonstrate that these two viruses can still infect cells containing fully glycosylated receptors. However, we have not ruled out the possibility that PS might induce subtle changes in receptor glycosylation, such as alterations in the structure or branching of the N-linked oligosaccharides, that might affect virus entry.
Instead of a block to virus binding, it is possible that PS affects the packing or mobility of the receptors in the plasma membrane. Several groups have suggested that receptor clusters, or multivalent Env-receptor complexes are required for retrovirus infection [17-21]. For example, an ASLV-A virion appears to require multiple contacts with receptors in order to enter a fusogenic state [21]. It is possible that glycosylated receptors are normally unable to pack as tightly, or move through the membrane as rapidly as their unglycosylated forms in order to facilitate virus infection. In this model, the disruption to the plasma membrane caused by PS treatment could allow sufficient concentrations of receptor to contact the viral Env proteins and initiate fusion. Exogenous PS has been shown to affect the curvature and stability of a lipid bilayer, providing a mechanism for this disruption [22,23]. On the other hand, fewer receptor contacts could be required by the virus to form a fusion pore if the activation energy for fusion to occur has been lowered by PS treatment [15]. Similarly, it is possible that the glycosylation of the receptors prevents the membranes from coming in close enough contact to fuse, but that the destabilization of the plasma membrane by PS increases the distance at which this fusion can occur. Further study will be required to understand the mechanism of glycosylation-specific enhancement of virus entry through PS treatment.
Conclusion
In summary, these results expand on our previous findings regarding the mechanism of enhancement of virus infection by PS treatment, and demonstrate an effect of PS treatment on cells containing glycosylation-blocked receptors. The ability to promote CHO-K1 and NIH 3T3 infection by MoMLV and RD114 vectors without tunicamycin treatment should be of interest to researchers studying these viruses and to those studying the nature of the glycosylation-induced block to retrovirus infection.
Methods
Cell culture and plasmids
NIH 3T3 thymidine kinase-deficient mouse embryo fibroblasts [24], and 293T human embryonic kidney cells [25] were maintained at 37°C and 5% CO2 in Dulbecco's modified Eagle medium with a high concentration of glucose (4.5 g per liter) and 10% FBS. CHO-K1 hamster cells (ATCC CCL-61) were maintained in Minimal Essential Medium Alpha at 37°C and 5% CO2. Clonal NIH 3T3 cells expressing an HA-tagged human ASCT2 (NIH 3T3/ASCT2 cells) were generated by transduction with the retroviral vector LNCRDRHA, that contains a human RDR (ASCT2) cDNA with a carboxy-terminal HA tag cloned into the LNCX retroviral vector [26]. The expression plasmid containing the myc-tagged murine ASCT1 was kindly provided by David Kabat [14].
Virus production
LAPSN is a Moloney murine leukemia virus (MoMLV)-based vector that encodes human placental alkaline phosphatase (AP) and neomycin phosphotransferase [16]. LAPSN containing viruses were generated from the following packaging lines expressing the indicated Env proteins; FlyRD (RD114) [27], and PE501 (MoMLV) [26]. All retroviral vectors used in these studies were harvested in medium exposed to producer cells and were centrifuged at 1,000 × g for 5 min to remove cells and debris.
Virus assays
All retrovirus vector infections were undertaken as follows. On day 0, cells were plated at 5 × 104 cells/well in 6-well dishes. On day 1, fresh phospholipid liposomes were generated and added to cells at 400 μM (unless otherwise noted). On day 2, the medium was replaced with fresh medium containing 4 μg/ml Polybrene and virus was added to the wells. On day 5 the cells were fixed with 0.5% glutaraldehyde and stained for AP expression. For the 24-h infection time courses, a large batch of PS liposomes was produced on day 1, and was added to cells every 4 h from 0–24 h. At 24 h, cells were either infected as described above or were prepared for annexin-V labeling.
Annexin-V labeling
Alexa Fluor 488-conjugated annexin-V, propidium iodide (PI), and annexin binding buffer were obtained from the Vybrant Apoptosis Assay Kit #2 (Molecular Probes, Eugene, OR). Annexin-V labeling was performed using a slight variation of the manufacturer's protocol as previously described [28]. The geometric mean fluorescence of 10,000 cells was obtained for the unlabeled and labeled cell populations, and the mean of the unlabeled cells was subtracted from the mean of the labeled cells to determine the relative amount of cell-surface PS for each sample. Dead cells were excluded from analysis on the basis of PI staining.
Generation of liposomes
L-α-phosphatidyl-L-serine was obtained as a 10 mg/ml solution in chloroform:methanol (95:5) (Sigma, St Louis, MO). To generate liposomes, phospholipid was dried in a glass tube under nitrogen, and resuspended in PBS to a final concentration of 5 mM. This solution was sonicated on ice 3 times for 5 min each, using a W-385 sonicator with a microtip on output level 3 (Heat Systems Ultrasonics). The liposomes were filtered through a 0.2 μm pore-size syringe filter and were used immediately unless otherwise described.
Western blot analysis
For analysis of the HA-tagged human ASCT2, washed cells were lysed for 30 min at 4°C in lysis buffer (50 mM Tris-HCL [pH 8.0], 150 mM NaCl, and 1% NP-40), and centrifuged at 970 × g for 10 min to remove nuclei and cell debris. The supernatant was boiled for 10 min after addition of SDS and β-mercaptoethanol to final concentrations of 0.5% and 1%, respectively. The sample was divided, an equal amount of either PNGase F (New England Biolabs) or lysis buffer was added to each half, and the samples were kept at 37°C for 3 h. The treated and untreated samples were analyzed by electrophoresis in a 10% polyacrylamide gel containing 0.1% SDS. The proteins were transferred to nitrocellulose membranes, blocked in 5% powdered milk, incubated with appropriate concentrations of anti-HA primary and secondary antibodies, and visualized using a chemiluminescence kit (Amersham Biosciences). Analysis of ASCT1 was performed following transient transfection of 293T cells with a myc-tagged expression vector for murine ASCT1 [14] using the calcium phosphate method [29]. Cell lysates were collected at 48 h post-transfection and were treated as described above, followed by incubation of Western blots with appropriate concentrations of anti-Myc tag primary and secondary antibodies.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
DAC helped design the study, carried out the experiments, analyzed the data, and drafted the manuscript. ADM helped design the study and write the manuscript.
Acknowledgements
We thank David Kabat for providing the myc-tagged murine ASCT1 expression vector and Neal Van Hoeven for providing the HA-tagged human ASCT2 retroviral expression vector. This study was supported by grants HL54881, DK47754, and HL36444 from the NIH.
==== Refs
Overbaugh J Miller AD Eiden MV Receptors and entry cofactors for retroviruses include single and multiple transmembrane-spanning proteins as well as newly described glycosylphosphatidylinositol-anchored and secreted proteins Microbiol Mol Biol Rev 2001 65 371 389 11528001 10.1128/MMBR.65.3.371-389.2001
Chabot DJ Chen H Dimitrov DS Broder CC N-linked glycosylation of CXCR4 masks coreceptor function for CCR5-dependent human immunodeficiency virus type 1 isolates J Virol 2000 74 4404 4413 10756055 10.1128/JVI.74.9.4404-4413.2000
Eiden MV Farrell K Wilson CA Glycosylation-dependent inactivation of the ecotropic murine leukemia virus receptor J Virol 1994 68 626 631 8289366
Lavillette D Marin M Ruggieri A Mallet F Cosset FL Kabat D The envelope glycoprotein of human endogenous retrovirus type W uses a divergent family of amino acid transporters/cell surface receptors J Virol 2002 76 6442 6452 12050356 10.1128/JVI.76.13.6442-6452.2002
Marin M Tailor CS Nouri A Kabat D Sodium-dependent neutral amino acid transporter type 1 is an auxiliary receptor for baboon endogenous retrovirus J Virol 2000 74 8085 8093 10933718 10.1128/JVI.74.17.8085-8093.2000
Wentworth DE Holmes KV Molecular determinants of species specificity in the coronavirus receptor aminopeptidase N (CD13): influence of N-linked glycosylation J Virol 2001 75 9741 9752 11559807 10.1128/JVI.75.20.9741-9752.2001
Tailor CS Lavillette D Marin M Kabat D Cell surface receptors for gammaretroviruses Curr Top Microbiol Immunol 2003 281 29 106 12932075
Rasko JE Battini JL Gottschalk RJ Mazo I Miller AD The RD114/simian type D retrovirus receptor is a neutral amino acid transporter Proc Natl Acad Sci USA 1999 96 2129 2134 10051606 10.1073/pnas.96.5.2129
Tailor CS Nouri A Zhao Y Takeuchi Y Kabat D A sodium-dependent neutral-amino-acid transporter mediates infections of feline and baboon endogenous retroviruses and simian type D retroviruses J Virol 1999 73 4470 4474 10196349
Kubo Y Ishimoto A Ono T Yoshii H Tominaga C Mitani C Amanuma H Yamamoto N Determinant for the inhibition of ecotropic murine leukemia virus infection by N-linked glycosylation of the rat receptor Virology 2004 330 82 91 15527836 10.1016/j.virol.2004.09.011
Albritton LM Tseng L Scadden D Cunningham JM A putative murine ecotropic retrovirus receptor gene encodes a multiple membrane-spanning protein and confers susceptibility to virus infection Cell 1989 57 659 666 2541919 10.1016/0092-8674(89)90134-7
Wang H Paul R Burgeson RE Keene DR Kabat D Plasma membrane receptors for ecotropic murine retroviruses require a limiting accessory factor J Virol 1991 65 6468 6477 1942243
Kim JW Closs EI Albritton LM Cunningham JM Transport of cationic amino acids by the mouse ecotropic retrovirus receptor Nature 1991 352 725 728 1652100 10.1038/352725a0
Marin M Lavillette D Kelly SM Kabat D N-linked glycosylation and sequence changes in a critical negative control region of the ASCT1 and ASCT2 neutral amino acid transporters determine their retroviral receptor functions J Virol 2003 77 2936 2945 12584318 10.1128/JVI.77.5.2936-2945.2003
Coil DA Miller AD Enhancement of enveloped virus entry by phosphatidylserine J Virol 2005 79 11496 11500 16103200 10.1128/JVI.79.17.11496-11500.2005
Miller DG Edwards RH Miller AD Cloning of the cellular receptor for amphotropic murine retroviruses reveals homology to that for gibbon ape leukemia virus Proc Natl Acad Sci USA 1994 91 78 82 8278411
Davey RA Zuo Y Cunningham JM Identification of a receptor-binding pocket on the envelope protein of friend murine leukemia virus J Virol 1999 73 3758 3763 10196270
Battini JL Danos O Heard JM Receptor-binding domain of murine leukemia virus envelope glycoproteins J Virol 1995 69 713 719 7815534
Salaun C Gyan E Rodrigues P Heard JM Pit2 assemblies at the cell surface are modulated by extracellular inorganic phosphate concentration J Virol 2002 76 4304 4311 11932396 10.1128/JVI.76.9.4304-4311.2002
Valsesia-Wittmann S Morling FJ Hatziioannou T Russell SJ Cosset FL Receptor co-operation in retrovirus entry: recruitment of an auxiliary entry mechanism after retargeted binding. EMBO J 1997 16 1214 1223 9135138 10.1093/emboj/16.6.1214
Damico R Bates P Soluble receptor-induced retroviral infection of receptor-deficient cells J Virol 2000 74 6469 6475 10864659 10.1128/JVI.74.14.6469-6475.2000
Farge E Increased vesicle endocytosis due to an increase in the plasma membrane phosphatidylserine concentration Biophys J 1995 69 2501 2506 8599656
Farge E Ojcius DM Subtil A Dautry-Varsat A Enhancement of endocytosis due to aminophospholipid transport across the plasma membrane of living cells Am J Physiol 1999 276 C725 33 10070001
Wei CM Gibson M Spear PG Scolnick EM Construction and isolation of a transmissible retrovirus containing the src gene of Harvey murine sarcoma virus and the thymidine kinase gene of herpes simplex virus type 1 J Virol 1981 39 935 944 6270359
DuBridge RB Tang P Hsia HC Leong PM Miller JH Calos MP Analysis of mutation in human cells by using an Epstein-Barr virus shuttle system Mol Cell Biol 1987 7 379 387 3031469
Miller AD Rosman GJ Improved retroviral vectors for gene transfer and expression Biotechniques 1989 7 980 990 2631796
Cosset FL Takeuchi Y Battini JL Weiss RA Collins MK High-titer packaging cells producing recombinant retroviruses resistant to human serum J Virol 1995 69 7430 7436 7494248
Coil DA Miller AD Phosphatidylserine is not the cell surface receptor for vesicular stomatitis virus J Virol 2004 78 10920 10926 15452212 10.1128/JVI.78.20.10920-10926.2004
Corsaro CM Pearson ML Enhancing the efficiency of DNA-mediated gene transfer in mammalian cells Somatic Cell Genet 1981 7 603 616 7292260 10.1007/BF01549662
|
16091143
|
PMC1201173
|
CC BY
|
2021-01-04 16:36:39
|
no
|
Retrovirology. 2005 Aug 9; 2:49
|
utf-8
|
Retrovirology
| 2,005 |
10.1186/1742-4690-2-49
|
oa_comm
|
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-951610517010.1186/1465-9921-6-95ResearchIncidence of asthma and mortality in a cohort of young adults: a 7-year prospective study de Marco Roberto [email protected] Francesca [email protected] Lucia [email protected] Massimilian [email protected] Aurelia [email protected] Alessandra [email protected] Unit of Epidemiology and Medical Statistics, University of Verona, Verona, Italy2 National Health Service, CPA-ASL 4 Unit of Respiratory Medicine, Turin, Italy3 Department of Applied Health Sciences, University of Pavia, Pavia, Italy2005 16 8 2005 6 1 95 95 28 6 2005 16 8 2005 Copyright © 2005 Roberto et al; licensee BioMed Central Ltd.2005Roberto 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
Few longitudinal data exist on the incidence of asthma in young adults and on the overall mortality risk due to asthma. A 7-year follow-up prospective study was performed to assess the incidence of asthma and mortality from all causes in a cohort of young adults.
Methods
The life status of a cohort of 6031 subjects, aged 20–44 years, who replied to a respiratory screening questionnaire between 1991 and 1992, was ascertained in 1999. A new questionnaire investigating the history of asthma was subsequently sent to the 5236 subjects who were still alive and residents in the areas of the study. 3880 subjects (74%) replied to the second questionnaire.
Results
The incidence of adult-onset asthma was 15.3/10,000/year (95%CI:11.2–20.8). The presence of asthma-like symptoms (IRR:4.17; 95%CI:2.20–7.87) and allergic rhinitis (IRR:3.30; 95%CI:1.71–6.36) at baseline were independent predictors of the onset of asthma, which was more frequent in women (IRR:2.32; 95%CI:1.16–4.67) and increased in the younger generations.
The subjects who reported asthma attacks or nocturnal asthma symptoms at baseline had an excess mortality risk from all causes (SMR = 2.05; 95%CI:1.06–3.58) in the subsequent seven years. The excess mortality was mainly due to causes not related to respiratory diseases.
Conclusion
Asthma occurrence is a relevant public health problem even in young adults. The likelihood of developing adult onset asthma is significantly higher in people suffering from allergic rhinitis, in women and in more recent generations. The presence of asthma attacks and nocturnal symptoms seems to be associated with a potential excess risk of all causes mortality.
==== Body
Introduction
Asthma is usually considered to be a chronic disease that starts in childhood, characterised by a low specific mortality risk, which can be completely avoidable with adequate management. However, recent epidemiological studies have pointed out [1] that a non negligible number of the current asthmatics in the general population developed the disease in adulthood and that adult-onset asthma has a worse prognosis than childhood asthma.
Unfortunately, few longitudinal data exist on the incidence of asthma in adults [2-4]. Consequently, knowledge of the natural history and the risk factors for adult asthma is limited and relies almost completely on prevalence data, which depend on both the incidence and the persistence of the disease. There are even fewer longitudinal studies investigating the outcome of adult asthma in terms of overall mortality [5-9], which suggest that asthma and asthma symptoms are associated with a significantly higher all-causes mortality compared to the general population.
In this paper we describe the incidence of asthma and all-causes mortality in a large, representative sample of young Italian adults, who participated in the European Respiratory Health Survey (ECRHS) stage I in 1991, and who were followed up in 1999/2000. In particular, our objective was to assess the incidence of adult onset asthma and to test : 1) whether a previous history of allergic rhinitis was a predictor for the onset of asthma, after adjusting for known risk factors; 2) whether a previous history of asthma attacks or nocturnal asthma-like symptoms was associated with a subsequent overall mortality excess in young adults.
Methods
i) Study design
A repeated survey of those who replied to the screening questionnaire [10] in the frame of the Italian branch of the ECRHS-stage1 in 1991/1992, was performed in three Italian centres (Verona, Pavia and Turin) during 1999/2000. The initial cohort was made up of 6031 subjects, randomly chosen from the general population. The mean age of the cohort in 1991/1992 was 32.7 years (range 20–44 years) and the percentage of women was 49.6%. All subjects were mailed a new questionnaire up to two times, in case of non response. The questionnaire administered in 1999/2000 included the same standard questions used in the first survey (enquiring about asthma attacks, wheezing, nocturnal dyspnoea, nocturnal tightness, allergic rhinitis in the last 12 months and current use of asthma drugs) with additional questions on the history of asthma (doctor diagnosis, age of the first/ last attack), the history of exposure to active smoking and social class. The questions used in the first and second surveys have been published elsewhere [11].
After the second postal wave, the life and residence status of all non responders to the mail survey were obtained from vital statistics records. Then all subjects who had neither moved nor died before or during the survey were sent a third postal wave and were finally interviewed by phone.
At the end of follow-up, 748 (12.4%) subjects had moved and 47 (0.8%) died before or during the study; 352 (5.8%) could not be traced through anagraphic records, 1004 (16.6%) did not reply or explicitly refused to answer the questionnaire and 3880 (64.3%) replied.
The protocol of the study was approved by the Italian Ethical Committees of the participating centres, and the individual information was collected in compliance with the Italian law (n°675/1996) concerning the protection of the privacy of individual health data.
ii) Incidence analysis
5236 (86.4%) subjects from the initial cohort were considered eligible for this analysis (anyone who had moved or died was excluded). Of these, 3880 replied to the second questionnaire in 1999/2000 (response rate 74%); 24 subjects were excluded because of the mismatching of age and/or sex. The population at risk for the incidence analysis included all the asthma-free subjects in the first study (1991/92). Accordingly, 302 out of 3856 valid respondents were excluded because they were considered asthmatics at baseline: that is, people who reported having current asthma in the first study (having had an attack of asthma in the last 12 months or currently taking any medicine for asthma) as well as those who, in the second study, reported having had the first asthma attack more than 3 years before the first survey. This was done considering that a discrepancy of less than three years could be due to recall bias when reporting the age of onset (Pattaro C. Long-term repeatability of a questionnaire for lifelong asthma assessment.Personal communication). New cases of asthma were considered those, among the population at risk, who gave a positive answer to the question "Have you ever had asthma?" in the second survey.
Incidence rates were computed by dividing the number of new cases by the total number of person-years. PY was computed as the time between the first and second survey for non asthmatics and the time between the first interview and the first asthma attack for new asthmatics. For 6 subjects, who reported in the second survey having had the first attack of asthma less than 3 years before the first survey, the time at risk was set at 1 day [4].
Subjects, reporting in the first study, having had 'wheezing or whistling in their chest at any time' and/or having 'woken up with a feeling of tightness in the chest' and/or having 'woken up with an attack of shortness of breath' in the past 12 months, were considered as having asthma symptoms at baseline.
All subjects who reported never having smoked in the second questionnaire were considered non smokers at baseline. Subjects who reported having smoked at one time in the second questionnaire were classified as smokers or ex-smokers at baseline, according to their answers to the following questions: 'How old were you when you started smoking?' and 'How old were you when you stopped smoking?'.
According to postal codes, each subject was considered a resident either in urban or suburban areas. Areas were classified as suburban when the inhabitants of the municipal district were less than 40,000, urban otherwise. Participants were divided into two social classes based on their profession: low (blue collars/unemployed/retired) and medium/high (manager/white collars/teachers...).
The Poisson regression model [12] was used to assess the association of baseline symptoms and individual characteristics with the incidence of asthma.
Statistical analysis was performed using STATA software [13], release 7.0 (Stata Corp 1997. Stata Statistical Software: release 5.0. Stata Corporation, College Station, TX).
iii) Mortality analysis
All subjects were considered eligible for this analysis. Mortality rates were computed by dividing the number of deaths during the follow-up by the person-years (PY) at risk [12]. PY were computed as: i) the time between the first and the second survey (follow-up time) for responders to the questionnaire; ii) the average follow-up time for non responders; iii) the time between the first survey and the date of death for deceased people; iv) half follow-up time for subjects who had moved before the second survey or were untraceable (censored alive at mid follow-up). Mortality rates were estimated according to symptoms reported at baseline. Standardised Mortality Ratios (SMR adjusted by age, sex and centre) were used to compare the mortality rates in subjects with and without symptoms at baseline.
The specific cause of death was ascertained using the mortality records of the health districts of the three centres involved in the survey; the underlying cause of death was coded according to the International Classification of Diseases, Ninth Revision. The specific cause of death was not found for one subject who had died outside the health district. As information on smoking habits was missing in the first ECRHS questionnaire, it was not possible to adjust the SMR for this variable. However, to verify the stability of our results, a sensitivity analysis was performed on all the subjects who replied to the second questionnaire (3856) and who had the information on smoking habits. In this subgroup, SMRs adjusted for smoking habits were computed assuming that the 47 deaths were 1) all non smokers; 2) all smokers; 3) had the same smoking distribution of the respondents (random assignment).
Results
Incidence
Subjects who were included in the incidence analysis were found to be slightly older (33.1 vs 32.5 years, p < 0.05) and there was a greater percentage of women (51.2% vs 47.8%, p < 0.005) than non responders to the second questionnaire; however, there was no statistically significant difference in symptoms at baseline (first study) between responders and non responders. The average time of follow-up for the incidence study was 7.72 years. Forty-one new cases of asthma out of 3554 people at risk occurred during the period between the two studies (Table 1), and the average annual incidence rate was 15.2/10000/year (95%CI: 11.2–20.7), ranging from 10.1/10000/year in the older generation to 22.8/10000/year in the younger generation. The incidence rate for asthma was greater in women (Figure 1) than men, higher in the younger groups and in urban than suburban areas, lower in active smokers than non smokers and peaked in subjects who reported asthma-like symptoms (wheezing and/or nocturnal dyspnoea and/or nocturnal tightness in the chest) and allergic rhinitis at baseline.
Table 1 Number of subjects at risk at start of follow-up (1991/92), number of new cases of asthma, person-years, crude incidence (95% C.I.) during the period 1991–2000 according to sex and birth cohort.
N° of subjects at risk N° of new cases (1991–2000) Person-years Incidence*10,000
Total 3554 41 26881 15.25 (11.23–20.71)
Sex
Men 1720 12 13040 9.20 (5.23–16.20)
Women 1834 29 13841 20.95 (14.56–30.15)
Birth Cohort
1946–1951 793 6 5935 10.11 (4.54–22.50)
1951–1956 738 6 5634 10.65 (4.78–23.71)
1956–1961 710 9 5379 16.73 (8.70–32.15)
1961–1966 732 10 5543 18.04 (9.71–33.52)
1966–1971 581 10 4389 22.78 (12.26–42.34)
Figure 1 Average annual incidence of asthma (*10,000/year) and 95% confidence interval during the period 1991–2000, according to baseline characteristics (reported in the first study:1991/1992).
After adjusting for all these factors, the incidence of asthma (Table 2) was significantly and independently associated with sex (Incidence rate ratio, IRR = 2.32, 95%CI:1.16–4.67), the presence of asthma-like symptoms (IRR = 4.17, 95%CI: 2.20–7.87) and allergic rhinitis (IRR = 3.30, 95%CI: 1.71–6.36); it was also significantly lower in smokers than non-smokers (IRR = 0.33; 95%CI:0.15–0.73). The incidence of asthma showed a significantly increasing trend (p < 0.01) according to the birth cohort.
Table 2 Mutually Adjusted Incidence Rate Ratios (95% C.I.) for the association of the incidence of asthma with the main baseline characteristics (1991/1992).
Relative risk (IRR) 95% C.I. p-value
Sex
Men 1
Women 2.32 1.16–4.67 0.017
Birth cohort *
1946–1951 1
1951–1956 1.12 0.36–3.49 0.84
1956–1961 1.69 0.60–4.77 0.32
1961–1976 1.68 0.60–4.69 0.32
1966–1971 2.28 0.82–6.34 0.11
Urban area
Suburban area 1
Urban area 2.52 0.77–8.19 0.12
Asthma-like symptoms**
No 1
Yes 4.17 2.20–7.87 <0.001
Allergic rhinitis
No 1
Yes 3.30 1.71–6.36 <0.001
Smoking habits
Non smokers 1
Ex-smokers 0.49 0.15–1.65 0.25
Smokers 0.33 0.15–0.73 <0.01
Social class
Medium /High 1
Low 1.65 0.77–3.52 0.20
* test for trend p ≤ 0.01
** wheezing and/or nocturnal shortness of breath and/or nocturnal tightness in the chest
Mortality
Average follow-up time for the mortality study was 7.05 years. Forty-seven subjects from the initial cohort died between 1991 and 2000. The average annual mortality rate was 11.0/10000/year (95%CI: 8.3–14.7), consistent with the official mortality data for the same age groups and the same areas (Table 3). Mortality rates were particularly high (Table 4) for people reporting having had at least one attack of asthma (32.2/10000/year), nocturnal dyspnoea (30.3/10000/year) and nocturnal tightness (23.5/10000/year). Subjects who reported no respiratory symptoms at baseline had very similar mortality rates to those who reported wheezing, allergic rhinitis or nocturnal cough.
Table 3 Number of deaths, person-years, average annual mortality rates (95% Confidence Interval) for the ECRHS-Italy cohort (1991–2000),and official mortality rates in northern Italy.
Number of deaths Person-years Mortality rate *10,000 (95% C.I.) Northern Italy mortality rates *10,000 $
Total 47 42523 11.0 (8.3–14.7) 10.9
Sex
Men 34 21592 15.7 (11.2–22.0) 14.7
Women 13 20931 6.2 (3.6–10.7) 6.9
$ ISTAT (National Institute of statistics) 1998
Table 4 Number of subjects who reported a specific symptom in the first survey (1991/1992), number of deaths observed during the follow-up (1991–2000), crude annual mortality rate (*10,000/year), and its 95% Confidence Interval.
Number of subjects at risk Deaths during 1991–2000 Annual mortality rate*10,000 95% Confidence Interval
No respiratory symptoms 3225 26 11.4 7.7–16.7
Wheezing 600 4 9.4 3.5–25.1
Allergic rhinitis 959 7 10.5 5.0–22.1
Nocturnal Cough 1771 10 8.0 4.3–14.8
Nocturnal Tightness 1771 8 23.5 11.7–46.9
Nocturnal Dyspnoea 429 9 30.3 15.8–58.3
Asthma attacks 221 5 32.2 13.4–77.3
The reporting of asthma attacks and/or nocturnal dyspnoea and/or nocturnal tightness in the first survey (1991/92) was associated with a two-fold increase in subsequent overall mortality (SMR = 2.05; 95%CI:1.06–3.58) compared to the rest of the population, after adjusting for age, sex and centre (Table 5). The increase in overall mortality for this group of subjects was mainly due to a statistically significant increase in mortality from accidents and a moderate increase in mortality from cancer and cardiovascular diseases. No deaths from respiratory diseases were observed in our cohort, while 4 out of 16 deaths from tumours were due to lung cancer (1 in the symptomatic group and 3 in the control group). When the one death with unknown cause was attributed to each specific cause of death, the interpretation of the results was the same (see last column of Table 5).
Table 5 Number of deaths (1991–2000), average annual mortality rates for overall and specific causes of death in subjects with or without asthma attacks/nocturnal dyspnoea/nocturnal tightness (A/D/T) in the first survey (1991/1992) and Standardized Mortality Ratios adjusted for sex, age and centre.
Subjects without A/D/T
(n = 5169) Subjects with A/D/T
(n = 862)
Number of
deaths Crude mortality rate
*10,000 Number of
deaths Crude mortality rate
*10,000 SMR
(95% C.I.) $ Simulated SMR
(95% C.I.)$$
All causes 35 9.58 12 19.96 2.05 (1.06–3.58) 2.05 (1.06–3.58)
All tumours (140–239) 12 3.29 4 6.65 2.09 (0.57–5.36) 1.80 (0.49–4.61)
- Lung cancer (160–165) 3 0.82 1 1.66 2.01 (0.05–11.18) 1.53 (0.04–6.52)
Cardiovascular diseases (390–459) 4 1.10 1 1.66 1.34 (0.03–7.47) 1.11 (0.03–6.18)
Accidents (800–999) 4 1.10 4 6.65 6.95 (1.89–17.80) 5.47 (1.49–14.01)
Other causes 14 3.83 3 4.99 1.29 (0.27–3.78) 1.21 (0.25–3.54)
Unknown 1 0.27 0 0.0 0.00
$ with respect to subjects without A/D/T
$$ SMR was computed attributing the one death with unknown cause to each of the specific causes of death
The sensitivity analysis confirmed the stability of the previous results. In fact, in the subgroup of subjects with smoking information, the SMR not adjusted for smoking was 1.99 (95CI:1.03–3.48) while those adjusted for smoking were: 1.91 assuming that all the deaths were smokers; 2.02 assuming that all the deaths were non smokers and 1.94 when the same distribution of smoking among the responders was assumed.
The mortality rates for subjects with asthma attacks/nocturnal symptoms and reporting to be or not to be under treatment were 14.4/10000/year (95% C.I.: 2.0–102.3) and 20.9/10000/year (95% C.I.: 11.6–37.8), respectively, and were not statistically different (p = 0.72).
Discussion
Our results provide information about incidence of adult asthma and all causes mortality risk related to asthma in northern Italy, an area with a relatively low prevalence of asthma [14]. The main findings of our longitudinal analysis were as follows:
i) The average incidence of adult onset-asthma in northern Italy is 15.2/10,000/year, which ranks in the lower part of the ranges reported in longitudinal studies in other countries. Allergic rhinitis and asthma-like symptoms are strong and independent predictors of adult-onset asthma, which occurs more frequently in women and in recent generations.
ii) The presence of asthma attacks and nocturnal asthma symptoms (tightness and dyspnoea) in young adults is associated with a two-fold increase in the risk of dying in the subsequent seven years, compared to asymptomatic subjects.
Incidence
In our analysis of the incidence of asthma, the population at risk and the new cases of the disease were defined according to the occurrence or absence of a relevant clinical event, such as an asthma attack, which is less likely to be influenced by recall bias and more reproducible than the reporting of asthma symptoms [15,16]. It has also been reported that, at least in Italy, the large majority (85%) of subjects who self-reported an attack of asthma had been diagnosed by a doctor [1]. Consequently, it is likely that our definition slightly underestimates the incidence of the disease compared to symptom-based definition [17] on one hand; but then on the other, it guarantees better specificity [18] and stability of our estimates than other definitions.
Our estimate of the average annual incidence of asthma in young adults in Italy was 15.2/10,000/year, ranging from 10.1/10,000/year, in the older generations, to 22.8/10,000/year in the younger generations. Our longitudinal study confirms that adult asthma occurs more frequently in women and in younger generations. The gender difference in the occurrence of asthma has been pointed out in several previous studies [17,19,20] from different countries and it has been attributed mainly to the role of female sex hormones [21]. The reasons for the generalized increase in asthma incidence in the youngest generations are still unknown. Changes in susceptibility to environmental stimuli leading to asthma (hygiene hypothesis) as well as changes in exposure to environmental factors (increase in air pollution) could have contributed to this increase [22,23]. Our finding that incidence of asthma is higher in urban than suburban areas support both the previous hypotheses.
The strongest independent predictor of the incidence of asthma was the presence of asthma-like symptoms, such as diurnal wheezing and nocturnal dyspnoea and tightness, in agreement with a recent longitudinal Spanish study [4]. This finding indicates that there is a group of subjects for which the clinical phase of the disease starts with mild symptoms which are probably not recognized as asthma, but which will be diagnosed in the following years when the disease gets worse and the exacerbations are labelled as asthma attacks in the end.
The presence of allergic rhinitis at baseline, regardless of other asthma symptoms, was another strong predictor of asthma. This result concords with other longitudinal studies showing that allergic rhinitis is an independent risk factor for the onset of asthma [24,2].
The association between allergic rhinitis and asthma has been traditionally interpreted as the progression of a common airway disease associated with atopy [25,26]. It has also been reported [24,27,28] that rhinitis is a significant risk factor for adult asthma in both atopic and non atopic subjects, suggesting that other mechanisms, besides atopy, should be invoked to explain the observed association [29,30]. Our study, based on questionnaire data and not on a measurement of atopy, cannot contribute to highlight this hypotheses. However, the strong association between early allergic rhinitis and onset of asthma in adulthood suggests that earlier treatment of allergic rhinitis symptoms [31,32] could modify the natural evolution of asthma and prevent the development of a more severe disease. Indeed, interventions such immunotherapy [33], pharmacologic therapy [34] and allergens avoidance [35] for allergic rhinitis seem to be effective in preventing complications and the development of asthma [29-31].
In agreement with other longitudinal studies [36,37], we found that active smoking does not increase the risk of asthma in adults, probably because susceptible people simply do not start smoking or quit smoking very quickly. This finding is in contrast with another longitudinal study on adolescents [38], where active smoking was a risk factor for the incidence of asthma. In any case, whether active smoking is a risk factor for the incidence of asthma or whether it is a selection factor (healthy smoker effect) continues to be an open problem.
Finally, in contrast to a recent paper [39] that found an association between asthma prevalence and socioeconomic status, our results do not support that being in a low social class increases the risk of developing asthma. This suggest that previously reported associations, based on prevalence studies, reflect more the association of social class with the severity and/or the persistence of the disease rather than its association with the occurrence of asthma.
Mortality
To our knowledge, few studies have been published on the overall mortality risks of adults with asthma or asthma-like symptoms. This is one of the few, that prospectively investigated asthma related mortality risk in a large cohort of young adults. We ascertained life status for all subjects still resident (82%) in the same areas where they lived in 1991, while we assumed that people who were untraceable or who had moved before the second survey were censored alive halfway through the follow-up. As there are no reasons to expect that censoring could be related to the outcome or presence of asthma symptoms, our estimates of the mortality risk are not biased by potential selection in the initial cohort.
The presence of asthma attacks and/or nocturnal asthma symptoms was associated with an increased mortality risk from all causes (SMR = 2.05; 95% CI:1.06–3.58). Our estimate of the relative mortality risk for asthmatics agrees with those previously reported, which ranged from 1.5 [9] to 2.4 [5].
Although the excess death due to chronic obstructive diseases accounted for a substantial part of the mortality risk in other studies [9,5,6], in our study it was mainly due to non respiratory causes of death, such as accidents, and to a lesser extent, to cardiovascular diseases and cancer, which concords with a French study [8]. This discrepancy can be explained by the fact that our cohort included much younger individuals than previous studies, and it is well known that young people < 45 years are at a much lesser risk of dying from chronic diseases than older people.
Death from accidents is the leading cause of death in the general population aged 15–44 years, at least in Italy. Our results showed that asthmatics belonging to this age group had a 5- to 6-fold risk of dying from accidents than the general population; this suggests that a previous history of asthma attacks and nocturnal symptoms made them even more susceptible to this kind of hazard. This could be due to the fact that asthmatic symptoms, especially nocturnal ones, may interfere with diurnal attentiveness, which could cause fatal occupational accidents [40] as well as car accidents [41]. Furthermore, it has been reported that asthma patients have high rates of anxiety disorders and major depression [42,43], which can also make them more prone to accidents. Another aspect to take into consideration is the fact that many asthmatic subjects also take antihistamines for their symptoms [45], which could have a sedative effect [46] and alter their diurnal concentration. However, in our study, no association was found between the use of drugs reported in the first study and subsequent mortality.
The moderate increase in mortality from cardiovascular diseases and cancer has already been reported [44,9] and could reflect the greater susceptibility of asthmatics to traditional risk factors, which could be fully appreciated in cohorts older than ours. No association was found between wheezing at baseline and mortality likely because this symptom alone has either a low specificity for asthma [18] or identifies a very mild form of the disease. Furthermore, nocturnal asthma-like symptoms or asthma attacks could be partially attributable to other diseases like psychiatric illnesses and heart diseases.
As the screening questionnaire used in the first ECRHS study did not include information on smoking habits, it was impossible to adjust for this factor in the mortality analysis. However, the sensitivity analysis carried out on a sizable sample showed that when smoking information was taken into account, the results were very similar, suggesting that the failure to adjust for this potential confounder could have biased our results only to a minor extent.
Nevertheless, some caveat in the interpretation of our mortality results should be taken into account due to the relatively short follow-up period (7 years) and the consequent low number of deaths. As a consequence, our analysis cannot establish a definite causal relationship between the presence of severe asthma symptoms and the subsequent overall mortality due to both the observational nature of our survey and the expected low number of deaths. The absence of a specific asthma related risk of death in our young cohort does not weaken our result: in fact, mortality from asthma is completely avoidable in this age group, and no deaths are expected to be found if health services are adequate [47]. Consequently, the suggestion emerging from our study of an increased non specific mortality risk even in young subjects reporting severe asthma symptoms should be considered as a warning signal, indicating that asthma symptoms may involve a more general risk than normally expected, and should at least promote other studies dealing with the overall mortality pattern in asthmatics.
Conclusion
The incidence of asthma is a relevant public health problem even in young adults. The likelihood of developing adult onset asthma is significantly higher in people suffering from allergic rhinitis and mild asthma-like symptoms, in women and in more recent generations. Furthermore, our results suggest that the presence of asthma attacks and nocturnal symptoms may be associated with an excess risk of all causes mortality. Greater medical attention should be paid to early asthma-like symptoms (particularly nocturnal ones) and allergic rhinitis.
List of Abbreviations
ECRHS: European Community Respiratory Health Survey
PY: Person-years
SMR: Standardised Mortality Ratio
IRR: Incidence Rate Ratio
Competing interests
RdeM has received a reimbursement for travel expenses to the ERS congress by GlaxoSmithKline Italia in 2003 and 2004.
Authors' contributions
RdM developed the idea for the study and was responsible for the study design.
FL was responsible for the data management and analysis.
All the authors were responsible for the data collection in local centres and participated in the interpretation and presentation of the results.
All authors read and approved the final manuscript.
Acknowledgements
All authors received funding from the Italian Ministry of University and Scientific Research.
RdM, FL, LC were supported by the University of Verona and the Veneto Region;
AM was supported by the National Health Service, AUSL Pavia;
MB and AC were supported by the National Health Service, ASL 4 Turin
All the funds were used during the collection of data. All the other phases, i.e. the study design, the analysis and interpretation of data, the preparation of the manuscript for the submission were completely independent of these funds.
==== Refs
de Marco R Locatelli F Cerveri I Bugiani M Marinoni A Giammanco G Incidence and remission of asthma: a retrospective study on the natural history of asthma J Allergy Clin Immunol 2002 110 228 235 12170262 10.1067/mai.2002.125600
Huovinen F Kaprio J Laitinen LA Koskenvuo M The incidence and prevalence of asthma among adult Finnish men and women of the Finnish Twin Cohort from 1975 to and their relation to hay fever and chronic bronchitis Chest 1990 115 928 36 10208188 10.1378/chest.115.4.928
Toren K Hermansson BA Incidence rates of adult-onset asthma in relation to age, sex atopy and smoking: a Swedish population based study of 15813 adults Int J Tuberc Lung Dis 1999 3 192 97 10094318
Basagana X Sunyer J Zock JP Kogevinas M Urrutias I Maldonado JA Almar E Payo F Anto JM Incidence of asthma and its determinants among adults in Spain Am J Respir Crit Care Med 2001 164 1133 1137 11673198
Ulrik CS Frederiksen J Mortality and markers of risk of asthma death among 1075 outpatient with asthma Chest 1995 108 10 15 7606941
Lange P Ulrik CS Vestbo J Mortality in adults with self reported asthma. Copenaghen City Heart Study Group Lancet 1996 347 1285 9 8622503
Connolly CK Mamun M Alcock SM Prescott RJ The Darlington and Northallerton Prospective asthma study: best function predicts mortality during the first 10 years Respir Med 1998 92 1274 80 9926140 10.1016/S0954-6111(98)90228-1
Dantzer C Tessier JF Nejjari C Barberger Gateau P Dartigues JF Mortality of elderly subjects with self-reported asthma in a French cohort, 1991–1996 Eur J Epidemiol 2001 17 57 63 11523577 10.1023/A:1010996718008
Huovinen E Kaprio J Vesterinen E Koskenvuo M Mortality of adults with asthma: a prospective cohort study Thorax 1997 52 49 54 9039240
de Marco R Verlato G Zanolin E Bugiani M Drane JW Nonresponse bias in EC Respiratory Health Survey in Italy Eur Respir J 1994 7 2139 4511 7713194 10.1183/09031936.94.07122139
de Marco R Zanolin ME Accordini S Signorelli D Marinoni A Bugiani M Lo Cascio V Woods R Burney P A new questionnaire for the repeat of the first stage of the European Respiratory Health Survey: a pilot study Eur Respir J 1999 14 1044 8 10596688 10.1183/09031936.99.14510449
Breslow NE Day NE Statistical methods in Cancer research: The design and analysis of Cohort Studies 1980 2 Lyon France: IARC Scientific
Stata Corp Stata Statistical Software release 50 1997 Stata Corporation, College Station, TX
European Community Respiratory Health Survey Variations in the prevalence of respiratory symptoms, self reported attack of asthma, and use of asthma medication in the European Community Respiratory Health Survey Eur Respir J 1996 9 687 695 8726932 10.1183/09031936.96.09040687
Burney PG Laitinen LA Perdrizet S Huckauf H Tattersfield AE Chinn S Poisson N Heeren A Britton JR Jones T Validity and repeatability of the IATLD bronchial symptom questionnaire: an international comparison Eur Respir J 1989 2 940 945 2606194
de Marco R Campello C Basso O Moi P Verlato E Zanolin ME Agreement between self-reported symptoms and clinical interview in the European respiratory Health Survey in Italy [abstract] Eur Resp J 1993 6 s499
Strachan DP Butland BK Anderson HR Incidence and prognosis of asthma and wheezing illness from early childhood to age 33 in a national British cohort BMJ 1996 312 1195 9 8634562
de Marco R Cerveri I Bugiani M Ferrari M Verlato G An undetected burden of asthma in Italy: the relationship between clinical and epidemiological diagnosis of asthma Eur Respir J 1998 11 599 605 9596109
de Marco R Locatelli F Sunyer J Burney P Differences in incidence of reported asthma related to age in men and women. A retrospective analysis of the data of the European Respiratory Health Survey Am J Respir Crit Care Med 2000 162 68 74 10903222
Thomsen SF Ulrik CS Kyvik OK Larsen K Skadhauge LR Steffensen I Backer V The incidence of asthma in young adults Chest 2005 127 1928 1934 15947304 10.1378/chest.127.6.1928
Svanes C Real FG Gislason T Jansson C Jogi R Norrman E Nystrom L Toren K Omenaas E Association of asthma and hay fever with irrefular menstruation Thorax 2005 60 445 450 15923242 10.1136/thx.2004.032615
Sunyer J Anto JM Tobias A Burney P for ECRHS Generational increase of self reported first attack of asthma in fifteen industrialized countries Eur Respir J 1999 14 885 891 10573237 10.1034/j.1399-3003.1999.14d26.x
Braun-Fahrlander C Environmental exposure to endotoxin and other microbial products and the decrease risk of childhood atopy: evaluating developments since april 2002 Curr Opin Allergy Clin Immunol 2003 3 325 9 14501429 10.1097/00130832-200310000-00001
Guerra S Sherril DL Martinez Fd Barbee RA Rhinitis as an independent risk factor for adult-onset of asthma J Allergy Clin Immunol 2002 109 419 425 11897985 10.1067/mai.2002.121701
Sporik R Holgate ST Platts-Mill TA Cogswell JJ Exposure to house dust mite allergen (Der p I) and the development of asthma in childhood New Engl J Med 1990 323 502 507 2377175
Litonjua A Sparrow D Weiss ST O'Connor GT Long AA Ohman JL Sensitization to cat allergen is associated with asthma in older man and predicts new onset airway hyperresponsiveness Am J Respir Crit Care Med 1997 156 23 27 9230721
Rusconi F Galassi C Corbo GM Forastiere F Biggeri A Ciccone G Renzoni E Risk factors for early, persistent and late onset wheezing in young children. SIDRIA Collaborative Group Am J Respir Crit Care Med 1999 160 1617 22 10556130
Leynaert B Bousquet J Neukirch C Liard R Neukirch F Perennial rhinitis: an independent risk factors for asthma in nonatopic subjects: results from the European Respiratory Health Survey J Allergy Clin Immunol 1999 104 301 4 10452748
Vignola AM Chanez P Godard P Bousquet J Relationship between rhinitis and asthma Allergy 1998 53 833 9 9788683
Bousquet J Allergic Rhinitis and Its Impact on Asthma J Allergy Clin Immunol 2001 108 Suppl 5 147 334 10.1067/mai.2001.118891
Passalacqua G Ciprandi G Canonica GW United airways disease: therapeutic aspects Thorax 2000 55 Suppl 2 26 27 10.1136/thorax.55.suppl_2.S26
Lack G Pediatric allergic rhinitis and comorbid disorders J Allergy Clin Immunol 2001 108 Suppl 1 9 15 10.1067/mai.2001.115562
Moller C Dreborg S Ferdousi HA Halken S Host A Jacobsen L Koivikko A Koller DY Niggemann B Norberg LA Urbanek R Valovirta E Wahn U Pollen immunotherapy reduces the development of asthma in children with seasonal rhinocongiuntivitis (the PAT-study) J Allergy Clin Immunol 2002 109 251 6 11842293 10.1067/mai.2002.121317
Beckman DB Grammer LC Pharmacotherapy to prevent the complications of allergic rhinitis Allergy Asthma Proc 1999 20 215 23 10476319
Skoner DP Complications of allergic rhinitis J Allergy Clin Immunol 2000 105 Suppl 6 605 9 10.1067/mai.2000.106150
Rasmussen F Siersted HC Lambrechtsen J Hansen HS Hansen NG Impact of airway lability, atopy, and tobacco smoking on the development of asthma-like symptoms in asymptomatic teenagers. The Odense Schoolchild Study Chest 2000 117 1330 35 10807819 10.1378/chest.117.5.1330
Vesterinen E Kaprio J Koskenvuo M Prospective study of asthma in relation to smoking habits among 14729 adults Thorax 1988 43 534 539 3269222
Larsson L Incidence of asthma in Swedish teenagers: relation to sex and smoking habits Thorax 1995 50 260 4 7660339
Basagana X Sunyer J Kogevinas M Zock JP Duran-Tauleria E Jarvis D Burney P Anto JM European Community Respiratory Health Survey Socioeconomic status and asthma prevalence in young adults: the European Community Respiratory Health Survey Am J Epidemiol 2004 160 178 88 15234940 10.1093/aje/kwh186
Akerstedt T Fredlund P Gillberg M Jansson B A prospective study of fatal occupational accidents-relationship to sleeping difficulties and occupational factors J Sleep Res 2002 11 69 71 11869429 10.1046/j.1365-2869.2002.00287.x
Barbe Pericas J Munoz A Findley L Anto JM Agusti AG Automobile accidents in patients with sleep apnea syndrome. An epidemiological and mechanistic study Am J Respir Crit Care Med 1998 158 18 22 9655701
Brown ES Khan DA Mahadi S Psychiatric diagnoses in inner city outpatients with moderate to severe asthma Int J Psychiatry Med 2000 30 319 327 11308036
Nascimento I Nardi AE Valenca AM Lopes FL Mezzasalma Ma Nascentes R Zin WA Psychiatric disorders in asthmatic outpatients Psychiatry Res 2002 110 73 80 12007595 10.1016/S0165-1781(02)00029-X
Rosero SZ Zareba W Moss AJ Robinson JL Hajj Ali RH Locati EH Benhorin J Andrews ML Asthma and the risk of cardiac events in the long QT syndrome. Long QT Syndrome Investigating Group Am J Cardiol 1999 84 1406 11 10606113 10.1016/S0002-9149(99)00586-X
Wilson AM Are antihistamines useful in managing asthma? Curr Opin Allergy Clin Immunol 2002 2 53 920 11964751 10.1097/00130832-200202000-00009
Van Ganse E Kaufman L Derde MP Yernault JC Delaunois L Vincken W Effects of antihistamines in adult asthma: a meta-analysis of clinical trials Eur Respir J 1997 10 2216 24 9387943 10.1183/09031936.97.10102216
Holland WW European Community Atlas of 'Avoidable Death' 1985-89 1997 Oxford: Oxford University Press
|
16105170
|
PMC1201174
|
CC BY
|
2021-01-04 16:23:25
|
no
|
Respir Res. 2005 Aug 16; 6(1):95
|
utf-8
|
Respir Res
| 2,005 |
10.1186/1465-9921-6-95
|
oa_comm
|
==== Front
J Transl MedJournal of Translational Medicine1479-5876BioMed Central London 1479-5876-3-281604280710.1186/1479-5876-3-28ReviewRNA amplification for successful gene profiling analysis Wang Ena [email protected] Immunogenetics Section, Department of Transfusion Medicine, Clinical Center, National Institutes of Health, Bethesda, MD, 20892, USA2005 25 7 2005 3 28 28 13 4 2005 25 7 2005 Copyright © 2005 Wang; licensee BioMed Central Ltd.2005Wang; 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 study of clinical samples is often limited by the amount of material available to study. While proteins cannot be multiplied in their natural form, DNA and RNA can be amplified from small specimens and used for high-throughput analyses. Therefore, genetic studies offer the best opportunity to screen for novel insights of human pathology when little material is available. Precise estimates of DNA copy numbers in a given specimen are necessary. However, most studies investigate static variables such as the genetic background of patients or mutations within pathological specimens without a need to assess proportionality of expression among different genes throughout the genome. Comparative genomic hybridization of DNA samples represents a crude exception to this rule since genomic amplification or deletion is compared among different specimens directly. For gene expression analysis, however, it is critical to accurately estimate the proportional expression of distinct RNA transcripts since such proportions directly govern cell function by modulating protein expression. Furthermore, comparative estimates of relative RNA expression at different time points portray the response of cells to environmental stimuli, indirectly informing about broader biological events affecting a particular tissue in physiological or pathological conditions. This cognitive reaction of cells is similar to the detection of electroencephalographic patterns which inform about the status of the brain in response to external stimuli. As our need to understand human pathophysiology at the global level increases, the development and refinement of technologies for high fidelity messenger RNA amplification have become the focus of increasing interest during the past decade. The need to increase the abundance of RNA has been met not only for gene specific amplification, but, most importantly for global transcriptome wide, unbiased amplification. Now gene-specific, unbiased transcriptome wide amplification accurately maintains proportionality among all RNA species within a given specimen. This allows the utilization of clinical material obtained with minimally invasive methods such as fine needle aspirates (FNA) or cytological washings for high throughput functional genomics studies. This review provides a comprehensive and updated discussion of the literature in the subject and critically discusses the main approaches, the pitfalls and provides practical suggestions for successful unbiased amplification of the whole transcriptome in clinical samples.
Gene profilingcDNA microarrayRNA amplificationpolymerasehigh throughput analysis.
==== Body
Introduction
Quantification of gene expression is a powerful tool for the global understanding of the biology underlying complex pathophysiological conditions. Advances in gene profiling analysis using cDNA or oligo-based microarray systems uncovered genes critically important in disease development, progression, and response to treatment [1-12]. While the expression of a single or a limited number of genes can be readily estimated using minimum amount of total or messenger RNA (mRNA) from experimental or clinic samples, gene profiling requires large amount of RNA which can only be generated from global RNA amplification when using often limited amount clinical material. Conventionally at least 50 – 100 μg of total RNA (T-RNA) or 2 – 5 μg poly(A)+ RNA are generally necessary for global transcript analysis studies though efforts to enhance signal intensity and fluorochrome incorporation have reduced the amount of total RNA needed for array analysis to 1–5 ug [13]. Large amounts of RNA are not usually obtainable from clinical specimens. Thus, they pertain to experimental endeavors where cultured cell lines or tissues from pooled experimental models are used while only occasionally they are obtainable from large excisional biopsies [14]. However, most biological specimens directly obtained ex vivo for diagnostic or prognostic purposes or for clinical monitoring of treatment are too scarce to yield enough RNA for high throughput gene expression analysis. Needle or punch biopsies provide the opportunity to serially sample lesions during treatment or to sample lesion to identify predictors of treatment outcome by observing the fate of the lesion left in place. In addition, the simplicity of the storage procedure associated with the collection of small samples which can be performed at the bed side provides superior quality of RNA with minimum degradation [15]. Finally, the hypoxia which follows ligation of tumor-feeding vessels before excision is avoided with these minimally invasive methods, therefore, obtaining a true snapshot of the in vivo transcriptional program. These minimally invasive sampling techniques yield generally few micrograms of total RNA and most often even less [15,16]. Similarly, breast and nasal lavages and cervical brush biopsies, routinely used for pathological diagnosis, generate insufficient material far below the detection limit of most assays. Acquisition of cell subsets by fluorescent or magnetic sorting or laser capture micro-dissection (LCM) for a more accurate portraying of individual cell interactions in a pathological process generate even less material, in most cases, nanograms of total RNA [17-20].
Efforts have been made to broaden the utilization of cDNA microarrays using two main strategies: intensifying fluorescence signal [13,21-24] or amplifying RNA. Signal intensification approaches have reduced the requirement of RNA few folds but cannot extend the utilization of microarray to sub-microgram levels. RNA amplification in turn has gained extreme popularity based on amplification efficiency, linearity and reproducibility lowering the amount of total RNA needed for microarray analysis to nanograms without introducing significant biases. Methods aimed at the amplification of poly(A)-RNA [25] via in vitro transcription (IVT) [26] or cDNA amplification via polymerase chain reaction (PCR) [27] have reduced the material needed for cDNA microarray application and extended the spectrum of clinical samples that can be studied. Nanograms of total RNA have been successfully amplified into micrograms of pure mRNA for the screening of the entire transcriptome without losing the proportionality of gene expression displayed by the source material. Curiously, the most important advances were made by Eberwine whose main goal was not to use clinical material for high-throughput studies but rather to amplify enough material from single cells for individual or few gene analysis [28,29]. His revolutionary contribution has, however, provided a striking opportunity to explore the function of the human genome ex vivo and has exponentially opened the frontiers of clinical investigation. Modifications, optimizations and validations of RNA amplification technology based on Eberwine's pioneering work are still actively explored.
In this chapter, we will summarize efforts to optimize RNA amplification and describe in detail current amplification procedures that have been validated and applied to cDNA microarray analysis.
Collection of source material and RNA isolation
Samples used for RNA isolation and amplification should always be collected fresh and immediately processed. Excisional biopsies should be handled within 20 min and stored at -80°C (for instance with RNAlater™, Ambion, Austin, TX) if RNA isolation cannot be performed right away. Material from FNA should be collected in 5 ml of ice cold 1 × PBS or other collection medium without serum at the patient's bedside to minimize RNA metabolism or degradation. After spinning at 1,500 rpm for 5 minutes at 4°C, 2.5 ml of ACK lysing buffer should be added with 2.5 ml of 1 × PBS and incubated for 5 minutes on ice to lyse red blood cells (RBC) in case of excessive contamination. Cell pellets should be washed in 10 ml 1 × PBS and then re-suspend in small volumes of RNAlater followed by snap freezing or prior lysis of the pellet in 350 μl of RLT buffer with fresh addition of 2-mercaptoethanol (2-ME) (RNeasy mini kit, QIAGEN Inc, Valencia, CA USA) before snap freezing at -80°C. For LCM, good results can be obtained by lysing cells directly in 50 μl RLT buffer with 2-ME. Total RNA (T-RNA) and poly A RNA can both be used as starting material for RNA amplification.
The RNA isolation method strongly affects the quality and quantity of RNA. T-RNA can be isolated using commercially available RNA isolation kits. The T-RNA content per mammalian cell ranges between 20 to 40 pg of which only 0.5 – 1.0 pg are constituted by messenger RNA (mRNA) [30,31]. Sample condition, viability, functional status and phenotype of the cells are the major reasons for differential yield of T-RNA. Sample handling with precaution for RNase contamination always improves the quality and quantity of the RNA obtained. Measurement of T-RNA concentration can be performed with a spectrophotometer at OD260. An OD260/280 ratio above 1.8 is to be expected. When a very limited number of cells is available such as from LCM or FNA, very low or even negative OD readings may be observed. In this case, OD reading can be omitted. When RNA is isolated from archived samples or from samples whose collection and storage conditions were not controlled and optimized, it is preferable to estimate RNA quality and quantity using Agilent Bioanalyzer (Agilent Technologies Inc. Palo Alto, CA) or RNA gels. Clear 28S and 18S ribosomal RNA bands indicate good quality of RNA. Since 28S rRNA degradation occurs earlier than 18S rRNA and mRNA degradation in most cases correlates with 28S ribosomal RNA, the ratio of 28S versus 18S rRNA is a good indicator of mRNA quality [32]. 28S/18S rRNA ratios equal or close to 2 suggest good RNA quality.
Single strand cDNA synthesis
A critical step in RNA or cDNA amplification is the generation of double stranded cDNA (ds-cDNA) templates. First strand cDNAs are reverse transcribed from mRNA using oligo dT or random primers. In order to generate full length first strand cDNA, oligo dT(15–24 nt) with an attachment of a bacterial phage T7 promoter sequence is commonly used to initiate the cDNA synthesis [25,29,33-36]. In case of degraded RNA [37], random primers with attachment of T3 RNA polymerase promoter (T3N9) have been used for first and second strand cDNA synthesis [38]. To prevent RNA degradation while denaturing and during the reverse transcription (RT) reaction, it is useful to denature the RNA (65°C for 5 minutes or 70°C for 3 minutes) in the presence of RNasin® Plus RNase Inhibitor (Promega, Madison, WI) which forms a stable complex with RNases and inactivates RNase at temperatures up to 70°C for at least 15 minutes.
To enhance the efficiency of the RT reaction and reduce incorporation errors, the temperature of the RT reaction can be maintained at 50°C [39,40] instead of 42°C to avoid the formation of secondary mRNA structures. This can be done by using thermo-stable reverse transcriptase (ThermoScript™ RNase H- Reverse Transcriptase, Invitrogen, Carlsbad, CA) or regular RTase [41] in the presence of disaccharide trehalose [42-44]. Disaccharide trehalose not only can enhance the thermo-stability of RTase but also posses thermo-activation functions. This modification greatly enhances the accuracy and the efficiency of RT with minimum impact on the DNA polymerase activity [39]. The utilization of DNA binding protein T4gp32 (USB, Cleveland) in RT reactions also improves cDNA synthesis [40,41,45,46]. T4gp32 protein may essentially contribute to the qualitative and quantitative efficiency of the RT reaction by reducing higher order structures of RNA molecules and hence reduce the pause sites during cDNA synthesis.
In Van Gelder and Eberwine's T7 based RNA amplification [28], the amount of oligo dT-T7 primer used in the first strand cDNA synthesis can affect the amplified RNA in quantity and quality. Excessive oligo dT-T7 in the RT reaction could lead to template independent amplification [47]. This phenomenon is not observed when the template switch approach is combined to in vitro transcription (Wang, E. unpublished data).
Double stranded cDNA (ds-cDNA) synthesis
RNA amplification methods differ according to the strategies used for the generation of ds-cDNA as templates for in vitro transcription or PCR amplification. There are two basic strategies that have been extensively validated and applied for high throughput transcriptional analysis. The first is based on Gubler-Hoffman's [48] ds-cDNA synthesis subsequently optimized by Van Gelder and Eberwine [28,29]. This technology utilizes RNase H digestion to create short fragments of RNA as primers to initiate the second strand cDNA elongation under DNA polymerase I. Fragments of second strand cDNA are then ligated to each other sequentially under E. Coli DNA ligase followed by polishment using T4 DNA polymerase to eliminate loops and to form blunt ends. Amplifications based on this methods have been widely used in samples obtained in physiological or pathological conditions and extensively validated for its fidelity, reproducibility and linearity compared to un-amplified RNA from the same source materials [29,33,47,49-52].
The alternative ds-cDNA synthesis approach utilizes retroviral RNA recombination as a mechanism for template switch to generate full length ds-cDNA. The method was initially invented for full length cDNA cloning and, therefore, the main targets of this method are undegredated transcripts. Gubler-Hoffman's ds-cDNA synthesis has the potential of introducing amplification biases because of a possible 5' under-representation. In addition, the low stringency of the temperature in which ds-cDNA synthesis occurs may introduce additional biases [33]. Although 5' under-representation could, in theory, be overcome by hairpin loop second-strand synthesis [53], the multiple enzymes (4) used in the reaction could also in turn cause errors.
To ensure generation of full-length ds-cDNA, [54] synthesis is performed taking advantage of the intrinsic terminal transferase activity and template switch ability of Moloney Murine Leukemia Virus RTase [55]. This enzyme adds non-template nucleotides at the 3' end of the first strand cDNA, preferentially dCTP oligo nucleotides. A template-switch oligonucleotide (TS primer) containing a short string of dG residues at the 3' end is added to the reaction to anneal to the dC string of the newly synthesized cDNA. This produces an overhang that allows the RTase to switch template and extend the cDNA beyond the dC to create a short segment of ds-cDNA duplex. After treatment with RNase H to remove the original mRNA, the TS primer initiates the second stranded cDNA synthesis by PCR. Since the terminal transferase activity of the RTase is triggered only when the cDNA synthesis is complete, only full-length single stranded cDNA will be tailed with the TS primer and converted into ds-cDNA. Using the TS primer, second strand cDNA synthesis is carried at 75°C after a 95°C denaturing and a 65°C annealing step in the presence of single DNA polymerase [35]. This technique, in theory, overcomes the bias generated by amplification methods depending only on 3' nucleotide synthesis and hence it is, in theory, superior to the Gubler-Hoffman's ds-cDNA synthesis. However, no significant differences in correlation coefficients of amplified versus non amplified RNA were observed when the Gubler-Hoffman's ds-cDNA method was compared with the TS ds-cDNA amplification using high throughput analysis [40,56] The fidelity of template switch-based amplification methods has been assessed by numerous gene profiling analyses on different type of microarray platforms, real time PCR and sophisticated statistical analyses and it has been well accepted for high throughput transcriptome studies.
RNA amplifications
Linear amplification
Amplification of mRNA without skewing relative transcript abundance remains a focus of research. Linear amplification methods have been developed that in theory should maintain the proportionality of each RNA species present in the original sample. IVT using ds-cDNA equipped with a bacteriophage T7 promoter [28] provides an efficient way to amplify mRNA sequences and thereby generate templates for synthesis of fluorescently-labeled single-stranded cDNA [25,26,28,29,33,53]. Depending upon the T7 or other (T3 or SP6) promoter sequence position on the ds-cDNA, amplified RNA can be either in sense or antisense orientation. Oligo dT attachments to the promoter sequence, for example oligo dT-T7, prime first strand cDNA positioned the promoter at the 3' end of genes (5' end of cDNA) and, therefore, lead to the amplification of antisense RNA (aRNA) or complement RNA (cRNA). Promoters positioned at the 5' end of genes by random [57] or TS primers (Wang E, unpublished observation) generate sense RNA (sRNA). Amplified sRNA can be also produced by tailing of oligo dT to the 3' of the cDNA followed by oligo dA-T7 priming for double stranded T7 promoter generation at the 5' end of genes [58]. The singularity of this approach resides in the utilization of a DNA polymerase blocker at the 3' of the oligo dA-T7 primer which prevents the elongation of second strand cDNA synthesis while priming for the elongation of the double stranded promoter. In this fashion, only sense amplification can be achieved by the presence of the 5' ds-T7 promoter followed by single strand cDNA templates.
IVT using DNA-dependent RNA polymerase is an isothermal reaction with linear kinetics. The input ds-cDNA templates are the only source of template for the complete amplification and, therefore, any errors created on the newly synthesized RNA will not be carried or amplified in the following reactions. Overall, RNA polymerase makes an error at a frequency of about once in 10,000 nucleotides corresponding to about once per RNA strand created . This contrasts with DNA-dependent DNA polymerase which incorporates an error once in every 400 nucleotides. Most importantly, these errors are exponentially amplified in the following reaction since the amplicons serve as templates. Thus, RNA polymerase catalyzes transcription robotically and efficiently without sequence dependent bias. Recombinant RNA polymerases have been engineered to enhance the stability of the enzyme interacting with templates and reduce the abortive tendency [59] of the wild type RNA polymerase which in turn improved the elongation phase resulting in complete mRNA transcripts. The length of amplified RNA ranges from 200 to 6,000 nucleotides for the first round of amplification and 100 to 3,000 nucleotides for the second round when random primers are used [36,60] The amplification efficiency is greater than 2,000 fold in the first round and 100,000 fold in the second round [35,60]
Two rounds of IVT are commonly required when sub micrograms of input total RNA are used. It has been estimated that after two rounds of amplification the frequency of only 10% of the genes in a specimen is reduced [61] and more than two rounds of amplification may still retain at least in part the proportionality of gene expression among different RNA populations [35]. However, we, generally, do not recommend going over two rounds of amplification unless necessary for extremely scant specimens such as when processing single or few cell specimens, to avoid unnecessary biases related to amplification. The fidelity of IVT has been extensively assessed by gene profiling analysis, quantitative real-time PCR and statistical testing by comparing estimates of gene expression in amplified versus non-amplified RNA [35].
Pitfalls have been also associated with IVT. The fidelity of the first round amplification decreases when the input starting material is less than 100 ng because of the intrinsic low abundance of transcripts (particularly those under represented in the biological specimen). This can be rescued by two rounds of IVT if sufficient RNA species are present in the input material [35]. In addition, two rounds of amplification tend to introduce a 3' bias due to the use of random primers in the cDNA synthesis for ds-cDNA template creation. This should not affect the usefulness of the technique for high throughput gene profiling analysis since cloned cDNA arrays are 3' biased and even oligo arrays are designed to target the 3' end of each gene. Sequence-specific biases introduced during amplification are generally reproducible and, although negligible, could mislead data interpretation only when amplified RNA is directly compared with non amplified RNA on the same array platform. This type of error can be easily circumvented by using samples processed in identical conditions. Degradation of amplified RNA during prolonged (more than 5 hours) IVT may result in lower average size of aRNA and decreased yields [37]. This results from residual RNase in the enzyme mixture used for IVT reaction and can be prevented by the addition of RNase inhibitor in the reaction if a prolonged amplification is needed.
PCR-based exponential amplification
IVT is burdensome time consuming and may, theoretically, produce a 3' bias especially when two rounds of amplification are employed. Exponential amplification (PCR-based) may avoid these drawback and it has shown promise since, contrary to the IVT, is simple and efficient. However, PCR-based amplification has its own drawbacks.
The limitations of PCR-based amplification stem from the characteristics of the DNA-dependent DNA polymerase enzymatic function. The function of this enzyme is biased towards a lower efficiency in the amplification of GC rich sequences compared with AT rich sequences. In addition, as previously discussed, not only creates errors more frequently than RNA polymerase but also amplifies these mistakes because the reaction utilizes the amplicons as templates for subsequent amplification [62]. In addition, due to the exponential amplification, the reaction could reach saturation in conditions in which excess input template quantities are used or because of the exhaustion of substrate. This would favor the amplification of high-abundance transcripts which would compete more efficiently for substrate in the earlier cycles of the amplification process resulting in loss of proportionality of the amplification process. Optimization of PCR cycle number to avoid reaching the saturation cycle and adjustments in the amount of template input could overcome the problems [63]. The utilization of DNA polymerase with proofreading function could eradicate errors created in the cDNA amplification [64]. This approach preserves the relative abundance of transcript [65] and it may outperform IVT when less than 50 ng of input RNA are available as starting material [66,67].
PCR-based cDNA amplification can be categorized as template switching (TS)-PCR [52,68,69], random PCR [70] and 3' tailing with 5' adaptor ligation PCR [71] based on the generation of a 5' anchor sequence which provides a platform for 5' primer annealing. TS-PCR employs the same template switch mechanism in ds-cDNA generation and in the amplification of ds-cDNA using 5' TS primer II (truncated TS primer) and 3' oligo dT or dT-T7 primers (depending upon the primer used in the first strand cDNA synthesis). Random PCR utilizes modified oligo dT primers (dT-T7 or dT-TAS (Target Amplification Sequence) or random primers with an adaptor sequence for the first strand cDNA initiation and random primers with an attachment of the same adaptor, for example dN10-TAS [70], for second strand cDNA synthesis. The attached sequence, such as TAS, generates a 5' anchor on the cDNA for subsequent PCR amplification with a single TAS-PCR primer. This approach is more suitable for RNA with partial degradation and with the risk of under representation of the 5' end. The third exponential amplification utilizes terminal deoxynucleotidyl transferase function to add a polymonomer, for example poly dA, tail to the 5' end of the gene. The tailed poly dA provides an annealing position for the oligo dT primer which lead the second strand cDNA synthesis. Ds-cDNA can then be amplified under one oligo dT primer or dT-adaptor primer if an adaptor sequence is attached [66]. Direct adaptor ligation is another alternative way to generate ds-cDNA with a known anchor sequence at the 5' end [71]. In this way, single strand cDNA is generated using oligo dT primers immobilized onto magnetic beads and second strand cDNA is completed by Van Gelder and Eberwine's ds-cDNA generation method. A ds-T7 promoter-linker is then unidirectionally ligated to the blunted ds-cDNA at the 5' end. PCR amplification can then be performed using the 5' promoter primer and the 3' oligo dT or dT-adapter primer, if an adapter is attached. PCR amplified ds-cDNA is suitable for either sense or antisense probe arrays.
The combination of PCR amplification to generate sufficient ds-cDNA template followed by IVT [70,71]. is an attractive strategy to amplify minimal starting material since it takes advantage of the efficiency of the PCR reaction and the linear kinetics of IVT while minimizing the disadvantage discussed above. Validations of PCR-based RNA amplification methods are fewer than those for IVT but have been so far persuasive in spite of the prevalent expectations. Skepticism concerning the reproducibility and linearity are still one of the key factors preventing the extensive application of this approach.
Target labeling for cDNA microarray using amplified RNA
The generation of high quality cDNA microarray data depends not only on sufficient amount and highly representative amplified target, but also on the target labeling efficacy and reproducibility. Steps involved in the targets preparation such as RNA amplification, target labeling, pre-hybridization, hybridization and slides washing are imperative in enhancing foreground signal to background noise ratios. Linear spectrum of signal intensity that correlates with gene copy numbers without having to compensate detection sensitivity is one of the key factors for high quality cDNA analysis. Therefore, target labeling is a critical step to achieve consistently high signal images.
Typically, fluorescently-labeled cDNA is generated by incorporation of conjugated nucleotide analogs during the reverse transcription process. Depending upon the detection system, labeled markers can be either radioactive, color matrix or florescent. Florescence labeling outperforms the other labeling methods because of the versatile excitation and emission wave length. In addition, it has the advantage of not being hazardous. Among the fluorochrome, Cy3 (N, N8-(dipropyl)-tetramethylindocarbocyanine) and Cy5 (N, N8-(dipropyl)-tetramethylindodicarbocyanine) are most commonly used in cDNA microarray applications due to their distinct emission (510 and 664 respectively). Cy5 labeled dUTP and dCTP are less efficient in incorporation during the labeling reaction compared to Cy3 labeled dUTP or dCTP and they are more sensitive to photo bleach because of their chemical structure. Therefore, labeling bias needs to be accurately analyzed and results should be normalized according to standard normalization procedures.
Target labeling can be divided into two major categories: direct fluorescence incorporation and indirect fluorescence incorporation. The first category utilizes fluorescence-labeled dUTP or dCTP to partially substitute unlabeled dTTP or dCTP in the RT reaction to generate Cydye-labeled cDNA. This label incorporation method is suitable for cDNA clone microarray using amplified aRNA as templates or oligo array using amplified sRNA as template.
A limitation of direct labeling consists in the fact that fluorescent nucleotides are not the normal substrates for polymerases and some may be particularly sensitive to the structural diversity of these artificial oligonucleotides. The fluorescent moieties associated with these nucleotides are often quite bulky and, therefore, the efficiency of incorporation of such nucleotides by polymerase tends to be much lower than that of natural substrates. An alternative is to incorporate, either by synthesis or by enzymatic activity, a nucleotide analog similar to the natural nucleotide in structure featuring a chemically reactive group, such as 5-(3-Aminoallyl)-2'-deoxyuridine 5'-triphosphate (aa-dUTP), to which a fluorescent dye, such as Cydye, may then be attached [72]. The reactive amine of the aa-dUTP can be incorporated by a variety of RNA-dependent and DNA-dependent DNA polymerases. After removing free nucleotides, the aminoallyl labeled samples can coupled to dye, purified again, and then applied to a microarray [73]. The optimized ratio of aa-dUTP versus dTTP in the labeling reaction should be 2 to 3 respectively.
In theory, indirect outperforms direct labeling by reducing of the cost and maximizing signal intensity through increases in incorporation of fluorochrome or through signal amplification using fluorescence-labeled antibody or biotin-streptavdin complexes. However, more steps are involved in the purification of the labeled target prior to hybridization which make this strategies less frequently used.
RNA amplification protocols
The protocols presented here are routinely used in our laboratory in response to several inquires by interested investigators, The protocol is based on a combination of strategies discussed in the previous section that have been used for RNA amplification that we have applied to optimize TS-IVT following the original Eberwine's RNA amplification protocols.
Material and reagents
Dilute stock solution to the appropriate working concentration.
dNTP mix solution (dATP, dCTP, dGTP, dTTP, 10 mM each) (Pharmacia Cat# 27-2035-02)
Low T dNTP (5 mM dA, dG and dCTP, 2 mM dTTP)
RNasin Plus (20 units/μl) (Promega Cat# N2611)
Advantage PCR buffer (come with Advantage cDNA polymerase)
Linear Acrylamide (0.1 ug/μl. Ambion; Cat# 9520)
Phenol: Chloroform: Isoamyl alcohol (25:24:1) (Boehringer Mannhem Cat #101001)
Phase Lock gel (heavy) (5 prime to 3 prime, Inc.; Cat# pl-188233)
7.5 M ammonium acetate (Sigma; Cat# A2706)
DEPC treated H2O
In Vitro Tanscription Kit (Ambion; T7 Megascript Kit #1334)
Cy-dUTP (1 mM Cy3 or Cy5)
1 M NaOH
500 mM EDTA
1 × TE
1 M Tris pH 7.5
50× Denhardt's blocking solution (Sigma; Cat# 2532)
Poly dA40–60 (8 mg/ml) (Pharmacia; Cat# 27-7988-01)
Human Cot I DNA (10 mg/ml) (Invitrogen; Cat# 15279-011)
20 × SSC
10% SDS
T4gp32 protein (8 mg/ul) (USB Cat# 74029Y)
Enzymes
RNase H- MMLV Reverse Transcriptase (Superscript II) (200 units/μl) (Invitrogen; Cat# 18064-071)
50× Advantage cDNA Polymerase mix (Clontech Cat# 8417-1)
RNase H (2 U/μl. Invitrogen; Cat# 18021-071)
10× T7 RNA polymerase mix (within the Megascript kit)
Primers
Oligo dT-T7 primer (5' AAA CGA CGG CCA GTG AAT TGT AAT ACG ACT CAC TAT AGG CGC T(15) 3') (0.125–0.25 μg/μl for the first round amplification depending on the amount of input total RNA and 0.5 μg/μl for the second round amplification) in RNase free water. Synthesized primer should be SDS-PAGE purified to insure the full length. The concentration of primer is varied according to the starting material used. This promoter sequence is much longer than the consensus sequence defined by Dunn and Studier (1983) and can be purchased from New England Biolabs and Stratagene Inc. In the extended sequence shown here, the consensus sequence is embedded in the between a 5'flanking region that provides space for the T7 RNA polymerase to bind and a 3'-flanking trinucleotide that stimulates transcription catalyzed by the enzyme.
TS primer (5' AAG CAG TGG TAA CAA CGC AGA GTA CGC GGG 3') (0.25 μg/μl) SDS-PAGE purified. According to the Chenchik's [74] data, ribouncleotide GGG at the 3' end should give the best TS effect instead of deoxinucleotide GGG. We have used TS primer with dGGG at the 3' end in multiple experiments and achieved satisfying results. The amount of TS primer used in the second strand synthesis can be varied according to the amount of starting material. We generally use 0.25 μg/μl when 3–6 ug of total RNA used and 0.125 μg/μl when less total RNA used.
Random hexamer (dN6) (8 μg/μl).
Columns
Micro Bio-Spin Chromatograph column (Bio-gel P-6) (Bio-Rad; Cat# 732–6222)
Microcon YM-30 column (Millipore; Cat# 42410).
Procedures
First strand cDNA synthesis
1. In PCR reaction tube, mix 0.01–5 μg total RNA in 9 μl DEPC H2O with 1 μl (0.1–0.25 μg/μl) oligo dT(15)-T7 primer (5' AAA CGA CGG CCA GTG AAT TGT AAT ACG ACT CAC TAT AGG CGC T(15) 3'), 1 ul of RNasin Plus and heat to 70°C for 3 min. Cool to room temperature then add the following reagents: (a master mix can be prepared for multiple samples)
2. 4 μl 5 × First strand buffer
3. 1 μl (0.1–0.25 μg/μl) TS (template switch) oligo primer
4. 2 μl 0.1 M DTT
5. 2 μl 10 mM dNTP
6. 1 μl Superscript II
7. 0.5 μl of T4 gp32 (8 μg/μl)
50°C for 90 min in thermal cycler.
Second strand cDNA synthesis
1. Add 106 μl of DEPC treated H2O to the cDNA reaction tube
2. 15 μl Advantage PCR buffer
3. 3 μl 10 mM dNTP mix
4. 1 μl of RNase H
5. 3 μl Advantage cDNA Polymerase mix
Cycle at 37°C for 5 min to digest mRNA, 94°C for 2 min to denature, 65°C for 1 min. for specific priming and 75°C for 30 min for extension.
Note: Since the TS primer which initiates the second strand cDNA synthesis is already present in the first strand cDNA synthesis reaction and has been primed to the extended part of the cDNA, no additional primer is required in this step.
Stop reaction with 7.5 μl 1 M NaOH solution containing 2 mM EDTA and incubate at 65°C for 10 min. to inactivate enzymes.
(Reaction can be stopped after this step and the reaction tube can be stored at -20°C.)
Double stranded cDNA cleanup
(This step is designed to prevent carry over of non-incorporated dNTP, primers and inactivated enzymes into the following in vitro transcription. Keep in mind that although the double stranded cDNAs are stable and will not be affected by RNase contamination, they will be used as template in the IVT reaction which is RNase free.)
Phenol-Chloroform-Isoamyl isolation and ethanol precipitation
Add 1 μl Linear Acrylamide (0.1 μg/μl) as DNA carrier to the sample to enhance double stranded-cDNA precipitation. Add 150 μl Phenol: Chloroform: Isoamyl alcohol (25:24:1) to the double stranded cDNA tube and mix well by pipetting (be careful not to spill or contaminate). Transfer the slurry solution to Phase lock gel tube and spin at 14,000 rpm for 5 min at room temperature. Transfer the aqueous phase to RNase/DNase-free 1.7 ml tube and add 70 μl of 7.5 M ammonium acetate first and then 1 ml 100% ethanol (EtOH). Mix well. Centrifuge right away at 14,000 rpm for 20 min at room temperature to prevent co-precipitation of oligos. (A visible small white pellet should be seen at the bottom of the tube even if nano grams of starting material have been used. This pellet suggests successful precipitation.) Wash pellet with 800 μl 100% EtOH and spin down at maximum speed for 8 min. Repeat this washing step one more time. Air dry or speedvac and re-suspend double stranded cDNA in 8 ul DEPC H2O.
In Vitro Tanscription (Ambion; T7 Megascript Kit #1334)
2 μl of each 75 mM NTP (A, G, C and UTP)
2 μl reaction buffer
2 μl enzyme mix (RNase inhibitor and T7 phage RNA polymerase)
8 μl double stranded cDNA
37°C for 5 hr.
According to Ambion, the incubation can be interrupted by storing reaction tube at -20°C and resuming the incubation later without losing efficiency.
Purification of amplified RNA
Any manufactured RNA isolation kit can be applied
Monophasic reagent such as TRIzol reagent from GibcoBRL, (Cat#15596) are used here based on the efficient recovery of aRNA (RNeasy mini kit could be used for aRNA purification instead of TRIzol but, in our experience, RNA recovering is about 50% of that recovered with the TRIzol method.).
a. Add 0.5 ml of TRIzol solution to the transcription reaction. Mix the reagents well by pipetting or gentle vortexing.
b. Add 100 μl chloroform. Mix the reagents by inverting the tube for 15 seconds. Allow the tube to stand at room temperature for 2 – 3 minutes.
c. Centrifuge the tube at 10,000 g for 15 min at 4°C.
d. Transfer the aqueous phase to a fresh tube and add 250 μl of isopropanol.
e. Store the sample on ice for 5 minutes and then centrifuge at 10,000 g for 15 minutes.
f. Wash the pellet twice with 800 μl 70% EtOH
g. Allow the pellet to dry in air on ice and then dissolve it in 20 μl DEPC H2O
h. Measure the quantity of RNA concentration spectrophotometrically.
Second round of amplification
Mix amplified aRNA (0.5–1 μg) in 9 ul DEPC H2O with 1 μl (2 μg/μl) random hexamer (i.e. dN6) and heat to 70°C for 3 min, cool to room temperature. Then add the following reagents:
1. 4 μl 5 × First strand buffer
2. 1 μl (0.5–1 μg/ul) oligo dT-T7 primer
3. 2 μl 0.1 M DTT
4. 1 μl RNAsin
5. 2 μl 10 mM dNTP
6. 1 μl Superscript II
42°C for 90 min.
(Note: More than 1 ug of aRNA is not suggested. Too much template in IVT reaction could cause the amplification to reach a plateau with loss of amplification linearity. Because of random primer used here, 42°C in stead of 50°C is used)
From here, follow the previously described procedure for second strand cDNA synthesis, double stranded cDNA cleanup. In the second IVT, 40 ul of IVT reaction mixture are suggested to use instead of 20 ul. RNA isolation is followed.
Target labeling by reverse transcription
4 μl First strand buffer
1 μl dN6 primer (8 μg/μl)
2 μl 10× low T – dNTP (5 mM A, C and GTP, 2 mM dTTP)
2 μl Cy-dUTP (1 mM Cy3 or Cy5)
2 μl 0.1 M DTT
1 μl RNasin
3–6 μg amplified aRNA in 8 μl DEPC H2O
Mix well and heat to 65°C for 5 min then cool down to 42°C.
Add 1.5 μl SSII. Incubate for 90 min at 42°C. Add 2.5 μl 0.5 M EDTA and heat to 65°C for 1 min. Add 5 μl 1 M NaOH and incubate at 65°C for 15 min to hydrolyze RNA. Add 12.5 μl 1 M Tris immediately to neutralize the pH. Bring volume to 70 μl by adding 35 μl of 1 × TE.
Note: The amounts of aRNA used for labeling depends on the size of the array. If the array with 2000–8000 genes, 3 ug aRNA will be sufficient while a larger chip such as 16–20 k will need 6 ug of aRNA. The labeling reaction components do not need to be changed.
Target clean up
Prepare Bio-6 column and run target solution through it. Collect flow through and add 250 μl 1 × TE to it. Concentrate target to ~20 μl using Microcon YM-30 column.
Hybridization
Combine Cy3 labeled reference sample and Cy5 labeled target sample (adjust the color to purple) and then complete dry the sample using speedvac. Resuspend sample in 37 μl volume (for 22 mm × 40 mm printing surface) containing 1 μl 50× Denhardt's blocking solution, 1 μl poly dA (8 μg/μl), 1 μl yeast tRNA (4 mg/ml), 10 μl Human Cot I DNA, 3 μl 20× SSC, 1 μl of 10% SDS and 20 μl of DEPC treated water. Heat sample for 2 min at 99°C and apply target mixture to array slide, add coverslip, place in humidified hyb chamber, and hybridize at 65°C over night.
Washing
1. Wash with 2 × SSC + 0.1% SDS to get rid of the cover slide.
2. Wash with 1 × SSC for 1 min.
3. Wash with 0.2 × SSC for 1 min.
4. Wash with 0.05 × SSC for 10 second
5. Centrifuge slide at 80–100 g for 3 min. (Slide can be put in slide rack on microplate carriers or in 50 ml conical tube and centrifuged in swinging-bucket rotor.)
Scan Slide
==== Refs
Bittner M Meltzer P Chen Y Jiang E Seftor E Hendrix M Molecular classification of cutaneous malignant melanoma by gene expression: shifting from a countinuous spectrum to distinct biologic entities Nature 2000 406 536 840 10952317 10.1038/35020115
Wang E Panelli MC Marincola FM Genomic analysis of cancer Princ Pract Oncol 2003 17 1 16 10.1159/000071805
Yeoh E-J Ross ME Shurtleff SA Williams WK Patel D Mahfouz R Classification, subtype discovery and prediction of outcome in pediatric acute lymphoblastic leukemia by gene expression profiling Cancer Cell 2002 1 133 136 12086872 10.1016/S1535-6108(02)00032-6
Wang E Miller LD Ohnmacht GA Mocellin S Petersen D Zhao Y Prospective molecular profiling of subcutaneous melanoma metastases suggests classifiers of immune responsiveness Cancer Res 2002 62 3581 3586 12097256
Panelli MC Wang E Phan G Puhlman M Miller L Ohnmacht GA Genetic profiling of peripheral mononuclear cells and melanoma metastases in response to systemic interleukin-2 administration Genome Biol 2002 3 RESEARCH0035 12184809 10.1186/gb-2002-3-7-research0035
Wang E Lichtenfels R Bukur J Ngalame Y Panelli MC Seliger B Ontogeny and oncogenesis balance the transcriptional profile of renal cell cancer Cancer Res 2004 64 7279 7287 15492247
Wang E Ngalame Y Panelli MC Deavers M Mueller P Ju W Peritoneal and sub-peritoneal stroma may facilitate regional spread of ovarian cancer Clin Cancer Res 2005 11 113 122 15671535
Singh D Febbo PG Ross K Jackson DG Manola J Ladd C Gene expression correlates of clinical prostate cancer behavior Cancer Cell 2002 1 203 209 12086878 10.1016/S1535-6108(02)00030-2
Mellick AS Day CJ Weinstein SR Griffiths LR Morrison NA Differential gene expression in breast cancer cell lines and stroma-tumor differences in microdissected breast cancer biopsies by display array analysis Int J Cancer 2002 100 172 180 12115566 10.1002/ijc.10451
Islam TC Lindvall J Wennborg A Branden LJ Rabbani H Smith CI Expression profiling in transformed human B cells: influence of Btk mutations and comparison to B cell lymphomas using filter and oligonucleotide arrays Eur J Immunol 2002 32 982 993 11920564 10.1002/1521-4141(200204)32:4<982::AID-IMMU982>3.3.CO;2-9
Sasaki H Ide N Fukai I Kiriyama M Yamakawa Y Fujii Y Gene expression analysis of human thymoma correlates with tumor stage Int J Cancer 2002 101 342 347 12209958 10.1002/ijc.10624
Skotheim RI Monni O Mousses S Fossa SD Kallioniemi OP Lothe RA New insights into testicular germ cell tumorigenesis from gene expression profiling Cancer Res 2002 62 2359 2364 11956097
Xiang CC Kozhich OA Chen M Inman JM Phan QN Chen Y Amine-modified random primers to label probes for DNA microarrays Nature Biotech 2002 20 738 742 10.1038/nb0702-738
DeRisi J Penland L Brown PO Bittner M Meltzer PS Ray M Use of cDNA microarray to analyse gene expression patterns in human cancer Nature Genetics 1996 14 457 460 8944026 10.1038/ng1296-457
Wang E Marincola FM A natural history of melanoma: serial gene expression analysis Immunol Today 2000 21 619 623 11114422 10.1016/S0167-5699(00)01724-2
Wang E Marincola FM cDNA microarrays and the enigma of melanoma immune responsiveness Cancer J Sci Am 2001 7 16 23
Bonner RF Emmert-Buck M Cole K Pohida T Chuaqui R Goldstein S Laser capture microdissection: molecular analysis of tissue Science 1997 278 1481 1483 9411767 10.1126/science.278.5342.1481
Pappalardo PA Bonner R Krizman DB Emmert-Buck MR Liotta LA Microdissection, microchip arrays, and molecular analysis of tumor cells (primary and metastases) Semin Radiat Oncol 1998 8 217 223 9634498
St Croix B Rago C Velculescu V Traverso G Romans KE Montgomery E Genes expressed in human tumor endothelium Science 2000 289 1121 1122 10970218
Tsuda H Birrer MJ Ito YM Ohashi Y Lin M Lee C Identification of DNA copy number changes in microdissected serous ovarian cancer tissue using a cDNA microarray platform Cancer Genet Cytogenet 2004 155 97 107 15571795 10.1016/j.cancergencyto.2004.03.002
Chen JJ Wu R Yang PC Huang JY Sher YP Han MH Profiling expression patterns and isolating differentially expressed genes by cDNA microarray system with colorimetry detection Genomics 1998 51 313 324 9721201 10.1006/geno.1998.5354
Rajeevan MS Dimulescu IM Unger ER Vernon SD Chemiluminescent analysis of gene expression on high-density filter arrays J Histochem Cytochem 1999 47 337 342 10026235
Zejie Y Xiaoyi W Yu T Huan H The method of micro-displacement measurement to improve the space resolution of array detector Med Engineer Phys 1998 20 149 151 10.1016/S1350-4533(97)00044-1
Yu J Othman MI Farjo R Zareparsi S MacNee SP Yoshida S Evaluation and optimization of procedures for target labeling and hybridization of cDNA microarrays Mol Vis 2002 8 130 137 12011805
Lockhart DJ Dong H Byrne MC Folliette MT Gallo MV Chee MS Expression monitoring of hybridization to high-density oligonucleotide arrays Nature Biotechnol 1996 14 1675 1680 9634850 10.1038/nbt1296-1675
Luo L Salunga RC Guo H Bittner A Joy KC Galindo JE Gene expression profiles of laser-captured adjacent neuronal subtypes Nat Med 1999 5 117 122 9883850 10.1038/4806
Trenkle T Welsh J Jung B Mathieu-Daude F McClelland M Non-stoichiometric reduced complexity probes for cDNA arrays Nucleic Acids Res 1998 26 3883 3891 9705494 10.1093/nar/26.17.3883
Van Gelder RN von Zastrow ME Yool A Dement WC Barchas JD Eberwine JH Amplified RNA synthesized from limited quantities of heterogeneous cDNA Proc Natl Acad Sci USA 1990 87 1663 1667 1689846
Eberwine JH Yeh H Miyashiro K Cao Y Nair S Finnell R Analysis of gene expression in single live neurons Proc Natl Acad Sci U S A 1992 89 3010 3014 1557406
Roozemond RC Ultramicrochemical determination of nucleic acids in individual cells using the Zeiss UMSP-I microspectrophotometer. Application to isolated rat hepatocytes of different ploidy classes Histochem J 1976 8 625 638 993053 10.1007/BF01003963
Uemura E Age-related changes in neuronal RNA content in rhesus monkeys (Macaca mulatta) Brain Res Bull 1980 5 117 119 6155181 10.1016/0361-9230(80)90182-3
Skrypina NA Timofeeva AV Khaspekov GL Savochkina LP Beabealashvilli RS Total RNA suitable for molecular biology analysis J Biotechnol 2003 105 1 9 14511905 10.1016/S0168-1656(03)00140-8
Phillips J Eberwine JH Antisense RNA amplification: a linear amplification method for analyzing the mRNA population from single living cells Methods 1996 10 283 288 8954839 10.1006/meth.1996.0104
Eberwine JH Amplification of mRNA populations using aRNA generated from immobilized oligo(dT)-T7 primed cDNA Biotechniques 1996 20 584 594 8800675
Wang E Miller L Ohnmacht GA Liu E Marincola FM High fidelity mRNA amplification for gene profiling using cDNA microarrays Nature Biotech 2000 17 457 459
Wang E Marincola FM Bowtell D, Sambrook J Amplification of small quantities of mRNA for transcript analysis DNA arrays – A Molecular Cloning Manual 2002 Cold Springs Harbor, NY: Cold Spring Harbor Laboratory Press 204 213
Spiess AN Mueller N Ivell R Amplified RNA degradation in T7-amplification methods results in biased microarray hybridizations BMC Genomics 2003 4 44 14606961 10.1186/1471-2164-4-44
Xiang CC Chen M Ma L Phan QN Inman JM Kozhich OA A new strategy to amplify degraded RNA from small tissue samples for microarray studies Nucleic Acids Res 2003 31 e53 12711698 10.1093/nar/gng053
Malboeuf CM Isaacs SJ Tran NH Kim B Thermal effects on reverse transcription: improvement of accuracy and processivity in cDNA synthesis Biotechniques 2001 30 1074 8 1080, 1082, passim 11355343
Kenzelmann M Klaren R Hergenhahn M Bonrouhi M Grone HJ Schmid W High-accuracy amplification of nanogram total RNA amounts for gene profiling Genomics 2004 83 550 558 15028277 10.1016/j.ygeno.2003.09.026
Schlingemann J Thuerigen O Ittrich C Toedt G Kramer H Hahn M Effective transcriptome amplification for expression profiling on sense-oriented oligonucleotide microarrays Nucleic Acids Res 2005 33 e29 15718295 10.1093/nar/gni029
Mizuno Y Carninci P Okazaki Y Tateno M Kawai J Amanuma H Increased specificity of reverse transcription priming by trehalose and oligo-blockers allows high-efficiency window separation of mRNA display Nucleic Acids Res 1999 27 1345 1349 9973624 10.1093/nar/27.5.1345
Spiess AN Mueller N Ivell R Trehalose is a potent PCR enhancer: lowering of DNA melting temperature and thermal stabilization of taq polymerase by the disaccharide trehalose Clin Chem 2004 50 1256 1259 15229160 10.1373/clinchem.2004.031336
Carninci P Nishiyama Y Westover A Itoh M Nagaoka S Sasaki N Thermostabilization and thermoactivation of thermolabile enzymes by trehalose and its application for the synthesis of full length cDNA Proc Natl Acad Sci U S A 1998 95 520 524 9435224 10.1073/pnas.95.2.520
Rapley R Enhancing PCR amplification and sequencing using DNA-binding proteins Mol Biotechnol 1994 2 295 298 7866882
Villalva C Touriol C Seurat P Trempat P Delsol G Brousset P Increased yield of PCR products by addition of T4 gene 32 protein to the SMART PCR cDNA synthesis system Biotechniques 2001 31 81 3 86 11464524
Baugh LR Hill AA Brown EL Hunter CP Quantitative analysis of mRNA amplification by in vitro transcription Nucleic Acids Res 2000 29 E29 E38 11222780 10.1093/nar/29.5.e29
Gubler U Hoffman BJ A simple and very efficient method for generating cDNA libraries Gene 1983 25 263 269 6198242 10.1016/0378-1119(83)90230-5
Li Y Li T Liu S Qiu M Han Z Jiang Z Systematic comparison of the fidelity of aRNA, mRNA and T-RNA on gene expression profiling using cDNA microarray J Biotechnol 2004 107 19 28 14687968 10.1016/j.jbiotec.2003.09.008
Rudnicki M Eder S Schratzberger G Mayer B Meyer TW Tonko M Reliability of t7-based mRNA linear amplification validated by gene expression analysis of human kidney cells using cDNA microarrays Nephron Exp Nephrol 2004 97 e86 e95 15292679 10.1159/000078642
Park JY Kim SY Lee JH Song J Noh JH Lee SH Application of amplified RNA and evaluation of cRNA targets for spotted-oligonucleotide microarray Biochem Biophys Res Commun 2004 325 1346 1352 15555575 10.1016/j.bbrc.2004.10.151
Li Y Ali S Philip PA Sarkar FH Direct comparison of microarray gene expression profiles between non-amplification and a modified cDNA amplification procedure applicable for needle biopsy tissues Cancer Detect Prev 2003 27 405 411 14585328 10.1016/S0361-090X(03)00105-3
Kacharmina JE Crino PB Eberwine JH Preparation of cDNA from single cells and subcellular regions Methods Enzymol 1999 303 3 19 10349635
Matz M Shagin D Bogdanova E Britanova O Lukyanov S Diatchenko L Amplification of cDNA ends based on template-switching effect and step-out PCR Nucleic Acids Res 1999 27 1558 1560 10037822 10.1093/nar/27.6.1558
Chenchik A Zhu YY Diatchenko L Li R Hill J Siebert PD Siebert P, Larrick J Generation and use of high-quality cDNA from small amounts of total RNA by SMART™ PCR Gene cloning and analysis by RT-PCR 1998 Natick, MA: Biotechniques Books 305 319
Zhao H Hastie T Whitfield ML Borresen-Dale AL Jeffrey SS Optimization and evaluation of T7 based RNA linear amplification protocols for cDNA microarray analysis BMC Genomics 2002 3 31 12445333 10.1186/1471-2164-3-31
Marko NF Frank B Quackenbush J Lee NH A robust method for the amplification of RNA in the sense orientation BMC Genomics 2005 6 27 15740627 10.1186/1471-2164-6-27
Goff LA Bowers J Schwalm J Howerton K Getts RC Hart RP Evaluation of sense-strand mRNA amplification by comparative quantitative PCR BMC Genomics 2004 5 76 15469607 10.1186/1471-2164-5-76
Cheetham GM Jeruzalmi D Steitz TA Transcription regulation, initiation, and "DNA scrunching" by T7 RNA polymerase Cold Spring Harb Symp Quant Biol 1998 63 263 267 10384290 10.1101/sqb.1998.63.263
Feldman AL Costouros NG Wang E Qian M Marincola FM Alexander HR Advantages of mRNA amplification for microarray analysis Biotechniques 2002 33 906 914 12398200
Gold D Coombes K Medhane D Ramaswamy A Ju Z Strong L A comparative analysis of data generated using two different target preparation methods for hybridization to high-density oligonucleotide microarrays BMC Genomics 2004 5 2 14709180 10.1186/1471-2164-5-2
Polz MF Cavanaugh CM Bias in template-to-product ratios in multitemplate PCR Appl Environ Microbiol 1998 64 3724 3730 9758791
Seth D Gorrell MD McGuinness PH Leo MA Lieber CS McCaughan GW SMART amplification maintains representation of relative gene expression: quantitative validation by real time PCR and application to studies of alcoholic liver disease in primates J Biochem Biophys Methods 2003 55 53 66 12559588 10.1016/S0165-022X(02)00177-X
Smith L Underhill P Pritchard C Tymowska-Lalanne Z Abdul-Hussein S Hilton H Single primer amplification (SPA) of cDNA for microarray expression analysis Nucleic Acids Res 2003 31 e9 12560512 10.1093/nar/gng009
Bettinotti MP Panelli MC Ruppe E Mocellin S Phan GQ White DE Clinical and immunological eveluation of patients with metastatic melanoma undergoing immunization with the HLA-C2*0702 associated epitope MAGE-A12:170–178 Int J Cancer 2003 105 210 216 12673681 10.1002/ijc.11045
Iscove NN Barbara M Gu M Gibson M Modi C Winegarden N Representation is faithfully preserved in global cDNA amplified exponentially from sub-picogram quantities of mRNA Nature Biotech 2002 20 940 943 10.1038/nbt729
Stirewalt DL Pogosova-Agadjanyan EL Khalid N Hare DR Ladne PA Sala-Torra O Single-stranded linear amplification protocol results in reproducible and reliable microarray data from nanogram amounts of starting RNA Genomics 2004 83 321 331 14706461 10.1016/j.ygeno.2003.08.008
Petalidis L Bhattacharyya S Morris GA Collins VP Freeman TC Lyons PA Global amplification of mRNA by template-switching PCR: linearity and application to microarray analysis Nucleic Acids Res 2003 31 e142 14602935 10.1093/nar/gng142
Rox JM Bugert P Muller J Schorr A Hanfland P Madlener K Gene expression analysis in platelets from a single donor: evaluation of a PCR-based amplification technique Clin Chem 2004 50 2271 2278 15472037 10.1373/clinchem.2004.035386
Klur S Toy K Williams MP Certa U Evaluation of procedures for amplification of small-size samples for hybridization on microarrays Genomics 2004 83 508 517 14962677 10.1016/j.ygeno.2003.09.005
Ohtsuka S Iwase K Kato M Seki N Shimizu-Yabe A Miyauchi O An mRNA amplification procedure with directional cDNA cloning and strand-specific cRNA synthesis for comprehensive gene expression analysis Genomics 2004 84 715 729 15475249 10.1016/j.ygeno.2004.06.012
Randolph JB Waggoner AS Stability, specificity and fluorescence brightness of multiply-labeled fluorescent DNA probes Nucleic Acids Res 1997 25 2923 2929 9207044 10.1093/nar/25.14.2923
DeRisi J Bowtell D, Sambrook J Stability, specificity and fluorescence brightness of mulitply-labeled fluorescent DNA probes DNA arrays – A molecular cloning manual 2002 Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press 204 213
Chenchik A Zhu Y Y Diatchenko L Li R Hill J Siebert PD Siebest P. and Larrick J. , pp. Generation and use of high-quality cDNA from small amounts of total RNA by SMARTTM PCR. Gene cloning and analysis by RT-PCR 1998 305 319
|
16042807
|
PMC1201175
|
CC BY
|
2021-01-04 16:39:28
|
no
|
J Transl Med. 2005 Jul 25; 3:28
|
utf-8
|
J Transl Med
| 2,005 |
10.1186/1479-5876-3-28
|
oa_comm
|
==== Front
J Transl MedJournal of Translational Medicine1479-5876BioMed Central London 1479-5876-3-321609113110.1186/1479-5876-3-32ResearchPredictors of primary breast cancers responsiveness to preoperative Epirubicin/Cyclophosphamide-based chemotherapy: translation of microarray data into clinically useful predictive signatures Modlich Olga [email protected] Hans-Bernd [email protected] Marc [email protected] Werner [email protected] Hans [email protected] Institute of Chemical Oncology, University of Düsseldorf, Düsseldorf, Germany2 Bayer Healthcare AG, Diagnostic Research Germany, Leverkusen, Germany3 Interdisciplinary Breast Center IBC, City Hospital, Düsseldorf, Germany2005 9 8 2005 3 32 32 2 5 2005 9 8 2005 Copyright © 2005 Modlich et al; licensee BioMed Central Ltd.2005Modlich 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
Our goal was to identify gene signatures predictive of response to preoperative systemic chemotherapy (PST) with epirubicin/cyclophosphamide (EC) in patients with primary breast cancer.
Methods
Needle biopsies were obtained pre-treatment from 83 patients with breast cancer and mRNA was profiled on Affymetrix HG-U133A arrays. Response ranged from pathologically confirmed complete remission (pCR), to partial remission (PR), to stable or progressive disease, "No Change" (NC). A primary analysis was performed in breast tissue samples from 56 patients and 5 normal healthy individuals as a training cohort for predictive marker identification. Gene signatures identifying individuals most likely to respond completely to PST-EC were extracted by combining several statistical methods and filtering criteria. In order to optimize prediction of non responding tumors Student's t-test and Wilcoxon test were also applied. An independent cohort of 27 patients was used to challenge the predictive signatures. A k-Nearest neighbor algorithm as well as two independent linear partial least squares determinant analysis (PLS-DA) models based on the training cohort were selected for classification of the test samples. The average specificity of these predictions was greater than 74% for pCR, 100% for PR and greater than 62% for NC. All three classification models could identify all pCR cases.
Results
The differential expression of 59 genes in the training and the test cohort demonstrated capability to predict response to PST-EC treatment. Based on the training cohort a classifier was constructed following a decision tree.
First, a transcriptional profile capable to distinguish cancerous from normal tissue was identified. Then, a "favorable outcome signature" (31 genes) and a "poor outcome signature" (26 genes) were extracted from the cancer specific signatures. This stepwise implementation could predict pCR and distinguish between NC and PR in a subsequent set of patients. Both PLS-DA models were implemented to discriminate all three response classes in one step.
Conclusion
In this study signatures were identified capable to predict clinical outcome in an independent set of primary breast cancer patients undergoing PST-EC.
breast cancerpreoperative chemotherapymicroarrayprognostic classification
==== Body
Introduction
Breast cancer is the most common neoplasia of women being diagnosed in approximately 211,000 women annually in the United States. In spite of earlier detection and improved treatment, it remains the second leading cause of cancer-related death in the United States and in other developed countries [1]. The genetic background of patients and the tumor's genetic and epigenetic anomalies create, in combination, molecularly distinct subtypes arising from distinct cell types within the ductal epithelium [2,3]. This genetic complexity underlies the clinical heterogeneity of breast cancer limiting a rational selection of treatment tailored to individual patient/tumor characteristics.
Standard therapeutic decision-making (i.e.., NIH; St. Gallen consensus) rely on several clinicopathological factors such as patients' age, tumor stage, grade, size, nodal status as well as hormone, and growth receptor status. The analysis of single molecular markers such as ki-67 and ERBB2 can also contribute to the therapeutic decision making. Although all of these factors have been correlated to patients' survival in general, the same prognostic profile often results in dissimilar clinical outcomes in individual patients. Thus, conventional prognostic factors provide insufficient information to evaluate the heterogeneity of this disease and to make treatment more effective for individual patients.
One problem faced by present cancer therapy is the over-treatment of patients with chemotherapy, which is associated with severe toxicity and increasing healthcare spending without clear survival benefit over untreated controls [4,5]. Because of the lack of adequate predictive markers (i.e., ER, HER2/neu), nearly all patients receive routinely standard treatment in spite of grim changes of deriving any benefit. Therefore, the identification of molecular markers predictive of patients' responsiveness to treatment is becoming a central focus of translational research.
Micro array technology offers insights about the simultaneous expression of thousands of genes providing global information about the transcriptional program associated with specific cellular or tissue conditions. This provides a high-throughput screening tool for the identification of molecular patterns of cancerous cell possibly associated with their sensitivity to therapy [6-9]. This strategy yielded significant contributions by dissecting beyond histopathologic features the molecular aspects of breast cancers, their association with lymph-nodal spread, metastatization and overall survival.
An important and so far seldom explored utilization of micro array technology is the identification of signatures predictive of responsiveness to chemotherapy. To get clear correlation between chemotherapeutic success and pre treatment gene expression we needed to rely in a model where chemotherapy is given before surgical resection so that its outcome could be evaluated. Beside the most common postoperative (adjuvant) chemotherapy, preoperative systemic therapy (PST) has been recently proposed for early-stage breast cancer. PST, uses cytotoxic drugs as the first modality of treatment allowing in vivo monitoring of the therapeutic responsiveness of a primary tumor over a given time period (e.g., 4 months). PST is offered preoperatively to patients with either large inoperable breast cancers or to patients interested in breast conserving surgery [10-16]. PST in general does not offer a survival advantage over standard adjuvant treatment but does identify those patients (up to 20%) with tumors reacting with a complete remission to the drug [17,18]. Complete tumor remission, as confirmed by pathological examination, is often associated with prolonged disease free survival [19-22]. Additionally, PST can reduce the growth rate of residual distant micrometastases compared with classical adjuvant therapy [17].
By predicting which subset of tumors may respond to PST transcriptional profiling of pre-treatment samples could represent a powerful tool for patient selection.
Patients, Materials and methods
Patients
This study was performed in collaboration with the Institute of Chemical Oncology, University of Düsseldorf, Germany, and Bayer Healthcare AG, Diagnostic Research, Leverkusen, Germany. All patients were recruited at the Interdisciplinary Breast Center IBC, City Hospital Düsseldorf. Patients signed an informed consent before any procedure. Study eligibility criteria required that participants presented with not previously treated primary breast cancer to be treated preoperatively.
Samples of primary breast carcinomas were collected between May 1999 and March 2003 from patients subjected to PST treatment with epirubicine/cyclophosphamide (EC). Since in several cases treatment modifications occurred or full pathological confirmation of response status was not conclusive not all samples were studied. Quality of samples related to delays in processing also limited the number of samples studied. In the end, a total of 56 tumor samples were identified from comparable treatment groups and were studied for marker discovery together with five normal breast samples excised from patient with benign pathology. Additionally, tumor samples removed from 27 patients treated with EC-based PST between December 2002 and September 2003 were analyzed as a second independent validation cohort. EC consisted of epirubicin 90 mg m2 per day 1 in a short i.v. infusion, and cyclophosphamide 600 mg m2 per day 1 in a short i.v. infusion. Four cycles of EC were administrated 14 days apart. Some patients received additionally Tamoxifen, Femara or seldom Zoladex for 4–5 weeks after EC course and before surgery. All tumor samples were collected as needle biopsies of primary tumors prior to any treatment. The biopsies were obtained under local anesthesia using Bard® MAGNUM™ Biopsy Instrument (C. R. Bard, Inc., Covington, U. S.) with Bard® Magnum biopsy needles (BIP GmbH, Tuerkenfeld, Germany) following ultrasound guidance. Samples were collected following routine conditions for pathological diagnosis following institutional review board guidelines. Pathological examination was carried out for all tumor samples by the same pathologist at the Interdisciplinary Breast Center IBC. The remainders of the samples were flash-frozen. After PST, all patients underwent a radical mastectomy or a lumpectomy and axillary node dissection at the discretion of the treating breast surgeon. Postoperative chemotherapy was administrated at the discretion of the treating medical oncologist. Breast or chest wall irradiation was administrated in selected patients. In addition, all women with ER-positive tumors were started on tamoxifen therapy. A detailed list of all samples and clinical data is presented in Tables 1 and 2 (see also Additional files 1 and 2). Additionally, five normal breast samples from reduction mammoplasties were analyzed.
Table 1 Clinical and molecular data on breast cancer patients (training set).
Case Response Tumor reduction,% Histology pre Age ER PR BCL2 P53 KI67,% CerbB2 Grading
BC1492 NC 0 invasive lobular 50 0 0 0 1 28 0 2
BC1426 NC 0 invasive ductal + intraductal(40%) 62 1 1 400 0 2 1+ to 2+ 1 a. 2
BC1257 NC 0 invasive lobular 69 1 1 180 0 16 1+ to 2+ 2
BC1176 NC 0 invasive lobular a. tubular-lobular 47 1 1 130 0 7 0 2
BC1092 NC 0 invasive lobular a. ductal 66 1 1 0 0 10 0 2
BC1050 NC 0 invasive tubular-lobular; multifocal 60 1 1 60 0 3 0 2
BC1034 NC 0 intraductal a. invasive 43 1 1 0 0 2 0 2
BC1044 NC 0 invasive tubular-lobular 68 1 1 70 0 6 0 1
BC1466 pCR 100 invasive ductal a. intraductal(30%) 57 0 0 0 0 26 3+ 2
BC1255 pCR 100 invasive lobular 57 1 0 60 1 26 0 2
BC1254 pCR 100 medullary 62 0 0 0 0 70 1+ 3
BC1180 pCR 100 invasive lobular a. ductal 32 0 0 0 0 20 1+ 2
BC1159 pCR 100 invasive lobular a. ductal 40 0 0 70 0 70 1+ to 2+ 2
BC1042 pCR 100 non-typical medullary a. intraductal (bifocal) 38 0 0 10 0 35 0 3
BC1032 pCR 100 invasive lobular a. ductal 58 0 0 0 0 15 3+ 2
BC1443 PR/CR 94 invasive lobular a. ductal 61 1 0 10 0 30 3+ 2
BC1167 PR 0 invasive lobular 71 1 1 400 0 8 0 2
BC1162 PR 0 invasive lobular 66 1 1 70 0 7 1+ 2
BC1143 PR 0 invasive ductal a. intraductal(5%) 54 1 1 210 0 17 1+ 2
BC1138 PR 0 invasive lobular; multifocal 57 1 1 80 0 13 1+ 2
BC1100 PR 0 invasive tubular-lobular 55 1 1 140 0 4 0 1
BC1040 PR 0 invasive ductal 40 1 1 294 0 28 0 2
BC1170 PR 0 invasive lobular a. ductal bifocal 67 1 0 210 0 4 2+ to 3+ 2
BC1140 PR 10 left: invasive lobular 73 1 1 180 0 14 0 2
BC1418 PR 12 left:bifocal invasive tubular-lobular 57 1 0 300 0 7 2+ 1
BC1420 PR 15 invasive ductal a. lobular; multifocal 63 1 1 400 0 3 2+ 2
BC1491 PR 18 invasive lobular 64 1 1 300 0 7 0 2
BC1515 PR 20 invasive ductal 64 1 1 300 1 10 0 2
BC1445 PR 20 right:invasive lobular; bifocal 64 1 0 160 0 15 1+ to 2+ 2
BC1036 PR 24 invasive ductal 58 1 1 392 0 15 0 2
BC1308 PR 25 invasive lobular; multifocal 74 0 0 0 0 16 2+ 2
BC1133 PR 25 invasive ductal 53 0 0 294 0 50 0 3
BC1259 PR 32 invasive ductal a. lobular (Herd1) a. invasive ductal (Herd2) 59 1 0 300 0 16 2+ to 3+ 2
BC1498 PR 33 invasive ductal; bifocal 62 1 0 15 0 19 3+ 3
BJ_40613 PR 35 invasive ductal 61 1 1 100 0 4 2+ 2
BC1166 PR 35 invasive ductal a. lobular 45 1 1 140 0 14 1+ to 2+ 2
BC1142 PR 35 invasive lobular; bifocal 53 1 1 20 0 18 1+ 2
BC1422 PR 40 invasive ductal a. lobular; multifocal 53 1 1 360 0 2 2+ 2
BC1132 PR 40 invasive ductal 41 1 1 30 1 18 2+ 3
BC1096 PR 40 invasive ductal 46 1 1 294 0 13 0 2
BC1129 PR 42 invasive tubular-lobular 52 0 0 0 0 15 2+ 2
BC1130 PR 45 invasive lobular 42 1 1 140 0 NA 0 2
BC1131 PR 45 invasive ductal (pulmonal, ossar) 63 1 0 300 0 45 0 2
BC1256 PR 50 invasive lobular 60 0 0 0 1 17 1+ 2
BC1446 PR 50 invasive lobular; bifocal 53 1 1 140 0 35 1+ 2
BC1116 PR 53 invasive lobular 49 1 1 180 0 3 0 2
BC1415 PR 55 invasive lobular; multifocal 53 0 0 15 1 40 0 2
BC1141 PR 55 invasive lobular 52 1 1 500 0 20 0 2
BC1495 PR 60 invasive tubular a. intraductal (10%); bifocal 58 1 0 300 0 2 0 2
BC1497 PR 75 invasive lobular 42 1 1 300 0 5 2+ 2
BC1160 PR 75 invasive lobular 47 0 1 30 0 35 0 2
BC1038 PR 75 invasive ductal 35 1 1 294 0 30 0 2
BC1095 PR 85 invasive tubular-lobular 60 1 0 294 0 1 0 1
BC1024 PR 88 invasive lobular 59 1 1 70 0 18 0 2
BC1101 PR 75–85 invasive lobular; multifocal 75 1 0 180 0 10 3+ 2
BC1139 PR 89–90 invasive lobular with DCIS parts; multifocal 55 1 1 NA 0 15 NA 1
ER and PgR status were determined by immunohistochemistry. 1 – positive (>7 fmol/mg protein); 0 – negative (<7 fmol/mg protein). HER2/neu status: copies of HER2/neu gene. Status of p53 oncogene: 0 - <180 score; 1 - >180 score. NA – not available.
Table 2 Clinical and molecular data on breast cancer patients (test set).
Case Response Tumor reduction, % Histology pre Age ER PR BCL2 P53 KI67, % CerbB2 Grading
BC1843 NC 0 invasive ductal; bifocal 63 1 1 200 0 18 1+ 2
BC1850 NC 0 invasive lobular 58 1 1 300 0 6 0 2
BC1862 NC 0 invasive lobular a. intraductal 59 1 1 196 0 11 1+ 2
BC1871 NC 10 invasive lobular 46 1 1 140 1 24 0 2
BC1869 pCR 100 invasive ductal 60 0 0 0 0 50 0 3
BC1864 pCR 100 invasive a. intraductal (DCIS; 80%) 55 0 0 0 0 50 0 2
BC1421 pCR 100 invasive lobular 71 0 0 0 1 26 0 2
BC1870 cCR 100 invasive ductal 43 0 0 0 1 35 1+ to 2+ 2
BC18611 PR 40 invasive ductal a. intraductal (very small) 36 0 0 0 0 35 0 3
BC1879 PR 47 invasive ductal 37 1 1 500 0 45 0 2
BC1866 PR 40 invasive lobular 52 1 1 300 1 24 2+ to 3+ 2
BC1837 PR 90 invasive ductal 69 0 0 0 0 48 1+ 3
BC1838 PR 80 invasive lobular 59 1 0 50 0 20 1+ to 2+ 2
BC1842 PR 92 invasive ductal 68 0 0 0 0 29 1+ 2
BC1834 PR 0 invasive ductal a. intraductal (very small) 60 1 1 30 0 10 3+ 3
BC1858 PR 0 invasive lobular 62 1 1 140 0 16 1+ 2
BC1880 PR 40 invasive ductal and intraductal (5%) 62 1 1 200 0 26 0 2
BC1881 PR 62 invasive ductal 72 0 0 0 1 22 0 2
BC1849 PR 22 invasive ductal 52 1 0 200 0 19 3+ 2
BC1839 PR 10 invasive ductal 62 0 0 45 1 16 0 2
BC1513 PR 33 invasive lobular; multicentr. 60 1 1 300 0 10 0 2
BC1877 PR/NC 50 left: invasive lobular a. CLIS Type-A 53 1 1 300 0 6 1+ 2
BC1853 PR/NC 0 invasive lobular 51 1 1 400 0 14 0 2
BC1448 PR 68 medullary invasive 50 1 1 30 1 38 3+ 3
BC1134 PR/NC 5 invasive lobular 73 1 0 70 0 14 0 2
BC18402 PR 25 invasive ductal 45 1 1 60 0 35 2+ to 3+ 3
BC1848 PR 85 invasive ductal; bifocal 42 1 0 500 1 28 3+ 2
Note: 1 – This patient has received 4 × EC and, additionally, 4 × Taxol; 2 – this patient has received 3 × EC and, additionally, 3 × FEC.
Immunohistochemistry (IHC)
Hematoxilin/eosin-stained sections from tumor specimens were examined to assess the relative amounts of tumor cells, benign epithelium, stroma, and lymphocytes. Standard clinical parameters, such as estrogen receptor-α (ER), progesterone receptor (PgR), proliferation marker (ki-67), tumor suppressor p53, regulator of apoptosis Bcl2, protooncogene cerbB2/HER2neu, epidermal growth factor receptor (EGFR) were assessed according to routine bio- and/or immunohistochemical methods.
Immunohistochemical staining was performed on 5-μm paraffin sections. Sections were deparaffinized in xylene and re-hydrated. Epitope retrieval was performed by heat induction in Target Retrival Solution pH6.1 (DAKO, DakoCytomation GmbH, Hamburg, Germany). Tissues were blocked for endogenous peroxidase in a 0.3% H2O2 solution for 15 min. Monoclonal antibodies (ERa: ER1D5 DAKO 1:35, PR: PGR636 DAKO 1:50, bcl-2: Clone 124 1:200 DAKO, EGFR: 31G7 CYTOMED 1:20, cerb2/Her-2/neu polyclonal DAKO 1:250, and ki67 (MIB-1) DAKO 1:200) were used for specific epitope detection. The ChemMate DAKO peroxidase/DAB Detection Kit was used for linking and staining. Slides were counterstained with methyl green and coverslipped with entelan. Histologic scores were calculated by multiplying color intensity (range 0 to 5) with proportion of cells staining positive.
Response Criteria
Response to the treatment followed the Unio Internationale Contra Cancrum criteria [23]. pCR (pathological diagnosis based complete responders), was defined as absence of invasive carcinoma in the breast by the examining pathologist and lack of lymph nodal involvement. cCR (clinical complete responders) was defined as clinical absence of invasive carcinoma of the breast. This parameter was used as a surrogate for pCR in one occasion when a patient declined post-PST surgical excision. PR (partial responders) was determined as a reduction in the tumor mass of both perpendicular dimensions ranging from 10% to 75% of the initially measured tumor size based on dynamic contrast-enhanced magnetic resonance imaging or magnetic resonance tomography (MRT), and sometimes on both, MRT and ultrasonography. NC (non-responders or no change) was defined as an absence of tumor reduction or an increase in tumor size (stable or progressive disease). The percent of tumor reduction (Table 1 and 2) was calculated as a ratio between pathologic tumor size [cm] after neoadjuvant chemotherapy at the time of surgical excision compared to the size of tumors defined clinically at the time of diagnosis. More details are available as Additional files 3 and 4 from the JTM Web page.
RNA Preparation and Microarray Analysis
Total RNA was extracted from cell lysates of ground tissue and subsequent purification with RNeasy mini spin columns (Qiagen, Hilden, Germany). Subsequent washing and elution steps were performed according to manufacturer's instructions. High-quality RNA was obtained as suggested by well-preserved 28S and 18S ribosomal RNA bands (present in an approximately 2:1 intensity ratio), along with A260/A280 ratios between 1.8 and 2.0. Quality and integrity of total RNA was tested with a Bioanalyzer 2100 (Agilent Technologies Inc., Palo Alto, CA, USA). Gene expression analysis was performed on an Affymetrix Human Genome U133A GeneChip platform containing 22,283 probes. Preparation and processing of labeled and fragmented cRNA targets, hybridization and scanning procedures were carried according to the manufacturer's protocol (Affymetrix, Santa Clara, CA, USA) [24]. Starting material for labeling consisted of 5 μg of total RNA from each tumor specimen. Labeling was limited to one cycle of in vitro transcription. Thus, starting with 5 μg of total RNA, approximately 50 to 60 μg of amplified RNA (cRNA) could be generated, which could be used in multiple microarray experiments. The cRNA was quantified by Agilent Nano Chip technology and evaluated for size relative to pure polyadenylated RNA. Fifteen micrograms of cRNA were subsequently used for hybridization. After washing and staining arrays were scanned by Gene Array scanner 2500 (Affymetrix). Hybridization intensity data were automatically acquired and processed by Affymetrix Microarray Suite 5.0 software. The expression level (average difference) of each gene was determined by calculating the average of differences in intensity (perfect match-mismatch) between its probe pairs as described elsewhere [25]. Scans were rejected if the scaling factor exceeded 2 or "chip surface scan" revealed scratches, specks or gradients affecting overall data quality (Refiner, GeneData AG, Basle, Switzerland).
Quantitative Real-Time PCR.
Aliquots of total RNA used for GeneChip expression analysis were used for quantitative RT-PCR with an ABI PRISM 7900 Sequence Detection System (Applied Biosystems, Foster City, CA, USA). cDNA for PCR amplification was generated by oligo dT primed reverse transcription (Superscript First Strand System, Invitrogen Corporation, Carlsbad, CA, USA) including DNAse I treatment. Primers and probes were designed with the Primer Express software (Applied Biosystems, Foster City, CA, USA) and spanned the same gene region of the respective Affymetrix probe set. Labeled oligonucleotides were obtained from Eurogentec s.a. (Liege, Belgium). Absolute copy numbers were normalized according to GAPDH as a reference gene. The primer/probes were prepared by mixing 25 μl of 100 μM stock solution "upper primer", 25 μl of 100 μM stock solution "lower primer" plus 12,5 μl of the 100 μM stock solution TaqMan-probe (FAM/TAMRA) and adjusted to 500 μl with H2O (Primer/probe-mix). PCR reactions using cDNA generated from 1.5 ng total RNA were performed in duplicates in a volume of 10 μl. This included TaqMan universal Mix (Eurogentec s.a.) according to manufacturer's protocol in a 384-well format and 1 μl of the P&P mix. Thermal cycler parameters were 2 min at 50°C, 10 min at 95°C and 40 cycles, each consisting of a 15 s denaturation step at 95°C and a 1 min annealing/extension step at 60°C. Relative abundance of a gene transcript was calculated either by the ΔΔCt method or by arbitrarily defined RNA copy number estimates at a Ct = 24 as 106 copies. Subsequent analysis included normalization steps such as median centering and per gene median division.
Data Filtering
Fifty-six primary breast cancer and 5 non cancerous breast tissue samples were analyzed as a training set for marker discovery. Raw data was acquired using Microsuite 5.0 software from Affymetrix and normalized following the standard practice of scaling the average of all gene signal intensities to a common arbitrary value (TGT = 100). Gene expression data were stored including P-value, as generated by Microsuite 5.0 software, for quality assessment of individual measurements for each transcript. The data-file was imported into Expressionist Analyst software package (GeneData AG) for further statistical analysis. To enhance quality we excluded gene probe sets for the following reasons. Fifty-nine probe sets corresponding to hybridization reference (housekeeping genes, etc.) as identified by Affymetrix were removed with the exception of GAPDH and β -actin, for which a 3' biased probe set was included. One hundred genes, whose expression levels are routinely used for normalization of the HG-U133A and HG-U133B GeneChip versions [26], were also removed from the analysis. These genes reflect a very homogenous expression pattern among several human tissues and could therefor be categorized as "house keeping" genes. Genes with potentially high levels of noise (81 probe sets) frequently observed with low absolute expression values (below 30 relative signal units (RSU) in all experiments) were also removed. The remaining genes were preprocessed to eliminate those (3,196) whose signal intensities were not significantly different (P > 0.04) from their background levels and thus labeled as "Absent" by MicroSuite 5.0. To apply a higher stringency to the data, we eliminated genes whose significance levels (P < 0.04) were only reached in 10% of the breast cancer samples (3,841 probe sets). Data for the remaining 15,006 probe sets were used for the subsequent analysis.
Statistical Analysis
For the analysis we applied a similar strategy to the one applied by Wang E. and colleagues [27] to predict immune responsiveness of melanoma metastases. Genes differentially expressed by lesions characterized by different responsiveness to PST were identified with the nonparametric Wilcoxon rank sum test, two-sample independent Students't-test and Welch test. Probes were ranked in order of significance (SUM-Rank test) combining the results of these tests using as a cut-off P-value < 0.05 and fold change between groups >2. The Kruskal-Wallis and ANOVA tests were applied when two distinct groups (i.e. pCR vs NC) with extreme response patterns where studied in the presence of a third intermediate group (PR). All statistical tests were two-tailed. Principal components analysis (PCA) and hierarchical clustering were applied for data display and structural analysis and in certain steps for dimensional (probe set) reduction. All these different tools were used as implemented in the GeneData Expressionist Analyst software package and were only modified by selection of starting parameters and appropriate distance weight matrices. Additionally, partial least squares discriminant analysis for multivariate data (PLS-DA) with SIMCA-P software (Umetrics, AB, Umea, Sweden) was used.
Results
Preliminary analysis about ER status and inflammation
Previous studies [28-31] reported that patients with negative estrogen receptor (ER) status respond better to PST compared to those with a positive one. In addition, PgR1 gene expression may affect outcome. Furthermore, patients with ER-negative tumors suffer shorter disease-free and overall survival [32-34]. ER status is also associated with a characteristic gene expression profile independent of other clinical/pathological parameters [35,36]. Therefore, we separately studied genes known to be associated with ER signaling. We analyzed previously the expression profile of breast cancer in two patient cohorts with positive and negative ER status (not part of this study). The complete gene list and expression data within the two cohorts is available (Additional file 5). We identified 828 Affymetrix probe sets by ANOVA and t-test (P < 0.005) with a median fold change of 1.2 or above. Analysis of the 828 ER-related signatures in the 56 tumors from the present study correlated well with ER-α status by immunohistochemistry. To avoid the influence on clinical outcome of ER-specific signatures and identify alternative, ER-independent predictors of response and survival, these genes were excluded
We also excluded genes related to immune function since we could not predict the effect that the heterogeneity of immune infiltrates might bear on the transcriptional profile of individual lesions. Immune genes (1,025) were identified and excluded. The complete list of excluded genes is available (Additional file 6). Many of the excluded genes are members of immunoglobulin families. The final data set contained 13,145 probe sets. Although there is currently plenty of interest about the impact of immunity as a predictor of clinical outcome, this was beyond the purpose of this work and will be considered in a subsequent manuscript.
Determination of predictor genes
Starting with the training cohort, we built response subclasses based on the post surgical clinicopathological examination. Eight of the 56 training cases experienced a pCR and eight progressed (NC). To identify the most predictive genes for each class we implemented a comparison schedule for the training set as follows:
(I) PR vs. NC (n = 40 vs. 8); (II) pCR vs. PR (n = 8 vs. 40), and (III) pCR vs. NC (n = 8 vs. 8). These comparisons were carried out by non-parametric t-test, Welch, Wilcoxon, Kolmogorov-Smirnov tests using the Expressionist Analyst software (GeneData AG). Differentially expressed genes were considered those reaching a significance cut off P-value of < 0.05 in all tests; 2,301 were identified. Additional restrictions were then applied (at least 2-fold change of median expression level and average expression more than 30 RSU (relative signal units) in all three groups) resulting in only 1,512 probe sets useful for further analyses.
For the "three-group tests" (pCR vs. PR vs. NC) statistical significance was measured with the Kruskal-Wallis and one-way ANOVA tests with a cut off P-value of < 0.05 identifying 414 probe sets. Overlap of the gene lists (1,512 probe sets and 414 probe sets) by Venn diagram analysis qualified 397 probe sets. This high stringency potentially eliminated genes of interest but decreased the false discovery rates of random selected genes at P-value cut off <0.05. PCA using all predefined tissue classes: non cancerous breast tissue pCR, cCR, PR and NC was applied to the 397 probe sets. Separation of pCR and cCR tumors on the one side and NC samples on the other was defined by 2 most distinguishing components. We applied a cutoff on the correlation matrix of the PCA and filtered genes at < -0.4 and > 0.4. This sorted out 325 by eliminating 72 probe sets.
We then excluded from the remaining 325 genes those known to be specifically expressed in blood vessels, adipocytes, and muscle tissues based on differential expression profiling of tumor cells and normal cells after their separation by laser capture microdissecction or by comparing breast tumor's gene expression profiles with expression profiles of normal blood vessels, adipose and muscle tissue samples reducing the number of genes by 61. The list of the excluded 61 genes is available (Additional file 7). Rank ordering of the remaining 264 genes' significance was determined by SUM-Rank test for all samples and compared to the original 13,145 genes.
In addition, two classifier genes were identified (FHL1 and CLDN5) highly discriminative between most "normal" tissue samples and all breast cancer samples analyzed. Whereas these genes are expressed at very high levels in normal breast tissue their low level expression was rarely detected in malignant breast samples. We combined these 2 genes with 57 most discriminative genes from 264 filtered probe sets (Additional file 8). Such combination allows simple and fast separation of normal tissue samples from malignant ones, which might be useful for routine clinical diagnostics. A detailed table containing raw data for 59 genes and 83 tumors is available as supplemental information (Additional files 9 and 10).
Validation on independent cases
The determined classifiers could be subdivided into three categories: those genes/probe sets capable to distinguish between (a) normal breast and breast cancer tissues (2 genes, FHL1 and CLDN5), (b) pCR or cCR from unfavorable outcomes (PR or NC) (31 probe sets or "favorable response signature"), and (c) NC and PR (26 probe sets or "poor response signature"). We expected that both signatures, favorable and poor, would separate the two most extreme classes pCR and NC and effectively recognize the respective expression patterns. These classifiers were challenged against samples from an independent test cohort (n = 27; 4 pCR, 4 NC, 19 PR; see Table 2 or Additional file 2). Classification was performed by k-NN (k = 3) following a three step decision tree based on the 59 genes listed above. All 27 tumor samples were correctly qualified as cancerous tissues using the two-gene signature (FHL1 and CLDN5). Whit the "favorable response signature" a group of 8 tumor samples was classified as CR or PR. Finally, the rest of the tumors were classified as NC or PR by the "poor response signature". There were four potentially wrong classified cases. Results of classification for the test cohort are shown in Table 3. Summarized results of validation, as well as sensitivity, specificity positive and negative predictive values (PPV and NPV, respectively) for each class are shown in Table 4.
Table 3 Comparison of predicted and pathologic response in test set.
Case Tumor reduction,% Response, pathologic Predicted response PCA cross validated by k-NN Predicted response PLS-DA model 1 Predicted response PLS-DA model 2
BC1843 0 NC NC NC PR
BC1850 0 NC NC NC NC
BC1862 0 NC NC NC PR
BC1871 10 NC NC NC PR
BC1869 100 pCR CR CR CR
BC1864 100 pCR CR CR CR
BC1421 100 pCR CR CR CR
BC1870 100 cCR CR CR CR
BC1861 40 PR PR CR CR
BC1879 47 PR PR NC NC
BC1866 40 PR PR CR CR
BC1837 90 PR CR CR CR
BC1838 80 PR PR NC NC
BC1842 92 PR PR PR PR
BC1834 0 PR PR PR CR
BC1858 0 PR NC NC NC
BC1880 40 PR PR PR PR
BC1881 62 PR NC PR PR
BC1849 22 PR NC NC PR
BC1839 10 PR NC NC PR
BC1513 33 PR PR NC PR
BC1877 50 PR PR NC NC
BC1853 0 NC NC NC NC
BC1448 68 PR CR CR CR
BC1134 5 NC NC NC NC
BC1840 25 PR NC NC NC
BC1848 85 PR CR CR CR
Note: complete pathologic response shown as CR when predicted.
Table 4 Summarized results of validation on the test cohort.
Predicted response k-NN Predicted response PLS-DA; model 1 Predicted response PLS-DA; model 2
predicted CR 7 9 10
other 20 18 17
predicted PR 9 4 9
other 18 23 18
predicted NC 11 14 8
other 16 13 19
Sensitivity CR 100 100 100
Specificity CR 87 78 74
PPV 57 44 40
NPV 100 100 100
Sensitivity PR 53 24 35
Specificity PR 100 100 100
PPV 100 100 67
NPV 56 43 39
Sensitivity NC 100 100 50
Specificity NC 76 62 76
PPV 55 43 38
NPV 100 100 84
PCA and Hierarchical Clustering
A PCA plot was created displaying the position of each tumor sample from training and test cohorts (83 tumors) using three main Eigenvectors (Fig. 1A). The PCA was performed with the set of 57 response predictive genes for illustration purpose. The two most disparate response groups (pCR and NC) are clearly separated with the exception of one NC case, BC1492, which clusters with pCR tumors. This plot is consistent with k-NN cross-validation results for training cohort, which defined that NC case BC1492 as complete response. Hierarchical clustering of all 83 tumors and 57 response predicting genes is shown in Fig. 1B and 1C: eleven of twelve pCR tumors are organized in one sub-branch of the sample dendrogram and NC tumors are placed into the separate dendrogram branch.
Figure 1 Clustering of gene expression data for 57 genes from 83 breast tumors corresponding training and test cohorts. A. PCA analysis of response groups and gene expression. The visualization of high-dimensional data in three-dimensional principal components. Individual samples from training and test cohorts are labeled according to three response groups: green and light green – pCR; yellow and light yellow – PR; red and light red – NC. The distance between samples reflects their approximate degree of correlation. B. Hierarchical clustering presents the clustered samples in columns and the clustered 57 genes in rows. A color representation of gene expression levels is shown with the scale on the left side. The 57 genes used fir both clustering methods were obtained by multi-step statistical approach, as described in 'A predictor gene set determination' section of Results. C. An enlarged version of sample dendrogram, which reflects similarities in their expression profiles.
Partial least squares discriminant analysis (PLS-DA)
Direct linear discriminant analysis was applied to compare the previous results and test the potential of our first classifier model. PLS-DA applies well to the large number of predictors and the multicollineality. Supervised PLS-DA analysis uses independent (expression levels) and dependent variables (classes) for class comparison applying multivariate statistical methods such as soft independent modeling of class analogy (SIMCA) and partial least squares modeling with latent variables to allow simultaneous analysis of all variables [37-42]. Additionally, PLS-DA provides a quantitative estimation of the discriminatory power of each descriptor by means of VIP (variable importance for the projection) parameters. VIP values represent an appropriate quantitative statistical parameter ranking descriptors (gene expression values) according to their ability to discriminate different classes.
PLS-DA was carried out on the original 13,145 probe sets that passed the QC filtering process in the training cohort. Although this process may lead to an over parameterized model with poor prediction properties, it provides a preliminary assessment of the most important discriminative variables. Two independent models were tested each consisting of two classes: model 1 (class 1 – pCR, class 2 – NC, and PR cases were excluded); model 2 (class 1 – pCR, class 2 – NC and PR together). The model with three classes (pCR, NC and PR) demonstrated rather poor prediction power being strongly dependent on the definition of partial response (Table 1). Possibly the comparison of pathological estimates (post treatment) compared to clinical measurements (pre-treatment) over estimated the tumor reduction measurements and biased the attribution of samples as PR rather than NC.
Those variables satisfying the criteria of expression levels above 60 RSU (as a mean value in at least one of each sample group, pCR and NC), ratio (pCR/NC) >1.9 or <0.55, and VIP of >1.9 were retained. Figure 2 shows a scatter plot of samples from the training set grouped according to the two components for either PLS in model 1 (96 probe sets; Fig. 2) or in the model 2 (90 probe sets; Fig. 3) after the second iteration. The numbers next to the symbols are the sample IDs as detailed in Table 1. It is apparent that pCR and NC samples are clearly discriminated. However, the results of permutation tests for both models (data not shown) demonstrated that both reduced models were still over-parameterized. Thus, we retained the 20 probe sets deduced from model 1 and 20 probe sets from model 2 with highest VIP values. In both cases, models performed much better than expected by chance.
Figure 2 PLS discrimination according to tumor response class using the variables selected by PLS (VIP > 1.9) and ratio (pCR/NC) > 1.9 or < 0.55. Model 1 (PR cases were deleted; class 1 – pCR, black dots; class 2 – NC, black squares); 96 probe sets (cDNAs) retained.
Figure 3 PLS discrimination according to tumor response class using the variables selected by PLS (VIP > 1.9) and ratio (pCR/NC) > 1.9 or < 0.55. Model 2 (class 1 – pCR, black dots; class 2 – NC, black squares and PR, gray triangles); 90 probe sets retained.
Two groups of selected probe sets were compared and nine probe sets were found to be represented in both lists, which were deduced from model 1 and 2. A combined list containing 31 probe sets was used for model validation (Table 5) by applying PLS-DA to the second, independent group of tumors (n = 27; Table 2) to test the discriminative power of the final gene list. The results are presented in Table 3. PLS-DA classified partially responding tumors with good (> 60% tumor shrinkage) or very poor response to therapy as complete response (e.g., BC1837, BC1848, BC1448) or no response (e.g., BC1877, BC1134, BC1840) respectively. This observation indicates that for further studies the monitoring of tumor shrinkage during PST is pivotal to correctly judge the final response classification and it might have been the major limitation of this study. Both statistical approaches, one that yielded the 59 gene and PLS-DA were compared and identified 19 genes in common. PLS-DA alone demonstrated a lower predictive power compared to the first multi-step analysis combined with k-NN classification.
Table 5 Top 31 genes extracted from two different models from PLS-DA SIMCA.
Gene Symbol Gene Description Ref. Sequences Unigene ID
KPNA2 nuclear localization sequence receptor hSRP1alpha, karyopherin alpha 2 (RAG cohort 1 importin alpha 1) NM_002266 4504896
HDAC2 transcriptional regulator homolog RPD3 histone deacetylase 2 similar to yeast RPD3 NM_001527 4557640
PRKAB1 5-AMP-activated protein kinase beta-1, non-catalytic subunit NM_006253 18602783
IMPDH2 inosine monophosphate dehydrogenase (IMPDH2) NM_000884 4504688
YR-29 hypothetical protein clone YR-29 NM_014886 7662676
CD2BP2 CD2 antigen (cytoplasmic tail)-binding protein 2 NM_006110 5174408
FHL2 heart protein (FHL-2) four and a half LIM domains 2 NM_001450 4503722
DDB2 damage-specific DNA binding protein p48 subunit (DDB2; 48 kD) NM_000107 4557514
ASNS asparagine synthetase NM_001673 4502258
XPA XPAC protein xeroderma pigmentosum complementation group A NM_000380 4507936
PLA2G7 LDL-phospholipase A2 phospholipase A2 group VII (platelet-activating factor acetylhydrolase plasma) NM_005084 4826883
BTBD2 BTB (POZ) domain containing 2 hypothetical protein FLJ20386 EST NM_017797 8923361
CCNG1 cyclin G1 clone MGC:6 NM_004060 -
PDHB pyruvate dehydrogenase E1-beta subunit d pyruvate dehydrogenase (lipoamide) beta NM_000925 4505686
MKI67 mki67a (long type)antigen of monoclonal antibody Ki-67 NM_002417 4505188
TNRC15 KIAA0642 protein trinucleotide repeat containing 15 AL_045800 18550089
RPL17 ribosomal protein L17 NM_000985 14591906
GNG12 DKFZp586B0918 (from clone DKFZp586B0918) NM_018841 -
RPL17 ribosomal protein L17 NM_000985 14591906
DKC1 Cbf5p homolog (CBF5) dyskeratosis congenita 1 dyskerin nucleolar protein NM_001363 15011921
DCTN4 dynactin p62 subunit dynactin 4 (p62) NM_016221 14733974
FLJ20273 RNA-binding protein NM_019027 9506670
FLJ11323 hypothetical protein EST NM_018390 8922994
MGC11242 hypothetical protein MGC11242 ESTs NM_024320 13236560
SRR serine racemase Homo sapiens cDNA NM_021947 8922495
ARL3 48c8 ADP-ribosylation factor-like 3 EST NM_004311 4757773
CCNB2 cyclin B2 NM_004701 10938017
MAD2L1 MAD2 protein MAD2 (mitotic arrest deficient yeast homolog)-like 1 NM_002358 6466452
LIG1 membrane glycoprotein LIG-1d NM_015541 18554950
PMSCL1 polymyositisscleroderma autoantigen 1 (75 kD) EST NM_005033 4826921
APBB2 amyloid beta (A4) precursor protein-binding, family B NM_173075 18557629
Note: – Genes in common with EC predictor (Table 3).
Confirmation of expression measurements by real-time RT-PCR
Real-time RT-PCR (qPCR) measurement of gene expression levels on the same RNAs used for GeneChip hybridization experiments obtained from 32 breast tumors from training and test cohorts was performed on 46 genes selected from those presented in Table 3. Primer and probes were designed in regions within or close to the target region of the GeneChip oligonucleotides. A Ct value of 24 was empirically considered to represent 106 RNA copies per well based on spiking experiments. Raw data from real-time RT-PCR are presented in Supplemental Data on the Web page, as above, along with Affymetrix GeneChip's data. Relative expression as measured by the GeneChip was compared with qPCR results adjusting the median expression of all 46 genes within one sample to 100 relative units. To detect the relative difference in expression between samples for each gene, all measurements were divided by the median expression of this gene. This median normalization was carried out for both platforms independently. Raw and normalized data for Affymetrix and TaqMan platforms are shown in Additional file 11. In order to compare the individual measurements and the relative abundance of each transcript we preformed hierarchical clustering with the data generated with the GeneChip system. We performed this clustering (Fig. 4) with a correlation matrix on the samples as well as on the genes while the distance measurement was carried out with an average weight matrix. Once having the cluster of the GeneChip data in place we ordered all samples and all genes for the qPCR data in the same order as derived from the previous clustering. This operation resulted in very similar heat-maps as depicted in Figure 4 with an overall correlation of R2 = 0.73. We also performed independent clustering of the qPCR data (Fig. 5), which resulted in similar correlation trees.
Figure 4 Confirmation of expression measurements by real-time RT-PCR. GeneChip median expression for 46 genes from Table 3 within one sample was adjusted to 100 RLU. Then, all measurements were median centered for each gene. Hierarchical clustering algorithm was applied to median normalized expression data of 46 genes from 39 tumor samples from training and test cohorts. Hierarchical clustering presents the clustered samples in columns and the clustered 46 genes in rows. A color representation of gene expression levels is shown with the scale on the left side (pCR represented in green, NC represented in red). Clustering of the data was performed according a correlation analysis with an average distance determination. The threshold Ct values obtained in real-time RT-PCR were converted into an arbitrary RNA-copy number Ct value of 24, which was then empirically settled to 106 RNA copies per well. These measurements were median centered, as for microarray data. All data for samples and genes were ordered according to the hierarchical structure of the microarray data set in Fig. 3A for Affymetrix platform.
Figure 5 Confirmation of expression measurements by real-time RT-PCR Independent clustering of the qPCR data for both, Affymetrix and TaqMan platforms resulted in similar correlation trees.
Discussion
The aim of this study was to identify a multigene predictor of response to EC in a PST. Several recent studies demonstrated that gene expression profiling can predict response in the neoadjuvant setting [43-47]. Since the patient-specificity of such predictors remain questionable [48], further attempts devoted to the understanding of the process (es) underlying responsiveness to systemic therapy are of obvious importance.
Primary systemic chemotherapy is often being used to downstage large and locally advanced breast tumors in patients prior to surgery. There is increasing evidence that response and, particularly, complete response to neoadjuvant chemotherapy predicts improved disease-free and overall survival [49-51]. Unquestionable, pathological complete response (pCR) is not a synonym for cure, since a risk remains for metastatic disease. But such risk is decreased in association with the down-staging of the primary tumor and the achievement of a node negative status confirmed at the time of surgery. Therefore, it is reasonable to suggest that a good response to neoadjuvant therapy may correspond to survival benefit.
The role of biological characteristics and/or molecular markers as predictor of sensitivity to specific treatments has been extensively studied [52-56]. However, their role in response prediction remains unclear. Results from different studies are often contradictory and, consequently, no individual biological marker can be reliably used clinically for prediction of response to chemotherapy [57,58].
The patients analyzed in this study were part of a much larger cohort (n = 319) receiving treatment with EC-based PST. We have observed in this patient population that age, histologic grade, estrogen receptor (ER), progesterone receptor (PgR), levels of oncogene B-cell leukemia 2 (Bcl2), proliferation-related Ki-67 antigen (ki-67), and epidermal growth factor receptor (EGFR) expression were related to response in a univariate analysis, also confirmed by Colleoni et al. [33] in preoperative settings. However, in a multivariate model it was only ki-67 expression that predicted a better pathological response (P = 0.011), and this factor was linked to the patient's age [59]. Thus, a true predictive marker that could be measured by routine methods (e.g., IHC) to identify patients likely to benefit from neo-adjuvant EC remains elusive.
Several studies on breast cancer assessed classifiers predictive of survival [60-65]. A Dutch group reported 70 genes predictive of disease recurrence in women with lymph-node-negative primary breast cancer and confirmed the findings in a second study comprising additional 198 patients [65]. This study could assign some women to a low-risk category beyond the discriminating power of conventional histopathological criteria.
However, the concordance among different studies on survival of breast cancer patients is low. Data inconsistency can be particularly explained by the use of different microarray technologies and different patients' demographics. In addition, subtleties in data analysis may explain some discrepancies since there is no standardize method for expression data analysis when a large number of data points per individual are studied in relatively low sample populations.
In this study, we accurately discriminated samples that had a high tumor content from normal breast tissue based on the previous demonstration that FHL1 and CLDN5 can serve as such predictors. Then, we identified predictors in cancer tissues from primary tumors by identifying genes capable of segregating two distinct classes of tumors according to response to treatment (pCR vs NC). "Favorable outcome signature" could predict complete remission of a primary tumor with >90% sensitivity. Some genes found to be highly expressed in pCR samples belong to the "biological topic" of mitosis and cell proliferation (e.g. MAD2L1, CCNB2). This is concordant with the observations we [59] and others made on the ki-67 expression and the negative ER status in responding tumors [66]. Possibly, actively dividing tumors, either driven by the lack of hormonal control or by other signals such as via the insulin receptor pathway may respond best. The "poor outcome signature" distinguishing tumors unlikely to respond to PST included DDB2 or XPA, involved in DNA damage repair which makes perfect logical sense. The highest predictive value was sought in a stepwise manner by comparing pCR to NC cases and comparing predictors of each group by multi-step statistical approaches and k-NN (k = 3) validation. This classifier could predict with a remarkable level of accuracy a pathological response in the subsequent cohort of 27 patients used for validation. It is also possible that there were mis-assignments of responding cases especially in borderline cases that responded with minor changes in size or bifocal tumors. Ultra-sound imaging applied for size determination prior to chemotherapy might not be comparable to the accurate measurements that pathologists can make on resected samples. Thus, 10 cases in the training set considered as PR might not have qualified if comparable measurements could be used before and after therapy. This undefined error might have partially affect our statistical analysis decreasing the sensitivity of the model adopted (i.e., predict many NC cases as PR and vice versa).
We also observed that application of different statistical algorithms to the data analysis lead to the extraction of overlapping predictor signatures (19 of 57 genes were in common). Although some of the genes identified by the PLS-DA could have been dismissed by the stringent filtering criteria applied, both analytical approaches could predict pCR. Accuracy of NC prediction could only be achieve through the stepwise identified signatures. Further in depth interpretation of the biological processes associated with the genes identified statistically will probably enhance the robustness of our findings in the future [67-69] [70].
We attempted to override the risk of overfitting of the model based on the training data (i.e., finding a mass of less relevant genes that may lead to the loss of a few relevant ones). The prediction accuracy was relatively high but was limited by the number of validation events (pCR or NC) so far analyzed suggesting that improved selection predictor genes among the ones identified based on a larger validation study may increase the accuracy of our findings and as a consequence their clinical value. We are currently collecting samples for a second validation cohort receiving EC based PST under similar conditions at an independent institution.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
OM and MM contributed equally to this work. All authors read and approved the final version of the manuscript.
Supplementary Material
Additional File 1
contains clinico-/pathological information on patients in training cohort.
Click here for file
Additional File 2
contains clinico-/pathological information on patients in test cohort.
Click here for file
Additional File 3
contains more clinico-/pathological information on patients in training cohort.
Click here for file
Additional File 4
contains more clinico-/pathological information on patients in test cohort.
Click here for file
Additional File 5
contains the complete gene list and expression data within the two cohorts of patients with ER positive and ER negative status.
Click here for file
Additional File 6
contains complete list of genes related to immune system (1,025), which were excluded from the analysis.
Click here for file
Additional File 7
contains list of 61 genes. Those genes were excluded from the analysis by comparing breast tumor's gene expression profiles with expression profiles of normal blood vessels, adipose and muscle tissue samples.
Click here for file
Additional File 8
contains EC predictor.
Click here for file
Additional File 9
contains raw data for 59 genes and 83 tumors in training and test cohorts, respectively.
Click here for file
Additional File 10
contains raw data for 59 genes and 83 tumors in training and test cohorts, respectively.
Click here for file
Additional File 11
contains Affymetrix and TaqMan raw and normalized data for 46 genes and 32 primary breast tumors (confirmation study of expression measurements by real-time RT-PCR).
Click here for file
Acknowledgements
We thank Dr. Michael Korenberg, an editor of a new volume, tentatively entitled "Microarray Data Analysis: Methods and Applications", in an ongoing series "Methods in Molecular Biology" by Humana Press, USA for reading and helpful discussion of the statistic methods applied for the microarray data analysis in the present manuscript.
==== Refs
Jemal A Tiwari RC Murray T Ghafoor A Samuels A Ward E Cancer statistics CA Cancer J Clin 2004 54 8 29 14974761
Waterworth A Introducing the concept of breast cancer stem cells Breast Cancer Res 2004 6 53 54 14680485 10.1186/bcr749
Perou CM Sorlie T Eisen MB van de Rijn M Jeffrey SS Rees CA Molecular portraits of human breast tumours Nature 2000 406 747 752 10963602 10.1038/35021093
Early Breast Cancer Trialists' Collaborative Group Multi-agent chemotherapy for early breast cancer Cochrane Database Syst Rev 1, CD000487 2002
Early Breast Cancer Trialists' Collaborative Group Tamoxifen for early breast cancer Cochrane Database Syst Rev 1, CD000486 2001
Russo G Zegar C Giordano A Advantages and limitations of microarray technology in human cancer Oncogene 2003 22 6497 6507 14528274 10.1038/sj.onc.1206865
Goldsmith ZG Dhanasekaran N The microrevolution: applications and impacts of microarray technology on molecular biology and medicine (review) Int J Mol Med 2004 13 483 495 15010847
Zhang W Laborde PM Coombes KR Berry DA Hamilton SR Cancer genomics: promises and complexities Clin Cancer Res 2001 7 2159 2167 11489789
Modlich O Prisack HB Munnes M Audretsch W Bojar H Immediate gene expression changes after the first course of neoadjuvant chemotherapy in patients with primary breast cancer disease Clin Cancer Res 2004 10 6418 6431 15475428
Garces CA Cance WG Neoadjuvant chemotherapy of breast cancer Am Surg 2004 70 565 569 15279176
Scholl SM Asselain B Palangie T Dorval T Jouve M Garcia Giralt E Neoadjuvant chemotherapy in operable breast cancer Eur J Cancer 1991 27 1668 1671 1782080
Green M Hortobagyi GN Neoadjuvant chemotherapy for operable breast cancer Oncology (Huntingt) 2002 16 871 898 12164555
Brenin DR Morrow M Breast-conserving surgery in the neoadjuvant setting Semin Oncol 1998 25 13 18 9566202
Fisher B Brown A Mamounas E Wieand S Robidoux A Margolese RG Effect of preoperative chemotherapy on local-regional disease in women with operable breast cancer: findings from National Surgical Adjuvant Breast and Bowel Project B-18 J Clin Oncol 1997 15 2483 93 9215816
Schwartz GF Hortobagyi GN Proceedings of the consensus conference on neoadjuvant chemotherapy in carcinoma of the breast Cancer 2004 100 2512 2532 15197792 10.1002/cncr.20298
Bonadonna G Veronesi U Brambilla C Ferrari L Luini A Greco M Primary chemotherapy to avoid mastectomy in tumors with diameters of three centimeters or more J Natl Cancer Inst 1990 82 1539 1545 2402015
Kuerer HM Newman LA Smith TL Ames FC Hunt KK Dhingra K Clinical course of breast cancer patients with complete pathologic primary tumor and axillary lymph node response to doxorubicin-based neoadjuvant chemotherapy J Clin Oncol 1999 17 460 469 10080586
Brittenden J Heys SD Miller I Sarkar TK Hutcheon AW Needham G Dietary supplementation with L-arginine in patients with breast cancer (> 4 cm) receiving multimodality treatment: report of a feasibility study Br J Cancer 1994 69 918 921 8180024
Chollet P Charrier S Brain E Cure H van Praagh I Feillel V Clinical and pathological response to primary chemotherapy in operable breast cancer Eur J Cancer 1997 33 862 866 9291806 10.1016/S0959-8049(97)00038-5
Chollet P Amat S Cure H de Latour M Le Bouedec G Mouret-Reynier MA Prognostic significance of a complete pathological response after induction chemotherapy in operable breast cancer Br J Cancer 2002 86 1041 1046 11953845 10.1038/sj.bjc.6600210
Pierga JY Mouret E Laurence V Dieras V Savigioni A Beuzeboc P Prognostic factors for survival after neoadjuvant chemotherapy in operable breast cancer: the role of clinical response Eur J Cancer 2003 39 1089 1096 12736108 10.1016/S0959-8049(03)00069-8
Cance WG Carey LA Calvo BF Sartor C Sawyer L Moore DT Long-term outcome of neoadjuvant therapy for locally advanced breast carcinoma: effective clinical downstaging allows breast preservation and predicts outstanding local control and survival Ann Surg 2002 236 295 302 12192316 10.1097/00000658-200209000-00006
Sotiriou C Powles TJ Dowsett M Jazaeri AA Feldman AL Assersohn L Gene expression profiles derived from fine needle aspiration correlate with response to systemic chemotherapy in breast cancer Breast Cancer Res 2002 4 3 10.1186/bcr433
Chang JC Wooten EC Tsimelzon A Hilsenbeck SG Gutierrez MC Elledge R Gene expression profiling for the prediction of therapeutic response to docetaxel in patients with breast cancer Lancet 2003 362 362 369 12907009 10.1016/S0140-6736(03)14023-8
Ayers M Symmans WF Stec J Damokosh AI Clark E Hess K Gene expression profiles predict complete pathologic response to neoadjuvant paclitaxel and fluorouracil, doxorubicin, and cyclophosphamide chemotherapy in breast cancer J Clin Oncol 2004 22 2284 2293 15136595 10.1200/JCO.2004.05.166
Buchholz TA Stivers DN Stec J Ayers M Clark E Bolt A Global gene expression changes during neoadjuvant chemotherapy for human breast cancer Cancer J 2002 8 461 468 12500855
Wang E Miller LD Ohnmacht GA Mocellin S Perez-Diez A Petersen D Zhao Y Simon R Powell JI Asaki E Alexander HR Duray PH Herlyn M Restifo NP Liu ET Rosenberg SA Marincola FM Prospective molecular profiling of melanoma metastases suggests classifiers of immune responsiveness Cancer Res 2002 62 3581 3586 12097256
Ahr A Karn T Solbach C Seiter T Strebhardt K Holtrich U Identification of high risk breast-cancer patients by gene expression profiling Lancet 2002 359 131 2 11809257 10.1016/S0140-6736(02)07337-3
Monfardini S Brunner K Crowther D Evaluation of the cancer patient and the response to treatment UICC-Manual of Adult and Pediatric Medical Oncology 1987 Berlin, Germany: Spinger 22 38
Lockhart DJ Dong H Byrne MC Follettie MT Gallo MV Chee MS Expression monitoring by hybridization to high-density oligonucleotide arrays Nat Biotechnol 1996 14 1675 1680 9634850 10.1038/nbt1296-1675
Alon U Barkai N Notterman DA Gish K Ybarra S Mack D 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 1999 96 6745 6750 10359783 10.1073/pnas.96.12.6745
Warrington JA Nair A Mahadevappa M Tsyganskaya M Comparison of human adult and fetal expression and identification of 535 housekeeping/maintenance genes Physiol Genomics 2000 2 143 147 11015593
Colleoni M Viale G Zahrieh D Pruneri G Gentilini O Veronesi P Chemotherapy is more effective in patients with breast cancer not expressing steroid hormone receptors: a study of preoperative treatment Clin Cancer Res 2004 10 6622 6628 15475452
Cocquyt VF Schelfhout VR Blondeel PN Depypere HAT Daems KK Serreyn RF The role of biological markers as predictors of response to preoperative chemotherapy in large primary breast cancer Med Oncol 2003 20 221 231 14514971 10.1385/MO:20:3:221
Takei H Horiguchi J Maemura M Koibuchi Y Oyama T Yokoe T Predictive value of estrogen receptor status as assessed by ligand-binding assay in patients with early-stage breast cancer treated with breast conserving surgery and radiation therapy Oncol Rep 2002 9 375 378 11836611
Colleoni M Minchella I Mazzarol G Nole F Peruzzotti G Rocca A Response to primary chemotherapy in breast cancer patients with tumors not expressing estrogen and progesterone receptors Ann Oncol 2000 11 1057 1059 11038046 10.1023/A:1008334404825
Lee Y Lee CK Classification of multiple cancer types by multicategory support vector machines using gene expression data Bioinformatics 2003 19 1132 1139 12801874 10.1093/bioinformatics/btg102
Liu Y Ringner M Multiclass discovery in array data BMC Bioinformatics 2004 5 70 15180908 10.1186/1471-2105-5-70
Hartigan JA Wong MA A K-means clustering algorithm Applied Statistics 1979 28 100 108
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 14863 14868 9843981 10.1073/pnas.95.25.14863
Kohonen T Self-Organizing Maps 2001 3rd Berlin:Springer
Nguyen DV Rocke DM Partial least squares proportional hazard regression for application to DNA microarray survival data Bioinformatics 2002 18 1625 1632 12490447 10.1093/bioinformatics/18.12.1625
Nguyen DV Rocke DM Multi-class cancer classification via partial least squares with gene expression profiles Bioinformatics 2002 18 1216 1226 12217913 10.1093/bioinformatics/18.9.1216
Perez-Enciso M Tenenhaus M Prediction of clinical outcome with microarray data: a partial least squares discriminant analysis (PLS-DA) approach Hum Genet 2003 112 581 592 12607117
Datta S Exploring relationships in gene expressions: a partial least squares approach Gene Expr 2001 9 249 255 11763996
Tan Y Shi L Tong W Gene Hwang GT Wang C Multi-class tumor classification by discriminant partial least squares using microarray gene expression data and assessment of classification models Comput Biol Chem 2004 28 235 243 15261154 10.1016/j.compbiolchem.2004.05.002
Schleyer PLS in Chemistry 2004 The Encyclopedia of Computational Chemistry, P.V.R. Chichester, UK: John Wiley & Sons
Ellis M Ballman K Trawling for genes that predict response to breast cancer adjuvant therapy J Clin Oncol 2004 22 2267 2269 15136594 10.1200/JCO.2004.03.950
Smith IE Lipton L Preoperative/neoadjuvant medical therapy for early breast cancer Lancet Oncol 2001 2 561 570 11905711 10.1016/S1470-2045(01)00490-9
Cleator S Parton M Dowsett M The biology of neoadjuvant chemotherapy for breast cancer Endocr Relat Cancer 2002 9 183 195 12237246 10.1677/erc.0.0090183
Bonnefoi H Diebold-Berger S Therasse P Hamilton A van de Vijver M MacGrogan G Locally advanced/inflammatory breast cancers treated with intensive epirubicin-based neoadjuvant chemotherapy: are there molecular markers in the primary tumour that predict for 5-year clinical outcome? Ann Oncol 2003 14 406 413 12598346 10.1093/annonc/mdg108
Petit T Wilt M Velten M Millon R Rodier JF Borel C Comparative value of tumour grade, hormonal receptors, Ki-67, HER-2 and topoisomerase II alpha status as predictive markers in breast cancer patients treated with neoadjuvant anthracycline-based chemotherapy Eur J Cancer 2004 40 205 211 14728934 10.1016/S0959-8049(03)00675-0
Faneyte IF Schrama JG Peterse JL Remijnse PL Rodenhuis S van de Vijver MJ Breast cancer response to neoadjuvant chemotherapy: predictive markers and relation with outcome Br J Cancer 2003 88 406 412 12569384 10.1038/sj.bjc.6600749
Martin-Richard M Munoz M Albanell J Colomo L Bellet M Rey MJ Serial topoisomerase II expression in primary breast cancer and response to neoadjuvant anthracycline-based chemotherapy Oncology 2004 66 388 394 15331926 10.1159/000079487
MacGrogan G Mauriac L Durand M Bonichon F Trojani M de Mascarel I Primary chemotherapy in breast invasive carcinoma: predictive value of the immunohistochemical detection of hormonal receptors, p53, c-erbB-2, MiB1, pS2 and GST pi Br J Cancer 1996 74 1458 1465 8912545
Sjostrom J Blomqvist C Heikkila P Boguslawski KV Raisanen-Sokolowski A Bengtsson NO Predictive value of p53, mdm-2, p21, and mib-1 for chemotherapy response in advanced breast cancer Clin Cancer Res 2000 6 3103 3110 10955790
Ogston KN Miller ID Schofield AC Spyrantis A Pavlidou E Sarkar TK Can patients' likelihood of benefiting from primary chemotherapy for breast cancer be predicted before commencement of treatment? Breast Cancer Res Treat 2004 86 181 189 15319570 10.1023/B:BREA.0000032986.00879.d7
Makris A Powles TJ Dowsett M Osborne CK Trott PA Fernando IN Prediction of response to neoadjuvant chemoendocrine therapy in primary breast carcinomas Clin Cancer Res 1997 3 593 600 9815725
Prisack HB Karreman Ch Modlich O Audretsch W Rezai M Bojar H Predictive biological markers for response of invasive breast cancer to anthracycline/cyclophosphamid-based primary (radio-) chemotherapy Anticancer Res (accepted for publication in June, 2005).
Sorlie T Perou CM Tibshirani R Aas T Geisler S Johnsen H Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications Proc Natl Acad Sci U S A 2001 98 10869 10874 11553815 10.1073/pnas.191367098
van 't Veer LJ Dai H van de Vijver MJ He YD Hart AA Mao M Gene expression profiling predicts clinical outcome of breast cancer Nature 2002 415 530 536 11823860 10.1038/415530a
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
Bertucci F Nasser V Granjeaud S Eisinger F Adelaide J Tagett R 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
van de Vijver MJ He YD van't Veer LJ Dai H Hart AA Voskuil DW A gene-expression signature as a predictor of survival in breast cancer N Engl J Med 2002 347 1999 2009 12490681 10.1056/NEJMoa021967
Bottini A Berruti A Brizzi MP Bersiga A Generali D Allevi G Aguggini S Bolsi G Bonardi S Tondelli B Vana F Tampellini M Alquati P Dogliotti L Cytotoxic and antiproliferative activity of the single agent epirubicin versus epirubicin plus tamoxifen as primary chemotherapy in human breast cancer: a single-institution phase III trial Endocr Relat Cancer 2005 12 383 92 15947110 10.1677/erc.1.00945
Simon R Radmacher MD Dobbin K McShane LM Pitfalls in the use of DNA microarray data for diagnostic and prognostic classification J Natl Cancer Inst 2003 95 14 18 12509396
Simon R Radmacher MD Dobbin K Design of studies using DNA microarrays Genet Epidemiol 2002 23 21 36 12112246 10.1002/gepi.202
Ben-Dor A Bruhn L Friedman N Nachman I Schummer M Yakhini Z Tissue classification with gene expression profiles J Comput Biol 2000 7 559 83 11108479 10.1089/106652700750050943
Ein-Dor L Kela I Getz G Givol D Domany E Outcome signature genes in breast cancer: is there a unique set? Bioinformatics 2004 21 171 178 15308542 10.1093/bioinformatics/bth469
|
16091131
|
PMC1201176
|
CC BY
|
2021-01-04 16:39:28
|
no
|
J Transl Med. 2005 Aug 9; 3:32
|
utf-8
|
J Transl Med
| 2,005 |
10.1186/1479-5876-3-32
|
oa_comm
|
==== Front
Virol JVirology Journal1743-422XBioMed Central London 1743-422X-2-581602950610.1186/1743-422X-2-58ResearchGenetic characterization of measles viruses isolated in Turkey during 2000 and 2001 Korukluoglu Gulay [email protected] Stephanie [email protected] Dalya [email protected] Fumio [email protected]_tokyo.ac.japRota Paul A [email protected] William J [email protected] Ali [email protected] Meliksah [email protected] National Measles/Rubella Laboratory, Refik Saydam National Hygiene Center, Ankara, Turkey2 Division of Viral and Rickettsial Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia, USA3 National Immunization Program, Centers for Disease Control and Prevention, Atlanta, Georgia, USA4 Biomedical Sciences Association, Tokyo, Japan5 Department of Public Health, Dicle University School of Medicine, Diyarbakir, Turkey2005 19 7 2005 2 58 58 20 6 2005 19 7 2005 Copyright © 2005 Korukluoglu et al; licensee BioMed Central Ltd.2005Korukluoglu 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
Molecular epidemiologic studies have made significant contributions to measles surveillance activities by helping to identify source and transmission pathways of the virus. This report describes the genetic characterization of wild-type measles viruses isolated in Turkey in 2000 and 2001.
Results
Wild-type measles viruses were isolated from 24 cases from five provinces in Turkey during 2001. The viruses were analyzed using the standard genotyping protocols. All isolates were classified as genotype D6, the same genotype that was identified in Turkey in previous outbreaks during 1998.
Conclusion
Turkey has begun implementation of a national program to eliminate measles by 2010. Therefore, this baseline genotype data will provide a means to monitor the success of the elimination program.
==== Body
Background
Measles virus (MV), an enveloped virus with a single-stranded, negative sense RNA genome, is a member of the genus Morbillivirus within the family Paramyxoviridae. MV is highly contagious and causes a disease characterized by high fever, cough, coryza, conjunctivitis and appearance of a maculopapular rash [1]. In many parts of the world, vaccination programs have controlled measles. However, despite the tremendous progress of global measles control, MV is still responsible for the deaths of approximately 700,000 thousand children each year, mostly in developing countries [2]. Measles remains the most common of vaccine-preventable childhood mortality.
Although MV is considered to be monotypic, genetic variability exists among wild type strains [3]. Genetic characterization of wild-type MVs is based on sequence analysis of a hypervariable region (450 nt) of the nucleoprotein (N) gene and the full-length hemagglutinin (H) gene. A standard nomenclature and analysis protocol for describing the genetic characteristics of wild-type MVs was established by the World Health Organization (WHO) [4-7]. WHO recommends that genetic analysis of MV isolates should be conducted during all phases of measles control. Genetic analysis of wild-type MVs has provided an increasingly comprehensive picture of the worldwide distribution of MV genotypes [8]. Molecular epidemiologic studies can help to measure transmission pathways and to clarify epidemiological links during outbreaks. Virologic surveillance can also help to measure the success of measles vaccination programs by documenting the interruption of transmission of the endemic viral genotype(s) [9,10].
In 2001, Turkey experienced a large measles epidemic and the number of reported measles cases was over 30,000 [11]. From October 2000 to August 2001, we isolated MVs from measles cases in five different provinces of Turkey. Since Turkey has recently initiated a program to eliminate measles, this report provides important baseline data that will allow future molecular epidemiologic studies to help measure the success of this program.
Results and Discussion
With the exception of one specimen that was collected in October 2000, the remaining specimens were collected between February and August in 2001 (Table 1). MV isolates were obtained from 24 specimens collected from widely dispersed areas of Turkey, including the provinces of Ankara, Sinop, Diyarbakir, Sirnak, and Ardahan (Figure 1, Table 1). Measles specific IgM antibody was detected in serum samples from 16 of 20 cases, while serologic results were not available for 4 cases. The serum samples from 3 of the 4 IgM negative cases were taken 2 days after rash onset when the sensitivity of IgM detection is low.
Table 1 Epidemiological and serological information on measles virus isolates from Turkey.
WHO Name [Genotype] Age Measles IgM Date of after rash Cell lines used for isolation Type of specimen Province Epi-link
MVi/Ankara.TUR/38.00 [D6] 7 y positive 3 B95 a urine Ankara sporadic
MVi/Ankara.TUR/05.01 [D6] 17 y negative 2 B95 a urine Ankara sporadic
MVi/Ankara.TUR/06.01-1 [D6] 24 y negative 2 B95 a urine Ankara sporadic
MVi/Ankara.TUR/06.01-2 [D6] 21 y negative 2 B95 a urine Ankara epidemic
MVi/Ankara.TUR/07.01 [D6] 2.5 y ? ? B95 a urine Ankara sporadic
MVi/Sinop.TUR/11.01-1 [D6] 13 y positive 4 B95 a urine Sinop epidemic
MVi/Sinop.TUR/11.01-2 [D6] 13 y positive 5 B95 a urine Sinop epidemic
MVi/Sinop.TUR/11.01-3 [D6] 13 y positive 4 B95 a throat swab Sinop epidemic
MVi/Sinop.TUR/11.01-4 [D6] 13 y positive 3 B95 a urine Sinop epidemic
MVi/Sinop.TUR/11.01-5 [D6] 13 y positive 4 B95 a urine Sinop epidemic
MVi/Sinop.TUR/11.01-6 [D6] 13 y positive 4 B95 a throat swab Sinop epidemic
MVi/Ankara.TUR/14.01 [D6] ? positive ? B95 a nasal swab Ankara sporadic
MVi/Ankara.TUR/19.01-1 [D6] 25 y. ? 3 COBL urine Ankara sporadic
MVi/Ankara.TUR/19.01-2 [D6] ? ? ? COBL urine Ankara sporadic
MVi/Ardahan.TUR/23.01 [D6] ? ? ? COBL urine Ardahan epidemic
MVi/Sirnak.TUR/29.01-1 [D6] 3 y. negative 5 COBL urine Şırnak epidemic
MVi/Sirnak.TUR/29.01-2 [D6] 3 y. positive 3. COBL throat swab Şırnak epidemic
MVi/Sirnak.TUR/29.01-4 [D6] 3 y. positive 7 COBL urine Şırnak epidemic
MVi/Sirnak.TUR/29.01-5 [D6] 4 y. positive 3 COBL urine Şırnak epidemic
MVi/Sirnak.TUR/29.01-6 [D6] 2 y. positive 5 COBL urine Şırnak epidemic
MVi/Sirnak.TUR/29.01-7 [D6] 8 mo. positive 6 COBL blood Şırnak epidemic
MVi/Diyarbakir.TUR/30.01-1 [D6] 7 y. positive 4 COBL blood Diyarbakır epidemic
MVi/Diyarbakir.TUR/30.01-2 [D6] 7 y. positive 2 COBL urine Diyarbakır epidemic
MVi/Ankara.TUR/30.01 [D6] 2 y. positive ? COBL urine Ankara sporadic
Figure 1 Map of Turkey showing province and number of measles virus isolates obtained during 2000–2001.
Comparison of the N gene sequences of the Turkish viruses with the sequences of the current of WHO reference strains showed that all 24 Turkish strains were members of genotype D6 (Figure 2). The sequences of the Turkish viruses were closely related to each other showing no more than 1.3% nucleotide heterogeneity overall. In fact, the N gene sequences of 21 of these MV isolates were identical, though they came from different regions of Turkey. Although the Turkish viruses were clearly in genotype D6, the sequences of the more recently isolated viruses formed a distinct group relative to other genotype D6 viruses recently isolated in Germany, Luxembourg, Brazil and the United States [10,18-20]. However, the nucleotide sequences from the Turkish cluster differed from the sequences of the non-Turkish viruses by no more than 1.1% overall. The sequence of a single isolate from Ankara in 2000, MVi/Ankara.TUR/38.00, and a genotype D6 isolate from the 1998 outbreak, MVi/Ankara/10-98-4 [21], were more closely related to the sequences of the European, and Brazilian genotype D6 viruses than the sequences of the Turkish cluster (Figure 2).
Figure 2 Phylogenetic analysis of the N gene sequences of wild-type MVs isolated in Turkey. Sequences of the Turkish viruses were compared to the sequence of the WHO reference strains (genotype shown in bold). Turkish viruses are indicated by arrows. Sequences of previously described genotype D6 viruses [10, 8–20] are also included in this un-rooted tree.
At present, genotype D7 appears to be the most frequently detected genotype in Western European countries; however, D6 genotype is still circulating in some European countries including the Russian Federation [6,19]. Genotype D6 viruses were imported to the United States from various European countries and Brazil on 13 occasions between 1997 and 2000; however, after 2000, only 2 genotype D6 viruses were detected in the United States (Rota, unpublished).
In some parts of Europe, measles is near elimination or has been eliminated, whereas in others measles is still endemic [22]. Despite an active vaccination program, measles has been an endemic disease in Turkey with epidemics occurring every 3–4 years. In 2001, the last epidemic year, over 30.000 cases were reported [11]. The previous epidemic year was 1998, when more than 27,000 cases were reported. The virologic surveillance data suggest that viruses in genotype D6 were responsible for both epidemics and continued to circulate during the inter-epidemic periods.
To reduce measles morbidity and mortality in Turkey, the Ministry of Health launched a National Measles Elimination Program in 2002. In parallel with the strategic plan of the European Regional Office of WHO, the Turkish national plan targets elimination of measles by 2010 [23]. The plan included a "catch-up" vaccination campaign targeting nearly 20 million children between 9 months and 14 years of age to be conducted in two phases during December 2003 and 2005 [24]. The National Measles Plan also includes activities for establishing a laboratory based surveillance system to monitoring the effectiveness of the measles elimination program [25]. In Turkey, sub-national laboratories from seven selected provinces will carry out laboratory-based surveillance, each representing a region of the country. These sub-national laboratories will perform serologic confirmation of suspected measles cases. Clinical specimens collected from laboratory-confirmed cases will be sent to the National Measles and Rubella Laboratory for virus isolation and genotyping.
Conclusion
Genetic analysis of MVs isolated after the measles vaccination campaigns will help to determine if the circulation of the endemic genotype D6 viruses is interrupted. This analysis would not be possible without the baseline data presented in this report. Turkey is in a unique geographic position to monitor transmission of measles virus between Europe, the Middle East and the rest of Asia. Strengthening virologic surveillance capacity in Turkey will benefit several WHO regions.
Materials and methods
Clinical specimens
Urine, nasopharyngeal secretions and blood samples were collected from 24 patients who had acute, febrile maculopapular rash from five different provinces in Turkey. All clinical samples were collected within six days of rash onset and transported to Refik Saydam Hygiene Center, National Measles and Rubella Laboratory in accordance with standard protocols (Table 1). Isolation of MV was performed using the B95a cell line (12) for 12 samples and the COBL cell line (IL-II treated human cord blood cells, 13) for 15 samples. Syncytia formation, the cytopathic effect (CPE) characteristic of MV infection, appeared within 1–7 days. When the CPE was advanced the cultures were harvested and stored at -80°C. All isolates were confirmed as measles by a neutralization test performed by using monospecific rabbit antibody to the H protein.
Sequence analysis
RNA was extracted from infected cells using the guanidinium acid-phenol technique [14]. The 450 nucleotides corresponding to the COOH-terminal 150 amino acids of the N protein were amplified by using a one-step RT-PCR kit according to manufacturer's protocol (Superscript, Invitrogen). Forward and reverse primers were: 5'GCTATGCCATGGGAGTAGGAGTGG and 5'CTGGCCCTCGGCCTCTCGCAC, respectively. Sequences of the PCR products were derived by automated sequencing with the BigDye terminator VI.I chemistry according to the manufacturer's protocol (Perkin Elmer-Applied Biosystems, Foster City, CA). Sequence reaction product results were analyzed on an automatic sequencer (ABI 3100, Perkin Elmer Applied Biosystems, Foster City, CA). Sequence data were analyzed by using version 10.0 of the Genetics Computer Group Sequence Analysis Software Package [15] and phylogenetic analyses were performed using PHYLIP ver 3.4 [16] and PAUP ver 4.0 [17]. All phenograms were drawn as unrooted trees. Sequence data were deposited in GenBank under accession numbers (AY899306-AY899329).
List of Abbreviations
MV: measles virus
N: nucleoprotein
COOH- carboxyl
WHO: World Health Organization
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
GK, FK, AC, ME collected specimens and performed virus isolation and measles IgM assays; GK, FK established COBL cell in the Ankara laboratory; GK, SL, PR performed RT-PCR and sequence analysis; GK, DG, PP, WB analyzed data and prepared draft manuscript. All authors revised manuscript and approved final draft.
Acknowledgements
The authors would like to thank the field staff in Turkey for obtaining appropriate clinical samples and for providing epidemiologic data for the cases. The CDC laboratory is a WHO Measles Strain Bank.
==== Refs
Griffin E Knipe DM, Howley PM Measles virus Fields Virology 2001 1 4 Philadelphia: Lippincott Williams and Wilkins 1401 1441
WHO Progress in reducing global measles deaths: 1999–2002 Weekly Epidemiol Rec 2004 79 13 24
Bellini WJ Rota PA Genetic diversity of wild type measles virus: implication for global measles elimination programs Emerg Infect Dis 1998 4 29 35 9452396
WHO Standardization of the nomenclature for describing the genetic characteristics of wild type measles viruses Wkly Epidemiol Rec 1998 73 265 272 9745371
WHO Nomenclature for describing the genetic characteristics of wild type measles viruses (update). Part 1 Wkly Epidemiol Rec 2001 76 241 247
WHO Nomenclature for describing the genetic characteristics of wild type measles viruses (update). Part 2 Wkly Epidemiol Rec 2001 76 249 251
WHO Update of the nomenclature for describing the genetic characteristics of wild-type measles viruses: new genotypes and reference strains Wkly Epidemiol Rec 2003 78 229 232 12866261
Rota PA Bellini WJ Update on the global distribution of genotypes of wild-type measles viruses J Inf Dis 2003 187 270 276 12552451 10.1086/368042
Rota JS Heath JL Rota PA King GE Celma ML Carabaña J Fernandez-Muñoz R Brown D Jin L Bellini WJ Molecular epidemiology of measles virus: Identification of pathways of transmission and the implications for measles elimination J Inf Dis 1996 173 32 37 8537679
Rota PA Rota JS Redd S Papania M Bellini WJ Genetic Analysis of Measles Viruses Isolated in the United States between 1989 and 2001: Absence of an Endemic Genotype Since 1994 J Inf Dis 2004 189 171 176 10.1086/377716
Guris D Bayazit Y Ozdemirer U Measles Epidemiology and Elimination Strategies in Turkey J Inf Dis 2003 187 230 234 10.1086/368115
Kobune Sakata H Sugiura A Marmoset lymphoblastoid cell as sensitive host for isolation of measles virus J Virol 1990 64 700 705 2153236
Kobune F 10–11 May 2001. WHO Meeting "1st National Measles Laboratory Network" Geneva, Switzerland
Chomczynski P Sacchi N Single step method of RNA isolation by acid guanidinium thiocyanate-phenol-cholorform extraction Anal Biochem 1987 162 156 159 2440339 10.1016/0003-2697(87)90021-2
Devereaux J Haeberli P Smithies O A comprehensive set of sequence analysis programs for the VAX Nucleic Acids Res 1984 12 387 395 6546423
Felsenstein J Phylogenies from molecular sequences: inferences and reliability Am Rev Genet 1988 22 512 565
Swofford DL PAUP:phylogenetic analysis using parsimony Version 311 1988 Champaign, IL: Illinois Natural History Survey, Champaign, IL
Oliveira MI Rota PA Curti SP Figueiredo CA Afonso AMS Genetic homogeneity of measles viruses associated with a measles outbreak, Sao Paulo, Brazil, 1997 Emerg Infect Dis 2002 8 808 813 12141966
Santibanez S Heider A Gerike E Agafonov A Schreier E Genotyping of measles virus isolates from central Europe and Russia J Med Virol 1999 58 313 320 10447429 10.1002/(SICI)1096-9071(199907)58:3<313::AID-JMV19>3.0.CO;2-P
Tischer A Santibanez S Siedler A Heider A Hengel H Laboratory investigations are indispensable to monitor the progress of measles elimination – results of the German Measles Sentinel 1999–2003 J Clin Virol 2004 31 165 178 15465408 10.1016/j.jcv.2004.05.007
Yılmaz N Rota PA Tasyurek T Genetic characterization of measles virus isolated in Turkey Proceedings of The National Conference of Moleculer Biology and Diagnostic Microbiology, Nevsehir, Turkey 2000
Spika J Steven W Richard P Measles and rubella in the World Health Organization European Region: diversity creates challenges J Inf Dis 2003 187 191 197 10.1086/368336
World Health Organization Database
World Health Organization Database
World Health Organization Database
|
16029506
|
PMC1201177
|
CC BY
|
2021-01-04 16:39:00
|
no
|
Virol J. 2005 Jul 19; 2:58
|
utf-8
|
Virol J
| 2,005 |
10.1186/1743-422X-2-58
|
oa_comm
|
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-521607900010.1186/1477-7819-3-52ReviewInvasive breast cancer following bilateral subcutaneous mastectomy in a BRCA2 mutation carrier: a case report and review of the literature Kasprzak Lidia [email protected] Benoit [email protected] Francine [email protected] Maria [email protected] Fawaz [email protected] William D [email protected] Department of Medicine, McGill University Health Centre, Montreal, Canada2 Department of Radiology, McGill University Health Centre, Montreal, Canada3 Department of Surgery, McGill University, Montreal, Canada4 Department of Pathology, McGill University, Montreal, Canada5 Program in Cancer Genetics, Departments of Oncology and Human Genetics, McGill University, Montreal, Canada2005 4 8 2005 3 52 52 25 5 2005 4 8 2005 Copyright © 2005 Kasprzak et al; licensee BioMed Central Ltd.2005Kasprzak 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 prevention of breast cancer through prophylactic mastectomy can reduce the risk of malignancy in high-risk individuals. No type of mastectomy completely removes all breast tissue, but a subcutaneous mastectomy leaves more tissue in situ than does a simple mastectomy.
Case presentation
We report a case of invasive breast cancer in a BRCA2-positive woman 33 years after bilateral subcutaneous mastectomy. To our knowledge, only one case of primary breast cancer after prophylactic mastectomy in a BRCA1-positive patient has been reported in the literature and none in BRCA2-positive individuals.
Conclusion
Careful documentation and long follow-up is essential to fully assess the benefits and risks of preventive surgical procedures in BRCA1 and BRCA2 mutation carriers.
==== Body
Background
The risk of breast cancer in women who inherit a germline mutation in the BRCA1 gene can be as high as 20% by the age of 40 and 50% by the age of 50 [1] and as high as 13% by the age of 40 and 60% by the age of 50 in BRCA2 mutation carriers [2]. These estimates apply to individuals who belong to very high risk, multiple-case breast cancer families. Prophylactic surgery to reduce cancer risk remains an option for carriers of BRCA gene mutations; however, its efficacy is likely to depend on the ability to remove nearly all breast tissue. Different surgical procedures, (subcutaneous or a simple mastectomy), limited patient follow-up and lack of adequate control population confound the numerous studies on this subject. Only three reports have specifically addressed the extent of risk reduction by prophylactic breast removal in BRCA1 and BRCA2-positive individuals [3-5]. Most surgeons believe that subcutaneous mastectomy (SCM) is not optimal for prophylaxis because a substantial amount of breast tissue remains in the nipple-areola complex and on the skin flaps, and therefore it has fallen into disuse. Here, we report a case of breast cancer occurring following a SCM. A literature review using PubMed, searching from 1994 to the present, revealed this to be the first reported case of breast cancer occurring post-SCM in a carrier of a BRCA2 gene mutation. We discuss the possible implications of this seemingly uncommon finding.
Case Presentation
A 49 year old G3P2A1 presented in 2002 with a six month history of a painless lump in her inner left breast. She had undergone bilateral SCM with immediate implantation of silicone prostheses at the age of 16 due to extensive fibrocystic breast disease with adenosis. On physical examination a firm, mobile, 1.5 cm nodule was palpated superficially. There was no associated skin retraction or thickening. No enlarged abnormal nodes were palpable. We performed mammographic and sonographic examinations. Mammography, of limited value because of the previous SCM, did not show any obvious abnormality. Sonographic examination revealed a 1.5 cm in size, hypoechoic, solid, non-calcified, circumscribed mass, grossly ovoid with a thin echoic rim, located in the subcutaneous fat at 9 o'clock in the left breast (Figure 1). Minimal vascular pole was identified on conventional color sonography. The imaging findings appeared to indicate a benign nature of the lesion; however, based on the family history, patient age, and the recent occurrence of the nodule, excisional biopsy was performed. An infiltrating ductal carcinoma, apocrine type, grade 2 of 3 of modified Bloom and Richardson, with associated ductal carcinoma in situ, cribriform type, nuclear grade 2 of 3, occupying 10% of tumor mass (Figures 2A and 2B) was identified. Estrogen receptor status was strongly positive on most malignant cells, progesterone receptor moderately positive on 50% of malignant cells and HER-2/neu was negative (score 0) with no membrane staining of malignant cells. There was no evidence of malignancy in 23 lymph nodes examined following the left axillary contents dissection.
Figure 1 Sonographic examination: 15 mm hypoechoic solid, non-calcified circumscribed mass, with a thin echoic rim, benign in appearance, located in subcutaneous fat at 9 o'clock in the left breast.
Figure 2 photomicrograph 2A) Invasive ductal carcinoma, apocrine type: tumor exhibits an irregular invasive border and forms glandular structures. Hematoxylin and eosin stain, original magnification × 40. 2B) Invasive ductal carcinoma, apocrine type: cytological characteristics of intermediate nuclear grade, prominent nucleoli, and eosinophilic granular cytoplasm. Hematoxylin and eosin stain, original magnification × 100.
Subsequently, the proband was referred for genetic counseling and found to carry BRCA2: 6503delTT, a mutation previously described in the French Canadian population [6]. Family history (Figure 3) was significant for breast cancer in proband's father (II-6) who was diagnosed at the age of 77 and mother (II-7) diagnosed at 79. A 52 years old paternal cousin (III-3), also affected with breast cancer, was previously identified as a BRCA2 mutation carrier at another institution. Of note, individual III-3's mother (II-5) and sister (III-2) both had breast cancer and died at 47 and 49 years of age, respectively. There were only two paternal aunts known to have been affected with breast cancer at the time of patient's bilateral SCM.
Figure 3 Pedigree of the family with germline BRCA2: 6503delTT mutation. All individuals affected with cancer (ca) are depicted by filled-in symbols. Individual ID numbers, age at the time of diagnosis (dx) and/or death are below each symbol.
Given the diagnosis of invasive breast cancer and BRCA2 mutation carrier status, our proband opted for prophylactic bilateral salpingo-oophorectomy with hysterectomy, as well as removal of the nipple-areola complex along with remaining breast tissue. The pathology examination identified the presence of bilateral adnexal stromal hyperplasia, several leiomyomata in the myometrium as well as simple ductal epithelial hyperplasia without atypia in the mastectomy specimen. There was no evidence of malignancy.
Discussion
Currently available management strategies for women who carry inherited predisposition to develop breast cancer are limited given the lack of prevention methods with proven efficacy. Furthermore, there have been no prospective, controlled trials of the breast cancer risk reduction associated with bilateral prophylactic mastectomy. Such studies are unlikely to take place due to ethical and practical considerations. Prophylactic mastectomy, whether subcutaneous or total, significantly reduces, but does not eliminate, the risk of breast cancer in high-risk individuals.
Several studies showed that breast reduction procedures substantially lower the risk of breast cancer. Recently, Brinton et al [7] confirmed that the magnitude of cancer risk reduction is directly related to the amount of tissue removed during the operation. With a SCM, the nipple-areola complex is preserved and some of the underlying breast tissue remains on the skin flaps. When SCM is intended as prophylaxis against breast cancer, the surgeon's aim is to remove as much tissue as possible. It is plausible that a less thorough removal of glandular tissue may have taken place given the indication for surgery in our patient's case. It is generally agreed that the prophylactic nature of bilateral mastectomy in an unaffected BRCA1 or BRCA2 mutation carrier calls for the most complete breast tissue removal. This viewpoint makes SCM a less desirable choice. Skin-sparing mastectomy [8] could be seen as a partial compromise and appears to be an increasingly popular option for women at high risk. More recently, geneticists have questioned the rejection of simple SCM as a viable procedure in such women. It is argued that the magnitude of the risk reduction offered by SCM, when combined with its greater cosmetic acceptability, is sufficient to keep this option available to women [9].
There are numerous reports in the literature describing the occurrence of breast cancer after SCM [10,11]. Subsequently, the perception exists that SCM fails to eliminate the risk of breast cancer. Although the extent of risk reduction achieved by SCM is limited given that about 5–10% of the mammary tissue remains in situ, it is thought to be of the order of >85% [4,12]. As stated above [9], at this level of risk reduction, SCM would have a greater effect on breast cancer rates in BRCA1/2 carriers than would total mastectomy if at least 50% of BRCA1/2 carriers chose preventive SCM. Currently, preventive bilateral total mastectomy rates are about 20% in most populations.
The first retrospective study of efficacy of prophylactic mastectomy carried out by Hartmann et al [12] included 18 subjects later confirmed to be carriers of deleterious mutations in the BRCA genes but, unfortunately, it had insufficient statistical power to detect a difference in the risk reduction between total and SCM. In this cohort, all breast cancers (n = 7) were diagnosed in women who had undergone SCM (total of 950). None were known to be BRCA1/2 mutation carriers. Of the seven cases, only one occurred in the nipple-areolar area. Not surprisingly, the majority of the high-risk women (n = 17) described in the subsequent report [3] underwent SCM. After a median of 13.4 years of follow-up, none of the BRCA1/2 germ-line mutation carriers has developed breast cancer. The authors concluded that at least 90% risk reduction could be expected among women with confirmed BRCA mutation status following prophylactic bilateral SCM. Meijers-Heijboer et al [4] report the initial results of a prospective study of 76 women with deleterious BRCA1 or BRCA2 mutations who chose to undergo bilateral simple mastectomy and no breast cancers were observed after a mean follow-up of 2.9 ± 1.4 years. The Prevention and Observation of Surgical End Points (PROSE) Study Group findings [5] support the notion that bilateral mastectomy results in approximately 90% breast cancer risk reduction. Of 105 BRCA mutation carriers, only two women (1.9%) developed breast cancer 2.3 and 9.2 years after SCM. The first breast cancer case was diagnosed at the age of 28 years in the BRCA2 mutation carrier who presented with a palpable axillary mass at 27 months post-SCM. Subsequently, metastatic adenocarcinoma in an axillary lymph node was identified and it was most likely consistent with a primary breast cancer already present at the time of SCM. It is important that this case is not considered as a failure of SCM, and it should be therefore classified as a recurrence which would have likely taken place despite the surgery. The second breast cancer case occurred in a BRCA1 carrier at the age of 41 years.
When hereditary predisposition to breast cancer is being assessed, it is important to consider the impact of the age-related penetrance of the BRCA1 and BRCA2 genes. BRCA1 has a higher penetrance than BRCA2 in the pre-menopausal years [1]. The benefits of preventive surgery will be proportionally greater for an older BRCA2 carrier than an older BRCA1 carrier, and therefore if a BRCA2 carrier discovers her mutation status when she is peri- or post-menopausal, the potential benefits of preventive mastectomy should not be understated, as the breast cancer risks do not significantly diminish following menopause.
In our review of the literature (see Table 1), only one primary breast cancer has been reported to occur in a cohort of 207 BRCA1/2 mutation carriers who opted for preventive surgery. It could be argued that the efficacy of the bilateral total mastectomy has not been studied adequately in the high-risk individuals to prove its absolute superiority over SCM. The total number of the BRCA1 and BRCA2 carriers who have undergone this type of mastectomy is relatively small and the mean follow-up is rather short. Furthermore, the presence of a microscopic primary lesion at the time of surgery may result in subsequent recurrence that would be impossible to differentiate from a new primary breast cancer. Based on the above data, the risk-reducing effect of SCM should not be ignored when presenting prophylactic mastectomy options to women at high risk who find total mastectomy unacceptable and would not otherwise have considered surgical prevention. Nevertheless, the lack of popularity of this procedure among surgeons will likely limit its use.
Table 1 Studies assessing efficacy of bilateral prophylactic mastectomy (PM) in BRCA1 and BRCA2 carriers
Hartmann et al [3] Meijers-Heijboer et al [4] Rebbeck et al [5]
Study recruitment centers USA Netherlands USA, Canada, UK, Netherlands
Median follow-up (yrs) 13.4 2.8 5.5
Mean age at surgery (yrs) 39 36 38
Number of PM patients 18 76 105
Number of controls - 63 378
Type of study Retrospective cohort Prospective cohort Case-control
Primary invasive breast cancer
- in cases (after PM) 0 0 1 (BRCA1carrier)
- in controls - 8 (13%) 184 (49%)
The existing literature on the mammographic and sonographic appearance of breast cancer in BRCA-positive patients' reconstructed breasts is rather scanty. Pathologic studies have demonstrated that tumors in BRCA1 and BRCA2 mutation carriers are associated with morphologic features of continuous pushing margins [13], with a reduced potential for stromal infiltration explaining that this appearence might mimic benign-looking lesions at mammography [14] and breast sonography as well. Indeed, sonographic criteria of the mass in our case – ovoid axis, thin pseudocapsule, posterior enhancement, and well defined margins – were in keeping with a benign nodule [15]. In addition, according to Giovagnorio criteria [16], the lesion described in our case, with a single vascular pole (type 2), was compatible with a benign lesion. Cconsistent with the Lamb et al study [17], this lesion appeared benign but was in fact a moderate to high-grade invasive cancer.
Conclusion
We report an unusual case of late occurrence of breast cancer after SCM in a BRCA2 mutation carrier. As these cases are so rare, the long-term risk of breast cancer following preventive mastectomy in BRCA1/2-positive individuals is likely to be very low. Nevertheless, vigilant, long-term surveillance based on clinical examination combined with breast sonography when indicated remains necessary, as delayed malignancy can occur.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
LK obtained family history, searched literature and drafted the manuscript, BM performed the sonographic assessment and assisted in manuscript preparation,FT initiated the report, provided patient history and referral for genetic counseling, managed the patient, MG carried out the molecular genetic studies, FH carried out the histopathological studies and provided diagnostic consultation, WDF obtained patient consent, assisted in literature search, helped to draft the manuscript and edited the final version. All authors read and approved the final manuscript.
Acknowledgements
Written consent was obtained from the patient for publication of the case report. The authors thank Dr. D. Gareth Evans for helpful comments.
==== Refs
Easton DF Ford D Bishop DT Breast and ovarian cancer incidence in BRCA1-mutation carriers. Breast Cancer Linkage Consortium Am J Hum Genet 1995 56 265 271 7825587
Easton DF Steele L Fields P Ormiston W Averill D Daly PA McManus R Neuhausen SL Ford D Wooster R Cannon-Albright LA Stratton MR Goldgar DE Cancer risks in two large breast cancer families linked to BRCA2 on chromosome 13q12-13 Am J Hum Genet 1997 61 120 128 9245992
Hartmann LC Sellers TA Schaid DJ Frank TS Soderberg CL Sitta DL Frost MH Grant CS Donohue JH Woods JE McDonnell SK Vockley CW Deffenbaugh A Couch FJ Jenkins RB Efficacy of bilateral prophylactic mastectomy in BRCA1 and BRCA2 gene mutation carriers J Natl Cancer Inst 2001 93 1633 1637 11698567 10.1093/jnci/93.22.1733
Meijers-Heijboer H van Geel B van Putten WL Henzen-Logmans SC Seynaeve C Menke-Pluymers MB Bartels CC Verhoog LC van den Ouweland AM Niermeijer MF Brekelmans CT Klijn JG Breast cancer after prophylactic bilateral mastectomy in women with a BRCA1 or BRCA2 mutation N Engl J Med 2001 345 159 164 11463009 10.1056/NEJM200107193450301
Rebbeck TR Friebel T Lynch HT Neuhausen SL van 't Veer L Garber JE Evans GR Narod SA Isaacs C Matloff E Daly MB Olopade OI Weber BL Bilateral prophylactic mastectomy reduces breast cancer risk in BRCA1 and BRCA2 mutation carriers: the PROSE Study Group J Clin Oncol 2004 22 1055 1062 14981104 10.1200/JCO.2004.04.188
Tonin PN Mes-Masson AM Futreal PA Morgan K Mahon M Foulkes WD Cole DE Provencher D Ghadirian P Narod SA Founder BRCA1 and BRCA2 mutations in French Canadian breast and ovarian cancer families Am J Hum Genet 1998 63 1341 1351 9792861 10.1086/302099
Brinton LA Persson I Boice JD JrMcLaughlin JK Fraumeni JF Jr Breast cancer risk in relation to amount of tissue removed during breast reduction operations in Sweden Cancer 2001 91 478 483 11169929 10.1002/1097-0142(20010201)91:3<478::AID-CNCR1025>3.0.CO;2-5
Simmons RM Adamovich TL Skin-sparing mastectomy Surg Clin North Am 2003 83 885 899 12875600 10.1016/S0039-6109(03)00035-5
Metcalfe KA Semple JL Narod SA Time to reconsider subcutaneous mastectomy for breast-cancer prevention? Lancet Oncol 2005 6 431 4 15925821 10.1016/S1470-2045(05)70210-2
Pennisi VR Capozzi A Subcutaneous mastectomy data: a final statistical analysis of 1500 patients Aesthetic Plast Surg 1989 13 15 21 2728994 10.1007/BF01570320
Slade CL Subcutaneous mastectomy: acute complications and long-term follow-up Plast Reconstr Surg 1984 73 84 90 6691079
Hartmann LC Schaid DJ Woods JE Crotty TP Myers JL Arnold PG Petty PM Sellers TA Johnson JL McDonnell SK Frost MH Jenkins RB Efficacy of bilateral prophylactic mastectomy in women with a family history of breast cancer N Engl J Med 1999 340 77 84 9887158 10.1056/NEJM199901143400201
Lakhani SR Jacquemier J Sloane JP Gusterson BA Anderson TJ van de Vijver MJ Farid LM Venter D Antoniou A Storfer-Isser A Smyth E Steel CM Haites N Scott RJ Goldgar D Neuhausen S Daly PA Ormiston W McManus R Scherneck S Ponder BA Ford D Peto J Stoppa-Lyonnet D Bignon YJ Struewing JP Spurr NK Bishop DT Klijn JG Devilee P Cornelisse CJ Lasset C Lenoir G Barkardottir RB Egilsson V Hamann U Chang-Claude J Sobol H Weber B Stratton MR Easton DF Multifactorial analysis of differences between sporadic breast cancers and cancers involving BRCA1 and BRCA2 mutations J Natl Cancer Inst 1998 90 1138 1145 9701363 10.1093/jnci/90.15.1138
Kaas R Kroger R Hendriks JH Besnard AP Koops W Pameijer FA Prevoo W Loo CE Muller SH The significance of circumscribed malignant mammographic masses in the surveillance of BRCA 1/2 gene mutation carriers Eur Radiol 2004 14 1647 53 15083333 10.1007/s00330-004-2307-3
Stavros AT Thickman D Rapp CL Dennis MA Parker SH Sisney GA Solid breast nodules: use of sonography to distinguish between benign and malignant lesions Radiology 1995 196 123 134 7784555
Giovagnorio F Andreoli C De Cicco ML Color Doppler sonography of focal lesions of the skin and subcutaneous tissue J Ultrasound Med 1999 18 89 93 10206814
Lamb PM Perry NM Vinnicombe SJ Wells CA Correlation between ultrasound characteristics, mammographic findings and histological grade in patients with invasive ductal carcinoma of the breast Clin Radiol 2000 55 40 44 10650109 10.1053/crad.1999.0333
|
16079000
|
PMC1201178
|
CC BY
|
2021-01-04 16:39:05
|
no
|
World J Surg Oncol. 2005 Aug 4; 3:52
|
utf-8
|
World J Surg Oncol
| 2,005 |
10.1186/1477-7819-3-52
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1614985010.1371/journal.pbio.0030322EssayScience PolicyNoneScience Star over Asia EssayTan Chris Y. H 9 2005 13 9 2005 13 9 2005 3 9 e322Copyright: © 2005 Chris Y. H. Tan.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 founding director of Singapore's Institute of Molecular and Cell Biology illustrates the rise of science in Asia.
==== Body
Recent news of human embryonic stem cell (hESC) research in Seoul has made headlines around the world. In 2004, Korea took the world by surprise when W. S. Hwang and his team published the isolation of hESCs from a cloned blastocyst [1]. One year later, the same Korean team published the establishment of 11 hESC lines, made from transplanting the nucleus of patients' skin cells into donated human oocytes [2]. But few in the Western science community are aware that hESC research has its root in Asia. Ariff Bongso and his associates at Singapore's National University Hospital in 1994 were the first to derive hESCs from a five-day-old discarded human embryo and discover that these cells were pluripotent and, henceforth, had therapeutic potentials as cell transplants [3]. Eight years later, they were able to substitute the use of mouse feeder layer cells with human cell feeders and human serum to grow hESCs, reducing the risk of introducing mouse and bovine pathogens [4] when hESCs are transplanted into patients.
Grit and imagination drove Asian science to compete with the best genomic science in the world.
The stem cell success in Korea comes out of a history in and growing investment in Asia–Pacific science that the West is slowly appreciating. It also comes at a time when independent events in the United States are having an impact on clinical research and health care. The ongoing debate of the Right to Life movement versus allowing hESC research to fight life-threatening diseases continues to polarize public opinion. The recent withdrawals of rofecoxib (Vioxx), methylphenidate (Ritalin), and a few other high-profile drugs from the market threatens the reputation of the pharmaceutical industry. Medicine of the 21st-century is at a crossroad, with immense changes ahead that could change the face of the pharmaceutical and health-care industry. With these changes, Asia is becoming a player in research outsourcing to discover new medicine. Amazingly, for a latecomer to modern drug development, there are now 140 drugs in China's pipeline alone, 60 (10% of the world's total) of which are biologically derived agents, including antibodies and vaccines. And in the near future, it is likely that numerous clinical trials will be performed in Asia as clinical trial costs escalate in the West. This rise of science in the Asia–Pacific region is not simply fortuitous. As I will discuss, it is the result of concentrated efforts to capitalize on its strengths and to form strategic partnerships.
Traditional Chinese Medicine and Natural Product Drug Discovery
Traditional Chinese medicine (TCM) became a national health-care delivery focus when China became the People's Republic of China in 1949. Medicinal factories were set up to extract TCMs from herbs and to manufacture TCM pills and medicinal powders for China's masses. It took another two decades, during Nixon's Ping-Pong Diplomacy, for the US to discover the wide use of acupuncture in China to treat chronic pain. But until very recently, TCM has remained an enigma for the West. In the 1980s Hong Kong became interested in TCM as a focus for biotechnology. Today Hong Kong researchers collaborate with their Chinese counterparts to search for herbs with medicinal properties by investigating ancient Chinese medical literature, and then finding the described herbs in the collection of herbal libraries in China (Table 1). For instance, in the search for compounds to treat pain, a Chinese biotechnology company (International Wex Technologies) has successfully applied the TCM principle of using one poison to counter another—they have used sublethal doses of tetrodotoxins for the treatment of pain associated with heroin withdrawal [5] and for the management of pain in patients with cancer [6].
Table 1 Treatments Derived from TCMs
Most of the active ingredients in TCM are uncharacterized, and its alleged efficacy depends on multiple active ingredients from herbal extracts, whereas Western medicine often relies on the reductive principle of one active ingredient to treat a given disease. But Western medicine of late has also had success in extending the practice of combinatorial drug therapy from cancer treatment to the combination of various anti-HIV drugs to inhibit the rapidly mutating virus in patients with HIV, and the combination of antibiotics and inhibitors of acid secretion for the treatment of gastrointestinal ulcers. It would appear that Western medicine and TCM are beginning to share some common ground in the use of more than one medicine to treat a disease.
In 1993, cell signalling research at Singapore's Institute of Molecular and Cell Biology (IMCB), where I served as the founding director, attracted GlaxoWellcome to set up a High-Throughput Drug Discovery Center in the IMCB to screen for lead compounds from natural products. This collaboration eventually spun off a drug discovery company, MerLion Pharmaceutical, a Singapore-based company that has the world's most comprehensive and diverse collection of natural products (600,000). MerLion is collaborating with several companies such as Abbott Laboratories, Merck, Athelas, NovImmune, Schering–Plough, KuDOS Pharmaceuticals, Fujisawa Pharmaceuticals, and Dow AgroSciences to search for lead compounds in its impressive collection of natural products. High-throughput screening, selection of highly specific cell-signalling intervention points, and computational indexing of libraries cover the value chain of modern drug discovery, and have been established for several years in Singapore. This is now bringing in companies from around the world to the Asia Pacific to discover new leads from natural products.
Genomics, Gene Therapy, and hESCs
As leadership in Singapore, Hong Kong, Taipei, and Seoul were considering avenues of biotechnology investment in the 1980s, the US and Europe had already decided to sequence the human genome. Apart from Japan, the Asia–Pacific region was generally unprepared for the coming of the age of genomics. By the 1990s, different pieces of the human genome were assigned to different countries for sequencing, with the US having the lion's share. Nevertheless, Singapore's IMCB resolved to make a meaningful contribution to genomics by sequencing the genome of the Japanese puffer fish Fugu rubripes. With such a small genome (365 megabases) and relatively little repetitive DNA, the puffer fish offered a potential tool to annotate and thereby study the human genome. Indeed the puffer fish genome sequence enabled the Singapore-led team to discover approximately 1,000 human putative genes that had not been described in the public annotation prediction databases [7].
In the same year, another genome feat was being accomplished in China, where two US-trained scientists, Gane Wong and Jun Yu, persuaded the Chinese Academy of Sciences to support the genomic sequencing of the most important agricultural plant in the history of Asia, rice. Wong and Yu established the Beijing Genomics Center and gathered 500 young researchers to sequence the rice genome. To beat the competition, it was not uncommon for young researchers to work around the clock, camping in the laboratory. And in 2002, the Beijing Genomics Center surprised the world by announcing that it had sequenced the rice genome [8]. With essentially no track record in the field, IMCB and the Beijing Genomics Center made significant breakthroughs, despite being late off the starting block in the genomics race.
Grit and imagination drove Asian science to compete with the best genomic science in the world. The same can be said of both human stem cell and gene therapy research in Asia, except that the impact is even bigger. The first licensing of a made-in-China gene therapy product surprised the world. SiBiono Gene Technologies, a gene therapy company in Shenzhen, announced the world's first gene therapy medicine, developed using the recombinant adenovirus-p53 tumour suppressor gene for the treatment of head and neck squamous-cell carcinoma. In clinical trials, 64% of patients with late-stage head and neck cancer experienced complete regression of their tumours. SiBiono's success was attributed to developing the right system for delivery of its adenoviral vector. It addressed safety by careful dosing and follow-up of the patients for up to five years.
China's large patient population base means it is easier to recruit patients and, hence, generate clinical data quickly. Clinical trials of gene therapy have not been delayed in China as they were in the West because of a few fatalities in the US and Europe during the initial clinical studies of gene therapy. So far, no deaths have been reported from gene therapy trials in China. The approval of the first gene therapy product has propelled China to the forefront of gene therapy medicine. There are now ten gene therapy products in development in China, compared to 43 in the US and ten in Europe. Patients with head and neck cancer from the US and Europe are opting to go to China for gene therapy treatment. No doubt the rest of the world will closely monitor China's experience with gene therapy when the interests of patients worldwide are at stake.
Can't Do, Me Too, or Can Do Economies
Some 20 years ago, few scientists in the West would have believed that good science could come out of Asia within their lifetime, since none of the Asian countries except Japan had invested much into basic biological sciences. Only in the early 1980s did the tiger economies of Singapore, Hong Kong, South Korea, and the Republic of China ponder the merit of graduating from a “me too” manufacturing economy to a “can do”, knowledge-based society. Could it be done, and, if so, how long would it take for these countries to build their respective critical mass of scientists?
At that time, Deputy Prime Minister of Singapore, Dr. Goh Keng Swee, the architect of the Singapore economic miracle, asked whether Singapore could build an equivalent of the Weizmann Institute, a state-funded center for scientific excellence. To address this question, the Singapore IMCB was set up in 1987, and it began recruiting young scientists from Europe, the US, and Canada. Within ten years, the IMCB established an infrastructure for intensive research in modern biology (http://www.imcb.a-star.edu.sg). It recruited 300 scientists working on cell signalling mechanisms during differentiation, proliferation, development, cell movement, and host–virus interactions. In as little as one decade, good science was fostered from essentially nothing. Within another decade, the IMCB experiment was expanded to a dozen more research institutes focusing on translation medicine, genomics, neuroscience, nanotechnology, bioinformatics, imaging, drug discovery, infectious diseases, and human disease model systems, all housed in the newly built science city Biopolis, as well as in universities and hospitals. The number of scientists working in Singapore has more than doubled to 3,000 in the past ten years, reflecting an apparent emigrational trend of scientists from the West to the Asia–Pacific region.
If the present is a glimpse of the future, there are several signs of things to come: Alex Matter, the discoverer of the well-known anticancer drug imatinib (Gleevec) has established the Novartis Institute for Tropical Disease Research in Biopolis; David Lane, the discoverer of p53 as a tumour suppressor, has been recruited to succeed myself as Director of IMCB; Sydney Brenner has great enthusiasm to get the Institute of Translation Research going in Singapore; and others like George Radda (United Kingdom), Axel Ullrich (Germany), and K. C. Nicolaou (US) are setting up laboratories in Singapore, while continuing to maintain their main laboratory in the US and Europe. Successful senior Asian scientists in the US such as neuroscientist M. M. Poo at the University of California at Berkeley, X. D. Wong at the University of Texas Southwestern, Yale University's geneticist T. Xu, Singapore IMCB's P. Li, C. W. Wong of Scripps Research Institute, and M. Lai of the University of Southern California have all recently set up corresponding research institutes in China or Taiwan (Taipei).
Movement is also occurring within Asia. In 2001, the recruitment of a leading cancer researcher at the University of Kyoto, Y. Ito, with his entire team to Singapore's IMCB created a debate in the Japanese news media. Shortly after Ito's move to Singapore, Japan's Minister of Science made a serious effort to accelerate reform of modern biology in Japan, an exercise its former Minister of Education started 20 years ago. Although Japan annually invests as much as the US on a per capita basis in biomedical research, by several measures, it is a distant third in the world of modern biology. The decision was made to build a graduate university for modern biology in Okinawa. The president of this graduate university shall be an eminent scientist from outside of Japan; the faculty and the students will consist of a cosmopolitan mix of non-Japanese and Japanese researchers, and the language of instruction will be English. The purpose is to reform modern biology in Japan by enticing promising scientists into Japan through Okinawa. Will this be a successful experiment for Japan even though some consider Okinawa too far away from the centers of excellence on the main island of Japan?
Good science will flourish wherever the conditions are most friendly, as it did in the US in the last century. Many factors, such as good funding, cost per discovery, the ready availability of a large pool of young talent, cultural acceptability of non-natives to Asian societies, and good living conditions must be in place. As knowledge-based societies become a reality, industry worldwide will outsource discovery to wherever the job can be done well and at the lowest cost. In the 20th century, the US was the Mecca of science, attracting the giants of science from the rest of the world. Economic conditions are different today; there is abundant wealth and young talent in Asia. For Asia to become a Mecca of science in the 21st century, countries in the region need to take a long-term view and be as eager to promote scientific collaboration with each other as they are to collaborate with the US.
Without a doubt, the science star is shining over Asia, but Asian science has a glaring deficiency in being highly “Balkanized”. In Europe, support of research crosses national boundaries and is facilitated by the European Molecular Biology Organization. Today research funding crosses few national borders in Asia. Notwithstanding this, eight years ago a small group of visionary scientists led by Kenichi Arai, the former Director of the Institute of Medical Sciences at the University of Tokyo, formed the equivalent of the European Molecular Biology Organization in Asia called the Asia–Pacific International Molecular Biology Network. To date the Asia–Pacific International Molecular Biology Network has 300 scientists in its network, representing key scientists from 16 countries in Asia. In spite of a lack of cross-border funding, the Asia–Pacific International Molecular Biology Network has managed to conduct workshops and symposia annually through innovative fundraising schemes. Good science is on the move in Asia, and when Asia's talents are mobilized, it will move the world as never seen before.
Citation: Tan CYH (2005) Science star over Asia. PLoS Biol 3(9): e322.
Chris Y. H. Tan is a senior counsellor at Asia–Pacific International Molecular Biology Network, Vancouver, British Columbia, Canada. E-mail: [email protected]
Abbreviations
IMCBInstitute of Molecular and Cell Biology
hESChuman embryonic stem cell
TCMtraditional Chinese medicine
==== Refs
References
Hwang WS Ryu YJ Park JH Park ES Lee EG Evidence of a pluripotent human embryonic stem cell line derived from a cloned blastocyst Science 2004 303 1669 1674 14963337
Hwang WS Roh SI Lee BC Kang SK Kwon DK Patient-specific embryonic stem cells derived from human SCNT blastocytes Science 2005 308 1777 1783 15905366
Bongso A Fong CY Ng SC Ratnam S Isolation and culture of inner cell mass cells from human blastocysts Hum Reprod 1994 9 2110 2117 7868682
Richards M Fong CY Chan WK Wong PC Bongso A Human feeders support prolonged undifferentiated growth of human inner cell masses and embryonic stem cells Nat Biotechnol 2002 20 933 936 12161760
Medical Research News Biotoxin from puffer fish may provide treatment for opioid withdrawal 2004 November 18 Sydney (Australia) News–Medical Net Available: http://www.news-medical.net/?id=6367 . Accessed 25 July 2005
[Anonymous] Tetrodotoxin is safe and effective for severe, refractory cancer pain J Support Oncol 2004 2 18 15330369
Aparicio A Chapman JE Putnam N Chia J Dehal P Whole-genome shotgun assembly and analysis of the genome of Fugu rubripes
Science 2002 297 1301 1310 12142439
Yu J Hu S Wang J Wong GK Li S A draft sequence of the rice genome Science 2002 296 79 92 11935017
Shen ZX Chen CQ Ni JH Li XS Use of arsenic trioxide in the treatment of acute promyelocytic leukemia Blood 1997 89 3354 3360 9129042
Bjorkman A Bhattarai A Public health impact of drug resistant Plasmodium faciparum malaria Acta Trop 2005 94 163 169 15893289
Liu L Zhao SP Cheng YC Li YL Xuezhikang decreases lipoproteins and C-reactive protein concentrations in patients with coronary heart diseases Clin Chem 2003 49 1347 1352 12881451
|
16149850
|
PMC1201306
|
CC BY
|
2021-01-05 08:21:27
|
no
|
PLoS Biol. 2005 Sep 13; 3(9):e322
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030322
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030325Book Reviews/Science in the MediaBiophysicsNoneA Night Out with the Nerds Book Review/Science in the MediaKnapp Sandra Mallet James [email protected] 2005 13 9 2005 13 9 2005 3 9 e325Wiseman R, Singh S (2005) Theatre of science [stage production]. London: Soho Theatre. Produced 4 July 2005-19 July 2005. Copyright: © 2005 Knapp and Mallet.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.Simon Singh and Richard Wiseman draw on examples from physics to psychology, to explore the extraordinary in the ordinary in their innovative new play Theatre of Science
==== Body
Add performance to science, and the equation results in Theatre of Science. Part scientific lecture, part magic show, and part music and dance, Theatre of Science is an innovative collaboration between Simon Singh and Richard Wiseman. The cosy atmosphere of the Soho Theatre, stuffed with a lively crowd, which appeared swelled by a smattering of family and friends, makes the show a personal interchange between performers and audience. Several in the audience must have been scientists, judging by appearances: one enthusiastic member of the audience was a dead ringer for the recently outgoing president of the Royal Society, only taller and more etiolated. Every show will be different, depending on the audience's reaction. Wiseman, in a warm-up session, explicitly “titrates” the audience for its sense of humour, before embarking on the rest of the show. Here is an example: “We were hoping that our contortionist Delia would be here tonight, but, I promise you, this is not a joke, she called in a few moments ago to say she was tied up in traffic”. With the audience ready, Singh demonstrates “the scientific method” by proving mathematically that Teletubbies = evil. As a hint, the proof involves the well-known law, “Money is the root of all evil”.
Using amusing and wacky examples from physics and psychology, Singh and Wiseman explore the extraordinary in the ordinary. Ground-breaking science often emerges from massive pieces of equipment, such as mile-long linear accelerators or radio telescopes, but science—much of it still unexplored—pervades the seemingly ordinary, everyday world as well. For example, we are now convinced of profound “holes in our understanding” of the physics of balloons. (How did he put that knitting needle through, and then withdraw it again, without the balloon popping?)
Singh is a well-known broadcaster, popular science writer, and a successful scientist who obtained his PhD in physics at Cambridge University and the European Organization for Nuclear Research (CERN) (see http://www.simonsingh.net). Wiseman is a magician (a failed magician, he says, as he deliberately drops a card he has palmed). Instead, he is now the world's only Professor of Public Understanding of Psychology, at the University of Hertfordshire (see http://www.richardwiseman.net). Wiseman's forte is optical illusion, the magician's stock-in-trade, but most impressive is his exploration of quirks of our perception of ordinary things. “Psychologists today earnestly debate if anything we see is real at all”, he says. If you think you are a great observer—and as scientists we are generally proud of our powers of objective observation—just relax and participate in Theatre of Science. You'll be amazed how easily you can be tricked by Wiseman's ruses.
Mixed in with Singh's three-minute explanation of the Big Bang (more or less, if you exclude his five-minute appendix), we see a clear demonstration of the inevitability of coincidence. In The Bible Code, Michael Drosnin argues that the Bible contains many prophecies of today's events; however, the Theatre of Science demonstrates that Herman Melville's Moby Dick has even more startlingly correct predictions…about Princess Diana's death, for instance. In another piece, an alarmingly glowing orange and sparking gherkin is followed by beautiful, weird sounds from theramin player Sarah Angliss and her assistant. The theramin is an electronic musical instrument consisting of a rod and a hoop, both attached to tuned circuits that are connected to speakers. The theramin is played with hand and body movements, which affect volume and pitch; it is as much dance as it is music, and the sound is ethereal.
Contortionists make us wince—but how do they do that with a body that looks just like ours? Is it their bones, their ligaments, or something quite different? Inserting contortionist Delia Du Sol into a whole-body magnetic resonance imaging scanner provides the answer—the images tell an amazing story. (Guess! We're not giving it away.) Then Du Sol's performance—accompanied by theramin music and overlaid with recordings of her talking about how she felt while performing, in particular how she hoped that people didn't perceive her as a freak—opened up a new dimension. The contrast between her matter-of-fact speech and the tension in the audience was unnerving.
Although the show doesn't end with a thunderous bang, it does have real lightning—every bit as good. This final experiment (or do we mean skit?) requires two six-foot-high, out-of-phase, coupled Tesla coils producing a combined total of a million volts, we are told. Initially, in spite of the big numbers, we aren't impressed by the voltage (don't combs generate similar voltages in dry hair?), but when we see and hear the arcs jumping six feet and smell the burnt air (thought to be caused by nitrogen/oxygen compounds forming in the plasma created by the arcs), we realize there are quite a few amps as well as volts. Audience tension is slowly and carefully building. The electrifying finale is no set-up job, and there is real potential for danger. Those wearing pacemakers are advised to leave the room to avoid certain death. This is supreme showmanship, part circus, part laboratory, and part drama. Do we, the audience, viscerally believe in scientific theory?
We vote on whether Singh or Wiseman will occupy the Faraday cage, also known as the “coffin of terror”, being inserted between the “coils of death”. We see and hear lightning strike the humanoid form (which looks like it was constructed to house the son of Frankenstein) now containing Wiseman. After the discharges, the sounds die away, and there follows a long period of silence and immobility. Singh then moves to slowly and deliberately open the Faraday cage – the tension mounts. Suddenly, Wiseman leaps out of the “coffin of terror”, saying “the physics was right, thank you!”
The equation works, and we can be proud to be nerds. It is a blend of science, humour, and performance that might not have worked at all, but it most emphatically did, and we recommend it as a great night out for scientists and nonscientists alike. We hope for a return engagement soon!
Citation: Knapp S, Mallet J (2005) A night out with the nerds. PLoS Biol 3(9): e325.
James Mallet is in the Department of Biology, University College London, London, United Kingdom. Sandra Knapp is at the Natural History Museum, London, United Kingdom.
==== Refs
Performance Reviewed
Wiseman R Singh S Theatre of science [stage production] 2005 London Soho Theatre Produced 4 July 2005–19 July 2005
|
0
|
PMC1201307
|
CC BY
|
2021-01-05 08:21:27
|
no
|
PLoS Biol. 2005 Sep 13; 3(9):e325
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030325
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1614985110.1371/journal.pbio.0030326Community PageScience PolicyNonePeer Review—The Newcomers' Perspective Community PageMainguy Gaell Motamedi Mohammad R Mietchen Daniel [email protected] 2005 13 9 2005 13 9 2005 3 9 e326Copyright: © 2005 Mainguy 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 World Academy of Young Scientists argue that double blind peer-review will generate a better perception of fairness and equality in global scientific funding and publishing
==== Body
Created under the auspices of the United Nations Educational, Scientific, and Cultural Organization (UNESCO) as an offspring of the “International Forum of Young Scientists,” the World Academy of Young Scientists (WAYS) was officially launched in November 2003 at the World Science Forum in Budapest, Hungary. Our organization represents a permanent global platform for young researchers, and presently gathers some 2,000 members in all disciplines from about 100 countries. WAYS benefits from the support of a number of distinguished senior scientists, including several Nobel laureates. Our objectives are to make science more attractive, comprehensible, and accessible, and to support career development opportunities for young scientists from around the world. WAYS encourages interdisciplinary collaboration and networking among scientists, irrespective of their age or institutional affiliations. We provide a global forum to communicate the opinions, concerns, and questions of young scientists to decision-makers in science policy.
At our first general assembly in December 2004 in Marrakech, Morocco, peer-review procedures in scientific publication and research funding were debated intensely. Even though peer review is universally accepted as an essential element of research, considerable debate persists on how to implement it. The vast majority of our members, especially from developing countries, were concerned about the apparent unfairness of the current procedure, a perception that is prone to generate frustration, fear of discrimination, and distrust. We reached a consensus that slight modifications to the current review process would help in getting more objective reviews based on the quality of the research rather than the age, affiliation, gender, or pedigree of the authors.
Single-blind peer review (SBPR), in which the reviewer knows the identity of the author but not vice versa, is the currently accepted practice. Because SBPR can be vulnerable to sexism and nepotism [1], its ethical foundations have come under criticism; the method is frequently recognized to be biased against new ideas, women, young scientists, career changers, and scholars from less prestigious universities and/or from developing countries (see [2] and references therein). Generally, two policies have been proposed to eliminate bias from the peer-review process: open peer review and double-blind peer review (DBPR).
We believe that current peer-review process, even though functional, can be, and should be, improved.
In open peer review, the identities of both authors and reviewers are revealed, affording the authors the ability to identify the reviewers' comments to a person. Even though this might be an equitable strategy to prevent unfair rejections, this process has no safeguard against unfair acceptance of papers—reviewers, and especially newcomers, may feel pressured into accepting a mediocre paper from a more established lab in fear of future reprisals.
DBPR, in which both the reviewers and the authors remain anonymous to each other, is thought to disentangle the peer-review process from non-scientific factors, thereby presenting an appealing alternative. The a priori case for masking and blinding is strong, and several studies have suggested that articles published in DBPR journals were cited significantly more often than articles published in non-DBPR journals [3,4]. However, other studies have been less convincing; critics of DBPR argue that it is difficult to hide the identity of the institution, laboratory, and/or authors of a paper from the reviewers, especially in smaller specializations. For instance, in a DBPR policy trial, despite explicit instructions to authors, 34% of prospectively evaluated manuscripts contained hints to unblind the authors, and editors correctly identified the authors or institutions of 25% of the manuscripts [5]. The disconnection between principle and practice is evident, and so far, few journals, and even fewer in biomedical sciences, have implemented DBPR policies. The reasons appear to be partly historical, as journals are used to SBPR, and partly intellectual, as the benefits of DBPR still remain controversial [6].
Maintenance of trust within the international scientific community is crucial, not only for future scientific development, but also to continue the dialogue of civilizations. We believe that the current peer-review process, even though functional, can be, and should be, improved to bolster a more even playing field for all scientists. In biomedical sciences, the effectiveness of DBPR is hotly debated. However—using data from computer science, philosophy, or economics, which have adopted and have been using DBPR for some time—the inescapable conclusion is that DBPR performs at least as well as the traditional peer-review process. We propose here that DBPR is a better system because, in addition to being a reasonably fair process, it also bears symbolic power that will go a long way to quell fears and frustrations, thereby generating a better perception of fairness and equality in global scientific funding and publishing. This will, in turn, help to keep research more accessible for future generations.
Citation: Mainguy G, Motamedi MR, Mietchen D (2005) Peer review—The newcomers' perspective. PLoS Biol 3(9): e326.
Gaell Mainguy and Mohammad R. Motamedi, who made an equal contribution to this work, and Daniel Mietchen, are all with the World Academy of Young Scientists (http://www.waysnet.org).
Abbreviations
DBPRdouble-blind peer review
SBPRsingle-blind peer review
WAYSWorld Academy of Young Scientists
==== Refs
References
Wenneras C Wold A Nepotism and sexism in peer-review Nature 1997 387 341 343 9163412
Abate T What's the verdict on peer review? 1996 Available: http://www.columbia.edu/cu/21stC/issue-1.1/peer.htm . Accessed 20 July 2005
McNutt RA Evans AT Fletcher RH Fletcher SW The effects of blinding on the quality of peer review. A randomized trial JAMA 1990 263 1371 1376 2304216
Laband DN Piette MJ A citation analysis of the impact of blinded peer review JAMA 1994 272 147 149 8015128
Katz DS Proto AV Olmsted WW Incidence and nature of unblinding by authors: Our experience at two radiology journals with double-blinded peer review policies Am J Roentgenol 2002 179 1415 1417 12438028
Davidoff F Masking, blinding, and peer review: The blind leading the blinded Ann Intern Med 1998 128 66 68 9424984
|
16149851
|
PMC1201308
|
CC BY
|
2021-01-05 08:21:28
|
no
|
PLoS Biol. 2005 Sep 13; 3(9):e326
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030326
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1614985210.1371/journal.pbio.0030327EditorialOtherReplication Publication EditorialPatterson Mark [email protected] Lon 9 2005 13 9 2005 13 9 2005 3 9 e327Copyright: © 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.
==== Body
In December 2003, PLoS Biology published a research article reporting evidence of an association between a gene called GAD2 and susceptibility to obesity (DOI: 10.1371/journal.pbio.0000068). Although the genetic data were suggestive rather than conclusive, the editors and reviewers supported publication because unambiguous evidence in association studies is notoriously difficult to obtain and obesity is such a major burden on public health. Enthusiasm was heightened because GAD2 lies in a region of human Chromosome 10 that is thought to contain a gene influencing obesity, and GAD2 itself is involved in the synthesis of a neurotransmitter implicated in the regulation of food intake. Publication would allow others to test just how important the connection between GAD2 and obesity might be—and the enticing prospect was that insight into the role of GAD2 in this disease might lead ultimately to new therapeutic approaches.
In this issue of PLoS Biology, we publish a follow-up to the original study, from a separate team of researchers based at the University of California at San Francisco (DOI: 10.1371/journal.pbio.0030315). Swarbrick et al. looked at different and much larger patient populations and analyzed variation within GAD2 more comprehensively, but disappointingly, the new study was not able to replicate the initial finding.
Despite the negative conclusions, however, the reviewers were no less enthusiastic about the publication of this paper than the first. One of the reasons for this is related to a problem that has beset the field of complex disease genetics for several years—a tendency towards the publication of studies that show an association.
Diseases are termed complex when their etiology is influenced by many factors, both genetic and environmental. These diseases include obesity, diabetes, mental illness, and many more common ailments. But finding the genes that are involved with these diseases is an extremely difficult problem: individual genes have only a limited effect, which means that large patient populations need to be studied; specific genes might have effects only in certain ethnic groups; and statistical artifacts can be hard to eliminate. These and other problems mean that success stories are few and far between, whereas false leads have been plentiful. High-profile journals, in particular, have therefore tended to publish the positive results, whereas negative data often end up in specialist literature or, much worse, don't get published at all.
Because the identification of a complex disease gene is so difficult, the first positive report is rarely definitive—replication studies that use independent populations with sample sizes large enough to detect the expected effects are necessary to bolster (or undermine) the case. A bias towards publishing positive results might therefore perpetuate the false impression that a gene is indeed associated with a disease. In a field this complex, such bias is entirely unhelpful.
Replication studies are therefore vital to the field of human genetics, and at PLoS Biology, we've taken the view that well-designed, and high-powered replication studies can be just as worthy of publication in the journal as the initial finding of a genetic association—and PLoS's community journal PLoS Genetics takes a similar view. The new paper on GAD2 demonstrates this point—the study was judged by the reviewers to be sufficiently large and well conducted to provide a robust test of the hypothesis that variation in GAD2 is involved—above a certain minimum effect—in susceptibility to obesity. The negative conclusion suggests that researchers now need to consider possible reasons for the variability in results such as sample or phenotype differences, or else examine the possibility that other variants in the Chromosome 10 region might be involved in obesity. It might ultimately be that the initial finding simply does not hold in larger samples. That replication papers are important for the field is supported by the publication of another large study in type 2 diabetes (DOI: 10.1371/journal.pbio.0000020), which included replication data, and has already been cited 28 times.
Would editorial opinion of the paper by Swarbrick et al. have been any different if the result had been positive? We think not, because the first study left sufficient room for doubt, and a positive result could have been just as significant. In general, then, we will judge a replication paper on its merits, and consider whether the submitted work provides a major advance on the previous genetic findings.
But PLoS Biology and PLoS Genetics are both highly selective journals. By themselves they cannot solve the problem, and will only publish the most significant association studies—whether the results are positive or negative. Furthermore, the accumulation of genetic association data is likely only to accelerate in the coming years, thanks to major collaborative efforts such as the SNP Consortium (http://snp.cshl.org) and the International HapMap Project (http://www.hapmap.org). Public availability of these rich resources will stimulate genome-wide efforts to map the genes underlying any number of complex disorders. Data will abound, and they need to be published and also made publicly available. Most of the individual studies will not provide definitive evidence of associations, but as the data build, meta-analyses will become more and more informative.
PLoS is currently exploring the possibility of providing further publication opportunities to extend the venues that are available for association studies. As with all PLoS articles, the data and the associated papers would be freely available to read and reanalyze. With the introduction of more such open-access publications and with support from the community, it would be possible to eliminate bias in the publication of association studies for good.
Citation: Patterson M, Cardon L (2005) Replication publication. PLoS Biol 3(9): e327.
Mark Patterson is Senior Editor for PLoS Biology. Lon Cardon is at the University of Oxford, Oxford, United Kingdom, and is a member of the PLoS Biology Editorial Board.
|
16149852
|
PMC1201309
|
CC BY
|
2021-01-05 08:21:27
|
no
|
PLoS Biol. 2005 Sep 13; 3(9):e327
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030327
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 1616339410.1371/journal.pcbi.001003605-PLCB-RA-0080R4plcb-01-04-01Research ArticleBioinformatics - Computational BiologyCell BiologyDiabetes - Endocrinology - MetabolismSystems BiologyEukaryotesA Biophysical Model of the Mitochondrial Respiratory System and Oxidative Phosphorylation Biophysical Model of Oxidative PhosphorylationBeard Daniel A Biotechnology and Bioengineering Center, Department of Physiology, Medical College of Wisconsin, Milwaukee, Wisconsin, United States of AmericaBourne Philip EditorUniversity of California at San Diego, United States of AmericaE-mail: [email protected] 2005 9 9 2005 1 4 e3619 4 2005 3 8 2005 Copyright: © 2005 Beard.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 computational model for the mitochondrial respiratory chain that appropriately balances mass, charge, and free energy transduction is introduced and analyzed based on a previously published set of data measured on isolated cardiac mitochondria. The basic components included in the model are the reactions at complexes I, III, and IV of the electron transport system, ATP synthesis at F1F0 ATPase, substrate transporters including adenine nucleotide translocase and the phosphate–hydrogen co-transporter, and cation fluxes across the inner membrane including fluxes through the K+/H+ antiporter and passive H+ and K+ permeation. Estimation of 16 adjustable parameter values is based on fitting model simulations to nine independent data curves. The identified model is further validated by comparison to additional datasets measured from mitochondria isolated from rat heart and liver and observed at low oxygen concentration. To obtain reasonable fits to the available data, it is necessary to incorporate inorganic-phosphate-dependent activation of the dehydrogenase activity and the electron transport system. Specifically, it is shown that a model incorporating phosphate-dependent activation of complex III is able to reasonably reproduce the observed data. The resulting validated and verified model provides a foundation for building larger and more complex systems models and investigating complex physiological and pathophysiological interactions in cardiac energetics.
Synopsis
Cells are able to perform tasks that consume energy (such as producing mechanical force in muscle contraction) by using chemical energy delivered in the form of a chemical compound called adenosine triphosphate, or ATP. Two Nobel Prizes were awarded (in 1978 to Peter D. Mitchell and in 1997 to Paul D. Boyer and John E. Walker) for the determination of how ATP is synthesized from the components adenosine diphosphate (ADP) and inorganic phosphate in a subcellular body called the mitochondrion. The operating theory, called the chemiosmotic theory, describes how a driving force called the proton motive force, which arises from the sum of contributions from the electrical potential and the hydrogen ion concentration difference across the mitochondrial inner membrane, is developed by reactions catalyzed by certain enzymes and consumed in generating ATP. Yet, to date, no computer model has successfully described the development and consumption of both the chemical and electrical components of the proton motive force in a thermodynamically balanced simulation. Beard introduces such a model, which is extensively validated based on previously published sets of data obtained on isolated mitochondria. The model is used to test hypotheses about how intracellular respiration is regulated; this model could serve as a foundation for investigating the control of mitochondrial function and for developing larger integrated simulations of cellular metabolism.
Citation:Beard DA (2005) A biophysical model of the mitochondrial respiratory system and oxidative phosphorylation. PLoS Comput Biol 1(4): e36.
==== Body
Introduction
As the key cellular organelle responsible for transducing free energy from primary substrates into the ATP potential that drives the majority of energy-consuming processes in a cell, the mitochondrion plays a central role in the majority of eukaryotic intracellular events. Therefore, the development of a quantitative mechanistic understanding of cellular function must rely on a reasonable quantitative description of mitochondrial function. Additionally, development of computational models of physiological systems that span multiple scales, from intracellular biochemistry to whole-organ function, requires a self-consistent integrated description of the biophysical processes observed at the molecular, cellular, tissue, and whole-organ levels of resolution.
Recognizing the need for a computational model of mitochondrial energetics and the need that such a model be available for integration with other physiological systems models, a computational model of the biophysics of the respiratory system and oxidative phosphorylation was developed to meet the following requirements: (1) the model must be consistent with the available experimental data, and (2) the model must be constrained by the relevant physics/biophysics. The utility of the first requirement is self-evident. The second requirement is that models must obey applicable physical laws (e.g., conservation of mass, charge, and energy; the Second Law of Thermodynamics). The alternative to the second requirement is to use empirically derived relationships that are often useful in developing data-driven models based on specific datasets. This approach is not used here because the resulting models often fail when combined together. Physics-based models, on the other hand—i.e., models built on principles including the laws of mechanics and thermodynamics, in which assumptions and approximations are made explicit—operate with a common currency of mass, charge, energy, and momentum [1–3]. Such models naturally integrate across disparate scales.
Previous models of oxidative phosphorylation fail to meet either one or both of the above requirements. For example, the widely used model of Korzeniewski and Zoladz [4–6] invokes an empirical linear relationship between the difference in pH across the mitochondrial inner membrane (matrix pH minus cytosol pH) and the magnitude of inner membrane potential. While the Korzeniewski model has been validated and verified based on a number of studies and is widely applied [7–9], the central empirical relationship between electrostatic potential and pH difference is not expected to apply under all conditions. For example, in the extensive study of isolated cardiac mitochondria published by Bose et al. [10], this relationship is not obeyed. In fact, not only is the linear relationship violated, it is observed in the study of Bose et al. [10] that the matrix pH is nearly constant and can even drop as the magnitude of membrane potential increases. For example, the matrix can become more acidic as the magnitude of the membrane potential increases. This phenomenon cannot be explained by a model that collapses the proton concentration gradient, the membrane potential, and the proton motive force into a single state variable.
The model developed by Magnus–Keizer [11] was integrated by Dudycha and Jafri and colleagues into a detailed model of mitochondrial metabolism, including the reactions of the TCA cycle [12,13]. The Dudycha–Jafri model has been adopted by Cortassa and co-workers and recently extended to account for the production of reactive oxygen species by the respiratory chain [14,15]. The Magnus–Keizer model, developed based on Hill's formalism for biochemical kinetics and free energy transduction [16], is self-consistent and thermodynamically balanced. While it has the advantage over the Korzeniewski model in that the electrostatic potential is treated as a state variable, charge is balanced, and bulk electroneutrality is obeyed in the steady state, the Magnus–Keizer model treats the pH gradient across the inner membrane as a constant. Also, the Magnus–Keizer model addresses certain aspects of mitochondrial ion transport—specifically the transport of calcium across the inner membrane—that are not included at the present stage in the current model. However, details that are unique to the present model—specifically pH buffering by potassium ion exchange [17,18]—are necessary to analyze the extensive experimental dataset considered here.
The goal of the current work is to introduce a quantitative model representing the chemiosmotic theory of the respiratory chain and ATP synthesis that treats the proton gradient and mitochondrial membrane potential as distinct state variables. To be of maximal utility, this biophysically based model of mitochondrial respiration must conserve charge, mass, and energy, and use a treatment of the electron transport system, ATP synthesis, and substrate transporters that does not violate the laws of thermodynamics. The model is developed based on the dataset of Bose et al. [10]. Values of 16 adjustable parameters are estimated based on fitting a relatively large number of experimentally measured data curves (nine data curves). The resulting parameterized model is further validated by comparison to data measured at low oxygen level in isolated mitochondria by Wilson et al. [19] and Gnaiger and Kuznetsov [20,21].
Results
This section is organized as follows. First, a thermodynamically balanced biophysical model of mitochondrial oxidative phosphorylation is developed, along with identification of model parameters based on the measured dependence of NADH and cytochrome c redox states, oxygen consumption, and matrix pH on levels of buffer phosphate. It is shown that the developed model cannot match data on the inner membrane electrostatic potential without incorporating phosphate-dependent control of oxidative phosphorylation. In the next section, the isolated mitochondrial model is modified to include phosphate-dependent activation of complex III and is fit to the data on inner membrane potential. In the final section, the behavior of the model at low oxygen level is shown to compare favorably to data reported by Gnaiger and Kuznetsov [20,21] and Wilson et al. [19].
Mitochondrial Model without Phosphate Control
The basic components of the mitochondrial model, which include the reactions at complexes I, III, and IV of the respiratory chain and ATP synthesis at F1F0 ATPase, are illustrated in Figure 1A. Substrate transporters, including adenine nucleotide translocase (ANT) and the phosphate–hydrogen co-transporter (PiHt) are illustrated in Figure 1B. Cation fluxes across the inner membrane, illustrated in Figure 1C, include fluxes through the K+/H+ antiporter and passive H+ and K+ permeation. The model includes two mitochondrial compartments (matrix and intermembrane [IM] space). In the following description of the model equations, subscripts “x”, “i”, and “e” denote concentrations of reactants in the matrix, IM space, and external space, respectively; the concentration variables [fATP]i and [fATP]x, and [fADP]i and [fADP]x, denote the magnesium-unbound components of ATP in the IM space and the matrix, and unbound ADP concentrations in the IM space and matrix, respectively; variables [mATP]i and [mATP]x, and [mADP]i and [mADP]x, denote the magnesium-bound components of ATP in the IM space and the matrix, and magnesium-bound ADP concentrations in the IM space and matrix, respectively. The external space corresponds to the extra-mitochondrial buffer used in the experiments simulated below.
Figure 1 Illustration of the Components Included in the Model of Mitochondrial Oxidative Phosphorylation
(A) The major components of the electron transport system, which transfers reducing potential from NADH to oxygen, and the F1F0 ATPase, which transduces energy from proton motive force to ATP, are illustrated. Complexes I, III, and IV are labeled C1, C3, and C4, respectively.
(B) The substrate transport process included in the model is shown, including the ANT and PiHt on the inner membrane, and passive permeation of ATP, ADP, AMP, and phosphate across the outer membrane. The AK reaction in the IM space is shown.
(C) Transporters for hydrogen and potassium ions on the inner membrane, including K+/H+ antiporter and passive proton and potassium fluxes, are included. It is assumed that these cations rapidly equilibrate across the outer membrane.
Note on units.
All concentrations in the following sections are expressed in molar units—specifically moles per liter of compartment volume. Fluxes are expressed in units of mass per unit time per unit mitochondrial volume. To avoid confusion, the units on flux variables are written as “mol s−1 (l mito volume)−1” and not simplified to “M s−1”, which would be ambiguous when referring to fluxes for membrane transporters.
Dehydrogenase flux.
In this model the TCA cycle and other NADH-producing reactions are not explicitly modeled. Instead, a phenomenological driving force is used to simulate the phosphate-dependent rate of reduction of NAD+ to NADH via the reaction NAD+
NADH + H+ in the mitochondrial matrix. The following expression is used to model the dehydrogenase flux:
where [NADH]x and [NAD]x denote the concentrations of NADH and NAD in the mitochondrial matrix; [Pi]x denotes the matrix inorganic phosphate concentration, and XDH, r, k
Pi,1, and k
Pi,2 are empirical parameters. Thus, the dehydrogenase flux drives the [NADH]x/[NAD]x ratio toward r, with a reaction rate dependent on the concentration of phosphate, which serves as a substrate for mitochondrial dehydrogenase enzymes.
Electron transport system fluxes.
The overall reaction for electron transfer from NADH to ubiquinol at complex I is expressed as
where the notation ΔH+ is used to indicate hydrogen ion transferred from the matrix to the cytosol, against the electrochemical gradient. The direction of positive flux for the reaction of equation 2 and for all reactions introduced below is left-to-right.
The flux through complex I is modeled using the following expression:
where X
C1 is an adjustable parameter, [Q] and [QH2] denote oxidized and reduced ubiquinol, RT is the gas constant multiplied by the absolute temperature; ΔGo,C1 = −69.37 kJ mol−1 is the standard free energy for the reaction H+ + NADH + Q
NAD+ + QH2 at pH = 7; and ΔG
H is the proton motive energy, or the free energy change associated with pumping a proton from the matrix side to the cytosol side of the mitochondrial inner membrane. The factor of four in the exponent of equation 3 arises from the four protons pumped from the matrix space to the IM space for each pair of electrons transferred from NADH to ubiquinol. The proton motive energy is computed as follows:
where F is Faraday's constant, ΔΨ is the membrane potential measured as the outer potential minus the inner potential, and [H+]e/[H+]x is the ratio of external hydrogen ion concentration to matrix concentration. Equation 3 represents a minimal thermodynamically balanced one-parameter model for the flux through the first step in respiratory system—i.e., the flux described by equations 3 and 4 drives concentrations towards thermodynamic equilibrium. Specifically, the flux is driven towards a thermochemical equilibrium defined by the effective equilibrium expression
which is an explicit function of the proton gradient and electrostatic gradient across the inner membrane.
For complex III, it is assumed that four protons are pumped for each pair of electrons transferred from ubiquinol to cytochrome c [22,23]:
where cytC(ox)3+ and cytC(red)2+ denote the oxidized and reduced forms of cytochrome c, respectively. As indicated in Figure 1, cytochrome c is assumed to be present in the IM space. Although four protons are pumped across the inner membrane for each unit flux through this reaction, the total number of charges transferred is two, owing to the redox transfer from ubiquinol to cytochrome c, which generates two matrix hydrogen ions for each turnover of the reaction. The flux through complex III takes a form similar to equation 3:
where X
C3 is an adjustable parameter, ΔGo,C3 = −32.53 kJ mol−1. Below, the expression for complex III flux is modified to test the hypothesis that phosphate modulates complex III activity. It will be shown that, while much of the available experimental data can be explained by the model developed in this section, which does not consider phosphate-dependent control of the respiratory chain enzymes, the observed data are better fit by a model that incorporates phosphate-dependent control of complex III activity.
The overall reaction of complex IV involves the transfer of two protons across the membrane and a total of four charges [24,25]:
with a flux computed as
where X
C4 is an adjustable parameter, ΔGo,C4 = −122.94 kJ mol−1, and cytCtot = [cytC(ox)3+] + [cytC(red)2+].
The complex IV flux of equation 9 is expressed in a form similar to those of complexes I and III, with a few key differences. While, as for complexes I and III, the flux is formulated to drive the system toward thermodynamic equilibrium, additional multiplicative factors have been included in order to successfully reproduce the observed data. The factor
is included in equation 9 to account for the observed dependence of the rates of oxygen consumption and ATP generation on oxygen concentration [6,19–21,26,27]. It will be shown that the factor [cytC(red)2+]/cytCtot is found to provide better fits to the observed data than are possible without it.
ATP synthesis.
ADP is phosphorylated to ATP in the matrix via the F1F0-ATPase reaction:
where nA ≈ 3 is the number of protons transported each time this reaction turns over. Since ATP synthesis requires magnesium as a cofactor, the flux through this complex is modeled using the thermodynamically balanced expression
where X
F1 is an adjustable parameter, ΔGo,ATP = −36.03 kJ mol−1 and K
Mg-ATP and K
Mg-ADP >are the equilibrium dissociation constants for ATP and ADP binding with Mg2+. The factor 1 M multiplying [mATP]x is used so that the term in parenthesis is balanced in terms of units.
Magnesium binding.
Binding between magnesium ion and ATP and ADP is driven via the following fluxes:
fATP and fADP denote the concentrations of ATP and ADP that are not bound to magnesium ion in the matrix and IM space; mATP and mADP denote the magnesium-bound species. The parameter X
MgA is the forward binding rate constant for these reactions; the effective unbinding constant is computed to satisfy the equilibrium dissociation relations for ATP and ADP binding with Mg2+.
Substrate transport.
Permeation of ATP, ADP, AMP, and inorganic phosphate between the external buffer and the IM space is governed by the following fluxes:
where the subscripts “e” and “i” denote external buffer and IM space, respectively. The buffer concentrations are set as constants in this study. The permeabilities of the outer membrane to adenine nucleotides and to inorganic phosphate are given by p
A and p
Pi, respectively; γ denotes the ratio of mitochondrial outer membrane area to total cardiomyocyte cell volume.
ANT flux involves the displacement of one negative charge from the matrix to the IM space, and is therefore coupled to the electrostatic membrane potential. The following empirical expression [4,6] is used to model the ANT flux:
where the ANT is assumed to operate on magnesium-unbound ATP and ADP in the two compartments.
Transport of inorganic phosphate between the matrix and IM space is coupled to the hydrogen ion gradient [28]. It is assumed that H+ and
are transported by a co-transport process, with H+ and
moving together across the membrane in a 1:1 ratio in a net electroneutral exchange. Hydrogen binding to inorganic phosphate via the reaction
is assumed to be in equilibrium on either side of the membrane with
and
, where k
dH is the dissociation constant for the reaction. In these expressions Pi represents the sum of species
and
. The phosphate-hydrogen co-transporter flux is modeled as reversible Michaelis–Menten flux:
where X
PiHt is an adjustable parameter, k
PiHt is the Michaelis–Menten constant for
on the outside of the membrane.
Adenylate kinase reaction.
High-energy phosphates are transferred between ATP, ADP, and AMP in the IM space via the Adenylate kinase (AK) reaction:
The AK flux in the IM space is computed as follows:
where K
AK = 0.4331 is the equilibrium constant for the reaction of equation 17, and X
AK is the AK enzyme activity.
Cation transport.
The present work assumes that calcium and sodium concentrations and fluxes have only secondary effects on membrane potential compared to the primary effects of currents associated with the respiratory chain, the ANT current, and the proton leak. Therefore, fluxes of sodium and calcium are not considered at this stage. Since K+ is required to buffer the matrix pH [17] and Mg2+ is required for ATP synthesis and the ANT flux, these ions are considered in the model. Expressions for K+ and Mg2+ channel and transporter fluxes are developed below.
The expression for the leak of H+ across the inner membrane is obtained by solving the one-dimensional Nernst–Planck equation, the differential equation for diffusion and drift of a charged species across a permeable membrane. The resulting flux is calculated from the Nernst–Goldman equation [29,30]:
Passive flux of potassium into the matrix is modeled using a similar expression:
While significant evidence exits for passive flux of potassium through various channels into the matrix [17,31], it is unclear exactly what transporters are present to prevent potassium concentrations from approaching thermochemical equilibrium across the inner membrane. It is assumed that the outflow of potassium ions from the matrix is coupled to the proton gradient, and outflow is modeled using a simple reversible antiporter with flux given by mass-action kinetics:
The above expressions for K+ and H+ transport assume that these ions rapidly equilibrate across the outer membrane. Therefore, the IM space concentrations are assumed to be equal to the external space concentrations.
Governing equations.
The flux expressions are used to construct a kinetic model for the system; the overall system is governed by the following set of 17 differential equations:
where V
x and V
i are the matrix and IM space water volumes, and r
buff is the buffering capacity of the matrix space, which is set to r
buff
−1 = (100 M−1) · [H+]x [32].
The stoichiometric coefficients multiplying the complex I, III, and IV, and F1F0-ATPase fluxes include two terms, one term representing the number of protons transported across the inner membrane for a given reaction, and one term representing the number of protons consumed by the associated biochemical reaction. For example, the complex III reaction pumps four H+ out of the matrix and consumes one matrix H+ for every turnover of the reference biochemical reaction H+ NADH + Q
NAD+ + QH2, resulting in a total of three matrix H+ consumed. Thus, the net stoichiometric coefficients multiplying J
C1, J
C3, J
C4, and J
F1 are −(4 − 1), −(2 + 2), −(4 − 2), and +(nA − 1).
The membrane potential kinetics depends on the effective membrane capacitance, C
IM, which is estimated below.
In addition to the 17 state variables treated in equation 22, the concentrations of oxidized matrix NAD, Q, and cytochrome c, and the matrix ADP concentration, are computed as follows:
where NADtot, Qtot, cytCtot, and Atot, are the total concentrations of NAD(H), ubiquinol, cytochrome c, and adenine nucleotide in the matrix, respectively.
Parameter values.
The values of the parameters used in this section are listed in Table 1. The units on activities are expressed as mass flux per unit time per unit total mitochondrial volume, specifically mol s−1 (l mito volume)−1. The parameters have been categorized into three classes, denoted classes A, B, and C. The meaning of these categories is as follows.
Table 1 Mitochondrial Model Parameter Values
Class A refers to free parameters with values determined by fitting model simulations to the data published by Bose et al. [10]. In total, there are 14 adjustable parameters, which are estimated by fitting to seven data curves (described below.) Of the 14 adjustable parameters, four correspond to the phenomenological model of dehydrogenase flux, while the remaining ten are associated with the biophysical model of oxidative phosphorylation and electron transport system in cardiac mitochondria.
Class B refers to 17 parameters for which values are established in the literature. These parameter values were fixed and not treated as adjustable. The values used for NADtot, Atot, Qtot, and cytCtot are obtained from the previous models of Vendelin et al. [8] and Korzeniewski and Zoladz [4,6]. The current model assumes cytochrome c to be distributed within the IM space, in contrast to previous models, in which cytochrome c is in the matrix. To keep the total mass of cytochrome c consistent, cytCtot is set to 2.7 mM, a value that is ten times greater than that used in the Vendelin and Korzeniewski models, since the IM volume is assumed to be 1/10 of the mitochondrial water volume. The value for the outer membrane permeability to adenine nucleotides is estimated from Lee et al. [33]. Assuming the mitochondrial inner membrane has a capacitance of 1 μF per square centimeter of surface area [34], the inner membrane capacitance is calculated to be 6.75 × 10−6 mol (l mito volume)−1 mV−1 for an inner membrane area of 60 μm2. It is observed that the steady-state model behavior presented below is not sensitive to the assumed value of mitochondrial membrane capacitance. The steady-state membrane potential is determined by bulk electroneutrality, which imposes the constraint that the sum of the various currents across the membrane is zero. Thus, 4J
C1 + 2J
C3 + 4J
C4 − nAJ
F1 −J
ANT − J
Hle − J
K − 2J
Mg = 0, where each of these fluxes depends on the membrane potential. The ratio of mitochondrial surface area to volume γ is estimated from morphological data [35] to be 5.99 μm−1. Given this value of γ and the assumed values of outer membrane permeability, the gradients obtained for ATP and ADP across the outer membrane at maximal respiration rate are 20 μM. In this range, the resistance to passive transport between the inner membrane space and buffer is not great enough to be significant and the model behavior is not sensitive to the assumed value of γ.
Class C refers to two parameters that are set to extreme values such that the simulated model behavior is not sensitive to the specific value chosen. The AK and magnesium-binding activities are set to values high enough that the corresponding reactions maintain equilibrium.
Table 2 lists the values for the standard free energies for the reactions of the respiratory chain at pH = 7.
Table 2 Standard Free Energies of Respiratory Chain Reactions
Simulation of isolated mitochondria.
The extensive dataset published by Bose et al. [10] is used to parameterize the mitochondrial model. The external K+, H+, ATP, ADP, AMP, and inorganic potassium concentrations were set as constants according to the buffer concentrations imposed in the experiment in order to compare model simulations to the experimental data. Specifically, [H+]e = 10−7.1, [ATP]e = 0, and [AMP]e = 0; [ADP]e was set at either 0 or 1.3 mM, as described below, and [Pi]e was varied from 0 to 10 mM. The total magnesium concentration in the buffer was fixed at 5.0 mM; buffer potassium concentration was fixed at 150 mM. The simulations described in this section were computed with the oxygen partial pressure in the matrix set to 20 mm Hg, or [O2] = 2.6 × 10−5 M.
The black curves plotted in Figure 2 illustrate comparisons between model-simulated and experimentally measured values for the dataset used to estimate the 14 adjustable parameters listed in Table 1. Parameters values were adjusted to obtain the best fit (least squares error) between model simulations and experimental measures for steady-state values of NADH concentration, rate of oxygen consumption, cytochrome c redox state, and matrix pH, as shown in the figure. Optimal parameter values were found using a global Monte-Carlo-based simulated annealing algorithm that searched for the optimal set of parameter values to simultaneously fit several data curves. In total, seven independent curves were used to estimate the 14 parameter values. It was found that best-fit model solutions are obtained by setting the passive potassium and magnesium fluxes to zero. Shown in Figure 2A are model simulations of steady-state NADH (normalized to NADtot) as a function of external inorganic phosphate, [Pi]e. The two curves correspond to two different values of external ADP concentration, 0 and 1.3 mM, as indicated in the figure. Also shown are data from Bose et al. [10], collected from isolated mitochondria suspensions, with buffer ADP concentrations of 0 mM (circles) and 1.3 mM (triangles). Figure 2B illustrates the experimentally measured and model-simulated values of the rate of oxygen consumption (MVO2) for the same conditions as described for the NADH curves in Figures 2A. When substrate concentration (either ADP or inorganic phosphate) goes to zero, the nonzero MVO2 corresponds to the basal oxygen consumption necessary to maintain the proton motive energy with a finite proton leak across the inner membrane. Figure 2C illustrates the model-simulated and experimentally measured values of cytochrome c redox state; Figure 2D illustrates the model-simulated and experimentally measured values of matrix pH. The matrix pH is buffered at a nearly constant value via H+/K+ exchange.
Figure 2 Comparison of Model Simulations to Experimental Data on NADH, MVO2, Cytochrome C Redox, and Matrix pH for Model without Phosphate Control
(A) Results for normalized matrix NADH as a function of buffer inorganic phosphate concentration are shown for the two experimental cases of resting mitochondria ([ADP]e = 0, state 4) and active state mitochondria ([ADP]e = 1.3 mM, state 3).
(B) Results for MVO2 (rate of oxygen consumption) are shown for the same experimental cases as in (A). Experimental data are not available for the resting state, in which a minimal flux through the electron transport system is maintained to compensate for cation flux across the inner membrane.
(C) Results for cytochrome C reduced fraction are shown for the experimental cases as in (A). The black curves correspond to the model equations developed in the text. The red curves correspond to the best-fit model simulations obtained with equation 9 modified to not include the factor [cytC(red)2+]/cytCtot multiplying the expression for J
C4.
(D) Matrix pH (model-simulated and experimentally measured) is plotted as a function of buffer phosphate for the experimental cases as in (A).
All computed results in this figure correspond to steady-state simulations of model described under “Mitochondrial Model without Phosphate Control.” Model simulations for [ADP]e = 1.3 mM, and [ADP]e = 0 mM are plotted as solid lines and dashed lines, respectively. Experimental data (circles and triangles) are obtained from [10].
Also shown as red curves plotted in Figure 2C is the best fit to the cytochrome redox state data obtained without the factor cytC(red)2+]/cytCtot multiplying the J
C4 flux expression of equation 9. It is observed that the model's fits to the cytochrome c redox data are improved by incorporating this multiplicative factor in the expression for complex IV flux. Thus the best-fit model solution assumes the flux through complex IV depends on [cytC(red)2+]2, which may be explained by the fact that two reduced cytochrome c molecules are require to donate a single electron pair one oxygen atom, generating H2O.
In contrast to the results illustrated in Figure 2, the model described in this section is unable to reproduce data on mitochondrial membrane potential. In Figure 3 are plotted the model-simulated and experimentally measured values of ΔΨ, as a function of [Pi]e for the cases of [ADP]e = 0 mM and [ADP]e = 1.3 mM.
Figure 3 Comparison of Model Simulations to Experimental Data on Membrane Potential for Model without Phosphate Control
The model without phosphate control is not able to fit the experimental data on mitochondrial membrane potential. Computed results in this figure correspond to steady-state simulations of the model described under “Mitochondrial Model without Phosphate Control.” Model simulations for [ADP]e = 1.3 mM, and [ADP]e = 0 mM are plotted as solid lines and dashed lines, respectively. Experimental data (circles and triangles) are obtained from [10].
In the inactive state ([ADP]e = 0), as the external (buffer) phosphate concentration is increased, the dehydrogenase flux governed by equation 1 increases, providing an increased thermodynamic driving force for electron transport and a corresponding increase in the magnitude of the membrane potential. However, the model-simulated rate of increase in ΔΨ with [Pi]e is much smaller than that observed experimentally. When the active state is simulated ([ADP]e = 1.3 mM), the model fit is even worse than for the inactive state. The addition of ADP to the external buffer, representing a sink for the free energy stored in the redox state in the matrix and the membrane potential, results in a drop in both redox and membrane potential. The simulated magnitude of the potential difference decreases with increasing [Pi]e, the opposite of the trend that is observed experimentally. With 14 adjustable parameters, it is not possible to reproduce the experimentally observed behavior of ΔΨ while maintaining reasonable fits to the curves plotted in Figure 2.
Mitochondrial Model with Phosphate Control
Reasonable fits to the observed ΔΨ require that phosphate-dependent control be incorporated into the model, as proposed by Bose et al. [10]. It is found that by including phosphate-dependent control of complex III it is possible to obtain improved fits to the data on membrane potential. The flux expression of equation 7 is modified as follows:
where two new parameters, k
Pi,3 and k
Pi,4, have been introduced. Thus, the total number of adjustable parameters in the model is increased from 14 in the previous section to 16. By varying model parameters, including these additional parameters, we are able to obtain fits to the nine data curves as illustrated in Figures 4 and 5. Thus, the large number of parameters is offset by a relatively large number of data curves to fit for estimation of parameter values. As for the model described in the previous section, it is found that best-fit model solutions are obtained by setting the passive potassium flux to zero. Therefore, in the model parameterization presented in Table 1 the activities of the corresponding channels are set to zero, effectively reducing the number of adjustable parameters. Sensitivity analysis reveals that the activities of the channels cannot be determined based on the present dataset. The activities of potassium channels are known to be modulated by ischemic and anesthetic preconditioning [31,36,37] and will be explored in future work.
Figure 4 Comparison of Model Simulations to Experimental Data on NADH, MVO2, Cytochrome C Redox, and Matrix pH for Model with Phosphate Control
(A) Results for normalized matrix NADH as a function of buffer inorganic phosphate concentration are shown for the two experimental cases of resting mitochondria ([ADP]e = 0, state 4) and active state mitochondria ([ADP]e = 1.3 mM, state 3).
(B) Results for MVO2 (rate of oxygen consumption) are shown for the same experimental cases as in (A).
(C) Results for cytochrome C reduced fraction are shown for the experimental cases as in (A).
(D) Matrix pH (model-simulated and experimentally measured) is plotted as a function of buffer phosphate for the experimental cases as in (A).
All computed results in this figure correspond to steady-state simulations of the model described under “Mitochondrial Model with Phosphate Control.” Model simulations for [ADP]e = 1.3 mM and [ADP]e = 0 mM are plotted as solid lines and dashed lines, respectively; experimental data are the same as plotted in Figure 2.
Figure 5 Comparison of Model Simulations to Experimental Data on Membrane Potential for Model with Phosphate Control
The model with phosphate control compares much more favorably to the experimental measurements than the model without phosphate control (see Figure 3). Computed results in this figure correspond to steady-state simulations of the model described under “Mitochondrial Model with Phosphate Control.” Model simulations for [ADP]e = 1.3 mM and [ADP]e = 0 mM are plotted as solid lines and dashed lines, respectively; experimental data are the same as plotted in Figure 3.
The fits obtained using equation 24 to model the complex III flux are plotted in Figures 4 and 5. Note that the model simulation of NADH, MVO2, cytochrome c state, and matrix pH plotted in Figure 4 produces values similar to those obtained using the previous model (see Figure 2). Of particular note is the behavior of the NADH redox state and MVO2 curves, plotted in Figure 4A and 4B. In these cases the mean squared error between observed data and the model fits is slightly lower than that obtained using the model with no phosphate control (see Figure 2A and 2B).
The major difference between the model of this section and that of the previous section is seen in the simulated membrane potential values, plotted in Figure 5. With the phosphate-modulated control described by equation 24, the model remains unable to reproduce the observed data on membrane potential as phosphate concentration goes to zero. Yet while the model's fits to the observed data remain imperfect, the agreement between simulation and experimental data is significantly improved by incorporating the expression of equation 24 to model the complex III flux. Possible mechanisms explaining differences between model simulations and the data observed in the limit [Pi]e → 0 at [ADP]e = 1.3 mM are outlined in the Discussion.
The values of the parameters used in generating Figures 4 and 5 are listed in Table 1. The sensitivities of the estimated values of model parameters are considered in the Discussion.
Behavior of Model at Low Oxygen Concentration
A key to understanding cellular energetics during hypoxia and ischemia is a mechanistic model of mitochondrial function at low oxygen concentration. In this section the behavior of the model is compared to measurements of oxygen consumption and cytochrome c reduction in isolated mitochondria as functions of the oxygen concentration of the medium.
The behavior of the model at low oxygen concentration is illustrated in Figure 6. Plotted are the rate of mitochondrial oxygen consumption and the predicted reduced fraction of cytochrome c as functions of the oxygen content of the medium.
Figure 6 Behavior of Model at Low Oxygen Concentration
Predicted rate of oxygen consumption (MVO2) normalized to maximal rate of oxygen consumption and fraction of cytochome c reduced are plotted against oxygen concentration, which is expressed in micromoles (lower axis) and oxygen partial pressure (upper axis). The oxygen consumption curve was computed for state-3 respiration, corresponding to experimental conditions reported in [18] and [19]. The cytochrome c curve corresponds to state-4 experimental conditions reported in [17]. Inset shows predicted curves for oxygen concentrations from 0 to 5 μM.
The oxygen consumption curve was computed for active state-3 respiration, with [ADP]e and [ATP]e set to 1.0 mM and 0 mM, respectively. This curve corresponds to the curves reported in Figure 7A of [20] and Figure 2A of [21]. The model-predicted P
50 for half-maximal oxygen consumption is 0.373 μM (or an oxygen partial pressure of 0.287 mm Hg), close to the reported value of 0.35 ± 0.07 μM [20,21]. The curve for the predicted reduced fraction of cytochrome c (dashed line in Figure 6) was computed for state-3 respiration ([ADP]e = 0.5 mM; [ATP]e = 0.87 mM), corresponding to the measurements of Wilson et al. [19]. The predicted curve can be compared to Figures 4B and 5A of [19]. Wilson et al. found that in mitochondria isolated from rat liver, cytochrome c is approximately 15%–18% reduced for oxygen concentration in the range of 40–50 μM. The current model predicts a slightly lower value of 14% reduced at [O2]e = 50 μM.
Thus, beyond the dataset used for parameterization, the model was further validated by comparison to additional datasets measured from mitochondria isolated from rat heart [20,21] and liver [19] and observed at low oxygen concentration. The model agrees quantitatively with the measured dependence of oxygen consumption rate on oxygen concentration in state-3 respiration. Although the observed cytochrome c redox state is slightly (1%–4%) more reduced in measurements reported for isolated rat liver mitochondrial at low oxygen concentrations compared with the model predictions, the predicted behavior of cytochrome c reduction relative to oxygen concentration is qualitatively similar to the corresponding experimental observations. Allowing for differences in the behaviors of hepatic and cardiac mitochondria, an exact quantitative agreement may not be expected.
Discussion
Model Development and Parameterization
The main contribution of the current study is the introduction of a self-consistent thermodynamically balanced model of oxidative phosphorylation and the electron transport system in mitochondria. A biophysical model incorporating all of the components illustrated in Figure 1 required the development of a system of 17 differential equations and the introduction of 16 adjustable parameters. To identify such a large number of parameters, it was necessary to make use of a large number of independent measurements made on mitochondria isolated from rat cardiac tissue [10]. This previously published dataset consists of measures of NAD(H) redox state, cytochrome c redox state, rate of oxygen consumption, mitochondrial membrane potential, and matrix pH, for a range of buffer conditions, including a range of concentrations of inorganic phosphate and for resting and active state mitochondria ([ADP]e = 0 and 1.3 mM, respectively). In total, 16 parameters were estimated by simultaneously fitted model-simulated steady states to nine independent data curves (see Figures 4 and 5), providing quantitative estimates of the model parameters.
While an exhaustive statistical analysis of the 16-dimensional parameter space is not computationally feasible, it is possible to compute the sensitivities of the mean squared error between the model solutions to the estimated parameter values. Parameter sensitivity can be estimated from the diagonal entries of the Hessian of the error function,
, where E is the mean squared difference between model simulations and experimental data and xi represents ith parameter. However, the partial derivatives of E with respect to parameter values represent local measures that do not necessarily reflect how the error changes with finite changes in the parameter values. To estimate the sensitivity to finite changes in parameter values, the sensitivity to each parameter was computed as the relative change in mean squared error due to a 10% change in a given parameter value:
where E* is the minimum mean squared difference between model simulations and experimental data, and
is the optimal value of the ith parameter. The term
is the error computed while setting parameter xi to 10% above and below its optimal value. The relative sensitivities to the adjustable parameters are listed in Table 1. These sensitivity values represent a measure of the degree to which the curves plotted in Figures 4 and 5 are sensitive to the value of the individual parameters. A high sensitivity value indicates that changing a given parameter results in significant changes to the simulated curves used to identify the set of adjustable parameter values. Note that six of the adjustable parameters show relative sensitivity of less than 5%, indicating that model solutions are not particularly sensitive to these parameters in the neighborhood of the reported values and that these parameters are not well estimated by the present analysis. In particular, in comparison to the dataset presented here, the model is relatively insensitive to the values of the activities of the potassium transporters and the F1F0 ATPase. The best model fits were obtained with the potassium channel activities set effectively to zero. Therefore, the relative sensitivity to this parameter is reported to be zero as well. The ATP synthesis activity X
F1 is determined to be large enough that the reaction is effectively maintained in equilibrium. To estimate the parameters that are poorly identified, it will be necessary to obtain appropriate data in the future. In fact, because the model is parameterized based solely on steady-state data, it may not accurately match time-dependent behavior of mitochondrial oxidative phosphorylation. Thus, it is expected that kinetic data, in particular, will be of great value in refining the parameter estimates, refining the model, and generally improving the ability of the model to predict observed behavior.
Yet while the model represents only one component of mitochondrial energy metabolism that one may build on and refine, it does represent the most complete model of the respiratory chain that is available to date. The model includes the major components of oxidative phosphorylation and the electron transport system and appropriately balances mass, charge, and free energy. By integrating the components illustrated in Figure 1 into a self-contained model, the observations of Bose et al. [10] have been explained based on a model that incorporates phosphate control at the dehydrogenase flux (perhaps via mass action) and phosphate-dependent activation of complex III. Of course, the model does not reproduce the experimental data with perfect fidelity. The major shortcomings of the current model analysis are the inability to reproduce the observed data in state-3 mitochondria as buffer phosphate concentration approaches zero and the inability to sensitively identify parameters for membrane ion transporters. As discussed below, the observations at [Pi]e → 0 may be influenced by an experimental artifact; detailed identification of the inner membrane ion transporters will require the design of further experiments sensitive to the kinetic behavior of these channels.
Phosphate-Dependent Control of the Respiratory Chain
To obtain the model solutions illustrated in Figures 4, 5, and 6, an expression for complex III flux was developed based on the hypothesis that inorganic phosphate level modulates the activity of the complex III. While no data directly measuring the activity of complex III in intact mitochondria as a function of phosphate concentration are available, the hypothesis is supported by the fact that the model's fits to the observed data are significantly improved when phosphate-dependent control is included compared to the case when it is not. In work not detailed, a number of similar control expressions were tested using phosphate and other species (ATP, ADP, Mg2+) as putative controllers of complexes I, III, and IV and of F1F0 ATPase and the ANT system. It was found that no other choice of controlled enzyme and controller species could provide fits to the data of Figure 5 as reasonable as that of the hypothetical control of complex III by inorganic phosphate. Thus, 20 independent hypotheses (each formulated as the activity of one of five enzymes dependent on one of four species) were quantitatively tested and 19 were excluded as not able to reproduce the observed data. Yet, by increasing the complexity of the model, in particular by hypothesizing phosphate-dependent control of a more than one of the enzymes in the system, it is possible to obtain improved fits to the observed data. However, doing so requires introducing additional free parameters that cannot be well identified. For this reason, a minimal model that satisfactorily explains the data within a reasonable error tolerance was developed.
An alternative hypothesis for the biophysical mechanism behind phosphate-dependent activation of the electron transport system is that phosphate modulates the redox coupling of cytochrome b and c, as proposed by Bose et al. [10]. It has been observed that binding to phosphate and other ions changes the apparent redox potential for cytochrome c [38]. These data on binding of cytochrome c and phosphate allow one to formalize this alternative hypothesis by appropriately modifying ΔGo,C3 and ΔGo,C4 to depend on phosphate concentration. However, a simple model (results not shown) using the linear relationship between the apparent cytochrome c redox potential and phosphate concentration observed by Gopal et al. [38] was unable to reproduce the phosphate-dependent behavior observed by Bose et al. [10]. Thus, the model detailed in this work represents the most parsimonious explanation that was found for the data illustrated in Figures 4 and 5.
In addition to reproducing the data used to parameterize the model, the mechanism of phosphate-dependent control of respiration matches observations in permeabilized cardiomyocytes indicating that phosphate represents the major feedback signal in low and medium work loads [7].
As noted above, the model remains unable to reproduce the observed data on membrane potential as phosphate concentration goes to zero. Note that as [Pi]e → 0, the observed [NADH]/NADtot ratio, which is the driving force for electron transport and proton pumping, is approximately 0.6 for both levels of buffer ADP concentration that were studied (see Figure 4A). Yet the observed membrane potential, which is ultimately driven by the matrix redox state, is approximately 25 mV lower at [ADP]e = 1.3 mM than at [ADP]e = 0. Since there can be no ATP synthesis when [Pi]e = 0, this behavior cannot be explained by a load on the F1F0 ATPase or the ANT.
It is expected that there always exists trace phosphate in the buffer and matrix and that nonviable mitochondria (e.g., mitochondria with compromised membranes) present in the experimental isolation act as ATPases, consuming ATP and generating phosphate [39]. Therefore, to obtain improved fits to the membrane potential data with ADP present at low phosphate concentration it may be appropriate to include an ATP-consuming reaction in the model of isolated mitochondria. However, introducing an ATP-consuming reaction into the model of isolated mitochondria has the effect of increasing the predicted oxygen consumption. With the current model it is not possible to reproduce both the difference in ΔΨ observed at [Pi]e = 0 and the oxygen consumption curve by introducing an ATP-consuming reaction into the model (results not shown). Likely, the observed drop in membrane potential for [ADP]e = 1.3 mM (and [Pi]e = 0) compared to at [ADP]e = 0 is due to residual phosphate in the bath and in the mitochondrial matrix, as proposed by Bose et al. [10].
Another mechanism that was considered as a possible explanation for the drop in ΔΨ observed when [ADP] is added to the buffer at [Pi]e = 0 is that the activity of one or more the of the components of the electron transport system is enhanced when [ADP]e = 0 compared to when [ADP]e ≠ 0. Using the current model, one can obtain improved fits to the observed data by reducing the complex I or III activity, or by reducing the dehydrogenase activity, when [ADP]e = 1.3 mM compared to the case when [ADP]e = 0. Thus, inhibition of the activity of respiratory complexes by ADP represents a possible mechanism to explain the observations of Bose et al. [10] at [Pi]e = 0. However, the present dataset does not provide enough information to determine which sites in the respiratory chain are modulated by the addition of ADP into the buffer. Further experiments are necessary to test competing hypotheses and to formulate a mechanistic model that explains the phenomenon in detail.
Future Directions
As indicated above, the model remains imperfect and there remains room for improvement to model's fits to observed data. One potential avenue for improving the scope and predictive power of the model is to extend the model to include additional components of cardiac energetic metabolism. This task must be undertaken under the guidelines outlined in the Introduction. This operating philosophy behind model development requires that the use of purely data-driven empiricisms be avoided wherever possible. While nonphysical empirical relationships were not introduced in modeling the central components of the current model illustrated in Figure 1, the model is driven by an arbitrary four-parameter function (equation 1) used to represent overall dehydrogenase flux in the isolated system of Bose et al. [10]. This expression invokes a phenomenological dependence of the dehydrogenase flux on phosphate concentration, required to reproduce the observed data. In general, it is often difficult to avoid invoking such driving functions at the boundaries of a given model. Since it is planned to integrate the current model of oxidative phosphorylation with other components of cardiac metabolism [40], the data-driven dehydrogenase flux will be replaced with realistic models of the TCA cycle [12,13,40] and other reactions generating intracellular reducing potential [40]. While the current model was developed to analyze data from isolated mitochondria respiring on complex I substrates, such an integrated model would require a biophysical treatment of FAD(H2) redox handling at complex II of the respiratory chain. Thermodynamically balanced flux expressions for complex II flux could be developed in a manner analogous to that for complexes I and III in this study. These steps represent planned progress toward a major long-term goal of the construction of a complete energetic model of the cell [41].
Large-scale models of cellular energetics can be used for a variety of applications. For example, the current model may be linked with excitation–contraction models and calcium-dependent control of energy metabolism by extending the model to include sodium and calcium ion exchangers, as has been done in previous models [11,13,40]. Simulation and analysis of cardiac energetic requires integrating metabolic models into models of substrate transport at the cellular and tissue levels [42,43]. Thus, the current model provides a basis for integrating a detailed model of mitochondrial function into multiple-scale models of the heart.
While it was determined that the passive potassium channel flux is effectively zero in the model parameterization developed here, modeling the K+/H+ exchanger is required to buffer the matrix pH when both ΔΨ and [H+]x are treated as state variables. Mitochondrial ATP-dependent potassium channel flux [17,18,37,44–47] will need to be incorporated to investigate the cardiac metabolic response to ischemia, hypoxia, and preconditioning. In addition, kinetic data must be obtained to effectively parameterize and validate the time-dependent behavior of the model. In sum, while a great number of extensions and improvements are possible, and many are planned, the current model represents the foundation for building larger and more complex systems and investigating complex physiological and pathophysiological interactions.
Materials and Methods
The model was implemented, simulated, and analyzed using the MATLAB (The Mathworks, Natick, Massachusetts, United States) computing environment. All calculations were performed on a desktop PC. Computer codes are available from the author upon request. In addition, the model is available in the CellML exchange format at http://www.cellml.org.
The author thanks Peter Hunter, Peter Villiger, and the Auckland Bioengineering Research Group for help with porting the model to the CellML model exchange format. I am grateful to James Bassingthwaighte, Robert Balaban, Saleet Jafri, and Marko Vendelin for valuable discussions. This work was supported by National Institutes of Health grant HL072011.
Competing interests. The author has declared that no competing interests exist.
Author contributions. DAB conceived and designed the experiments, performed the experiments, analyzed the data, and wrote the paper.
A previous version of this article appeared as an Early Online Release on August 4, 2005 (DOI: 10.1371/journal.pcbi.0010036.eor).
Abbreviations
AKadenylate kinase
ANTadenine nucleotide translocase
IMintermembrane
MVO2rate of oxygen consumption
PiHtphosphate–hydrogen co-transporter
==== Refs
References
Qian H Beard DA Liang SD 2003 Stoichiometric network theory for nonequilibrium biochemical systems Eur J Biochem 270 415 421 12542691
Beard DA Qian H Bassingthwaighte JB 2004 Stoichiometric foundation of large-scale biochemical system analysis Ciabanu G Rozenberg G Modelling in molecular biology New York Springer 1 20
Qian H Beard DA 2005 Thermodynamics of stoichiometric biochemical networks in living systems far from equilibrium Biophys Chem 114 213 220 15829355
Korzeniewski B 1998 Regulation of ATP supply during muscle contraction: Theoretical studies Biochem J 330 1189 1195 9494084
Korzeniewski B 2000 Regulation of ATP supply in mammalian skeletal muscle during resting state→intensive work transition Biophysical Chemistry 83 19 34 10631477
Korzeniewski B Zoladz JA 2001 A model of oxidative phosphorylation in mammalian skeletal muscle Biophys Chem 92 17 34 11527576
Saks VA Kongas O Vendelin M Kay L 2000 Role of the creatine/phosphocreatine system in the regulation of mitochondrial respiration Acta Physiol Scand 168 635 641 10759600
Vendelin M Kongas O Saks V 2000 Regulation of mitochondrial respiration in heart cells analyzed by reaction-diffusion model of energy transfer Am J Physiol Cell Physiol 278 C747 C764 10751324
Matsuoka S Sarai N Jo H Noma A 2004 Simulation of ATP metabolism in cardiac excitation-contraction coupling Prog Biophys Mol Biol 85 279 299 15142748
Bose S French S Evans FJ Joubert F Balaban RS 2003 Metabolic network control of oxidative phosphorylation: Multiple roles of inorganic phosphate J Biol Chem 278 39155 39165 12871940
Magnus G Keizer J 1997 Minimal model of beta-cell mitochondrial Ca2+ handling Am J Physiol 273 C717 C733 9277370
Dudycha SJ 2000 A detailed model of the tricarboxylic acid cycle in heart cells [thesis] Baltimore Johns Hopkins University 141 p.
Nguyen MHT Ramay HR Dudycha SJ Jafri MS 2004 Modeling the cellular basis of cardiac excitation-contraction coupling and energy metabolism Recent Res Dev Biophys 3 125 145
Cortassa S Aon MA Marban E Winslow RL O'Rourke B 2003 An integrated model of cardiac mitochondrial energy metabolism and calcium dynamics Biophys J 84 2734 2755 12668482
Cortassa S Aon MA Winslow RL O'Rourke B 2004 A mitochondrial oscillator dependent on reactive oxygen species Biophys J 87 2060 2073 15345581
Hill TL 1977 Free energy transduction in biology: The steady-state kinetic and thermodynamic formalism New York Academic Press 229 p.
Garlid KD Paucek P 2003 Mitochondrial potassium transport: The K(+) cycle Biochim Biophys Acta 1606 23 41 14507425
Garlid KD Dos Santos P Xie ZJ Costa AD Paucek P 2003 Mitochondrial potassium transport: The role of the mitochondrial ATP-sensitive K(+) channel in cardiac function and cardioprotection Biochim Biophys Acta 1606 1 21 14507424
Wilson DF Rumsey WL Green TJ Vanderkooi JM 1988 The oxygen dependence of mitochondrial oxidative phosphorylation measured by a new optical method for measuring oxygen concentration J Biol Chem 263 2712 2718 2830260
Gnaiger E 2001 Bioenergetics at low oxygen: Dependence of respiration and phosphorylation on oxygen and adenosine diphosphate supply Respir Physiol 128 277 297 11718759
Gnaiger E Kuznetsov AV 2002 Mitochondrial respiration at low levels of oxygen and cytochrome c Biochem Soc Trans 30 252 258 12023860
Trumpower BL 1990 The protonmotive Q cycle. Energy transduction by coupling of proton translocation to electron transfer by the cytochrome bc1 complex J Biol Chem 265 11409 11412 2164001
Crofts AR 2004 The cytochrome bc1 complex: Function in the context of structure Annu Rev Physiol 66 689 733 14977419
Michel H 1998 The mechanism of proton pumping by cytochrome c oxidasex127e Proc Natl Acad Sci U S A 95 12819 12824 9788998
Michel H 1999 Cytochrome c oxidase: Catalytic cycle and mechanisms of proton pumping—A discussion Biochemistry 38 15129 15140 10563795
Wilson DF Erecinska M 1986 The oxygen dependence of cellular energy metabolism Adv Exp Med Biol 194 229 239 3529859
Wilson DF Rumsey WL 1988 Factors modulating the oxygen dependence of mitochondrial oxidative phosphorylation Adv Exp Med Biol 222 121 131 2834920
Greenbaum NL Wilson DF 1985 The distribution of inorganic phosphate and malate between intra- and extramitochondrial spaces. Relationship with the transmembrane pH difference J Biol Chem 260 873 879 2857172
Lakshminarayanaiah N 1969 Transport phenomena in membranes New York Academic Press 517 p.
Fall CP Wagner J Marland E 2002 Computational cell biology New York Springer 468 p.
O'Rourke B 2000 Pathophysiological and protective roles of mitochondrial ion channels J Physiol 529 23 36 11080248
Korzeniewski B 2001 Theoretical studies on the regulation of oxidative phosphorylation in intact tissues Biochim Biophys Acta 1504 31 45 11239483
Lee AC Zizi M Colombini M 1994 Beta-NADH decreases the permeability of the mitochondrial outer membrane to ADP by a factor of 6 J Biol Chem 269 30974 30980 7983033
Gentet LJ Stuart GJ Clements JD 2000 Direct measurement of specific membrane capacitance in neurons Biophys J 79 314 320 10866957
Munoz DR de Almeida M Lopes EA Iwamura ES 1999 Potential definition of the time of death from autolytic myocardial cells: A morphometric study Forensic Sci Int 104 81 89 10581715
da Silva MM Sartori A Belisle E Kowaltowski AJ 2003 Ischemic preconditioning inhibits mitochondrial respiration, increases H2O2 release, and enhances K+ transport Am J Physiol Heart Circ Physiol 285 H154 H162 12623788
Mironova GD Negoda AE Marinov BS Paucek P Costa AD 2004 Functional distinctions between the mitochondrial ATP-dependent K+ channel (mitoKATP) and its inward rectifier subunit (mitoKIR) J Biol Chem 279 32562 32568 15138282
Gopal D Wilson GS Earl RA Cusanovich MA 1988 Cytochrome c: Ion binding and redox properties. Studies on ferri and ferro forms of horse, bovine, and tuna cytochrome c J Biol Chem 263 11652 11656 2841331
Vendelin M Lemba M Saks VA 2004 Analysis of functional coupling: Mitochondrial creatine kinase and adenine nucleotide translocase Biophys J 87 696 713 15240503
Jafri MS Dudycha SJ O'Rourke B 2001 Cardiac energy metabolism: Models of cellular respiration Annu Rev Biomed Eng 3 57 81 11447057
Bassingthwaighte JB 2001 The modelling of a primitive ‘sustainable' conservative cell Philos Trans R Soc London A 359 1055 1072
Beard DA Schenkman KA Feigl EO 2003 Myocardial oxygenation in isolated hearts predicted by an anatomically realistic microvascular transport model Am J Physiol Heart Circ Physiol 285 H1826 H1836 12869375
Schenkman KA Beard DA Ciesielski WA Feigl EO 2003 Comparison of buffer and red blood cell perfusion of guinea pig heart oxygenation Am J Physiol Heart Circ Physiol 285 H1819 H1825 12869374
Sasaki N Sato T Marban E O'Rourke B 2001 ATP consumption by uncoupled mitochondria activates sarcolemmal K(ATP) channels in cardiac myocytes Am J Physiol Heart Circ Physiol 280 H1882 H1888 11247805
Belisle E Kowaltowski AJ 2002 Opening of mitochondrial K+ channels increases ischemic ATP levels by preventing hydrolysis J Bioenerg Biomembr 34 285 298 12392192
Peart JN Gross GJ 2002 Sarcolemmal and mitochondrial K(ATP) channels and myocardial ischemic preconditioning J Cell Mol Med 6 453 464 12611635
Gross ER Peart JN Hsu AK Grover GJ Gross GJ 2003 K(ATP) opener-induced delayed cardioprotection: Involvement of sarcolemmal and mitochondrial K(ATP) channels, free radicals and MEK1/2 J Mol Cell Cardiol 35 985 992 12878485
Tomashek JJ Brusilow WS 2000 Stoichiometry of energy coupling by proton-translocating ATPases: A history of variability J Bioenerg Biomembr 32 493 500 15254384
Alberty RA 2003 Thermodynamics of biochemical reactions Hoboken (New Jersey). John Wiley-Interscience 397 p.
Vinnakota KC Bassingthwaighte JB 2004 Myocardial density and composition: A basis for calculating intracellular metabolite concentrations Am J Physiol Heart Circ Physiol 286 H1742 H1749 14693681
|
16163394
|
PMC1201326
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Sep 9; 1(4):e36
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010036
|
oa_comm
|
==== Front
Nutr Metab (Lond)Nutrition & Metabolism1743-7075BioMed Central London 1743-7075-2-171607639310.1186/1743-7075-2-17ResearchAn evaluation of the metabolic syndrome in a large multi-ethnic study: the Family Blood Pressure Program Kraja Aldi T [email protected] DC [email protected] Alan B [email protected] Thomas H [email protected] Stephen T [email protected] Chao Agnes [email protected] Thomas [email protected] Richard [email protected] J David [email protected] Michael A [email protected] Division of Biostatistics, Washington University School of Medicine, St. Louis, MO, USA2 University of Michigan Hospitals, Ann Arbor, MI, USA3 Division of Geriatrics, Department of Medicine, University of Mississippi Medical Center, Jackson MS, USA4 Mayo Clinic, College of Medicine, Rochester, MN, USA5 National Health Research Institutes, Division of Biostatistics, Taipei, Taiwan6 Stanford University School of Medicine, Stanford, CA, USA7 Loyola University Medical Center, Maywood, IL, USA8 Pacific Health Research Institute, Honolulu, HI, USA2005 2 8 2005 2 17 17 17 5 2005 2 8 2005 Copyright © 2005 Kraja et al; licensee BioMed Central Ltd.2005Kraja 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 Family Blood Pressure Program is an ongoing, NHLBI-sponsored, multi-center program to study the genetic determinants of high blood pressure. The goal of this particular study was to study patterns of metabolic syndrome (MetS) in four ethnic groups: African Americans, Caucasians, Hispanics, and Asians.
Methods
A major part of participants in three networks GENOA, HyperGEN and SAPPHIRe were recruited mainly through hypertensive probands. MetS was defined as a categorical trait following the National Cholesterol Education Program definition (c-MetS). MetS was also characterized quantitatively through multivariate factor analyses (FA) of 10 risk variables (q-MetS). Logistic regression and frequency tables were used for studying associations among traits.
Results
Using the NCEP definition, the Hispanic sample, which by design was enriched for type 2 diabetes (T2D), had a very high prevalence of MetS (73%). In contrast, its prevalence in Chinese was the lowest (17%). In African Americans and Hispanics, c-MetS was more prevalent in women than in men. Association of c-MetS with type 2 diabetes (T2D) was prominent in the Hispanics and African Americans, less pronounced in the Whites and Japanese, (although still significant), and weakest in the Chinese sample.
Using FA without rotation, we found that the main factor loaded obesity (OBS) and blood pressure (BP) in African Americans; OBS and insulin (INS) in Hispanics, in Japanese, and in Whites; and OBS alone in Chinese. In Hispanics, Whites, and Japanese, BP loaded as a separate factor. Lipids in combination with INS also loaded in a separate factor. Using FA with Varimax rotation, 4 independent factors were identified: "Obesity-INS," "Blood pressure," "Lipids-INS," and "Central obesity." They explained about 60% of the variance present in the original risk variables.
Conclusion
MetS ethnic differences were identified. Ascertaining for hypertension or T2D increased the MetS prevalence in networks compared with the one in the US general population. Obesity was the most prominent risk factor contributing to both c-MetS and q-MetS. INS contributed in two important factors (obesity and lipids). The information imbedded into c-MetS trait /q-MetS factors scores can contribute in future research of the MetS, especially its utilization in the genetic analysis.
==== Body
Background
Metabolic syndrome (MetS) is defined as a clustering of cardiovascular and type 2 diabetes risk factors including obesity, insulin resistance, dyslipidemia, and hypertension (HT). The genetic control of MetS is expected to be complex since it represents a syndrome of multifaceted abnormalities. Categorical and clinically applicable criteria were developed by the National Cholesterol Education Program (NCEP), which defined MetS as the presence in an individual of at least 3 out of 5 risk factors (increased waist circumference (WAIST), increased level of triglycerides (TG), low levels of high density lipoprotein cholesterol (HDL), HT, and fasting glucose (GLUC) ≥ 110 mg/dl) [1].
We designated this dichotomous definition of MetS (presence or absence) as c-MetS (see also Material and Methods). Quantitative factor analytic treatment of MetS was designated as q-MetS (see Material and Methods). Several earlier studies have employed multivariate techniques such as factor analysis (FA) to investigate MetS. This method transforms a set of MetS risk variables to a smaller set of latent factors. Most studies have reported 2 to 4 underlying factors, depending on the number of risk factors included, whether or not the Varimax rotation was used, and statistical decisions made [2-7].
This study is an investigation of MetS in the Family Blood Pressure Program (FBPP) [8]. The FBPP is comprised of 4 different networks: GenNet, GENOA, HyperGEN, and SAPPHIRe. These networks were established to study the genetic determinants of high blood pressure.
The goal of this particular study was to evaluate MetS in the rich FBPP database using both the c-MetS and q-MetS definitions. Common features and differences among the major ethnic groups were explored. Finally, the relationships of c-MetS with T2D and vascular heterogeneous atherosclerotic (VHA) events were also investigated.
Materials and methods
Participants
The FBPP pooled database (version 3) of 13,592 participants from 4 different networks represents one of the largest compilations of ethnically diverse data. GenNet had only partial data for defining c-MetS and therefore, data from GenNet were excluded from analysis. GENOA (Genetic Epidemiology Network of Atherosclerosis) includes 3 field centers: the Jackson, MS center recruited African Americans; the one in Starr County, TX recruited Hispanics; the one in Rochester, MN, recruited Whites. HyperGEN (Hypertension Genetic Epidemiology Network) included field centers in Birmingham AL, which recruited African Americans; the rest of centers in Forsyth County, NC, in Framingham, MA, in Minneapolis, MN, and in Salt Lake City, UT recruited Whites. SAPPHIRe (Stanford Asian Pacific Program in Hypertension and Insulin Resistance), with 3 major field centers, recruited Asian Pacific populations of Chinese origin residing in Taiwan and of Japanese origin residing in Hawaii, and California. GENOA recruited African American, Hispanic, and White sibships with at least 2 hypertensive sibs (with HT onset before the age of 60). The Hispanic sibships were recruited with at least 2 sibs who were each diagnosed as type 2 diabetic (T2D). HyperGEN recruited African American and White hypertensive sibships with 2 or more hypertensive sibs, at least 1 of them having severe HT. In addition, HyperGEN recruited random samples of African American and White participants, and parents. SAPPHIRe recruited sib-pairs concordant and/or discordant for hypertension. For all participants the diet was uncontrolled and reflective of the "free-living" dietary habits of these populations.
We analyzed data from the FBPP where participants with data for defining c-MetS included: 1857 African Americans, 1799 Hispanics, and 1578 Whites in GENOA; 2010 African Americans, and 1888 Whites in HyperGEN; 1630 Chinese, and 581 Japanese participants in SAPPHIRe). In all networks, subjects with unknown ethnicity were excluded. As a result, information on a total of 11,343 participants was considered in the c-MetS study (Tables 1, 2, 3). In contrast, the sample sizes for q-MetS were considerably smaller because there were only 7,562 participants with no missing values for any of the 10 risk factors (see Material and Methods for risk factors analyzed and Table 5 for the exact sample sizes).
Table 1 Variables Analyzed in the FA (GENOA)
African Americans (N = 1312) Hispanics (1160) Whites (1073)
Variables Mean St. Dev. Mean St. Dev. Mean St. Dev.
AGE 58 10 55 12 55 11
BMI 31 7 31 6 30 6
WAIST 103 17 107 14 99 16
WHR 0.91 0.08 0.97 0.08 0.91 0.09
INS 11 10 14 14 9 7
GLUC 108 42 142 64 98 26
LDL 121 40 116 35 122 34
HDL 57 18 47 13 54 16
TG 126 43 162 47 156 47
SBP 130 23 129 22 133 19
DBP 71 11 70 10 76 10
Table 2 Variables Analyzed in the FA (HyperGEN)
African Americans (1731) Whites (1255)
Variables Mean St. Dev. Mean St. Dev.
AGE 48 13 56 13
BMI 32 7 30 6
WAIST 102 17 103 15
WHR 0.90 0.08 0.94 0.08
INS 11 9 8 6
GLUC 109 45 103 30
LDL 120 37 118 32
HDL 54 15 49 15
TG 104 56 152 75
SBP 130 22 123 20
DBP 75 12 69 11
Table 3 Variables Analyzed in the FA (SAPPHIRe)
Chinese (747) Japanese (284)
Variables Mean St. Dev. Mean St. Dev.
AGE 50 9 55 9
BMI 25 3 27 4
WAIST 84 11 90 12
WHR 0.87 0.08 0.91 0.09
INS 8 5 8 5
GLUC 92 18 101 21
LDL 124 37 121 34
HDL 43 12 48 14
TG 124 69 166 78
SBP 132 25 136 21
DBP 78 14 79 11
Table 5 Correlation Matrices Among Risk Factors Prepared for MetS by Network
GENOA N = 1160 Hispanics
N = 1312 BMI WAIST WHR INS GLUC LDL HDL TG SBP DBP
A. BMI - 0.84‡ 0.33‡ 0.40‡ -0.15‡ -0.03 -0.17‡ 0.12‡ 0.16‡ 0.03
A WAIST 0.87‡ - 0.63‡ 0.37‡ -0.19‡ -0.03 -0.18‡ 0.12‡ 0.19‡ 0.06*
m WHR 0.38‡ 0.68‡ - 0.24‡ -0.21‡ 0.03 -0.16‡ 0.13‡ 0.18‡ 0.10‡
e INS 0.45‡ 0.45‡ 0.31‡ - -0.13‡ -0.08† -0.18‡ 0.19‡ 0.10 0.03
r GLUC †† -0.28‡ -0.30‡ -0.26‡ -0.38‡ - -0.03 0.15‡ -0.18‡ -0.17‡ -0.10‡
I LDL 0.03 0.04 0.05 0.05 -0.03 - 0.01 0.19‡ 0.04 0.07*
c HDL -0.18‡ -0.21‡ -0.20‡ -0.33‡ 0.18‡ -0.12 - -0.30‡ 0.01 0.05
a TG 0.13‡ 0.19‡ 0.21‡ 0.30‡ -0.20‡ 0.20‡ -0.34‡ - 0.14‡ 0.09†
n SBP 0.13‡ 0.14‡ 0.14‡ 0.08† -0.12‡ 0.04 0.00 0.06* - 0.72‡
s DBP -0.02 0.00 0.06* -0.01 -0.02 0.06* 0.03 0.04 0.75‡ -
GENOA
N = 1073 BMI WAIST WHR INS GLUC LDL HDL TG SBP DBP
BMI -
WAIST 0.90‡ -
WHR 0.51‡ 0.74‡ -
INS 0.57‡ 0.57‡ 0.42‡ -
W GLUC -0.32‡ -0.31‡ -0.24‡ -0.41‡ -
h LDL 0.00 0.02 0.01 0.00 -0.02 -
i HDL -0.28‡ -0.30‡ -0.27‡ -0.37‡ 0.18‡ -0.12 -
t TG 0.28‡ 0.29‡ 0.21‡ 0.31‡ -0.17‡ 0.16‡ -0.37‡ -
e SBP 0.24‡ 0.22‡ 0.14‡ 0.16‡ -0.15‡ 0.00 -0.04 0.12‡ -
s DBP 0.07* 0.09† 0.09† 0.03 -0.02 0.05 0.02 0.12‡ 0.68‡ -
HyperGEN N = 1255 Whites
N = 1731 BMI WAIST WHR INS GLUC LDL HDL TG SBP DBP
A. BMI - 0.89‡ 0.45‡ 0.51‡ -0.31‡ 0.04 -0.19‡ 0.22‡ 0.17‡ 0.01
A WAIST 0.90‡ - 0.68‡ 0.50‡ -0.32‡ 0.06* -0.19‡ 0.24‡ 0.13‡ 0.00
m WHR 0.45‡ 0.70‡ - 0.35‡ -0.24‡ 0.06* -0.21‡ 0.25‡ 0.12‡ 0.07*
e INS 0.49‡ 0.51‡ 0.43‡ - -0.30‡ -0.03 -0.35‡ 0.35‡ 0.17‡ 0.07*
r GLUC -0.27‡ -0.30‡ -0.29‡ -0.40‡ - -0.05 0.19‡ -0.21‡ -0.13‡ -0.02
I LDL 0.12‡ 0.11‡ 0.09‡ 0.09‡ -0.08 - -0.01 0.08† 0.05 0.06*
c HDL -0.21‡ -0.24‡ -0.27‡ -0.37‡ 0.24‡ -0.15‡ - -0.43‡ -0.01 0.05
a TG 0.18‡ 0.23‡ 0.29‡ 0.36‡ -0.29‡ 0.18‡ -0.41‡ - 0.11‡ 0.04
n SBP 0.18‡ 0.16‡ 0.11‡ 0.05 -0.06 0.02 0.04 0.04 - 0.68‡
s DBP -0.03 -0.03 0.01 -0.06* 0.05* 0.00 0.07† -0.02 0.74‡ -
SAPPHIRe N = 284 Japanese
N = 747 BMI WAIST WHR INS GLUC LDL HDL TG SBP DBP
BMI - 0.83‡ 0.31‡ 0.64‡ -0.33‡ 0.15* -0.29‡ 0.22‡ 0.18‡ 0.03
WAIST 0.78‡ - 0.62‡ 0.58‡ -0.28‡ 0.17† -0.24‡ 0.20‡ 0.09 -0.02
WHR 0.38‡ 0.73‡ - 0.30‡ -0.14 0.08 -0.13* 0.19‡ 0.06 0.02
C INS 0.54‡ 0.51‡ 0.31‡ - -0.41‡ 0.11 -0.37‡ 0.32‡ 0.19‡ 0.09
h GLUC -0.24‡ -0.23‡ -0.15‡ -0.41‡ - -0.07 0.35‡ -0.18† -0.10 0.03
i LDL 0.03 0.02 -0.03 0.01 0.05 - -0.10 0.08 0.09 0.10
n HDL -0.23‡ -0.23‡ -0.16‡ -0.31‡ 0.14‡ 0.02 - -0.41‡ -0.15* -0.09
e TG 0.33‡ 0.30‡ 0.25‡ 0.40‡ -0.25‡ 0.01 -0.37‡ - 0.21‡ 0.12*
s SBP 0.20‡ 0.20‡ 0.16‡ 0.07 -0.04 0.01 -0.01 0.13‡ - 0.69‡
e DBP 0.17‡ 0.17‡ 0.12‡ 0.08* -0.02 0.02 -0.02 0.14‡ 0.84‡ -
* p < 0.05; † p < 0.01 ; ‡ p < 0.001 ; †† A negative correlation of GLUC and INS is result of the inverse squared transformation of GLUC
It is important to mention that each network had different exclusion criteria when recruiting participants: GENOA had excluded any case of HT secondary to other diagnoses or HT onset after age 60; HyperGEN had excluded type I diabetics, secondary hypertensives, or HT onset after age 60; SAPPHIRe had excluded participants with the following conditions: if they were using insulin or other prescription for diabetes, with cancer diagnosis, cirrhosis of the liver, terminal illness, and body mass index (BMI) > 35 kg/m2.
Metabolic Syndrome, T2D and the VHA Definition
The categorical trait c-MetS was created employing the NCEP definition. c-MetS was defined by the presence of 3 or more of the following abnormalities in an individual: WAIST > 102 cm in men or > 88 cm in women, TG ≥ 150 mg/dl, HDL < 40 mg/dl in men or < 50 mg/dl in women, systolic blood pressure (SBP) ≥ 130 mm Hg and/or diastolic blood pressure (DBP) ≥ 85 mm Hg, or on treatment for HT, and GLUC ≥ 110 mg/dl or on treatment for diabetes [1]. The quantitative trait q-MetS was defined by the clustering patterns of 10 risk factors. The following risk factors were included in FA: BMI (kg/m2), WAIST (cm), waist-to-hip ratio (WHR), fasting insulin (INS, μU/ml), fasting GLUC (mg/dl), SBP and DBP, mm Hg), low density lipoprotein cholesterol (LDL, mg/dl), HDL (mg/dl) and fasting triglycerides (TG, mg/dl). To maintain consistency among the three networks, the minimum fasting time required was set at 8 hours.
Type 2 diabetes was defined by a fasting GLUC ≥ 126 mg/dl, or current use of hypoglycemic medication or insulin that was documented at examination in the clinic, or diabetes reported on questionnaires. An age of onset ≥ 40 years was also required to diagnose T2D [9].
In the FBPP pooled database, three important VHA variables were available from questionnaires: stroke or transient ischemic attacks, heart attack, and bypass or angioplasty. If any of these were reported in an individual, it was used to define the VHA status.
Statistical analysis
The 10 risk variables of MetS were checked for normality and outliers. Each variable was adjusted for age, age2, age3, and field center within each gender-by-race-by-Network group. INS, TG, and HDL were transformed by natural logarithm to render them approximately normal. Likewise, GLUC was transformed as the inverse of the squared value (1/GLUC2). These transformations, together with standardization to zero mean and unit variance, prepared the data for FA. FA was performed by employing the FACTANAL function in S-plus version 6.1, Insightful Corp., Seattle. We applied an exploratory factor analysis where the extraction of the latent factors was performed based on the maximum likelihood estimation [10]. The statistical details of FA may be found elsewhere [11,12]. In short, FA explains the relationships among the risk variables in terms of a fewer number of underlying latent factors. The data were analyzed with and without Varimax rotation. When no rotation is applied, the first few factors can be considered the most important for MetS. Kaiser (1958) proposed the Varimax rotation that maximizes the sum of variances (of squares of loadings) for latent factors [13]. The two options provide ways of identifying the structure of the latent factors for MetS. One may consider "No rotation" for understanding how risk factors cluster to represent MetS; Varimax rotation may be used to identify distinct latent factors. This can be useful in genetic analyses [14]. Factor loadings are correlation coefficients between the original risk variables and the latent factors [12]. A loading ≥ 0.4 is interpreted as representing an important contribution of an original variable to the latent factor (marked in bold in Tables 6 and 7).
Table 6 Loadings of the Original Risk Factors in the Latent Factors by Network, Ethnicity (No Rotation)
Rotation "NONE" Network/ Ethnicity Factor BMI WAIST WHR INS GLUC LDL HDL TG SBP DBP SS loadings
GENOA (African Americans) Factor 1 0.75 † 0.68 0.34 0.36 -0.26 -0.12 0.12 0.75 0.48 2.17
Factor 2 0.66 0.55 0.18 0.28 -0.12 -0.14 -0.66 -0.58 1.66
Factor 3 0.38 0.92 0.15 -0.16 -0.15 0.17 1.10
Factor 4 0.45 -0.32 0.20 -0.50 0.52 0.87
GENOA (Hispanics) Factor 1 1.00 0.84 0.33 0.40 -0.15 -0.17 0.13 0.16 2.07
Factor 2 0.16 0.37 -0.21 0.28 0.89 0.72 1.61
Factor 3 0.35 0.76 0.13 -0.13 -0.20 0.23 -0.29 -0.25 0.98
Factor 4 -0.13 -0.22 0.11 0.18 -0.27 0.85 0.92
GENOA (Whites) Factor 1 1.00 0.92 0.57 0.59 -0.33 -0.30 0.28 0.19 2.81
Factor 2 0.12 0.68 1.00 1.50
Factor 3 0.35 -0.25 0.19 -0.53 0.47 0.73
Factor 4 0.27 0.75 0.12 -0.16 0.70
HyperGEN (African Americans) Factor 1 0.77 0.69 0.37 0.35 -0.21 -0.11 0.15 0.77 0.46 2.21
Factor 2 0.64 0.58 0.27 0.35 -0.16 -0.20 0.11 -0.64 -0.60 1.80
Factor 3 0.36 0.80 0.28 -0.23 -0.25 0.30 1.05
Factor 4 0.39 -0.34 0.18 -0.51 0.54 0.87
HyperGEN (Whites) Factor 1 0.96 0.97 0.61 0.53 -0.33 -0.21 0.25 0.17 2.75
Factor 2 0.11 0.84 0.80 1.37
Factor 3 0.37 -0.21 -0.65 0.56 0.92
Factor 4 -0.23 0.16 0.58 0.14 0.45
SAPPHIRe (Chinese) Factor 1 0.38 0.73 1.00 0.32 -0.15 -0.17 0.25 0.16 0.12 1.93
Factor 2 0.53 0.39 0.30 -0.14 -0.11 0.23 0.77 0.78 1.82
Factor 3 0.61 0.48 0.44 -0.23 -0.21 0.18 -0.45 -0.47 1.36
Factor 4 0.11 -0.42 0.36 0.37 -0.49 0.70
SAPPHIRe (Japanese) Factor 1 0.83 1.00 0.62 0.58 -0.28 0.17 -0.24 0.20 2.63
Factor 2 0.19 0.18 -0.14 0.15 0.70 0.96 1.54
Factor 3 0.10 0.34 -0.41 -0.60 0.50 0.12 0.93
Factor 4 -0.52 0.36 -0.24 0.18 0.11 0.27 0.59
†The loadings > = 0.40 are in bold.
Table 7 Loadings of the Original Risk Factors in the Latent Factors by Network, Ethnicity (Varimax Rotation)
Rotation "Varimax" GENOA
African Americans Whites Hispanics
Factors / Variables Factor 1 Factor 2 Factor 3 Factor 4 Factor 1 Factor 2 Factor 3 Factor 4 Factor 1 Factor 2 Factor 3 Factor 4
BMI 0.99 † 0.02 0.14 0.07 0.97 0.11 0.14 0.12 0.99† 0.05 -0.01 0.04
WAIST 0.83 0.03 0.19 0.43 0.83 0.10 0.46 0.15 0.82 0.07 0.01 0.44
WHR 0.28 0.06 0.22 0.93 0.38 0.06 0.85 0.15 0.29 0.08 -0.11 0.88
INS 0.38 0.01 0.53 0.10 0.51 0.03 0.18 0.44 0.39 0.04 0.15 0.12
GLUC -0.21 -0.07 -0.38 -0.12 -0.28 -0.01 -0.09 -0.30 -0.14 -0.14 -0.18 -0.15
LDL 0.00 0.03 0.20 0.01 -0.03 0.04 -0.01 0.19 -0.03 0.04 0.20 0.01
HDL -0.10 0.05 -0.54 -0.06 -0.20 0.05 -0.12 -0.58 -0.17 0.06 -0.31 -0.10
TG 0.04 0.01 0.56 0.08 0.20 0.10 0.07 0.52 0.13 0.07 0.91 -0.01
SBP 0.10 0.99 0.07 0.04 0.16 0.68 0.03 0.09 0.11 0.94 0.07 0.07
DBP -0.04 0.75 0.02 0.02 -0.05 0.99 0.04 0.05 -0.01 0.76 0.04 0.04
SS Loadings 1.95 1.57 1.18 1.10 2.22 1.50 1.02 0.99 1.99 1.51 1.05 1.03
Cumulative Variance (%) 19.5 35.1 47 57.9 22.2 37.2 47.4 57.3 19.9 35.1 45.5 55.8
Rotation "Varimax" HyperGEN
African Americans Whites
Factors/Variables Factor 1 Factor 2 Factor 3 Factor 4 Factor 1 Factor 2 Factor 3 Factor 4
BMI 0.97 0.04 0.24 0.09 0.95 0.06 0.17 0.14
WAIST 0.83 0.03 0.25 0.44 0.83 0.02 0.15 0.51
WHR 0.32 0.05 0.31 0.81 0.32 0.05 0.21 0.75
INS 0.35 -0.02 0.56 0.17 0.42 0.11 0.47 0.15
GLUC -0.15 -0.02 -0.45 -0.13 -0.25 -0.07 -0.27 -0.13
LDL 0.07 0.00 0.21 0.01 0.03 0.06 0.02 0.06
HDL -0.06 0.07 -0.06 -0.08 -0.08 0.05 -0.68 -0.05
TG 0.02 0.02 0.64 0.12 0.09 0.07 0.61 0.13
SBP 0.13 0.99 0.03 0.02 0.11 0.85 0.06 0.04
DBP -0.06 0.76 -0.06 0.02 -0.04 0.80 -0.02 0.05
SS Loadings 1.90 1.56 1.54 0.93 1.97 1.40 1.23 0.90
Cumulative Variance (%) 19.0 34.6 50.1 59.3 19.7 33.7 45.9 54.9
Rotation "Varimax" SAPPHIRe
Chinese Japanese
Factors / Variables Factor 1 Factor 2 Factor 3 Factor 4 Factor 1 Factor 2 Factor 3 Factor 4
BMI 0.48 0.48 0.08 0.58 0.94 0.06 0.25 0.24
WAIST 0.80 0.38 0.08 0.35 0.66 -0.01 0.17 0.73
WHR 0.93 0.18 0.11 -0.29 0.12 0.00 0.13 0.71
INS 0.25 0.69 0.01 0.15 0.49 0.11 0.48 0.25
GLUC -0.07 -0.47 0.00 0.00 -0.20 0.03 -0.47 -0.08
LDL -0.01 -0.01 0.02 0.07 0.09 0.10 0.09 0.13
HDL -0.08 -0.46 0.01 0.03 -0.10 -0.09 -0.65 -0.09
TG 0.12 0.60 0.12 -0.07 0.05 0.12 0.54 0.10
SBP 0.09 0.03 0.90 0.11 0.10 0.70 0.16 0.01
DBP 0.05 0.05 0.92 0.10 -0.05 0.99 0.00 0.04
SS Loadings 1.85 1.68 1.68 0.60 1.64 1.53 1.32 1.20
Cumulative Variance (%) 18.5 35.3 52.0 58.0 16.4 31.6 44.8 56.8
†The loadings > = 0.40 are in bold.
Odds ratios, prevalence rates, and their confidence intervals were estimated by utilizing the FREQ and LOGISTIC regression procedures of SAS. Means and standard deviations were estimated with SAS software v. 9.0., SAS Institute, NC [15].
Results
Mean age of participants ranged from 48 years in the HyperGEN African Americans to 58 years in the GENOA African Americans (Tables 1, 2, 3). Mean BMI and WAIST were higher for African Americans, Hispanics, and Whites than for Japanese and Chinese. A mean of approximately 130 mm Hg with a standard deviation of 20 mm Hg was evident for SBP across the networks and ethnicities.
c-MetS
Among African Americans, twice as many were hypertriglyceridemic in GENOA than in HyperGEN, even though they were very comparable for all other MetS risk factors. This difference contributed to a relatively higher percentage of c-MetS in GENOA African Americans (41%) than in HyperGEN African Americans (34%). Whites in both networks were similar with respect to all NCEP thresholds (Figure 1).
Figure 1 Percent of Participants by Network and Ethnicity that Pass Thresholds per Each Risk Factor as Defined by NCEP and Percent of Participants that had 3, 4, and 5 MetS Risk Factor Combinations Beyond the NCEP Thresholds. Footnote. MS3, MS4 and MS5: three, four and five risk factors beyond the NCEP thresholds; AACo-African Americans combined; AAG-Genoa African Americans; AAH-HyperGEN African Americans; WCo-Whites combined; WG-GENOA Whites; WH-HyperGEN Whites; ACo-Asians combined; AC-Chinese Asians; AJ-Japanese Asians; His-Hispanics
The percentages of participants beyond the NCEP thresholds differed by ethnicity. The percentage of Japanese above the NCEP threshold for WAIST was half that of Hispanics, African Americans, or Whites, but twice that of the Chinese. The percentage of Japanese participants above the NCEP threshold for TG was similar to that in Whites, but about twice as large as in the Chinese; twice as many Whites were above the TG threshold than African Americans. The percentage of Japanese above the GLUC threshold, or on hypoglycemic medications, was similar to that in Whites, but about 3 times lower than in Hispanics and about 3 times higher than in the Chinese sample. The primary risk factors for c-MetS were WAIST, HDL, and blood pressure (BP) in African Americans, Hispanics, and Whites; TG had smaller weight in African Americans, but higher in Hispanics and Whites; in the Japanese, TG, HDL, BP, and to a smaller degree WAIST; and in the Chinese, HDL, BP, and less of TG. Hispanics also had a higher contribution of GLUC. Hispanics had a total of 73% of participants with at least 3 risk factors beyond the thresholds (MS3: 24%; MS4: 30%; MS5: 19%). Levels lower than the NCEP thresholds were more frequent in the Chinese and in the Japanese samples than in Hispanics, African Americans, or Whites. As expected, all networks and ethnicities selected had a high percentage of participants above the NCEP threshold for SBP/DBP or using hypertensive medications. For a better comparison, similar ethnicities were combined across networks (Figure 1).
Performing logistic regression on c-MetS risk factors, by using dummy variables defined as 1 if a risk factor was equal or beyond the NCEP thresholds, 0 otherwise, showed significant differences among ethnicities for most of the 5 NCEP MetS risk factors (Results not shown). Looking at the q-MetS patterns, Asians had more a contribution of hypertension and dyslipidemia. Whites and African Americans showed a classical MetS as a combination of obesity, dyslipidemia, and hypertension, much like the Hispanics (Figure 1).
c-MetS and T2D
The presence of c-MetS was associated with T2D. This association was highly expressed in the Hispanics, where c-MetS and T2D were present simultaneously in 57% of the sample. The lowest percentage of participants with both c-MetS and T2D was 2% in the Chinese (Figure 2). All the association tests between c-MetS and T2D were significant with p-values < 0.0001 (Table 4). The odds of having T2D were 10.8 higher (8.3 – 14.1 a 95% odds CI) in GENOA African Americans if participants had c-MetS. The odds of having T2D when having c-MetS for Chinese and Japanese were about 5 with 95% odds CI (3.1 – 11.0) and (2.8 – 8.8), respectively.
Figure 2 Frequency of Subjects by Network Within Ethnicity Given MetS and T2D Affection Status.
Table 4 Association Between MetS and T2D in Networks Within Ethnicity
Odds Ratio Prevalence Ratio
Network: Ethnicity (95% Confidence Interval) (95% Confidence Interval)
GENOA: African Americans 10.8 (8.3 – 14.1) 6.2 (4.9 – 7.8)
GENOA: Hispanics 7.3 (5.8 – 9.2) 2.5 (2.2 – 2.8)
GENOA: Whites 8.0 (5.3 – 12.2) 6.5 (4.4 – 9.6)
HyperGEN: African Americans 9.7 (7.6 – 12.4) 5.7 (4.6 – 5.9)
HyperGEN: Whites 6.7 (4.9 – 9.1) 5.1 (3.9 – 6.8)
SAPPHIRe: Chinese 5.8 (3.1 – 11.0) 5.4 (2.9 – 9.9)
SAPPHIRe: Japanese 5.0 (2.8 – 8.8) 4.0 (2.4 – 6.6)
VHA
The highest prevalence of VHA events was 17% in Hispanics, followed by the HyperGEN Whites with 15%; the lowest prevalence was 4% in the Chinese. The rest of the groups had intermediate VHA prevalence: GENOA African Americans, 10%; GENOA Whites, 10%; HyperGEN African Americans 13%; and Japanese 7%.
q-MetS
Table 5 presents correlation matrices and sample sizes for 10 risk variables for MetS. Each upper or lower triangle of a matrix corresponds to correlation coefficients and the associated significance levels for a specified network and ethnicity. All networks and ethnicities studied showed a high correlation of BMI with WAIST (from 0.78 in the Chinese to 0.9 in the GENOA Whites and the HyperGEN African Americans), and a lower correlation of BMI with WHR (from 0.31 in the SAPPHIRe Japanese to 0.51 in the GENOA Whites). INS was correlated with BMI, WAIST, GLUC, and TG, but displayed significantly lower correlation with WHR. As expected TG was negatively correlated with HDL. SBP and DBP were highly correlated to each other (from 0.68 in the GENOA Whites to 0.84 in the SAPPHIRe Japanese). These patterns of correlations were likely to shape the latent factors created by analyzing the 10 MetS risk variables investigated.
q-MetS with no Rotation
Tables 6 and 7 illustrate how groups of the original risk variables contribute (load) on latent factors from FA with and without Varimax rotation. The sums of squares of loadings (SS Loadings) correspond to the proportion of the variance of the original variables explained by each latent factor identified. When no rotation was performed on the factors, the main factor of African Americans was composed of BMI, INS, SBP, DBP, WAIST, and WHR. In GENOA Hispanics and Whites, BMI, INS, WAIST, and WHR contributed in the first factor. For the Chinese, the first factor loaded only obesity and explained 19% of the original variance; the third factor loaded primarily obesity, INS, and negative SBP and DBP, and explained about 14% of the original variance. The blood pressure components loaded in general on a separate factor with the exception of African Americans where the loading tended to be on the first factor. Lipids also contributed on a separate factor.
q-MetS with Varimax Rotation
After performing the Varimax rotation, Factor 1 in the GENOA African Americans explained about 20% of the variance in the original 10 risk variables. Four factors explained from 55% of the variance in the original risk variables in the HyperGEN Whites to 60% in the HyperGEN African Americans. The four factors identified were not identical among the ethnicities in each network. Factor 1 loaded essentially BMI, WAIST, and INS in African Americans, Hispanics, Whites, and Japanese, but less BMI and INS compared to WAIST and WHR in Chinese. SBP and DBP contributed in a separate factor for each network and ethnicity. TG, HDL, INS, and GLUC contributed in a separate latent factor. The remaining fourth factors loaded WAIST and WHR in most of the networks' ethnicities, but in the Chinese loaded BMI and WAIST, with negative loading for WHR.
Discussion
One of the major contributors to MetS, as can be seen in Figure 1, was high BP. About 70% of African Americans, Whites, or Asians, and 58% of Hispanics sampled had BP above the NCEP threshold or used anti-HT medications. These findings coincided with the main ascertainment in the sampled populations, reflecting the main goal of FBPP, to study the genetic causes of high blood pressure. The ascertainment schemes within each network may have played a role in the observed associations of the features of MetS and the prevalence of c-MetS. However, the characteristics described in the results stress that there are important ethnic differences, which need to be taken into consideration when evaluating / diagnosing MetS.
If we compare the prevalence of MetS in our study and a 23–24% of U.S. MetS prevalence reported by Ford et al. (2004) using data from the National Health and Nutrition Examination Survey (NHANES), it is evident that our US samples have a higher prevalence of MetS than the general US population [16]. In our study of 3,867 African Americans, 3,466 Whites, 2,211 Asians and 1,799 Hispanics, 37%, 46%, 21%, and 73%, respectively, were classified with c-MetS. This trend emphasizes the fact that selection for hypertension in most cases was associated with higher prevalence of MetS. Another example emphasizing that selection for a disorder part of the MetS, increases the prevalence of MetS, comes from a multinational study, Genetic Epidemiology of Metabolic Syndrome Project. This study has revealed a prevalence of 76% of MetS out of 1,436 participants, as result of selecting for atherogenic dyslipidemia [17].
The prevalence of MetS was comparable across Networks within the same ethnicity. However, there are ethnic differences in the prevalence of MetS. Prevalence of c-MetS is high in GENOA Hispanics (Figure 1). They also show high association of c-MetS with T2D (Figure 2). Although we believe that the prevalence of MetS is influenced by selection for type 2 diabetes, these results are in accordance with a large body of literature that illustrates that Hispanics have a trend for being more susceptible to MetS. Simon et al (2003) have reported that the prevalence of T2D was approximately two times higher among Hispanics than non-Hispanics [18]. McNeely and Boyko (2004) have reported that odds ratios for diabetes, compared to Whites, were 1 for Asians, 2.3 for African Americans, 2 for Hispanics, 2.2 for Native Americans, and 3.1 for Pacific Islanders [19]. Sanchez-Castillo et al. (2004) reported that in excess of 50% of adult population in Mexico are overweight and obese [20]. Furthermore, in our data, we found that the VHA events were highest in Hispanics. Our findings are in accord with the literature reporting that Mexican Americans had a 70% greater risk of cardiovascular mortality, and a 60% greater risk of coronary heart disease mortality than non-Hispanic Whites [21]. A higher incidence of hospitalized myocardial infarction in Mexican Americans than non-Hispanic Whites was also reported [22].
Conversely, Asians (and especially the Chinese) are leaner than others. We recognize that SAPPHIRe exclusion criteria biased the obesity findings. They also had lower T2D prevalence (Figure 2), because specifically the treated type 2 diabetics were excluded earlier than the clinical visit. They had lower prevalence of c-MetS. It is suggested that the NCEP criteria for obesity may not be suitable for the Japanese [23]. Tan et al. (2004) suggested that the NCEP definition of MetS underestimates its prevalence in Asian populations, because it embodies an unsuitable threshold of central obesity for Asians [24]. For example, in the FBPP Chinese sample (which represented individuals of Chinese origin living in Taiwan), if one would have lowered the threshold for WAIST as Tan et al. (2004) suggested, the prevalence of c-MetS in them would have increased.
Among African Americans and Hispanics, men had significantly lower odds of having c-MetS than women (Results not shown). Other authors have concluded that African-American women and Hispanic men and women have the highest prevalence of MetS. They attributed this to higher BP, obesity, and diabetes in African Americans, and the high prevalence of obesity and diabetes in Hispanics [25]. In the FBPP, more Whites had TG and HDL beyond the NCEP threshold as compared to African Americans.
In our study, each of the ethnicities considered showed significant MetS and T2D associations. Young et al. (2003), in a longitudinal cohort study of 429,918 veterans with diabetes, found that African Americans and Native Americans had a higher odds ratio (1.3 and 1.5 respectively) for having early diabetic nephropathy than Whites [26]. In the FBPP, the Hispanic sample exhibits a high occurrence of MetS along with T2D (57%) in association with a constellation of several risk factors for MetS beyond the NCEP thresholds. Our data (Figure 2), demonstrate also a small group of subjects with T2D, not classified as having MetS. This group is intriguing, because three or more risk factors are under the NCEP threshold, and it represents a deviation from the general notion that a cluster of risk factors of MetS may lead to T2D development. Is it possible that the scale for classifying T2D is error prone? Is there any genetic factor in this group that affects GLUC levels in the blood, without interfering with obesity and dyslipidemia pathways? A genetic analysis of this group in contrast with one having concurrently MetS and T2D, may identify important genetic differences related to MetS.
Four independent factors were identified when factor analysis was performed with Varimax rotation. Their pattern was very similar in African Americans, Hispanics, Whites, and Japanese, but not entirely so in Chinese. BMI, WAIST, and INS contributed together mainly in a factor labeled by us as "Obesity-INS." SBP and DBP contributed in a separate "BP" factor. A "Lipids-INS" factor was constructed mainly from contributions of LDL, HDL, TG, and INS. The last, "Central obesity" factor, was mostly an involvement of WAIST and WHR. These 4 factors were persistent also by gender in the HyperGEN data [14]. When no rotation was employed, the main MetS factor represented primarily a contribution of obesity together with INS in Hispanics, Whites, and Japanese; obesity and BP in African Americans; and obesity in Chinese. These patterns are quite important for a geneticist, because they show possible underlying trait combinations. The known interactions among traits grant ways to investigate the underlying genes, proteins and their substructures involved in these communications. For a clinician, the traits groupings shed light on the most important factors to be tackled when combating MetS. For the pharmacological research, these patterns can help in envisioning new medications intended to tackle the excess expression of risk variables in one, two, and/or three factors at once.
In general, our results about the structure of the factors, which reflect multivariate correlations of the variables studied, are supported by the literature. However, there are also differences that could be the result of variations in recruitment. In a study of Japanese Americans, it was found that visceral fat was a significant correlate of hypertension and independent of fasting INS [27]. In contrast, we found that correlations of WAIST/WHR with INS were highly significant in the Japanese, but not correlated with BP components.
In conclusion, patterns of the MetS were relatively similar across networks within ethnicity, but were statistically different among ethnicities. Overall, obesity was the most prominent compound risk factor expressed in both c-MetS and q-MetS. However, the degree of consistency in factor structures observed across ethnicities and networks is remarkable given that there are considerable differences in the Network-specific study designs. The notable exception of Hispanics in GENOA is quite understandable since the sample was also enriched for T2D. Thus, some of the differences especially in the prevalence of MetS are, at least in part, attributable to the study design differences. Nevertheless, the increase of MetS prevalence in our U.S. samples compared to the U.S. general population confirmed that there is an important link between HT and MetS. Together, our results underline that MetS is a compound phenotype, where obesity, dyslipidemia, and hypertension enable MetS. If we assume that obesity and dyslipidemia have separate biochemical pathways for their expression, it appears that the presence of INS in both latent obesity and lipids factors may be an indication that INS is an important contributor and possibly a connector of pathways in the development of MetS.
The reported findings will be useful if they lead to innovations. One application of these results can be the genetic analysis of the new MetS created data. It is well known that a categorical trait has less power in detecting genetic linkage as compared to a quantitative trait for a complex phenotype. Two types of q-MetS factor scores (with and without Varimax rotation) provide ample opportunity to discover quantitative trait loci for MetS. Parallel with this work, we have undertaken a detailed genetic analysis of the MetS factors that will be reflected in another publication (unpublished observations). Qualitative and quantitative characterization of MetS in the rich Family Blood Pressure Program pooled data will help in getting a better understanding of the genetic inheritance underlying MetS and its interaction with the environmental causes.
List of Abbreviations Used
FBPP, Family Blood Pressure Program; NCEP, National Cholesterol Education Program; MetS, metabolic syndrome; c-MetS, qualitative MetS; q-MetS, quantitative MetS; OBS, obesity; T2D, type 2 diabetes; FA, factor analysis; VHA, vascular heterogeneous atherosclerotic events; HT, hypertension; BMI, body mass index; WAIST, waist circumference; WHR, waist to hip ratio; INS, insulin; GLUC, glucose; TG, triglycerides; LDL, low density lipoprotein cholesterol; HDL, high density lipoprotein cholesterol; SBP, systolic blood pressure; DBP, diastolic blood pressure; BP, blood pressure, CI-confidence interval.
Competing interests
The author(s) declare that they have no competing interests.
Acknowledgements
The entire FBPP is supported by a series of cooperative agreements with the NHLBI. GenNet: (U10s) HL54466, HL54485, HL54508, HL54512, and HL64777; GENOA: (U10s) HL54457, HL54463, HL54464, HL54481, HL54504, and HL54526; HyperGEN: (U10s) HL54471, HL54472, HL54473, HL54495, HL54496, HL54497, HL54509, and HL54515; SAPPHIRe: (U01s) HL54527, and HL54498.
The complete list of FBPP investigators and sources of support can be found at
We thank Prof. Steven C. Hunt, Prof. John Grove, Prof. Richard A. Olshen, three anonymous reviewers, and the editors of the Nutrition & Metabolism journal for their suggestions in improving a previous version of this article.
==== Refs
Third report of the National Cholesterol Education Program (NCEP) expert panel on detection, evaluation and treatment of high blood cholesterol in adults (Adult Treatment Panel III) Final report Circulation 2002 106 3144 3421
Ford ES Factor analysis and defining the metabolic syndrome Ethn Dis 2003 13 429 437 14632262
Choi KM Lee J Kim KB Kim DR Kim SK Shin DH Kim NH Park IB Choi DS Baik SH South-west Seoul Study Factor analysis of the metabolic syndrome among elderly Koreans – the South-west Seoul Study Diabet Med 2003 20 99 104 12581260 10.1046/j.1464-5491.2003.00890.x
Novak S Stapleton LM Litaker JR Lawson KA A confirmatory factor analysis evaluation of the coronary heart disease risk factors of metabolic syndrome with emphasis on the insulin resistance factor Diabetes Obes Metab 2003 5 388 396 14617224 10.1046/j.1463-1326.2003.00289.x
Ramos R Marrugat J Basagana X Sala J Masia R Elosua R REGICOR Investigators The role of age in cardiovascular risk factor clustering in non-diabetic population free of coronary heart disease Eur J Epidemiol 2004 19 299 304 15180099 10.1023/B:EJEP.0000024697.55346.c2
Schutte AE van Rooyen JM Huisman HW Kruger HS de Ridder JH Factor analysis of possible risks for hypertension in a black South African population J Hum Hypertens 2003 17 339 348 12756407 10.1038/sj.jhh.1001553
Shen BJ Todaro JF Niaura R McCaffery JM Zhang J Spiro A 3rdWard KD Are metabolic risk factors one unified syndrome? Modeling the structure of the metabolic syndrome X Am J Epidemiol 2003 157 701 711 12697574 10.1093/aje/kwg045
The FBPP Investigators Multi-center genetic study of hypertension. The Family Blood Pressure Program (FBPP) Hypertension 2002 39 3 9 11799070 10.1161/hy1201.100415
Hunt KJ Heiss G Sholinsky PD Province MA Familial history of metabolic disorders and the multiple metabolic syndrome: the NHLBI family heart study Genet Epidemiol 2000 19 395 409 11108648 10.1002/1098-2272(200012)19:4<395::AID-GEPI10>3.0.CO;2-3
S-PLUS 6 for Windows Guide to Statistics 2003 2 Insightful Corporation, Seattle, WA 65 82
Johnson RA Wichern DW Applied multivariate statistical analysis 1992 3-d New Yersey 396 458 1313637
Khattree R Naik DN Multivariate data reduction and discrimination with SAS software 2000 Kari, NC 111 210
Kaiser HF The varimax criterion for analytic rotation in factor analysis Psychometrika 1958 23 187 200
Kraja AT Hunt SC Pankow JS Myers RH Heiss G Lewis CE Rao DC Province MA An evaluation of the metabolic syndrome in the HyperGEN study Nutrition & Metabolism 2005 2 2 15656912 10.1186/1743-7075-2-2
Stokes ME Davis CS Koch GG 2-nd Categorical data analysis using the SAS system 2000 Cary, NC
Ford ES Giles WH Mokdad AH Increasing prevalence of the metabolic syndrome among U.S. adults Diabetes Care 2004 27 2444 2449 15451914
Wyszynski DF Waterworth DM Barter PJ Cohen J Kesäniemi YA Mahley RW McPherson R Waeber G Bersot TP Sharma SS Nolan V Middleton LT Sundseth SS Farrer LA Mooser V Grundy SM Relation between atherogenic dyslipidemia and the Adult Treatment Program-III definition of metabolic syndrome (Genetic Epidemiology of Metabolic Syndrome Project) Am J Cardiol 2005 95 194 198 15642551 10.1016/j.amjcard.2004.08.091
Simon PA Zeng Z Wold CM Fielding JE Burrows NR Diabetes among Hispanics – Los Angeles County, California, 2002–2003 MMWR Morb Mortal Wkly Rep 2003 52 1152 1155 14647017
McNeely MJ Boyko EJ Type 2 diabetes prevalence in Asian Americans: results of a national health survey Diabetes Care 2004 27 66 69 14693968
Sanchez-Castillo CP Pichardo-Ontiveros E Lopez RP The epidemiology of obesity Gac Med Mex 2004 140 S3 S20 15641467
Hunt KJ Resendez RG Williams K Haffner SM Stern MP Hazuda HP All-cause and cardiovascular mortality among Mexican-American and non-Hispanic White older participants in the San Antonio Heart Study – evidence against the "Hispanic paradox" Am J Epidemiol 2003 158 1048 1057 14630600 10.1093/aje/kwg249
Goff DC Nichaman MZ Chan W Ramsey DJ Labarthe DR Ortiz C Greater incidence of hospitalized myocardial infarction among Mexican American than non-Hispanic whites Circulation 1997 95 1433 1440 9118510
Zhang H Tamakoshi K Yatsuya H Murata C Wada K Otsuka R Nagasawa N Ishikawa M Sugiura K Matsushita K Hori Y Kondo T Toyoshima H Long-term body weight fluctuation is associated with metabolic syndrome independent of current body mass index among Japanese men Circ J 2005 69 13 18 15635195 10.1253/circj.69.13
Tan C Ma S Wai D Chew S Tai ES Can we apply the National Cholesterol Education Program Adult Treatment Panel definition of the metabolic syndrome to Asians? Diabetes Care 2004 27 1182 1186 15111542
Hall WD Clark LT Wenger NK Wright JT JrKumanyika SK Watson K Horton EW Flack JM Ferdinand KC Gavin JR 3rdReed JW Saunders E O'Neal W Jr The Metabolic Syndrome in African Americans: a review Ethn Dis 2003 13 414 428 14632261
Young BA Maynard C Boyko EJ Racial differences in diabetic nephropathy, cardiovascular disease, and mortality in a national population of veterans Diabetes Care 2003 26 2392 2399 12882868
Hayashi T Boyko EJ Leonetti DL McNeely MJ Newell-Morris L Kahn SE Fujimoto WY Visceral adiposity and the prevalence of hypertension in Japanese Americans Circulation 2003 108 1718 1723 12975250 10.1161/01.CIR.0000087597.59169.8D
|
16076393
|
PMC1201342
|
CC BY
|
2021-01-04 16:37:46
|
no
|
Nutr Metab (Lond). 2005 Aug 2; 2:17
|
utf-8
|
Nutr Metab (Lond)
| 2,005 |
10.1186/1743-7075-2-17
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1614442610.1371/journal.pbio.0030335Research ArticleGenetics/Genomics/Gene TherapyRecombination Hotspots and Population Structure in Plasmodium falciparum
Recombination Variation in P. falciparumMu Jianbing
1
Awadalla Philip [email protected]
2
Duan Junhui
1
McGee Kate M
2
Joy Deirdre A
1
McVean Gilean A. T
3
Su Xin-zhuan [email protected]
1
1 Laboratory of Malaria and Vector Research, National Institutes of Health, Rockville, Maryland, United States of America,2 Department of Genetics, North Carolina State University, Raleigh, North Carolina, United States of America,3 Department of Statistics, University of Oxford, Oxford, United KingdomHey Jody Academic EditorRutgers UniversityUnited States of America10 2005 13 9 2005 13 9 2005 3 10 e3358 4 2005 26 7 2005 Copyright: © 2005 Mu 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.
Clues to the Evolution of the Malarial Chromosome
Understanding the influences of population structure, selection, and recombination on polymorphism and linkage disequilibrium (LD) is integral to mapping genes contributing to drug resistance or virulence in Plasmodium falciparum. The parasite's short generation time, coupled with a high cross-over rate, can cause rapid LD break-down. However, observations of low genetic variation have led to suggestions of effective clonality: selfing, population admixture, and selection may preserve LD in populations. Indeed, extensive LD surrounding drug-resistant genes has been observed, indicating that recombination and selection play important roles in shaping recent parasite genome evolution. These studies, however, provide only limited information about haplotype variation at local scales. Here we describe the first (to our knowledge) chromosome-wide SNP haplotype and population recombination maps for a global collection of malaria parasites, including the 3D7 isolate, whose genome has been sequenced previously. The parasites are clustered according to continental origin, but alternative groupings were obtained using SNPs at 37 putative transporter genes that are potentially under selection. Geographic isolation and highly variable multiple infection rates are the major factors affecting haplotype structure. Variation in effective recombination rates is high, both among populations and along the chromosome, with recombination hotspots conserved among populations at chromosome ends. This study supports the feasibility of genome-wide association studies in some parasite populations.
==== Body
Introduction
The interaction between Plasmodium falciparum and humans has been a potent selective force in the evolution of both species. The high mortality rate—an estimated 1.1–2.7 million people die each year from malaria [1]—leads to strong selection, both on host genes that contribute to resistance [2,3], and on parasite genes involved in the infection process [4]. Colonization of novel populations and the use of anti-malarial drugs have imposed additional selective forces on the parasite genome [5–7]. The signatures of selection can potentially be used to map genes associated with drug resistance, local adaptation, or antigenic interactions [8]. Two key factors, however, limit the power of population-based approaches to gene mapping. First, population structure, caused either by geographic isolation or epidemiologic stratification, can generate spurious association between phenotypic and genetic variation. Second, the extent of linkage disequilibrium (LD) within populations, which determines both the number of markers required for association studies and the mapping resolution, is unknown. LD can be influenced by diverse factors, including the recombination rate, the local parasite effective population size, population differentiation, and the extent of inbreeding [9]. Determining the extent to which geography, recombination, and host-parasite interactions have shaped genomic structure requires analysis of global-scale patterns of genetic variation.
For P. falciparum, data describing population structure and LD have come mostly from nucleotide polymorphisms in a number of genes encoding candidate malaria vaccines that may be under strong host immune selection [10–12]. More “neutral” microsatellite markers sampled from across the genome have also been employed to study P. falciparum populations. In a study using 12 microsatellites, parasite populations were clustered into major groups based on their continental origins, with the majority of variation found within locations in Africa and Papua New Guinea (PNG), but also between subpopulations from different sampling sites in South America (Brazil, Colombia, and Bolivia) [13]. Because the microsatellites were located on different chromosomes, the study could not provide information about the chromosomal scale of LD and variation in population recombination rates. Similarly, significant microsatellite allele-sharing was found among South American isolates, but not among those from Africa, suggesting distinct population structure in different continents [5]. Although variation in cross-over rates was observed in a meiotic map constructed for P. falciparum [14,15], the power to detect recombination variation using genetic crosses is limited. In contrast, population genetic data provide greater resolution in detecting historical recombination events, which can be estimated from events occurring over generations ago.
In both the yeast and human genomes, recombination principally occurs at specific regions called “hotspots” [16–19]. However, very limited information on recombination hotspot locations at a genome-wide scale is available in other organisms, particularly for important human pathogens such as P. falciparum. The molecular mechanisms determining hotspot location and activity are largely unknown. In yeast, chromatin structure influences the initiation of double-strand breaks at hotspots [18], but no specific sequence or motif has been identified as causing recombination hotspots. The observation of meiotic drive associated with hotspots has led some to suggest that they may be short-lived because of selection against sites that initiate double-strand breaks [20]. Identification of recombination hotspots and a better understanding of the mechanism underlying hotspot activity will not only facilitate association studies, but also shed light on P. falciparum genome evolution.
Recombination allows sites to evolve independently, and thus it may act as a diversifying force, generating new variants for the parasite to evade host immunity. The malaria parasite has asexual and sexual stages in its life history. In natural P. falciparum populations, recombination events will be observed only when a host is infected with multiple genotypes. As a result, “effective” population recombination rates are directly correlated with frequencies of multiple infections, endemicity, and the out-crossing rate [21]. Understanding the patterns of variation in recombination rates along the chromosome is critical for detecting regions that have been potentially subjected to selective sweeps [22]. To study the nature and scale of LD and recombination variation in the parasite genome, we assayed 99 worldwide P. falciparum isolates (Figure S1) and one chimpanzee parasite (P. reichenowi) for single nucleotide polymorphisms (SNPs) spanning Chromosome 3, at an average interval of one SNP per ∼5.5 kilobases (kb) [15]. We show high variation in recombination rates, among different parasite populations, and along the length of Chromosome 3. The presence of LD in some populations, and selection by anti-malarial agents, support the possibility of genome-wide association for genes affecting drug response of malaria parasites.
Results/Discussion
SNP Collection and Distribution
The SNPs were ascertained by re-sequencing based on variant sites identified from a panel of five global isolates [15]. Of the 183 SNPs that produced genotypes for most of the 99 isolates, 138 (75.4%) are SNPs with a global minor allele frequency greater than or equal to 0.02. Although 105 SNPs segregate in more than one of the four major regions surveyed—Africa, Southeast (SE) Asia, South and Central America, and PNG—only 39 segregate in all populations; 21 SNPs segregate in Africa alone, 27 in America, 16 in SE Asia, and 9 in PNG. Average pair-wise nucleotide diversity for the four major geographic regions revealed similar per-site heterozygosity values (relative values to Africa equal 1; America, 1.13; SE Asia, 0.87; and PNG, 0.91). The continent-specific SNPs suggest potential population structure in this worldwide parasite collection.
Parasite Populations Genetically Clustered According to Their Continental Origins
Because our parasites were collected from various locations at different times, population differentiation may affect our inferences of population parameters [23]. We therefore estimated potential population structure using a Bayesian model-based clustering algorithm in Structure 2 [24]. The program assumes a model with K populations, with each population characterized by a set of allele frequencies at each locus. The likelihood of a series of K values is calculated and a best K value, based on likelihood tests, is selected after a series of runs with different K values. Genotype data can be analyzed using different ancestry models, such as an admixture model for individuals with mixed ancestry, or a linkage model for recently mixed populations with linked markers. For the worldwide parasite populations, both admixture and linkage models produced similar results, with the most positive likelihood values (or best statistical support) when the number of populations (K), a priori, was assumed to be five (Figure 1A). Most individuals clustered according to continent, regardless of their location or time of isolation, with the exception of a few isolates (569, T2/C6, Camp, K1, JAV, and 106/1) possibly due to either contamination during in vitro culture or human migration. Although parasites from SE Asia and PNG grouped together, some differences can be seen (Figure 1A). The 3D7 parasite (widely believed to be from Africa) clustered with 105/7 from Sudan, differing by only four SNPs, and yet separately from the rest of the African isolates. This result suggests that the two parasites are closely related, but that the origin of 3D7 requires further investigation. Similarly, parasite isolate 106/1 is closely related to the SE Asian parasite FCB, as the two have identical chromosomal haplotypes (with the exception of one SNP), similar to results from microsatellite genotypes [5]. Parasites from Central America (HB3 and Haiti) and coastal South America (ECU and JAV) showed some differences from those of Brazil and Peru. The clustering of JAV with African parasites, despite having part of its genome shared with those of South American parasites, suggests that the parasite might have migrated from Africa more recently, and recombination had allowed its genome to be mixed with those of South American parasites. The clustering of HB3 and P. reichenowi may simply reflect their relatively unique genetic backgrounds, as compared with the rest of the isolates.
Figure 1 Inferred Population Structure of Global Parasite Isolates
Each vertical bar represents an individual parasite with its names given at the top of the panels. Predefined numbers of populations (K = 2–8) were run ten times each, using Structure 2.
(A) Population partitions using SNPs from Chromosome 3 (K = 4–6, linkage model). At K = 5, the most consistent membership coefficients were obtained.
(B) Population partition using SNPs from 37 putative transporters (admixture model). Similar clustering was obtained with that of Chromosome 3, including Camp and T2/C6 (no data for K1) with African parasites and 106/1 with SE Asian parasites. However, African parasites were partitioned into chloroquine-resistant and -susceptible parasites. Note: Only 81 are available for the transporter data.
*, parasites resistant to chloroquine; ?, parasites from which drug data are not available.
The continental genetic partitions were supported by analyses of population variance FST, a fixation index describing genetic variation among populations [25] (Table 1), confirming significant population differentiation among parasites from different continents, which accounts for 24% of the total genetic variation. Although Structure 2 did not resolve differences between isolates from SE Asia and PNG, significant heterogeneity between isolates was observed when these populations were treated as separate, a priori, and hence, as individual populations, in the recombination analyses.
Table 1 Pairwise FST of Subpopulations from Different Regions of the World
Effects of Drug Selection on the Inference of Population Structure
Chromosome-wide variants provide not only a means for inferring population structure, but also a genomic control for comparison with other variants that are functionally significant, with respect to evasion of host immunity or drug resistance. We compared the inferred population structure from Chromosome 3 variants with estimates using 102 SNPs from 37 putative transporter genes on different chromosomes (Table S1), some of which were shown to be associated with chloroquine resistance [26]. The membership partition of parasite isolates using the transporter SNPs is largely concordant with those from Chromosome 3 variants, grouping the parasites into major geographic boundaries; however, obvious exceptions are the presence of two clear subpopulations in Africa, and the separation of PNG parasites from SE Asian parasites (Figure 1B). Interestingly, the subpopulations in Africa are partitioned according to parasite responses to chloroquine, i.e., one group consisting of chloroquine-resistant parasites (red bars in Figure 1B) and the other of chloroquine-sensitive parasites, regardless of sampling location within Africa, including parasites simultaneously isolated from one location, such as those from Ghana (9013, 9016, 9020, 9021). In contrast, clustering the African parasites using Chromosome 3 SNPs at K = 2 did not separate the African parasites into resistant and sensitive groups (unpublished data). Similarly, the separation of PNG isolates from SE Asian parasites could be due to the independent origin of chloroquine-resistance founder mutations in PNG. These results illustrate how selection on specific genes can greatly influence estimates of population structures.
High Recombination Rate Variation among Different Parasite Populations
A key factor in the success of association studies is the detection of variation in LD and population recombination rates, within and between populations. Here we examined the evidence for variable recombination rates, among populations and along chromosomes, using recently developed parametric [27] and nonparametric [28] methods (see Materials and Methods), respectively. The parametric methods use a model to infer the effective population recombination rate [2Ner(1 − f)], where Ne is the effective population size, r is the per-generation recombination rate, and f is the inbreeding coefficient; the nonparametric method estimates the minimum number of recombination events (Rh). We cannot separate the effects of Ne, r, and f on individual estimate of a sample, but we can show the relationship of f on estimates of the effective population recombination parameter nonetheless. This is because the variation in estimates of recombination rates among populations is due to variation in transmission frequencies and/or f. The parametric methods have been shown to be robust in the context of alternative mutation models and have high power to detect recombination, even with SNP ascertainment bias [19,27].
Chromosome-wide estimates of the population recombination rate parameter ranged from ∼400 per Mb in the American to over 105 per Mb in the African populations (Table 2). Although the Structure 2 analyses did not point to a distinction between PNG and SE Asia populations, FST estimates suggest significant differentiation between the two. We therefore calculated the rates for PNG (n = 11) alone and in combination with SE Asia (n = 29, combined n = 40). The PNG population has the smallest Rh estimate but a large population recombination rate estimate; however, the rate estimate has a large confidence interval, due in part to the small sample size. When the PNG and SE Asian populations were combined, estimates of the effective population recombination rate (1,667) were more similar to SE Asia estimates alone (Table 2) with narrower confidence intervals (877–1,916). These estimates point to extreme differences in the effective population recombination rates, which can result from differences in transmission rates, inbreeding, and effective population sizes, an observation that is itself unparalleled.
Table 2 Population Recombination Rate Variation on Chromosome 3 among P. falciparum Populations
In addition to measuring differences in recombination events among populations, the chromosome-wide data are also informative for evaluating the influence of genetic drift and recombination on genetic variation. In partially out-crossing species, population genetic parameters can be scaled by the inbreeding coefficients, f [21,29]. In P. falciparum, inbreeding coefficients have been shown to be directly correlated with rates of transmission and/or frequencies of multiple infection [30]. Here we compared relative values of drift, Ne/(1 + f) (see Materials and Methods), estimated from nucleotide variation in a Bayesian fashion, through the use of a beta-binomial model [23,31], and Ne(1 − f) from our population recombination rate estimates (Table 2). The beta-binomial model measures the difference of each population from a hypothetical average, or ancestral, population by a parameter, cj, for each population, j [31]. These parameters can be thought of as a generalization of FST [23]. We can show relative estimates (to the American population) only, not the actual estimates of Ne, as inferences would be affected by SNP ascertainment bias in an unknown manner. All populations are similar in terms of relative rates of drift. For example, SE Asia and PNG parasites have similar population mutation rates and therefore similar rates of genetic drift, but the SE Asian population has a significantly reduced recombination rate relative to the African population, indicating a higher rate of inbreeding. Although the African population has a comparable rate of genetic drift with that of American isolates, it has a recombination rate at least two orders of magnitude larger. These results show that differences in transmission frequencies and rates of multiple infections among populations, rather than effective population sizes, are most important in shaping haplotype variation among populations and in determining local levels of LD.
Recombination Hotspots Conserved at the Middle and at the Ends of the Chromosome
We next examined whether recombination rates were uniform along the chromosome, because variation in recombination rates will dramatically affect the effectiveness of association studies [32]. The majority of recombination events cluster near the chromosome ends and in the middle of the chromosome (Figure 2A and 2B). Nonparametric estimates revealed many recombination events, as well as recombination hotspots, among African parasites (Figure 2A). Similar recombination hotspots were also found in the remaining parasite populations, except American, where the hotspot in the middle of the chromosome is absent. Parametric methods, based on coalescent models, also detected significant recombination rate variation (SE Asian, p < 0.001; American, p < 0.001; PNG, p < 0.05) in all populations except African, for which the high levels of historic recombination invalidated the test. Figure 2C shows the recombination map along the chromosome for the four populations, as estimated by the Reversible Jump Markov Chain Monte Carlo (RJMCMC) method [19]. Even though the sample size for PNG is relatively small for these types of rate estimates [27,33] and the confidence intervals for the overall PNG rate is large, the inferred location of recombination hotspots using the two approaches generally concur (Table 2).
Figure 2 The Distribution of Detectable Recombination Events on Chromosome 3 of P. falciparum
In (A) and (B), each panel shows, for two populations, a minimum number of recombination events (assuming an infinite-sites model) between each pair of segregating sites, scaled by physical distance to identify regions of high and low recombination.
(A) African (upper) and American (lower) populations.
(B) SE Asian (upper) and PNG (lower) populations. The color bar unit is recombination event/kb.
(C) Estimates of population recombination rate variation for African (red line), SE Asian (blue), American (black), and PNG (purple) samples using the RJMCMC method with a jump penalty of five. Shown on top are the locations of genes on the plus (top) and minus (below) strands, with known cell-surface genes in red [34].
Although the overall population recombination rate is highly variable among populations, the chromosomal locations of major recombination hotspots were conserved. Subtelomeric regions in P. falciparum clearly exhibit elevated crossing over, similar to those observed in human males (but not females) [32]. The conservation is likely due, in part, to the shared evolutionary ancestry of P. falciparum populations. Additionally, these regions contain a high density of genes such as var, rifin, and stevor, whose products are implicated in cell-surface interactions [34] and are consequently under strong immune-mediated diversifying selection (as demonstrated by the high rate of amino acid evolution) [35]. Increases in recombination can be indirectly selected, either because recombination generates genetic variation at these genes or because it allows sites that are targeted by selection to freely evolve without interfering with each other [36]. Alternatively, these gene families themselves may contribute to increased recombination through concerted evolutionary processes, such as unequal rates of cross-over. Regardless, these observations suggest that elevated recombination rates may play a significant role in generating multiple haplotypes at genes important for P. falciparum's evasion of host immunity.
Variation in LD and Haplotype Blocks among Parasite Populations
Population genetic models predict that the extent of LD is inversely proportional to the population recombination rate [37,38]. As expected, we observed less LD in African populations. Indeed, LD decays with increasing physical distance between pairs of segregating sites in all populations, but more slowly in SE Asian and American populations relative to parasites from Africa, where LD decays rapidly over very short distances (Figure 3A).
Figure 3 LD and Haplotype Blocks across Chromosome 3 of P. falciparum
(A) LD decays with increasing distances between variable sites along Chromosome 3 in populations from Africa, SE Asia, S America, and PNG. The LD index r2 (square of correlation coefficient) were calculated considering all pair-wise values for SNPs with minor allele frequency > 5%. Parasite isolates K1, Camp, T2/C6, 105/7, 3D7, HB3, Haiti, JAV, ECP, 569, and 1905 are excluded from the LD analyses because these parasites were placed in clusters different from locations they were isolated from or have different genetic backgrounds.
(B) Haplotype blocks defined as regions where all pairs of sites have D′ ≥ 0.8. Values below each block are the number of htSNPs required to capture all haplotype variants when haplotype blocks are characterized by complete association among variants (r2 = 1). Bars on the top are locations of assayed SNPs. Triangles denote a site in the middle of a recombination hotspot, and the width of the triangles represents the region hotspot spans.
Critical for association studies is the identification of haplotype blocks and the minimal set of haplotype tagging SNPs (htSNPs) required to capture haplotype variation in a population sufficiently, which will reduce cost and effort. Haplotype blocks of various sizes were identified for the four populations (Figure 3B). The African population, with its high inferred population recombination rate, clearly had the smallest average block length (11.2 kb) and the greatest number of blocks (n = 46), whereas the average block size for PNG was 56 kb, with only 11 blocks defined. Again, the relatively large blocks in PNG could be due to a small sample size and/or sampling from a small isolated area or population. Relatively high inbreeding frequency (0.915) [12] could also contribute to the large haplotype blocks. Among African populations, the number of htSNPs required to capture haplotype variation sufficiently was 53% of the SNPs polymorphic in African parasites. Whereas major hotspots are clearly breaking up block structure in all populations near the chromosome ends (Figure 2B), other recombination events are also disrupting LD and haplotype blocks (Figures 2 and 3) in different regions.
A primary objective in studying population recombination rate variation and haplotype maps in P. falciparum is to facilitate identification of genes responsible for important parasite traits, such as drug resistance and virulence. There is currently tremendous interest in using LD to map human disease genes. The present results suggest that LD mapping may, in some circumstances, be more effective in studying partially selfing species such as P. falciparum than out-crossing species, such as humans or Drosophila, because LD can persist over extensive genomic regions. The overall estimated population recombination rates clearly vary among populations, primarily due to different inbreeding and transmission rates, but the chromosomal locations of major hotspots are conserved. Population structure among sub-Saharan African parasites appears to be minimal because of high estimated recombination rates; however, very high marker densities may be required for mapping even newly arisen traits, because traces of LD between loci will be lost quickly. The presence of chromosomal segments with low population recombination rates in SE Asian and American parasites suggests it is possible to conduct genome-wide association mapping for certain phenotypes in these geographic locations, using relatively low densities of marker loci. This approach is likely to be particularly effective for genes involved in drug resistance, since the mutations involved have occurred recently, allowing little time for LD between marker and trait loci to be broken down. However, as shown in this study, recombination rates are not uniform across the chromosome, and, therefore, the location of the genes of interest will be highly relevant. Studying these hotspots will ultimately illuminate the as yet mysterious factors that direct the location and frequency of recombination in this and other species. The presence of LD in at least some populations, the recent appearance of mutations conferring drug resistances, and the use of high-density genetic maps make it practical to conduct genome-wide association studies in this relatively small genome.
Materials and Methods
DNA sequences and SNP ascertainment
Predicted coding sequences of the 3D7 parasite were downloaded from PlasmoDB (http://www.plasmodb.org/). DNA sequences covering SNP sites identified from five isolates in our previous study [15] were amplified and sequenced from additional 94 isolates collected from different regions of the world (Figure S1). After exclusion of those SNPs that were difficult to amplify from all isolates due to high AT content or other technical reasons, 183 SNPs were obtained for analysis (see Figure 3B for physical locations on the chromosome). DNA sequences were trimmed and aligned using Phred/Phap (http://www.phrap.org/) and Sequencher 3.1 (Gene Codes, Ann Arbor, Michigan, United States) to identify SNPs. All potential SNPs and discrepancies were verified by visually inspecting chromatogram traces.
Population structure analysis
Population structures were analyzed using the Structure 2 package [24]. We ran the program ten times at each K value (K = 2–8) with 50,000 burn-ins and 100,000 iterations. For the SNPs from Chromosome 3, both admixture and linkage models were used (admixture only for SNPs from the putative transporters). FST values were calculated in Arlequin [25]. Inferences of effective population size assuming a beta-binomial distribution were estimated in a Bayesian fashion as in Marchini et al. [23] using the R package popgen (http://www.stats.ox.ac.uk/~marchini/software.html).
Inference of population recombination rate
Nonparametric estimates of the number of recombination events (Rh) were calculated using the methodology of Myers and Griffiths [28]. Model-based parametric estimates of the recombination rate were calculated using the LDhat programs, pairwise and interval (http://www.stats.ox.ac.uk/~mcvean/LDhat/LDhat1.0/LDhat1.0.html). The method extends the composite likelihood approach of Hudson [33], to allow for recurrent mutation [27], and adopts a Bayesian implementation that uses a RJMCMC scheme to fit a recombination map composed of a series of intervals of constant rate [19]. Under simple models of demographic history, the key quantity in determining the extent of LD between alleles at linked loci is the product of the effective population size, Ne, which is inversely related to the rate of genetic drift, the per-generation rate of recombination, r, and for P. falciparum, the rate of outcrossing, 1 − f, where f is the inbreeding coefficient [9,21]. Without an absolute estimate of the out-crossing rate and Ne in a population, we can only estimate the compound parameter of the recombination rate, 2N
e
r(1 − f). Such methods have also been used to develop an “effective recombination rate” for malaria parasites [39]. It has also been shown that the coalescent with partial selfing looks like the standard coalescent, but with rates of coalescence (1 + f) times faster [21,29]. If time is rescaled in units of 2N/(1 + f) generations, the coalescence process is identical to the standard coalescent. Similarly, ancestral recombination graphs with inbreeding have rates of coalescence that are (1 + f) times faster and rates of recombination that are (1 − f) times slower than in a completely out-crossing situation [29]. Again, if time is rescaled in units of 2N/(1 + f) generations, the process looks like the out-crossing case but with a recombination rate that is (1 − f)/(1 + f) times slower than in the out-crossing case. In the case of inbreeding, 4Nμ is therefore replaced by 4Nμ/(1 + f), and 4Nr by 4Nr(1 − f).
The compound population recombination rate parameter can be estimated from population genetic data using coalescent methods adapted specifically to account for the possibility of recurrent or back mutation and for an AT-rich genome such as that of P. falciparum [27,33,40]. By estimating this parameter, we can evaluate how historical population size and out-crossing rate affect LD in populations. These approaches have been shown to be robust in evaluating population recombination even when the underlying nucleotide variation is variable along a chromosome [27] and with respect to SNP ascertainment bias [19]. Because of the non-independence introduced by the composite likelihood, the variance for RJMCMC is unknown, and we therefore report the marginal posterior mean recombination rate (Figure 2C) for each SNP interval as a point estimate while performing the statistical tests (below) to obtain confidence intervals for local recombination rate variation.
To test the hypothesis of constant recombination rate in each population, we used coalescent simulations to generate the distribution of a test statistic, T,
which is the sum, over all pairs of sites Xij, of the difference in log likelihood between the marginal maximum likelihood estimate of the recombination parameter ρ and the global maximum likelihood estimate (assuming a constant recombination rate), scaled by the ratio of the physical distance between the sites, dij, to the total length of the sequence analyzed, L. Coalescent simulations (10,000) were carried out as described previously [19]. Because of the high level of recombination estimated for the African population, coalescent simulations were not feasible.
A parametric bootstrap was performed to obtain confidence intervals for the rate estimates for each population (Table 2). Simulations were carried out as above conditioning on the estimates of per-chromosome recombination rate, population mutation rate, and frequency spectrum of variation at each SNP for each population. The simulations were sorted for the top and bottom 97.5 percentiles.
Estimates of LD and haplotype blocks
Standard LD estimates r2 and D′ were calculated for all pairs of sites, and significance was assessed by randomizing the positions of segregating sites 10,000 times [27]. Plots of LD decline with distance were constructed using DnaSP 4.0 [41]. Haplotype blocks were obtained using methods described [42]. We used two different LD classifications (D′ and r2) for haplotype blocks and htSNP identification, and the major differences in numbers of blocks and htSNPs between populations remained the same regardless: D′ is a measure of LD normalized by the largest LD value at the given allele frequencies, and r2 is the square of correlation coefficient of LD normalized by the variance in allele frequencies at the two loci. Haplotype blocks of various sizes were defined here as genomic regions with D′ greater than or equal to 0.8 [42].
Supporting Information
Figure S1 Worldwide Distribution of P. falciparum Isolates Used in this Study
Parasites are color-coded as resistant to chloroquine (red), sensitive (black), or not tested (green).
(79 KB PDF).
Click here for additional data file.
Table S1 Putative Transporter Genes and Their Chromosomal Locations
(29 KB DOC).
Click here for additional data file.
We thank Drs. Tim Anderson and Matthew Hahn for critical reading of the manuscript, Ms. Brenda Rae Marshall for editorial assistance, and the Wellcome Trust for financial support for PA.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. XS conceived and designed the experiments. JM and JD performed the experiments. JM, PA, KMM, GATM, and XS analyzed the data. PA, DAJ, and XS wrote the paper.
Citation: Mu J, Awadalla P, Duan J, McGee KM, Joy DA, et al. (2005) Recombination hotspots and population structure in Plasmodium falciparum. PLoS Biol 3(10): e335.
Abbreviations
htSNPhaplotype tagging SNP
kbkilobase
LDlinkage disequilibrium
PNGPapua New Guinea
RJMCMCReversible Jump Markov Chain Monte Carlo
SESoutheast
SNPsingle nucleotide polymorphism
==== Refs
References
WHO WHO Expert Committee on Malaria World Health Organ Tech Rep Ser 2000 892 1 74
Tishkoff SA Varkonyi R Cahinhinan N Abbes S Argyropoulos G Haplotype diversity and linkage disequilibrium at human G6PD: Recent origin of alleles that confer malarial resistance Science 2001 293 455 462 11423617
Ohashi J Naka I Patarapotikul J Hananantachai H Brittenham G Extended linkage disequilibrium surrounding the hemoglobin E variant due to malarial selection Am J Hum Genet 2004 74 1198 1208 15114532
Conway DJ Cavanagh DR Tanabe K Roper C Mikes ZS A principal target of human immunity to malaria identified by molecular population genetic and immunological analyses Nat Med 2000 6 689 692 10835687
Wootton JC Feng X Ferdig MT Cooper RA Mu J Genetic diversity and chloroquine selective sweeps in Plasmodium falciparum
Nature 2002 418 320 323 12124623
Nair S Williams JT Brockman A Paiphun L Mayxay M A selective sweep driven by pyrimethamine treatment in southeast asian malaria parasites Mol Biol Evol 2003 20 1526 1536 12832643
Roper C Pearce R Bredenkamp B Gumede J Drakeley C Antifolate antimalarial resistance in southeast Africa: A population-based analysis Lancet 2003 361 1174 1181 12686039
Anderson TJ Mapping drug resistance genes in Plasmodium falciparum by genome-wide association Curr Drug Targets Infect Disord 2004 4 65 78 15032635
Hill WG Babiker HA Ranford-Cartwright LC Walliker D Estimation of inbreeding coefficients from genotypic data on multiple alleles, and application to estimation of clonality in malaria parasites Genet Res 1995 65 53 61 7750746
Paul RE Packer MJ Walmsley M Lagog M Ranford-Cartwright LC Mating patterns in malaria parasite populations of Papua New Guinea Science 1995 269 1709 1711 7569897
Babiker HA Lines J Hill WG Walliker D Population structure of Plasmodium falciparum in villages with different malaria endemicity in east Africa Am J Trop Med Hyg 1997 56 141 147 9080871
Conway DJ Roper C Oduola AM Arnot DE Kremsner PG High recombination rate in natural populations of Plasmodium falciparum
Proc Natl Acad Sci U S A 1999 96 4506 4511 10200292
Anderson TJ Haubold B Williams JT Estrada-Franco JG Richardson L Microsatellite markers reveal a spectrum of population structures in the malaria parasite Plasmodium falciparum
Mol Biol Evol 2000 17 1467 1482 11018154
Su X Ferdig MT Huang Y Huynh CQ Liu A A genetic map and recombination parameters of the human malaria parasite Plasmodium falciparum
Science 1999 286 1351 1353 10558988
Mu J Duan J Makova K Joy DA Huynh CQ Chromosome-wide SNPs reveal an ancient origin for Plasmodium falciparum
Nature 2002 418 323 326 12124624
Chakravarti A Buetow KH Antonarakis SE Waber PG Boehm CD Nonuniform recombination within the human beta-globin gene cluster Am J Hum Genet 1984 36 1239 1258 6097112
Jeffreys AJ Kauppi L Neumann R Intensely punctate meiotic recombination in the class II region of the major histocompatibility complex Nat Genet 2001 29 217 222 11586303
Petes TD Meiotic recombination hot spots and cold spots Nat Rev Genet 2001 2 360 369 11331902
McVean GA Myers SR Hunt S Deloukas P Bentley DR The fine-scale structure of recombination rate variation in the human genome Science 2004 304 581 584 15105499
Jeffreys AJ Neumann R Reciprocal crossover asymmetry and meiotic drive in a human recombination hot spot Nat Genet 2002 31 267 271 12089523
Pollak E On the theory of partially inbreeding finite populations. I. Partial selfing Genetics 1987 117 353 360 3666446
Pritchard JK Przeworski M Linkage disequilibrium in humans: Models and data Am J Hum Genet 2001 69 1 14 11410837
Marchini J Cardon LR Phillips MS Donnelly P The effects of human population structure on large genetic association studies Nat Genet 2004 36 512 517 15052271
Pritchard JK Stephens M Donnelly P Inference of population structure using multilocus genotype data Genetics 2000 155 945 959 10835412
Schneider S Roessli D Excoffier L Arlequin: A software for population genetics data analysis. Version 2.000. Genetics and Biometry Lab, Department of Anthropology, University of Geneva 2000 Available: http://lgb.unige.ch/arlequin/software/2.000/doc/faq/faqlist.htm . Accessed 4 August 2005
Mu J Ferdig MT Feng X Joy DA Duan J Multiple transporters associated with malaria parasite responses to chloroquine and quinine Mol Microbiol 2003 49 977 989 12890022
McVean G Awadalla P Fearnhead P A coalescent-based method for detecting and estimating recombination from gene sequences Genetics 2002 160 1231 1241 11901136
Myers SR Griffiths RC Bounds on the minimum number of recombination events in a sample history Genetics 2003 163 375 394 12586723
Nordborg M Linkage disequilibrium, gene trees and selfing: An ancestral recombination graph with partial self-fertilization Genetics 2000 154 923 929 10655241
Hill WG Babiker HA Estimation of numbers of malaria clones in blood samples Proc R Soc Lond B Biol Sci 1995 262 249 257
Nicholson G Smith AV Jónsson F Gústafsson Ó Stefánsson K Assessing population differentiation and isolation from single-nucleotide polymorphism data J Royal Stat Soc Series B 2002 64 695 716
Kong A Gudbjartsson DF Sainz J Jonsdottir GM Gudjonsson SA A high-resolution recombination map of the human genome Nat Genet 2002 31 241 247 12053178
Hudson RR Two-locus sampling distributions and their application Genetics 2001 159 1805 1817 11779816
Bowman S Lawson D Basham D Brown D Chillingworth T The complete nucleotide sequence of Chromosome 3 of Plasmodium falciparum
Nature 1999 400 532 538 10448855
Volkman SK Hartl DL Wirth DF Nielsen KM Choi M Excess polymorphisms in genes for membrane proteins in Plasmodium falciparum
Science 2002 298 216 218 12364807
Otto SP Barton NH The evolution of recombination: Removing the limits to natural selection Genetics 1997 147 879 906 9335621
Hill WG Disequilibrium among several linked neutral genes in finite population. II. Variances and covariances of disequilibria Theor Popul Biol 1974 6 184 198 4445974
Weir BS Hill WG Nonuniform recombination within the human beta-globin gene cluster Am J Hum Genet 1986 38 776 781 3013006
Dye C Williams BG Multigenic drug resistance among inbred malaria parasites Proc Biol Sci 1997 264 61 67 9061961
Awadalla P The evolutionary genomics of pathogen recombination Nat Rev Genet 2003 4 50 60 12509753
Rozas J Sanchez-DelBarrio JC Messeguer X Rozas R DnaSP, DNA polymorphism analyses by the coalescent and other methods Bioinformatics 2003 19 2496 2497 14668244
Zhang K Deng M Chen T Waterman MS Sun F A dynamic programming algorithm for haplotype block partitioning Proc Natl Acad Sci U S A 2002 99 7335 7339 12032283
|
16144426
|
PMC1201364
|
CC BY
|
2021-01-05 08:21:27
|
no
|
PLoS Biol. 2005 Oct 13; 3(10):e335
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030335
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030361SynopsisGenetics/Genomics/Gene TherapyClues to the Evolution of the Malarial Chromosome Synopsis10 2005 13 9 2005 13 9 2005 3 10 e361Copyright: © 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.
Recombination Hotspots and Population Structure in Plasmodium falciparum
==== Body
Understanding the recombination patterns across a chromosome—determining the positions and frequency of genetic exchanges between homologous chromosomes—is crucial for understanding and tracking inheritance of traits. Mapping genes that affect parasites' traits, such as responses to various antimalarial agents, is possible because, during meiosis, homologous chromosomes line up and may exchange segments. Genes—or any polymorphic bits of DNA—that are close together tend to remain linked during this process, while those far apart tend to become separated. Identifying and following polymorphic markers through multiple generations is a key technique for genetic mapping.
For Plasmodium falciparum, the microbe that causes malaria, chromosomal mapping is necessary for understanding the evolution of the parasite and development of drug resistance, but multiple factors make this a complex task. In this issue, Jianbing Mu and colleagues use single nucleotide polymorphisms (SNPs) to evaluate some of these factors, and set the stage for further mapping of this important parasite's genome.
The authors began by locating 183 SNPs spaced across Chromosome 3 in 99 P. falciparum populations from throughout the world. Not all SNPs were found in all populations, indicating a more recent evolutionary origin for some SNPs; these differences were then used to track evolution and migration in parasites. Statistical analysis of the SNPs allowed the populations to be parsed into five groups, largely corresponding to continents. More refined analysis of the SNPs revealed possible migratory history, including a recent migration of an African variety to coastal South America.
Plasmodium falciparum, the microbe that causes malaria, infects red blood cells. By analyzing different populations of the pathogen from around the world, researchers found clues to its genome structure that will be important for identifying genes that contribute to drug resistance and virulence
Mu and colleagues also showed for the first time that the historical rate of recombination varies widely—over 20-fold—among different populations. A large part of the variation is due to a combination of the frequency of infections with multiple parasite strains (because sexual recombination occurs only within an infected mosquito) and the degree of inbreeding within a parasite population. Inbreeding tends to lower the extent of detectable recombination events, while multiple infections by different strains increase it.
Despite the wide differences in recombination rates, all populations had a similar clustering of recombination “hot spots” at the middle and ends of the chromosome. Recombination is most likely to occur at these spots, and the similar localization reflects either the common evolutionary history of all the populations or localization of crossover events to particular genomic regions.
The authors compared their results from population structure analysis with those using SNPs from genes that might be influenced by drug pressure. Their results showed that misleading inferences about the parasite population structures could be derived using information from genes that are potentially under drug selection.
These results are important because they provide information on the multiple complex factors that must be considered in understanding the genomic structure of P. falciparum, which is critical for identifying genes that contribute to phenotypes such as drug resistance and virulence. Reseachers conducting future mapping studies will be able to draw on the important findings and caveats revealed by this work to refine their own methods and interpret their results. —Richard Robinson
|
0
|
PMC1201365
|
CC BY
|
2021-01-05 08:21:27
|
no
|
PLoS Biol. 2005 Oct 13; 3(10):e361
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030361
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1614641410.1371/journal.pmed.0020252Research ArticleObstetrics/GynecologyObstetricsPregnancyPredicting Cesarean Section and Uterine Rupture among Women Attempting Vaginal Birth after Prior Cesarean Section Predicting Outcome of VBACSmith Gordon C. S
1
*White Ian R
2
Pell Jill P
3
Dobbie Richard
4
1Department of Obstetrics and Gynaecology, Cambridge University, Cambridge, United Kingdom,2Medical Research Council Biostatistics Unit, Institute of Public Health, Cambridge, United Kingdom,3Department of Public Health, Greater Glasgow NHS Board, Glasgow, United Kingdom,4Information and Statistics Division, Common Services Agency, Edinburgh, United KingdomFisk Nicholas M. Academic EditorImperial College LondonUnited Kingdom*To whom correspondence should be addressed. E-mail: [email protected]
Competing Interests: The authors have declared that no competing interests exist.
Author Contributions: GCSS designed the study. RD collected the data. IRW developed methods. GCSS and JPP analyzed the data. GCSS, IRW, JPP, and RD contributed to writing the paper.
9 2005 13 9 2005 2 9 e25226 1 2005 13 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.
Does the Maxim "Once a Caesarean, Always a Caesarean" Still Hold True?
A Tool to Estimate the Risks of Repeat Cesarean Section
Background
There is currently no validated method for antepartum prediction of the risk of failed vaginal birth after cesarean section and no information on the relationship between the risk of emergency cesarean delivery and the risk of uterine rupture.
Methods and Findings
We linked a national maternity hospital discharge database and a national registry of perinatal deaths. We studied 23,286 women with one prior cesarean delivery who attempted vaginal birth at or after 40-wk gestation. The population was randomly split into model development and validation groups. The factors associated with emergency cesarean section were maternal age (adjusted odds ratio [OR] = 1.22 per 5-y increase, 95% confidence interval [CI]: 1.16 to 1.28), maternal height (adjusted OR = 0.75 per 5-cm increase, 95% CI: 0.73 to 0.78), male fetus (adjusted OR = 1.18, 95% CI: 1.08 to 1.29), no previous vaginal birth (adjusted OR = 5.08, 95% CI: 4.52 to 5.72), prostaglandin induction of labor (adjusted OR = 1.42, 95% CI: 1.26 to 1.60), and birth at 41-wk (adjusted OR = 1.30, 95% CI: 1.18 to 1.42) or 42-wk (adjusted OR = 1.38, 95% CI: 1.17 to 1.62) gestation compared with 40-wk. In the validation group, 36% of the women had a low predicted risk of caesarean section (<20%) and 16.5% of women had a high predicted risk (>40%); 10.9% and 47.7% of these women, respectively, actually had deliveries by caesarean section. The predicted risk of caesarean section was also associated with the risk of all uterine rupture (OR for a 5% increase in predicted risk = 1.22, 95% CI: 1.14 to 1.31) and uterine rupture associated with perinatal death (OR for a 5% increase in predicted risk = 1.32, 95% CI: 1.02 to 1.73). The observed incidence of uterine rupture was 2.0 per 1,000 among women at low risk of cesarean section and 9.1 per 1,000 among those at high risk (relative risk = 4.5, 95% CI: 2.6 to 8.1). We present the model in a simple-to-use format.
Conclusions
We present, to our knowledge, the first validated model for antepartum prediction of the risk of failed vaginal birth after prior cesarean section. Women at increased risk of emergency caesarean section are also at increased risk of uterine rupture, including catastrophic rupture leading to perinatal death.
By studying a large group of Scottish women giving birth, Gordon Smith and colleagues developed and validated a tool to assess the risk of failed vaginal birth after prior caesarean section.
==== Body
Introduction
Encouraging women with a prior cesarean delivery to attempt vaginal birth in subsequent pregnancies is a strategy that has been employed to address rising rates of cesarean delivery. However, a series of retrospective studies published in the last five to ten years have indicated an increased risk of serious adverse outcomes among women who attempted vaginal birth compared with those who had a planned repeat cesarean delivery [1–3]. A recent large-scale prospective study has shown that among women with a prior cesarean delivery, the rates of maternal complications are highest among women who attempt vaginal birth and fail (14.1%), intermediate among women who have a planned cesarean delivery (3.6%), and lowest among women who attempt vaginal birth and succeed (2.4%) [4]. Therefore, the balance of risks and benefits of trial of labor versus planned repeat cesarean section itself depends on the risk of emergency cesarean section should labor be attempted.
Many studies have addressed methods for identifying women at low and high risk of failure of an attempted vaginal birth after a prior cesarean. A recent systematic review reported that only two of the six available tools had been validated [5]. Both of these incorporated data that would only be available when a women presented in labor, such as the results of electronic fetal monitoring and cervical dilatation on admission [6,7]. Currently, therefore, there is no validated antepartum tool to predict the risk of a failed attempt at vaginal birth among women with a prior cesarean delivery. Moreover, there are no data on whether women at increased risk of cesarean section are also at increased risk of uterine rupture. We sought to develop a simple, validated model to predict the risk of emergency cesarean section among women attempting vaginal birth and to determine whether women at increased risk of cesarean section were also at increased risk of uterine rupture, including catastrophic rupture leading to death of the infant.
Methods
Population
The Scottish Morbidity Record (SMR2) collects information on clinical and demographic characteristics and outcomes for all patients discharged from Scottish maternity hospitals. The register is subjected to regular quality assurance checks and has been more than 99% complete since the late 1970s [8]. A quality assurance analysis compared 1,414 records in 1996–1997 with the clinical notes. This analysis demonstrated that the register was free from significant errors in more than 98% of records in all the specific fields used in the present analysis, with the exception of postcode (94.0%), height (96.2%), estimated gestation (94.4%), and method of induction of labor (93.6%). The previous cesarean section field was 99.7% accurate. International classification of disease (ICD) diagnostic codes were found to be 80%–90% accurate for the first four diagnoses and 70%–80% accurate for the remainder [9]. SMR2 records were linked to records from the Scottish Stillbirth and Neonatal Death Enquiry. This national register has routinely classified all perinatal deaths in Scotland since 1983. It is virtually 100% complete and has been described in detail elsewhere [10]. The predictors of cesarean section employed in the current study were those that were recorded in the SMR2 and which had been identified in previous studies as possible risk factors for emergency cesarean delivery.
Study Group
The population was drawn from all term singleton births to women with one prior cesarean section in Scotland between 1985 and 2001, inclusive. The exclusion criteria for the study group were preterm birth, perinatal deaths due to congenital anomaly, antepartum stillbirth due to any cause, deliveries by planned cesarean section, and women documented as being primigravid despite also being documented as having had a prior cesarean delivery. The primary analysis was confined to women who delivered at or after 40-wk gestation.
Definitions
Emergency cesarean section was defined as any non-planned cesarean delivery. Maternal age was defined as the age of the mother at the time of birth. Maternal height was measured in centimeters, and the value used was that documented in each woman's clinical record. Gestational age at birth was defined as completed wk of gestation on the basis of the estimated date of delivery in each woman's clinical record. Gestational age has been confirmed by ultrasound in the first half of pregnancy in more than 95% of women in the United Kingdom since the early 1990s [11]. Hospital throughput was defined as the total number of births recorded in the SMR2 database for the given hospital over the given year and was categorized into above or below the national median (3,000 births).9
Perinatal deaths were classified on the basis of data from the Scottish Stillbirth and Neonatal Death Enquiry [10]. Death caused by congenital anomaly was defined as any structural or genetic defect incompatible with life or potentially treatable but causing death. The registry subclassifies stillbirths into antepartum (deaths before the onset of labor) and intrapartum (deaths during labor). A death was taken to be due to intrapartum uterine rupture if it was an intrapartum stillbirth or neonatal death, when the cause of death was documented as intrapartum anoxia, when the obstetric cause of death was coded as “mechanical” under the modified Wigglesworth classification [12], and when the ICD9 code for intrapartum uterine rupture (665.1) was listed as a specific diagnosis. Intrapartum uterine ruptures not resulting in perinatal death were identified using ICD9 and ICD10 diagnostic codes 665.1 and O711, respectively, from the diagnostic fields in the SMR2 record related to hospital discharge following delivery.
Statistical Analyses
Continuous variables were summarized by the median and interquartile range (IQR), and comparisons between groups were performed using the Mann-Whitney U test. Univariate comparisons of categorical data were performed using Fisher's exact test. The p values for all hypothesis tests were two-sided. The risk of adverse outcomes was modeled using multivariate logistic regression [11]. First order interactions were assessed using the likelihood ratio test and significance assumed at p < 0.05 after correction for the number of comparisons using the Bonferroni method. The goodness of fit of logistic regression models was assessed using the Hosmer and Lemeshow test. Assessment of linearity of age and height in logistic models was performed using fractional polynomials. Cases with extreme values of age or height (≤0.1 percentile and ≥99.9 percentile) were excluded. Out-of-sample validation of the model was performed by dividing the cohort into model development and model validation groups. Models were constructed for the development group and the predicted numbers of cesarean sections were related to the observed number of events in the validation group when categorized into deciles of predicted probability. Selection of model development and validation groups was initially random and the process was then repeated selecting the groups on specific characteristics (hospital throughput, deprivation category and year of delivery). Random allocation into two groups was performed using a pseudo-random number. The predictive ability of models was assessed by the area under the receiver operating characteristic (ROC) curve, and curves were compared using the algorithm described by De Long et al [12]. The final logistic regression model fitted to the entire cohort was expressed as adjusted log likelihood ratios (ALLRs) using a modification of our recently described method [13] (see Supporting Information for details). Logistic regression analysis of the risk of perinatal death was performed using exact logistic regression due to the rarity of the event. All statistical analyses were performed using the Stata software package version 8.2 (Stata Corporation, College Station, Texas, United States), except exact logistic regression which was performed using LogExact version 5.0.1 (Cytel Software Corporation, Cambridge, Massachusetts, United States).
Results
Between 1985 and 2001, 68,380 women delivered who had one prior cesarean delivery. We excluded 150 (0.2%) births outside the range 24 to 43 wk, 4,700 (6.9%) preterm births, 366 (0.5%) antepartum stillbirths, 21,677 (31.7%) women delivered by planned cesarean section, 124 (0.2%) women whose infant was a perinatal death attributed to a congenital abnormality, and 76 (0.1%) women documented as being primigravid. A total of 25,836 (37.8%) women had one or more of these exclusions, leaving 42,544 (62.2%) women. Among the 25,964 (61.0%) women who delivered at or after 40-wk gestation, 2,585 (10.0%) had a missing value for height, one (<0.1%) had a missing value for age, 51 (0.2%) had an extreme value of height, and 41 (0.2%) had an extreme value of maternal age, leaving 23,286 women eligible for study. These women were randomly allocated to a model development or model validation group, and the demographics and basic outcome data for the cohort were tabulated (Table 1). Women who had previously had a vaginal birth were older than those with no previous vaginal birth (median IQR: 30 [27–34] versus 29 [25–32], respectively, p < 0.001).
Table 1 Characteristics of Population by Allocation to Development or Validation Group
In univariate and multivariate analysis in the model development group, all factors were significantly associated with the risk of emergency cesarean section except induction of labor using a means other than prostaglandin (Table 2). The area under the ROC curve in the development group was 0.706, which was significantly greater than for any of the individual predictors (all p < 0.001). There were no statistically significant first order interactions between the predictors. When the model was applied to the validation group, the area under the ROC curve was 0.708 (Table 3). The observed proportion of emergency cesarean deliveries and the proportion predicted by the multivariate model derived from the model development group were similar (Figure 1A). In the validation group, 36% of the women had a low predicted risk of caesarean section (<20%) and 16.5% of women had a high predicted risk (>40%); 10.9% and 47.7% of these women, respectively, actually had deliveries by caesarean section.
Figure 1 Observed and Expected Proportion of Cesarean Deliveries in the Model Validation Group by Decile of Predicted Probability
The white bars indicate the observed proportion and the black bars indicate the expected proportion of cesarean deliveries, based on estimates derived from logistic regression model fitted to the development group. Different graphs represent different procedures for selecting development and validation groups: (A) random selection, (B) selected on hospital throughput, (C) selected on deprivation category (Carstairs score), and (D) selected on year of delivery. Area under the ROC curve for each model is listed in Table 3.
Table 2 Univariate and Multivariate Analysis of Predictors of Emergency Cesarean Section in the Model Development Group (n = 11,643)
Table 3 Assessment of the Modeling Approach in Development and Validation Samples
The process of model development and validation was then repeated with nonrandom selection of the development and validation samples. Three nonrandom procedures for selection were evaluated, namely, hospital throughput (<3,000 births per year and ≥3,000 births per year), deprivation category (Carstairs category <5 and Carstairs category ≥5), and year of birth (1985–1992 and 1993–2001). The area under the ROC curve was similar when the development and validation samples were compared, (Table 3) and when the data were plotted according to the predicted risk, the expected and observed number of cesarean deliveries were similar in the validation samples (Figure 1B–1D).
A logistic regression model was then fitted for the whole cohort. The area under the ROC curve was 0.707 and the global goodness-of-fit test showed no evidence of poor fit (p = 0.95). The output was converted to ALLRs (Table 4) using a modification of our previously described method [13]. The calculation of a summary ALLR for a series of maternal characteristics is illustrated in the box. The summary ALLR could also be used in combination with a published nomogram to generate a predicted probability [14]. Assuming a prior probability of emergency cesarean delivery of 26%, a summary ALLR of 0.71 or less was associated with a less than 20% chance of emergency cesarean section, and a summary ALLR of 1.91 or more was associated with a greater than 40% chance of emergency cesarean section.
Table 4 ALLRs for Maternal Characteristics and Fetal Sex Derived from Logistic Regression Model Fitted for the Whole Population
The probability of cesarean section was calculated for each woman in the cohort using the multivariate model. We then analyzed the risk of uterine rupture in relation to the predicted risk of emergency cesarean section. The predicted probability of cesarean section was also associated with the risk of all uterine rupture (Figure 2; odds ratio for a 5% increase in predicted risk = 1.22, 95% confidence interval [CI]: 1.14 to 1.31) and uterine rupture associated with perinatal death (odds ratio for a 5% increase in predicted risk = 1.32, 95% CI: 1.02 to 1.73). Among women with a predicted cesarean section risk of less than 20%, the incidence of uterine rupture was 2.0 (95% CI: 1.1 to 3.2) per 1,000, and among women with a cesarean section risk of greater than 40%, the incidence of uterine rupture was 9.1 (95% CI: 6.4 to 12.6) per 1,000, relative risk 4.5, (95% CI: 2.6 to 8.1).
Figure 2 Proportion of Uterine Ruptures in Relation to the Quintile of Predicted Probability of Emergency Cesarean Delivery for the Whole Population
n = 23,286; p < 0.001 (Chi square test for trend).
The population studied had excluded women who delivered at 37- to 39-wk gestation. A model (excluding week of gestation) was fitted for women delivering between 40- and 42-wk and was evaluated among women delivered at 37- to 39-wk gestation in whom the documented duration of labor was greater than or equal to 4 h but otherwise applying the same inclusion and exclusion criteria as the main study cohort. The area under the ROC curve was 0.692. Among the 10,147 eligible women who delivered at 37 to 39 wk, there were 1,826 cesarean deliveries (18.0% compared with 26.0% in the rest of the population), giving a pretest odds of 0.22. When the probability of cesarean section was estimated using a prior odds of 0.22 and the ALLRs listed in Table 4 (excluding week of gestation), the observed and expected number of cesarean deliveries were similar (Figure 3).
Figure 3 Use of ALLRs to Predict Probability of Cesarean Section among Women Delivered 37 to 39 wk with a Documented Duration of Labor of Greater than or Equal to 4 h
Probability estimated using likelihood ratios in Table 4 (excluding gestational age) and the prior odds of 0.22 (equivalent to background risk of cesarean section in this group).
Discussion
Women who have had a prior cesarean delivery need to choose whether to have a planned repeat cesarean section or to attempt vaginal birth in subsequent pregnancies. The risk of maternal morbidity depends on whether the attempt at vaginal birth is successful [1,4]. An informed discussion of this decision requires an assessment of the risk of emergency cesarean section. However, there is, at present, no validated method that allows antepartum assessment of the risks of emergency cesarean section [5], and counseling of women is, at best, semiquantitative. In the present study we provide a validated model that classifies over half this population as being at low or high risk of emergency cesarean section, on the basis of thresholds suggested by a previous systematic review [5]. When the model was validated, 36% of women had a predicted risk of cesarean section of less than 20%, and their overall cesarean section rate was 10.9%. Conversely, 16.5% of women had a predicted risk of cesarean section of greater than 40%, and their overall cesarean section rate was 47.7%.
One of the other principal concerns among women who have had a prior cesarean section is the risk of intrapartum uterine rupture. Uterine rupture is associated with an increased risk of severe maternal complications, such as hysterectomy and hemorrhage [1,4] and with an increased risk of severe effects on the infant, including hypoxic ischemic encephalopathy[4] and perinatal death [9]. Even if a woman had a low risk of emergency cesarean section, she may choose to have a planned repeat cesarean section due to concerns about the possibility of uterine rupture. However, we found that women who were at low risk of emergency cesarean section were also at low risk of uterine rupture, including catastrophic rupture leading to perinatal death. Among women with a predicted cesarean section risk of less than 20%, the incidence of uterine rupture was 2.0 per 1,000, whereas among women with a cesarean section risk of greater than 40%, the incidence of uterine rupture was 9.1 per 1,000. This is the first study, to our knowledge, to demonstrate a direct association between the risk of a failed attempt at vaginal birth and the risk of uterine rupture. This cannot be explained by ascertainment bias because the association was still apparent when the analysis was confined to catastrophic uterine rupture leading to death of the infant, which would be ascertained irrespective of the mode of delivery.
In order for this model to be clinically useful, it needs to be presented in a way that practicing clinicians can understand and apply. To this end, we have employed a method for converting the logistic regression model into ALLRs. These can be used like conventional likelihood ratios and an example is given in the box. The prior odds are multiplied by the appropriate ALLRs to give the posterior odds from which the probability of cesarean section can be derived. Because this method is very similar to that used for Down syndrome screening, we feel that it is likely to be generally understood by practicing clinicians. A previous method has been described in detail to convert logistic regression models into a Bayesian format [15]. This method provides identical results to our method for simple models. However, for models with categorical variables containing three or more groups or in which the scaling of a continuous variable changes between the univariate and multivariate analysis, the previously described method does not generate identical estimates of probability. Our method always generates estimates of probability that are identical to the logistic regression model. It can be thought of, therefore, as a simple format for the presentation of logistic regression models.
The present study has a number of strengths over previous studies. First, we had a population of over 23,000 women. The largest previous study included approximately 5,000 women [6]. Second, we were able to ascertain uterine rupture in our population, including uterine rupture leading to perinatal death. Studies using registry-based data have the profound weakness that uterine rupture may be inconsistently defined and include cases of avascular wound dehiscence detected at the time of cesarean delivery [2]. This study design could lead to ascertainment bias because women having cesarean delivery would be more likely to have avascular wound dehiscence identified. However, our data sources allowed us to identify uterine ruptures that led to death of the infant. Third, because of the large numbers, we could confine the analysis to women delivered at or after 40-wk gestation. Large-scale registries lack details such as whether an attempt at vaginal birth was planned. Planned cesarean sections are typically performed at 38–39 wk in the United Kingdom [16]. By confining the analysis to births at or after 40 wk, we could exclude women who were not truly attempting vaginal birth. Fourth, the risk of cesarean section could be estimated using information available in the antepartum period. Counseling of women regarding vaginal birth frequently involves the distinction between attempting vaginal birth if the onset of labor is spontaneous but not attempting it if labor needs to be induced, particularly if prostaglandins are used to ripen the cervix [17]. For this reason, we included week of delivery and method of induction of labor in the model, and, therefore, the current model can inform such decisions. Excluding gestation and mode of induction had very little practical effect on the predictive ability of the model (data not shown). Fifth, we analyzed continuous variables continuously rather than categorizing them, which increases statistical power. This analysis may explain why we observed a positive association between maternal age and risk of cesarean section, whereas some other studies have not [7]. Interestingly, the association with age became much stronger in multivariate analysis. This result may reflect negative confounding by previous vaginal birth. This factor was strongly protective against cesarean delivery, and these women were significantly older than other women.
The present study has some weaknesses. First, the data were obtained from Scotland and there may be concerns in applying this model to other populations. However, we assessed the robustness of the predictors employed by selecting records for the development and validation groups on the basis of factors that might reflect variation in other populations. We found the model was similarly predictive in and out of sample when these categorizations were performed by hospital throughput, socio-economic deprivation category, and year of birth. This finding suggests that the maternal and obstetric characteristics used in the model are likely to be robust even when applied to populations with different obstetric practices. Second, we lacked data on other factors that might be predictive of the risk of emergency cesarean delivery, such as body mass index, the indication for the previous cesarean section, and whether a previous vaginal birth preceded or followed the previous cesarean section. Nevertheless, we report the first validated model to give useful discrimination of risk to greater than 50% of the population [5].
A further potential weakness with the model is that it was derived from women delivering at or after 40-wk gestation. As discussed above, we confined the primary analysis to this group in order to identify women who were truly attempting vaginal birth. Some women who were scheduled for planned cesarean section will have attended prior to this date in labor. Such women would be documented as an intrapartum emergency cesarean section but did not truly attempt vaginal delivery. However, we needed to assess the validity of the model for women who deliver at earlier week of gestation at term. Another means to identify women who were truly attempting vaginal birth is to confine analysis to those with a documented duration of labor of at least 4 h. We evaluated the model in women who were delivered between 37 and 39 wk who had labor for 4 h or longer. The discriminative power of the model was comparable, with an area under the ROC curve of 0.692. These women had a lower prior risk of cesarean section (18%) than women at or after 40-wk gestation (26%). When ALLRs were employed and the lower overall rate of cesarean delivery was accounted for by using the prior odds of 0.22, the observed and expected numbers of cesarean deliveries were similar (Figure 3). We conclude that the ALLR-based model is appropriate for births between 37–39 wk if lower prior odds of cesarean delivery are employed. Moreover, this analysis highlights one advantage of an ALLR-based approach, namely, that it is simple to adjust the estimate of probability for a lower or higher prior odds of the outcome.
In conclusion, we present a simple, validated model for clinical estimation of the risk of emergency cesarean section among women with a prior cesarean delivery attempting vaginal birth. Women at high risk of cesarean delivery are also at increased risk of uterine rupture, including catastrophic rupture leading to perinatal death.
Supporting Information
The full logistic regression model for calculating ALLRs is
where x
1, x
2,…, xn are the predictor variables, β1, β2,…, βn are their regression coefficients, and α is the constant. Let the fitted values of α, β1, β2,…, βn be .
The log likelihood ratio for x
1, for example, may be defined as the log odds of the outcome conditional on x
1, x
2,…, xn minus the log odds of the outcome conditional on x
2,…, xn. The latter odds cannot in general be derived from equation 1 because the effects of the omitted x
1 are partly picked up by x
2,…, xn: the true likelihood ratio for x
1 therefore depends on the values of x
2,…, xn. We have created ALLRs that do not depend on the values of x
2,…, xn.
To create the ALLRs, we force the coefficients in the second model to be the same as those estimated in the first model, but allowing a different intercept:
In this model, only the parameter is to be estimated; the other parameters take their fitted values from equation 1. We can then calculate the ALLR for x
1 as
This procedure is repeated for each variable x
2,…,Xn to calculate ALLR
2,…, ALLRn
A small correction factor must be added to the ALLRs in order to ensure that the sum of the overall log odds and all the ALLRs is exactly equal to the log odds computed from equation 1. The appropriate correction factor is cid, where , α̂ is the overall log odds, and Σi
ci = 1. In this paper, d = −0.021 and all correction factors were smaller than 0.01 in magnitude. To ensure that values of each ALLRi straddle 1, ci is calculated as mi/(m
1 +…+ mn) where mi is the sample minimum or maximum (depending on whether d is positive or negative) of ALLRi.
At the end of this procedure, the sum of the overall log odds and all the ALLRs exactly equals the log odds computed from equation 1. Our procedure is therefore nothing more than a restatement of the results of the logistic regression in an easily interpretable format.
Patient Summary
Background
The number of cesarean sections performed is increasing. Many women who have had a previous cesarean section want to try to have a vaginal birth in the next pregnancy, but they and their doctors may be worried about whether or not it is safe for them and the baby to attempt the vaginal birth.
What Did the Researchers Do and Find?
They looked at a large number of women in Scotland who had had one previous cesarean section and who were about to have another baby. Altogether they studied 23,286 women who had attempted to give birth vaginally between 1985 and 2001. They split the women into two groups; using one group, they developed a way of predicting the outcome (whether or not the women were going to need an emergency cesarean section) by looking at various risk factors including mother's age, height, sex of baby, gestation, and whether and how the birth was induced. Then, using the second group of women, they tested the model they had developed. They discovered that they could identify half of the women as being at high or low risk of needing emergency cesarean section, with the remainder being at intermediate risk. The things that increased risk were older maternal age, smaller height, male sex of baby, labor induced by prostaglandin, not having had a previous vaginal birth, and later birth. They also found that the risk of having a ruptured uterus went up as the risk of emergency cesarean section went up.
What Do These Findings Mean?
Obstetricians will be able to use the model developed to try to give women a more accurate estimate of whether they will need to have a cesarean section once they have had one in a previous pregnancy.
Where Can I Get More Information Online?
The following Web sites have relevant information.
MedlinePlus has a selection of pages with information for patients:
http://www.nlm.nih.gov/medlineplus/cesareansection.html
The United Kingdom's OMNI gateway has links to sites about cesarean section:
http://omni.ac.uk/browse/mesh/D002585.html
Box 1.
Sample Calculation
Background risk of cesarean section = 26%. Convert into odds if prior odds of cesarean section = 26/74 = 0.35
Example
A 37-y-old woman, 160 cm tall, with no previous vaginal birth, and with a male infant wishes to know probability of cesarean section if she requires induction of labor at 41 wk gestation using prostaglandin.
Summary
ALLR = 1.40 × 1.05 × 1.51 × 1.10 × 1.13 × 1.37 = 3.78. Posterior odds = 0.35 × 3.78 = 1.32. Chance of cesarean delivery = 1.32/(1 + 1.32) = 0.57 or 57%. (This is identical to the estimated risk using the logistic regression equation in the footnote of Table 4).
The authors received no external funding.
Citation: Smith GCS, White IR, Pell JP, Dobbie R (2005) Predicting cesarean section and uterine rupture among women attempting vaginal birth after prior cesarean section. PLoS Med 2(9): e252.
Abbreviations
ALLRadjusted log likelihood ratio
CIconfidence interval
ICDInternational classification of disease
IQRinterquartile range
ORodds ratio
ROCreceiver operating characteristic
SMR2Scottish Morbidity Record
==== Refs
References
McMahon MJ Luther ER Bowes WA Olshan AF Comparison of a trial of labor with an elective second cesarean section N Engl J Med 1996 335 689 695 8703167
Lydon-Rochelle M Holt VL Easterling TR Martin DP Risk of uterine rupture during labor among women with a prior cesarean delivery N Engl J Med 2001 345 3 8 11439945
Smith GC Pell JP Cameron AD Dobbie R Risk of perinatal death associated with labor after previous cesarean delivery in uncomplicated term pregnancies JAMA 2002 287 2684 2690 12020304
Landon MB Hauth JC Leveno KJ Spong CY Leindecker S Maternal and perinatal outcomes associated with a trial of labor after prior cesarean delivery N Engl J Med 2004 351 2581 2589 15598960
Hashima JN Eden KB Osterweil P Nygren P Guise JM Predicting vaginal birth after cesarean delivery: A review of prognostic factors and screening tools Am J Obstet Gynecol 2004 190 547 555 14981405
Flamm BL Geiger AM Vaginal birth after cesarean delivery: An admission scoring system Obstet Gynecol 1997 90 907 910 9397100
Troyer LR Parisi VM Obstetric parameters affecting success in a trial of labor: designation of a scoring system Am J Obstet Gynecol 1992 167 1099 1104 1415398
Cole S Chalmers I McIlwaine GM Scottish maternity and neonatal records Perinatal audit and surveillance 1980 London Royal College of Obstetricians and Gynaecologists 39 51
Smith GC Pell JP Pasupathy D Dobbie R Factors predisposing to perinatal death related to uterine rupture during attempted vaginal birth after caesarean section: Retrospective cohort study BMJ 329 375
Information and Statistics Division NHS Scotland Scottish perinatal and infant mortality report 2000 2001 Edinburgh ISD Scotland Publications 74
Campbell S Soothill P Chervenak FA Isaacson GC Campbell S Detection and management of intrauterine growth retardation: A British approach Ultrasound in obstetrics and gynecology 1993 Boston Little Brown and Company 1431 1435
Hey EN Lloyd DJ Wigglesworth JS Classifying perinatal death: Fetal and neonatal factors Br J Obstet Gynaecol 1986 93 1213 1223 3801351
Smith GCS Dellens M White IR Pell JP Combined logistic and Bayesian modeling of cesarean section risk Am J Obstet Gynecol 2004 191 2029 2034 15592287
Fagan TJ Letter: Nomogram for Bayes theorem N Engl J Med 1975 293 257 1143310
Spiegelhalter DJ Knill-Jones RP Statistical and knowledge-based approaches to clinical decision support systems, with an application in gastroenterology J R Stat Soc [Ser A] 1984 147 35 77
Morrison JJ Rennie JM Milton PJ Neonatal respiratory morbidity and mode of delivery at term: Influence of timing of elective caesarean section Br J Obstet Gynaecol 1995 102 101 106 7756199
ACOG Committee on Obstetric Practice Committee opinion. Induction of labor for vaginal birth after cesarean delivery Obstet Gynecol 2002 99 679 680 12039139
|
16146414
|
PMC1201366
|
CC BY
|
2021-01-05 10:40:31
|
no
|
PLoS Med. 2005 Sep 13; 2(9):e252
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020252
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 10.1371/journal.pmed.0020325SynopsisObstetrics/GynecologyObstetricsPregnancyA Tool to Estimate the Risks of Repeat Cesarean Section Synopsis9 2005 13 9 2005 2 9 e325Copyright: © 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.
Predicting Cesarean Section and Uterine Rupture among Women Attempting Vaginal Birth after Prior Cesarean Section
==== Body
Cesarean section can be a life-saving technique for both mother and infant; however, it is a major abdominal operation that poses medical risks to a mother's health, including infections, hemorrhage, need for transfusion, injury to other organs, anesthetic complications, psychological complications, and a maternal mortality two to four times greater than that for a vaginal birth. The World Health Organization (WHO) has said that no country can justify having a cesarean rate greater than 10%–15%. Despite this advice, in the past 20 years, cesarean section rates have risen to nearly 25% in some countries.
To address rising rates of cesarean delivery, health authorities have encouraged women with a previous cesarean to attempt vaginal birth in subsequent pregnancies. However, studies have indicated an increased risk of serious adverse outcome among such women who attempt vaginal birth compared with a planned repeat cesarean delivery. This is due to a greatly increased risk of complications among women who attempt vaginal birth but ultimately are delivered by emergency cesarean section. Consequently, researchers have tried to identify women at low and high risk of failure for an attempted vaginal birth after a prior cesarean, but currently there is no validated antepartum tool to predict the risk of a failed attempt at vaginal birth among women with a prior cesarean delivery.
Now Gordon Smith and colleagues describe the development of a simple, validated model to predict the risk of emergency cesarean section among women attempting vaginal birth after a previous cesarean delivery. They also try to determine whether women at increased risk of cesarean were also at increased risk of uterine rupture, including catastrophic rupture leading to death of the infant.
The team studied 23,286 women with one prior cesarean delivery who attempted vaginal birth at or after 40 weeks of gestation. The population was randomly split into two groups, one on which the model was developed and the second on which it was validated.
The researchers found that the following factors were associated with emergency cesarean section: increased maternal age, lower maternal height, male fetus, no previous vaginal birth, prostaglandin induction of labor, and birth at 41 weeks or 42 weeks gestation compared with 40 weeks. In the validation group, 36% of the women had a low predicted risk of caesarean section and 16.5% of women had a high predicted risk; 10.9% and 47.7% of these women, respectively, were actually delivered by caesarean.
The predicted risk of caesarean was also associated with the risk of uterine rupture in general, and of uterine rupture associated with perinatal death, and women who were at low risk of emergency cesarean section were also at low risk of uterine rupture, including catastrophic rupture leading to perinatal death—one of the principal concerns among women who have had a previous cesarean birth.
Despite the strengths of the present study, including the very large population size, studies using registry-based data have the weakness of inconsistent definitions, admitted the authors. Also the study lacked data on other risk factors for emergency cesarean delivery, such as body mass index, the indication for the previous cesarean section, and whether a previous vaginal birth preceded or followed the previous cesarean section.
Nonetheless, the findings offer a validated model for estimating the risk of emergency cesarean section among women with a prior cesarean delivery who attempt vaginal birth. The true worth of the model will become clear when other researchers test it.
Erratum note: The figure that appeared with this synopsis (Figure DOI:10.1371/journal.pmed.0020325.g001) was incorrect and has been removed from the HTML version of the article. We are unable to remove it from the PDF version. Corrected 10/10/05
|
0
|
PMC1201367
|
CC0
|
2021-01-05 10:40:32
|
no
|
PLoS Med. 2005 Sep 13; 2(9):e325
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020325
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1616339510.1371/journal.pgen.001003305-PLGE-RA-0134R2plge-01-03-03Research ArticleBioinformatics - Computational BiologyEvolutionStatisticsGenetics/GenomicsGenetics/Genome ProjectsEukaryotesAnimalsVertebratesMammalsMus (Mouse)Evidence of a Large-Scale Functional Organization of Mammalian Chromosomes Mammalian Chromosome OrganizationPetkov Petko M Graber Joel H Churchill Gary A DiPetrillo Keith King Benjamin L Paigen Kenneth *The Jackson Laboratory, Bar Harbor, Maine, United States of AmericaClark Andy EditorCornell University, United States of America* To whom correspondence should be addressed. E-mail: [email protected] 2005 9 9 2005 1 3 e3316 5 2005 3 8 2005 Copyright: © 2005 Petkov 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.Evidence from inbred strains of mice indicates that a quarter or more of the mammalian genome consists of chromosome regions containing clusters of functionally related genes. The intense selection pressures during inbreeding favor the coinheritance of optimal sets of alleles among these genetically linked, functionally related genes, resulting in extensive domains of linkage disequilibrium (LD) among a set of 60 genetically diverse inbred strains. Recombination that disrupts the preferred combinations of alleles reduces the ability of offspring to survive further inbreeding. LD is also seen between markers on separate chromosomes, forming networks with scale-free architecture. Combining LD data with pathway and genome annotation databases, we have been able to identify the biological functions underlying several domains and networks. Given the strong conservation of gene order among mammals, the domains and networks we find in mice probably characterize all mammals, including humans.
Synopsis
The arrangement of genes along chromosomes affects their function as well as the likelihood that particular combinations of genes will be inherited together, and evolution has had many millions of years to optimize these arrangements. Because the arrangements are nearly identical in all mammals, one can use the powerful techniques of mouse genetics to explore their roles in our own genomes. The authors find that genes that cooperate in bringing about various cellular and physiological functions, such as immune responses, are often clustered together on chromosomes, and that detailed maps of these relationships can be built. The new techniques have proven so powerful that they can identify functional interactions among genes that are not even on the same chromosome. Beyond illuminating the evolutionary pressures that brought them about, mapping these arrangements will be of great utility in the ongoing searches in many laboratories for the genes underlying our common diseases, such as cancer, heart disease, and diabetes.
Citation:Petkov PM, Graber JH, Churchill GA, DiPetrillo K, King BL, et al. (2005) Evidence of a large-scale functional organization of mammalian chromosomes. PLoS Genet 1(3): e33.
==== Body
Introduction
The physical and functional organizations of eukaryotic chromosomes are correlated outcomes of evolution potentially reflecting interactions among structural, regulatory, and functional factors. As typified by the α and β globin gene clusters, tandem duplications can give rise to gene families whose members develop divergent, but still related, functions over time. Gene clusters may arise as a means of promoting their coregulation through regional controls of chromatin structure and expression, and there is now considerable evidence, well summarized by Hurst et al. [1], that for variety of eukaryotes, including yeast, Caenorhabditis, Drosophila, higher plants, and mammals, genes sharing expression patterns are more likely to be in proximity than would be expected by chance. And finally, Fisher [2] and later Nei [3,4] have argued on theoretical grounds that when genes interact epistatically, evolutionary selection will promote their genetic linkage as a means of enhancing the coinheritance of favorable allelic combinations. Dobzhansky and others have provided experimental evidence on the importance of coadapted sets of alleles in their studies analyzing the fitness of chromosomal inversions in Drosophila [5]. Although there is limited molecular evidence in this regard, we can presuppose that allelic coadaptation is most likely to be effective when the various gene products participate in the same biological function. In yeast, proteins within the same macromolecular complex are about twice as likely to be encoded by genes on the same chromosome as would be expected by chance [6], and for a variety of eukaryotes, including humans, the genes encoding the enzymes of some pathways of intermediary metabolism occur in multiple clusters [7]. Additionally, several very specific, albeit isolated, examples of functional clustering are known in mammals, including the major histocompatibility complex, reviewed in [8], and the four HOX clusters [9].
Obviously, structural, regulatory, and functional factors are not mutually exclusive possibilities underlying selection, and while they could act in concert, that is not a requirement. Genes coexpressed in the same tissue may participate in distinct functions, the multiple functions of hepatocytes being a prime example, and a particular biological function may involve interaction among multiple cell types, as occurs in the mammalian immune response.
To address the question of functional clustering further, we have turned to inbred mice, which provide a unique and readily accessible experiment on evolutionary selection. Nearly a million years ago the species Mus musculus diverged into three geographically distinct subspecies, M. m. domesticus, M. m. musculus, and M. m. castaneus, along with a fourth subspecies, M. m. molossinus, that is an ancient fusion of the latter two subspecies. Over the last few centuries amateur mouse fanciers intercrossed these subspecies, and beginning less than a hundred years ago these genomic mixtures were used as the source of many of our present laboratory inbred strains. The process of inbreeding to homozygosity imposes intense selective pressures; all efforts among some species have failed, and with mice, only a fraction of initial attempts succeeded. Accordingly, we can expect that if clustering of functionally related genes is a common feature of mammalian genomes, there is likely to be selection for coadapted allelic combinations among the genes encoding functions that influence fitness and survival during inbreeding. This would result in regions of linkage disequilibrium (LD) among inbred strain genomes; i.e., some allelic combinations should occur more often than expected by chance.
This prediction can be tested using the large numbers of single-nucleotide polymorphism (SNP) genomic markers that have been typed on multiple mouse strains. In doing so, we found a substantial fraction of the mouse genome present in LD domains averaging several megabases in size, along with experimental evidence that the preferred configurations of these domains confer a considerable selective advantage during inbreeding. Moreover, the LD data show that domains identified in this way interact in complex networks across the genome. Correlating domain and network maps with information from pathway and Gene Ontology (GO) databases has made it possible to identify some of the underlying functions on which selection has acted.
Results
Strains and Single-Nucleotide Polymorphisms
The starting point for this analysis was a dataset describing the distribution of alleles at 1,456 SNPs, chosen for their high information content, among a set of 60 common and wild-derived inbred mouse strains chosen for their genetic diversity. They represent all of the SNPs and strains meeting a set of minimum requirements within a larger set (1,638 SNPs, 102 strains) originally developed as a mouse mapping panel [10]. The final dataset was characterized by a median minor allele frequency of 0.32 (minimum 0.1) to provide statistical robustness; a median frequency of allelic differences among pairs of strains of 42.8% (minimum 20%) to remove “sibling” strains that would otherwise introduce artifactual LD; and a median frequency of successful SNP determination of 98.3% (minimum 90%) to avoid possible biases from failed typings. The identity of these strains and the phylogenetic relationships among them are indicated in Figure 1, which was constructed using neighbor-joining methodology [11].
Figure 1 Neighbor-Joining Distance Tree of the Mouse Strains Used in This Study
The length and angles of the branches have been optimized for printing and do not represent actual phylogenetic distances. Group 1, Bagg albino, 129, and DBA-related strains; group 2, Swiss mice and Asian strains; group 3, wild-derived strains.
Estimating LD
Historically, the term linkage disequilibrium has been applied to populations, referring to the nonrandom association of alleles at linked genes; i.e., over time, recombination has failed to establish a random assortment of alleles, some combinations occurring more often than expected by chance. Although in mammalian genetics LD has been traditionally applied to linked genes, we are reluctant to introduce new terminology, and so have extended the definition to any pair of markers, regardless of location. Following the discussions and recommendations of Hedrick [12] and Devlin and Risch [13] on the use of alternative methods, we have estimated LD using D′, the difference between the observed frequency of an allelic combination and its random expectation, relative to the maximum deviation possible given the allele frequencies of the two markers [14,15]. D′ corrects for differences in allele frequencies and describes LD equally well when there is selection for or against the combination of majority alleles. A cumulative Fisher's exact test (FET) was used to compute the probability (pFET) of obtaining an equally or more extreme distribution under the null hypothesis of random allelic association between pairs of SNPs. This approach has the advantage of providing separate estimates of the extent of disequilibrium and the likelihood of its being due to chance, and is especially valuable for markers with lower minor allele frequencies [12,16,17]. In addition to D′, we considered X
2, r
2 and mutual information measures of disequilibrium on the same data. All of these measures were highly correlated, above 0.9, and here we report our results in terms of D′.
LD Patterns Reveal the Presence of Domains
The observed data for the 60 strains (Figure 2, solid squares) shows that among marker pairs less than 1 Mb apart, a considerable excess—approximately 44%—are in disequilibrium at pFET < 0.001, and that disequilibrium decays slowly as the distance between pairs increases, reaching a lower bound of approximately 1% at distances above 20 Mb. We estimated the false discovery rate at pFET < 0.001 to be 0.09 via the conservative method of Benjamini and Hochberg [18]. Tests with two randomized datasets indicate that these observations of high LD are highly significant and not a consequence of either marker location or allele frequency distributions. In one set, marker locations were randomized while maintaining the assignments of alleles to strains (Figure 2, red triangles), and in the other set the assignments of alleles to strains were randomized while preserving allele ratios and marker locations (Figure 2, solid circles). Neither control set indicated dependence on marker separation, with a uniform 1% of the pairs in LD for set 1, a value similar to that observed in interchromosomal marker pairs (unpublished data), and less than 0.1% of the pairs in LD for set 2, approximately conforming with random expectation.
Figure 2 Dependence of the Fraction of Markers in LD from the Distance between Them
Fisher's exact test was used, pFET < 0.001 unadjusted. Solid squares, actual data; red triangles, randomized genomic positions of the markers; solid circles, randomized alleles and strains.
Almost certainly, these calculations underestimate the true extent of LD, as the requirement of pFET < 0.001, chosen to keep the false discovery rate low, also increases the false negative rate.
Figure 3 displays the interactions between all pairs of markers on Chromosome 14, a chromosome exhibiting extensive LD. Marker coordinates on the chromosome, expressed in Mb, form the two axes; where the coordinates for a pair of markers intersect, D′ and the base-ten logarithm of 1/pFET are plotted above and below the diagonal, respectively. Points along the diagonal represent closely spaced markers. We used a dynamic-programming method (Materials and Methods) to identify the most probable domains of LD, which appear as blocks along the diagonal. Depending on the exact parameters chosen for dynamic programming, the mouse genome contains several hundred LD domains, occupying one-fourth to one-third of the total length.
Figure 3 A Representation of LD between Marker Pairs on Mouse Chromosome 14 Reveals a Domain Structure
LD is plotted as D′ and log10 (1/pFET) above and below the diagonal, respectively. The x- and y-coordinates are NCBI Build 33 genome positions for SNPs. Black regions reflect genomic sequence not covered in this SNP set (i.e., missing data). To highlight pairs of interest, D' values have been suppressed (plotted as gray) for marker pairs with pFET > 10−3. White boxes represent LD domains, identified as described in the text. Yellow boxes represent regions of synteny identified through mouse-rat-human-chicken comparison [19].
At the present level of resolution, domains defined in this manner appear related to the evolutionarily conserved syntenic blocks previously identified in a joint analysis of the mouse, rat, human, and chicken genomes [19]. It will be of considerable interest to see if this relationship persists when more extensive data become available.
To confirm that these domain maps are independent of our SNP panel, Chromosome 14 was tested using two other panels with denser SNP coverage, albeit on fewer strains. These panels were generously provided by Tim Wiltshire at GNF-Novartis [20] and Eric Schadt at Rosetta-Merck. The results are almost identical to the patterns in Figure 3 (unpublished data). Although there is no reason to believe it will vitiate our results, we should note as a caution that all presently available mouse SNP panels are limited in their representation of mouse genome diversity, in that they are derived by comparing the genomic sequences of a small number of strains.
Domains Reflect Inbreeding Selection
Recombinant inbred (RI) lines of mice provide a direct means of testing whether domains result from inbreeding selection. We can ask whether LD domain regions whose allelic composition has been scrambled by recombination have a reduced ability to survive further inbreeding, using as the control regions of the genome that do not show LD. RI lines are nearly ideal for this approach, as they are created by crossing two genetically defined progenitor, inbred strains of mice, obtaining an F2 population in which all allele ratios are 50:50, and then inbreeding a set of new lines from pairs of F2 mice. The result is a new set of inbred mouse strains created from genetically defined progenitors. Within LD domains, if inbreeding during RI line formation favors the survival of preexisting allelic combinations over new ones, the number of lines showing recombination across these regions should be fewer than expected. This effect should be absent from the nondomain regions. The number of RI lines expected to show recombination across a region if there is no adverse selection can be calculated from the rates of recombination seen among the F2 progeny obtained from crosses of the two parental strains. This is provided by the classic Haldane-Waddington equations, in which the fraction of lines recombinant between two autosomal markers is given by 4c/(1 + 6c), where c is the recombination fraction in a single generation. This equation, which was originally derived theoretically, has recently been validated by computer simulation [21].
Rates of recombination in crosses between C57BL/6J and either A/J or DBA/2J mice were measured among F2 progeny and then, using the same sets of markers, across the large BXA and BXD sets of RI lines created from these progenitor strains. Domain and nondomain regions were chosen for comparison solely on the basis of the availability of flanking markers and without prior knowledge of recombination frequencies. Comparing a set of domains totaling 108.0 Mb with the set of nondomains totaling 83.8 Mb, we found nearly equivalent single generation recombination rates, both very similar to the overall genome wide recombination rate of 0.55 cM/Mb (Table 1). After subsequent inbreeding, there were markedly fewer than expected RI lines containing recombinant LD domains, but this was not the case for the nondomain regions (p = 0.0000003 and 0.57, respectively). We conclude that LD domains are not deficient in normal recombination, and that selection against less favorable allelic combinations is a strong factor generating LD. These results from RI lines are noteworthy in that, unlike the data from inbred strains, there are no issues reflecting common origins, biases of marker selection among multiple strains, or possible effects of allele frequencies, as all input allele ratios are 50:50.
Table 1 Comparison of Recombination in F2 Crosses and Recombinant Inbred Lines
Confirmation that selection reducing the survival of recombinants occurs on a broad scale during RI line inbreeding was provided by genomewide measures of recombination among sets of RI lines. A comprehensive dataset of 1,575 markers used to describe recombination among 109 RI lines bred from various parental combinations has been assembled, expanded and quality controlled by Williams et al. [22]. The expected apparent genome length among these strains can be calculated using the Haldane-Waddington equations by multiplying the autosome lengths by 4, the X chromosome by 8/3 [21], and reducing the expected length by 1/1.075 (the Williams et al. correction for the average distance between markers). If there is no selection reducing survival of recombinants, the single generation genome length of 1,465 cM should generate an apparent genome length of 5,362 cM. For the 109 strains, this predicts 5,845 recombination events, however, only 4,786 were observed, a deficiency of 18.1% (X2 = 192, p < 10−47). Every autosome was deficient in recombination, and the deficiency was particularly marked for the X chromosome, 32%. In further evidence of selective forces during inbreeding, the authors noted multiple regions of residual heterozygosity persisting long after they would be expected to be lost by chance.
If the 40% reduction in recombinant survival seen among the LD domains described in Table 1 is typical of LD domains in general, an 18.1% genome wide lack of recombinant survival among RI lines suggests that a substantial fraction of the mouse genome, as much as one-third to one-half, may lie within the LD domains defined by selection during inbreeding. This would agree with our probable underestimate of the true extent of LD in Figure 2.
Domains Are Not the Result of Deficient Recombination, Gene Content, or Strain Origins
Although reduced recombination is a commonly accepted mechanism generating LD in human populations, as Table 1 shows, recombination frequencies per Mb were not significantly different between LD domain and nondomain regions, or from the genome-wide average of 0.55 cM/Mb. Reduced recombination across LD domains was seen only as a consequence of later inbreeding selection, and can best be described as “lack of recombinant survival” rather than “lack of recombination.”
LD domains are also not regions of particularly high or low gene density; domains and nondomains being quite similar in this regard (Table 1).
Although many of the older inbred mouse strains had their origins in limited populations of domesticated mice, this factor is not sufficient to explain the LD observations. The 60 mouse strains tested (excluding the branch anchored by C57BL/6J which provided the reference sequence for SNP discovery) separate into three phylogenetically distinct groups that conform well to our historical knowledge of their origins (see Figure 1) in which groups 1 and 2 consist of strains derived from domesticated mice and group 3 contains wild-derived strains. If LD is due to commonality of origin, the three groups should behave independently. However, when compared using a rank-order test (see Materials and Methods), there was an appreciable excess over chance of marker pairs showing LD in more than one group (Figure 4). This was also true when any two of the three groups were compared pairwise (unpublished data). These results do not mean that common origin effects are entirely absent from inbred mouse strains, only that selection appears to be a significant mechanism generating LD.
Figure 4 Distribution of Sum of Rank Scores for Marker Pairs
Distribution is shown for the three groups of mouse strains as described in Materials and Methods. Deviation from the random simulation indicates sharing of LD pairs between groups.
Functional Networks Are Contained and Organized within Domains
To investigate the hypothesis that LD domains contain biologically related genes with coadapted alleles, genes from the ten most significant domains (based on dynamic programming) were analyzed for an excess of functionally related genes using the VLAD program (http://proto.informatics.jax.org/prototypes/vlad) that relies on GO [23] annotations. Membership in gene pathways and networks was tested using Ingenuity's Pathways Analysis software (http://www.ingenuity.com), which relies on literature annotation of physical, metabolic, and regulatory interactions between gene products, and, importantly for this purpose, ignores information on gene locations. Genes with common GO annotations were found in five of the ten domains. When analyzed for their pathway content, in addition to a Hox cluster on Chromosome 6, four of these domains contained a total of 13 additional Ingenuity pathways, each containing eight to 21 colocalized genes. Collectively, 35% (157/455) of the genes in these pathways were located in single domains.
The densest clustering occurred in the Chromosome 1 domain between 167.2 and 174.2 Mb, a region containing a total of 119 predicted genes (via the ENSEMBL Nucleotide Sequence Database [http://www.ebi.ac.uk/embl]). When all 119 genes were tested for functional relationships, 21 genes proved to be components of an Ingenuity pathway of 35 genes containing two distinct subnetworks linked by the gene Crp (Figure 5). The lymphocyte subnetwork spans from Crp to Eat2 and includes genes coding for 15 proteins, among them five Fc receptors and the cell surface antigens Cd48, Cd244, and Ly9. The general functions of the genes in this network include immune response, inflammation, inflammatory disease, phagocytosis, movement of lymphatic system cells, activation and proliferation of leukocytes, and costimulation of T lymphocytes. The Myc subnetwork contains ten genes acted on by Myc and three genes whose products act on Myc. The general functions of this subnetwork include apoptosis (including apoptosis of lymphatic system cells), proliferation of tumor cell lines, transformation of cells, inactivation of mast cells, and cardiovascular system development and function.
Figure 5 An Example of a Gene Network that Is Largely Contained within an LD Domain Located between 167.2 and 174.2 Mb on Mouse Chromosome 1
Highlight colors on the network plot correspond with the regions shown on the genomic map. The genes in grey boxes are positioned in the same LD domain but not clustered. In this example, eight of the 11 genes in the lymphocyte subnetwork that are in the block (Eat2, Fcgr2b, Fcgr3a, Fcer1g, Cd244, Ly9, Slamf1, and Cd84) map to within 1.4 Mb. One gene, Adamts4, is located within this 1.4 Mb region, but is part of the Myc subnetwork. Three additional genes, Crp, Apcs, and Fcer1a, are within 600 kb of one another and 900 kb away from the other group of eight genes. One set of four genes, Cd244, Cd84, Ly9, and Slamf1, each bind Sh2d1a [44] and are organized sequentially along the chromosome. In humans, missense mutations in SH2D1A are associated with X-linked lymphoproliferative disease (XLP) [44], and homozygous targeted mutations of Sh2d1a in mice yield immune system abnormalities [45]. Another set of three genes, Fcgr2b, Fcgr3a, and Fcer1g, that are organized sequentially are Fc receptor subunits. A second set of three genes, Crp, Apcs, and Fcer1a, are also sequentially organized and directly bind with Fcer1g in the case of Fcer1a and Apcs [46] or bind Fcgr1a and Apcs in the case of Crp [47]. In the Myc subnetwork, two groups of genes are located in close proximity to one another. Mgst3 and Lmx1a are within 300 kb of one another. Myc increases expression of Mgst3 [26], and LMX1A regulates Ins1 [48], which has its expression downregulated by Myc [49]. Pex19, Pea15, Casq1, and Tagln2 are located within 400 kb. Myc decreases the expression of Pex19 [26] and Tagln2 [50]. Mtpn increases the expression of Casq1 and Myc [51]. Myc decreases the expression of Akt1 [26], and Akt1 increases serine phosphorylation of Pea15 [52].
The probability of finding a pathway with 21 of 35 genes clustered within a contiguous block of 119 genes is very small (p = 10−25, based on a hypergeometric calculation) even after correcting for the presence of six likely gene duplications, and after Bonferroni correction for the number of contiguous blocks of 119 genes in the genome. The chance of finding 13 such pathways with at least eight clustered genes spread across five LD blocks is even smaller.
The 21 pathway genes colocalized on Chromosome 1 are not all coregulated within a single tissue, indicating that this clustering reflects functional interactions among multiple cell types. Using the Mouse Gene Atlas database of gene expression patterns [24], many of the genes in the immune function branch of the pathway showed broad expression, albeit with an emphasis on expression in bone marrow and lymphocytes. However, two genes in the pathway, Apcs and Crp, are liver specific in expression, coding for secreted peptides that react with Fc receptors (other members of the pathway) on lymphocytes in the acute phase inflammatory response.
The physical locations of the pathway genes along the chromosome are correlated with the topology of the functional pathway (Figure 5); neighboring pathway genes are clustered in chromosomal subdomains, providing strong support for assembly of LD domains as a result of functional interactions. The two genes sharing liver specific expression, Apcs and Crp, are located within 200 Kb of each other, further suggesting that a small regulatory domain may be nested within a considerably larger functional domain.
One of the two remaining pathways represented in this domain is connected to the immune function pathway through the gene CD244; the joined pathways contain a total of 69 genes, of which 31 are located within the Chromosome 1 domain.
Four noncontiguous markers in the Chromosome 1 domain are also in LD with a single marker on Chromosome 6 that is located near the Cdkn1b, Bcl2l14, Emp1, and Csda genes, which are involved in apoptosis, cell cycle arrest, and cell growth. Myc decreases the expression of Cdkn1b and increases the expression of Emp1 [25] and Csda [26], and although there is no evidence that Bcl2l14 is regulated by Myc, Bcl2 protein family members are known critical regulators of apoptosis [27]. The LD observed between functionally related genes on Chromosomes 1 and 6 provides additional evidence for selection of coadaptive alleles, albeit on different chromosomes.
Domains and Markers Associate in Scale-Free Networks
The observation of LD between Chromosomes 1 and 6 is typical; domains on one chromosome are often in LD with domains on other chromosomes or with distant domains on the same chromosome, even when the two are separated by domains that are not in LD with either (Figure 6). Figure 6A expands Figure 3 to include Chromosomes 14–17, illustrating the fact that domains can be in disequilibrium with other distant domains or individual markers on the same or different chromosomes, suggesting the existence of interacting networks. Using the LD data, a network graph was constructed in which the markers are nodes, and edges are created between all marker pairs with D′ > 0.8 and pFET < 0.001. To identify the most highly connected subnets, the display was restricted to include only nodes that were part of a fully connected subnet (clique) of at least six nodes. As shown, the highlighted nodes in Figure 6B correspond to the same interchromosome networks highlighted in Figure 6A. Relaxing any of these very stringent requirements (D′, pFET, or connectivity) vastly expands the network.
Figure 6 Interchromosomal Plots of LD Reveal the Presence of Putative Interaction Networks
(A) A plot of the disequilibrium between pairs of SNP markers on mouse Chromosomes 14–17. Plot parameters are identical to Figure 2. The members of two mutually exclusive, and completely connected, putative interaction networks are highlighted with red and blue circles, chosen to correspond with the highly connected network cores shown in (B).
(B) A representation of two reduced, highly connected networks was created by restricting the edges to marker pairs with D′ ≥ 0.8 and pFET ≤ 10−3. To highlight only the most connected markers (nodes), the graph was reduced to show only nodes that were part of biconnected components (cliques) consisting of six or more nodes, and only components that include markers from Chromosomes 14–17, as shown in (A). Highlighted nodes correspond to the highlighted regions in (A).
Further confirmation of the existence of interchromosomal networks is provided by the RI lines data referred to above [22] which show substantial LD between markers on separate chromosomes, and, importantly, these associations form networks (e.g., chromosomes 8,10,12,13). These results extend earlier observations among RI lines reporting LD between several pairs of markers on separate chromosomes [28].
Metabolic and regulatory networks in lower organisms form scale-free networks [29–32]. In these networks, the frequency of molecules (nodes in network terminology) with n connections to other molecules is a negative exponential function of n; that is, as n increases, there is a constantly declining fraction of nodes with that number of connections. The distant interactions detected by LD might also show scale-free properties when the nodes correspond to genes coding for particular macromolecular gene products (protein or RNA) and possibly DNA binding sites, and the connections correspond to metabolic or physical interactions among these gene products. Scale-free behavior would also demonstrate that the LD networks are nonrandom and hence not a consequence of chance associations. To avoid the complications of local LD, the analysis only considered distant interactions, i.e., those between pairs of markers on separate chromosomes or at least 20 Mb apart on the same chromosome. With this limitation, interactions among the 1,456-marker set tested conformed well to a scale-free network, where fC(n), the fraction of markers with n associations, is fC(n) = 0.79n
−1.73, and this fraction is clearly different from the chance expectation that would be given by a Poisson distribution with the same average number of connections per marker (Figure 7).
Figure 7 The Connectivity among Pairs of Markers Shows a Scale-Free Character
The graph plotting the frequency of markers having n connections was created by designating each SNP marker as a potential node in a network, and considering a pair of markers to be connected if D′ ≥ 0.8 and pFET ≤ 10−3. To eliminate local effects, all pairs of markers separated by less than 20 Mb on a common chromosome were excluded from the analysis. The regression line was calculated for the best fit to the observed data. The deviation from the theoretical straight line at low connectivity is expected for a finite population when the average connectivity is greater than one, and the observed deviation agrees in magnitude with that obtained in computer simulations. The open squares are the results expected for the same average number of connections per marker if the frequency of markers with n connections conformed to a random Poisson distribution.
Biological Functions Correlate with LD Networks
A search for biological functions that might underlie the LD networks was carried out using every term in the GO annotation database [23] that had more than 50 and less than 500 genes assigned to it. Each gene set was tested for excess LD by counting the number of gene pairs in LD when judged by the stringent requirements of D′ > 0.8 and pFET < 0.001. SNPs within 2 Mb of each gene (based on their NCBI Build 33 genome coordinates) were used as markers for the gene, and the random expectation determined by carrying out the same analysis on 2,000 sets of an equal number of genes randomly chosen from the panel of annotated genes in ENSEMBL. The results show that after using the conservative Benjamini-Hochberg correction for multiple testing, there are a number of biological functions with an excess of distant genes in LD (Table 2).
Table 2 A Summary of Systematic GO Analysis of the LD Network
The validity of this test is supported by prior expectation of a positive result for the term “eye morphogenesis.” It is well known among mouse geneticists that mouse handlers picking mice from a cage inadvertently, but invariably, pick any visually impaired or blind mice first; simply put, the other mice do a better job of running away from the forceps. The result is strong selection for visual impairment.
Discussion
The relation between LD domains and the “haplotype blocks” [33–38] or “haplotype networks” [38] described in the literature requires clarification. A haplotype is a particular sequence, either of base pairs or allelic markers, in a defined DNA segment. A haplotype block is defined as a contiguous segment of DNA in which the number of observed haplotypes in a population is a small fraction of the total number possible. Such reduced sequence variability will necessarily lead to LD such as we have observed; thus, at their core, LD domains and haplotype blocks reflect the same phenomena.
Various computational approaches have been used to identify and characterize haplotype blocks. These include using a confidence interval restriction on all adjacent marker pairs within the block [37], and using a dynamic programming algorithm based on D′ values [35,38,39]. Haplotype blocks have also been defined purely by their lack of recombination hotspots or unlinked sequences within the block. Phillips et al. [35] have used extensive simulation to demonstrate that haplotype block structures could arise in the absence of selection simply through extremes of marker density and minor allele frequencies, two pitfalls we have been careful to avoid. Several additional, potential sources of haplotype block structure have been postulated, including (i) heterogeneous recombination, (ii) natural selection, (iii) population bottlenecks, and (iv) population admixtures. Haplotype networks resemble LD domains in that they do not explicitly forbid the inclusion of unlinked markers, a restriction commonly found in neighbor-based definitions [33].
Our statistical definition of LD domains relies on a dynamic programming approach, which explicitly allows contributions from all marker pairs in a putative domain and does not limit the analysis to adjacent or close neighboring pairs of markers (see Materials and Methods). In effect, our method identifies the most probable assignment of domains by maximizing a local sum based on the mutual information for all pairs of markers contained within a defined domain. We have retained the term “LD domain” for our own data for several reasons. First and foremost, our domains are operationally defined by LD, and this is the critical parameter in their definition. Additionally, (i) the literature is not entirely consistent in its definitions of haplotype blocks; (ii) unlike the human case, mouse LD domains do not differ from nondomains in recombination activity or gene content; (iii) the LD domains we observe are an order of magnitude larger than previously reported haplotype blocks in mice and even more so for those reported in the human genome, and (iv) the markers within domains associate across chromosomes in scale-free networks. Finally, and very importantly, the LD domains of mice appear to have a functional basis, arising as a consequence of inbreeding selection for compatible sets of genetically linked, functional elements, an association that, as far as we know, has not yet been made for haplotype blocks.
It is difficult to escape the conclusion that the selective factors acting to generate LD domains and networks during inbreeding reflect clustering and/or interaction of functionally related elements along chromosomes, thereby providing an opportunity for expanding our limited knowledge of the forces that drive molecular evolution in general, and coadaptation of alleles in particular. Chromosome maps and pathway networks are reflections of each other, with the potential of being mutually informative.
These observations are consonant with the theoretical suggestions of Fisher [2] and Nei [3,4] that evolution promotes the development of genetic linkage as a means of enhancing the coinheritance of favorable allelic combinations, and with the experimental work of Dobzhansky and others emphasizing the existence of coadapted gene sets in explaining the population genetics of Drosophila [5]. Inbred mice have provided a unique evolutionary experiment confirming these concepts in that they are derived from progenitor populations that had over a million generations of prior evolution in which to develop coadapted sets of alleles within local populations or subspecies. Laboratory matings arbitrarily scrambled these combinations. The resulting progeny were then subjected to intensive selection during inbreeding for the many epistatic interactions among genes, reinforced by the imposition of homozygosity, processes that effectively selected particular allelic combinations.
The LD domain on Chromosome 1 (167–174 Mb), which contains a large functional network, is a microcosm of the multiple factors—structural, regulatory, and functional—that give rise to genomic organization. It includes separate examples of apparent gene duplications, tissue-specific coregulation of adjacent genes, and functional signaling between cell types. Eight of the 21 functionally related genes within this domain might be related by gene duplications. Four, located within a 300-kb interval (Slamf1, Cd84, Lys, and Cd244), are related members of the immunoglobulin superfamily, and almost certainly arose as gene duplications. The genes Fcgr2B and Fcgr3A both code for Fc family receptors and are less than 100 Kb apart, suggesting that this pair also likely arose by gene duplication, and two other genes, Crp and Apcs, which share 50% sequence identity, are less than 200 Kb apart, again suggesting an ancient gene duplication. At the regulatory level, Crp and Apcs show a liver-specific expression, and may well be coregulated [40]. Thus, the Chromosome 1 domain is likely the outcome of a complex pattern of evolution reflecting structural, regulatory and functional selective factors, all acting to create this functional domain.
Comparative genomics has shown that gene order is a highly conserved feature of mammalian and even more evolutionarily distant chromosomes [19,41]. The selective pressures that originally drove the clustering of functionally related elements on mammalian chromosomes must have acted prior to the first divergence of mammalian orders more than 75 million years ago, and are not unique to any one order of mammals, much less the genus Mus. They almost certainly involved functions essential for mammalian existence. We can expect that elucidating these functional connections will be revelatory for all mammals, including ourselves, and suggestive evidence that LD domains resulting from selection might also occur in human populations has been reported recently [35,38,42].
What is notable about our present results is the extent to which functional clustering appears to be present in the mammalian genome. However, what must be emphasized is that while the LD domains of mice have been valuable in leading to this view, their sharply delimited edges probably reflect the particular selection pressures of inbreeding mice in a laboratory environment, rather than evolutionary forces in general. It is likely that functional clustering is as common in the non-domain regions, but less apparent, as it is not under inbreeding selection. Accepting this broader role for evolutionary selection suggests the possibility that all, or nearly all, of the mammalian genome is a linear continuum of functionally related elements and that clusters of functionally related genes may well be interdigitated among each other. Indeed, virtually every search for the many loci underlying complex traits not under laboratory selection, such as disease susceptibility, has revealed multiple epistatic interactions. The functional anatomy of the mammalian genome must be more complex than the fraction we have been able to observe so far.
Materials and Methods
Definition of domain boundaries through dynamic programming.
We defined LD domains of putatively functionally related elements, using a one-dimensional dynamic programming algorithm based on the mutual information (MI) content of all pairs of markers. A similar method was used previously with D′ as the basis of comparison [35,38]. The treatment below can be equivalently implemented with D′ replacing MI. The mutual information is defined by:
where the x and y summations are over all possible bases at markers i and j, respectively. The terms fx and fy are the observed frequencies of the indicated base at markers i and j, respectively, and fxy is the observed joint frequency for simultaneous occurrence of base x at marker i and base y at marker j.
MI varies between 0 and 1, but for dynamic programming to be stable, the average value of the sum term must be negative, therefore we used an offset of the mutual information, as defined by:
As shown, the term Aij is set identically to zero for marker pairs that are not on the same chromosome, since our domains are defined by proximity on a single chromosome. The offset Y(i,j) is dependent on markers i and j only through a sigmoid function of the distance separating the markers, dij. At large separations, the offset function asymptotically approaches a value of 1.0, making all such marker pairs noncontributing to a domain.
Y(i,j) has two free parameters: Y0, the minimum value of the offset, and D0, the center of the sigmoid function. In practice, Y0 was set as a quantile value (labeled as q) of the distribution of observed mutual information values for all interchromosomal pairs of markers. We typically set q between 0.95 and 0.999, however, other values were tested, as described below. The center of the sigmoid was typically set between 3 and 10 Mb (in practice, both D0 and dij were expressed in megabases), based on our observation (see Figure 1) that intrachromosomal markers become indistinguishable from interchromosomal markers, which necessarily cannot be part of a domain, for separations greater than 15–20 Mb.
All possible domain definitions were assessed by computing the triangular sum of all pairwise marker terms implied by the bounding markers i and j, again identically setting the term Sij to 0 if the bounding markers were not on the same chromosome:
Finally, the optimal local domain structure was obtained in a one-dimensional Smith-Waterman [43] dynamic programming, based on the recursion relationship:
Local traceback (with subsequent reconstruction of the B-vector) was performed in decreasing order on all peak values greater than zero until all domains of at minimum two markers were identified.
To assess the robustness of the LD domain definitions, we implemented the algorithm for all combinations of D0 = 1, 3, 5, 10, and 20 Mb, and q = 0.85, 0.9, 0.95, 0.99, 0.995, and 0.999, generating 30 different versions of the domains.
Testing the effect of strain origins by rank-order test.
Groups 1 and 2 (Figure 1) are the two major groups of strains derived from domesticated mice; Group 3 includes wild-derived strains not influenced by domestication. The branch containing C57BL/6J was omitted to avoid marker bias, as this was the primary comparison strain for marker development. If the LD observed between marker pairs resulted from commonality of strain origins, there should be little similarity between the identities of the gene pairs in LD in the various groups.
The resulting group sizes are 22, 16, and 14, which greatly reduces the statistical power of a D′/FET analysis. To overcome this limitation, the marker pairs in each group were put in rank order on the basis of their calculated pFET values for LD, and the sum of the three rank orders was calculated. To assure comparability, this analysis was restricted to the 1,031 markers that are polymorphic within every group of Figure 1. If the LD among marker pairs is unrelated in the three groups, the sum of rank orders should be distributed as the sum of three integers that are randomly distributed between 1 and 530,965, the number of marker pairs.
We thank A. Sia and S. Sheehan for technical help in DNA preparation; W. Zhang for preparing neighbor-joining tree; F. Pardo-Manuel de Villena, J. Flint, J. Blake, and W. Frankel for critical reading of the manuscript and helpful comments; T. Wiltshire and E. Schadt for kindly providing SNP databases; and R. W. Williams and his group for their prior excellent analysis of RI lines. JHG is partially supported by NIH/NCRR INBRE Maine contract 2 P20 RR16463-04.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. PMP and KP conceived and designed the experiments. PMP performed the experiments. GAC and JHG generated the statistical analysis. JHG developed the analytical and image-generating software. PMP, JHG, GAC, BLK, and KP analyzed the data. KD contributed reagents/materials/analysis tools. KP wrote the paper.
Abbreviations
FETFisher's exact test
GOGene Ontology
LDlinkage disequilibrium
RIrecombinant inbred
SNPsingle-nucleotide polymorphism
==== Refs
References
Hurst LD Pal C Lercher MJ 2004 The evolutionary dynamics of eukaryotic gene order Nat Rev Genet 5 299 310 15131653
Fisher RA 1930 The genetical theory of natural selection Oxford, United Kingdom Clarendon Press 318 p.
Nei M 1967 Modification of linkage intensity by natural selection Genetics 57 625 641 5583732
Nei M 2003 Genome evolution: Let's stick together Heredity 90 411 412 12764410
Dobzhansky T 1970 Genetics of the evolutionary process New York Columbia University Press 505 p.
Teichmann SA Veitia RA 2004 Genes encoding subunits of stable complexes are clustered on the yeast chromosomes: An interpretation from a dosage balance perspective Genetics 167 2121 2125 15342545
Lee JM Sonnhammer EL 2003 Genomic gene clustering analysis of pathways in eukaryotes Genome Res 13 875 882 12695325
Kelley J Walter L Trowsdale J 2005 Comparative genomics of major histocompatibility complexes Immunogenetics 56 683 695 15605248
Ferrier DE Minguillon C 2003 Evolution of the Hox/ParaHox gene clusters Int J Dev Biol 47 605 611 14756336
Petkov PM Ding Y Cassell MA Zhang W Wagner G 2004 An efficient SNP system for mouse genome scanning and elucidating strain relationships Genome Res 14 1806 1811 15342563
Saitou N Nei M 1987 The neighbor-joining method: A new method for reconstructing phylogenetic trees Mol Biol Evol 4 406 425 3447015
Hedrick PW 1987 Gametic disequilibrium measures: Proceed with caution Genetics 117 331 341 3666445
Devlin B Risch N 1995 A comparison of linkage disequilibrium measures for fine-scale mapping Genomics 29 311 322 8666377
Lewontin RC 1964 The interaction of selection and linkage. II. Optimum models Genetics 50 757 782 14221879
Lewontin RC 1995 The detection of linkage disequilibrium in molecular sequence data Genetics 140 377 388 7635301
Zapata C Alvarez G 1997 On Fisher's exact test for detecting gametic disequilibrium between DNA polymorphisms Ann Hum Genet 61 71 77 9066929
Lewontin RC 1988 On measures of gametic disequilibrium Genetics 120 849 852 3224810
Benjamini Y Hochberg Y 1995 Controlling the false discovery rate: A practical and powerful approach to multiple testing J R Stat Soc Ser B 85 289 300
Bourque G Zdobnov EM Bork P Pevzner PA Tesler G 2005 Comparative architectures of mammalian and chicken genomes reveal highly variable rates of genomic rearrangements across different lineages Genome Res 15 98 110 15590940
Pletcher MT McClurg P Batalov S Su AI Barnes SW 2004 Use of a dense single nucleotide polymorphism map for in silico mapping in the mouse PLoS Biol 2 e393. DOI: 10.1371/journal.pbio.0020393 . 15534693
Broman KW 2005 The genomes of recombinant inbred lines Genetics 169 1133 1146 15545647
Williams RW Gu J Qi S Lu L 2001 The genetic structure of recombinant inbred mice: High-resolution consensus maps for complex trait analysis Genome Biol 2 RESEARCH0046 11737945
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
Su AI Cooke MP Ching KA Hakak Y Walker JR 2002 Large-scale analysis of the human and mouse transcriptomes Proc Natl Acad Sci U S A 99 4465 4470 11904358
Oster SK Ho CS Soucie EL Penn LZ 2002 The myc oncogene: MarvelouslY Complex Adv Cancer Res 84 81 154 11885563
Guo QM Malek RL Kim S Chiao C He M 2000 Identification of c-myc responsive genes using rat cDNA microarray Cancer Res 60 5922 5928 11085504
Reed JC Doctor K Rojas A Zapata JM Stehlik C 2003 Comparative analysis of apoptosis and inflammation genes of mice and humans Genome Res 13 1376 1388 12819136
Guenet JL 1985 Do non-linked genes really reassort at random? Ann Inst Pasteur Immunol 136C 85 90 3994305
Jeong H Mason SP Barabasi AL Oltvai ZN 2001 Lethality and centrality in protein networks Nature 411 41 42 11333967
Wagner A 2001 The yeast protein interaction network evolves rapidly and contains few redundant duplicate genes Mol Biol Evol 18 1283 1292 11420367
Alon U 2003 Biological networks: The tinkerer as an engineer Science 301 1866 1867 14512615
Bray D 2003 Molecular networks: The top-down view Science 301 1864 1865 14512614
Wang N Akey JM Zhang K Chakraborty R Jin L 2002 Distribution of recombination crossovers and the origin of haplotype blocks: The interplay of population history, recombination, and mutation Am J Hum Genet 71 1227 1234 12384857
Yalcin B Fullerton J Miller S Keays DA Brady S 2004 Unexpected complexity in the haplotypes of commonly used inbred strains of laboratory mice Proc Natl Acad Sci U S A 101 9734 9739 15210992
Phillips MS Lawrence R Sachidanandam R Morris AP Balding DJ 2003 Chromosome-wide distribution of haplotype blocks and the role of recombination hot spots Nat Genet 33 382 387 12590262
McVean GA Myers SR Hunt S Deloukas P Bentley DR 2004 The fine-scale structure of recombination rate variation in the human genome Science 304 581 584 15105499
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
Dawson E Abecasis GR Bumpstead S Chen Y Hunt S 2002 A first-generation linkage disequilibrium map of human chromosome 22 Nature 418 544 548 12110843
Yalcin B Fullerton J Miller S Keays DA Brady S 2004 Unexpected complexity in the haplotypes of commonly used inbred strains of laboratory mice Proc Natl Acad Sci U S A 9734 9739 15210992
Russell AI Cunninghame Graham DS Shepherd C Roberton CA Whittaker J 2004 Polymorphism at the C-reactive protein locus influences gene expression and predisposes to systemic lupus erythematosus Hum Mol Genet 13 137 147 14645206
Bourque G Pevzner PA Tesler G 2004 Reconstructing the genomic architecture of ancestral mammals: Lessons from human, mouse, and rat genomes Genome Res 14 507 516 15059991
Hinds DA Stuve LL Nilsen GB Halperin E Eskin E 2005 Whole-genome patterns of common DNA variation in three human populations Science 307 1072 1079 15718463
Smith TF Waterman MS 1981 Identification of common molecular subsequences J Mol Biol 147 195 197 7265238
Morra M Simarro-Grande M Martin M Chen AS Lanyi A 2001 Characterization of SH2D1A missense mutations identified in X-linked lymphoproliferative disease patients J Biol Chem 276 36809 36816 11477068
Czar MJ Kersh EN Mijares LA Lanier G Lewis J 2001 Altered lymphocyte responses and cytokine production in mice deficient in the X-linked lymphoproliferative disease gene SH2D1A/DSHP/SAP Proc Natl Acad Sci U S A 98 7449 7454 11404475
Bharadwaj D Mold C Markham E Du Clos TW 2001 Serum amyloid P component binds to Fc gamma receptors and opsonizes particles for phagocytosis J Immunol 166 6735 6741 11359830
Christner RB Mortensen RF 1994 Specificity of the binding interaction between human serum amyloid P-component and immobilized human C-reactive protein J Biol Chem 269 9760 9766 8144569
Jurata LW Gill GN 1997 Functional analysis of the nuclear LIM domain interactor NLI Mol Cell Biol 17 5688 5698 9315627
Kaneto H Sharma A Suzuma K Laybutt DR Xu G 2002 Induction of c-Myc expression suppresses insulin gene transcription by inhibiting NeuroD/BETA2-mediated transcriptional activation J Biol Chem 277 12998 13006 11799123
Shiio Y Donohoe S Yi EC Goodlett DR Aebersold R 2002 Quantitative proteomic analysis of Myc oncoprotein function EMBO J 21 5088 5096 12356725
Sarkar S Leaman DW Gupta S Sil P Young D 2004 Cardiac overexpression of myotrophin triggers myocardial hypertrophy and heart failure in transgenic mice J Biol Chem 279 20422 20434 14970239
Trencia A Perfetti A Cassese A Vigliotta G Miele C 2003 Protein kinase B/Akt binds and phosphorylates PED/PEA-15, stabilizing its antiapoptotic action Mol Cell Biol 23 4511 4521 12808093
|
16163395
|
PMC1201368
|
CC BY
|
2021-01-05 08:00:59
|
no
|
PLoS Genet. 2005 Sep 9; 1(3):e33
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010033
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1617041010.1371/journal.pgen.001003405-PLGE-RA-0124R2plge-01-03-05Research ArticleBioinformatics - Computational BiologyEvolutionGenetics/GenomicsGenetics/Comparative GenomicsGenetics/Gene ExpressionMammalsProtein Modularity of Alternatively Spliced Exons Is Associated with Tissue-Specific Regulation of Alternative Splicing Tissue-Switched Alternative SplicingXing Yi Lee Christopher J *Molecular Biology Institute, and Institute for Genomics and Proteomics, Department of Chemistry and Biochemistry, University of California, Los Angeles, California, United States of AmericaGibson Greg EditorNorth Carolina State University, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 16 9 2005 1 3 e342 6 2005 2 8 2005 Copyright: © 2005 Xing and Lee.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.Recent comparative genomic analysis of alternative splicing has shown that protein modularity is an important criterion for functional alternative splicing events. Exons that are alternatively spliced in multiple organisms are much more likely to be an exact multiple of 3 nt in length, representing a class of “modular” exons that can be inserted or removed from the transcripts without affecting the rest of the protein. To understand the precise roles of these modular exons, in this paper we have analyzed microarray data for 3,126 alternatively spliced exons across ten mouse tissues generated by Pan and coworkers. We show that modular exons are strongly associated with tissue-specific regulation of alternative splicing. Exons that are alternatively spliced at uniformly high transcript inclusion levels or uniformly low levels show no preference for protein modularity. In contrast, alternatively spliced exons with dramatic changes of inclusion levels across mouse tissues (referred to as “tissue-switched” exons) are both strikingly biased to be modular and are strongly conserved between human and mouse. The analysis of different subsets of tissue-switched exons shows that the increased protein modularity cannot be explained by the overall exon inclusion level, but is specifically associated with tissue-switched alternative splicing.
Synopsis
Alternative splicing is a biological process that generates multiple mRNA and protein variants through alternative combinations of protein-coding exons. It is a widespread mechanism of gene regulation in higher eukaryotes. In recent years, scientists have found that when an exon is observed to be alternatively spliced in multiple species, its length is much more likely to be an exact multiple of three nucleotides. Since each amino acid is encoded by three nucleotides, these exons can be inserted or removed from the transcript as a “modular” protein-coding unit, without affecting the downstream protein translation. However, the precise roles of these modular exons in gene regulation and genome evolution remain unclear.
Xing and Lee have now investigated these modular exons using high-throughput genomics data. They analyzed the mouse splicing microarray data from the research group of Dr. Benjamin Blencowe at University of Toronto. Exons whose alternative splicing levels vary dramatically across multiple tissues are much more likely to be modular exons and are highly conserved during human and mouse evolution. This study establishes a strong link between protein modularity of alternatively spliced exons and tissue-specific regulation of alternative splicing. It provides new insights into the function and regulation of alternative splicing and how it evolves.
Citation:Xing Y, Lee CJ (2005) Protein modularity of alternatively spliced exons is associated with tissue-specific regulation of alternative splicing. PLoS Genet 1(3): e34.
==== Body
Introduction
Recently, there has been great interest in characterizing the functional selection pressures for alternative splicing by evolutionary genomics [1–4]. Ancestral alternative splicing events (i.e., alternative splicing events observed in multiple organisms) show evidence for strong functional constraints: these exons are more likely to be multiples of 3 nt in length [4–6], so the inclusion or exclusion of an exon does not disrupt the downstream protein reading frame or cause premature protein truncation; they are often flanked by highly conserved intronic sequences, suggesting increased selection for preserving important splicing regulatory signals [7,8]; the exon sequences are more conserved, indicated by increased nucleotide sequence identity [8–10]. These features have been explored extensively to discern functional alternative splicing events from splice variants generated by random spliceosomal errors and have been used successfully for predicting alternative splicing from raw genomic sequences [9,11,12].
In many of these studies, protein reading frame preservation has emerged as a valuable criterion for “functional” alternative splicing events [4,5,9,11], and it is interesting to ask what the significance of this “modular” class of exons is: that is, what general role they play in regulating the proteome and what distinguishes them from other alternative splicing events. Clearly, by maintaining the same protein reading frame regardless of whether the alternatively spliced exon is included or skipped, they enable a modular segment of amino acid sequence to be added or deleted from the protein product without altering the rest of the protein or inducing nonsense-mediated decay [13]. This modular pattern is not seen in constitutive exons and in alternatively spliced exons that are included in the majority of a gene's transcripts. Instead, it is strikingly associated with ancestral alternative splicing events (i.e., exons that are observed to be alternatively spliced in two or more species), most particularly those exons that are only included in a minority of a gene's transcripts [5]. However, the precise role of these modular alternative splicing events in genome evolution remains unclear.
To better characterize this interesting class of exons, we have analyzed microarray data for a large set of mouse genes containing 3,126 alternatively spliced exons, generated by Pan and coworkers [14]. Whereas previous analyses of modular exons used expressed sequence tag (EST) data to estimate each exon's “inclusion level” (fraction of the gene's transcripts that include that exon) as a single value summed over all tissue types (often based on only a small number of EST counts [3]), these microarray data permit accurate measurements of its varying inclusion levels in ten different tissue types [14]. This opens up the interesting question of tissue-specific regulation of alternative splicing [15,16].
Exons that are of particular interest in terms of tissue variation of alternative splicing are those that are strongly included in the transcripts in some tissues, but also are strongly excluded in some other tissues—suggesting that the alternative splicing of these exons is under significant change across different tissues and shows strong tissue specificity. In this paper, we refer to these exons as “tissue-switched” exons. We show that modular exons are strongly associated with tissue-switched alternative splicing. Comparing the exon sequences between human and mouse genomes, we demonstrate that many tissue-switched alternative exons are ancient and were under elevated selection pressure for both protein modularity (frame preservation) and sequence conservation during the recent mammalian evolution.
Results
Protein Reading Frame Preservation Is Associated with Tissue-Switched Exons
Using the mouse microarray data of Pan and colleagues [14], we analyzed the exon inclusion levels across ten tissues for 3,126 alternatively spliced exons in mouse. From the total set, 2,171 exons were assigned with a confident inclusion level in at least three tissues. We identified a total of 237 (11%) tissue-switched alternative exons according to our criteria (see Defining Categories of Tissue-Switched Exons from Microarray Data, in Materials and Methods), 605 exons that were major-form in all the tissues (always major), and 120 exons that were minor-form in all the tissues (always minor) (Table 1). These 962 exons were included in our further analyses. The size of our dataset is comparable to or substantially larger than a few other recent human–mouse comparative studies of alternative splicing [6,7,10,17,18], with information about these exons' splicing patterns at a much higher resolution (on average a confident inclusion level in seven tissues) compared to EST-based studies [3,5].
Table 1 Frame Preservation Analysis for the Total and Conserved Sets of Always Major, Tissue-Switched, and Always Minor Exons
Examining the frame-preservation ratio for these exons, we observed that only tissue-switched exons had an overall association with protein frame preservation. Previous studies by Gilbert and colleagues [19,20] have shown that exons in the human genome are slightly more likely to be frame-preserving than expected by the random chance (0.5), yielding a background frame-preservation ratio of 0.64 (i.e., 39% of exons in the genome are frame-preserving [5]). The frame-preservation ratios measured for always major and always minor alternatively spliced exons were 0.68 and 0.69, respectively, almost the same as the background ratio observed for constitutive exons. By contrast, the frame-preservation ratio for tissue-switched exons nearly doubled (Table 1), a statistically significant result (p < 0.001 for tissue-switched exons versus always major exons; p = 0.017 for tissue-switched exons versus always minor exons; one-sided Fisher exact test).
Tissue-Switched Exons Are Strongly Conserved
Since frame preservation has been observed to be associated primarily with ancestral alternative splicing events (i.e., alternative splicing events that have been observed in more than one species [5]), we tested the conservation of these exons between the mouse and human genomes (Table 1). Whereas always minor exons were conserved in only 10% of cases (in agreement with previous analyses [3,14]), tissue-switched exons showed a high rate of conservation (54%) similar to that of always major exons (64%). To control for the effect of exon length on our BLAST search, we restricted our analysis to a set of exons longer than 90 nt, and obtained similar results (data not shown).
To assess whether this pattern of conservation simply reflects the overall inclusion level of an exon (summed over all tissues), we further subdivided tissue-switched exons into three classes: usually major, observed to be the major-form in the majority of tissues; usually minor, observed to be a minor form in the majority of tissues; intermediate, observed to be neither the major-form in the majority of tissues, nor minor form in the majority of tissues. It should be emphasized that by definition all tissue-switched exons were major-form in at least one tissue and minor form in at least one other tissue. Among the 237 tissue-switched exons, we identified 40 usually major exons, and 37 usually minor exons by these criteria.
Analysis of the conservation of these exon types shows that tissue-switched alternative splicing, and not just a high overall inclusion level, is associated with a high rate of conservation (Figure 1A). Whereas always minor exons had a low rate of conservation (10%), usually minor exons had a high rate of conservation (62.2%) similar to that of always major exons (64%), despite the fact that they had a low overall inclusion level, only marginally higher than that of always minor exons. Overall, all types of tissue-switched exons had high rates of conservation similar to those of always major and constitutive exons.
Figure 1 Increased Functional Constraints Are Associated with All Types of Tissue-Switched Exons
(A) The fraction of exons conserved between human and mouse for each exon category. All types of tissue-switched exons (usually major, intermediate, usually minor) showed a high rate of conservation.
(B) The frame-preservation ratio for each category of exons. Protein reading frame preservation is associated with all types of tissue-switched exons and conserved, always minor exons. Conserved, always major exons showed no evidence for increased protein reading frame preservation.
(C) The percent nucleotide substitution density for each exon category. All types of tissue-switched exons showed a reduced nucleotide-sequence substitution density between human and mouse, similar to what is observed for conserved, always minor exons. Error bars represent 95% confidence intervals obtained from nonparametric bootstrapping.
(D) The nucleotide substitution rate calculated for synonymous sites (Ks) from human and mouse orthologous exon sequences (see Materials and Methods). All types of tissue-switched exons showed reduced Ks rates between human and mouse, similar to what is observed for conserved, always minor exons. Error bars represent 95% confidence intervals obtained from nonparametric bootstrapping.
Protein Frame Preservation Is Strongly Associated with All Types of Conserved Tissue-Switched Exons
We evaluated the frame-preservation ratio for conserved tissue-switched exons, subdivided by different classes (Figure 1B). These data show that frame preservation is strongly associated with all the different types of tissue-switched exons. Intermediate and usually minor exons had a 3-fold higher frame-preservation ratio than always major exons (p < 0.001 for intermediate versus always major exons; p = 0.009 for usually minor versus always major exons; one-sided Fisher exact test). The largest increase was observed in usually major exons (a 7-fold increase; p < 0.001 for usually major versus always major exons), which differ in the microarray data only slightly from always major exons (indeed, observation of the exon to be minor-form in only a single tissue sample is sufficient to move it from the always major to the usually major category). These data suggest that frame preservation, like conservation, cannot be explained by the overall inclusion level, but instead is strongly associated with tissue-switched exons. Although 90% of the always minor exons were not conserved between human and mouse, those that were conserved also had a high frame-preservation ratio, consistent with previous studies using EST data [5].
Unusually high sequence identity in conserved exons has been observed to be another valuable indicator of functional alternative splicing [8–11], indicative of the presence of splicing regulatory elements within the exon sequence. We therefore examined the level of sequence identity within the different classes of conserved exons. Tissue-switched exons displayed a dramatic decrease in the density of nucleotide substitutions compared with always major exons, and similar to what is observed in a small number of conserved always minor exons (Figure 1C). To assess whether this difference might be attributable to amino acid-level selection pressure, we measured the nucleotide substitution density specifically at synonymous sites (where substitutions cause no change to the amino acid sequence). Tissue-switched exons displayed a greater than 2-fold decrease in substitution density (relative to always major exons), even at synonymous sites (Figure 1D). These data indicate that the protein-level functional selection pressure demonstrated by frame preservation is accompanied in tissue-switched exons by an additional selection effect that cannot be explained by amino acid selection, consistent with an increased abundance of splicing regulatory elements as previously demonstrated [8,21,22].
Discussion
Frame-preserving alternative splicing events are of great functional interest because they produce a modular alteration of the protein product—adding or removing a single peptide segment without altering the rest of the protein sequence. Frame preservation has been proposed as evidence that an alternative splicing event is functional [4–6] and has proved valuable for predicting which exons in a genomic sequence are likely to be alternatively spliced [9,11,12]. Such alternative splicing events can have surprisingly sophisticated effects on protein structure, protein interactions, and function, as illustrated recently in the Piccolo C2A domain [23].
Our data suggest that this pattern of modular alternative splicing is strongly associated with tissue-switched exons. Analysis of microarray data from ten mouse tissues indicates that tissue-switched exons have the highest frame-preservation ratio, even for relatively subtle tissue-switching events. For example, whereas always major exons had a frame-preservation ratio near background (i.e., the ratio for constitutive exons in the mouse genome), exons that were usually major but observed to become the minor form in at least one tissue showed a 7-fold increase in frame preservation. Overall, the vast majority of nonrandom frame-preservation events (i.e., those above the number expected by chance) displayed tissue-switched alternative splicing even in the small panel of tissues (ten) analyzed here.
We have performed several control tests to evaluate the possibility of bias or artifacts due to the confidence-rank cutoff (recommended by Pan and colleagues) and classifiers (e.g., inclusion-level cutoffs for major versus minor form) applied to the dataset. Pan et al. recommended a cutoff of top-16,000 confidence ranks to identify confident exon inclusion levels [14]. To further exclude possible artifacts due to noise in the microarray experiment, we tested a more stringent filtering criterion (a cutoff of top-10,000 confidence ranks). These data robustly reproduced our original results. We also tested several different inclusion-level cutoffs for defining major versus minor forms (60% versus 40%, 66% versus 34%, and 75% versus 25%). These different cutoffs yielded consistent results. These control analyses demonstrate that our results are robust and are not artifacts of microarray noise or arbitrary cutoff values.
Tissue-switched exons combine several interesting features. On the one hand, they are strongly conserved, like constitutive and always major exons. Even usually minor tissue-switched exons showed a high frequency of conservation, similar to that of always major exons. On the other hand, tissue-switched exons display strong patterns of functional selection characteristic of ancestral minor-form alternative splicing, including strong frame preservation and reduced nucleotide substitution density. Even usually major exons had a frame-preservation ratio seven times that of always major exons, and a nucleotide substitution density 2-fold less than that of always major exons. This reduced level of substitution cannot be attributed to amino acid selection pressure, since it is also observed at synonymous codon positions. The fact that this pattern is observed specifically in tissue-switched exons suggests that it may reflect the presence of conserved regulatory motifs important for tissue-specific regulation of alternative splicing [24,25]. Increased sequence identity at alternatively spliced exons compared to constitutive exons has been reported as a predictive characteristic of alternative splicing [9,11,12] and has been shown to be associated with exonic splicing enhancer and splicing silencer sites [7,8,10,22]. This pattern of reduced nucleotide substitution appears to correlate quantitatively with the overall inclusion level for each exon (Figure 1C and 1D). Usually major exons had a synonymous substitution density of 0.223, intermediate exons 0.20, and always minor exons 0.06. This implies that restriction of an exon's expression to fewer and fewer tissues may require more regulatory sites.
These data also suggest several questions about ancestral alternative splice forms previously characterized by many groups [4–6,10,17,26]. Ancestral alternative splicing events (defined as alternative splicing observed in more than one species, and thus likely to be inherited from the common ancestor) show a similar profile of strong frame preservation [4,5], particularly for ancestral minor-form exons [5]. While the previous analysis pooled ESTs from all tissues to estimate the overall inclusion level of an alternative exon [5], in this study we used microarray data to obtain tissue-specific inclusion levels measured in ten different mouse tissues. These data show that it is not simply the overall inclusion level, but, more importantly, tissue-switched regulation of alternative splicing that is highly correlated with protein modularity. Therefore, our study reveals a strong link between protein modularity of alternative exons and tissue-specific regulation of alternative splicing. Consistent with previous studies, we observed that a small fraction of always minor mouse exons were conserved in the human genome and had a high frame-preservation ratio (3.0). One obvious question is whether many of these apparently always minor exons might actually be tissue-specific. Since only a small panel of ten mouse tissues was analyzed in the Pan et al. [14] microarray data, it is possible that some of these exons might be expressed as a major-form in other mouse tissues or individual cell types. A second possibility is that they are associated with transient regulatory events (i.e., a specific cellular activation state), rather than an individual tissue. Finally, the fact that these exons are conserved between human and mouse and have a high frame-preservation ratio similar to that previously reported for ancestral alternative splicing events [5] suggests that they may also be alternatively spliced in other species (such as human).
A minority of tissue-switched exons was not conserved between mouse and human, and these exons did not exhibit an elevated frame-preservation ratio. This raises several questions. What is the function of these mouse-specific tissue-switched exons, and why do they not show a bias for protein frame preservation as is seen in the conserved tissue-switched exons? One possibility is that the evolution of frame preservation for a given exon may be a slow process, so recently created, mouse-specific exons might not have had time to be converted in substantial numbers. Another possibility is that some of these exons may regulate function by inducing nonsense-mediated decay, as has been proposed by Brenner and colleagues [13,27].
Materials and Methods
Identification of alternative exons from mouse splicing microarray profile.
We identified tissue-switched alternative exons using data from a recent microarray analysis of alternative splicing in mouse [14]. Starting from 4,892 candidate alternative splicing events detected in ESTs, Pan and colleagues applied a set of filters to exclude errors and artifacts in the EST libraries. A total of 3,126 candidate alternative splicing events were included in their microarray design. Their dataset is thus a large and comprehensive collection of exon skipping events in the mouse genome, representing the vast majority of such events for which there was acceptable evidence. The exon inclusion level for these alternatively spliced exons was determined by microarray experiments across ten tissues [14]. Pan and colleagues assigned a confidence rank to each exon inclusion level, based on their statistical analyses of the splicing microarray data. According to their subsequent RT-PCR validation of the inclusion levels, they recommended a confidence rank of top 16,000 as a cutoff for confident exon inclusion levels. We followed this recommendation throughout this study, although our tests show that the use of a more stringent filter does not change our results significantly (see Discussion). We restricted our analysis to inclusion-level measurements within the top-16,000 confidence-rank cutoff and excluded exons with less than three tissue measurements meeting this criterion.
Defining categories of tissue-switched exons from microarray data.
An exon was defined as the major-form in a tissue if its inclusion level was greater than 66% in that tissue, or as the minor form if its inclusion level was less than 34% [3]. Because we were interested in the variations of the exon inclusion levels across multiple tissues, we referred to an exon as always major if it was the major-form in every tissue (with a confident exon inclusion level). Similarly, we referred to an exon as always minor if it was a minor-form exon everywhere. We defined an exon to be a tissue-switched exon if its inclusion level was higher than 66% in some tissues and less than 34% in other tissues. For tissue-switched exons, we further defined an exon as usually major if it was a major-form exon in the majority of the tissues. Similarly, we defined an exon as usually minor if it was a minor-form exon in the majority of the tissues. Finally, we defined a tissue-switched exon as intermediate if it was neither usually major nor usually minor.
Frame-preservation ratio analysis.
We defined an exon as frame-preserving if the length of the exon was a multiple of 3 nt, and as frame switching if not [5]. Inclusion or exclusion of a frame-preserving exon by alternative splicing leaves the downstream protein reading frame unchanged; for this reason, frame preservation has been proposed by several groups as evidence that an alternative splicing event is functional [5,6]. We calculated the frame-preservation ratio for a given set of exons as the number of frame-preserving exons divided by the number of frame-switching exons. We performed the Fisher exact test to assess whether the frame-preservation ratios for two groups of exons were significantly different.
Comparative analysis of tissue-switched exons in human and mouse genomes.
To determine whether an exon was conserved between human and mouse, we searched the human genome using nucleotide BLAST [28]. We defined an exon as conserved in another genome if we obtained a significant hit (BLAST expectation value less than 10−4) from BLASTN, aligning to the full length of the mouse exon, with no more than 12 nt deletion. It should be emphasized that this differs somewhat from the criteria of Modrek and Lee, whose dataset was constrained to the subset of genes where the exons adjacent to the alternatively spliced exon were successfully mapped to the orthologous gene in humans [3]. Since the dataset presented here lacks that extra constraint, it gives somewhat lower conservation estimates than Modrek & Lee and similar conservation estimates to Pan and colleagues [14].
For alternative exons conserved across genomes, we calculated their percent nucleotide sequence identity between human and mouse. We also calculated their rates of synonymous divergence (Ks), following the protocol of Nekrutenko and colleagues [29]. Briefly, orthologous exon sequences from human and mouse were translated and then aligned using CLUSTALW under default parameters [30]. This protein alignment was used to seed an alignment of corresponding nucleotide sequences, and gaps in the alignment were trimmed. We estimated the Ks rate from the codon-based nucleotide sequence alignment using the yn00 program of the PAML package [31,32]. This method takes into account the transition/transversion bias and codon usage bias for estimating Ks. We performed the Wilcoxon rank sum test to assess whether the nucleotide sequence identity or Ks rate for different groups of exons showed a statistically significant difference.
We wish to thank Q. Pan and B. Blencowe for providing mouse sequence data. We wish to thank B. Blencowe, A. Resch, M. Roy, and Q. Wang for their helpful discussions and comments. This work was supported by National Institutes of Health Grant U54-RR021813, a Teacher–Scholar award to CJL from the Dreyfus Foundation, and Department of Energy Grant DE-FC02-02ER63421. YX is supported by a University of California, Los Angeles Dissertation Year Fellowship.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. YX and CJL conceived and designed the experiments. YX performed the experiments. YX and CJL analyzed the data and wrote the paper.
Abbreviations
ESTexpressed sequence tag
Ksrate of synonymous divergence
==== Refs
References
Kan Z States D Gish W 2002 Selecting for functional alternative splices in ESTs Genome Res 12 1837 1845 12466287
Boue S Letunic I Bork P 2003 Alternative splicing and evolution Bioessays 25 1031 1034 14579243
Modrek B Lee C 2003 Alternative splicing in the human, mouse and rat genomes is associated with an increased rate of exon creation/loss Nat Genet 34 177 180 12730695
Thanaraj TA Clark F Muilu J 2003 Conservation of human alternative splice events in mouse Nucleic Acids Res 31 2544 2552 12736303
Resch A Xing Y Alekseyenko A Modrek B Lee C 2004 Evidence for a subpopulation of conserved alternative splicing events under selection pressure for protein reading frame preservation Nucleic Acids Res 32 1261 1269 14982953
Sorek R Shamir R Ast G 2004 How prevalent is functional alternative splicing in the human genome? Trends Genet 20 68 71 14746986
Sorek R Ast G 2003 Intronic sequences flanking alternatively spliced exons are conserved between human and mouse Genome Res 13 1631 1637 12840041
Fairbrother WG Holste D Burge CB Sharp PA 2004 Single nucleotide polymorphism-based validation of exonic splicing enhancers PLoS Biol 2 e268. DOI: 10.1371/journal.pbio.0020268 15340491
Sorek R Shemesh R Cohen Y Basechess O Ast G 2004 A non-EST-based method for exon-skipping prediction Genome Res 14 1617 1623 15289480
Itoh H Washio T Tomita M 2004 Computational comparative analyses of alternative splicing regulation using full-length cDNA of various eukaryotes RNA 10 1005 1018 15208437
Philipps DL Park JW Graveley BR 2004 A computational and experimental approach toward a priori identification of alternatively spliced exons RNA 10 1838 1844 15525709
Yeo GW Van Nostrand E Holste D Poggio T Burge CB 2005 Identification and analysis of alternative splicing events conserved in human and mouse Proc Natl Acad Sci U S A 102 2850 2855 15708978
Lewis BP Green RE Brenner SE 2003 Evidence for the widespread coupling of alternative splicing and nonsense-mediated mRNA decay in humans Proc Natl Acad Sci U S A 100 189 192 12502788
Pan Q Shai O Misquitta C Zhang W Saltzman AL 2004 Revealing global regulatory features of mammalian alternative splicing using a quantitative microarray platform Mol Cell 16 929 941 15610736
Xu Q Modrek B Lee C 2002 Genome-wide detection of tissue-specific alternative splicing in the human transcriptome Nucleic Acids Res 30 3754 3766 12202761
Yeo G Holste D Kreiman G Burge CB 2004 Variation in alternative splicing across human tissues Genome Biol 5 R74 15461793
Nurtdinov RN Artamonova II Mironov AA Gelfand MS 2003 Low conservation of alternative splicing patterns in the human and mouse genomes Hum Mol Genet 12 1313 1320 12761046
Kol G Lev-Maor G Ast G 2005 Human-mouse comparative analysis reveals that branch-site plasticity contributes to splicing regulation Hum Mol Genet 14 1559 1568 15857856
Long M Rosenberg C Gilbert W 1995 Intron phase correlations and the evolution of the intron/exon structure of genes Proc Natl Acad Sci U S A 92 12495 12499 8618928
de Souza SJ Long M Klein RJ Roy S Lin S 1998 Towards a resolution of the introns early/late debate: Only phase zero introns are correlated with the structure of ancient proteins Proc Natl Acad Sci U S A 95 5094 5098 9560234
Fairbrother WG Yeh RF Sharp PA Burge CB 2002 Predictive identification of exonic splicing enhancers in human genes Science 297 1007 1013 12114529
Wang Z Rolish ME Yeo G Tung V Mawson M 2004 Systematic identification and analysis of exonic splicing silencers Cell 119 831 845 15607979
Garcia J Gerber SH Sugita S Sudhof TC Rizo J 2004 A conformational switch in the Piccolo C2A domain regulated by alternative splicing Nat Struct Mol Biol 11 45 53 14718922
Han K Yeo G An P Burge CB Grabowski PJ 2005 A combinatorial code for splicing silencing: UAGG and GGGG motifs PLoS Biol 3 e158. DOI: 10.1371/journal.pbio.0030158 15828859
Minovitsky S Gee SL Schokrpur S Dubchak I Conboy JG 2005 The splicing regulatory element, UGCAUG, is phylogenetically and spatially conserved in introns that flank tissue-specific alternative exons Nucleic Acids Res 33 714 724 15691898
Pan Q Bakowski MA Morris Q Zhang W Frey BJ 2005 Alternative splicing of conserved exons is frequently species-specific in human and mouse Trends Genet 21 73 77 15661351
Hillman RT Green RE Brenner SE 2004 An unappreciated role for RNA surveillance Genome Biol 5 R8 14759258
Altschul SF Madden TL Schäffer AA Zhang J Zhang Z 1997 Gapped BLAST and PSI-BLAST: A new generation of protein database search programs Nucleic Acids Res 25 3389 3402 9254694
Nekrutenko A Chung WY Li WH 2003 ETOPE: Evolutionary test of predicted exons Nucleic Acids Res 31 3564 3567 12824366
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 Acids Res 22 4673 4680 7984417
Yang Z 1997 PAML: A program package for phylogenetic analysis by maximum likelihood Comput Appl Biosci 13 555 556 9367129
Yang Z Nielsen R 2000 Estimating synonymous and nonsynonymous substitution rates under realistic evolutionary models Mol Biol Evol 17 32 43 10666704
|
16170410
|
PMC1201369
|
CC BY
|
2021-01-05 08:00:57
|
no
|
PLoS Genet. 2005 Sep 16; 1(3):e34
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010034
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1617041110.1371/journal.pgen.001003505-PLGE-RA-0077R2plge-01-03-06Research ArticleEvolutionMolecular Biology - Structural BiologyGenetics/GenomicsGenetics/Comparative GenomicsPrimatesMammalsVertebratesAnimalsEukaryotesPervasive Adaptive Evolution in Primate Seminal Proteins Evolution of Primate Seminal ProteinsClark Nathaniel L *Swanson Willie J Department of Genome Sciences, University of Washington, Seattle, Washington, United States of AmericaGojobori Takashi EditorNational Institute of Genetics, Japan*To whom correspondence should be addressed. E-mail: [email protected] 2005 16 9 2005 1 3 e3513 4 2005 4 8 2005 Copyright: © 2005 Clark and Swanson.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.Seminal fluid proteins show striking effects on reproduction, involving manipulation of female behavior and physiology, mechanisms of sperm competition, and pathogen defense. Strong adaptive pressures are expected for such manifestations of sexual selection and host defense, but the extent of positive selection in seminal fluid proteins from divergent taxa is unknown. We identified adaptive evolution in primate seminal proteins using genomic resources in a tissue-specific study. We found extensive signatures of positive selection when comparing 161 human seminal fluid proteins and 2,858 prostate-expressed genes to those in chimpanzee. Seven of eight outstanding genes yielded statistically significant evidence of positive selection when analyzed in divergent primates. Functional clues were gained through divergent analysis, including several cases of species-specific loss of function in copulatory plug genes, and statistically significant spatial clustering of positively selected sites near the active site of kallikrein 2. This study reveals previously unidentified positive selection in seven primate seminal proteins, and when considered with findings in Drosophila, indicates that extensive positive selection is found in seminal fluid across divergent taxonomic groups.
Synopsis
Proteins found in seminal fluid accompanying sperm show dramatic effects on reproduction, such as manipulating female behavior. Even in primates they participate in competition between sperm of different males, and serve to protect sperm from infection by pathogens. These types of roles require the proteins to constantly adapt to stay ahead of the competition. Such adaptive pressures on proteins leave characteristic signatures in the DNA sequences that encode them. The authors used these signatures to identify adaptive evolution in primate seminal proteins and found extensive signs of adaptation when comparing thousands of seminal genes between human and chimpanzee. They further characterized outstanding genes in several primate species, including a diversity of apes and monkeys. Several of these proteins have no known function, yet by visualizing the adaptation on their three-dimensional surfaces, the authors uncovered clues to what is driving their evolution. In addition, they found several cases in which certain species lost their functional copies of these genes. Interestingly, species that showed loss of function do not participate in sperm competition. Past studies found widespread adaptation in fruit fly seminal fluid, and this study reveals extensive adaptation in primate seminal proteins. Could this be a phenomenon common among animals?
Citation:Clark NL, Swanson WJ (2005) Pervasive adaptive evolution in primate seminal proteins. PLoS Genet 1(3): e35.
==== Body
Introduction
Studies of adaptive evolution have revealed multiple classes of reproductive proteins under positive selection, including those involved in gamete recognition, seminal fluid factors, and proteins in the female reproductive tract [1–5]. The unknown pressures driving this adaptive evolution may be shared among taxonomic groups. For example, evidence of positive selection in gamete recognition proteins is found across divergent taxonomic groups, including mollusks, echinoderms, green algae, and mammals [1,4–7]. Positive selection in seminal proteins is observed in Drosophila and in primate semenogelin proteins [2,8–10]. However, the extent of selection in primates remains unknown, and it has not been determined whether seminal fluid proteins in divergent taxa experience such adaptive pressures.
Seminal fluid proteins in Drosophila initiate striking reproductive responses in females [11]. Inseminated proteins have been shown to affect sperm storage in the female reproductive tract, copulatory plug formation, ovulation, oogenesis, female receptivity to re-mating, and female lifespan. These can be important effects for sperm competition and sexual conflict, both of which may drive adaptive evolution. Additional seminal factors show antibacterial activity and may serve in pathogen defense, another adaptive driving force.
There is reason to believe that similar forces act on primate seminal fluid, leaving signatures of positive selection. As in drosophilids, several primate species form a post-mating copulatory plug, which could serve in sperm competition by excluding subsequent ejaculates from competing males. Plugs are present in diverse primates [12], including prosimians, New World and Old World monkeys, and in the chimpanzee, the closest living relative to humans. Consistent with adaptation, positive selection is seen in semenogelin proteins functioning in this pathway [9,10,13].
To identify positive selection in primate proteins, we used a measure of selective pressure, the d
N/d
S ratio [14,15]. Positive selection for amino acid diversification results in the rate of nonsynonymous substitutions exceeding that of synonymous substitutions. This effect is measured on coding sequences as the nonsynonymous substitution rate divided by the synonymous substitution rate— the d
N/d
S ratio. A value greater than one is indicative of positive selection, and a value less than one indicates purifying selection. In the absence of selection (neutral evolution), a value of one is expected. When measured over the entire length of a gene, the d
N/d
S ratio is a conservative measure of positive selection; in the presence of strong positive selection at some sites, conservation at others will lower the ratio. For this reason, when measured over the entire gene, we consider an elevated d
N/d
S to be suggestive of positive selection acting on a portion of the gene [3]. We then measure statistical significance using multiple species alignments through a likelihood method (CODEML) allowing different d
N/d
S values for codon sites [5,16,17]. Criticisms of this method include the argument that it gives false positive results under certain parameter combinations [18]. A more extensive study conducted by Wong et al. [19] found that this problem was limited to an early version of the program and to problems with convergence, and that the maximum likelihood method has good power and accuracy in detecting positive selection.
In this study we aimed to determine the extent of positive selection in primate seminal fluid proteins and to characterize outstanding candidates in further detail. Eight candidates were chosen from a pairwise human-chimpanzee d
N/d
S screen of seminal proteins. More detailed analysis with several species sequences provided strong evidence that positive selection acts on several primate seminal proteins.
Results
A Selective Pressure Screen
A list of proteins present in human seminal fluid was compiled from mass spectrometry studies of seminal plasma and prostasomes [20,21]. A total of 161 proteins were identified, 129 in prostasomes and 43 in seminal plasma, with 11 found in both studies. Human coding regions for these genes were aligned with chimpanzee orthologous sequences in order to estimate selective pressure between these two lineages as indicated by the d
N/d
S ratio. Estimates of pairwise d
N/d
S ratios were calculated using both the Nei and Gojobori method and a maximum likelihood method [14,16]. Both gave similar estimates (Table S1).
Rates of nonsynonymous versus synonymous substitution in these genes revealed several coding sequences with elevated d
N/d
S ratios (Figure 1A). Of 161 seminal fluid proteins, 17 had a d
N/d
S greater than one, and 36 greater than 0.5. The median d
N/d
S value was 0.19. A study of Drosophila seminal proteins showed similar variation of selective pressure (Figure 1B) [2]. Primate genes with elevated d
N/d
S ratios were involved in immune response (complement component 7, interleukin 1 receptor-like 2), semen coagulum (semenogelins I and II, prostate-specific transglutaminase 4, prostatic acid phosphatase), cellular structure (desmoglein 1, profilin I), and other roles, including several proteins of unknown function. Results for all 161 genes are shown in Table S1.
Figure 1 Plots of d
N Versus d
S for Primate and Drosophila Seminal Fluid Genes
(A) Genes encoding seminal fluid proteins identified by mass spectrometry in human versus chimpanzee.
(B) Drosophila simulans male-specific accessory gland genes versus D. melanogaster [2].
The diagonal represents neutral evolution, a d
N/d
S ratio of one. Most genes are subject to purifying selection and fall below the diagonal, while several genes fall above or near the line suggesting positive selection. Comparison of the two plots shows elevated d
N/d
S ratios in seminal fluid genes of both taxonomic groups.
Secreted proteins may tend to have higher d
N/d
S values, as they encounter adaptive pressures from exterior forces, such as interactions in the female reproductive tract. The subset of 43 proteins with secretion signal sequences showed a higher mean d
N/d
S (0.30) than those without (0.15). This difference is significant as determined by permutations (p = 0.0091). This 2-fold increase for secreted proteins was also observed in Drosophila seminal fluid [2].
Since the mass spectrometry studies are not expected to provide an exhaustive catalog of seminal fluid proteins, a similar screen was performed on prostate-expressed genes identified from an expression study of noncancerous human prostate [22]. Of 2,858 prostate-expressed genes, 290 showed a d
N/d
S greater than one, while the median value was 0.15. Secreted proteins again showed a higher mean d
N/d
S (0.24 versus 0.17, p = 0.00038).
The pairwise estimates were aimed at predicting candidate genes under selection, and eight were selected for in-depth analysis. Criteria used to select these candidates were a high d
N/d
S value, a high d
N, and evidence for high or specific prostate expression. Seven of eight candidates were taken from the mass spectrometry set because of the direct evidence that they are present in ejaculate. One gene, kallikrein 2 (KLK2), was taken from the prostate-expressed set, since it has a known role in seminal fluid dynamics. An additional gene, prostate-specific antigen (PSA), was analyzed due to its importance in copulatory plug dissolution, despite its low pairwise d
N/d
S value (Table 1). Since it was not chosen from the screen results, PSA is not considered a candidate. Overall, we chose three candidate genes involved in semen coagulation and five candidates whose functions are unknown.
Table 1 Seven Candidate Genes from the Screen Show Signs of Positive Selection
Positive Selection in Candidate Genes
To assess statistical significance of positive selection in candidate genes, we sequenced primate coding regions to provide eleven species sequences on average. We then assessed the selective pressure acting on these sequences using d
N/d
S ratios. Using a method that predicts a uniform d
N/d
S ratio across all codon sites, several pairwise comparisons of prolactin-induced protein (PIP) and β-microseminoprotein (MSMB) sequences have d
N/d
S ratios significantly greater than one, suggesting positive selection (unpublished data) [23]. This is a conservative approach, since it is unlikely that all codon sites are subject to the same selective pressure during evolution. More sensitive methods allow testing for variation in d
N/d
S at codon sites by comparing neutral models to selection models of codon evolution. Model parameters were estimated using a maximum likelihood method employed in the CODEML program of the PAML package [5,16,17]. For each gene, three different comparisons of neutral and selection models gave similar results (M1 versus M2, M7 versus M8, and M8A versus M8). From these comparisons, significant signs of positive selection were found in seven of eight candidate genes (Table 1). Since candidates were chosen based on high human-chimpanzee d
N/d
S values, there could be a statistical bias when sequences from the initial screen are included in the multiple alignments. When human and chimpanzee sequences were removed, six of the seven remained statistically significant, showing positive selection. The analysis that failed this conservative test, that of prostate-specific transglutaminase 4 (TGM4), may have suffered a lack of power, because the total tree length (0.47) was below optimal (~1) for this maximum likelihood method, due to the removal of two taxonomic groups [24].
The codon classes predicted to be under positive selection had d
N/d
S values ranging from 2 to 14 and were estimated to contain large proportions of codons for some genes (MSMB, PIP) and smaller proportions for others (TGM4) (Table 1). The rapid evolution of MSMB was noted in past studies of primate, rodent, and bird sequences [25,26], and was attributed to either low selective constraint or positive selection. We found highly significant signs of positive selection within primates (p < 0.001), with an estimated 42% of codons showing a d
N/d
S ratio of 2.90. Three diversified paralogs of the MSMB gene exist in New World monkeys [27], and their functions are unknown. When only Old World monkey and ape sequences are analyzed, significant positive selection is still observed (p = 0.029), and selection is predicted at similar codon sites.
We looked for lineage-specific variation in selective pressure by estimating d
N/d
S along phylogenetic lineages. For TGM4, a model estimating independent d
N/d
S ratios for each lineage fit the data better than a model with a uniform ratio (p = 0.0031). This indicates that variable selective pressure acted on TGM4 during its evolution, with branch-specific d
N/d
S values ranging from 0.1 to 1.95 (Figure 2). Prostatic acid phosphatase (ACPP) also shows significant variation in d
N/d
S, with elevated values in the chimpanzee and rhesus macaque lineages—1.16 and 0.64, respectively (p = 0.016). Finally, PSA does not have a high pairwise human-chimpanzee d
N/d
S, but it shows significant variation in selective pressure during its evolution. A branch model shows PSA lineages with d
N/d
S ratios exceeding one and was a significantly better fit than a model with uniform ratios for all lineages (p = 0.004). The extreme values in all three of these genes could be due to either positive selection or a reduction in functional constraint.
Figure 2 Variable Selective Pressure is Seen Between Lineages for Semen Coagulum Protein TGM4
This primate phylogeny shows selective pressure on TGM4 with estimated d
N/d
S ratios indicated on branches. Ratios greater than one are suggestive of either relaxed constraint or positive selection. Ratios are only shown for long branches, those with at least eight substitutions. A null model with a uniform d
N/d
S ratio across all lineages is rejected in favor of these estimates (p = 0.003). Branch lengths are estimated from TGM4 coding sequences. NWM, New World monkeys; OWM, Old World monkeys.
Spatial Distributions of Selected Sites on Three-Dimensional Structures
Positively-selected codon sites were predicted by a Bayes Empirical Bayes method for all genes showing significant positive selection [28]. Observed levels of divergence and number of sequences were appropriate for accurate prediction of sites according to a power analysis of Bayes prediction [29]. The spatial relationship of these selected sites was evaluated by mapping them onto three-dimensional protein structures. This analysis was done to find connections between positive selection and functional sites, because previous studies of MHC, lysin, and ZP3 proteins showed that predicted sites of positive selection fall into regions or binding clefts where diversification is biologically relevant [4,5]. We mapped selected sites onto five primate seminal proteins employing either solved crystal structures or threaded structural models, and used only predicted sites with a high level of support (p > 0.9). The spatial patterns and locations yielded intriguing patterns of positive selection.
Positively selected sites of the KLK2 protein fall near the active site residues and are found in known functional regions (Figure 3A). One selected site is in a known substrate binding cleft and two are in the kallikrein loop [30]. These locations and the pattern of clustering suggest that there was selective pressure for KLK2 to change substrate binding affinity. To assess statistical significance of surface clustering, we compared the mean pairwise distance between positively selected sites and a null distribution generated from randomly drawn surface (solvent exposed) sites (Figure 3B). Comparing the observed mean to the null distribution (10,000 permutations) lends statistical support to the hypothesis that these positively selected sites are clustered on the surface of KLK2 (p = 0.0043). The spatial distribution of selected sites on KLK2 provides an example of how positive selection can lead to inferences about evolution of protein function. In this case, a change in substrate is suggested.
Figure 3 Positive Selection at Sites Involved in Substrate Binding in KLK2
(A) Several amino acid sites predicted to be under positive selection (red) are near the protease active site (yellow). Three selected sites are found in known structural components of kallikrein proteins (light blue residues): Gly191 is part of the S1 substrate binding pocket, and His89 and Gln90 are part of the kallikrein loop [30]. Selected sites are labeled with the human residue on this threaded model.
(B) Positively selected sites are significantly clustered on the surface of KLK2. The observed mean pairwise distance between predicted positively selected sites is significantly lower than random sets of surface sites (p = 0.0043). This spatial clustering suggests that positive selection acted during KLK2 evolution to alter substrate binding.
MSMB is one of the most abundant proteins in human seminal plasma, yet its function remains unknown. It is also evolving rapidly, with an estimated 42% of codons under positive selection (Table 1). Positively selected sites are found all over the exterior of a threaded structure of MSMB (Figure 4), in contrast to the clustering seen on KLK2. When a clustering test is performed on MSMB selected sites, the observed mean distance falls just short of being significantly dispersed (p = 0.066). This dispersed pattern suggests that selection has acted uniformly on the surface of MSMB and no distinct functional regions can be inferred.
Figure 4 Positively Selected Sites on MSMB are Spread across the Protein Surface.
According to sites models of codon evolution, 42% of MSMB residues experienced adaptive pressure to alter their amino acids. Those predicted with high support are shown in red on this threaded structural model of human MSMB. Blue and purple residues demarcate two structural domains of the protein [62]. The amino acid sites show no clustering and are almost significantly dispersed throughout the protein (p = 0.066). This pattern is quite different from that shown by KLK2 (Figure 3). Although MSMB is one of the most abundant human seminal proteins, its function remains unknown.
Because few sites could be mapped onto structural models of transmembrane serine protease 2 (TMPRSS2), ACPP, and acyl-coA-binding protein (DBI), spatial distributions were less distinct, and clustering was not seen; however, some functional hypotheses may be made. Selected sites were predicted in two domains of TMPRSS2, the serine protease and the low-density lipoprotein receptor. When selected sites in the protease domain are mapped onto a threaded three-dimensional structure, they all occupy exterior positions on the same face, opposite of the protease active site. No interactions are confirmed in this region, but TMPRSS2 is thought to be activated through cleavage at a site located on this face [31]. The selective pressure on ACPP may be related to its substrate; a selected surface site (V77) neighbors two active site residues (R79 and H257) in the solved crystal structure [32]. Although this selected site was only moderately supported in Bayes Empirical Bayes analysis (p = 0.824), it is intriguing because of its proximity to the active site in an otherwise conserved region.
Structural analysis of selected sites and biochemical characterization are complementary approaches for elucidating the biological roles of these proteins. For example, our evidence of selection in KLK2 implies a testable change in substrate binding during primate evolution. As more coding sequences are determined, prediction of selected sites will improve, allowing site-specific selective pressures to be evaluated in functional contexts.
Loss of Function
Evidence for loss of function in several species was seen for two candidate genes, TGM4 and KLK2. Interestingly, both of these genes are involved primate semen coagulation. Prostate-specific TGM4 forms semen coagulum and copulatory plugs through cross-linking by its transglutaminase (TG) domain. Sequence from gorilla showed a homozygous, 11-basepair deletion in exon 7 at the start of the TG domain, a frameshift that would lead to early termination at amino acid 293 of 684. This deletion is likely fixed in gorilla populations, since four additional gorillas showed the same homozygous deletion. Abrogation of transglutaminase activity is likely since this exon contains ~20% of the TG domain and the remaining 80% falls downstream. Similarly, the sequenced Hylobates lar individual was homozygous for an early stop at codon position 411 downstream of the TG domain and before the first transglutaminase C-terminal domain.
In KLK2, the Macaca mulatta individual showed a homozygous change altering the active site residue D120 to alanine, which would eliminate proteolytic activity. This change was not seen in the four other Old World monkeys examined, including Macaca nigra. This suggests that abrogation of KLK2 activity occurred in those macaques closely related to M. mulatta or in that species alone.
Other evidence suggests loss of the KLK2 gene from gorilla and lesser apes. Although KLK2 sequence was obtained in several divergent New and Old World monkeys, only the first and last exons (1 and 5) were obtained from gorilla, despite several amplification conditions and primer combinations. When conditions were relaxed, PCR products from exons 2 through 4 of the paralog PSA were obtained instead, suggesting that KLK2 was lost in the gorilla lineage. Similar difficulties were encountered in all three analyzed species of genus Hylobates, suggesting a similar loss in lesser apes. PSA and KLK2 are paralogous genomic neighbors and likely arose through tandem duplication, so that unequal crossing-over could lead to deletion of one of the paralogs.
Discussion
Several seminal fluid proteins show dynamic evolutionary histories, significant positive selection, and variable selective pressure between lineages. Multiple instances of loss of function also hint at changing selective pressure. Seminal protein adaptation could result from several potential pressures, including sexual selection, pathogen response, and coevolution with changing binding partners and substrates. It is hypothesized that sexual selection, namely sperm competition and sexual conflict, is a major driving force behind the adaptive evolution of Drosophila seminal fluid proteins [33], and could be responsible for primate divergence as well.
Copulatory Plug Candidate Genes
In some primate species, the degree of semen coagulation is high enough to form a firm copulatory plug, a mechanism of sperm competition. Four prostate-specific candidate genes (TGM4, KLK2, PSA, and ACPP) participate in formation or dissolution of human seminal coagulum [34,35]. Significant positive selection is seen in TGM4, KLK2, and ACPP, along with significant variation in selective pressure between lineages for TGM4, ACPP, and PSA. Additionally, both of the genes showing loss of function participate in the formation (TGM4) or dissolution (KLK2) of semen coagulum. Loss of function of gorilla TGM4 is consistent with the lack of semen coagulation in gorilla [12] and with past evidence of early stop codons in alleles of gorilla semenogelins I and II [9,10]. Degeneration of semen coagulation may also be occurring in the lar gibbon, as its TGM4 coding sequence shows an early stop codon. Loss of semen coagulation is consistent with the mating systems of gorillas and gibbons, since both species are considered monoandrous, so males are not competitive postmating.
After a copulatory plug has been set, breaking it down is a strategy for competing males to win fertilizations. Positive selection seen in ACPP and KLK2 could be due to optimization of this function. KLK2 proteolytically activates PSA, a protease that breaks down semen coagulum. The likely loss of function of KLK2 observed in the rhesus macaque, gorilla, and lesser apes could result in reduced ability to dissolve semen coagulum. This change could reflect either lack of constraint for this function or adaptive value.
Conflict over Sperm Levels
Human seminal fluid factors, such as prostaglandin E, can locally suppress female immune response [36]. This function may be related to conflict between males and females over sperm levels. As sperm competition leads to higher sperm levels, chances of polyspermy increase, causing females to limit sperm numbers and strengthen barriers to fertilization. Candidate genes TGM4 and MSMB could serve to protect sperm from immune attack in the reproductive tract; evidence suggests that they both bind to sperm surfaces. TGM4 may deter attack by altering the sperm surface [37,38], and MSMB was found to be the main immunoglobulin binding factor in human seminal plasma [39,40]. Hence, these proteins may play roles in suppressing immune response against sperm, resulting in positive selection. Although highly expressed in human prostate, MSMB is also found in other mucous tissues [41], so a role in general pathogen defense must also be considered.
Pathogen Response
Like other secretions, seminal fluid contains protective antipathogenic factors. One candidate, PIP, has a likely role in host defense. PIP shows strong signs of positive selection, with 25% of codons estimated at a high d
N/d
S of 7.56. Notably, when just apes and one Old World monkey are analyzed, PIP shows highly significant positive selection (p = 0.00024). This secreted aspartyl proteinase is expressed at high levels in prostate and other exocrine glands. It is thought to play a role in host defense by binding bacteria, and it may suppress T-cell apoptosis [42,43]. Protection of sperm and the male reproductive tract from pathogens may also drive divergence of other seminal fluid proteins.
Antagonistic Pleiotropy
The major source of seminal fluid proteins is the prostate, a common site of male cancer. Disease research may benefit from studies of selection, since positive selection is often associated with human disease genes. In an analysis of 7,645 genes, those showing signs of positive selection were overrepresented in genes associated with disease in the Online Mendelian Inheritance in Man catalog [44]. In addition, Nielsen et al. found several genes involved in tumor suppression and apoptosis among those showing the strongest signs of positive selection between human and chimpanzee [45]. Also, the cancer susceptibility genes BRCA1 and angiogenin show signs of positive selection [46–49]. Such selection could result in antagonistic pleiotropy, a phenomenon in which adaptation in one respect brings deleterious effects in another. It is important to explore the possibility that adaptive evolution of seminal fluid factors contributes to disease through pleiotropic effects. Adaptation in prostate-expressed genes may benefit primates during their reproductive lifespan, but could lead to damaging side effects in later life.
Screen Utility
Overall, this human-chimpanzee selective pressure screen was successful in identifying seminal fluid genes with significant signs of positive selection. This is notable because the screen compared two closely related species with relatively few nucleotide differences per gene. Since seven of eight candidate genes showed statistically significant positive selection, we expect a fraction of other genes with elevated pairwise d
N/d
S ratios to be under positive selection.
Conclusion
The lower limit for primate seminal fluid proteins under positive selection is nine, from seven proteins in this study plus semenogelins I and II. Given the rate of support in this study, we speculate that there are others from the set of screened genes, as well as from genes not included in this screen. In conclusion, primate seminal fluid contains several proteins that exhibit dynamic evolutionary histories involving positive selection and loss of function. Extensive adaptive evolution in seminal proteins may be common to internally fertilizing taxa, since evidence of positive selection is seen in both Drosophila and primates.
Materials and Methods
Selective pressure screen.
A list of 161 proteins identified in human seminal fluid was compiled from mass spectrometry studies of seminal plasma and prostasomes [20,21]. Prostate-expressed genes were identified from an expression study of whole normal prostate (NCI CGAP Pr22) from the Prostate Expression Database (http://www.pedb.org) [22]. Of 4,277 unique ESTs from the study, 2,858 were traced to unique accession numbers in the reference sequence (RefSeq) database. Human exons encoding these seminal fluid proteins and prostate-expressed genes were retrieved from the UCSC Table Browser (http://genome.ucsc.edu/cgi-bin/hgTables). Each human exon was aligned to the best BLASTN hit over a threshold of 1 × 10−10 from chimpanzee whole genome shotgun contigs [50], and coding sequence alignments were created for evolutionary analysis.
Pairwise values of d
N/d
S for each human-chimpanzee coding sequence alignment were estimated by CODEML of the PAML package [17]. It was noted that some perceived substitutions resulted from poor-quality chimpanzee sequence, so substitution base calls in all gene candidates were manually verified in raw sequence chromatograms found in the sequence reads database [http://www.ncbi.nlm.nih.gov/Traces/trace.cgi/]). Presence of a signal sequence was predicted using SignalP [51]. Statistical significance of the difference between secreted versus nonsecreted d
N/d
S values was evaluated by a permutation test comparing the differences between average d
N/d
S values for random subsets through 100,000 permutations.
Sequencing of candidate genes.
Coding portions of nine genes were sequenced in divergent primates. Total DNA from the following primates were obtained from the Coriell Institute for Medical Research (Camden, New Jersey, United States): Pan troglodytes, P. paniscus, Gorilla gorilla, Pongo pygmaeus abelii, Macaca mulatta, M. nemestrina, Erythrocebus patas, Saguinus labiatus, and Ateles geoffroyi. DNA samples from the following species were obtained through the Integrated Primate Biomaterials and Information Resource (IPBIR; http://www.ipbir.org/ ): Hylobates gabriellae, H. lar, H. syndactylus, Cercopithecus cephus, and Papio anubis. Gorilla DNA samples were a generous gift from Evan Eichler at the University of Washington, Seattle, Washington, United States. Sequences of the following genes were obtained from GenBank: M. fascicularis PSA, Papio hamadryas MSMB, Saguinus oedipus MSMB, and M. fuscata PIP. Human coding sequences were taken from RefSeq entries in the UCSC genome browser (http://genome.ucsc.edu). Lemur KLK2 exons were retrieved from a BAC clone (GenBank accession number AC153325) sequenced by the NIH Intramural Sequencing Center (http://www.nisc.nih.gov/).
PCR was used to amplify exon-containing fragments from total DNA of various primates. PCR primers were designed from human introns, and clade-specific primers were designed when possible. PCR conditions and primer sequences are available from the authors upon request. Single-band PCR products were sequenced using Big Dye v.3.1 (Applied Biosystems, Foster City, California, United States). Sequence analysis was done using Phred, Phrap, and Consed [52,53]. High-quality sequence was used to generate coding sequences for each species based on human splice sites. Splice acceptor and donor sites were systematically checked for preservation of GT and AG nucleotides. Multiple alignments were made for each gene using ClustalW [54]. The close relationship between primates allowed for confident multiple alignments with few gaps. For estimation of d
N/d
S at sites or lineages, we removed secretion signal sequences and those species sequences showing loss of function.
Evolutionary analysis.
Phylogenetic relationships between the studied primates were taken from published studies [55–57]. Pairwise differences in d
N and d
S were calculated by MEGA version 3.0 [23], using a modified Nei-Gojobori (Jukes-Cantor) codon model with standard error computed analytically. Maximum likelihood evolutionary analysis was done with CODEML of the PAML 3.14 package [17], which estimates parameters for codon models of evolution. In order to ensure correct estimation of model parameters, we checked for convergence by running the optimization multiple times with different starting values of the omega parameter. The omega parameter estimates the d
N/d
S ratio and is used to determine selective pressure on codon sites, where a value greater than one is indicative of positive selection. Statistical significance is determined by a likelihood ratio test comparing a neutral model, where omega is limited to the interval (0, 1), to a selection model with an additional class of codons whose omega value is allowed to be greater than one. Different codon models were used for testing variation in d
N/d
S between sites; the models were compared as follows: (neutral to selection) M1 to M2, M7 to M8, and M8A to M8. All three comparisons gave similar results; we report those for M8A to M8 in Table 1. Model M1 (neutral) allows two classes of codons, one with omega over the interval (0,1) and the other with an omega value of one. Model M2 (selection) is similar to M1 except that it allows an additional class of codons with a freely estimated omega value. Model M7 (neutral) estimates omega with a beta-distribution over the interval (0, 1), while model M8 (selection) adds parameters to M7 for an additional class of codons with a freely estimated omega value. M8A (neutral) is a special case of M8 that fixes the additional codon class at an omega value of one [58]. Significance of positive selection found among codon sites was estimated both with and without the sequences from the initial screen (human and chimp). Such exclusion avoids a bias of selecting lineages from the screen with high numbers of nonsynonymous substitutions. When significant signs of positive selection were found, specific codon sites subjected to positive selection were predicted using a Bayes Empirical Bayes approach employed in CODEML [28]. Such an approach gives more reliable probability calculations than past methods, since it takes into account sampling errors in estimates of model parameters. To evaluate variation in selective pressure over a phylogeny, the branch model of CODEML estimated d
N/d
S values for each branch. The branch model is compared to the null hypothesis, model M0, in which all lineages have the same d
N/d
S value.
Structural analysis.
Threaded protein structures for KLK2, DBI, TMPRSS2, and MSMB were created with SwissModel using human primary sequence [59]. Structural analysis of ACPP sites was done on a solved crystal structure [32]. Protein structure images were produced using RasMol 2.7.2.1 [60].
Statistical significance of spatial clustering of amino acids was assessed by comparing the mean pairwise physical distance between positively selected sites to the mean distance between an equal number of random surface sites. A p-value was obtained by making this comparison 10,000 times. Surface sites were defined as those amino acids that are at least 20% solvent-exposed over their surface area. Solvent exposure was calculated using GETAREA 1.1 [61].
Supporting Information
Table S1 Selective Pressure Screen Comparing 161 Human and Chimpanzee Seminal Protein Genes
This table shows pairwise d
N/d
S estimates from two different methods.
(79 KB XLS)
Click here for additional data file.
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/) accession numbers of the genes discussed in this paper are M. fascicularis PSA (AY647976), Papio hamadryas MSMB (U49786), Saguinus oedipus MSMB (AJ010154, AJ010155, AJ010158), and M. fuscata PIP (AB098481).
The sequences generated in this study have been submitted to GenBank under accession numbers DQ150438 through DQ150526.
We would like to thank Dr. Evan Eichler's lab for providing quality gorilla and siamang DNA. We also thank Dr. Joshua Akey, Chris Saunders, Dr. Dick Hwang, Dr. Peter Nelson, and the members of the Swanson lab for valuable advice. We appreciate the helpful comments made by two anonymous reviewers. NLC is supported by the University of Washington National Institutes of Health Genetics Training Grant. WJS is supported by National Institutes of Health grant HD42563 and NSF grant #DEB 0410112.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. NLC and WJS conceived and designed the experiments. NLC performed the experiments. NLC analyzed the data. NLC and WJS contributed reagents/materials/analysis tools. NLC wrote the paper.
Abbreviations
TGtransglutaminase
==== Refs
References
Galindo BE Vacquier VD Swanson WJ 2003 Positive selection in the egg receptor for abalone sperm lysin Proc Natl Acad Sci U S A 100 4639 4643 12676995
Swanson WJ Clark AG Waldrip-Dail HM Wolfner MF Aquadro CF 2001 Evolutionary EST analysis identifies rapidly evolving male reproductive proteins in Drosophila
Proc Natl Acad Sci U S A 98 7375 7379 11404480
Swanson WJ Wong A Wolfner MF Aquadro CF 2004 Evolutionary expressed sequence tag analysis of Drosophila female reproductive tracts identifies genes subjected to positive selection Genetics 168 1457 1465 15579698
Swanson WJ Yang Z Wolfner MF Aquadro CF 2001 Positive Darwinian selection drives the evolution of several female reproductive proteins in mammals Proc Natl Acad Sci U S A 98 2509 2514 11226269
Yang Z Swanson WJ Vacquier VD 2000 Maximum-likelihood analysis of molecular adaptation in abalone sperm lysin reveals variable selective pressures among lineages and sites Mol Biol Evol 17 1446 1455 11018152
Ferris PJ Pavlovic C Fabry S Goodenough UW 1997 Rapid evolution of sex-related genes in Chlamydomonas
Proc Natl Acad Sci U S A 94 8634 8639 9238029
Metz EC Palumbi SR 1996 Positive selection and sequence rearrangements generate extensive polymorphism in the gamete recognition protein bindin Mol Biol Evol 13 397 406 8587504
Tsaur SC Wu CI 1997 Positive selection and the molecular evolution of a gene of male reproduction, Acp26Aa of Drosophila
Mol Biol Evol 14 544 549 9159932
Jensen-Seaman MI Li WH 2003 Evolution of the hominoid semenogelin genes, the major proteins of ejaculated semen J Mol Evol 57 261 270 14629036
Kingan SB Tatar M Rand DM 2003 Reduced polymorphism in the chimpanzee semen coagulating protein, semenogelin I J Mol Evol 57 159 169 14562960
Wolfner MF 2002 The gifts that keep on giving: Physiological functions and evolutionary dynamics of male seminal proteins in Drosophila
Heredity 88 85 93 11932766
Dixson AL Anderson MJ 2002 Sexual selection, seminal coagulation and copulatory plug formation in primates Folia Primatol (Basel) 73 63 69 12207054
Dorus S Evans PD Wyckoff GJ Choi SS Lahn BT 2004 Rate of molecular evolution of the seminal protein gene SEMG2 correlates with levels of female promiscuity Nat Genet 36 1326 1329 15531881
Nei M Gojobori T 1986 Simple methods for estimating the numbers of synonymous and nonsynonymous nucleotide substitutions Mol Biol Evol 3 418 426 3444411
Hughes AL Nei M 1988 Pattern of nucleotide substitution at major histocompatibility complex class I loci reveals overdominant selection Nature 335 167 170 3412472
Nielsen R Yang Z 1998 Likelihood models for detecting positively selected amino acid sites and applications to the HIV-1 envelope gene Genetics 148 929 936 9539414
Yang Z 1997 PAML: A program package for phylogenetic analysis by maximum likelihood Comput Appl Biosci 13 555 556 9367129
Suzuki Y Nei M 2004 False-positive selection identified by ML-based methods: Examples from the Sig1 gene of the diatom Thalassiosira weissflogii and the tax gene of a human T-cell lymphotropic virus Mol Biol Evol 21 914 921 15014169
Wong WS Yang Z Goldman N Nielsen R 2004 Accuracy and power of statistical methods for detecting adaptive evolution in protein coding sequences and for identifying positively selected sites Genetics 168 1041 1051 15514074
Fung KY Glode LM Green S Duncan MW 2004 A comprehensive characterization of the peptide and protein constituents of human seminal fluid Prostate 61 171 181 15305340
Utleg AG Yi EC Xie T Shannon P White JT 2003 Proteomic analysis of human prostasomes Prostate 56 150 161 12746840
Hawkins V Doll D Bumgarner R Smith T Abajian C 1999 PEDB: The Prostate Expression Database Nucleic Acids Res 27 204 208 9847181
Kumar S Tamura K Nei M 2004 MEGA3: Integrated software for Molecular Evolutionary Genetics Analysis and sequence alignment Brief Bioinform 5 150 163 15260895
Anisimova M Bielawski JP Yang Z 2001 Accuracy and power of the likelihood ratio test in detecting adaptive molecular evolution Mol Biol Evol 18 1585 1592 11470850
Nolet S St-Louis D Mbikay M Chretien M 1991 Rapid evolution of prostatic protein PSP94 suggested by sequence divergence between rhesus monkey and human cDNAs Genomics 9 775 777 2037304
Lazure C Villemure M Gauthier D Naude RJ Mbikay M 2001 Characterization of ostrich (Struthio camelus) beta-microseminoprotein (MSP): Identification of homologous sequences in EST databases and analysis of their evolution during speciation Protein Sci 10 2207 2218 11604528
Makinen M Valtonen-Andre C Lundwall A 1999 New World, but not Old World, monkeys carry several genes encoding beta-microseminoprotein Eur J Biochem 264 407 414 10491085
Yang Z Wong WS Nielsen R 2005 Bayes empirical Bayes inference of amino acid sites under positive selection Mol Biol Evol 22 1107 1118 15689528
Anisimova M Bielawski JP Yang Z 2002 Accuracy and power of Bayes prediction of amino acid sites under positive selection Mol Biol Evol 19 950 958 12032251
Laxmikanthan G Blaber SI Bernett MJ Scarisbrick IA Juliano MA 2005 1.70 Å X-ray structure of human apo kallikrein 1: Structural changes upon peptide inhibitor/substrate binding Proteins 58 802 814 15651049
Afar DE Vivanco I Hubert RS Kuo J Chen E 2001 Catalytic cleavage of the androgen-regulated TMPRSS2 protease results in its secretion by prostate and prostate cancer epithelia Cancer Res 61 1686 1692 11245484
Ortlund E LaCount MW Lebioda L 2003 Crystal structures of human prostatic acid phosphatase in complex with a phosphate ion and alpha-benzylaminobenzylphosphonic acid update the mechanistic picture and offer new insights into inhibitor design Biochemistry 42 383 389 12525165
Chapman T 2001 Seminal fluid-mediated fitness traits in Drosophila
Heredity 87 511 521 11869341
Balk SP Ko YJ Bubley GJ 2003 Biology of prostate-specific antigen J Clin Oncol 21 383 391 12525533
Brillard-Bourdet M Rehault S Juliano L Ferrer M Moreau T 2002 Amidolytic activity of prostatic acid phosphatase on human semenogelins and semenogelin-derived synthetic substrates Eur J Biochem 269 390 395 11784334
Kelly RW Critchley HO 1997 Immunomodulation by human seminal plasma: A benefit for spermatozoon and pathogen? Hum Reprod 12 2200 2207 9402282
Mukherjee DC Agrawal AK Manjunath R Mukherjee AB 1983 Suppression of epididymal sperm antigenicity in the rabbit by uteroglobin and transglutaminase in vitro Science 219 989 991 6130601
Paonessa G Metafora S Tajana G Abrescia P De Santis A 1984 Transglutaminase-mediated modifications of the rat sperm surface in vitro Science 226 852 855 6149619
Kamada M Mori H Maeda N Yamamoto S Kunimi K 1998 Beta-microseminoprotein/prostatic secretory protein is a member of immunoglobulin binding factor family Biochim Biophys Acta 1388 101 110 9774712
Hirano M Kamada M Maeda N Yamamoto S Aono T 1996 Presence of immunoglobulin binding factor on human sperm surface as sperm coating antigen Arch Androl 37 163 170 8939294
Weiber H Andersson C Murne A Rannevik G Lindstrom C 1990 Beta microseminoprotein is not a prostate-specific protein. Its identification in mucous glands and secretions Am J Pathol 137 593 603 2205099
Schenkels LC Walgreen-Weterings E Oomen LC Bolscher JG Veerman EC 1997 In vivo binding of the salivary glycoprotein EP-GP (identical to GCDFP-15) to oral and non-oral bacteria detection and identification of EP-GP binding species Biol Chem 378 83 88 9088536
Gaubin M, Autiero M, Basmaciogullari S, Metivier D, Mis hal Z, et al. 1999 Potent inhibition of CD4/TCR-mediated T cell apoptosis by a CD4-binding glycoprotein secreted from breast tumor and seminal vesicle cells J Immunol 162 2631 2638 10072505
Clark AG Glanowski S Nielsen R Thomas PD Kejariwal A 2003 Inferring nonneutral evolution from human-chimp-mouse orthologous gene trios Science 302 1960 1963 14671302
Nielsen R Bustamante C Clark AG Glanowski S Sackton TB 2005 A scan for positively selected genes in the genomes of humans and chimpanzees PLoS Biol 3 e170 15869325
Huttley GA Easteal S Southey MC Tesoriero A Giles GG 2000 Adaptive evolution of the tumour suppressor BRCA1 in humans and chimpanzees. Australian Breast Cancer Family Study Nat Genet 25 410 413 10932184
Zhang J Rosenberg HF 2002 Diversifying selection of the tumor-growth promoter angiogenin in primate evolution Mol Biol Evol 19 438 445 11919285
Pavlicek A Noskov VN Kouprina N Barrett JC Jurka J 2004 Evolution of the tumor suppressor BRCA1 locus in primates: Implications for cancer predisposition Hum Mol Genet 13 2737 2751 15385441
Fleming MA Potter JD Ramirez CJ Ostrander GK Ostrander EA 2003 Understanding missense mutations in the BRCA1 gene: An evolutionary approach Proc Natl Acad Sci U S A 100 1151 1156 12531920
Altschul SF Gish W Miller W Myers EW Lipman DJ 1990 Basic local alignment search tool J Mol Biol 215 403 410 2231712
Nielsen H Engelbrecht J Brunak S von Heijne G 1997 Identification of prokaryotic and eukaryotic signal peptides and prediction of their cleavage sites Protein Eng 10 1 6
Gordon D Abajian C Green P 1998 Consed: A graphical tool for sequence finishing Genome Res 8 195 202 9521923
Ewing B Hillier L Wendl MC Green P 1998 Base-calling of automated sequencer traces using phred. I. Accuracy assessment Genome Res 8 175 185 9521921
Chenna R Sugawara H Koike T Lopez R Gibson TJ 2003 Multiple sequence alignment with the Clustal series of programs Nucleic Acids Res 31 3497 3500 12824352
Purvis A 1995 A composite estimate of primate phylogeny Philos Trans R Soc Lond B Biol Sci 348 405 421 7480112
Muller S Hollatz M Wienberg J 2003 Chromosomal phylogeny and evolution of gibbons (Hylobatidae) Hum Genet 113 493 501 14569461
Hayasaka K Fujii K Horai S 1996 Molecular phylogeny of macaques: Implications of nucleotide sequences from an 896-base pair region of mitochondrial DNA Mol Biol Evol 13 1044 1053 8752012
Swanson WJ Nielsen R Yang Q 2003 Pervasive adaptive evolution in mammalian fertilization proteins Mol Biol Evol 20 18 20 12519901
Schwede T Kopp J Guex N Peitsch MC 2003 SWISS-MODEL: An automated protein homology-modeling server Nucleic Acids Res 31 3381 3385 12824332
Milner-White RSaEJ 1995 RasMol: Biomolecular graphics for all Trends Biochem Sci 20 374 7482707
Franzkiewicz R Braun W 1998 Exact and efficient analytical calculation of the accessible surface areas and their gradients for macromolecules J Comput Chem 19 319 333
Wang I Lou YC Wu KP Wu SH Chang WC 2005 Novel solution structure of porcine beta-microseminoprotein J Mol Biol 346 1071 1082 15701518
|
16170411
|
PMC1201370
|
CC BY
|
2021-01-05 08:00:58
|
no
|
PLoS Genet. 2005 Sep 16; 1(3):e35
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010035
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1617041210.1371/journal.pgen.001003605-PLGE-RA-0097R2plge-01-03-07Research ArticleInfectious DiseasesPharmacology - Drug DiscoveryGenetics/Gene DiscoveryGenetics/Disease ModelsYeast and FungiSaccharomycesYeast Model Uncovers Dual Roles of Mitochondria in the Action of Artemisinin Mode of Action of ArtemisininLi Wei 1Mo Weike 1Shen Dan 1Sun Libo 1Wang Juan 1Lu Shan 1Gitschier Jane M 2Zhou Bing 1*1 The State Key Laboratory of Biomembrane and Membrane Biotechnology, Department of Biological Sciences and Biotechnology, Tsinghua University, Beijing, China
2 Departments of Medicine and Pediatrics and Howard Hughes Medical Institute, University of California, San Francisco, California, United States of America
Dutcher Susan EditorWashington University, United States of America*To whom correspondence should be addressed: [email protected] 2005 16 9 2005 1 3 e369 5 2005 8 8 2005 Copyright: © 2005 Li, 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.Artemisinins, derived from the wormwood herb Artemisia annua, are the most potent antimalarial drugs currently available. Despite extensive research, the exact mode of action of artemisinins has not been established. Here we use yeast, Saccharamyces cerevisiae, to probe the core working mechanism of this class of antimalarial agents. We demonstrate that artemisinin's inhibitory effect is mediated by disrupting the normal function of mitochondria through depolarizing their membrane potential. Moreover, in a genetic study, we identify the electron transport chain as an important player in artemisinin's action: Deletion of NDE1 or NDI1, which encode mitochondrial NADH dehydrogenases, confers resistance to artemisinin, whereas overexpression of NDE1 or NDI1 dramatically increases sensitivity to artemisinin. Mutations or environmental conditions that affect electron transport also alter host's sensitivity to artemisinin. Sensitivity is partially restored when the Plasmodium falciparum NDI1 ortholog is expressed in yeast ndi1 strain. Finally, we showed that artemisinin's inhibitory effect is mediated by reactive oxygen species. Our results demonstrate that artemisinin's effect is primarily mediated through disruption of membrane potential by its interaction with the electron transport chain, resulting in dysfunctional mitochondria. We propose a dual role of mitochondria played during the action of artemisinin: the electron transport chain stimulates artemisinin's effect, most likely by activating it, and the mitochondria are subsequently damaged by the locally generated free radicals.
Synopsis
Malaria kills at least 1 million people worldwide a year. Recent years saw the rapid emergence of drug-resistant malaria strains. Artemisinins, derived from the Chinese wormwood herb Artemisia annua, are the most potent antimalarials currently available. Despite extensive research, the exact mode of action of artemisinins has not been established. In this article, Li et al. investigated yeast as a model to probe the core working mechanism of this class of antimalarials. They showed that artemisinin can disrupt the normal function of mitochondria by depolarizing its membrane potential, and that artemisinin's effect can be affected by its interaction with the mitochondrial electron transport chain, an apparatus that couples oxygen oxidation and energy generation in the cell. They proposed a dual role of mitochondria played during the action of artemisinin: the electron transport chain likely activates artemisinin, and the mitochondria are subsequently damaged by the locally generated free radicals associated with this activation. The research has provided a fine tool for the study of the mechanism of artemisinin in a model organism (yeast), and laid the framework for a set of possible future experiments to be conducted in yeast and malaria parasites.
Citation:Li W, Mo W, Shen D, Sun L, Wang J, et al (2005) Yeast model uncovers dual roles of mitochondria in the action of artemisinin. PLoS Genet 1(3): e36.
==== Body
Introduction
Malaria, the most prevalent and pernicious parasitic disease of humans, is estimated to kill between 1 million and 2 million people, mainly children, each year. Resistance has emerged to all classes of antimalarial drugs except artemisinin, an endoperoxide antimalarial drug derived from an ancient Chinese herbal remedy, qinghaosu. The mechanism of the action of artemisinin remains a mystery, although iron appears to be involved in activating this endoperoxide to generate cytotoxic free radicals [1]. Several candidates have been hypothesized as targets of artemisinins, including haem, a translationally controlled tumor protein and some parasite membrane proteins [1–3], but none of these have been convincingly shown to be functionally relevant. Recently, Eckstein-Ludwig et al. [4] proposed PfATP6, a sarco/endoplasmic reticulum Ca2+-ATPase, as artemisinin's target and inferred that artemisinin might act by mobilizing intracellular Ca2+ stores; however, this conclusion has also been debated [5].
One important reason for the lack of satisfying progress in understanding artemisinin is that it is difficult to carry out genetic analysis of malaria parasites. In contrast, the yeast Saccharamyces cerevisiae is an ideal model organism to uncover a variety of molecular mechanisms that otherwise might be difficult to address with other systems. In this study, we developed a yeast model and used it to probe the fundamental mechanisms of artemisinin's action.
Results
Artemisinin Inhibits Yeast Growth in Nonfermentable Media
In a pilot experiment, we tested whether artemisinin is toxic to the growth of yeast cells. As demonstrated in Figure 1A, the inhibitory effect of artemisinin on yeast growth in YPD (a fermentable medium with dextrose as the carbon source) liquid was not detectable at 8 μM, and only started to be observable at 80 μM, a concentration far above the level needed to inhibit the growth of Plasmodium falciparum (a 50% growth inhibition concentration [IC50] of approximately a few nM) [6,7]. On YPD plates, no significant inhibition of growth is even noticeable in the presence of 80 μM artemisinin.
Figure 1 Artemisinin Inhibits Yeast Respiratory Growth by Depolarizing the Mitochondrial Membrane
(A) Artemisinin (Art) inhibits yeast growth in nonfermentable media. In YPD the effect of artemisinin is minimal, whereas in YPG, artemisinin is highly effective.
(B) Yeast growth is inhibited by artemisinin in YPG with an IC50 that is comparable to that required to kill cultured malaria parasites. Relative growth in the presence of artemisinin was measured against to that of the yeast grown in the absence of artemisinin. Experiments shown were performed three times in liquid YPG media. Error bars represent standard errors of the mean for each assay.
(C) Artemisinin depolarizes mitochondrial membrane. The peak shift toward the left represents a decrease of fluorescence signal indicating the loss of membrane potential. Cells were grown in YPG with or without artemisinin (Art) for 2 h.
This level of insensitivity to artemisinin led us to consider whether use of the YPD media precluded our assessment of one important organelle, the mitochondrion, in the action of artimisinin. Yeast can grow either anaerobically or aerobically by utilizing fermentable or nonfermentable carbon sources. In the presence of a fermentable carbon source such as glucose (dextrose), as in YPD, yeast preferentially adopt glycolysis to generate energy even under aerobic conditions, and can grow normally even when mitochondrial respiration level is minimal. In order to determine whether artemisinin interferes with mitochondrial function, yeast was grown in YPG or YPE media in which glucose was replaced with the nonfermentable carbon source glycerol or ethanol. Amazingly, in YPG(E) media, yeast growth was severely inhibited when artemisinin was present. Figure 1A shows the dramatic reduction of cell growth in the presence of 8 μM. Indeed, for yeast grown in YPG, the IC50 (defined here as 50% growth inhibition in 48 h) was determined to be approximately 10 nM (Figure 1B), a concentration comparable to what is effective in P. falciparum. The dramatic contrast of sensitivity between yeast grown on artemisinin-supplemented YPD and YPG(E) media indicates that artemisinin affects mitochondrial function in YPG(E).
Artemisinin Depolarizes the Mitochondrial Membrane
The observation of growth inhibition of yeast on artemisinin-supplemented YPG(E) media prompted us to investigate whether the mitochondrial membrane potential of yeast is affected by artemisinin, because maintenance of the membrane potential is key to mitochondrial functions, including oxidative phosphorylation and metabolism of amino acids, lipids, and haem, as well as intracellular Ca2+ homeostasis. As shown in Figure 1C, treatment of yeast with artemisinin caused substantial loss of membrane potential, as evidenced by the decreased rodamine uptake into the mitochondria. This loss of membrane potential was also observed with yeast grown in YPD media, but required the presence of a slightly higher concentration of artemisinin or a longer treatment time.
Genetic Studies Reveal Electron Transport Chain Plays Important Roles in the Action of Artemisinin
To gain further insights into the mechanism of action of artemisinin, we screened for yeast genes that, when inactivated, can confer artemisinin resistance. A library comprised of 4,757 homozygous deletion strains [8,9] was plated on YPG(E) agar plates with 3–5 μM of artemisinin. Three genes whose deletion resulted in artemisinin resistance were identified. Two of these, NDE1 and NDI1, encode yeast NADH dehydrogenases in the mitochondrial electron transport chain [10–12]. ndi1Δ and nde1Δ displayed similar level of growth on artemisinin-supplemented YPGE plates (Figure 2A). Overexpression of NDE1 and NDI1 in their corresponding deletion strains or in the parental BY4742 background (wild type) resulted in significantly increased sensitivity to the drug (Figure 2B, first three rows). These data suggest that NADH dehydrogenase activity is positively correlated with yeast sensitivity to artemisinin.
Figure 2 The Genetic Screen for Artemisinin-Resistant Mutations Identified Genes in the Electron Transport Chain or in the Pathway of Respiratory Control
(A) The three mutants isolated display increased resistance to artemisinin. YPGE plates with or without 4 μM artemisinin were used. nde1Δ ndi1Δ exhibited severe growth defect in nonfermentable media.
(B) Increased activities of NADH dehydrogenases exacerbate artemisinin sensitivity, and Sip5 may be positioned upstream of NADH dehydrogenases. Plates are all SG-Ura (with or without 4 μM artemisinin) to prevent plasmid loss. ADH1-NDE1 and ADH1-SIP5 here denote constructs that express NDE1 and SIP5 under the control of ADH1 promoter. The results of ADH1-NDI1 are similar to that of ADH1-NDE1 and are not shown on the two plates.
(C) Expression of PfNDI1 in ndi1Δ restores yeast sensitivity to artemisinin. Plates used here are SG-Ura (with or without 8 μM artemisinin).
Art, artemisinin; SG, synthetic yeast media with glycerol as the carbon source; WT, wild type.
In addition to NDE1 and NDI1, a third homologous yeast gene, NDE2, is known to provide NADH dehydrogenase activity in yeast. These NADH dehydrogenases oxidize NADH at either the cytosolic side (Nde1p and Nde2p) or the matrix side (Ndi1p) of the inner mitochondrial membrane. Although deletion of NDE1 leads to a major loss of NADH dehydrogenase activity at the cytosolic side, deletion of both NDE1 and NDE2 results in complete loss of NADH dehydrogenase activity at the cytosolic side [11–13]. Because of sequence and function similarity between Nde1p, Ndi1p, and Nde2p, we asked whether NDE2 interacts with artemisinin. Compared to that of nde1Δ or ndi1Δ, nde2Δ manifested only marginal resistance to artemisinin, and the double mutant nde1Δ nde2Δ was not substantially more resistant to artemisinin than nde1Δ. This is consistent with the observation that at the cytosolic side Nde1p is the major NADH dehydrogenase and Nde2p only plays a minor role [10–12].
Next, we investigated whether the double mutant nde1Δ ndi1Δ is more resistant to the action of artemisinin or not. nde1Δ ndi1Δ yeast can grow well in glucose media. However, their growth on nonfermentable media is greatly affected (see Figure 2A), indicating severely compromised mitochondrial function of the strain. This growth defect precludes our comparative analysis of nde1Δ ndi1Δ resistance against artemisinin.
Although three NADH dehydrogenases are present in yeast, only one such dehydrogenase, PfNDI1 (PFI0735c), with closest homology to Ndi1p, was found in the P. falciparum genome database [13]. Similarly, Nde1p and Nde2p do not have counterparts in mammalian cells, which utilize redox shuttle mechanisms to couple cytosolic NADH oxidation to NADH dedydrogenase (complex I) in the matrix side. To test whether PfNDI1 expression in yeast confers artemisinin's sensitivity, PfNDI1 was amplified from a cDNA pool of P. falciparum. As expected, expression of PfNDI1 in ndi1Δ background partially restored its sensitivity to artemisinin (Figure 2C). This indicates that PfNDI1 expressed in yeast interacts with artemisinin in a similar fashion to Ndi1p.
The third artemisinin-resistant strain is sip5Δ (Figure 2A). Little is known about Sip5p except that it biochemically interacts with Snf1 protein kinase and Reg1/Glc7 protein phosphatase [14], both of which are involved in the regulation of alternative carbon source utilization and respiration. Thus, it is possible that SIP5 somehow indirectly affects energy generation, or, more specifically, mitochondrial respiration. To test the hypothesis that respiration control is involved in artemisinin resistance, mutants of seven other components involved in carbon-source regulation and respiration [15] were examined. Most of them are variably resistant to the action of artemisinin (see Materials and Methods).
The epistatic relationship of SIP5 with NDI1 or NDE1 was also explored. Overexpression of SIP5 in the control strain, ndi1Δ or nde1Δ did not have any effect on sensitivity or resistance to artemisinin. In contrast, when either NDI1 or NDE1 is overexpressed in sip5Δ background, yeast still exhibit hypersensitivity to artemisinin. These findings suggest that sip5Δ affects artemisinin resistance indirectly by acting on the electron transport chain. The identification of these three genes that are directly or indirectly involved in respiration reinforced our previous interpretation that artemisinin acts through mitochondria.
Because the electron transport chain is involved in the action of artemisinin, we asked further whether other components downstream of NADH dehydrogenase interact with artemisinin. We have checked available deletion mutants (25 strains altogether) of other known genes in the electron transport chain for possible artemisinin resistance. Although some of these strains cannot grow on nonfermentable media, of those that can, none appears to be resistant to the action of artemisinin (see Materials and Methods). Although it is likely some of these mutants have impaired electron transport activity, the fact that all 25 deletion strains do not manifest either artemisinin hypersensitivity or artemisinin resistance (compared to the wild-type strain) implies that not all components of the pathway interact with artemisinin.
Artemisinin's Action Generates Reactive Oxygen Species
Generation of oxidative stress is believed to be involved in the antiparasitic effects of artemisinin [16]. Figure 3A shows that the levels of reactive oxygen species (ROS) produced in three strains after treatment with artemisinin positively correlated with their levels of sensitivity to artemisinin. We thus investigated whether the artemisinin-resistant mutants that we isolated were cross-resistant to other oxidative agents. Unexpectedly, these mutants did not show resistance to peroxide or paraquat, and, if anything, manifested hypersensitivity to these oxidative agents (Figure 3B). On the other hand, although sod1Δ and sod2Δ mutants display increased sensitivity to oxidative stress, neither of them exhibited more sensitivity to artemisinin than the wild type (result not shown). This result, combined with the observation that yeast cells are unaffected by artemisinin in YPD, suggests that artemisinin is not simply a promiscuous oxidant. Instead, we suggest that artemisinin may have local effects on the mitochondria.
Figure 3 Artemisinin Generates ROS in Yeast
When applicable, 8 μM artemisinin was used.
(A) Artemisinin-resistant strains generate fewer ROS. Yeast untreated with artemisinin was used as the control. The experiment was performed three times with similar results. NDE1 denotes the overexpressor strain of NDE1 driven by ADH1 promoter.
(B) Isolated artemisinin-resistant strains are not cross-resistant to paraquat or peroxide. Shown here are the wild-type (WT) parental strain (BY4742), nde1Δ and ndi1Δ on YPD plates without or with 0.02% paraquat.
(C) Iron is possibly involved in artemisinin (Art) activation. Addition of BPS to the medium reduces yeast's sensitivity to artemisinin, whereas BPS alone does not enhance general yeast survival on drug-free plates. We did not use a higher amount of BPS to further reduce the iron level because a severe reduction in iron dramatically affects yeast growth on YPG.
Because activation of artemisinin is known to depend upon the cleavage of a peroxide bridge via an iron-dependent mechanism [17], we investigated the interaction between iron and artemisinin in yeast. When bathophenanthrolinedisulfonic acid (BPS), an iron chelator, was added into YPG media together with 8 μM artemisinin, the inhibitory effect of artemisinin on growth was attenuated (Figure 3C). In a separate series of experiments to determine whether extra iron could aggravate the effect of artemisinin, we added between 1 and 100 μM Fe2+ or Fe3+ to the plates. At these concentrations iron-enhanced toxicity was not observed. In addition, incubation with iron before artemisinin exposure does not exacerbate yeast sensitivity to artemisinin. These results suggest iron plays an important role in the activation of artemisinin, but additional iron above a certain threshold level has no further effect. This is in agreement with the report that artemisinin can also effectively kill parasites at stages that lack haemozoin [18].
Discussion
In this report we developed a sensitive yeast model to analyze the action of artemisinin. We demonstrated that artemisinin is able to depolarize the mitochondrial membrane and that increased electron transport activity increases the sensitivity of the cell to artemisinin. Further, we showed artemisinin produces an increase in ROS by a mechanism that differs from that of general oxidants.
Although low amounts of artemisinin affect yeast growth only in nonfermentable media, in both YPG(E) and YPD media artemisinin depolarizes the mitochondrial membrane. The fact that YPD slightly attenuates the effect of artemisinin in depolarizing mitochondrial membrane potential is possibly due to the partial repression of the activity of respiration enzymes (including NDI1 and NDE1) by glucose [19]. Likewise, mutations that indirectly affect respiration will possibly also affect artemisinin sensitivity. The positive correlation observed between respiration activity and sensitivity to artemisinin suggests that the electron transport chain acts to stimulate the activity of artemisinin. As such, the mitochondrial electron transport chain per se does not seem to be the primary target of artemisinin; instead, the electron transport chain seems to play a role in activating artemisinin. The activated artemisinin may then locally depolarize the mitochondrial membrane, impeding many functions dependent upon its potential. Although the disruption of mitochondrial function does not dramatically affect yeast growth in YPD media, for P. falciparum, loss of membrane potential will additionally affect pyrimidine biosynthesis [20], a key metabolic process for the survival of the parasite. This proposed mode of action for artemisinin differs from that of atovaquone, a broad-spectrum antiparasitic drug, which has been shown to inhibit complex III of the electron transport chain and, consequently, collapse mitochondrial membrane potential and kill P. falciparum [21].
Our findings indicate that iron is involved in the action of artemisinin. Mitochondria are well known to be a rich source of transition metals, including iron and copper. Conceivably, an active electron transport chain provides the electron source to Fe-S in catalyzing artemisinin and produces carbon-centered free radicals. Although we cannot exclude the possibility that other components in the electron transporting chains may interact with artemisinin, our current findings suggest that limited components, including NADH dehydrogenases, are involved in this process.
Distinctive ultrastructural changes have been noted in artemisinin-treated P. falciparum, including marked swelling of the mitochondria, followed by the appearance of electron-dense chromatin materials in the nuclei, clumping of ribosomes, nuclear membrane blebbing, and segregation of the nucleoplasm [22–26]. At high artemisinin concentration, even mammalian cells partially lose their mitochondria function [27,28]. These observations are consistent with a primary role of mitochondria in the action of artemisinin, whereas other physiological changes are probably secondary. Not surprisingly, P. falciparum has a functional mitochondrial electron transport system and an oxygen-requiring system that is necessary for growth and survival [29–31]. The relative insensitivity of mammalian cells to artemisinins may be due to the fundamental structural difference of their electron transport chains; for example, human complex I is composed of at least 43 components whereas its counterparts in yeast and P. falciparum are much simpler (one polypeptide chain). Indeed, a BLAST search with Nde1p or Ndi1p as a query revealed no significant homolog in the human genome. Even more, no significant homologs could be found in available National Center for Biotechnology Information (NCBI) databases across the whole Metazoan kingdom, indicating NADH dehydrogenase is not evolutionarily highly conserved between yeast and Metazoa.
The yeast model provides us with insights into mode of action of artemisinin: It interacts with the electron transport chain, generates local ROS, and causes the depolarization of the mitochondrial membrane. This scheme agrees with what we know about the chemical properties of artemisinin and findings in malaria studies. Although some physiological differences exist between yeast and malarial parasites, the pathway or basic principles that we have derived in our studies of artemisinin in yeast should aid future malaria research. The yeast model that we have developed provides a new way to probe the action of artemisinin. Our findings suggest many future genetic and biochemical analyses of this natural, powerful, and mysterious drug.
Materials and Methods
Yeast growth and library screen.
Standard yeast media and growth conditions were used. The yeast deletion library used is a collection of 4,757 homozygous diploid S. cerevisae strains (BY4743: MATa/MATá his3Δ1/his3Δ1 leu2Δ0/leu2Δ0 lys2Δ0/+ met15Δ0/+ ura3Δ0/ura3Δ0) in which each strain has a single open reading frame replaced with the KanMX4 module. We used homozygous diploid strain library for the original screen because haploid strain appears to have a relatively higher rate of screening background, presumably due to spontaneous mutation. An aliquot of the pooled yeast library was plated on YPG (2% glycerol as carbon source) or YPE (3% ethanol as carbon source) agar plates supplemented with 3–5 μM artemisinin. (Altogether about 100,000 total colonies were plated). Artemisinin-resistant colonies were isolated, and serial dilutions were made and spotted on artemisinin plates to confirm the original phenotype. About a dozen relatively more resistant colonies were chosen, and the corresponding genes were identified by bubble PCR or inverse PCR and DNA sequencing analysis. Phenotypes of these mutants were further confirmed in haploid background (BY4742: MATα his3Δ1 leu2Δ0 lys2Δ0 ura3Δ0).
Artemisinin sensitivity/resistance assay.
For growth in liquid, yeast were grown overnight in YPD medium, spun down, washed three times with YPG medium, and diluted to an A600 of 0.5. Then, 10 μl was inoculated into YPD or YPG media with or without artemisinin, and A600 was measured over time. For growth on agar plates, yeast previously grown in liquid YPD was spotted with 10-fold serial dilutions.
Besides SIP5, genes that may be involved in carbon-source utilization were individually checked for artemisinin sensitivity/resistance. These include GAL83, SIP1, −2, −3, and −4, and SNF1 and −4. snf1 cannot grow on YPG. sip1Δ, sip3Δ, sip4Δ, and gal83Δ display artemisinin resistance.
Mutants of genes in the electron transport chain, including SDH1, −2, −3, and −4, COR1, CYT1, QCR2, −6, −7, −8, −9, and -10, RIP1, INH1, STF1 and -2, and COX4, −5A, −5B, −6, −7, −8, −9, −12, and -13 were individually examined for artemisinin resistance. Among these, SDH1 and -3, CYT1, QCR9, and COX4, −6, −7, and -13 are required for yeast growth on YPG media, precluding analysis of their involvement in artemisinin's action.
Gene deletion and expression.
Gene replacement was achieved by homologous recombination with URA3 or LEU2 as the replacement marker in strain BY4742 background. Gene deletions were verified by PCR. For expression, SIP5, NDI1, NDE1, and PfNDI1 were PCR amplified from BY4743 genomic DNA or a cDNA library of P. falciparum. Amplified target DNAs were then cloned into an ADH1-driven expression vector (pADH1-YES2, a vector modified from pYES2 from Invitrogen (Carlsbad, California, United States), in which the GAL1 promoter was replaced with an ADH1 promoter). Yeast transformation was done by the lithium acetate method. Transformed cells were selected on SD-Ura. Sequences for PCR primers used for SIP5, NDI1, NDE1, and PfNDI1 are available by request.
Analysis of ROS production.
Flow cytometric analysis was used to assay the production of free intracellular radicals. Briefly, cells were incubated with dihydrorhodamine 123 for 2 h and then analyzed by a FACS Calibur (Becton Dickinson, San Jose, California, United States) at a low flow rate with excitation and emission settings of 488 and 525–550 nm (filter FL1), respectively.
Assay of the electrochemical potential.
After treatment in 20 mM HEPES buffer (pH 7.4) containing 50 mM glucose, 1 ml of the cell suspension was incubated with 2 μM Rh123 (rhodamine 123) for 30 min, washed, and then resuspended in 100 μl PBS. Mitochondrial electrochemical potential was expressed as the fluorescence intensity of Rh123, which was read through a FACS Calibur (Becton Dickinson) with excitation at 480 nm and emission at 530 nm.
We thank C. Vulpe for reading the paper. The P. falciparum cDNA was a kind gift from P. Rosenthal's lab. JMG is an investigator with Howard Hughes Medical Institute. This work is supported by grants from Tsinghua 985 (to BZ), NSFC (30470973 and 30330340 to BZ), China Postdoctoral Science Foundation (to WL), and the Excellent Young Teacher Program of MOE, People's Republic of China (to BZ).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. WL and BZ conceived and designed the experiments. WL, WM, DS, JW, LS, and SL performed the experiments. BZ analyzed the data. JMG contributed reagents/materials/analysis tools. WL, JMG, and BZ wrote the paper.
Abbreviations
BPSbathophenanthrolinedisulfonic acid
IC5050% growth inhibition concentration
ROSreactive oxygen species
SGsynthetic yeast media with glycerol as the carbon source
YPG(E)yeast media with glycerol or ethanol as the carbon source
YPGEyeast media with glycerol and ethanol as carbon sources
==== Refs
References
Meshnick SR 2002 Artemisinin: mechanisms of action, resistance and toxicity Int J Parasitol 32 1655 1660 12435450
Asawamahasakda W Benakis A Meshnick SR 1994 The interaction of artemisinin with red cell membranes J Lab Clin Med 123 757 762 8195681
Bhisutthibhan J Pan XQ Hossler PA Walker DJ Yowell CA 1998 The Plasmodium falciparum translationally controlled tumor protein homolog and its reaction with the antimalarial drug artemisinin J Biol Chem 273 16192 16198 9632675
Eckstein-Ludwig U Webb RJ Van Goethem ID East JM Lee AG 2003 Artemisinins target the SERCA of Plasmodium falciparum
Nature 424 957 961 12931192
Mercereau-Puijalon O Fandeur T 2003 Antimalarial activity of artemisinins: identification of a novel target? Lancet 362 2035 2036 14697798
Reed MB Saliba KJ Caruana SR Kirk K Cowman AF 2000 Pgh1 modulates sensitivity and resistance to multiple antimalarials in Plasmodium falciparum
Nature 403 906 909 10706290
Price RN Cassar C Brockman A Duraisingh M van Vugt M 1999 The pfmdr1 gene is associated with a multidrug-resistant phenotype in Plasmodium falciparum from the western border of Thailand Antimicrob Agents Chemother 43 2943 2949 10582887
Giaever G Chu AM Ni L Connelly C Riles L 2002 Functional profiling of the Saccharomyces cerevisiae. Genome Nature 418 387 391 12140549
Steinmetz LM Scharfe C Deutschbauer AM Mokranjac D Herman ZS 2002 Systematic screen for human disease genes in yeast Nat Genet 31 400 404 12134146
Luttik MA Overkamp KM Kotter P de Vries S van Dijken JP 1998 The Saccharomyces cerevisiae NDE1 and NDE2 genes encode separate mitochondrial NADH dehydrogenases catalyzing the oxidation of cytosolic NADH J Biol Chem 273 24529 24534 9733747
Marres CA de Vries S Grivell LA 1991 Isolation and inactivation of the nuclear gene encoding the rotenone-insensitive internal NADH: Ubiquinone oxidoreductase of mitochondria from Saccharomyces cerevisiae
Eur J Biochem 195 857 862 1900238
Small WC McAlister-Henn L 1998 identification of a cytosolically directed NADH dehydrogenase in mitochondria of Saccharomyces cerevisiae
J Bacteriol 180 4051 4055 9696750
Sanz P Ludin K Carlson M 2000 Sip5 interacts with both the Reg1/Glc7 protein phosphatase and the Snf1 protein kinase of Saccharomyces cerevisiae
Genetics 154 99 107 10628972
Gardner MJ Hall N Fung E White O Berriman M 2002 Genome sequence of the human malaria parasite Plasmodium falciparum
Nature 419 498 511 12368864
Schuller HJ 2003 Transcriptional control of nonfermentative metabolism in the yeast Saccharomyces cerevisiae
Curr Genet 43 139 160 12715202
Krungkrai SR Yuthavong Y 1987 The antimalarial action on Plasmodium falciparum of qinghaosu and artesunate in combination with agents which modulate oxidant stress Trans R Soc Trop Med Hyg 81 710 714 3329778
Meshnick SR Yang YZ Lima V Kuypers F Kamchonwongpaisan S 1993 Iron-dependent free radical generation from the antimalarial agent artemisinin (qinghaosu) Antimicrob Agents Chemother 37 1108 1114 8517699
ter Kuile F White NJ Holloway PH Pasvol G Krishna S 1993
Plasmodium falciparum: In vitro studies of the pharmacodynamic properties of drugs used for the treatment of severe malaria Exp Parasitol 76 85 95 8467901
DeRisi JL Iyer VR Brown PO 1997 Exploring the metabolic and genetic control of gene expression on a genomic scale Science 278 680 686 9381177
Krungkrai J 2004 The multiple roles of the mitochondrion of the malarial parasite Parasitology 129 511 524 15552397
Srivastava IK Rottenberg H Vaidya AB 1997 Atovaquone, a broad spectrum antiparasitic drug, collapses mitochondrial membrane potential in a malarial parasite J Biol Chem 272 3961 3966 9020100
Ellis DS Li ZL Gu HM Peters W Robinson BL 1985 The chemotherapy of rodent malaria XXXIX. Ultrastructural changes following treatment with artemisinine of Plasmodium berghei infection in mice, with observations of the localization of [3H]-dihydroartemisinine in P. falciparum in vitro Ann Trop Med Parasitol 79 367 374 3907556
Kawai S Kano S Suzuki M 1993 Morphologic effects of artemether on Plasmodium falciparum in Aotus trivirgatus
Am J Trop Med Hyg 49 812 818 8279648
Maeno Y Toyoshima T Fujioka H Ito Y Meshnick SR 1993 Morphologic effects of artemisinin in Plasmodium falciparum
Am J Trop Med Hyg 49 485 491 8214279
Jiang JB Jacobs G Liang DS Aikawa M 1985 Qinghaosu-induced changes in the morphology of Plasmodium inui
Am J Trop Med Hyg 34 424 428 3890573
Ye ZG Li ZL Li GQ Fu XQ Liu HP 1983 Effects of Qinghaosu and chloroquine on the ultrastructure of the erythrocytic stage of P. falciparum in continuous cultivation in vitro
J Tradit Chin Med 3 95 102 6353081
Reungpatthanaphong P Mankhetkorn S 2002 Modulation of multidrug resistance by artemisinin, artesunate and dihydroartemisinin in K562/adr and GLC4/adr resistant cell lines Biol Pharm Bull 25 1555 1561 12499639
Fishwick J Edwards G Ward SA McLean WG 1998 Morphological and immunocytochemical effects of dihydroartemisinin on differentiating NB2a neuroblastoma cells Neurotoxicology 19 393 403 9621345
Ginsburg H Divo AA Geary TG Boland MT Jensen JB 1986). Effect of mitochondrial inhibitors on intraerythrocytic Plasmodium falciparum in in vitro cultures J Protozool 33 121 125 3514901
Uyemura SA Luo S Moreno SN Docampo R (2000 Oxidative phosphorylation, Ca2+transport, and fatty acid-induced uncoupling in malaria parasites mitochondria J Biol Chem 275 9709 9715 10734123
Krungkrai J Burat D Kudan S Krungkrai S Prapunwattana P 1999 Mitochondrial oxygen consumption in asexual and sexual blood stages of the human malarial parasite, Plasmodium falciparum
Southeast Asian J Trop Med Public Health 30 636 642 10928353
|
16170412
|
PMC1201371
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Sep 16; 1(3):e36
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010036
|
oa_comm
|
==== Front
Clin Mol AllergyClinical and molecular allergy : CMA1476-7961BioMed Central London 1476-7961-3-101599626610.1186/1476-7961-3-10ResearchLysis with Saponin improves detection of the response through CD203c and CD63 in the basophil activation test after crosslinking of the high affinity IgE receptor FcεRI Hoffmann Hans Jürgen [email protected]øgebjerg Mette [email protected] Lars Peter [email protected] Ronald [email protected] Department of Pulmonary Medicine, Aarhus University Hospital, Aarhus University, DK 8000 Aarhus C, Denmark2 Institute of Pharmacology, Aarhus University, DK 8000 Aarhus C, Denmark2005 4 7 2005 3 10 10 11 4 2005 4 7 2005 Copyright © 2005 Hoffmann et al; licensee BioMed Central Ltd.2005Hoffmann 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 basophil activation test (BAT), in which translocation of markers to the surface of blood basophils is measured in response to allergen by flow cytometry, is a rapid assay that is gaining popularity. Two markers are currently being evaluated for the BAT; CD63 and the lineage-specific CD203c. In a recent report, detection of CD203c after lysis with Saponin was shown to be superior to detection of CD63 after lysis with formic acid. We wanted to compare a) lysis with formic acid and lysis with Saponin, b) the response through CD203c and CD63, and c) the definition 10% activated cells above background with the probability binning metric T(χ) > 4, on sets of data generated with blood basophils stimulated with varying concentrations of anti-FcεRI antibody.
Methods
Blood from volunteers was incubated with serial logarithmic dilutions of anti-FcεRI and subsequently with antibodies to CD203c PE and CD63 FITC. Sets of samples set up in parallel were lysed with either Saponin based Whole Blood Lysing reagent or with formic acid based Immunoprep/Q-prep. Samples were acquired on a FACS Calibur, but were compensated and analysed offline. Responders were defined as persons who had 10% or more activated basophils above background, or a T(χ) > 4, for two consecutive dilutions of anti-FcεRI antibody.
Results
More basophils (median 1164 vs. median 397) and better discrimination of upregulated CD203c and CD63 amongst responders were obtained after lysis with Saponin than after lysis with formic acid. We suggest that CD203c may be a more sensitive marker for the BAT than CD63, as 6/11 responders were found with CD203c, compared with 3/11 with CD63. Most responders (7/11) were identified with probability binning.
Conclusion
A combination of lysis with Saponin and the markers CD203c and CD63 computed by probability binning may be the most sensitive method of detecting activation of basophils after stimulation through FcεRI.
==== Body
Background
The basophil activation test (BAT), in which an allergen-specific response is measured by flow cytometry (reviewed in Ebo et al [1]), is gaining popularity as an ex vivo diagnostic tool. It is a rapid test with relatively high sensitivity and specificity that relies on surface translocation of transmembrane markers by regulated exocytosis in response to a stimulus through the high affinity IgE receptor (FcεRI). Crosslinking by anti-IgE of IgE bound to FcεRI [2,3], or stimulation with fMLP [4] serve as positive control. A third option is to crosslink FcεRI with a monoclonal antibody [5]. Concentrations of allergens selected to elicit a graded response are used to test for response to allergen. We regard the BAT as an attractive tool in the arsenal of the allergologist to identify culprit allergens.
Two markers are currently evaluated for the BAT – CD63 with a broad expression profile [6] and more recently CD203c, a lineage marker for CD34+ progenitor cells, mast cells and basophil granulocytes [7]. As CD203c is a unique marker for basophils and mast cell precursors, it may be sufficient for identification and detection of activation of basophils. When using CD63 as a metric, it is common to use antibodies to IgE [2,8-10], sometimes with CD45 [11,12] to identify basophils. An alternative method uses CD123 and HLA DR [13].
Most reports on the test have used either one of the markers, but in a recent publication [14] these markers were directly compared – with the caveat that response through CD63 was evaluated after lysis with Q-prep (based on formic acid), and the response through CD203c was evaluated after lysis with Whole Blood Lysing reagent (WBL, based on Saponin), both from Coulter. Although Hauswirth et al [7] have shown that there is good concordance between CD63 and CD203c, authors that established their experience base with CD63 contested the publication of data suggesting that CD203c is superior to CD63 [5,15]. We have compared the two markers CD63 and CD203c after lysis with WBL or Immunoprep/Q-prep, the manual kit from Coulter using the same chemistry as Q-prep, and find that lysis with the Saponin-based WBL is superior to lysis with Immunoprep/Q-prep, and that the response through CD203c after lysis with Saponin is stronger and more distinct than that through CD63. We have also tested probability binning condition T(χ) > 4 as an algorithm to identify a response, and found it comparable to "baseline + 10% activated cells", the method we used to define positive events [14].
Methods
Stimulation and flow cytometry
The method used was designed to be rapid for use in routine diagnosis. Heparinised blood (4 ml) was obtained from 11 informed volunteers, of which 2 had allergic airway disease. The procedure had been approved by the Ethics Committee of Aarhus County. Aliquots (100 μl) were incubated at 37°C for 5 minutes with increasing amounts of antibody to FcεRI CRA1 (Kyoto Pharmaceutical Industry Co., Japan) [16](spanning 7 orders of magnitude from 0,01 pg/μl to 10 ng/μh). CD203c PE (Immunotech, France) and CD63 FITC (Caltag, USA) were added to the same tube (titrated to 5 μl for each antibody) and incubation at 37°C continued for 10 minutes. The time point at 15 minutes was selected on the basis of published optimal times of response for CD203c [7,17] and CD63 [17]. The reaction was stopped by addition of lysing reagent, and after lysis, fixation and a wash, the samples were analysed on a FACS Calibur (Becton-Dickinson, Irvine, CA, USA) without hardware compensation. Samples were lysed with either WBL or Immunoprep/Q-prep, (both from Coulter Corporation, Hialeah, FL, USA) according to the manufacturer's instructions. Standards for software compensation were acquired by labelling one drop of Comp beads (Becton-Dickinson, Irvine, CA, USA) with 5 μl of antibody.
Data analysis and statistics
Data files were compensated and analysed with FlowJo version 6.1 (Treestar Corp, USA, Figure 1). The lymphocyte region containing CD203c+ basophils was confirmed by the dynamic backgating function of FlowJo (Figure 2), and basophils were identified as CD203c+ cells. In a separate dot plot, basophil expression of CD63 and CD203c were plotted. Thresholds were set at 2% on histograms of CD203c and CD63 (Figure 3). Figures 1, 2, 3 were generated on the same representative dataset. For probability binning analysis [18] of cells in the basophil gate, unstimulated samples were set as reference, and all samples stimulated with CRA1 were compared to that sample. Normal distribution of the data sets (% positive cells) was confirmed (SPSS v 10), and data was analysed with the Students t test. P < 0,05 was assumed to be significant.
Figure 1 Analysis of Basophil activation. Analysis of basophils after lysis by Immunoprep/Q-prep (a–f) or WBL (g–l) of a representative donor. By dynamic backgating (illustrated in figure 2), the region in a forward scatter vs side scatter plot in which basophils are located was optimised (a & g). CD203c+ cells were identified in this gate (b & h), and the expression of CD203c and CD63 was evaluated (c & i). CD203c vs CD63 expression at differend concentrations of CRA1 is shown after both lysis conditions (1 ng/ul panels c, g, 0,001 ng/ul in panels f & l, 0, 0,0001 in panels e & k; pbs in panels d & j).
Figure 2 Backgated basophil populations. Backgated basophils identified after lysis with Immunoprep/Q-prep (a) and WBL (b).
Figure 3 Histograms of CD63 and CD203c expression on basophils and control cells. Histograms of expression of CD203c (a & c) and CD63 (b & d) on basophils at baseline (red line) and after maximal stimulation (green line), and on lymphocytes (blue line) after lysis with Immunoprep/Q-prep (a & b) and WBL (c & d). Markers were set to include ~2% of unstimulated basophils.
Results
More basophils are detected with WBL than with Immunoprep/Q-prep
The yield of basophils from the WBL lysis (median 1164 cells for 250 000 events acquired) was significantly better than the yield with Immunoprep/Q-prep (median 397 cells for 250 000 events acquired) for 7/11 data sets (Table 1). In the four sets where the difference was not significant, the yield of basophils in the WBL assay was lower than the median, but still better than with Immunoprep/Q-prep. When plotting the cell number against the amount of CRA1 added, there was a trend toward an increased yield at high concentrations of CRA1 after lysis with Immunoprep/Q-prep, suggesting that basophils were detected more easily when they express high levels of CD203c. This trend was much less pronounced after lysis with WBL.
Table 1 Basophil cell yields. Average of numbers detected after lysis at 7 different concentrations of CRA1 from 250000 (normalized) events after lysis with either WBL or Immunoprep/Q-prep (average ± SD, tested with a paired t test). + = atopics, - = non atopics
Donor Allergy WBL ± SD Immunoprep ± SD p-value
1 - 1714 ± 314 1108 ± 449 <0,015
2 - 1979 ± 132 369 ± 274 <0,001
3 + 1638 ± 66 406 ± 262 <0,001
4 - 568 ± 84 182 ± 107 <0.001
5 - 1646 ± 282 371 ± 248 <0,001
6 + 483 ± 113 436 ± 138 0,494
7 - 1164 ± 70 213 ± 160 <0.001
8 - 1132 ± 372 582 ± 150 0,003
9 - 1397 ± 207 306 ± 133 <0,001
10 - 995 ± 481 577 ± 415 0,107
11 - 407 ± 128 397 ± 183 0,904
Median 1164 397
CD203c is more sensitive than CD63 at detecting signalling through FcεRI
When defining a positive response as two consecutive responses of more than 10% above baseline [14], fewer responders were recorded with CD63 (3/11 data sets) than with CD203c (6/11 data sets). All responders through CD63 respondent also through CD203c. Lysis procedure had no effect on CD63, but there was one more response detected with CD203c after lysis with WBL than after lysis with Immunoprep/Q-prep (Table 2). The participants were split into three groups on the basis of >10% positive cells at two consecutive dilution (Table 2): responders with both CD63 and CD203c (n = 3), responder with CD203c only (4 for WBL, n = 3 for Immunoprep/Q-prep) and non responders (n = 5).
Table 2 Responders as defined by Boumiza et al [14] or by T(χ) > 4 for two consecutive dilutions of CRA1. Y = responder
CD63 CD203c T(χ)63,203c
Donor WBL Immunoprep WBL Immunoprep WBL Immunoprep
1
2
3
4 Y
5
6 Y Y Y
7 Y Y Y Y
8 Y Y Y
9 Y Y Y Y Y
10 Y Y Y Y Y Y
11 Y Y Y Y Y Y
Probability binning offers an integrative alternative to using either only CD203c or only CD63
When analyzing the same dataset by defining that the probability binning metric T(χ)CD203c, CD63 > 4 for two consecutive dilutions as a response, a similar classification emerged for the data set obtained after lysis with WBL (7/11 data sets), and to some part with Immunoprep/Q-prep (4/11 data sets, Table 2). Discrimination of T(χ)CD203c, CD63 was significantly better after lysis with WBL than after lysis with Immunoprep/Q-prep (4/11 data sets).
The ratio of activation was higher for CD203c than for CD63
The degree of activation detected through CD203c and CD63 after lysis with either Immunoprep/Q-prep or WBL was compared by dividing the fraction of positive cells in CRA1-activated samples of responders by the fraction of positive cell at baseline (Figure 4). The signal was better with WBL (Figure 4b & 4d) than with Immunoprep/Q-prep (Figure 4a & 4c) and was slightly better with CD203c (Figure 4c & 4d) than with CD63 (Figure 4a & 4b). Lysis with WBL was significantly better for both CD203c (3/6 data sets) and for CD63 (1/6 data sets) (Table 3). Detection with CD203c was significantly better after lysis with WBL (1/6 data sets) and Immunoprep/Q-prep (2/6 data sets). Detection of CD203c was significantly better in 4/6 experiments when comparing the lysis condition used by Boumiza et al (Immunoprep/Q-prep for detecting CD63 and WBL for detection of CD203c) [14].
Figure 4 Comparison of response of CD63 and CD203c with different lysis methods. Degree of activation at varying concentrations of CRA1. The % positive cells at a given concentration are plotted against the amount of anti-FceRI antibody. The lowest data point is labelled 1e-8 for the purpose of representation on a log scale. (a) Immunoprep/Q-prep lysis, detection with CD63, (b) Immunoprep/Q-prep lysis, detection with CD203c, (c) WBL lysis, detection with CD63, (d) WBL lysis, detection with CD203c. In panel d, one responder achieved significantly higher activation ratios than 35.
Table 3 Difference in activation for combinations of WBL, Immunoprep/Q-prep and CD63 and CD203c amongst responders The p-value was calculated with the Students t-test. ns = not significant
WBL vs Immunoprep CD203c vs CD63
CD63 CD203c Immunoprep WBL, CD203c WBL vs CD63 Immunoprep
6 ns 0,025 ns ns ns
7 ns 0,016 0,027 0,016 0,012
8 0,016 ns 0,004 ns 0,019
9 ns ns ns ns ns
10 ns 0,018 ns ns 0,019
11 ns ns ns ns 0,043
Discussion
The BAT is an exiting development in applied functional flow cytometry, and a number of laboratories have developed independent procedures to use it The first common approach to standardization is a EAACI working paper . We chose to use stimulation of blood basophils through FcεRI with the antibody CRA1 [16] to compare different lysis methods. Recently, Boumiza published a controversial comparison of responses through CD63 after lysis with Immunoprep/Q-prep and CD203c after lysis with WBL [14] that was contested by groups with experience in detecting CD63. Other reports that so far have compared CD63 and CD203c [7,17] give anecdotal evidence of a similar response through the markers, but have not compared them stringently. We had noticed that lysis with Saponin (on which WBL is based) gives appreciably better results than lysis with ammonium chloride (unpublished), and chose to compare the lysis methods (WBL, with Saponin, and Immunoprep/Q-prep lysing reagent, with formic acid) and markers (CD63 FITC and CD203c PE) used for the report)[14].
The yield was remarkably better for WBL (involving Saponin) than for Immunoprep/Q-prep (involving formic acid). Although the minimum number of basophils for a useful test has been reported to be 100 [19], we prefer to have more than 500, which is within reason as we could obtain approximately 500 basophils from 100 μl blood from most donors using lysis with WBL (Table 1).
The threshold for a positive response has in the past been set by an empirically determined fraction of basophils detected above background. The threshold for detection of allergens by a BAT should be set using Receiver Operating Curves [20]. For the present study it was deemed to be stringent to set it to be 10% above the unstimulated control experiment in two consecutive dilutions of allergen or, in this case, antibody to FcεRI to be comparable to the previous study comparing CD63 and CD203c [14]. Using this threshold, 6 of 11 persons mounted a positive response to cross linking of FcεRI. In other studies, the threshold is set empirically between 6% and 17% (EAACI Position paper at ).
As there is evidence that CD203c and CD63 are translocated to the basophil cell surface by different mechanisms and with different kinetics [17], it may be of interest to monitor them simultaneously. Probability binning [21] is an algorithm by which variance in the control distribution is minimised by varying bin size before a normalised chi-square value is computed for each sample distribution using the same bins. This has two advantages: 1. it combines the information residing in cell number in a given bin with median fluorescence intensity and 2. the bins constructed during the analysis can be extended from a univariate analysis to be rectangles on a dot plot of CD203c vs CD63 containing an even number of events of the unstimulated control, and a single chi-square value can then be computed that incorporates change in both markers [18]. We show that the result of probability binning with CD63 and CD203c as dimensions is similar to the result of the method of assigning a threshold at baseline + 10%. This suggests that T(χ)CD203c CD63 is as sensitive as the conditions defined by Boumiza et al [14] (background + 10% positive).
The relative sensitivity of CD203c and CD63 was compared by calculating the relative increase in ratio of positive cells over baseline conditions at each concentration of FcεRI antibody. As lysis with Immunoprep/Q-prep results in a lower basophil yield and immunofluorescence than WBL, and CD63 appears to be upregulated to a lesser extent than CD203c, the combination used in [14], detecting CD63 after lysis with Immunoprep/Q-prep (Figure 4a) and detection of CD203c (Figure 4d) after lysis with WBL, accentuated differences between the markers.
Conclusion
We present data that supports the claim that WBL is a better lysis method than the automated Immunoprep/Q-prep (shown here), and that CD203c is more sensitive than CD63 at detecting FcεRI-mediated activation and uniquely identifies basophils in human blood. As the presented data were obtained with an antibody to FcεRI, the results need to be confirmed after stimulation with allergen. Probability binning offers an approach that combines CD63 and CD203c into one metric that has a high response. A combination of lysis with Saponin and the markers CD203c and CD63 [17] computed by probability binning may be the most sensitive method of detecting activation of basophils after stimulation through FcεRI.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
HJH conceived the project, analysed the data and wrote the manuscript. BMH recruited patients and did the experiments, LPN contributed to project design and writing of the manuscript. RD contributed to the design of the study and writing of the manuscript. All authors read and approved the final manuscript.
Acknowledgements
This study was financed by The Danish Velux Foundation, Hørslev-Fonden, Augustinus-fonden, C. C. Klestrup og Hustru Henriette Klestrups Mindelegat and Danish Medical Research Council.
==== Refs
Ebo DG Hagendorens MM Bridts CH Schuerwegh AJ De Clerck LS Stevens WJ In vitro allergy diagnosis: should we follow the flow? Clin Exp Allergy 2004 34 332 339 15005724 10.1111/j.1365-2222.2004.01891.x
Sanz ML Sanchez G Gamboa PM Vila L Uasuf C Chazot M Allergen-induced basophil activation: CD63 cell expression detected by flow cytometry in patients allergic to Dermatophagoides pteronyssinus and Lolium perenne Clin Exp Allergy 2001 31 1007 1013 11467990 10.1046/j.1365-2222.2001.01122.x
Sainte-Laudy J Sabbah A Drouet M Lauret MG Loiry M Diagnosis of venom allergy by flow cytometry. Correlation with clinical history, skin tests, specific IgE, histamine and leukotriene C4 release Clin Exp Allergy 2000 30 1166 1171 10931125 10.1046/j.1365-2222.2000.00863.x
Ebo DG Lechkar B Schuerwegh AJ Bridts CH De Clerck LS Stevens WJ Validation of a two-color flow cytometric assay detecting in vitro basophil activation for the diagnosis of IgE-mediated natural rubber latex allergy Allergy 2002 57 706 712 12121189 10.1034/j.1398-9995.2002.23553.x
De Weck AL Sanz ML For allergy diagnostic flow cytometry, detection of CD203c instead of CD63 is not at all an improvement in other hands Clin Exp Allergy 2003 33 849 852 12848168 10.1046/j.1365-2222.2003.00691.x
Nieuwenhuis HK van Oosterhout JJ Rozemuller E van Iwaarden F Sixma JJ Studies with a monoclonal antibody against activated platelets: evidence that a secreted 53,000-molecular weight lysosome-like granule protein is exposed on the surface of activated platelets in the circulation Blood 1987 70 838 845 3620703
Hauswirth AW Natter S Ghannadan M Majlesi Y Schernthaner GH Sperr WR Recombinant allergens promote expression of CD203c on basophils in sensitized individuals J Allergy Clin Immunol 2002 110 102 109 12110828 10.1067/mai.2002.125257
Moneret-Vautrin DA Sainte-Laudy J Kanny G Fremont S Human basophil activation measured by CD63 expression and LTC4 release in IgE-mediated food allergy Ann Allergy Asthma Immunol 1999 82 33 40 9988204
Schuerwegh AJ Ebo DG Bridts CH De Clerck LS Stevens WJ CD63 expression on basophils of nonallergic controls and patients allergic to wasp J Allergy Clin Immunol 2001 108 150 152 11447407 10.1067/mai.2001.116125
Erdmann SM Heussen N Moll-Slodowy S Merk HF Sachs B CD63 expression on basophils as a tool for the diagnosis of pollen-associated food allergy: sensitivity and specificity Clin Exp Allergy 2003 33 607 614 12752589 10.1046/j.1365-2222.2003.01660.x
Abuaf N Rajoely B Ghazouani E Levy DA Pecquet C Chabane H Validation of a flow cytometric assay detecting in vitro basophil activation for the diagnosis of muscle relaxant allergy J Allergy Clin Immunol 1999 104 411 418 10452764
Monneret G Benoit Y Debard AL Gutowski MC Topenot I Bienvenu J Monitoring of basophil activation using CD63 and CCR3 in allergy to muscle relaxant drugs Clin Immunol 2002 102 192 199 11846462 10.1006/clim.2001.5156
Sturm GJ Bohm E Trummer M Weiglhofer I Heinemann A Aberer W The CD63 basophil activation test in Hymenoptera venom allergy: a prospective study Allergy 2004 59 1110 1117 15355471 10.1111/j.1398-9995.2004.00400.x
Boumiza R Monneret G Forissier MF Savoye J Gutowski MC Powell WS Marked improvement of the basophil activation test by detecting CD203c instead of CD63 Clin Exp Allergy 2003 33 259 265 12580920 10.1046/j.1365-2222.2003.01594.x
Ebo DG Lechkar B Schuerwegh AJ Bridts CH De Clerck LS Stevens WJ Comments regarding 'Marked improvement of the basophil activation test by detecting CD203c instead of CD63' by Boumiza et al Clin Exp Allergy 2003 33 849 3 12801323 10.1046/j.1365-2222.2003.00691.x
Jensen BM Hansen JB Dissing S Gerwien J Skov PS Poulsen LK Monomeric immunoglobulin E stabilizes FcepsilonRIalpha from the human basophil cell line KU812 by protecting it from natural turnover Clin Exp Allergy 2003 33 655 662 12752595 10.1046/j.1365-2222.2003.01653.x
Buhring HJ Streble A Valent P The basophil-specific ectoenzyme E-NPP3 (CD203c) as a marker for cell activation and allergy diagnosis Int Arch Allergy Immunol 2004 133 317 329 15031605 10.1159/000077351
Roederer M Moore W Treister A Hardy RR Herzenberg LA Probability binning comparison: a metric for quantitating multivariate distribution differences Cytometry 2001 45 47 55 11598946 10.1002/1097-0320(20010901)45:1<47::AID-CYTO1143>3.0.CO;2-A
Erdmann SM Sachs B Hoffmann-Sommergruber K Scheiner O Merk H Regarding Ebo DG, Hagendorens MM, Bridts CH, Schuerwegh AJ, De Clerck LS & Stevens WJ. In vitro allergy diagnosis: should we follow the flow? Clin Exp Allergy 2004; 34:332–9 Clin Exp Allergy 2004 34 1498 1499 15347386 10.1111/j.1365-2222.2004.02060_1.x
Hemery ML Arnoux B Dhivert-Donnadieu H Rongier M Barbotte E Verdier R Confirmation of the diagnosis of natural rubber latex allergy by the Basotest method Int Arch Allergy Immunol 2005 136 53 57 15591814 10.1159/000082585
Roederer M Treister A Moore W Herzenberg LA Probability binning comparison: a metric for quantitating univariate distribution differences Cytometry 2001 45 37 46 11598945 10.1002/1097-0320(20010901)45:1<37::AID-CYTO1142>3.0.CO;2-E
|
15996266
|
PMC1201566
|
CC BY
|
2021-01-04 16:36:24
|
no
|
Clin Mol Allergy. 2005 Jul 4; 3:10
|
utf-8
|
Clin Mol Allergy
| 2,005 |
10.1186/1476-7961-3-10
|
oa_comm
|
==== Front
Mol CancerMolecular Cancer1476-4598BioMed Central London 1476-4598-4-271608349910.1186/1476-4598-4-27ResearchPancreatic Stellate Cells (PSCs) express Cyclooxygenase-2 (COX-2) and pancreatic cancer stimulates COX-2 in PSCs Yoshida Seiya [email protected] Michael [email protected] Xian-Zhong [email protected] Carolyn [email protected] Mark S [email protected] Richard H [email protected] Woody [email protected] Thomas E [email protected] Department of Surgery and Robert H Lurie Comprehensive Cancer Center, Northwestern University Feinberg School of Medicine, 333 East Superior 4–713, Chicago, IL 60611, USA2005 5 8 2005 4 27 27 20 8 2004 5 8 2005 Copyright © 2005 Yoshida et al; licensee BioMed Central Ltd.2005Yoshida 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
Cyclooxygenase 2 (COX-2), the inducible form of prostaglandin G/H synthase, is associated with several human cancers including pancreatic adenocarcinoma. Pancreatic stellate cells (PSCs) play a central role in the intense desmoplasia that surrounds pancreatic adenocarcinoma. The present study examined COX-2 expression in PSCs. PSCs isolated from normal rats, were cultured and exposed to conditioned medium (CM) from the human pancreatic cell line, PANC-1.
Methods
COX-2 expression was evaluated by immunostaining and western blotting. Proliferation of PSCs was determined by thymidine incorporation and cell counting.
Results
COX-2 was found to be constitutively expressed in PSCs, and COX-2 protein was up-regulated by PANC-1 CM. Moreover, the induction of COX-2 by PANC-1 CM was prevented by U0126, an extracellular signal-regulated kinase (ERK) 1/2 inhibitor suggesting that activation of ERK 1/2 is needed for stimulation of COX-2. Finally, NS398, a selective COX-2 inhibitor, reduced the growth of PSCs by PANC-1 CM, indicating that activation of COX-2 is required for cancer stimulated PSC proliferation.
Conclusion
The results suggest that COX-2 may play an important role in the regulation of PSC proliferation in response to pancreatic cancer.
==== Body
Background
Vitamin A-containing cells were first reported in 1982 by Watari et al. in vitamin A loaded mice using fluorescence and electron microscopy [1]. This cell type was subsequently identified by electron microscopy in normal rat and human pancreatic tissues [2]. These cells were identified as pancreatic stellate cells (PSCs) by Apte et al and Bachem et al in 1998 [3,4]. In the normal pancreas, stellate cells are quiescent and can be identified by the presence of vitamin A-containing lipid droplets in the cytoplasm. In response to pancreatic injury or inflammation, PSCs are transformed ("activated") from quiescent phenotypes into highly proliferative myofibroblast-like cells which express the cytoskeletal protein α-smooth muscle actin (α-SMA), and produce type I collagen and other extracellular matrix components. Many of the morphological and metabolic changes associated with the activation of PSCs in animal models of fibrosis also occur when these cells are cultured on plastic in serum-containing medium.
Activated PSCs have also been implicated in the deposition of extracellular matrix components in pancreatic adenocarcinoma [5]. In patients with pancreatic cancer, an intense, interstitial, fibrillar staining for PSCs is evident in the peritumoral fibrous regions. Procollagen I staining colocalized with α-SMA to these fibroblast-shaped cells suggests that they are responsible for the deposition of matrix components and the desmoplastic reaction that surrounds the pancreatic tumor [5].
Cyclooxygenases (COXs) are key rate-limiting enzymes involved in the conversion of arachidonic acid to prostaglandin (PG) H2, the precursor of a variety of compounds including PGs, prostacyclin, and thromboxanes. Two isozymes are found in mammalian tissues, COX-1 and COX-2. COX-1 is expressed constitutively in a wide variety of tissues, where it is involved in the maintenance of tissue homeostasis. In contrast, COX-2, which is not expressed in resting cells, is the inducible form of the enzyme responsible for PG production at sites of inflammation. Growth factors, cytokines, tumor promoters, and other inflammatory mediators can induce COX-2 expression [6,7]. COX-2 expression and activity is up-regulated in pancreatic cancer, but absent in normal pancreatic acinar and duct cells [8-10]. Some scattered cells in normal pancreatic tissues express COX-2 [11,12].
The current study revealed that COX-2 is expressed in primary cultured PSC. Furthermore, conditioned media from pancreatic cancer stimulates PSC proliferation and COX-2 expression. The increase in PSC proliferation in response to conditioned media is prevented by inhibition of COX-2.
Results
COX-2 in primary cultured PSCs
In early primary PSCs, cytoplasmic COX-2 staining was detected (Figure 1). However, early primary cultured PSCs (quiescent cells) were α-SMA negative (Figure 1). After passage, PSCs flattened and developed long cytoplasmic extensions (activated PSCs), and showed positive immunostaining for COX-2 and α-SMA (Figure 2).
Figure 1 Immunostaining of COX-2 and α-smooth muscle actin (α-SMA) in pancreatic stellate cells (PSCs) after one day in culture. (A) Immunostaining of COX-2 in quiescent PSCs. All PSCs stained for COX-2. (B) Immunostaining of α-SMA in quiescent PSCs. PSCs did not stain for α-SMA. Magnification 400×.
Figure 2 Immunostaining of COX-2 and α-smooth muscle actin (α-SMA) in pancreatic stellate cells (PSCs) after 10 days in culture. (A) Immunostaining of COX-2 in activated PSCs. (B) Immunostaining of α-SMA in activated PSCs. Magnification 400×. All PSCs stained for both COX-2 and α-SMA.
COX-2 protein in culture-activated PSCs
On days one and four in primary culture, PSCs expressed low levels of α-SMA (Figure 3). Between day 7 and day 20, α-SMA expression increased substantially (Figure 3). In. contrast, the COX-2 protein was detected in primary cultured PSC from day 1 through day 20 (Figure 3).
Figure 3 Induction of COX-2 and α-smooth muscle actin (SMA) protein in pancreatic stellate cells (PSCs). After isolation of PSCs, equal amounts of protein from the cell lysates were loaded by SDS-PAGE and immunoblotted with COX-2 or α-SMA antibodies. Upper panels show representative western blots and lower panels show the densitometry data from all experiments. PSCs expressed α-SMA after seven days in culture. In contrast, PSCs expressed COX-2 throughout this time period.
Expression of COX-2 protein in PSCs was increased by PANC-1 CM
PSCs were treated with PANC-1 CM for 0.5, 1, 3, 6, 12, 24, 48, and 72 hours. PANC-1 CM caused sustained up-regulation of the COX-2 protein, which was maximally increased after 12 hours and remained elevated for at least 24 hours (Figure 4).
Figure 4 The expression of COX-2 protein in pancreatic stellate cells (PSCs) was increased by cancer conditioned medium (PANC-1 CM). Stellate cells were isolated and cultured in media containing 10% serum for 12 days. Then, following 18-hour culture in 1% serum medium, cells were treated with PANC-1 CM for the indicated times. Upper panels are representative western blots showing the effect of PANC-1 CM on COX-2 expression over two different time-courses. The lower panels show that densitometric analysis western blots from the separate experiments. PANC-1 CM caused a rapid and sustained up-regulation of COX-2 protein, which was maximally increased after 12 hours and remained elevated.
The increase in expression of COX-2 protein in PSCs by PANC-1 CM was inhibited by U0126
PSCs were treated with PANC-1 CM and control medium and PANC-1 CM with the mitogen-activated protein kinase kinases (MEK) inhibitor U0126 (10 μM) for 0.5, 1, 3, 6, 12, 24, 48, 72 hours. U0126 significantly inhibited PANC-1-induced expression of COX-2 (Figure 5).
Figure 5 Effects of U0126, a specific inhibitor of ERK activation on cancer conditioned medium (PANC-1 CM)-induced COX-2 expression in pancreatic stellate cells (PSCs) by Western blot. Following 18-hour culture in 1% serum medium, cells were treated with control medium and PANC-1 CM in the absence and presence of 10 μM U0126 for the indicated times. Equal amounts of protein from the cell lysates were loaded by SDS-PAGE and immunoblotted with COX-2 antibody. The upper panel is a representative western blot and the lower panel shows densitometric analysis of the western blots from the separate experiments. The increase in expression of the COX-2 protein in PSCs in response to PANC-1 CM was abolished by U0126. This figure is representative of three separate experiments.
NS398 inhibits cell proliferation of PSCs stimulated by PANC-1 CM
Since COX-2 is increased by PANC-1 CM, the role of COX-2 in PANC-1 CM-induced PSC proliferation was investigated using a specific COX-2 inhibitor, NS398. PANC-1 CM increased PSC thymidine incorporation as well as cell number compared to control medium (Figure 6). Inhibition of COX-2 with NS398 resulted in a concentration-dependent decrease in thymidine incorporation and cell number.
Figure 6 Effect of cancer all conditioned medium (PANC-1 CM) with or without NS398. (A) Inhibition of COX-2 activity with NS398 decreased DNA synthesis (thymidine incorporation) in pancreatic stellate cells (PSCs). Results are expressed as percent of control. (B) NS398 inhibited cell growth in PSCs. Results are expressed as mean ± SEM from three separate experiments. * P < 0.001; ** P < 0.01; *** P < 0.05.
Discussion
There is accumulating evidence that PSCs play a role in the development of pancreatic fibrosis [13,14]. Little is known regarding the relationship between PSCs and pancreatic cancer, or the role of COX-2.
The present study revealed that PSCs express COX-2 constitutively and when activated. The two isoforms of COX, COX-1 and COX-2, differ in many respects. COX-1 is a housekeeping gene that is expressed in most tissue, while COX-2 is not detected in most normal tissues. In the pancreas, islet cells display a strong expression of COX-2 [9]; however, some scattered basal cells in normal pancreas express COX-2 as well, though less than seen in islet cells [11,12]. In hepatic stellate cells (HSCs) which are similar to PSCs, COX-2 expression is virtually undetectable by Western blot analysis in protein extracts obtained from freshly isolated HSC [15]. However, serum-deprived unstimulated HSC express low levels of the COX-2 protein and expression is dramatically enhanced in response to IL-1α, TNF α or endothelin-1 [15,16]. The present study suggests that COX-2 expression is independent of the activation status in isolated PSCs. While there is no marked expression of COX-2 in desmoplastic areas of pancreatic cancers [8-10], it is possible that the enzyme is up-regulated early in the activation of stellate cells in vivo but increased expression may not be required for maintenance of stellate function once activated.
Stimulation of PSC by PANC-1 CM increased the expression of COX-2. Oncogenes, growth factors, cytokines, chemotherapy and tumor promoters stimulate COX-2 transcription via protein kinase C and RAS-mediated signaling. Stimulation of either protein kinase C or RAS-mediated signaling enhances mitogen-activated protein (MAP) kinase activity, which in turn, activates transcription of COX-2 [17]. We have previously reported that PANC-1 CM enhances ERK 1/2 activation and growth of PSCs [18]. We speculate that a growth factor is responsible for these effects, however, our attempts to identify the candidate using receptor antagonists and immunoneutralization have not been successful. Inhibition of ERK1/2 phosphorylation by U0126 prevented the PANC-1 CM-stimulated increase in PSC COX-2 protein production. In previous studies U0126 alone had no effect on ERK1/2 or COX-2 expression [19,20]. This suggests that the MAP kinase pathway plays a role in cancer-induced stimulation of COX-2 in PSCs. The reported biological consequences of COX-2 up-regulation include growth stimulation inhibition of apoptosis [21], increased metastatic potential [22] and promotion of angiogenesis [23]. Increased expression of COX-2 in PSCs by PANC-1 CM may contribute to tumor progression.
Finally, the proliferation of PSCs was inhibited by treatment with NS398, a COX-2 inhibitor. In pancreatic carcinomas, COX-2 is overexpressed and NS398 inhibits tumor growth [8,9,24]. This COX-2 inhibitor alone has no effect on expression of COX-2 or ERK1/2 and shows no toxicity at the concentration used in the present studies [20,24]. Recent studies have demonstrated a role for the COX-2 enzyme and PGE2 in the regulation of epithelial cell growth and angiogenesis [23,25,26]. These properties will need to be studied further in pancreatic adenocarcinoma and stellate cells. NS-398 has been previously shown to inhibit cell proliferation of colorectal carcinoma by inducing apoptosis in a COX-2-independent fashion [27]. More studies are needed to confirm the mechanism of inhibition by NS398.
Conclusion
The COX-2 protein is up-regulated in pancreatic stellate cells by pancreatic cancer-conditioned media. The induction of COX-2 by pancreatic cancer cells is mediated by extracellular signal-regulated kinases 1/2 (ERK1/2). The COX-2 induction by pancreatic cancer cells is involved in mediating PSC proliferation. Therefore, COX-2 may play an important role in the regulation of desmoplasia in pancreatic cancer and inhibition of this enzyme may prevent or reduce this response.
Materials and methods
Materials
Iscove's modified Dulbecco's medium (IMDM), Dulbecco's modified Eagle's medium (DMEM), albumin, and pronase were purchased from Sigma Chemical (St Louis, MO). Fetal bovine serum (FBS), glutamine, and antibiotics were purchased from Mediatech, Inc. (Herndon, VA). Collagenase P was purchased from the Roche Diagnostics Corporation (Indianapolis, IN), and deoxyribonuclease from Amersham Biosciences (Piscataway, NJ). Nycodenz was obtained from Nycomed Pharma AS (Oslo, Norway). U0126, a specific inhibitor of extracellular signal-regulated kinase (ERK) activation, was obtained from Calbiochem (San Diego, CA). NS398, COX-2 inhibitor was obtained from Cayman Chemicals (Ann Arbor, MI). 3H-methyl thymidine was purchased from ICN Pharmaceuticals (Costa Mesa, CA).
Animals
Male Sprague-Dawley rats weighing 200–250 g were used in accordance with standard institutional animal welfare guidelines, and protocols were approved by the Institutional Animal Care and Use Committee, Northwestern University School of Medicine.
Isolation and culture of PSCs
Rat PSCs were isolated as previously described [3]. Briefly, the pancreas was digested with a mixture of collagenase P and pronase and deoxyribonuclease in Gey's balanced salt solution. The resulting suspension of cells was centrifuged in a 28.7% Nycodenz gradient at 1400 g for 23 minutes. Stellate cells then separated into a hazy band just above the interface of the Nycodenz solution and the aqueous buffer. Cells were harvested, washed, and resuspended in IMDM containing 10% FBS, 4 mmol/l glutamine, and antibiotics. PSCs were all used within two passages following isolation.
Conditioned medium
The poorly differentiated pancreatic adenocarcinoma cell line PANC-1 (American Type Tissue Culture, Rockville, MD) was grown in DMEM in 75 cm2 flasks. When the cells reached confluence, the serum-containing medium was removed and the cells were cultured in 20 ml of serum-free medium. After 24 hours, the medium was collected and the peptide containing fraction obtained by semi-purifying on a Sep-Pak Plus C18 Cartridge (Waters, Milford, MA). After washing, Sep-Paks were eluted with 50% acetonitrile (EM Science, Gibbstown, NJ) with 0.1% trifluoroacetic acid (J. T. Baker, Phillipsburg, NJ). The eluates were lyophilized and reconstituted in fresh IMDM with 1% FBS, forming what we refer to as PANC-1 conditioned medium (CM). In total extracts of 100 ml pf PANC-1 conditioned media were purified and reconstituted in 20 ml of media for PSC culture. Control medium consisted of only serum-free medium without PANC-1 cells which underwent are same Sep-Paking procedure. The eluates were also lyophilized, and then reconstituted in fresh IMDM with 1% FBS.
Immunostaining
COX-2 and α-SMA expression in PSCs was evaluated by immunohistochemical staining. Cultured PSCs were grown directly on glass coverslips in six-well plates, and immunostained for COX-2 using peroxidase-labeled streptavidin for immunohistochemistry (KPL, Inc., Maryland) according to the manufacturer's instructions. Cells were fixed for 30 minutes in acetone at -20°C. Thereafter, glass coverslips were air-dried and stored at 4°C until the cells were stained. Endogenous peroxidase activity was blocked by incubation in methanol with 0.3% hydrogen peroxidase for 30 minutes. After immersion in normal goat serum for 30 minutes, the slides were incubated with COX-2 (murine) polyclonal antibody (Cell Signaling Technology, Inc., Beverly, MA) diluted 1:200 in tris-buffered saline (TBS) 1× bovine serum albumin and stored in a humid chamber overnight at 4°C. The slides were incubated with anti-mouse immunoglobulins for 10 minutes at 37°C, followed by peroxidase conjugated streptavidin for 30 minutes at room temperature. Finally, color was developed incubating the slides for 8 minutes, with diaminobenzine (DAB Reagent Set; KPL, Inc., Maryland). Expression of α-SMA was examined in a similar manner by using monoclonal anti-α-SMA antibody (Sigma, St Louis, MI).
Protein extraction
Protein concentrations in the cell lysates were measured by the method of Lowry et al. [28] using bovine serum albumin as the standard.
3H-methyl thymidine incorporation
Following the treatment of cells with PANC-1 CM for 48 hours, DNA synthesis was measured by adding to each well 0.5 μCi of 3H-methyl thymidine and incubating these plates for the final 24 hours. The cell protein was precipitated with 10% trichloroacetic acid overnight, washed twice with phosphate buffered saline (PBS) and then dissolved by adding 0.25 ml of 0.5 mol/l NAOH to each well. Incorporation of 3H-methyl thymidine into DNA was measured by adding 1 ml of scintillation cocktail (ScintiSafe Plus 50%, S × 25-5, Fischer Scientific, Pittsburgh, PA) followed by count measurements using the Wallac WinSpectral liquid scintillation counter (Wallac Turku, Finland).
Cell counts
Cells were washed with PBS, harvested by trypsinization using 0.5% trypsin-0.2% EDTA, resuspended in 100 μl culture medium, and counted using the Guava Personal Cytometer (Guava Technologies, Inc., Burlingame, CA).
Western blotting
Expression of COX-2 and α-SMA were detected by Western blotting. Protein extracts (5 μg) from each sample were separated by gel electrophoresis using a 10% sodium dodecyl sulfate-polyacrylamide gel (SDS-PAGE). Known molecular weight protein standards were run alongside the samples. Separated proteins were then transferred to a nitrocellulose membrane (Bio-Rad, Hercules, CA), which was incubated for one hour at room temperature in blocking buffer (TBS and 0.1% Tween 20 with 5% nonfat dry milk). A murine COX-2 polyclonal antibody was diluted 1:2000 buffer (TBS and 0.1% Tween 20 with 5% nonfat dry milk). After incubation with the primary antibody overnight at 4°C, the membrane was exposed to the secondary antibody with gentle agitation for one hour at room temperature. Western Blots were visualized using Chemiluminescence Luminol Reagent (Santa Cruz Biotechnology, Inc., CA). α-SMA expression was examined in a similar manner by using monoclonal anti-α-SMA antibodies. Bands from individual western blots were quantified densitometrically and the mean ± SEM for each time point or concentration calculated for presentation.
Statistical analysis
All experiments were repeated at least twice. Data are expressed as mean ± SEM. Statistical analysis was performed using ANOVA with the Prism software package (GraphPad, San Diego, CA).
Authors' contributions
SY, MU, CP and WD participated in the PSC isolation and western blot experiments, cell culture experiments and drafted the manuscript. TA, XD, RB, MT and WD participated in the design of the study and trouble-shooting of the experiments. All authors read and approved the final manuscript.
Acknowledgements
Supported by NCI SPORE program (P50 CA72712); the American Institute for Cancer Research (00B056); and the Michael Rolfe Foundation for Pancreatic Cancer.
==== Refs
Watari N Hotta Y Mabuchi Y Morphological studies on a vitamin A-storing cell and its complex with macrophage observed in mouse pancreatic tissues following excess vitamin A administration Okajimas Folia Anat Jpn 1982 58 837 858 7122019
Ikejiri N The vitamin-A storing cells in the human and rat pancreas Kurume Med J 1990 37 67 81 2255178
Apte MV Haber PS Applegate TL Norton ID McCaughan GW Korsten MA Pirola RC Wilson JS Periacinar stellate shaped cells in rat pancreas: identification, isolation, and culture Gut 1998 43 128 133 9771417
Bachem MG Schneider E Gross H Weidenbach H Schmid RM Menke A Siech M Beger H Grunert A Adler G Identification, culture, and characterization of pancreatic stellate cells in rats and humans Gastroenterology 1998 115 421 432 9679048
Yen TW Aardal NP Bronner NP Thorning DR Savard CE Lee SP Bell RH Jr Myofibroblasts are responsible for the desmoplastic reaction surrounding human pancreatic carcinomas Surgery 2002 131 129 134 11854689 10.1067/msy.2002.119192
Ristimäki A Garfinkel S Wessendorf J Maciag T Hla T Induction of cyclooxygenase-2 by interleukin-1α; evidence for post-transcriptional regulation J Biol Chem 1994 269 11769 11775 8163473
Smith WL DeWitt DL Garavito RM Cyclooxygenases: structutal, cellular, and molecular biology Annu Rev Biochem 2000 69 145 182 10966456 10.1146/annurev.biochem.69.1.145
Molina MA Sitja-Arnau M Lemoine MG Frazier M Sinicrope FA Increased cyclooxygenase-2 expression in human pancreatic carcinomas and cell lines: growth inhibition by nonsteroidal anti-inflammatory drugs Cancer Res 1999 59 4356 4362 10485483
Okami J Yamamoto H Fujiwara Y Tsujie M Kondo K Noura S Oshima S Nagano H Dono K Umeshita K Ishikawa O Sakon M Matsuura N Nakamori S Monden M Overexpression of cyclooxygenase-2 in carcinoma of the pancreas Clin Cancer Res 1999 5 2018 2024 10473081
Tucker ON Dannenberg AJ Yang EK Zhang F Teng L Daly JM Soslow RA Masferrer JL Woerner BM Koki AT Fahey TJ III Cyclooxygenase-2 expression is up-regulated in human pancreatic cancer Cancer Res 1999 59 987 990 10070951
Merati K Siadaty MS Andea A Sarkar F Ben-Josef E Mohammad R Philip P Shields AF Vaitkevicius V Grignon DJ Adsay NV Expression of inflammatory modulator COX-2 in pancreatic ductal adenocarcinoma and its relationship to pathologic and clinical parameters Am J Clin Oncol 2001 24 447 452 11586094 10.1097/00000421-200110000-00007
Maitra A Ashfaq R Gunn CR Rahman A Yeo CJ Sohn TA Cameron JL Hruban RH Wilentz RE Cyclooxygenase 2 expression in pancreatic adenocarcinoma and pancreatic intraepithelial neoplasia: an immunohistochemical analysis with automated cellular imaging Am J Clin Pathol 2002 118 194 201 12162677 10.1309/TPG4-CK1C-9V8V-8AWC
Wells RG Crawford JM Pancreatic stellate cells: the new stars of chronic pancreatitis? Gastroenterology 1998 115 491 493 9758536
Haber PS Keogh GW Apte MV Moran CS Stewart NL Crawford DH Pirola RC McCaughan GW Ramm GA Wilson JS Activation of pancreatic stellate cells in human and experimental pancreatic fibrosis Am J Pathol 1999 155 1087 1095 10514391
Efsen E Bonacchi A Pastacaldi S Valente AJ Wenzel UO Tosti-Guerra C Pinzani M Laffi G Abboud HE Gentilini P Marra F Agonist-specific regulation of monocyte chemoattractant protein-1 expression by cyclooxygenase metabolites in hepatic stellate cells Hepatology 2001 33 713 721 11230753 10.1053/jhep.2001.22761
Gallois C Habib A Tao J Moulin S Maclouf J Mallat A Lotersztajn S Role of NF-κB in the antiproliferative effect of endothelin-1 and tumor necrosis factor-α in human hepatic stellate cells. Involvement of cyclooxygenase-2 J Biol Chem 1998 273 23183 23190 9722548 10.1074/jbc.273.36.23183
Subbaramaiah K Dannenberg AJ Cyclooxygenase 2: a molecular target for cancer prevention and treatment Trends Pharmacol Sci 2003 24 96 102 12559775 10.1016/S0165-6147(02)00043-3
Yoshida S Yokota T Ujiki M Ding XZ Pelham C Adrian TE Talamonti MS Bell RH Denham W Pancreatic cancer stimulates pancreatic stellate cell proliferation and TIMP-1 production through the MAP kinase pathway Biochem Biophys Res Commun 2004 323 1241 1245 15451430 10.1016/j.bbrc.2004.08.229
Ding XZ Tong WG Adrian TE Multiple Signal Pathways Are Involved in the Mitogenic Effect of 5(S)-HETE in Human Pancreatic Cancer Oncology 2003 65 285 294 14707447 10.1159/000074640
Gao J Niwa K Takemura M Sun W Onogi K Wu Y Seishima M Mori H Tamaya T Significant anti-proliferation of human endometrial cancer cells by combined treatment with a selective COX-2 inhibitor NS398 and specific MEK inhibitor U0126 Int J Oncol 2005 26 737 744 15703831
Tsujii M DuBois RN Alterations in cellular adhesion and apoptosis in epithelial cells overexpressing prostaglandin endoperoxide synthase 2 Cell 1995 83 493 501 8521479 10.1016/0092-8674(95)90127-2
Tsujii M Kawano S DuBois RN Cyclooxygenase-2 expression in human colon cancer cells increases metastatic potential Proc Natl Acad Sci USA 1997 94 3336 3340 9096394 10.1073/pnas.94.7.3336
Tsujii M Kawano S Tsuji S Sawaoka H Hori M Dubois RN Cyclooxygenase regulates angiogenesis induced by colon cancer cells Cell 1998 93 705 716 9630216 10.1016/S0092-8674(00)81433-6
Ding XZ Tong WG Adrian TE Blockade of cyclooxygenase-2 inhibits proliferation and induces apoptosis in human pancreatic cancer cells Anticancer Res 2000 20 2625 2631 10953335
Leahy KM Ornberg RL Wang Y Zweifel BS Koki AT Masferrer JL Cyclooxygenase-2 inhibition by celecoxib reduces proliferation and induces apoptosis in angiogenic endothelial cells in vivo Cancer Res 2002 62 625 631 11830509
Jabbour HN Boddy SC Prostaglandin E2 induces proliferation of glandular epithelial cells of the human endometrium via extracellular regulated kinase 1/2-mediated pathway J Clin Endocrinol Metab 2003 88 4481 4487 12970327 10.1210/jc.2003-030297
Elder DJ Halton DE Hague A Paraskeva C Induction of apoptotic cell death in human colorectal carcinoma cell lines by a cyclooxygenase-2 (COX-2)-selective nonsteroidal anti-inflammatory drug: independence from COX-2 protein expression Clin Cancer Res 1997 3 1679 1683 9815550
Lowry OH Rosebrough NJ Farr AL Randall RJ Protein measurement with the folin phenol reagent J Biol Chem 1951 193 265 275 14907713
|
16083499
|
PMC1201567
|
CC BY
|
2021-01-04 16:36:35
|
no
|
Mol Cancer. 2005 Aug 5; 4:27
|
utf-8
|
Mol Cancer
| 2,005 |
10.1186/1476-4598-4-27
|
oa_comm
|
==== Front
Nutr Metab (Lond)Nutrition & Metabolism1743-7075BioMed Central London 1743-7075-2-151598515510.1186/1743-7075-2-15ResearchRegulation of mouse hepatic genes in response to diet induced obesity, insulin resistance and fasting induced weight reduction Raab R Michael [email protected] John [email protected] Joanne [email protected] Christos [email protected] Gregory [email protected] Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA2 Beth-Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, USA3 Department of Physiology, The George Washington School of Medicine and Health Sciences, Washington, DC, USA2005 28 6 2005 2 15 15 25 4 2005 28 6 2005 Copyright © 2005 Raab et al; licensee BioMed Central Ltd.2005Raab 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
Obesity is associated with insulin resistance that can often be improved by caloric restriction and weight reduction. Although many physiological changes accompanying insulin resistance and its treatment have been characterized, the genetic mechanisms linking obesity to insulin resistance are largely unknown. We used DNA microarrys and RT-PCR to investigate significant changes in hepatic gene transcription in insulin resistant, diet-induced obese (DIO)-C57/BL/6J mice and DIO-C57/BL/6J mice fasted for 48 hours, whose weights returned to baseline levels during these conditions.
Results
Transcriptional profiling of hepatic mRNA revealed over 1900 genes that were significantly perturbed between control, DIO, and fasting/weight reduced DIO mice. From this set, our bioinformatics analysis identified 41 genes that rigorously discriminate these groups of mice. These genes are associated with molecular pathways involved in signal transduction, and protein metabolism and secretion. Of particular interest are genes that participate in pathways responsible for modulating insulin sensitivity. DIO altered expression of genes in directions that would be anticipated to antagonize insulin sensitivity, while fasting/ weight reduction partially or completely normalized their levels. Among these discriminatory genes, Sh3kbp1 and RGS3, may have special significance. Sh3kbp1, an endogenous inhibitor of PI-3-kinase, was upregulated by high-fat feeding, but normalized to control levels by fasting/weight reduction. Because insulin signaling occurs partially through PI-3-kinase, increased expression of Sh3kbp1 by DIO mice may contribute to hepatic insulin resistance via inhibition of PI-3-kinase. RGS3, a suppressor of G-protein coupled receptor generation of cAMP, was repressed by high-fat feeding, but partially normalized by fasting/weight reduction. Decreased expression of RGS3 may augment levels of cAMP and thereby contribute to increased, cAMP-induced, hepatic glucose output via phosphoenolpyruvate carboxykinase (PCK1), whose mRNA levels were also elevated.
Conclusion
These findings demonstrate that hepatocytes respond to DIO and weight reduction by controlling gene transcription in a variety of important molecular pathways. Future studies that characterize the physiological significance of the identified genes in modulating energy homeostasis could provide a better understanding of the mechanisms linking DIO with insulin resistance.
==== Body
Background
Obesity is a growing concern in the industrialized world. It is estimated that over 61% of adult Americans are overweight or obese [1] and an alarming number of children and adolescents are following suit [2]. Of primary concern are the associated complications stemming from obesity's growing prevalence, among which type 2 diabetes is reaching epidemic proportions.
The aetiology of type 2 diabetes is complex because of its heterogeneous origins that result in the commonly observed hyperglycemia and hyperinsulinemia, which are characteristic of insulin resistance. While an enormous number of investigations have resulted in identifying some of the relevant molecular pathways, particularly in muscle and adipose tissue, more research is required to fully understand genetic susceptibility to type 2 diabetes and insulin resistance.
In the liver, hepatic glucose output (HGO) increases during insulin resistance and several key molecules contributing to this phenotype have been widely studied [3-6]. Despite these extensive efforts, the genes identified thus far do not alone account for all of the variability in HGO. Further insight may be obtained by conducting genome wide transcriptional studies during diet induced obesity (DIO) and its associated insulin resistant physiological state. This approach is a critical step towards further defining the molecular processes that regulate the phenotype and thereby augment the discovery of new potential therapeutic targets.
C57/BL/6J mice fed a high-fat diet become obese, hyperglycemic, and hyperinsulinemic, reflecting an insulin resistant metabolic state [7-11] that resembles the human condition. Although it has been demonstrated that short-term caloric restriction can improve insulin resistance [12], the regulatory pathways that control hepatic metabolism during DIO and associated insulin resistance, and the improvement of insulin resistance with caloric restriction, are the focus of intense research efforts. The molecular mechanisms underlying these pathways rely upon alterations in gene transcription [13], which can be monitored using DNA microarrays [14,15].
To investigate hepatic gene regulation in response to DIO and insulin resistance, whole genome microarrays containing 17,280 gene probes were used to examine transcription in two groups of C57/BL/6J mice : 1) the "control mice" received a normal diet for 10 weeks, 2) the "high-fat mice" received a high-fat diet for 10 weeks. In addition, to assess hepatic gene regulation in response to caloric restriction, which is a commonly recommended treatment for DIO and insulin resistance, a third group of mice was used, the "fasted/ weight reduced mice", which was fed the same high-fat diet for ten weeks followed immediately by 48 hours of fasting, returning their weights to baseline levels prior to tissue harvest. Fasting/ weight reduction data provides further differentiation among genes that not only respond to DIO and insulin resistance, but are also normalized by caloric restriction.
An extensive bioinformatics analysis led to the identification of 41 discriminatory genes participating in key molecular pathways in DIO, insulin resistance, and fasting/ weight reduction. The implicated pathways involve signal transduction and protein metabolism and secretion. In addition, the 41 genes identified can accurately classify the three groups of mice ("control", "high-fat", and "fasted/ weight reduce"), and importantly, they represent a set of candidate genes that may influence hepatic function during periods of insulin resistance and sensitivity.
Methods
Animals
Three to five week old C57/BL/6J mice were obtained from Jackson Laboratories (Bar Harbor, ME). All animals were allotted a seven day acclimation period with access to food and water ad libitum, and were maintained at 25°C with a 12-hour light/ dark cycle (lights on from 06:30–18:30) for the duration of the study. A normal chow (Purina Rodent Chow; Harlan Teklad #5008; 6.5% fat, 49% carbohydrate, 23% protein, 3.5 kcal/g) and high-fat diet (Harlan Teklad #TD88137, 42.16% fat, 42.81% carbohydrate, 15.02% protein, 4.53 kcal/g) were fed to respective mice, as outlined below.
This report explored alterations in hepatic gene mRNA levels in C57/BL/6J mice fed either a control or high-fat diet for 10 weeks, as well as alterations in mRNA levels of C57/BL/6J mice fasted for 48 hours following 10 weeks of high-fat feeding. Fasted animals were allowed access to water during the fasting period. All animals were sacrificed by CO2 asphyxiation, followed by immediate collection of liver tissue, which were stored at -80°C as previously described [16].
The control group consisted of C57/BL/6J mice fed normal chow diet for 10 weeks. The experimental group consisted of C57/BL/6J mice fed a high-fat diet for 10 weeks (n = 9/group). The ten week high-fat dietary treatment has been demonstrated to be long enough for C57/BL/6J mice to develop insulin resistance and a condition that resembles type 2 diabetes [7,8]. Two days before tissue harvest, the C57/BL/6J mice on the high-fat diet were divided into two groups, with one group remaining on the high-fat diet (n = 5; to be used in the first study) and one group fasting for the final 48 hours (n = 4; to be used in the second study). Mouse weights were recorded two days prior to, and on the day of tissue harvest. All animals were handled in accordance with the principles and guidelines established by the National Institutes of Health. The protocol was approved by the Institutional Review Board at Beth Israel Deaconess Medical Center, Boston, MA.
Preparation of total RNA and cDNA for microarray hybridization
Total RNA was purified from liver tissue samples using STAT-60 (Tel-Test, Inc., Friendswood, TX) according to the manufacturer's instructions, and stored at -80°C. Labeled control cDNA was made from Total RNA control samples (Universal Mouse Reference RNA, catalog #740100, Stratagene) using Cy3 dCTP (Perkin-Elmer), and labeled liver cDNA was made from total RNA experimental samples using Cy5 dCTP (Perkin-Elmer) during reverse transcription, as described previously [17].
Microarrays were prepared using GAPS glass slides (Corning) and a Virtek arrayer (Bio-Rad). Arrays contained 17,280 features, printed from a synthesized oligonucleotide mouse library (Operon) as described previously [17].
RT-PCR analysis of IL6st, PTP4a2, G6P, PCK1, and malic enzyme
A two-step RT-PCR protocol was performed to confirm the mRNA levels of several genes. In this procedure the cDNA synthesis was performed as detailed previously [17] except the Cy-labeled nucleotides were replaced with unlabeled nucleotides such that all dNTPs were at the same final concentration during the reaction. PCR was conducted in 94-well plates using the iQ SYBR Green Supermix Kit (Bio-Rad), according to the manufacturer's instructions on an iCycler RT-PCR machine (Bio-Rad). Briey, 1 μL of the final, diluted cDNA template was mixed with 19 μL of RNase free water, 25 μL of Bio-Rad RT-PCR Supermix (Bio-Rad), 2 μL of sense and antisense primers, and 1 μL of 12.5 mM dNTPs. The final primer concentration was 0.25 μM. The PCR cycle used a single three minute hot-start at 95°C, followed by 50 cycles of 30 seconds at 95°C, one minute at 60°C, and two minutes at 72°C during which time the reaction fluorescence was measured. Each mouse sample was measured in either triplicate or quadruplicate. The sense and antisense primer sequences were: for interleukin 6 signal transducer IL6st 5'-GCGGCTCGAACTTCACTGC-3', and 5'-CACGATGTAGCTGGCATTCACG-3'; for protein tyrosine phosphatase 4a2 PTP4a2 5'-TTTCTGCTGCGGAACATTTCAAG-3', and 5'-GCGTGCGTGTGTGAGTGTG-3'; for regulator of g-protein signalling 3 RGS3 5'-GCACATCCCGCATTCCAGTTAC-3', and 5'-AGGGAACACCAGGACTTTAGGG-3'; for glucose-6-phosphatase G6P 5'-GTGATTGCTGACCTGAGGAACG-3', and 5'-TGCCACCCAGAGGAGATTGATG-3'; for phosphoenolpyruvate carboxykinase PCK1 5'-CAGAGAGACACAGTGCCCATCC-3', and 5'-AAGTCCTCTTCCGACATCCAGC-3'; for malic enzyme 5'-GCCAGAGGATGTCGTCAAGG-3', and 5'-ATTACAGCCAAGGTCTCCCAAG-3', respectively. These primers each gave specific fragments of the correct length when viewed upon a 4% agarose gel (data not shown). As an internal control β-Actin mRNA levels were also measured. The sense and antisense sequences were 5'-AATAAGTGGTTACAGGAAGTC-3' and 5'-ATGAAGTATTAAGGCGGAAG-3', respectively.
Gene specific standards were developed by amplifying the entire mRNA coding sequence of each gene by PCR, gel purifying the resulting band, and then diluting it to concentrations from 104 μg/μL to 10 9 μg/μL. The R2 value of the standard curve, relating the threshold cycle to the amount of standard template, was always greater than 0.97. The mRNA levels of β-actin measured were not significantly (p > 0.05) different between the dietary treatments for any of the groups.
Array validation
Microarray protocols have been extensively validated in our laboratory as described previously [17]. For validation, we prepared arrays containing an approximately 13,000 gene sub-set of our oligonucleotide mouse library, printed in triplicate. Total RNA from skeletal muscle and brain tissue were used for validation comparisons, and each sample was analyzed in duplicate and prepared and processed as described above. Matlab was used to calculate basic statistics.
The arrays' ability to detect differential transcription between muscle and brain RNA was evaluated by two different methods. In the first, we examined the number of genes that were up- or down-regulated by a factor greater than two (i.e., whose mean ratio was either greater than two, or less than 0.5) in the muscle versus muscle and the muscle versus brain RNA comparisons. This criterion has been used as a basis for assessing differential transcription in a number of studies [18-20]. In the second method, we defined a threshold for differential expression by using the 95% confidence interval determined from the muscle versus muscle control arrays. Table 1 summarizes the results, where the p-values reported were from two-tailed student t-tests.
Table 1 Differential gene transcription validation data. This table summarizes the results of the array validation with respect to the study of differential expression.
Array Condition # of Probes Detected # of Genes >2-Fold Different Differentially Genes at the Expressed 95% Confidence Level
Muscle vs. Muscle 7574 438 429
Muscle vs. Muscle 6417 314 302
Average 6996 376 366
Muscle vs. Brain 7143 1201 1161
Muscle vs. Brain 8318 981 931
Average 7731 1091 1046
P-value 0.47 0.03 0.03
Although there are only about 370 genes exceeding the threshold in the muscle versus muscle arrays, more than 1000 genes were differentially expressed in the muscle versus brain arrays. This result supports the assertion that the assaying method and selection criterion are significantly more likely to identify differentially expressed genes.
The coefficient of variation, CV, was calculated for each replicated gene expression and the distribution across all genes is plotted in Figure 1. For the muscle versus muscle control arrays, the median CV across all probes was 10.2%. For the muscle versus brain arrays the median coefficient of variation across all probes was 9.8%. This indicates that for a gene transcription ratio of 1, we might expect the true value to lie between 0.9 and 1.1; similarly for a gene transcription ratio of 3, we might expect the true value to lie between 2.7 and 3.3. Although the median CV across all probes for the muscle versus muscle control arrays was 10.2%, the median CV for the 314 genes common to both muscle versus muscle arrays that had a fold difference greater than two, was 24.7%. Because of their increased CV and high fold change, none of these genes were included in our subsequent analysis.
Figure 1 Distribution of the coefficient of variation for DNA microarrays. The coefficient of variation was calculated for every gene in the experiment, and plotted for the muscle versus muscle and muscle versus brain.
In duplicate arrays, 76% of the genes observed on one muscle versus muscle array were also observed on the duplicate; likewise 77% of the genes found on one muscle versus brain array were conserved on the duplicate. These data demonstrate the inter-array reproducibility by showing the majority of genes are reproducibly found in multiple replicate arrays.
RT-PCR was also used to verify the array results for IL6st, PTP4a2, and RGS3. The variation in the ratios of the mRNA levels was less than 30% for each of these genes whether measured using the arrays or RT-PCR as shown in Table 2.
Table 2 Comparison of array results and RT-PCR results for selected genes. Gene expression percentages are reported relative to the control values. F/ WR: Fasting Weight Reduction.
Genes Assay High Fat vs. Control F/ WR vs. Control
IL6st Array 154 ± 21%* 144 ± 21%*†
RT-PCR 167 ± 19%* 185 ± 15%*†
PTP4a2 Array 71 ± 4%* 89 ± 3%*
RT-PCR 75 ± 16% 94 ± 18%
RGS3 Array 35 ± 5%* 54 ± 8%*
RT-PCR 38 ± 9%* 59 ± 8%*
G6P RT-PCR 476 ± 72% 769 ± 216%*
PCK1 RT-PCR 132 ± 28% 217 ± 80%*
Malic Enzyme RT-PCR 9.1 ± 1.5%* 0.1 ± 0.1%*
*Indicates that the measurements were significantly different from control values at P < 0.01.
†Indicates that the measurements made on the micro array were significantly different from the RT-PCR measurement at P < 0.05.
Computational methods
A combination of statistical and data mining methods were used to extract information from the microarray data. Statistical methods rigorously quantify the reliability of differences in the microarray data [21] and can objectively evaluate changes in gene transcription ratios and derivative quantities. Data mining is particularly useful for uncovering patterns and structure in microarray data that might have otherwise been difficult to detect through manual inspection and intuition alone [22,23]. Applying statistics and data mining methods to microarray data in unison enables rapid and reliable analysis without a priori assumptions that may bias expectations about the data set.
A t-test [24] was used to evaluate whether a gene exhibited statistically significant expression differences in pairwise comparisons between the control, high-fat, and fasting/ weight reduced groups. The t-test results showed that 1981 genes had at least one statistically significant (p < 0.05) change between the treatments. Wilks-λ based ranking [25] was used to identify discriminatory genes that differentiated the three groups. This technique is particularly appropriate for multi-class comparisons, ranking genes on the basis of their within group, and between group variances. Thus, a gene exhibiting a small variation within each of the three groups, but a large variation between groups would rank highly; conversely, a gene that had a high level of variation within each group and a low level of variation among the groups would be ranked low. The Wilks-λ score can be transformed into an F statistic, which can be compared with the F distribution to assess the statistical significance of the observation [25]. In this analysis a Wilks-λ threshold value of 0.47 was used, which is equivalent to a p value of 0.05. From the 1981 genes selected by the p < 0.05 cutoff, we retained the 1169 genes that had a Wilks-λ value below 0.47.
Fischer Discriminant Analysis [26] (FDA) was used to identify not just individual genes, but combinations of genes whose expression levels are capable of correctly classifying the control mice, high-fat mice, and fasting/ weight reduced mice. FDA is based on linear combinations of gene expressions and considers the discriminatory power of gene groups as opposed to individual genes. Samples are scored based on the weighted contribution of each gene's expression level to a newly defined metric called a "canonical variable" (CV). Because each gene's contribution to a sample's score is weighted by a coefficient called a "loading," genes with very small loadings do not significantly contribute to the sample's score and classification, and can therefore be eliminated from further consideration. This technique can be used as a tool to visualize the gene transcription results in a lower dimensional space defined by the canonical variables. As shown in Figure 2, using expression data of the selected gene combinations allows accurate classification of the dietary treatments suggesting that the genes in Table 3 (See Additional file: 1) are discriminatory of the conditions examined when sample classification is used as a criterion. On the basis of the successful classification afforded by the FDA projection, discriminatory genes were selected using the magnitude of the loading coefficients. Principle Component Analysis [27] was used as an unsupervised classification procedure to complement FDA. The results of the PCA analysis largely mirrored the FDA results (data not shown).
Figure 2 Fisher discriminant analysis plot of mouse liver samples. Samples were scored according to the canonical variables determined by Fisher Discriminant Analysis (FDA). Each canonical variable is defined as a weighted sum of 100 specific genes, including each of the 41 genes contained in Table 3 (See Additional file: 1). To score a sample, the gene expression value is multiplied by an FDA coefficient, called a loading, and the products from the 100 genes used in the analysis are summed to give the canonical variable score for the sample. F/ WR: Fasting/ Weight Reduced.
Methods used here, along with the data set, are available for public use at our laboratory's web-site [28]. The entire data set is also available through the National Center for Biotechnology Information's Gene Expression Omnibus database [29].
Results
The effect of 10 weeks of high-fat feeding and 48 hours of caloric restriction on body weight in C57/BL/6J mice
C57/BL/6J mice significantly increased their body weight by 32% after 10 weeks of high-fat feeding (p < 0.001; Table 4). After 48 hours of fasting, their weights returned to baseline levels and were not significantly different from the control mice, but were significantly less than mice maintained on the high-fat diet (p < 0.001; Table 4).
Table 4 Experimental treatments and mouse weights.
Diet Feeding Regimen Weight 48 hours Prior to Harvest (Average ± St. Dev., n) Weight at Harvest (Average ± St. Dev., n)
Normal Chow Ad libitum 35.6 ± 1.8, 9 35.6 ± 1.5, 9
High-Fat Ad libitum 47.1 ± 5.8*, 9 51.7 ± 4.4*†,5
High-Fat Restricted 37.3 ± 2.6,4
*Indicates that the weight was statistically different from the control at P < 0.001.
† Indicates that the weight of the high-fat and fasted mice was different at P < 0.001.
Microarray analysis of hepatic genes after 10 weeks of high-fat feeding and 48 hours of fasting/ weight reduction in C57/BL/6J mice
Employing statistical and data mining methods we searched the transcription data set for hepatic genes that direct the biological response during DIO, associated insulin resistance, and fasting/ weight reduction. We used the t-test to determine the statistical significance of every pairwise gene difference between the treatments. The t-test showed that 1981 genes had at least one statistically significant (p < 0.05) change between the treatments. Within this gene set, 113 genes were significantly changed between the high-fat fed mice and the control mice, 169 genes were significantly changed between the fasting/ weight reduced mice and the control mice, and 260 genes were significantly changed between the high-fat fed and fasting/ weight reduced mice, all at p < 0.01. From the 1981 genes selected by the p < 0.05 cutoff, we retained the 1169 genes that had a Wilks-λ value below our cutoff criterion of 0.47, which is equivalent to a p-value of less than 0.05 [25]. From these genes we selected those with the greatest Fisher Discriminant Analysis (FDA) and Principle Component Analysis (PCA) loading coefficients [27], resulting in the 41 genes reported in Table 3 (See Additional file: 1).
The 41 discriminating genes contributed to the classification observed in Figure 2. In Figure 2, each sample is given a canonical variable (CV) score, based on the weighted sum of its gene expression values. The genes with the largest contributions to CV1 and CV2 are given in Table 3 (See Additional file: 1), suggesting these genes underlie the biological differences between the samples. Figure 2 shows that 10 weeks of high-fat feeding altered the transcriptional levels of genes composing CV1 and CV2 so as to separate the control and high-fat mice in the CV1 and CV2 space. However, while 48 hours of fasting/ weight reduction normalized many of the genes contributing to CV2, resulting in a return to control levels for that variable, the genes contributing to CV1 remained perturbed, resulting in the observed separation between the fasted/ weight reduced mice and control mice. This suggests that while some genes, and their associated pathways that differentiate DIO and insulin resistance from normal physiology, return to control levels as weight is reduced, other genes remain perturbed, reflecting further physiological adaptations that occur during these treatments. To show individual gene responses to the dietary treatments, the 41 genes were clustered according to changes in the p-values from pairwise comparisons between the control mice, the high-fat fed mice, and the fasting/ weight reduced mice. This classification arranges the genes according to their transcript levels during the physiological states examined. For example, Group A in Table 3 (See Additional file: 1) comprises genes that were significantly elevated or repressed (p < 0.05) by high-fat feeding, but then normalized to (insignificant, p > 0.05) control levels by fasting and weight reduction. Similarly, group B genes were significantly elevated or repressed (p < 0.05) by high-fat feeding and partially normalized to control levels by fasting/ weight reduction: the expression differences are still significant (p < 0.05) when comparing both the high-fat and control mice with the fasted/ weight reduced mice. The genes of each group along with their normalized expression levels are given in Table 3 (See Additional file: 1).
Among the 41 discriminatory genes identified in this study, interleukin 6 signal transducer (IL6st), protein tyrosine phosphatase 4a2 (PTP4a2), SH3-domain kinase binding protein 1 (Shk3bp1), and regulator of g-protein signaling 3 (RGS3) are of special interest because, based on known biology, they may contribute to the physiological changes that accompany DIO, insulin resistance, and increased insulin sensitivity due to fasting/ weight reduction. Both IL6st and Sh3kbp1 are significantly upregulated after 10 weeks of high-fat feeding (p < 0.001), but only Sh3kbp1 is normalized to baseline levels after 48 hours of fasting and weight reduction (Table 3: See Additional file: 1). Both PTP4a2 and RGS3 are significantly downregulated after 10 weeks of high-fat feeding (p < 0.01), and both are partially normalized after 48 hours of fasting/ weight reduction (p < 0.01 for fasted/ weight reduced versus high-fat and fasted/ weight reduced versus control; Table 3: See Additional file: 1).
RT-PCR analysis of IL6st, PTP4a2, RGS3, G6P, PCK1, and malic enzyme
We compared the transcript levels measured by RT-PCR with the ratios measured using DNA microarrays by dividing RT-PCR expression values observed in high-fat fed mice and fasted/ weight reduced mice by the expression values measured in the control mice. Liver mRNA levels for each mouse in the study were determined by RT-PCR for IL6st, PTP4a2, and RGS3. The values measured by RT-PCR were not significantly different from the results observed by hepatic microarray analysis (p > 0.05; Table 2) for all genes except IL6st between the fasting/ weight reduced mice and control mice. Notably, in this single case, both microarray analysis and RT-PCR show significant increases (p < 0.001) in the levels of IL6st mRNA, demonstrating similar qualitative changes between the measurement methods. The close agreement between the micoarray results and RT-PCR results thus validates the specificity and accuracy of our microarray measurements. The difference in the ratios between the values determined by RT-PCR and those determined by microarray analysis was less than 30% for each of these genes (Table 2).
Although several commonly studied genes, such as glucose-6-phosphatase (G6P), phosphoenolpyruvate carboxykinase (PCK1), and malic enzyme, did not make it into our bioinformatics analysis, we evaluated their expression by RT-PCR because of their considerable effects on hepatic glucose output. G6P and PCK1 were upregulated following 10 weeks of high-fat feeding, but only the change observed in G6P achieved statistical significance (p = 0.09 for PCK1 and p < 0.01 for G6P in the high-fat versus control comparison; Table 2). Fasting/ weight reduction resulted in even larger increases in mRNA levels for both G6P and PCK1 (p < 0.01 versus controls; Table 2). In contrast, malic enzyme exhibited significant underexpression following 10 weeks of high-fat feeding, with further down-regulation following fasting/ weight reduction (Table 2).
Discussion
Diet induced obesity (DIO) in C57/BL/6J mice is a commonly used animal model for the development of insulin resistance in humans [7-11], which results in simultaneous hyperglycemia and hyperinsulinemia. Although short-term caloric restriction and weight loss can improve insulin resistance [12,30,31], the regulatory mechanisms in the liver that lead to insulin resistance in response to DIO, as well as the improvement of insulin sensitivity in response to short-term caloric restriction and weight reduction, remain largely unknown. To identify genes involved in hepatic physiology during DIO and short-term caloric restriction, we used DNA microarrays to measure genome-wide transcript abundance.
The 41 most discriminating genes determined by our bioinformatics analysis lie essentially within two large groups (Table 3: See Additional file: 1): 1) Genes that are significantly induced or repressed by 10 weeks of high-fat feeding and completely (Group A) or partially (Group B) normalized by 48 hours of fasting/ weight reduction, 2) Genes that are significantly induced or repressed by 10 weeks of high-fat feeding, but are not normalized by 48 hours of fasting/ weight reduction (Group D). Both of these groups contain genes involved in signal transduction pathways, as well as protein metabolism and secretion, highlighting the importance of these molecular pathways in the hepatic response to DIO and fasting/ weight reduction.
Because genes in Group A and B (Table 3: See Additional file: 1) were perturbed by DIO, their expression levels correlate with observed physiological differences that develop during this condition. These differences include elevated concentrations of serum triglycerides, leptin, and tumor necrosis factor-α, as well as changes in the levels of other factors that have been previously demonstrated to play a physiological role during DIO in C57/BL/6J mice [7-9,11,32]. Notably, Group A and B genes are either completely (Group A, Table 3: See Additional file: 1) or partially (Group B, Table 3: See Additional file: 1) normalized following 48 hours of fasting/ weight reduction, when insulin sensitivity has increased, suggesting they may be important to the development of hepatic insulin resistance during DIO. Several relevant signal transduction pathways are influenced by the genes within Group A and B (Table 3: See Additional file: 1), particularly Sh3kbp1, PTP4a2, and RGS3. While Sh3kbp1 and PTP4a2 may be directly involved with insulin signaling, by respectively binding PI-3-kinase and dephosphorylating protein tyrosine residues, RGS3 interacts directly with G-proteins and some evidence suggests RGS family members may also indirectly affect proteins in the MAPK signal transduction pathways [33] as well as certain tyrosine phosphatases [34].
Sh3kbp1 (SH3-domain kinase binding protein, also called Ruk) belongs to the CD2AP/CMS family of adapter-type proteins, which mediate a number of different cellular mechanisms including signal transduction [35]. Insulin signaling occurs via phosphorylation of insulin receptor substrates (IRSs) that interact with signal transduction molecules including PI-3-kinase, Grb2, nck, and SHP2 [36]. Sh3kbp1 has been shown to directly inhibit PI-3-kinase signaling by binding the p85α regulatory subunit in vivo and in vitro, and interacts with Grb2 in vitro [37]. Therefore, increased levels of Sh3kbp1 mRNA in the high-fat fed mice relative to both the control and fasted/ weight reduced mice, suggests that Sh3kbp1 may mediate DIO associated insulin resistance in hepatocytes via a mechanism described in Figure 3.
Figure 3 Inhibition of PI-3-Kinase signaling by Sh3kbp1. In this figure, insulin, I, binds to its receptor, activating the receptor's tyrosine kinase activity. Insulin receptor substrates, IRS, are activated by phosphorylation. IRS phosphorylates PI-3-kinase, which migrates to the cell membrane where it generates phosphatidylinositol, PI, second messengers, which alters physiological processes. Shown here, Sh3kbp1 is capable of binding the regulatory subunit of PI-3-kinase, inhibiting its ability to generate PI second messengers, and thereby attenuating insulin signaling.
PTP4a2 (Protein tyrosine phosphatase 4a2) dephosphorylates tyrosine residues in proteins. When insulin binds its receptor it activates the receptor's tyrosine kinase activity [38], leading to autophosphorylation and subsequent tyrosine phosphorylation of molecules containing Src homology 2 (SH2) or phosphotyrosine binding (PTB) domains. Therefore PTPs can in fluence insulin signaling by dephosphorylating protein tyrosine residues. Although it would be anticipated that PTPs would attenuate insulin signaling, they have been implicated in both positive and negative regulation of this pathway [39]. A definitive role for many PTPs in glucose homeostasis and insulin signaling has not been established, however, PTP1B knock-out mice have enhanced insulin sensitivity and are resistant to DIO [40]. Therefore if PTP4a2 also negatively regulates insulin signaling, its significant downregulation (p < 0.01) following 10 weeks of high-fat feeding may be a physiological adaptation that protects hepatocytes against insulin resistance, which is normalized by fasting/ weight reduction.
RGS3 (Regulator of G-protein coupled receptor (GPCR) signaling 3) has been primarily studied in neurons [41-43] and cells in culture [44,45]. RGS proteins bind Gα subunits and generally increase the GTPase activity [46]. We found that hepatic RGS3 mRNA levels are significantly decreased (p < 0.01) after 10 weeks of high-fat feeding, but partially normalized by fasting/ weight reduction. These findings are particularly relevant because hepatocytes express a truncated form of RGS3 that has been shown to directly inhibit Gsα stimulated cAMP production and Gqα stimulated IP production [47], in addition to interacting with, Giα [48]. Glucagon signals via a GPCR that stimulates adenyl cyclase and increases cAMP levels [49]. Because the truncated form of RGS3 inhibits cAMP production, lowering RGS3 concentration may augment basal cAMP levels and thereby promote hepatic glucose output resulting from cAMP induced phosphoenolpyruvate carboxykinase (PCK1) expression and cAMP repressed glucokinase transcription. Although glucokinase expression levels were not measured, PCK1 mRNA levels were increased by both 10 weeks of high-fat feeding and fasting/ weight reduction (Table 2).
While genes in Group D (Table 3: See Additional file: 1) were also significantly induced or repressed following 10 weeks of high-fat feeding, unlike genes in Group A and B, they do not respond to 48 hours of fasting/ weight reduction. Therefore hepatic regulation of Group D genes may not be as directly linked to changes resulting from DIO and insulin resistance and sensitivity. Despite this, it is interesting that a number of Group D genes are also implicated in several signal transduction pathways that may be activated by DIO. These genes include BMP2, Fosb, Gabrr1, IL6st, and 4833414G15Rik.
BMP2 (Bone morphogenetic protein 2), is a highly conserved member of the transforming growth factor-β (TGF-β) gene family. BMP2 is related to BMP9, which was the first reported hepatic factor shown to decrease blood glucose levels by increasing insulin release and decreasing food intake [50]. While these mechanisms may be a compensating response to DIO, they oppose the physiological adaptations that accompany 48 hours of fasting/ weight reduction, and therefore additional studies are required to determine the effects of BMP2 upregulation in mice following these dietary treatments.
FosB is a member of the AP-1 family of transcription factors [51]. These molecules are considered immediate early genes, because they initiate responses to environmental stimuli [52]. The Fos family of transcription factors form either homodimers with one another, or heterodimers with the Jun family of transcription factors, which then bind DNA to alter gene transcription [53]. Because insulin affects the expression of members of the AP-1 family of transcription factors [54], it is not surprising that during DIO and fasting/ weight reduction, conditions that perturb insulin signaling, significantly increase transcription of FosB.
IL6st (Interleukin 6 signal transducing subunit, also called gp130) is a key component in cytokine signal transduction that occurs during inflammation through the JAK (Janus kinase)/ STAT (signal transducers and activators of transcription) pathway. IL6st forms homo- and heterodimers with other signal transducing subunits in response to binding by an assortment of ligands including IL-6, IL-11, LIF, CT-1, CNTF, and OSM [55]. Among these, IL-6 knockout mice develop mature-onset obesity [56], and treatment of hepatocytes with IL-6 reduces the expression of PCK1 [6], thus implicating IL-6 in the regulation of hepatic glucose output. There are at least four different Jaks (Jak1, Jak2, Jak3, and Tyk2) and seven different STAT factors (STAT1, 2, 3, 4, 5a, 5b, and 6) that can interact with IL6st. Of particular relevance to DIO and insulin resistance is STAT3. The liver-specific STAT3 knockout mouse is insulin resistant and develops glucose intolerance when fed a high-fat diet, due in part to increased expression of PCK1 and G6P [57]. Adenoviral mediated reconstitution of STAT3 signaling ameliorated glucose intolerance in both L-ST3KO and Lepr-/- mice [57] by lowering PCK1 and G6P levels, demonstrating the importance of STAT3 signalling to hepatic glucose output. Because IL6st is significantly upregulated (p < 0.001) by 10 weeks of high-fat feeding and 48 hours of fasting/ weight reduction, when PCK1 and G6P were also induced relative to control levels (Table 2), it may be that IL6st performs a sensitizing function that contributes to feedback control of hepatic glucose output via IL6 and STAT3 signaling. In addition to the cellular signaling pathways that contained differentially expressed genes identified in this study, a number of genes involved in protein metabolism and secretion were also identified. Although a direct link between protein metabolism/ secretion and DIO/ insulin resistance is not as well established, in other insulin sensitive tissues the release of hormones and trafficking of receptors clearly plays a role in regulating tissue specific responses to insulin and glucose. Group A and B genes involved in protein metabolism and secretion pathways include Kcnk8, Pmm1, Serpina5, and Eif4a2. Group D genes that were identified include Copz2, Rab3c, and 4933432M07Rik.
Serpina5, encodes a serine protease inhibitor. Serine protease inhibitors represent a family of glycoproteins that are known to inactivate serine proteases by forming stoichiometric enzyme-inhibitor complexes. Among the proteases known to be inhibited by Serpins are trypsin, chymotrypsin, the sperm protease acrosin, and a variety of proteases involved in hemostasis [58]. Copz2 encodes a vesicle coating protein that helps to mediate vesicle trafficking, while Rab3c is a member of the Ras oncogene family that encodes a monomeric GTP-binding protein that is implicated in regulated exocytosis and vesicle transport, and has been suggested to play a role in GLUT4 translocation in rat cardiac muscle cells [59]. Hence, Copz2 and Rab3c may synergistically influence protein trafficking in response to 10 weeks of high-fat feeding and 48 hours of fasting/ weight reduction.
Conclusion
Using DNA microarrays we have investigated the effects of DIO and fasting/ weight reduction on liver gene transcription. We have analyzed this data set using four computational methods that represent a rigorous approach to analysis requiring no a priori assumptions about the data. This has enabled us to infer the importance of any given gene change among a multitude of gene differences resulting from DIO and fasting/ weight reduction. Our results lead us to focus on 41, out of an initial 1981 genes.
Although many of the genes resulting from our analysis have not yet been studied extensively in the context of energy homeostasis, several are related to important molecular pathways that have been previously identified in the literature. Those pathways include different signal transduction cascades, as well as pathways involved in protein metabolism and secretion. Given the diverse functions of the liver, identifying genes involved in signaling and protein metabolism pathways in response to DIO and fasting/ weight reduction is not surprising. Among the genes involved in signaling are Sh3kbp1, Rgs3, PTP4a2, BMP2, IL6st, Fosb, Gabrr1, and possibly Rab3c. Genes implicated in protein metabolism and secretion pathways include Crym, Serpina5, Eif4a2, Ctrl, Snrpg, Kcnk8, Copz2, and Rab3c.
While the link between many of these genes and DIO will require further investigations, their identification here is an important contribution to understanding how the hepatic response to DIO and fasting/ weight reduction is mediated through a variety of molecular pathways. These genes all share a consistent set of attributes that made them stand out in the data set. They demonstrate significant differences between the dietary treatments, are individually discriminatory of each treatment, and are members of a set that classifies each sample using both supervised and unsupervised algorithms. Genes that satisfy all of these criteria represent good candidates for influencing the liver's response to DIO and fasting/ weight reduction, and therefore warrant more detailed investigations.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RMR helped with the tissue harvest, conducted the RNA purification, prepared the DNA microarrays, conducted the sample labeling and hybridization, conducted the RT-PCR, conducted the data analysis, and wrote the manuscript. JB handled the animals used in the study, helped conduct the tissue and RNA harvests, and helped write the manuscript. JK consulted on the project and reviewed the manuscript. CM helped design the experiment and set-up the animal handling procedures. GS organized the project, helped design the experiments, and helped write the manuscript.
Note
Table 3: (See Additional file: 1) Percent control expression of genes found common to all analysis methods. Included are genes identified using t-test, Wilks ranking, fisher discriminate analysis, and principle component analysis. These genes are organized by their pairwise t-test results, and the relation between their expression levels. F/ WR: Fasting/ Weight Reduced.
Supplementary Material
Additional file 1
Table 3
Click here for file
Acknowledgements
We thank Dr. Depi Sanoudo at the Harvard Medical School for her insightful comments and review of the manuscript. This work was supported by a grant from the National Institutes of Health Bioengineering research partnership, DK-585331. The Virtek arrayer was generously provided by Virtek, and supported by Bio-Rad.
==== Refs
Keller K Lemberg L Obesity and the Metabolic Syndrome American Journal of Critical Care 2003 12 167 170 12625176
Goran MI Ball GD Cruz ML Obesity and Risk of Type 2 Diabetes and Cardiovascular Disease in Children and Adolescents Journal of Clinical Endocrinology and Metabolism 2003 88 1417 1427 12679416 10.1210/jc.2002-021442
youn Kim S il Kim H Park SK Im SS Li T Cheon HG ho Ahn Y Liver Glucokinase can be Activated by Peroxisome Proliferator-Activated Receptor-gamma Diabetes 2004 53 S66 S70 14749268
Kerstein S Desvergne B Wahli W Roles of PPARs in Health and Disease Nature 2000 405 421 424 10839530 10.1038/35013000
Odom DT Zizlsperger N Gordon DB Bell GW Rinaldi NJ Murray HL Volkert TL Schreiber J Rolfe PA Gifford DK Fraenkel E Bell GI Young RA Control of Pancreas and Liver Gene Expression by HNF Transcription Factors Nature 2004 303 1378 1381 10.1126/science.1089769
Christ B Yazici E Nath A Phosphatidylinositol 3-kinase and Protein Kinase C Contribute to the Inhibition by Interleukin 6 of Phosphoenolpyruvate Carboxykinase Gene Expression in Cultured Hepatocytes Hepatology 2000 31 461 468 10655271 10.1002/hep.510310228
Surwit R Kuhn C Cochrane C McCubbin J Feinglos M Diet-induced Type II Diabetes in C57BL/6J Mice Diabetes 1988 37 1163 1167 3044882
Parekh P Petro A Tiller J Feinglos M Surwit R Reversal of Diet-induced Obesity and Diabetes in C57BL/6J Mice Metabolism 1998 47 1089 1096 9751238 10.1016/S0026-0495(98)90283-9
Frederich R Hamann A Anderson S Lollmann B Flier J Leptin Levels Reflect Body Lipid Content in Mice: Evidence for Diet-Induced Resistance to Leptin Action Nature Medicine 1995 1 1311 1314 7489415 10.1038/nm1295-1311
El-Haschimi K Pierroz D Hileman S Bjorbaek C Flier J Two Defects Contribute to Hypothalamic Leptin Resistance in Mice with Diet-Induced Obesity Journal of Clinical Investigation 2000 105 1827 1832 10862798
Lin S Thomas T Storlien L Huang X Development of High Fat Diet-Induced Obesity and Leptin Resistance in C57Bl/6J Mice International Journal of Obesity and Related Metabolic Disorders 2000 24 639 646 10849588 10.1038/sj.ijo.0801209
Belfiore F Ianneloo S Rabuazzo A Campione R Metabolic Effects of Short-Term Fasting in Obese Hyperglycaemic Humans and Mice International Journal of Obesity 1987 11 631 640 2831163
O'Brien RM Granner DK Regulation of Gene Expression by Insulin Physiological Reviews 1996 76 1109 1161 8874496
Jr AWF Thearle M Liao T Leibel RL Effects of Leptin Deficiency and Short-term Repletion on Hepatic Gene Expression in Genetically Obese Mice Diabetes 2001 50 2268 2278 11574408
Tobe K Suzuki R Aoyama M Terauchi Y Increased Expression of SREBP-1 Gene in IRS-2(-/-) Mice Liver Diabetes 2001 50 A324 A324
Ziotopoulou M Mantzoros C Hileman S Flier J Differential Expression of Hypothalamic Neuropeptides in the Early Phase of Diet-Induced Obesity in Mice American Journal of Physiology: Endocrinology and Metabolism 2000 279 E838 E845 11001766
Bullen JW JrZiotopoulou M Ungsunan L Misra J Alevizos I Kokkotou E Maratos-Flier E Stephanopoulos G Mantzoros CS Short-term Resistance to Diet-Induced Obesity in A/ J Mice is Not Associated with Regulation of Hypothalamic Neuropeptides American Journal of Physiology: Endocrinology and Metabolism 2004
Jelinsky S Samson L Global Response of Saccharomyces cerevisiae to an Alkylating Agent Proceedings of the National Academy of Sciences – USA 1999 96 1486 1491 10.1073/pnas.96.4.1486
Wang Y Rea T Bian J Gray S Sun Y Identification of the Genes Responsive to Etoposide-induced Apoptosis: Application of DNA Chip Technology FEBS Letters 1999 445 269 273 10094470 10.1016/S0014-5793(99)00136-2
Ruan H Hacohen N Golub TR Parijs LV Lodish HF Tumor Necrosis Factor-α Suppresses Adipocyte-Specific Genes and Activates Expression of Preadipocyte Genes in 3T3-L1 Adipocytes Diabetes 2002 51 1319 1336 11978627
Nadon R Shoemaker J Statistical Issues with Microarrays: Processing and Analysis TRENDS in Genetics 2002 18 265 271 12047952 10.1016/S0168-9525(02)02665-3
Quackenbush J Computational Analysis of Microarray Data Nature Reviews Genetics 2001 2 418 427 11389458 10.1038/35076576
Kuruvilla FG Park PJ Schreiber SL Vector Algebra in the Analysis of Genome-wide Expression Data Genome Biology 2002 3 0011.1 0011.11 10.1186/gb-2002-3-3-research0011
Thomas JG Olson JM Tapscott SJ Zhao LP An Efficient and Robust Statistical Modeling Approach to Discover Differentially Expressed Genes using Genomic Expression Profiles Genome Research 2001 11 1227 1236 11435405 10.1101/gr.165101
Hwang D Schmitt WA Stephanopoulos G Stephanopoulos G Determination of Minimum Sample Size and Discriminatory Expression Patterns Bioinformatics 2002 18 1184 1193 12217910 10.1093/bioinformatics/18.9.1184
Stephanopoulos G Hwang D Schmitt WA Misra J Stephanopoulos G Mapping Physiological States from Microarray Expression Measurements Bioinformatics 2002 18 1054 1063 12176828 10.1093/bioinformatics/18.8.1054
Dillon WR Goldstein M Multivariate Analysis: Methods and Applications 1984 New York: John Wiley & Sons
Bioinformatics and Metabolic Engineering Laboratory at MIT
Edgar R Domrachev M Lash AE Gene Expression Omnibus: NCBI Gene Expression and Hybridization Array Data Repository Nucleic Acids Research 2002 30 207 210 11752295 10.1093/nar/30.1.207
Escalante-Pulido M Escalante-Herrera A Milke-Najar M Alpizar-Salazar M Effects of Weight Loss on Insulin Secretion and in vivo Insulin Sensitivity in Obese Diabetic and Non-diabetic Subjects Diabetes Nutrition and Metabolism 2003 16 277 283
Heilbronn LK Ravussin E Calorie Restriction and Aging: Review of the literature and implications for studies in humans American Journal of Clinical Nutrition 2003 78 361 369 12936916
Ahima RS Prabakaran D Mantzoros C Qu D Lowell B Maratos-Flier E Flier JS Role of Leptin in the Neuroendocrine Response to Fasting Nature 1996 383 250 252 8805699 10.1038/382250a0
Koch W Hawes B Allen L Lefkowitz R Direct Evidence that Gi-Coupled Receptor Stimulation of Mitogen-Activated Protein Kinase is Mediated by Gβγ Activation of p21ras Proceedings of the National Academy of Sciences: U S A 1994 91 12706 12710
Kullander K Klein R Mechanisms and Functions of EPH and EPHRIN Signalling Nature Reviews Molecular Cell Biology 2002 3 475 486 12094214 10.1038/nrm856
Verdier F Valovka T Zhyvoloup A Drobot LB Buchman V Waterfiedl M Gout I Ruk is Ubiquitinated but not Degraded by the Proteasome European Journal of Biochemistry 2002 269 3402 3408 12135478 10.1046/j.1432-1033.2002.03031.x
White MF Jr MGM Diabetes Mellitus, Carbohydrate Metabolism, and Lipid Disorders 2000 Canada: Harcourt Canada
Gout I Middleton G Adu J Ninkina NN Drobot LB Filonenko V Matsuka G Davies AM Waterfield M Buchman VL Negative Regulation of PI 3-kinase by Ruk, a Novel Adaptor Protein EMBO Journal 2000 19 4015 4025 10921882 10.1093/emboj/19.15.4015
Cheng A Dube N Gu F Tremblay ML Coordinated Action of Protein Tyrosine Phosphatases in Insulin Signal Transduction European Journal of Biochemistry 2002 269 1050 1059 11856336 10.1046/j.0014-2956.2002.02756.x
Asante-Appiah E Kennedy BP Protein Tyrosine Phosphatases: The quest for negative regulators of insulin action American Journal of Physiology, Endocrinology and Metabolism 2003 284 E663 E670 12626322
Elchebly M Payette P Michaliszyn E Cromlish W Collins S Loy A Normandin D Cheng A Himms-Hang J Chan C Ramachandran C Gresser M Tremblay M Kennedy B Increased Insulin Sensitivity and Obesity Resistance in Mice Lacking the Protein Tyrosine Phosphatase-1B Gene Science 1999 283 1544 1548 10066179 10.1126/science.283.5407.1544
Arshavsky V Jr EP Lifetime Regulation of G Protein-Effector Complex: Immerging Importance of RGS Proteins Neuron 1998 20 11 14 9459437 10.1016/S0896-6273(00)80430-4
Jeong S Ikeda S Endogenous Regulator of G-Protein Signalling Proteins Modify N-type Calcium Channel Modulation in Rat Sympathetic Neurons Journal of Neuroscience 2000 20 4489 4496 10844018
Sinnarajah S Dessauer CW Srikumar D Chen J Yuen J Yilma S Dennis JC Morrison EE Vodyanoy V Kehrl JH RGS2 Regulates Signal Transduction in Olfactory Neurons by Attenuating Activation of Adenylyl Cyclase III Nature 2001 409 1051 1055 11234015 10.1038/35059104
Chan R Otte C Isolation and Genetic Analysis of Saccharomyces cerevisiae Mutants Supersensitive to G1 Arrest by a Factor and α-factor Pheromones Molecular Cellular Biology 1982 2 11 20
Vries LD Zheng B Fischer T Elenko E Farquhar M The Regulator of G Protein Signalling Family Annual Reviews in Pharmacology and Toxicology 2000 40 235 271 10.1146/annurev.pharmtox.40.1.235
Siderovski D Strockbine B Behe C Whither goest the RGS proteins? Critical Reviews in Biochemistry and Molecular Biology 1999 34 215 251 10517644 10.1080/10409239991209273
Chatterjee TK Eapen AK Fisher RA A Truncated Form of RGS3 Negatively Regulates G Protein-coupled Receptor Stimulation of Adenylyl Cyclase and Phosphoinositide Phospholipase C The Journal of Biological Chemistry 1997 272 15481 15487 9182581 10.1074/jbc.272.24.15481
Neubig RR Siderovski DP Regulators of G-Protein Signalling as New Central Nervous System Drug Targets Nature Reviews Drug Discovery 2002 1 187 197 12120503 10.1038/nrd747
Berne RM Levy MN Physiology 1998 fourth St. Louis, MO: Mosby
Chen C Grzegorzewski KJ Barash S Zhao Q Schneider H Wang Q Singh M Pukac L Bell AC Duan R Coleman T Duttaroy A Cheng S Hirsch J Zhang L Lazard Y Fischer C Barber MC Ma ZD Zhang YQ Reavey P Zhong L Teng B Sanyal I Ruben SM Blondel O Birse CE An Integrated Functional Genomics Screening Program Reveals a Role for BMP-9 in Glucose Homeostasis Nature Biotechnology 2003 21 294 301 12598908 10.1038/nbt795
Angel P Imagawa M Chiu R Stein B Imbra R Rahmsdorf H Jonat C Herrlich P Karin M Phorbol Ester-inducible Genes Contain a Common cis Element Recognized by a TPA-modulated trans-acting Factor Cell 1987 49 729 739 3034432 10.1016/0092-8674(87)90611-8
Segal R Greenberg M Intracellular Signalling Pathways Activated by Neurotrophic Factors Annual Reveiws in Neuroscience 1996 19 463 489
Nakabeppu Y Nathans D A Naturally Occurring Truncated for of FosB that Inhibits Fos/Jun Transcriptional Activity Cell 1991 64 751 759 1900040 10.1016/0092-8674(91)90504-R
O'Brien RM Granner DK Regulation of Gene Expression by Insulin Biochemistry Journal 1991 278 609 619
Heinrich PC Behrmann I Muller-Newen G Schaper F Graeve L Interleukin-6-type Cytokine Signalling through the gp130/Jak/STAT Pathway Biochemistry Journal 1998 334 297 314
Wallenius V Wallenius K Ahren B Rudling M Carlsten H Dickson S Ohlsson C Jansson J Interleukin-6-deficient Mice Develop Mature-Onset Obesity Nature Medicine 2002 8 75 79 11786910 10.1038/nm0102-75
Inoue H Ogawa W Ozaki M Haga S Matsumoto M Furukawa K Hashimoto N Kido Y Mori T Sakaue H Teshigawara K Jin S Iguchi H Hiramatsu R LeRoith D Takeda K Akira S Kasuga M Role of STAT-3 in Regulation of Hepatic Gluconeogenic Genes and Carbohydrate Metabolism in vivo Nature Medicine 2004 10 168 174 14716305 10.1038/nm980
Zechmeister-Machhart M Hufnagl P Uhrin P Korschineck I Binder B Geiger M Molecular Cloning and Sequence Analysis of the Mouse Protein C Inhibitor Gene Gene 1997 186 61 66 9047345 10.1016/S0378-1119(96)00681-6
Uphues I Chern Y Eckel J Insulin-dependent Translocation of the Small GTP-binding Protein RAB3C in Cardiac Muscle: Studies on Insulin-Resistant Zucker Rats FEBS Letters 1995 377 109 112 8543030 10.1016/0014-5793(95)01315-6
|
15985155
|
PMC1201568
|
CC BY
|
2021-01-04 16:37:46
|
no
|
Nutr Metab (Lond). 2005 Jun 28; 2:15
|
utf-8
|
Nutr Metab (Lond)
| 2,005 |
10.1186/1743-7075-2-15
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1614641510.1371/journal.pmed.0020263Research ArticleGenetics/Genomics/Gene TherapyNeuroscienceMental HealthPathologyPsychiatryGeneticsPathologyPsychiatrySchizophrenia and Other Psychotic DisordersIncreased Expression in Dorsolateral Prefrontal Cortex of CAPON in Schizophrenia and Bipolar Disorder CAPON Expression in Psychotic IllnessXu Bin
1
Wratten Naomi
1
Charych Erik I
2
Buyske Steven
1
3
Firestein Bonnie L
2
Brzustowicz Linda M
1
4
*1Department of Genetics, Rutgers University, Piscataway, New Jersey, United States of America,2Department of Cell Biology and Neuroscience, Rutgers University, Piscataway, New Jersey, United States of America,3Department of Statistics, Rutgers University, Piscataway, New Jersey, United States of America,4Department of Psychiatry, University of Medicine and Dentistry of New Jersey, New Jersey Medical School, Newark, New Jersey, United States of AmericaMcGuffin Peter Academic EditorKings College LondonUnited Kingdom*To whom correspondence should be addressed. E-mail: [email protected]
Competing Interests: The authors have declared that no competing interests exist.
Author Contributions: BX, NW, EIC, BLF, and LMB designed the study. BX, NW, and EIC performed experiments. BX, SB, and LMB analyzed the data. BX, NW, EIC, SB, BLF, and LMB contributed to writing the paper.
10 2005 13 9 2005 2 10 e2634 1 2005 29 6 2005 Copyright: © 2005 Xu 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.
CAPON and Schizophrenia--Does Size Matter?
Does the CAPON Gene Confer Susceptibility for Schizophrenia?
Background
We have previously reported linkage of markers on chromosome 1q22 to schizophrenia, a finding supported by several independent studies. Within this linkage region, we have identified significant linkage disequilibrium between schizophrenia and markers within the gene for carboxyl-terminal PDZ ligand of neuronal nitric oxide synthase (CAPON). Prior sequencing of the ten exons of CAPON failed to reveal a coding mutation associated with illness.
Methods and Findings
We screened a human fetal brain cDNA library and identified a new isoform of CAPON that consists of the terminal two exons of the gene, and verified the expression of the predicted corresponding protein in human dorsolateral prefrontal cortex (DLPFC). We examined the expression levels of both the ten-exon CAPON transcript and this new isoform in postmortem brain samples from the Stanley Array Collection. Quantitative real-time PCR analysis of RNA from the DLPFC in 105 individuals (35 with schizophrenia, 35 with bipolar disorder, and 35 psychiatrically normal controls) revealed significantly (p < 0.005) increased expression of the new isoform in both schizophrenia and bipolar disorder. Furthermore, this increased expression was significantly associated (p < 0.05) with genotype at three single-nucleotide polymorphisms previously identified as being in linkage disequilibrium with schizophrenia.
Conclusion
Based on the known interactions between CAPON, neuronal nitric oxide synthase (nNOS), and proteins associated with the N-methyl-D-aspartate receptor (NMDAR) complex, overexpression of either CAPON isoform would be expected to disrupt the association between nNOS and the NMDAR, leading to changes consistent with the NMDAR hypofunctioning hypothesis of schizophrenia. This study adds support to a role of CAPON in schizophrenia, produces new evidence implicating this gene in the etiology of bipolar disorder, and suggests a possible mechanism of action of CAPON in psychiatric illness.
Comparative analysis of CAPON expression in postmortem brain samples from patients with schizophrenia, bipolar disorder, and from healthy controls.
==== Body
Introduction
Schizophrenia (SCZD) is a serious neuropsychiatric illness estimated to affect approximately 1% of the general population. Family, twin, and adoption studies have demonstrated that schizophrenia is predominantly a genetic disorder, with a high heritability (reviewed in [1]). Multiple genetic and nongenetic factors are likely to be involved [2]. As part of a genome-wide search for loci contributing to risk for schizophrenia, we previously reported linkage, with a maximum heterogeneity lod score of 6.5, to chromosome 1q21-1q22 (SCZD9) in a group of 22 medium-sized Canadian families that were selected for study because multiple relatives were clinically diagnosed with schizophrenia or schizoaffective disorder [3]. We have also reported the results of fine linkage mapping of this 1q21-1q22 region in the same sample of individuals, narrowing the region most likely to harbor this susceptibility locus to approximately 1 Mb between APOA2 and D1S2675, again with a maximum multipoint heterogeneity lod score of 6.5 [4]. Other studies have also reported linkage [5–8] and linkage disequilibrium (LD) [9] of schizophrenia to this region. More recently, we have tested markers from this region for evidence of LD to schizophrenia, identifying significant LD with several markers within the gene for carboxyl-terminal PDZ ligand of neuronal nitric oxide synthase (CAPON; also termed nitric oxide synthase 1 [neuronal] adaptor protein [NOS1AP]) [10]. Association of single-nucleotide polymorphisms (SNPs) within CAPON to schizophrenia has recently been replicated in the Chinese Han population [11], although with association detected in the Chinese sample to SNPs located more distal in the gene than the SNPs associated in the Canadian sample.
CAPON is an attractive candidate for schizophrenia susceptibility. CAPON was first identified in the rat as a neuronal nitric oxide synthase (nNOS) binding protein, capable of disrupting the association of nNOS with the postsynaptic density scaffolding proteins PSD93 and PSD95 through the binding of the carboxyl terminus of CAPON to nNOS [12]. The interaction of nNOS with PSD93 and PSD95 is important in targeting nNOS to the postsynaptic N-methyl-D-aspartate receptor (NMDAR) complex and facilitates the tight coupling between activation of the NMDAR and nNOS, allowing nNOS activation by Ca2+ influx through the NMDAR and producing NMDAR-mediated NO release into the synaptic structures [13,14]. This places CAPON at the scene of NMDAR glutamate neurotransmission, long proposed to be involved in schizophrenia (reviewed in [15]). CAPON can also serve as an nNOS adaptor protein, with the amino terminus binding either to a direct target of NO-mediated activation by S-nitrosylation [16] or to Synapsin [17], resulting in the localization of nNOS to the presynaptic terminals.
Sequencing of the coding region of CAPON in individuals from the Canadian linkage sample failed to identify any coding mutations associated with illness [10], consistent with current results for other candidate genes for schizophrenia [18]. CAPON has a large, approximately 300-kb genomic extent, only 1.5 kb of which represents coding sequence. Therefore, there are many potential sites for regulatory sequences that could be disrupted and lead to altered gene expression. In this study, we screened a human cDNA library to identify possible alternative splice forms of CAPON, documented CAPON mRNA and protein expression in postmortem tissue from the dorsolateral prefrontal cortex (DLPFC) of human brains, and investigated the expression of CAPON by quantitative real-time PCR in the Stanley Array Collection, derived from DLPFCs of individuals with schizophrenia and with bipolar disorder, and a set who were psychiatrically normal controls.
Methods
Identification and Characterization of CAPON Transcripts
A human fetal brain arrayed cDNA library (Origene Technologies, Rockville, Maryland, United States; #LLFB-1001) was screened using the supplier's protocol. PCR amplification was performed with primers from CAPON exon 10 (Ex10-F, 5′-
AAATCAACAACCTTGCCTAACG-3′; Ex10-R, 5′-
GAAAGCACTCCAGCTTCACC-3′). Exon 10 sequence was chosen for screening purposes to identify alternative transcripts that also contained the nNOS-binding PDZ domain. Individual positive clones were sequenced on a CEQ 8000 (Beckman Coulter, Fullerton, California, United States). The transcript for the previously undescribed short isoform was further characterized by 3′ and 5′ RACE, performed with RACE-ready cDNA from human brain (Ambion, Austin, Texas, United States). For each reaction, a pair of nested PCR primers were designed from short-form 5′ UTR sequences (5RACES, 5′-
TTAGAGGTTCCTGGAGGGTGGTGC-3′; 5RACESn, 5′-
TTGAGTCCAAGGAGAGGGTAGTGG-3′) and 3′ UTR sequence (3RACE1, 5′-
AATGAATGCAAGCTGATAGCTGAGACTG-3′; 3RACE1n, 5′-
TGAATCACTGCCACTTGGGTCAGG-3′; 3RACE2, 5′-
AGAAGGAAGAACCAGGAAAGTGAGATCC-3′; 3RACE2n, 5′-
ATCCAGTGTGGCTGAGCCTACCTAGC-3′; and 3RACE3, 5′-
ATGTGGATGGAGAGGGCTTGT-3′; 3RACE3n, 5′-
GTGAGGAAGGCCGCTTCTAAAT-3′) and were used in conjunction with a set of universal nested primers (supplied). Products were cloned into TOPO TA cloning vector (Invitrogen, Carlsbad, California, United States) and sequenced on a CEQ 8000.
The sequence for the novel CAPON transcript has been deposited in GenBank (http://www.ncbi.nlm.nih.gov/entrez/) under accession number AY841899.
Human Postmortem Samples
RNA and DNA samples from the Stanley Array Collection of the Stanley Brain Collection (http://www.stanleyresearch.org/programs/brain_collection.asp) were analyzed. This is a collection of biomaterials derived from postmortem brain specimens from 35 individuals with schizophrenia, 35 individuals with bipolar disorder, and 35 psychiatrically normal controls. Diagnoses were made by two senior psychiatrists, applying DSM-IV criteria to medical records, and, when necessary, telephone interviews with family members. Diagnoses of unaffected controls were based on structured interviews by a senior psychiatrist with family member(s) to rule out DSM-IV Axis I diagnoses of psychiatric clinical syndromes.
Specimens were collected, with informed consent from next of kin, by participating medical examiners between January 1995 and June 2002. The specimens were collected, processed, and stored in a standardized way. Exclusion criteria for all specimens included: (1) significant structural brain pathology on postmortem examination by a qualified neuropathologist, or by premortem imaging; (2) history of significant focal neurological signs premortem; (3) history of central nervous system disease that could be expected to alter gene expression in a persistent way; (4) documented IQ under 70; or (5) poor RNA quality. RNA integrity and purity were determined with an Agilent 2100 Bioanalyzer (Agilent, Palo Alto, California, United States). Degradation was defined as a shift in the RNA size distribution toward smaller fragments and a decrease in fluorescence signal of ribosomal peaks. Additional exclusion criteria for unaffected controls included age less than 30 (thus, still in the period of maximum risk) and substance abuse within 1 y of death or evidence of significant alcohol-related changes in the liver.
DNA and RNA from Brodmann's area 46 (in the DLPFC) was available for all individuals in the collection. Genotyping and expression analyses were conducted with the samples coded to keep investigators blind to diagnostic status. After the blind was broken, diagnostic status and a range of clinical variables were provided for analysis. These included gender, race, age at time of death, age of onset, postmortem interval (PMI), brain pH, total brain weight, hemisphere used for RNA extraction, smoking status at time of death (coded as nonsmoking for individuals who smoked previously but had quit), antipsychotic status at time of death, and lifetime antipsychotic exposure in fluphenazine milligram equivalents. In addition, lifetime alcohol and substance use were each rated on a scale of 0 to 5 using the categories “little or none,” “social,” “moderate past,” “moderate present,” “heavy past,” and “heavy present.” Total sample storage time was calculated from date of death to date of experiment, so it represents the sum of storage time of the brain tissue prior to RNA extraction and storage time of the extracted RNA. Overall, the sample was 66% male, with a mean age at death of 44 y (standard deviation [SD] 8.9 y, range 19–64 y), and was predominantly Caucasian (97%). The mean PMI was 33 h (SD 16 h, range 9–84 h), and the mean storage time was 5.0 y (SD 1.9 y, range 2.1–9.7 y). Smoking status at time of death was available for 67 individuals, with 72% of the sample smokers. Lifetime alcohol use estimates were available on all but one patient, with 57% of the sample reporting no, little, or social use, 17% reporting past or present moderate use, and 26% reporting past or present heavy use. Lifetime substance use estimates were available on all but two individuals, with 65% of the sample reporting no, little, or social use, 16% reporting past or present moderate use, and 19% reporting past or present heavy use. The mean age of onset for the schizophrenia group was 21.3 y (SD 6.1 y, range 9–34 y) and for the bipolar group was 25.1 y (SD 9.1 y, range 14–48 y). Further information about the Stanley Array Collection is available from the Stanley Medical Research Institute (http://www.stanleyresearch.org/programs/brain_collection.asp).
Studies of CAPON protein were conducted using brain specimens from five normal control individuals (three male and two female) obtained from the Human Brain and Spinal Fluid Resource Center (Los Angeles, California, United States). Individuals ranged in age from 53 to 90 years at time of death (mean 67.4 y, SD 16.4 y). Samples were recovered and frozen after a mean PMI of 21.5 h (SD 3.0 h, range 17.8–26.0 h). Specimens were shipped frozen for protein and RNA extraction in our laboratory. Mean storage time prior to extraction for these specimens was 3.0 y (SD 1.4 y, range 1.9–5.3 y); protein and RNA was quantified within 1 wk of extraction. To assure a similar composition of each sample for protein and RNA extractions, samples were first ground while still frozen into a fine powder using liquid nitrogen and a mortar and pestle. Aliquots of this homogeneous powder were then used for protein and RNA extraction as below.
Preparation and Western Blotting of Human Brain Homogenates
Tissue from Brodmann's area 46 (DLPFC) from the Human Brain and Spinal Fluid Resource Center samples was homogenized with 10 times the equivalent volume per weight of TEE (25 mM Tris-HCl [pH 7.4], 1 mM EDTA, 1 mM EGTA, and 1 mM PMSF) using a serrated Teflon pestle. Ten micrograms of protein were loaded and resolved on a 12% SDS polyacrylamide gel and transferred to polyvinylidene difluoride membrane (Immobilon P; Millipore, Billerica, Massachusetts, United States) in transfer buffer lacking SDS, which was determined to interfere with detection of the short isoform of CAPON (CAPON-S). Blots were probed with a rabbit antibody to CAPON (Santa Cruz Biotechnology, Santa Cruz, California, United States; #R-300) and visualized using ECL plus (Amersham Biosciences, Piscataway, New Jersey, United States) with a secondary antibody coupled to horseradish peroxidase. For actin normalization, blots were incubated in strip buffer (62.5 mM Tris-HCl [pH 7.4], 2% SDS, and 90 μM β-mercaptoethanol) at 65 °C for 20 min with agitation, followed by probing with a rabbit antibody to actin (the antigen corresponds to amino acids 20–33, conserved in all actin isoforms; Sigma, Saint Louis, Missouri, United States) using the procedure described above. To ensure that quantification used an exposure that was within the linear range of the film used to image the blot, a series of images with increasing exposure time to the blot were developed. The intensity of the bands on each image was quantified using Image Pro image analysis software (Media Cybernetics, Silver Spring, Maryland, United States) and plotted to determined at what exposure time the darkest band intensities were no longer in the linear range. Nonsaturated exposures were then used to determine the ratio of CAPON to actin. Using ImagePro image analysis software, the relative immunoreactivities of anti-CAPON bands were calculated as (a − b)/(c − d), where a = sum of pixels in each CAPON immunoreactive band; b = sum of pixels in an area equivalent to that occupied by a within a nonreactive portion of the lane; c = sum of pixels in each actin immunoreactive band; and d = sum of pixels in an area equivalent to that occupied by c within a nonreactive portion of the lane.
COS-7 Cell Culture and Transfection
COS-7 cells were plated at 70%–80% confluence and maintained in Dulbecco's Modified Eagle Medium (Invitrogen), supplemented with 7.5% fetal bovine serum and 7.5% horse serum in a 7% CO2 atmosphere. Cells were transfected with 0.5 μg of plasmid from the Origene library containing the CAPON short-form RNA, cloned behind the CMV promoter in pCMV6-XL4, using the LipofectAMINE 2000 method (Invitrogen) following the manufacturer's instructions. After allowing 48 h for expression, cells were lysed in TEE, and lysates were analyzed by Western blotting as described above.
RNA Extraction
Total RNA was extracted from the Human Brain and Spinal Fluid Resource Center samples using RNeasy Lipid Tissue Mini Kits (Qiagen, Valencia, California, United States), following the manufacturer's instructions. RNA was quantified with a Pharmgena UV spectrometer. Quality of the samples was verified by fractionation on a 2% agarose gel at 100 V for 40 min, staining with ethidium bromide, and visualization under UV light.
RNA Quantification
Total RNA (5 μg) was treated with the DNA-free Kit (Ambion) in accordance with the manufacturer's protocol to eliminate DNA contamination. The resulting DNase-treated RNA was used in a 40-μl reverse transcriptase reaction to synthesize cDNA following the SuperScript II First-Strand cDNA Synthesis protocol (Invitrogen), including optional RNaseOUT treatment. Samples were then quantified using real-time PCR. The previously described primer pair NN05224/NN05225 (Human Unidentified Gene-Encoded database [http://www.kazusa.or.jp/huge/index.html] clone KIAA0464) was used for quantification of full-length CAPON [19,20]. This primer pair produces a 338-bp product that spans the boundary of exons 7 and 8. The primers specific for the short form were designed using Primer Express Software Version 2.0 (Applied Biosystems, Foster City, California, United States). The forward primer (Short-F: 5′-
CATTCATGTCCCTCTCTTCTCTC-3′) is located in the 5′ coding sequence that is unique to the short form transcript, the reverse primer (Short-R, 5′-
AATGCAGGTCCTCTGGCTTAG-3′) is located within exon 10, and the pair produces a 321-bp product that spans the boundary of exons 9 and 10. The housekeeping gene ACTB, the gene encoding beta-actin, was used as reference gene to normalize the total RNA input. The forward (b-actin-F, 5′-
CATCCTCACCCTGAAGTACCC-3′) and reverse (b-actin-R, 5′-
GAGAAGATGACCCAGATCATGTTT-3′) primers produce a 184-bp product that spans the boundary of exons 3 and 4. Real-time PCR analysis was conducted using 1 μl of a 1:5 dilution of cDNA, 0.1 μM of each primer, and 5 μl of SYBR Green Master Mix (Applied Biosystems) in a total reaction volume of 10 μl in a 384-well plate on an ABI Prism 7900HT sequence detector system (Applied Biosystems). Samples were initially warmed to 50 °C for 2 min, then the AmpliTaq Gold DNA polymerase was activated by heating to 95 °C for 10 min. PCR amplification was performed with 40–50 cycles of 95 °C for 30 s, 58 °C (full-length experiments) or 61 °C (short-form experiments) for 40 s, and 72 °C for 1 min. Each real-time PCR assay was repeated three times. The standard curve used for determining the relative quantity of each isoform in each sample was constructed by the amplification of serial dilutions of pooled brain cDNA. In each experiment, the R
2 value of the standard curve was greater than 0.98, and the no-template control produced no detectable signal. Dissociation curve analysis was conducted on all PCR products to ensure that only a single product was present in the reaction. Real-time PCR data acquisition and analysis were performed using SDS v2.0 software (Applied Biosystems).
SNP Genotyping
DNA samples from the Stanley Array Collection were genotyped for three SNPs from dbSNP, rs1415263, rs4145621, and rs2661818, that were previously identified as being in linkage disequilibrium with schizophrenia by a primer extension strategy (Pyrosequencing, Uppsala, Sweden) using the automated PSQ HS96A platform as previously described [10].
Statistical Analyses
RNA amounts were quantified by the ABI Relative Quantitation of Gene Expression protocol (Applied Biosystems; http://docs.appliedbiosystems.com/pebiodocs/04303859.pdf). The results from three repeat assays were averaged to produce a single mean quantity value for each mRNA for each individual. The quantity values of the target gene were then normalized over the quantity values of the reference gene ACTB (encoding beta-actin) to produce normalized expression quantities. These are unitless measures of the relative amount of transcript that is present in each individual.
Normalized expression quantities and patient variables were analyzed with SAS v8.2 for UNIX software (SAS Institute, Cary, North Carolina, United States) and with the R statistical environment [21]. Correlation between protein and mRNA levels in control brain samples from the Human Brain and Spinal Fluid Resource Center was tested with Pearson's product moment correlation. For the samples from the Stanley Array Collection, potentially confounding variables (Table 1) were tested for association with CAPON expression in all individuals. As CAPON expression levels correlated with time of storage, subsequent analyses included storage time as a covariate. Effect of exposure to antipsychotics (either as a dichotomous trait or a quantitative estimate of total lifetime exposure) was examined only in patients with bipolar disorder or schizophrenia. Continuous variables (age at death, postmortem interval, brain pH, and brain weight) were tested by regression; dichotomous variables (gender, brain hemisphere analyzed, smoking status at time of death, and history of exposure to antipsychotic medication) and ordinal variables (lifetime alcohol use and lifetime substance abuse) were tested by ANOVA. The strength of the association with CAPON expression was measured with Pearson's product moment correlation for continuous variables (except for quantitative lifetime antipsychotic exposure, for which Spearman's correlation was used) and Kendall's rank correlation for binary and ordinal variables. The relationship between gene expression levels and diagnostic group was summarized with Kendall's rank correlation and tested with ANOVA, and the relationship between gene expression levels and age of onset within the bipolar and schizophrenia groups was summarized with Pearson's product moment correlation and tested with regression. The relationship between genotype and CAPON expression was analyzed using ANOVA with a term for years of storage (as in all of the ANOVAs), a term coded for a dominance effect of the SNP, and a term for a recessive effect of the SNP.
Table 1 Correlations between CAPON Isoform Expression and Possible Confounding Variables
Results
Identification and Characterization of Two Human CAPON Isoforms
Screening of a human fetal brain cDNA library resulted in the isolation of four distinct clones. Three clones (“CAPON full-length”) contained inserts of approximately 2.5 kb, corresponding to exons 1–10, as previously described in the NCBI reference sequence for CAPON [12,22], except that exon 4 from these clones was missing the first 15 bp listed in the reference sequence. The fourth clone (“CAPON short-form”) contained an insert of 4 kb with a unique 5′ untranslated region (UTR) corresponding to genomic sequence from intron 8, followed by exons 9 and 10, and a 3′ UTR longer than that contained by the other clones or in the reference sequence. This transcript is predicted to produce a 211 amino acid protein, including 18 novel amino acids at the amino terminus, and will contain the CAPON PDZ-binding domain.
The previously undescribed short-form transcript of CAPON was further characterized by 5′ and 3′ RACE. The 5′ UTR was 2,367 bp, while the longest 3′ RACE product ended 2,556 bp downstream of the stop codon. From these results, the total length of the short-form mRNA was calculated to be 5,559 bp. Due to the much longer 5′ and 3′ UTRs, the CAPON transcript encoding the short-form of the protein is actually significantly longer than the transcript that encodes the full-length protein. The gene, transcripts, and predicted protein structures of the two forms are shown in Figure 1.
Figure 1 CAPON Gene Structure, Isoforms, and Protein Functional Domains
CAPON full-length transcript and protein illustrations are based on NCBI reference sequences.
(A) Genomic organization of CAPON. Exons are represented by numbered boxes; 5′ and 3′ UTR are represented by half-height boxes.
(B) CAPON transcripts. 5′ and 3′ UTR are represented by yellow shaded boxes, exons by white boxes.
(C) CAPON proteins. The phosphotyrosine-binding domain (PTB) is shaded green, and the PDZ-binding domain is shaded blue.
Protein and total RNA were extracted from Brodmann's area 46 (DLPFC) of postmortem samples from five normal controls, COS-7 cell that had been transfected with cDNA encoding the CAPON short-form, and untransfected COS-7 cells that normally express the CAPON long form. Western blots of these samples were probed with a rabbit polyclonal antibody raised against the carboxyl terminus of CAPON. This antibody is predicted to interact with both the full-length and short-form of CAPON, since the antigen used to generate the antibody is in the carboxyl termini of both forms. Bands were observed at the expected sizes, near the 75-kDa marker for the full-length protein and near the 30-kDa marker for the short-form (Figure 2). There appears to be two smaller forms of the full-length CAPON (CAPON-L bracket, Figure 2), which could be due to posttranslational modification (i.e., phosphorylation) of CAPON. For example, analysis by software designed to identify potential sites for phosphorylation by protein kinase C (http://mpr.nci.nih.gov/mpr/ScanProteinForPKCSitesPage.aspx) identified five such sites in the full-length CAPON sequence. The short form of CAPON appears as a doublet, although the presence of the lower band (CAPON-S′) in untransfected COS-7 cells may indicate that this band is caused by recognition by the CAPON antibody of a cross-reacting protein. The appearance of the higher band (CAPON-S) in the transfected cells, which clearly comigrates with a band in human brain tissue, indicates that the short-form transcript is translated into a protein of the expected size in DLPFC.
Figure 2 Western Blot of CAPON Protein Isoforms in DLPFC from Normal Control Individuals
Tissue from Brodmann's area 46 (DLPFC) from five individuals was homogenized in TEE. Proteins were resolved by SDS-PAGE and transferred to PVDF membrane. Blots were probed with rabbit polyclonal antibodies to CAPON and actin, and proteins were detected using chemiluminescence. Band intensities for CAPON-L, CAPON-S, and CAPON-S + CAPON-S′ were calculated and normalized to the intensities of the corresponding actin bands. Untransfected COS-7 cells expressing CAPON-L and CAPON-S′ (COS-7/NT) and COS-7 cells expressing recombinant CAPON-S (COS-7/CAPON-S) and were used as controls for these proteins. CAPON-L appears to include multiple bands, possibly due to phosphorylation.
As protein samples were not available for the Stanley Array Collection samples, we tested the correlation between protein and RNA expression using the normal control brain samples to determine if RNA levels could be used as a reasonable indicator of protein expression for CAPON. CAPON protein levels were quantified from Western blot image analysis and were normalized to levels of actin protein, while RNA levels were quantified by reverse-transcription real-time PCR normalized to levels of ACTB (beta-actin). For full-length CAPON, the correlation between protein and RNA levels was significant (p = 0.019) with r = 0.94. For the short form, the correlation was also significant (p = 0.0049) with r = 0.97 between the RNA and protein levels of the S band, which corresponds to the size of the cloned CAPON product. While it seems likely that the S′ band represents a cross-reacting protein, given its presence in untransfected COS-7 cells, it is possible that it could represent a modified form of the short-form protein. Considering the CAPON short-form product as the sum of the S and S′ bands, the correlation with RNA levels remained significant (p = 0.030), with r = 0.91.
Analysis of CAPON Isoform Expression by Diagnosis
Expression levels of both CAPON isoforms were determined by reverse-transcription real-time PCR for all 105 samples from the Stanley Array Collection. Expression levels were normalized to ACTB (beta-actin), and these normalized relative expression levels were used for all subsequent analyses. No significant correlations were detected between mRNA levels of either CAPON isoform and the potentially confounding variables of age at death, PMI, brain pH, brain weight, gender, hemisphere, smoking status at time of death, lifetime alcohol use, or lifetime substance abuse (Table 1). CAPON expression levels were found to be significantly (p < 0.001) correlated with length of sample storage for both isoforms (Table 1). Therefore, for all subsequent analyses storage time was used as covariate.
The potential confounding effects of antipsychotic medication treatment on CAPON expression levels were also very important to examine, but since all of the patients with schizophrenia and none of the controls had been treated with such medications, the effects of treatment and diagnosis could not be separated by analyses that included these groups. Within the 35 individuals in the bipolar group, however, 18 individuals were on antipsychotic medication at the time of death, 11 individuals had never received antipsychotic medication, and six individuals were not on antipsychotic medication at time of death, but had been treated with these medications at some point in the past. CAPON levels were compared between antipsychotic-treated and untreated individuals with bipolar disorder. Neither a positive history of lifetime antipsychotic use nor antipsychotic use at time of death was significantly correlated with CAPON short-form expression within the bipolar group (Table 2). In contrast, expression of full-length CAPON was significantly correlated with treatment (Table 2), with a 40% decrease in mean expression in the patients (n = 24) with bipolar disorder and a history of treatment with antipsychotics in the past or at time of death (p = 0.003), and a 45% decrease in patients (n = 18) receiving antipsychotics at time of death (p = 0.0007), when compared to antipsychotic-untreated individuals (n = 11) with bipolar disorder. An estimate of total lifetime antipsychotic medication was available for all but one individual in the bipolar and schizophrenia groups with a positive history of antipsychotic treatment (n = 58). No significant correlations were found between levels of lifetime antipsychotic exposure and expression of either CAPON isoform (Table 2).
Table 2 Effect of Antipsychotic Treatment on CAPON Isoform Expression
Overall, there was no significant difference in CAPON full-length mRNA expression across diagnostic categories (Figure 3). Since treatment with antipsychotics may influence expression of the full-length isoform, we examined expression of this transcript in antipsychotic-naïve patients with bipolar disorder. While mean CAPON full-length mRNA levels were increased by 24% in patients with bipolar disorder but no history of exposure to antipsychotic medication (n = 11) as compared to normal controls, this increase did not reach statistical significance (p = 0.11). Results were similar when comparing bipolar patients not receiving antipsychotic medication at time of death, regardless of past treatment history, (n = 17) to normal controls, with a 18% increase in full-length CAPON levels (p = 0.14).
Figure 3 ACTB (Beta-Actin)–-Normalized CAPON mRNA Full-Length Expression by Diagnosis
Expression levels are least squares means. Mean values per category are plotted with 95% confidence intervals. The number of individuals per sample is indicated within each bar. Level of expression does not differ significantly by diagnostic group. The mean (95% confidence interval lower bound, upper bound) for the control, schizophrenia, and bipolar groups are 1.28 (1.12, 1.45), 1.33 (1.17, 1.49), and 1.16 (0.99, 1.32), respectively.
Mean CAPON short-form mRNA levels were significantly increased by 48% in the schizophrenia group (p = 0.0035) and 50% in the bipolar group (p = 0.0002) as compared to the control group (Figure 4). The schizophrenia and bipolar groups did not differ significantly from each other in CAPON short-form expression (p = 0.94). CAPON short-form expression was significantly correlated with the age of onset in the schizophrenia group (Pearson's r = 0.53, p = 0.0008), but not in the bipolar group (r = −0.02, p = 0.92). This significance (or lack thereof) is unchanged when age of death is included as a covariate. The majority of samples were from individuals of European decent (97%), with one African American individual with bipolar disorder, one Native American individual with bipolar disorder, and one Hispanic individual with schizophrenia. None of these individuals exhibited extreme values for expression of either CAPON isoform, and re-analysis with these patients excluded did not change which comparisons reached statistical significance (unpublished data).
Figure 4 ACTB (Beta-Actin)–Normalized CAPON mRNA Short-Form Expression by Diagnosis
Expression levels are least squares means. Mean values per category are plotted with 95% confidence intervals. The number of individuals per sample is indicated within each bar. Expression is significantly higher in patients with schizophrenia (p = 0.0013) and bipolar (p = 0.0009) as compared to controls. The mean (95% confidence interval lower bound, upper bound) for the control, schizophrenia, and bipolar groups are 1.34 (1.05, 1.62), 2.02 (1.73, 2.30), and 2.05 (1.77, 2.34), respectively.
Analysis of CAPON Isoform Expression by Genotype
All individuals were genotyped at rs1415263, rs4145621, and rs2661818, three SNPs within CAPON that were previously identified as being in significant linkage disequilibrium with schizophrenia [10]. For each SNP, individuals with one or two copies of the previously identified associated allele were observed to have higher group mean CAPON short-form expression than the group of individuals homozygous for the unassociated allele (Figure 5). All three SNPs individually showed significant differences among means for the short-form expression (rs1415263, p = 0.019; rs4145621, p = 0.022; rs2661818, p = 0.019), while none showed significantly different means for the full-length expression (rs1415263, p = 0.67; rs4145621, p = 0.52; rs2661818, p = 0.50). Genotypes with one or two copies of the associated alleles had higher mean short-form expression (rs1415263, 30%; rs4145621, 32%; rs2661818, 34%). None of the SNPs showed significant expression differences between individuals heterozygous or homozygous for the associated allele. Given that the prior demonstrated correlation between CAPON full-length expression and antipsychotic treatment could represent a medication treatment effect, the correlation analysis between CAPON full-length expression and genotype was rerun using only individuals not receiving antipsychotic medications at time of death (35 controls and 17 individuals with bipolar disorder). Again, there were no significant differences in mean CAPON full-length expression among genotypes for any of these SNPs (rs1415263, p = 0.41; rs4145621, p = 0.82; rs2661818, p = 0.58).
Figure 5 ACTB (Beta-Actin)–Normalized CAPON mRNA Short-Form Expression by Genotype
Expression levels are least squares means. Individuals from all three diagnostic classifications are included, grouped only by genotype. Mean values per genotype for each SNP are plotted with 95% confidence intervals. The number of individuals per genotype is indicated within each bar. SNP alleles are given for forward strand sequence. All three SNPs exhibit significantly (p < 0.05) different levels of CAPON expression by genotype, with a dominant effect. Higher levels of CAPON are seen in individuals with one or two copies of alleles previously identified as associated with schizophrenia (T, rs1415263; C, rs4145621; and C, rs2661818). The mean (95% confidence interval lower bound, upper bound) for the three genotypes for each SNP are as follows. For rs1415263: 1.54 (1.29, 1.80) for CC; 2.07 (1.81, 2.34) for CT; and 1.83 (1.36, 2.29) for TT. For rs4145621: 1.51 (1.25, 1.78) for TT; 2.01 (1.74, 2.27) for TC; and 2.01 (1.61, 2.41) for CC. For rs2661818: 1.49 (1.21, 1.76) for GG; 1.96 (1.70, 2.23) for CG; and 2.04 (1.68, 2.41) for CC.
Discussion
Our screening of a human fetal total brain cDNA library resulted in the identification of two isoforms of CAPON mRNA corresponding to two forms of CAPON protein. Our screen used only primers from exon 10, so we would not have detected isoforms lacking this portion of the gene. One of the two identified transcripts encompasses ten exons and encodes a 501 amino acid protein containing two known functional domains, an amino-terminal phosphotyrosine-binding domain and a carboxyl-terminal PDZ-binding domain. This full-length form corresponds to transcripts previously identified in the rat and human [12,22]. The second transcript contains the last two exons of CAPON and is predicted to produce a short form of the protein, 211 amino acids long and containing the PDZ-binding domain. Prior work has demonstrated that the terminal 125 amino acids of the full-length protein are sufficient to bind the PDZ-domain of nNOS and interfere with the binding between nNOS and PSD93 or PSD95 [12]. In addition to the ability of CAPON to bind to nNOS, the terminal 125 amino acids also appear to be able to directly bind to the second PDZ domain of PSD95 [23], the normal site of nNOS binding to PSD95 [13]. As the first 180 amino acids of CAPON have been previously demonstrated to contain the domain needed to bind to the amino-terminal targets Dexras1 and Synapsin [16,17], it would seem that only the full-length form of CAPON would be able to serve as an adaptor protein between nNOS and these targets. A physiological role of the short form would likely be limited to the competitive inhibition of binding of other ligands to the PDZ domains of nNOS and PSD93 or PSD95.
There are significant obstacles to the study of gene expression in the human brain. Obtaining high-quality postmortem samples suitable for RNA extraction is difficult and labor-intensive. Obtaining appropriate matched control groups is also a challenge. While it may be possible to collect samples with relative consistency across some variables, such as PMI or brain pH, many factors that may potentially affect gene expression, such as treatment history and substance use, are beyond the control of investigators. The rate of collection of individuals with too many clinical restrictions (e.g., treatment-naïve individuals with schizophrenia and no history of substance abuse, alcohol use, or smoking) would be too slow to produce a useful number of samples. Added to these clinical variables is likely etiological heterogeneity, with only a subset of affected individuals expected to harbor a primary causative mutation in any given gene. All of these factors may lessen the chance that significant differences in gene expression can be demonstrated using a particular sample.
We chose to conduct our expression studies using the Stanley Array Collection, as this collection contains samples from more individuals than other postmortem collections, and the samples were collected in a standardized fashion with an emphasis on obtaining high-quality RNA for expression studies. Limitations of this collection include the facts that protein samples were not available for parallel analysis, and that only one brain region, the DLPFC, was available for study. However, this brain region has long been hypothesized to be involved in schizophrenia, implicated by evidence from neuropsychological, neuroimaging, histopathological, and neurochemical studies (reviewed in [24]).
Our results suggest that mRNA expression of the short-form of CAPON is significantly (p < 0.005) increased in patients with either schizophrenia or bipolar disorder. If the short-form protein behaves as predicted, it would disrupt the binding of nNOS to PSD95 through competitive inhibition and remove nNOS from the NMDAR complex, thereby decoupling NO generation from NMDAR activation. This could produce a picture consistent with the NMDAR hypofunction hypothesis of schizophrenia. Based on our data, expression of short-form mRNA does not appear sensitive to treatment with antipsychotic medication. Full-length CAPON mRNA expression, in contrast, appears to be highly influenced by treatment with antipsychotic medication, at least in bipolar disorder. It will be of interest to further investigate and confirm this effect in subsequent studies. Pre- and postexposure expression studies in animals may be helpful in determining if the relationship between antipsychotic treatment and decreased CAPON mRNA expression is causal. While we found no significant group differences in expression levels between patients with schizophrenia or bipolar disorder and normal controls, it is possible that this is due to the normalization of full-length CAPON mRNA expression by antipsychotic treatment. Additional expression studies in individuals with schizophrenia not receiving antipsychotic medication would be of great interest to assess this possibility.
The Stanley Array Collection consists of samples collected from several locations within the United States, and therefore represents a sample that is independent from our Canadian familial schizophrenia collection. Nonetheless, there is significant evidence for association between affection phenotypes and the same alleles at three different SNPs in both samples. Consistent with the hypothesis that CAPON short form overexpression is associated with schizophrenia, the alleles observed associated with schizophrenia in our Canadian sample are significantly (p < 0.05) associated with higher short form expression in the Stanley Array Collection.
The three SNPs investigated span nearly 98 kb and are located in introns 2 and 3 of CAPON, the most proximal being 70 kb upstream from the short-form transcription start site. Although there is evidence for LD spanning large regions within CAPON, it is unlikely that these SNPs are in tight disequilibrium with polymorphisms in the short-form basal promoter [10,25]. Our prior work on this gene revealed that the three SNPs used in this study are in significant (p < 0.0001) linkage disequilibrium with each other (rs1415263 and rs4145621, D′ = 0.748; rs1415263 and rs2661818, D′ = 0.801; rs4145621 and rs2661818, D′ = 0.491), while being in much weaker LD (D′ values ranging from 0.074 to 0.432) with SNPs located in intron 8 (rs7521206) and exon 9 (rs348624) [10]. More probable is that the SNPs used in the present study are in LD with a mutation in an enhancer region that is located at some distance upstream of the short-form transcript. Enhancers can regulate gene expression from distances of up to 1 Mb [26], and mutations in enhancer sequences have been shown to be responsible for a number of human diseases [27].
Additional studies are needed to further examine the level of CAPON protein among individuals with different psychiatric diagnoses, in both the DLPFC and other regions of the brain. The implication that CAPON may influence schizophrenia susceptibility through disruption of NMDAR functioning adds to the list of candidate genes that may act at this receptor system, including Neuregulin 1 [28], D-amino acid oxidase and G72 [29], Dysbindin [30], and PPP3CC [31]. Additional work on the interaction of these different candidates may also further our understanding of the genetic component of schizophrenia susceptibility.
Accession Numbers
The Online Mendelian Inheritance in Man (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=OMIM) accession numbers for some loci discussed in this paper are SCZD (181500) and SCZD9 (604906). The GenBank (http://www.ncbi.nlm.nih.gov/Genbank/index.html) accession numbers for other loci and proteins discussed in this paper are ACTB (NM_001101), APOA2 (X02619), CAPON/NOS1AP (RNA, NM_014697; protein, NP_055512), and D1S2675 (Z52679).
Patient Summary
Background
Although the exact cause of schizophrenia remains unknown, people who inherit certain variations in genes are thought to be at higher risk. Scientists have been trying to get a handle on which genes are involved. They hope that this will help them understand the disease better and eventually lead to treatments and maybe even prevention. So far, a number of genes have been implicated, but it is important to confirm such findings, and to find out what the genes actually do that could lead to abnormalities in the brain.
Why Was This Study Done?
The scientists who did this study had previously discovered that a gene called CAPON might be involved in schizophrenia. The scientists had found that certain variations in the region of the gene were more common in patients with schizophrenia. In this study, the researchers examined CAPON in more detail and asked two main questions: (1) Are there differences in the activity of the CAPON gene in brains from patients with schizophrenia and bipolar disorder compared with those from healthy individuals? (2) Could they confirm the link between variations in the CAPON gene and the disease in a second study of different patients with schizophrenia?
What Did the Researchers Do and Find?
They looked for variations in the CAPON gene and measured the level of CAPON gene expression in samples from a collection of postmortem brain specimens. The collection contained brain tissue samples from 35 patients with schizophrenia, 35 patients with bipolar disorder, and 35 patients without psychiatric illness (known as “controls”). Genes are templates for proteins (which make up the majority of active components in cells and body), and the researchers found that the CAPON gene is a template for two different proteins, a short form and a long form. Brain samples from patients with schizophrenia and bipolar disorder had higher levels of the short form than brain samples from patients without psychiatric illness. Moreover, these higher levels of the short version were predominantly seen in people with versions of the CAPON gene that had been previously linked to schizophrenia.
What Does This Mean?
These results lend more support to the idea that CAPON is somehow involved in schizophrenia. Given that studies by other groups also point to a link between CAPON and schizophrenia, it seems clear that further study of CAPON is justified. The findings here suggest that we need to learn more about the short version of CAPON, and specifically what exactly the CAPON protein does in the brain. Because of the limitations of working with human postmortem samples, it is likely that most of the next round of experiments will be done in cell culture and in animal models.
Where Can I Find More Information Online?
The following Web sites provide information on schizophrenia.
US National Institutes of Mental Health (search for “schizophrenia”): http://www.nimh.nih.gov/
UK National Institute for Health and Clinical Excellence pages on schizophrenia: http://www.nice.org.uk/page.aspx?o=42770
The National Alliance for Research on Schizophrenia and Depression (NARSAD): http://www.narsad.org/
The National Alliance for the Mentally Ill (NAMI):
http://www.nami.org/
The Schizophrenia Society of Canada:
http://www.schizophrenia.ca/
This work was supported by grant R01 MH62440 to LMB and BLF from the National Institute of Mental Health (NIMH) and K25 AA015346 to SB from the National Institute on Alcohol Abuse and Alcoholism. EIC is supported by a Pharmaceutical Research and Manufacturers of America Foundation Postdoctoral Fellowship in Pharmacology and Morphology. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. RNA and DNA specimens were donated by the Stanley Medical Research Institute Brain Collection courtesy of Drs. Michael B. Knable, E. Fuller Torrey, Maree J. Webster, and Robert H. Yolken. Control tissue specimens were obtained from the Human Brain and Spinal Fluid Resource Center, Veterans Administration West Los Angeles Healthcare Center, 11301 Wilshire Boulevard, Los Angeles, CA 90073, United States, which is sponsored by the National Institute of Neurological Disorders and Stroke/NIMH, National Multiple Sclerosis Society, and the Department of Veterans Administration. Thanks to Kenyatta Lucas for preliminary studies on CAPON protein expression and to Dawn Little for technical assistance.
Citation: Xu B, Wratten N, Charych EI, Buyske S, Firestein BL, et al. (2005) Increased expression in dorsolateral prefrontal cortex of CAPON in schizophrenia and bipolar disorder. PLoS Med 2(10): e263.
Abbreviations
CAPONcarboxyl-terminal PDZ ligand of neuronal nitric oxide synthase
DLPFCdorsolateral prefrontal cortex
LDlinkage disequilibrium
NMDARN-methyl-D-aspartate receptor
nNOSneuronal nitric oxide synthase
PMIpostmortem interval
SDstandard deviation
SNPsingle-nucleotide polymorphism
UTRuntranslated region
==== Refs
References
McGuffin P Asherson P Owen M Farmer A The strength of the genetic effect. Is there room for an environmental influence in the aetiology of schizophrenia? Br J Psychiatry 1994 164 593 599 7921708
Bassett AS Chow EW Waterworth DM Brzustowicz L Genetic insights into schizophrenia Can J Psychiatry 2001 46 131 137 11280081
Brzustowicz LM Hodgkinson KA Chow EW Honer WG Bassett AS Location of a major susceptibility locus for familial schizophrenia on chromosome 1q21–q22 Science 2000 288 678 682 10784452
Brzustowicz LM Hayter JE Hodgkinson KA Chow EW Bassett AS Fine mapping of the schizophrenia susceptibility locus on chromosome 1q22 Hum Hered 2002 54 199 209 12771552
Shaw SH Kelly M Smith AB Shields G Hopkins PJ A genome-wide search for schizophrenia susceptibility genes Am J Med Genet 1998 81 364 376 9754621
Gurling HM Kalsi G Brynjolfson J Sigmundsson T Sherrington R Genomewide genetic linkage analysis confirms the presence of susceptibility loci for schizophrenia, on chromosomes 1q32.2, 5q33.2, and 8p21–22 and provides support for linkage to schizophrenia, on chromosomes 11q23.3–24 and 20q12.1–11.23 Am J Hum Genet 2001 68 661 673 11179014
Hwu HG Liu CM Fann CS Ou-Yang WC Lee SF Linkage of schizophrenia with chromosome 1q loci in Taiwanese families Mol Psychiatry 2003 8 445 452 12740602
Lewis CM Levinson DF Wise LH DeLisi LE Straub RE Genome scan meta-analysis of schizophrenia and bipolar disorder, part II: Schizophrenia Am J Hum Genet 2003 73 34 48 12802786
Rosa A Fañanás L Cuesta MJ Peralta V Sham P 1q21–q22 locus is associated with susceptibility to the reality-distortion syndrome of schizophrenia spectrum disorders Am J Med Genet 2002 114 516 518 12116186
Brzustowicz LM Simone J Mohseni P Hayter JE Hodgkinson KA Linkage disequilibrium mapping of schizophrenia susceptibility to the CAPON region of chromosome 1q22 Am J Hum Genet 2004 74 1057 1063 15065015
Zheng Y Li H Qin W Chen W Duan Y Association of the carboxyl-terminal PDZ ligand of neuronal nitric oxide synthase gene with schizophrenia in the Chinese Han population Biochem Biophys Res Commun 2005 328 809 815 15707951
Jaffrey SR Snowman AM Eliasson MJ Cohen NA Snyder SH CAPON: A protein associated with neuronal nitric oxide synthase that regulates its interactions with PSD95 Neuron 1998 20 115 124 9459447
Brenman JE Chao DS Gee SH McGee AW Craven SE Interaction of nitric oxide synthase with the postsynaptic density protein PSD-95 and alpha 1-syntrophin mediated by PDZ domains Cell 1996 84 757 767 8625413
Brenman JE Christopherson KS Craven SE McGee AW Bredt DS Cloning and characterization of postsynaptic density 93, a nitric oxide synthase interacting protein J Neurosci 1996 16 7407 7415 8922396
Coyle JT Tsai G Goff D Converging evidence of NMDA receptor hypofunction in the pathophysiology of schizophrenia Ann N Y Acad Sci 2003 1003 318 327 14684455
Fang M Jaffrey SR Sawa A Ye K Luo X Dexras1: A G protein specifically coupled to neuronal nitric oxide synthase via CAPON Neuron 2000 28 183 193 11086993
Jaffrey SR Benfenati F Snowman AM Czernik AJ Snyder SH Neuronal nitric-oxide synthase localization mediated by a ternary complex with synapsin and CAPON Proc Natl Acad Sci U S A 2002 99 3199 3204 11867766
Harrison PJ Owen MJ Genes for schizophrenia? Recent findings and their pathophysiological implications Lancet 2003 361 417 419 12573388
Kikuno R Nagase T Nakayama M Koga H Okazaki N HUGE: A database for human KIAA proteins, a 2004 update integrating HUGEppi and ROUGE Nucleic Acids Res 2004 32 D502 504 14681467
Ohara O Nagase T Ishikawa K Nakajima D Ohira M Construction and characterization of human brain cDNA libraries suitable for analysis of cDNA clones encoding relatively large proteins DNA Res 1997 4 53 59 9179496
R Development Core Team R: A language and environment for statistical computing 2005 Vienna R Foundation for Statistical Computing Available: http://www.R-project.org . Accessed 08 August 2005
Seki N Ohira M Nagase T Ishikawa K Miyajima N Characterization of cDNA clones in size-fractionated cDNA libraries from human brain DNA Res 1997 4 345 349 9455484
Tochio H Hung F Li M Bredt DS Zhang M Solution structure and backbone dynamics of the second PDZ domain of postsynaptic density-95 J Mol Biol 2000 295 225 237 10623522
Bunney WE Bunney BG Evidence for a compromised dorsolateral prefrontal cortical parallel circuit in schizophrenia Brain Res Brain Res Rev 2000 31 138 146 10719142
Hinds DA Stuve LL Nilsen GB Halperin E Eskin E Whole-genome patterns of common DNA variation in three human populations Science 2005 307 1072 1079 15718463
Lettice LA Heaney SJ Purdie LA Li L de Beer P A long-range Shh enhancer regulates expression in the developing limb and fin and is associated with preaxial polydactyly Hum Mol Genet 2003 12 1725 1735 12837695
Kleinjan DA van Heyningen V Long-range control of gene expression: Emerging mechanisms and disruption in disease Am J Hum Genet 2005 76 8 32 15549674
Stefansson H Sigurdsson E Steinthorsdottir V Bjornsdottir S Sigmundsson T Neuregulin 1 and susceptibility to schizophrenia Am J Hum Genet 2002 71 877 892 12145742
Chumakov I Blumenfeld M Guerassimenko O Cavarec L Palicio M Genetic and physiological data implicating the new human gene G72 and the gene for D-amino acid oxidase in schizophrenia PNAS 2002 182412499
Straub RE Jiang Y MacLean CJ Ma Y Webb BT Genetic variation in the 6p22.3 gene DTNBP1, the human ortholog of the mouse dysbindin gene, is associated with schizophrenia Am J Hum Genet 2002 71 337 348 12098102
Gerber DJ Hall D Miyakawa T Demars S Gogos JA Evidence for association of schizophrenia with genetic variation in the 8p21.3 gene, PPP3CC, encoding the calcineurin gamma subunit Proc Natl Acad Sci U S A 2003 100 8993 8998 12851458
|
16146415
|
PMC1201690
|
CC BY
|
2021-01-05 10:38:12
|
no
|
PLoS Med. 2005 Oct 13; 2(10):e263
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020263
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1615669410.1371/journal.pmed.0020271EssayEpidemiology/Public HealthVaccinesHealth PolicyInternational healthResource allocation and rationingMaking Practical Markets for Vaccines Why I decided that the Center for Global Development Report, Making Markets for Vaccines, offers poor advice to government and foundation leadersEssayLight Donald W Donald W. Light was a member of the CGD/Gates Foundation “Pull” Mechanisms Working Group for Making Markets for Vaccines. He is a professor of comparative health care systems at the University of Medicine and Dentistry of New Jersey (Stratford, New Jersey, United States of America), a fellow of the Center for Bioethics, University of Pennsylvania (Philadelphia, Pennsylvania, United States of America), and an adjunct fellow at the Leonard Davis Institute for Health Economics, University of Pennsylvania. E-mail: [email protected]
Competing Interests:The author declares that he has no competing interests
10 2005 13 9 2005 2 10 e271Copyright: © 2005 Donald W. Light.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.
Introduction of New and Underutilized Vaccines: Sustaining Access, Disease Control, and Infrastructure Development
Donald Light raises concerns over the current proposal to fund vaccine development through advanced market commitments to companies.
==== Body
This is the first of two articles in the October 2005 issue on ways to create practical markets for vaccines.
“In the time it takes you to read this preface, 100 people will die of diseases that can already be prevented with vaccines, and 150 more will die of malaria, HIV or tuberculosis [1].” So begins Making Markets for Vaccines, a report from the Center for Global Development (CGD) in Washington, D.C. that is being vigorously promoted to leaders of the G8 and foundations as a blueprint for how to spend billions of dollars in donations to end the economic and personal burdens of so much suffering and loss.
In order to stop children “dying at the rate of an average-sized high school every hour,” the report offers a plan that is “simple and practical”—make an “advanced market commitment” (AMC) to purchase the equivalent of the revenues that the big multinational firms would receive from major Western drugs, once the vaccine or vaccines are discovered, tested, and brought to market. This commitment, says the report, will induce companies to start investing serious research funds and unleash the creative powers of their large research teams to discover new, effective vaccines. In return for a large buyout of a few billion dollars, the company or companies that win the windfall contract(s) would commit thereafter to sell all doses at a very low, cost-plus price (i.e., basic cost plus a small profit margin). Contracts would also depend on poor countries participating, paying a co-payment of about a dollar a dose, and meeting other requirements. The result will be that hundreds of millions of children will get immunized against deadly diseases so that they can learn, create, and unleash the productive potential in poor nations to transform themselves.
Vaccines won't reach their targets
(Illustration: Emma Burns)
The AMC, which is not like a market in most ways, becomes a long-term contract best aimed at late-stage and existing vaccines, not at research for nonexistent vaccines. An advanced commitment is a slower, less efficient way to incentivize research to discover an effective new vaccine than direct research support. As a complement to public and charitable funding of research and development, an advanced commitment can buy many more million doses, to save millions more lives, at a much lower price because the risk and cost of research and development are being borne by the funders. While advanced commitments are a good idea for overcoming long delays due to patent enforcements, what leaders need is a different kind of report on how to make a big splash and a real difference with between US$1 billion and $5 billion, a report that outlines how advanced commitments can be most effective in saving lives and how the key issues in the manufacturing, organization, and delivery of vaccines in poor regions can be addressed. I will first identify the problems in what could be called the “core draft” of the CGD model by Michael Kremer, who holds the Gates professorship at Harvard, and then comment briefly on how the final report to leaders fudges and qualifies Kremer's model with add-ons to please all readers, so that what is being proposed becomes less coherent and more difficult to pin down.
The Context
As a professor of comparative health-care systems, I served on the “Pull” Mechanisms Working Group (the Group) that the Gates Foundation funded the CGD to manage. Over the past several years, the Gates Foundation has transformed vaccine and drug research and development (R&D) for global diseases through bold funding and institution-building. These are called “push” efforts, because they use the direct force of contracts, funds, and grants to push along leading projects and programs. As a result, over a dozen promising new vaccines are entering clinical trials, or soon will be. But Kremer had proposed using the “pull” of a large financial commitment as the way to induce R&D in the private sector for neglected diseases (parts 1 and 2 of [2]). Though this is called an “advanced market commitment,” it is not a market, but one or a few donors making a large purchase. The Group should have explored and assessed the pros and cons of various pull mechanisms, but I felt it increasingly became a cheering squad for Kremer's model, which then was applied to malaria as the way to supersede the many previous efforts by government- and foundation-sponsored scientists to discover an effective vaccine. The more I learned as a neophyte about how weak the evidence was that this appealing idea would work and the ways it might make things worse, the more doubtful I became.
Little Bang for Big Bucks
Two studies were featured in the report as proof that advanced commitments are a revolutionary technique to launch a new era of innovation. The study by Finkelstein provides a systematic analysis of how advanced commitment funding for vaccines has affected investments in R&D [3]. Finkelstein finds that only large firms respond to the inducement, by taking an already-discovered vaccine off the shelf and testing it, such as GlaxoSmithKline's (GSK's) vaccine candidate for malaria with modest, short-term efficacy [4]. Ironically, this is the only candidate mentioned in the Group's report, Making Markets…—yet it was push funding for testing, not pull inducement, that apparently got GSK to take it off the shelf after 15 years, and start trials. Finkelstein found that small firms, where most innovation is taking place, begin to participate later in a sustained larger market, which this advanced commitment model is not designed to create. Finkelstein concludes from her large sample that “for every $1 permanent increase in expected annual market revenue from vaccines against a particular disease [the CGD design], the pharmaceutical industry will spend an additional 6 cents annually in present discounted value on R&D for vaccines against that disease” (p. 543 of [3]).
The other major study cited in the CGD report as evidence that a $3 billion advanced commitment would have long, deep pull back to basic research (rather than short, shallow pull to fund clinical testing) comes to the implausible conclusion that just a 1% increase in market size leads to a 4%–6% increase in new drugs [5]. Not a 4- to 6-fold increase in research funding (also implausible) but a 4- to 6-fold increase in actual new drugs! This miraculous conclusion is stated as if it were fact, when it is based on a highly artificial econometric model. The model assumes that all individuals live indefinitely, that there is only one firm at any one time with the best-practice technology, that anticipated future market size (not actual size) prompts more innovation over long periods, and that “new drugs” include all generics and all newly approved drugs, even though less than 15% of the latter are therapeutically superior to existing drugs [6,7]. Like Finkelstein, the authors sensibly note that “pharmaceutical companies may respond more to profit incentives at the later stages of the research process than at the earlier stages.” Thus, both studies support using advanced commitments to encourage late-stage development, not basic research to discover new drugs or vaccines.
The CGD model creates a one-time market and does not address sustainability.
The studies cited to prove that donor-pull will spur companies to invest in basic research that might or might not discover an effective vaccine 10–15 years down the road in fact offer dubious evidence. Further, the vaccine business is technically different from drugs, and most of the big companies decided years ago to get out of it. Is an advanced commitment for one vaccine (or one disease—an ambiguity that creates further problems) enough to get them back into the vaccine business? A central problem is that the CGD model creates a one-time market and does not address sustainability. Meanwhile, the few companies that have vaccine research teams are already being funded directly or through public–private partnerships (PPPs), often by the Gates Foundation, so that an AMC for research is unnecessary. Finally, going after a big contract designed not to pay a penny until a company has invested a decade or more in discovery, development, testing, and approval is a less cost-effective way to commit billions of dollars than to do what Gates and others are doing already: funding the best basic research ideas (including from private-sector teams), creating PPPs and other bridging organizations, and bringing the best experts together in a global research community.
Market-induced basic research is still less plausible in the CGD model, because the more closely one reads the text, the less clear it becomes how much a company would actually get if it were to gamble hundreds of millions so that it could discover an effective vaccine. The core Kremer model comes up with $3 billion to match the average sales of an individual top-selling drug in order to make investing in research as attractive as for other products. But then it makes room for second or third successful vaccines by other companies, among whom the total amount of money has to be shared. Contracts also depend on the governments of each participating poor country agreeing to terms as subsidized purchasers. Then the final report shifts the argument from an advanced commitment for a vaccine to an advanced commitment for all vaccines for a given disease. In sum, these provisions make it unclear how much a company would get after years of R&D investments.
Table 1 Contrasting Models of Advanced Purchase Commitments
These same provisions also make a binding contract impossible, because the donor cannot specify what it would pay a company if it invests in research to discover a new vaccine. And what is a company to make of the assurance that an advanced commitment will not cost the donor a penny until an effective vaccine meets the contractual criteria? If the advanced commitment requires no set-aside, why should investors and companies think it's real and not subject to executive or political change? If donors' financial commitment is real, why not save real lives by committing to make an effective vaccine available to the world's poor now, rather than possibly save hypothetical lives years from now?
The Scientific Barriers to Vaccines
Besides weak evidence that a $3 billion advanced commitment would induce basic research, nothing is mentioned about the daunting scientific barriers to developing a vaccine for either malaria or HIV-AIDS. The Kremer model assumes that creating a large purchase will induce a solution; but scientists who have done the research say that the scientific obstacles may be insurmountable because the targets are multiple and evolving. This observation leads to a more serious weakness in a global competition for a big contract: it rewards scientific secrecy rather than sharing, whereas the cooperative push efforts in recent years have fostered partnerships and sharing. Here is a stark trade-off. Which is more likely to lead to better vaccines faster—fierce competition for a big future payoff or cooperative sponsorship and PPPs? The more cooperative government, university and nonprofit research teams will probably get nothing under the advanced commitment model.
The other big trade-off question was (and is): will committing large sums to the deep, long pull of an advanced commitment mean less money for grants and contracts to push vaccine development forward? The report asserts it would not. I find that suspect. I was told, in support of this assertion, that wealthy countries are ready to commit billions, and then billions more, to eradicating global diseases of the poor beyond the multinational scheme to buy and administer existing but underused vaccines. Is that true? If so, why have at least two studies concluded that foundations and governments (especially European) have not yet adequately funded R&D for neglected diseases [8,9]? Three billion dollars more for research will foster more innovation than $3 billion committed as an inducement for more research.
The big trade-off question gets buried by emphasizing that advanced commitments are to be added to current push efforts to “complement” them, as if committing a few billion dollars to “pull” funding has no effect on “push” funding. But if it does, the CGD report itself documents how much more progress has been made, for a fraction the cost, through directly funded grants and programs. Ironically, complementary uses of pull mechanisms with push ones were little discussed by the Group over the months of deliberation. Criticisms of the Kremer draft led to softening the final report but not to substantive development of synergistic combinations. Those are still waiting to be done.
To summarize, the rationale for Kremer's model, which still lies behind all the add-ons and qualifiers of the CGD final report, assumes that a large purchase will unleash innovative research to discover effective vaccines for the world's most intransigent diseases [10]. It is promoted, as Farlow notes of Kremer's book, “in much the same way that some pharmaceutical companies promote ‘;wonder drugs’: emphasizing the positives, burying the negatives, and ending up suggesting that we now have all the answers…” [11]. Neither evidence nor logic support the Kremer and CGD model, and advanced commitments for early-stage research can crowd out faster, more effective efforts both politically and economically. The CGD model belies its president's call for a “global commons” in which the best minds and teams work together for “a global social contract” to benefit humanity [1,7]. Why is there such a discrepancy between the rhetoric and the reality of the CGD model?
Designed for Big Pharma
As drafts of the CGD report progressed, the number of contractual features and one-sided passages that favored the multinational corporations made me increasingly uncomfortable. Here are several examples:
Why is the advanced commitment contract designed so that competing firms get no money until a new vaccine is fully tested and approved? Only big Western firms have the cash reserves to sink hundreds of millions into research to discover and develop new vaccines, shutting out smaller companies in Asia, the Americas, and Africa. Interim and milestone payments were suggested but rejected as part of push grants, not pull AMCs. There are good reasons for using such payments in both initiatives. The final report keeps repeating that the process is open to all, but the contractual terms allow only cash-rich corporations to gamble for years for a possible big payoff and exclude future biotech companies that discover a vaccine after the initial contacts are signed.
Why do the winners get to keep patent rights, when these patent rights are the principal reason for the long delays in getting vaccines to poor countries at low prices? Drug companies with patent rights do not have a good record for sharing and building a global commons. Sharing and combining vaccines for malaria is especially important. A $3 billion advanced commitment is supposed to be a windfall buyout to shortcut access to poor nations, and it should include the rights and technical know-how needed for flexible capacity-building for that price. In fact, in many cases those rights could probably be bought for a tenth of that price.
What happened to the early goal of building up technical and manufacturing capacity in each continent? Several design features of the CGD model mitigate against it.
Why were principal legal advisers to big pharma chosen to do all the legal work, rather than a more neutral source? They are now coauthors with Group members as part of the promotional push for the “one true answer.” And why are the contractual term sheets drawn up by these advisers so vague in all the critical places? Innovative firms in Korea, India, China, Cuba, Brazil, or elsewhere outside the big pharma US–UK club are unlikely to trust this contractual process. Finally, why launch the report in the offices of principal legal advisers to big pharma?
Why is the cost of an advanced commitment set to the sales curves of drugs rather than to the sales curves of better-selling vaccines? Why does the report draw almost exclusively on industry-supported data and studies for the “facts” on which the advanced commitment is based? The result, when combined with the other points, is a bonanza for big pharma, and the text indicates that $3 billion is only a starting price, which is likely to increase rapidly to between $5 billion and $8 billion.
After the big payoff for 200 million courses, little is said (or was discussed) about how to sustain the vaccine effort. Sustainability is a major issue in vaccines for the poor; yet all the focus here appeared to be on a multi-billion dollar payment to big pharma.
Almost no time was spent analyzing the organizational, regulatory, and financial causes of past delays in making new vaccines available in poor countries. Will a $3 billion buyout solve all the sources of delay? Learning from the past did not seem to be the point.
Likewise, no time was spent understanding the organizational, political, and cultural barriers to effective delivery of the vaccines, only purchasing them. Rather than actually delivering vaccines to people, is a windfall purchase the real goal here? As an expert in health-care delivery, I could not endorse a report that ignored these issues.
Why is GSK's marginally effective vaccine candidate mentioned by name in the report—and why are the terms of contract then made loose enough so that a small, hand-picked committee is permitted to lower (but not raise!) the minimal thresholds for a vaccine to be acceptable?
Answers to such questions were brought into focus by the comment in Europe of a senior, international expert on vaccines and their markets. He explained that the major companies are running out of markets to sustain their rapid growth. That's why they're turning sexual performance or shyness into medical problems. They have been looking for years for a way to make a profitable market out of global vaccines, and in the CGD group's proposal it looks as if they have found a way: “Why don't they just say they want to give GSK $3 billion for their marginally effective vaccine?”
Were members of the CGD group being used as agents for this agenda?
Making Markets for Sustainable Cheap Vaccines
The reasonable doubts here that led me to withhold my endorsement of the CGD report do not address a number of other serious concerns: how difficult, for instance, it is to get the buyout price, and especially the post-buyout price, right years in advance (Box 1). There are also problems with the contracts, the oversight committee, and liability issues; problems of inequities; and problems with the increasingly confused terms of what is being proposed—issues taken up in more detail elsewhere [12–14]. The G8 finance ministers have been misadvised to write that advanced purchase commitments are a potentially powerful mechanism to incentivize research [15]. But none of these problems detracts from my thinking that advanced purchase commitments are a good idea when applied where they work best: on existing vaccines that could save millions from suffering and dying now. It seems morally dubious for a foundation or nation to do otherwise. The singular omission in the Grand Challenges in Global Health is that they do not call for the eradication of all the diseases for which effective vaccines already exist [16,17]. When millions of lives could be saved now, why give priority to future lives that might or might not be saved?
Box 1. The Price Is Wrong
Remember the quiz show, “The Price is Right,” where contestants guessed how much things actually cost? A big problem with advanced purchase or market commitments is that guessing both the payoff price and the post-payoff price is very difficult when attempted years before one even knows the kind of effective vaccine that will be discovered and how it is to be administered.
To be fair to competitors, the payoff price should be adjusted for how much R&D was paid for by governments and foundations, because some risk much less of their own money, net of tax subsidies, than others. The CGD price of $15 per course and $3 billion might be much too high, or too low, by 2015. And no adjusters are mentioned for external subsidies and other factors.
The permanent post-payoff cost of manufacturing most vaccines in volumes above 10 million units, according to information given to our Group, is very low, between one and five cents. But a new-age biologic vaccine might cost much more. The CGD price of $1 per course is quite high compared to many generic vaccines and for many poor governments. But then, it might not cover the costs of a technically expensive vaccine. Both prices can be right, however, if an advanced commitment can be made for an existing vaccine to eradicate a dread disease.
An advanced commitment as a complement to paying for R&D could be designed to establish a sustainable, long-term market for an effective vaccine to eradicate a global disease. With little risk or private investment to pay off, one could commit to 600 million doses for $3 billion rather than 200 million doses. The terms should build in financial support as well as expert help to strengthen the public health delivery systems of recipient nations and their capacity to build the new vaccine into their budgets and planning. The donor could announce honestly that it is eradicating a global scourge, instead of saying that it might do so ten years from now. Licenses for low-income markets as well as manufacturing know-how would be part of the deal, and favoring regional manufacturers would be a related goal. Through this kind of flexible, long-term contracting focused on delivery and capacity-building, an advanced commitment could create sustainable, whole-systems markets for new vaccines that current R&D efforts are pushing forward. This is one idea, but we need the kind of report I described at the beginning, which assesses this model along with other forms of advanced commitments and push–pull combinations.
I am indebted to enlightening conversations with, and comments from, Owen Barder, Andrew Farlow, Michael Kremer, Steve Landry, Ruth Levine, Richard Mahoney, Pauline Rosenau, Raj Shah, and Roy Widdus. They are, of course, not responsible for the content of this essay.
Citation: Light DW (2005) Making practical markets for vaccines. PLoS Med 2(10): e271.
Abbreviations
AMCadvanced market commitment
CGDCenter for Global Development
the Groupthe “Pull” Mechanisms Working Group of the Center for Global Development
GSKGlaxoSmithKline
PPPpublic–private partnership
R&Dresearch and development.
==== Refs
References
Center for Global Development Making markets for vaccines: Ideas into action 2005 Washington (D. C.) Center for Global Development 115
Kremer M Jaffe AB Lerner J Stern S Creating markets for new vaccines: Parts 1 & 2 Innovation policy and the economy 2001 1 Cambridge (Massachusetts) MIT Press 300
Finkelstein A Static and dynamic effects of health policy: Evidence from the vaccine industry Q J Econ 2004 119 527 564
Alonso PL Sacarlal J Aponte JJ Leach A Macete E Efficacy of the RTS,S/ASO2A vaccine against Plasmodium falciparum infection and disease in young African children: Randomized controlled trial Lancet 2004 364 1411 1420 15488216
Acemoglu D Linn J Market size in innovation: Theory and evidence from the pharmaceutical industry Q J Econ 2004 119 1049 1090
National Institute for Health Care Management Research and Education Foundation Changing patterns of pharmaceutical innovation 2002 Washington (D. C.) National Institute for Health Care Management Available: http://www.nihcm.org/innovations.pdf . Accessed 17 July 2005
Prescrire International A review of new drugs and indications in 2002: Financial speculation or better patient care? Prescrire Int 2003 12 74 77 12674131
Towse A The initiative on public-private partnerships for health: Estimates of the medium term financial resource needs for development of pharmaceutical to combat neglected diseases 2004 London Office of Health Economics 47
Moran M New EU approaches to funding R&D for neglected diseases 2005 Available: http://www.who.int/intellectualproperty/submissions/Mary.Moran.pdf . Accessed 15 July 2005
Kremer M Glennester R Strong medicine: Creating incentives for pharmaceutical research on neglected diseases 2004 Princeton (New Jersey) Princeton University Press 152
Farlow A Over the rainbow: the pot of gold for neglected diseases Lancet 2004 364 2011 2012
Farlow A Light DW Mahoney RT Widdus R Concerns regarding the Center for Global Development report, “Making Markets for Vaccines.” 2005 Geneva Commission on Intellectual Property Rights, Innovation and Public Health Available: http://www.economics.ox.ac.uk/members/andrew.farlow/CIPIH1May2005.pdf . Accessed 15 July 2005
Barder O Kremer M Levine R Answering concerns about Making Markets for Vaccines 2005 Geneva Commission on Intellectual Property Rights, Innovation and Public Health (CIPIH) Available: http://www.who.int/intellectualproperty/submissions/BarderSubmission.pdf . Accessed 15 July 2005
Maurer S The right tool(s): Designing cost-effective strategies for neglected disease research 2005 Berkeley (California) University of California at Berkeley Goldman School of Public Policy Available: http://www.who.int/intellectualproperty/studies/S.Maurer.pdf . Accessed 15 July 2005
G8 Finance Ministers G8 Finance Ministers' conclusions on development. June 2005 London HM Treasury Available: http://www.hm-treasury.gov.uk./otherhmtsites/g7/news/conclusions_on_development_110605.cfm . Accessed 1 August 2005
Gates Foundation Grand challenges in global health: Gates Foundation, 2005 2005 Available: http://www.grandchallengesgh.org/subcontent.aspx?SecID=408 . Accessed 15 July 2005
Birn AE Gates's grandest challenge: Transcending technology as public health ideology 2005 Available: http://image.thelancet.com/extras/04art6429web.pdf . Accessed 15 July 2005
|
16156694
|
PMC1201691
|
CC BY
|
2021-01-05 10:38:14
|
no
|
PLoS Med. 2005 Oct 13; 2(10):e271
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020271
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1615669510.1371/journal.pmed.0020276EssayInfectious DiseasesEpidemiology/Public HealthHIV/AIDSHIV Infection/AIDSMedicine in Developing CountriesHealth PolicyResource allocation and rationingFree Antiretrovirals Must Not Be Restricted Only to Treatment-Naive Patients Experience in Uganda suggests that restricting access is not the way forwardEssayColebunders Robert Kamya Moses Semitala Fred Castelnuovo Barbara Katabira Elly McAdam Keith *Robert Colebunders, Moses Kamya, Fred Semitala, Barbara Castelnuovo, Elly Katabira, and Keith McAdam are at the Infectious Diseases Institute, Faculty of Medicine, Makerere University, Kampala, Uganda, and Robert Colebunders is also at the Prince Leopold Institute of Tropical Medicine, Antwerp, Belgium and at the University of Antwerp, Antwerp, Belgium.
*To whom correspondence should be addressed. E-mail: [email protected]
Competing Interests: A substantial grant to build the Infectious Diseases Institute, where the authors are based, was given by the Philanthropy Division of Pfizer to the Academic Alliance for AIDS Care and Prevention in Africa (http://www.ifpma.org/Health/hiv/health_academic_hiv.aspx). However, the institute does not receive any antiretroviral medications from Pfizer.
10 2005 13 9 2005 2 10 e276Copyright: © 2005 Colebunders 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.In some antiretroviral drug distribution programmes, free drugs are provided only, or preferentially, to patients who are treatment-naive. Colebunders and colleagues argue that such a restricted policy is highly problematic.
==== Body
Despite the “3 by 5” initiative [1], which aims to treat 3 million people with antiretrovirals (ARVs) by the end of 2005, access to ARVs in resource-poor settings remains limited. Moreover, in some ARV-access programmes, free ARVs are provided only, or preferentially, to patients who are ARV naive.
Treatment restricted to patients who are ARV naive was initially the case in Uganda, with the free ARVs provided by the Multi-Country AIDS Programme and the Presidential Emergency Plan for AIDS Relief. While in the policy documents of both projects it was not stated that free ARVs should be given preferentially to ARV-naive patients, many physicians involved in the roll-out of ARVs in Uganda felt that this was what these projects were recommending. Indeed, the Ministry of Health's “Antiretroviral Treatment Policy for Uganda” stated that “those who are clinically eligible and can afford to pay for ART will be encouraged to do so. Those already in privately provided and privately paid ART should be encouraged to remain in this situation. Others who become clinically eligible over time and have the ability to pay or have a third party able to pay in their place or cost share with them, should pay full cost for ART, whether they avail themselves of treatment provided in the public or private sector” [2]. These policy statements were interpreted by many clinicians to mean that patients who are already paying for their drugs can afford to do so and are not a priority for free drugs.
In this essay, we explain the rationale for restricting free access to treatment-naive patients, and then we outline the reasons why such restriction is highly problematic.
The pharmacy at the Infectious Diseases Institute
(Photo: R. Colebunders)
Counting pills during a clinical trial at the Infectious Disease Institute, where the authors are based
(Photo: R. Colebunders)
Why Programmes Restrict Access
There are four principal reasons why free ARVs are being restricted to treatment-naive patients. Firstly, ARV treatment regimens are more predictable in ARV-naive patients [3]. Programmes limited to this population will provide what donors most desire: good outcomes. Secondly, ARV-naive patients will usually respond readily to less expensive first-line ARV regimens, and only a few will require more complex and costly second-line ARVs. Thirdly, results of programmes are easier to compare, because similar ARV regimens are started in more homogeneous, comparable populations. And finally, it is often believed that patients who in the past have been able to buy ARVs should continue to do so.
Problems with a Restricted-Access Strategy
There are, however, serious problems with this strategy of restricted access. To start, patients are quick to share information, and rumours spread fast regarding ways to obtain free ARVs. When patients learn that ARVs are being given at no cost only to treatment-naive patients, they may not disclose that they have taken ARVs in the past, even to experienced counsellors. We discovered such patients in our centre when they were found to have an undetectable viral load or a CD4+ lymphocyte count higher than expected, when tested prior to starting ARVs. Only when confronted about these results did they acknowledge prior ARV experience. Such withholding of information may result in clinicians choosing inappropriate ARV regimens, thereby placing patients at risk of adverse effects or of development of resistance.
Second, forcing some patients to pay for ARVs can have disastrous consequences for adherence to long-term treatment. Our experience in Uganda has shown that if ARVs are provided for free through an international organisation or a clinical trial, patients achieve excellent adherence. In contrast, if patients or their relatives pay for drugs themselves, treatment failure and the development of resistance is frequently observed [4]. A study in Senegal found that user fees had a negative effect on adherence and were associated with frequent interruptions in treatment [5], and there have been similar findings in Kenya [6], Botswana [7], Cote d'Ivoire [8], and Nigeria [9]. Data from West Africa showed that the level of adherence in the treated population was inversely proportional to the amount of co-payment [10].
In Uganda, currently about 56,000 patients are being treated with ARVs, and about 52% of these treatment regimens are paid for by patients or their relatives (E. Namagala [Uganda Ministry of Health], personal communication). Most of these patients are relatively poor and most have had life-threatening HIV complications. Almost all are struggling with the expense. Many in the near future will be forced, from lack of resources, to either stop their ARVs or take them irregularly. Patients and their families are more willing to sacrifice scarce resources to pay for ARVs when patients are ill, but once they achieve a better health status, additional priorities, such as school fees, tend to take precedence, resulting in poor adherence. It is, therefore, essential that free ARVs be offered to all patients if they are to remain well and if development of resistance is to be thwarted.
Selecting only treatment-naive patients for free ARVs raises a human rights issue.
Third, programmes that offer ARVs only to ARV-naive patients may be relatively slow to enrol patients. All patients beginning ARV therapy for the first time must be carefully selected and prepared, and this can take several clinic visits, including counselling sessions. Moreover, once patients have been started on ARVs, they must be followed closely for side effects and to ensure adherence to the regimen. In contrast, patients already taking ARVs and known to be adherent to their regimen could be enrolled quickly, with less effort.
The Human Rights Dimension
Finally, we also believe that selecting only ARV-naive patients for free ARVs raises a human rights issue. Can treatment be denied to those who have somehow found money to initiate therapy—often forestalling their demise—and who are now struggling to pay for ARVs?
Most of these patients have made great sacrifices to access ARV treatment. They have kept appointments at the clinic, and they have been adherent to their treatment regimen as long as money was found to purchase ARVs. In contrast, other individuals may have delayed being tested for HIV, while continuing risky behaviour, and only accepted HIV testing because of the promise of free ARVs. Many of them have not been as ill as those who have managed to somehow find resources to initiate treatment.
Since the increased access to free ARVs, a sharp increase in HIV testing has been noted at voluntary testing and counselling sites in Kampala, Uganda, and also in many other countries [11]. In Uganda in the Mulago–Mbarara Teaching Hospitals' Joint AIDS Programme supported by the President's Emergency Plan for AIDS Relief, 15,000 patients were tested for HIV in the last seven months, and 40% of these patients were HIV positive. Although these newly diagnosed ARV-naive patients who meet the biological criteria for ARV treatment should be treated, we must be able to also prioritise individuals for ARVs who are “HIV veterans” and are now barely finding money or have already exhausted their resources to pay for their drugs.
Today in Uganda, ARVs provided by the Multi-Country AIDS Programme, the Global Fund to Fight AIDS, Tuberculosis, and Malaria, and the President's Emergency Plan for AIDS Relief are also used for patients that previously paid for the drugs themselves. Nevertheless, many patients still have not been taken in by these free ARV programmes. It is important that ARV roll-out projects in other countries learn from this experience in Uganda.
Citation: Colebunders R, Kamya M, Semitala F, Castelnuovo B, Katabira E, et al. (2005) Free antiretrovirals must not be restricted only to treatment-naive patients. PLoS Med 2(10): e276.
Abbreviation
ARVantiretroviral
==== Refs
References
World Health Organization Scaling up antiretroviral therapy in resource-limited settings: Treatment guidelines for a public health approach 2003 Geneva World Health Organization Available: http://www.who.int/3by5/publications/documents/arv_guidelines/en/index.html . Accessed 28 July 2005
Ministry of Health Antiretroviral treatment policy for Uganda 2003 Kampala Ministry of Health Available: http://www.aidsuganda.org/pdf/ART_Policy_draft_June.pdf . Accessed 1 August 2005
Losina E Islam R Pollock AC Sax PE Freedberg KA Effectiveness of antiretroviral therapy after protease inhibitor failure: An analytic overview Clin Infect Dis 2004 38 1613 1622 15156451
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 Acquir Immune Defic Syndr 2003 34 281 288 14600572
Laniéce 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
African Woman and Children Feature Service AIDS patients quitting treatment 2004 September 30 Nairobi: African Woman and Child Feature Service
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 Acquir Immune Defic Syndr 2003 34 281 288 14600572
Laguide R Elengua N Fassinou P Moatti JP Coriat B Souteyrand Y Barnett T Dumoulin J Direct costs of medical care for HIV-infected children before and during HAART in Abidjan Cote D'Ivoire 20002002. Economics of AIDS and access to HIV/AIDS care in developing countries: Issues and challenges 2003 ANRS Paris: Agence Nationale de Recherches sur le Sida 311 327
Daniel OJ Ogun SA Odusoga OL Falola RL Ogundahunsi OA Adherence pattern to ARV drugs among AIDS patients on self-purchased drugs and those on free medications in Sagamu Nigeria [abstract] XV International AIDS Conference; 2004 July 1116; Bangkok Thailand 2004 Available: http://www.iasociety.org/ejias/show.asp?abstract_id=2171173 . Accessed 5 August 2005
Kazatchkine M Antiretroviral treatment for HIV-infected patients in developing countries [abstract] Sixth International Congress on Drug Therapy in HIV Infection; 2002 November 1721; Glasgow United Kingdom 2002 Available: http://www.hiv6.com/sci_prog/pdf/hiv6.pdf . Accessed 1 August 2005
Joint United Nations Programme on HIV/AIDS 2004 report on the global AIDS epidemic: 4th global report. Geneva: Joint United Nations Programme on HIV/AIDS 2004 Available:http://www.unaids.org/bangkok2004/GAR2004_pdf/UNAIDSGlobalReport2004_en.pdf . Accessed 3 August 2005
|
16156695
|
PMC1201692
|
CC BY
|
2021-01-05 10:39:15
|
no
|
PLoS Med. 2005 Oct 13; 2(10):e276
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020276
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1615669310.1371/journal.pmed.0020286EssayVaccinesHealth PolicyInternational healthResource allocation and rationingIntroduction of New and Underutilized Vaccines: Sustaining Access, Disease Control, and Infrastructure Development EssayAndrus Jon Kim *Fitzsimmons John 10 2005 13 9 2005 2 10 e286Copyright: 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.
Making Practical Markets for Vaccines
For vaccines to reach the greatest number of people requires access, accelerated regional disease control, and the development of public health infrastructure.
==== Body
This is the second of two articles in the October 2005 issue on ways to create practical markets for vaccines.
Thinking of ways to create practical markets for vaccines poses some interesting challenges, particularly in the context of introducing new and underutilized vaccines. In a complex, ever-changing world, making vaccines, whether they are new or old, available to children and families who need them most should be a top priority [1]. Models for creating markets have been recently developed that might benefit from the experience gained from the regional and national immunization approaches used in the Americas. From the perspective of the Pan American Health Organization (PAHO), strategies to optimize the market and, ultimately, the use of vaccines include three important, overriding guiding principles: access, accelerated regional disease control, and the development of a public-health infrastructure.
Access Is Essential
Access to preventive and other health services covers several factors, including access to health units and hospitals, trained staff providing quality services, cost-effective technologies, the best use of available technologies, and information to improve the community's knowledge base and best practices.
Inherently, improved access also addresses issues of equity. Equity is a critical cross-cutting concept that drives much of the work of PAHO and member countries. For vaccines, the challenge has always been, and will continue to be, to ensure that all communities benefit from the potential impact of these technologies [2].
The Example of Rubella
The initiative to eliminate rubella and congenital rubella syndrome in the Americas serves as an example of PAHO's member countries seizing the opportunity to ensure access, while promoting the ultimate goal of disease control (virtual elimination of a disease), and the improvement of the public-health infrastructure [3]. These efforts help ensure that all families receive the benefit of vaccine technology because disease elimination requires that all communities be reached, regardless of ethnic background, religion, or income. Although the rubella vaccine was only recently introduced in many countries of the Americas, many of them simultaneously accelerated the rubella disease control strategies by strengthening surveillance and conducting mass vaccination campaigns to rapidly reduce the pool of rubella-susceptible individuals in the community. Such efforts eventually led to the adoption of a regional rubella-elimination initiative in the Americas by the year 2010 [4].
Public-sector capacity development creates regional independence.
Targeting High-Risk Districts
In the Americas, targeting high-risk districts with low immunization coverage for special attention also promotes improved access and accelerated disease control. Efforts to strengthen leadership, management, and supervision of program activities have been essential for the success of this strategy. Districts that have less than 95% coverage are targeted for special training, outreach, and follow-up. This targeting helps ensure that even in middle-income countries the substantial disparities that exist in health are addressed. To that end, PAHO's member countries have conducted the annual Vaccination Week in the Americas, during which all countries attempt to reach and vaccinate marginalized, poor populations [5].
Global Approaches to Vaccination
In the context of access and equity, there is concern about global approaches to support only countries with average annual per capita income of less than $1,000, such as the approach taken by the Global Alliance for Vaccines and Immunization, thus limiting the scope of work and benefits that impoverished children and families could otherwise receive in the Americas. Currently, only six countries are eligible to receive support from the Global Alliance for Vaccines and Immunization: Bolivia, Cuba, Guyana, Haiti, Honduras, and Nicaragua—representing only 7% of the population of the Caribbean and Latin America. However, 11% of people in the Caribbean and Latin America live below the international poverty line ($365 annual per capita income) [6]. In addition, huge discrepancies exist within countries, and these countries urgently need extra resources to ensure access of vaccines to marginalized populations.
Disease Control in the Americas
In the Americas, efforts to control disease over the last 20 years have led to significant improvements in the public-health infrastructure, particularly in the area of program management, surveillance, and public-health laboratories. In some countries, such as Brazil, the development of public-sector capacity for vaccine production was also a top priority. Such public-sector capacity development creates regional independence, competition to reduce vaccine prices among private-sector producers, and dependable supply chains that are sustained and that contribute to intercountry cooperation [7]. Currently, Brazil produces a yellow fever vaccine that has been used in (and in some cases donated to) neighboring countries suffering from deadly yellow fever outbreaks, for example, the yellow fever outbreak in Colombia in 2004.
Future Approaches to Practical Markets for Vaccines
Future approaches that attempt to enhance practical markets for vaccines and that enhance the introduction of new and underutilized vaccines should consider prioritizing the following: access and equity for as much of the population as is possible, well-implemented accelerated disease-control and prevention strategies, and development of a public-health infrastructure.
The general approach also requires high-level political commitment, adequate attention to management, supervision, and logistics, and sound technical strategies. The opportunity to address widespread disparities in health that exist in low- and low-middle-income countries that otherwise do not benefit from Global Alliance for Vaccines and Immunization's marketing approaches will be a huge challenge for the future. PAHO will continue to make the support of immunization to member countries a priority.
Abbreviations: PAHO, Pan American Health Organization
Citation: Andrus JK, Fitzsimmons J (2005) Introduction of new and underutilized vaccines: Sustaining access, disease control, and infrastructure development. PLoS Med 2(10): e286.
Jon Kim Andrus is Chief of Immunization Unit, Family and Community Health, Pan American Health Organization, Washington, D. C., United States of America. John Fitzsimmons is Senior Program Officer of Immunization Unit, Pan American Health Organization, Washington, D. C., United States of America.
*To whom correspondence should be addressed. E-mail: [email protected]
Competing Interests: The authors have declared that no competing interests exist.
==== Refs
References
Andrus JK Tambini G di Fabio JL Roses Periago M Anticipating new vaccines in the Americas Rev Panam Salud Publica 2004 16 369 370 15673478
Ropero AM Danovaro-Holiday MC Andrus JK Progress in vaccination against hepatitis B in the Americas J Clin Virol 2005 In press
Andrus JK Roses Periago M Elimination of rubella and congenital rubella syndrome in the Americas: Another opportunity to address inequities in health Rev Panam Salud Publica 2004 15 145 146 15096283
Castillo-Solorzano C Andrus JK Rubella elimination and improving health care for women Emerg Infect Dis 2004 10 2017 2021 15550217
Pan American Health Organization 3rd annual vaccination week in the Americas EPI Newsl 2004 27 8
Pan American Health Organization Health situation in the Americas: Basic indicators PAHO/AIS/04.0 2004 Washington (D. C.) Pan American Health Organization Available: http://www.paho.org/english/dd/ais/BI-brochure2004.pdf . Accessed 2 August 2005
Andrus JK de Quadros CA Kahn P Gust I Koff W Global access: Deployment, use, and acceptance Accelerating AIDS vaccine development: Challenges and opportunities 2005 Norfolk (United Kingdom) Horizon Scientific In press
|
16156693
|
PMC1201693
|
CC0
|
2021-01-05 10:38:16
|
no
|
PLoS Med. 2005 Oct 13; 2(10):e286
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020286
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1615669210.1371/journal.pmed.0020291Policy ForumHealth PolicyMental HealthPsychiatryMedicine in Developing CountriesInternational healthHealth PolicyAchieving the Millennium Development Goals: Does Mental Health Play a Role? Policy ForumMiranda J. Jaime Patel Vikram *J. Jaime Miranda is a Wellcome Trust Research Training Fellow, and Vikram Patel is a reader in International Mental Health, London School of Hygiene and Tropical Medicine, London, United Kingdom. Vikram Patel is author of Where There Is No Psychiatrist—A Mental Health Care Manual.
Competing Interests: The authors declare that no competing interests exist.
*To whom correspondence should be addressed: E-mail: [email protected] 2005 13 9 2005 2 10 e291Copyright: © 2005 Miranda and Patel.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.Miranda and Patel argue that mental disorders are among the most important causes of disability and premature mortality in developing countries.
==== Body
The Millennium Development Goals (MDGs) have captured the attention of the international health and development community in recent years [1–6], and in 2003 two world reports—the Human Development Report and the World Health Report—concentrated specifically on these goals [7,8]. The MDGs provide a vision for development in which health and education are squarely at the centre [1,9]. Three of the eight goals, eight of the 16 targets, and 18 of the 48 indicators relate directly to health [9,10]. Health is also an important contributor to several other goals.
Intriguingly, the health goals almost entirely ignore noncommunicable diseases, including mental disorders. Yet there is compelling evidence that in developing countries mental disorders are amongst the most important causes of sickness, disability, and, in certain age groups, premature mortality. Mental health–related conditions, including depressive and anxiety disorders, alcohol and drug abuse, and schizophrenia, contribute to a significant proportion of disability-adjusted life years (DALYs) and years lived with disability (YLDs), even in poor countries [11]. Apart from causing suffering, mental illness is closely associated with social determinants, notably poverty and gender disadvantage, and with poor physical health, including having HIV/AIDS and poor maternal and child health. Yet mental health remains a largely ignored issue in global health, and its complete absence from the MDGs reinforces the position that mental health has little role to play in major development-related health agendas.
This article seeks to question this assumption. Using evidence on mental health in developing countries, we argue that addressing mental health problems is an integral part of health system interventions aimed at achieving some of the key MDGs.
Ammanuel Psychiatric Hospital in Addis Ababa, Ethiopia
(Photo: © WHO/P.Virot)
Mental Health and the MDGs
Below, we consider the evidence linking mental health with three MDGs: eradicating poverty, reducing child mortality, and improving maternal health. However, the relevance of mental health is not limited to these goals alone. For example, a major reason why children are not able to either enrol in schools or complete primary education (MDG 2) is related to developmental and mental disorders, for example, learning disabilities [12]. There are several areas of confluence between HIV/AIDS (MDG 6) and mental health [13]—for example, people with HIV/AIDS are much more likely to suffer mental health problems, and these problems in turn can affect their overall health outcomes.
Goal 1: Eradicate extreme poverty and hunger
Stressful life experiences such as exposure to violence and poor physical health, which are well-recognized risk factors for mental disorders, are more likely to be experienced by poor people. Thus, it is not surprising that virtually all population-based studies of the risk factors for mental disorders, particularly depressive and anxiety disorders, consistently show that poor and marginalized people are at greater risk of suffering from these [14]. We also know that mental disorders impoverish people because of both increased costs of health care—often being sought through private providers—and lost employment opportunities. Most mental illnesses are relatively simple, and cheap, to treat, and evidence from clinical trials shows that efficacious treatment is associated with significant reductions in overall health-care costs [15]. Thus, treating mental disorders, particularly in the poor, who bear a disproportionate burden of suffering, would help people with mental disorders work more productively and reduce their health-care expenditures, facilitating the conditions necessary to rise out of poverty.
Goal 4: Reduce child mortality
A series of studies from South Asia have shown that early childhood failure to thrive, as indicated by undernutrition and stunting in babies under a year old, is independently associated with depression in mothers [16]. For example, a recent population-based cohort study from Pakistan has shown that babies of mothers who were depressed during pregnancy and in the postnatal period had a risk more than five times greater of being underweight and stunted at six months than babies of nondepressed mothers, even after adjustment for other known confounders such as maternal socioeconomic status [17]. Childhood failure to thrive is a major risk factor for child mortality; thus, it would be plausible to hypothesize that depression in mothers is also associated with increased child mortality. Indeed, evidence shows that depressed mothers are more likely to cease breast-feeding, and that their babies are significantly more likely to suffer diarrhoeal episodes or to not have their complete immunizations [17], all of these being recognized risk factors for childhood mortality. This study also showed that depression during pregnancy was strongly associated with low birth weight, an association that has been replicated in studies in India [18] and Brazil (S. Mitsuhiro, C. Ferri, V. Patel, M. Barros, E. Chalen, et al., unpublished data).
Goal 5: Improve maternal health
One of the most common health problems affecting mothers during pregnancy and after childbirth is depression. A large number of studies from most regions of the developing world show that 10%–30% of mothers will suffer from depression [19–21]. This condition is typically missed, not least because many of its core features such as fatigue and poor sleep are also commonly associated with motherhood itself. However, it is no trivial condition. Apart from its effect on the child, as described above, there is evidence that maternal depression can profoundly affect mothers themselves. Depressed mothers are much more disabled and less likely to care for their own needs. Suicide is a leading cause of maternal death in developed countries [22]. Suicide is now a leading cause of death in young women in the reproductive age group in the world's two most populous countries, India and China [23,24]. It is plausible that depression in mothers may also lead to increased maternal mortality, both through adversely affecting physical health needs as well as more directly through suicide.
Challenges to Acknowledging Mental Health in the MDGs
There are five major challenges to acknowledging mental health in the MDGs. The first, and perhaps the greatest, challenge lies in the very nature of the MDGs themselves. Although the MDGs have been portrayed as a consensus view of international development, it has been questioned whether it is worthwhile to have ambitious goals of this nature, given the patchy record of implementation of previous international declarations [1]. Examples of this patchy record include the failure of the international community to respect and fulfil the values expressed in the Universal Declaration of Human Rights [25], the failure to achieve the goal of the Declaration of Alma-Ata [26], and the failure to meet the international targets for sexual and reproductive health promoted during the last decade [27].
Second, national ownership of the goals is an important issue [1,8]. The power and purpose of the MDGs is that they are supposed to represent a means by which people can hold authorities accountable. There is a risk, however, that the MDGs are seen by some developing countries as being of primary concern to donors; they may be perceived as a new form of conditionality and as too restrictive in their scope to cover the multifaceted nature of health and development. In countries that have already achieved some of the MDGs, such as many countries in Latin America and Asia where targets for child mortality and maternal mortality are already met, the one-size-fits-all prescription suggested by the MDGs may not have local validity. Even though communicable diseases remain virtually the sole priority for global health policy, they do not constitute the major contributor to burden of disease in any region of the world apart from sub-Saharan Africa [28]. Even as mental health is now being prioritised as a major health problem in several developing countries, ironically their concerns do not find a place in global health targets and agendas.
Third, if mental health has a role to play towards meeting the MDG targets and health development goals, its role is likely to be more evident at the local health service level than at the level of international discourse. Nevertheless, the MDGs do not address strengthening of health systems [1,2,5,29,30]. This failure to address health systems raises important concerns because it risks diverting resources in under-resourced and overstretched services towards activities aimed at achieving specific targets. As a consequence, mental health and a host of other health problems, particularly those of a chronic and noncommunicable nature—which require a strong health system to deliver effective, multicomponent interventions—fall further by the wayside.
Fourth, as in any other plan with specific targets, the ultimate aim is to achieve a particular set of indicators that are expressed as national averages, but these averages may end up masking ongoing inequities. Significant progress in groups other than the poor can, for example, result in the achievement of the targets, with only minor improvements in the health status of the poor [1,31]. In terms of mental health, it is necessary to be conscious about this limitation because it is known that the least advantaged groups in society are the ones that carry the greater burden of mental illness. The stigma associated with mental illness that already serves to hide the suffering of countless millions is further compounded by being altogether ignored in the new programs focused on achieving the MDG targets.
Finally, the MDGs have been outlined with a specific number of objectives, targets, and indicators, which serve as standards for comparability purposes. Unfortunately, none of the targets or indicators devised for the MDGs [10] have a specific connection with mental health, nor do they enable development of monitoring methods that address mental health.
Implications for Global Policy
It is surprising that, while the developed world is investing substantial funds into mental health care and mental health promotion programs for its own populations, the leaders of the MDG project, international donors, and multilateral agencies, all of which are heavily represented by the developed world, have chosen to completely ignore mental health in the agenda for the health of the developing world. They have chosen this course of action, despite evidence of the burden of mental disorders, their association with the MDGs, and, perhaps most importantly, evidence that they can be effectively treated using locally available and affordable resources [32].
Are poor people in developing countries less deserving of mental health care? It is commonly argued that poor people in developing countries have more serious physical health problems to contend with and, therefore, the scarce resources that are available should be allocated to such problems. However, the evidence, some of which we have briefly outlined above, clearly shows that mental health has an integral role to play in achieving many of the MDGs. Can we provide effective health care for mothers or people living with HIV/AIDS, for example, without addressing their mental health needs?
Our prescription for global policy is to urge those involved with implementing and funding programs aimed at achieving the MDGs to take a broad and holistic approach to the targets. This approach would imply an explicit focus on strengthening basic health-care systems, for example, by strengthening the availability and skills of health workers, not only to deliver babies in hygienic circumstances but also to counsel mothers about stresses and provide effective psychological interventions. Another example of strengthening health-care systems is to ensure that while district health managers are sourcing antiretrovirals for people with HIV/AIDS, effective treatments for depression are also being made available for those who need them.
A number of mental health indicators can be developed and used to monitor the mental health of target populations, ranging from individual-level indicators—such as rates of depression measured using simple, short questionnaires—to population-level indicators—such as suicide rates and alcohol-related mortality. These prescriptions do not translate into the need for substantial additional resources. In many instances, it is only a broader orientation that is required. Where new resources are needed, they are likely to be cheap and cost-effective.
But perhaps most important of all is to advocate effectively to challenge the nihilism of global health planners, regarding the role of mental health. The acknowledgment of the importance of maternal depression to maternal and newborn health in this year's World Health Report is a welcome step in this direction [33]. Undoubtedly, stigma plays a key role in explaining the lack of acceptance of mental health as a legitimate health concern of people in developing countries. We must challenge this both through research evidence and through ensuring more opportunity for local voices from developing countries to acknowledge their needs and agendas. Above all, the global mental health advocacy discourse needs to reinforce its key message: there is no health without mental health [34].
JJM is supported by a Wellcome Trust Research Training Fellowship. VP is supported by a Wellcome Trust Career Development Fellowship in Tropical Medicine.
Citation: Miranda JJ, Patel V (2005) Achieving the Millennium Development Goals: Does mental health play a role? PLoS Med 2(10): e291.
Abbreviations
DALYdisability-adjusted life-year
MDGMillennium Development Goal
YLDyear lived with disability
==== Refs
References
Haines A Cassels A Can the millennium development goals be attained? BMJ 2004 329 394 397 15310610
Lee K Walt G Haines A The challenge to improve global health: Financing the Millennium Development Goals JAMA 2004 291 2636 2638 15173157
Sachs JD Health in the developing world: Achieving the Millennium Development Goals Bull World Health Organ 2004 82 947 949 15654410
Sachs JD McArthur JW The Millennium Project: A plan for meeting the Millennium Development Goals Lancet 2005 365 347 353 15664232
Task Force on Health Systems Research Informed choices for attaining the Millennium Development Goals: Towards an international cooperative agenda for health-systems research Lancet 2004 364 997 1003 15364193
Torres C Mujica OJ [Health, equity, and the Millennium Development Goals] Rev Panam Salud Publica 2004 15 430 439 15272989
United Nations Development Programme Human development report 2003—Millennium Development Goals: A compact among nations to end human poverty 2003 New York United Nations Development Programme Available: http://hdr.undp.org/reports/global/2003/ . Accessed 27 July 2005
World Health Organization The world health report 2003—Shaping the future 2003 Geneva World Health Organization Available: http://www.who.int/whr/2003/en/index.html . Accessed 27 July 2005
World Health Organization WHO and the Millennium Development Goals 2005 Geneva World Health Organization Available: http://www.who.int/mdg/en/ . Accessed 27 July 2005
United Nations Statistics Division Millennium indicators database 2005 New York United Nations Statistics Division Available: http://millenniumindicators.un.org/unsd/mi/mi_goals.asp . Accessed 27 July 2005
World Health Organization The world health report 2001—Mental health: New understanding, new hope 2001 Geneva World Health Organization Available: http://www.who.int/whr/2001/en/index.html . Accessed 27 July 2005
Patel V De Souza N School drop-out: A public health approach for India Natl Med J India 2000 13 316 318 11209489
Freeman M Patel V Collins PY Bertolote J Integrating mental health in global initiatives for HIV/AIDS Br J Psychiatry 2005 187 1 3 15994563
Patel V Kleinman A Poverty and common mental disorders in developing countries Bull World Health Organ 2003 81 609 615 14576893
Patel V Chisholm D Rabe-Hesketh S Dias-Saxena F Andrew G Efficacy and cost-effectiveness of drug and psychological treatments for common mental disorders in general health care in Goa, India: A randomised, controlled trial Lancet 2003 361 33 39 12517464
Patel V Rahman A Jacob KS Hughes M Effect of maternal mental health on infant growth in low income countries: New evidence from South Asia BMJ 2004 328 820 823 15070641
Rahman A Iqbal Z Bunn J Lovel H Harrington R Impact of maternal depression on infant nutritional status and illness: A cohort study Arch Gen Psychiatry 2004 61 946 952 15351773
Patel V Prince M Maternal psychological morbidity and low birth weight in developing countries Br J Psychiatry 2005 In press
Chandran M Tharyan P Muliyil J Abraham S Post-partum depression in a cohort of women from a rural area of Tamil Nadu, India. Incidence and risk factors Br J Psychiatry 2002 181 499 504 12456520
Cooper PJ Tomlinson M Swartz L Woolgar M Murray L Post-partum depression and the mother-infant relationship in a South African peri-urban settlement Br J Psychiatry 1999 175 554 558 10789353
Fisher JR Morrow MM Ngoc NT Anh LT Prevalence, nature, severity and correlates of postpartum depressive symptoms in Vietnam BJOG 2004 111 1353 1360 15663118
Oates M Suicide: The leading cause of maternal death Br J Psychiatry 2003 183 279 281 14519602
Aaron R Joseph A Abraham S Muliyil J George K Suicides in young people in rural southern India Lancet 2004 363 1117 1118 15064031
Phillips MR Li X Zhang Y Suicide rates in China, 1995–99 Lancet 2002 359 835 840 11897283
United Nations General Assembly Universal declaration of human rights 1948 New York United Nations Available: http://www.un.org/Overview/rights.html . Accessed 27 July 2005
International Conference on Primary Health Care Declaration of Alma-Ata 1978 Copenhagen World Health Organization Regional Office for Europe Available: http://www.euro.who.int/AboutWHO/Policy/20010827_1 . Accessed 27 July 2005
El Feki S The birth of reproductive health: a difficult delivery PLoS Med 2004 1 e9 10.1371/journal.pmed.0010009 15526057
Ollila E Global health priorities—Priorities of the wealthy? Global Health 2005 1 6 15847685
Travis P Bennett S Haines A Pang T Bhutta Z Overcoming health-systems constraints to achieve the Millennium Development Goals Lancet 2004 364 900 906 15351199
Wyss K An approach to classifying human resources constraints to attaining health-related Millennium Development Goals Hum Resour Health 2004 2 11 15238166
Gwatkin DR How much would poor people gain from faster progress towards the Millennium Development Goals for health? Lancet 2005 365 813 817 15733726
Patel V Araya R Bolton P Treating depression in the developing world Trop Med Int Health 2004 9 539 541 15117296
World Health Organization The world health report 2005: Make every mother and child count 2005 Geneva World Health Organization Available: http://www.who.int/whr/2005/en/ . Accessed 27 July 2005
Patel V No health without mental health 2005 Brighton id21 Insights Available: http://www.id21.org/insights/insights-h06/insights-issh06-art00.html . Accessed 27 July 2005
|
16156692
|
PMC1201694
|
CC BY
|
2021-01-05 10:39:14
|
no
|
PLoS Med. 2005 Oct 13; 2(10):e291
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020291
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1615669610.1371/journal.pmed.0020318Policy ForumInfectious DiseasesOtherEpidemiology/Public HealthHealth PolicyObstetrics/GynecologyPediatricsWomen's HealthInternational healthPublic HealthWomen's HealthMedicine in Developing CountriesInfectious DiseasesAn Immeasurable Crisis? A Criticism of the Millennium Development Goals and Why They Cannot Be Measured Policy ForumAttaran Amir Amir Attaran is Associate Professor and Canada Research Chair in Law, Population Health, and Global Development Policy, University of Ottawa, Ottawa, Ontario, Canada, and Associate Fellow, Chatham House, London, United Kingdom. E-mail: [email protected]
Competing Interests: AA has held small contracts or been paid per diem by the World Bank, United Nations Development Program, and the Roll Back Malaria Partnership in the last five years. None of these agencies was consulted in the development of this manuscript. Research funding was provided exclusively by the Canada Research Chairs program.
10 2005 13 9 2005 2 10 e318Copyright: © 2005 Amir Attaran.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.Attaran argues that five years into the Millenium Development Goals project, problems with measurement mean that often we cannot know if true progress towards these goals is occurring.
==== Body
In September 2000, 147 heads of state met at the United Nations (UN) headquarters—the largest such gathering ever—to resolve action on the most pressing problems of humanity and nature [1]. To underscore their commitment, they set numerical targets and deadlines to measure performance. These are the Millennium Development Goals (MDGs), and they span a large range of topics, including poverty, infectious disease, education, and gender equality (Box 1).
Box 1. The MDGs and Targets
By the year 2015, UN member states have pledged to meet eight goals; each goal subsumes one or more targets, as reproduced verbatim here (quoted from [40]). Details of the targets subsumed by goal eight and the various indicators for all the goals or targets can be found in [40,41].
Goal 1: Eradicate extreme poverty and hunger
• Reduce by half the proportion of people living on less than a dollar a day
• Reduce by half the proportion of people who suffer from hunger
Goal 2: Achieve universal primary education
• Ensure that all boys and girls complete a full course of primary schooling
Goal 3: Promote gender equality and empower women
• Eliminate gender disparity in primary and secondary education preferably by 2005, and at all levels by 2015
Goal 4: Reduce child mortality
• Reduce by two thirds the mortality rate among children under five
Goal 5: Improve maternal health
• Reduce by three quarters the maternal mortality ratio
Goal 6: Combat HIV/AIDS, malaria, and other diseases
• Halt and begin to reverse the spread of HIV/AIDS
• Halt and begin to reverse the incidence of malaria and other major diseases
Goal 7: Ensure environmental sustainability
• Integrate the principles of sustainable development into country policies and programmes; reverse loss of environmental resources
• Reduce by half the proportion of people without sustainable access to safe drinking water
• Achieve significant improvement in lives of at least 100 million slum dwellers, by 2020
Goal 8: Develop a global partnership for development
This September, the heads of state will gather again for the Millennium +5 Summit to assess the five-year progress of the MDGs. They will find that the MDGs have become all-important, not just within the UN, but also as the zeitgeist of the global development enterprise. As Professor Jeffrey Sachs, Director of the UN's Millennium Project, has declared, “To the extent that there are any international goals, they are the Millennium Development Goals”[2].
But is it wise to elevate the MDGs to the pedestal where they now sit? Could it be, despite an appearance of firm targets, deadlines, and focused urgency, that the MDGs are actually imprecise and possibly ineffective agents for development progress?
In this article, I argue that many of the most important MDGs, including those to reduce malaria, maternal mortality, or tuberculosis (TB), suffer from a worrying lack of scientifically valid data. While progress on each of these goals is portrayed in time-limited and measurable terms, often the subject matter is so immeasurable, or the measurements are so inadequate, that one cannot know the baseline condition before the MDGs, or know if the desired trend of improvement is actually occurring. Although UN scientists know about these troubles, the necessary corrective steps are being held up by political interference, including by the organisation's senior leadership, who have ordered delays to amendments that could repair the MDGs [3]. In short, five years into the MDG project, in too many cases, one cannot know if true progress towards these very important goals is occurring. Often, one has to guess.
The MDGs and Principles of Measurement
What makes the MDGs attractive is their concreteness. For example, the MDG to eradicate extreme poverty subsumes a “target” to “halve, between 1990 and 2015, the proportion of people whose income is less than $1 a day”, which in turn subsumes “indicators”, one of which is to measure income based on purchasing power.
Knowing that, worldwide, 28% of people in 1990 had purchasing power below $1 a day gives rise to a benchmark: that in 2015, fewer than 14% of people should be so destitute [4,5]. Currently, East Asia is on track; sub-Saharan Africa is not [6]. Such definitive statements about the benchmark or the trend are possible because non-stop effort goes into measuring incomes and prices—the UN, governments, and businesses all do it—so there are sufficient and reliable data.
It is harder to get sufficient and reliable data for the health MDGs. Even the most basic life indicators, such as births and deaths, are not directly registered in the poorest countries. Within this decade, only one African country (Mauritius) registers such events according to UN standards [7]. Without reliable vital registration systems to track even the existence of births or deaths, naturally the data for the medical circumstances of those births or deaths—or the lives in between—are unreliable.
Accordingly, most of the available data on the health MDGs come from methods of estimation, censuses, specialised household surveys, or all of these together.
There are many—too many—household surveys. In the public-health field, the best known are the Demographic and Health Survey (DHS) and the Multiple Indicator Cluster Survey (MICS), funded mainly by the United States and United Nations Children's Fund (UNICEF), respectively [8]. In addition to those household surveys, the Centers for Disease Control and Prevention, the World Health Organization (WHO), the United Nations Population Fund, the World Bank, and other organisations contribute surveys, making a rich alphabet soup—RHS, WHS, CWIQ, LSMS, PAPFAM, and so on. The proliferation is so excessive that there is now an International Household Survey Network, the rationale for which reads:
Donor's [sic] support is not always appropriately coordinated. There are many examples of duplicated or conflicting data collection activities. This lack of coordination does not only causes [sic] a huge waste of funds, it also put [sic] a high burden on national statistics offices. In the past few years, significant progress has been made to identify synergies among different survey programs or to develop common questionnaire modules, and to conduct joint data collection activities. But there is certainly room for much more cooperation. [9]
All of this is true, but even within the UN, different agencies jostle counterproductively for data. For example, in 2002, the WHO launched a new World Health Survey in over 70 countries to compete with the longer-running DHS and MICS [10]. Justified as a “sound basis for evaluating progress towards the millennium development goals”, instead the WHO's new survey tied up the few qualified statistical staff in the poorest countries [11]. Three years later (at the time of going to press), the new project has yet to publish a single dataset. (Ironically, the WHO has since created a new project called the Health Metrics Network, for “reducing overlap and duplication” caused by a “plethora of separate and often overlapping [data] systems” [12]. One cannot yet say whether the Health Metrics Network will succeed at this important goal, or add a further layer to the problem.)
Figure 1 shows the number of reported DHS and MICS surveys since 1990, which is the most common MDG baseline year. To generalise, most countries have had two or three such surveys, each gathering data on perhaps 5,000–10,000 households. Together with other surveys or national censuses, DHS and MICS are the backbone of measuring progress on the MDG health indicators.
Figure 1 Map of DHS and MICS Surveys
The map shows the number of DHS and MICS surveys by country, 1990–2005, according to completed reports made available to the public in June 2005. These reports are top-level summaries of the underlying micro-level survey data. Note, however, that UNICEF has not publicly disclosed micro-level data for 13 countries (Afghanistan, Algeria, Botswana, Cambodia, Cuba, Georgia, India, the Maldives, Somalia, Syria, Tunisia, Ukraine, and Federal Republic of Yugoslavia), making independent verification of those reports impossible (see http://www.childinfo.org; http://www.measuredhs.com/).
(Illustration: Bang Wong, www.clearscience.info)
Yet household surveys are serviceable but crude tools. Even with a simple question, such as about a child's birth weight, people's answers only roughly approximate the truth, as would be measured by weighing on a scale [13]. Other survey questions are so technical that no layperson can answer them accurately. MICS, for example, asks parents if their child's anti-malaria bed net was “ever treated with a product to kill mosquitoes”: an accurate answer depends on the type, dose, and date of insecticide treatment, and whether the local mosquito species carry insecticide resistance genes [14]. Because household surveys do not announce these or other sources of error, one can easily have false confidence in them. For example, many MICS survey reports present their findings as single-point estimates, without any of the usual qualifiers of data inaccuracy or quality, such as statistical confidence intervals or significance tests (see India's report for example; [15]).
In short, there are many sources of data on the MDGs. When those sources suffice to reveal statistically significant trends in the MDGs, then all is well, and it is possible to make conclusive statements: that the MDGs are being met, or that the MDGs are being missed. But, as the case studies below illustrate, such certainty is highly elusive.
Malaria
MDG 6, Target 8, pledges to “have halted by 2015 and begun to reverse the incidence of malaria”. The malaria MDG overlaps with a somewhat earlier (1998) WHO-led goal known as Roll Back Malaria (RBM), which aims “to halve malaria-associated mortality by 2010 and again by 2015” [16]. Even though the MDG and the RBM goal are only quasi-consistent with one another, the UN allows them to coexist, and UN communications often mention both [16]. Accordingly, both are discussed here.
Yet with double attention on malaria, and the head start afforded by RBM, the UN still is unable to make an official pronouncement on the progress of its malaria goals. The WHO and UNICEF write that it is “too soon to determine whether the global burden of malaria”, meaning both incidence and mortality, “has increased or decreased since 2000” [16].
Too soon? RBM is in its seventh year, and past the halfway mark of its 2010 deadline. The only two possible reasons not to know if malaria has increased or decreased are that the UN either (i) did not encourage timely measurements or (ii) chose indicators—malaria incidence and mortality—that are essentially immeasurable.
Actually, both are true. What follows is a cautionary history.
In 2002, the British government commissioned an independent evaluation of the UN's malaria efforts. It did so because it was the largest financier of RBM, and because of a perception that there was insufficient alignment between the efforts of the UN agencies and malarious countries. On the subject of measuring progress, the evaluators wrote:
The main problem affecting…data collection efforts…has been that an overly complex and insufficiently prescriptive approach has been taken. There has been a failure to clearly define goals and priorities of the [measurement] strategy at the global and regional levels....Too many indicators are proposed. Too many sources of data are suggested. Insufficient guidance is given to countries on data collection and methodology….Some countries are measuring one thing, some countries are measuring another….In some cases, data are being collected without any systematic and scientific sampling methodology, and so are essentially meaningless and impossible to interpret. [17]
This unsparing criticism points to two problems, which although they pertain to RBM, often apply with equal force to the malaria and other MDGs. The first problem concerns the lack of a baseline: it is impossible to retrospectively measure worldwide (or regional, or national) malaria incidence and mortality existing at the inception of the RBM goal or the MDG, when the data from that era are universally acknowledged to be poor [18]. Without knowing the original condition, it is futile to stipulate either “to halve” malaria mortality by 2010 or “to halt” malaria incidence by 2015. Such words have no meaning where the baseline is mysterious.
The second problem concerns the unsuitability of the indicators: both malaria incidence and mortality are so crudely measured by household surveys and most countries' health records that, essentially, they are immeasurable. The UN's malaria monitoring group agrees, writing that “malaria-specific mortality should not be monitored routinely, as this can not be measured easily in malaria-endemic Africa” [19]. Yet the UN often ignores such warnings, even when they are timely, explicit, and the opinions of its own scientists. It was only two months after WHO scientists wrote that “it will not, in general, be possible to measure the overall incidence rate of malaria” that the UN chose the incidence rate as the mainstay of the malaria MDG [20].
The legacy of unfortunate decisions now leaves malaria risk mapping as the only feasible way to estimate (not measure) malaria incidence and mortality. The principle is to superimpose a map of a population onto a map of malaria intensity, although, in practice, the limitations include malaria maps from the 1960s and too few demographic surveillance sites to accurately measure and calibrate incidence and mortality risks [21,22]. The WHO has been slow to use risk mapping, probably because it fears public criticism when, inevitably, the current estimates of malaria severity must be revised upward [23,24].
Accordingly, years after the withering external evaluation, the UN neither has achieved convincing measurement or estimation of malaria incidence and mortality, nor has it abandoned those as the key indicators of progress. Both the RBM goal and the malaria MDG are today immeasurable.
Maternal Mortality
MDG 5, Target 6, pledges to “reduce by three quarters, between 1990 and 2015, the maternal mortality ratio” [1]. As such, this MDG target echoes a 1994 UN goal set at the Cairo Conference on Population and Development to halve maternal mortality by 2000, and again by 2015 [25].
The UN Millennium Project reports that at about 530,000 deaths annually, “overall levels of maternal mortality are believed to have remained unchanged” in the last 15 years [26]. Both the number of such deaths and the number of births are used to calculate the maternal mortality ratio (MMR; the number of women dying through complications of pregnancy and delivery per 100,000 live births). However, it is exactly in the poorest countries where the maternal mortality problem is severest that the data about deaths and births are least satisfactory. Vital registration would help, but few developing countries, accounting for 24% of the world's live births, have complete data [7]. Directly measuring MMR in the whole population is not today an option.
Therefore MMR must be estimated. The current method is crude, and uses regression modelling based on partial vital registration, censuses, household surveys, and other inputs [27]. The outputs are a point estimate for MMR in each geographic region, surrounded by an educated guess (not the same as a valid statistical confidence interval) of the lower and upper range in which the point estimate could lie.
Accordingly, the most recent (2000) published estimate for MMR worldwide is 400 maternal deaths per 100,000 births, within an unscientific, best-guess range of perhaps 210 (low) to 620 (high) [28]. Estimates for the MDG baseline year (1990) are similarly vague [29].
Without a statistically robust estimate for MMR in the baseline year, or in later years, nobody knows whether worldwide MMR has increased or decreased since 1990, other than in a “handful of countries” [26]. The limitations of current estimation techniques are so profound that UNICEF and WHO scientists warn that “it would be inappropriate to compare the 2000 estimates with those for 1990…and draw conclusions about trends” [28].
Thus, 11 years after the Cairo Conference first set an explicit target to reduce MMR by 75%, the UN neither has achieved measurement of MMR, nor has it heeded the warnings of its own scientists that MMR is basically immeasurable. The MDG carries that mistaken goal forward to 2015, and the impossibility of measuring and demonstrating success is certainly preordained.
Tuberculosis
MDG 6, Target 8, pledges to “have halted by 2015 and begun to reverse the incidence of…major diseases”, which the UN has interpreted to include TB [1]. The provenance of the TB MDG is it neither reiterates an earlier (1991) goal, nor is it obviously a purposeful improvement [30].
As with malaria, measuring TB incidence is notoriously difficult. It requires counting the annual number of new patients with TB disease (i.e., not just new TB infections). Currently, no country measures TB incidence regularly, as the MDG target stipulates [31].
Fortunately, the MDG indicators provide for some simpler alternatives: TB disease prevalence and deaths (Indicator 23), and the proportion of TB disease cases detected and cured using a WHO-recommended treatment called “directly observed therapy—short course” (DOTS; Indicator 24). The TB prevalence and case detection indicators are directly measurable, but, ironically, the WHO does not actually measure them. Instead, it uses a unique, arguably outdated estimation method.
Nobody can say with scientific confidence what the actual trends for TB are.
In the WHO's method, the only true measurement is the number of new, sputum-positive TB cases that are detected and notified to the authorities for treatment with DOTS. To estimate the case detection rate, the WHO divides that number of notified TB cases (the numerator) by an estimate of at-large case incidence (the denominator) [32]. Further, the WHO obtains case incidence from “an independent estimate of the case detection rate” [33]. In effect, the WHO's two estimates are circular and lack definite meaning, for each estimate draws upon the other estimate. Further, the WHO bases this estimation process on inputs that are not always rigorous, and the inputted data are often obtained from collective opinion rather than measurement [33].
Accordingly, it is impossible to state the actual trends in TB disease with any degree of statistical confidence. The WHO's best guess is that its estimates “typically range from −20% to [+]40%” in accuracy [32].
Others have criticised the circular estimation technique. The WHO's former director for evidence argues that “essentially no empirical basis exists to assess the trend in case detection in regions where tuberculosis is most prevalent, including sub-Saharan Africa” [34]. He calls the WHO's trend estimates “serial guessing” [34]. Certainly, the WHO's leading assumption (known as the “Styblo rule” [35]) has infrequently been tested in Africa, where TB is accelerated by an unparalleled HIV/AIDS epidemic. The WHO's own scientists concede that it may no longer apply there [32].
Nevertheless, the WHO maintains that where access to DOTS treatment is extensive—that is, not in Africa—its estimated case detection rates are an adequate guide to true TB trends. This is debatable: in China, which is the WHO's finest DOTS success, actual measurements (not estimations) of TB prevalence corroborated the WHO's case detections less well than expected [36].
The best solution now proposed in the scientific literature would redefine the case detection rate, based on measuring true TB prevalence by widespread radiographic or microscopic surveys [31]. Although similar prevalence measurements have been the cornerstone of East Asia's successful attack on TB, the WHO resists changing from estimation to true measurement [37]. As a result, nobody can say with scientific confidence what the actual trends for TB are or whether the TB MDG is on track.
Child Mortality
The above case studies could leave the dismal impression that all time-limited development goals are immeasurable, lack baseline data, and imply trends having no scientific meaning. Not quite. There is a happy exception: MDG 4, Target 5, which reads to “reduce by two thirds, between 1990 and 2015, the under-five [child] mortality rate” [1].
The under-five child mortality (U5M) rate is an excellent MDG indicator because it is easily measured. For most parents the birth or death of a child is highly memorable; ask properly about these events in a household survey and their recollection is likely to be accurate. If the survey asks enough parents in a population, and continues to ask at regular intervals, a statistically significant trend emerges with time—the very point of the MDGs.
The best proof of this concept comes from Africa. Using data from sequential DHS cycles, in Ghana during 1988–1998, the U5M rate improved 30% [38]. Conversely, in Zimbabwe during 1988–1999, the U5M rate deteriorated 44% [38]. Unlike other MDGs where such changes are, to put it bluntly, only guessed at, these trends in the U5M rate are properly measured and, importantly, are scientifically meaningful, with confidence intervals that reveal the accuracy and quality of the underlying data. Just by keeping the current DHS technique, and interviewing about 7,000 women per country every five years, it is possible to reliably detect either a 15% gain or loss in the U5M rate with scientific confidence.
There is an invaluable and gratifying lesson to draw from the U5M case study: if the UN sets an MDG target that is practical to measure (most are not), and the measurement technique for that MDG target is suitable (most are not), and measurements are taken at the baseline year and in subsequent years (they rarely are), it is then possible to measure the state of the world's health reliably and accurately, and with excellent scientific confidence regarding the trend. In short, it becomes possible to know, not just to guess, if the MDGs are on track or not—even in Africa.
Discussion
I did not write this paper to doubt the moral necessity of investing more money and political capital in global development; that is unarguable, and it would be reprehensible to use these arguments to seed those doubts.
Instead, I hope to open an important debate, unable to be fully answered by this paper, on a hitherto almost unexplored question: is the world better off with or without the MDGs and similar UN-sponsored, time-limited, quantitative development goals? The answer to that question must be sought without pro-UN or anti-UN ideology, but with awareness that there are two prongs to consider: (i) whether such goals are interpreted so as to advance the dignity and well-being of the large number of people who live in extreme poverty , and (ii) whether such goals advance the reputation of the UN and the global development establishment. I believe the MDGs risk trouble on both fronts.
Viewed objectively, it must be agreed that the MDGs palter. The health goals for 2015 sound quantitative, but for most of them, their quantification is irretrievably flawed. The trends that the health goals allude to are either immeasurable or were not measured properly from the 1990 baseline year onward. This is not an extraordinarily controversial conclusion: recall that in each of the cautionary examples discussed—malaria, maternal mortality, and TB—the UN's own current or former staff have said that the trends are immeasurable or lack baseline data.
Short of abandoning the MDGs, the better option is to amend the goals, targets, or indicators—all three levels of the hierarchy—to be feasibly measurable.
Unfortunately, the UN leadership has, to date, delayed this option. In a September 2004 memo, one year ahead of the Millennium +5 Summit, the UN's Deputy Secretary General instructed the organisation's experts in charge of the MDG statistics with the following:
The [Millennium +5 Summit]…should not be distracted by arguments over the measurement of the MDGs—or worse, over different numbers being used by different agencies for the same indicator…. [P]roposals for modifications of definitions or new indicators will only be considered formally after the [Millennium +5 Summit]… as any changes at this stage would only distract from the result that we would like to achieve. [3]
The Deputy Secretary General's order interferes with and shows a profound disrespect for the scientific process—a process that fundamentally is not “distracted by arguments” nor disturbed by “different numbers”. On the contrary, intellectual arguments between scientists are essential for devising new methods of measurements for the MDGs, so that they in turn yield more accurate numbers about the extent and causes of extreme poverty.
By suppressing proposals to amend the MDGs ahead of the Millennium +5 Summit, the UN leadership discarded the only timely opportunity to win high-level political support for truly measurable, scientifically meaningful goals. While the Deputy Secretary General plans “a process that will consider recommendations regarding refinements” to the MDGs, the process will commence only after this September's summit [3]. As a result, any recommendations to amend the MDGs that may arise must await ratification at the next heads-of-state summit—presumably, the Millennium +10 Summit in 2010 (to date, summits occur every five years). In that case, there would remain only five years to the MDGs' final reckoning in 2015. Such extreme delay is illogical and sabotages the MDGs' chances of success.
Some may disagree with my emphasis on measurement and timelines. One anonymous peer reviewer of this paper wrote that while measuring the MDGs is “of concern for epidemiologists and others”, my interpretation “misses the point” because the purpose of the MDGs is merely to be exhortatory. “The MDGs are not a measuring exercise”, wrote the reviewer, but instead are a “common vision of what matters most for improving the lives of people in poor countries”.
This sort of thinking, although widespread among development professionals, is neglectful towards people living in extreme poverty. Neglect occurs when one touts the MDGs for the “common vision” of, say, reducing maternal mortality, while being indolent about measurements to prove mortality is genuinely decreasing. That formulation values consensus about helping pregnant women, ahead of certainty about helping pregnant women—an outcome that, if they knew about it, the women could easily find ideological and dehumanising.
Further, the notion that the MDGs are merely exhortatory discriminates against the world's poorest people. Imagine if European or American leaders, taking aim at poverty in their own countries, set quantitative goals to reduce unemployment or teen pregnancy—only to declare the unemployment and teen pregnancy rates were “not a measuring exercise”. Most people would abhor the dishonesty, for obvious reasons.
But if it is shameful, as I believe, to interpret the MDGs as merely exhortatory, imparting no standards of performance, the converse error also exists: to interpret the MDGs as all-encompassing and imparting too many standards of performance.
The latest fashion, exemplified by the UN Millennium Project, is to treat the MDGs as catch-alls or tautologies for development itself. In a list entitled “Interventions by MDG Target”, the UN Millennium Project recommends to build “roads” or “transport infrastructure” for all of the following MDG targets: primary education, hunger, gender equality, water and sanitation, child mortality, and, of course, malaria, maternal mortality, and TB [39]. Electricity, slum upgrading, and education are similar panaceas.
Definitely roads or electricity matter to holistic development, but justifying those under the cover of goals expressly for child mortality or malaria, makes goal-setting seem pointless. Worse, such justification sounds dishonest—a camouflage job. It is no wonder that with the MDGs subordinated into empty vessels for tenuously related interventions—subordinated into, as Professor Jeffrey Sachs says, just “any international goals”—there is resistance to measure the progress of the specific goals, targets, and indicators with rigor and precision [2].
The MDGs could turn from opportunity to liability.
I believe that without thoroughgoing action to change the current scenario (see Box 2), the MDGs could turn from opportunity to liability. As 2015 nears, the UN becomes increasingly vulnerable to criticism if it still lacks data to prove whether the MDGs are or are not being met. A stream of embarrassing disclosures, similar to the external evaluation of RBM, will likely ensue. Certainly journalists will report the embarrassments, and opponents of foreign aid may use them to discredit further generosity to poor countries. These unhappy events are entirely foreseeable, and for that reason, must give pause to anyone who naively believes that measuring the MDGs is an occupation only scientists need care about. Anyone wishing to preserve the credibility of the UN and the global development enterprise ten years from now also must care.
Box 2. Five Recommendations to Make the MDGs Truly Time-Limited and Quantitative
• Convene an external (non-UN) scientific peer review to examine the goals, targets, and indicators to ascertain whether the desired trend of improvement in each is, with current data, measurable or estimable at scientifically accepted levels of accuracy and statistical significance.
• For those goals, targets, or indicators measurable by household surveys, choose only a single survey instrument; determine the minimum sample size needed to detect favourable or adverse trends with statistical significance; conduct the survey at regular intervals; and make all the micro-level data fully public, so independent scientists can replicate the UN's conclusions. Eliminate the many superfluous household surveys now in use.
• For those goals, targets, or indicators not measurable by household surveys, institute sample surveys (“mini censuses”) by creating a large number of new demographic surveillance sites in various countries. The Canadian-funded Tanzania Essential Health Interventions Project is a superb example (see [18]; http://video.idrc.ca/tehip/tehip_dss_e_1000.asx; http://www.economist.com/displaystory.cfm?story_id=1280587).
• For those goals, targets, or indicators that are not measurable by any practical means, first consider to amend them, and if that is not possible, abandon them (bearing in mind that any feasible amendment to the goals, targets, or indicators can only modestly deviate from the political consensus that underpins the MDGs now).
• Within 18 months, hold a high-level UN-sponsored event at which governments ratify final actions for all the above. Have those actions be developed by external scientists and given to the Deputy Secretary General directly.
More thoughtful and timely action for the sake of these institutions, and, needless to say, the millions of people who shall live—or die—with the success or failure of the MDGs, is only wise.
I wish to thank Prof. Martien Borgdorff and Prof. Bob Snow for discussions on TB and malaria epidemiology, respectively. Thanks also to the peer reviewers (Prof. Tom Novotny, Prof. Ron Waldman, and two anonymous persons) for helpful comments.
Citation: Attaran A (2005) An immeasurable crisis? A criticism of the Millennium Development Goals and why they cannot be measured. PLoS Med 2(10): e318.
Abbreviations
DOTSdirectly observed therapy—short course
DHSDemographic and Health Survey
MDGMillennium Development Goal
MICSMultiple Indicator Cluster Survey
MMRmaternal mortality ratio
RBMRoll Back Malaria
TBtuberculosis
U5Munder-five child mortality
UNUnited Nations
UNICEFUnited Nations Children's Fund
WHOWorld Health Organization
==== Refs
References
United Nations General Assembly Road map towards the implementation of the United Nations Millennium Declaration Report of the Secretary-General. Report number A/56/326 2001 September Available: http://www.un.org/documents/ga/docs/56/a56326.pdf . Accessed 1 August 2005
Eviatar D Spend $150 billion per year to cure world poverty New York Times 2004 November 7 44 Sect 6
Deputy Secretary General United Nations Message to the inter-agency and expert meeting on MDG indicators Geneva 29 September–1 October 2004 2004 New York United Nations Available: http://unstats.un.org/unsd/mi/techgroup/Sept2004/message_to_inter_agency_mdg.pdf . Accessed 1 August 2005
World Bank MDG 1: Eradicate extreme poverty and hunger 2004 Washington D. C World Bank Available: http://ddp-ext.worldbank.org/ext/MDG/gdmis.do . Accessed 1 August 2005
Reddy SG Pogge TW How notto count the poor 2003 New York Columbia University Available: http://www.columbia.edu/~sr793/count.pdf . Accessed 1 August 2005
World Bank Global monitoring report 2005. Millennium Development Goals: From consensus to momentum 2005 Washington D. C World Bank Available: http://siteresources.worldbank.org/GLOBALMONITORINGEXT/Resources/complete.pdf . Accessed 1 August 2005
AbouZahr C Wardlaw T Maternal mortality at the end of a decade: Signs of progress? Bull World Health Organ 2001 79 561 568 11436479
David P Haberlen S 10 best resources for measuring population health Health Policy Plan 2005 20 260 263 15965038
International Household Survey Network Rationale 2005 Washington D. C International Household Survey Network Available: http://www.internationalsurveynetwork.org/home/index.php?option=content&task=view&id=3&Itemid=26 . Accessed 1 August 2005
Diallo K Zurn P Gupta N Dal Poz M Monitoring and evaluation of human resources for health: An international perspective Hum Resour Health 2003 1 3 12904252
World Health Organization Statement by the Director-General to the executive board at its 109th session Report number EB109/2 2002 January Geneva World Health Organization Available: http://www.who.int/gb/ebwha/pdf_files/EB109/eeb1092.pdf . Accessed 10 August 2005
World Health Organization Health Metrics Network: What it is what it will do and how countries can benefit 2005 Geneva World Health Organization Available: http://www.who.int/entity/healthmetrics/about/concept_paper.doc . Accessed 1 August 2005
Robles A Goldman N Can accurate data on birthweight be obtained from health interview surveys? Int J Epidemiol 1999 28 925 931 10597993
United Nations Chilren's Fund Multiple indicator cluster survey 2 questionnaire: Children under-5 questionnaire 2000 New York United Nations Chilren's Fund Available: http://www.childinfo.org/MICS2/finques/M2finQ.htm . Accessed 1 August 2005
Government of India United Nations Children's Fund Multiple indicator survey—2000 India summary report. New York: United Nations Children's Fund 2001 November Available: http://www.childinfo.org/MICS2/newreports/india/india.pdf . Accessed 2 August 2005
World Health Organization United Nations Children's Fund World malaria report 2005. Geneva: World Health Organization 2005 Available: http://rbm.who.int/wmr2005/pdf/WMReport_lr.pdf . Accessed 1 August 2005
World Health Organization Final report of the external evaluation of Roll Back Malaria: Achieving impact: Roll Back Malaria in the next phase 2002 Geneva World Health Organization Available: http://www.rbm.who.int/cmc_upload/0/000/015/905/ee_toc.htm . Accessed 1 August 2005
De Savigny D Binka F Monitoring future impact on malaria burden in sub-Saharan Africa Am J Trop Med Hyg 2004 71 224 231 15331841
World Health Organization Second meeting of the RBM partnership Monitoring and Evaluation Reference Group 2003 November Geneva World Health Organization Available: http://rbm.who.int/partnership/wg/wg_monitoring/docs/MERG_Nov03minutes270104final.doc . Accessed 1 August 2005
Watt C Dye C Indicators to measure the impact of malaria control Report number WHO/CDS/RBM/2000.22 2000 July Geneva World Health Organization Available: http://whqlibdoc.who.int/hq/2000/WHO_CDS_RBM_2000.22.pdf . Accessed 1 August 2005
Korenromp EL Williams BG Gouws E Dye C Snow W Measurement of trends in childhood malaria mortality in Africa: An assessment of progress toward targets based on verbal autopsy Lancet Infect Dis 2003 3 349 358 12781507
Snow RW Guerra CA Noor AM Myint HY Hay SI The global distribution of clinical episodes of Plasmodium falciparum malaria Nature 2005 434 214 217 15759000
BBC News Global toll of malaria ‘doubled' 2005 March 10 London BBC News Available: http://newswww.bbc.net.uk/1/hi/health/4332921.stm . Accessed 1 August 2005
[Anonymous] Reversing the failures of Roll Back Malaria Lancet 2005 265 1439
United Nations Population Fund Part 1: Programme of action of the United Nations International Conference on Population and Development 1994 New York United Nations Population Fund Available: http://www.unfpa.org/icpd/icpd_poa.htm . Accessed 2 August 2005
UN Millennium Project Who's got the power? Transforming health systems for women and children 2005 London Earthscan 224
Hill K AbouZahr C Wardlaw T Estimates of maternal mortality for 1995 Bull World Health Organ 2001 79 182 193 11285661
World Health Organization United Nations Children's Fund United Nations Population Fund Maternal mortality in 2000: Estimates developed by WHO UNICEF and UNFPA 2003 Geneva World Health Organization Available: http://www.who.int/reproductive-health/publications/maternal_mortality_2000/mme.pdf . Accessed 10 August 2005
World Health Organization United Nations Children's Fund Revised 1990 estimates of maternal mortality: A new approach by WHO and UNICEF 1996 April Geneva World Health Organization Available: http://whqlibdoc.who.int/hq/1996/WHO_FRH_MSM_96.11.pdf . Accessed 10 August 2005
World Health Organization Forty-fourth world health assembly Resolution WHA 44.8 1991 May Geneva World Health Organization Available: http://policy.who.int/cgi-bin/om_isapi.dll?hitsperheading=on&infobase=wha&record=2496&softpage=Document42 . Accessed 1 August 2005
Borgdorff MW New measurable indicator for tuberculosis case detection Emerg Infect Dis 2004 10 1523 1528 15498151
Dye C Watt CJ Bleed DM Hosseini SM Raviglione MC Evolution of tuberculosis control and prospects for reducing tuberculosis incidence prevalence and deaths globally JAMA 2005 293 2767 2775 15941807
Dye C Scheele S Dolin P Pathania V Raviglione MC Global burden of tuberculosis: Estimated incidence prevalence and mortality by country JAMA 1999 282 677 686 10517722
Murray CJ Lopez AD Wibulpolprasert S Monitoring global health: Time for new solutions BMJ 2004 329 1096 1100 15528624
Styblo K The relationship between the risk of tuberculous infection and the risk of developing infectious tuberculosis Bull Int Union Tuberc Lung Dis 1985 60 117 119
China Tuberculosis Control Collaboration The effect of tuberculosis control in China Lancet 2004 364 417 422 15288739
Borgdorff MW Nagelkerke NJD Dye C Nunn P Gender and tuberculosis: Comparison of prevalence surveys with notification data to explore sex differences in case detection Int J Tuberc Lung Dis 2000 4 123 132 10694090
Korenromp EL Arnold F Williams BG Nahlen BL Snow RW Monitoring trends in under-5 mortality rates through national birth history surveys Int J Epidemiol 2004 33 1293 1301 15319411
United Nations Millennium Project Interventions by MDG target 2005 New York United Nations Available: http://www.unmillenniumproject.org/documents/Interventions%20by%20MDG%20Target.doc . Accessed 1 August 2005
United Nations UN millennium development goals 2005 New York United Nations Available: http://www.un.org/millenniumgoals . Accessed 12 August 2005
United Nations Statistical Division Millennium development goal indicators database [database] 2005 New York United Nations Statistical Division Available: http://unstats.un.org/unsd/mi/mi_goals.asp . Accessed 12 August 2005
|
16156696
|
PMC1201695
|
CC BY
|
2021-01-05 10:38:15
|
no
|
PLoS Med. 2005 Oct 13; 2(10):e318
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020318
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 10.1371/journal.pmed.0020329SynopsisGenetics/Genomics/Gene TherapyMental HealthPathologyPsychiatrySchizophrenia and Other Psychotic DisordersGeneticsPsychiatryPathologyCAPON and Schizophrenia—Does Size Matter? 10 2005 13 9 2005 2 10 e329Copyright: © 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.
Increased Expression in Dorsolateral Prefrontal Cortex of CAPON in Schizophrenia and Bipolar Disorder
==== Body
Schizophrenia and bipolar disease are complex diseases, with multiple genes and environmental factors thought to be responsible for their manifestation. Many reports have implicated changes in certain regions of the human genome in schizophrenia. An area on Chromosome 1 has been associated with the disease in different studies and populations. Linda Brzustowicz and colleagues had previously described association of several single nucleotide polymorphisms (SNPs) within a gene called CAPON (for carboxyl-terminal PDZ ligand of neuronal nitric oxide synthase) with schizophrenia in a set of Canadian families. A separate study in a Chinese population found an association between schizophrenia and a separate group of SNPs within CAPON. CAPON is an attractive candidate for a “schizophrenia gene”: CAPON was first identified as a protein binding to neuronal nitric oxide synthase (nNOS), and indirect evidence suggests that it might be linked to the regulation of glutamate neurotransmission. However, so far, no coding sequence mutations in CAPON have been found in patients with schizophrenia.
Brzustowicz and colleagues now report results from a study of CAPON expression in postmortem brain samples from patients with schizophrenia, from patients with bipolar disorder, and from control individuals without psychiatric illness. Initially screening a human fetal brain cDNA library for potential alternative splice forms of CAPON, they found, in addition to the predicted full-length transcript, a shorter isoform that consists of the last two exons of the gene. They also confirmed that both long and short versions of the protein are present in human brain. (The short isoform would still be able to bind nNOS and possibly disrupt its interaction with other proteins.)
The authors then examined CAPON mRNA expression in postmortem brain samples from 35 patients with schizophrenia, from 35 patients with bipolar disorder, and from 35 controls. They looked for transcripts encoding the long and short forms and also compared expression levels (relative to beta-actin) across the diagnostic groups. In the dorsolateral prefrontal cortex (thought to be involved in schizophrenia and the only area for which samples were available), expression levels of the long isoform did not differ between patient and control samples. However, the short isoform was expressed at higher levels in the patients with mental illness than in the controls. They also analyzed DNA from these individuals at three SNPs associated with schizophrenia and found, for each SNP, that the group of individuals who carried one or two copies of the schizophrenia-associated allele had overall higher levels of the short form transcript than those homozygous for the non-associated CAPON allele. They saw no differences in levels of the full-length transcript between the groups with different SNP genotypes.
These are intriguing but preliminary results. Getting reliable results from studies on postmortem samples is extremely difficult because of numerous confounding factors. Brzustowicz and colleagues tried to control for some of them (such as sex of the individual, substance abuse by the individual, and storage time of the sample), but others are impossible to determine or control for. Moreover, this particular collection of samples had only tissue from the dorsolateral prefrontal cortex available for study. Sample preparation was geared toward high-quality RNA for expression studies and not suitable for parallel protein analysis. Future studies, with at least some of them in animal models that allow controlled conditions and experimentation, will need to determine the functions of the long and short isoforms of CAPON and their interaction with other proteins involved in postsynaptic neurotransmission (some of which have also been linked to schizophrenia), and elucidate a possible role for CAPON in psychiatric disorders.
|
0
|
PMC1201696
|
CC BY
|
2021-01-05 10:38:15
|
no
|
PLoS Med. 2005 Oct 13; 2(10):e329
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020329
|
oa_comm
|
==== Front
Ann Clin Microbiol AntimicrobAnnals of Clinical Microbiology and Antimicrobials1476-0711BioMed Central London 1476-0711-4-121614304410.1186/1476-0711-4-12ResearchIn vitro activities of 11 fluoroquinolones against 816 non-typhoidal strains of Salmonella enterica isolated from Finnish patients with special reference to reduced ciprofloxacin susceptibility Kotilainen Pirkko [email protected]änen Susa [email protected] Anja [email protected] Pentti [email protected] Antti J [email protected] Antimicrobial Research Laboratory, Department of Bacterial and Inflammatory Diseases, National Public Health Institute, Turku, Finland2 Department of Medicine, Turku University Hospital and Turku University, Turku, Finland3 Enteric Bacteria Laboratory, Department of Bacterial and Inflammatory Diseases, National Public Health Institute, Helsinki, Finland2005 5 9 2005 4 12 12 7 7 2005 5 9 2005 Copyright © 2005 Kotilainen et al; licensee BioMed Central Ltd.2005Kotilainen 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 number of Salmonella strains with reduced susceptibility to fluoroquinolones has increased during recent years in many countries, threatening the value of this antimicrobial group in the treatment of severe salmonella infections.
Methods
We analyzed the in vitro activities of ciprofloxacin and 10 additional fluoroquinolones against 816 Salmonella strains collected from Finnish patients between 1995 and 2003. Special attention was focused on the efficacy of newer fluoroquinolones against the Salmonella strains with reduced ciprofloxacin susceptibility.
Results
The isolates represented 119 different serotypes. Of all 816 Salmonella strains, 3 (0.4%) were resistant to ciprofloxacin (MIC ≥ 4 μg/ml), 232 (28.4%) showed reduced susceptibility to ciprofloxacin (MIC ≥ 0.125 – 2 μg/ml), and 581 (71.2%) were ciprofloxacin-susceptible. The MIC50 and MIC90 values of ciprofloxacin for these strains were 0.032 and 0.25 μg/ml, respectively, being lower than those of the other fluoroquinolone compounds presently on market in Finland (ofloxacin, norfloxacin, levofloxacin, and moxifloxacin). For two newer quinolones, clinafloxacin and sitafloxacin, the MIC50 and MIC90 values were lowest, both 0.016 and 0.064 μg/ml, respectively. Moreover, clinafloxacin and sitafloxacin exhibited the lowest MIC50 and MIC90 values, 0.064 and 0.125 μg/ml, against the 235 Salmonella strains with reduced susceptibility and strains fully resistant to ciprofloxacin.
Conclusion
Among the registered fluoroquinolones in Finland, ciprofloxacin still appears to be the most effective drug for the treatment salmonella infections. Among the newer preparations, both clinafloxacin and sitafloxacin are promising based on in vitro studies, especially for strains showing reduced ciprofloxacin susceptibility. Their efficacy, however, has not been demonstrated in clinical investigations.
==== Body
Background
Fluoroquinolones generally have a good in vitro and clinical activity against isolates of the Salmonella species [1]. Although the Clinical and Laboratory Standards Institute (CLSI; formerly the National Committee for Clinical Laboratory Standards, NCCLS) guidelines recommend the MICs of ≤1 and ≥4 μg/ml as respective breakpoints of susceptibility and resistance to ciprofloxacin [2], it is now commonly accepted that Salmonella isolates with MICs between ≥0.125 and 2 μg/ml are characterized by reduced susceptibility i.e. low-level resistance to fluoroquinolones. The selection of MICs ≥0.125 μg/ml as a breakpoint of reduced ciprofloxacin susceptibility is justified based on histogram and scatterblot analyses combined with sequencing data [3-8]. This selection is further supported by reports on several treatment failures with ciprofloxacin and other fluoroquinolones in patients with infections caused by Salmonella strains showing reduced susceptibility to fluoroquinolones [9-17].
It is of concern that in recent years, the number of Salmonella isolates with reduced fluoroquinolone susceptibility has increased in many countries, including countries of the European Union [6,18-21]. In Finland, reduced fluoroquinolone susceptibility emerged between 1995 and 1997 among Salmonella isolates of domestic origin [21], and a significant increase in the annual proportion of reduced fluoroquinolone susceptibility was observed between 1995 and 1999 among both the domestic and foreign isolates [5].
Several fluoroquinolone preparations are on market all over the world, while a number of new compounds are presently being developed or undergoing clinical studies. The purpose of the present study was to determine the in vitro activities of various older and newer fluoroquinolones towards the Salmonella species, focusing special attention on the isolates with reduced ciprofloxacin susceptibility. In so doing, we determined the susceptibilities of 816 epidemiologically unrelated Salmonella isolates to ciprofloxacin and 10 additional fluoroquinolones including two novel extended-spectrum compounds: clinafloxacin and sitafloxacin.
Methods
Salmonella strains
Antimicrobial susceptibility of Salmonella enterica isolates has been surveyed in the National Public Health Institute, Finland, since 1995 by analyzing yearly 200–400 strains collected from Finnish patients seeking medical assistance for gastroenteritis. Starting in January each year, we consecutively collect 100–200 foreign strains and 100–200 domestic (i.e., Finnish) strains for susceptibility testing. An isolate is designated as of foreign origin, if the patient has reported travel abroad during one month before the specimen day. All other isolates are designated as of domestic origin. Epidemiological information regarding potential travelling and the travel destination is collected from the forms which accompany each isolate sent to the Enteric Bacteria Laboratory of the National Public Health Institute, Helsinki, which serves as the National Salmonella Reference Centre in Finland.
We included in this study a total of 816 clinical Salmonella strains collected between 1995 and 2003 during the annual surveys. All strains were isolated from stool. Of these strains, 365 were designated as of domestic origin and 451 as of foreign origin. The strains were considered to be epidemiologically unrelated based on their recovery from distinct sources. For each Salmonella outbreak recognized, only one isolate representing the epidemic strain was included. The Salmonella collection consisted of 119 different serotypes. All strains belonged to non-typhoidal Salmonella enterica. The most prevalent serotypes were S. Enteritidis, S. Typhimurium, and S. Hadar, accounting for 27.2%, 19.0%, and 7.1% of the isolates, respectively.
Susceptibility testing
The minimum inhibitory concentrations (MIC) of the isolates were determined by the standard agar plate dilution method according to the NCCLS guidelines [22]. Mueller-Hinton II agar (BBL, Becton Dickinson and Company, Cockeysville, Md.) was used as the culture medium. Altogether 11 fluoroquinolone compounds were analyzed. Reagent powder of each of these agents was provided by its manufactorer: ciprofloxacin (Bayer, Wuppertal, Germany), clinafloxacin (Pfizer, Ann Arbor, MI, Unites States), enrofloxacin (Bayer, Elberfeld, Germany), gatifloxacin (Grunenthal BHBH, Aachen, Germany), gemifloxacin (GlaxoSmithKline, Worthing, United Kingdom), levofloxacin (Hoechst Marion Roussel, Romainville Cedex, France), lomefloxacin (Sigma, St. Luis, MO, United States), moxifloxacin (Bayer, Wuppertal, Germany), norfloxacin and ofloxacin (Sigma, Steinheim, Germany), and sitafloxacin (Daiichi Pharmaceuticals, Tokyo, Japan). The Salmonella isolates were also tested for susceptibility to nalidixic acid.
Of the fluoroquinolone compounds analyzed, ciprofloxacin, ofloxacin, norfloxacin, levofloxacin, and moxifloxacin are presently on market in Finland.
Staphylococcus aureus ATCC 29213, Escherichia coli ATCC 25922, E. coli ATCC 35218, and Pseudomonas aeruginosa ATCC 27853 were used as controls in susceptibility testing.
Data analysis
The susceptibility data were analyzed using the WHONET5 computer program (available from ).
Results
Of all 816 Salmonella strains, 3 (0.4%) were resistant to ciprofloxacin (MIC ≥ 4 μg/ml), 232 (28.4%) showed reduced susceptibility to ciprofloxacin (MIC ≥ 0.125 – 2 μg/ml), and 581 (71.2%) were ciprofloxacin-susceptible. Of the 451 foreign Salmonella strains, 2 (0.4%) were ciprofloxacin-resistant and 193 (42.8%) showed reduced susceptibility to ciprofloxacin. Of the 365 domestic Salmonella strains, 1 (0.3%) was ciprofloxacin-resistant and 39 (10.7%) showed reduced susceptibility to ciprofloxacin.
For all 816 strains, the MIC50 and MIC90 values of ciprofloxacin were 0.032 and 0.25 μg/ml, respectively (Table 1). For ofloxacin, levofloxacin, norfloxacin, and moxifloxacin, the MIC50 values varied between 0.064 and 0.125 μg/ml; and the MIC90 values, between 0.5 and 1 μg/ml. The MIC50 and MIC90 values of enrofloxacin, lomefloxacin, gatifloxacin, and gemifloxacin were similar to or higher than those of ciprofloxacin. For both clinafloxacin and sitafloxacin, the MIC50 and MIC90 values were lower than for the other agents tested, 0.016 and 0.064 μg/ml, respectively. The histograms illustrating the MICs of the fluoroquinolones studied are presented in Figure 1. The MIC50 and MIC90 values of nalidixic acid were 4 and 512 μg/ml, respectively (range, <0.5 to >512 μg/ml).
Table 1 MICs of 11 fluoroquinolones for 816 non-typhoidal strains of Salmonella enterica isolated from Finnish patients between 1995 and 2003
MIC (μg/ml)
Fluoroquinolone Number of isolates MIC50 MIC90 Range
Ciprofloxacin 816 0.032 0.25 0.008 – 16
Clinafloxacin 816 0.016 0.064 0.002 – 2
Enrofloxacin 808 0.064 0.5 0.032 – >32
Gatifloxacin 816 0.032 0.25 0.002 – 8
Gemifloxacin 815 0.032 0.25 0.002 – >16
Levofloxacin 808 0.064 0.5 ≤0.008 – 16
Lomefloxacin 807 0.25 2 0.064 – >32
Moxifloxacin 816 0.125 0.5 0.008 – >16
Norfloxacin 816 0.125 1 0.016 – >32
Ofloxacin 816 0.125 1 0.016 – >16
Sitafloxacin 816 0.016 0.064 0.002 – 2
Figure 1 MIC histograms of fluoroquinolones. Minimum inhibitory concentration (MIC) histograms of 11 fluoroquinolones against 816 Salmonella strains collected from Finnish patients between January 1995 and January 2003.
The most prevalent serotypes among the 235 Salmonella strains with reduced susceptibility or resistance to ciprofloxacin were S. Hadar, S. Enteritidis, and S. Virchow accounting for 22.6%, 20.9%, and 12.3% of the strains, respectively. For these 235 strains, the MIC50 values of ofloxacin, levofloxacin, norfloxacin and moxifloxacin varied between 0.5 and 1 μg/ml; and the MIC90 value was 1 μg/ml for all (Table 2). The MIC50 and MIC90 values were lowest, 0.064 and 0.125 μg/ml, for both clinafloxacin and sitafloxacin. For the additional 4 fluoroquinolones examined, the MIC50 varied between 0.25 and 2 μg/ml; and the MIC90 values, between 0.5 and 2 μg/ml. The MIC50 and MIC90 values of nalidixic acid for these strains were 512 and >512 μg/ml, respectively (range, 4 to >512 μg/ml).
Table 2 MICs of 11 fluoroquinolones for 235 non-typhoidal strains of Salmonella enterica with reduced ciprofloxacin susceptibility1 or ciprofloxacin resistance2 isolated from Finnish patients between 1995 and 2003
MIC (μg/ml)
Fluoroquinolone Number of isolates MIC50 MIC90 Range
Ciprofloxacin 235 0.25 0.5 0.125 – 16
Clinafloxacin 235 0.064 0.125 0.008 – 2
Enrofloxacin 233 0.5 1 0.032 – >32
Gatifloxacin 235 0.25 0.5 0.032 – 8
Gemifloxacin 235 0.25 0.5 0.064 – >16
Levofloxacin 233 0.5 1 0.064 – 16
Lomefloxacin 232 2 2 0.25 – >32
Moxifloxacin 235 0.5 1 0.25 – >16
Norfloxacin 235 1 1 0.25 – >32
Ofloxacin 235 1 1 0.125 – >16
Sitafloxacin 235 0.064 0.125 0.016 – 2
1Ciprofloxacin MIC ≥0.125 – 2 μg/ml
2Ciprofloxacin MIC ≥4 μg/ml
The scattergrams correlating the MICs of ciprofloxacin to those of clinafloxacin, levofloxacin, moxifloxacin, norfloxacin, ofloxacin, and sitafloxacin for the 816 Salmonella strains are presented in Figure 2. These pictures illustrate that there is a distinct correlation between the ciprofloxacin susceptibility and the susceptibility to other fluoroquinolones. The MICs of levofloxacin, moxifloxacin, norfloxacin, and ofloxacin were generally higher than those of ciprofloxacin, but for each, the fully susceptible population was separate from the population with reduced susceptibility. Of the 232 Salmonella strains with reduced ciprofloxacin susceptibility, the MICs of levoxacin, moxifloxacin, norfloxacin, and ofloxacin were 1 or 2 dilution steps higher than those of ciprofloxacin. In contrast, the MICs of clinafloxacin and sitafloxacin were generally 2 dilution steps lower than those of ciprofloxacin.
Figure 2 Scattergrams correlating the MICs of ciprofloxacin to those of other fluoroquinolones. Scattergrams for 816 Salmonella strains correlating the minimum inhibitory concentrations (MIC) of ciprofloxacin to those of clinafloxacin, levofloxacin, moxifloxacin, norfloxacin, ofloxacin, and sitafloxacin. The numbers within the graphs indicate the numbers of Salmonella strains. The vertical solid lines indicate the Clinical and Laboratory Standards Istitute (CLSI) breakpoint recommendations for susceptibility and resistance, respectively, to ciprofloxacin (MIC ≤1 and ≥4 μg/ml); and the vertical dashed lines, the breakpoint for reduced susceptibility (MIC ≥0.125 μg/ml). The horizontal solid lines indicate the respective CLSI breakpoint recommendations to those fluoroquinolones for which such recommendations are available: ofloxacin (MIC ≤2 and ≥8 μg/ml); norfloxacin (MIC ≤4 and ≥16 μg/ml); and levofloxacin (MIC ≤2 and ≥8 μg/ml).
Discussion
Of all 816 Salmonella strains included in the present study, 28.8% showed resistance or reduced susceptibility to ciprofloxacin. The MIC50 and MIC90 values of ciprofloxacin for these isolates were 0.032 and 0.25 μg/ml, respectively, being similar to or lower than those of all older fluoroquinolone compounds studied. Thus, we show that ciprofloxacin still is the most effective fluoroquinolone drug registered in Finland for the treatment of salmonellosis. In the entire Salmonella collection, the MIC50 and MIC90 values were lowest for two new quinolones: clinafloxacin and sitafloxacin. Moreover, clinafloxacin and sitafloxacin exhibited the lowest MIC50 and MIC90 values for the 235 ciprofloxacin-resistant Salmonella strains or strains showing reduced ciprofloxacin susceptibility. Based on these in vitro results, both clinafloxacin and sitafloxacin appear promising drugs for the treatment of salmonella infections, but their efficacy has not been demonstrated in clinical investigations.
The introduction of fluoroquinolones in the 1980's had an almost revolutionary effect on the treatment of salmonellosis, since they offered an effective per oral alternative to treat clinical infections as well as to eradicate long-term carriage. So far, the emergence and rapid increase of reduced fluoroquinolone susceptibility in non-typhoidal salmonellas has not led to major consequences, since these microbes characteristically cause gastroenteritis, for which antimicrobial treatment is not always indicated. The situation is totally different in invasive salmonella infections, in which administration of an effective antimicrobial agent is vitally important. If such an infection is caused by a Salmonella strain with reduced fluoroquinolone susceptibility, treatment with a fluoroquinolone compound may not be a safe alternative. There are several reports showing that the use of fluoroquinolones in infections caused by Salmonella isolates with reduced susceptibility may lead to treatment failures [6-17]. The majority of the cases have involved Salmonella enterica serotype Typhi [8,10-12,15]. In non-typhoidal salmonellosis, the causative strain has often been initially fully susceptible, but after fluoroquinolone treatment failure, the MICs for these strains have been ≥0.125 μg/ml [13,14,17].
Clinafloxacin and sitafloxacin are the most interesting fluoroquinolone compounds studied here, since earlier data on their activity against the Salmonella species are limited. We found only one previous study, in which the in vitro activity of sitafloxacin against 326 Salmonella isolates was tested, with a finding that its activity was equal to or slightly better than that of ciprofloxacin [23].
One of the main purposes of the present study was to analyze the in vitro activities of these two novel fluoroquinolone compounds towards Salmonella isolates with reduced ciprofloxacin susceptibility. Our results show that the MIC50 and MIC90 of clinafloxacin and sitafloxacin against these strains were generally 2 dilution steps lower than those of ciprofloxacin and even 4 to 5 dilution steps lower than some of the older fluoroquinolones. Nevertheless, this does not necessarily mean that they should have a superior in vivo activity. The fluoroquinolone antimicrobial group is characterized by cross resistance, indicating that when a bacterial isolate is resistant to one fluoroquinolone, it is also resistant to other members of the same group [24].
The development of fluoroquinolone resistance most commonly involves a mutation in the quinolone resistance determining region (QRDR) of the gyrA gene [4,5,14,25,26]. One single point mutation usually leads to nalidixic acid resistance and reduced fluoroquinolone susceptibility, while additional mutations or accumulation of several resistance mechanisms are needed to produce high-level fluoroquinolone resistance. On this basis, one can assume that in the majority of the Salmonella strains with reduced fluoroquinolone susceptibility analyzed here, quinolone resistance is associated with one point mutation. Based on their low MIC values, both clinafloxacin and sitafloxacin may potentially have useful clinical activity against Salmonella strains with a single mutation in the gyrA gene. It must be borne in mind, however, that having undergone one point mutation, these isolates are potentially inclined to a second mutation, leading to higher MIC values. Thus, it is probable that resistance will develop during fluoroquinolone treatment also against these newer quinolones, despite their initially low MIC values.
It is also of note that the clinical relevance of various MIC values of clinafloxacin and sitafloxacin is not known. Further, no breakpoins of resistance or susceptibility have been given by the CLSI for these newer fluoroquinolone compounds. Neither do we know anything about their potential breakpoints for reduced fluoroquinolone susceptibility.
Many previous studies have focused on in vitro efficacy of clinafloxacin and sitafloxacin on microbes other than salmonellas [23,27-29]. In these studies, the two new fluoroquinolones have had a better activity than the older preparations towards a variety of bacterial species, even including the ciprofloxacin-resistant strains. For example, clinafloxacin exhibited greater activity than older fluoroquinolones against ciprofloxacin-resistant Klebsiella pneumoniae and Enterobacter aerogenes isolates [27], and sitafloxacin proved superior to the others against ciprofloxacin-resistant isolates of several enterobacterial species [23]. Moreover, a number of studies have demonstrated the clinical efficacy of clinafloxacin in the treatment of systemic infections, including severe skin and soft tissue diseases and infective endocarditis, and in empirical therapy of febrile granulocytopenic patients [30-33]. These data are encouraging and suggest that the clinical efficacy of these novel fluoroquinolones should be tested also in salmonellosis.
The possible clinical usefulness of the two newer quinolones against Salmonella strains will also depend on the levels of the drug reaching the infecting organism. It has been shown in previous studies that in general, newer fluoroquinolones have equal or greater bioavailability compared with ciprofloxacin, which varies between 55 to 88% [34]. Limited data suggest that at least 70% of sitafloxacin is absorbed after an oral dose [35]. In one study, the absolute bioavailability of orally administered clinafloxacin was approximately 90% [36]. Thus, it is expected that these drugs will prove effective also in the oral treatment of salmonellosis.
Conclusion
Among the registered fluoroquinolone compounds in Finland, ciprofloxacin still appears to be the most effective drug for the treatment salmonella infections. Among the newer preparations, both clinafloxacin and sitafloxacin are promising based on in vitro studies, since they exhibited the lowest MIC50 and MIC90 values for all Salmonella isolates as well as for those with reduced fluoroquinolone susceptibility.
Authors' contributions
PK, SP, PH and AH planned and carried out the design of the study. SP and AH participated in laboratory studies and interpretation of the data. AS participated in the collection of Salmonella isolates and data, and was responsible for serotyping of Salmonella isolates. PK and AH wrote the first draft of the manuscript. All authors had intellectual contribution, and all read and approved the final manuscript.
Acknowledgements
We thank Erkki Nieminen for technical assistance and Tarja Heiskanen, Liisa Immonen, and Minna Lamppu for laboratory assistance.
==== Refs
Asperilla MO Smego RA Scott LK Quinolone antibiotics in the treatment of Salmonella infections Rev Infect Dis 1990 12 873 889 2237130
Clinical and Laboratory Standards Institute Performance Standars for Antimicrobial Susceptibility Testing; Fifteenth Informational Supplement M100-S15 2005 Clinical and Laboratory Standards Institute. Wayne, PA, USA
Crump JA Barrett TJ Nelson JT Angulo FJ Reevaluating fluoroquinolone breakpoints for Salmonella enterica serotype Typhi and for non-typhi Salmonellae Clin Infect Dis 2003 37 75 81 12830411 10.1086/375602
Hakanen A Kotilainen P Jalava J Siitonen A Huovinen P Detection of decreased fluoroquinolone susceptibility in salmonellas and validation of nalidixic acid screening test J Clin Microbiol 1999 37 3572 3577 10523554
Hakanen A Kotilainen P Huovinen P Helenius H Siitonen A Reduced fluoroquinolone susceptibility in Salmonella enterica serotypes in travelers returning from Southeast Asia Emerg Infect Dis 2001 7 996 1003 11747728
Mølbak K Baggesen DL Aarestrup FM Ebbesen JM Engberg J Frydendahl K Gerner-Smidt P Petersen AM Wegener HC An outbreak of multidrug-resistant, quinolone-resistant Salmonella enterica serotype typhimurium DT104 N Engl J Med 1999 341 1420 1425 10547404 10.1056/NEJM199911043411902
Ouabdesselam S Tankovic J Soussy CJ Quinolone resistance mutations in the gyrA gene of clinical isolates of Salmonella Microb Drug Resist 1996 2 299 302 9158789
Wain J Hoa NT Chinh NT Vinh H Everett MJ Diep TS Day NP Solomon T White NJ Piddock LJ Parry CM Quinolone-resistant Salmonella typhi in Viet Nam: molecular basis of resistance and clinical response to treatment Clin Infect Dis 1997 25 1404 1410 9431387
Helms MP Vastrup P Gerner-Smidt P Mølbak K Excess mortality associated with antimicrobial drug-resistant Salmonella typhimurium Emerg Infect Dis 2002 8 490 495 11996684
Kadhiravan T Wig N Kapil A Kabra SK Renuka K Misra A Clinical outcomes in typhoid fever: adverse impact of infection with nalidixic acid-resistant Salonella typhi BMC Infect Dis 2005 5 37 15904505 10.1186/1471-2334-5-37
Launay O Van J-CN Buu-Hoi A Acar JF Typhoid fever due to a Salmonella typhi strain of reduced susceptibility to fluoroquinolones Clin Microbiol Infect 1997 3 541 543 11864178
Le Lostec Z Fegueux S Jouve P Cheron M Mornet P Boisivon A Reduced susceptibility to quinolones in Salmonella typhi acquired in Europe: a clinical failure of treatment Clin Microbiol Infect 1997 3 576 577 11864187
Pers C Sogaard P Pallesen L Selection of multiple resistance in Salmonella enteritidis during treatment with ciprofloxacin Scand J Infect Dis 1996 28 529 531 8953688
Piddock LJ Griggs DJ Hall MC Jin YF Ciprofloxacin resistance in clinical isolates of Salmonella typhimurium obtained from two patients Antimicrob Agents Chemother 1993 37 662 666 8494360
Rowe B Ward LR Threlfall EJ Ciprofloxacin-resistant Salmonella typhi in the UK Lancet 1995 346 1302 7475749 10.1016/S0140-6736(95)91906-6
Slinger R Desjardins M McCarthy AE Ramotar K Jessamine P Guibord C Toye B Suboptimal clinical response to ciprofloxacin in patients with enteric fever due to Salmonella spp. with reduced fluoroquinolone susceptibility: a case series BMC Infect Dis 2004 4 36 15380025 10.1186/1471-2334-4-36
Vasallo FJ Martin-Rabadan P Alcala L Garcia-Lechuz JM Rodriguez-Creixems M Bouza E Failure of ciprofloxacin therapy for invasive nontyphoidal salmonellosis Clin Infect Dis 1998 26 535 536 9502506
Frost JA Kelleher A Rowe B Increasing ciprofloxacin resistance in salmonellas in England and Wales 1991–1994 J Antimicrob Chemother 1996 37 85 91 8647778
Piddock LJ Ricci V McLaren I Griggs DJ Role of mutation in the gyrA and parC genes of nalidixic-acid-resistant salmonella serotypes isolated from animals in the United Kingdom J Antimicrob Chemother 1998 41 635 641 9687102 10.1093/jac/41.6.635
Threlfall EJ Frost JA Ward LR Rowe B Increasing spectrum of resistance in multiresistant Salmonella typhimurium Lancet 1996 347 1053 1054 8606603 10.1016/S0140-6736(96)90199-3
Hakanen A Siitonen A Kotilainen P Huovinen P Increasing fluoroquinolone resistance in salmonella serotypes in Finland during 1995–1997 J Antimicrob Chemother 1999 43 145 148 10381114 10.1093/jac/43.1.145
National Committee for Clinical Laboratory Standards Methods for Dilution Antimicrobial Susceptibility Tests for Bacteria that Grow Aerobically, 6th ed. Approved standard M7-A6 2003 National Committee for Clinical Laboratory Standards, Wayne, PA, USA
Milatovic D Schmitz F-J Brisse S Verhoef J Fluit AC In vitro activities of sitafloxacin (DU-6859a) and six other fluoroquinolones against 8,796 clinical bacterial isolates Antimicrob Agents Chemother 2000 44 1102 1107 10722524 10.1128/AAC.44.4.1102-1107.2000
Piddock LJV Fluoroquinolone resistance in Salmonella serovars isolated from humans and food animals FEMS Microbiology Reviews 2002 26 3 16 12007640 10.1016/S0168-6445(01)00076-6
Griggs DJ Gensberg K Piddock LJ Mutations in gyrA gene of quinolone-resistant Salmonella serotypes isolated from humans and animals Antimicrob Agents Chemother 1996 40 1009 1013 8849216
Reyna F Huesca M Gonzalez V Fuchs LY Salmonella typhimurium gyrA mutations associated with fluoroquinolone resistance Antimicrob Agents Chemother 1995 39 1621 1623 7492118
Brisse S Milatovic D Fluit AC Verhoef J Martin N Scheuring S Köhrer K Schmitz F-J Comparative in vitro activities of ciprofloxacin, clinafloxacin, gatifloxacin, levofloxacin, moxifloxacin, and trovafloxacin against Klebsiella pneumoniae, Klebsiella oxytoca, Enterobacter cloacae, and Enterobacter aerogenes clinical isolates with alterations in GyrA and ParC proteins Antimicrob Agents Chemother 1999 43 2051 2055 10428935
Heinemann B Wisplinghoff H Edmond M Seifert H Comparative activities of ciprofloxacin, clinafloxacin, gatifloxacin, gemifloxacin, levofloxacin, moxifloxacin, and trovafloxacin against epidemiologically defined Acinetobacter baumannii strains Antimicrob Agents Chemother 2000 44 2211 2213 10898706 10.1128/AAC.44.8.2211-2213.2000
Vila J Ribera A Marco F Ruiz J Mensa J Chaves J Hernandez G Jiminez De Anta MT Activity of clinafloxacin, compared with six other quinolones, against Acinetobacter baumannii clinical isolates J Antimicrob Chemother 2002 49 471 477 11864947 10.1093/jac/49.3.471
Glauser MP Brennscheidt U Cornely O Grigg A Figuera A Keyserling C Trostmann U Welling L Tack K Clinafloxacin monotherapy (CI-960) versus ceftazidime plus amikacin for empirical treatment of febrile neutropenic cancer patients Clin Microbiol Infect 2002 8 14 21 11906496 10.1046/j.1198-743x.2001.00338.x
Levine DP Holley HP Eiseman I Willcox P Tack K Clinafloxacin for the treatment of bacterial endocarditis Clin Infect Dis 2004 38 620 631 14986244 10.1086/381670
Siami G Christou N Eiseman I Tack KJ the Clinafloxacin severe skin and soft tissue infections study group Clinafloxacin versus piperacillin-tazobactam in treatment of patients with severe skin and soft tissue infections Antimicrob Agents Chemother 2001 45 525 531 11158750 10.1128/AAC.45.2.525-531.2001
Winston DJ Lazarus HM Beveridge RA Hathorn JW Gucalp R Ramphal R Chow AW Ho WG Horn R Feld R Louie T Territo MC Blumer JL Tack KJ Randomized, double-blind, multicenter trial comparing clinafloxacin with imipenem as empirical monotherapy for febrile granulocytopenic patients Clin Intect Dis 2001 32 381 390 10.1086/318500
Aminimanizani A Beringer P Jelliffe R Comparable pharmacokinetics and pharmacodynamics of the newer fluoroquinolone antimicrobials Clin Pharmacokinet 2001 40 169 187 11327197
Nakashima M Uematsu T Kosuge K Umemura K Hakusui H Tanaka M Pharmacokinetics and tolerance of DU-6859a, a new fluoroquinolone, after single and multiples oral doses in healthy volunteers Antimicrob Agents Chemother 1995 39 170 174 7695301
Randinitis EJ Brodfuehrer JI Eiseman I Vassos AB Pharmacokinetics of clinafloxacin after single and multiple doses Antimicrob Agents Chemother 2001 45 2529 2535 11502525 10.1128/AAC.45.9.2529-2535.2001
|
16143044
|
PMC1208849
|
CC BY
|
2021-01-04 16:38:21
|
no
|
Ann Clin Microbiol Antimicrob. 2005 Sep 5; 4:12
|
utf-8
|
Ann Clin Microbiol Antimicrob
| 2,005 |
10.1186/1476-0711-4-12
|
oa_comm
|
==== Front
Aust New Zealand Health PolicyAustralia and New Zealand Health Policy1743-8462BioMed Central London 1743-8462-2-191612020710.1186/1743-8462-2-19Research(Re)form with Substance? Restructuring and governance in the Australian health system 2004/05 Rix Mark [email protected] Alan [email protected] Kathy [email protected] Centre for Health Service Development, Faculty of Commerce, University of Wollongong, NSW, 2515, Australia2005 24 8 2005 2 19 19 11 7 2005 24 8 2005 Copyright © 2005 Rix et al; licensee BioMed Central Ltd.2005Rix 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 Australian health system has been the subject of multiple reviews and reorganisations over the last twenty years or more. The year 2004–2005 was no different.
This paper reviews the reforms, (re)structures and governance arrangements in place at both the national and state/territory levels in the last year. At the national level some progress has been made in 2004/05 through the Australian Health Ministers' Council and there is now a national health reform agenda, albeit not a comprehensive one, endorsed by the Council of Australian Governments (COAG) in June 2005. Quality and safety was an increasing focus in 2004–2005 at both the national and jurisdictional levels, as was the need for workforce reform. Although renewed policy attention was given to the need to better integrate and coordinate health care, there is little evidence of any real progress this last year. More progress was made on a national approach to workforce reform.
At the jurisdictional level, the usual rounds of reviews and restructuring occurred in several jurisdictions and, in 2005, they are organisationally very different from each other. The structure and effectiveness of jurisdictional health authorities are now more important. All health authorities are being expected to drive an ambitious set of national and local reforms. At the same time, most have now blurred the boundary between policy and service delivery and are devoting significant resources to centrally 'crisis managing' their service systems. These same reasons led to decentralisation in previous restructuring cycles. While there were many changes in 2004–2005, and a new national report to COAG on health reform is expected at the end of 2005, based on current evidence there is little room for optimism about the prospects for real progress.
==== Body
Review
The Council of Australian Governments' (COAG) 15th meeting on 3 June 2005 in Canberra endorsed a national health reform agenda with an unusual level of national consensus. The heads of Governments agreed that Australia has one of the best health systems in the world, albeit with room for improvement, particularly in areas where governments' responsibilities intersect. After several years of apparent stalemate, it seemed that the discussion had re-opened on ways to improve Australia's health system. In its most ambitious section, the COAG 2005 Communique "agreed that where responsibilities between levels of government need to change, funding arrangements would be adjusted so that funds would follow function." [1]
The governments stated in their Communique that the health system can be improved by clarifying roles and responsibilities, and by reducing duplication and gaps in services. They recognised that many Australians, including the elderly and people with disabilities, still face problems at the interfaces of different parts of the health system. They restated the aim of integration and a smooth transition between acute care, home and residential care, and helping younger people with disabilities. The necessary themes of workforce supply and flexibility, prevention, electronic records, a national call centre, and rural and remote services were reinforced. Senior officials were given till December 2005 to come up with a plan of action to progress these reforms.
The core ideas behind this change of climate have been well rehearsed in both health bureaucracies and the submissions of industry groups. In 2004 the Australian Healthcare Association (AHA) released five policies, the first of which called for a National Health System so that, by 2008, all Australian governments will have adopted a nation-wide approach to health policy and service delivery. AHA argued that "a National Health System is fundamental to successful health system reform in Australia and will provide access to health care services for Australians irrespective of borders or payers" [2].
The rest of the policy agenda was a call for a National Package of Healthcare Services, so that the next Australian Healthcare Agreement (2008–2013) would govern all public sector health programs and services administered by all Australian governments in partnership [3]. This was accompanied by a National Approach to Quality and Safety in Health [4], and policies for better integration and coordination of health care [5], and a national approach to workforce reform [6].
While no doubt ambitious, there is little dispute about the merit of these policies and their potential impact on health in Australia. Given the renewed relevance of larger scale reform under the 2005 COAG announcements, these policy signposts form a useful framework by which to assess the state of the Australian health care system and its attempts at reform in 2004/05.
The state of play in Australian-state/territory government relations in 2004/05 – ritual or reform?
Without a clear strategy to move to a more truly national system, all reform is going to be counterbalanced by the inbuilt tendencies of the current system to move towards increased fragmentation. In a recent overview and assessment of Australian health system restructuring, Dwyer lamented that 'Unfortunately, the Commonwealth: state responsibility split, the one structural barrier most central to the systemic weakness of Australian primary care (and therefore most important for the capacity to develop and support new models of care for chronic diseases), is one that a state can't address, at least not alone [7].'
Before the COAG announcement in June 2005, there was little prospect of progress. The Prime Minister had announced in October 2004 a Task Force headed by Andrew Podger, the previous head of the Department of Health and Ageing (DHA). Consisting of officers from the Departments of Prime Minister and Cabinet, Health and Ageing and the Treasury, its role was to review the operation of health policy to examine how to improve the delivery of health services. Finalised in early 2005, its report has not been released. Federal Health Minister Tony Abbott referred briefly to it earlier in 2005 by commenting that it has been commissioned so that "the Government can respond to any state proposal" [8].
At that time Minister Abbott had outlined the problems of the health system from the point of view of the Australian Government when he addressed the Committee for Economic Development of Australia Conference in February 2005. In his address, Minister Abbott was not concerned with primary care, but with the 'big health issue' for 2005 – hospitals. For the Minister, this was primarily a matter for the states and territories rather than the Commonwealth, and a good opportunity for some "free-kicks". He pointed out that Section 51 of the Constitution relegates the Commonwealth to little more than a funding authority having no operational control of public hospital systems in the subordinate jurisdictions. While he did concede that it would make more sense for one level of government to be responsible for the entire health system, the real issue of the day was not so much who funds hospitals but how they are managed.
The Minister complained that 'years of poor management mean that public hospital patients now face long waits for essential as well as elective treatment (Abbott 2005).' Private hospitals, in stark contrast, were in the business of providing patients 'with what they want, when they want it'. The challenge for the federal government was to exercise effective leadership over the public hospital systems that are run by the states and the territories and private sector hospitals which, the minister remarked, 'aren't run by the government at all [8].'
Fortunately, the national health reform agenda is too important to be left to health ministers alone. Continuing its previous calls for more reform, the Productivity Commission recommended in February 2005 an independent public review of Australia's health care system, as the first step in the development of an integrated reform program. "The review should include consideration of: the future determinants of demand for and supply of health services; health financing (including Federal/State responsibilities and their implications); coordination of individual services (including with aged care); the interface between public and private services; information management; and the appropriate balance of resourcing between prevention and treatment" [9].
If a revived impetus for a national approach to reform through heads of government can broaden the debate beyond who runs hospitals, and look at primary care and the relationships to residential aged care and community care, then the latest review under COAG will be able to build on progress that has been made under the auspices of the health ministers' conferences at national level over the last year.
Toward a national reform agenda – small steps in the right direction?
The perennial issue of the Commonwealth: state/territory split of responsibility for health is hardly the only matter of concern in assessing the effectiveness of public health policy in Australia – even though, as Dwyer comments, it is 'probably the single most significant problem in health system design [[7], p 4].' The treatment and prevention of chronic disease are also of great concern with chronic disease accounting for '80% of the total burden of disease' and approximately '40% of total health expenditure' [[7], p 6].
Chronic disease and related issues were high on the agenda of the meeting between Australian Health Ministers and clinicians in Hobart in July 2004. The meeting was reviewing the progress of the Health Reform Agenda agreed to at the November 2003 meeting of Health Ministers (the Health Ministers had first agreed to the Reform agenda at their April 2003 meeting). Instead of a large systematic process of reform as suggested by the Productivity Commission and others, health reform in Australia is to be progressed in a series of small steps.
The first step was the establishment of a Health Reform Agenda Working Group (HRAWG) to progress the Health Reform Agenda. It reports to the Australian Health Ministers' Advisory Council (AHMAC). As shown in Table 1, there are optimistic signs in 2005 of some progress toward genuine reform, albeit in small steps.
Table 1 Outcomes of Australian Health Ministers' Conferences 2004–2005 in relation to structural reform
Agreement / Outcome Press release date
Agreement to take "immediate action to progress reform of the Australian health care system in the areas of after hours GP services; aged care; chronic disease and cancer services; medical workforce planning; and, renal disease services" [10] 28 November 2003
Establishment of a national nursing taskforce to drive major nursing education and workforce reforms [11] 28 November 2003
Release of Australia's first national health workforce strategic framework [12] 23 April 2004
Agreement to take further steps "to progress reform of the Australian health care system in the areas of after hours GP services; aged care; chronic disease and cancer services; medical workforce planning; and, renal disease services" [13] 23 April 2004
Agreement on a nationally consistent approach to medical registration [14] 23 April 2004
Agreement on the first National Health Workforce Action Plan [15] 29 July 2004
Agreement to continue the Health Reform Agenda and the future priorities for reform [16] 29 July 2004
Agreement to establish a Review of the Future Governance Arrangements for Safety and Quality in Health Care [17] 29 July 2004
Agreement to establish a new national entity to drive critical e-health initiatives – NEHTA [18] 28 January 2005
Endorse development of a National Framework for Action on Dementia [19] 28 January 2005
At the April 2004 meeting, the Health Ministers acknowledged the need for immediate action to ensure progress in reforming after hours GP services, aged care, chronic disease and cancer services, medical workforce planning, and renal disease services. Among other matters, the Health Ministers agreed to establish a 'set of principles' that would allow jurisdictions to work together towards improving delivery of after hours GP services in certain regions and building collaborative working relationships with emergency departments in public hospitals.
They also agreed to a range of initiatives such as enhanced transition care, rehabilitation and step-down care that would improve the transition between acute and aged care services. In addition, the Ministers reached agreement to finalise an integrated Chronic Disease Strategy and Service Improvement Framework for Cancer services [13].
At their July 2004 meeting, Health Ministers were asked by clinicians to consider three issues that they regarded as important:
• integration and coordination of services at the community-based and hospital-based services interface;
• improving the community's access to better health outcomes, in particular, for children, people with chronic care needs, older Australians, and indigenous Australians; and
• the need for a sustainable, skilled and flexible workforce to enable the adequate provision of health services into the future.
As these issues were already on their Health Reform Agenda, Health Ministers agreed to endorse child health and well being as a specific area for reform. In turn, the clinicians recommended that a way to progress a number of the important items on the Reform agenda was to conduct a trial in each state and territory of specific services that integrate community-based and hospital based-services, suggesting coordinated chronic care and integrated aged care as possible cases [16]. Those trials are yet to eventuate.
At the same meeting, the ministers agreed to establish a Review of the Future Governance Arrangements for Safety and Quality in Health Care. The review is to advise on future arrangements for the effective leadership and national coordination of safety and quality initiatives in health care. It is to report before the Australian Council for Safety and Quality in Health Care completes its current term in June 2006. National governance arrangements for leadership and coordination of safety and quality in health care were accordingly included in the Terms of Reference of the health care safety and quality governance arrangements review [17]. In parallel, several states and territories introduced their own initiatives, reflecting the increasing priority being given to governance arrangements for quality and safety in 2004–2005.
Meeting in Sydney in January 2005, the Australian Health Ministers' Conference agreed to establish the National E-Health Transition Authority (NEHTA). This is a new entity in the form of a company limited by guarantee and governed by a board of directors comprising the CEOs of the national, state and territory Health Departments. Its core activities include 'the development of timelines for the urgent advancement of the e-health agenda; option assessment and business case development; standards development and implementation support; and provision of advice and resources to assist implementation of already agreed solutions [18].' It is also expected to advance other significant national priorities in key areas including clinical data standards and terminologies, consent models, electronic health record (EHR) standards, and health informatics industry reform.
Incremental and crisis-driven reforms at state/territory level – just more change or so me real progress?
Australian states and territories have a long history of independent reviews leading, cyclically, to the centralisation and decentralisation of management and governance at various times. In 2004–05, Australian jurisdictions are, in the main, in a centralisation phase. Queensland is the subject of an independent review at the time of writing while Western Australia has a Health Reform Implementation Taskforce in progress. Dwyer [7] reviewed the round of reviews in the Australian health system between 2002 and 2004. These resulted in restructuring in New South Wales, South Australia, the Northern Territory, Western Australia, and the ACT. There is a strong tendency towards increasing centralisation so that, in 2005, 6 of 8 jurisdictions now directly manage public sector health services, with Victoria and South Australia having mixed models. With the recent centralisation of management in New South Wales, Dwyer calculated that two thirds of Australians now live in areas under centralised control [7].
Given this, the structure and effectiveness of jurisdictional health authorities is becoming increasingly important in determining whether reforms are achieved in areas such as clinical governance, quality and safety and others included in the National Reform Agenda. This is especially the case given that, at the same time, the centralised authorities will continue to devote considerable resources to responding to each new 'crisis' in their service system.
Table 2 summarises the management structures in place in each jurisdiction in May 2005. As this summary indicates, there are significant differences in the role and scope of the various health authorities. This has important implications in relation to structural opportunities to reform and improve the coordination and planning of service delivery, particularly for those with complex and continuing care needs. At the same time, there is little similarity in how the various departments are organised, as reflected in their executive level divisional structures (listed in alphabetical order in the table).
Table 2 Management of health and human services by jurisdiction – the state of play in 2004–2005
Jurisdiction Scope Organisational divisions Regions
Australian Government Health and Ageing
Separate authorities for Family and Community Services and Veterans' Affairs Acute Care
Ageing and Aged Care
Business
Health Services Improvement
Medical and Pharmaceutical Services
Office of Aboriginal and Torres Strait Islander Health
Population Health
Portfolio Strategies
Primary Care Each states and territory is a region
ACT Health
Separate authorities for Disability, Housing and Community Services. ACT Emergency Services Authority provides the ambulance service
Separate Community and Health Services Complaints Commissioner established in late 2004 Allied Health Adviser
Clinical operations
Financial and Risk Management
Government Relations and Planning
Human Resource Management
Information Services
Nursing and Midwifery Office
Policy
Population health None. All services directly managed by the Department
Northern Territory Health and community services
Separate authorities for Community Development, Sport and Cultural Affairs
Separate Health and Community Services Complaints Commission Aboriginal Health, Family & Social Policy
Acute Care
Community Services
Corporate Management Services
Health Services
Information
Strategy & Quality None. All services directly managed by the Department
St John Ambulance Service is separately incorporated, as are some Aboriginal Health Services
NSW Health
Separate authorities for Ageing, Disability and Home Care, Housing, Community Services and Medical Research
Separate Health Care Complaints Commission Health System Performance
Health System Support
Population Health
Strategic Development 8 Area Health Services (no Boards) reporting directly to the department plus:
Ambulance Service of NSW
Children's Hospital at Westmead (with Board of Directors)
Justice Health
Clinical Excellence Commission
NSW Cancer Institute
Queensland Health
Separate authorities for Child Safety, Communities, Emergency Services [including Ambulance Service], Housing, Disability Services
Separate Health Rights Commission of Queensland Health Services
Information
Innovation and Workforce Reform
Resource Management.
Strategic Policy and Government Liaison 3 Zones
37 Districts within zones
All services directly managed by the Department
South Australia Health
Department of Families and Communities manages other human services, including Aged and Community Care
Separate Health and Community Services Complaints Commission announced in 2004
Separately incorporated bodies deliver ambulance services
Veterans Repatriation Hospital separate
Hospitals and Dom care separate incorporation Population and Environmental
Healthy SA
Service Planning
State Dental managed in a region
Mental Health managed in a region
Drug and Alcohol managed in a region
SA Health Reform 2 metropolitan health regions and Children, Youth and Women's Health Service with own Boards.
4 country regional health services
Tasmania Health and Human Services
Separate Health Complaints Commissioner Children and Families
Community, Population and Rural Health
Corporate Services
Hospitals and Ambulance
Housing Tasmania None. All services directly managed by the Department
Victoria Health and Human Services
Separate Office of the Health Services Commissioner
Office for Children that now reports to Minister for Children Disability Services
Financial & Corporate Services
Housing & Community Building
Metropolitan Health & Aged Care Services
Operations
Policy & Strategic Projects
Rural & Regional Health & Aged Care Services 8 departmental Regions
12 Melbourne networks with own boards within metropolitan regions
71 agencies with own boards in rural regions
Victorian Ambulance Service
Western Australia Health
Separate authorities for Community Development, Disability Services and Housing, Office of Health Review.
Separate Office of Safety and Quality Clinical Policy Division.
Statewide Health Support
Population Health Division
Country Health Services
Central Wait List Bureau 3 Area Health Services, 1 Country Health Service and Women's and Children's Health Service, all directly managed by the Department
St John Ambulance Service is separately incorporated
At a national level, the department is responsible for both health and ageing. But the health care needs of war veterans are the responsibility of the Department of Veterans Affairs (DVA) and not the Department of Health and Ageing (DHA). The 'ageing and aged care' functions of DHA include community care programs and services such as the Home and Community Care (HACC) program that are managed by the health authority in all but two jurisdictions. New South Wales has a separate Department of Ageing, Disability and Home Care. In South Australia, the previous Department of Human Services was split on 1 July 2004 into two, with a new Department of Families and Communities taking responsibility for, among other portfolios, community care and disability.
In 2005, an authority with broader human and community services responsibilities is managing health care in the Northern Territory, Tasmania and Victoria. These other responsibilities include, among some others, disability services and housing. Neither function is now within scope of the health departments in the other jurisdictions.
Tasmania has the broadest role and is responsible for both the policy and direct operations of its ambulance service. This is not the case in either Victoria or the Northern Territory where the department manages policy but ambulance services are separately incorporated. In other states with a narrower 'health department', ambulance services are managed by departments of emergency services (ACT, Queensland), by the health department (New South Wales) or are separately incorporated services (South Australia).
All jurisdictions now have independent authorities (however named) to review health care complaints. The ACT and South Australia established theirs in 2004. However, important differences in the philosophy and role of such bodies were identified in evidence given to the Special Commission of Inquiry into Campbelltown and Camden Hospitals [20], particularly in relation to their role in 'blaming' those responsible for errors.
Western Australia and New South Wales have gone further. An independent Office for Safety and Quality in Health Care was established in Western Australia in 2002. It is responsible for supporting the establishment of effective quality and safety systems, as well as investigating complaints. New South Wales established a separate Clinical Excellence Commission in 2005 (replacing its previous Institute of Clinical Excellence). Not surprisingly, both initiatives followed major media coverage of 'hospital scandals' in those two states.
Organisational and executive structures differ between jurisdictions. As one example, public (or population) health is its own division, and reports directly to the departmental head, in the ACT, WA and NSW. In Victoria, it is an office within the Rural and Regional Health and Aged Care Services Division while in Queensland it is a branch within the Health Services Division. Population health functions in the Northern Territory also sit in a Health Services Division, but not in one branch. Instead, population health is the responsibility of several sections including the Centre for Disease Control and a Health Development and Oral Health Branch. In Tasmania, population health is a subdivision of the Community, Population and Rural Health Division. At least in part, these differences reflect the scope of the various departments. However, there is no evidence to suggest whether any of these structures produce more effective policy than others. Nor is there evidence on what structure is best able to manage the health system and its reform.
As one further example, workforce reform (one of the five 2004 AHA policies and also on the Australian Health Ministers Health Reform Agenda) is managed differently across the jurisdictions. In 2005, Queensland has a new Innovation and Workforce Reform Directorate while Western Australia announced in May 2005 the creation of a new Clinical Reform and Policy Division. In other jurisdictions, there is either no organisational unit responsible for workforce reform or it is incorporated in the functions of other sections such as human resource departments. As before, there is no evidence to suggest whether any of these structures will be more effective than others in delivering on the workforce reform agenda.
One reason for the differences between jurisdictions appears to be the circumstances that triggered each of their latest restructures. As Dwyer [7] notes, all but one (NSW) arose from an independent review with the reorganisation of NSW coming in the aftermath of a hospital 'scandal' that attracted much media coverage. On the same basis, a number of reviews are now underway in Queensland.
In response to the so-called 'Doctor Death' scandal in relation to the appointment of Dr Jayant Patel in Bundaberg (Queensland), the Queensland Premier (not the Health Minister) announced in April 2005 the Queensland Health Systems Review. Its establishment had been the suggestion of the major doctors lobby group, the Australian Medical Association (AMA) [21]. This major review of Queensland Health's administration, management and performance systems is due for public release on 30 September 2005.
At the same time, three other inquiries have been commissioned. A Commission of Inquiry has been established to investigate events at Bundaberg Hospital, including the role and conduct of the Queensland Medical Board in relation to overseas trained medical practitioners. Like the Queensland Health Systems Review, it has also been asked to consider changes to recruitment, employment and supervision of medical practitioners, management of complaints and measures to increase the availability of medical practitioners across the State. In parallel, the Crime and Misconduct Commission is also conducting a public inquiry into Queensland Health's handling of complaints regarding care at Bundaberg Hospital and a Queensland Health review of clinical services at Bundaberg Hospital is also underway [22].
The Queensland Health Systems Review has broad terms. The administrative systems to be examine include (among other matters) district and corporate organisational structures and layers of decision making; corporate planning and budgeting systems; the effectiveness of performance reporting and monitoring systems; quality and safety systems; and clinical audit and governance systems. On the workforce front, it will examine recruitment; retention; training and clinical leadership. It will also review performance management systems including asset management and planning, information management and monitoring systems.
Regardless of the detail, it seems unlikely that the status quo will remain in Queensland in 2006. No doubt other jurisdictions will be watching in an attempt to learn the lessons.
Conclusion
We noted that the five AHA policies released in 2004 form a useful framework by which to assess the state of the Australian health care system and its attempts at reform in 2004/05. In 2005, Australia does not have a National Health System. Some progress has been made in 2004/05 and there is now a national health reform agenda under COAG. However, the current evidence suggests it is still a reform agenda in separate bits and will not be systemic. Little or no progress was made toward a National Package of Healthcare Services and there are no indications of any progress in the near future on that front. A National Approach to Quality and Safety is emerging, with significant advances in some jurisdictions. Better integration and coordination of health care remained a fashionable idea in 2004/05 but this goal has been acknowledged as important for decades and real progress is dependent on more systemic change. More progress was made on a national approach to workforce reform with the release of a national "strategic framework to guide national health workforce policy and planning throughout the decade". But a framework is still a long way from a strategy.
At the state and territory level, reviews and restructuring continued in several jurisdictions. In 2005, there are significant organisational differences between them, with little evidence of the strengths and weaknesses of the different approaches. What is becoming increasingly apparent is that the structure and effectiveness of jurisdictional health authorities is now more important. All health authorities are being expected to drive an ambitious set of national and local reforms. At the same time, most have now blurred the boundary between policy and service delivery and are devoting significant resources to 'crisis managing' their service systems. These same reasons led to decentralisation in previous restructuring cycles.
With scandals, public criticism and concern with rising costs increasingly being the impetus to restructure, the prospects for 2006 are for more of the same. At the same time, delivering on the reform promises of 2004/05 will become increasingly difficult but more important than ever.
Competing interests
The author(s) declare that they have no competing interests.
==== Refs
Council of Australian Governments 2005 Communique, June 3
Australian Healthcare Association AHA 2004 Policy 1. A National Health System
Australian Healthcare Association AHA 2004 Policy 2. A National Package of Healthcare Services
Australian Healthcare Association AHA 2004 Policy 3. A National Approach to Quality and Safety in Health
Australian Healthcare Association AHA 2004 Policy 4. Better integration and coordination of healthcare
Australian Healthcare Association AHA 2004 Policy 5. A National Approach to Workforce Reform
Dwyer JM Australian health system restructuring – what problem is being solved? Australia and New Zealand Health Policy 2004 1 6 15679935 10.1186/1743-8462-1-6
Abbott T 2005 Minister for Health and Ageing The Hon Tony Abbott MHR, addresses the Committee for Economic Development of Australia conference in Sydney Press release 25 February 2005
Productivity Commission Review of National Competition Policy Reforms, Report no. 33 Canberra 2005
Australian Health Ministers' Conference 2003 Joint Communique Australian Health Ministers agree to Reform Agenda 28 November 2003
Australian Health Ministers' Conference 2003 Joint Communique National Nursing and Nursing Education Taskforce
Australian Health Ministers' Conference 2004 Joint Communique National health workforce strategic framework
Australian Health Ministers' Conference 2004 Joint Communique Health Ministers Agree to Reform Agenda
Australian Health Ministers' Conference 2004 Joint Communique Australian Health Ministers agree on nationally consistent approach to medical registration
Australian Health Ministers' Conference 2004 Joint Communique National health workforce strategic framework
Australian Health Ministers' Conference 2004 Joint Communique Health Ministers Agree to continue reform agenda 29 July 2004
Australian Health Ministers' Conference 2004 Joint Communique Review of Future Governance Arrangements for Safety and Quality in Health Care 29 July 2004
Australian Health Ministers' Conference 2005 Joint Communique National entity to drive E-Health 28 January 2005
Australian Health Ministers' Conference 2004 Joint Communique Australian Health Ministers endorse development of a National Framework for Action on Dementia 28 January 2005
Walker B Final Report of the Special Commission of Inquiry into Campbelltown and Camden Hospitals 2004 30 July 2004
Beattie P 2nd Inquiry Will Check Health Systems To Aim For Better Results Press Release 26 April 2005
Queensland Health Other reviews 2005
|
16120207
|
PMC1208850
|
CC BY
|
2021-01-04 16:38:28
|
no
|
Aust New Zealand Health Policy. 2005 Aug 24; 2:19
|
utf-8
|
Aust New Zealand Health Policy
| 2,005 |
10.1186/1743-8462-2-19
|
oa_comm
|
==== Front
Behav Brain FunctBehav Brain FunctBehavioral and Brain Functions : BBF1744-9081BioMed Central 1744-9081-1-121606096310.1186/1744-9081-1-12ResearchMoment-to-moment dynamics of ADHD behaviour Aase Heidi [email protected] Terje [email protected] Norwegian Centre for the Studies of Conduct Problems and Innovative Practice, UNIRAND, University of Oslo, P.O.Box 1565 Vika, 0118 Oslo, Norway2 Department of Physiology, University of Oslo, Norway3 Centre for Advanced Studies (CAS) at the Norwegian Academy for Science and Letters, Oslo, Norway2005 1 8 2005 1 12 12 23 5 2005 1 8 2005 Copyright ©2005 Aase and Sagvolden; licensee BioMed Central Ltd.2005Aase and Sagvolden; 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 properly cited.Background
The behaviour of children with Attention-Deficit / Hyperactivity Disorder is often described as highly variable, in addition to being hyperactive, impulsive and inattentive. One reason might be that they do not acquire complete and functional sequences of behaviour. The dynamic developmental theory of ADHD proposes that reinforcement and extinction processes are inefficient because of hypofunctioning dopamine systems, resulting in a narrower time window for associating antecedent stimuli and behaviour with its consequences. One effect of this may be that the learning of behavioural sequences is delayed, and that only short behavioural sequences are acquired in ADHD. The present study investigated acquisition of response sequences in the behaviour of children with ADHD.
Methods
Fifteen boys with ADHD and thirteen boys without, all aged between 6–9 yr, completed a computerized task presented as a game with two squares on the screen. One square was associated with reinforcement. The task required responses by the computer mouse under reinforcement contingencies of variable interval schedules. Reinforcers were cartoon pictures and small trinkets. Measures related to response location (spatial dimension) and to response timing (temporal dimension) were analyzed by autocorrelations of consecutive responses across five lags. Acquired response sequences were defined as predictable responding shown by high explained variance.
Results
Children with ADHD acquired shorter response sequences than comparison children on the measures related to response location. None of the groups showed any predictability in response timing. Response sequencing on the measure related to the discriminative stimulus was highly related to parent scores on a rating scale for ADHD symptoms.
Conclusion
The findings suggest that children with ADHD have problems with learning long sequences of behaviour, particularly related to response location. Problems with learning long behavioural sequences may ultimately lead to deficient development of verbally governed behaviour and self control. The study represents a new approach to analyzing the moment-to-moment dynamics of behaviour, and provides support for the theory that reinforcement processes are altered in ADHD.
Reinforcementresponse sequenceserial behaviourvariabilitymotor controlautocorrelationsbehavioural units
==== Body
Background
Attention-Deficit/Hyperactivity Disorder (ADHD) [1] is a behavioural disorder characterized by developmentally inappropriate levels of hyperactive, inattentive, impulsive, and variable behaviour. Impulsiveness is increasingly considered as a major behavioural symptom. A recent comprehensive theory of ADHD, the dynamic developmental theory (DDT), suggests two processes, altered reinforcement processes and inefficient extinction, as being causative of several of the behavioural symptoms in ADHD [2,3]. Specifically, the DDT suggests that delayed learning of complete and functional behavioural sequences may be causing the frequent shifts between activities, non-completion of tasks, lack of long-term planning, and deficient self-control that often are described as outcomes of impulsivity.
There is some support for the notion that ADHD behaviour may be characterized by hampered acquisition of complete and functional sequences of behaviour. First, children with ADHD did not perform sequences of arm movements as one functional unit, but were slower, showed greater variability in movement timing, and demonstrated longer inter-segment intervals than children without ADHD, who appeared to program the entire arm movements and executed the sequence as one functional unit that was temporally coordinated [4]. The children without ADHD in this study showed age adequate planned movement, while the children with ADHD resembled the performance of younger children using "on-line" or immediate-feedback monitoring [5]. Second, in a serial choice button-press task where advance information about the next steps in the sequence was gradually reduced, children with ADHD (and children with Tourette syndrome) showed increasing movement sequencing deficits compared to healthy controls as the level of advance information was reduced [6]. Third, on a task requiring high-level controlled processing (follow a target that randomly moves across the computer screen), preschool children at risk for ADHD were disproportionately more inaccurate and variable compared to healthy controls, children with borderline ADHD, and children with other psychopathology [7]. On a task requiring low-level processing (trace the mouse cursor within the limits of two lines), though, the difference between the groups was not significant. The authors concluded that deficits in self-control and self-regulation seemed to be present very early in the development of ADHD [7]. Finally, in a study investigating multitasking in ADHD and community controls, children with ADHD appeared to have a specific deficit in monitoring their ongoing behaviours and generating useful strategies for task completion [8].
Reinforcement and behavioural sequences
The DDT suggests that dysfunctioning reinforcement and extinction processes can explain why symptomatic ADHD behaviour is acquired through dynamic interaction between the child and the environment throughout development [2,3]. Reinforcement and extinction are the main selection mechanisms of behaviour, and they are associated with dopaminergic activity [9]. According to the DDT, these mechanisms may operate constantly to reprogram neuronal connections by strengthening (reinforcing or potentiating) connections associated with reinforced behaviour, and at the same time weakening (extinguishing or depressing) other neuronal connections associated with nonreinforced behaviour [2].
On a behavioural level, reinforcers select responses by increasing the probability of repeating responses that produce reinforcers. Reinforcement as a process operates within a limited time window from the occurrence of the behaviour to the perception of the consequences of this behaviour. Altered reinforcement processes in ADHD may be described as a narrower time window than normal for associating behaviour with its consequences. A narrow time window may theoretically be depicted as a shorter and steeper delay-of-reinforcement gradient (called delay gradient from here onwards) (Figure 1). The delay gradient describes that the effect of a reinforcer is largest when it is delivered immediately after the response has been emitted, and wanes as a function of the delay in reinforcer delivery. The delay gradient thus depicts the relation between reinforcers and responses as an effect of time.
Figure 1 Theoretical delay-of-reinforcement gradients. The shorter and steeper delay gradient for ADHD (solid red line) implies that the relation between the response R5 and R will not be reinforced, while this relation will be reinforced with a normal delay gradient (broken blue line). The relation between R1 and R is close enough to be reinforced both when the delay gradient is short and when it is normal. A reinforcer will have almost the same effect on responses occurring immediately before reinforcer delivery with both a short and a long delay gradient, increasing the probability of repeating R with almost the same amount irrespective of the shape of the gradient.
There may be many behavioural consequences of a shorter delay gradient in ADHD (see [2,3] for details); one of them being that it will only allow for short behavioural sequences to be associated with reinforcement. Thus, when there is a short time interval passing from the occurrence of a particular response to the presentation of a reinforcer, or a short sequence of responses is quickly and contingently followed by a reinforcer, this sequence will be strengthened by reinforcement with equal probability in ADHD behaviour and in normal behaviour (given that the delay gradients are at the same height at the time of reinforcement, see Figure 1). However, when reinforcers are delayed or follow after a long behavioural sequence, only the behaviour occurring within the restricted time window will be associated with the reinforcement and thus be strengthened. This may affect the establishment of serial ordering of behavioural units, which is fundamental to all forms of skilled action, from speech to typing to reaching and grasping [10,11].
Skilled performance involves hierarchically organized units of behaviour, where the higher levels are controlled by longer-term consequences, and lower levels are controlled by short-term consequences of individual movements [12]. The hierarchical organization of behavioural units seems to combine autonomous functions at low levels with the possibility of learning new operations at higher control levels: "If the 'vital' centers of the lowest levels were not strongly organized at birth, life would not be possible; if the centers on the highest levels ('mental centers') were not little organized and therefore very modifiable we could only with difficulty and imperfectly adjust ourselves to the circumstances and should make few acquirements" (Taylor, 1932 [13], p 437, cited in [12], p. 701). Hypothetically, a narrow time window for associating actions with its consequences may, throughout ontogenesis, detain the natural evolution of hierarchically controlled behavioural units of increasing complexity. Further, with a short delay gradient, a discriminative stimulus will not systematically be associated with reinforcement and the establishment of stimulus control will be slowed, resulting in impulsivity and increased behavioural variability [2]. This style of learning may ultimately impede the development of verbally governed behaviour and "self control", and will have consequences for how the child with ADHD understands and behaves within his or her environment.
Purpose of the present study
The aims of the present study is to investigate the hypothesis that a short and steep delay gradient in ADHD will result in shorter and less predictable sequences in the behaviour of children with ADHD compared to controls, and to explore an untraditional way of investigating details in behavioural change. Traditional ways of analyzing behaviour in terms of means and standard deviations are too crude to identify moment-to-moment changes in behaviour, and may not reveal the behavioural dynamics underlying more global concepts like impulsivity and variability. The present study of behavioural sequences applies autocorrelations of consecutive responses as a means of studying moment-to-moment dynamics in responding. The data set was obtained from a study presented previously [14] and is presently analyzed in a different way.
The task was a computerized game where mouse clicks on one of two squares on the screen resulted in the presentation of a reinforcer. Reinforcers were delivered according to variable interval (VI) schedules, where responses result in a reinforcer after varying time intervals. With VI schedules a possible confusion of reinforcement effects with timing problems is avoided, as reinforcers are presented at unpredictable times [15]. All mouse clicks were recorded both in terms of where on the screen responses were placed (response location; spatial dimension) and response timing (temporal dimension). Thus, the present task allowed for analysis of both spatial and temporal aspects of behaviour.
Methods
Participants
The present study analyzed response data from 28 boys in the age range of 6:2–9:0 (yr:mo), 15 of whom had an ADHD diagnosis and 13 were healthy controls. These boys represented the young age group from the previous study. The older children (aged 9.5–12 yr) from that study were not included, as the comparison group showed to be inadequate (see [14], for discussion). The details of the recruitment and assessment procedures are presented elsewhere [14]; only an outline of group characteristics is provided here.
Participants in the ADHD group were referred from different clinical sources (school psychologists, child and adolescent psychiatric units, habilitation services, and a private specialist centre) and were included if they met the following criteria: 1) DSM-IV diagnosis of ADHD, of either three subcategories; 2) Full Scale IQ of at least 80; 3) no evidence of neurological disorder, psychosis, or pervasive developmental disorder; and 4) not taking any medication within the last 48 hours prior to testing. A diagnosis was confirmed after thorough clinical evaluation. Eight of the children had ADHD combined type, six had ADHD hyperactive / impulsive type, and one child had ADHD inattentive type. In addition, inclusion in the ADHD group warranted a score at or above the 95th percentile on the Disruptive Behaviour Rating Scale (DBRS) [16] home or school version.
Comparison children were recruited from schools in urban and suburban areas of Oslo, the capital of Norway. Inclusion criteria were the same as for the ADHD group, except that no DSM-IV diagnosis should be present. In addition, they had to score below the sub-clinical range on the DBRS.
Intellectual ability was assessed by screening all children with four subtests (information, similarities, block design, and picture completion) of the WISC-R [17] (demographic variables outlined in Table 1).
Table 1 Means, Standard Deviations, and t-Tests for Age, IQ, and Questionnaire Scores
Groups
Measure ADHD group
N = 15 Normal controls
N = 13 Group Comparisonf
Mean SD Mean SD
Age yr:mo (SD in mo) 7:6 9,5 7:10 9,4 p > .282, n.s.
- range yr : mo 6:2 – 8:9 6:4 – 9:0
IQ Full scale WISC-R 104.5a 10.5 114.5 12.5 p < .04
DBRSc Teacher
- inattention items 16.1 6.5 3.0 2.5 p < .001
- hyperactive/impulsive 18.4 8.0 2.0 2.5 p < .001
DBRSc Parents
- inattention items 16.7 6.0 4.1 2.1 p < .001
- hyperactive/impulsive 17.4 4.4 3.0 2.7 p < .001
CBCLd
- externalised T-score 68.2 8.9 38.9 7.5 p < .001
- internalised T-score 57.2 9.5 40.0 5.7 p < .001
- attention factor T-score 63.8 9.9 51.1 2.1 p < .001
TRFe
- externalised T-score 69.5b 10.9 46.1 6.8 p < .001
- internalised T-score 57.5b 9.5 45.0 7.2 p < .001
- attention factor T-score 60.5b 6.5 50.4 1.0 p < .001
aOne case missing
bTwo cases missing
cDisruptive Behaviour Rating Scale [16]
dChild Behaviour Checklist – parent form [30].
eTeacher Report Form [30].
fTwo-tailed t-test for equality of means. Bonferroni adjustment of the p-level is set to p = .039. Results are highlighted after Bonferroni adjustment
Procedure
The study was approved of by the Regional Medical Committee of Research Ethics. The parents of all the participants received written information about the study and gave written consent for their child to take part. All children were tested in quiet rooms, using the same tasks, apparatus, and test procedures (see [14] for details).
Reinforcement Task
The task was designed as a computer game, and was presented to the child with the following instruction (translated from Norwegian): "This is a game you may play now. It is a little strange, because I will not tell you how to play the game. Your task is to find out how the game works. You may use this mouse and move the arrow across the screen like this (experimenter demonstrates how to move the mouse). If you want to point, you can click with one of these buttons (experimenter points to the mouse buttons). You may talk while you are playing, but I will not answer any questions about the game. I will sit back here and write a little while you play. Do you understand your task? You may start now."
The task was run on a Toshiba Pentium 300 CDT laptop connected to a colour monitor (see [14] for details). In brief, response squares were two same-sized, aligned squares on the screen, one in a light and the other in a dark shade of grey. The computer mouse was the response device. Clicks with either right or left button on one of the squares induced a brief change in the grey shade as feedback. Responses outside the squares were recorded, but did not result in any feedback. The dark grey square was the "correct" target. Clicks within this square would, with varying time intervals, result in a cartoon picture (reinforcer) appearing on the screen for 1.5 s together with a sound. Responses on the light grey square never resulted in cartoon presentations. Following reinforcer delivery, the squares switched sides at random, keeping the total number of presentations on each side the same.
Variable interval (VI) schedules of reinforcement, where responses may produce reinforcers after the passage of varying time intervals [15], were used. Two VI schedules alternated, each signalled by a separate screen background colour. The background colour functioned as the conditioned discriminative stimulus for the specific condition in operation, while the dark grey colour of the correct square was the discriminative stimulus for the reinforcer. The two schedules were a short VI (VI 2s) signalled by a navy blue background and a long VI (VI 20s) signalled by a bright yellow background.
There were two sessions, each of five segments. Each segment consisted of four short and four long intervals, and was terminated by a response and the delivery of a reinforcer. In the first session, the child would see a total of 40 reinforcers (cartoons). In the second session, the child received a small tangible reinforcer (trinket, coin or sweet) in addition to the cartoon picture. This was done in order to maintain reinforcer value. The entire task, including instruction, break between sessions, and a final, short interview with the child, took less than 30 min to complete.
Data recording and statistics
Data were recorded by the laptop. Response side (left or right), response coordinates (i.e. the horizontal and vertical pixel that the tip of the arrow-shaped cursor touched when the child clicked a mouse button), and interresponse times (IRTs) were the recorded dependent measures. The individual IRT distributions were highly skewed with a long tail towards long IRTs. IRTs were therefore normalized by log transformations prior to analysis (logIRT = log10 (IRT/1000 + 0.001)).
Behavioural measures
Data from the VI 20s condition was used to study response sequences, as the short schedule only allowed for one or a few responses before a reinforcer was delivered. Predictability of responses over long sequences could theoretically be found according to different aspects of the behaviour, and in order to explore the different possibilities we computed three measures related to the spatial dimension and one measure related to timing. The first measure was a general side response pattern, i.e., whether consecutive responses were on the left or right side of the screen. Highly predictable responding would probably be related to the side where the correct target was positioned, and would thus be a complementary measure of stimulus control. Likewise, low predictability implies that responses are equally distributed on the two sides and is thus also a measure of low stimulus control. Next, due to the fact that reinforcers affect more responses than the one that produces it (Figure 1), predictable patterns in other aspects of response locations were explored. The square response pattern measure was based on the distance from the centre of the selected square, whether correct or not, to the spot where the response was placed. The target response pattern measure was based on the distance from the centre of the correct response target to where the response was placed. Both distance scores were in terms of pixels, with the centre of the square defined as 0,0. Finally, response sequences might be predicted by patterns in timing. Thus, timing response patterns were analyzed based on consecutive interresponse times (IRTs).
The ordering of responses spatially and temporally was assessed by autocorrelations. Autocorrelations (serial correlations) of each measure were correlations of consecutive values over five lags (correlations between n and n+1 response is the first lag, between n and n+2 response is the second lag, and so on up to correlations between n and n+5 response being the fifth lag). The autocorrelations were computed for each individual over sessions and segments. We predicted that the behaviour of children with ADHD would be characterized by lower autocorrelations and that their autocorrelation curves across lags would be steeper compared to the behaviour of healthy comparison children.
Statistics
Data were analyzed by means of SPSS 11.0 for Windows (SPSS) and Statistica 6.1 [18] program packages. The distance scores were computed as the square root of the sum of squared horizontal and vertical distances. Explained variance (autocorrelations squared) was analyzed using repeated measures ANOVA over sessions, segments, and lags. The ANOVA was supplemented with MANOVA. A multivariate approach to repeated measures of more than two levels is recommended because it bypasses the assumption of compound symmetry and sphericity [18]. Clinical group (2) was the between-group variable; and session (2), segment (5), and lag (5) were within-group variables.
Demographic data
Demographic and diagnostic measures of the ADHD and the comparison group were tested with two-tailed t-tests for equality of means and are displayed in Table 1. There were significant differences between groups on all measures including IQ, but not age. Whether to control for IQ difference has been debated (e.g., [19]) as undue weight may be put on the impact of IQ and remove variance that is a result of ADHD itself. Running the analyses with IQ as a covariate and without gave a similar overall picture of results. Thus, IQ was not included as a covariate in the reported analyses. (All non-published results may be obtained from the first author upon request.)
Results
In general, acquisition of predictable response sequences was found in the spatial measures and not in the temporal measure. The ADHD group had significantly lower autocorrelations than the comparison group on two of the three spatial measures. Both groups had very low autocorrelations related to response timing. There was no Session effect on the three spatial measures, but the main effect of Lag was significant for all four measures. Results of the planned analyses with ANOVA and MANOVA are shown in Table 2.
Table 2 Results from repeated measures ANOVA and multivariate tests for repeated measures, of explained variance (squared autocorrelations)1
Measure Variable ANOVA Multivariate
Df F Df F
Side Response Pattern Group (G) 1, 26 14,700***
Session (Ses) 1, 26 1,136 1, 26 1,136
Segment (Seg) 4, 104 11,684*** 4, 23 5,317**
Lag 4, 104 143,735*** 4, 23 42,369***
G * Seg 4, 104 4,756*** 4, 23 2,824*
G * Seg * Lag 16, 416 1,810* 16, 11 1,675
Square Response Pattern G 1, 26 10,981**
Ses 1, 26 2,505 1, 26 2,505
Seg 4, 104 1,290 4, 23 0,721
Lag 4, 104 91,581*** 4, 23 28,576***
G * Lag 4, 104 8,779*** 4, 23 5,097**
G * Ses * Lag 4, 104 2,812* 4,023 1,681
Target Response Pattern G 1, 26 3,083
Ses 1, 26 1,061 1, 26 1,061
Seg 4, 104 4,127** 4, 23 3,131*
Lag 4, 104 201,232*** 4, 23 60,921***
G * Lag 4, 104 1,232 4, 23 3,172*
Timing Response Pattern G 1, 26 0,155
Ses 1, 26 6,793** 1, 26 6,793**
Seg 4, 104 2,125 4, 23 1,253
Lag 4, 104 41,613*** 4, 23 13,364***
G * Ses 1, 26 5,116* 1, 26 5,116*
G * Ses * Lag 4, 104 4,326** 4, 23 2,499
* : p < .05; **: p < .01; ***: p < .001
1Due to a very large number of possible interactions, only main effects and significant interactions involving the group variable are reported. All non-published results may be obtained from the first author.
Side response pattern
This measure assessed whether responding predictably would continue on the same side or vary unpredictably between sides, irrespective of on which side the correct response target was displayed (see Behavioural measures in Methods for a detailed description of the variables). Highly predictable responding over lags would imply that behaviour was ordered in sequences related to side. Less variance was accounted for in the ADHD group than in the comparison group. For the ADHD group, explained variance in the first lag across segments was low (range 0.33 > mean R2 > 0.22, median R2 = 0.26), while it was in the upper range for the comparison group (0.62 > mean R2 > 0.35; median R2 = 0.50) (Figure 2). Actually, in some segments, explained variance over all five lags was higher for the comparison group than in the first lag for the ADHD group. In addition, explained variance in these segments did not descend much from the first to the fifth lag (not shown), indicating highly predictable response sequences for up to six responses for the comparison group.
There were significant main effects of Group, Segment, and Lag (Table 2). In addition, there was a significant two-way interaction between Group and Segment. In the comparison group there was a general within-session upward trend from segment 1 to 4, indicating a learning effect, but this trend did not continue into the last segment of each session (see Figure 2). The ADHD group did not show a similar pattern; explained variance did not improve during the task. This group difference was supported by a significant three-way ANOVA interaction between Group, Segment, and Lag. However, this interaction was not confirmed by the MANOVA.
Figure 2 Response pattern according to side of the screen. Predictability of which side of the screen consecutive responses were placed, depicted as mean explained variance (autocorrelations squared), by segments (1–5) and lags (1–5 per segment), for ADHD and comparison groups. Graphs show means of session 1 and session 2.
Square response pattern
This measure assessed to what degree the children tended to respond in any predictable pattern in terms of the distance between responses, anchored to the centre of the square that responses were placed within (whether it was the correct response target or not). Highly predictable responding would imply that behaviour was ordered in sequences of similar distances between responses. Again, the explained variance for the ADHD group was lower than for the comparison group (Figure 3). Explained variance in the first lag for the ADHD group were in the low range (0.25 > mean R2 > 0.13; median R2 = 0.18), indicating that there was rather low predictability from one response to the next. For the comparison group, explained variance in the first lag was higher (0.51 > mean R2 > 0.22; median R2 = 0.42).
There were significant main effects of Group and Lag, but not of Session and Segment (Table 2). A significant ANOVA interaction between Group and Lag was confirmed by the multivariate analysis, while the significant interaction between Group, Session, and Lag was not confirmed by the MANOVA.
Figure 3 Response pattern according to distance from the centre of a square. Predictability of distance from the centre of the chosen square, whether correct or not, to where on the screen consecutive responses were placed. Curves show mean explained variance (autocorrelations squared) by segments (1–5) and lags (1–5 per segment), for ADHD and comparison groups. Graphs show means of session 1 and session 2.
Target response pattern
This measure assessed patterns of response placements in terms of distance from the centre of the correct square. Highly predictable responding would imply that behaviour was ordered in sequences of similar distances between responses, specifically related to the centre of the correct square. Again, low explained variance indicated high variability in responding. Explained variance in the first lag across segments of the ADHD group was in the middle range (0.46 > mean R2 > 0.27; median R2 = 0.40), and slightly higher for the comparison group (0.60 > mean R2 > 0.36; median R2 = 0.41) (Figure 4).
The main effects of Segment and Lag were statistically significant, but not the main effects of Group or of Session (Table 2). The MANOVA, but not the ANOVA, showed a two-way interaction between Group and Lag. There were no other significant interactions. As can be seen in Figure 4, the curves are quite similar for the two groups, but explained variance tends to be lower in lags 2–5 in the ADHD group compared to the comparison group.
Figure 4 Response pattern according to distance from the centre of the correct target. Predictability of distance from the centre of the correct target to where on the screen consecutive responses were placed. Curves show mean explained variance (autocorrelations squared) by segments (1–5) and lags (1–5 per segment), for ADHD and comparison groups. Graphs show means of session 1 and session 2.
Timing response pattern
The development of patterns in response timing was investigated by means of consecutive interresponse times (IRTs). Explained variance of the first lag was generally very low and not significantly different between the groups (0.06 > mean R2 > 0.12; median R2 = 0.06 for ADHD, and 0.16 > mean R2 > 0.02; median R2 = 0.1 for comparisons). The significant main effect of Session showed that explained variance was lower in the second session than in the first, particularly for the comparison group. The main effect of Lag was significant, while the main effect of Segment was not. There was a significant interaction effect between Group and Session, showing that while the ADHD group had lower explained variance than the comparison group in the first session, the comparison group had lower explained variance than the ADHD group in the second session. The significant interaction between Group, Session, and Lag in the ANOVA was not confirmed by the MANOVA.
Relation to clinical scores
The dynamic developmental theory (DDT) argues that a shortened delay gradient probably relates more to the hyperactive / impulsive and the combined subtypes of ADHD than to the inattentive subtype [2]. The present sample did not allow for a differential analysis of clinical subtypes. However, the relation between the individual scores on the sub-dimensions of ADHD and explained variance on the spatial measures could be computed and would indicate if either the hyperactive / impulsive dimension or the inattentive dimension were better predictors of the explained variance. Thus, mean explained variance across lags for each spatial measure was correlated with sum scores of the inattentive and the hyperactive / impulsive dimensions on the DBRS parent form [16]. The correlations showed an inverse relationship between scores on the rating scale and predictability of responses, indicated by negative values (Figure 5). Correlations between the explained variance and the two sub-dimensions on the DBRS were not very different, but explained variance from session 2 was better than session 1 as a predictor of scores on the DBRS. The explained variance of the side response pattern was the best predictor of scores on the DBRS, with increasing correlation over lags (Figure 5, solid line).
Figure 5 Relation between explained variance of responding over lags and clinical scores. A correlogram showing the relation between mean explained variance (autocorrelations squared) by lags and scores on the hyperactive / impulsive items of the Parent form of the Disruptive Behaviour Rating Scale (DBRS) for the three spatial measures in session 2. The relation was negative; i.e. high scores on the DBRS predicted low scores on the autocorrelations.
Discussion
The present study investigated the predictability of behavioural sequences in ADHD and in comparisons. The aims of the study was both to investigate the hypothesis that a shortened delay gradient in ADHD would result in short and less predictable response sequences in the behaviour of children with ADHD [2], and to explore the use of autocorrelations as a way of analyzing details in behavioural dynamics. Consecutive responding was studied in terms of three spatial and one temporal response dimensions. The results showed that predictable response sequences did develop according to the spatial response dimensions, but not according to the temporal dimension. Importantly, on the spatial dimensions, predictability of response sequences was considerably lower for the ADHD group than for the comparison group, as shown by significant interactions involving group and lag (number of consecutive responses). In addition, the overall explained variance of the ADHD group responding was significantly lower than that of the comparison group according to side response pattern and square response pattern, but not according to target response pattern or timing response pattern. This was supported by the high correlations between variance accounted for in consecutive response lags of these two spatial measures and the scores on both hyperactive/impulsive and inattentive clinical dimensions (Figure 5). Thus, the disorganized behaviour observed in the ADHD group may be a general behavioural feature captured by the clinical scoring by teachers and parents.
The side response pattern assessed whether responding could be predicted according to side (left or right) of the screen, irrespective of on which side the correct target was displayed. The significant group difference implied a highly predictable pattern in the comparison group, indicating that these children varied their responding between sides almost only to the extent that the correct target square switched sides on the screen. This was supported by their mean percent correct responding at 87% during stable state [14], demonstrating good discriminative control. The ADHD group, however, varied their responding between sides even though the correct target square was still on the same side, and they never exceeded 61% correct responses [14]. The comparison group showed predictable sequences of up to six responses where variance accounted for was larger for the n+5 response than for n+1 response in the ADHD group (e.g. Segments 2 and 3 in Figure 2).
However, there was a drop in explained variance in the comparison group at the end of each session (Figure 2, segment 5). This may have been an effect of the schedule, as the VI schedule was made up of predefined interval lengths, with more of the short intervals in the beginning of the session and more of the longer intervals towards the end. Hence, the disappearance of regular, predictable responding seen in the comparison group may have been the result of inter-reinforcement intervals being very long. This effect was not seen in the ADHD group.
The square response pattern and the target response pattern were both computed in order to explore possible patterns related to the spatial distance between responses, anchored to the centre of the squares. Organizing responding within the squares was not differentially reinforced. However, a shortened time window available for strengthening connections between events, as suggested by the DDT [2], predicts less systematic response patterns in ADHD compared to normal, which was found. The square response pattern measured distances from the centre of the chosen square to where the response was placed. The comparison group showed significantly more predictable responding than the ADHD group, both overall and across lags, indicated by the significant interaction between group and lag. The target response pattern measured the distance from the centre of the correct square to where the response was placed. The similar magnitude of explained variance (about 40–50%) in the first lag of the two groups indicates high predictability from response n to n+1 when in the correct square, while the ADHD group varied more on consecutive responses as indicated by the significant interaction between group and lag.
Increased variability is a consequence of reduced stimulus control. This is seen in the ADHD group in terms of the side response pattern. Therefore, it might be argued that the reduced predictability found in both square response pattern and target response pattern is a consequence of larger arm movements in the ADHD group because of more varied responding from side to side, rather than an effect of inefficient reinforcement of response placement within a square. The present analysis did not allow identifying predictability of response placements when consecutive responding was within the same square, which involves smaller movements. However, disentangling the variability related to larger arm movements and the variability related to decreased stimulus control (revealed as more varied responding) may not be feasible. Further, it may be speculated that, since the striatum, which is involved in the planning and execution of motor actions, and the nucleus accumbens, which is involved in learning and reinforcement, both receive important dopaminergic afferents, these functional processes may both be impaired in an individual with ADHD [2].
Acquisition of functional behavioural sequences may be related to processes involved in habit learning. Habit learning is characterized by a transition from response-consequence associations that are flexible and sensitive to reinforcement devaluation, to stimulus-response associations that are less flexible and sensitive (e.g. [20]). Thus, the initial part of habit learning may mainly involve activity in the mesolimbic dopaminergic branch, while the established habit and the skilled execution of the motor sequence may mainly involve the nigrostriatal dopaminergic branch. It may be argued that the present study is mainly concerned with acquisition, since the task was new and relatively short. On the other hand, animal model studies have indicated that operant learning rapidly become habitual when the contingency between the response and reinforcer is weakened by using interval schedules [20], as used in the present study. Hence, it might be speculated that the control group rapidly developed a habit, while this process was hampered in the ADHD group. Whether the present findings are due to impairments in habit formation or motor control related to the striatum, or to learning deficits related to nucleus accumbens cannot be settled, but the DDT predicts dysfunction of both processes [2].
Explained variance of first lag (and following lags) was too low to conclude that there was any predictability in IRTs. Thus, there was no timing response pattern. Speculatively, this lack of a predictable timing pattern could be related to the visuo-spatial nature of the task, both in terms of response alternatives (where, not when) and in terms of the reinforcer. There is some evidence that striatal neurons involved in sequential habit learning may encode visuo-spatial information rather than temporal [21].
The present findings suggest that the learning of coherent and predictable behavioural sequences will be difficult in children with ADHD, probably because the time window available for reinforcers to work is narrower in ADHD compared to normal. Skilled performance is characterized by responses emitted with brief interresponse intervals, smooth transitions between responses, and efficient coordination of consecutive movements so that a whole sequence is conducted in a predictable manner (e.g. [22]). Reinforcers are actively involved in the selection and shaping of responses, in chunking discrete responses into larger behavioural units, and in establishing relations between antecedent stimuli and behavioural units into sequences that together constitute a complete action (e.g. sequences of behavioural units comprising the entire action of typing a word, writing your signature, or playing an arpeggio on the piano). With a shorter delay gradient, the whole process of organizing hierarchical structures of actions comprised of functional units of behaviour may be hampered in ADHD. Although speculative, the present results indicate a different style of learning in ADHD, probably brought about by inefficient dopaminergic processes, which might be regarded as a separate endophenotype of ADHD [3,23] that forms the development of behavioural characteristics of variability, impulsiveness, lack of goal-directed behaviour, and hampered development of self-control. Such a learning style may explain the heterogeneity in symptom presentations among individuals with ADHD, because the behaviour of different individuals will be the result of interactions with different environmental contingencies.
Other interpretations are possible, however. It has been suggested that the length of delay gradients may be dependent on working memory (WM) capacity because the ability to relate sensory information to responses and reinforcing stimuli seems to correlate with ongoing neuronal activity in prefrontal cortex [24]. Behavioural sequences or serial movement has been related to WM capacity, as it has been argued that serial sensory information is stored in WM and converted into a movement program with the help of visual stimuli [25]. There is obviously some kind of memory process involved in reinforcement, and the delay gradient may as well be described as the result of pairing the reinforcer with the fading of precursors, e.g., the fading of memory traces of the behaviour [3,26]. For the present analysis, it is not critical if a shorter delay gradient in ADHD is caused by reduced WM capacity or if both are expressions of underlying dysfunctioning dopamine systems. Further, WM capacity is not necessarily a fixed entity and may also be modified by learning: a recent study reports that WM was improved in ADHD children by computerized training [27]. The present findings indicate that the delayed learning of a new task is related to reduced predictability in consecutive responding of ADHD individuals and not, for instance, to increased activity in general [14] or to increased perseveration, which would be the opposite of low predictability.
There might be alternative motivational explanations (than a shorter delay gradient) for the behaviour observed in the ADHD group. Rather than a reduced effect of positive reinforcement (i.e., the delivery of a reinforcer), the behaviour may be a result of reduced negative control (i.e., reduced compliance). However, non-compliance would have involved refusal to complete the task, which none of the children did, and most of them even reported that the task was fun. In addition, the instruction was non-directive, excluding the possibility of non-compliance to any instructed response pattern. A group of children aged 9–12 yr also participated in the original study, but the results of the comparison group indicated that their responding was mainly compliant and not controlled by reinforcers, as shown by lack of schedule control [14]. The behaviour of the younger children that participated in the present analysis did show schedule control, but the response patterns were different.
The present study provides new insights into the recently started discussion of intra-individual ADHD-related variability (ARV; [28]). Castellanos et al [28] calls for systematic studies of moment-to-moment changes in ongoing behaviour and to integrate such studies into causal models. The present study shows that ARV is found in other domains than reaction times, that it may be influenced by reinforcement contingencies, and that it might be explained by a short and steep delay gradient. Although the present study is purely behavioural, it might provide a framework within which to analyze several parallel processes, including cardiovascular and neurophysiologic, as suggested by Castellanos et al [28].
Some limitations of the present study could be improved in future studies. First, the findings need to be replicated in a larger sample of children. New samples, including girls and children of other ages, need to be analyzed in order to verify if this behavioural style is general to ADHD, whether it may be identified at an earlier age, and whether it continues to constitute an important factor in the behaviour of older children. Children with other psychopathologies should be added in order to establish the specificity of this learning style. Second, the present task may be criticized for its lack of ecological validity. On the other hand, the present task may have been optimal for studying ordering and sequencing of responses, as no earlier experience or learning history would interfere. In the future, a wide range of serial or sequential tasks should be investigated because the dynamic developmental theory of ADHD predicts that the presently-observed learning style should be found in the behaviour of people with ADHD across tasks and activities. Finally, various serial tasks could be conducted during brain imaging in order to investigate the brain areas involved.
Conclusion
The present findings provide support for the dynamic developmental theory of ADHD predicting that a short and steep delay-of-reinforcement gradient will result in fewer responses in a predictable sequence than when the delay gradient is normal [2]. The hypothesized difference in timing patterns did not appear in the present study, as none of the groups showed any predictability in timing patterns.
The present study represents a new approach to analyzing the micro-dynamics of consecutive responses in a moment-to-moment manner. A majority of the responses were emitted with IRT < 1s. The study of behavioural processes and their environmental correlates may thus approach the time-scale of brain processes and we may get closer to directly measure brain-behavioural interactions. In this perspective, concepts like working memory, inhibition, and even timing problems may be too wide concepts to identify the underlying learning style, because a learning style characterized by inefficient chunking or development of entire behavioural sequences may be manifest in different domains in different individuals. One might speculate that sequencing of motor action and sequencing of speech and thought may be of the same functional origin (cf., [29]). Thus, sequencing deficits in ADHD may cause problems with rule-governed behaviour and self-control typical in ADHD behaviour.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
HA participated in the development of the study design, development of the reinforcement task used, carried out the data collection, prepared the data, performed the statistical analyses, and wrote the manuscript. TS participated in the development of the study design, development of the reinforcement task, wrote the programs for statistical analyses, participated in data analyses, read the manuscript and approved the final draft.
Acknowledgements
The present study was supported by grants from The National Council for Mental Health – Norway (Heidi Aase) and from The University of Oslo. We thank Professor Edmund Sonuga-Barke, University of Southampton, for valuable discussions during the development of the reinforcement task, and Mr. Martin Hall, University of Southampton, for programming it. Professor Peter Killeen contributed with valuable insights during data analyses.
==== Refs
American Psychiatric Association Diagnostic and statistical manual of mental disorders: DSM-IV 1994 4 Washington, D.C.: Author
Sagvolden T Johansen EB Aase H Russell VA A dynamic developmental theory of Attention-Deficit /Hyperactivity Disorder (ADHD) predominantly hyperactive/impulsive and combined subtypes Behav Brain Sci 2005 28 397 419 16209748
Johansen EB Sagvolden T Aase H Russell VA The dynamic developmental theory of attention-deficit/hyperactivity disorder (ADHD): Present status and future perspectives Behav Brain Sci 2005 28 451 454 16209770
Yan JH Thomas JR Arm movement control: differences between children with and without attention deficit hyperactivity disorder Res Q Exerc Sport 2002 73 10 18 11926479
Yan JH Thomas KT Stelmach GE Thomas JR Developmental differences in children's ballistic aiming movements of the arm Percept Mot Skills 2003 96 589 598 12776841
Sheppard DM Bradshaw JL Georgiou N Bradshaw JA Lee P Movement sequencing in children with Tourette's syndrome and attention deficit hyperactivity disorder Mov Disord 2000 15 1184 1193 10.1002/1531-8257(200011)15:6<1184::AID-MDS1018>3.0.CO;2-N 11104203
Kalff AC de Sonneville LM Hurks PP Hendriksen JG Kroes M Feron FJ Low- and high-level controlled processing in executive motor control tasks in 5-6-year-old children at risk of ADHD J Child Psychol Psychiatry 2003 44 1049 1057 10.1111/1469-7610.00189 14531587
Siklos S Kerns KA Assessing multitasking in children with ADHD using a modified Six Elements Test Arch Clin Neuropsychol 2004 19 347 361 10.1016/S0887-6177(03)00071-4 15033221
Schultz W Dayan P Montague PR A neural substrate of prediction and reward Science 1997 275 1593 1599 10.1126/science.275.5306.1593 9054347
Lashley KS Jeffress LA The problem of serial order in behavior Cerebral mechanisms in behavior 1951 New York: Wiley 112 136 (Reprinted in The neuropsychology of Lashley. Edited by Beach FA, Hebb DO, Morgan CT, & Nissen HW. New York: McGraw-Hill 1960:506–528)
Averbeck BB Chafee MV Crowe DA Georgopoulos AP Parallel processing of serial movements in prefrontal cortex Proc Natl Acad Sci U S A 2002 99 13172 13177 10.1073/pnas.162485599 12242330
Rhodes BJ Bullock D Verwey WB Averbeck BB Page MP Learning and production of movement sequences: behavioral, neurophysiological, and modeling perspectives Hum Mov Sci 2004 23 699 746 10.1016/j.humov.2004.10.008 15589629
Taylor J Ed Selected writings of John Hughlings Jackson 1932 London: Hodder and Stoughton
Aase H Sagvolden T Infrequent, but not frequent, reinforcers produce more variable responding and deficient sustained attention in young children with Attention-Deficit/Hyperactivity Disorder (ADHD) J Child Psychol Psychiatry in press 10.1111/j.1469-7610.2005.01468.x
Catania AC Learning 1998 4 N.J. Englewoods Cliffs: Prentice Hall, Upper Saddle River
Barkley RA Murphy KR Attention-Deficit Hyperactivity Disorder: A Clinical Workbook 1998 2 New York: Guilford Press
Wechsler D Wechsler Intelligence Scale for Children – Revised 1974 Trondheim, Norwegian Psychology Association
StatSoft Inc STATISTICA for Windows [Computer software] 1974 Tulsa, OK: StatSoft Inc
Seidman LJ Biederman J Faraone SV Weber W Ouellette C Toward defining a neuropsychology of attention deficit-hyperactivity disorder: performance of children and adolescents from a large clinically referred sample J Consult Clin Psychol 1997 65 150 160 10.1037//0022-006X.65.1.150 9103744
Yin HH Knowlton BJ Balleine BW Lesions of dorsolateral striatum preserve outcome expectancy but disrupt habit formation in instrumental learning Eur J Neurosci 2004 19 181 189 10.1111/j.1460-9568.2004.03095.x 14750976
Schmitzer-Torbert N Redish AD Neuronal activity in the rodent dorsal striatum in sequential navigation: separation of spatial and reward responses on the multiple T task J Neurophysiol 2004 91 2259 2272 10.1152/jn.00687.2003 14736863
Matsumoto N Hanakawa T Maki S Graybiel AM Kimura M Role of [corrected] nigrostriatal dopamine system in learning to perform sequential motor tasks in a predictive manner J Neurophysiol 1999 82 978 998 10444692
Coghill D Nigg J Rothenberger A Sonuga-Barke E Tannock R Whither causal models in the neuroscience of ADHD? Dev Sci 2005 8 105 114 10.1111/j.1467-7687.2005.00397.x 15720368
Tannock R Hypodopaminergic function influences learning and memory as well as delay gradients Behav Brain Sci 2005 28 419 451 16209749
Ohbayashi M Ohki K Miyashita Y Conversion of working memory to motor sequence in the monkey premotor cortex Science 2003 301 233 236 10.1126/science.1084884 12855814
Killeen P Gradus ad Parnassum Behav Brain Sci 2005 28 419 451 16209749
Klingberg T Fernell E Olesen PJ Johnson M Gustafsson P Dahlstrom K Computerized training of working memory in children with ADHD – a randomized, controlled trial J Am Acad Child Adolesc Psychiatry 2005 44 177 186 10.1097/00004583-200502000-00010 15689731
Castellanos FX Sonuga-Barke EJ Scheres A Di Martino A Hyde C Walters JR Varieties of attention-deficit/hyperactivity disorder-related intra-individual variability Biol Psychiatry 2005 57 1416 1423 10.1016/j.biopsych.2004.12.005 15950016
Aldridge JW Berridge KC Coding of serial order by neostriatal neurons: a "natural action" approach to movement sequence J Neurosci 1998 18 2777 2787 9502834
Achenbach TM Howell CT Quay HC Conners CK National survey of problems and competencies among four- to sixteen- year-olds: parents' reports for normative and clinical samples Monogr Soc Res Child Dev 1991 56 1 131 1770964
|
16060963
|
PMC1208851
|
CC BY
|
2021-01-04 22:25:38
|
no
|
Behav Brain Funct. 2005 Aug 1; 1:12
|
utf-8
|
Behav Brain Funct
| 2,005 |
10.1186/1744-9081-1-12
|
oa_comm
|
==== Front
Behav Brain FunctBehavioral and brain functions : BBF1744-9081BioMed Central London 1744-9081-1-131609113810.1186/1744-9081-1-13ResearchFast, visual specialization for reading in English revealed by the topography of the N170 ERP response Maurer Urs [email protected] Daniel [email protected] Bruce D [email protected] Sackler Institute, Weill-Cornell Medical College, Box 140, 1300 York Ave. New York, NY, 10021, USA2 Department of Child and Adolescent Psychiatry, Brainmapping Research, University of Zurich, Neumunsterallee 9, CH-8032 Zurich, Switzerland2005 9 8 2005 1 13 13 18 3 2005 9 8 2005 Copyright © 2005 Maurer et al; licensee BioMed Central Ltd.2005Maurer 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
N170 effects associated with visual words may be related to perceptual expertise effects that have been demonstrated for faces and other extensively studied classes of visual stimuli. Although face and other object expertise effects are typically bilateral or right-lateralized, the spatial topography of reading-related N170 effects are often left-lateralized, providing potential insights into the unique aspects of reading-related perceptual expertise.
Methods
Extending previous research in German [1], we use a high-density channel array to characterize the N170 topography for reading-related perceptual expertise in English, a language with inconsistent spelling-to-sound mapping. N170 effects related to overall reading-related expertise are defined by contrasting responses to visual words versus novel symbol strings. By contrasting each of these conditions to pseudowords, we examined how this reading-related N170 effect generalizes to well-ordered novel letter strings.
Results
A sample-by-sample permutation test computed on word versus symbol ERP topographies revealed differences during two time windows corresponding to the N170 and P300 components. Topographic centroid analysis of the word and symbol N170 demonstrated significant differences in both left-right as well as inferior-superior dimensions. Words elicited larger N170 negativities than symbols at inferior occipito-temporal channels, with the maximal effect over left inferior regions often unsampled in conventional electrode montages. Further contrasts produced inferior-superior topographic effects for the pseudoword-symbol comparison and left-lateralized topographic effects for the word-pseudoword comparison.
Conclusion
Fast specialized perception related to reading experience produces an N170 modulation detectable across different EEG systems and different languages. Characterization of such effects may be improved by sampling with greater spatial frequency recordings that sample inferior regions. Unlike in German, reading-related expertise effects in English produced only partial generalization in N170 responses to novel pseudowords. The topographic inferior-superior N170 differences may reflect general perceptual expertise for orthographic strings, as it was found for words and pseudowords across both languages. The topographic left-right N170 difference between words and pseudowords was only found in English, and may suggest that ambiguity in pronunciating novel pseudowords due to inconsistency in spelling-to-sound mapping influences early stages of letter string processing.
==== Body
Background
The N170 is a component of the event-related potential (ERP) peaking between 150 and 200 ms and showing an occipito-temporally negative and fronto-centrally positive topography. It is strongly elicited by certain classes of visual stimuli, such as faces [2,3], relative to other visual control stimuli. Investigations of the psychological principles that drive the N170 to respond more strongly to some classes of stimuli over others have demonstrated perceptual expertise effects across several classes of stimuli, including enhanced N170 responses (relative to other object control stimuli) for bird experts viewing birds [4], car experts viewing cars [5], and has even been demonstrated for laboratory-induced expertise with 3D novel figures ("greebles" [6]). These results support a potential relationship between extensive visual experience with a stimulus domain and alterations in visual processes within the first 200 ms of perceptual identification. This framework of perceptual expertise may also account for experience-dependent changes in reading skill – a domain in which extensive practice develops considerable visual expertise at the level of letter-strings and the pattern by which letters typically are combined to create visual word forms [7].
Neurophysiological studies have shown that skilled adult readers develop fast, perceptual identification processes that are specialized for words and other letter strings, reflected by differences in N170 responses compared to control stimuli, such as symbol strings, that control for visual features [1,8-10].
Unlike findings of right-lateralized or bilateral N170 responses for faces, N170 responses to word stimuli showed a left-lateralized topography [1,3,11-13]. Across studies, however, the degree of the left-lateralization varied between strong [3,12] and moderate [1,13].
Some studies also showed that the N170 is sensitive for linguistic processing [14,15]. Consonant strings had larger N170 amplitudes than words [14,15], and sublexically irregular pseudowords were in between [14]. Other studies, however, did not find N170 differences between words and pseudowords [1,8,16]. In one study, the differences between consonant strings and words were only found for lexical and semantic tasks, but not during implicit reading [8], whereas it was found across semantic, passive, and implicit viewing in another [14]. These results suggest that N170 responses are somewhat variable across experiments, which might be due to different task demands and presentation modes.
Word frequency effects in the N170 were more consistently found across studies, with low frequency words producing more negative N170 amplitudes [17-21] (but see also [22]).
One recent study of native German speaking adults demonstrated sublexical N170 effects by contrasting responses to strings of novel symbols to letters strings grouped into common orthographic patterns (familiar or novel word forms in German), demonstrating that specialization of processing as assessed by either lexical or sublexical strings is strongest at inferior occipito-temporal channels [1]. However, the region that demonstrated the peak effect was at the edge of the montage of electrodes applied, suggesting that this effect might be better characterized by an electrode array, which covers regions inferior to the classical 10–20 or 10-10 electrode montages [23,24].
The aim of the current study was to further characterize the nature of reading-related N170 expertise effects by applying a 129-channel array (geodesic sensor net, Electrical Geodesics, Inc.) that extends the coverage of the classical 10-10 montage [23,24] to more inferior regions, thereby providing adequate spatial sampling of the peak effect of interest (Fig. 1). This study adopts a paradigm used in a previous N170 study conducted with German speaking subjects [1] to explore potential replicability of effects across languages that differ in the level of consistency of how letters map onto word sounds [25], as well as to examine the topography of the N170 with a greater spatial sampling of inferior regions that might be critical to capturing effects produced near ventral posterior brain regions (see [7] for review). EEG was recorded continuously as participants actively monitored for an occasional target, defined as an immediately repeated item (i.e. "one-back"), among a series of word, pseudoword, or symbol string stimuli. Advantages of this paradigm include relatively equal engagement for all classes of stimuli, which can be assessed via behavioral responses to targets, as well the segregation of response-free trials from infrequent target trials within ERP analyses.
Figure 1 High-density 129-channel montage. Filled black dots indicate electrodes corresponding to the 10–20 system positions [34]. Inferior occipito-temporal channel groups used for waveform illustration and additional lateralization analyses are marked with dotted circles. Note that the high-density montage extends both the 10-10 and 10-5 montages [24] for an additional inferior row (approx. 5% of the Nasion-Inion distance).
Results
Behavior
Participants detected targets with a high accuracy across all conditions (>90% in each condition, see table 1), indicating that all conditions were relatively easy. Subtle condition differences were detectable, however, via repeated measure ANOVA analyses, which revealed a main effect of stimulus condition for accuracy (F(2,13) = 6.23, p < 0.05), but not for reaction time (F(2,13) = 0.61, p = ns). Symbol strings were detected slightly less accurately than the other two conditions.
Table 1 Behavioral results for detecting targets
Words Pseudowords Symbols
Accuracy (% correct) 98.4 95.5 90.1
Reaction time (ms) 609 599 542
Word-symbol differences in consecutive ERP maps
To assess differential processing of words and symbol strings over time, a Topographic Analysis of Variance (TANOVA, [26]) on non-normalized (raw) ERP maps was computed for each time point. TANOVA on raw maps detects all systematic amplitude differences between two maps (i.e. including all 129 electrodes). Accordingly, word and symbol processing differed (p < 0.01, to adjust for multiple comparisons) during two separate time windows, from 160–244 ms and from 324–512 ms, largely overlapping with the Global Field Power (GFP [27]) peaks of the N170 and P300 ERP components (Fig. 2).
Figure 2 A. Point-to-point differences (TANOVA) between word and symbol ERP maps superimposed with Global Field Power. Significant differences (black bars) between word and symbol processing were found in two time windows corresponding to the N170 and P300 GFP components. B. N170 maps for words and symbols and their difference t-map. Words elicited larger N170 negativity than symbols at inferior occipito-temporal channels, especially over the left hemisphere. Note that the channels with the largest negative difference are located inferior to the classical 10-10 montage system.
N170 time window
Our approach to analyzing topographic effects in the N170 time window follows an ERP mapping approach [28,29], designed to take full advantage of information from all the channels in the high-density channel array. According to this approach, ERPs are seen as a series of maps changing in Global Field Power (GFP [27]) and topography over time. Moreover, ERP topographies tend to remain stable for short periods of time, typically changing at time points with low GFP. To get a robust measure for the N170 component, we averaged samples across the time segment between the two GFP minima (based on the average of word and symbol grandmeans) that marked the beginning and end of the N170 component as in [1]. The N170 segment maps (144–248 ms) of the word, pseudoword, and symbol conditions were subsequently analyzed to characterize GFP and topography effects across these stimulus groups. Topographic effects were tested using difference map t-statistics, centroid analyses, and analyses of selected channels. First, overall reading-related N170 effects were analyzed, comparing words vs. symbol strings. Second, generalization of reading-related N170 specialization to novel word forms was tested comparing pseudowords both to symbols and to words.
Statistical difference map analyses
The word N170 topography showed the largest negativity at occipito-temporal electrodes with a maximum over the left hemisphere and the largest positivity at fronto-central electrodes. The symbol N170 topography showed the largest negativity at occipito-parietal electrodes with a maximum over the right hemisphere and the largest positivity at fronto-polar electrodes. The two topographies clearly differed, as shown by the large t-values in the statistical difference map (Fig. 2). The maximal effects were found at parietal electrodes and at left inferior temporal and occipito-temporal electrodes at the edge of the channel array (Fig. 2).
N170 responses to pseudowords and symbols clearly differed in the statistical difference map (Fig. 3). The maximal effects were found at parietal electrodes and bilaterally at inferior electrodes at the edge of the electrode array.
Figure 3 Generalization of reading-related N170 expertise to novel word forms. The pseudoword N170 differs from both the symbol N170 and the word N170, but the two effects show distinct topographies. The pseudoword-symbol effect shows large differences at inferior (surrounding negative difference) and superior (central positive difference) locations. The word-pseudoword effect is most pronounced over left occipito-temporal electrodes. Critical t-values are 2.14 (p<0.05), 2.98 (p<0.01), and 4.14 (p<0.001).
The statistical difference map also indicated clear N170 effects between responses to words and pseudowords (Fig. 3). These effects were left-lateralized, as they were found at many occipito-temporal channels over the left hemisphere, but hardly over the right hemisphere. The pseudoword N170 topography resembled the word N170, but showed a more bilateral negativity and a positivity centered around the fronto-central midline electrodes, in contrast to the more left-lateralized negativity and positivity of the word N170 (Fig. 3).
Global Field Power analysis
To assess overall map strength, we ran a repeated measure Analysis of Variance (ANOVA) on the N170 GFP value separately for the overall reading-related contrast (word vs. symbol), and for the two contrasts testing generalization of reading-related N170 specialization to novel word forms (pseudowords vs. symbols and words vs. pseudowords).
There was no significant overall reading-related effect in the N170 GFP measure, although map strength was somewhat larger for symbols than for words (F(1,14) = 3.00, p = ns, see also Fig. 2). The additional comparisons revealed that N170 GFP was larger in response to symbols than to pseudowords (F(1,14) = 6.07, p < 0.05), but did not differ between words and pseudowords (F(1,14) = 2.13, p = ns), although map strength was slightly larger for words than pseudowords.
Topographic centroid analyses
We used centroid measures (centers of gravity) of the positive and negative fields on the scalp surface to characterize the ERP topography [30-32]. The 3D locations of the positive and negative centroids were computed from all 129 electrode positions (in x-, y-, and z-Talairach space [33]) weighted by their positive or negative values, respectively. Repeated measure ANOVAs were run on the centroid positions, separately for the overall reading-related contrast (words vs. symbols), and the two contrasts testing for generalization to pseudowords (pseudowords vs. symbols and words vs. pseudowords). Positive and negative centroids were grouped in a factor "polarity", as positive and negative poles are often systematically related in ERP maps, and the three spatial coordinates were treated as multivariate dependent measures. Contrast main effects and polarity interactions are only reported if they differ significantly (p < 0.05) at the multivariate level. For multivariate significant effects, univariate tests were computed for the x-, y-, and z-axes, to characterize the nature of the multivariate effect in 3D space. Contrast main effects are referred to as "mean centroids" (positive and negative centroids showed a similar pattern), and contrast-by-polarity interaction effects are referred to as "centroid distribution" (positive and negative centroids showed a different pattern).
As summarized in Table 2, the centroid analysis revealed clear overall reading-related effects in the N170, indicated by a different centroid distribution between word and symbol maps (contrast x polarity, F(3,12) = 10.76, p < 0.01). These differences appeared in both the analysis of the inferior-superior z coordinate axis (F(1,14) = 37.30, p < 0.001) and the analysis of left-right x coordinate axis (F(1,14) = 7.07, p < 0.05, table 2). As illustrated in Fig. 4, this interaction captures differences between negative centroids appearing as inferior and left-lateralized for words, yet superior and right-lateralized for symbols. The positive centroids showed a reversed pattern, as they were located more superior for words and more inferior for symbols (Fig. 4).
Table 2 Effects of the N170 topographic centroid analyses.
Topographic effects (multivariate significant) x-axis y-axis z-axis
Contrast (words vs. symbols) x polarity (positive vs. negative) p < 0.05 ns p < 0.001
Contrast (pseudowords vs. symbols) x polarity (positive vs. negative) ns* ns p < 0.001
Contrast (words vs. pseudowords) p < 0.05 ns Ns
* non-significant trend (F(1,14) = 3.73, p < 0.1)
Figure 4 Positive and negative centroids of the N170 word, pseudoword, and symbol topographies. Symbol centroids show a different pattern from word and pseudoword centroids, with a reversed polarity in the inferior-superior direction. In addition, the negative centroid is left-lateralized for words, but right-lateralized for symbols. The centroids are also more left-lateralized for words than for pseudowords. Note that the centroids represent the ERP topography on the scalp surface and are by no means estimations of the underlying sources.
Clear topographic effects for the N170 responses to pseudowords and symbols were also found in the centroid analysis, as indicated by different centroid distributions between pseudoword and symbol conditions (contrast x polarity F(3,12) = 7.10, p < 0.01). This difference was mainly found on the z-axis (F(1,14) = 20.94, p < 0.001, table 2). The negative centroids were located more inferior for pseudowords and more superior for symbols, whereas the positive centroids showed a reversed pattern (Fig. 4). There was an additional non-significant trend on the x-axis with the centroids more lateralized for symbols than for pseudowords (F(1,14) = 3.74, p < 0.1).
Topographic effects between the N170 in response to words and pseudowords were indicated by different mean centroid locations for words and pseudowords (F(3,12) = 3.59, p < 0.05). These effects appeared on the left-right x axis (F(1,14) = 4.73, p < 0.05, table 2). The mean centroids were more left-lateralized for words than for pseudowords (Fig. 4).
Selected waveform analyses
In order to allow comparisons with more conventional ERP analysis approaches, we also performed an analysis on left and right inferior occipito-temporal channel groups which have been shown to be most sensitive to word-symbol differences [1]. Figure 1 illustrates the specific channels included in the left and right groups, respectively. Repeated measure ANOVAs were run with the hemisphere factor (left vs. right channel group) separately for the overall reading-related contrast (words vs. symbols), and the two contrasts testing for generalization to pseudowords (pseudowords vs. symbols and words vs. pseudowords).
A clear overall reading-related effect was seen in the selected waveform analysis comparing word and symbol N170. The N170 amplitude was larger for words than for symbols (contrast, F(1,14) = 9.93, p < 0.01), and this difference was larger over the left hemisphere (contrast x hemisphere, F(1,14) = 10.78, p < 0.01; Fig. 5).
Figure 5 Waveforms at left and right inferior occipito-temporal channels. The N170 is larger for words than for pseudowords and symbols, especially at the left hemisphere channels. Pseudowords have a larger N170 than symbols at both hemispheres, especially during the late part of the N170.
The pseudoword-symbol contrast also revealed a significant effect. Pseudowords elicited larger N170 amplitudes than symbols (F(1,14) = 7.18, p < 0.05; Fig. 5). Although this difference was somewhat larger over the left hemisphere, the interaction with hemisphere failed to reach significance (F(1,14) = 3.26, p < 0.1).
The word-pseudoword contrast also revealed a significant effect in N170 amplitudes at inferior occipito-temporal channels. The amplitudes were larger for words than pseudowords (contrast, F(1,14) = 8.35, p < 0.05), which were more pronounced at the channels over the left than over the right hemisphere (contrast x hemisphere, F(1,14) = 7.57, p < 0.05; Fig. 5).
Discussion
Summary of the results
The present high-density ERP study clearly shows reading-related expertise effects that occur early during processing of visual words. Processing differences between words and novel symbol strings that control for basic visual features emerged in a time window corresponding to the N170 ERP component.
One of the central goals of this study involved extensive analysis of the topography of the N170 reading-related perceptual expertise effect. The N170 responses elicited by word and symbol stimulus blocks demonstrated significantly distinct topographies. The strongest topographic effect was found in a different centroid distribution on the inferior-superior coordinate axis. This effect reflected the occipito-temporal negativity and fronto-central positivity in the word maps and the occipito-parietal negativity and fronto-polar positivity in the symbol maps. This difference also led to maximal effects in the t-map at inferior and superior electrodes. Indeed, the largest negative effects were at the inferior edge of the channel array and might be missed by many conventional montages, which typically do not sample these regions [23,24,34]. An additional topographic effect was found in a different lateralization of the word and the symbol N170. The negative centroids were left-lateralized for words, but right-lateralized for symbols. This lateralization difference was corroborated in the selected waveform analysis, which showed larger N170 amplitudes for words than symbols especially over the left inferior occipito-temporal channels.
The results also inform the generalization of reading-related expertise in the N170 to novel word forms. The pseudoword vs. symbols N170 contrast led to large t-values in the statistical difference map and demonstrated strong evidence for perceptual expertise elicited by novel word forms. The topography of this N170 effect played out primarily as a shift in the inferior-superior dimension of the positive and negative centroids, very similar to the inferior-superior topographic N170 difference found between word and symbol centroids. In contrast, the topographic effect in the left-right dimension only partially generalized to pseudowords, showing just a non-significant trend for lateralization differences, which was corroborated in the selected waveform analysis.
Overall, N170 reading expertise effects do not appear to fully generalize to pseudoword probes in English, as a generalization was found for the topographic inferior-superior effect, but only partially for the topographic lateralization effect.
The strongest evidence that generalization to novel word forms differs in lateralization came from the word-pseudoword comparison, in which the N170 centroids were more left-lateralized for words than for pseudowords. This left-lateralization was corroborated in the additional analysis at inferior occipito-temporal channel groups and in the statistical difference map. This demonstrates that the left-lateralized topographic effect of reading-related visual expertise in the N170 does not generalize well to pseudowords.
Behavioral results, overall, served to ensure that the participants demonstrated roughly equivalent levels of engagement with the different classes of stimuli, although subtle behavioral differences were revealed in the case of the symbol condition, in which slightly lower accuracy suggests that detecting symbol strings was slightly more difficult compared to the other conditions. Interestingly, the equivalent speed and accuracy for target detection across words and pseudowords demonstrates that the left-lateralized word-pseudoword effect in the N170 is not dependent on processing differences assessed by behavioral measures.
Replication of effects across languages
The present study adopted a paradigm based on an earlier study in Zurich [1] that used a different EEG system with participants speaking a different language. The paradigm was identical in the two studies, except for language-specific word and pseudoword stimuli, and the same analysis strategy was used. This allows us to examine the two sets of results regarding similarities and differences.
Overall, the basic findings regarding reading-related specialization were successfully replicated, such as major word-symbol differences during the N170 and the P300 time windows, and similar topographic N170 differences between words and symbols with larger negativities for words than for symbols at inferior occipito-temporal channels especially over the left hemisphere. Particularly, the centroid analyses of the N170 maps showed the same robust differences between words and symbols in the inferior-superior dimension in the two studies. The left-lateralized topographic effect for the word-symbol comparison in the present study also appeared in the Zurich data, where it reached significance in the last two thirds of the N170. Overall, the results suggest that word-symbol differences in the N170 time range are robust markers for rapid specialization for reading, which can be detected across different EEG systems and languages.
Although the two studies showed similar topographic N170 effects for the word-symbol comparison, reading-related specialization differed between the two languages when probed with novel word forms. For the pseudoword-symbol comparison, both studies found significant differences in the inferior-superior dimension, similar to the one for the word-symbol comparison. In the left-right dimension, however, the difference was significant in Zurich, but not in the present study. The strongest evidence for a difference in generalization to pseudowords between the two studies was found in the word-pseudoword comparison, where in contrast to the significant lateralization difference in the present study, no difference at all was found in the Zurich study. This reflects the fact that the pseudoword topographies were left-lateralized for German speakers, but bilateral in the present study. As we will discuss below, differences in orthographic depth between the two languages may explain this effect in particular and may shed some light on the characteristic left-lateralized topography of reading-related N170 specialization in general.
Behavioral results were very similar in both studies. Repetition detection was high (>90%) in all conditions, with a slight advantage for detecting words and pseudowords compared to symbols in both studies. There was no difference in reaction time between conditions in any of the two studies. This shows that the difference in generalization of reading-related expertise to pseudowords between the two studies was not due to differences in overt behavioral responses.
In addition to the generalization difference of reading-related N170 expertise to novel word forms, the two studies also differed in relative map strength of the N170 between the conditions. In the Zurich study, words and pseudowords had larger GFP than symbols, but in the present study this relation tended to be reversed. Smaller overall amplitudes in the EEG for electrolyte-net systems compared to electrogel-cap systems, as reported earlier [35], should affect language and symbol stimuli equally. Although the different word-symbol GFP ratios between the studies could potentially result from differing symbol GFP, this seems unlikely because the two studies mainly differed with respect to the language stimuli. Thus, this may suggest that words and pseudowords elicit relatively smaller N170 amplitudes in English than in German which could be due to differences in orthographic depth between the two languages. However, to exclude a possible influence of the recording system, such language-related N170 effects on GFP may be investigated in future studies using the same system for the two language groups.
In contrast, reading-related N170 effects basically replicated their general topographies across languages, which suggests that topography (rather than GFP) is a robust marker of reading-related specialization in the N170 across EEG systems and language backgrounds. This, in turn, suggests that the topographic N170 differences between the two studies found in the word-pseudoword comparisons are valid markers for the influence of different languages.
The topographic centroid analyses not only revealed similar effects for reading-related N170 specialization in the word-symbol comparison between the two studies, the centroid analysis also appeared as a more fruitful analysis strategy than the selected waveform analysis in the present study. Although the waveform analysis confirmed the lateralization effects of the centroids, the pre-selected channel array prevented the detection of the strong topographic inferior-superior effects in the word-symbol and pseudoword-symbol N170 contrasts. This illustrates, how the selection of particular channels in multi-channel recordings could lead to biased results and conclusions. The centroid analysis method is a means for unbiased topographic ERP analyses, and has been proven useful to detect topographic differences in earlier studies (e.g. [30-32]).
General Discussion
The differences between word and symbol processing that appeared in the N170 component in the present study, corroborate findings from studies using MEG [10] or conventional EEG systems [1,8,9,36,37] in supporting the general conclusion that processes in posterior brain regions that are activated within the first 200 msec are sensitive to reading-related experience. The present study extends earlier findings in two important ways: showing that the maximal effect of the negative difference appears at scalp locations that were not sampled previously, and showing that reading-related expertise in the N170 has a distinct functional organization in English which can be detected when the expertise system is probed with pseudowords.
Reading-related expertise in the N170 appeared in two topographic dimensions: the inferior-superior dimension and the left-right dimension. These two topographic effects generalized to a different degree to pseudowords in the present study, suggesting that the two effects may be associated with different functional properties of reading-related expertise.
The inferior-superior modulation of the N170 word-symbol difference fully generalized to pseudowords in the present study. The same inferior-superior effect was also found for the word-symbol and pseudoword-symbol contrast in the Zurich study with participants speaking a different language [1]. Since the inferior-superior topographic effect is robust across languages and generalizes to novel word forms, it may reflect visual expertise for the familiarity of letters within strings. Visual expertise reflected by the inferior-superior topographic N170 effect may be related to expertise in other visual domains, such as expertise for faces or objects [2,3].
Such speculation regarding various forms of expertise, however, requires additional investigation, as studies on face or object expertise have not analyzed topographic effects beyond lateralization [4-6]. It remains to be tested whether the inferior-superior topographic effect in the N170 relates to a functional property that is shared across different domains of visual expertise, or whether it represents a functional property characteristic of reading.
Earlier findings suggested that the main characteristic feature of reading-related N170 specialization lies in its left-lateralized topography [1,3,8], contrasting the typically bilateral or right-lateralized N170 topographies for faces and objects of expertise [2,3]. The present results add further evidence for this notion, but extend earlier findings, by showing that this left-lateralized topographic effect does not fully generalize to novel word forms in English, as it did in German [1]. Thus, the left-lateralized topographic effect may be associated with functional properties that can be inferred from language differences between English and German, especially with respect to pseudoword reading.
English and German differ in the degree of orthographic depth, which is deep in English and shallow in German. Orthographic depth refers to the level of consistency whith which spelling maps onto word sounds (feedforward consistency) and word sounds map onto spelling (feedbackward consistency). In the case of pseudoword reading, the former plays an important role. Due to inconsistency in spelling-to-sound mapping in English, the pronunciation of pseudowords is much more ambiguous in English than in German. Thus the left-lateralized topographic N170 effect may specifically relate to processes involved in mapping letters onto word sounds. Mapping consistency has been demonstrated to be a central factor modulating the rise of automaticity in information processing [38]. The lack of a left-lateralization for English pseudowords may suggest that such processes are less automatic in English [39], and are engaged to a lesser degree while detecting pseudoword repetitions, because repetition detection does not require explicit pronunciation of the stimuli.
Such an interpretation also fits with the lack of N170 word-pseudoword differences in earlier studies in Finnish and French [8,16]. Whereas Finnish orthography is shallow, French orthography has some inconsistencies in sound-to-spelling mapping, but is rather consistent in spelling-to-sound mapping, which renders pseudoword pronunciation less ambiguous in French [40].
One study in English also did not find N170 differences between words and pseudowords, but the participants performed a lexical task that encouraged deeper language processing of the pseudowords, which is in agreement with the automaticity hypothesis [21].
Another study in English found larger N170 amplitudes for irregular pseudowords than for words, which may suggest that the irregularity of the pseudowords led to an enhancement of the N170 similar to findings for consonant strings [14,15]. The effect of larger N170 amplitudes for consonant strings compared to words may be related to the well-replicated findings of larger N170 amplitudes for low-frequency words than for high-frequency words [18-21]. For both the consonant strings and the word frequency effects, no lateralization differences have been reported and the inferior-superior topographic effect has not been investigated. Future studies may show whether this effect is related to the inferior-superior topographic effect in the present study, or whether it represents an additional reading-related modulation of the N170 component.
Different levels of engagement in orthographic-to-phonological processing might also explain the variable left-lateralization of N170 related to reading in the literature [3,12]. Thus, the degree of left-lateralization may vary according to language, task, and stimulus factors that impact the degree to which visual, orthographic and phonological codes are engaged.
Evidence for language-specific effects on pseudoword processing also comes from a PET study with English and Italian subjects. During explicit and implicit pseudoword processing, left posterior inferior temporal regions were more activated in English subjects, whereas in Italian subjects left superior temporal regions were more activated [25]. These results corroborate that pseudowords are processed differently in languages that differ in consistency of spelling-to-sound mapping. However, future studies combining hemodynamic and electrophysiological methods are needed to clarify the relation between metabolic activation and N170 amplitude modulation for pseudoword processing in English.
Combined hemodynamic and electrophysiological studies can also help localize the sources of the reading-related N170 specialization. Studies combining fMRI with MEG [41] and EEG [42] support the view that the word N170 originates predominantely from inferior occipito-temporal regions, in agreement with source localization from studies using MEG and EEG alone [1,3,10]. The posterior left-lateralized effect for words in the current study is consistent with sources in the left inferior occipito-temporal region, including the general region of the "Visual Word Form Area", and suggests that such a left posterior region demonstrates different patterns of neuronal responses to words vs. visual control stimuli within the first 200 msec of processing.
The notion of the Visual Word Form Area was first inspired by neuropsychological observations of "pure" alexia, or letter by letter reading, characterized by an inability to read entire words, typically following damage to left-inferior-temporal regions with a maximal probability over fusiform gyrus (see [7] for review). Left fusiform gyrus regions are also activated in metabolic studies contrasting words and visual control stimuli, and this area has been termed the Visual Word Form Area [7,42], suggesting a structure-function linkage between this region and early cognitive perceptual processes proposed in models of word recognition [38]. Although the left fusiform gyrus is typically activated in visual word tasks, this specialization does not necessarily exclude the participation of this region in other forms of processing, such as picture recognition, nor does it exclude the participation of additional regions in visual word processing (for review see [7,43]).
Previous neuroimaging studies of the putative "Visual Word Form Area" in the left fusiform gyrus have shown sensitivity for orthographic regularity, with more activation for words and pseudowords than for nonwords (for a review see [7]). In contrast, sensitivity for familiarity of word forms was small (for a review see [7]), although a more recent study suggests that activation may increase for words with low frequency and for pseudowords [44]. In the present ERP study the N170 showed sensitivity for the familiarity of word forms suggesting that this sensitivity may be language-dependent, which might also apply for the fMRI results. However, there are additional reasons that could explain different results between ERP and fMRI studies. Some N170 effects may be too transient to be captured by the low temporal resolution of fMRI. Moreover, it is also possible that the N170 effects for the word-pseudoword comparison do not originate from the left fusiform region, but from other left posterior regions contributing to the N170. Future research combining fMRI and ERP in the same study, and examining factors such as spelling-to-sound consistency patterns across languages, and within words and pseudowords, may help to elucidate the nature of different neural contributions to the N170 related to reading expertise.
Conclusion
The present study provides further evidence that there are rapid perceptual processes in the brain that are specialized for reading. The N170 topography is a robust neurophysiological marker for this specialization showing more inferior and left-lateralized negativity for words compared to symbols. The results extend this general finding via a dense array and extended inferior coverage, demonstrating that the maximal negative effect is more inferior than reported previously. Further characterization of the N170 response in the current study suggests that reading-related perceptual expertise in the N170 can be characterized by at least two topographic effects which generalize to novel word forms to different degrees. The inferior-superior topographic effect in the N170 fully generalized to novel word forms, and may reflect expertise for letters or well-ordered letter strings. Unlike in German, the left-lateralized topographic effect in the N170 did not generalize to novel word forms in English. Inference from language differences between English and German suggests that the left-lateralized topographic effect in reading-related N170 specialization may reflect spelling-to-sound conversion, which might be less automaticly engaged in pseudoword processing in English due to more ambiguous pronunciation of novel word forms.
Methods
Participants
The data of 15 right-handed, native English speakers (19 to 29 years old) are presented. All subjects had normal or corrected-to-normal vision and their word reading and pseudoword decoding abilities [45] were within the normal range (within 2 SD of the norm mean). Although EEG data of 20 subjects were obtained, data of 5 subjects were discarded due to low signal-to-noise ratios (3 subjects), bad net fit (1 subject), and outlier values in the ERP (>3 SD, 1 subject). All subjects provided informed consent approved by the Weill-Cornell Institutional Review Board Committee.
Procedure
To investigate rapid specialization for print, we used a paradigm that was used in an earlier study in Zurich, Switzerland [1]. The main differences between the two studies are the different EEG systems (geodesic net vs. electrode caps) and the language of the stimuli and participants (English vs. German). The experiment used the same stimulus string conditions (words, pseudowords, symbol strings) as [1], but while the symbol-strings were identical, the German words were translated to English (high-frequency words in both languages), and the German pseudowords were replaced by regular English pseudowords. Both words and pseudowords were printed with an initial capital letter to match the visual characteristics of German nouns. Words, pseudowords, and symbol-strings were matched for string length and contained 4.5 letters/symbols on average (range: 3–7), which also equaled the German string length. The experimental parameters were kept identical to the Zurich study: stimuli were shown every 2050 ms for 700 ms in black on a white background 100 cm away from the subject at a visual angle of 1.6–3.6 degrees (shorter distance and smaller print size to keep the same visual angle). In each condition, 72 stimuli were presented in 2 blocks containing 17% repetitions, which served as targets. To keep the experiment context the same as in the Zurich study, 2 blocks of pictures were also presented within the same session, but these data are not reported, as the different stimulus size and stimulus contrast of the pictures would confound condition effects in the N170 component. In all blocks, participants were instructed to press a button with their right thumb whenever they detected an immediate repetition.
Electrophysiological Recording and Analyses
The 129 channel ERPs were recorded using a geodesic sensor net [46] with a Cz reference. Data were sampled at 250 Hz/channel with filter settings 0.1–100 Hz and with calibrated technical zero baselines. Impedance was kept below 50 kΩ [47]. Using BESA software, channels with excessive artifacts were spline interpolated (in average 3.5 channels per subject), and eye blinks were corrected (multiple source eye correction method [48], as applied in the Zurich study). The data then were digitally bandpass filtered (0.3–30 Hz), segmented (-150–850 ms), artifact rejected (± 100 uV), and averaged according to non-target stimuli separately for the four conditions. Using Brain Vision Analyzer software, the averaged data were re-referenced to average reference ([27]), and highpass filtered (1 Hz) to further reduce slow wave drifts. After computing Global Field Power (GFP) [27], grandmeans were computed for all four stimulus conditions.
To assess differential processing of word and symbol strings, a Topographic Analysis of Variance (TANOVA, [26], part of the LORETA-Key software package, available at on non-normalized (raw) ERP maps was computed for each time point. TANOVA on raw maps detects all systematic amplitude differences between two maps running a nonparametric randomization test [49] on the GFP of difference maps [26,27]. Note that differences resulting from TANOVA on raw maps can be due to different topographies, as well as due to different map strengths.
For the N170 analysis, a time segment was selected between the two GFP minima before and after the N170 peak (144–248 ms) of the averaged word and symbol GFP grandmeans, as in [1]. For the N170 segment maps, GFP and 3D centroids were computed. GFP is the root mean square of the values at all electrodes. The positive 3D centroid is the voltage-weighted average of the locations of all electrodes showing positive values; the negative 3D centroid is the analogous computed for electrodes with negative values. Centroid locations are shown in Talairach space [33]. GFP and centroid measures were analyzed in repeated measure ANOVAs with a "contrast" factor (either words vs. symbols, pseudowords vs. symbols, or words vs. pseudowords). For the centroid analyses "polarity" (positive vs. negative) was included as an additional factor, and the x-, y-, and z-axes were treated as multivariate dependent measures. The selected waveform analysis in the N170 segment was computed with values from inferior occipito-temporal channel groups over the left (channels 57, 58, 63, 64, 65, 69, and 70) and right (channels 90, 91, 95, 96, 97, 100, 101) hemisphere, as inferior occipito-temporal regions were most sensitive to word-symbol differences in earlier work [1]. This analysis was similar to the GFP analysis with the additional "hemisphere" factor (left vs. right).
For the behavioral analysis two repeated measure ANOVAs were computed for accuracy and reaction time with the "condition" factor (words vs. pseudowords vs. symbols).
List of abbreviations
3D: three-dimensional
ANOVA: Analysis of Variance
EEG: Electroencephalography
ERP: Event-related potential
GFP: Global Field Power
TANOVA: Topographic Analysis of Variance
Authors' contributions
UM, DB, and BDM designed the study and the data analysis strategy. UM and BDM shared the elaboration of the paper. UM was responsible for data collection and data analysis.
Acknowledgements
This research was supported by the Swiss National Science Foundation (Fellowship for Prospective Researchers: UM) and the US National Science Foundation (NSF 529112). We thank Sela Han for helping collecting the data and Jason D. Zevin for helpful discussions.
==== Refs
Maurer U Brem S Bucher K Brandeis D Emerging neurophysiological specialization for letter strings J Cogn Neurosci
Bentin S Allison T Puce A Perez E Electrophysiological studies of face perception in humans J Cogn Neurosci 1996 8 551 565
Rossion B Joyce CA Cottrell GW Tarr MJ Early lateralization and orientation tuning for face, word, and object processing in the visual cortex Neuroimage 2003
Tanaka JW Curran T A neural basis for expert object recognition Psychol Sci 2001 12 43 47 11294227 10.1111/1467-9280.00308
Gauthier I Curran T Curby KM Collins D Perceptual interference supports a non-modular account of face processing Nat Neurosci 2003 6 428 432 12627167 10.1038/nn1029
Rossion B Gauthier I Goffaux V Tarr MJ Crommelinck M Expertise training with novel objects leads to left-lateralized facelike electrophysiological responses Psychol Sci 2002 13 250 257 12009046 10.1111/1467-9280.00446
McCandliss BD Cohen L Dehaene S The visual word form area: expertise for reading in the fusiform gyrus Trends Cogn Sci 2003 7 293 299 12860187 10.1016/S1364-6613(03)00134-7
Bentin S Mouchetant-Rostaing Y Giard MH Echallier JF Pernier J ERP manifestations of processing printed words at different psycholinguistic levels: time course and scalp distribution J Cogn Neurosci 1999 11 235 260 10402254 10.1162/089892999563373
Brem S Lang-Dullenkopf A Maurer U Halder P Bucher K Brandeis D Neurophysiological signs of rapidly emerging visual expertise for symbol strings Neuroreport 2005 16 45 48 15618888 10.1097/00001756-200501190-00011
Tarkiainen A Helenius P Hansen PC Cornelissen PL Salmelin R Dynamics of letter string perception in the human occipitotemporal cortex Brain 1999 122 2119 2132 10545397 10.1093/brain/122.11.2119
Brandeis D Lehmann D Michel CM Mingrone W Mapping event-related brain potential microstates to sentence endings Brain Topogr 1995 8 145 159 8793125 10.1007/BF01199778
Koenig T Kochi K Lehmann D Event-related electric microstates of the brain differ between words with visual and abstract meaning Electroencephalogr Clin Neurophysiol 1998 106 535 546 9741753 10.1016/S0013-4694(97)00164-8
Skrandies W Evoked potential correlates of semantic meaning--A brain mapping study Brain Res Cogn Brain Res 1998 6 173 183 9479069 10.1016/S0926-6410(97)00033-5
McCandliss BD Posner MI Givon T Brain plasticity in learning visual words Cogn Psychol 1997 33 88 110 10.1006/cogp.1997.0661
Compton PE Grossenbacher P Posner MI Tucker DM A cognitive-anatomical approach to attention in lexical access J Cogn Neurosci 1991 3 304 312
Wydell TN Vuorinen T Helenius P Salmelin R Neural correlates of letter-string length and lexicality during reading in a regular orthography J Cogn Neurosci 2003 15 1052 1062 14614815 10.1162/089892903770007434
Assadollahi R Pulvermuller F Early influences of word length and frequency: a group study using MEG Neuroreport 2003 14 1183 1187 12821805 10.1097/00001756-200306110-00016
Hauk O Pulvermuller F Effects of word length and frequency on the human event-related potential Clin Neurophysiol 2004 115 1090 1103 15066535 10.1016/j.clinph.2003.12.020
Neville HJ Mills DL Lawson DS Fractionating language: Different neural subsystems with different sensitive periods Cereb Cortex 1992 2 244 258 1511223
Sereno SC Brewer CC O'Donnell PJ Context effects in word recognition: evidence for early interactive processing Psychol Sci 2003 14 328 333 12807405 10.1111/1467-9280.14471
Sereno SC Rayner K Posner MI Establishing a time-line of word recognition: evidence from eye movements and event-related potentials Neuroreport 1998 9 2195 2200 9694199
Proverbio AM Vecchi L Zani A From orthography to phonetics: ERP measures of grapheme-to-phoneme conversion mechanisms in reading J Cogn Neurosci 2004 16 301 317 15068599 10.1162/089892904322984580
Chatrian GE Lettich E Nelson PL Ten percent electrode system for topographic studies of spontaneous and evoked EEG activity Am J EEG Technol 1985 25 83 92
Oostenveld R Praamstra P The five percent electrode system for high-resolution EEG and ERP measurements Clin Neurophysiol 2001 112 713 719 11275545 10.1016/S1388-2457(00)00527-7
Paulesu E McCrory E Fazio F Menoncello L Brunswick N Cappa SF Cotelli M Cossu G Corte F Lorusso M Pesenti S Gallagher A Perani D Price C Frith CD Frith U A cultural effect on brain function Nat Neurosci 2000 3 91 96 10607401 10.1038/71163
Strik WK Fallgatter AJ Brandeis D Pascual-Marqui RD Three-dimensional tomography of event-related potentials during response inhibition: evidence for phasic frontal lobe activation Electroencephalogr Clin Neurophysiol 1998 108 406 413 9714383 10.1016/S0168-5597(98)00021-5
Lehmann D Skrandies W Reference-free identification of components of checkerboard-evoked multichannel potential fields Electroencephalogr Clin Neurophysiol 1980 48 609 621 6155251 10.1016/0013-4694(80)90419-8
Lehmann D Skrandies W Spatial analysis of evoked potentials in man--a review Prog Neurobiol 1984 23 227 250 6395186 10.1016/0301-0082(84)90003-0
Brandeis D Lehmann D Event-related potentials of the brain and cognitive processes: approaches and applications Neuropsychologia 1986 24 151 168 3703232 10.1016/0028-3932(86)90049-7
Brandeis D Vitacco D Steinhausen HC Mapping brain electric micro-states in dyslexic children during reading Acta Paedopsychiatr 1994 56 239 247 8079644
Maurer U Bucher K Brem S Brandeis D Development of the automatic mismatch response: from frontal positivity in kindergarten children to the mismatch negativity Clin Neurophysiol 2003 114 808 817 12738427 10.1016/S1388-2457(03)00032-4
Maurer U Bucher K Brem S Brandeis D Altered responses to tone and phoneme mismatch in kindergartners at familial dyslexia risk Neuroreport 2003 14 2245 2250 14625456 10.1097/00001756-200312020-00022
Talairach J Tournoux P Co-planar Stereotaxic Atlas of the Human Brain. 1988 Stuttgart , Thieme
Luu P Ferree T Determination of the Geodesic Sensor Nets' Average Electrode Positions and their 10-10 International Equivalents Technical Note 2000 Eugene, Oregon , Electrical Geodesics, Inc 1 11
Kayser J Tenke CE Bhattacharya N Stuart BK Hudson J Bruder GE A direct comparison of Geodesic Sensor Net (128-channel) and conventional (30-channel) ERPs in tonal and phonetic oddball tasks Psychophysiology 2000 37 S17
Schendan HE Ganis G Kutas M Neurophysiological evidence for visual perceptual categorization of words and faces within 150 ms Psychophysiology 1998 35 240 251 9564744 10.1017/S004857729897010X
Zhang XL Begleiter H Porjesz B Litke A Visual object priming differs from visual word priming: an ERP study Electroencephalogr Clin Neurophysiol 1997 102 200 215 9129576 10.1016/S0013-4694(96)95172-3
Schneider W Shiffrin RM Controlled and automatic human information processing: I. detection, search, and attention Psychological Review 1977 84 1 66 10.1037//0033-295X.84.2.127
Zevin JD Balota DA Priming and attentional control of lexical and sublexical pathways during naming J Exp Psychol Learn Mem Cogn 2000 26 121 135 10682293 10.1037//0278-7393.26.1.121
Ziegler JC Perry C Coltheart M Speed of lexical and nonlexical processing in French: the case of the regularity effect Psychon Bull Rev 2003 10 947 953 15000543
Dale AM Liu AK Fischl BR Buckner RL Belliveau JW Lewine JD Halgren E Dynamic statistical parametric mapping: combining fMRI and MEG for high-resolution imaging of cortical activity Neuron 2000 26 55 67 10798392 10.1016/S0896-6273(00)81138-1
Cohen L Dehaene S Naccache L Lehericy S Dehaene-Lambertz G Henaff MA Michel F The visual word form area: spatial and temporal characterization of an initial stage of reading in normal subjects and posterior split-brain patients Brain 2000 123 291 307 10648437 10.1093/brain/123.2.291
Price CJ Devlin JT The pro and cons of labelling a left occipitotemporal region: "the visual word form area" Neuroimage 2004 22 477 479 15110041 10.1016/j.neuroimage.2004.01.018
Kronbichler M Hutzler F Wimmer H Mair A Staffen W Ladurner G The visual word form area and the frequency with which words are encountered: evidence from a parametric fMRI study Neuroimage 2004 21 946 953 15006661 10.1016/j.neuroimage.2003.10.021
Torgesen JK Wagner RK Rashotte CA TOWRE - Test of word reading efficiency 1999 Austin, Texas , PRO-ED
Tucker DM Spatial sampling of head electrical fields: the geodesic sensor net Electroencephalogr Clin Neurophysiol 1993 87 154 163 7691542 10.1016/0013-4694(93)90121-B
Ferree TC Luu P Russell GS Tucker DM Scalp electrode impedance, infection risk, and EEG data quality Clin Neurophysiol 2001 112 536 544 11222977 10.1016/S1388-2457(00)00533-2
Berg P Scherg M A multiple source approach to the correction of eye artifacts Electroencephalogr Clin Neurophysiol 1994 90 229 241 7511504 10.1016/0013-4694(94)90094-9
Holmes AP Blair RC Watson JD Ford I Nonparametric analysis of statistic images from functional mapping experiments J Cereb Blood Flow Metab 1996 16 7 22 8530558 10.1097/00004647-199601000-00002
|
16091138
|
PMC1208852
|
CC BY
|
2021-01-04 16:39:20
|
no
|
Behav Brain Funct. 2005 Aug 9; 1:13
|
utf-8
|
Behav Brain Funct
| 2,005 |
10.1186/1744-9081-1-13
|
oa_comm
|
==== Front
Behav Brain FunctBehavioral and brain functions : BBF1744-9081BioMed Central London 1744-9081-1-141609114110.1186/1744-9081-1-14ResearchNeuropsychological measures of attention and memory function in schizophrenia: relationships with symptom dimensions and serum monoamine activity Oades Robert D [email protected]öpcke Bernd [email protected] Uwe [email protected] Ansgard [email protected] Biopsychology Research Group, University Clinic for Child and Adolescent Psychiatry, Virchowstr. 174, 45147 ESSEN, Germany2 University Clinic for Psychiatry and Psychotherapy, Bergische Landstr. 2,40629 Düsseldorf, Germany2005 9 8 2005 1 14 14 10 3 2005 9 8 2005 Copyright © 2005 Oades et al; licensee BioMed Central Ltd.2005Oades 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
Some clinical symptoms or cognitive functions have been related to the overall state of monoamine activity in patients with schizophrenia, (e.g. inverse correlation of the dopamine metabolite HVA with delusions or visual-masking performance). However, profiles (as presented here) of the relations of the activity of dopamine, noradrenaline and serotonin to neuropsychologic (dys)functions in major patient sub-groups with their very different symptomatic and cognitive characteristics have not been reported.
Methods
Serum measures of dopamine, noradrenaline and serotonin turnover were examined by regression analyses for the prediction of performance on 10 neuropsychological measures reflecting left- and right-hemispheric and frontal-, parietal- and temporal-lobe function in 108 patients with schizophrenia and 63 matched controls. The neuropsychological battery included tests of verbal fluency, Stroop interference, trail-making, block-design, Mooney faces recognition, picture-completion, immediate and delayed visual and verbal recall. Paranoid and nonparanoid subgroups were based on ratings from the Positive and Negative Syndrome Scale (PANSS). Groups with high and low ratings of ideas-of-reference and thought-disorder were formed from a median split on the Scale for Assessment of Positive Symptoms (SAPS).
Results
Verbal-fluency and Stroop-interference (left frontal and fronto-cingulate function) were negatively associated with noradrenergic turnover in nonparanoid and thought-disordered patients. High dopamine turnover related to speeded trail-making (frontal modulation of set switching) in those with many ideas-of-reference. In contrast, low dopamine turnover predicted poor recall in nonparanoid patients and those with little thought disorder. Serotonin metabolism did not independently contribute to the prediction any measure of cognitive performance. But, with regard to the relative activity between monoaminergic systems, increased HVA/5-HIAA ratios predicted visual-reproduction and Mooney's face-recognition performance (right-hemisphere functions) in highly symptomatic patients. Decreased HVA/MHPG predicted non-verbal recall.
Conclusion
Clinical state and function are differentially sensitive to overall levels of monoamine activity. In particular, right-lateralised cerebral function was sensitive to the relative activities of the monoamines. Increased noradrenergic activity was associated with enhanced frontal but impaired temporal lobe function in nonparanoid syndromes. Low dopaminergic activity predicted poor attentional set control in those with ideas-of-reference, but poor recall in nonparanoid patients. These data, especially the HVA/5-HIAA ratios, provide a basis for planning the nature of antipsychotic treatment aimed at patient specific symptom dimensions and cognitive abilities.
schizophreniacognitiondisorganisedideas of referenceparanoiddopaminenoradrenalineserotonin
==== Body
Background
The idea that there may be a "group of schizophrenias" among the psychotically ill is largely attributable to Bleuler [1]. Yet, only in the last decade or so have studies of the biological concomitants of schizophrenia explicitly studied the 3–4 symptom dimensions making up the syndrome and taken account of their variation across patient samples or within individuals over time. Carpenter et al. noted that "Studies comparing a schizophrenia group with a comparison group often show differences without clarifying whether the difference relates only to a subgroup of the subjects" [2]. The reverse also occurs. Many studies report there is no difference between subject groups without considering the potentially opposing influences in the constituent subgroups (e.g. paranoid vs. nonparanoid schizophrenia, [3]).
Few clinicians would expect measures of attentive abilities in thought-disorganised patients to resemble those from patients with nonparanoid negative symptoms, yet few report on the biological variations that might underlie, contribute to or modulate these differences. Here we report on some indicators of such modulation (monoamine transmitters) and the association of their activity with measures of cognitive abilities in subgroups of schizophrenia.
Such contrasts are illustrated in reviews by Amin et al.[4,5]. Negative relations of the dopamine (DA) metabolite homovanillic acid (HVA) with numerous negative symptoms and positive relationships with productive symptoms have been reported frequently. Yet for summed ratings of negative symptoms [6] no correlations with plasma levels of HVA have been reported [7-9]. Just occasionally, when a general diagnosis of schizophrenia has been used, some biochemical associations with groups of symptoms have been described (e.g. HVA correlated positively with anhedonia-asociality ratings [9], or negatively with depression and hostility [10]).
Peripheral measures of monoamine metabolism may be weak indicators of activity affecting the central systems: but the changes that are registered are strong enough to be relevant to the study of behaviour [12]. In particular, the ratio of DA to serotonin (5-HT) metabolites (HVA/5-HIAA: [5-hydroxyindole-acetic acid]) characterises some patient groups. Low ratios are evident in thought-disordered patients but high ratios in those with prominent ideas-of-reference [11]. The amelioration of (usually) positive symptoms parallels changes in plasma HVA [13,14] and these changes can be reflected in both plasma and cerebrospinal fluid (CSF) samples [15-17]. For CSF, similar correlations for the HVA/5-HIAA ratio with both symptoms and cognitive performance have been reported [18,19]. Low CSF HVA/5-HIAA ratios were associated with a persistent impairment and a poor outcome [20], and increased ratios with a good response to treatment [21]. As would be predicted from these CSF results, plasma HVA/5-HIAA ratios rose in responders to treatment with antipsychotic drugs [22] and approached normal levels after atypical vs. typical neuroleptic medication [11]. Thus, as extensively argued before [4,5], plasma/serum measures of monoamine metabolism and especially the ratio between their metabolites serve as a useful proxy for widespread and general changes associated with illness and medication.
There are several findings from studies of cognitive performance that parallel the symptom-transmitter relationships. Catecholamine metabolism affects early automatic information processing in schizophrenia. Thus, DA metabolism modulates the duration of the masking influence in backward-masking tasks [23], and sensory gating of the startle response and event-related potentials [24,25]. The improvement of gating was associated with reductions of DA turnover (HVA/DA, [26,27]) and the noradrenaline (NA) metabolite (MHPG, [28]). Subsequent controlled information processing can also be influenced by each of the monoamines. Low levels of CSF HVA (but not other metabolites) correlate with poor executive functions, such as card-sorting and visuo-spatial recall [29]. Csernansky et al. [30] reported that CSF levels of HVA correlated with those of 5-HIAA, but only the latter related to performance on WAIS tests reflecting attention-related abilities (digit-span, digit-symbol, picture arrangement). With serum measures increased 5-HT turnover predicted not only the better gating of evoked potentials, but improved Tower-of-London executive function, especially for patients with symptoms of disorganisation [26].
In this study the cognitive tasks were chosen to reflect attentional and memory functions of the frontal and temporal lobes frequently impaired in schizophrenia [31]. We have reported on the neuropsychological performance and on the characterisation of subgroups in terms of monoamine metabolism: [11,32,33]. These are summarised in the appropriate sections. Here we look at the associations of monoamine activity with neuropsychological abilities, and whether such associations relate to major symptom dimensions (e.g. paranoid, thought-disorder and ideas-of-reference). The aim is to improve understanding of the bases modulating cognition and the expression of symptoms in patients with schizophrenia and with such a profile to improve the monitoring and titration of therapeutic measures for patients with differing symptom dimensions.
Based on the results reviewed above, we hypothesise 1) that some features apparently characteristic of the whole patient group (e.g. catecholamine activity) would be largely attributable to a given sub-group (e.g. paranoid vs. nonparanoid), with consequences for these patients' neuropsychological abilities and antipsychotic drug responses. More specifically, we suggest 2) that ratios of DA to 5-HT activity (e.g. HVA/5-HIAA) would predict cognitive impairments in disorganised patients, and that 3) ideas-of-reference, (rarely the target of investigation but with some similarities to productive symptoms) would be associated with catecholamine activity. It is important to bear in mind that the results reflect interactions of the type of illness and treatment and do not reflect exclusively the one or the other. Thus we discuss the results (i.e. the relationship of monoamine activity to task performance in subgroups of patients) separately from three perspectives, – profiles for symptom dimensions, for neuropsychological function and for monoaminergic activity, in turn.
Results
Clinical Groups
Patients had been ill on average for 9.5y after onset of the disorder at 23.6y and a first admission at 25.2y. The groups did not differ in age, socio-economic status of the parents or years in education. But the patients did show a lower non-verbal IQ (F1,161 = 49.4, P < 0.0001: table 1).
Table 1 Demographic and clinical data (means ± standard deviation) for 101 patients and 63 controls providing biochemical data on the left and neuropsychological performance scores on the right.
Demographic & Clinical Data Neuropsychological Tasks & Performance
Schizophrenics Controls Task Schizophrenics Controls
Age (years) 33.6 (11.0) 32.8 (11.0) Verbal fluency 28.6 (10.8) 35.0 (9.3)
Gender (m/f) 68 / 40 34 / 29 Mooney faces (hits) 9.8 (5.7) 8.6 (4.3)
Socio-economic group1 4.5 (1.9) 4.9 (1.6) Picture completion 11.8 (2.5) 14.3 (2.5)
Education (years) 13.3 (3.8) 13.7 (3.3) Block design 24.1 (9.5) 31.4 (7.7)
IQ (short APM) 6.9 (2.9) 9.9 (1.9) Vis Reprod (VR) 31.1 (5.0) 38.0 (3.5)
Handedness (Edinburgh) 17.5 (8.5) 18.9 (5.3) VR + delay 24.1 (9.1) 34.8 (6.0)
Onset-Age (years) 23.6 (8.4) Prose Reprod (PR) 17.7 (6.9) 29.8 (6.7)
First admission (years) 25.1 (9.6) PR + delay 13.1 (6.8) 26.3 (6.3)
Duration of illness (years) 9.5 (8.0) Trails B-A (sec) 79 (56) 35 (20)
Episode duration (days) 44.0 (40.3) Stroop Interfere (sec) 112 (38) 80 (20)
Diagnosis-
paranoid 70
disorganised 24
catatonic/residual 7
Symptoms,
PANSS positive 16.3 (6.0)
negative 18.6 (8.1)
general 36.8 (9.3)
SAPS ideas/reference 3.2 (3.8)
thought disorder 8.4 (6.5)
Extrapyramidal symptoms 5.8 (5.6)
AIMS 8.4 (2.9)
Antipsychotic drug dose
(CPZ: n = 99)2 665 (328)
typical + risperidone 572 (n = 49, 63% male, incl. 53% of paranoid, 29% of nonparanoid groups)
clozapine/olanzapine/sertindole 718 (n = 43, 67% male, incl. 33% of paranoid, 58% of nonparanoid groups)
Both 678 (n = 15, 47% male, incl. 14% of paranoid, 13% of nonparanoid groups)
Biperidene(mg/day: n = 15) 4.2 (1.8)
1. Scale 1–7, [84]: AIMS, Abnormal involuntary movement scale (n = 68/108 and 41/72); APM, Advanced progressive matrices; 2. CPZ, chlorpromazine equivalents, excluding 2 patients without medication; Biperidene is an anticholinergic drug (n = 15/108 and 9/72); PANSS, Positive and negative syndrome scale; SAPS, scale for assessment of positive symptoms; all neuropsychological tasks showed a significant group difference (p < 0.01) on point scores (low for patients) or latency (high for patients) except the Mooney faces test.
Group and Sub-Group Biochemical Measures
For the patient group as a whole, turnover for 5-HT was higher, for NA lower, and for DA there was no difference compared to controls. The HVA/5-HIAA ratio was lower in patients but along with catecholamine metabolites, levels increased and normalised more on atypical than typical antipsychotic drug treatment (Fig. 1). There was a trend for the HVA/MHPG ratio to be lower in the patient group (Fig. 2 left).
Figure 1 (Top) Serum monoamine and metabolite levels in controls (CON) and in patients treated with typical (TYP), atypical (ATYP) or both types of antipsychotic drug. Atypical > Typical, * P < 0.1; ** P < 0.08; *** P < 0.04; # P < 0.006 (covaried for age, IQ and CPZ). (Bottom) Turnover rates for 3 monoamines, and the ratio of dopamine (DA) to noradrenaline (NA) and serotonin (5-HT) metabolism (HVA/MHPG, HVA/5-HIAA), respectively. Controls (vs. whole patient group) showed less 5-HT TR (P < 0.05) and more HVA vs. 5-HIAA and MHPG (both ## P < 0.002).
Figure 2 Metabolite/monoamine turnover ratios for the whole subject group (bar diagram, left) and for patient sub-groups (tabular, right: PN, paranoid; NP, nonparanoid; ThD, thought disorder [+/-, high/low]; IoR, ideas of reference [+/-, high/low]). #P < 0.08, *P < 0.05, ** P < 0.01. 1 = levels of metabolite. Units: inter-amine ratios ×10; 5-HT TR ×103; DA and NA TR ×10.
Thought-disordered (disorganised) patients were characterised by low levels of HVA, HVA/5-HIAA and HVA/MHPG ratios, whereas high DA and 5-HT turnover and a high HVA/5-HIAA ratio were features of those with ideas-of-reference (IoR: Fig. 2 right). Nonparanoid patients showed a much higher NA turnover than the paranoid group. This was reflected in a trend for an increased DA turnover, where the HVA may have partly derived from NA metabolism. The increased NA turnover parallels plasma MHPG levels in patients with negative symptoms and the deficit-syndrome [55]. The non-paranoid group also had a larger HVA/5-HIAA ratio (Fig. 2 centre). We reported previously that increased DA D2 occupancy was predicted by decreases of DA metabolism and of HVA/5-HIAA ratios, especially in paranoid patients. DA D2 occupancy and catecholamine metabolism were unrelated in nonparanoid patients [32].
Neuropsychology: Noradrenergic Activity
Patients were impaired in the performance of all tests except the Mooney-faces-closure test [[33]: table 1]. Low NA turnover, attributable to low levels of the metabolite MHPG, was characteristic of the patient group as a whole (Fig. 1), but was a significant feature for the paranoid vs. the nonparanoid subgroup ([11] Fig. 2 centre).
A regression analysis showed significant effects of NA turnover on Stroop-interference, and verbal fluency, with opposite effects evident for controls and patients. For controls (n, 60) increases of Stroop-interference were associated with increases of NA turnover (F1,58 = 7.4, P = 0.009: r = +0.34, R2 = 11.3%: Fig. 3, left). The inverse effect, (increasing interference was associated with decreasing NA turnover), was evident for the patients (n, 95: partial correlation r = -.26, P < 0.01, Fig. 3 middle). However, increased verbal fluency also remained in the last step of this regression, and was clearly associated with decreases of NA turnover (F2,92 = 5.5, P = 0.005; r = -0.29, P = 0.004, Fig. 3 right). Together this explained 10.7% of the variance (R2). In other words, stimulation of the low levels of NA TR in patients would be expected to improve (reduce) interference in the incongruent Stroop condition but to impair (reduce) verbal fluency.
Figure 3 Partial correlations (left) for increased noradrenaline turnover (NA TR: Ln MHPG/NA) with increased Stroop-test interference scores in healthy controls (r = +0.34, P = 0.009), and (middle) for decreased NA TR with increased Stroop interference scores in patients with schizophrenia (r = -0.26, P = 0.01). But, (right) decreases of NA TR are associated with better verbal fluency in patients with schizophrenia (r = -0.29, P = 0.004).
There were no significant regression models predicting relationships for NA turnover with neuropsychological performance in paranoid patients (n 67) or the sub-groups with marked IoR (n 39) and with little thought disorder (n 43).
However, the non-paranoid (n 28), thought-disordered subgroups (n 51), and patients with few IoR (n 55) showed very similar negative associations for NA activity with verbal fluency. Thus, it was the nonparanoid, thought disordered patients who were responsible for the same effect seen with the patient group as a whole (Table 2 left). Nonparanoid patients with few IoR also showed improving immediate verbal recall as delayed verbal recall deteriorated with decreasing NA activity (Table 2 middle).
Table 2 Partial correlations for performance on 4 neuropsychological tasks reflecting left hemisphere (and cingulate) and two reflecting right hemisphere function with NA turnover in 3 patient sub-groups with few paranoid symptoms
Left hemisphere function Right hemisphere function
Verbal Stroop Immediate Delayed Trails B – A Delayed
Fluency Interference Verbal Recall Attention Set Visual Reproduction
p r p r p r p r p r p r p R2
NP n 28
F(3,24) = 8.5, 0.0005 -0.60 0.001 -- -- -0.50 0.01 +0.60 0.001 -- -- -- -- 51.4%
High-ThD n 51
F(3,47) = 5.1, 0.004 -0.44 0.002 -0.34 0.017 -- -- -- -- -- -- +0.29 0.04 24.6%
Low-IoR n 55
F(5,49) = 4.7, 0.0014 -0.39 0.015 -0.33 0.005 -0.35 0.013 +0.42 0.002 +0.34 0.015 -- -- 32.3%
PN = paranoid, NP = Nonparanoid, High-ThD = above median summed thought disorder ratings, Low-IoR = below median of summed ideas-of-reference ratings
Like the patient group as a whole, thought-disordered patients with few IoR showed increasing Stroop interference with decreasing NA turnover. Their right hemisphere functions such as set shifting latency (trails) and nonverbal recall were positively related to NA activity (Table 2). As patients were showing lower than normal NA activity, this means that decreases of NA turnover were associated with poorer delayed recall. Short-term information processing appeared to be enhanced by low NA activity (e.g. verbal productivity and immediate verbal recall in nonparanoid and thought disordered patients), but Stroop and Trails indicators of set-shifting were associated in opposite ways by changes of NA activity in patients with few IoR.
Neuropsychology: Dopaminergic Activity
The patients as a whole did not differ significantly from the healthy subjects on measures of DA turnover (Fig. 1: [11]). But a trend for less DA activity in patients became significant for the paranoid group (vs. nonparanoid). In contrast, patients with marked IoR showed a higher DA turnover than the other subgroups (Fig. 2).
For healthy controls (n 50) only the performance on the picture completion task predicted DA turnover. Better performance was associated with less DA activity (F1,48 = 6.8, P = 0.012, r = -0.35, R2 = 12.4%: Fig. 4 left). For the patient group as a whole (n 84) the dominant relationship for DA turnover was with the immediate recall of stories. The effect was the opposite of that described for NA turnover where immediate recall improved with decreasing NA activity (above). Here, poor immediate recall was associated with low, decreasing DA activity (F1,82 = 7.4, P = 0.008, r = +0.29, R2 = 8.2%: Fig. 4 middle).
Figure 4 Partial correlations (left) show that reduced dopamine turnover (DA TR: Ln HVA/DA) is associated with enhanced picture completion scores in healthy subjects (r = -0.35, P = 0.01). Increasing DA TR (middle) is associated with improved verbal recall in patients with schizophrenia (r = +0.29, P = 0.008). For patients with high ratings for ideas-of reference (IoR: right) increased DA activity was associated with promoting the speed of switching between sets on the trail-making test (r = +0.34, P = 0.015).
A further relationship of low DA turnover with poor verbal recall was characteristic of those patients expressing little thought-disorder and few IoR (Table 2). The one striking relationship for increased DA turnover was in the IoR subgroup. They tended to show high levels of DA activity. Here increased DA activity was associated with increased set-switching as indicated by decreased Trails B-A differences (Fig. 4, right).
Neuropsychology: Ratio of Catecholamine Metabolites (Activity)
In controls, relative catecholamine activity (HVA/MHPG) was unrelated to task performance. For the patients as a whole (n 85) the opposing nature of the associations of cognitive measures with DA and NA activity noted above was reflected in a trend (F7,77 = 1.8, P = 0.09, R2 = 14.1%). This showed opposite influences on the right hemisphere functions of immediate and delayed visual reproduction (r = +0.23, P = 0.04, r = -0.25, P = .03) and block-design (r = + 0.22, P = 0.05).
Interestingly these right hemisphere functions reflected the contributions of activity in the high IoR but nonparanoid subgroups (Table 3, right). In addition, in the high IoR group negative relationships emerged for the left hemisphere functions of verbal fluency and cingulate mediation of Stroop interference (Table 3, left). As the DA activity was relatively high in the high-IoR group, increases of the ratio HVA/MHPG may be considered as related to diminishing fluency, yet improved reduction of interference.
Table 3 Partial correlations for performance on 4 neuropsychological tasks reflecting left hemisphere (and cingulate) function and 4 tasks reflecting right hemisphere function with DA turnover (A: DA TR) and the ratio of DA to NA activity (B: HVA/MHPG) in patient subgroups with few paranoid symptoms
Left hemisphere function Right hemisphere function
Verbal Stroop Immediate Delayed Block Trails B-A Mooney Delayed
Fluency Interference Verbal Recall Design Attention Set Faces Visual Reproduction
p r p r p r p r p r p r p r p r p R2
a/ DA TR
NP n 28
F(1,26) = 5.6, 0.026 -- -- -- -- -- -- -- -- -- -- -- -- -0.42 0.026 -- -- 16.3%
Low-ThD n 41
F(1,39) = 8.7, 0.005 -- -- -- -- +0.43 0.005 -- -- -- -- -- -- -- -- -- -- 18.3%
High-ThD n 43 N.S. -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
Low-IoR n 48
F(1,46) = 3.9, 0.05 -- -- -- -- +0.28 0.05 -- -- -- -- -- -- -- -- -- -- 7.9%
High-IoR n 36
F(1,34) = 8.3, 0.007 -- -- -- -- -- -- -- -- -- -- -0.44 0.007 -- -- -- -- 19.6% -
b/HVA/MHPG
NP n 25
F(1,23) = 5.0, 0.036 -- -- -- -- -- -- -- -- +0.42 0.036 -- -- -- -- -- -- 17.8%
Low-ThD n 45
F(1,48) = 4.7, 0.037 -- -- -- -- -- -- +0.33 0.037 -- -- -- -- -- -- -- -- 10.9%
High-ThDn 45 N.S. -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
Low-IoR n 47
F(3,43) = 2.3, 0.095 +0.30 0.047 -- -- -- -- -- -- -- -- -- -- -- -- -- -- 13.6%
High-IoR n 38
F(4,33) = 4.7, 0.004 -0.35 -0.35 -0.50 0.002 -- -- -- -- +0.50 0.002 -- -- -- -- -0.36 0.03 36.3%
PN = paranoid, NP = Nonparanoid, Low-/High-ThD/IoR = below /above a median split of thought disorder/ideas-of reference ratings. NS = Model not significant
Neuropsychology: Serotonergic Activity
The patient group showed a slight but significantly higher level of 5-HT activity than the control group ([11] Fig. 2). But among the subgroups only those expressing many IoR showed more 5-HT activity than patients with few IoR. Neither the patient nor the control group, neither the paranoid nor the nonparanoid subgroup showed a model where a significant part of the 5-HT turnover was explained by neuropsychological task performance.
The only patients for whom 5-HT played a significant role in task performance were those who expressed little thought-disorder. This is of interest, firstly as 5-HT metabolism has been reported to play a role in the task performance of disorganised patients elsewhere (see discussion), and secondly because the task was also influenced by catecholamine metabolism. For patients without much thought disorder immediate verbal recall related positively to 5-HT activity (Table 4). There was no significant regression model for patients showing much thought disorder nor for those with a rather high 5-HT turnover (the high IoR subgroup).
Table 4 Partial correlations for performance on 4 neuropsychological tasks reflecting left hemisphere (and cingulate) function and 2 tasks reflecting right hemisphere function with 5-HT turnover (A: 5-HT TR) and the ratio of DA to 5-HT activity (B: HVA/5-HIAA) in patient sub-groups
Left hemisphere function Right hemisphere function
Verbal Immediate Picture Mooney Trails B – A Immedate Delayed
Fluency Verbal Recall Completion Faces Attention Set Visual Reproduction
p r p r p r p r p r p r p r p R2
a/ 5-HT TR
NP n 26 N.S. -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
PN n 70 N.S. -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
Low-ThD n 46
F(1,44) = 4.3 0.04 -- -- +0.30 0.04 -- -- -- -- -- -- -- -- -- -- 8.8%
High-ThD n 49 N.S. -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
Low-IoR n 55 N.S. -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
High-IoR n 40 N.S. -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
b/ HVA/5-HIAA
NP n 26 N.S. -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
PN n 59
F(1,57) = 4.1, 0.04 -- -- -- -- -- -- +0.26 0.04 -- -- -- -- -- -- 6.7%
Low-ThD n 41
F(5,35) = 3.5, 0.01 +0.36 0.028 -- -- -0.40 0.01 -- -- -- -- +0.45 0.005 -0.29 0.07 33.3%
High-ThD n 43
F(1,41) = 4.7, 0.035 -- -- -- -- -- -- +0.32 0.035 -- -- -- -- -- -- 10.4%
Low-IoR n 53 N.S. -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
High-IoR n 31
F'F(4,26) = 5.7, 0.002 -- -- -- -- -- -- +0.51 0.005 -0.50 0.006 +0.53 0.004 -0.48 0.009 46.7%
PN = paranoid, NP = Nonparanoid, Low-/High-ThD/IoR = below /above a median split of thought disorder/ideas-of reference ratings. NS = Model not significant
Neuropsychology: Ratio of DA to 5-HT Metabolites (Activity)
Was the relative activity of DA to 5-HT (HVA/5-HIAA) associated with any type of cognitive performance? In the controls (n 51), with a higher ratio than in the patients ([11] Fig. 1), HVA/5-HIAA was associated negatively with delayed visual reproduction (F1,49 = 10.2, P = 0.002, r = -0.41, R2 = 17.2%). Picture-completion, a measure of right hemisphere function, was also negatively related to DA activity, but there was no direct influence of 5-HT activity. The implication is that normal right hemisphere recall mechanisms are driven by DA activity and mildly modulated by 5-HT activity.
In the whole patient group, where low HVA/5-HIAA ratios tended to normalise on atypical (vs. typical) antipsychotic medication (Fig. 1), the relative activity of DA to 5-HT systems was related clearly to the performance on no less than 4 tasks (F5,79 = 3.5, P = 0.006). The one is characteristic of left hemisphere function (verbal fluency: r = +0.23, P = 0.04), while the other 3 tasks reflect right hemisphere function (Mooney-faces; r +0.26, p = 0.02: immediate visual reproduction; r = +0.31, P = 0.005: delayed visual reproduction; r -0.24, P = 0.03).
Analyses of the sub-groups showed that marked positive associations with the Mooney-faces task were contributed by patients who were paranoid, with many IoR (Table 4, bottom). Other right hemisphere functions where HVA/5-HIAA was important (e.g. visual reproduction and picture completion) were found in patients with many IoR but little thought disorder (Table 4). Striking is the large degree of variance explained for non-verbal memory processes, Mooney faces recognition and trails' measures of set-shifting in patients with marked IoR (whose HVA/5-HIAA ratios were significantly above those with few IoR).
Discussion
This is the first comprehensive report of potential functionally-related associations between neuropsychological performance and plasma markers of the overall activity of 3 monoamines in normal subjects and those with schizophrenia. Novel is that the regression analyses take account of the heterogeneity of schizophrenia, whereby 3 major dimensions of the clinical presentation of schizophrenia were examined: paranoid-nonparanoid, disorganisation/thought-disorder and ideas-of-reference. The tasks represented cognitive contributions of the frontal or temporal lobes in the left or right hemispheres. We now draw together profiles with respect to the variance explained from the perspective of nosological, neuropsychological and biochemical descriptors.
On the variance explained
First we consider the main results from the point of view of the degree of variance explained in the regression analyses. Considering the whole groups of patients and controls, there was no example of neuropsychological performance relating to the variance of 5-HT turnover. This may reflect an over-proportional contribution from peripheral organs [56]. Theoretically, opposite effects in mutually exclusive subgroups could also be responsible. In practice it should be noted that 5-HT metabolism did have a modulatory effect in some subgroup analyses. In contrast, there were examples in both healthy and patient groups of cognitive performance having a predictive relationship for 8–12% of the variance of NA and DA turnover. A correlation with frontal lobe function was evident with NA activity (Stroop and verbal fluency), while DA activity related to posterior temporo-parietal function (picture completion and immediate verbal recall).
With respect to the subgroups of schizophrenia, the greatest amount of variance was explained in analyses showing similar contributions of the performance on 3–4 tasks to the monoaminergic ratio considered. This is not surprising as the widespread innervation by monoaminergic systems would be expected to contribute to functions sensitive to several tasks. The most striking example was with non-paranoid NA turnover as the independent variable. Here 3 left-hemisphere functions contributed over 50% of the variance (verbal fluency, immediate and delayed verbal recall). Three more large contributions to the variance related to the relative activity of the monoamines. Two concerned the patients with marked ideas-of-reference (IoR). For HVA/5-HIAA, right-hemisphere function predominated (trails, Mooney-faces, immediate and delayed visual recall: 47% of the variance). This may implicate a more dominant contribution of DA activity in view of the sparse associations for 5-HT turnover (see below). HVA/MHPG also correlated with performance on 4 tasks (36% of the variance). The two right hemisphere tasks may reflect the contribution of DA activity to more posterior parietal and temporal lobe function (block-design and visual delayed recall, see previous paragraph). Tasks reflecting the left-hemisphere (verbal fluency) and midline cingulate function (Stroop) may reflect the dominant role of NA activity, prominent in nonparanoid subjects (see above).
Profiles – nosological
In paranoid patients the single significant association concerned HVA/5-HIAA and Mooney-faces performance. This positive association contrasts with the negative relationship with DA activity seen in nonparanoid subjects for Mooney faces. The contrast points to the potential negative influence of increasing DA activity (levels were higher in the nonparanoid group), and the positive influence of increasing HVA/5-HIAA ratios with atypical antipsychotic drugs (that benefited the paranoid patients in particular). Further the paranoid/nonparanoid contrast illustrates well a point made in the introduction: namely, the whole patient group did not show impaired performance on the Mooney faces task (Table 1) despite clearly opposing tendencies in the constituent sub-groups. Otherwise nonparanoid patients were characterised by a) the absence of a relationship between cognitive function and 5-HT activity, where a previous report found low activity related to slow discrimination learning [57], and b) a strong relationship with NA activity (nonparanoid> paranoid, Fig. 3), where increases related to improved delayed verbal recall, but to less verbal productivity – that is usually poor in this subgroup [58].
Like the nonparanoid group, patients with little thought-disorder showed a relationship between frontal functions and NA activity. But they also showed a relationship of the performance on 5 tasks to an aspect of 5-HT activity (turnover or the ratio to DA activity). Indeed one might posit a similar role for the relative strength of DA/5-HT activity in immediate verbal and visual recall, picture completion and even verbal fluency. In contrast those with high levels of thought-disorder had a limited set of associations with left frontal (NA) function and right parietal involvement (HVA/5-HIAA). The relatively high levels of 5-HT activity found in these thought-disordered patients tends to be supported by drug-challenge, genetic and plasma measures [59-61].
For subjects with few or with marked IoR there was a remarkable set of associations between cognition and monoamine activity. For those with few IoR left hemisphere (frontal and temporal lobe) cognitive function related to NA metabolism. But in patients with strong IoR, right hemisphere functions (trails, Mooney-faces, immediate and delayed visual recall) were related to HVA/5-HIAA. It is notable that some of the latter functions incur parietal activity that was reported to be high in imaging studies of patients with marked IoR [62].
Profiles – monoamines
NA activity correlated with neuropsychologic function in patients with much thought disorder and those who were nonparanoid with few IoR. Decreasing NA metabolism was favourably related to improved verbal fluency, and less Stroop interference. Further the decreases of NA metabolism in nonparanoid patients with few IoR related to better immediate verbal recall. These normalising effects of decreasing NA activity in thought disordered patients concur with reports on plasma levels of MHPG [63], and results achieved on trials of NA antagonists [64,65].
Increases of DA activity related positively to improved immediate verbal recall in those without much thought disorder or IoR. On the other hand increasing DA activity was related to reduced Mooney-faces scores in nonparanoid patients and faster switching of set in the trails task. This is consistent with decreases of switching performance registered after treatment with DA antagonists (e.g. antisaccades, trails, Stroop; [66]). The relative degree of DA to NA activity (i.e. HVA/MHPG) was important for left and right hemisphere function (4 tasks) especially in those with marked IoR (Table 2, bottom).
For 5-HT activity, the important issue for cognitive function was its activity relative to that of DA (HVA/5-HIAA). This may reflect the opposing influences of DA and 5-HT on the processes underlying efficient cognitive function (e.g. working memory, [67]). The importance of the ratio was evident in patients with productive symptoms (paranoid), much thought-disorder and IoR. The direction of the association could be negative or positive, but remarkable is that the functions largely belong to the domain of the right hemisphere (Table 4 bottom).
Profiles – Neuropsychology
Here we point out only the strongest relationships for the 10 neuropsychological tasks. For verbal fluency and Stroop interference (signs of frontal and cingulate functions) correlations were seen in highly symptomatic patients. Verbal fluency was reduced with increasing NA activity in the thought-disordered and those with negative symptoms (nonparanoid). This seems to parallel the sharp rise (then fall) of negative symptoms and plasma NA levels in a study of the effects of ketamine-induced psychosis [68]. However, they reported associations with impaired verbal memory rather than fluency – although it is difficult to be sure what the state of the NA response was in each task. NA activity was also implicated in those with many IoR where decreasing HVA/MHPG ratios were associated with increased Stroop interference. For those with high IoR, decreasing latencies on the trails B-A task, a frontal sign of the ability to shift between set, related to increasing DA turnover and HVA/5-HIAA ratios. In the same patients worsening block-design and Mooney-faces performance (signs of parietal function) related to decreasing HVA/MHPG and HVA/5-HIAA, respectively. Among the few studies of this patient subgroup, others have noted an association between IoR symptoms and performance on tests of self-monitoring abilities such as these [69], and for ketamine-induced IoR with DA binding activity [70].
A remarkable feature across analyses is that the signs of the correlations for immediate verbal or visual recall usually changed for the delayed recall version (Table 2 left, Table 4 right). Nonetheless poorer immediate verbal recall was associated with decreasing DA turnover in the low thought-disorder group, while poorer delayed verbal recall related to decreasing NA turnover in the nonparanoid patients. One could perhaps generalise this result: poor recall abilities are associated with low catecholamine activity, especially in patients not showing positive and productive symptoms.
General: methods – advantages/disadvantages
Many previous reports have deliberated over the interpretation of biochemical data collected remotely from the source; CSF, blood (plasma) and urinary measures provide peripheral data remote from the CNS monoamine pathways producing them. Nonetheless, all the products of monoamine metabolism will spill over and be reflected in peripheral catchments on the way to breakdown and excretion. Major psychological or psychiatric states and drug treatment are the most likely agents for changes of transmitter activity from the norm. Schizophrenia and antipsychotic drugs are here widely documented cases in point, where changes of monoaminergic activity clearly affect peripheral and central functions alike.
We argue that the obvious disadvantages of this study in not being able to measure at source are partly counteracted by two innovations. The first is the more precise description of the clinical state. We suggested in the introduction that clinicians easily distinguish the confused thought processes of disorganised patients from those who hallucinate frequently. Further, the clearly separate substrates responsible are not only reflected in different neuropsychological performance and state but reflect different levels of monoaminergic activity. By separating these factors it should be possible to find relationships between cognition and monoamine activity in the direction of those reported here. We limited ourselves to extreme states on the dimensions of paranoia, disorganisation and ideas-of reference (6 groups). Others would do well to study the relations for negative symptoms: (e.g. are plasma HVA levels higher in deficit vs. nondeficit schizophrenia [55] or vice versa due to symptom profile [71] or typical antipsychotic treatment [72], or is the type of onset critical (abrupt/gradual) where the age of onset was unrelated to plasma HVA levels [10]?).
The second innovation was the use of turnover measures of metabolism and inter-amine activity ratios. This was intended to resolve the following issue. Reliance on amine measures may, in the case of (say) high levels, reflect increased release resulting from increased metabolic demand, and increased impulse flow [73]. But they may also reflect production over the need, as often seen when circumstances change or the ongoing state inhibits an adaptive reduction of release. With these innovations we proposed to find changes and relationships that have otherwise been masked by the study of subgroups with opposing tendencies within the population under investigation. This may have been the case with reports of the absence of associations found in relatively chronic adult patients [74] and adolescents with a more recent onset [75]. A third important feature of this study helped add precision to a close description of psychiatric state and the general state of monoaminergic activity. This has been the ability through recent neuroimaging and brain-damage studies to relate the above changes of biochemical and psychological state to the function of gross divisions of the CNS in performing well-studied neuropsychological tasks [76].
General: conflict and consistency in CSF monoamine findings
CSF levels of NA were reported to be unrelated to verbal abilities, arithmetic or memory function [74,75,77]. CSF MHPG levels were not associated with performance on tasks testing memory and executive function [29]. Such negative results are unexpected considering that high NA (potentially reflecting a low turnover) and decreasing MHPG levels in the CSF have been associated with increasing psychopathology [77,78]. But, some apparently conflicting findings [77] are consistent with our results in one aspect. The psychopathology they described related to positive symptoms: it was in such paranoid patients that we found a lower NA turnover. Further the absence of a relationship for NA levels with verbal abilities and memory [77] is explicable by our finding that decreasing NA activity related to improved verbal fluency and recall. The subtle difference is that we found this relationship most evident in patients who suffered much thought disorder or few IoR or paranoid symptoms – those with higher NA turnover.
CSF levels of 5-HIAA have been related to performance on tests of attention, working memory and executive function as measured by the digit-symbol, digit-span and picture arrangement tasks [19,30]. Both groups noted a correlation between HVA and 5-HIAA levels, and the relationship of HVA/5-HIAA to both cognitive measures and symptom ratings [19]. So it is not surprising and supportive of our methods that we also found that a clinical improvement following atypical antipsychotic drug-treatment was associated with increased HVA/5-HIAA ratios [11], and that several cognitive functions (associated with the right hemisphere) improved or worsened with increases and decreases of the HVA/5-HIAA ratio, respectively. Novel was the finding that this relationship was marked for patients with many IoR and little thought disorder in the trails-measure of attention-shift and in visual memory. In contrast, those with thought-disorder and paranoia showed similar relationships but restricted to the Mooney-faces-closure test (Table 4).
It was surprising to find relatively few relationships between DA activity and task performance, considering there was a broad range of high to low values among the healthy subjects and suppressed levels among patients – to which both the illness and medication would have contributed. However, it may be noted that even with other patient groups (e.g. HIV: [79]), where CSF HVA levels were <50% of the non-affected controls, correlations were restricted to response slowing on tests of executive attention and concentration and unrelated to memory performance. For elderly subjects an inverse relationship between plasma HVA and bradykinesia was reported [80], which to a degree is consistent with studies of schizophrenia [29] and schizotypal disorder [17] where more neuropsychological impairment was observed with decreasing levels of CSF HVA. Our finding of a positive relationship with HVA for immediate verbal recall in patients without thought disorder and IoR (Table 2) fits into this pattern, but contrasts with plasma HVA findings in a sample of chronically ill patients [81]. But we agree with a report from the same research group [19], that where there was a lack of association with HVA alone, the ratio HVA/5-HIAA explained performance better. (Note: typical antipsychotic drugs impair performance on many tests of memory [82], but some of the above results were obtained with non-medicated subjects [17] and, in this report with subjects on atypical antipsychotic medication.)
The one intriguing exception to these reports on the effects of decreases of HVA concerned patients showing many IoR. These patients not only showed unusually high levels of DA metabolism but also a correlation of shorter set-shift latencies on the trails' test with increasing DA turnover. This fits well with an early formulation of the role of DA in switching between information processing channels [83] and could form the basis of a working hypothesis for future study of the bases of these patients' illusions. Namely, that the increased DA activity permits switching between multiple associations for a given sensory input, permitting in the pathological case a grossly inappropriate interpretation of incoming information.
In conclusion, the profiles of the associations of neuropsychological performance with changes of activity of dopamine, noradrenaline and serotonin (and their activities relative to each other), can assist with an understanding of the modulation by monoaminergic pathways of frontal and temporal lobe function. This pertains even though the origin of the peripheral monoamine measures cannot be attributed to any particular anatomical source. Since this report concentrates on patients showing strong or few symptoms on three major dimensions of psychopathology, the profiles for these sub-groups of schizophrenia should help to inform goal-directed interventions in the treatment of patients with these particular features.
Methods
Subjects
Using DSM-IV criteria [34] 108 patients from the University Psychiatry Clinics were diagnosed with schizophrenia by the senior ward physician and subsequently by 2 senior physicians associated with this research. Affective, schizoaffective and schizophreniform psychoses were excluded. The undifferentiated subtype was regarded as a residual category that contrasts with the paranoid, disorganised and catatonic subtypes. Patients were screened to exclude other major psychiatric or somatic illness, alcohol abuse in the last 5 years and substance abuse other than nicotine at the time of testing. A group of 63 healthy controls, recruited by advertisement and paid for their participation, was closely matched for age, education, socio-economic family status and handedness. Exclusion criteria for controls were the same as for patients, and based on a semi-structured interview: they reported no family history of psychosis, nor that they had ever consulted with a psychiatrist or psychologist. Following approval from the Hospital Ethics Committee, in accord with the Declaration of Helsinki informed signed consent was obtained from each patient and the responsible care-giver, and from each healthy participant. Clinical and demographic data are given in table 1.
The study took place in the post-acute phase, on average 6 weeks after the patient's admission (range 4–8 weeks) and following stabilisation of treatment. Of 101 patients providing biochemical data 99 were treated with antipsychotic drugs according to their clinical requirements. Drug doses were converted to chlorpromazine equivalents (CPZ: [[35-38], correspondence with the suppliers of olanzapine and sertindole]). Patients receiving conventional, atypical or both categories of antipsychotic drug were considered separately and together [[11]: table 1]. Symptoms were rated with the Positive and Negative Syndrome Scale (PANSS: [39], along with Schneiderian ideas-of-reference (ego-disturbance) and thought-disorder items from the Scale for Assessment of Positive Symptoms (SAPS: [6]) that are under-represented in the PANSS. The subtypes of schizophrenia was based on a factor analysis of the symptom ratings. This resulted in 4 dimensions (varimax rotation, eigen values >1, excluding those contributing <5% to the variance: [3]): (1) disorganised (especially thought/concept disorder), (2) nonparanoid (largely negative symptoms), (3) ideas-of-reference and (4) paranoid (positive symptoms). Four categorical groupings were made on the basis of this factor analysis. Thought disorder (ThD) and ideas-of-reference (IoR) ratings were split at the median to provide groups with high vs. low levels of the respective symptom clusters. Paralleling the paranoid/nonparanoid (PN/NP) diagnostic split, two mutually exclusive groups with PN and NP symptoms were formed. This procedure resulted in a comparison of 6 subgroups of patients with schizophrenia.
Neuropsychology
Ten tasks were administered. The verbal-fluency test [40] required the generation of as many words as possible starting with the letter F, A or S (1 minute each). In the trail-making test subjects were asked to join up in sequence first a series of numbers (form-A), then an alternating series of letters and numbers (form-B, e.g. 1-A-2-B-3; the score used was the form-B-minus-A latency; [41]). These tests reflect functions in the left and right frontal lobes, respectively [33]. The Stroop test interference score was the increased latency to name the print colour of a word that named a different colour compared to the latencies to name colours and words naming colours. Performance activates the frontal and cingulate cortices [42,43].
The block-design and Mooney faces closure tests reflect broadly parietal function. Block-design required that a given square form was reconstructed out of 4 or 9 pieces [44]. The modified Mooney faces closure test asked for the classification of the age of degraded images of faces [45,46]. In the picture-completion test (a reflection of temporo-parietal function) the subject identified the missing feature on a picture of an everyday scene [44]. Visual reproduction and logical memories under conditions of immediate- and delayed-recall [47] involved a presentation of a series of visual patterns or two short stories for recall, and reflect right- and left-sided temporal lobe function in visuo-spatial and verbal memory, respectively. In addition the short 12-item form of the Advanced Progressive Matrices (APM) was used as a measure of IQ, where scores <6 are below and scores of 12 are above average [[48]: table 1].
Serum assessments and assays for monoamines and their metabolites
A 30 ml blood sample was taken at 08.00 (± 30 min) after 10 h of fast and rest before smoking, medication, exercise and breakfast. The sample was centrifuged for 10 min at 2000 g and stored at -70°C until analysis. All samples were analysed blind to their origin by reversed phase high performance liquid chromatography with a glassy carbon electrochemical detector using internal standards. Separate isocratic determinations were run for a) DA and NA, b) 5-HT with organic-sodium dihydrophosphate mobile phases [after [49]], c) HVA, 5-HIAA, [50] and d) MHPG with organic sodium-acetate mobile phases [51]. Intra- and inter-assay coefficients of variation were, DA 11.6/8.9%, NA 11.6/9.1%, 5-HT 2.3/9.2%, HVA 7.1/19.8%, 5-HIAA 6.0/7.3% and MHPG 6.9/18.1%, respectively. Recovery ranged from ca. 74–80% with the following sensitivities (ng/mL) for DA 0.5, NA 0.01, 5-HT 1.0, HVA and 5-HIAA 1.25, and MHPG 1.0.
Serotonergic measures allowed for the taking of comparable metabolic measures between monoamines. Circulating levels can be expressed in terms of platelet numbers, their protein content, or per unit volume of blood, serum or platelet-rich plasma. The first two methods may be contaminated by non-platelet derivatives and their proteins, while the latter 3 measures can vary with efflux from the platelets. Nonetheless, a high correlation between platelet and circulating levels with excellent intra-individual replicability at 3 months has been repeatedly reported for healthy subjects [52,53]. Our values match those of Jernej [53] and 5 studies discussed (98–312 ng/ml), and are a bit lower than in 3 reports on platelet-rich plasma and serum (269–271 ng/ml). Turnover rates were similar to plasma values [54]. The scarcity of outliers (± 2 SD) was similar between the monoamines, does not point to irregular platelet-release and emphasises the lack of variability.
Data treatment
The biochemical data from 5 subjects were above a cut-off criterion for outliers of the mean ± 2 SD and removed from the analysis. For technical reasons some data were missing from individual measures and covariates for 15 subjects. This resulted in full sets of data across all measures in 73 patients and 37, less a further 10 patients for the metabolite ratios. Their demographic and clinical characteristics did not differ from the original sample [32]. The number of subjects used in each analysis is cited below in parentheses. Normal distributions of the biochemical data were obtained after natural logarithmic transformations.
Separate multivariate analyses of covariance were used to compare task performance or biochemical measure between subject groups. Age and IQ were related to some biochemical measures, and therefore used as covariates. Other demographic variables (e.g. education, status) were not used: they were matched between groups, their variance was not large and their communal contribution overlapped with age and IQ. Biochemical data were analysed in two stages. The first considered the 3 monoamines and their 3 metabolites [11]. The second, forming the basis for the present report, concerned the 3 turnover ratios and two inter-amine metabolite ratios (e.g. HVA/5-HIAA, HVA/MHPG). The differential effect of conventional and atypical medication was controlled for the effect of dose with the use of chlorpromazine equivalents (CPZ) as covariate. The locus of effect was sought by t-tests (patient data) and one-way analyses of variance (neuropsychology and biochemistry). The relationships of clinical subgroups to biochemical measures, and neuropsychological performance were reported in Oades et al. [11,32,33] and are described here only in summary form.
The emphasis in the present report lies with an explanation of the contribution to test performance by the groups and subgroups of patients and the controls made by measures of plasma monoamine activity with the use of standard, linear regression models. First task performances (n 10) were entered into equations for turnover measures in the patient and control groups. Following significant standard regression models at α = 0.01, the relationships for the 6 sub-groups of patients for explanation by 5 measures of biochemical ratios were explored with backward, step-wise, linear regression models. We emphasise an α-correction to 0.001% for significance. However, in view of the exploratory nature of the study and the inherent uncertainties associated with the models, we describe trends up to levels of <0.05%.
Competing interests
The author(s) declare that they have no competing interests
Abbreviations
DA, dopamine, CSF cerebrospinal fluid, HVA homovanillic acid, IoR ideas-of-reference, MHPG 3-methyl-4-hydroxyphenylglycol, NA noradrenaline, NP nonparanoid, PN paranoid, ThD thought-disorder, 5-HT serotonin, 5-HIAA 5-hydroxyindoleacetic acid,
Acknowledgements
We thank S. Bender, A. Czwink, A. Dolberg, I. Eisenbeis, L. Fierro, M. Geissler, C. Heeper, A. Hesse, B. Müller, G. Sartory and J. Wolstein for their help in various aspects of this work and the DFG for grants Oa 4/1-1, 1–2.
==== Refs
Bleuler E Dementia Praecox oder Gruppe der Schizophrenien Vienna, Deuticke 1911
Carpenter WT Kirkpatrick B Buchanan RW Schizophrenia: syndromes and diseases J Psychiatr Res 1999 33 473 475 10628522 10.1016/S0022-3956(99)00015-1
Bender S Müller B Sartory G Oades RD Conditioned blocking and schizophrenia: a replication and study of the role of symptoms, age, onset-age of psychosis and illness-duration Schizophr Res 2001 49 157 170 11343874 10.1016/S0920-9964(00)00040-2
Amin F Davidson M Kahn RS Schmeidler J Stern R Knott PJ Apter SH Assessment of the central dopaminergic index of plasma HVA in schizophrenia Schizophr Bull 1995 21 53 66 7770741 10.1016/0920-9964(95)95169-A
Amin F Silverman JM Siever LJ Smith CJ Knott PJ Davis KL Genetic antecedents of dopamine dysfunction in schizophrenia Biol Psychiatry 1999 45 1143 1150 10331106 10.1016/S0006-3223(98)00262-5
Andreasen NC Olsen S Negative versus positive schizophrenia: definition and validation Arch Gen Psychiatry 1982 39 789 794 7165478
Galinowski A Poirier MF Aymard N Leyris A Beauverie P Bourdel MC Loo H Evolution of plasma homovanillic acid (HVA) in chronic schizophrenic patients treated with haloperidol Acta Psychiatr Scand 1998 97 458 466 9669519
Garver DL Steinberg JL McDermott BE Yao JK Ramberg JE Lewis S Kingsbury SJ Etiologic heterogeneity of the psychoses: is there a dopamine psychosis? Neuropsychopharmacology 1997 16 191 201 9138435 10.1016/S0893-133X(96)00186-8
Zhang ZJ Peet M Ramchand CN Shah S Reynolds GP Plasma homovanillic acid in untreated schizophrenia – relationship with symptomatology and sex J Psychiat Res 2001 35 23 28 11287053 10.1016/S0022-3956(01)00008-5
Sharma RP Javaid JI Davis JM Janicak PG Pretreatment plasma homovanillic acid in schizophrenia and schizoaffective disorder: the influence of demographic variables and the inpatient drug-free period Biol Psychiatry 1998 44 488 492 9777181 10.1016/S0006-3223(97)00451-4
Oades RD Klimke A Henning U Rao ML Relations of clinical features, subgroups and medication to serum monoamines and schizophrenia Hum Psychopharmacol 2002 17 15 27 12404703 10.1002/hup.368
Pliszka SR Maas JW Javors MA Rogeness GA Baker J Urinary catecholamines in attention-deficit hyperactivity disorder with and without comorbid anxiety J Am Acad Child Adolesc Psychiatry 1994 33 1165 1173 7982867
Bowers MB Swigar ME Jatlow PI Goicoechea N Plasma catecholamine metabolites and early response to haloperidol Clin Psychiatry 1984 45 248 251
Bowers MB Swigar ME Jatlow PI Hoffman FJ Goicoecha N Early neuroleptic response: clinical profiles and plasma catecholamine metabolites J Clin Psychopharmacol 1987 7 83 86 2884237
Sharma RP Javaid JI Janicak PG Faull K Comaty J Davis JM Plasma and CSF HVA before and after pharmacological treatment Psychiat Res 1989 28 97 104 10.1016/0165-1781(89)90201-1
Siever LJ Tamminga CA, Schulz SC The biology of the boundaries of schizophrenia Schizophrenia: Advances in Neuropsychiatry and Psychopharmacology 1991 New York, Raven Press
Siever LJ Biologic factors in schizotypal personality disorders Acta Psychiat Scand 1994 90 45 50
Scheepers FE Gispen-De-Wied CC Westenberg HG Kahn RS The effect of olanzapine treatment on monoamine metabolite concentrations in the cerebrospinal fluid of schizophrenic patients Neuropsychopharmacology 2001 25 468 475 11557160 10.1016/S0893-133X(01)00250-0
Forrest TJ Allen DN Weiner C Yao JK Gurklis JA Van Kammen DP The HVA/5-HIAA ratio and cognitive function in schizophrenia Schizophr Res 2001 49 52 53
Wieselgren I-M Lindström LH CSF levels of HVA and 5-HIAA in drug-free schizophrenic patients and healthy controls: a prospective study focused on their predictive value for outcome in schizophrenia Psychiat Res 1998 81 101 110 10.1016/S0165-1781(98)00090-0
Kahn RS Davidson M Knott P Stern RG Apter SH Davis KL Effect of neuroleptic medication on cerebrospinal fluid monoamine metabolite concentrations in schizophrenia: serotonin-dopamine interactions as a target for treatment Arch Gen Psychiatry 1993 50 599 605 7688208
Maas JW Bowden CL Miller AL Javors MA Funderburg LG Berman N Weintraub ST Schizophrenia, psychosis and cerebral spinal fluid homovanillic acid concentrations Schizophr Bull 1997 23 147 154 9050120
Butler PD Harkavy-Friedman JM Amador XF Gorman JM Backward masking in schizophrenia: relationship to medication status, neuropsychological functioning, and dopamine metabolism Biol Psychiatry 1996 40 295 298 8871777 10.1016/0006-3223(96)00007-8
Waldo MC Adler LE Leonard S Olincy A Ross RG Harris JG Freedman R Familial transmission of risk factors in the first-degree relatives of schizophrenic people Biol Psychiatry 2000 47 231 239 10682220 10.1016/S0006-3223(99)00272-3
Bender S Schall U Wolstein J Grzella I Zerbin D Oades RD A topographic event-related potential follow-up study on 'prepulse inhibition' in first and second episode patients with schizophrenia Psychiat Res Neuroimaging 1999 90 41 53 10320210 10.1016/S0925-4927(98)00053-5
Oades RD Bender S Schall U Klimke A Balcar A Thienel R Improved indices of attention in schizophrenia predicted by increased serotonin metabolism: admission /discharge comparisons in first/second episode patients Psychophysiology 1999 36 86 10098383 10.1017/S0048577299961565
Oades RD Müller B Schall U Bender S Wolstein J Automatic vs. controlled attention in schizophrenia: conditioned blocking and sensory gating Behav Pharmacol 2000 11 346
Waldo MC Gerhardt G Baker NJ Drebing C Adler LE Freedman R Auditory sensory gating and catecholamine metabolism in schizophrenic and normal subjects Psychiat Res 1992 44 21 32 10.1016/0165-1781(92)90066-C
Kahn RS Harvey PD Davison M Keefe RSE Apter SH Neale JM Mohs RC Davis KL Neuropsychological correlates of central monoamine function in chronic schizophrenia: relationship between CSF metabolites and cognitive function Schizophr Res 1994 11 217 224 7514887 10.1016/0920-9964(94)90015-9
Csernansky JG King RJ Faustman WO Moses JA Poscher ME 5-HIAA in cerebrospinal fluid and deficit schizophrenic characteristics Br J Psychiatry 1990 156 501 507 1696842
Zakzanis KK Kaplan E Leach L Neuropsychological differential diagnosis Lisse, The Netherlands, Swets & Zeitlinger 1999
Oades RD Rao ML Bender S Sartory S Müller BW Neuropsychological and conditioned blocking performance in patients with schizophrenia: assessment of the contribution of neuroleptic dose, serum levels and dopamine D2-receptor occupancy Behav Pharmacol 2000 11 317 330 11103886
Oades RD Müller BW Bender S Sartory G Neuropsychological indicators of heteromodal cortex (dys)function relevant to the expression and impairment of conditioned blocking measures of selective attention in schizophrenia Cogn Neuropsychiatry 2001 6 41 61 10.1080/13546800042000043
American Psychiatric Association Diagnostic and statistical manual of mental disorders: DSM-IV 1994 4 Washington, D.C
Benkert O Hippius H Psychiatrische Pharmakotherapie 1986 4 Heidelberg, Springer-Verlag
Rey M-J Schulz P Costa C Dick P Tissot R Guidelines for the dosage of neuroleptics. 1: Chlorpromazine equivalents of orally administered neuroleptics Int Clin Psychopharmacol 1989 4 95 104 2568378
Schulz P Rey M-J Dick P Tissot R Guidelines for the dosage of neuroleptics. II: Changing from daily oral to long-acting injectable neuroleptics Int Clin Psychopharmacol 1989 4 105 114 2568377
Kane JM Drug therapy: Schizophrenia New England J Med 1996 334 34 41 7494570 10.1056/NEJM199601043340109
Kay SR Opler LA Fizbein A The Positive and Negative Syndrome Scale (PANSS) 1992 Toronto, Canada, Manual Multi-Health Systems
Benton AL Hamsher K Multilingual Aphasia Examination 1989 Iowa City, IA, AJA Associates
Reitan RM The validity of the Trail Making Test as an indicator of organic brain damage Percept Mot Skills 1958 8 271 276
Leung H-C Skudlarski P Gatenby JC Peterson BS Gore JC An event-related functional MRI study of the Stroop color word interference task Cereb Cortex 2000 10 552 560 10859133 10.1093/cercor/10.6.552
MacDonald AW Cohen JD Stenger VA Carter CS Dissociating the role of the dorsolateral prefrontal and anterior cingulate cortex in cognitive control Science 2000 288 1835 1838 10846167 10.1126/science.288.5472.1835
Wechsler D Wechsler Adult Intelligence Scale-revised (WAIS-R) 1981 New York, NY, The Psychological Corporation
Lansdell H Relation of extent of temporal removals to closure and visuomotor factors Percept Mot Skills 1970 31 491 498 5492324
Mooney CM Ferguson GA A new closure test Canad J Psychol 1951 5 129 133 14870072
Wechsler D The Wechsler Memory Scale-Revised San Antonio, TX 1987 The Psychological Corporation, Harcourt Brace Jovanovich
Raven JC Kratzmeier H, Horn R Raven-Matrizen-Test: Advanced Progressive Matrices German language manual 1976 Stuttgart, Germany, Beltz-Verlag
Eriksson B-M Persson B-A Determination of catecholamines in rat heart tissue and plasma samples by liquid chromatography with electrochemical detection J Chromatogr 1982 228 143 154 7076741
Gypta RN Whelton C Determination of plasma homovanillic acid by liquid chromatography with electrochemical detection J Chromatogr 1992 582 236 241 1491047
Minegishi A Ishizaki T Determination of free 3-methoxy-4-hydroxy-phenylglycol with several other monoamine metabolites in plasma by high-performance liquid chromatography with amperometric detection J Chromatogr 1984 311 51 57 6520169
Croonenberghs J Delmeire L Verkerk R Lin AH Meskal A Neels H Van der Plancken M Scharpe S Deboutte D Pison G Maes M Peripheral markers of serotonergic and noradrenergic function in post-pubertal, Caucasian males with autistic disorder Neuropsychopharmacology 2000 22 275 283 10693155 10.1016/S0893-133X(99)00131-1
Jernej B Banovic M Cicin-Sain L Hranilovic D Balija M Oreskovic D Folnegovic-Smalc V Physiological characteristics of platelet/circulatory serotonin: study on a large human population Psychiat Res 2001 94 153 162 10.1016/S0165-1781(00)00129-3
Dursun SM Whitaker RP Andrews H Reveley MA Effects of natural ageing on plasma 5-HT turnover in humans Hum Psychopharmacol 1997 12 365 367 10.1002/(SICI)1099-1077(199707/08)12:4<365::AID-HUP879>3.0.CO;2-2
Thibaut F Ribeyre J-M Dourmap N Menard J-F Dollfus S Petit M Plasma 3-methoxy-4-hydroxyphenylglycol and homovanillic acid measurements in deficit and non-deficit forms of schizophrenia Biol Psychiatry 1998 43 24 30 9442341 10.1016/S0006-3223(97)00023-1
Lambert GW Kaye DM Cox HS Vaz M Turner AG Jennings GL Esler MD Regional 5-hydroxyindoleacetic acid production in humans Life Sci 1995 57 255 267 7541101 10.1016/0024-3205(95)00269-C
Oades RD Stimulus dimension shifts in patients with schizophrenia, with and without paranoid hallucinatory symptoms, or obsessive compulsive disorder: strategies, blocking and monoamine status Behav Brain Res 1997 88 115 132 9401715 10.1016/S0166-4328(97)02304-8
Himmelhoch S Taylor SF Goldman RS Tandon R Frontal lobe tasks, antipsychotic medication and schizophrenia syndromes Biol Psychiatry 1996 39 227 229 8837987 10.1016/0006-3223(95)00387-8
Soper HV Elliott RO Rejzer AA Marshall BD Effects of fenfluramine on neuropsychological and communicative functioning in treatment-refractory schizophrenic patients J Clin Psychopharmacol 1990 10 168 175 2198295
Malhotra AK Goldman D Mazzanti C Clifton A Breier A Pickar D A functional serotonin transporter (5-HTT) polymorphism is associated with psychosis in neuroleptic-free schizophrenics Mol Psychiat 1998 3 328 332 10.1038/sj.mp.4000412
Duval F Mokrani MC Monreal J Bailey P Valdebenito M Crocq MA Macher JP Dopamine and serotonin function in untreated schizophrenia: clinical correlates of the apomorphine and d-fenfluramine tests Psychoneuroendocrinol 2003 28 627 642 10.1016/S0306-4530(02)00047-1
Frith CD Blakemore S-J Wolpert DM Explaining the symptoms of schizophrenia: abnormalities in the awareness of action Brain Res Rev 2000 31 357 363 10719163 10.1016/S0165-0173(99)00052-1
Kaneko M Honda K Kanno T Horikoshi R Manome T Watanabe A Kumashiro H Plasma free 3-methoxy-4-hydroxyphenylglycol in acute schizophrenics before and after treatment Neuropsychobiol 1992 25 126 129 1357582
Yorkston NJ Zaki SA Weller MP Gruzelier JH Hirsch SR DL-propranolol and chlorpromazine following admission for schizophrenia: a controlled comparison Acta Psychiat Scand 1981 63 13 27 7015789
Freedman R Kirch DG Bell J Adler LE Pecevich M Pachtman E Denver P Clonidine treatment of schizophrenia: double blind comparison to placebo and neuroleptic drugs Acta Psychiat Scand 1982 65 35 45 6138923
Nieman DH Bour LJ Linszen DH Goede J Koelman JHTM Gersons BPR Ongeboer de Visser BW Neuropsychological and clinical correlates of antisaccade task performance in schizophrenia Neurology 2000 54 866 871 10690978
Luciana M Collins PF Depue RA Opposing roles for dopamine and serotonin in the modulation of human spatial working memory functions Cereb Cortex 1998 8 218 226 9617916 10.1093/cercor/8.3.218
Newcomer JW Farber NB Jevtovic-Todorovic V Selke G Melson AK Hershey T Craft S Olney JW Ketamine-induced NMDA receptor hypofunction as a model of a memory impairment and psychosis Neuropsychopharmacol 1998 20 106 118 10.1016/S0893-133X(98)00067-0
Stirling JD Hellewell JSE Qurashi N Self-monitoring dysfunction and the schizophrenic symptoms of alien control Psychol Med 1998 28 675 683 9626723 10.1017/S0033291798006679
Vollenweider FX Vontobel P Oye I Hell D Leenders KL Effects of (S)-ketamine on striatal dopamine: a [11C]raclopride PET study of a model psychosis in humans J Psychiatr Res 2000 34 35 43 10696831 10.1016/S0022-3956(99)00031-X
Nibuya M Kanba S Sekiya U Suzuki E Matsuo Y Kinishita N Shintani F Yagi G Asai M Schizophrenic patients with deficit syndrome have higher plasma homovanillic acid concentrations and ventricular enlargement Biol Psychiatry 1995 38 50 56 7548472 10.1016/0006-3223(94)00216-P
Suzuki E Kanba S Koshikawa H Nbuya M Yagi G Asai M Negative symptoms in nondeficit syndrome respond to neuroleptic treatment with changes in plasma homovanillic acid concentrations J Psychiat Neurosci 1996 21 167 171
Amin F Davidson M Davis KL Homovanillic acid measurement in clinical research: a review of methodology Schizophr Bull 1992 18 123 148 1553492
Nyman H Nybäck H Wiesel F-A Oxenstierna G Schalling D Neuropsychological test performance, brain morphological measures and CSF monoamine metabolites in schizophrenic patients Acta Psychiat Scand 1986 74 292 301 3788656
Jacobsen LK Rapaport JL Research update: Childhood onset schizophrenia: implications of clinical and neurobiological research J Child Psychol Psychiat 1998 39 101 113 9534088 10.1017/S002196309700173X
Cabeza R Nyberg L Imaging cognition II: an empirical review of 275 PET and fMRI studies J Cogn Neurosci 2000 12 1 47 10769304 10.1162/08989290051137585
Kemali D Maj M Galderisi S Ariano MG Starace F Factors associated with increased noradrenaline levels in schizophrenic patients Prog Neuropsychopharmacol Biol Psychiatry 1990 14 49 59 2300679 10.1016/0278-5846(90)90063-M
Van Kammen DP Kelley ME Yao JK Gilbertson MW Gurklis JA Inosaka T Saito H Peters JL Sato M Predicting haloperidol treatment response in chronic schizophrenia Psychiat Res 1996 64 47 58 10.1016/0165-1781(96)02906-X
Di Rocco A Bottiglieri T Dorfman D Werner P Morrison C Simpson D Decreased homovanillic acid in cerebrospinal fluid correlates with impaired neuropsychologic function in HIV-1-infected patients Clin Neuropharmacol 2000 23 190 194 11020122 10.1097/00002826-200007000-00004
Amin F Seeman TE Mohs RC Davidson M Knott P Berkman LF Albert M Blazer D Plasma homovanillic acid and performance on motor and cognitive tasks in community dwelling elderly Neuropsychopharmacol 1994 10 29 35 8179792
Gilbertson MW Yao JK van Kammen DP Memory and plasma HVA changes in schizophrenia: are they episode markers? Biol Psychiatry 1994 35 203 206 8173021 10.1016/0006-3223(94)91153-3
Medalia A Gold J Merriam A The effects of neuroleptics on neuropsychological test results of schizophrenics Arch Clin Neuropsychol 1988 3 249 271 14589696 10.1016/0887-6177(88)90018-2
Oades RD The role of noradrenaline in tuning and dopamine in switching between signals in the CNS Neurosci Biobehav Rev 1985 9 261 283 3925395 10.1016/0149-7634(85)90050-8
Brauns H Haun D Steinmann S The construction of an internationally comparable classification by class (Erwerbsstatistische Besonderheiten am Beispiel von Labour Force Surveys der Bundesrepublik Deutschland, Frankreichs, Großbritanniens und Ungarns) Arbeitspapiere Arbeitsbereich 1/22 1997 Mannheim Center for European Social Research (MZES), Mannheim
|
16091141
|
PMC1208853
|
CC BY
|
2021-01-04 16:39:21
|
no
|
Behav Brain Funct. 2005 Aug 9; 1:14
|
utf-8
|
Behav Brain Funct
| 2,005 |
10.1186/1744-9081-1-14
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2001608684110.1186/1471-2105-6-200SoftwareGenomic multiple sequence alignments: refinement using a genetic algorithm Wang Chunlin [email protected] Elliot J [email protected] Department of Microbiology, University of Alabama at Birmingham, Birmingham, Alabama 35294-2170, USA2005 8 8 2005 6 200 200 21 1 2005 8 8 2005 Copyright © 2005 Wang and Lefkowitz; licensee BioMed Central Ltd.2005Wang and Lefkowitz; 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
Genomic sequence data cannot be fully appreciated in isolation. Comparative genomics – the practice of comparing genomic sequences from different species – plays an increasingly important role in understanding the genotypic differences between species that result in phenotypic differences as well as in revealing patterns of evolutionary relationships. One of the major challenges in comparative genomics is producing a high-quality alignment between two or more related genomic sequences. In recent years, a number of tools have been developed for aligning large genomic sequences. Most utilize heuristic strategies to identify a series of strong sequence similarities, which are then used as anchors to align the regions between the anchor points. The resulting alignment is globally correct, but in many cases is suboptimal locally. We describe a new program, GenAlignRefine, which improves the overall quality of global multiple alignments by using a genetic algorithm to improve local regions of alignment. Regions of low quality are identified, realigned using the program T-Coffee, and then refined using a genetic algorithm. Because a better COFFEE (Consistency based Objective Function For alignmEnt Evaluation) score generally reflects greater alignment quality, the algorithm searches for an alignment that yields a better COFFEE score. To improve the intrinsic slowness of the genetic algorithm, GenAlignRefine was implemented as a parallel, cluster-based program.
Results
We tested the GenAlignRefine algorithm by running it on a Linux cluster to refine sequences from a simulation, as well as refine a multiple alignment of 15 Orthopoxvirus genomic sequences approximately 260,000 nucleotides in length that initially had been aligned by Multi-LAGAN. It took approximately 150 minutes for a 40-processor Linux cluster to optimize some 200 fuzzy (poorly aligned) regions of the orthopoxvirus alignment. Overall sequence identity increased only slightly; but significantly, this occurred at the same time that the overall alignment length decreased – through the removal of gaps – by approximately 200 gapped regions representing roughly 1,300 gaps.
Conclusion
We have implemented a genetic algorithm in parallel mode to optimize multiple genomic sequence alignments initially generated by various alignment tools. Benchmarking experiments showed that the refinement algorithm improved genomic sequence alignments within a reasonable period of time.
==== Body
Background
One of the primary goals in analyzing complete genomes is to identify all of the functional regions in the sequences, including genes and regulatory regions. However, this interpretive work is not keeping pace with the avalanche of raw sequence data. This disparity is due in part to the fact that algorithm development for genomic annotation has been relatively slow, and annotation of completely sequenced genomes inevitably depends on human expert knowledge. The most effective method to understand genomic content is to compare multiple genomes of various phylogenetic distances. The coding regions of a large set of common genes can be identified by comparing genomic sequences that are distantly related phylogenetically. In addition, comparing the genomic sequences of divergent non-coding regions that show some degree of conservation can yield important information related to regulation of gene expression, structural organization of the genome, and possibly other yet unknown functions [1]. Finally, functional and evolutionary inferences can be made from comparative genomic analysis. For example, orthologous relationships can suggest the function of a genetic sequence when the function of a similar sequence in another species is known.
One of the major challenges when comparing two or more genomic sequences is producing a high-quality multiple sequence alignment of all the genomes. In recent years, a number of computer programs have been developed for the alignment of large genomic sequences including CHAOS/DIALIGN [2], MUMmer [3], WABA [4], VISTA [5], BLASTZ [6], MAVID [7], and Multi-LAGAN [8]. In general, these tools utilize heuristic algorithms that provide an approximate solution to the problem of generating multiple sequence alignments. These heuristic tools are based on many different paradigms, but they all fall into the "anchor-extension" strategy, wherein a series of strong sequence similarities (anchors) are identified first and gaps are filled in (extension) by aligning the regions between the anchor points. The resulting alignment is globally correct, but in many cases is suboptimal locally.
In this paper, we describe the development of a program, GenAlignRefine, which improves the overall quality of global multiple sequence alignments by using a genetic algorithm to improve alignments in local regions.
Implementation
Multiple sequence alignment (MSA) is one of the most difficult problems in computational biology and there are only approximate solutions for all but the smallest alignments [9]. Therefore a number of novel heuristic algorithms have been proposed [10]. There are at least two distinct technical problems remaining: the choice of an objective function (OF) that assesses the quality of an alignment, and the design of an appropriate algorithm to optimize the score from that objective function. Ideally, a biologically meaningful alignment should produce a better OF score than a suboptimal alignment. One popular OF is the sum-of-pair (SP) function, which is a direct extension of the scoring method used in pair-wise alignments [11]. The SP score for an aligned column in the MSA is computed by scoring all of the pair-wise comparisons between each residue in each column of an alignment and adding the scores together. The major shortcoming of the SP function is that the alignment quality is extremely dependent on the choice of a score matrix and gap penalties. In general, this choice is problem-specific. For instance, for closely related input sequences, it might be better to use a score matrix reflecting a short phylogenetic distance. On the other hand, for divergent sequences, it might be more appropriate to choose a score matrix reflecting a longer phylogenetic distance.
One alternative to using the SP function is the COFFEE [12] function, which evaluates the consistency between a multiple alignment and libraries of optimal pair-wise alignments of the same sequences. Although COFFEE does not completely overcome the problems related to the choice of score matrix and gap penalties, it can reduce them. (There are some algorithms such as Dialign, in which the OF does not consider gapped regions. Instead, the OF of Dialign uses the sum of weights of gap free segment pairs [13]). The optimal pair-wise alignment is inevitably affected by the choice of score matrix and gap penalty, but a correct choice will be more likely since it is possible to choose different score matricies and gap penalties adaptively based on the distance detected during the pair-wise alignment process. The COFFEE OF has been shown to be more robust and to lead to better alignments [14]. In our application, we chose to utilize the COFFEE OF as a measure of the optimization of the multiple sequence alignment. Our COFFEE score is calculated based on the following formula:
where N is the number of sequences to be aligned; Wij is the percent identity between sequence i and j in the optimal pair-wise alignment library; Cij is the number of aligned character pairs that are shared between the multiple alignment and the optimal pair-wise alignment; and Lij is the length of the optimal pair-wise alignment of sequences i and j. The optimal pair-wise alignment library is constructed by aligning every pair of input sequences with an implementation of the Needleman-Wunsch algorithm [15].
To optimize an alignment by attempting to maximize its COFFEE score, we chose to implement a genetic algorithm. A genetic algorithm is a stochastic search method based on the concept of biological evolution; i.e., in simulating an evolutionary process in a population of potential solutions, a better solution will evolve [16]. Biological terms are used to describe the evolutionary process. Each potential solution is called a chromosome; a set of chromosomes refers to a population; and successive populations are called generations. To create new chromosomes (or offspring), two types of operators are generally used: mutation, which changes a single chromosome, and crossover, which exchanges information from two or more chromosomes. Based on Darwin's principle of survival of the fittest, chromosomes that perform well on certain fitness functions will have a greater likelihood of producing more offspring. Since the best performing individual in each generation is always selected for the next generation, the solutions in each generation are at least as good as those provided in previous runs. In this way, the genetic algorithm is able to optimize solutions from any source. Using genetic algorithms to solve MSA problems is not a new idea. SAGA [17] successfully applies a genetic algorithm to MSAs by attempting to optimize the weighted sum-of-pairs with natural or quasi-natural affinity gap penalties. Further attempts based on SAGA include SAGA-COFFEE, which tries to optimize the consistency-based objective function [14]. Our approach differs from SAGA in several ways. First, GenAlignRefine is designed to optimize multiple sequence alignments without regards to their length. SAGA is not optimized to align genome-length sequences. Second, in order to reduce the number of genetic operators that need to be utilized, GenAlignRefine pre-aligns each fuzzy region using T-Coffee. This allows us to optimize the application of the genetic operators by using a combination of only 3 operators rather than the full set of 22 operators used in SAGA (see below).
There are two considerations when designing a genetic algorithm. The first is how quickly a genetic algorithm can converge to an optimal solution. The second is the risk of misguiding an optimization process to a solution that appears to be optimal, but in fact resulted from convergence to a local optimum. Genetic algorithms are known to be extremely slow, with some MSA implementations being hundreds of times slower than ClustalW [18,19]. Genomic sequences may be megabases in length, and depending on the similarities between the sequences to be aligned, there may be thousands of poorly aligned regions. For that reason speed is a critical factor. To improve the overall performance of this application, we implemented the entire optimization process as a parallel, cluster-based program.
Based on manual inspection of the multiple genome alignments produced by various tools, we have found that regions encompassing and surrounding gaps are where most of the discrepancies between alignment methods occur. In this study, we concentrated our attention on these "fuzzy" regions. Fuzzy regions were defined as columns in an alignment that contain a gap adjacent to a gap-free region of at least 20 nucleotides. The gap-free regions of 20 nucleotides in length provided a constrained space in the multiple alignment that allowed the refining algorithm to place gaps between the constrained positions. Experimenting with different lengths of gap-free regions showed that a length of 20 was sufficient to allow for reasonable constraint. Longer gap-free regions did not increase the quality of final alignment, but did increase the length of time required for the optimization process.
GenAlignRefine was developed based on the assumption that the original starting alignment is globally correct. With this rationale, the overall genome alignment is shaped by those anchor regions that show strong similarity and are therefore obvious orthologs. These optimal anchor regions are kept intact. Optimization of the fuzzy regions between these anchor points will therefore not reshape the overall alignment, but will improve the overall quality of the alignment by improving each individual local region. One possible reason that the starting alignment might not be globally correct would be if one or more genomes contained large sequence rearrangements in comparison to the other genomes. Regions containing such rearrangements would need to be removed from the analysis since these regions will not be directly alignable. GenAlignRefine handles this problem by removing from consideration fuzzy regions longer than 1000 bases that may contain, in addition to rearrangements, large numbers of repeat sequences. Manual inspection of whole-genome pair-wise dotplots [3,20] were also used to identify these unalignable regions.
Using a Linux cluster in master-slave mode, fuzzy regions from an initial rough global MSA were identified. These fuzzy regions were then sent by the master to one of the slave nodes where a genetic algorithm (described below) was used to optimize the fuzzy region. The process continued until all fuzzy regions were optimized.
It is important for the efficiency of the search process to utilize a reasonable alignment for each fuzzy region as the starting point for optimization. This avoids local optima and yields better results in shorter periods of time. Therefore, in GenAlignRefine, we first used T-Coffee [14] on each of the fuzzy regions to produce an initial alignment, which is then used as the starting point for the genetic algorithm, thereby becoming the first chromosome from which successive generations originate. This strategy has been proven successful in other studies [21] and we have found it to be very efficient. Furthermore, since each potential solution in a population is derived from the T-Coffee alignment, the simplest way to optimize the COFFEE score is to re-arrange gaps in the alignment. For that reason, we implemented a subset of the mutation operators from SAGA [17] that perform gap re-arrangement: random_gap, local_gap_shuffle, and block_gap_shuffle (see Fig. 1). The random_gap operator randomly inserts a gap into every sequence of the alignment. The local_gap_shuffle operator shuffles one gap in every sequence by randomly moving it to a different position in that sequence. The block_gap_shuffle randomly moves gaps in selected columns of every sequence of the alignment to different positions in that alignment. In addition, any columns containing only gaps are deleted in each generation. A combination of random_gap, local_gap-shuffle, and block_gap_shuffle was able to effectively simulate any single mutation operator in SAGA. Previous studies have shown that operator schedules (setting a probability and frequency for use of each genetic operator) did not improve the performance of SAGA compared to the uniform selection of SAGA operators [21]. After experimenting with different operator scheduling strategies, we chose to use the three operators at equal frequency.
Figure 1 Mutation operators used in GenAlignRefine. a) Random_gap operator. Random gaps (shaded hypens) are inserted into the parent alignment to produce the offspring sequences. b) Local_gap_shuffle operator. Gaps in the parent are randomly moved to produce new offspring. c) Block_gap_shuffle operator. Contiguous blocks of gaps are randomly moved to new positions.
In summary, the overall process used to generate refined multiple sequence alignments starts with a set of sequences that are initially aligned using a genome sequence MSA tool such as Multi-LAGAN. Fuzzy regions within these alignments are identified and individually realigned using T-Coffee. These realigned regions are then refined using our implementation of the indicated set of genetic algorithm operators. All computation following generation of the initial multiple sequence alignment takes place on a Linux cluster.
GenAlignRefine was implemented in Perl to take advantage of the convenience in manipulation of biological sequences provided by the Bioperl [22] application programming interface (API). To bridge the gap between Perl and the message passing interface (MPI) [23] API which is implemented in C/C++ and is required for our parallel implementation, we also provide a wrapper module that ports to Perl only those MPI C/C++ procedures necessary for this application. The implementation of an efficient Needleman-Wunsch algorithm [15] coded in C/C++ and ported to PERL was used to construct pair-wise alignment libraries.
Results and discussion
We performed our benchmarking experiment on a 32-node Linux cluster in the Department of Computer and Information Science at the University of Alabama at Birmingham. All machines have 1.6 GHz Dual AMD Opteron™ Processors, 2 GB of RAM, and are connected via Gigabit Ethernet.
The lack of a "gold standard" for assessment of multiple sequence genome alignments makes it difficult to assess the performance of multiple genome alignment tools. In this study, we chose a simulation-based approach to benchmark the results produced by GenAlignRefine. Multiple sequences along with the correct, "optimal" multiple alignment of these sequences as generated by the software tool Rose (random model of sequence evolution) [24] have been widely used to benchmark the performance of multiple alignment tools [25] and phylogenetic analyses [26]. We started with a sequence alignment created by Rose which contains 9 sequences comprising about 100,000 nucleotides each. Sequence generation began with 1 randomly generated ancestral sequence composed of equal nucleotide base frequencies. From that sequence, the 9 test sequences were generated based on the HKY evolutionary model of point substitution [27] using a transition/transversion bias of 2.5. The insertion and deletion threshold was set to allow insertions and/or deletions 5% of the time. The mean base substitution rate was set to 0.05 substitutions per site. The tree was set to ((a:.2,b:.5):.1,(c:.4,d:.5,e:.4):.2,(f:.3,g:.4):.3,(h:.4,i:.5):.1). For each sequence, the mutation probability of each nucleotide position was set to either 0.0, 0.3, 0.6, 0.9, or 1. Using this set of 9 sequences, we generated two new alignments using the programs Multi-LAGAN [8] which was run locally using the default options, and CHAOS/DIALIGN [2] which was run using the available web application [28]. Each of the two new alignments was then subjected to refinement by GenAlignRefine. We then measured the consistencies between the alignments by comparing each of the four new alignments to the original simulated alignment produced by Rose. The consistency between any two alignments, A and B is defined as the ratio between the number of identical character pairs between the two alignments, and the total number of character pairs in alignment B. The results from this comparison are provided in table 1. In each case, GenAlignRefine was able to improve the quality of each multiple genome sequence alignment by at least 7% as measured by an increase in the number of pair-wise matches to the "optimal" alignment as constructed by Rose. It was also apparent that the quality of the final alignment was dependent on the quality of the original alignment prior to refinement.
Table 1 Performance of GenAlignRefine on simulated data.
Program Before Refinement After Refinement
CHAOS/DIALIGN 78.0%* 85.1%
Multi-LAGAN 86.3% 93.0%
* The numbers indicate the consistency between the alignment generated with the genome alignment tool and the "correct" alignment generated by Rose (see text).
In addition to the above benchmarking experiments, we also conducted a study to demonstrate the usefulness of the program GenAlignRefine by refining the genome alignment of complete genome sequences available for the virus genus, Orthopoxvirus that includes variola virus, the agent responsible for causing smallpox. Genome sequences used in this analysis were obtained mostly from GenBank and included vaccinia virus strains Copenhagen [GenBank: M35027], Western Reserve (WR) [GenBank: AY243312], and Tian Tan [GenBank: AF095689] (with updates from Dr. Chris Upton of the University of Victoria); variola major virus strains Bangladesh [GenBank: L2579] and India [GenBank: X69198]; variola minor virus strain Garcia [GenBank: Y16780]; camelpox virus strains CMS [GenBank: AY009089] and M-96 [GenBank: AF438165]; cowpox virus strains Brighton Red [GenBank: AF482758] and GRI-90 [GenBank: X94355]; ectromelia virus strain Moscow [GenBank: AF012825]; and monkeypox virus strain Zaire [GenBank: AF380138]. Genome sequences of monkeypox virus strain WRAIR 7–61 [GenBank: AY603973] and rabbitpox virus strain Utrecht [GenBank: NC_005858] were kindly provided by Dr. Mark Buller of Saint Louis University. The genome sequence of ectromelia virus strain Naval was obtained from [29]. An initial alignment of these 15 Orthopoxvirus genomic sequences was created using Multi-LAGAN [8]. The overall length of the alignment was approximately 260,000 nucleotides.
Two considerations in implementing our genetic algorithm were when to stop the optimization process and how to breed the next generation. In our benchmarking experiment, if there was no improvement in an alignment's COFFEE score for 500 generations, the optimization process was stopped and the alignment with the best COFFEE score was returned to the master node. As the population size increases, the risk of falling into a local, suboptimal alignment decreases, but so does the speed of the optimization process. Since our starting alignments are derived from T-Coffee, it is assumed that they are close to the global optimum. For that reason, and for the sake of efficiency, we try to keep the population size relatively small. From each pool of 1000 individuals (alignments generated by application of the genetic algorithm), only the top 100 individuals with the best COFFEE scores are allowed to breed. In addition, we use an elite selection strategy in which some of the fittest individuals from the first generation are allowed to carry over unaltered into the second generation. We also permit individuals having the best COFFEE scores to have more offspring. All of these parameters are adjustable.
In general, regions with gaps are most likely to be discordant and therefore in need of improvement. However, even for closely related species, not all regions in their genomic sequences can be aligned. For instance, the regions at the ends of poxvirus genome sequences contain variable numbers of repeat sequences and some of these repeating units are species-specific [30] and thus cannot be aligned. In our study, fuzzy regions longer than 1000 bases were considered unalignable, so these regions were not subject to optimization. In our orthopoxvirus alignment, there were 18 fuzzy regions longer than 1000 bases, all of which occurred at the ends of the original alignment. Between these unalignable genomic regions, there were about 400 gapped (fuzzy) regions that were then subjected to the genetic algorithm. Using 40 processors, it took 150 minutes to optimize all of these regions. Only some 200 regions were actually improved by the genetic algorithm with the remaining 200 regions already showing optimal COFFEE scores. Figure 2 displays the improvement in the 200 fuzzy regions based on the COFFEE scores. It is apparent that, in general, regions with lower initial COFFEE scores showed more improvement while regions with higher initial COFFEE scores showed less improvement.
Figure 2 Improvement of COFFEE score for fuzzy regions. The 200 fuzzy regions derived from the starting Orthopoxvirus alignment that showed improvement following application of GenAlignRefine are displayed. For clarity, regions are sorted according to the overall improvement in COFFEE score. Vertical bars connect dots that show the improvement in COFFEE score for each region at each step in the refinement process. Red dots plot the original COFFEE score of the Multi-LAGAN-generated alignment for each region; green dots plot the COFFEE score of the same region after realignment by T-Coffee; blue dots indicate the COFFEE score of the same region after optimization by the genetic algorithm. The small magenta squares plot the overall improvement in COFFEE score for each region.
An improvement in COFFEE score is only one possible measure that might reflect an actual improvement in alignment quality. And since for these poxvirus sequences, there is no "correct" alignment for comparison as there was for the simulation, we chose to measure improvement by simply demonstrating an increase in the overall percentage identity calculated between all pair-wise sequence comparisons, along with a decrease in the length of the overall alignment due to the introduction of fewer gaps. An increase in the percentage identity can be achieved by simply inserting greater numbers of gaps into the alignment without necessarily improving the overall quality of the alignment. However, when the overall percentage identity is increased concomitant with a decrease in the length of the alignment, then in general, we would argue that the overall quality of the alignment has been improved. The alignment of the 15 Orthopoxvirus genomic sequences produced by Multi-LAGAN was 259817 nucleotides in length and showed 96.0% identity [see Additional file 1]. After refinement by GenAlignRefine, the length of the alignment was 258593 with 96.2% identity [see Additional file 2]. The improvement in the percentage identity was marginal for the overall alignment, but for each fuzzy region, the improvement was much higher. Significantly, this occurred at the same time that the overall alignment length decreased – through the removal of gaps – by approximately 200 gapped regions representing roughly 1,300 gaps. Therefore, the increase in the overall alignment quality was substantial.
The three mutation operators used in the genetic algorithm were effective after using T-Coffee to initially realign each fuzzy region of the starting alignment thus producing a seed for subsequent improvement. This is similar to what has been seen in previous studies [21]. As the resulting alignment from T-Coffee can be viewed as an approximation to the optimal result, starting with this alignment, which is presumably close to the optimal result in the multiple alignment search space, should decrease the chance of becoming trapped into a local optimum. And although there is still a risk that the optimization process will be misguided to a local optimum, the chance of this occurring should be small.
Genetic algorithms are known to be slow and computationally intensive compared to other methods [19]. However by using appropriate design parameters along with a large computing cluster, we have shown that GenAlignRefine can be used to efficiently and effectively improve multiple sequence alignments of whole genome sequences.
Availability and requirements
GenAlignRefine is freely available under the Artistic License described by the Open Source Initiative [31]. The source code can be downloaded via ftp [32]. Contact [email protected] for information on obtaining the software. It has been tested on an AMD Opteron™ Processor-based Linux cluster with LAM/MPI and should be compatible with other implementations of MPI.
Authors' contributions
CW was responsible for the conception, design, implementation, and testing of the GenAlignRefine. EJL contributed to its conception and testing and provided overall project coordination. Both authors have read and approved the final manuscript.
Supplementary Material
Additional File 1
Orthopoxvirus multiple genome alignment prior to optimization. Original Orthopoxvirus alignment produced by Multi-LAGAN in Clustal .aln format.
Click here for file
Additional File 2
Orthopoxvirus multiple genome alignment after optimization. Final Orthopoxvirus alignment following improvement by GenAlignRefine in Clustal .aln format.
Click here for file
Acknowledgements
We would like to acknowledge the High Performance Computing Laboratory [33] in the Department of Computer and Information Sciences at the University of Alabama at Birmingham for giving us the opportunity to use their Linux cluster. We thank Vijay Velusamy and Zhijie Guan for assistance with the benchmark experiments. We would also like to thank Dr. Mark Buller of St. Louis University for providing us with the genomic sequences of monkeypox virus strain WRAIR 7–61 and rabbitpox virus strain Utrecht that were used in the benchmarking experiments prior to their GenBank release. We gratefully acknowledge Dr. Purushotham V. Bangalore and Ms. Catherine B. Galloway for helpful comments on the manuscript. This work was supported by NIH/NIAID/DARPA Grant No. U01 AI48706 and NIH/NIAID Contract No. HHSN266200400036C to EJL.
==== Refs
Ovcharenko I Nobrega MA Loots GG Stubbs L ECR Browser: a tool for visualizing and accessing data from comparisons of multiple vertebrate genomes Nucleic Acids Res 2004 32 W280 6 15215395
Brudno M Steinkamp R Morgenstern B The CHAOS/DIALIGN WWW server for multiple alignment of genomic sequences Nucleic Acids Res 2004 32 W41 4 15215346
Delcher AL Phillippy A Carlton J Salzberg SL Fast algorithms for large-scale genome alignment and comparison Nucleic Acids Res 2002 30 2478 2483 12034836 10.1093/nar/30.11.2478
Baillie DL Rose AM WABA success: a tool for sequence comparison between large genomes Genome Res 2000 10 1071 1073 10958624 10.1101/gr.10.8.1071
Frazer KA Pachter L Poliakov A Rubin EM Dubchak I VISTA: computational tools for comparative genomics Nucleic Acids Res 2004 32 W273 9 15215394
Schwartz S Kent WJ Smit A Zhang Z Baertsch R Hardison RC Haussler D Miller W Human-mouse alignments with BLASTZ Genome Res 2003 13 103 107 12529312 10.1101/gr.809403
Bray N Dubchak I Pachter L AVID: A global alignment program Genome Res 2003 13 97 102 12529311 10.1101/gr.789803
Brudno M Do CB Cooper GM Kim MF Davydov E Green ED Sidow A Batzoglou S LAGAN and Multi-LAGAN: efficient tools for large-scale multiple alignment of genomic DNA Genome Res 2003 13 721 731 12654723 10.1101/gr.926603
Gusfield D Algorithm on Strings, Trees, and Sequences 1997 Cambridge University Press 212
Hirosawa M Totoki Y Hoshida M Ishikawa M Comprehensive study on iterative algorithms of multiple sequence alignment Comput Appl Biosci 1995 11 13 18 7796270
Nicholas HBJ Ropelewski AJ Deerfield DW Strategies for multiple sequence alignment Biotechniques 2002 32 572 4, 576, 578 passim 11911661
Notredame C Holm L Higgins DG COFFEE: an objective function for multiple sequence alignments Bioinformatics 1998 14 407 422 9682054 10.1093/bioinformatics/14.5.407
Morgenstern B DIALIGN 2: improvement of the segment-to-segment approach to multiple sequence alignment Bioinformatics 1999 15 211 218 10222408 10.1093/bioinformatics/15.3.211
Notredame C Higgins DG Heringa J T-Coffee: A novel method for fast and accurate multiple sequence alignment J Mol Biol 2000 302 205 217 10964570 10.1006/jmbi.2000.4042
Needleman SB Wunsch CD A general method applicable to the search for similarities in the amino acid sequence of two proteins J Mol Biol 1970 48 443 453 5420325 10.1016/0022-2836(70)90057-4
Holland JH Adaptation in natural and artificial systems : an introductory analysis with applications to biology, control, and artificial intelligence 1975 Ann Arbor , University of Michigan Press viii, 183 p.
Notredame C Higgins DG SAGA: sequence alignment by genetic algorithm Nucleic Acids Res 1996 24 1515 1524 8628686 10.1093/nar/24.8.1515
Thompson JD Higgins DG Gibson TJ CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice Nucleic Acids Res 1994 22 4673 4680 7984417
Wang Y Li KB An adaptive and iterative algorithm for refining multiple sequence alignment Comput Biol Chem 2004 28 141 148 15130542 10.1016/j.compbiolchem.2004.02.001
Upton C Hogg D Perrin D Boone M Harris NL Viral genome organizer: a system for analyzing complete viral genomes Virus Res 2000 70 55 64 11074125 10.1016/S0168-1702(00)00210-0
Thomsen R Boomsma W Multiple Sequence Alignment Using SAGA: Investigating the Effects of Operator Scheduling, Population Seeding, and Crossover Operators. 2004 113 122
Stajich JE Block D Boulez K Brenner SE Chervitz SA Dagdigian C Fuellen G Gilbert JG Korf I Lapp H Lehvaslaiho H Matsalla C Mungall CJ Osborne BI Pocock MR Schattner P Senger M Stein LD Stupka E Wilkinson MD Birney E The Bioperl toolkit: Perl modules for the life sciences Genome Res 2002 12 1611 1618 12368254 10.1101/gr.361602
LAM/MPI
Stoye J Evers D Meyer F Rose: generating sequence families Bioinformatics 1998 14 157 163 9545448 10.1093/bioinformatics/14.2.157
Pollard DA Bergman CM Stoye J Celniker SE Eisen MB Benchmarking tools for the alignment of functional noncoding DNA BMC Bioinformatics 2004 5 6 14736341 10.1186/1471-2105-5-6
Muller T Rahmann S Dandekar T Wolf M Accurate and robust phylogeny estimation based on profile distances: a study of the Chlorophyceae (Chlorophyta) BMC Evol Biol 2004 4 20 15222898 10.1186/1471-2148-4-20
Hasegawa M Kishino H Yano T Dating of the human-ape splitting by a molecular clock of mitochondrial DNA J Mol Evol 1985 22 160 174 3934395
CHAOS/DIALIGN [http://dialign.gobics.de/chaos dialign-submission]
Ectromelia virus strain Naval - Sanger Institute
Gubser C Smith GL The sequence of camelpox virus shows it is most closely related to variola virus, the cause of smallpox J Gen Virol 2002 83 855 872 11907336
Open Source Initiative
FTP Site
HPCL
|
16086841
|
PMC1208854
|
CC BY
|
2021-01-04 16:27:23
|
no
|
BMC Bioinformatics. 2005 Aug 8; 6:200
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-200
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2041611531710.1186/1471-2105-6-204Research ArticleTandem machine learning for the identification of genes regulated by transcription factors Dinakarpandian Deendayal [email protected] Venetia [email protected] Saumil [email protected] Erin G [email protected] Peter K [email protected] School of Computing and Engineering, University Of Missouri-Kansas City, Kansas City, Missouri, USA2 Pharmaceutical Sciences, St. Jude's Children's Research Hospital, Memphis, Tennessee, USA3 Laboratory of Human Molecular Genetics, Children's Mercy Hospital, Kansas City, Missouri, USA2005 22 8 2005 6 204 204 9 3 2005 22 8 2005 Copyright © 2005 Dinakarpandian et al; licensee BioMed Central Ltd.2005Dinakarpandian 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 identification of promoter regions that are regulated by a given transcription factor has traditionally relied upon the identification and distributions of binding sites recognized by the factor. In this study, we have developed a tandem machine learning approach for the identification of regulatory target genes based on these parameters and on the corresponding binding site information contents that measure the affinities of the factor for these cognate elements.
Results
This method has been validated using models of DNA binding sites recognized by the xenobiotic-sensitive nuclear receptor, PXR/RXRα, for target genes within the human genome. An information theory-based weight matrix was first derived and refined from known PXR/RXRα binding sites. The promoter region of candidate genes was scanned with the weight matrix. A novel information density-based clustering algorithm was then used to identify clusters of information rich sites. Finally, transformed data representing metrics of location, strength and clustering of binding sites were used for classification of promoter regions using an ensemble approach involving neural networks, decision trees and Naïve Bayesian classification. The method was evaluated on a set of 24 known target genes and 288 genes known not to be regulated by PXR/RXRα. We report an average accuracy (proportion of correctly classified promoter regions) of 71%, sensitivity of 73%, and specificity of 70%, based on multiple cross-validation and the leave-one-out strategy. The performance on a test set of 13 genes showed that 10 were correctly classified.
Conclusion
We have developed a machine learning approach for the successful detection of gene targets for transcription factors with high accuracy. The method has been validated for the transcription factor PXR/RXRα and has the potential to be extended to other transcription factors.
==== Body
Background
Nucleic acid binding sites recognized by transcription factors are comprised of families of short, related, often degenerate sequences that share a common function. This degeneracy may be represented in the form of a position-specific weight matrix (PWM) [1,2]. In fact, PWMs have been widely applied [2,3] and several databases host them [4,5].
Using information theory, the degree of conservation of an individual member (both known and predicted) of that family and its corresponding weight matrix may be quantified in terms of bits of information [1]. The strength of experimental binding has been shown to correlate with the predicted value of binding strength in bits for individual transcription factor binding sites [6]. We have used this approach, for example, to successfully find potential splice sites and to predict the probability that a given splice site will be used [7,8].
Regions upstream of transcription initiation sites typically contain multiple potential heterogeneous binding sites recognized by the same transcription factor. However, such sites can be found in the promoter regions of both genes that are, and those that aren't, regulated by the same factor [9], suggesting that additional sequence or structural properties are needed to discriminate true target genes from those which are not regulated by a particular factor. This paper addresses the problem of the identification of target genes for xenobiotic-sensitive transcription factors in the human genome by using a combination in tandem of information theory, a novel information density-based clustering (IDBC) algorithm and machine learning approaches for classification.
As proof of concept, binding sites recognized by the nuclear receptor transcription factor, the pregnane X receptor (PXR), are used to develop an algorithm that identify genomic target genes regulated by this factor. The algorithm exploits a PWM that accurately models the affinity of the protein for these sequences [6]. PXR is a ligand-activated transcription factor that heterodimerizes with the 9-cis retinoic acid receptor X (RXR) to form PXR/RXRα. Following exposure to xenobiotics like rifampin, clotrimazole, ritonavir, phenobarbital and hyperforin, and endogenous compounds like lithocholic acid steroids [10], PXR/RXRα binds response elements in the promoters of regulated genes to induce gene expression.
Results
An information theoretic approach has high recall in identifying promoter elements bound by PXR/RXRα. Given a representative weight matrix, sensitivity is essentially 100%. However, information theory weight matrices are not sufficient to discriminate between true and false positives, since binding sites were also found within the promoter regions of genes known not to be regulated by PXR/RXRα. The analysis of total information content in positive versus negative sites shows that, on average, the information is concentrated more in stronger sites in the case of regulated genes (Fig. 1). Figure 1 shows the ratios between the fractions of the sum of the information content found in regulated versus unregulated genes at each binding strength. For example, if we consider sites that have a binding strength of 17 bits, we first count the number of such sites found in regulated genes and multiply this by 17. The resulting number is then divided by the total number of bits for all sites, irrespective of strength, and this yields the average fraction of information content in a promoter region that is found in sites with a bit strength of 17. If the same calculation is carried out for genes known to be unrelated to PXR/RXRα, the ratio of the fraction computed for regulated genes to that for unrelated/unregulated genes gives the corresponding y-axis value in Fig. 1. Thus, the graph shows that a 10-fold higher proportion of total information content is found within sites having a strength of 17 bits in regulated versus unregulated promoters. However, note that this is an observation based on an average quantity, and no significant differences are evident at sites that are 18, 20 or 21 bits in strength. It was apparent from our preliminary studies (of CYP3A4, CYP3A7 and other genes) that multiple binding sites were necessary for transcriptional activation. None of the promoters examined to date contain a single strong binding site, rather they contain multiple sites. Putatively, a few moderately strong sites might increase the odds of a transcription factor recognizing a promoter that is then bound to the strongest site(s) in the region. Therefore, in this paper, we have outlined a tandem machine-learning approach for the characterization of PXR/RXRα genomic signatures that takes into account the density, strength, and location of the sites predicted by information theory. In effect, the unknown function for successful regulation of a gene that we are trying to approximate may be expressed as F(B, D, S) where B = the frequency distribution of binding site strengths, D = the set of distances of binding sites from the transcription start site and S = the spacing (or density-based clusters) between the sites for a given promoter region.
Figure 1 Higher proportion of information content lies within stronger sites in promoter regions of regulated genes. A histogram representing strengths of putative binding sites for the transcription factor RXR/RXRα is shown. The x-axis represents the binding strength of a site in bits. The y-axis represents the relative ratio between the proportions of total information found at the corresponding strength in regulated and unregulated promoter regions (10 kb upstream of respective genes).
The information gain ratio with respect to classification of a region as positive or negative was computed for each attribute (see Feature Selection under Methods). The gain ratio rather than information gain was used to accommodate the fact that the number of possible states for the distribution differs from attribute to attribute. The attributes with the highest information gain ratios were considered for further analysis. Based on this analysis, the attributes with the highest gain ratios were found to be the number of clusters (0.79), total information (0.75), number of sites (0.7), information within clusters (0.67), information in top three clusters (0.62), and information over (x = 6) bits (0.5). Hence, these attributes were used for analysis with classification algorithms.
The results of the learning from the classification algorithms are summarized in Tables 1, 2. We expected to get progressively better performance in the sequence Naïve Bayesian, Decision trees, Neural Networks (Log-Sigmoid) and Neural Networks (Radial Basis Function). This is in keeping with the limiting assumption of Naïve Bayes learning that all attributes have independent probability distributions and the progressively flexible allowance for boundaries in dimensional hyperspace. For example, decision trees are limited to the use of hyperplanes that are perpendicular to some attribute or variable. Neural networks based on backpropagation relax this requirement to allow hyperplanes to adopt any orientation. The use of radial basis functions allows further flexibility in the partitioning of hyperspace. Contrary to expectation, all these methods exhibited comparable average performance. However, the NNs had the most consistent performance, i.e., they showed the least variance in performance with respect to choice of training set.
Table 1 Results of cross-validation by Decision Trees and the Naïve Bayes Classifier
Method/Performance DECISION TREES NAïVE BAYES CLASSIFIER
3-way CROSS VALIDATION
ACCURACY 71 72
SENSITIVITY 63 83
SPECIFICITY 71 70
LEAVE-ONE-OUT STRATEGY
ACCURACY 63 68
SENSITIVITY 70 79
SPECIFICITY 63 68
Table 2 Results of cross-validation by Neural Networks
Method/Performance NEURAL NETWORKS NEURAL NETWORKS
Log Sigmoid Radial Basis Function
3-way CROSS VALIDATION
ACCURACY 62 73
SENSITIVITY 71 77
SPECIFICITY 62 73
LEAVE-ONE-OUT STRATEGY
ACCURACY 77 78
SENSITIVITY 75 67
SPECIFICITY 77 78
We also tested the performance of the prediction methods on a test set of 13 genes (Table 3) that had not been included in either training or validation sets. The methods individually classified between 8 and 11 of the 13 genes correctly and a jury prediction correctly classified 10 of the 13 genes. The proportion of correct classification (77%) on this test set is in line with the performance noted in cross-validation of the training data (Tables 1, 2).
Table 3 Predictive performance on a test set of genes
Test Gene Neural Network Naïve Bayes Decision Tree True Class
CYP3A4 P P P P
CYP3A7 N P P P
CYP3A5 N N N P
SRP P P P P
CYP51A1 N N N P
CRYZ P P P P
SMN2 P P P P
HOXA9 N N N N
CDC2L5 N N N N
AKAP9 P P N N
VIK N N N N
ATP5J2 P N N N
PFKB4 N N N N
One of the questions raised by the IDBC analysis was to what extent did the genomic organization of predicted binding sites determine whether a particular gene was classified as a target for PXR/RXRα. We first considered the possibility that treating DNA strands separately might for promoter regulation, however experimental studies have shown that functional PXR/RXRα sites can occur on either strand and enhance transcription of genes regulated by this factor [6]. We then looked for trends in inter-site spacing. There is a bias (Fig. 2) in the periodicity of the separations of sites within clusters, indicating a preference for separation of sites by helical turn length, ie. 10 bp (and multiples thereof), which is consistent with multimeric protein recognition across the same face of the helix. The information maxima within PXR/RXRα binding sites are also separated by 10 bp, also consistent with major groove recognition [11]. However, the length of the binding site is considerably longer than this spacing, suggesting that binding site cluster recognition may in some way be mediated by interactions involving overlapping or alternating half sites. Surprisingly, this distribution is evident in promoter regions of both genes that are regulated PXR/RXRα (solid lines, Fig 2) and those whose expression is unchanged in response to rifampin (dotted lines, Fig. 2). There was no evidence of higher order chromatin accessibility to binding sites, since there was no preference for nucleosome phasing (160–200 bp) of binding sites (not shown).
Figure 2 Plot of inter-site distance and information content. The x-axis represents the spacing between a pair of sites expressed in number of bases, whereas the y-axis represents the corresponding pair-wise sum of information for all occurrences at a given spacing. The y-axis value is expressed in terms of a Z-score – units of standard deviation from the mean. The solid line represents the curve for a set of genes known to be regulated by PXR/RXRα, while the dotted line represents genes known to be unaffected by PXR/RXRα.
Discussion
The importance of clustered binding sites as an indicator of a regulatory region has been noted in several studies. One of the first [12] modeled the occurrence of clusters as a Poisson process in order estimate a p-value. However, this study required exact matches to consensus sequences, rather than PWMs and did not identify putative binding sites. The objective was to maximize specificity at the cost of recall. Berman et al. [13] used a program called CIS-ANALYST for the recognition of clusters of binding sites in the Drosophila genome. CIS-ANALYST uses a window-based approach to cluster sites solely based on physical location, without regard to strength of the sites. Further, cluster boundaries are coarse because they depend on the simple rule of collapsing overlapping windows. Therefore, the size of the cluster is on the order of a multiple of the parametric window size. MSCAN [14] aims to detect regulatory regions in genomes by clustering binding sites for all transcription factors. Both CIS-ANALYST and MSCAN use PWMs [1,2], clearly a superior alternative to consensus sequence detection of binding sites. However, a fixed window size is used for the detection of clusters by MSCAN, and the computed p-values represent an upper bound. Several other studies also use a fixed-size window [15,16]. Cluster-Buster [17] scans whole genomes with PWMs using dynamic programming to efficiently compute the log likelihood ratio of a clustered model to that of a random background model. It is not clear what threshold should be used to determine if the results obtained are significant.
The algorithm, the underlying information PWMs, and the nature of the present study distinguish the current approach from previous work. IDBC uses the metric of information density for delineating clusters of binding sites. Thus, the criterion for clustering is not the distances separating the binding sites per se, but is proportional to the number of bits of information. This implies that the size of a cluster may be highly variable, being influenced by both binding strength and the number of constituent sites. Neither is a rigid requirement for the number of sites imposed (a single site could also potentially constitute a cluster), nor is the boundary of a cluster constrained. Some studies [14,17] claim to obviate the need for training data. This is strictly not true as a PWM implicitly represents a trained model, but such studies do offer the advantage of not needing a known list of regulated genes. Such methods may be valuable as preliminary screening tools, especially when there is paucity of training data. But approaches that use training data have the potential to yield higher specificity and sensitivity such as that reported in the present study. Better discrimination between regulated and unregulated genes can be achieved by having suitable positive and negative examples from which to automate learning. This will help to map a transcription factor to a comprehensive set of cognate gene targets.
Despite the high level of performance we report, it is necessary to consider why our accuracy is limited. Possible explanations are:
i) The effect of 3D higher order structure of DNA has not been taken into account in the study, other than looking for periodicity in location of the sites. This might result in a difference to accessibility of different promoter regions by transcription factors.
ii) For some of the genes, the concomitant presence of binding sites for other transcription factors might be important. Or, in general, other additional factors may be important. In other words, PXR/RXRα might be necessary, but not sufficient, for activation/repression of some of the target genes.
iii) The negative training set may be confounded by cryptic PXR/RXRα target genes whose expression did not change in response to rifampin treatment in the HepG2 hepatic cell line. Since PXR/RXRα appears to be more broadly expressed [18], it is quite plausible that some target genes containing binding site signatures may not have been activated or repressed in the HepG2 background. These false negative assignments may be rectified by analysis of expression in appropriate tissues treated with PXR ligands.
iv) The analysis has been limited to the 10 kb region upstream of each gene. Though this is likely to be the most representative region for the vast majority of genes, this may ignore the presence of control elements in other locations in a few cases.
v) The classification boundaries might be highly complex. This is supported by the fact that we noticed a slight improvement upon changing the neural network architecture from standard backpropagation with log-sigmoid functions to RBF learning. The former is theoretically limited by its use of hyperplanes while the latter uses Gaussian distributions to divide multidimensional space.
Conclusion
We have presented a tandem machine learning approach for the computational identification of target genes for a given transcription factor. The locations and organization of binding sites alone are insufficient to discriminate genes regulated by a transcription factor from other gene targets. The strength of the approach is based on the tandem use of information-theoretic weight matrices, a novel density-based clustering approach and machine learning methods for classification. The method has been validated for the transcription factor PXR/RXRα, and has the potential for the improved identification of transcriptional regulatory targets across the entire genome [19].
Methods
An overview of the general approach to the problem is given in Fig. 3. Each stage in the process was performed as follows.
Figure 3 Overview of tandem machine learning. For each gene, the PWM representing binding sites for PXR/RXRα was used to scan the 10 kb region upstream of the transcription start site to generate a list of the location and strength of individual binding sites. This list was used to generate summary features, e.g., the total number of sites, total information content. It was also used as input for IDBC to generate clusters. A second set of summary features was extracted from the clustering obtained, e.g., total number of clusters, total information content within clusters. The combined list of features for each promoter region constituted a single data item for input to one of several machine-learning algorithm.
Search for potential binding sites
The information-theoretic approach for refinement of binding site models of known and predicted binding sites has been described previously [1,7,8]. The initial binding site model, derived from 15 previously reported PXR binding sites, was found to be significantly biased towards consensus sequence-like recognition sites based on their high individual information contents (Ri) [6]. The corresponding PWM derived from this data failed to detect weak and intermediate strength binding sites and inaccurately predicted their binding affinities. With progressive model refinement incorporating newly identified, experimentally-validated sites, the Ri values tended towards a Gaussian distribution. Using the refined PWM, the Ri values more accurately measured the affinity of known regulatory sequences and more comprehensively identified predicted sites, consistent with the expectations from information theory [1] and the findings seen in refined models of other genome-wide protein-nucleic acid interactions [6]. This motivated our selection of the PXR/RXRα information PWM [6] for the present study.
We scanned the promoter regions (10 kb upstream of transcription start sites (TSS)) on both strands of 24 genes known to be regulated by PXR/RXRα (Table 4) for binding sites and 288 genes known to be negative based on literature and microarray data. The location and strength of binding sites for each promoter region were recorded for use in subsequent steps.
Table 4 Genes known to be regulated by PXR/RXRα based on [25] and/or microarray analysis
ABCB11 LTB4R
ABCB4 SLC17A4
ABCC2 SLC21A14
AV6993471 SLC21A8
CHST7 SLC21A9
CYP2A3 SLC2A10
CYP2B6 UGT1A1
CYP2C8 UGT1A3
CYP2C9 UGT1A4
CYP3A43 UGT1A6
CYP4F3 UGT1A9
GSTA2 UGT2B15
1This is a spliced EST, possibly an incomplete gene.
Information density-based clustering (IDBC)
IDBC is a modified version of the DBSCAN density-based clustering algorithm [20]. Implicitly, all clustering algorithms are based on the notion of density of data points in multi-dimensional space. Approaches such as DBSCAN explicitly use density of the distribution of data as a metric for clustering. IDBC supplements the consideration of spatial density with the additional contribution of information density. We define information density as being proportional to the magnitude of information content that is packed into small areas of the promoter region (see 'neighborhood information content' below). This is used as the basis for finding clusters of information rich sites. The steps of the algorithm are:
1. For each site s, calculate the neighborhood information content (nic) as being the total of pairwise sums of the information content for the site s and each site lying within distance d (number of bases) of s.
2. For every site s that has an nic exceeding a threshold parameter I, create a cluster by promoting s to the role of a cluster center c and including all sites within its neighborhood as members.
3. In the first phase of merging clusters (Fig. 4), consider all pairs of clusters with centers ci and cj. If ci is a member of the cluster with cj as its center and vice versa, then merge the two clusters and replace the center with the stronger of the sites ci and cj. If they are equal in strength, the center containing more sites is made the center of the new cluster, while the other center is relegated to being just a site. The process is iterated until no ci occurs in more than a single cluster.
Figure 4 Information Density Based Clustering (IDBC) Algorithm. The steps of IDBC are described in the Methods section. Panel A shows the location of putative binding sites upstream of the transcription start site. The vertical height of each bar indicates the strength of the respective binding site. Panel B shows the initial list of 4 clusters derived from the first iteration of the algorithm. This includes an example of an overlap where one of the sites is shared between clusters 3 and 4. Panel C shows the result of a refining step where the overlapping point is resolved, exclusively, to cluster 3. Since the single site in cluster 4 is not strong enough to be a cluster, the final clustering has only 3 clusters.
4. In the second phase of merging, all s that belong to more than one cluster are exclusively allocated to the cluster with the stronger c.
5. In the re-evaluation phase, a final check is made to ensure that each cluster fulfils the criterion of minimum information density (as in step 2) after the possible reallocation of sites in the preceding step. Clusters failing the check are dissolved into individual sites.
Note that the redundant counting in the pair-wise sums of step 1 is intentional in favoring stronger sites towards becoming cluster centers. For example, given two sites si and sj in the same neighborhood, such that si has higher information content than sj, si is more likely to qualify as a cluster center as it will have a larger nic.
The parameters d and I were determined as follows. A randomly chosen training sample of positive (regulated) and negative (unregulated/unrelated) genes was used for several rounds of IDBC with a range of values for d and I. At the end of each IDBC round, the total clustered information content, i.e., the sum of information content of all sites found within any cluster across the entire promoter region was computed. The final values chosen (d = 370; I = 24) were those that gave the largest difference in the mean values of total information content within clusters between the positive and negative training sets. The final clustering resulting from this algorithm allows for high total information content within clusters in more ways than one. Cluster membership can be attained by either the proximity of a few strong sites or several closely packed sites of moderate strength. Also, note that since the IDBC algorithm operates on the notion of information density and not an absolute inter-site distance, there is no constraint placed on a requirement for symmetric clusters. In other words, there is no enforced arbitrary bound on the location of ci with respect to the edge of a cluster.
Feature selection
The following summarized features were derived as attributes for further analysis.
a) The sum of information content of all binding sites with positive Ri in a promoter region.
b) The total number of binding sites in a promoter region.
c) Information over x bits – This is the same as "a" except that sites with Ri less than x bits are not included in the sum. Based on training data, x was set to 6 bits.
d) The total number of clusters rich in information content found in each region by the IDBC algorithm.
e) The total information within clusters. This is the same as "a" except that sites not lying within any cluster are not included in the sum. This is a metric of how clustered the information is.
f) Information in top three clusters – The sum of information in the three highest information bearing clusters.
Experimental data used for training and validation
Genes whose expression is regulated by PXR/RXRα (defined as "positive" in machine learning algorithms) or which are unchanged in response to treatment with the PXR ligand, rifampin, (defined as "negative") were identified by microarray analysis and from published literature. Microarray studies were carried out using HepG2-PXR cell lines were generated that stably expressed PXR. HepG2 cells were grown in MEM-alpha medium containing 10% fetal bovine serum and 1% penicillin/streptomycin at at 37C in 5% CO2. HepG2 cells were plated in P60 dishes at 50,000 cells per well. Twenty-four hours later cells were transfected with 5000 ng of human PXR-pcDNA3 (to create HepG2-PXR) or pcDNA3 (InVitrogen) (to create HepG2-NEO) by calcium phosphate co-precipitation and individual clones selected with 1000 μg/ml of G418. Clones were screened for protein expression of PXR and the clone with the highest expression of both was chosen for further study (HepG2-PXR). HepG2-PXR cells were treated with 10 uM rifampin for 48 hrs. HepG2-NEO transfected cells treated for 48 hrs with DMSO served as the control. RNA was isolated from the transfected cell lines from two independent experiments. RNA quality was verified with the "Lab-On-A-Chip" system (Agilent Technologies), reverse transcribed, and cRNA was labeled with Cy3 and and Cy5, respectively for the HepG2-PXR and HepG2-NEO lines. HG-U95 oligonucleotide microarrays (Affymetrix) were hybridized with a mixture of control and HepG2-PXR cRNA and analyzed with the Agilent Gene Array Scanner. Gene expression values were calculated using Affymetrix Microarray Suite software to compare changes in gene expression between the HepG2-NEO treated with DMSO vs. HepG2-PXR cells treated with rifampin. Genes whose expression is unchanged in response to rifampin are interpreted to be unregulated by PXR/RXRα. This assumption may only be valid for the present transfected HepG2 cell lines and it is conceivable that the regulatory status of these genes may be different in other tissues that normally express PXR/RXRα.
Classification algorithms
The summary attributes described in the previous section were separately used as input for the construction of decision trees. Decision trees, neural networks (NN) and the naïve Bayes classifier were used separately, and as part of jury prediction. The J48 algorithm implemented in the WEKA suite of machine learning algorithms [21] based on the C4.5 decision tree builder algorithm [22] was used in this experiment.
The Stuttgart Neural Network Simulator (SNNS) [23] was used for neural network analysis. All attribute values were normalized by the value two standard deviations higher than the respective means observed in the training set, in order to constrain all values to lie in the range between 0 and ~1. For standard backpropagation, a single hidden layer with 2 neurons was used and the log-sigmoid function used for all layers. The latter gives a smooth output in the range of 0–1 that lends itself to probabilistic interpretation. A receiver operating curve (ROC) plot (Fig. 5) based on the training set was used to select the output value for optimal partition. Neural networks were also evaluated by using radial basis functions (RBF) [24] for the neurons in the hidden layer. In this case, the hidden layer had 3 neurons. In both cases, early stopping with keeping the size of the NN to the minimum necessary was used to reduce the possibility of overfitting.
Figure 5 ROC plot for Neural Network cross-validation. The training data was divided into multiple (n = 4 in this figure) non-overlapping sets. Each of the n sets was used to train a different neural network (NN) and tested on the remaining data. A Receiver Operating Curve was generated for each trained network by calculating specificity and sensitivity for different values of the cut-off for the output value to discriminate between regulated and unregulated gene targets. The ideal curve would be collinear with the y-axis for x = 0, and then run parallel to the x-axis as the line y = 1.
In addition, the naïve Bayes classifier was used. The training set was used to compute a frequentist estimate of the probabilities of observing each value of the attribute in either positive or negative genes. Then, given a set of attributes for a gene in a test set, the assignment to the positive or negative class was made by considering the relative probabilities for finding such a promoter region in positive versus negative genes.
Data-partitioning for training, validation, and testing
The data was partitioned n-ways (n = 3 or 4), with each partition being left out of the training and being validated in turn. In addition, given the small number of known positives, leave-out-one validation was also carried out, where each example was left out of the training and then subjected to prediction. Balanced training was carried out by having the same number of positive and negative examples being presented to the classification algorithms. The test set of 13 genes was not part of the training set.
The PWM was derived from 48 validated binding sites [6], including 3 individual binding sites that are present in the CYP3A4 and CYP3A5 promoter regions used to train the model. The neural network should exhibit little if any bias towards recognition of genes containing these 3 specific sequences, because it depends on multiple orthogonal features, and does not imply circularity for the following reasons: i) The sites included from CYP3A4 and CYP3A5 represent only 6% of sites in the model and less than 0.1% of those that are predicted within the respective promoter regions used to train the model. ii) The strength of these individual binding sites (see Fig. 3) is just one of the metrics used as input for the learning algorithm. Most of the training data are based on the locations of the binding sites and their spatial relationships. These criteria are unrelated to binding site strength per se. iii) Finally, the approach fails to yield the correct prediction for CYP3A5, which would had been expected should the validated binding site information contents had disproportionate effect on the NN outcome.
Authors' contributions
DD designed the IDBC algorithm and conceived the overall machine learning strategy: feature representation and selection, choice and parameterization of machine learning algorithms, and validation of results. PKR defined the problem, suggested the use of neural networks, provided the data, the PWM, and scanned the promoter regions. SM implemented the IDBC algorithm and carried out the neural network analysis. VR experimented with other neural network architectures, performed Decision Tree and the Naïve Bayesian Classifier analyses, and carried out the cross-validation and analysis of the test set. EGS performed and provided the results of microarray expression studies. The manuscript was written by DD and PKR. Figures and tables were prepared by VR and DD.
Acknowledgements
We would like to acknowledge support for this project from the University of Missouri Research Board [UMRB Round 2, 2004] to DD and the NIH to PKR [PHS ES 10855-02] and EGS [R01 GM60346].
==== Refs
Schneider TD Information content of individual genetic sequences J Theor Biol 1997 189 427 441 9446751 10.1006/jtbi.1997.0540
Stormo GD DNA binding sites: representation and discovery Bioinformatics 2000 16 16 23 10812473 10.1093/bioinformatics/16.1.16
Altschul SF Madden TL Schaffer 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
Matys V Fricke E Geffers R Gossling E Haubrock M Hehl R Hornischer K Karas D Kel AE Kel-Margoulis OV Kloos DU Land S Lewicki-Potapov B Michael H Munch R Reuter I Rotert S Saxel H Scheer M Thiele S Wingender E TRANSFAC: transcriptional regulation, from patterns to profiles Nucleic Acids Res 2003 31 374 378 12520026 10.1093/nar/gkg108
Marchler-Bauer A Anderson JB Cherukuri PF DeWeese-Scott C Geer LY Gwadz M He S Hurwitz DI Jackson JD Ke Z Lanczycki CJ Liebert CA Liu C Lu F Marchler GH Mullokandov M Shoemaker BA Simonyan V Song JS Thiessen PA Yamashita RA Yin JJ Zhang D Bryant SH CDD: a Conserved Domain Database for protein classification Nucleic Acids Res 2005 33 D192 6 15608175 10.1093/nar/gki069
Vyhlidal CA Rogan PK Leeder JS Development and Refinement of Pregnane X Receptor (PXR) DNA Binding Site Model Using Information Theory: INSIGHTS INTO PXR-MEDIATED GENE REGULATION J Biol Chem 2004 279 46779 46786 15316010 10.1074/jbc.M408395200
Rogan PK Svojanovsky S Leeder JS Information theory-based analysis of CYP2C19, CYP2D6 and CYP3A5 splicing mutations Pharmacogenetics 2003 13 207 218 12668917 10.1097/00008571-200304000-00005
Nalla VK Rogan PK Automated splicing mutation analysis by information theory Hum Mutat 2005 25 334 342 15776446 10.1002/humu.20151
Podvinec M Kaufmann MR Handschin C Meyer UA NUBIScan, an in silico approach for prediction of nuclear receptor response elements Mol Endocrinol 2002 16 1269 1279 12040014 10.1210/me.16.6.1269
Goodwin B Redinbo MR Kliewer SA Regulation of cyp3a gene transcription by the pregnane x receptor Annu Rev Pharmacol Toxicol 2002 42 1 23 11807162 10.1146/annurev.pharmtox.42.111901.111051
Schneider TD Reading of DNA sequence logos: prediction of major groove binding by information theory Methods Enzymol 1996 274 445 455 8902824
Wagner A Genes regulated cooperatively by one or more transcription factors and their identification in whole eukaryotic genomes Bioinformatics 1999 15 776 784 10705431 10.1093/bioinformatics/15.10.776
Berman BP Nibu Y Pfeiffer BD Tomancak P Celniker SE Levine M Rubin GM Eisen MB Exploiting transcription factor binding site clustering to identify cis-regulatory modules involved in pattern formation in the Drosophila genome Proc Natl Acad Sci U S A 2002 99 757 762 11805330 10.1073/pnas.231608898
Alkema WB Johansson O Lagergren J Wasserman WW MSCAN: identification of functional clusters of transcription factor binding sites Nucleic Acids Res 2004 32 W195 8 15215379
Markstein M Markstein P Markstein V Levine MS Genome-wide analysis of clustered Dorsal binding sites identifies putative target genes in the Drosophila embryo Proc Natl Acad Sci U S A 2002 99 763 768 11752406 10.1073/pnas.012591199
Rebeiz M Reeves NL Posakony JW SCORE: a computational approach to the identification of cis-regulatory modules and target genes in whole-genome sequence data. Site clustering over random expectation Proc Natl Acad Sci U S A 2002 99 9888 9893 12107285 10.1073/pnas.152320899
Frith MC Li MC Weng Z Cluster-Buster: Finding dense clusters of motifs in DNA sequences Nucleic Acids Res 2003 31 3666 3668 12824389 10.1093/nar/gkg540
Lamba V Yasuda K Lamba JK Assem M Davila J Strom S Schuetz EG PXR (NR1I2): splice variants in human tissues, including brain, and identification of neurosteroids and nicotine as PXR activators Toxicol Appl Pharmacol 2004 199 251 265 15364541 10.1016/j.taap.2003.12.027
Gadiraju S Vyhlidal CA Leeder JS Rogan PK Genome-wide prediction, display and refinement of binding sites with information theory-based models BMC Bioinformatics 2003 4 38 12962546 10.1186/1471-2105-4-38
Ester M Kriegel HP Sander J Xu X A density-based algorithm for discovering clusters in large spatial databases Proceedings of the 1996 International Conference on Knowledge Discovery and Data Mining (KDD '96) 1996 226 231
Witten IH Frank E Data mining: Practical machine learning tools and techniques with Java implementations 1999 1st San Francisco, Morgan Kaufmann 416
Quinlan JR C4.5: Programs for machine learning 1993 San Francisco, Morgan Kaufmann 302
Zell AKTMNST Recent Developments of the Neural Network Simulator Proceedings of the Applications of Neural Networks Conference, SPIE 1991 1294
Poggio T Girosi F Networks for approximation and learning Proceedings of the IEEE 1990 78 1481 1497 10.1109/5.58326
Handschin C Meyer UA Induction of drug metabolism: the role of nuclear receptors Pharmacol Rev 2003 55 649 673 14657421 10.1124/pr.55.4.2
|
16115317
|
PMC1208855
|
CC BY
|
2021-01-04 16:27:26
|
no
|
BMC Bioinformatics. 2005 Aug 22; 6:204
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-204
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2111612239510.1186/1471-2105-6-211Softwarestam – a Bioconductor compliant R package for structured analysis of microarray data Lottaz Claudio [email protected] Rainer [email protected] Max Planck Institute for Molecular Genetics and Berlin Center for Genome Based Bioinformatics, Ihnestr. 73, D-14195 Berlin, Germany2005 25 8 2005 6 211 211 16 3 2005 25 8 2005 Copyright © 2005 Lottaz and Spang; licensee BioMed Central Ltd.2005Lottaz and Spang; 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
Genome wide microarray studies have the potential to unveil novel disease entities. Clinically homogeneous groups of patients can have diverse gene expression profiles. The definition of novel subclasses based on gene expression is a difficult problem not addressed systematically by currently available software tools.
Results
We present a computational tool for semi-supervised molecular disease entity detection. It automatically discovers molecular heterogeneities in phenotypically defined disease entities and suggests alternative molecular sub-entities of clinical phenotypes. This is done using both gene expression data and functional gene annotations.
We provide stam, a Bioconductor compliant software package for the statistical programming environment R. We demonstrate that our tool detects gene expression patterns, which are characteristic for only a subset of patients from an established disease entity. We call such expression patterns molecular symptoms. Furthermore, stam finds novel sub-group stratifications of patients according to the absence or presence of molecular symptoms.
Conclusion
Our software is easy to install and can be applied to a wide range of datasets. It provides the potential to reveal so far indistinguishable patient sub-groups of clinical relevance.
==== Body
Background
Microarray analysis is among the most promising clinical applications of modern genomics. It opens perspectives for more reliable and efficient diagnosis of established tumor entities [1,2], risk group determination [3,4], and the prediction of response to treatment [5]. In the supervised setting, various software tools implementing algorithms from statistical learning theory are available and have been evaluated in the context of microarray data (e.g. [6-10]).
All these methods aim for reproducing or predicting predefined clinical phenotypes. However, often clinical phenotypes will not be homogeneous from a molecular point of view. For example, when distinguishing between recurrent and non-recurrent disease, it is of course possible that recurrence has various molecular backgrounds. If this is the case, one will expect different molecular changes in different patients, and purely supervised analysis is unsatisfactory.
In several studies, unsupervised clustering algorithms have been applied to patient profiles, with the aim to define novel disease entities [11-14]. However, clustering of patients is not straightforward, since the clinical relevance of a clustering result is often unclear. It is quite possible that a given clustering reflects unimportant covariates like gender and age or even experimental artifacts. This is usually avoided by visual inspection of the clustered data and an educated manual selection of interesting genes. Automated software tools for this problem are not available so far.
We have recently suggested a novel algorithm for semi-supervised analysis called structured analysis of microarrays [15]. We consider the setting where a disease group is to be distinguished from a set of patients with a different clinical phenotype (controls). Instead of determining a single global signature to detect all disease cases, we generate several local signatures, which identify only subsets. We call the local signatures molecular symptoms. A special feature of the method is that it produces multiple candidate symptoms and characterizes each by a functional annotation, like patients with poor prognosis and altered expression of apoptosis related genes. The functional annotations stored in the Gene Ontology (GO) are used to ensure biological focus.
In GO [16], terms describing biological processes, molecular functions and cellular localizations are organized in a directed acyclic graph, where each node represents a biological process and child-terms are either members or representatives of their parent-terms. Genes are attributed to nodes according to the knowledge the biological research community has gathered so far. Molecular symptoms found by stam exclusively contain genes associated with one node of the Gene Ontology and therefore have a biological focus.
For each patient and each node classifier stam calculates a value between 0 and 1 for each relevant molecular symptom. These values indicate how likely it is that a patient belongs to the disease class according to the corresponding molecular symptom. In addition, we suggest to use patterns of absence or presence of molecular symptoms to stratify patients into subclasses. An overview of the algorithm is given in Figure 1.
Figure 1 Overview on structured analysis of microarrays. Input of stam is a gene expression dataset, the structure of the Gene Ontology and associations of genes with GO terms. The output of the method is a resolved diagnosis per patient according to molecular symptoms, arid thus a molecular stratification of patients according to absence and presence of these symptoms.
Implementation
A detailed description of structured analysis of microarrays is given in [15]. Here we only give a brief review of the method.
We use Gene Ontology's hierarchical structure. Based on the GO graph of biological terms, stam generates a classifier graph holding one classifier for each node. The classifiers only depend on genes annotated to corresponding nodes or their descendants. In a nutshell, stam consists of the following steps as illustrated in Figure 2:
Figure 2 stam's algorithm in a nutshell. The algorithm for structured analysis of microarrays splits into three phases: classification in leaf nodes, propagation through inner nodes and shrinkage of the classifier graph. Calibration by the user allows for fine-tuning of specificity versus sensitivity in classifier evaluation and redundancy versus performance tradeoff in graph shrinkage.
• generate a rooted, directed classifier graph according to the gene Ontology,
• construct leaf-node classifiers based on expression values of genes, which are directly annotated to the leaf nodes,
• propagate the results through inner nodes to the root,
• and shrink the classifier graph to determine a concise set of molecular symptoms.
We have implemented the algorithm based on the R package for statistical computing [17]. Time-consuming parts of the method are written in C to improve computational performance. Furthermore, we rely on packages from the Bioconductor suite of bioinformatics tools [18].
The raw classifier graph
Starting from a node of interest specified by the user, stam generates a graph of classifiers according to the structure of the Gene Ontology. The graph is generated anew for each chip type. Any GO node can be chosen to start the procedure with this node as root of the graph. The default is the root of the biological process branch of the gene ontology. Our implementation uses Bioconductor meta-data packages to obtain chip-specific associations of probe-sets with genes as well as the generic GO structure. Table 1 summarizes the annotation data, which is currently available for Affymetrix-GeneChip® microarrays.
Table 1 Gene Ontology annotations available in Bioconductor – For the microarrays listed in this table, Bioconductor meta data packages are available. The second column gives the number of leaf nodes the third column the number of inner nodes considered when generating classifier graphs. The last column reports the ratio of probe-sets being associated with any leaf node.
Species Probe-sets Leaf nodes Inner nodes Annotated
hgu133a human 22283 1649 1049 67.6%
hgu133b human 22645 1136 783 30.7%
hgul33plus2 human 54675 1725 1094 44.1%
hu6800 human 7129 1300 872 84.7%
hgu95av2 human 12625 1492 966 76.2%
hgu95b human 12620 972 669 33.3%
hgu95c human 12646 895 633 27.7%
hgu95d human 12644 866 603 22.9%
hgu95e human 12639 935 641 32.1%
mgu74av2 mouse 12488 1379 934 63.5%
mgu74bv2 mouse 12411 975 696 33.7%
mgu74cv2 mouse 11934 826 590 26.0%
moe430a mouse 22690 1538 1017 61.4%
moe430b mouse 22575 997 719 21.0%
rgu34a rat 8799 974 689 46.0%
rgu34b rat 8791 403 332 7.7%
rgu34c rat 8789 439 345 8.5%
rae230a rat 15866 1032 718 23.5%
rae230b rat 15276 319 297 3.1%
yg98 yeast 6777 1028 667 80.8%
YEAST yeast 5799 1030 668 99.9%
Leaf-node classifiers
Each leaf node contains a set of associated genes. The classifiers for leaf nodes are constructed using only these genes. For each patient, it returns a number between zero and one. Zero indicates clear evidence for the control group, one indicates clear evidence for the disease group and intermediate values represent levels of uncertainty. In the current implementation we use the shrunken centroid classifiers [9] implemented in the Bioconductor package pamr for leaf node predictions.
Propagation of classifier results
For propagating leaf node results to inner nodes, weighted sums of child classifications are used. Children with good classification performance receive more weight than those with poor performance. Thereby, stam measures performance according to the desired properties of molecular symptoms by punishing low specificity more severely than lack of sensitivity. Prediction results are propagated from the leaf nodes towards the root in a postorder traversal of inner nodes. Hence, stam always computes results for all children before it computes results for the parent node. The root naturally displays an overall classification result.
Classifier graph shrinkage
Many biological processes are not involved with the investigated phenotype. Therefore, stam simplifies the classifier graph by eliminating irrelevant branches. This is done in analogy to gene shrinkage in the shrunken centroid algorithm [9]. stam controls the shrinkage process by calibrating a shrinkage parameter in a cross validation setting. We define an objective function considering two independent goals: good predictive performance in the root and a set of molecular symptoms for patient stratification. For the second goal aggressive shrinkage is counterproductive, since it eliminates too many inherently heterogeneous molecular symptoms.
The program's output is a classifier graph, where each node represents a molecular symptom. We have shown in [15] that the collection of these classifiers yields state-of-the-art predictive performance and allows for a resolved diagnosis. A stam-diagnosis is more resolved than the classification provided for training because molecular symptoms are usually absent in some of the disease patients. Patterns of absence and presence of molecular symptoms identify smaller groups of patients and thus provide an additional molecular stratification of patients. Due to this unsupervised aspect within our supervised method, we call our approach semi-supervised.
Results
Installing stam works like any other Bioconductor package either by downloading and installing from a local copy or directly through the internet. We provide packaged versions ready for download on the Bioconductor web site [19] as well as on our own web page [20].
Computing with stam is done on a command-line level. Gene expression matrices can either be provided as plain R matrices or as exprSet Bioconductor objects. R can read tab-delimited files written by any other software. stam provides functions for cross validation, model fit, and prediction. First, cross validation is applied on training data to find the appropriate shrinkage level. The second function computes a classifier model given this shrinkage level. This model can than be used by the prediction function to diagnose new patients and assign them to novel molecular disease entities. For convenience all three steps can be performed by one call of an evaluation function. This function can also randomly split patients into a training and a test set.
For further illustration, we use a data set from a microarray study on lung cancer [1]. The investigators have analyzed gene expression profiles from 186 lung cancer as well as 17 non-tumor lung biopsies using hierarchical and probabilistic clustering with the goal to uncover novel molecular lung cancer entities. The study uses the HG-U95Av2 microarray from Affymetrix and contains samples from various subtypes of lung cancer. For illustration, we apply stam with the squamous cancers forming the disease group of interest and all other cancers as controls. In the dataset there are 21 squamous carcinomas. The 203 samples are randomly split into a training set (135 samples containing 14 squamous) and a test set (68 samples containing 7 squamous).
Automatic and manual calibration of graph shrinkage
Graph shrinkage can be calibrated automatically by cross validation or manually. To this end, stam provides two performance scores and corresponding plots. The first score is root performance measured as what would be the log-likelihood in a probabilistic setting. The second score, called mean redundancy, represents the diversity of molecular symptoms in the graph. It is the mean of pairwise redundancies. Here as well we interpret classifier outputs as probabilities. Our definition of pairwise redundancy is then the negative logarithm of the probability for unequal class prediction, stam aims for small values for both scores. While automated optimization uses an affine combination of the scores as objective function [15], manual calibration allows for a problem specific adjustment of the performance versus diversity trade off. The left pane of Figure 3 displays both scores for the whole classifier graph together with the error rate in the root node depending on the graph shrinkage level. The right pane of Figure 3 shows the number of nodes in the graph, and the number of genes accessible through these nodes. Figure 4 resolves the scores node per node. It contains scatter plots displaying sensitivity versus specificity and redundancy versus performance of single nodes. The redundancy of single nodes is defined in [15]. It is high if the corresponding classifier provides results which are similar to those of other classifiers in the shrunken graph.
Figure 3 Cross validation evaluation. For several graph shrinkage candidates error rates in the root node, the root performance and the mean redundancy (top panel), as well as the number of nodes remaining in the shrunken graph and the number of genes accessible through these (bottom panel) are shown for the task to identify squamous lung cancers. The vertical lines show the automatically chosen shrinkage level.
Figure 4 Nodewise evaluation. Classifiers for all nodes representing molecular symptoms have different performance. The left panels oppose performance to redundancy (to all other nodes remaining in the shrunken classifier graph). The right panels contrast sensitivity to specificity. The upper panels are drawn using training data while the lower ones are generated based on the test data.
Browsable results
Results are written on interlinked HTML pages. Links allow navigation along the edges of the classifier graph. The pages contain classification results and performance evaluation for all nodes as well as overall information on cross-validation, model fit and root diagnosis of patients. For inner nodes the propagation weights are provided and for leaf nodes the genes used for classification can be displayed. The user can further explore term definitions and probe-set annotations through external links to the Gene Ontology and the Affymetrix web sites.
If the package graphviz [21] is installed, an interactive graphical representation of the classifier graph is included. Links on the nodes lead to the corresponding node-specific result page. Figure 5 shows an example using the lung cancers dataset.
Figure 5 Shrunken classifier graph. For the squamous lung cancer identification this classifier graph containing 90 nodes is generated. Its 24 leaf nodes use 1614 probes (12.8% of all probes on the HG-U95Av2 microarray).
For visualizing patient stratification a molecular symptoms image is generated, as illustrated in Figure 6. Classifier outputs are color coded with bright colors representing presence, and dark color absence of a molecular symptom. Columns represent patients, and rows molecular symptoms. Rows and columns are arranged such that similar rows and columns are placed together.
Figure 6 Molecular symptoms image. Resolved diagnoses for all squamous lung cancers are represented by the columns in this image. Rows represent molecular symptoms. The seven samples from the test set are marked with capital letters on the x-axis. Only a few samples including the test sample E present all molecular symptoms. Several samples lack some of the molecular symptoms. For instance, the symptom attributed to "intra-cellular protein transport" is not present in some of the training samples as well as samples B, C, F and G from the test set.
Interactive use of stam
For the interactive exploration of the graph shrinkage level and other parameters affecting visual output, we have implemented a WWW based solution. stam can write HTML forms for these parameters directly on the HTML result pages. We provide CGI scripts with the package, which collect the user's entries and pass them to the stam server. This server is also included in the R package. It consists of an R function which, is constantly polling for user requests submitted via the internet. The architecture is illustrated in Figure 7. The user's WWW browser is redirected to a progress page which reloads automatically every second. As soon as the stam server has finished treating the request, the browser is redirected to an updated result page.
Figure 7 Client-server architecture for interactive exploration. Architecture of stam's server feature. CGI scripts deposit requests in a task repository from where the stam-server reads and executes them.
Discussion
In this paper we present a software package to integrate biological annotation into statistical class prediction analysis of microarray data in an a priori fashion. We use the functional annotation collected in the Gene Ontology database to construct structured classifiers. Class predictions are computed for each term in the Gene Ontology which is related to the disease. Our method allows for biologically resolved diagnosis of patients. It is thus able to stratify complex clinical phenotypes, where different patients who show the phenotype may display different molecular characteristics.
We suggest structured analysis of microarrays for different applications. In addition to predictive performance we also aim for making underlying disease mechanisms transparent. We do this by identifying molecular symptoms associated to subsets of patients in the disease group. Molecular symptoms are always restricted to well defined biological processes. Patients who are positive for a molecular symptom display specific gene expression in the corresponding process. Not all patients in the disease group are positive for every identified molecular symptom, but some patients can be positive for more than one of them. Using patterns of absence and presence of molecular symptoms, we define an additional molecular stratification of patients.
Conclusion
In summary, stam is a novel algorithm for uncovering previously unknown molecular disease sub-entities. The R package is easily accessible to all researchers working with Affymetrix® oligo chips.
Availability and requirements
The Bioconductor compliant R package stam is available through the Bioconductor web site [19]. Alternatively we also make it available on the Computational Diagnostics Software Page at the Max Planck Institute for Molecular Genetics in Berlin [20]. There, the source package is available for download and we also run a Bioconductor compliant package repository.
Our software is written for the R package for statistical computing downloadable from [22]. An installation of version 2.0.0 or later of the R software is needed to run stam. Our software is based on other Bioconductor packages, namely the meta data packages for the Gene Ontology annotations. We recommend to install release 1.5 of the Bioconductor suite from [19]. For the layout of classifier graphs, we rely on the graphviz package versions 1.10 or later available at [23].
We have extensively used stam on Linux installations on i686 based machines as well as alpha based UNIX machines running OSF1 and True64 operating systems.
Acknowledgements
The authors are grateful to Florian Markowetz, Jörn Tödling, Jochen Jäger, Stefanie Scheid and Stefan Bentink from our work group as well as to our partners Renate Kirschner-Schwabe, Christian Hagemeier and Karl Seeger from the Charité Medical Center for fruitful discussions. This research has been supported by BMBF grants 031U117/217 of the German Federal Ministry of Education and the National Genome Research Network.
==== Refs
Bhattacharjee A Richards W Staunton J Li C Monti S Vasa P Ladd C Beheshti J Bueno R Gillette M Loda M Weber G Mark E Lander E Wong W Johnson B Golub T Sugarbaker D Meyerson M Classification of human lung carcinomas by mRNA expression profiling reveals distinct adenocarcinoma subclasses Proc Natl Acad Sci U S A 2001 98 13790 5 11707567 10.1073/pnas.191502998
Yeoh EJ Ross ME Shurtleff SA Williams WK Patel D Mahfouz R Behm FG Raimondi SC Relling MV Patel A Cheng C Campana D Wilkins D Zhou X Li J Liu H Pui CH Evans WE Naeve C Wong L Downing JR Classification, subtype discovery, and prediction of outcome in pediatric acute lymphoblastic leukemia by gene expression profiling Cancer Cell 2002 1 133 45 12086872 10.1016/S1535-6108(02)00032-6
Huang E Cheng SH Dressman H Pittman J Tsou MH Horng C Bild A Iversen E Liao M Chen CM West M Nevins JR Huang AT Gene expression predictors of breast cancer outcomes Lancet 2003 361 1590 6 12747878 10.1016/S0140-6736(03)13308-9
van't Veer L Dai H van de Vijver M He Y Hart A Mao M Peterse H van der Kooy K Marton M Witteveen A Schreiber G Kerkhoven R Roberts C Linsley P Bernards R Friend S Gene expression profiling predicts clinical outcome of breast cancer Nature 2002 415 530 6 11823860 10.1038/415530a
Cheok MH Yang W Pui CH Downing JR Cheng C Naeve CW Relling MV Evans WE Treatment-specific changes in gene expression discriminate in vivo drug response in human leukemia cells Nature Genet 2003 34 85 90 12704389 10.1038/ng1151
Ben-Dor A Bruhn L Friedman N Nachman I Schummer M Yakhini Z Tissue classification with gene expression profiles J Comp Biol 2000 7 559 83 10.1089/106652700750050943
Dudoit S Fridlyand J Speed T Comparison of Discrimination Methods for the Classification of Tumors Using Gene Expression Data J Amer Stat Assoc 2002 97 77 87 10.1198/016214502753479248
Slonim DK Tamayo T Mesirov JP Golub TR Lander ES Class Prediction and Discovery Using Gene Expression Data Proc Internatl Conf Comp Biol 2000 263 72
Tibshirani R Hastie T Narasimhan B Chu G Diagnosis of multiple cancer types using shrunken centroids of gene expression Proc Natl Acad Sci 2002 99 6567 72 12011421 10.1073/pnas.082099299
West M Blanchette C Dressman H Huang E Ishida S Spang R Zuzan H Olson JA Marks JR Nevins JR Predicting the clinical status of human breast cancer by using gene expression profiles Proc Natl Acad Sci 2001 98 11462 7 11562467 10.1073/pnas.201162998
Alizadeh AA Eisen MB Davis RE Ma C Lossos IS Rosenwald A Boldrick JC Sabet H Tran T Yu X Powell JI Yang L Marti GE Moore T Hudson J JrLu L Lewis DB Tibshirani R Sherlock G Chan WC Greiner TC Weisenburger DD Armitage JO Warnke R Levy R Wilson W Grever MR Byrd JC Botstein D Brown PO Staudt LM Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling Nature 2000 403 503 11 10676951 10.1038/35000501
Monti S Savage KJ Kutok JL Feuerhake F Kurtin P Mihm M Wu B Pasqualucci L Neuberg D Aguiar RC Cin PD Ladd C Pinkus GS Salles G Harris NL Dalla-Favera R Habermann TM Aster JC Golub TR Shipp MA Molecular profiling of diffuse large B-cell lymphoma identifies robust subtypes including one characterized by host inflammatory response Blood 2005 105 1851 1861 15550490 10.1182/blood-2004-07-2947
Perou CM Sorlie T Eisen MB van de Rijn M Jeffrey SS Rees CA Pollack JR Ross DT Johnsen H Akslen LA Fluge O Pergamenschikov A Williams C Zhu SX Lonning PE Borresen-Dale AL Brown PO Botstein D Molecular portraits of human breast tumours Nature 2000 406 747 752 10963602 10.1038/35021093
Rosenwald A Wright G Chan WC Connors JM Campo E Fisher RI Gascoyne RD Muller-Hermelink HK Smeland EB Giltnane JM Hurt EM Zhao H Averett L Yang L Wilson WH Jaffe ES Simon R Klausner RD Powell J Duffey PL Longo DL Greiner TC Weisenburger DD Sanger WG Dave BJ Lynch JC Vose J Armitage JO Montserrat E Lopez-Guillermo A Grogan TM Miller TP LeBlanc M Ott G Kvaloy S Delabie J Holte H Krajci P Stokke T Staudt LM The use of molecular profiling to predict survival after chemotherapy for diffuse large-B-cell lymphoma N Engl J Med 2002 346 1937 1947 12075054 10.1056/NEJMoa012914
Lottaz C Spang R Molecular decomposition of complex clinical phenotypes using biologically structured analysis of microarray data Bioinformatics 2005 21 1971 8 15677704 10.1093/bioinformatics/bti292
The Gene Ontology Consortium Gene ontology: Tool for the unification of biology Nature Genet 2000 25 25 9 10802651 10.1038/75556
R Development Core Team R: A language and environment for statistical computing 2004 R Foundation for Statistical Computing, Vienna, Austria
Gentleman RC Carey VJ Bates DM Bolstad B Dettling M Dudoit S Ellis B Gautier L Ge Y Gentry J Hornik K Hothorn T Huber W Iacus S Irizarry R Leisch F Li C Maechler M Rossini AJ Sawitzki G Smith C Smyth G Tierney L Yang JY Zhang J Bioconductor: open software development for computational biology and bioinformatics Genome Biol 2004 5 R80 15461798 10.1186/gb-2004-5-10-r80
The BioConductor Home Page
The Computational Diagnostics Software Page
Gansner ER North SC An open graph visualization system and its applications to software engineering Software Practice and Experience 2000 30 1203 33 10.1002/1097-024X(200009)30:11<1203::AID-SPE338>3.0.CO;2-N
The R Project for Statistical Computing
Graphviz – Graph Visualization Software
|
16122395
|
PMC1208856
|
CC BY
|
2021-01-04 16:27:26
|
no
|
BMC Bioinformatics. 2005 Aug 25; 6:211
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-211
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2131612487610.1186/1471-2105-6-213Methodology ArticleImproved prediction of critical residues for protein function based on network and phylogenetic analyses Thibert Boris [email protected] Dale E [email protected] Rio Gabriel [email protected] Buck Institute For Age Research, 8001 Redwood Blvd, Novato, CA 94945, USA2 University of California, San Francisco, San Francisco, CA 94143, USA3 Instituto de Fisiología Celular, UNAM, Circuito Exterior, Ciudad Universitaria, 04510, México, D.F2005 26 8 2005 6 213 213 21 12 2004 26 8 2005 Copyright © 2005 Thibert et al; licensee BioMed Central Ltd.2005Thibert 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
Phylogenetic approaches are commonly used to predict which amino acid residues are critical to the function of a given protein. However, such approaches display inherent limitations, such as the requirement for identification of multiple homologues of the protein under consideration. Therefore, complementary or alternative approaches for the prediction of critical residues would be desirable. Network analyses have been used in the modelling of many complex biological systems, but only very recently have they been used to predict critical residues from a protein's three-dimensional structure. Here we compare a couple of phylogenetic approaches to several different network-based methods for the prediction of critical residues, and show that a combination of one phylogenetic method and one network-based method is superior to other methods previously employed.
Results
We associate a network with each member of a set of proteins for which the three-dimensional structure is known and the critical residues have been previously determined experimentally. We show that several network-based centrality measurements (connectivity, 2-connectivity, closeness centrality, betweenness and cluster coefficient) accurately detect residues critical for the protein's function. Phylogenetic approaches render predictions as reliable as the network-based measurements, although, interestingly, the two general approaches tend to predict different sets of critical residues. Hence we propose a hybrid method that is composed of one network-based calculation – the closeness centrality – and one phylogenetic approach – the Conseq server. This hybrid approach predicts critical residues more accurately than the other methods tested here.
Conclusion
We show that network analysis can be used to improve the prediction of amino acids critical for protein function, when utilized in combination with phylogenetic approaches. It is proposed that such improvement is due to the complementary nature of these approaches: network-based methods tend to predict as critical those residues that are highly connected and internal (i.e., non-surface), although some surface residues are indeed identified as critical by network analyses; whereas residues chosen by phylogenetic approaches display a lower overall probability of being surface inaccessible.
==== Body
Background
This article deals with the problem of predicting critical amino acid residues for a given protein's function, i.e., those residues that, if mutated, result in a loss of protein function by the lack of proper folding and/or the inability to perform a biochemical function (hereafter referred to simply as critical residues). A common approach to this problem consists of aligning multiple orthologous protein sequences, and predicting as critical residues the most highly conserved ones. This approach is known as the phylogenetic approach, and requires that a significant number of distinct protein sequences be aligned. This approach also assumes that the aligned sequences from different species serve the same function in the different species. However, for an increasing number of proteins studied by the Structural Genomic Consortium, a different trend is observed: for a given protein there may be a lack of a significant number of homologous sequences, whereas the three-dimensional structure (3D structure) of the protein of interest may already have been described. For this set of proteins, having an accurate way to identify critical residues in the absence of known orthologues – i.e., simply from the protein's three-dimensional structure – may be valuable. Identifying critical residues from protein structures can also be extended to those proteins for which there are a significant number of homologous sequences [1]. Furthermore, the prediction of critical residues from protein structures has been shown to be useful in validating protein structure predictions [2,3].
Over the past few decades, network analyses have been used to model diverse systems such as the World Wide Web, social systems, and biological systems (e.g., protein-protein interactions or the cellular metabolic network) [4]. Only very recently has such an approach been used to predict critical amino acids from protein structures [[5,6] and [7]].
By definition, a network is composed of two sets: a set of vertices and a set of edges (an edge being a pair of connected vertices). A protein can be modelled in a network as follows: each vertex of the network represents an amino acid residue, and each edge represents a chemical interaction between any two amino acid residues. In the particular case in which the 3D structure of the protein is known, it can be assumed, for the sake of the model, that there is a chemical interaction between two amino acids if they are sufficiently close to one another (less than 5 Angstroms apart in our case; see Methods). This gives a way of building a network of amino acid interactions [[5,6] and [7]], with the obvious caveat that the accuracy of such a model will be affected by the accuracy of the assumption of chemical interactions between different residues.
Previous studies have related specific centrality measurements of these amino acid interaction networks to the tolerance for replacement (i.e., critical nature) of the various amino acids within the network. In particular, the most traversed vertices (also referred as vertices with the greatest betweenness) have been shown to be important for folding [5] and for the functions of the modelled proteins [7]. These vertices, when detected from several structures for a given protein, relate more accurately with critical residues than the most conserved residues detected from some phylogenetic approaches [7]. Hence, betweenness, a network centrality measurement, is associated with the capacity of any given amino acid residue to play a critical role in a protein's function. These results indicate that critical residues are central to the interactions among the residues in a protein structure. However, whether or not the experimentally determined critical residues are best predicted by measures of betweenness, as opposed to any other centrality measurement, has not been tested. Evaluation of other centrality measurements may also help to understand the significance of the observed relationship.
In the current work we compare five different network centrality measurements, applying them to the study of the structure/function relationship of proteins. Previous studies on the structure/function relationship of proteins have shown that critical residues tend to display specific geometrical characteristics, such as distorted dihedral angles or central location, buried within the proteins' cores ([8] and references therein). However, whether these geometrical properties are related to network centrality (e.g., betweenness) has not been studied. In this work, we analyze some of these relationships, as well.
It is possible to define several centrality measurements at a vertex p of a network. Some of these centrality measurements have already been used for modelling different systems. One of the most direct and widely used centrality measurements is the connectivity cp ([9], for example), which is the number of neighbours of p (i.e., the number of vertices that share an edge with p). Another, the clustering coefficient [9], defined by 2np/cp(cp - 1), where np is the number of connections between the neighbours of p, estimates the degree of interconnectivity among the neighbours of p. These two measurements are local, in the sense that their values only depend on the vertex p and on its neighbours.
There are also global centrality measurements assignable to every vertex in a network (i.e., whose values depend on the structure of the whole network). Many of these are based on the notion of shortest path. We say that a path connects two vertices p and q if there exists a sequence of edges that connect p and q. The length of a path is its number of edges. The distance l(p, q) between p and q is the length of the shortest path connecting p and q. Hence, one of these global centrality measurements is the closeness centrality – also referred as chemical distance [[10,11] and [6]] – and is defined as the inverse of the average distance value between p and all the other vertices of the network. Another, the eccentricity [12], is the inverse of the maximum distance from p to any other vertex. We can also define the betweenness at a vertex p [5,7] as the number of all shortest paths connecting two vertices of the network that pass through p normalized by the total number of paired vertices.
Here, we show that closeness centrality correlates more accurately with critical residues than betweenness or any other centrality measurement tested. Also, we show that the sets of critical residues predicted by any of the centrality measurements (e.g., betweenness, closeness centrality) do not completely overlap with those based on phylogeny. Hence, we present evidence that combined predictions based on closeness centrality and phylogeny improve the predictions achieved by any of these approaches alone. Finally, we observe that the critical residues detected by centrality measurements correlate with low-accessibility solvent residues, although these are not exclusively in the protein's core. These results are used to suggest an explanation for the improvement achieved by the combined approach to detect critical residues.
Results and discussion
Five methods based on network analyses
We propose and compare five different centrality measurements to predict critical amino acids from a protein structure: connectivity, cluster coefficient, closeness centrality, betweenness and 2-connectivity. The 2-connectivity at a vertex p is the number of vertices connected to p with fewer than 2 edges. This new measurement is local and can be seen as a generalization of the connectivity.
Let us describe now how we proceed to determine the critical residues from one of these centrality measurements, for example the closeness centrality (the method described for closeness centrality can be extended to the other four measurements). We assume that the amino acids that have the largest closeness centrality are critical for the function of the protein. Since the set of predicted critical residues depends on the cut-off value used to define the "largest closeness centrality", we evaluated all possible cut-off values.
Hence, we define five methods to predict critical residues, each one based on a different centrality measurement. For each, the set of critical amino acids is defined to be composed of the vertices:
• Method 1: with the largest closeness centrality,
• Method 2: with the largest betweenness,
• Method 3: with the largest 2-connectivity,
• Method 4: with the largest connectivity,
• Method 5: with the smallest cluster coefficient.
We compared these methods on a set of five well-characterized proteins in terms of their 3D structures and biochemical identification of critical residues. These include TEM1 beta-lactamase, T4 lysozyme and HIV-1 protease [[13,14] and [15]]. Exhaustive mutagenesis and analysis has been utilized to identify every residue critical for the functioning of these three proteins, and therefore they serve as excellent test proteins for the computational methods described here. We also include Barnase [16] and the bacteriophage f1 gene V protein [17]. Two differences are relevant for these two proteins with respect to the previous set of three proteins mentioned before: the number of homologous sequences available is reduced and the sensitivity of the biological assay is augmented. For the beta-lactamase and the protease of the HIV-1, more than 40 unique homologous sequences were found by the Conseq server, while for the bacteriophage f1 gene V protein only 6 homologous sequences were found on the Hssp database. For Barnase, fewer than 5 homologous sequences were found by the Conseq server and 20 homologous sequences were found within the Hssp database (see Methods). Although our choice of these proteins was not simply because of the relative lack of homologous sequences, this limitation of the current experimental data serves to illustrate the point that a method that does not require a large number of homologous sequences is desirable as a complement to phylogenetic analyses. Finally, the sensitivity of the biological assay employed to identify critical residues in Barnase (inactive mutants were defined as those having < 1% of wild-type activity) was much higher than the one employed for the other four proteins, thus reducing the total number of critical residues detected experimentally.
In order to evaluate the quality of our predictions, we use the sensitivity and the specificity statistical measurements. However, one of these two criteria alone is not sufficient to compare the different methods (see Methods). Therefore, in order to obtain a single value to evaluate the predictions of our methods that includes both sensitivity and specificity, we propose three different strategies:
1. For a given centrality measurement and cut-off value, we calculate the error dist(centr. meas., cut-off) (see Methods). This error considers simultaneously the specificity and the sensitivity. It is small if both the sensitivity and the specificity are good. This error allows us to sort all the predictions obtained by any centrality measurement/cut-off value used (this is used in Table 1 and 4).
Table 1 Sorting of the methods according to the sensitivity and specificity criteria (5 proteins set)
Method Proportion of predicted residues Error Sen. Spe.
Union set (10) 38% 0,41 71% 73%
2-connectivity (3) 42% 0.43 74% 67%
Conseq 42% 0,44 76% 67%
Closeness (1) 38% 0.44 68% 71%
Intersection set (9) 38% 0,44 68% 71%
Connectivity (4) 38% 0,45 67% 71%
1/cluster (5) 38% 0.48 63% 70%
Betweenness (2) 38% 0.50 62% 69%
1/eccentricity 45% 0.53 68% 60%
Hssp 38% 0.54 62% 66
For every method, results are obtained in average over five 3D structures: 1HIV, 2LZM, 1BTL, 1A2P and 1GVP. Numbers are obtained the following way: For a given cut-off value (between 0% and 100% of residues that are predicted), we calculate the average (over the five proteins) of the dist(method, cut-off) error, the sensitivity and the specificity. This is done for several cut-off values. Then we choose the cut-off for which the average of the dist(method, cut-off) error is minimal. Note that in most of the cases, the error is the smallest if we predict around 40% of the amino acids to be essential. Furthermore the smallest error for 1/eccentricity is worse than the smallest errors for the five other centrality measurements (1 to 5).
Table 4 Sorting of different methods according to the sensitivity and specificity criteria (128 proteins set)
Method Proportion of amino acids predicted Error Sen. Spe.
Hssp 24% 0,41 72% 77%
Conseq 27% 0,44 71% 74%
Union set (10) 25% 0,50 61% 76%
Closeness 36% 0,56 64% 64%
Intersection set (9) 32% 0,58 60% 68%
Betweenness (2) 33% 0,58 58% 67%
2-connectivity (3) 37% 0,60 60% 64%
1/cluster (5) 49% 0,63 68% 51%
1/eccentricity 24% 0,64 44% 77%
Connectivity (4) 49% 0,65 65% 51%
Results are obtained in average other the set of 128 proteins. Numbers are obtained by calculating the average as in Table 1.
2. For a given centrality measurement and cut-off value, we also compare the sensitivity value to the proportion of predicted amino acids (number of predicted residues/total number of residues in a protein). This is relevant, because the sensitivity is greater than the proportion of predicted amino acids if and only if the method is better than random (see Methods, Figs. 1, 2 and 3). Furthermore, by plotting the sensitivity versus the proportion of predicted amino acids (see Figs 1, 2 and 3), we can compare all the methods with all of their different cut-off values.
Figure 1 Comparison of 6 different network-based methods for essential amino acids prediction. For each method (Methods 1–6 and "Random <1%") the abscissa x is the proportion of amino acids predicted (by the appropriate centrality measurement) and y is the proportion of essential amino acids predicted (i.e. sensitivity). For example, x = 0.2 means that we select 20% of the amino acids. If we consider the closeness centrality, it means that we select 20% of the amino acids that have the largest closeness centrality. We can notice in this case that we select around 38% of the essential amino acids. The slide of the "Maximum value" curve is the proportion of amino acids that are essential. All the curves have to be under this curve. The better a method is, the closer to the "Maximum value" its associated curve is. A method is better than random if its curve is over the "Average value: y = x" curve. A method is worse than random if its curve is under the "Average value: y = x" curve. The calculations (of the curves 1–6) are made as an average over the five networks representing HIV-1 protease, TEM1 beta-lactamase, T4 lysozyme, Barnase and bacteriophage f1 gene V protein. The curve "Random <1%" only depends on the average number of amino acids of the proteins. There is a probability less than 1% that a random selection of the amino acids will produce a curve over the "Random <1%" curve.
Figure 2 Comparison between phylogenetic, random and network methods. We compare method 1 (based on closeness centrality) with two phylogenetic methods (method 7 -Hssp database- and method 8 -Conseg server-) and with a random selection of the amino acids. There is a probability less than 1% (resp. 0.5%) that a random selection of the amino acids will produce a curve over the "Random <1%" curve (resp. "Random <0.5%" curve). We notice that if we predict between 10% and 50% of the amino acids, there is a probability less than 0.5% that a random selection of the amino acids will give better results than methods 1 and 7. The calculations are made as an average over the five networks representing HIV-1 protease, TEM1 beta-lactamase, T4 lysozyme, Barnase and the bacteriophage f1 gene V protein.
Figure 3 A new method for essential amino acid prediction. We compare the new method (based on the two sets of amino acids Union and Intersection) with method 1 (based on closeness centrality) and the Conseq server. This method is better than the others: indeed, if we predict fewer than 35% of the amino acids, then the curve (9) is, over most of that range, the highest one; if we predict more than 60% of the amino acids, the curve (10) is, over most of that range, the highest. The calculations are made as an average over the five networks representing HIV-1 protease, TEM1 beta-lactamase, T4 lysozyme, Barnase and the bacteriophage f1 gene V protein.
3. We also compare the specificity values of the different methods when the sensitivity is equal to 50% (see Tables 5 and 6).
Table 5 Specificity value for each method at the sensitivity = 50% level (5 proteins set)
Method Specificity Proportion of predicted residues
Union set (10) 89% 18%
Conseq 85% 22%
Intersection set (9) 85% 23%
Closeness 85% 22%
Connectivity (4) 85% 22%
2-connectivity (3) 83% 28%
Betweenness 80% 27%
1/cluster (5) 80% 27%
Hssp 80% 24%
1/eccentricity 76% 30%
For every method, results are obtained in average over the set of five proteins. Numbers are obtained by calculating the average as in Table 1. For each method, we indicate the value of the specificity and the proportion of residues that are predicted when the sensitivity is equal to 50%.
Table 6 Specificity value for each method at the sensitivity = 50% level (128 proteins set)
Method Specificity Proportion of predicted residues
Union set (10) 87% 15%
Conseq 87% 14%
Intersection set (9) 77% 24%
Closeness 77% 23%
Connectivity (4) 65% 35%
2-connectivity (3) 73% 28%
Betweenness 74% 26%
1/cluster (5) 68% 33%
Hssp 87% 14%
1/eccentricity 68% 32%
Results are obtained in average over the set of 128 proteins. Numbers are obtained by calculating the average as in Table 1. For each method, we indicate the value of the specificity and the proportion of residues that are predicted when the sensitivity is equal to 50%.
We show that the five methods give better results than a random prediction of critical residues, at any given cut-off value selected (in Fig. 1, the curves of Methods 1 to 5 represent the different methods and are always more accurate than the "random selection" that is represented by the line "y = x"). That is, if we randomly select between 10% and 75% of the amino acids in a given protein, there is a probability of less than 1% to have better results than with Methods 1–4 (indeed, the curves of Methods 1 to 4 are above the "Random < 1%" curve for 0,1 < x < 0,75).
Furthermore, if we predict as critical residues fewer than 30% of the protein's amino acids, the best method is based on the global centrality measurement closeness centrality (Method 1). Alternatively, if we predict as critical residues 30% to 55% of the protein's amino acids, the method based on the local centrality measurement 2-connectivity (Method 3) gives the best results (See Fig. 1 and Table 1). Method 1 (based on closeness centrality) always gives better results than Method 2 (based on betweennes (see Fig. 1).
We also tried other network centrality measurements (e.g., eccentricity and 1/eccenticity), However, the results obtained indicated much less accuracy than with these five methods: the method using the eccentricity measurement is worse than random (see Fig. 1) and the method using 1/eccentricity measurement is worse than Methods 1–5 (see Table 1).
Our results improve the reliability of predictions based on the network centrality measurement betweenness described in [7], in which one structure was used to identify critical residues. This is probably due to the fact that we considered several cut-off values. For example, if we want to maximize both the specificity and the sensitivity for the predictions based on betweenness, we need to predict around 40% of a protein's amino acids (Table 1), while in [7] less than 20% of a protein's amino acids were predicted as critical.
Different centrality measurements do not recognize the same set of critical amino acids
The comparison of the level of accuracy in predicting critical residues of the five network centrality-based methods and two phylogenetic approaches reveals that these phylogenetic methods (see Methods) are as reliable as the best network centrality-based methods (Fig. 2). We analyzed whether the sets of critical residues identified by any of these approaches display overlap. If there is little or no overlap, then combining the predictions from these methods may improve the reliability of their individual predictions.
There is no strong "functional relationship" between any two different network-based centrality measurements. In other words, we cannot express precisely one of these properties as a mathematical function of any other properties. For example, plotting the betweenness versus closeness centrality (Fig. 4) shows a cloud of points that are not disposed on a curve. More generally, we do not observe any strong "functional relationship" between any two of these network-based centrality measurements: the connectivity, the cluster coefficient, the eccentricity, the closeness centrality, the betweenness and the 2-connectivity. Hence, the centrality-based methods (Methods 1–5) may not predict the same set of critical amino acids.
Figure 4 Relationship between the betweenness and the closeness centrality for TEM1 beta-lactamase. For each amino acid of TEM1 beta-lactamase, we plot the betweenness versus the closeness centrality. We notice that the points are not approximately disposed along a "natural curve", which suggests that there is no natural "functional relationship" between these two centrality measurements. Therefore, we cannot deduce one centrality measurement from the other one. However, we notice that the cloud of points is located under the curve "y = 0.058x-0.09". This implies that the set of amino acids with the highest betwenness is included in a set of amino acids with highest closeness centrality (for a given cut-off).
Alternatively, looking at a less rigorous relationship among centrality measurements, we did observe some overlap. As shown in Fig. 4, the critical amino acids detected by two network centrality-based measurements present some overlap: if the betweenness of an amino acid is high, then its closeness centrality is typically high, as well. Hence, the set of vertices with high betweenness is completely included in a set of vertices with high closeness centrality. A similar trend is also observed for other centrality measurements, as expected from the fact that these all share some common factors (e.g., shortest path, see Background section).
There is also no strong "functional relationship" between the set of amino acids predicted with any of the five network methods and the set of amino acids predicted with phylogenetic approaches. This also implies that a network method and a phylogenetic method do not predict the same set of residues, although they both present similar reliability in their abilities to predict critical residues. For example, Fig. 5 shows that the score from one of the phylogenetic approaches used (Conseq server, [18]) is not related to the closeness centrality score (data shown only for 1BTL). Therefore, since these two methods predict with the best confidence critical residues and do not predict the same set of amino acids, a combination of the two methods is expected to improve these results (at least to improve sensitivity; whether the overall accuracy will be improved depends on the effect on specificity as well as sensitivity).
Figure 5 Relationship between the closeness centrality and the Conseq value for TEM1 beta-lactamase. For each amino acid of TEM1 beta-lactamase, we plot the closeness centrality versus the Conseq value. We notice that the points are not approximately disposed along a "natural curve", which suggests that there is no strong natural "functional relationship" between these two centrality measurements and that we cannot deduce precisely one centrality measurement from the other one.
An improved method using the Conseq server and the closeness centrality
We propose a new method that improves the accuracy of the predictions achieved by the phylogenetic approaches tested here (see Methods) or the five methods based on network analyses. This method is based on combining critical residues predicted by closeness centrality and by one phylogenetic approach (ConSeq server, [18]). The method is now briefly described.
For a given number k, the set Union is obtained by combining both the k amino acids that have the largest Conseq value and the k amino acids that have the largest closeness centrality value. The set Intersection is obtained by taking the common amino acids between the k amino acids that have the largest Conseq value with the k amino acids that have the largest closeness centrality value.
If we predict less than 35% of a protein's amino acids to be critical for protein function, then the set Intersection gives better results than both phylogenetic approaches and centrality-based methods; if we predict more than 60% of the amino acid residues, the Union set gives the best results (Fig. 3). Furthermore, the sensitivity and the specificity considered simultaneously are better for the Union set than for the other methods (Table 1). If we compare the different methods for a sensitivity value of 50%, we notice that the Union set is also giving the best results (Table 5). Therefore, we have a method that improves both the phylogenetic methods and the centrality-based methods.
Testing the method in a different set of proteins
In order to evaluate our combined method in a different set of proteins, we used a set of 128 proteins (see Methods). These proteins all have their three-dimensional structures solved, and information is available about some but not all of their critical residues (information available from the SITE annotation in the PDB files). This set was used to evaluate the reliability of the predictions of the closeness centrality (see Table 2). We observed as with the previous set of proteins, that for a sensitivity value of 50%, the Union set predicts more true critical residues than phylogenetic approaches (i.e., presents a higher specificity value) (see Table 6).
Table 2 Estimated Sensitivity and Specificity for closeness centrality
Protein(s) Sensitivity Specificity
1BTL 67.4% 72.7%
1HIV 58.7% 83.0%
2LZM 68.3% 78.8%
1A2P 75.0% 89.4%
1GVP 87.5% 64.5%
Average(lBTL, lHIV, 2LZM, lA2P, lGVP) 71.4% 77.7%
Average(128 proteins set) 63.9% 64.1%
The sensitivity and specificity values are reported for each individual protein (1BTL, 1HIV, 2LZM, 1A2P and 1GVP), and the average values obtained for these proteins and the set of 128 proteins. These sensitivity and specificity values correspond to the best cut-off value of the error (i.e. smallest dist(centr. meas., cut-off)) estimated for the closeness centrality. Results in average are obtained by taking the best cut-off value for every protein (and not by taking a fixed cut-off as in Table 1).
To compare closeness centrality against phylogenetic approaches, we noticed that the SITE annotation in this 128 proteins, mostly included residues annotated to be involved in catalysis (218 residues), ligand binding (156 residues) and/or metal-binding (273 residues) sites (see Table 3). Many of the metal binding sites were identified from crystallization conditions, as opposed to be truly involved in any biochemical function. For instance, 77 out of 273 residues involved in metal binding sites participated in catalysis. Our method identified with similar or better reliability than phylogenetic methods active site residues, ligand binding sites and metal binding sites involved in catalysis (see Table 3). Overall, phylogenetic methods predict with better reliability the set of annotated sites for these 128 proteins, mainly because these methods detected better metal binding sites than our method.
Table 3 Type of residues predicted by network and phylogenetic analyses
SITE Closeness Centrality HSSP ConSeq
Metal binding 75 126 134
Metal binding in active site 52 24 38
Ligand binding 57 63 59
Active site 110 104 116
The number of predicted residues that were annotated as SITE in 128 selected proteins is reported for columns Closeness Centrality, HSSP and ConSeq. These last two columns correspond to the results obtained with phylogenetic methods (see Methods section). Four classes of SITE annotations were distinguished and included in column SITE.
Relationship between central amino acids and protein surface-accessible area
Looking for an explanation for why the combination of phylogeny with network centrality renders improved results in the prediction of critical residues, we decided to analyze the nature of the residues detected by these two approaches. It has been previously shown that critical amino acids present special geometrical properties [8]. For instance, critical residues have been proposed to have a tendency to be in low surface-accessibility areas [19]. Since proteins are compact molecules and the methods described here aim to detect residues central to a protein's amino acid interactions, it is expected that central residues may be at the protein's core (i.e., present low-accessibility surfaces). Indeed, TEM1 beta-lactamase, HIV-1 protease, T4 lysozyme, Barnase and bacteriophage f1 gene V protein have their critical residues at a low-surface area (data are shown for TEM1 beta-lactamase in Fig. 6): if we assume that residues above 100 square Angstroms of surface accessibility are exposed, and buried otherwise, all critical residues for this protein are buried (low-accessibility area), as well as all central residues. Furthermore, phylogenetically highly conserved residues also present this trend: conserved residues have low surface accessibility.
Figure 6 Relationship between some network-based centrality measurements and the surface area for TEM1 beta-lactamase. For each amino acid of TEM1 beta-lactamase (1BTL), we plot four network centrality measurements (closeness centrality, betweenness, connectivity and 2-connectivity) versus the surface area. The "best cut-off curve corresponds to the cut-off that minimizes the dist(centr. meas., cut-off) error. In other words, for the four centrality measurements, the best result is obtained by predicting an amino acid to be critical if it is above this line (the corresponding set of predicted amino minimizes the error dist(centr. meas.,cut-off)).
However, it is clear that the surface accessibility measurement alone is not a good predictor of critical residues (see Figs. 6 and 7). Also, it is important to note that the cut-off value chosen to define buried residues is an arbitrary one. In this sense, critical residues are not all "buried" inside a protein; some of them tend to be "exposed" (e.g., catalytic sites, cofactor binding sites). For instance, the critical residues detected in the five proteins by closeness centrality include active site residues that are not core residues (see Table 3). In any case, changing the criterion that defines core residues does not change the conclusions reached here: network centrality measurements are good predictors of the criticality of an amino acid in protein structure and function.
Figure 7 Relationship between the Conseq value and the surface area for TEM1 beta-lactamase. For each amino acid of TEM1 beta-lactamase (1BTL), we plot the Conseq value versus the surface area. The best cut-off for the Conseq value is given by the "best cut-off curve: the set of residues above this curve is the set that minimizes the error dist(conseq value, cut-off).
On the other hand, there is a relationship between the surface accessibility and the centrality measurement for a given residue (data are shown for 1BTL in Fig. 6): the most central residues tend to have a low-accessibility surface. This relationship is not observed for the conserved residues (data only shown for 1BTL in Fig. 7). Hence, in order to explain our results showing centrality measurements as good predictors of critical residues, we propose that these measurements represent more accurately the property that establishes a residue as one critical for protein function: critical residues are central to the residue-residue interactions and these tend to have low-accessibility surfaces.
While revising this manuscript, we became aware that Amitai and colleagues [20] reported that closeness centrality was effective in identifying critical residues from protein structures in a different set of proteins than ours. This may constitute an additional validation that centrality is a new feature of protein structures that is related to the function of residues in a protein. However, we cannot compare directly our results with theirs. The main difference is the procedure used to build the network of contacts in these two studies. Based on this difference, Amitai and colleagues reported a correlation between conserved residues and central ones, a feature we did not observe. Another difference is that the specificity of the method reported by Amitai and colleagues (<10%) is much lower than ours (>70%). Also, the sensitivity of both methods differs (Amitai and col. ~40%, ours >70%). However, Amitai and colleagues found that closeness centrality is related to low surface accessibility, just as we did.
Conclusion
We compared several methods for prediction of critical residues for a given protein function. In the case in which the protein cannot be aligned with a significant number of homologues, we provide five network-based methods that require the proteins' 3D structures but do not require homologous proteins. Although these five methods do not predict the same sets of critical residues, they all give results much better than random.
In the case in which a protein with known 3D structure has enough protein sequence homologues, phylogenetic approaches are as reliable as the five network-based methods. More importantly, there is little overlap in the set of predicted critical residues by any compared method. Hence we propose a new method based on both a network centrality measurement (e.g., closeness centrality) and a phylogeny approach (e.g., Conseq server, [18]) to predict critical residues for protein function based on the 3D structure of proteins and multiple sequence alignments. This hybrid approach improves upon the results of any of the methods compared here.
We observed that there is a trend in plotting centrality measurements and surface accessibility: the most central residues are also those with least likelihood of surface exposure. Such a trend is not as striking for conservation scores and surface accessibility. Since critical residues tend to be more within or near the core (as noted above, central residues identify not only core residues but also active site residues), we propose that the improvement achieved by combining phylogeny with network centrality measurements is due to the complementary nature of these two approaches.
Methods
Set of proteins
We studied three proteins: TEM1 beta-lactamase, HIV-1 protease and T4 lysozyme. We chose this set of proteins because these have been systematically mutated and their 3D structures are available. The Protein Data Bank (PDB) files used for each of these three proteins are respectively 1BTL, 1HIV and 2LZM [[13,14] and [15]]. For this set of proteins, saturation single-point mutation experiments were performed, thus allowing for the identification of every single residue being important for each protein's function. In our study, a critical residue is defined as one important for the function of a protein as a whole (i.e., folding and biochemical function). For these three proteins, the biological assay identified critical residues as those mutants with less than 20% of activity with respect to the wild-type protein [[13,14] and [15]].
Two more proteins were analyzed in this study: Barnase [16] and the bacteriophage f1 gene V protein [17], for in these two cases the proteins have been systematically mutated and their 3D structures are available. The Protein Data Bank (PDB) files used for each of these three proteins are respectively 1A2P and 1GVP. In the case of the 1GVP protein, the biological assays evaluated the activity in both E. coli survival and the bacteriophage f1 propagation ability. So, two different sets of critical residues were identified for these two separate essays. Here we considered as critical residues those that were critical in both assays.
Finally, we used a set of 128 proteins from the PDB database that contained information about critical residues (i.e., SITE annotations on the PDB file) and were defined as representative folds in the FSSP database. The 128 PDB names used in this set are: 1a3c, 1a6q, 1a7j, 1aac, 1ac5, 1ah7, 1ak1, 1ako, 1amj, 1an8, 1apq, 1arv, 1atg, 1auz, 1ayl, 1ayx, 1az9, 1b64, 1bag, 1bdb, 1bea, 1bfd, 1bia, 1bif, 1bix, 1bk0, 1bli, 1bn5, 1bor, 1boy, 1bp1, 1bqk, 1brt, 1btl, 1c25, 1ca1, 1cby, 1cex, 1cfb, 1chc, 1chd, 1csh, 1ctn, 1ctt, 1cvl, 1dmr, 1drw, 1dxy, 1ecl, 1eh2, 1emn, 1esl, 1eut, 1far, 1fnc, 1gca, 1htn, 1hyt, 1iba, 1ido, 1iow, 1iyu, 1kcw, 1kpf, 1lam, 1lay, 1lbu, 1lgr, 1lml, 1lox, 1mfs, 1mla, 1mrp, 1mup, 1nif, 1opc, 1pda, 1pdc, 1pfo, 1phd, 1phm, 1pii, 1pkp, 1poa, 1poc, 1rfs, 1rie, 1rkd, 1rlw, 1skf, 1snc, 1sra, 1thx, 1uch, 1uox, 1ush, 1whi, 1wod, 1xbd, 1xpa, 1ytw, 2abk, 2adr, 2af8, 2cba, 2cmd, 2dkb, 2dri, 2fha, 2fua, 2liv, 2mcm, 2mnr, 2rn2, 2sas, 2vil, 3dfr, 3dni, 3ebx, 3gcb, 3pte, 3ssi, 3tgl, 4enl, 4icb, 4pah, 5eat and 7rsa.
Network associated to a protein
We created one network per protein structure in the following way: we calculated the distance d(a1, a2) between two amino acids a1 and a2 by:
where a1, k denotes all the different positions of the atoms of ak. Then, we connected all pairs with d(a1, a2) < 5oA by an edge.
Calculating global centrality measurements
The closeness centrality, the betweenness and the eccentricity are based on shortest paths. We used the Dijkstra algorithm for tracing the shortest path between two vertices [21]. Hence, every vertex p is assigned a betweenness value obtained by counting the number of times each node is traversed in this process. The closeness centrality ccp and the eccentricity ep are obtained by calculating:
where n is the number of vertices in the network and l(p, q) is the length of a shortest path between p and q.
The hypergeometric distribution
A random selection of amino acids follows a hypergeometric distribution. That is, let N be the total number of amino acids of a given protein and let K be the number of amino acids that are essential for the protein. If we select randomly n amino acids, then the probability P(X = k) of having k amino acids that are essential follows a hypergeometric distribution and satisfies , where are the combinatorial. Therefore, the probability of predicting more than k essential amino acids is . The "Random <c" curves are determined the following way: for every x = K/N, y is the smallest number k/n, such that P(X ≥ k) ≤ c. It is also worth noting that the average value of this distribution is equal to n K/N.
Phylogenetic approaches
Protein sequences from 2LZM, 1BTL, 1HIV, 1GVP, 1A2P and the set of 128 PDB proteins were aligned with their homologues using the Conseq server [18] or using the alignments provided for these three proteins in the Hssp database [22,23].
The parameters used to run the Conseq server for 2LZM, 1BTL and 1HIV were: Maximum likelihood method used to calculate the conservation scores, PSI-BLAST E-value = 0.001, maximum number of homologous sequences = 50 and the number of PSI-BLAST iterations = l (except for 2LZM, for which it was 3). Alignments with 15, 50 and 43 unique sequences were analyzed for 2LZM, 1BTL and 1HIV respectively. For 1GVP and 1A2P, we could not run the Conseq server only by using the parameters (there were too few homologues identified). Therefore, we used the homologues determined by the HSSP alignments (6 homologues for 1GVP and 20 homologues for 1A2P). The conservation score is a number (called color in this website) between 1 and 9. The score 9 refers to the most conserved residues. The conservation score at a site corresponds to the site's evolutionary rate.
Alternatively, the HSSP alignments were used to calculate the percentage of conservation of the amino acids.
Amino acids on the protein's surface/core
The contribution of the surface area of every amino acid to the protein surface area was calculated using the ASC package [24]. The bigger the surface area of an amino acid is, the more exposed it is. If the surface area of an amino acid is 0, it means that this amino acid is completely buried in the core of the protein.
Evaluating the reliability of the predictions
Every method depends on a cut-off value which is the proportion of predicted amino acids. In order to make a more precise study, we decided to analyse the evolution of the set (that contains between 0% and 100% of the amino acids).
We use the classical notions of sensitivity and specificity to evaluate the reliability of the predictions. Sensitivity is defined as the proportion of truly predicted residues (TP) divided by the number of residues experimentally determined (T) to be essential (Sensitivity = TP/T). Specificity is defined by the ratio of non-predicted essential residues (residues experimentally determined not to be essential (F) – false predicted essential residues (FP)) divided by the number of residues experimentally determined not to be essential (F) (Specificity = (F-FP)/F). The best values are obtained when specificity and sensitivity are equal to 1. However, if we predict no amino acids, the specificity is always equal to 1; similarly, if we predict all the amino acids, then the sensitivity is equal to 1. This means one of these two criteria alone is not sufficient to analyse the reliability of a method. That is why we decided to calculate the following criterion for every method/cut-off:
Using this combined criterion, we sorted the different centrality measurements (Table 1).
We also compare the sensitivity to the proportion of amino acids that are predicted. This is relevant, because the sensitivity (i.e. k/K with the previous notations) is larger than the proportion of amino acids that are predicted (i.e. n/N) if and only if the method is better than random (i.e. the number k of truly predicted amino acids is larger than the average value of the distribution n K/N).Therefore, we plot the sensitivity versus the proportion of amino acids that are predicted (see Figs 1, 2 and 3) to analyze/compare all the methods with all the different cut-offs. The higher a curve is, the better its associated method is.
Authors' contributions
BT participated in the conception of the study, carried out the programming, participated in the design of the study and in the drafting of the manuscript. DB participated in the design and coordination of the study and helped to draft the manuscript. GDR initially conceived the study, participated in its design and computer programming, its coordination, and in the drafting of the manuscript. All authors read and approved the final manuscript.
Acknowledgements
We thank Keith Ball (Buck Institute for Age Research) and Diana Escalante Alcalde (National University of Mexico, Instituto de Fisiologia Celular) for their helpful comments on the manuscript. This work was supported by a grant to the Buck Institute from American Bioscience, Inc.
==== Refs
Glaser F Pupko T Paz I Bell RE Bechor D Martz E Ben-Tal N ConSurf: Identification of Functional Regions in Proteins by Surface-Mapping of Phylogenetic Information Bioinformatics 2003 19 163 164 12499312 10.1093/bioinformatics/19.1.163
Del Rio G Garciarrubio A Cusack M Bredesen DE Functional network analyses as an effective scoring system for protein structure prediction
Olmea O Rost B Valencia A Effective use of sequence correlation and conservation in fold recognition J Mol Biol 1999 293 1221 39 10547297 10.1006/jmbi.1999.3208
Oltvai ZN Barabasi AL Systems biology. Life's complexity pyramid Science 2002 298 763 4 12399572 10.1126/science.1078563
Vendruscolo M Dokholyan NV Paci E Karplus M Small-world view of the amino acids that play a key role in protein folding Phys Rev E Stat Nonlin Soft Matter Phys 2002 65 061910 12188762
Dokholyan NV Lewyn L Ding F Shakhnovich EI Topological determinants of protein folding PNAS 2002 99 8637 8641 12084924 10.1073/pnas.122076099
Del Rio G Del Rio H Bartley T Castro-Obregon S Bredesen D Functional assessment of protein structures as biological networks
Elcock AH Prediction of functionally important residues based solely on the computed energetics of protein structure J Mol Biol 2001 312 885 96 11575940 10.1006/jmbi.2001.5009
Watts DJ Strogatz ST Collective dynamics of 'small-world' network Nature 1998 393 440 442 9623998 10.1038/30918
Ma HW Zeng AP The connectivity structure, giant strong component and centrality of metabolic networks Bioinformatics 2003 19 1423 1430 12874056 10.1093/bioinformatics/btg177
Sabidussi G The centrality index of a graph Psychometrica 1966 31 58 603
Hage P Harary F Eccentricity and centrality in networks Social networks 1995 17 57 63 10.1016/0378-8733(94)00248-9
Rennell D Bouvier SE Hardy LW Poteete AR Systematic mutation of bacteriophage T4 lysozyme J Mol Biol 1991 222 67 88 1942069 10.1016/0022-2836(91)90738-R
Huang W Petrosino J Hirsch M Shenkin PS Palzkill T Amino acid sequence determinants of beta-lactamase structure and activity J Mol 1996 258 688 703 10.1006/jmbi.1996.0279
Loeb DD Swanstrom R Everitt L Manchester M Stamper SE Hutchison CA Complete mutagenesis of the HIV-1 protease Nature 1989 340 397 400 2666861 10.1038/340397a0
Axe DD Foster NW Fersht AR A search for single substitutions that eliminate enzymatic function in a bacterial ribonuclease Biochemistry 1998 37 7157 7166 9585527 10.1021/bi9804028
Terwilliger TC Zabin HB Horvath MP Sandberg WS Schlunk PM In vivo characterization of mutants of the bacteriophage f1 gene V protein isolated by saturation mutagenesis J Mol Biol 1994 236 556 571 8107140 10.1006/jmbi.1994.1165
Berezin C Glaser F Rosenberg J Paz I Pupko T Fariselli P Casadio R Ben-Tal N ConSeq: the identification of functionally and structurally important residues in protein sequences Bioinformatics 2004 20 1322 1324 14871869 10.1093/bioinformatics/bth070
Bartlett GJ Porter CT Borkakoti N Thornton JM Analysis of catalytic residues in enzyme active sites J Mol Biol 2002 324 105 121 12421562 10.1016/S0022-2836(02)01036-7
Amitai G Shemesh A Sitbon E Shklar M Netanely D Venger I Pietrokovski S Network analysis of protein structures identifies functional residues J Mol Biol 2004 344 1135 1146 15544817 10.1016/j.jmb.2004.10.055
Del Rio G Bartley TF Del Rio H Rao R Jin KL Greenberg DA Eshoo M Bredesen DE Mining DNA microarray data using a novel approach based on graph theory FEBS Lett 2001 509 230 4 11741594 10.1016/S0014-5793(01)03165-9
Sander C Schneider R Database of homology derived protein structures and the structural meaning of sequence alignment Proteins 1991 9 56 68 2017436 10.1002/prot.340090107
Hssp database
The Analytic Surface Calculation package
|
16124876
|
PMC1208857
|
CC BY
|
2021-01-04 16:27:26
|
no
|
BMC Bioinformatics. 2005 Aug 26; 6:213
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-213
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2161613524810.1186/1471-2105-6-216Methodology ArticleJACOP: A simple and robust method for the automated classification of protein sequences with modular architecture Sperisen Peter [email protected] Marco [email protected] Swiss Institute of Bioinformatics, Computational Cancer Genomics Group – ISREC, Ch. des Boveresses 155, 1066 Epalinges, Switzerland2 Swiss Institute of Bioinformatics, Vital IT Group, BEP-UNIL, 1015 Lausanne, Switzerland2005 31 8 2005 6 216 216 2 2 2005 31 8 2005 Copyright © 2005 Sperisen and Pagni; licensee BioMed Central Ltd.2005Sperisen and Pagni; 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
Whole-genome sequencing projects are rapidly producing an enormous number of new sequences. Consequently almost every family of proteins now contains hundreds of members. It has thus become necessary to develop tools, which classify protein sequences automatically and also quickly and reliably. The difficulty of this task is intimately linked to the mechanism by which protein sequences diverge, i.e. by simultaneous residue substitutions, insertions and/or deletions and whole domain reorganisations (duplications/swapping/fusion).
Results
Here we present a novel approach, which is based on random sampling of sub-sequences (probes) out of a set of input sequences. The probes are compared to the input sequences, after a normalisation step; the results are used to partition the input sequences into homogeneous groups of proteins. In addition, this method provides information on diagnostic parts of the proteins. The performance of this method is challenged by two data sets. The first one contains the sequences of prokaryotic lyases that could be arranged as a multiple sequence alignment. The second one contains all proteins from Swiss-Prot Release 36 with at least one Src homology 2 (SH2) domain – a classical example for proteins with modular architecture.
Conclusion
The outcome of our method is robust, highly reproducible as shown using bootstrap and resampling validation procedures. The results are essentially coherent with the biology. This method depends solely on well-established publicly available software and algorithms.
==== Body
Background
Whole-genome sequencing projects are currently producing an enormous amount of new sequences. As a consequence, protein sequence databases are rapidly increasing in size, thus resulting in severe practical consequences. For example, a simple database search can now yield hundreds of matches. An automated but sensible grouping of those proteins appears as an indispensable solution to analyse such an output in a timely manner.
Proteins are often described as the assembly of several structural/functional units called domains. Isolation of the domain sub-sequences renders a multiple alignment possible, from which domain descriptors are built based on efficient methods for remote homology detection (PSSM [1], generalised profiles [2], hidden Markov models (HMM) [3]). This led to the thriving development of protein-domain databases such as PROSITE [4], Pfam [5], Blocks [6], PRINTS [7], IDENTIFY [8], ProDom [9], Domo [10], SMART [11] and ADDA [12]. Classification of domain sub-sequences is relatively straightforward through direct sequence comparison but does not address the problem of whole multi-domain protein classification.
As a possible alternative to direct sequence comparison, proteins could be classified according to their domain architecture [13,14]. Although no general methodology has yet emerged, many review articles on particular families are available [15-17]. The highly modular proteins that are involved in signalling pathways are a typical example where domain architecture usually is diagnostic for the protein function (such an example is treated below). Unfortunately, in the course of evolution, the linear and modular organisation of proteins is not always preserved because of rare genetic events that are responsible for domain swapping, duplication or deletion. One must realise that the number of these particular cases increases with the number of genome sequences [18-20] and renders automated classification difficult.
Finally, the emerging picture of the human proteome provides evidence that alternative splicing is not anecdotal [21]. Indeed, this mechanism is a potential source of sequence variation and the proper handling of splice variants by clustering protocols is a challenge.
Several recent contributions addressed the clustering of very large sets of possibly unrelated protein sequences (e.g. TRIBE-MCL [22], Picasso [23], Systers [24], COG [25], ProtoMap [26], ProtoNet [27], ClusTr [28] and ProClass [29]). All these approaches are based on pairwise or multiple alignments of the sequences to be analysed. Alternatively alignment-free sequence comparison methods were proposed ([30] and references therein). However, they have not yet been widely used for the clustering of very large sets.
Here we present a novel method called JACOP (just another classification of proteins), which stands somewhere between pairwise alignment methods and alignment-free methods. We will employ a collection of unordered short sub-sequences as an intermediary layer in the comparison of two sequences. As a consequence, the linearity of the domain architecture – present in the investigated protein set – no longer plays an important role. In comparison to existing methods, our protocol is remarkably simple but nevertheless appears to be robust and highly reproducible on difficult test sets.
Results and discussion
Protocol
The different steps to partition a set of unaligned protein sequences using the JACOP protocol are presented below and summarised in Figure 1. The parameters used for the reference protocol are those established after intense testing leading to the most consistent results. The rationale behind the selection of these parameters is presented further down. The only prerequisite for a given sequence is a length of at least 50 residues. Otherwise no a priori knowledge about the protein sequences is needed.
Figure 1 Schematic representation of the reference protocol. The numbers correspond to the steps explained in the main text.
1. Random probe generation
p subsequences of 50 residues length (probes) are randomly sampled from the set of N input sequences. All probes that can possibly be generated out of the input sequences are equally likely. The sampling process is carried out until the cumulated length of the probes exceeds three times the cumulated length of the input sequences.
2. Pruning probes
An all-versus-all comparison of the probes is performed using the SW algorithm with a Blosum62 [31] similarity matrix and gap opening/extension penalties set to -12/-1. The list of probes is then pruned as follows. The first probe is kept. The second probe is compared with the first and kept only if its SW score is below a threshold of 160. The subsequent probes are successively compared with the list of already kept probes and added to this list if their SW scores are below 160, thus leaving a list of m probes.
3. Scoring input sequences with probes
A comparison of all probes of the pruned collection versus all input sequences is performed using the SW algorithm with a Blosum62 scoring matrix and gap opening/extension penalties set to -12/-1. The SW scores S are normalised to obtain bitscores Sbit [32] using:
The parameter values for the above used scoring system are λ = 0.312 and K = 0.076. These parameters were obtained by simulation with random sequences (see Materials and Methods). The bitscores are arranged in a matrix of dimensions Nxm.
4. Protein distances
The above matrix is first transformed to a binary matrix based on a threshold of 27 (equivalent to a p-value of 0.01 [33,34]). All bitscores below this threshold are set to 0 and those above, to 1 (match). Since the binary variables are asymmetric, a distance matrix between any pair of input sequences represented by the rows of the binary matrix is calculated using the Jaccard distance [35]:
where n11 is the number of probes which match both sequences, and n10 and n01 are the number of probes, which only match one of the two sequences. n00 is excluded from the calculation of the distance (see below). This distance measure ranges from 0 (closely related sequences with all probes in common) to 1 (sequences with no probe in common).
5. Identification of independent groups
One can identify groups that are separated by a distance of 1 (no match in common). These groups are called independent groups. The set of N input sequences is split into n independent groups containing li proteins (1≤i≤n).
6. Partitioning of proteins within an independent group
This is done using the PAM (Partitioning Around Medoids) algorithm [36] which, based on the above distance matrix, calculates all possible partitions ranging from 2 to li-1 subgroups. For each partition the overall average silhouette width [37] is calculated and the partition that maximises it is considered optimal.
7. Hierarchical clustering
The protein sequences are clustered using average-linkage agglomerative hierarchical clustering based on the Jaccard distances. The resulting tree is particularly useful to establish the relations between subgroups.
8. Identification of diagnostic probes
For each probe the number of subgroups, to which it matches, is determined. As a result one can distinguish probes that only match one group, probes that match all groups (conserved regions found in all members of an independent group) and those in between.
Case 1: Prokaryotic lyases
A set of protein sequences that can legitimately be arranged into a multiple sequence alignment is first considered to facilitate the comparison of the outcome of the JACOP protocol with other classification methods. All prokaryotic sequences retrieved with the Pfam (Version 7.3) HMM Lyase_1 (PF00206) in Swiss-Prot release 36 were retained as a test set. These are enzymes involved in double bond isomerisation, and catalyse five different reactions. A short sequence flanking a conserved methionine, described by the PROSITE pattern PS00163, is also present in all of these sequences.
Applying the JACOP protocol to this set of 53 sequences with an average length of 463 residues, resulted in the generation of 1473 probes to obtain three-time coverage. Only 154 probes were left after pruning, owing to the relatively high similarity among these sequences. The sequences were hierarchically clustered and the resulting tree is shown in Figure 2b. Two independent groups were obtained. The two class I fumarate hydratases are separated from all other sequences by a distance of one, i.e. the two groups do not share any probe and this despite the matches determined with the Pfam HMM and the PROSITE pattern. It turned out that these class I fumarate hydratases are biochemically different [38] (contain a [4Fe-4S] cluster, form homodimers and work under aerobic conditions) from the class II fumarate hydratases found among the remaining 51 sequences that form the second independent group (work under anaerobic conditions, do not contain an iron-sulfur cluster and form homotetramers). This latter group is partitioned into four sub-groups.
Figure 2 This figure shows the trees obtained either with ClustalW/PHYLIP (a) or with the JACOP protocol (b). The families of enzymes with different activities are presented in different colours. The resulting independent groups and subgroups found by JACOP are indicated by frames. In the case of the tree obtained with the JACOP protocol, the resampling/bootstrap values above 95% are indicated. The separation (*) between the three subgroups of homologues received a relatively low value because of comparable distances that induce a competition of sub-tree topology at that node.
1. The argininosuccinate lyases form a dense cluster with a distinctive enzymatic activity.
2. The adenylosuccinate lyases (PUR8_*) with the exception of PUR8_ECOLI/BUCAI/HAEIN, which appear to be a separate sub-group, and the 3-carboxy-cis, cis-muconate cycloisomerase (PCAB_*) are clustered together.
3. The adenylosuccinate lyases PUR8_ECOLI/BUCAI/HAEIN.
4. The class II fumarate hydratases and aspartate ammonia-lyases are clustered together.
Two sets of probes match to conserved regions of the second independent group (Figure 3). The first region corresponds to the active site according to Swiss-Prot annotation. The second region is the one identified with the Pfam HMM Lyase_1 (PF00206). In addition to these probes that identify conserved regions, other probes that are specific for the different subgroups were found.
Figure 3 Representative proteins of the prokaryotic lyases are shown together with matching probes at their respective positions. The probes mapping to the representative of the first independent group (FUMA_ECOLI) are represented as closed dark grey bars. Probes that map to the active sites of proteins of the second independent group are shown as closed red bars, and those that map the region identified by the Pfam HMM Lyase_1 (PF00206) as closed green bars. Probes that map specifically to the different subgroups are depicted as open bars using the following colours: blue, orange, magenta and yellow. Probes that are not specific for any particular subgroup are depicted as grey open bars. Features were obtained from the original Swiss-Prot annotation.
The JACOP results were compared to the classification obtained using other approaches. Thus the sequences were aligned using ClustalW with default settings, and a tree was derived from this alignment using the PROTDIST and FITCH programs of the PHYLIP package [39] (Figure 2a). The trees are comparable, with differences implicating the problematic PUR8_ECOLI/BUCAI/HAEIN. However, the ClustalW/PHYLIP approach cannot provide any indication that the class I fumarate hydratases are unrelated to all other sequences. A bootstrap analysis on the multiple sequence alignment was also performed [40-42]. The results were rather confusing. Indeed a classical bootstrap strategy is designed to handle multiple sequence alignments of related sequences and is therefore unsuited to deal with such divergent sequences. As an alternative, the sequences were aligned using T-Coffee [43] with default parameters and the tree was established as before. In contrast to ClustalW, this tree confirmed the result found with JACOP (details not shown).
Furthermore, the classification obtained with JACOP was compared with other publicly available large-scale protein clusters. The COG (release 3) classification [25] was fully consistent with the enzymatic nomenclature and correctly separated the fumarate hydratases into two families. More surprisingly, it distinguished the 3-carboxy-cis, cis-muconate cycloisomerases from all the adenylosuccinate lyases. Unfortunately, details leading to this distinction were not available at the time of writing.
In the case of SYSTERS (release 4) [24], ProtoNet (version 4.0) [27] as well as ClusTr [28], the same 5 clusters were found.
Parameter selection for JACOP
The rationale for our choices was the following:
i. Rigorous and well-described methods (SW algorithm, PAM algorithm) were preferred over faster but intricate heuristics for the sake of reproducibility by others. Nevertheless we tested BLAST (version 2.2.5) to compute the scores. However, this heuristic often failed to provide the correct SW scores, possibly due to the short length of the probes (details not shown).
ii. The main reason for choosing the PAM algorithm for partitioning the proteins was that this method, which is based on the minimization of the sum of dissimilarities, is more robust than methods that minimize the error sum of squares like k-means [36]. As an alternative to the PAM algorithm we have tested the fast TRIBE-MCL [22] algorithm but it failed to identify the relevant groups.
iii. Silhouette widths allow a good characterisation of all clusters that are not too elongated and make it possible to identify outliers in most situations. Another advantage of silhouette widths is their independence of the used partitioning algorithm. Silhouette widths s(i) [37] are calculated for each object i and range from -1 to +1. Values of s(i) close to one, indicate that the average dissimilarity of i to the other objects of the same cluster is much smaller than the smallest average dissimilarity to other clusters. If the value s(i) is about zero, then the two dissimilarities are approximately equal and hence it is not clear to which cluster the object i should be assigned. The worst situation takes place, when s(i) is negative indicating that object i has been misclassified. Furthermore, the overall average silhouette width over all objects can be used to objectively identify the most consistent partitioning for which it is the largest. The result of the partitioning is a list of protein sequences with the number of the subgroup to which they belong. Additionally, for each protein sequence, the number of the closest alternative subgroup and the corresponding silhouette width s(i) is given.
iv. Due to the random generation of the probes, one has to sample a sufficient number of them to "cover" the complete sequence. Sampling was stopped once the cumulated length of the probes exceeded three times the cumulated length of the input sequences (coverage 3x) because, in average, higher numbers do not further change the final number of probes after pruning.
v. The main reason for choosing the Jaccard distance measure was that it did not take into account non-significant matches shared by two proteins. As a consequence, proteins with no similarity other than noise are not grouped together.
The choice of the following parameters was based on two different validation procedures. The first one consists of re-sampling, i.e. the whole protocol is repeated 100 times from probe sampling to the partitioning of the independent groups, each time with a different seed for the random number generator. The second test is a classical bootstrap on the pruned probes, i.e. the first 3 steps of the protocol are run once and the resulting bitscore matrix is bootstrapped 100 times followed by the partitioning of the independent groups. The reproducibility of the 100 obtained partitions was evaluated based on the adjusted rand index [44]. It is a statistic designed to assess the degree of agreement between two partitions. An adjusted Rand index of 1 indicates identical partitions, whereas an adjusted rand index close to zero indicates random partitioning. After the simulations, each of the 100 partitions was compared to the other 99 partitions and the average was taken. The average adjusted rand indexes are given in Table 1 for the different parameter sets tested.
Table 1 This table summarizes the outcome of 100 resampling or bootstrapping tests done using the two data sets (prokaryotic lyases and SH2 containing proteins) with different parameter combinations. For each simulation the average adjusted rand index [44] has been calculated. The row marked in italic bold corresponds to the parameters used in the reference protocol. The values marked in bold indicate the changes made compared to the reference protocol.
Prokaryotic lyases SH2 containing proteins
Probe length Pruning threshold Coverage Scoring system Resampling Bootstrap Resampling Bootstrap
25 160 3 Blosum62/-12/-1 0.959 0.991 0.844 0.854
50 160 3 Blosum62/-12/-1 1 1 0.932 0.937
100 160 3 Blosum62/-12/-1 0.969 0.887 0.689 0.735
50 40 3 Blosum62/-12/-1 0.971 0.810 0.777 0.732
50 80 3 Blosum62/-12/-1 1 0.982 0.753 0.803
50 120 3 Blosum62/-12/-1 1 1 0.895 0.897
50 160 3 Blosum62/-12/-1 1 1 0.932 0.937
50 200 3 Blosum62/-12/-1 1 1 0.948 0.935
50 160 1 Blosum62/-12/-1 1 0.991 0.919 0.812
50 160 2 Blosum62/-12/-1 1 0.996 0.878 0.937
50 160 3 Blosum62/-12/-1 1 1 0.932 0.937
50 160 3 Blosum45/-13/-3 1 0.985 0.939 0.839
50 160 3 Blosum62/-12/-1 1 1 0.932 0.937
50 160 3 Blosum80/-10/-1 0.999 0.991 0.909 0.859
vi. Sampling with three times coverage results in a lot of redundancy. Eliminating probes that are too similar significantly reduces this redundancy, while keeping the information. Several threshold scores (raw scores of 40, 80, 120, 160 and 200) for the elimination of similar probes were tested for the pruning step. At a threshold of 40, most probes were eliminated and consequently most of the information. The resulting set of probes did not allow the robust and reproducible identification of subgroups. On the other hand, starting from threshold 160, the partitions were reproducible. In all cases the gain in resolution and information was marginal when the threshold was increased to 200. In addition, pruning has the desirable side effect that it preferentially removes probes with low complexity regions [45,46] – which are known to be a nuisance in sequence comparison.
vii. Because of the use of a local alignment algorithm, the average length of a match was shorter than the length of the probes (average length of about 19 for the above set of sequences). This implies that relatively long probes could document short conserved motifs. The use of probes of length 25 or 100 resulted in a substantially decreased reproducibility.
viii. The similarity matrices and gap penalties for the SW algorithm parameters were selected to ensure that the scoring system produces true local alignments (logarithmic phase [47]; this allowed the use of well established statistics [47-53].In addition to the scoring system Blosum62/-12/-1, the Blosum45/-13/-3 and the Blosum80/-10/-1 scoring systems were also evaluated with the appropriate values for λ and K. Surprisingly this only had a marginal effect on the outcome.
ix. Hierarchical clustering was also applied to the data sets generated during the validation procedure. The resulting 100 trees were compared to determine the stability of the nodes. In the case of prokaryotic lyases, counts larger than 95 are reported on the tree of Figure 2b. The outcomes of the two validation procedures agree to a large extent. Interestingly, the separation between the three major sub-groups of homologues received a relatively low value because comparable distances separated them. This results in a competition of sub-tree topology at that node. This incidentally indicates the superiority of the PAM algorithm over hierarchical clustering for our purpose.
Case 2: SH2 containing proteins
Here a set of protein sequences that cannot be arranged as a meaningful multiple sequence alignment is considered. This set contains all proteins from Swiss-Prot (release 40) with at least one Src homology 2 (SH2) domain as predicted by the Pfam HMM PF00017. The 203 proteins of this set belong to the super-family of intracellular signal-transducing proteins and represent a case study of modular architecture [17]. Indeed, together with one or two SH2 domains, many other domains were found. Amongst them are RhoGAP, RhoGEF, protein-tyrosine kinases (PTK), protein-tyrosine phosphatase or phosphatidylinositol specific phospholipase C X or Y domains, as well as the promiscuous SH3 domain [22]. In addition, 5% of all residues were found to be part of low complexity regions by the SEG program [45].
The JACOP reference protocol was applied to this set of 203 sequences of an average length of 628 residues. 7647 probes were extracted to obtain three-time coverage and 1799 probes remained after pruning. JACOP identified one single independent group containing 75 subgroups, which correlate very well with the Swiss-Prot IDs. Also the domain architectures – as identified by the Pfam HMM – correlate well with the subgroups identified in that every subgroup is reflected by one single domain architecture. However, different subgroups may share the same architecture. At this stage the result of the hierarchical clustering becomes helpful to delineate relationships among subgroups. That way one can distinguish three larger superfamilies (together with two singletons, three pairs and one quadruplet) (not shown).
• 24 signal transduction and activators of transcription (STAT) proteins are present in the first superfamily. They are subdivided into 7 subgroups. However, all 24 proteins share the same function and domain architecture together with a single SH2 domain (details not shown).
• 98 proteins containing a protein-tyrosine kinase (PTK) motif (PF00069) make up the second superfamily and are subdivided into 36 subgroups. The number of different architectures was limited, the two most frequent being SH3-SH2-PTK and SH2-PTK-PTK (details not shown).
• The third superfamily, containing 69 proteins, is the most complex one as the domain architectures found are extremely diverse. 15 subgroups are identified and represent 14 different architectures (Figure 4). The domain architectures – as identified by the Pfam HMM – correlate well with the subgroups identified. This superfamily is functionally diverse and contains enzymes, adaptor proteins, docking proteins and regulatory proteins [17]. Two sets of proteins deserve further discussion:
Figure 4 Classification of 69 SH2-containing protein sequences that represent various domain architectures. a: domain architecture as predicted by the Pfam HMMs; b: partitioning obtained by JACOP; c: Swiss-Prot IDs; d: the bitscore matrix with a colour code based on heat colours ranging from white for high values to red for low values, plus black for all bitscores that are below 27; e: the tree obtained by hierarchical clustering of the proteins based on Jaccard distances.
1. The phosphatidylinositol 3-kinase regulatory alpha (P85A_*), beta P85B_*) or gamma (P55G_*) subunits share two SH2 domains at the C-terminus [54]. In addition, the P85 subunits contain an SH3 and a RhoGAP domain at the N-terminus. Despite the different architecture, these sequences were clustered together due to the presence of highly conserved tandem SH2 domains (sequence identity greater than 70%), which were found to be quite distinct from tandem SH2 domains found in other subgroups.
2. The SH2/SH3 containing adaptors (NCK*_*, DRK_DROME, GRAP_HUMAN, GRB2_*, SEM5_CAEEL, GRP2_*, GAGC_AVIS and CRK*_*) are subdivided into 4 subgroups corresponding to different arrangements of the SH2 and SH3 domains. Their SH2 or SH3 domain sequences are more similar to each other than to the SH2/SH3 domain sequences found in other subgroups (details not shown). This strongly suggests, that these proteins were subject to a recent reshuffling event. Interestingly, the common denominator of these proteins is their role in regulating tyrosine kinase signalling. They serve to recruit proline-rich effector molecules to tyrosine-phosphorylated kinases or their substrates [55] and references therein).
As for the prokaryotic lyases, the probes were analysed. The grouping of the probes can be seen in Figure 4d. Several groups of probes that are specific for particular subgroups can easily be distinguished from probes that are of more general nature and map to regions that are conserved amongst all proteins of this superfamily. However, probes that are specific for one subgroup and that appear as a block in Figure 4d do not necessarily map to adjacent regions but can be distant from one another. The only common denominator of such probes is the fact that they match the same proteins.
The results for the third superfamily were compared to SYSTERS, ProtoMap, ProtoNet as well as ClusTr. The classifications of those approaches were found to be very similar to the partitioning obtained by JACOP.
Conclusion
The key point of the JACOP protocol is the random sampling of relatively short sub-sequences (probes) out of the sequences to be analysed. After a normalisation step, the probes are compared with the initial set of proteins and the resulting scores are used to classify the proteins based on a p-value of 0.01. The method produces meaningful and robust partitions of proteins with related functions out of a set of input sequences, even when the sequences cannot be arranged in a multiple sequence alignment due to their modular architecture and despite the method's stochastic nature. It also allows the identification of regions conserved amongst all sequences of an independent group or, alternatively, regions that are specific (diagnostic) to certain subgroups.
In our opinion one of the reasons for the robustness of JACOP is the use of the complete information present in the pool of pruned probes. In contrast to other methods that are based on direct pairwise comparisons, JACOP also uses information on protein sequence similarities outside the protein's own group. Hence a distance between two proteins is based on how similar the two proteins are with respect to some features present in the pool represented by the pruned probes and how dissimilar they are to other features.
It would seem appealing to define probes using the natural boundaries of the protein domains, in an attempt to describe the protein sequences by a systematic tiling with probes, in the same spirit as ProDom [9]. Unfortunately, there is no reliable algorithm for detecting domain boundaries. Also, whether they can be defined unambiguously is still an open question. However, it is obvious that a false definition of the boundaries of a domain has direct consequences on the definition of the boundaries of the adjacent domains if probes do not overlap. In this perspective, random sampling of potentially overlapping probes – in contrast to a systematic tiling of probes – appears to be a simple way to produce a set of sub-sequences with unbiased positions and boundaries. In addition, bootstrapping and/or re-sampling may be performed to estimate the stabilities of the resulting partitions.
Residue substitutions are certainly, and by far, more frequent in the course of evolution than insertions or deletions, which themselves are far more frequent than domain architecture re-organization. However, these different types of events co-occur and some domain re-organizations may be expected to have occurred more recently than many residue substitutions. This usually causes major problems in most methods that use direct pairwise alignments of two sequences to measure their similarity. The introduction of a collection of unordered probes as an intermediary layer in the comparison of two sequences elegantly solves the problem. Hence, when comparing a pair of sequences that exemplify a case of domain swapping, the sequences are locally co-linear through the probes, everywhere but in the swap region itself. This strategy results in a partial uncoupling of the domain architecture present in the proteins.
A Jaccard distance of 0.5 between two sequences (eq. 2) can actually correspond to different cases. The sequences can be globally homologous but sufficiently divergent to share only one half of the probes. Alternatively, one of the sequences can be a perfectly conserved fragment of about half the length of the other sequence. Hence, the JACOP method includes an implicit weighing scheme that relates the similarity measure at the sequence level to the architecture similarity. A better understanding and control of the implicit weighing scheme is the subject of future work.
When comparing the results of JACOP with other publicly available automated classifications, our results closely resembled the ones proposed by reference classifications such as Systers and ProtoMap. However – though simple – JACOP is a robust, efficient and reproducible approach for the classification of protein sequences. Also, JACOP can easily be applied since it only requires software (Perl [56], ssearch [57] and R) and algorithms (SW, PAM) available to every one.
Materials and methods
Implementation
The first 3 steps of the JACOP protocol, presented below, were performed using scripts written in Perl 5 [56]. Calculation of the raw Smith-Waterman (SW) scores [58] was either done using the program ssearch [57] or hardware accelerated using a GENEMATCHER (Paracel, Pasadena, CA). All statistical calculations were done using the statistics software R [59]. A package for R was specifically written for JACOP and the source code is available upon request.
Statistical parameters
Statistics for the scores of local alignments, unlike those of global alignments, are well understood [33,34,47-52,60-62]. The parameters for the underlying extreme value distribution (EVD) for the three scoring systems (Blosum62/-12/-1, Blosum45/-13/-3 and Blosum80/-10/-1) were estimated using random protein sequences of appropriate lengths. The SW scores obtained were subsequently used to estimate the EVD parameters by maximum likelihood [63].
Availability and requirements
Project name: JACOP; Project home page: Operating system(s):Platform independent.
Authors' contributions
PS developed and implemented the JACOP approach. MP conceived the method and provided guidance.
Acknowledgements
PS is supported by a grant from the Swiss Federal Office for Education and Health (OFES): 01.0101. MP gratefully acknowledges GlaxoSmithKline Research and Development, Geneva for financial support. We thank Drs Philipp Bucher, Kay Hofmann, Monique Zahn and Vivienne Baillie Gerritsen for carefully reading the manuscript and helpful discussion. We thank Dr Nadine Zangger for valuable input to the analysis of the simulations. We also thank the Swiss-Prot team for their very encouraging feedback.
==== Refs
Gribskov M McLachlan AD Eisenberg D Profile analysis: detection of distantly related proteins Proc Natl Acad Sci U S A 1987 84 4355 4358 3474607
Bucher P Karplus K Moeri N Hofmann K A flexible motif search technique based on generalized profiles Comput Chem 1996 20 3 23 8867839 10.1016/S0097-8485(96)80003-9
Krogh A Brown M Mian IS Sjolander K Haussler D Hidden Markov models in computational biology. Applications to protein modeling J Mol Biol 1994 235 1501 1531 8107089 10.1006/jmbi.1994.1104
Hulo N Sigrist CJ Le Saux V Langendijk-Genevaux PS Bordoli L Gattiker A De Castro E Bucher P Bairoch A Recent improvements to the PROSITE database Nucleic Acids Res 2004 32 D134 7 14681377 10.1093/nar/gkh044
Bateman A Coin L Durbin R Finn RD Hollich V Griffiths-Jones S Khanna A Marshall M Moxon S Sonnhammer EL Studholme DJ Yeats C Eddy SR The Pfam protein families database Nucleic Acids Res 2004 32 D138 141 14681378 10.1093/nar/gkh121
Henikoff JG Henikoff S Pietrokovski S New features of the Blocks Database servers Nucleic Acids Res 1999 27 226 228 9847186 10.1093/nar/27.1.226
Attwood TK The PRINTS database: a resource for identification of protein families Brief Bioinform 2002 3 252 263 12230034
Nevill-Manning CG Wu TD Brutlag DL Highly specific protein sequence motifs for genome analysis Proc Natl Acad Sci USA 1998 95 5865 5871 9600885 10.1073/pnas.95.11.5865
Bru C Courcelle E Carrere S Beausse Y Dalmar S Kahn D The ProDom database of protein domain families: more emphasis on 3D Nucleic Acids Res 2005 33 D212 205 15608179 10.1093/nar/gki034
Gracy J Argos P DOMO: a new database of aligned protein domains Trends Biochem Sci 1998 23 495 497 9868374 10.1016/S0968-0004(98)01294-8
Letunic I Copley RR Schmidt S Ciccarelli FD Doerks T Schultz J Ponting CP Bork P SMART 4.0: towards genomic data integration Nucleic Acids Res 2004 32 D142 4 14681379 10.1093/nar/gkh088
Heger A Wilton CA Sivakumar A Holm L ADDA: a domain database with global coverage of the protein universe Nucleic Acids Res 2005 33 D188 91 15608174 10.1093/nar/gki096
Junier T Pagni M Bucher P mmsearch: a motif arrangement language and search program Bioinformatics 2001 17 1234 1235 11751236 10.1093/bioinformatics/17.12.1234
Ponting CP Schultz J Milpetz F Bork P SMART: identification and annotation of domains from signalling and extracellular protein sequences Nucleic Acids Res 1999 27 229 232 9847187 10.1093/nar/27.1.229
Aravind L Dixit VM Koonin EV The domains of death: evolution of the apoptosis machinery Trends Biochem Sci 1999 24 47 53 10098397 10.1016/S0968-0004(98)01341-3
Aravind L Walker DR Koonin EV Conserved domains in DNA repair proteins and evolution of repair systems Nucleic Acids Res 1999 27 1223 1242 9973609 10.1093/nar/27.5.1223
Pawson T Gish GD Nash P SH2 domains, interaction modules and cellular wiring Trends Cell Biol 2001 11 504 511 11719057 10.1016/S0962-8924(01)02154-7
Holland IB Blight MA ABC-ATPases, adaptable energy generators fuelling transmembrane movement of a variety of molecules in organisms from bacteria to humans J Mol Biol 1999 293 381 399 10529352 10.1006/jmbi.1999.2993
Quentin Y Fichant G ABCdb: an ABC transporter database J Mol Microbiol Biotechnol 2000 2 501 504 11075924
Higgins CF ABC transporters: from microorganisms to man Annu Rev Cell Biol 1992 8 67 113 1282354 10.1146/annurev.cb.08.110192.000435
Modrek B Resch A Grasso C Lee C Genome-wide detection of alternative splicing in expressed sequences of human genes Nucleic Acids Res 2001 29 2850 2859 11433032 10.1093/nar/29.13.2850
Enright AJ Van Dongen S Ouzounis CA An efficient algorithm for large-scale detection of protein families Nucleic Acids Res 2002 30 1575 1584 11917018 10.1093/nar/30.7.1575
Heger A Holm L Picasso: generating a covering set of protein family profiles Bioinformatics 2001 17 272 279 11294792 10.1093/bioinformatics/17.3.272
Meinel T Krause A Luz H Vingron M Staub E The SYSTERS Protein Family Database in 2005 Nucleic Acids Res 2005 33 D226 9 15608183 10.1093/nar/gki030
Tatusov RL Fedorova ND Jackson JD Jacobs AR Kiryutin B Koonin EV Krylov DM Mazumder R Mekhedov SL Nikolskaya AN Rao BS Smirnov S Sverdlov AV Vasudevan S Wolf YI Yin JJ Natale DA The COG database: an updated version includes eukaryotes BMC Bioinformatics 2003 4 41 12969510 10.1186/1471-2105-4-41
Yona G Linial N Linial M ProtoMap: automatic classification of protein sequences and hierarchy of protein families Nucleic Acids Res 2000 28 49 55 10592179 10.1093/nar/28.1.49
Kaplan N Sasson O Inbar U Friedlich M Fromer M Fleischer H Portugaly E Linial N Linial M ProtoNet 4.0: a hierarchical classification of one million protein sequences Nucleic Acids Res 2005 33 D216 218 15608180 10.1093/nar/gki007
Kriventseva EV Servant F Apweiler R Improvements to CluSTr: the database of SWISS-PROT+TrEMBL protein clusters Nucleic Acids Res 2003 31 388 389 12520029 10.1093/nar/gkg035
Huang H Xiao C Wu CH ProClass protein family database Nucleic Acids Res 2000 28 273 276 10592245 10.1093/nar/28.1.273
Vinga S Almeida J Alignment-free sequence comparison-a review Bioinformatics 2003 19 513 523 12611807 10.1093/bioinformatics/btg005
Henikoff S Henikoff JG Amino acid substitution matrices from protein blocks Proc Natl Acad Sci USA 1992 89 10915 10919 1438297
Altschul SF The Statistics of Sequence Similarity Scores.
Karlin S Altschul SF Methods for assessing the statistical significance of molecular sequence features by using general scoring schemes Proc Natl Acad Sci USA 1990 87 2264 2268 2315319
Dembo A Karlin S Zeitouni O Limit distribution of maximal non-aligned two-sequence segmental score Ann Prob 1994 22 2022 2039
Jaccard P Nouvelles recherches sur la distribution florale. Bull Soc Vaudoise Sci Nat 1908 44 223 227
Kaufman L Rousseeuw PJ Finding groups in data: An introduction to cluster analysis 1990 New York , John Wiley and Sons
Rousseeuw PJ Silhouettes: a graphical aid to the interpretation and validation of cluster analysis J Comput App Mat 1987 20 53 65 10.1016/0377-0427(87)90125-7
Woods SA Schwartzbach SD Guest JR Two biochemically distinct classes of fumarase in Escherichia coli Biochim Biophys Acta 1988 954 14 26 3282546
Felsenstein J PHYLIP -- Phylogeny Inference Package (Version 3.2) Cladistics 1989 5 164 166
Efron B The Jackknife, the Bootstrap, and Other Resampling Plans CBMS-NSF Regional Conference Series in Applied Mathematics 1982 Philadelphia, Pennsylvania , Soc. Ind. Appl. Math.
Felsenstein J Confidence limits on phylogenies: An approach using the bootstrap Evolution 1985 39 783 791
Efron B Halloran E Holmes S Bootstrap confidence levels for phylogenetic trees Proc Natl Acad Sci U S A 1996 93 7085 7090 8692949 10.1073/pnas.93.14.7085
Notredame C Higgins DG Heringa J T-Coffee: A novel method for fast and accurate multiple sequence alignment J Mol Biol 2000 302 205 217 10964570 10.1006/jmbi.2000.4042
Hubert L Arabie P Comparing partitions. J Classif 1985 2 193 218 10.1007/BF01908075
Wootton JC Federhen S Statistics of local complexity in amino acid sequences and sequence databases Comput Chem 1993 17 149 163 10.1016/0097-8485(93)85006-X
Claverie JM Slates DJ Information enchancement methods for large scale sequence analysis. Comp Chem 1993 17 191 201 10.1016/0097-8485(93)85010-A
Waterman MS Gordon L Arratia R Phase transitions in sequence matches and nucleic acid structure Proc Natl Acad Sci U S A 1987 84 1239 1243 3469666
Altschul SF Gish W Local alignment statistics Methods Enzymol 1996 266 460 480 8743700
Altschul SF Madden TL Schaffer 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
Smith TF Waterman MS Burks C The statistical distribution of nucleic acid similarities Nucleic Acids Res 1985 13 645 656 3871073
Collins JF Coulson AF Lyall A The significance of protein sequence similarities Comput Appl Biosci 1988 4 67 71 3383005
Mott R Maximum-likelihood estimation of the statistical distribution of Smith-Waterman local sequence similarity scores Bull Math Biol 1992 54 59 75 10.1016/S0092-8240(05)80176-4
Pearson WR Lipman DJ Improved tools for biological sequence comparison Proc Natl Acad Sci U S A 1988 85 2444 2448 3162770
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
Buday L Membrane-targeting of signalling molecules by SH2/SH3 domain-containing adaptor proteins Biochim Biophys Acta 1999 1422 187 204 10393272
Wall L Christiansen T Orwant J Mui L Programming Perl 2002 O'Reilly & Associates, Inc., 101 Morris Street, Sebastopol, CA 95472
Pearson WR Effective protein sequence comparison Methods Enzymol 1996 266 227 258 8743688
Smith TF Waterman MS Identification of common molecular subsequences J Mol Biol 1981 147 195 197 7265238 10.1016/0022-2836(81)90087-5
Ihaka R Gentleman R R: A language for data analysis and graphics. J Comp Graph Stat 1996 5 299 314
Gumbel EJ Statistics of extremes 1958 New York , Columbia University Press
Pearson WR Empirical statistical estimates for sequence similarity searches J Mol Biol 1998 276 71 84 9514730 10.1006/jmbi.1997.1525
Arratia R Waterman MS A phase transition for the score in matching random sequences allowing deletions. Ann App Prob 1994 4 200 225
Eddy SR Maximum likelihood fitting of extreme value distributions
|
16135248
|
PMC1208858
|
CC BY
|
2021-01-04 16:27:26
|
no
|
BMC Bioinformatics. 2005 Aug 31; 6:216
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-216
|
oa_comm
|
==== Front
BMC BiotechnolBMC Biotechnology1472-6750BioMed Central London 1472-6750-5-221610916610.1186/1472-6750-5-22Methodology ArticleA simplified immunoprecipitation method for quantitatively measuring antibody responses in clinical sera samples by using mammalian-produced Renilla luciferase-antigen fusion proteins Burbelo Peter D [email protected] Radoslav [email protected] Thomas L [email protected] Lombardi Comprehensive Cancer Center, Department of Oncology, Georgetown University Medical Center, Washington, D.C. 200572005 18 8 2005 5 22 22 3 3 2005 18 8 2005 Copyright © 2005 Burbelo 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
Assays detecting human antigen-specific antibodies are medically useful. However, the usefulness of existing simple immunoassay formats is limited by technical considerations such as sera antibodies to contaminants in insufficiently pure antigen, a problem likely exacerbated when antigen panels are screened to obtain clinically useful data.
Results
We developed a novel and simple immunoprecipitation technology for identifying clinical sera containing antigen-specific antibodies and for generating quantitative antibody response profiles. This method is based on fusing protein antigens to an enzyme reporter, Renilla luciferase (Ruc), and expressing these fusions in mammalian cells, where mammalian-specific post-translational modifications can be added. After mixing crude extracts, sera and protein A/G beads together and incubating, during which the Ruc-antigen fusion become immobilized on the A/G beads, antigen-specific antibody is quantitated by washing the beads and adding coelenterazine substrate and measuring light production.
We have characterized this technology with sera from patients having three different types of cancers. We show that 20–85% of these sera contain significant titers of antibodies against at least one of five frequently mutated and/or overexpressed tumor-associated proteins. Five of six colon cancer sera tested gave responses that were statistically significantly greater than the average plus three standard deviations of 10 control sera. The results of competition experiments, preincubating positive sera with unmodified E. coli-produced antigens, varied dramatically.
Conclusion
This technology has several advantages over current quantitative immunoassays including its relative simplicity, its avoidance of problems associated with E. coli-produced antigens and its use of antigens that can carry mammalian or disease-specific post-translational modifications. This assay should be generally useful for analyzing sera for antibodies recognizing any protein or its post-translational modifications.
==== Body
Background
Although it is clear that a normal host immune system recognizes and responds to tumors, we understand very little about these complex tumor-host interactions. For example, it is not clear why tumor-associated proteins elicit humoral responses, although it is often speculated that such proteins can become antigenic when they are overexpressed or represent an unusual or modified form of a protein (e.g. altered spliced form), or are encoded by mutant genes [1,2]. Efforts to identify antibody responses to tumor antigens are motivated primarily by their diagnostic potential. Unfortunately, the immunoassay formats available to most laboratories are less than ideal.
Most immunoassays use bacterial-expressed proteins for detecting antigen-specific antibodies in human sera [2]. However, since such antigens do not carry post-translational modifications or may fold incorrectly, some immunoassays employ antigens produced in either yeast or insect cells. While these antigens may fold correctly and carry post-translational modifications, they will not carry either mammalian- or disease-specific posttranslational modifications. Tests employing bacterial-produced proteins can produce high backgrounds because it is difficult to completely eliminate or block serum antibodies reactive with trace amounts of bacterial contaminants present in most antigen preparations, even in pharmaceutical grade preparations [3]. Therefore to overcome the biological limitations and technical problems associated with bacterially and non-mammalian-produced antigens, we have developed a simple immunoassay that combines conventional immunoprecipitation techniques with a novel approach for the production of tumor antigens. The tumor antigens are fused to an enzymatic reporter, Ruc, and produced in mammalian cell cultures, where mammalian-specific post-translational modifications can be added. This technology is based on our previously published studies showing that a Ruc fusion with a human protein retain the biological activities of both the reporter and the human protein and can be used to detect weak protein-protein interactions [4]. In the present application, we utilized such fusions to detect protein-antibody interactions.
Our immediate interest in this technology is that we believe it can be used to systematically test the hypothesis that, in sporadic cancers, mutant or overexpressed tumor-associated proteins frequently induce humoral responses. While variations of this hypothesis have be proposed, it has not been vigorously tested for mainly historical and technological reasons. Until recently, only a few frequently mutated tumor-associated proteins were known [5]. For example, in sporadic breast and colon cancers, mutations in only a few proteins, p53 and SMAD4 or p53, k-Ras and APC, respectively, had been identified prior to 2001. Recent molecular genetic studies have greatly increased the number of genes known to be mutant in different types of sporadic cancers. For example, in colon cancers over 15 different genes are known to be frequently mutated [6-10], although individual patient tumors are highly heterogeneous in their mutant gene spectrum. In light of the fact that accurate classification of patient tumors into well-defined subtypes by gene expression profiling requires a panel of genes, each of which may be specifically up- or down-regulated in only a small percentage of tumors [11-15], we hypothesize that monitoring humoral immune responses to a panel of frequently mutated and/or overexpressed tumor-associated proteins in cancer patient sera can be used in an analogous manner, but with the added advantage of not requiring tumor tissue. In addition, existing data suggests that cancer patients' antibody responses to these mutant proteins are generally not limited to the mutated region of the protein. For example, colon cancer patient sera containing anti-Ras antibodies were equally reactive with either wild type or mutant K-Ras recombinant proteins [16]. Epitope mapping experiments showed that these sera always reacted with the C-terminus of K-Ras, although the mutated amino acid is almost always at the N-terminus of Ras. Similar results studying antibody responses to p53 were obtained with sera from patients with breast, colorectal and lung cancer [17]. While p53 mutations are in the central region, the majority of immunodominant epitopes are in the N- and/or C-termini of p53 [18,19]. For both mutant CDX2 in colon cancer, and mutant B-Raf in melanoma, patient antibodies react with both the wild type protein and mutant epitopes [20,21]. Here we describe a simple practical quantitative immunoprecipitation assay that has a number of practical advantages including that it is inexpensive, easy-to-perform and can be used for detecting antigen-specific antibodies in clinical sera samples. The proteins used here, as antigens are frequently mutated or overexpressed in the types of tumors carried by the patients whose sera are used to demonstrate the usefulness of this new immunoassay format.
Results and discussion
Description of the immunoprecipitation assay
We used Ruc-tagged proteins to develop an immunoprecipitation assay that can quantitatively measure serum antibody reactivity with protein antigens. Briefly, crude extract containing the Ruc-antigen fusions, sera and protein A/G beads are mixed together and incubated, during which the antigen fusions become immobilized; antigen-specific antibody is then quantitated by washing the beads and adding the colenterazine substrate. In these assays the amount of light produced is proportional to the amount of soluble fusion protein captured, directly or indirectly by the antibody-bound beads. It should be noted that the binding capacity of the protein A/G beads (Pierce Biochemical) used to capture either purified monoclonal antibodies or immunoglobulins from crude human or animal antisera is quite high (24 μg of immunoglobulins/μl of packed beads).
The immunoprecipitation assay shows a linear range of detection with commercial antibodies
To illustrate this technology we generated Ruc fusion protein constructs for p53, K-Ras, c-Myc, β-catenin and Smad4 by fusing cDNAs encoding these proteins (in frame) to DNA encoding the C-terminus of Ruc in a mammalian expression vector, pREN2, which also encodes a FLAG epitope tag at the N-terminus of Ruc. Transfections into Cos1 cells of these different constructs yielded crude extracts with 3–10 × 108 Ruc light units per 100 mm2 plate. We developed a standard assay format by using a commercial anti-FLAG monoclonal antibody and Cos1 cell extracts containing Ruc-p53. When crude extract, antisera and protein A/G beads were incubated together in a single tube, the amount of immunoprecipitated Ruc-p53 was directly proportional to the amount of anti-FLAG antibody over a 1000-fold range of concentrations, with a lower limit of detection of less than 5 picrograms (Figure 1). A commercial anti-p53 polyclonal antibody had a similar capture capacity as reflected by a similar dose-response curve, whereas a commercial polyclonal antibody against an unrelated antigen was unable to immunoprecipitate Ruc-p53 (Figure 1). Experiments using commercial polyclonal antibodies for serine-15 phosphorylated p53 and acetylated p53 (lysine-373 and lysine-382) also immunoprecipitated significant amounts of Ruc-p53 (data not shown). Since the ability of these modification-specific antibodies to immunoprecipitated Ruc-p53 was not competed by bacterially produced recombinant p53 protein (data not shown), this fusion protein appears to contain at least two types of post-translational modifications.
Human cancer patient sera contain antigen-specific antibodies
Since commercial antibodies can immunoprecipitate Ruc-antigen fusions from crude Cos1 extracts, we tested whether our simple assay format could also detect antigen-specific antibodies in clinical sera samples. Our motive for developing this technique was to have an improved method for detecting cancer patient antibody responses to tumor-associated proteins. Thus, we initially tested it with a small number of clinical sera samples taken from patients having three types of cancers, breast, colon and head and neck. In order to maximize our chances of detecting positive responses with these clinical sera samples we chose to use p53 and four other tumor-associated proteins (K-Ras, c-Myc, β-catenin and Smad4) that are either frequently mutated and/or overexpressed in various tumors. Wild type proteins were used as antigens because several studies show that cancer patient sera humoral immune responses are not restricted to or even preferential for the epitopes that usually contain the altered amino acids [16,18-21]. Cos1 extracts containing Ruc-antigen fusions were used to test a total of 36 sera, 10 controls and 26 cancer patients (Table 1). Negative and positive controls consisting of protein A/G beads alone and 0.1 μg of anti-FLAG monoclonal antibody with protein A/G beads, respectively, were used for each experiment. As expected, all sera had low reactivity with the non-specific binding control protein, Ruc-alone (Table 1). The positive control, anti-FLAG antibody, immunoprecipitated significant amounts of each of the Ruc-antigen fusions. However, the fraction of the total Ruc activity that could be captured varied amongst the different Ruc-antigen fusions, possibly reflecting reduced accessibility to the N-terminal FLAG epitope in some constructs (data not shown). At least one cancer patient sera had statistically significant antibody responses to each of the five Ruc fusions, where significance is defined as a response greater than the average plus three standard deviations of the 10 control sera (Table 1). Two of 10 head and neck, five of 10 breast, five of six colon cancer sera, but none of 10 healthy control sera gave positive responses. Six of the 12 positive tests were clustered in the six colon cancer patient sera and two antigens, p53 and K-Ras (Table 1). The significance of the relative response rates between different cancer-type sera cannot be calculated because the sample sizes are small and because no effort was made to match the control and patient sera by any criteria. Similarly, we cannot conclude that either K-Ras and/or p53 may be more antigenic in colon cancers than either β-catenin or c-Myc. Interestingly, the only multiple sample from any of the patients, head and neck samples 11 and 12, are sequential samples of which only the more recent sample showed significant levels of anti-p53 antibodies. Since the proteins used to test for antibodies in these 26 cancer patient sera are often mutated and/or overexpressed in the three types of cancer, our results are consistent with studies indicating that these categories of proteins are often antigenic in cancer patients [2]. Our results with colon cancer patient sera also support proposals that humoral immune responses to panels of tumor-associated antigens may be clinically useful when single antibody responses are not informative [22,23]. In any case, the detection of cancer sera from head and neck (20% sensitivity and 100% specificity), breast (40% sensitivity and 100% specificity), and colon (86% sensitivity and 100% specificity) obtained using a panel of five antigens are encouraging, given that our assay is not yet optimized and the sample sizes are small.
To determine whether patient antibody responses behave in the same linear manner as the commercial antibodies, we used the most reactive combination of patient sera and fusion antigen in our small sample set, colon cancer sera 34 and the Ruc-p53 fusion. Although the amount of Ruc-p53 captured by this serum is roughly linear with incubation time in the presence of protein A/G beads, reaching a plateau by 30–60 minutes (Figure 2A), the relative amount of immunoprecipitated Ruc-p53 was not completely linear with increasing amounts of sera (Figure 2B). Since the two commercial antibodies used in Figure 1 are highly purified, the non-linear dose-response curve of the clinical sera sample could be due to interfering agents such as anti-p53-specific IgA and IgM antibodies that recognize epitopes also recognized by IgG's but which bind poorly to protein A/G beads [24]. We are exploring modifications of the assay format in order to produce a more linear dose-response curve with clinical sera samples, which if found, would facilitate assay standardization and might increase sensitivity (note that 1–2 of the colon cancer responses to one of the antigens tested, β-catenin-Δ1, are barely below the cut-off value used to judge statistical significance). If the presumed interfering agents also affected ELISA tests, ELISA tests may significantly underestimate positive antibody responses and antibody titers unless the sera are sufficiently diluted.
Competition experiments with unmodified proteins
While human humoral immune responses to post-translational modifications are often ignored and/or undetectable with existing technologies, recent studies demonstrate that disease-related antibody responses can occur to post-translational protein modifications [25]. In at least one case, rheumatoid arthritis, antibody responses to a post-translational modification, citrullination, is now being intensely investigated as a potentially reliable disease indicator [26,27]. In light of these observations, we asked whether each positive sera response seen in Table 1 could include antibodies that were directed toward post-translational modifications by doing competition experiments with unmodified E. coli-produced antigens. These competition experiments (Table 2) show that 0–100% of the immunoprecipitated Ruc-antigen fusions were blocked by preincubating sera with 5 μg of the corresponding E. coli-produced antigens fused to maltose binding protein (MBP). These differences occur even between sera containing antibodies that recognize the same antigen (e.g. p53 or K-Ras), proteins known to contain post-translational modifications. These differences could mean that some tumors tend to produce proteins having more post-translational modifications or that some cancer patient's immune system tend to produce significantly more antibodies that recognize post-translational modifications. However, this data does not exclude the possibility that some or all of each positive antibody response detected is not even specific for the antigen listed, since the apparent anti-p53 or anti-K-Ras antibodies could be directed toward proteins that are in complexes with these tumor antigens. If the tumor antigens in these complexes were easily replaced by the MBP fusions, one would see higher competition values than if they were inefficiently replaced. Quantitative evaluation of different competition results requires, at a minimum, equal amount of reactive antibodies in each sera, a condition unlikely to be satisfied here, especially for the p53-reactive sera. In addition, when we compared the dose-response competition curves of sera 34 and the commercial polyclonal anti-p53, adjusted to similar capacities for immunoprecipitating Ruc-p53, we found a greater difference than indicated by the end-point values alone (Figure 3). Nevertheless, it is clear that our assay identifies patient sera having qualitatively different humoral immune responses to the same antigen. Additional tools, especially antigens bearing only a single type of modification, will be required to determine whether some or all of the presumed "low affinity" antibodies prefer epitopes bearing post-translational modifications because many such antibodies are likely to cross-react with unmodified epitopes.
We have preliminary observations suggesting that our approach of making antigen-enzyme fusions and producing these fusions in mammalian cells may be superior to conventional ELISA assays for detecting antigen-specific antibody responses in human sera. Specifically, we have tested the six colon cancer patient sera used here in a standard sandwich type ELISA where the antigen were fused to E. coli MBP and immobilized on ELISA plates with a monoclonal anti-MBP antibody. In these ELISA tests only two of the six colon cancer sera gave positive responses with any of the five tumor-associated proteins listed in Table 1 (data not shown). In any case, the immunoprecipitation assay described here offers a practical approach for identifying post-translational modification-specific antibody responses and studying their medical relevance.
Conclusion
These results demonstrate that a simple quantitative immunoprecipitation assay can identify human clinical sera samples containing disease-related antigen-specific antibodies. Quantitative results were obtained by using easily prepared crude cell extracts containing post-translationally modified antigens fused to a light-producing enzyme reporter. While the immunodetection of antigen-enzymes is not new [28,29], by combining a robust reporter, such as Ruc with the production of recombinant enzyme-antigen fusions in mammalian cells, we have created a highly sensitive user friendly assay. This assay requires fewer manipulations for reagent preparation and less time than other immunoprecipitation methods including avoiding having to purify and then radiolabel the purified proteins or having to perform additional analysis such as Western blotting after the immunoprecipitations [30]. Producing the target antigens in mammalian cells offers several potential advantages, including having mammalian-specific and/or disease-specific post-translational modifications added to these antigens. Thus, this immunoprecipitation assay provides a simple, accessible, reliable and reproducible tool for investigations aimed at documenting the role of post-translational modification in disease. Although altered post-translationally modified proteins occur in cancer [31,32], future studies are needed to explore whether there are detectable cancer patient-specific antibodies to post-translationally-modified tumor proteins. The levels and kinds of post-translational modifications on the Ruc-antigen fusions can be manipulated by exploiting mutant proteins, unique human cell lines (e.g. cell lines overexpressing tyrosine kinases) and various culture conditions. Mammalian-produced antigens have additional advantages over bacterial produced antigens including facilitating the study of antibody responses to very large proteins (>100 kDa) that are difficult or impossible to produce as intact proteins in E. coli. Our assay also avoids false positives caused by variable amounts of anti-E. coli antibodies present in patient sera that react with the minor amounts of E. coli proteins that co-purify with bacterial recombinant proteins; such contaminants are even present in some pharmaceutical-grade recombinant protein preparations [3]. These advantages, along with the possibility of improving the assay format, suggest that it may be worthwhile to use this assay to reevaluate the frequency with which known tumor-associated proteins are detectably antigenic in cancer patients. It is encouraging, although of limited significance, that the frequencies of significant antibody responses for two of the cancers are roughly comparable to reports in the literature. Thus, in colon cancer patients we detected statistically significant antibody responses to Ras and p53 in 50% and 33% of the sera, respectively, compared to published reports of 33% for Ras [16] and 26% for p53 [33]. In contrast, we did not find any statistically significant antibody responses to p53 in breast cancer sera, which have been reported to occur with 9% of patient sera [34]. Studies with much larger sample numbers are clearly needed to make statistically useful comparisons between our method and existing methods.
This assay format and high throughput modifications (e.g. magnetic A/G beads in a microtiter plate format) are obviously directly applicable to detecting human sera antibodies specific for any protein antigen of interest and is likely to be useful for non-human sera, such as sera obtained from animal models of disease, as well as for antibodies in other bodily fluids including from ascites and saliva. Variations of this immunoprecipitation assay format might also be useful for studying other types of protein-protein interactions.
Methods
Biochemical reagents and antibodies
Ultralink™ immobilized protein A/G beads were obtained from Pierce Biotechnology Inc. Commercially available antibodies were: mouse monoclonal anti-FLAG™ M2 from Sigma; rabbit anti-acetylated p53 from Upstate Biochemicals and polyclonal rabbit anti-p53, polyclonal rabbit phosphoserine p53 and polyc lonal anti-WASP from Santa Cruz Biotechnology.
Patient sera
The breast and colon cancer patient sera were obtained from the University of Wisconsin collection, now kept at Georgetown University Medical Center. Sera samples from head and neck cancer patients and control sera were collected by Dr. Radoslav Goldman at Georgetown University Medical Center (Washington, DC). The sex, age and disease stages of these samples were not examined until after the reactivities for all antigens were measured.
Generation of constructs encoding Ruc fused to tumor-associated antigens
pREN2, a FLAG-epitope-tagged mammalian expression vector, similar to the previously described pREN1 [4], was used to generate all plasmids encoding Ruc fusions. The tumor antigens are at the C-terminus and a single FLAG tag is at the N-terminus of Ruc. A map of pREN2 is shown in Figure 4. The cloned human cDNA fragments, amplified by PCR specific linker-primer adapters, were obtained from Dr. E. Chang (p53), Dr. R. Lechleider (Smad4), Dr. S. Byers (β-catenin), Dr. R. Dickson (c-Myc) and a publicly available cDNA clone (IMAGE ID 6714574) for K-Ras. Full-length coding sequences (excluding the initial methionine) were used for the tumor antigens, with the exception of the β-catenin, which encodes amino acids 2–277. In every case a stop codon was included after the C-terminal coding sequences of the tumor antigens. The primer adapter sequences used for cloning each antigen are as follows: p53, 5'-GAGGGATCCGAGGAGCCGCA GTCAGAT-3' and 5'-GAGCTCGAGTCAGTCTGAGTCAGGCC-3'; K-Ras, 5'-GAGGGATCCACTGAATATAAACTTGTG-3' and 5'-GAGCTCGAGTTACATAATTACACACTT; Smad4, 5'-GAGGGATCCGACAATATGTCTATTACG-3' and 5'-GAGCTCGAGTCAGTCTAAAGGTTGTGG-3'; β-catinin-Δ, 5'-GAGGGATCCGCTACTCA AGCTGATTTG-3' and 5'-GAGGTCGACTCAACCAGCTAAACGCACTGC-3'; and c-Myc, 5'-GAGG GATCCCTCAACGTTAGCTTCACC-3' and 5'-GAGCTCGAGTTACGCACAAGAGTTCCG-3'. For Ruc alone, a separate construct was prepared containing a stop codon at the end of the luciferase coding sequence in place of the polylinker present in pREN2.
Immunoprecipitation assays with Ruc fusion proteins
Forty-eight hours after Fugene-6 transfection, Cos1 cells in 100 mm2 plates were washed twice with PBS, scraped with 1.0 ml of Buffer A (20 mM Tris, pH 7.5, 150 mM NaCl, 5 mM MgCl2, 1% Triton X-100) plus 50% glycerol and protease inhibitors (10 μg/mL each of leupeptin, aprotinin and pepstatin), sonicated, centrifuged at 13,000 × g for 4 min, supernatants collected and used immediately or stored at -20°C. Total luciferase activity in 1 μl of each crude extract was measured by adding it to 100 μl of assay buffer and substrate mixture (Renilla Luciferase Reagent Kit, Promega) in a 12 × 75 mm glass tube, vortexing and immediately measuring light-forming units with a luminometer (GeneProbe) for 10 sec. Lysate prepared from each 100 mm2 plate of transfected Cos1 cells typically provides enough extract for 60–200 assays. These crude Cos1 extracts containing these Ruc fusions were stable for at least a few weeks when stored in 50% glycerol at -20°C.
Immunoprecipitation assays were performed in 100 μl volumes containng 6 μl of a 30% suspension of protein A/G beads (in PBS), 1–10 μl sera (undiluted or diluted in Buffer A plus 100 μg/ml BSA), sufficient Cos1 cell extract to generate 1–5 million light units (usually 5 μl to 10 μl) and Buffer A and incubated at 4°C with tumbling for 5–120 minutes, washed 4–5 times with 1.2 ml of cold Buffer A and once with 1.0 ml of PBS. After the final wash, the beads, in a volume of about 10 μl, were added to the Ruc substrate and light units measured as described above. Since the capacity of these protein A/G is 24–32 mg/ml of packed beads, 2 μl of packed beads should be sufficient to immobilize most or all of the IgG in 1 μl of undiluted sera (assumed to be 10 mg/ml IgG). The amount of IgG in 2 μl of each sera that actually bound to protein A/G beads was estimated by measuring the amount of bead-bound sera released by a low pH glycine elution buffer and measured using the BCA Protein Assay kit (Pierce Biotechnology Inc.). The protein values varied from 2.0 μg to 7.3 μg/μl of patient sera (see Additional file 3).
Competition experiments were performed using MBP fusion proteins. Bacterial expression vectors were constructed by subcloning cDNA fragments into the pMAL-c2 vector (New England Biolabs). Recombinant MBP fusion proteins were produced in bacteria, purified by amylose-agarose affinity and eluted with maltose as described by the manufacturer and stored frozen or in 50% glycerol at -20°C. An MBP fusion containing the SPEC2 cDNA [35] was produced and used as a non-specific inhibitor. The integrity of the proteins was confirmed by SDS-PAGE electrophoresis and protein concentration determined. Diluted patient sera (10 μl used of sera diluted 1:10 in buffer A containing 100 μg/ml BSA) were used in the competition experiments described in Table 2, while only 5 μl of 1:10 diluted colon patient sera 34 was used in the experiments described in Figure 3.
Authors' contributions
RG provided the sera samples for the controls and head and neck cancers. TLM produced the recombinant fusion proteins used for the competition experiments and assisted in reducing the concept to practice and manuscript writing. PDB* first conceived of the concept, generated the constructs and performed the immunoprecipitation experiments.
Supplementary Material
Additional File 3
Table 3. Amount of protein (IgG) bound to A/G bead (μg/1 μl) from different sera used in this study
Click here for file
Additional File 1
Table 1. Mean and standard deviations of sera reactivity
Click here for file
Additional File 2
Table 2. Competition of antibody responses by unmodified antigens
Click here for file
Acknowledgements
We would like to thank Nicholas Madian for technical help and Kathryn Ching for numerous helpful suggestions and assistance. We are also grateful to Dr. R. Parniak and Dr. A. Uren for helpful comments on the manuscript. We also thank the Friends You can Count On Foundation for funding our initial work of tumor antigens. This study was funded through a grant from the Susan G. Komen Breast Cancer Foundation (BCTR02-1017) awarded to PDB and in part by American Cancer Society Grant CRTG-02-245-01-CCE awarded to RG. Additional support was in part by the Lombardi Comprehensive Cancer Center Tissue Culture and Biomarkers Shared Resources, U.S. Public Health Service Grant 2P30-CA-51008 and 1S10 RR15768-01.
Supported in part by grant M01 RR-020359 from the National Center for Resereach Resources, National Institues of Health
Figures and Tables
Figure 1 Immunoprecipitation experiments with commercial antibodies. Various amounts of anti-FLAG monoclonal, anti-p53 polyclonal or control (anti-WASP) polyclonal antibodies were mixed with 5 μl of a Cos1 extract containing Ruc-p53 for 1 h in the presence of protein A/G beads, processed and light units measured. The data shown is from one of three independent experiments giving similar results.
Figure 2 The immunoprecipitation assay with Ruc-p53 and a clinical serum sample. A: The immunoprecipitation activity is proportional to incubation time. Tubes containing identical amounts of Ruc-p53 fusion protein extract (5 μl), patient 34 sera (1 μl) and protein A/G beads were incubated for 5, 30, 60, 90 and 120 min and processed for luciferase activity. B: Immunoprecipitation activity with various amounts of total crude patient 34 sera. Different amounts of patient sera (0.002 to 2 μl) were mixed with 5 μl of the Ruc-p53 fusion protein extract and incubated for 1 hour in the presence of protein A/G beads, processed and light units measured. The data shown is from one of three independent experiments giving similar results.
Figure 3 Competition assays blocking Ruc-p53 immunoprecipitation using bacterially-produced antigen. Different amounts of E. coli-produced MBP-p53 were incubated with patient sera 34 (0.5 μl) or commercial anti-p53 antibody (25 ng) for 1 h. Protein A/G beads and Ruc-p53 extract were then added and incubated for an additional 1 h, processed and light units measured. The data shown is from one of two independent experiments giving similar results.
Figure 4 Structure of the pREN2 mammalian expression vector. Features indicated are CMV (cytomegalovirus) promoter, the N-terminal FLAG epitope and Ruc. Sequences for Ruc are in bold. cDNAs for tumor antigens were cloned downstream of Ruc between the BamH1-Xho1 sites.
Table 1 Immunoprecipitation capacity of 1 μl of human sera for Ruc-tumor antigen fusion proteinsa
Ruc p53 K-Ras Smad4 β-CAT-Δ1 c-Myc
Controls 1 194 19,319 480 10,582 269 4,752
2 9 9,830 1,064 3,575 835 2,913
3 8 5,236 445 1,773 211 2,006
4 38 3,187 477 1,919 530 1,831
5 14 11,908 795 6,884 161 3,346
6 31 5,390 823 1,724 235 2,050
7 76 22,526 1,909 6,996 259 11,816
8 29 15,338 943 8,043 445 3,475
9 10 12,282 1,162 19,380 215 3,623
10 9 11,130 1,109 4,429 501 5,060
+ 3 SDb 214 30,234 2,237 22,788 997 12,874
Head and Neck 11 0 10,904 508 2,721 196 2,193
12 0 31,593c 738 4,822 465 3,801
13 0 12,367 840 1,868 673 4,407
14 13 14,705 1,012 5,666 195 1,837
15 33 31,733c 1,189 5,264 552 4,107
16 121 4,828 621 980 279 1,974
17 0 8,517 1,160 8,396 336 2,958
18 0 19,240 1,283 9,485 327 1,814
19 0 11,224 1,517 4,454 410 4,370
20 28 7,322 554 2,261 723 2,343
Breast 21 44 13,211 960 10,219 308 5,988
22 10 18,814 696 42,970c 302 5,450
23 38 14,598 608 8,484 339 4,336
24 77 11,587 1,655 17,297 2363c 3,431
25 17 19,954 532 10,184 772 15,650c
26 25 9,538 195 5,962 300 1,646
27 10 7,815 2,561c 20,628 426 3,524
28 21 15,607 308 7,380 284 1,579
29 0 18,058 160 6,790 304 2,333
30 245c 25,479 1,919 9,727 495 3,787
Colon 31 4 6,656 1,204 3,252 267 1,763
32 40 20,928 4,293c 5,567 962 6,143
33 42 34,703c 1,472 10,830 716 4,906
34 51 300,943c 6,439c 2,610 992 3,789
35 35 5,670 3,306c 3,860 477 1,772
36 44 6,516 695 37,344c 371 2,395
aSera, FLAG-Ruc-fusion extracts, protein A/G beads and buffer were mixed together, incubated for 60 minutes and processed. The data, light units, is the average of two experiments and is corrected for background (beads plus extract, but no sera). The standard deviation for each value is also available (see Additional file 1).
bValues of the averages of the 10 control sera plus 3 standard deviations.
cNumbers in bold are statistically significant: greater than the average plus 3 standard deviations of the 10 control sera.
Table 2 Competition of antibody responses by unmodified antigensa
Antigen/sera Control p53 K-Ras Smad4 β-CAT-Δ1 c-Myc
p53/12 21% 32%
p53/15 20% 60%
p53/33 7% 88%
p53/34 11% 72%
K-Ras/27 5% 91%
K-Ras/32 25% 82%
K-Ras/34 4% 0%
K-Ras/35 16% 100%
Smad4/22 4% 92%
Smad4/36 0% 93%
β-catenin-Δ1/24 23% 96%
c-Myc/25 0% 22%
aSera (1 μl), buffer and 5 μg competitor were incubated together for 60 min before adding the fusion extracts and protein A/G beads for an additional 60 minutes and processed. Background light units (beads plus extract but no sera) were subtracted before calculating percent competition. The first column identifies the antigen-sera combination tested. The other columns give the amount of competition obtained for each competitor antigen. All competitors, including the control (SPEC2), are MBP fusion proteins. Values are the averages plus from two independent experiments. The standard deviation for each value is also available (see Additional file 2).
==== Refs
Old LJ Chen YT New paths in human cancer serology J Exp Med 1998 187 1163 1167 9547328 10.1084/jem.187.8.1163
Preuss KD Zwick C Bormann C Neumann F Pfreundschuh M Analysis of the B-cell repertoire against antigens expressed by human neoplasms Immunol Rev 2002 188 43 50 12445280 10.1034/j.1600-065X.2002.18805.x
Wadhwa M Skog AL Bird C Ragnhammar P Lilljefors M Gaines-Das R Mellstedt H Thorpe R Immunogenicity of granulocyte-macrophage colony-stimulating factor (GM-CSF) products in patients undergoing combination therapy with GM-CSF Clin Cancer Res 1999 5 1353 1361 10389919
Burbelo PD Kisailus AE Peck JW Detecting protein-protein interactions using Renilla luciferase fusion proteins Biotechniques 2002 33 1044 1048 1050. 12449382
Vogelstein B Kinzler KW Cancer genes and the pathways they control Nat Med 2004 10 789 799 15286780 10.1038/nm1087
Bardelli A Parsons DW Silliman N Ptak J Szabo S Saha S Markowitz S Willson JK Parmigiani G Kinzler KW Vogelstein B Velculescu VE Mutational analysis of the tyrosine kinome in colorectal cancers Science 2003 300 949 12738854 10.1126/science.1082596
Wang Z Shen D Parsons DW Bardelli A Sager J Szabo S Ptak J Silliman N Peters BA van der Heijden MS Parmigiani G Yan H Wang TL Riggins G Powell SM Willson JKV Markowitz S Kinzler KW Vogelstein B Velculescu VE Mutational analysis of the tyrosine phosphatome in colorectal cancers Science 2004 304 1164 1166 15155950 10.1126/science.1096096
Wang Z Cummins JM Shen D Cahill DP Jallepalli PV Wang TL Parsons DW Traverso G Awad M Silliman N Ptak J Szabo S Willson JKV Markowitz S Goldberg ML Karess R Kinzler KW Vogelstein B Velculescu VE Langauer C Three classes of genes mutated in colorectal cancers with chromosomal instability Cancer Res 2004 64 2998 3001 15126332
Rajagopalan H Jallepalli PV Rago C Velculescu VE Kinzler KW Vogelstein B Lengauer C Inactivation of hCDC4 can cause chromosomal instability Nature 2004 428 77 81 14999283 10.1038/nature02313
Samuels Y Wang Z Bardelli A Silliman N Ptak J Szabo S Yan H Gazdar A Powell SM Riggins GJ Willson JKV Markowitz S Kinzler KW Vogelstein B Velculescu VE High frequency of mutations of the PIK3CA gene in human cancers Science 2004 304 554 15016963 10.1126/science.1096502
Perou CM Sorlie T Eisen MB van de Rijn M Jeffrey SS Rees CA Pollack JR Ross DT Johnsen H Akslen LA Fluge O Pergamenschikov A Williams CL Xhu SX Lonning PE Borresen-Dale AL Brown PO Botstein D Molecular portraits of human breast tumours Nature 2000 406 747 752 10963602 10.1038/35021093
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 Schreiber GJ Kerkhoven RM Roberts C Lisley PS Bernards R Friend SH Gene expression profiling predicts clinical outcome of breast cancer Nature 2002 415 530 536 11823860 10.1038/415530a
Sorlie T Perou CM Tibshirani R Aas T Geisler S Johnsen H Hastie T Eisen MB van de Rijn M Jeffrey SS Thorsen T Quist H Matese JC Brown PO Botstein D Lonning PE Borresen-Dale AL Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications Proc Natl Acad Sci U S A 2001 98 10869 10874 11553815 10.1073/pnas.191367098
Sorlie T Tibshirani R Parker J Hastie T Marron JS Nobel A Deng S Johnsen H Pesich R Geisler S Demeter J Perou CM Lonning PE Brown PO Borresen-Dale AL Botstein D Repeated observation of breast tumor subtypes in independent gene expression data sets Proc Natl Acad Sci U S A 2003 100 8418 8423 12829800 10.1073/pnas.0932692100
Sotiriou C Neo SY McShane LM Korn EL Long PM Jazaeri A Martiat P Fox SB Harris AL Liu ET Breast cancer classification and prognosis based on gene expression profiles from a population-based study Proc Natl Acad Sci U S A 2003 100 10393 10398 12917485 10.1073/pnas.1732912100
Takahashi M Chen W Byrd DR Disis ML Huseby ES Qin H McCahill L Nelson H Shimada H Okuno K Antibody to ras proteins in patients with colon cancer Clin Cancer Res 1995 1 1071 1077 9815896
Soussi T p53 Antibodies in the sera of patients with various types of cancer: a review Cancer Res 2000 60 1777 1788 10766157
Lubin R Schlichtholz B Bengoufa D Zalcman G Tredaniel J Hirsch A de Fromentel CC Preudhomme C Fenaux P Fournier G. Analysis of p53 antibodies in patients with various cancers define B-cell epitopes of human p53: distribution on primary structure and exposure on protein surface Cancer Res 1993 53 5872 5876 8261396
Vennegoor CJ Nijman HW Drijfhout JW Vernie L Verstraeten RA von Mensdorff-Pouilly S Hilgers J Verheijen RH Kast WM Melief CJ Kenemans P Autoantibodies to p53 in ovarian cancer patients and healthy women: a comparison between whole p53 protein and 18-mer peptides for screening purposes Cancer Lett 1997 116 93 101 9177463 10.1016/S0304-3835(97)00168-7
Ishikawa T Fujita T Suzuki Y Okabe S Yuasa Y Iwai T Kawakami Y Tumor-specific immunological recognition of frameshift-mutated peptides in colon cancer with microsatellite instability Cancer Res 2003 63 5564 5572 14500396
Fensterle J Becker JC Potapenko T Heimbach V Vetter CS Brocker EB Rapp UR B-Raf specific antibody responses in melanoma patients BMC Cancer 2004 4 62 15361259 10.1186/1471-2407-4-62
Stockert E Jager E Chen YT Scanlan MJ Gout I Karbach J Arand M Knuth A Old LJ A survey of the humoral immune response of cancer patients to a panel of human tumor antigens J Exp Med 1998 187 1349 1354 9547346 10.1084/jem.187.8.1349
Zhang JY Casiano CA Peng XX Koziol JA Chan EK Tan EM Enhancement of antibody detection in cancer using panel of recombinant tumor-associated antigens Cancer Epidemiol Biomarkers Prev 2003 12 136 143 12582023
Bjorck L Kronvall G Purification and some properties of streptococcal protein G, a novel IgG-binding reagent J Immunol 1984 133 969 974 6234364
Utz PJ Gensler TJ Anderson P Death, autoantigen modifications, and tolerance Arthritis Res 2000 2 101 114 11094420 10.1186/ar75
Schellekens GA de Jong BA van den Hoogen FH van de Putte LB van Venrooij WJ Citrulline is an essential constituent of antigenic determinants recognized by rheumatoid arthritis-specific autoantibodies J Clin Invest 1998 101 273 281 9421490
Yamada R Suzuki A Chang X Yamamoto K Citrullinated proteins in rheumatoid arthritis Front Biosci 2005 10 54 64 15574347
Koenen M Ruther U Muller-Hill B Immunoenzymatic detection of expressed gene fragments cloned in the lac Z gene of E. coli Embo J 1982 1 509 512 6329687
Ruther U Koenen M Sippel AE Muller-Hill B Exon cloning: immunoenzymatic identification of exons of the chicken lysozyme gene Proc Natl Acad Sci U S A 1982 79 6852 6855 6184717
Phizicky EM Fields S Protein-protein interactions: methods for detection and analysis Microbiol Rev 1995 59 94 123 7708014
Bode AM Dong Z Post-translational modification of p53 in tumorigenesis Nat Rev Cancer 2004 4 793 805 15510160 10.1038/nrc1455
Lim YP Mining the tumor phosphoproteome for cancer markers Clin Cancer Res 2005 11 3163 3169 15867208
Hammel P Boissier B Chaumette MT Piedbois P Rotman N Kouyoumdjian JC Lubin R Delchier JC Soussi T Detection and monitoring of serum p53 antibodies in patients with colorectal cancer Gut 1997 40 356 361 9135525
Crawford LV Pim DC Bulbrook RD Detection of antibodies against the cellular protein p53 in sera from patients with breast cancer Int J Cancer 1982 30 403 408 6292117
Pirone DM Fukuhara S Gutkind JS Burbelo PD SPECs, small binding proteins for Cdc42 J Biol Chem 2000 275 22650 22656 10816584 10.1074/jbc.M002832200
|
16109166
|
PMC1208859
|
CC BY
|
2021-01-04 16:02:58
|
no
|
BMC Biotechnol. 2005 Aug 18; 5:22
|
utf-8
|
BMC Biotechnol
| 2,005 |
10.1186/1472-6750-5-22
|
oa_comm
|
==== Front
BMC BiotechnolBMC Biotechnology1472-6750BioMed Central London 1472-6750-5-231613733010.1186/1472-6750-5-23Methodology ArticleDescription of a PCR-based technique for DNA splicing and mutagenesis by producing 5' overhangs with run through stop DNA synthesis utilizing Ara-C Ailenberg Menachem [email protected] Neil M [email protected] Mel [email protected] Department of Medicine, Medical Science Building, Room 7207, University of Toronto Toronto, Ontario M5S 1A8, Canada2005 1 9 2005 5 23 23 22 6 2005 1 9 2005 Copyright © 2005 Ailenberg 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
Splicing of DNA molecules is an important task in molecular biology that facilitates cloning, mutagenesis and creation of chimeric genes. Mutagenesis and DNA splicing techniques exist, some requiring restriction enzymes, and others utilize staggered reannealing approaches.
Results
A method for DNA splicing and mutagenesis without restriction enzymes is described. The method is based on mild template-dependent polymerization arrest with two molecules of cytosine arabinose (Ara-C) incorporated into PCR primers. Two rounds of PCR are employed: the first PCR produces 5' overhangs that are utilized for DNA splicing. The second PCR is based on polymerization running through the Ara-C molecules to produce the desired final product. To illustrate application of the run through stop mutagenesis and DNA splicing technique, we have carried out splicing of two segments of the human cofilin 1 gene and introduced a mutational deletion into the product.
Conclusion
We have demonstrated the utility of a new PCR-based method for carrying out DNA splicing and mutagenesis by incorporating Ara-C into the PCR primers.
==== Body
Background
Splicing of DNA molecules is an important task in molecular biology that facilitates cloning, mutagenesis and the creation of chimeric genes. While the advent of restriction enzymes substantially advanced DNA splicing techniques, they cannot be applied universally, and their use is limited to enzyme-specific loci. Other techniques like site-directed mutagenesis by overlap extension [SOE; [1]], insertional mutagenesis with the megaprimer technique [2] and staggered reannealing [3,4] have further improved DNA mutagenesis and splicing. Each method offers advantages and inherent drawbacks. Another cloning approach involving the formation of 5' overhangs utilizes incorporation of nucleotide derivatives into PCR primers [5-7] that stall polymerization. These techniques are dependent on an established set of optimal conditions for strong polymerization arrest, including the correct choice of polymerase or the incorporation of three ribonucleotide derivatives in the primer [7]. Furthermore, chimeric DNA/RNA primers need to be removed and reverse-transcribed in order for the splicing to be completed.
Although in the past we have successfully used the SOE technique for mutagenesis and splicing, we encountered difficulties while constructing larger genes. That led us to develop the staggered reannealing method [3,4]. This method proved to be useful as well, however, its efficiency declined as the gene to be mutagenized exceeded 1000 bp. Although these techniques allow splicing of any two DNA fragments without the need for restriction enzymes, their efficiency is inversely related to the length of the DNA fragments involved, since these techniques rely on the successful melting and reannealing of DNA to create matching overhangs. We sought to offer an alternative approach to facilitate the splicing of any two DNA segments for mutagenesis and construction of chimeric genes. Our technique utilizes two rounds of PCR, and is based on moderate template-dependent polymerization arrest using cytosine arabinose (Ara-C).
Ara-C is a nucleotide derivative (Fig 1) that is widely used in cancer therapeutics [8]. It is a competitive inhibitor of DNA polymerase and also affects polymerization initiation [9,10]. Ara-C exerts it therapeutic action on cellular DNA polymerase after phosphorylation by an endogenous kinase. Once phosphorylated, Ara-C facilitates inhibition of DNA replication in cancer cells. Sanger et al, in their search for polymerization terminating agents for use in sequencing techniques, found that while dideoxynucleotides were strong polymerase terminators, Ara-C only weakly halted polymerization [11]. Therefore, even today, dideoxynucleotides remain the terminators of choice in sequencing reactions. Previous studies have shown that while Ara-C could serve as a substrate for mammalian polymerases, it terminates polymerization by some prokaryotic polymerases [11].
Here we used Ara-C both as DNA polymerase inhibitor and template for DNA mutagenesis and splicing.
Results and discussion
We were searching for a mild template-dependent polymerization terminator. The rationale for mild termination is as follows: Termination must be strong enough to create 5' overhangs in the first PCR reaction, but weak enough to allow the polymerase to continue through the modified nucleotide during the second round of PCR (Fig 2). For the reasons mentioned above, Ara-C was chosen for use in the present study.
As proof of principal, a 20 bp deletion in the human cofilin 1 gene was created. We tried to splice together two segments of the gene: one 5' (237 bp) and one 3' (309 bp) segment (Fig 2). Primers were designed with one or two Ara-C molecules replacing native deoxycytidine nucleotides. When two Ara-C molecules were incorporated into the primer (Hospital for Sick Children, Toronto, Canada), template-dependent termination can potentially occur before, after one, or after two Ara-C molecules. Therefore, to determine the termination location, we designed one side of the overhang to accommodate termination after two Ara-C molecules, and the other side of the overhang to accommodate termination after one Ara-C molecule (Fig 3).
There were a total of 8 PCR reactions that included two Ara-C primers for each of the two segments, and the two polymerases (Taq and Pfu) for each set of primers. PCR products were gel-isolated. At this stage, gel-isolation is essential in order to remove any of the original plasmid that might serve as a template in the second PCR reaction. Alternatively, the original plasmid may be eliminated by digestion with DpnI, although this option is less recommended, since traces of undigested plasmid could affect the outcome of the second PCR reaction. Corresponding segments to be spliced were combined (total of four tubes) and ligated. As indicated above, the rationale for this technique is that Ara-C is a mild polymerization terminator, and therefore it will produce a mixture of cohesive and blunt ends. Hence, the reaction is expected to both terminate (producing sticky ends essential for the splicing phase) and run through the Ara-C (producing blunt ends; this feature is essential to the second PCR reaction). Therefore, lowered concentration of ligase and reduced ligation time were used to optimize conditions to favor cohesive end ligation. The products of the ligase reaction were amplified by the second PCR with Taq or Pfu polymerases using the sense primer A, and the anti-sense primer B, which span the cofilin 1 gene. This PCR reaction produced the expected 552 bp product (blunt end ligation, is expected to produce an extra duplicated piece of DNA of 15 bp). The PCR products were either sequenced directly, or cloned into a plasmid and then sequenced. Based on sequencing results, we observed that incorporation of two Ara-C nucleotides into the PCR primers yielded the expected product. This suggests that the two molecules of Ara-C provided the desired mild termination to produce a product with 5' overhangs, but also allowed the polymerase to run through during the 2nd PCR. Furthermore, based on the design of the primers, the polymerization stalled both after the first and the second Ara-C molecule. Both 5 and 30 min incubations with DNA Ligase were sufficient to preferentially ligate the cohesive ends. This further suggests that two adjacent molecules of Ara-C produce 5' overhangs. Even though both 5 and 30 min ligations were successful in producing the desired product, it is not recommended to allow the reaction to proceed for a prolonged time, nor is it recommended to use high levels of ligase, since these conditions may facilitate blunt end ligation that may produce a mixture of the blunt and cohesive end products. Both Pfu and Taq polymerases were equally capable of producing termination products in the first PCR, while still running through the Ara-C in the second PCR. When one molecule of Ara-C was incorporated in the PCR primers, no termination could be observed, as seen by the addition of a 15 bp segment in the PCR product indicative of blunt end ligation. Even ligation for 5 min in reduced concentration of ligase failed to produce cohesive end ligation when only one Ara-C was employed.
The run through stop method utilizes a novel approach for DNA splicing and mutagenesis. While other mutagenesis techniques like SOE, megaprimer and staggered reannealing create matching overhangs after melting and reannealing, the run through stop method creates matching overhangs by polymerization arrest with Ara-C. We were motivated to design the Ara-C approach because we were not successful in creating gene mutations with the other techniques. Hence, the run through stop offers a good alternative to these techniques.
It has been previously demonstrated that utilizing abasic or RNA nucleotides like tetrahydrofuran derivative or 2-o-methyl ribonucleotide in PCR primers produced 5' overhangs that facilitated cloning of DNA fragments into plasmids [5-7]. These approaches were dependent on strong polymerization termination by the nucleotides. Our technique established the conditions for mild termination of DNA polymerization with two Ara-C molecules. This enables us to use the Ara-C-containing primers in two steps of PCR for DNA splicing and mutagenesis. Although in the present study we used relatively short segments of DNA for proof of principal (~500 bp of the human cofilin gene), this technique, unlike the staggered reannealing technique, is not limited to short DNA fragments. Since both rounds of PCR in the present study are based on conventional PCR, the length limit of the DNA fragments to be mutagenized is that of the PCR technique.
Conclusion
The run through stop method can be summarized in four steps:
1. Amplify two segments of DNA to be spliced using PCR, with phosphorylated primers containing two adjacent molecules of Arabinose nucleotide with overlapping sequence.
2. Gel-isolate the two DNA products, combine and ligate.
3. Amplify the spliced product with flanking primers using PCR.
4. Clone the product into a plasmid.
Methods
First PCR
For the first PCR, 4 primers were designed: Primers A and B flanking the human cofilin 1 gene (Fig 2) and two primers containing Ara-C molecules (Figs 2 and 3)). Primer A-5'-ATActgcagATGGCCGCTGGTGTGGCTGTCTGTG-3'-sense primer of human cofilin 1. Lower case letters represent Pst I sequence and bold letters represent Ala to Ser mutagenesis for down stream usage. Primer Ara-C2-A-5'-GGCATAGCGGCAGTCXXAAAGGTGGCGTAGGGATCG-3'-anti-sense primer that contains two Ara-C molecules (XX) and designed to delete a 20 bp segment from the human cofilin 1 gene (Fig 3). Primer Ara-C2-B-5'-ACTGCCCGTTATGCXXTCTATGATGCAACCTATGAG-3'-sense primer that contains two Ara-C molecules (Fig 3). Additionally, two primers containing only one Ara-C molecule insertion were synthesized. Primer B-5'-CAActcgagGGCTGCCAGATGCTCCAGGCAGG-3'-anti-sense primer of the 3' end of human cofilin 1 gene. Lower case letters represent the sequence for the Xho I gene. In the first PCR, Primer A was used with primer Ara-C2-A, and Primer Ara-C2-B was used with primer B. In the second PCR, primer A was used with primer B (see also Fig 2).
Ara-C primers were phosphorylated for 30 min at room temperature using T4 polynucleotide kinase (Invitrogen, Burlington, ON), followed by inactivation at 65°C for 10 min, and used for PCR with no further purification. PCR was performed with corresponding primers (see above and Fig 2, 3) using 1 U Pfu polymerase (Stratagene, La Jolla, CA) or 1 U of Taq polymerase (Sigma, Oakville, ON), and plasmid pOTB7 containing the human cofilin 1 gene as template (ATCC, Manassas, VA). PCR conditions were as follows: heating to 94°C for 5 min; 40 cycles of: 94°C, 55°C and 72°C each for 30 seconds; final elongation for 7 min. PCR products were gel-isolated using MinElute Plasmid Purification Kits (Qiagen, Mississauga, ON). Corresponding segments to be spliced were combined (total of four tubes) and ligated for 30 min with 400 U, or five min with 200 U of T4 ligase (NEB, Pickering, ON) followed by inactivation for 10 min at 65°C.
Second PCR
Two μl of the ligase reaction were amplified by the second PCR with Taq or Pfu polymerase using the sense primer A, and the anti-sense primer B. Conditions for the second PCR were similar to those of the first PCR. The PCR products were purified (Qiagen). Alternatively, the PCR products were subjected to double digestion with PstI and XhoI followed by ligation into plasmid pcDNA3.1Zeo+ (Invitrogen). One μl of ligation reaction was used to transform 20 μl competent cells (DH5α; Invitrogen), using a short procedure: competent cells were incubated for 5 min on ice, and heat-shocked by immediate plating on pre-warmed (37°C) agar plates. Plasmids were prepared using Fast Plasmid Mini Kit (Eppendorf, Mississauga, ON), and sequenced using the T7 primer.
DNA sequencing
The products of PCR, as well as the products that were cloned into plasmid pcDNA3.1Zeo+ were sequenced in both directions, utilizing primers A and B, or primer T7, respectively (Hospital for Sick Children).
Authors' contributions
MA conceived and designed the study, performed the experiments and drafted the manuscript. NMG carried out some of the experiments, participated in critical evaluation and drafted the manuscript. MS provided general guidance, coordination and funding for the study, and drafted the manuscript.
Acknowledgements
This study was supported by CIHR grant MOP-15071. NMG was supported by a CIHR M.D./Ph.D. studentship.
Figures and Tables
Figure 1 Structure of cytidine and its derivatives. The derivatives featured in this figure vary in their sugar substitutes. Note that in Cytosine Arabinose (Ara-C), the arabinose sugar contains hydroxyl groups in positions 3 and 5 in a similar orientation to native ribose, thus permitting reaction with other nucleotides in DNA synthesis. However, the hydroxyls in positions 2 and 3 are in the trans orientation. Comparing position 2 on the arabinose ring to that of 2-deoxyribose reveals that the hydrogen in 2-deoxyribose, that is in trans configuration to hydroxyl 3, is replaced by the hydroxyl group found on arabinose. It should be emphasized that there are two types of polymerization arrest: a. Chain termination- the nucleotide is incorporated into the nascent DNA strand and synthesis is stalled because no new nucleotide is added. Dideoxy derivatives stall elongation after incorporation into the nascent DNA strand because they do not have hydroxl in position 3. Arabinose nucleotides also belong to this group, but they offer only partial stalling [11]. b. Template-dependent termination-nucleotides already incorporated in the DNA (e.g. in primers) are able to stall polymerization when the polymerase reads the template. It is believed that due to stereo restraints, the polymerase falls off the template. The frequency of this event determines the efficiency of the stalling. Arabinose derivatives belong to this group. The property of template-dependent termination of Ara-C was utilized in this study to create 5' overhangs in the first PCR. However, since the template-dependent termination by Ara-C is moderate, it was utilized for the amplification in the second PCR.
Figure 2 Schematic representation of run through stop DNA mutagenesis and splicing technique with Ara-C. In this example two pieces of DNA are to be spliced (5' and 3' DNA segments) and mutated with an insertion of additional DNA. The 5' segment is amplified using PCR primers A (sense) and Ara-C2-A (anti-sense). Primer Ara-C2-A is designed to contain hybrid DNA with two adjacent molecules of Ara-C to stall polymerisation and produce a 5' overhang. Mutational addition is also incorporated into this primer. (Note that in this paper we created a mutational deletion in the human cofilin 1 gene, but here for illustration purpose, we describe a mutational addition). The 3' segment is amplified using PCR primers Ara-C2-B (sense) and B (anti-sense). Primer Ara-C2-B contains overlapping sequence with primer Ara-C2-A, and 2 molecules of Ara-C are incorporated to stall polymerization and produce a 5' overhang that is complementary to the overhang in Ara-C2-A. Both Ara-C primers are phosphorylated for down stream ligation. Since two adjacent Ara-C molecules produce moderate termination, PCR products contain a mixture of 5' overhang and blunt end DNA. Each PCR product is gel-isolated and subjected to short ligation, where cohesive end ligation is predominant. A portion of the ligation reaction is then subjected to a second PCR reaction, using primers A and B that span the entire mutated chimeric DNA. As mentioned above, 2 Ara-C molecules are moderate polymerization terminators. This assures that at the first round of the second PCR, the polymerase will run through the Ara-C in the template and incorporate native dGMP, that will ensure in turn proper polymerization in the next rounds and a product that will contain the native dCMP. For cloning purposes of the final PCR product, primers A abd B can include restriction sites (as used in this study). Alternatively, by using Taq polymerase in the second PCR reaction, the product can be cloned into TA cloning plasmids. Another alternative is to design primers A and B to contain at least 2 molecules of Ara-C to produce 5' overhangs to match cloning plasmids.
Figure 3 Ara-C primer assignment. Shown is the double-stranded DNA segment of human cofilin 1 gene that was used for mutagenesis. Capital letters and arrows represent primers containing Ara-C molecules. Lower case letters represent deleted nucleotides, achieved with primer Ara-C2-A (broken line). Xs in primers denote Ara-C molecules that replace the original deoxy cytidine molecules. Note that the 5' end of primer Ara-C2-A was designed to produce an overhang, that restricts ligation to the 3' segment of the PCR product (see also Fig 2) only if termination occurred after the first Ara-C molecule. The 5' segment of primer Ara-C2-B was designed to produce an overhang that restricts ligation to the 5' segment of the PCR product (see Fig 2) only if termination occurred after two Ara-C molecules. Additionally, two primers containing only one Ara-C molecule insertion were synthesized (not shown).
==== Refs
Ho SN Hunt HD Horton RM Pullen JK Pease LR Site-directed mutagenesis by overlap extension using the polymerase chain reaction Gene 1989 77 51 59 2744487 10.1016/0378-1119(89)90358-2
Kammann M Laufs J Schell J Gronenborn B Rapid insertional mutagenesis of DNA by polymerase chain reaction (PCR) Nucleic Acids Res 1989 17 5404 2548160
Ailenberg M Silverman M Description of a one step staggered reannealing method for directional cloning of PCR-generated DNA using sticky-end ligation without employing restriction enzymes Biochem Molec Biol Int 1996 39 771 779 8843346
Ailenberg M Silverman M Site-directed mutagenesis using a one step staggered reannealing method without employing restriction enzymes BioTechniques 1997 22 624 630 9105609
Gal J Schnell R Szekeres S Kalman M Directional cloning of native PCR products with performed sticky ends (Autosticky PCR) Mol Gen Genet 1999 260 569 573 9928936 10.1007/s004380050930
Coljee VW Murray HL Donahue WF Jarrel KA Seamless gene enginnering using RNA- and DNA- overhang cloning Nat Biotechnol 2000 18 789 791 10888851 10.1038/77363
Donahue WF Turczyk BM Jarrell KA Rapid gene cloning using terminator primers and modular vectors Nucleic Acids Res 2002 30 e95 12235397 10.1093/nar/gnf094
Gee TS Yup K-P Clarkson BD Treatment of adult acute leukemia with arabinosylcytosine and thioguanine Cancer 1969 23 1019 1032 5252232
Furth JJ Cohen SS Inhibition of mammalian DNA polymerase by the 5'-triphosphate of 1-beta arabinofuranosylcytosine and the 5'-triphosphate of 9-beta-d-arabinofuranoxyladenine Cancer Res 1968 28 2061 2067 5754840
Fridland A Inhibition of deoxyribonucleic acid chain initiation: a new mode of action for 1-b-D arabinofuranosylcytosine in human lymphoblasts Biochemistry 1977 16 5308 5312 921933 10.1021/bi00643a023
Sanger F Nicklen S Coulson AR DNA sequencing with chain terminating inhibitors Proc Natl Acad Sci USA 1977 74 5463 5467 271968
|
16137330
|
PMC1208860
|
CC BY
|
2021-01-04 16:02:58
|
no
|
BMC Biotechnol. 2005 Sep 1; 5:23
|
utf-8
|
BMC Biotechnol
| 2,005 |
10.1186/1472-6750-5-23
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-1001609296510.1186/1471-2407-5-100DebateModern classification of neoplasms: reconciling differences between morphologic and molecular approaches Berman Jules [email protected] U.S. National Cancer Institute, Cancer Diagnosis Program, Bethesda, USA2005 10 8 2005 5 100 100 14 3 2005 10 8 2005 Copyright © 2005 Berman; 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
For over 150 years, pathologists have relied on histomorphology to classify and diagnose neoplasms. Their success has been stunning, permitting the accurate diagnosis of thousands of different types of neoplasms using only a microscope and a trained eye. In the past two decades, cancer genomics has challenged the supremacy of histomorphology by identifying genetic alterations shared by morphologically diverse tumors and by finding genetic features that distinguish subgroups of morphologically homogeneous tumors.
Discussion
The Developmental Lineage Classification and Taxonomy of Neoplasms groups neoplasms by their embryologic origin. The putative value of this classification is based on the expectation that tumors of a common developmental lineage will share common metabolic pathways and common responses to drugs that target these pathways. The purpose of this manuscript is to show that grouping tumors according to their developmental lineage can reconcile certain fundamental discrepancies resulting from morphologic and molecular approaches to neoplasm classification.
In this study, six issues in tumor classification are described that exemplify the growing rift between morphologic and molecular approaches to tumor classification: 1) the morphologic separation between epithelial and non-epithelial tumors; 2) the grouping of tumors based on shared cellular functions; 3) the distinction between germ cell tumors and pluripotent tumors of non-germ cell origin; 4) the distinction between tumors that have lost their differentiation and tumors that arise from uncommitted stem cells; 5) the molecular properties shared by morphologically disparate tumors that have a common developmental lineage, and 6) the problem of re-classifying morphologically identical but clinically distinct subsets of tumors. The discussion of these issues in the context of describing different methods of tumor classification is intended to underscore the clinical value of a robust tumor classification.
Summary
A classification of neoplasms should guide the rational design and selection of a new generation of cancer medications targeted to metabolic pathways. Without a scientifically sound neoplasm classification, biological measurements on individual tumor samples cannot be generalized to class-related tumors, and constitutive properties common to a class of tumors cannot be distinguished from uninformative data in complex and chaotic biological systems. This paper discusses the importance of biological classification and examines several different approaches to the specific problem of tumor classification.
==== Body
Background
Classifications provide a simplified view of a knowledge domain, with members of the domain grouped in a class hierarchy. Members of a class share one or more common features and inherit the class features of their ancestral classes [1,2]. Scientists use classifications to discover and test generalizable methods and properties that may apply to members of a class and their descendants. Much of the following discussion relating to the purpose and properties of classifications was inspired by Ernst Mayr, one of the leading evolutionary biologists of the past century [1]. Readers may have somewhat different views on the general subject of classification, but the views presented here are representative of a widely accepted approach developed by biologists.
The subject of tumor classification is made confusing by a variety of commonly held notions about the meaning and purposes of modern classifications [1]. Pathologists typically refer to anatomic tumor classifications when they are more accurately referring to lists of primary tumors that are known to occur at a particular location [3-7]. A list of tumors occurring at a body site is not a classification because it includes tumors that are biologically, clinically, and histologically unrelated. Although often referred to as World Health Organization (WHO) "classifications", the WHO accurately titles their organ-based lists of neoplasms as "Histologic Typings" for the different organs [8-12].
In the past decade, molecular biologists have tried to classify tumors based on grouping together tumor samples that share similar gene expression profiles [13-16]. The ability to separate tumors into groups is not equivalent to separating tumors into classes because the groups may represent expected variations of behavior within a single tumor population. For example, tumor samples of a particular type of tumor may contain groups that are separable based on proliferation rate, cell death rate, size, invasiveness, dominance of glycolytic enzyme pathways, etc. Variant groups within a population do not qualify as classes if it can be shown that the differences between the groups can be accounted for by transient differences in a tumor's biology. If a slow-growing tumor becomes a fast-growing tumor over time, or if a single tumor has foci of slow growth and fast growth, then the tumor cannot be classed by its rate of growth. A key principle in classification is that classes are intransitive (i.e. instances of a class never change their classs) [1]. Carcinomas never become lymphomas and lymphomas never become carcinomas. Grouping tumors by shared gene expression profiles may indicate that a certain tumor shares a similar profile with another tumor (for a chosen set of expressed genes), but it does not guarantee an intransitive classification of neoplasms.
2004 marked the introduction of a classification of tumors based on developmental lineage [2] similar to that proposed by pathologists in the mid-20th century [17]. The rationale for the new classification is that tumor cells will tend to use metabolic pathways inherited from their ancestral cells within their developmental lineage [see: Figure 1]. As pathway-specific targets for cancer treatment become available for clinical trials, it may prove efficacious to test these agents on tumors that have a common lineage [18]. Also in 2004, a comprehensive taxonomy of neoplasms was created by expanding the NCI-Thesaurus [19-21]. The new taxonomy contains over 140,000 names of neoplasms and is included as a supplemental file with this manuscript [see Additional file 1]. This taxonomy was ported into the new classification of neoplasms to create "The developmental lineage classification and taxonomy of neoplasms," hereinafter called "The developmental classification"[19].
In addition to providing a useful nomenclature for neoplasms, the taxonomy reconciles differences between morphology-based classifications of neoplasms (favored by pathologists) and newly emerging classifications based on genomic characterizations of tumors [22-24]. Specific features of the developmental classification and taxonomy are [2]:
1. The classification is comprehensive (e.g. every tumor of man can be placed somewhere within the classification, which is the largest listing of neoplasm terms [19]. It is available at no cost in XML and flat-file formats. In either format, each term is annotated with its complete ancestral lineage. The most recent version of the nomenclature is made available at the Association for Pathology Informatics download page [25].
2. The classification is simple. One of the purposes of a classification is to drive down the complexity that exists when the domain taxonomy is large. The entire classification is described by under 40 classifiers.
3. The classification is based on biologic principle and is not determined by an artificial construct such as medical specialty (e.g. dermatologic neoplasms) or by anatomically vague regions (e.g. head and neck tumors) or by any single taxon (e.g. epithelial versus spindled morphology)
4. The classification is represented as an eXtensible Markup Language (XML) document that permits data integration between heterogeneous biomedical databases.
5. The classification does not invalidate existing diagnoses found in pathology reports. The medicolegal importance of this feature cannot be exaggerated. This relieves pathologists from reviewing all their prior cases and re-diagnosing them in conformance with a new classification.
The developmental classification is not based on morphologic or genomic properties of tumors. It is based on developmental lineage. However, the class instances (i.e., tumors) all have morphologic and genetic properties. Genetic changes that control particular pathways leading to a particular morphologic phenotype may have class specificity within the developmental classification [26]. Consequently, the developmental classification can be examined to determine how it deals with apparent discrepancies between the morphologic and molecular classifications of neoplasms.
In this study, six fundamental issues in tumor classification are described that exemplify the growing rift between morphologic and molecular approaches to tumor classification: 1) the separation between epithelial and non-epithelial tumors; 2) the grouping of tumors based on shared cellular functions; 3) the distinction between germ cell tumors and pluripotent tumors of non-germ cell origin; 4) the distinction between tumors that have lost their differentiation and tumors that arise from uncommitted stem cells; 5) the molecular properties shared by morphologically disparate tumors that have a common developmental lineage, and 6) the problem of re-classifying morphologically identical but clinically distinct subsets of tumors. The discussion of these issues in the context of describing different methods of tumor classification is intended to underscore the clinical value of a robust tumor classification.
The purpose of this manuscript is to show that grouping tumors according to their developmental lineage can reconcile certain fundamental discrepancies resulting from morphologic and molecular approaches to neoplasm classification.
Discussion
The separation between epithelial and non-epithelial tumors
The separation of tumors into two large groups, epithelial and non-epithelial, is a popular device [27-31]. It has certain practical advantages for clinicians. If a patient has a brain mass that is stereotactically biopsied, the pathologist can make a quick, intra-operative consultation based on a small sampling of the specimen, reporting that the tumor is epithelial or non-epithelial. If it's epithelial, the tumor is unlikely to be a primary brain tumor. It is much more likely to be a metastatic lesion from a primary carcinoma that arose somewhere else in the body. The treatment of a primary brain tumor is different from the treatment of a metastatic tumor to the brain. Hence, the intra-operative assessment of epithelial morphology can be very helpful to the surgeon.
Despite the clinical utility, the separation of tumors into epithelial and non-epithelial subtypes does not provide a consistent dichotomy for tumor classification. Experts in the field of classification consider it poor form to create a class based on criteria of exclusion. Classifications are intended to identify properties common to classes [1]. There is seldom much value in assigning class membership based on the absence of a property found in another class.
In the specific case of a morphologic separation of tumors into epithelial and non-epithelial classes, a problem arises when experts do not agree on the qualifications of epithelial tumors. Some might argue that epithelial tumors are tumors that arise from an epithelium (e.g., squamous cell carcinoma) or an epithelial glandular lining (e.g., adenocarcinoma). The qualifying features of an epithelium are controversial. Are mesothelial cells (which cover the serosal surfaces of the pleura and peritoneum) epithelial cells? Mesotheliomas arising from coelomic epithelium are classed as [non-epithelial] sarcomas. Are specialized mesothelial cells that cover the serosal surface of the ovaries epithelial? Ovarian tumors arising from specialized coelomic mesothelium, including papillary cystadenocarcinoma, are classed as epithelial tumors.
Some pathologists divide tumors into epithelial and non-epithelial classes based purely on morphology without regard to their tissues of origin. An epithelial cell is a round or polyhedral cell that adheres closely to other tumor cells with desmosomes (cell junctions). Pathologists subscribing to a purely morphologic separation of tumors face an assortment of problems when they try to maintain a dichotomous classification.
1. Some tumors are mixed epithelial and non-epithelial (mixed tumor of salivary gland, fibroadenoma of breast, synovial sarcoma, carcinosarcoma of uterus) [29,32].
2. Some tumors that are typically epithelial can have non-epithelial variants (e.g., sarcomatoid squamous cell carcinoma, spindle cell nevus, sarcomatoid melanoma).
3. Some tumors that are typically non-epithelial can have epithelial morphology (epithelioid leiomyoma) or morphologic features in common with epithelial tumors (clear cell sarcoma of soft parts) [33]
4. Biochemical markers intended to distinguish epithelial fron non-epithelial cells often co-express in epithelial and non-epithelial tumors [30].
Aside from the failure of the epithelial/non-epithelial dichotomy to provide a consistent basis for tumor classification, this approach to tumor classification creates discrepancies with molecular classifications of tumors. Genetic markers characteristic of tumors can occur in both epithelial and non-epithelial cancers. Furthermore, a single oncogene or molecular marker may characterize both epithelial and non-epithelial tumors [15,34,35]. This means that the epithelial/non-epithelial tumor dichotomy conflicts with a molecular classification of tumors.
The developmental classification assigns morphologically diverse tumors to a single class if they have the same developmental lineage. Morphologic attributes (such as epithelial or non-epithelial appearance) can be preserved as tumor annotations.
Classifying tumors based on shared cellular functions
The endocrine tumors exemplify the confusion that occurs when tumors are classified by common function. Examples of endocrine organs include the thyroid gland and the adrenal gland. The U.S. National Cancer Institute lists 8 major endocrine glands and several other tissues that secrete physiologically active hormones [36].
Pituitary gland
Pineal gland
Thyroid gland
Parathyroid gland
Adrenal gland
Pancreatic islets of langerhans
Gonads (testes and ovaries)
Other endocrine glands – thymus, stomach, small intestines, heart, and placenta.
The problem with considering the endocrine tumors as a single class of tumors is that other than a common function (i.e. hormone secretion), tissues of the endocrine system my have nothing in common morphologically, genetically, or developmentally. Some of the endocrine glands have epithelial morphology (e.g. thyroid) others have spindle morphology (e.g. ovarian stroma). Any single endocrine gland is likely to contain several specialized cell types. The pituitary gland is divided into adenohypophysis (glandular part of the pituitary) and neurohypophysis (neural part of pituitary). The thyroid contains follicular cells and small nests of so-called C-cells that produce calcitonin. Thyroid follicular cells have an endodermal origin and thyroid C-cells have a neural crest origin. The adrenal gland has an epithelial cortex and a morphologically distinctive medullary portion. The adrenal cortex derives from mesoderm. The adrenal medulla consists of paraganglia cells derived from neural crest. As judged by embryologic origin and by morphology, the endocrine glands have no common [class] properties.
Within any endocrine organ are a variety of specialized cells that give rise to morphologically and clinically diverse neoplasms. The following list of tumors comprises a published "Classification of thyroid neoplasms" [37].
Follicular adenoma
Papillary adenoma
Atypical adenoma
Papillary carcinoma
Follicular carcinoma
Undifferentiated anaplastic carcinoma
Medullary carcinoma
Squamous cell carcinoma
Metastatic carcinoma
Fibrosarcoma
Angiosarcoma
Lymphoma
Teratoma
Other sarcomas
(oxyphil and clear cell variants listed as variants, not as members of the primary classification)
This classification of thyroid neoplasia includes tumors from several different lineages, including endodermal (e.g., follicular carcinomas of thyroid), neural crest (e.g. medullary carcinoma) mesenchymal (e.g., fibrosarcoma) and germ cell (e.g., teratoma). It seems unlikely that a single carcinogenic process would be responsible for the diverse tumors occurring in the thyroid or that a single metabolic pathway would provide a rational treatment target for all these tumors.
The advantages of classing endocrine neoplasms by lineage, rather than by function, become apparent when examining the co-occurrence of endocrine neoplasias in inherited tumor syndromes. Inherited tumor syndromes have germline aberrations that lead to multiple types of tumors occurring in affected individuals [38]. In many cases, the specific germline mutation or the involved metabolic pathway of an inherited syndrome has been characterized. Inherited disorders often involve tissues derived from a single developmental lineage (the "One disorder – One developmental lineage" rule). The following three generalizations demonstrate the concept:
1. Tumor syndromes that involve endocrine tumors and non-endocrine tumors recruit both endocrine tumors and non-endocrine tumors from a single developmental lineage
Example:
MEN2B – OMIM # 162300 (neural crest)
A single identical point mutation (exon 16) in the catalytic core of the tyrosine kinase domain of the ret gene has been found to be associated with both inherited and de novo MEN2B
Ganglioneuroma (ganglioneuromatosis);
Pheochromocytoma;
Calcitonin secreting medullary thyroid carcinoma;
*Parathyroid hyperplasia/adenoma (probably secondary to calcitonin secretion)
2. Tumor syndromes that exclusively involve endocrine tumors tend to recruit tumors from a single developmental lineage
Example:
MEN1 – OMIM # 131100 (endodermal origin)
The MEN1 gene contains 10 exons and encodes a ubiquitously expressed 2.8-kb transcript. The predicted 610-amino acid protein product is termed menin, a putative tumor suppressor protein inactivated by MEN1 mutations.
Pancreatic islet cell adenoma;
Parathyroid adenoma;
Pituitary adenoma;
Prolactinoma
Glucagonoma
Insulinoma
Vasointestinal peptide tumor
Gastrinoma
Carcinoid tumors
*Adrenocortical adenoma (secondary to Cushing syndrome [39])
3. Tumor syndromes associated with a mutation of a single metabolic pathway tend to recruit endocrine and non-endocrine tumors from a single developmental lineage.
Example:
Complex II mitochondrial pathway-associated tumors (neural crest [40])
Complex II, of which succinate dehydrogenase (EC 1.3.99.1) is a component, has 4 subunits:
The flavoprotein (SDHA; 600857),
No tumors, but is associated with Leigh syndrome of infantile subacute necrotizing encephalopathy [OMIM record 256000]
The iron sulfur protein (SDHB; 185470)
Carotid body tumors (cervical paraganglioma), and multiple extra-adrenal pheochromocytomas [41,42] [OMIM record 115310]
Pheochromocytoma [OMIM record 171300]
The 2 integral membrane proteins
SDHC (602413)
Familial nonchromaffin paragangliomas type 3 [OMIM recrod 605373][43]
SDHD (602690)
Paragangliomas, chemodectomas, carotid body tumors, glomus jugulare tumors [OMIM 168000]
Exceptions occur, and this is to be expected considering the multi-factor and multi-step nature of carcinogenesis. Tumors occurring in the von Hippel-Lindau syndrome (VHL) pose a seeming exception to the "One disorder – One lineage" rule. VHL is is a dominantly inherited familial cancer syndrome predisposing to a variety of malignant and benign neoplasms, most frequently retinal, cerebellar, and spinal hemangioblastoma, renal cell carcinoma, pheochromocytoma, and pancreatic endocrine tumors. It is caused by mutation in the VHL gene [OMIM record 608537]. VHL is characterized by tumors of the mesenchyme consisting of unusual angiomata, including angioma of retina and spinal cord, and hemangioblastoma of cerebellum. Also seen are pheochromocytomas (neural crest), renal cell carcinomas (mesoderm), and pancreatic islet cell tumors (endoderm). There are several possible explanations for this multilineage tumor syndrome. The mutation may involve a general cancer gene that triggers carcinogenic events independent of cell lineage. Alternately, VHL may be a complex disease that can involve multiple mutations targeting different developmental lineages. There is a tendency for specific combinations of tumors to cluster in different VHL families [44]. Both angiomatosis retinae and hemangioblastoma of the CNS occurred in most families, while renal cancer did not occur in VHL families with pheochromocytoma. This may mean that VHL is not a single mutational disorder. Perhaps the simplest hypothesis is that the VHL mutation targets [for carcinogenic transformation] a stem cell that precedes differentiation of the embryonic layers (endoderm, ectoderm, mesoderm and neuroectoderm). The tumors that occur in VHL families may therefore arise from any of the lineages that descend from the [hypothetical] VHL cancer stem cell.
A molecular classification of neoplasms tied to function (e.g., endocrine function) does not serve as a consistent class scaffold. Germline mutations that predispose to cancer do not select for target tissues that share a common function. Selection seems to favor tissues that share a common developmental lineage.
The distinction between germ cell tumors and pluripotent tumors of non-germ cell origin
Nowhere is the dissonance between morphology, molecular biology and developmental biology more striking than in the realm of the germ cell tumors. The germ cell tumors include tumors of diverse morphology, including seminomas (male), dysgerminomas (female), teratomas, embryonal carcinomas, endodermal sinus tumors, and some (but not all!) cases of choriocarcinoma [45]. Germ cell tumors also have diverse molecular markers with cytogenetic changes correllating with age of patient rather than type of germ cell tumor. Pediatric germ cell tumors show imbalances in of chromosome 1 and loss of 6q, while adult germ cell tumors often have an isochromosome 12p or amplification of 12p [45]. Germ cell tumors arise in almost any part of the body (particularly midline locations) and any age (with a pediatric dominance). It was difficult to imagine a way of placing all these tumors into a single class of neoplasms until Teilum suggested that all these tumors had the same cell of origin, the primordial germ cell [46]. The biologic mechanism by which one cellular progenitor can give rise to these seemingly unrelated tumors is still unfolding as a fascinating saga of developmental biology [45].
Germ cells normally follow a narrow developmental path. Male germ cells give rise to spermatocytes. Female germ cells give rise to oocytes. Under normal circumstances, germ cells are not pluripotent and are not related to embryonic stem cells. The only tumors arising directly from neoplastic germ cells are pure seminomas and dysgerminomas (the female equivalent of seminomas). At an early embryonic stage, primordial germ cells undergo a complete erasure of epigenetic programming, a phenomenon that uniquely characterizes these cells. This phenomenon precedes the transformation of primordial germ cells into embryonic stem cells in culture, is the putative mechanism that operates in animals to provide neoplastic precursors of totipotent tumors, and provides a biological rationale for separating germ cell tumors from other embryonic tumors [47-49]. It permits us to think of germ cells and their normal descendants (ova and sperm) as tumors occupying their own sub-class. We can think of the totipotent and embryonic tumors as tumors of embryonic stem cells (not germ cells). These embryonic stem cells may have derived from the phase of germ cell development when epigenetic programming is erased [48,49], but the resulting cells can be separately classified based on their biology and their shared epigenetic properties characteristic of a totipotent phenotype. By separating tumors of embryonic cells from tumors derived from germ cells, certain incongruities are avoided. In the developmental classification, a gestational choriocarcinoma (arising from cytotrophoblasts and syncytiotrophoblasts in the placenta and having no derivation from germ cells) can now be classed separately from seminomas. Furthermore, new drugs that target the germ cell tumors may find that differences in the genomic state of germ cell tumors and embryonic tumors (i.e., methylation patterns) may uncover vulnerabilities that provide new treatment options. Finally, a developmental classification that distinguishes germ cell tumors and embryonic tumors reconciles the fundamental differences between the morphologic and cytogenetic incongruities among these unique tumors.
The distinction between tumors that have lost their differentiation and tumors that arise from embryonic stem cells
Tumors that have grown into large tumor masses that deeply invade surrounding tissues and metastasize to many distant organs tend to have different morphology than tumors that are small and pre-invasive. Tumors in so-called late lesions tend to acquire morphologic features that are not found in non-neoplastic cells. The nucleus is markedly different from the normal nucleus with angulations of the nuclear membrane, splotchy chromaticity within the nucleus, and marked variation in nuclear size and shape from one tumor cell to the next. Cellularity is often high (i.e. more cells in a microscopic field compared with normal tissues). Mitotic figures, which may be abundant in a late tumor, tend to have abnormal mitotic spindles. Cytogenetics performed on late tumors often show aneuploidy with multiple, complex karyotypic abnormalities. In general, late lesions of widely different tumor types often resemble one another more than they may resemble the early cancer from which they arose. Pathologists often refer to tumors that have lost differentiated morphologic features [found in early lesions] as undifferentiated or poorly differentiated malignancies, implying that the tumor cells have lost properties characteristic of differentiated cell lineages.
A number of tumors, particularly in childhood, consist of cells that have no apparent developmental cell lineage. The small round cell tumors of childhood are prototypical examples. These tumor cells tend to have a uniform size within the tumor. They tend to be characterized by very specific balanced translocations leading to the creation of specific fusion genes [50]. These tumors are sometimes referred to as undifferentiated or primitive neoplasms.
The pathologist's concept of an "undifferentiated" tumor has been applied to such biologically diverse tumors as Ewing's tumor and late stage colon carcinomas. This exemplifies the limitations of grouping tumors using a morphologic feature (such as loss of differentiation) that does not distinquish pathogenetic mechanisms. By classifying tumors by their developmental lineage, useful morphologic features (including extent of differentation) can be applied without obscuring the class features that define the different tumor types.
The relationship among morphologically disparate tumors that share a developmental lineage
Epithelial tumors of the kidney and of the uterus have biologic features that separate them from epithelial tumors in other organs. Furthermore, many of the properties of epithelial tumors arising from kidney or from uterus are specifically associated with mesenchymal tumors.
The gene expression profiles of renal epithelial tumors are much more closely related to the gene expression profiles of sarcomas than of epithelial tumors derived from other organs [15]. Several epithelial tumor variants are characterized by specific tranlocations resulting in fusion genes [34,35,51]. Fusion genes are much more characteristic of sarcomas than epithelial tumors [24]. Mesoblastic nephroma, shares an identical fusion marker (TV6- NTRK3) with congenital fibrosarcoma [52]. Several variants of renal epithelial tumors have the same gene fusion marker as alveolar soft part sarcoma [34,35]. A diverse array of epithelial, stromal and mixed epithelial-stromal tumors are known to arise from kidney parenchyma [53]. How is it possible for an epithelial organ to give rise to epithelial and stromal tumors with epithelial tumors characterized by sarcoma molecular markers?
The kidney is an epithelial organ that has mesodermal lineage. No cells in the kidney arise from endoderm or ectoderm, the major embryonic lineages that give rise to most epithelial tumors. In a sense, the kidney is a stromal organ that masquerades as an epithelial organ. The only way to reconcile the discrepancy between morphologically epithelial renal tumors and their sarcomatous molecular features is to recognize that renal tumors can be classed according to their mesodermal lineage [2].
A similar story holds for uterine tumors. The uterus gives rise to adenocarcinomas, sarcomas, and variously named mixed tumors including carcinosarcomas, adenosarcomas, and mixed mullerian tumors with heterologous components. The histogenesis of mixed epithelial and stromal tumors of the uterus has always presented a special intellectual challenge. Both epithelial and non-epithelial tumors of the uterus seem to have an association with tamoxifen therapy [54], suggesting the possibility that a single stem cell targeted by tamoxifen can give rise epithelial and stromal tumors of the uterus.
In most other organs, epithelial cells do not share their genesis with mesenchymal cells. But the uterus, like the kidney, is derived entirely from mesoderm. The uterus is formed from a duct that forms within the mesoderm (the paramesonephric duct). This duct gives rise to the endometrial epithelium as well as the underlying stroma. Consequently, tumors of endometrial and stromal cells share the same lineage in the developmental classification (sub-coelomic ductal). Like the kidney, this classification ignores morphologic differences (epithelial versus mesenchymal) and creates a grouping in concordance with the observed mixed epithelial/stromal manifestations of some uterine tumors. This is another example wherein the developmental classification accommodates morphologic class ambiguity.
The problem of determining when morphologically identical but clinically distinct subsets of tumors constitute new slots in the classification
Whenever new subpopulations of tumors can be delineated, the question of classification arises. If a report shows that a fraction of a certain type of tumor has a special morphologic feature, does the morphologic variant qualify as a new subclass of the tumor? If a tumor can be divided into two distinct clusters based on gene expression profiles, is each cluster a new subclass of the tumor? If patients with a certain tumor can be divided into good prognosis and bad prognosis groups, can the tumor be subclassified based on clinical prognosis? The problem with classifying based on morphologic, molecular or clinical features is that any population can be expected to contain members with features that vary from the population norm. There is a difference between segregating population variants (e.g. fast runners and slow runners, chocolate lovers and chocolate haters) and creating a self-consistent classification. A classification has certain properties that distinguish it from other ways of organizing data. The general rules for classifications have been summarized [1,2]:
1. A classification is a hierarchical grouping, with each group defined by the greatest number of taxa (informative features) that can apply to every instance of the class.
2. Subclasses inherit the properties (shared taxons) of their ancestor classes.
3. Every instance of the knowledge domain must fit into the classification, and every instance and class must have exactly one slot in the classification.
4. Instances of one class are intransitive (e.g. an instance in one class cannot migrate to a different class, but must remain in the same class or a subclass of the same class).
5. A classification is a hypothesis about the fundamental properties of a knowledge domain. The hypotheses must be tested and re-tested and changed when the facts do not fit the model.
If a tumor lacks a morphologic or molecular feature at one point in its development and gains the feature at a later point, the feature cannot determine a new class of tumor. If a tumor has a good prognosis at one point (e.g. before it has metastasized) and a bad prognosis later (e.g. after it has metastasized), then prognostic features associated with metastasis cannot be used to determine a new class of tumor. In general, all new findings about subpopulations of tumors can be considered candidate taxa (i.e., features that characterize a class and distinguish the class from other classes). The full list of class features (items 1–5 above) must be satisfied before a candidate taxon can be used to define a new tumor class.
The future role of morphology and molecular analysis in tumor classification
Morphologic pathology has dominated tumor diagnosis for over 150 years [55]. The demise of morphologic pathology is a long-anticipated event that may never occur. In fact, the morphologic pathologist is a key developer of emerging technologies that promise the end of our dependence on histologic evaluation of lesions. It is the pathologist who prepares, describes and diagnoses the tissue samples used by the molecular biologist. It is the pathologist who collects, organizes and integrates the information in the patient's surgical pathology report with demographic information (age, gender), medical history, tumor staging, and ancillary hospital tests (radiology reports, hematology reports, and past/future tissue reports). In most cases, researchers would have no tumor samples for gene expression profiling if the pathologists had not been able to distinguish the tumor samples by careful morphologic evaluation. In all cases, the clinico-pathologic annotations used by the molecular biologist are generated in whole or in part by surgical pathologists. It has been noted that, "the pathologist's understanding of anatomic, physiologic, biochemical, immune, and other underlying factors that drive mechanisms of tissue responses to noxious agents turns a bewildering array of gene expression data into focused research programs"[56].
If molecular classification is to replace the morphologic classification of tumors, several seemingly intractable problems must be solved. Cytogenetic abnormalities and gene alterations in tumors co-occur with other abnormalities, and the complex state of molecular abnormalities in tumors makes it very difficult to settle on a set of alterations characteristic of classes of tumors. As an example, balanced translocations play biologic roles in several dozen tumors [24]. Although certain translocations are characteristic of individual tumors, it has proven difficult to generalize that translocations occur in any particular class of tumors. Certainly, characteristic tumor translocations occur more commonly in mesenchymal tumors [24], but such translocations have also been observed in secretory carcinoma of breast [57] and in midline [lung] carcinoma of children and young adults [58]. The notable exception wherein a class of tumors is characterized by a set of translocations is the Ewing's tumor family of tumors [50].
Mitelman has argued that translocations are tissue non-specific, occurring at a frequency related to the overall number of cytogenetic abnormalities found in tumors [59]. If it is difficult to assign classes of tumor to a single type of cytogenetic abnormality, it may be impossible to reach scientific consensus on complex sets of molecular signatures that define groups of tumors. It can be noted that despite numerous projects aimed at classifying tumors with gene expression profiles, no comprehensive classification based on this technology has emerged.
Much of what passes for neoplasm "classification" in the bioinformatics literature is actually the algorithmic ranking of expressed genes that can discriminate one tumor variant from another [13,14,60]. Once candidate molecules (i.e., genes, proteins, and other macromolecules or patterns of these molecules) are found to associate with a particular tumor variant, the pathologist gets a second chance to determine if a morphologic pattern correlates with the molecular property. An example comes from the study of gastrointestinal stromal tumors (GIST). Most GIST tumors have a c-Kit mutation that results in c-Kit protein overexpression [61]. Some GIST tumors lacking c-Kit mutations have a mutation in the platelet-derived growth factor receptor alpha gene [62]. Sakurai and coworkers have examined GIST tumors that stain negatively for CD-117, a marker for c-Kit protein overexpression. Many of these tumors have mutations in platelet-derived growth factor receptor alpha gene and a distinctive histomorphology characterized by myxoid epithelioid tumor cells and tumor infiltration by mast cells [63]. This newly recognized subtype of GIST involved the morphologic re-examination of the tumors following a molecular discovery.
Other examples abound. Secretory carcinoma of breast is an uncommon variant of breast cancer that occurs most frequently in young women. It is characterized by the ETV6-NTRK3 fusion gene [57]. The search and discovery of this molecular marker was accomplished through asynchronous contributions from three biomedical realms: 1) pathologists, who found defined the morphologic subset of breast carcinoma known as secretory carcinoma of breast; 2) oncologists who validated the clinically distinct features of the tumor, and 3) molecular biologists who discovered the translocation that characterized the tumor.
It is a basic assumption of the developmental classification that morphologic and molecular features of tumors will both fall sensibly into classes determined by tumor cell lineage. It is further assumed that pathways with molecular alterations producing a tumor phenotype will tend to operate in all tumors of a developmental class. Finally, it is hoped that morphologic properties associated with the altered pathway will be visible in all class members. Because classifications are hypotheses about the fundamental nature of a knowledge domain, the foundational assumptions of any classification must be continually evaluated and challenged.
Some classifications can be challenged more easily than others. A classification built on a set of continually changing parameters is constantly changing and difficult to evaluate. This is certainly true of a molecular classification, because our knowledge of the field changes almost daily. A few years ago, it was safe to say that all recurrent balanced translocations were a phenomenon of mesenchymal tumors. New findings of recurrent balanced translocations in non-mesenchymal tumors have nullified this class assertion [59]. Morphologists once classified clear cell sarcoma as a type of malignant melanoma, based on finding melanosomes within tumor cells. Recent molecular classification of these tumors clearly distinguish them from cutaneous melanoma. Clear cell sarcomas have characteristic EWS-AFT1 fusion transcript not found in cutaneous melanomas [64]. In addition, BRAF mutations, commonly found in cutaneous melanomas, are absent from clear cell sarcomas [65]. The rapid accumulation of new knowledge about the molecular characteristics of tumors can quickly change classifications built on morphology or molecular biology. Pathologists seem to be putting this tumor back into the mesenchymal class of neoplasms [66].
The developmental classification is built on a foundation of developmental biology that was improved over many decades by thousands of scientists. Our understanding of embryologic lineage has changed very little over the past half century, and a classification based on developmental biology permits tumors to be assigned to well-defined classes. Recent advances in embryology have shown that somatic DNA has lineage-specific epigentic modifications that occur throughout development [47,67]. This means that the developmental lineage of tumors may be measurable and refinable with new techniques that correlate patterns of epigenetic modifications (e.g. methylation) with lineage. In a recent paper by Kho and coworkers [68], the authors developed a method that projects gene expression profiles of tumors onto a mouse developmental sequence. Human medulloblastoma most closely matched the gene expression profile of postnatal day 5 mouse cerebellum. Although this study examined only a few tumors, it described a method that allows any human tumor to be matched against a library of gene expression profiles collected from normal tissues at different stages of development.
Summary
A scientifically sound classification of neoplasms will serve as a guide to selecting a new generation of cancer medications targeted to metabolic pathways specific for particular classes of tumors. Without a classification of tumors, biological measurements on individual tumor samples cannot be generalized to other tumors, and constitutive properties common to a class of tumors cannot be distinguished from uninformative data collected from a complex and chaotic biological system. Morphology, even in the post-genomic era, has enormous value in the realm of taxon discovery. Using morphologic examination, pathologists have discovered previously unrecognized morphologic features that are diagnostic for new tumors or new clinical variants of known tumors that have characteristic molecular profiles. By classifying tumors by lineage, problems arising from molecular and morphologic tumor classifications can be resolved or posed as testable hypotheses.
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:
Supplementary Material
Additional File 1
Neoplasia classification structure (XML version) Neoclxml.gz is a compressed (gzipped) XML file. The downloaded file should be renamed neoclxml.gz so that the .gz suffix can be recognized by unzip utilities. Unzip the file (using a free, open source utility such as gunzip.exe [69], or a proprietary utility such as Winzip). Once unzipped, the file should be renamed neocl.xml, so that it will have an .xml suffix. If the file is too large for viewing on your web browser, it can be viewed on plain-text word processors.
Click here for file
Acknowledgements
The author wrote this paper as part of his activities as an employee of the U.S. Government. The opinions expressed herein are those of the author and do not represent the opinions of the U.S. Government or any of its agencies.
Figures and Tables
Figure 1 Developmental Lineage Classification Schema [2].
==== Refs
Mayr E The growth of biological thought: diversity, evolution and inheritance 1982 Cambridge: Belknap Press
Berman JJ Tumor classification: molecular analysis meets Aristotle BMC Cancer 2004 4 10 15113444 10.1186/1471-2407-4-10
Willemze R Jaffe ES Burg G Cerroni L Berti E Swerdlow SH Ralfkiaer E Chimenti S Diaz-Perez JL Duncan LM Grange F Harris NL Kempf W Kerl H Kurrer M Knobler R Pimpinelli N Sander C Santucci M Sterry W Vermeer MH Wechsler J Whittaker S Meijer CJ WHO-EORTC classification for cutaneous lymphomas Blood 2005 3768 3785 Feb 3; 2005
Jaffe ES Harris NL Stein H Vardiman JW World Health Organization Classification of Tumours Pathology and Genetics of Tumours of the Haematopoietic and Lymphoid Tissues 2001 Lyon: IARC Press
Kleihues P Sobin LH World Health Organization classification of tumors Cancer 2000 88 2887 2887 10870076 10.1002/1097-0142(20000615)88:12<2887::AID-CNCR32>3.0.CO;2-F
Younossian AB Brundler MA Totsch M Feasibility of the new WHO classification of pulmonary neuroendocrine tumours Swiss Med Wkly 2002 132 535 540 12508138
International Classification of Diseases for Oncology, (ICD-O-3)
Travis WD Colby TV Corrin B Shimosato Y Brambilla E Shimosato Y Brambilla E World Health Organization, Histological Typing of lung and pleural tumours 1999 3 Berlin: Springer
Mostofi FK Sobin LH Torloni H Histological Typing of Urinary Bladder Tumours 1973 Geneva: Springer
Wahi PN Cohen B Luthra UK Torloni H Histological Typing of Oral and Oropharyngeal Tumours 1971 Geneva: Springer
Poulsen HE Taylor CW Sobin LH Histological Typing of Female Genital Tract Tumours 1993 Geneva: Springer
Williams ED Siebenmann RE Sobin LH Histological Typing of Endocrine Tumours 1980 Geneva: Springer
Ramaswamy S Tamayo P Rifkin R Mukherjee S Yeang C Angelo M Ladd C Reich M Latulippe E Mesirov JP Poggio T Gerald W Loda M Lander ES Golub TR Multiclass cancer diagnosis using tumor gene expression signatures Proc Natl Acad Sci USA 2001 98 15149 15154 11742071 10.1073/pnas.211566398
Golub TR Slonim DK Tamayo P Huard C Gaasenbeek M Mesirov JP Coller H Loh ML Downing JR Caligiuri MA Bloomfield CD Lander ES Molecular classification of cancer: class discovery and class prediction by gene expression monitoring Science 1999 286 531 537 10521349 10.1126/science.286.5439.531
Wang E Lichtenfels R Bukur J Ngalame Y Panelli MC Seliger B Marincola FM Ontogeny and oncogenesis balance the transcriptional profile of renal cell cancer Cancer Res 2004 64 7279 7287 15492247
Hans CP Weisenburger DD Greiner TC Gascoyne RD Delabie J Ott G Muller-Hermelink HK Campo E Braziel RM Jaffe ES Pan Z Farinha P Smith LM Falini B Banham AH Rosenwald A Staudt LM Connors JM Armitage JO Chan WC Confirmation of the molecular classification of diffuse large B-cell lymphoma by immunohistochemistry using a tissue microarray Blood 2004 103 275 282 14504078 10.1182/blood-2003-05-1545
Willis RA Borderland of Embryology and Pathology 1958 London: Butterworth
Liu MC Marshall JL Pestell RG Novel strategies in cancer therapeutics: targeting enzymes involved in cell cycle regulation and cellular proliferation Curr Cancer Drug Targets 2004 4 403 24 15320717 10.2174/1568009043332907
Berman JJ Tumor taxonomy for the developmental lineage classification of neoplasms BMC Cancer 2004 4 88 15571625 10.1186/1471-2407-4-88
De Coronado S Haber MW Sioutos N Tuttle MS Wright LW NCI Thesaurus: Using Science-based Terminology to Integrate Cancer Research Results Medinfo 2004 2004 33 37
Golbeck J Fragoso G Hartel F Hendler J Oberthaler J Parsia B The National Cancer Institute's Thesaurus and Ontology J of Web Semantics 2003 1 75 80
Director's challenge: toward a molecular classification of tumors
The Wellcome Trust Sanger Institute Census of Cancer Genes
Futreal A Coin L Marshall M Down T Hubbard T Wooster R Rahman N Stratton MR A Census of Human Cancer Genes Nature Reviews Cancer 2004 4 177 183 14993899 10.1038/nrc1299
Association for Pathology Informatics informatics resources
Xu G Zhang W Bertram P Zheng XF McLeod H Pharmacogenomic profiling of the PI3K/PTEN-AKT-mTOR pathway in common human tumors Int J Oncol 2004 24 893 900 15010827
Huang RL Chao CF Ding DC Yu CP Chang CC Lai HC Yu MH Liu HS Chu TY Multiple epithelial and nonepithelial tumors in hereditary nonpolyposis colorectal cancer: characterization of germline and somatic mutations of the MSH2 gene and heterogeneity of replication error phenotypes Cancer Genet Cytogenet 2004 153 108 114 15350299 10.1016/j.cancergencyto.2004.01.003
Treilleux T Mignotte H Clement-Chassagne C Guastalla P Bailly C Tamoxifen and malignant epithelial-nonepithelial tumours of the endometrium: report of six cases and review of the literature Eur J Surg Oncol 1999 25 477 482 10527595 10.1053/ejso.1999.0682
Suzuki S Kasashima F Nakanishi I Oda Y An unusual case of benign thyroid tumour consisting of epithelial and nonepithelial components Virchows Arch 1999 434 235 239 10190303 10.1007/s004280050333
Azumi N Battifora H The distribution of vimentin and keratin in epithelial and nonepithelial neoplasms. A comprehensive immunohistochemical study on formalin- and alcohol-fixed tumors Am J Clin Pathol 1987 88 286 296 2443000
Hittelman WN Genetic instability in epithelial tissues at risk for cancer Ann N Y Acad Sci 2001 952 1 12 11795428
Nishio J Iwasaki H Ishiguro M Ohjimi Y Isayama T Naito M Kikuchi M Identification of syt-ssx fusion transcripts in both epithelial and spindle cell components of biphasic synovial sarcoma in small tissue samples isolated by membrane-based laser microdissection Virchows Arch 2001 439 152 157 11561755 10.1007/s004280100428
Ewing CA Miller MJ Chhieng D Lin O Nonepithelial malignancies mimicking primary carcinoma of the breast Diagn Cytopathol 2004 31 352 357 15468118 10.1002/dc.20147
Argani P Lae M Hutchinson B Reuter VE Collins MH Perentesis J Tomaszewski JE Brooks JS Acs G Bridge JA Vargas SO Davis IJ Fisher DE Ladanyi M Renal Carcinomas With the t(6;11)(p21;q12): Clinicopathologic Features and Demonstration of the Specific Alpha-TFEB Gene Fusion by Immunohistochemistry, RT-PCR, and DNA PCR Am J Surg Pathol 2005 29 230 240 15644781 10.1097/01.pas.0000146007.54092.37
Argani P Antonescu CR Illei PB Lui MY Timmons CF Newbury R Reuter VE Garvin AJ Perez-Atayde AR Fletcher JA Beckwith JB Bridge JA Ladanyi M Primary renal neoplasms with the ASPL-TFE3 gene fusion of alveolar soft part sarcoma: a distinctive tumor entity previously included among renal cell carcinomas of children and adolescents Am J Pathol 2001 159 179 92 11438465
Endocrine glands and their hormones
Johnson RL Hartmann WH Silverberg SG The thyroid Principles and practice of surgical pathology 1983 New York: Wiley 1415 1441
OMIM online Mendelian inheritance in man
Skogseid B Larsson C Lindgren PG Kvanta E Rastad J Theodorsson E Wide L Wilander E Oberg K Clinical and genetic features of adrenocortical lesions in multiple endocrine neoplasia type 1 J Clin Endocrinol Metab 1992 75 76 81 1352309 10.1210/jc.75.1.76
Baysal BE Willett-Brozick JE Lawrence EC Drovdlic CM Savul SA McLeod DR Yee HA Brackmann DE Slattery WH Myers EN Ferrell RE Rubinstein WS Prevalence of SDHB, SDHC, and SDHD germline mutations in clinic patients with head and neck paragangliomas J Med Genet 2002 39 178 183 11897817 10.1136/jmg.39.3.178
DeAngelis LM Kelleher MB Post KD Fetell MR Multiple paragangliomas in neurofibromatosis: a new neuroendocrine neoplasia Neurology 1987 37 129 133 2879257
Astuti D Latif F Dallol A Dahia PLM Douglas F George E Skoldberg F Husebye ES Eng C Maher ER Gene mutations in the succinate dehydrogenase subunit SDHB cause susceptibility to familial pheochromocytoma and to familial paraganglioma Am J Hum Genet 2001 69 49 54 11404820 10.1086/321282
Niemann S Muller U Engelhardt D Lohse P Autosomal dominant malignant and catecholamine-producing paraganglioma caused by a splice donor site mutation in SDHC Hum Genet 2003 113 92 94 12658451
Neumann HPH Wiestler OD Clustering of features of von Hippel-Lindau syndrome: evidence for a complex genetic locus Lancet 1991 337 1052 1054 1673491 10.1016/0140-6736(91)91705-Y
Schneider DT Schuster AE Fritsch MK Hu J Olson T Lauer S Gobel U Perlman EJ Multipoint imprinting analysis indicates a common precursor cell for gonadal and nongonadal pediatric germ cell tumors Cancer Res 2001 61 7268 7276 11585765
Teilum G Classification of endodermal sinus tumour (mesoblatoma vitellinum) and so-called "embryonal carcinoma" of the ovary Acta Pathol Microbiol Scand 1965 64 407 429 5318716
Allegrucci C Thurston A Lucas E Young L Epigenetics and the germline Reproduction 2005 129 137 149 15695608 10.1530/rep.1.00360
Donovan PJ de Miguel MP Turning germ cells into stem cells Curr Opin Genet Dev 2003 13 463 471 14550410 10.1016/j.gde.2003.08.010
Matsui Y Zsebo K Hogan BL Derivation of pluripotential embryonic stem cells from murine primordial germ cells in culture Cell 1992 70 841 847 1381289 10.1016/0092-8674(92)90317-6
Riley RD Burchill SA Abrams KR Heney D Sutton AJ Jones DR Lambert PC Young B Wailoo AJ Lewis IJ A systematic review of molecular and biological markers in tumours of the Ewing's sarcoma family Eur J Cancer 2003 39 19 30 12504654 10.1016/S0959-8049(02)00500-2
Sidhar SK Clark J Gill S Hamoudi R Crew AJ Gwilliam R Ross M Linehan WM Birdsall S Shipley J Cooper CS The t(X;1)(p11.2;q21.2) translocation in papillary renal cell carcinoma fuses a novel gene PRCC to the TFE3 transcription factor gene Human Molecular Genetics 1996 5 1333 1338 8872474 10.1093/hmg/5.9.1333
Knezevich SR Garnett MJ Pysher TJ Beckwith JB Grundy PE Sorensen PH ETV6- NTRK3 gene fusions and trisomy 11 establish a histogenetic link between mesoblastic nephroma and congenital fibrosarcoma Cancer Res 1998 58 5046 5048 9823307
Michal M Hes O Bisceglia M Simpson RH Spagnolo DV Parma A Boudova L Hora M Zachoval R Suster S Mixed epithelial and stromal tumors of the kidney. A report of 22 cases Virchows Arch 2004 445 359 67 15322873 10.1007/s00428-004-1060-y
Hubalek M Ramoni A Mueller-Holzner E Marth C Malignant mixed mesodermal tumor after tamoxifen therapy for breast cancer Gynecol Oncol 2004 95 264 266 15385144 10.1016/j.ygyno.2004.06.039
Virchow R Die Cellularpathologie in ihrer Begrundung auf physiologische und pathologische Gewebelehre 1858 Berlin: August Hirschwald
Boorman GA Anderson SP Casey WM Brown RH Crosby LM Gottschalk K Easton M Ni H Morgan KT Toxicogenomics, drug discovery, and the pathologist Toxicol Pathol 2002 30 15 27 11890469 10.1080/01926230252824671
Makretsov N He M Hayes M Chia S Horsman DE Sorensen PH Huntsman DG A fluorescence in situ hybridization study of ETV6-NTRK3 fusion gene in secretory breast carcinoma Genes Chromosomes Cancer 2004 40 152 157 15101049 10.1002/gcc.20028
French CA Kutok JL Faquin WC Toretsky JA Antonescu CR Griffin CA Nose V Vargas SO Moschovi M Tzortzatou-Stathopoulou F Miyoshi I Perez-Atayde AR Aster JC Fletcher JA Midline carcinoma of children and young adults with NUT rearrangement J Clin Oncol 2004 22 4135 4139 15483023 10.1200/JCO.2004.02.107
Mitelman F Johansson B Mertens F Fusion genes and rearranged genes as a linear function of chromosome aberrations in cancer Nat Genet 2004 36 331 334 15054488 10.1038/ng1335
Peterson C Ringner M Analyzing tumor gene expression profiles Artif Intell Med 2003 28 59 74 12850313 10.1016/S0933-3657(03)00035-6
Berman J O'Leary TJ Gastrointestinal stromal tumor workshop Hum Pathol 2001 32 578 582 11431711 10.1053/hupa.2001.25484
Medeiros F Corless CL Duensing A Hornick JL Oliveira AM Heinrich MC Fletcher JA Fletcher CD KIT-negative gastrointestinal stromal tumors: proof of concept and therapeutic implications Am J Surg Pathol 2004 28 889 894 15223958
Sakurai S Hasegawa T Sakuma Y Takazawa Y Motegi A Nakajima T Saito K Fukayama M Shimoda T Myxoid epithelioid gastrointestinal stromal tumor (GIST) with mast cell infiltrations: a subtype of GIST with mutations of platelet-derived growth factor receptor alpha gene Hum Pathol 2004 35 1223 1230 15492989 10.1016/j.humpath.2004.07.008
Rakheja D Weinberg AG Tomlinson GE Partridge K Schneider NR Translocation (10;17)(q22;p13): a recurring translocation in clear cell sarcoma of kidney Cancer Genet Cytogenet 2004 154 175 179 15474157 10.1016/j.cancergencyto.2004.02.024
Panagopoulos I Mertens F Isaksson M Mandahl N Absence of mutations of the BRAF gene in malignant melanoma of soft parts (clear cell sarcoma of tendons and aponeuroses) Cancer Genet Cytogenet 2005 156 74 76 15588860 10.1016/j.cancergencyto.2004.04.008
Covinsky M Gong S Rajaram V Perry A Pfeifer J EWS-ATF1 fusion transcripts in gastrointestinal tumors previously diagnosed as malignant melanoma Hum Pathol 2005 36 74 81 15712185 10.1016/j.humpath.2004.10.015
Shiota K Kogo Y Ohgane J Imamura T Urano A Nishino K Tanaka S Hattori N Epigenetic marks by DNA methylation specific to stem, germ and somatic cells in mice Genes Cells 2002 7 961 969 12296826 10.1046/j.1365-2443.2002.00574.x
Kho AT Zhao Q Cai Z Butte AJ Kim JY Pomeroy SL Rowitch DH Kohane IS Conserved mechanisms across development and tumorigenesis revealed by a mouse development perspective of human cancers Genes Dev 2004 18 629 40 15075291 10.1101/gad.1182504
The gzip home page
|
16092965
|
PMC1208861
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Aug 10; 5:100
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-100
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-1031610916710.1186/1471-2407-5-103Research ArticleEnhanced efficacy of gemcitabine in combination with anti-CD20 monoclonal antibody against CD20+ non-Hodgkin's lymphoma cell lines in vitro and in scid mice Smith Mitchell R [email protected] Indira [email protected] Fang [email protected] Coleman [email protected] Department of Medical Oncology Fox Chase Cancer Center 333 Cottman Avenue Philadelphia, PA19111, USA2 Lilly Research Laboratories, Indianapolis, IN, USA2005 18 8 2005 5 103 103 10 3 2005 18 8 2005 Copyright © 2005 Smith 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
Despite exciting new targeted therapeutics against non-Hodgkin's lymphoma (NHL), chemotherapy remains a cornerstone of therapy. While purine nucleoside analogs have significant activity in low grade NHL, the pyrimidine nucleoside analog gemcitabine has been less extensively studied, but has important activity. Use of the anti-CD20 monoclonal antibody rituximab in combination with chemotherapy for B-NHL is becoming prevalent in clinical practice, but has not been extensively studied in pre-clinical models.
Methods
We have tested the activity of gemcitabine ± rituximab in vitro and in scid/human NHL xenograft models. We used two t(14;18)+, CD20+ follicular B cell NHL cell lines, DoHH2 a transformed NHL line and WSU-FSCCL isolated from pleural fluid of a patient with indolent NHL.
Results
Gemcitabine is cytotoxic to DoHH2 and WSU-FSCCL cells in vitro, and the IC50 is 2–3 fold lower in the presence of rituximab. Apoptosis is also enhanced in the presence of rituximab. Clearance of NHL cells from ascites in scid mice is prolonged by the combination, as compared with either agent alone. Most importantly, survival of scid mice bearing human NHL cells is significantly prolonged by the combination of gemcitabine + rituximab.
Conclusion
Based on our pre-clinical data showing prolonged survival of mice bearing human lymphoma cell line xenografts after treatment with gemcitabine + anti-CD20 antibody, this combination, expected to have non-overlapping toxicity profiles, should be explored in clinical trials.
==== Body
Background
Non-Hodgkin's lymphoma (NHL) is increasing in incidence and is now the fifth most common malignancy in the U.S. Despite novel targeted biologic treatment options, chemotherapy remains an important component of therapy. Furthermore, most patients with indolent lymphoma and at least half of all patients with aggressive NHL are not cured [1]. Improved therapeutic approaches are needed.
Gemcitabine is a pyrimidine nucleoside analog with clinical anti-cancer activity. Purine nucleoside analogs such as fludarabine, cladribine and pentostatin have been extensively studied and have significant activity against certain non-Hodgkin's lymphoma subtypes, particularly indolent forms. Though less well studied, an increasing body of data indicates activity of gemcitabine against lymphoma, both Hodgkin's and NHL [2-7]. The precise place of gemcitabine in the therapeutic armamentarium for NHL remains to be elucidated.
The chimeric anti-CD20 monoclonal antibody rituximab is active as a single agent in B cell NHL [8]. In addition, it may sensitize cells to the action of chemotherapeutic and other biologic agents pre-clinically [9-12], as well as in patients [13,14]. While mechanisms of rituximab action include direct apoptotic induction, complement activation and antibody dependent cytotoxicity, which of these is important may depend on the experimental conditions, and the relative importance in patients remains to be determined (reviewed in [15]). The same mechanisms, as well as intracellular signaling [16], may account for chemosensitization, but again the exact means by which this occurs in patients remains to be fully elucidated. One previous report has demonstrated in vitro sensitization of aggressive B cell NHL cell lines to gemcitabine by rituximab [10]. Here we extend these in vitro results to additional human CD20+ lymphoma cell lines that carry the t(14:18) translocation, perhaps more analogous to the clinical use of rituximab. More importantly, we demonstrate that gemcitabine + rituximab enhances survival in vivo in a human B-NHL cell line/scid mouse xenograft model.
Methods
Cell culture and growth assay
Cells are incubated under standard conditions and cell numbers are determined by methyl thiazol tetrazoliumbromide (MTT) assay as before [17]. Briefly, DoHH2 [18], a t(14;18)+ transformed lymphoma cell line, was obtained from Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (DMSZ, German Collection of Microorganisms and Cell Cultures, Braunschweig, Germany). WSU-FSCCL, as previously reported [19], was isolated from pleural fluid of a patient with follicular grade I lymphoma, contains t(14;18), is EBV negative and behaves in scid mice as a more indolent disease [17], although it does contain a c-myc translocation.
Anti-CD20 monoclonal antibody
Chimeric anti-CD20 (clone C2B8, rituximab) obtained from IDEC (San Diego, CA), and gemcitabine (generously supplied by Lilly, Indianapolis, IN) were injected intraperitoneally (ip).
Apoptosis and cell cycle
Cell cycle was analyzed by DNA content per cell, by propidium iodide (PI) staining of nuclei from hypotonically lysed cells [20]. Apoptosis was determined by dual staining of 1 × 105 intact cells in 100 μl calcium binding buffer containing 5 μl of fluoroscein isothiocyanate (FITC)-labeled annexin V (Pharmingen) and 5 μg/ml PI for 15 minutes in the dark, followed by analysis by flow cytometry (FACScan).
Poly(ADP-ribose) polymerase (PARP) cleavage assay
DoHH2 and WSU-FSCCL cells were lysed in cold radioimmunoprecipitation assay (RIPA) buffer [21] containing 100 μg phenylmethanesulfonyl fluoride (PMSF)/ml and 1 μg aprotinin/ml for 30 min on ice, pelleted and the supernatant separated by 4–20% SDS-PAGE. Transfer was to Immobilon-P, blocked with 1% casein-0.04% Tween-20 and probed with anti-PARP C2–10 antibody (Trevigen, Gaithersburg, MD). Secondary antibody was horseradish peroxidase labeled anti-mouse IgG (1:3000) for 30 min, detected by chemiluminescence and Hyperfilm ECL (Amersham, Piscataway, NJ). Densitometry utilized NIH Image 1.61 software.
Scid/human xenograft
Female CB17 scid mice were bred, housed and treated in the Fox Chase Cancer Center Laboratory Animal Facility under an approved protocol. Mice 4–8 weeks old were injected ip with either 1 × 107 WSU-FSCCL cells or 5 × 106 DoHH2 cells. Mice were observed daily and euthanized when they appeared ill. Lymphoma involving diffuse adenopathy, splenomegaly, infiltration of liver and bone marrow, along with ascites developed in untreated mice with each model, at 8–11 weeks with WSU-FSCCL and 4–6 weeks with DoHH2. Cells were collected sequentially from ascites by peritoneal washings and analyzed, or mice were followed for survival. For ascites clearance, mice bearing DoHH2 received 20 μg rituximab intraperitoneally (ip) on day 3 after lymphoma cell injection and/or gemcitabine 120 μg/gm ip on day 4, while mice bearing WSU-FSCCL received 100 μg rituximab intraperitoneally (ip) on day 7 after lymphoma cell injection, and/or gemcitabine 120 μg/gm ip on day 8. For survival, mice bearing DoHH2 received 5 μg rituximab on days 2, 9 and 16 and/or gemcitabine 120 μg/gm ip on days 3, 10 and 17.
Results
Cell growth inhibition by gemcitabine and rituximab
Both DoHH2 and WSU-FSCCL cells are growth inhibited by gemcitabine with an IC50 of 1 nM after 72 hr incubation (Figure 1). Addition of rituximab alone or in combination with gemcitabine had little effect on WSU-FSCCL cells in vitro. In contrast, DoHH2 cells are growth inhibited about 35% by 20 μg/ml rituximab. The maximal rituximab effect is seen at 1 μg/ml. This is accord with data that CD20 sites are saturated at this level, while effective serum levels in patients are felt to be 25 μg/ml. Thus, there is essentially no dose response at clinically achievable levels. This precludes the calculation of synergy by the standard approach [22], as has been concluded by other investigators as well [23]. Because of our prior experience that oligonucleotides can interfere with drug and antibody uptake, we separated addition of the two agents by 4 hours to preclude direct physical interaction of the antibody and drug, however, no effect of order of addition within 4 hours was observed with rituximab and gemcitabine (data not shown).
Cell cycle alteration by gemcitabine and rituximab
Continuous exposure to 2.5 nM gemcitabine led to accumulation of DoHH2 cells in the S phase, as has been generally reported for gemcitabine [24]. WSU-FSCCL cells, however, accumulated in the G2/M phase of the cell cycle (Figure 2). Although rituximab alone had no significant cell cycle effects on either cell line, nor on the S phase accumulation of DoHH2 cells, the combination of gemcitabine + rituximab led to S phase accumulation of WSU-FSCCL cells. Statistical analysis of S phase block in WSU-FSCCL cells comparing 2.5 nM gemcitabine alone versus gemcitabine plus rituximab revealed p < 0.001. Thus, the combination had different cell cycle effects on WSU-FSCCL than did either agent alone. The precise basis of this change is unclear, though presumably involves an alteration in the S phase DNA damage sensor, which appears to involve the ATM gene [25].
Apoptosis induced by gemcitabine and rituximab
Apoptosis was assayed by annexin V staining, which detects the altered location of phosphatidyl serine to the outer surface of the cell membrane. Apoptosis was induced at modest levels by gemcitabine or rituximab, but significantly more apoptosis was induced by the combination in both cell lines studied (Figure 3). When apoptosis was separated into early apoptosis, in which cells still exclude propidium iodide (PI), and late apoptosis where cells are PI permeable, early apoptosis is induced by the combination in DoHH2 cells, whereas both early and late apoptosis are demonstrated in WSU-FSCCL cells.
Apoptosis was also assessed by cleavage of PARP, which occurs when the apoptotic pathway is activated, eventually leading to cleavage of caspase 3 and of other downstream proteins including PARP. In each cell line, rituximab and gemcitabine result in PARP cleavage, with additional cleavage using the combination (Fig 4).
In vivo efficacy of gemcitabine ± rituximab
We initially screen for efficacy of therapy in our model by assessing the prevention of growth of cells in ascites fluid [11,17]. The lymphomas grow as bulky mesenteric nodes with development of hepatosplenomegaly, as well as diffuse adenopathy elsewhere, but not with measurable disease. While ascites represents only a small part of the animals' disease burden, it can be repeatedly sampled as an indicator of overall tumor in a mouse. Given the differing rates of growth of the cell lines in mice, treatment was started on day 3 after DoHH2 cell injection and day 7 after WSU-FSCCL cell injection. For DoHH2 (Figure 5A), gemcitabine delayed growth, so that at day 18 there were fewer lymphoma cells in the ascites fluid, however, by day 32 there was no longer a difference. Rituximab is effective in this model, although cells do eventually reaccumulate in ascites. There is a trend to longer time to recurrence of lymphoma cells with combined therapy. In the WSU-FSCCL model (Figure 5B), while gemcitabine alone had no effect, it did enhance the rituximab-mediated delay in lymphoma cell re-growth.
The most important endpoint for treatment efficacy, since there is not an externally measurable lesion, is survival of mice bearing the human lymphoma cell lines. We have performed duplicate experiments using mice injected with DoHH2 cells (Figure 6). At the gemcitabine dose of 120 μg/gm (~ 2.4 mg/mouse), used in previous reports [26] weekly for 3 doses, there was modest prolongation of median survival, with 0/5 and 2/6 long-term survivors in the two experiments. We used sub-maximally tolerated doses of rituximab, 5 μg per mouse, which had modest therapeutic effect (1/11 long-term survivors combined). Combination therapy with these two agents at the same dose and schedule, however, markedly prolonged survival (p = 0.04, left; p = 0.01, right for rituximab + gemcitabine vs rituximab), and cured 9 of 11 mice. The surviving mice were euthanized at the end of the experiment and found to be histologically negative for lymphoma.
Discussion
This report demonstrates the anti-lymphoma activity of the pyrimidine analog gemcitabine in vitro and in a scid mouse human lymphoma xenograft model. Rituximab is an active therapy for CD20+ NHL as a single agent and in combination with some biologic and chemotherapeutic agents [9-12]. The generalizability and mechanism of chemosensitization by rituximab has not been fully explored. Evidence for several mechanisms of rituximab mediated lymphoma cell death has been presented including direct induction of apoptosis, complement mediated killing and antibody-dependent cell mediated cytotoxicity (ADCC). Much of the data addressing these processes comes from in vitro systems, and the relevance to activity in patients remains uncertain. Our in vitro data in this report rely only on direct apoptosis, as we do not add human serum as a complement source nor effector cells. Direct apoptosis can be affected by degree of crosslinking of the antibody, which in turn can be controlled by humoral or cell surface molecules. Complement activity may be altered by inhibitory factors and may differ between human serum in vitro and mouse complement in murine models. We have found that crosslinking does enhance the degree of apoptosis after rituximab treatment in our two cell lines, while addition of complement has little effect (data not shown). ADCC depends on effector cells, and even the precise effector cells remain uncertain. Scid mice have residual granulocytes and NK cells, and depletion of these cells can abolish rituximab efficacy [27]. To understand the mechanisms of resistance to rituximab will require more complete knowledge of which of these mechanisms of action is or are most important in patients [15]. Recent data suggests that rituximab can alter intracellular signaling, even without inducing apoptosis, in ways that can sensitize cells to chemotherapy effects [16].
Our results demonstrate that the combination of gemcitabine and rituximab inhibits NHL cell growth, induces apoptosis in these cells, and, most importantly, is effective in prolonging survival of mice bearing human t(14;18)+ lymphoma cells. Prior reports have shown additive effects of gemcitabine and rituximab in aggressive NHL cell lines in vitro [10]. In that report, aggressive NHL cell lines that were relatively resistant to both gemcitabine and rituximab were pre-treated with rituximab for 24 hours and then treated with gemcitabine for 18 hours and found to have modest increases in hypodiploid cells and PI positive cells after PI staining.
The cell cycle changes seen after gemcitabine treatment alone, and in combination with rituximab, are cell line dependent. WSU-FSCCL cells, in contrast to most cells, are not blocked in S phase by gemcitabine alone, but are after rituximab is added. This may reflect altered intracellular signaling that could restore a putative S phase DNA damage sensor, such as the ATM pathway [25]. Similarly, induction of early versus late apoptosis is also cell line dependent. Further exploration of the biochemical basis of these changes is warranted to better understand which of the many potential targets of a pyrimidine nucleoside analog are the important determinants of gemcitabine activity against NHL cells, and how these targets are affected by rituximab.
Concurrent treatment of B cell lymphoproliferative disorders (NHL and CLL) with rituximab and chemotherapy is becoming more common, and data suggests benefit in time to disease progression in indolent disease [14] and also in overall survival in aggressive disease [13]. Questions remain regarding the optimal way to combine rituximab and chemotherapy and whether therapeutic efficacy of specific chemotherapeutic agents is enhanced by rituximab. Pre-clinical studies of the interaction of chemotherapy agents and rituximab may provide insight to guide the development of appropriate clinical trials.
Competing interests
There were no competing interests at the time this work was carried out. Subsequently, Dr. Obasaju has become employed by Lilly, maker of gemcitabine, but Lilly provided no support for this work.
Authors' contributions
MS conceived the study, provided oversight of the lab and wrote the manuscript. IJ designed and carried out the experiments and critiqued the manuscript. FJ assisted in carrying out the experiments and critiqued the manuscript. CO helped in the conception of the studies and critiqued the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was supported by NIH RO1CA71552, Fox Chase Cancer Center Core Grant NIH CA06927 and grants from the Mary L. Smith Charitable Lead Trust and the Martha Rogers Charitable Trust. Additional support came from the Janice Charach Epstein Research Fund and the Lester I. Smith Research Fund. Kathy Zenszer provided excellent secretarial assistance.
Figures and Tables
Figure 1 Cytotoxicity of rituximab and gemcitabine. WSU-FSCCL or DoHH2 cells incubated with varying concentrations of gemcitabine, and 4 hr later ± 20 μg/ml rituximab. Cytotoxicity is expressed as percentage of untreated control cells at 72 hr.
Figure 2 Effect of gemcitabine and rituximab on cell cycle. Cells (0.25 × 106/ml) incubated ± 20 μg/ml rituximab for 4 hr, then ± 2.5 nM gemcitabine for 24 hr. Cells stained with propidium iodide, analyzed by flow and cell cycle parameters calculated (ModFit LT program). A typical flow result is also shown (Figure 2A).
Figure 3 Apoptosis induced by gemcitabine and rituximab. Cells treated as in figure 2, except with 10 nM gemcitabine, then stained with annexin V and propidium iodide. Top left panel (A) is total annexin V positive cells. Top right panel (B) shows early apoptosis (annexin+/PI-, left) and late apoptosis (annexin+/PI+, right). A typical dot plot (C) is also shown.
Figure 4 PARP cleavage induced by gemcitabine and rituximab. Cells treated as in figure 2, but for 48 hr with either 2.5 or 5 nM gemcitabine, the cell lysates analyzed by Western using anti-PARP (C2–10, Trevigen, Inc), which detects both intact PARP (116 kD) and the apoptotic marker PARP cleavage fragment (85 kD), or G3PDH as a loading control.
Figure 5 Clearance of NHL cells from ascites by treatment with gemcitabine and rituximab. Scid mice were injected intraperitoneally (ip) with 5 × 106 DoHH2 cells (A, top) or 10 × 106 WSU-FSCCL cells (B, bottom). Mice received 20 μg rituximab (ip) on day 3 and 120 μg gemcitabine/gm body weight (ip) on day 4 after injection of DoHH2 cells. Ascites was collected on day 18, 32 and 42. Untreated scid/DoHH2 mice died prior to day 42. Because of the slower development of lymphoma with WSU-FSCCL, mice were treated with 100 μg rituximab (ip) on day 7 and 120 μg gemcitabine/gm body weight (ip) on day 8 after injection of WSU-FSCCL cells. Ascites was collected on days 22, 34 and 51. Cells from ascites were stained with PE anti-human CD45 antibody and analyzed by flow cytometry.
Figure 6 Survival of scid/DoHH2 mice after treatment with gemcitabine and rituximab. Two experiments in which scid mice injected with DoHH2 cells were treated ± 5 μg rituximab on days 2, 9 and 16 and ± gemcitabine (120 μg/gm) on days 3, 10 and 17. Mice were followed for survival and euthanized when ill. For comparison of rituximab + gemcitabine to rituximab alone, p = 0.04 (left) and p = 0.01 (right). Overall, the combination cured 9/11 mice, compared with 2/11 for gemcitabine and 1/11 with rituximab.
==== Refs
Smith MR Non-Hodgkin's Lymphoma Curr Probl Cancer 1996 20 6 77 8919170
Chau I Harries M Cunningham D Hill M Ross PJ Archer CD Norman AR Wotherspoon AC Koh DM Gill K Uzzell M Prior Y Catovsky D Gemcitabine, cisplatin and methylprednisolone chemotherapy (GEM-P) is an effective regimen in patients with poor prognostic primary progressive or multiply relapsed Hodgkin's and non-Hodgkin's lymphoma British Journal of Haematology 2003 120 970 977 12648066 10.1046/j.1365-2141.2003.04226.x
Crump M Baetz T Couban S Belch A Marcellus D Howson-Jan K Imrie K Myers R Adams G Ding K Paul N Shepherd L Iglesias J Meyer R Gemcitabine, dexamethasone, and cisplatin in patients with recurrent or refractory aggressive histology B-cell non-Hodgkin lymphoma: A Phase II study by the National Cancer Institute of Canada Clinical Trials Group (NCIC-CTG). Cancer 2004 101 1835 1842 15386331 10.1002/cncr.20587
Dumontet C Morschhauser F Solal-Celingy P Bouafia F Bourgeois E Thieblemont C Leleu X Hequet O Salles G Coiffier B Gemcitabine as a single agent in the treatment of relapsed or refractory low-grade non-Hodgkin's lymphoma British Journal of Haematology 2001 113 772 778 11380469 10.1046/j.1365-2141.2001.02795.x
Emmanouilides C Colovos C Pinter-Brown L Hernandez L Schiller G Territo M Rosen P Pilot study of fixed-infusion rate gemcitabine with Cisplatin and dexamethasone in patients with relapsed or refractory lymphoma Clin Lymphoma 2004 5 45 49 15245607
Fossa A Santoro A Hiddemann W Truemper L Niederle N Buksmaui S Bonodonna G Seeber S Nowrousian MR Gemcitabine as a Single Agent in the Treatment of Relapsed or Refractory Aggressive Non-Hodgkin's Lymphoma Journal of Clinical Oncology 1999 17 3786 3792 10577850
Nabhan C Krett N Gandhi V Rosen S Gemcitabine in hematologic malignancies Curr Opin Oncol 2001 13 514 521 11673693 10.1097/00001622-200111000-00015
McLaughlin P Grillo-Lopez AJ Link BK Levy R Czuczman MS Williams ME Heyman MR Bence-Bruckler I White CA Cabanillas F Jain V Ho AD Lister J Wey K Shen D Dallaire BK Rituximab chimeric anti-CD20 monoclonal antibody therapy for relapsed indolent lymphoma: half of patients respond to a four-dose treatment program. Journal of Clinical Oncology 1998 16 2825 2833 9704735
Alas S Ng CP Bonavida B Rituximab Modifies the Cisplatin-mitochondrial Signaling Pathway, Resulting in Apoptosis in Cisplatin-resistant Non-Hodgkin's Lymphoma Clinical Cancer Research 2002 8 836 845 11895917
Emmanouilides C Jazirehi AR Bonavida B Rituximab-mediated sensitization of B-non-Hodgkin's lymphoma (NHL) to cytotoxicity induced by paclitaxel, gemcitabine, and vinorelbine Cancer Biother Radiopharm 2002 17 621 630 12537665 10.1089/108497802320970226
Smith MR Fang J Joshi I Enhanced efficacy of therapy with antisense BCL-2 oligonucleotides plus anti-CD20 monoclonal antibody in scid mouse/human lymphoma xenografts Molecular Cancer Therapeutics 2004 3 1693 1699 15634664
Demidem A Lam T Alas S Hariharan K Hanna N Bonavida B Chimeric anti-CD20 (IDEC_C2B8) monoclonal antibody sensitizes a B cell lymphoma cell line to cell killing by cytotoxic drugs Cancer Biother Radiopharm 1997 12 177 186 10851464
Coiffier B Lepage E Briere J Herbrecht R Tilly H Bouabdallay R Morel P Van Den Neste E Salles G Gaulard P Reyes F Gisselbrecht C CHOP chemotherapy plus rituximab compared with CHOP alone in elderly patients with diffuse large-B-cell lymphoma. N England J Med 2002 346 235 242 11807147 10.1056/NEJMoa011795
Byrd JC Rai KR Peterson BL Appelbaum FR Morrison VA Kolitz JE Shepherd L Hines JD Schiffer CA Larson RA Addition of rituximab to fludarabine may prolong progression-free survival and overall survival in patients with previously untreated chronic lymphocytic leukemia: an updated retrospective comparative analysis of CALGB 9712 and CALGB 9011 Blood 2005 105 49 53 15138165 10.1182/blood-2004-03-0796
Smith MR Rituximab (monoclonal anti-CD20 antibody): mechanisms of action and resistance Oncogene 2003 22 7359 7368 14576843 10.1038/sj.onc.1206939
Jazirehi AR Bonavida B Cellular and molecular signal transduction pathways modulated by rituximab (rituxan, anti-CD20 mAb) in non-Hodgkin's lymphoma: implications in chemosensitization and therapeutic intervention Oncogene 2005 24 2121 2143 15789036 10.1038/sj.onc.1208349
Smith MR Xie T Zhao-Zong Z Joshi I Efficacy of treatment with antisense oligonucleotides complementary to immunoglobulin sequences of bcl-2/immunoglobulin fusion transcript in a t(14;18) human lymphoma-scid mouse model. Clinical Cancer Research 2001 7 400 406 11234896
Kluin-Neilemans HC Limpens J Meerabux J Beverstock GC Jansen JH de Jong D Kluin PM A new non-Hodgkin's B-cell line (DoHH2) with a chromosomal translocation t(14;18) (q32;q21) Leukemia 1991 85 221 224
Mohammad RM Mohamed AN Smith MR Jawadi NS Al-Katib A A unique EBV-negative low-grade lymphoma line (WSU-FSCCL) exhibiting both t(14;18) and t(8;11) Cancer Genet Cytogenet 1993 70 62 67 8221615 10.1016/0165-4608(93)90132-6
Freid J Perez AG Clarkson BD Rapid hypotonic method for flow cytofluorometry of monolayer cell cultures. Some pitfalls in staining and data analysis Journal of Histochemical Cytochemistry 1978 26 921 933
Brugge JS Erikson RL Identification of a transformation-specific antigen induced by an avian sarcoma virus. Nature 1977 269 346 348 198667 10.1038/269346a0
Chou TC Talalay P Quantitative analysis of dose-effect relationships: the combined effects of multiple drugs or enzyme inhibitors Adv Enzyme Regul 1984 22 27 55 6382953 10.1016/0065-2571(84)90007-4
Chow KU Sommerlad WD Boehrer S Schneider B Seipelt G Rummel MJ Hoelzer D Mitrou PS Weidmann E Anti-CD20 antibody (IDEC-C2B8, ribuximab) enhances efficacy of cytotoxic drugs on neoplastic lymphocytes in vitro: role of cytokines, complement and caspases. Haematologica 2002 87 33 43 11801463
Shi Z Azuma A Sampath D Li YX Huang P Plunkett W S-Phase arrest by nucleoside analogues and abrogation of survival without cell cycle progression by 7-hydroxystaurosporine Cancer Research 2001 61 1065 1072 11221834
Bartek J Lukas C Lukas J Checking on DNA damage in S phase Nat Rev Mol Cell Biol 2004 5 792 804 15459660 10.1038/nrm1493
Nowak AK Robinson BW Lake RA Gemcitabine exerts a selective effect on the humoral immune response: implications for combination chemo-immunotherapy. Cancer Research 2002 62 2353 2358 11956096
Hernandez-Ilizaliturri FJ Jupudy V Ostberg J Oflazoglu E Huberman A Repasky E Czuczman MS Neutrophils contribute to the biologican antitumor activity of rituximab in a non-Hodgkin's lymphoma severe combined immunodeficiency mouse model. Clinical Cancer Research 2003 9 5866 5873 14676108
|
16109167
|
PMC1208862
|
CC BY
|
2021-01-04 16:03:07
|
no
|
BMC Cancer. 2005 Aug 18; 5:103
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-103
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-1041610917010.1186/1471-2407-5-104Research ArticleExpression of G-protein inwardly rectifying potassium channels (GIRKs) in lung cancer cell lines Plummer Howard K [email protected] Madhu S [email protected] Maria [email protected] Hildegard M [email protected] Molecular Cancer Analysis Laboratory, Department of Pathobiology, College of Veterinary Medicine, University of Tennessee, Knoxville, TN 37996-4542, USA2 Molecular Cancer Analysis Laboratory, Department of Pathobiology, and Department of Large Animal Clinical Sciences, College of Veterinary Medicine, University of Tennessee, Knoxville, TN 37996-4542, USA3 Experimental Oncology Laboratory, Department of Pathobiology, College of Veterinary Medicine, University of Tennessee, Knoxville, TN 37996-4542, USA2005 18 8 2005 5 104 104 31 3 2005 18 8 2005 Copyright © 2005 Plummer 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
Previous data from our laboratory has indicated that there is a functional link between the β-adrenergic receptor signaling pathway and the G-protein inwardly rectifying potassium channel (GIRK1) in human breast cancer cell lines. We wanted to determine if GIRK channels were expressed in lung cancers and if a similar link exists in lung cancer.
Methods
GIRK1-4 expression and levels were determined by reverse transcription polymerase chain reaction (RT-PCR) and real-time PCR. GIRK protein levels were determined by western blots and cell proliferation was determined by a 5-bromo-2'-deoxyuridine (BrdU) assay.
Results
GIRK1 mRNA was expressed in three of six small cell lung cancer (SCLC) cell lines, and either GIRK2, 3 or 4 mRNA expression was detected in all six SCLC cell lines. Treatment of NCI-H69 with β2-adrenergic antagonist ICI 118,551 (100 μM) daily for seven days led to slight decreases of GIRK1 mRNA expression levels. Treatment of NCI-H69 with the β-adrenergic agonist isoproterenol (10 μM) decreased growth rates in these cells. The GIRK inhibitor U50488H (2 μM) also inhibited proliferation, and this decrease was potentiated by isoproterenol. In the SCLC cell lines that demonstrated GIRK1 mRNA expression, we also saw GIRK1 protein expression. We feel these may be important regulatory pathways since no expression of mRNA of the GIRK channels (1 & 2) was found in hamster pulmonary neuroendocrine cells, a suggested cell of origin for SCLC, nor was GIRK1 or 2 expression found in human small airway epithelial cells. GIRK (1,2,3,4) mRNA expression was also seen in A549 adenocarcinoma and NCI-H727 carcinoid cell lines. GIRK1 mRNA expression was not found in tissue samples from adenocarcinoma or squamous cancer patients, nor was it found in NCI-H322 or NCI-H441 adenocarcinoma cell lines. GIRK (1,3,4) mRNA expression was seen in three squamous cell lines, GIRK2 was only expressed in one squamous cell line. However, GIRK1 protein expression was not seen in any non-SCLC cells.
Conclusion
We feel that this data may indicate that stimulation of GIRK1 or GIRK2 channels may be important in lung cancer. Stimulation of GIRK channels and β-adrenergic signaling may activate similar signaling pathways in both SCLC and breast cancer, but lead to different results.
==== Body
Background
Recent studies in human cancer cell lines or in animal models have shown that the growth of adenocarcinomas of the lungs, pancreas and colon are under β-adrenergic control [1-5]. The tobacco carcinogenic nitrosamine 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK) has recently been identified as a high affinity β-adrenergic agonist that stimulated the growth of pulmonary and pancreatic adenocarcinomas in vitro and in animal models [1,3,5]. Expression of mRNA that encodes a G-protein coupled inwardly rectifying potassium channel (GIRK1) has been shown in tissue samples from approximately 40% of primary human breast cancers tested [6], and this expression of GIRK1 was associated with a more aggressive clinical behavior. Increases in GIRK currents by β-adrenergic stimulation have been reported in adult rat cardiomyocytes and in Xenopus laevis oocytes coexpressing β2-adrenergic receptors and GIRK1/GIRK4 subunits [7]. In addition, in rat atrial myocytes transiently transfected with β1 or β2 adrenergic receptors, the β-adrenergic agonist isoproterenol (Iso) stimulated GIRK currents, whereas this stimulation was not seen in non-transfected cells [8]. Previous data from our laboratory has indicated that there is a functional link between the β-adrenergic receptor pathway and the G-protein inwardly rectifying potassium channel (GIRK1) in breast cancer cell lines and these pathways were involved in growth regulation of these cells [9,10]. Additional data from our laboratory has also indicated that the normal breast epithelial cell line MCF 10A lacks GIRK1 expression [10].
Established risk factors for breast cancer include age, increased hormone exposure, alcohol consumption and family history as well as many other factors [11]. Smoking is a controversial risk factor for the development of this malignancy [12-14]. However, increases in pulmonary metastatic disease have been seen in smokers with breast cancer [15]. In addition, a study of 141,000 women showed a significantly increased risk of developing lung cancer for breast cancer patients, possibly due to interactions between radiotherapy and smoking [16]. An eight-fold increase in breast cancer risk has been seen in women who were smokeless tobacco users [17]. Smokeless tobacco has higher levels of nicotine and the tobacco carcinogen NNK than cigarette smoke [18]. However, many types of lung cancer are smoking related [19].
In order to investigate possible similarities between breast cancer and lung cancer, we wanted to determine if GIRK channels are expressed in human lung cancers and if β-adrenergic signaling is also involved in lung cancer. Voltage gated K+, Na+ and Ca2+ channels have been shown in the small cell lung cancer (SCLC) cell lines NCI-H128, NCI-H69 and NCI-H146 by measurement of currents [20]. These voltage activated K+ channels in the SCLC cell lines have a role in modulating cell proliferation [21]. A classical inwardly rectifying potassium channel not linked with G-proteins has been shown in a subclone of NCI-H69, H69AR, a subline with overexpressed multidrug resistance associated protein, but these currents were not seen in the original H69 cell line [22]. Inwardly rectifying potassium currents have also been measured in RERF-LC-MA SCLC cells, and these cells express the Kir2.1 mRNA and protein [23]. GIRK1 expression has been shown in tissue specimens from patients with non-small lung cancer and this expression was associated with lymph node metastasis [24]. Prior to this study, GIRK channels have not been identified in lung cancer cell lines or SCLC cell lines and clinical samples. In the present research we have screened small cell and non-small cell lung cancer cell lines and several normal lung cell types for GIRK mRNA and protein expression.
Methods
Cell culture
The human SCLC cell lines NCI-H69 (H69), NCI-H146 (H146), NCI-H187 (H187), NCI-H209 (H209), and NCI-H526 (H526), the human adenocarcinoma cell lines NCI-H322 (H322), NCI-H441 (H441), and A549, the carcinoid cell line NCI-H727 (H727), and the squamous cell lines NCI-H226 (H226), NCI-H2170 (H2170), and NCI-H520 (H520) were purchased from the American Type Culture Collection (Manassas, VA). The human SCLC cell line WBA [25] was a gift of Dr. G. Krystal, Medical College of Virginia. All cancer cell lines except A549 were maintained in RPMI medium supplemented with fetal bovine serum (10% v/v), L-glutamine (2 mM), penicillin (100 U/ml) and streptomycin (100 μg/ml) at 37°C in an atmosphere of 5% CO2. A549 cells were grown in Hams F12 media with supplements as above. Human small airway epithelia cells (SAEC) were purchased from Clonetics/BioWhittaker (Walkersville, MD). These primary cells were maintained in SAEC basal medium with supplements (Clonetics) at 37°C in an atmosphere of 5% CO2. Fresh surgical tissue samples were collected from patients at the University of Tennessee (UT) Graduate School of Medicine's Cancer Center and processed for reverse transcription polymerase chain reaction (RT-PCR). The collection of tissue was approved by the UT Institutional Review Board, and the authors have been certified by the NIH Office of Human Subjects Research. Cultures of fetal hamster pulmonary neuroendocrine cells (PNEC) were established from fetal lung periphery harvested on day 15 of gestation as previously described [26,27]. Studies with fetal hamster cells were approved by the UT Institutional Animal Care and Use Committee. Exposure of cells to isoproterenol (Iso), the GIRK inhibitor U50488H (U5) [28] (Sigma, St. Louis, MO), or ICI 118,551 (ICI) (Tocris, Ballwin, MO) for experiments was as detailed in the figure legends or results.
RT-PCR
RNA from cell cultures, adult hamster brain, PNEC and from fresh surgical tissue samples was isolated by guanidine isothiocyanate/cesium chloride ultracentrifugation [29] or by an Absolutely RNA kit (Stratagene, La Jolla, CA). RT-PCR was done as previously described [10]. The GIRK3 primers are forward 5'-gtgaccagcttcctccagac-3' and reverse 5'-gctaccatcttcccatccaa-3' which amplifies a 317 bp fragment (bases 1421–1737, Genbank Acession # NM_004983). PCR conditions are 94°C, 30 sec; 55°C, 30 sec; 72°C, 45 sec for 40 cycles. Cyclophilin primers were used an internal control (Ambion, Austin, TX).
Real-time PCR
Real-time PCR was done as previously described [10]. GIRK1 primers-forward 5'-ctctcggacctcttcaccac-3' and reverse 5'-gccacggtgtaggtgagaat-3' (bases 398–477, Genbank Acession # NM002239) and the internal TaqMan probe is 6-FAM-tcaagtggcgctggaacctc-TAMRA (bases 429–449), annealing 62°. GIRK2 primers-forward 5'-gacctgccaagacacatcag-3' and reverse 5'-cggtcaggtagcgataggtc-3' (bases 766–886, Genbank Acession # U52153) and the internal TaqMan probe is 6-FAM-gtgcaatgttcatcacggcaac-TAMRA (bases 837–859), annealing 56°. GIRK4 primers-forward 5'-agcgctacatggagaagagc-3' and reverse 5'-aagttgaagcgccacttgag-3' (bases 241–358, Genbank Acession # L47208) and the internal TaqMan probe is 6-FAM-accggtacctgagtgacctcttca-TAMRA (bases 301–324), annealing 62°. Reactions were run on a Cepheid SmartCycler (Sunnyvale, CA). Reaction conditions are 200 μM dNTPs, 0.3 μM gene specific primers, 0.2 μM TaqMan probe, 4 mM (GIRK1) or 6 mM (GIRK2or4) magnesium acetate, 2 μl cDNA and 1.5 U MasterTaq (Eppendorf, Westbury, NY) and MasterTaq buffer in a final volume of 25 μl. In some experiments, 18S primers were used an internal control (Qiagen, Valencia, CA).
Proliferation assay
Cells were counted and plated in RPMI without serum or phenol red at a density of 50,000 cells/well with appropriate treatments in a volume of 500 μl. Cells were allowed to grow for 48 hours, and proliferation was measured by a cell proliferation assay for 5-bromo-2'-deoxyuridine (BrdU) according to the manufacturer's instructions (Roche, Indianapolis, IN), N = 5 for each treatment group.
Western blots
Cell pellets were collected and membrane protein was isolated with the ReadyPrep protein extraction kit (signal) (Biorad, Hercules, CA). Protein levels were determined using the RCDC kit (Biorad). Aliquots of 20–30 μg protein were boiled in 3× loading buffer (New England Biolabs, Beverly, MA) for two minutes, then loaded onto 12% Tris-glycine-polyacrylamide gels (Cambrex, Rockland, ME), and transferred electrophoretically to nitrocellulose membranes. Membranes were incubated with the primary antibody (GIRK1; Upstate Biotechnology, Lake Placid, NY). In all western blots, membranes were additionally probed with an antibody for actin (Sigma) to ensure equal loading of protein between samples. The membranes were then incubated with appropriate secondary antibodies (Rockland, Gilbertsville, PA or Molecular Probes, Eugene OR). The antibody-protein complexes were detected by the LiCor Odyssey infrared imaging system (Lincoln, NE).
Results
In order to investigate possible expression similarities between breast cancer and lung cancer, we wanted to determine if GIRK channels are expressed in human lung cancers. Expression of mRNA for the GIRK1 channel was seen in three of six SCLC cell lines. GIRK 1 was expressed in the SCLC cell lines H69, H146 and WBA (Figure 1). Since GIRK1 cannot form channels alone, it must assemble with GIRK2, 3 or 4 [reviewed in [30]]. GIRK2 & 4 expression was determined in the six SCLC cell lines using real-time PCR. GIRK 4 expression was found in all six SCLC cell lines, and GIRK2 was found in all six SCLC cell lines with the exception of H146 and H187 (Table 1). In these experiments, real-time PCR was used as a second method of determining gene expression. Since gene expression was determined by only one sample (similar to gene expression studies by RT-PCR) no comparisons can be made between CT values for the samples. We wished to compare GIRK1 expression in SCLC to normal primary cells. GIRK1 was not expressed in the normal human SAE cells (Figure 1). Normal SAE cells express GIRK4 but do not express GIRK2 (data not shown). SCLC shares phenotypic and functional features with pulmonary neuroendocrine cells (PNEC), one of the possible origins of this cancer type [31]. Neither GIRK1 nor GIRK2 mRNA expression was found in normal tissue from PNEC cells isolated from prenatal hamster lungs using the primers designed for human GIRK1 and 2 (Figure 2). In order to determine if this result was due to species differences in GIRK1 and 2 between humans and hamsters, tissue was isolated from hamster brain. In both cases, mRNA from GIRK1 and 2 were found to be expressed in hamster brain (Figure 2). However, our GIRK4 primers were found not to work in hamster tissue (data not shown).
We wanted to determine if GIRK channels were also expressed in human non-SCLC (NSCLC) cancers. GIRK1 was not expressed in two adenocarcinoma cell lines of Clara cell phenotype, H322 and H441 (Figure 1). However, three GIRK channels (1,2,4) were expressed in both the human adenocarcinoma cell line with alveolar type II cell phenotype (A549) and human carcinoid cell line expressing neuroendocrine features (H727) (Figure 3). GIRK1 and 4 are expressed in three squamous cell lines, H2170, H226, and H520 (Figure 4). GIRK2 is only expressed in the H520 cell line, but not in either H2170 or H226 cell lines (Figure 4).
Although the predominant GIRK heterotetramers seem to be GIRK1/2 and GIRK1/4 [reviewed in [30]], GIRK3 expression may also be important in lung cancer. GIRK3 expression in normal cells, SCLC cell lines and NSCLC cell lines was examined. GIRK3 expression was seen in all six SCLC cell lines (Figure 5). GIRK3 was expressed in normal cells (SAEC, PNEC) and in the five NSCLC cell lines (A549, H727, H2170, H226, H520) (Figure 6). A summary of all the above experiments indicating gene expression data for the cell lines and normal cells is shown in Table 2. We also wanted to determine if GIRK1 was expressed in fresh surgical tissue samples. A limited number of samples of pulmonary adenocarcinomas (three) or pulmonary squamous carcinomas (two) were collected from patients at the UT Graduate School of Medicine's Cancer Center. None of these tissue samples expressed GIRK1 (Figure 7).
For functional GIRK channels in lung cells, protein expression is needed as well as gene expression. We determined GIRK1 protein expression in cell lines that express GIRK1 mRNA. Isolating enriched membrane protein using a Biorad kit, we found GIRK1 protein expression in the three SCLC cell lines that expressed GIRK1 mRNA, H69, H146 and WBA (Figure 8). In addition, in the NSCLC cell lines that express GIRK1 mRNA, GIRK1 protein expression was determined. GIRK1 protein expression was not seen in A549, H727, H2170, H226, and H520 (data not shown).
Since β-adrenergic ligands affected both gene expression of GIRK1 and cell proliferation for breast cancer cell lines in our laboratory [10], we wanted to determine if β-adrenergic agonists and antagonists had effects on either gene expression or cell proliferation in SCLC cells. H69 cells were used for these experiments because it is one of the two cell lines that expressed mRNA for all four GIRK channels as well as GIRK1 protein. The second cell line, WBA, is not as well characterized in the literature. Exposure of MDA-MB-453 breast cancer cells for six days to the β-adrenergic antagonist propranolol increased the GIRK1 mRNA levels [10]. In the present experiments, we used two specific β-adrenergic antagonists to possibly induce changes in GIRK1 mRNA levels. The specific β1 antagonist atenolol (100 μM) had no effect on gene expression when H69 cells were treated for seven days (Table 3). However, when H69 cells were treated with 100 μM of the β2 adrenergic antagonist ICI daily for seven days, differences in gene expression were seen as determined by real-time quantitative PCR (Table 3). Calculating differences in gene expression using the 2-ΔΔCT Formula [32], ICI treatment caused a 1.47× decrease in GIRK1 gene expression p < 0.0003 (Table 3). H69 cells express mRNA for both β1 and β2 (data not shown). As indicated above, we wanted to determine if β-adrenergic ligands altered proliferation in H69 cells. We determined the effects of the β-adrenergic agonist Iso and the GIRK channel inhibitor U5 [28] on proliferation in H69 cells. Iso (10 μM) inhibited proliferation (p < 0.001) (Figure 9). The GIRK inhibitor U5 (2 μM) also inhibited proliferation (p < 0.001), and this decrease was potentiated by Iso (p < 0.001) (Figure 9).
Discussion
This is the first report that describes the expression of mRNA for G-Protein Inwardly Rectifying Potassium Channels (GIRKs) in lung cancer cell lines. Previous reports have indicated that voltage gated K+, Na+ and Ca2+ channels have been shown in SCLC cell lines [20], and these voltage activated K+ channels have a role in modulating cell proliferation [21]. Classical inwardly rectifying potassium channels not linked with G-proteins have been shown in a few SCLC cell lines [22,23]. Since GIRK1 cannot form functional channels by itself, other GIRK channels are needed [30]. All six SCLC cell lines tested express mRNA for either GIRK2 or GIRK4 indicating that functional GIRK potassium channels are possible in these SCLC cancer cell lines. GIRK1 has been shown in tissue samples from approximately 40% of primary human breast cancers tested [6], and this expression of GIRK1 was associated with a more aggressive clinical behavior. GIRK1 has also been shown to contribute to tumor progression in NSCLC human tumors [24].
Differences in GIRK channel mRNA expression was seen in different types of lung cancer. The three adenocarcinomas from the lung cancer patient samples did not express GIRK1, and a subset of lung adenocarcinoma cell lines (H322, H441) also did not express GIRK1. However, GIRK1 expression was seen in an adenocarcinoma cell line with alveolar type II cell phenotype (A549). These variations may be due to differences between the Clara cell and the alveolar type II lineage, or it may be due to the small sample size. The cell linage of the lung cancer patient samples is unknown. Data from our laboratory has also indicated differences in responses to anticancer agents between alveolar type II and Clara cells in the lungs [33]. Further research with additional cell lines and tissue samples is needed to determine if the differences in GIRK expression between alveolar type II cells and Clara cells is significant.
The two samples from squamous cancer patients also did not express GIRK1 mRNA. However, squamous carcinoma cells lines and a carcinoid cell lines do express GIRK1. In all NSCLC cell lines that express GIRK1, either GIRK2 or GIRK4 mRNA was expressed, indicating that functional GIRK potassium channels are possible in these non-SCLC cancer cell lines. Our data showing lack of GIRK1 expression in tumors from human adenocarcinomas and squamous cell carcinomas in is in contrast to Takanami et al. [24]. They found that in 72 NSCLC patients, 69% had high GIRK1 levels and 31% had low GIRK1 levels [24]. The differences may be due to our small sample size, or there may be other factors in cancer patients that affect GIRK1 expression. The expression of GIRK2 or GIRK4 has not been determined in previous studies.
The mRNA for GIRK1 & GIRK2 channels have not been detected in the normal cell lines tested, although the mRNA for GIRK4 channels was found in SAE normal cells. Our data indicates more differential expression of GIRK1 and GIRK2 than of GIRK4 in both SCLC and NSCLC cell lines. It is our hypothesis that GIRK4 is less important in lung cancer due to the fact it was expressed in all normal and cancerous cell lines. Further study is needed to determine the importance of GIRK1 and GIRK2 in lung cancer. In addition GIRK3 was found in all cell lines tested, including SAEC and normal hamster PNEC. Since GIRK3 expression was seen in all cells tested, we also feel it is unlikely to be a factor in tumor growth or progression. This hypothesis is supported by data indicating that one of the functions of GIRK3 is to inhibit plasma membrane expression of other GIRK subunits [34]. Further studies are needed to determine the role of GIRK3 in lung cancer.
Since protein expression would also be necessary for functional GIRK channels, we determined GIRK1 protein expression in cell lines that expressed GIRK1 mRNA. Expression of GIRK1 protein was seen in the three SCLC cell lines that express GIRK1 mRNA. This is the first report of GIRK1 protein expression in SCLC cell lines. We also determined GIRK1 protein expression in NSCLC cell lines. None of the NSCLC cell lines that expressed GIRK1 showed protein expression for GIRK1. We feel that this indicates that GIRK1 expression may be a larger factor in SCLC than in NSCLC. Further study is needed to elucidate these differences.
In the present study, we demonstrated a small decrease in GIRK1 mRNA expression (1.5×) in the H69 SCLC cell line after treatment for one week with the β2 adrenergic antagonist ICI daily for seven days. In contrast to this data, exposure of the breast cancer cell line MDA-MB453 for six days to the β-adrenergic antagonist propranolol (1 μM) increased the GIRK1 mRNA levels [10]. Treatment with the β1 antagonist, atenolol, for the same time period had no effect, indicating that the β2-adrenergic receptor is more important for changes in GIRK mRNA. GIRK currents have been shown to be increased in cells stimulated with the β-adrenergic agonist Iso in rat atrial myocytes transfected with β1 or β2 receptors [8]. Heterologous facilitation of GIRK currents by β-adrenergic stimulation was also seen in rat cardiomyocytes [7]. The differences in signaling pathways stimulated by β-adrenergic ligands in breast cancer and SCLC remain to be elucidated.
The tobacco carcinogen 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK), a high affinity agonist for β-adrenergic receptors, stimulated DNA synthesis in two breast cancer cell lines, an effect inhibited by ICI [9]. In the present report the β-adrenergic agonist, Iso, inhibited proliferation in H69 SCLC cells, also indicating opposite effects between breast cancer and small cell lung cancer with β-adrenergic agonists. The GIRK channel inhibitor U5 [28] also inhibited proliferation. It is our contention that the same receptors and channels (β-adrenergic, GIRK) are in both breast cancer and lung cancer, and stimulation of these pathways can lead to different results. This could be important in devising therapies in women breast cancer patients that are also smokers. Another study has indicated down-regulation of protein synthesis and mRNA expression of voltage-dependent and calcium-activated potassium channels bronchial and bronchiolar smooth muscle cells in rats by chronic smoking [35]. No differences were found in protein extracted from the lung, however.
Further studies are needed with RNA interference (RNAi) or small interfering RNA (siRNA) and additional β-adrenergic agonists and antagonists to further determine the effects of GIRK1 and GIRK2 on β-adrenergic signaling in human lung cancer, especially studies on differences in GIRK protein expression. However, it appears that β-adrenergic signaling and GIRK channels are important in both SCLC and breast cancer. These findings may be important in devising therapies based on β-adrenergic signaling and GIRK channel expression in lung cancers that express GIRK channels or β-adrenergic receptors.
Conclusion
We feel that this data may indicate that stimulation of GIRK1 or GIRK2 channels may be important in lung cancer. Stimulation of GIRK channels and β-adrenergic signaling may activate similar signaling pathways in both SCLC and breast cancer, but lead to different results.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
HP carried out the majority of experiments, designed the study, and helped draft the manuscript. MD carried out the GIRK western blots. MD and MC were involved in proliferation studies. MC was also involved in RT-PCR. HS helped draft the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Research described in this article was supported in part by Philip Morris USA Inc. and by Philip Morris International. This research was also supported by a State of Tennessee Center of Excellence Fund grant to Dr. H.K. Plummer III. The WBA SCLC cell line was provided by Dr. G. Krystal, Medical College of Virginia, Richmond, VA. Expert editorial assistance was provided by Dr. T. Masi. Cell culture support for BrdU assays was provided by Ms A. Gardner.
Figures and Tables
Figure 1 Expression of GIRK1 in some SCLC cell lines but not in normal SAE cells, nor in adenocarcinoma Clara cell phenotype cell lines. Top panel: GIRK1 is expressed in WBA, H69, and H146 SCLC cell lines but not in H187, H209 and H526 SCLC cell lines. GIRK1 is also not expressed in normal SAE cells, nor is it expressed in H322 or H441 adenocarcinoma Clara cell lines. Bottom panel: Cyclophilin, used as a positive reaction control was seen in all samples. For all gene expression experiments, negative control reactions were performed and found to be negative. The bands on the agarose gels were consistent with the expected sizes: GIRK1-441 bp; cyclophilin-216 bp; M-100 bp.
Figure 2 Lack of GIRK1 and 2 expression in hamster PNEC cells. Top panel: Pulmonary neuroendocrine cells were isolated from fetal hamsters on the 15th day of gestation. Neither GIRK1 nor GIRK2 was expressed in normal PNEC. Bottom panel: Tissue was isolated from adult hamster brain to determine if the lack of expression in PNEC was not due to species differences in PCR primers. Both GIRK1 and GIRK2 were expressed in hamster brain. Cyclophilin, used as a positive reaction control was seen in the samples. For all gene expression experiments, negative control reactions were performed and found to be negative. The bands on the agarose gels were consistent with the expected sizes: GIRK1-441 bp; GIRK2-438 bp; cyclophilin-216 bp, M-100 bp.
Figure 3 Expression of GIRK channels in adenocarcinoma alveolar type II cell phenotype (A549) and carcinoid (H727) cell lines. Top panel: GIRK1 was expressed in both A549 and H727 cell lines. Bottom panel: Both A549 and H727 expressed GIRK2 and GIRK4. Cyclophilin, used as a positive reaction control was seen in both samples. For all gene expression experiments, negative control reactions were performed and found to be negative. The bands on the agarose gels were consistent with the expected sizes: GIRK1-441 bp; GIRK2-438 bp; GIRK4-401 bp; cyclophilin-216 bp; M-100 bp.
Figure 4 Differences in GIRK expression in three squamous carcinoma cell lines. Top panel: GIRK1 was expressed in all three squamous cell lines, H2170, H226, and H520. Middle panel: GIRK2 is only expressed in H520 cell line, but not in H2170 or H226. GIRK4 was expressed in all three squamous cell lines, H2170, H226, and H520. Bottom panel: Cyclophilin, used as a positive reaction control was seen in all samples. For all gene expression experiments, negative control reactions were performed and found to be negative. The bands on the agarose gels were consistent with the expected sizes: GIRK1-441 bp; GIRK2-438 bp; GIRK4-401 bp; cyclophilin-216 bp; M-100 bp.
Figure 5 Expression of GIRK3 in SCLC cell lines. Top panel: GIRK3 expression was seen in all six SCLC cell lines, in WBA, H69, H146, H187, H209, and H526. Bottom panel: Cyclophilin, used as a positive reaction control was seen in all samples. For all gene expression experiments, negative control reactions were performed and found to be negative. The bands on the agarose gels were consistent with the expected sizes: GIRK3-317 bp; cyclophilin-216 bp; M-100 bp.
Figure 6 GIRK3 expression in NSCLC cell lines and in normal cells. Top panel: GIRK3 was expressed in normal human SAEC and hamster PNEC as well as being expressed in all NSCLC cell lines tested. These NSCLC cell lines are: A549, H727, H226, H520, and H2170. Bottom panel: Cyclophilin, used as a positive reaction control was seen in all samples. For all gene expression experiments, negative control reactions were performed and found to be negative. The bands on the agarose gels were consistent with the expected sizes: GIRK3-317 bp; cyclophilin-216 bp; M-100 bp.
Figure 7 Lack of GIRK1 expression in fresh surgical tissue samples of pulmonary adenocarcinomas or pulmonary squamous carcinomas. Top panel: GIRK1 was not expressed in two pulmonary squamous carcinomas or three adenocarcinomas obtained from cancer patients. Bottom panel: Cyclophilin, used as a positive reaction control was seen in all samples. For all gene expression experiments, negative control reactions were performed and found to be negative. The bands on the agarose gels were consistent with the expected sizes: GIRK1-441 bp; cyclophilin-216 bp; M-100 bp.
Figure 8 GIRK1 protein expression in SCLC cells that express GIRK1 mRNA as assessed by western blot analysis. In the three SCLC cell lines that expressed GIRK1 mRNA (Figure 1), membrane protein expression was determined. Top panel: All three SCLC cell lines (WBA, H69, H146) expressed GIRK1 membrane protein. Bottom panel: Actin was used as a control for equal sample loading. The bands are consistent with the expected size: GIRK1-62 kDa; actin-42 kDa.
Figure 9 Inhibition of proliferation in H69 cells by the β-adrenergic agonist isoproterenol and the GIRK channel inhibitor U50488H. H69 cells were grown in the presence of isoproterenol or U50488H or a combination of Iso and U5 for 48 hours. Proliferation was determined by a BrdU assay. Both 10 μM Iso (p < 0.001) and 2 μM U5 (p < 0.001) inhibited proliferation with Iso having the greater effect. The effect of the GIRK inhibitor U5 on proliferation was potentiated by Iso (p < 0.001). N = 5.
Table 1 Expression of GIRK2 or GIRK4 in small cell lung cancer cell lines. Expression of either GIRK2 or GIRK4 was determined by real time PCR. Threshold values (CT) values are listed below.
Cell Line GIRK2 GIRK4
WBA 19.82 30.56
H69 23.54 32.18
H146 NF 40.54
H187 NF 31.60
H209 24.53 24.64
H526 27.55 23.21
NF – No CT value calculated due to lack of expression. Gene expression was determined by one sample. No comparisons can be made between CT values for the samples.
Table 2 Summary of GIRK mRNA expression in lung cells. This is a compilation of data shown in Figures 1–6 & Table 1. Some of the indicated data was not shown.
GIRK1 GIRK2 GIRK3 GIRK4
Normal primary
SAEC NF NF E E
PNEC NF NF E X
SCLC
WBA E E E E
H69 E E E E
H146 E NF E E
H187 NF NF E E
H209 NF E E E
H526 NF E E E
NSCLC
H322 NF ND ND ND
H441 NF ND ND ND
H549 E E E E
H727 E E E E
H2170 E NF E E
H226 E NF E E
H520 E E E E
E – mRNA expression
NF – mRNA expression not found
ND – mRNA expression not determined
X – human primer not compatible with hamster
Table 3 Changes in gene expression of GIRK1 after seven day treatment of H69 cells with the β2 adrenergic antagonist ICI 118, 551 (ICI) (100 μM) and lack of changes after seven day treatment with β1 adrenergic antagonist atenolol (AT) (100 μM).
Control ICI AT
GIRK1 21.186 ± 0.13 22.040 ± 0.06a 21.238 ± 0.06
18S 12.028 ± 0.17 12.326 ± 0.03 12.090 ± 0.11
Changes in gene expression were determined by real time quantitative PCR, and are expressed as threshold values (CT), N = 5. a p < 0.0003. Threshold/CT cycle values will be higher for samples with less mRNA expression.
==== Refs
Schuller HM Tithof PK Williams M Plummer HK III The tobacco-specific carcinogen 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone is a β-adrenergic agonist and stimulates DNA synthesis in lung adenocarcinoma via β-adrenergic receptor-mediated release of arachidonic acid Cancer Res 1999 59 4510 4515 10493497
Schuller HM Plummer HK IIIBoschler PN Dudrick P Bell JL Harris RE Co-expression of β-adrenergic receptors and cyclooxygenase-2 in pulmonary adenocarcinoma Int J Oncol 2001 19 445 449 11494020
Schuller HM Porter B Riechert A Beta-adrenergic modulation of NNK-induced lung carcinogenesis in hamsters J Cancer Res Clin Oncol 2000 126 624 630 11079726
Masur K Niggerman B Zanker KS Entschladen F Norepinphrine-induced migration of SW 480 colon carcinoma cells is inhibited by beta-blockers Cancer Res 2001 61 2866 2869 11306460
Weddle DL Tithoff PK Williams M Schuller HM Beta adrenergic growth regulation of human cancer cell lines derived from pancreatic ductal carcinomas Carcinogenesis 2001 22 473 479 11238189 10.1093/carcin/22.3.473
Stringer BK Cooper AG Shepard SB Overexpression of the G-protein inwardly rectifying potassium channel (GIRK1) in primary breast carcinomas correlates with axillary lymph node metastasis Cancer Res 2001 61 582 588 11212253
Mullner C Vorobiov D Bera AK Uezono Y Yakubovich D Frohnwieser-Steinecker B Dascal N Schreibmayer W Heterologous facilitation of G protein-activated K+ channels by β-adrenergic stimulation via cAMP-dependent protein kinase J Gen Physiol 2000 115 547 557 10779313 10.1085/jgp.115.5.547
Wellner-Kienitz MC Bender K Pott L Overexpression of β1 and β2 adrenergic receptors in rat atrial myocytes. Differential coupling to G protein inward rectifier K+ channel via Gs and Gi/o J Biol Chem 2001 276 37347 37354 11495921 10.1074/jbc.M106234200
Cakir Y Plummer HK IIISchuller HM Beta-adrenergic and arachidonic acid-mediated growth regulation of human breast cancer cell lines Int J Oncology 2002 21 153 157
Plummer HK IIIYu Q Cakir Y Schuller HM Expression of inwardly rectifying potassium channels (GIRKs) and beta-adrenergic regulation of breast cancer cell lines BMC Cancer 2004 4 93 15603589 10.1186/1471-2407-4-93
Baselga J Norton L Focus on breast cancer Cancer Cell 2002 1 319 322 12086846 10.1016/S1535-6108(02)00066-1
Greenlee RT Hill-Harmon MB Murray T Thun M Cancer Statistics CA Cancer J Clin 2001 51 15 36 11577478
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 76 12725422 10.1023/A:1022988918755
Morabia A Smoking (active and passive) and breast cancer: epidemiologic evidence up to June 2001 Environ Molec Mutagen 2002 39 89 95 11921174 10.1002/em.10046
Murin S Cigarette smoking and the risk of pulmonary metastasis from breast cancer Chest 2001 119 1635 1640 11399684 10.1378/chest.119.6.1635
Prochazka M Granath F Ekbom A Shields PG Hall P Lung cancer risks in women with previous breast cancer Eur J Cancer 2002 38 1520 1525 12110499 10.1016/S0959-8049(02)00089-8
Spangler JG Michielutte R Bell RA Dignan MB Association between smokeless tobacco use and breast cancer among native-american women in North Carolina Ethn Dis 2001 11 36 43 11289249
Hoffmann D Djordjevic MV Chemical composition and carcinogenicity of smokeless tobacco Adv Dent Res 1997 11 322 329 9524432
Vineis P Alavanja M Buffler P Fontham E Franceschi S Gao YT Gupta PC Hackshaw A Matos E Samet J Sitas F Smith J Stayner L Straif K Thun MJ Wichmann HE Wu AH Zaridze D Peto R Doll R Tobacco and cancer: recent epidemiological evidence J Natl Cancer Inst 2004 96 99 106 14734699
Pancrazio JJ Viglione MP Tabbara IA Kim YI Voltage-dependent ion channels in small-cell lung cancer cells Cancer Res 1989 49 5901 5906 2477149
Pancrazio JJ Tabbara IA Kim YI Voltage-activated K+ conductance and cell proliferation in small-cell lung cancer Anticancer Res 1993 13 1231 1234 8394676
Jirsch J Deeley RG Cole SPC Stewart AJ Fedida D Inwardly rectifying K+ channels and volume-regulated anion channels in multidrug-resistant small cell lung cancer cells Cancer Res 1993 53 4156 4160 8395978
Sakai H Shimizu T Hori K Ikari A Asano S Takeguchi N Molecular and pharmacological properties of inwardly rectifying K+ channels of human lung cancer cells Eur J Pharmacol 2002 435 125 133 11821018 10.1016/S0014-2999(01)01567-9
Takanami I Inoue Y Gika M G-protein inwardly rectifying potassium channel (GIRK1) gene expression correlates with tumor progression in non-small cell lung cancer BMC Cancer 2004 4 79 15541182 10.1186/1471-2407-4-79
Plummer HK IIICatlett J Leftwich J Armstrong B Carlson P Huff T Krystal G c-myc expression correlates with suppression of c-kit protooncogene expression in small cell lung cancer cell lines Cancer Res 1993 53 4337 4342 7689933
Linnoila IR Gazdar AF Funa K Becker KL Long-term selective culture of hamster pulmonary neuroendocrine cells Anat Rec 1993 236 231 240 8389531 10.1002/ar.1092360128
Schuller HM Carbon dioxide potentiates the mitogeneic effects of nicotine and its carcinogenic dervivative, NNK, in normal and neoplastic neuroendocrine lung cells via stimulation of autocrine and protein kinase C-dependent mitogenic pathways Neurotoxicology 1994 15 877 886 7715858
Shankar H Murugappan S Kim S Jin J Ding Z Wickman K Kunapuli SP Role of G protein-gated inwardly-rectifying potassium channels in P2Y12 receptor-mediated platelet functional responses Blood 2004 104 1335 1343 15142872 10.1182/blood-2004-01-0069
MacDonald RJ Swift GH Przybyla AE Chirgwin JW Isolation of RNA using quanidinium salts Methods Enzymol 1987 152 219 227 2443794
Mark MD Herlitze S G-protein mediated gating of inward-rectifier K+ channels Eur J Biochem 2000 267 5830 5836 10998041 10.1046/j.1432-1327.2000.01670.x
Johnson DE Georgieff MK Pulmonary perspectives: neuroendocrine cells in health and disease Amer Rev Resp Dis 1989 140 1807 1812 2574971
Livak KJ Schmittgen TD Analysis of relative gene expression data using real-time quantitative PCR and the 2-ΔΔCT method Methods 2001 25 402 408 11846609 10.1006/meth.2001.1262
Adissu HA Schuller HM Antagonistic growth regulation of cell lines derived from human lung adenocarcinomas of Clara cell and alveolar type II cell lineage: Implications for chemoprevention Int J Oncology 2004 24 1467 1472
Ma D Zerangue N Raab-Graham K Fried SR Jan YN Jan LY Diverse trafficking patterns due to multiple traffic motifs in G protein-activated inwardly rectifying potassium channels from brain and heart Neuron 2002 33 715 729 11879649 10.1016/S0896-6273(02)00614-1
Ye H Ma WL Yang ML Liu SY Wang DX Effect of chronic cigarette smoking on large-conductance calcium-activated potassium channel and Kv1.5 expression in bronchial smooth muscle cells of rats Sheng Li Xue Bao 2004 56 573 578 15497036
|
16109170
|
PMC1208863
|
CC BY
|
2021-01-04 16:03:07
|
no
|
BMC Cancer. 2005 Aug 18; 5:104
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-104
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-1051610917110.1186/1471-2407-5-105Research ArticleIn B-CLL, the codon 72 polymorphic variants of p53 are not related to drug resistance and disease prognosis Sturm Isrid [email protected] Andrew G [email protected] Michael [email protected]örken Bernd [email protected] Peter T [email protected] Department of Hematology, Oncology and Tumor Immunology, University Medical Center Charité, Campus Berlin-Buch and Department of Hematology and Oncology, University Medical Center Charité, Campus Virchow Klinikum, Humboldt University, 13353 Berlin, Germany2 Clinical and Molecular Oncology, Max Delbrück Center for Molecular Medicine, 13125 Berlin, Germany3 Bath Cancer Research, Royal United Hospital, Bath, BA1 3NG, UK4 Institute of Pathology, University Medical Center Charité, Campus Benjamin Franklin, Humboldt University, 12300 Berlin, Germany2005 18 8 2005 5 105 105 6 5 2005 18 8 2005 Copyright © 2005 Sturm 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 common sequence polymorphism at codon 72 of the p53 gene encoding either arginine or proline was recently shown to be functionally relevant for apoptosis induction in vitro. In B-type chronic lymphocytic leukemia (B-CLL), p53 gene mutations occur in a subset of patients and are associated with impaired survival and drug resistance. Here, we address the functional relevance of the codon 72 single nucleotide (SNP) polymorphism for cell death sensitivity following exposure to clinically employed cytotoxic drugs and γ-irradiation.
Methods
138 B-CLL samples were analysed by SSCP-PCR and sequencing for single nucleotide polymorphism at codon 72 of the p53 gene. The in vitro cytotoxicity assay (DiSC-assay) was performed with 7 drugs (chlorambucil, mafosfamide, fludarabine phosphate, methylprednisolone, doxorubicin, vincristine) or γ-irradiation.
Results
Of the138 B-CLL samples, 9 samples were homozygous for proline (Pro/Pro), 78 samples homozygous for arginine (Arg/Arg), and 49 samples heterozygous (Arg/Pro). No differences were found for patient survival and cell death triggered by 7 cytotoxic drugs or γ-irradiation.
Conclusion
These data indicate that polymorphic variants of p53 codon 72 are not clinically relevant for apoptosis induction or patient survival in B-CLL.
==== Body
Background
The tumor suppressor gene p53 plays a central role in the induction of cell cycle arrest, senescence and apoptosis [1-5]. The polyproline domain (PP domain) of p53 spanning amino acids 62–91 is involved in apoptosis induction and facilitates transactivation of pro-apoptotic genes by p53 [6].
Located in this PP domain is at codon 72 a common single nucleotide polymorphism (SNP), resulting in either a proline residue (p53Pro) or an arginine residue (p53Arg). Thus, each individual inherits a p53 genotype that can be heterozygous (Arg/Pro) or homozygous for either arginine (Arg/Arg) or proline (Pro/Pro). The polymorphism is balanced, varies with latitude and race, and is maintained at different allelic frequencies across the population [7]. These two SNPs appear to be different both biochemically and biologically [8-11]. Differences in apoptosis susceptibility to cytotoxic drugs were described [12,13], and the response and survival to radiochemotherapy in clinical samples of squamous cell carcinomas was found to be increased in case the arginine allele is retained [13].
Chronic lymphocytic B-cell lymphoma is still an incurable disease and may be addressed as a disease of intrinsic apoptosis deficiency. It is a disease where the mutational status of the p53 gene is linked to patient survival [14] (and references herein). We therefore asked whether the codon 72 polymorphism is of clinical relevance for in vitro resistance to cytotoxic drugs, γ-irradiation and patient prognosis.
Methods
Patients
Samples from 138 B-CLL patients (99 male, 39 female, age 63.2 ± 0.5 (mean ± SEM) were analysed. Peripheral blood was analysed for drug sensitivities using fresh cells. The same samples were analysed for mutations in the p53 DNA binding domain and of the p53 codon 72 SNP, using snap frozen cells from the same specimens. Only patients with high peripheral blood leukocyte count were included in this analysis (median WBC 120.8/nl, range 20.7–1262.2/nl). The Binet stage was A in 29 cases, B in 24 cases, A/B in 15 cases (due to insufficient information on clinical lymph node status) and C in 62. Of the 138 patients, 80 were pretreated (58%) with one to six drug regimens (mean number of pretreatments for these 80 patients ± SEM: 1.93 ± 1.005). Median survival was 30.1 months; median follow up for the 17 censored patients was 97.9 months. This study was performed in accordance with local ethical standards and the declaration of Helsinki.
Analysis of p53 codon 72 polymorphism
SNP in the p53 codon 72 were analysed by genomic SSCP-PCR analysis and DNA sequencing. The primers were CGG ACG ATA TTG AAC AAT GG (sense) and CGT TTT CTG GGA AGG GAG AG (antisense), resulting in a PCR product of 167 bp. A standard PCR reaction in 50 μl with a final concentration of 0.6 mM MgCl2 was performed (40 cycles, annealing temperature 56°C). PCR products were denatured and separated on a nondenaturating 10% polyacrylamide gel at 500 V and 50 mA for 2 h at 22°C and analysed by silver staining [15,16]. Sequence polymorphism was confirmed by DNA sequencing, as described [14]. A result was obtained in 136 samples (98.6%). Mutations in the DNA-binding domain of the p53 gene (exon 5 to 8) has been analyzed previously by the use of SSCP-PCR and sequencing as described in detail elsewhere [14].
Chemosensitivity assay
B-CLL cells were exposed to cytotoxic drugs (chlorambucil, mafosfamide, fludarabine phosphate, methylprednisolone, doxorubicin, vincristine) or γ-irradiation (2 Gy) and cultured for 92 h. Percentages of drug-induced cell death and LD90 doses were determined by the use of a standardized morphometric test, the DiSC assay [17,18].
Data analysis
For intervariable assessment, the non-parametric Mann-Whitney U-test (for 2 groups) or the Kruskal-Wallis test (for 3 groups) or the χ2-test or Fisher's exact test for categorical parameters were applied. Overall survival was estimated by the Kaplan-Meier product-limit method. For the chemosensitivity data, LC90 doses were determined by calculating the log dose at which the fitted survival probability was equal to 0.1. LC90 values were logged (base 10) before calculation of mean and SEM, as described [19].
Results
Codon 72 SNP and clinical data
Of the 138 samples analysed, 136 could be amplified by genomic PCR : 78 samples were Arg/Arg, 9 samples were Pro/Pro, and 49 samples were Arg/Pro. Patients age and gender was not different in the subgroups (63.3 ± 1.07 years; male/female 55/23 in Arg/Arg, 65.2 ± 2.7 years, male/female 6/3 in Pro/Pro, 63.2 ± 1.5; male/female 37/12 in Pro/Arg). In the Binet's stage, the distribution was: stage A: 18 Arg/Arg, 1 Pro/Pro, 10 Pro/Arg; stage B: 10 Arg/Arg, 1 Pro/Pro, 13 Pro/Arg; stage C: 37 Arg/Arg, 6 Pro/Pro, 18 Pro/Arg; stage A or B: 9 Arg/Arg, 1 Pro/Pro, 5 Pro/Arg (p = 0.6).
78 patients were pretreated with a median of 2 treatment regimens. There was, however, no association between pretreatment status and codon 72 SNP (p = 0.6).
Codon 72 SNP and IgVH and p53 gene mutational analysis
In 113 samples, the IgVH-gene status was known, and in all 136 cases the p53 genotype was known [14]. Concerning the IgVH-gene status, of the hypermutated samples, 19 were Arg/Arg, 0 Pro/Pro, 10 Pro/Arg and of the pre-germinal center samples, 50 were Arg/Arg, 6 Pro/Pro, 28 Pro/Arg (p = 0.3). Concerning the p53 genotype of the B-CLL samples, of the 22 p53-mutated samples, 17 were Arg/Arg, 2 Pro/Pro, 3 Pro/Arg and of the 114 p53-wildtype samples, 61 were Arg/Arg, 7 Pro/Pro, 46 Pro/Arg (p = 0.06).
Codon 72 SNP and survival
There was no difference in overall survival for the 3 different groups (p = 0.885) (Fig. 1B). This was also true when only the p53 wild type samples were analysed (p = 0.46).
Codon 72 SNP and cell death sensitivity
For none of the tested drugs (chlorambucil, mafosfamide, doxorubicine, vincristine, fludarabine, caldribine and methylprednisolone), a correlation to the p53 codon 72 SNP was seen. This was the case irrespective of the p53 mutational status (Table 1). For γ-irradiation, there was a trend (p = 0.099) for better survival of the B-CLL cells when the sample is homozygous for proline. This finding is in accordance with a recent report on response rates and survival of head and neck carcinoma patients treated with radiochemotherapy where patients with a homozygous proline genotype had an impaired outcome to treatment [13]. This association of the Pro/Pro genotype failed, however, to attain statistical significance. This, together with the lack of correlation with patient survival indicates a rather limited clinical relevance of such codon 72 polymorphisms.
Discussion
Recently, a sequence polymorphism at codon 72 the p53 gene (exon 4) encoding either arginine (CGC) or proline (CCC) was suggested to result in a drastically altered biological and biochemical behaviour of p53 in vitro. Compared to the proline encoding allele, the arginine allele appeared to trigger a more pronounced apoptosis response, whereas the proline allele induced significantly more G1 arrest [8,10,13].
The question of clinical relevance was addressed in a few case control studies. There, the proline allele was associated with urothelial [20], thyroid [21,22], and colorectal carcinomas [23] and chronic myeloid leukemia [24], whereas homozygosity for arginine was associated with advanced lung cancer [25]. Concerning patient outcome, in Italian breast cancer patients, the retention of an arginine allele was correlated with a reduction of survival in one study [26]. Another study addressed the clinical relevance of this SNP for treatment response in solid tumors. There, the homozygous proline genotype was correlated with impaired response to radiochemotherapy and reduced survival in head and neck carcinoma [13]. Furthermore, p53 protein encoded by the arginine allele appears to be more susceptible to HPV-E6 protein-induced degradation [27]. Although many studies investigated the relationship of the allelic distribution and susceptibility to HPV-associated cervical carcinoma, two recent meta-analyses could not establish such a correlation with the p53 gene codon 72 SNP [28,29].
In B-CLL, the p53 gene is known to be of clinical relevance concerning survival and treatment response [14]. We therefore investigated the potential clinical relevance of the p53 gene codon 72 SNP in a cohort of 138 samples from patients with B-CLL with respect to survival and drug sensitivity.
The distribution of the proline and arginine alleles in the B-CLL samples was, however, not different of frequency distributions observed in the general population [7]. Moreover, we found no correlation between codon 72 p53 SNP and p53 Exon 5–8 mutations or the IgVH hypermutation status or clinical Binet stage. Likewise, survival was not different in the p53 codon 72 SNP subgroups. Furthermore, no correlation with in vitro drug sensitivities was seen. The trend for reduced sensitivity to irradiation in case of homozygosity for proline is in accordance with a previous report in head and neck cancer [7], but fails to reach statistical significance. These data indicate that the codon 72 SNPs of p53 have per se no clinical relevance, at least in B-CLL. This may indicate either that B-CLL differs significantly from other tumors with regard to the regulation of cell death induced by cytotoxic anticancer therapies or that positive reports from other studies in clinical samples are false positive, eg. due to small sample size.
Conclusion
Although p53 gene mutations in the DNA-binding domain are clinically relevant for patients with B-CLL, the p53 gene codon 72 SNP was not found to be of clinical relevance in patients with B-CLL, neither for patient survival nor for ex vivo sensitivity for cytotoxic drugs and γ-irradiation.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
IS compiled the study, carried out the p53 genetic analysis, performed the statistical analysis and drafted the manuscript. AB carried out the cell death assays and provided the clinical samples and data, MH analysed the IgH mutational status, BD participated in the design of the study, PTD participated in the 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:
Acknowledgements
We thank Sylvia Scheele, Alison Burlton, Philip Bell, and David Head for expert technical assistance. The following drug companies kindly supplied drugs: Asta (mafosfamide), Schering (fludarabine). This work was funded in part by the following grants: "Schwerpunktprogramm Apoptose" by the Deutsche Krebshilfe and by the Deutsche Forschungsgemeischaft (grants to IS and PTD).
Figures and Tables
Figure 1 Analysis of p53 codon 72 SNP in B-CLL. Silver-stained polyacrylamide gel of 6 B-CLL samples analysed by SSCP-PCR for the codon 72 SNP. Sequencing of the 138 B-CLL samples confirmed the polymorphisms in p53 gene, codon 72 (exon 4): 6.6% of the samples were homozygous for proline (Pro/Pro), 57.4% were homozygous for arginine (Arg/Arg), and 36% were heterozygous with both an arginine and a proline allele (Arg/Pro).
Figure 2 Kaplan-Meier-survival estimates for overall survival for B-CLL patients with an Arg/Arg, Pro/Pro and Arg/Pro p53 codon 72 genotype. A: All patients. Squares and circles indicate censored patients. Log rang test p = 0.9 B: Subgroup of chemotherapy naïve patients ("no pretreatment") (n = 58). Squares and circles indicate censored patients. Log rang test p = 0.3.
Table 1 Resistance to ionising irradiation and cytotoxic drugs in relation to the p53 codon 72 status
all patients p53 wild type
p53 codon 72 Arg/Arg p53 codon 72 Pro/Pro p53 codon 72 Arg/Pro p p53 codon 72 Arg/Arg p53 codon 72 Pro/Pro p53 codon 72 Arg/Pro p
γ-irradiation
(% surviving cells) 36.3 ± 4.7
(n = 49) 50.0 ± 10.1
(n = 7) 24.2 ± 4.8
(n = 26) 0.099 30.6 ± 4.8
(n = 39) 54.7 ± 13.4
(n = 5) 22.9 ± 4.9
(n = 25) 0.14
log10 LC90 chlorambucil 0.73 ± 0.07
(n = 78) 0.76 ± 0.19
(n = 9) 0.62 ± 0.08
(n = 49) 0.63 0.6 ± 0.08
(n = 61) 0.76 ± 0.22
(n = 7) 0.57 ± 0.08
(n = 46) 0.70
log10 LC90 mafosfamide 0.32 ± 0.05
(n = 75) 0.37 ± 0.12
(n = 9) 0.21 ± 0.05
(n = 47) 0.39 0.3 ± 0.06
(n = 59) 0.38 ± 0.16
(n = 7) 0.21 ± 0.05
(n = 44) 0.42
log10 LC90 fludarabine 0.18 ± 0.08
(n = 78) 0.13 ± 0.18
(n = 9) -0.32 ± 0.1
(n = 49) 0.32 0.11 ± 0.1
(n = 61) 0.22 ± 0.23
(n = 7) -0.06 ± 0.01
(n = 46) 0.28
log10 LC90 cladribine -0.76 ± 0.1
(n = 78) -0.85 ± 0.18
(n = 9) -1.0 ± 0.11
(n = 49) 0.10 -0.89 ± 0.1
(n = 61) -0.79 ± 0.22
(n = 7) -1.04 ± 0.11
(n = 46) 0.22
log10 LC90 vincristine 0.13 ± 0.1
(n = 55) 0.01 ± 0.27
(n = 8) -0.22 ± 0.11
(n = 39) 0.71 -0.14 ± 0.11
(n = 40) 0.05 ± 0.33
(n = 6) -0.25 ± 0.11
(n = 36) 0.59
log10 LC90 doxorubicin -0.55 ± 0.04
(n = 78) -0.52 ± 0.06
(n = 9) -0.61 ± 0.05
(n = 49) 0.64 -0.55 ± 0.05
(n = 61) -0.49 ± 0.07
(n = 7) -0.63 ± 0.05
(n = 46) 0.35
log10 LC90 methylprednisolone 0.99 ± 0.14
(n = 78) 1.36 ± 0.45
(n = 9) 0.95 ± 0.18
(n = 48) 0.70 0.99 ± 0.16
(n = 61) 1.52 ± 0.47
(n = 7) 0.95 ± 0.18
(n = 45) 0.57
For the cytostatic drugs (chlorambucil, mafosfamide, fludarabine, cladribine, vincristine, doxorubicin or methylprednisolone) log10 LC90 concentrations were compared. In case of ionizing γ-irradiation with a fixed dose of 2 Gy, percentages of surviving cells were compared. p53 WT: wild type p53 gene. Mean +/- SEM is given. Statistical significance was calculated by means of the Kruskal-Wallis test.
==== Refs
Hemmati PG Gillissen B von Haefen C Wendt J Starck L Güner D Dörken B Daniel PT Adenovirus-mediated overexpression of p14(ARF) induces p53 and Bax-independent apoptosis Oncogene 2002 21 3149 3161 12082630 10.1038/sj.onc.1205458
Hemmati PG Normand G Verdoodt B von Haefen C Hasenjager A Güner D Wendt J Dörken B Daniel PT Loss of p21 disrupts p14ARF-induced G1 cell cycle arrest but augments p14ARF-induced apoptosis in human carcinoma cells Oncogene 2005 24 4114 4128 15750619
Mrozek A Petrowsky H Sturm I Kraus J Hermann S Hauptmann S Lorenz M Dörken B Daniel PT Combined p53/Bax mutation results in extremely poor prognosis in gastric carcinoma with low microsatellite instability Cell Death Differ 2003 10 461 467 12719723 10.1038/sj.cdd.4401193
Normand G Hemmati PG Verdoodt B von Haefen C Wendt J Güner D May E Dörken B Daniel PT p14ARF induces G2 cell cycle arrest in p53- and p21-deficient cells by down-regulating p34cdc2 kinase activity J Biol Chem 2005 280 7118 7130 15582998 10.1074/jbc.M412330200
Rau B Sturm I Lage H Berger S Schneider U Hauptmann S Wust P Riess H Schlag PM Dörken B Daniel PT Dynamic expression profile of p21WAF1/CIP1 and Ki-67 predicts survival in rectal carcinoma treated with preoperative radiochemotherapy J Clin Oncol 2003 21 3391 3401 12885834 10.1200/JCO.2003.07.077
Baptiste N Friedlander P Chen X Prives C The proline-rich domain of p53 is required for cooperation with anti-neoplastic agents to promote apoptosis of tumor cells Oncogene 2002 21 9 21 11791172 10.1038/sj.onc.1205015
Sjalander A Birgander R Kivela A Beckman G p53 polymorphisms and haplotypes in different ethnic groups Hum Hered 1995 45 144 149 7615299
Dumont P Leu JI Della Pietra AC George DL Murphy M The codon 72 polymorphic variants of p53 have markedly different apoptotic potential Nat Genet 2003 33 357 365 12567188 10.1038/ng1093
Thomas M Kalita A Labrecque S Pim D Banks L Matlashewski G Two polymorphic variants of wild-type p53 differ biochemically and biologically Mol Cell Biol 1999 19 1092 1100 9891044
Pim D Banks L p53 polymorphic variants at codon 72 exert different effects on cell cycle progression Int J Cancer 2004 108 196 199 14639602 10.1002/ijc.11548
Marin MC Jost CA Brooks LA Irwin MS O´Nicons J Tidy JA James N McGregor JM Harwood CA Yulug IG Vousden KH Allday MJ Gusterson B Ikawa S Hinds PW Crook T Kaelin WGJ A common polymorphism acts as an intragenic modifier of mutant p53 behavior Nat Genet 2000 25 47 54 10802655 10.1038/75586
Bonafe M Salvioli S Barbi C Mishto M Trapassi C Gemelli C Storci G Olivieri F Monti D Franceschi C p53 codon 72 genotype affects apoptosis by cytosine arabinoside in blood leukocytes Biochem Biophys Res Commun 2002 299 539 541 12459171 10.1016/S0006-291X(02)02691-8
Sullivan A Syed N Gasco M Bergamaschi D Trigiante G Attard M Hiller L Farrell PJ Smith P Lu X Crook T Polymorphism in wild-type p53 modulates response to chemotherapy in vitro and in vivo Oncogene 2004 23 3328 3337 15077186 10.1038/sj.onc.1207428
Sturm I Bosanquet AG Hermann S Güner D Dörken B Daniel PT Mutation of p53 and consecutive selective drug resistance in B-CLL occurs as a consequence of prior DNA-damaging chemotherapy Cell Death Differ 2003 10 477 484 12719725 10.1038/sj.cdd.4401194
Sturm I Köhne CH Wolff G Petrowsky H Hillebrand T Hauptmann S Lorenz M Dörken B Daniel PT Analysis of the p53/BAX pathway in colorectal cancer: low BAX is a negative prognostic factor in patients with resected liver metastases J Clin Oncol 1999 17 1364 1374 10334520
Sturm I Petrowsky H Volz R Lorenz M Radetzki S Hillebrand T Wolff G Hauptmann S Dörken B Daniel PT Analysis of p53/BAX/p16(ink4a/CDKN2) in esophageal squamous cell carcinoma: high BAX and p16(ink4a/CDKN2) identifies patients with good prognosis J Clin Oncol 2001 19 2272 2281 11304781
Bosanquet AG Bell PB Enhanced ex vivo drug sensitivity testing of chronic lymphocytic leukaemia using refined DiSC assay methodology Leuk Res 1996 20 143 153 8628013 10.1016/0145-2126(95)00127-1
Bosanquet AG Bell PB Novel ex vivo analysis of nonclassical, pleiotropic drug resistance and collateral sensitivity induced by therapy provides a rationale for treatment strategies in chronic lymphocytic leukemia Blood 1996 87 1962 1971 8634445
Bosanquet AG Bosanquet MI Ex vivo assessment of drug response by differential staining cytotoxicity (DiSC) assay suggests a biological basis for equality of chemotherapy irrespective of age for patients with chronic lymphocytic leukaemia Leukemia 2000 14 712 715 10764159 10.1038/sj.leu.2401727
Kuroda Y Tsukino H Nakao H Imai H Katoh T p53 Codon 72 polymorphism and urothelial cancer risk Cancer Lett 2003 189 77 83 12445680 10.1016/S0304-3835(02)00518-9
Granja F Morari J Morari EC Correa LA Assumpcao LV Ward LS Proline homozygosity in codon 72 of p53 is a factor of susceptibility for thyroid cancer Cancer Lett 2004 210 151 157 15183530 10.1016/j.canlet.2004.01.016
Boltze C Roessner A Landt O Szibor R Peters B Schneider-Stock R Homozygous proline at codon 72 of p53 as a potential risk factor favoring the development of undifferentiated thyroid carcinoma Int J Oncol 2002 21 1151 1154 12370767
Gemignani F Moreno V Landi S Moullan N Chabrier A Gutierrez-Enriquez S Hall J Guino E Peinado MA Capella G Canzian F A TP53 polymorphism is associated with increased risk of colorectal cancer and with reduced levels of TP53 mRNA Oncogene 2004 23 1954 1956 14647431 10.1038/sj.onc.1207305
Bergamaschi G Merante S Orlandi E Galli A Bernasconi P Cazzola M TP53 codon 72 polymorphism in patients with chronic myeloid leukemia Haematologica 2004 89 868 869 15257943
Papadakis ED Soulitzis N Spandidos DA Association of p53 codon 72 polymorphism with advanced lung cancer: the Arg allele is preferentially retained in tumours arising in Arg/Pro germline heterozygotes Br J Cancer 2002 87 1013 1018 12434294 10.1038/sj.bjc.6600595
Bonafe M Ceccarelli C Farabegoli F Santini D Taffurelli M Barbi C Marzi E Trapassi C Storci G Olivieri F Franceschi C Retention of the p53 codon 72 arginine allele is associated with a reduction of disease-free and overall survival in arginine/proline heterozygous breast cancer patients Clin Cancer Res 2003 9 4860 4864 14581358
Storey A Thomas M Kalita A Harwood C Gardiol D Mantovani F Breuer J Leigh IM Matlashewski G Banks L Role of a p53 polymorphism in the development of human papillomavirus-associated cancer Nature 1998 393 229 234 9607760 10.1038/30400
Koushik A Platt RW Franco EL p53 codon 72 polymorphism and cervical neoplasia: a meta-analysis review Cancer Epidemiol Biomarkers Prev 2004 13 11 22 14744727
Jee SH Won SY Yun JE Lee JE Park JS Ji SS Polymorphism p53 codon-72 and invasive cervical cancer: a meta-analysis Int J Gynaecol Obstet 2004 85 301 308 15145278 10.1016/j.ijgo.2003.08.017
|
16109171
|
PMC1208864
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Aug 18; 5:105
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-105
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-1061611148410.1186/1471-2407-5-106Case ReportA case repot of Merkel cell carcinoma on chronic lymphocytic leukemia: differential diagnosis of coexisting lymphadenopathy and indications for early aggressive treatment Papageorgiou KI [email protected] MG [email protected] St Andrews Center of Burns and Plastic Surgery, Broomfield Hospital, Court Road, Chelmsford, CM1 7ET, UK2 Plastic and Reconstructive Surgery Department, "G. Gennimatas" 6th IKA Oncological Hospital, 11473 Athens, Greece2005 19 8 2005 5 106 106 30 3 2005 19 8 2005 Copyright © 2005 Papageorgiou and Kaniorou-Larai; 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
Chronic lymphocytic leukemia (CLL) is a monoclonal disorder, characterized by a progressive proliferation of functionally incompetent B lymphocytes. There is increased evidence of association between CLL and skin cancers, including the uncommon Merkel cell carcinoma (MCC).
Case presentation
A case report of an 84-year old male, who presented with an aggressively recurrent form of MCC on the lower lip, on the background of an 8-year history of untreated CLL. During the recurrences of MCC, coexisting regional lymphadenopathy, posed a problem in the differential diagnosis and treatment of lymph node involvement. Histopathology and immunoistochemistry showed that submandibular lymphadenopathy coexisting with the second recurrence of MCC, was due to B-cell small lymphocytic lymphoma. The subsequent and more aggressive recurrence of the skin tumor had involved the superficial and deep cervical lymph nodes. Surgical excision followed by involved field radiation therapy has been proven effective for both malignancies.
Conclusion
MCC has a high incidence of regional lymphadenopathy at presentation (12–45%) and even when it arises on the background of chronic leucemia, lymphadenopathy at presentation should be managed agressively with elective lymph node dissection. We overview the postulated correlation between Merkel tumor and CCL, the differential diagnosis of regional lymphadenopathy during the recurrences of the skin tumor and the strategies of treatment
==== Body
Background
MCC, is an uncommon tumor, which mostly occurs as an asymptomatic, solitary, firm and red-pink nodule. It has been linked to increased sun exposure [1], both in its anatomic and geographical distribution. Usually is nonulcerated and ranges from 0.8 to 4 cm in diameter [2,3].
Predominantly involves the head and neck region (65%) [4,5], followed by the upper extremities (18%), and the lower extremities (13%) [3]. The precice origin is still controversial. However, the Merkel cell (assumed to be a touch receptor) and the melanocyte are the cutaneous counterparts of the Amine Precurcor Uptake and Decarboxylation (APUD) cells, which are of neuroectodermal origin [6,7].
The tumor is composed of small blue cells with hyperchromatic nuclei and minimal cytoplasm. Mitoses, nuclear fragments and lymphovascular invasion are almost invariable features [8]. Immunoistochemistry for cytokeratin species more typical of simple epitelial cells than keratinocytes, permits a differential identification of Merkel cells in tissue sections [9]. The immunoistochemical profile is characterized of positivity for Neuron-Specific Enolase (NSE), Neurofilament Protein (NFP) and CD57- CD99 [10]. A single punctate zone of cytoplasmic immunoreactivity for cytokeratins especially CK20 is more characteristic [11]. The reaction for CK20 has been used as a finding against metastatic small cell carcinoma of the lung (Cytokeratin 7+ and Thyroid Transcription Factor 1+), small cell melanoma (S100+) and lymphoma (Leucocyte Common Antigen +) [12].
MCC appears to metastasize principally via the lymphatics in a predictable stepwise fashion, with an initial involvement of the regional lymph nodes and subsequent systemic spread [13]. If the lymph nodes are not palpable, the pathological examination of the primary tumor (larger than 2 cm, high mitotic rate and lymphatic invasion) serves as a parameter for the need of lymph node biopsies [5]. Surgical excision with tumor free margins is the primary therapy for stage I-localized disease, with a 64% of survival in 5 years. However recent studies have shown that there is no clear evidence of a difference in survival [2,14], when the margins of resection are less than the 2–3 cm as generally recommended [2,13]
In fact, Mohs micrographic surgery may be more efficacious than wide excision as it inspects all major borders, including the deep margin (MCC often shows extensive vertical growth) and also allows maximal sparing of normal additional tissue, especially in such cosmetically sensitive anatomic areas as the face [15]. A useful adjunct in the treatment of MCC is the sentinel node mapping, which identifies the status of the first draining lymph node and allows to avoid unnecessary lymphadenectomies and the resulting postoperative morbidity [16].
MCC cells are radiosensitive, and several studies have argued for the benefits of radiation therapy not only after resection for local recurrence and palliation but as well as adjuvant treatment after initial surgery with curative intent [12]. Chemotherapy is the least studied theraupetic component and propably is mandatory in unresectable or unmanagable by radiotherapy tumors, as well as in metastatic disease [8,17].
CLL occurs primarily in middle-aged and elderly individuals, with increasing frequency in successive decades of life. It involves slow proliferation and accumulation of incompetent B-lymphocytes and concurrent abnormalities in both humoral and cellular immunity. The majority of patients live 5–10 years, with an initial course that is relatively benign but followed by a terminal, progressive and resistant fase lasting 1–2 years. During the later phase, morbidity is considerable both from the disease and the associated incidence of malignant neoplasms (especially skin cancers as squamous cell carcinoma, Kaposi sarcoma and melanoma) [19]. CLL shares a clinical and morphological overlap with Small Lymphocytic Lymphoma (SLL). In fact if a patient has an absolute lymphocytosis of > 5000/mm3 in the peripheral blood, CLL is diagnosed, regardless of the findings in the lymph node [19].
Treatment of CCL ranges from periodic observation with treatment of infectious, hemorrhagic or immunologic complications to a variety of therapeutic options, including steroids, alkylating agents, purine analogues, combination chemotherapy, monoclonal antibodies and transplant options [20]. As it occurs in an elderly population, progresses slowly and generally is not curable, it is usually treated in a conservative fashion [21]. Involved-field radiation therapy with relatively low doses of radiation can effect an excellent response for both CLL and Small Lymphocytic Lymphoma, especially when the lymphoma cells are contained in one or two areas of lymph nodes in the same part of the body.
Case presentation
An 84-year old white male, presented at the Department of Dermatology (6th IKA Oncological Hospital of Athens) with a pale, ulcerated lesion (1.3 cm in diameter), on the middle of his lower lip. There was no associated lymphadenopathy and an excisional biopsy was performed. Histopathological and immunoistochemical features revealed a Merkel cell carcinoma (MCC) but as the excision was incomplete the patient was sheduled for a wider excision in the following 2 months. In the meantime, the lesion recurred and the patient returned with a protruding white lesion of 1.1 cm in diameter. There was no associated lynphadenopathy and a wider excision, with an 8 mm margin, was performed. Histopathology confirmed the nature of MCC and the second excision was within healthy margins.
Two months later, the patient was referred to the Depatment of Plastic Surgery, for another protruding ulcerated lesion, 3 cm in diameter, on his lower lip (Fig 1). On examination, multiple palpable lymph nodes in the submandibular and cervical area (superficial and deep cervical lymphadenopathy) were present.
Past medical history revealed that 8 years earlier, the patient had been diagnosed as having chronic lymphocytic leukemia (CLL), (nodular and intermediate type), but he didn't receive any treatment. CT scan of the head and neck area, showed a soft tissue lobular mass, 3 cm in diameter, on the lower lip, with a possible extension to the mandible. The past medical history of CLL with the recent occurence of MCC posed a problem in the differential diagnosis of the patient's lymphadenopathy. A W-excision (4 × 3.5 × 1 cm) of the lip lesion was performed, with an open biopsy of one submandibular lymph node.
Histopathology, confirmed recurrence of MCC. An undifferentiated small cell carcinoma with hyaluronated stroma was identified. The cells arranged in nodules or rosettes, had dense nuclear chromatin, with mitoses and nuclear debris which are regular features of MCC (Fig. 2, 3).
Immunoistochemical procedures showed Neuron Specific Enolase (NSE) positivity, while antibodies for Epithelial Membrane Antigen (EMA) and Chromogranin were negative. The excision was described as complete. The submandibular lymph node was positive for malignacy but was associated with the CLL (non Hodgkin's, B cell small lymhocytic lymphoma). There was no evidence of metastatic infiltration by MCC and this was confirmed immunoistochemically with the positive expression of CD5 and CD20 antibody and negative expression of CD10 antibody and NSE. Due to the age of the patient, chemotherapy was not considered.
A month later, a new CT scan of the head and neck, depicted a soft tissue mass, consistent with recurrence of MCC, between the left angle of the mandible and the hyoid bone. Multiple enlarged superficial and deep cervical lymph nodes were present. Fine needle aspiration (FNA) of the submandibular swelling was performed and confirmed MCC (small atypical cells were found, isolated or forming rosettes and exhibited dense core granules of chromatin and scanty cytoplasm). The patient underwent one month of neck radiotherapy with Cobalt 60, a total dose of 4600cGy in 23 days. A further boost of 600cGy in 2 days, on the left submandibular area was administered.
The treatment was successfully completed with full remission of the cervical lymphadenopathy. Two months following the radiotherapy, a new CT scan of the head and neck showed reduction and obscurrence of the pre-existing mass on the left mandibular area, while the lymph nodes were smaller too. The patient is on regular follow up and CLL status is stable with no evidence of progression or further recurrence of MCC 9 months post- radiotherapy.
Conclusion
Our patient represents another case of MCC arising on CLL and this ocurrence re-inforces the postulated correlation between these 2 malignancies. There are two main aetiological factors associated with increased risk of skin cancers: Ultraviolet radiation and immunosupression. CLL is thought to cause immunosupression and alcylating agents and other immunosuppressant therapies may be involved [22]. A higher incidence of MCC is seen in patients with organ transplantation, with human immunodeficiency virus-1 infection or with advanced cancer and anergic status [23].
On the other side, CLL is relatively common and the association with skin cancers, including the rarer MCC, may be coincidental [24]. The submandibular lymphadenopathy on the second recurrence was unexpectedly due to lymphoma and this complicated the decision on the modality of treatment and further management of the patient. Fortunately both tumors responded to field radiation therapy and on the last follow up there was no evidence of deterioration of the leukemic status or MCC metastatic spread. Probably, a wider excision during the initial diagnosis of the skin tumor (with at least a 1.5 cm margin) [13], should have led to an earlier regression of MCC. However, Gillenwater et al [25] demonstrated no difference in outcome based on margins < 1 cm, 1 to 2 cm and > 2 cm. On the other sider, considering the agressiveness of MCC, radiation therapy should have been involved during the second recurrence, although lymphadenopathy was then positive only for CLL related lymphoma. However, as the subsequent lymph node recurrence of MCC occurred only one month later, it raises the possibility that it was already present when the biopsy was performed and the biopsy result was falsely reassuring.
Complete spontaneous regression constitutes 1.67% of the approximately 600 reported cases of MCC[26]. Therefore, MCC once recognized has to be treated agressively. Studies have shown that the best outcome is obtained in patients with regional disease following lymph node dissection with or without subsequent radiation. Our case report highlights the importance of aggressive treatment of MCC with elective lymph node dissection at presentation, even if the coexisting lymphadenopathy could be related to the coexisting CLL
Certainly, on the background of CLL, clinicopathological features and treatment modalities are more complicated and effectiveness depends on the activity status of the leukemia, which could facilitate recurrence and regional or distant metastatic spread.
Abreviations
Small Lymphocytic Lymphoma (SLL).
Merkel Cell Carcinoma (MCC)
Chronic Lymphocytic Leucemia (CLL)
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
KP conceived of the case, participated in the sequence alignment, drafted and revised the manuscript before and following the peer review. M K-l conceived of the case, had access to the data, participated in the sequence alignment and drafted the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Figure 1 Clinical appearance of second recurrence of Merkel tumor on lower lip.
Figure 2 Histopathology showing undifferentiated small cell carcinoma with paucicellular stroma and cells forming nodules or rosettes (haematoxylin and eosin; × 100).
Figure 3 Histopathology showing Merkel cells arranged in rosettes with mitoses present (haematoxylin eosin; × 100).
==== Refs
Miller RW Rabkin CS Merkel cell carcinoma and melanoma: etiological similarities and differences Cancer Epidemiol Biomarkers Prev 1999 8 153 158 10067813
Ratner D Nelson BR Brown MD Johnson TM Merkel cell carcinoma J Am Acad Dermatol 1993 29 143 156 8335732
Raaf JH Urmacher C Knapper WK Shiu MH Cheng EW Trabecular (Merkel cell) carcinoma of the skin Cancer 1986 57 178 3940617
Goepfert H Remmer D Silva E Wheeler B Merkel cell carcinoma (Endocrine carcinoma of the skin) of the head and neck Arch Otolaryngol 1984 110 707 712 6487123
Sibley R Dehner L Rosai J Primary neuroendocrine (Merkel cell?) carcinoma of the skin Am J Surg Pathol 1985 9 95 108 2579594
O'Connor WJ Roenigk RK Broadland DG Merkel cell carcinoma: comparison of Mohs micrographic surgery and wide excision in eighty-six patients Dermatol Surg 1997 23 929 33 9357504 10.1016/S1076-0512(97)00292-6
Tong CK Toker C Trabecular carcinoma of the skin: An ultrastructural study Cancer 1978 42 2311 2321 719609
Gollard R Weber R Kosty MP Greenway HT Massullo V Humberson C Merkel cell carcinoma: Review of 22 cases with surgical, pathologic and theraupetic considerations Cancer 2000 88 1842 51 10760761 10.1002/(SICI)1097-0142(20000415)88:8<1842::AID-CNCR13>3.0.CO;2-P
Moll R Moll I Franke WW Identification of Merkel cell in human skin by specific cytocheratin antibodies: Changes in cell density and distribution in fetal and adult plantar epidermis Differentiation 1984 28 136 6084624
Michels S Swanson PE Robb JA Wick MR Leu-7 in small cell neoplasms: An immunoistochemical study with ultrastructural correlations Cancer 1987 60 2958 3499971
Merot Y Margolis RG Dahl D Saurat JH Mihm MC Jr Coexpression of neurofilament and keratin proteins in cutaneous neuroendocrine carcinoma cells J Invest Dermatol 1986 86 74 2427595 10.1111/1523-1747.ep12283862
Goesling W McKee PH Mayer RJ Merkel cell carcinoma J Clin Oncol 2002 20 588 598 11786590 10.1200/JCO.20.2.588
Yiengpruksawan A Coit DG Thaler HT Urmacher C Knapper WK Merkel cell carcinoma: prognosis and management Arch Surg 1991 126 1514 9 1842182
Ott MJ Tanabe KK Gadd MA Stark P Smith BL Finkelstein DM Souba WW Multimodality management of Merkel cell carcinoma Arch Surg 1999 134 388 93 10199311 10.1001/archsurg.134.4.388
Rubsamen PE Tanenbaum E Grove AS Gould E Merkel cell carcinoma of the eyelid and the periocular tissues Am J Ophthalmol 1992 113 674 80 1598958
Messina JL Reintgen DS Cruse CW Rappaport DP Berman C Fenske NA Glass LF Selective lymphadenopathy in patients with Merkel cell cutaneous neuroendocrine carcinoma Ann Surg Oncol 1997 4 389 95 9259965
Fening E Brenner B Katz A Rakovsky E Hana MB Sulkes A The role of radiation therapy and chemotherapy in the treatment of Merkel cell carcinoma Cancer 1997 80 881 85 9307187 10.1002/(SICI)1097-0142(19970901)80:5<881::AID-CNCR8>3.0.CO;2-O
Barnerji R Byrd CJ Update in the biology of chronic lymphocytic leukemia Curr Opin Oncol 2000 12 22 9 10687725 10.1097/00001622-200001000-00004
Levi F Randimbison L Te V-C La Vecchia C Non Hodgkin's lymphomas, chronic lymphocytic leukemias and skin cancers Br J Cancer 1996 74 1847 1850 8956805
Keating MJ Chronic Lymphocytic Leukemia Semin Oncol 1999 26 107 14 10561025
Faguet GB Chronic Lymphocytic Leukemia: an updated review J Clin Oncol 1994 12 1974 90 8083719
Adami J Frisch M Yuen J Glimelius B Melbye M Evidence of an association between non-Hodgkin's lymphoma and skin cancer BMJ 1995 310 1491 5 7787593
Brenner B Sulkes A Rakowsky E Feinmesser M Yukelson A Bar-Haim E Katz A Idelevich E Neuman A Barhana M Fenig E Second neoplasm in patients with Merkel cell carcinoma Cancer 2001 91 1358 62 11283937 10.1002/1097-0142(20010401)91:7<1358::AID-CNCR1139>3.0.CO;2-C
Ziprin P Smith S Salerno G Rosin RD Two cases of Merkel cell tumor arising in patients with chronic lymphocytic leukaemia Br J Dermatol 2000 142 525 28 10735964 10.1046/j.1365-2133.2000.03370.x
Gillenwater AM Hessel AC Morrison WH Burgess M Silva EG Roberts D Goepfert H Merkel cell carcinoma of the head and neck: Effect of surgical excision and radiation on recurrence and survival Arch Otolaryngol Head Neck Surg 2001 127 149 154 11177031
Brown TJ Brooke AJ Macfarlane DF Goldberg LH Merkel cell carcinoma: spontaneous resolution and management of metastatic disease Dermatol Surg 1999 25 23 5 9935088 10.1046/j.1524-4725.1999.08142.x
|
16111484
|
PMC1208865
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Aug 19; 5:106
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-106
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-1071611148810.1186/1471-2407-5-107Research ArticlePhosphorylation states of cell cycle and DNA repair proteins can be altered by the nsSNPs Savas Sevtap [email protected] Hilmi [email protected] Fred A. Litwin Centre for Cancer Genetics, Samuel Lunenfeld Research Institute, Mount Sinai Hospital, Toronto, M5G 1X5, ON, Canada2 Department of Pathology and Laboratory Medicine, Mount Sinai Hospital, Toronto, M5G 1X5, ON, Canada3 Department of Laboratory Medicine and Pathobiology, University of Toronto, M5G 1L5, Toronto, ON, Canada2005 19 8 2005 5 107 107 24 11 2004 19 8 2005 Copyright © 2005 Savas and Ozcelik; 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
Phosphorylation is a reversible post-translational modification that affects the intrinsic properties of proteins, such as structure and function. Non-synonymous single nucleotide polymorphisms (nsSNPs) result in the substitution of the encoded amino acids and thus are likely to alter the phosphorylation motifs in the proteins.
Methods
In this study, we used the web-based NetPhos tool to predict candidate nsSNPs that either introduce or remove putative phosphorylation sites in proteins that act in DNA repair and cell cycle pathways.
Results
Our results demonstrated that a total of 15 nsSNPs (16.9%) were likely to alter the putative phosphorylation patterns of 14 proteins. Three of these SNPs (CDKN1A-S31R, OGG1-S326C, and XRCC3-T241M) have already found to be associated with altered cancer risk. We believe that this set of nsSNPs constitutes an excellent resource for further molecular and genetic analyses.
Conclusion
The novel systematic approach used in this study will accelerate the understanding of how naturally occurring human SNPs may alter protein function through the modification of phosphorylation mechanisms and contribute to disease susceptibility.
==== Body
Background
Phosphorylation is a common, reversible post-translational modification that occurs at serine (S), threonine (T), and tyrosine (Y) residues in proteins [1]. Overall, phosphorylation can alter the structure, function, interaction, stability, and the sub-cellular location of the proteins [2-4], and therefore play an indispensable role in regulation of the cellular processes such as signal transduction, gene expression, cytoskeletal regulation, apoptosis, homeostasis, cell cycle, and DNA damage recognition and repair [5-11]. The phosphorylation state of a protein is determined by the opposing actions of kinases and phosphatases [12]. Proteins may contain multiple phosphorylation sites, which may be targeted by different kinases/phosphatases [2]. The activity of kinases and phosphatases at different times and/or upon different stimuli provides a means of powerful control over the protein phosphorylation state and thus the biological processes the protein is involved in.
In the post-genomic era, there is an expanding interest in identification of the single nucleotide polymorphisms (SNPs) that might affect the protein function and thus contribute to the disease susceptibility. The non-synonymous SNPs (nsSNPs) substitute encoded amino acids in proteins, and therefore are good candidates as disease-modifiers. A variety of approaches have been developed and applied, based on criteria such as the evolutionary conservation status or structural parameters, to characterize and select the nsSNPs that are most likely to have functional consequences [13-19].
In this report, we predicted the potential effect of a set of nsSNPs [20,21] in altering the phosphorylation status of DNA repair and cell cycle proteins using the NetPhos tool [22], which is an artificial neural network method that predicts the phosphorylation sites with a sensitivity of 69–96%. DNA repair and cell cycle pathways interact during the cell growth and division to maintain the genomic stability of dividing cells. Abnormalities in the DNA repair and/or the cell cycle pathways can lead to abnormal cell growth/division or cellular death [23], and are implicated in many human diseases, including cancer [24-30]. Functional significance of many phosphorylated residues of several DNA repair and cell cycle proteins has already been evaluated. For example, phosphorylation of STATα residue S727 is required for its maximal transcriptional activation [31] and enhances its binding to the BRCA1 protein [32]. Similarly, phosphorylation of S383 and S387 are required for the FANCG function during mitosis [33]. Likewise, mutations of the phosphorylated residues Ser366 and Thr387 of p53 affect its transactivation function [34]. To our knowledge, although SNPs of DNA repair and cell cycle proteins have already been shown to contribute to cancer risk [35-37], the potential role of nsSNPs in alteration of phosphorylation patterns of proteins has not been evaluated before. Therefore, the novel approach described in this study will accelerate the formation of a bridge between variations in DNA repair/cell cycle function and predisposition to disease.
Methods
The nsSNPs extracted from public SNP databases were previously reported [20,21], however, only the nsSNPs that were found in ≥2 chromosomes in a sample panel of ≥46 chromosomes were included into that manuscript. A total of 89 nsSNPs from 47 genes involved in DNA repair and cell cycle constituted the final data set. The NetPhos [22] algorithm was utilized to predict putative phosphorylation sites for both the wild type and the variant protein sequences. Only the predictions that remove or create a site at either the SNP location or at kinase recognition motifs are included into this manuscript. Please note that the BRCA1 and NFKB1 proteins were initially identified as cell cycle protein interacting proteins [21]. However, in this manuscript, we classified the BRCA1 as a DNA repair and the NFKB1 as a cell cycle protein. The mouse orthologues were retrieved from the LocusLink resource of NCBI [38] and aligned with the human proteins using the ClustalW program [39] to identify the corresponding mouse residue.
Results and discussion
We utilized the NetPhos algorithm to predict putative phosphorylation sites along the DNA repair and cell cycle proteins, and studied whether 89 naturally occurring nsSNPs (64 from 28 DNA repair and 25 from 19 cell cycle genes) might alter the phosphorylation patterns in these proteins. The sensitivity of NetPhos prediction has been reported to be 69–96% with a false-positive prediction rate of 0–26% for Y, 0–11% for S, and 0–14% for T [22]. The results obtained using the NetPhos software are shown in Table I, and are summarized in Table II. Our results have shown that 16.9% (15/89) of the nsSNPs studied are likely to abolish or create 17 putative phosphorylation sites in 44.0% (14/32) of the proteins. As summarized in Table II, five nsSNPs (ERCC5-S311C, OGG1-S326C, XRCC3-T241M, CCND3-S259A, and CDKN1A-S31R) were predicted to abolish putative phosphorylation sites, whereas four nsSNPs were predicted to create putative phosphorylation sites in the proteins (ERCC2-H201Y, ERCC4-P379S, LIG4-P231S, and XRCC1-P309S). These nsSNPs resulted in the addition or removal of a S, T or Y residue at the predicted phosphorylation site.
The kinase recognition/interaction motif involves 7–12 amino acids around the phosphorylated residue [40], and the physicochemical characteristics of these amino acids determine the specificity of the protein kinases [41,42]. Thus, the amino acid substitutions within the kinase recognition motifs are likely to influence the substrate recognition and the subsequent phosphorylation by kinases. Accordingly, we have identified six nsSNPs (Table I, II) located within the phosphorylation motif of six proteins (within 4 amino acids on either side of the putative phosphorylated residue based on NetPhos outputs) that abolished eight putative phosphorylation sites (BRCA1-P871L at S868, BRCA1-S1040N at S1041, ERCC5-S311C at S310, IGHMBP2-T671A at S672, WRN-S1079L at S1083 and at S1084, CCNI-V207I at S208, and NFKB1-H712Q at T716). Interestingly, NetPhos predicts two overlapping phosphorylation motifs for the ERCC5-S311C nsSNP (S311 SLPSSSKMH and S310 ESLPSSSKM), which are both completely abolished by the substitution of the serine residue (position 311) with a cysteine (Table I). Similarly, the WRN-S1079L nsSNP was also predicted to remove 2 putative overlapping phosphorylation motifs (S1083 SKTVSSGTK and S1084 KTVSSGTKE) simultaneously.
The Swiss-Prot [43], HPRD [44], PhosphoBase [45], and Phospho.ELM [46] databases and the existing literature did not reveal any experimentally verified phosphorylation at the predicted sites. Analysis of the mouse orthologues showed that the corresponding amino acids at the BRCA1-S1041, CCNI-S208, ERCC5-S310, IGHMBP2-S672, WRN-S1083 and XRCC3-T241 residues were also predicted to be phosphorylated, suggesting that these motifs/sites might have been evolutionarily conserved between two species. On the other hand, the remaining phosphorylation sites, which are not detected in mouse proteins, may represent the newly evolved phosphorylation motifs in human. However, considering the false-positive rate of NetPhos as well as the possibility that the negative selection acting on the nsSNP sites can result in higher false-positive rates, we cannot totally rule out that all predictions in Table 2 are false. Yet these predictions are still of a great value and suggest possible phosphorylation sites that can be experimentally evaluated. In future, when sufficient molecular data regarding the phosphorylation status of orthologous proteins is available, more systematic analyses can be performed to maximize the accuracy of phosphorylation predictions.
We have also performed an extensive literature review to investigate the role of the reported nsSNPs (minor allele frequencies ≥5%) in human cancer predisposition (Table III). Supporting our hypothesis, three SNPs (CDKN1A-S31R, OGG1-S326C, and XRCC3-T241M) have already found to be associated with altered cancer risk. XRCC3-T241M nsSNP was reported to be associated with increased breast cancer [47,48] and melanoma risk [49], and was also found to be protective against bladder cancer in heavy smokers [50]. XRCC3 is a key DNA repair protein involved in base excision repair [29] and is involved in repairing the alterations caused by many DNA damaging agents. Recently, the XRCC3-M241 variant has been associated with increased risk of incidence of tetraploid cells, frequently observed in cancers, through affecting the function of the XRCC3- and Rad52-associated RPA protein [51]. Similarly, the OGG1-S326C SNP was found to be associated with increased lung [52], orolaryngeal and esophageal cancer risk [53,54]. OGG1 is a DNA repair protein that is protective against the mutations induced by the 8-hydroxyguanine. Yamane et al., [55] suggested that OGG1-C326, when compared to OGG1-S326, was associated with a lower repair capacity for 8-hydroxyguanine induced mutations in human cells. In the case of CDKN1A-S31R, the CDKN1A-S31 was suggested to be associated with increased endometrial cancer [56] whereas CDKN1A-R31 was associated with increased primary open-angle glaucoma [57] and esophageal cancer risk [58]. The CDKN1A-R31 form of the protein was not significantly different than the CDKN1A-S31 form in terms of its ability to suppress colony formation [59]. However, it is not clear whether this result would suggest that the CDKN1A-R31 would be functionally equivalent to the wild type allele in other diverse cellular mechanisms that the CDKN1A protein is involved in, such as apoptosis, cell migration, and senescence [60,61].
In addition to the SNPs already implicated in cancer risk, we identified one relatively common nsSNP potentially altering the phosphorylation pattern of a major breast and ovarian cancer susceptibility gene, BRCA1. The BRCA1-P871L SNP was not found to be associated with either breast [62] or ovarian cancer risk [63], however, further analyses is required to see whether this nsSNP or the other nsSNPs in Table III play a role in susceptibility to other cancer types.
How can we explain that commonly occurring nsSNPs (minor allele frequencies ≥5%) are likely to affect the phosphorylation and thus the function of the proteins? If the phosphorylation site is necessary for the function of the protein and the protein is necessary for the fitness of the organism (indispensable/essential protein), then we would expect such nsSNPs (deleterious alleles) to be either removed from the population or be kept at low allele frequency by means of the purifying selection. Thus, in this case, one can conclude that the common nsSNPs presented in this report can be falsely predicted as removing/creating putative phosphorylation sites by NetPhos program. However, the allele frequencies of the deleterious alleles from proteins that are essential for fitness get higher than expected when the nsSNPs are a) created by hot-spot mutation mechanism(s), b) subject to balancing selection, too [64]. Alternatively, even though the nsSNPs (and the abolished/created phosphorylation sites) have important impact on the protein function, the protein and/or the altered protein function may not affect the fitness, which can also explain the lack of purifying selection against such nsSNPs and their relatively high minor allele frequencies. Besides, the biological consequences of altered protein function may only be exerted under certain environmental conditions.
Conclusion
Here we report a set of nsSNPs in DNA repair and cell cycle genes that are predicted to alter the phosphorylation motifs of the encoded proteins, with possible consequences on protein function, structure, interaction, and stability. If the nsSNPs with a ≥5% minor allele frequency listed in Table III do indeed alter the phosphorylation state of the corresponding proteins, they then represent important candidates for disease susceptibility studies, especially relating to cancer risk. We conclude with the suggestion that our approach and the resulting data indicate a novel mechanism of SNP action: alteration of the functional characteristics of the proteins through phosphorylation may significantly contribute to our understanding of the molecular basis of complex diseases, such as cancer. This study is unique in the sense that it systematically links the possible post-translational modification functional effects of SNPs to disease (cancer) predisposition.
List of abbreviations
SNP: single nucleotide polymorphism; nsSNP: non-synonymous SNP.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SS participated in design of the study, collected and analyzed the data and prepared the draft of the manuscript. HO participated in the design and coordination of the study, and helped to draft the manuscript. Both authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Authors thank M. Farhan Ahmad, Mehjabeen Shariff, and David Y. Kim for technical assistance. We are indebted to Dr. Nikolaj Blom for his valuable communication about the false-positive rate of the NetPhos program. This work was supported by grants from the Susan Komen Breast Cancer Foundation, USA (BCTR0100627), and the Canadian Breast Cancer Foundation (HO). S. Savas is supported, in part, by a "CIHR Strategic Training Program Grant – The Samuel Lunenfeld Research Institute Training Program: Applying Genomics to Human Health" fellowship.
Figures and Tables
Table 1 nsSNPs that abolish or create putative phosphorylated residues in DNA repair and cell cycle proteins. Only the NetPhos [22] predictions that remove or create a site at either the SNP location or at kinase recognition motifs are shown. The nsSNPs that create or abolish putative phosphorylation sites at the nsSNP position are shown in bold. Under the wild type and variant columns are the NetPhos outputs with the location of the amino acid, the phosphorylation motif (the putative phosphorylated residue is underlined), the score, and the residue being phosphorylated. 1 and 2 under the frequency column represents the nsSNP minor allele frequencies <5% and ≥5%, respectively [20-21]. Please note that the BRCA1 and NFKB1 proteins were initially identified as cell cycle protein interacting proteins [21]. However, in this manuscript, we classified the BRCA1 as a DNA repair and the NFKB1 as a cell cycle protein. §The putative phosphorylation sites that are also predicted in mouse proteins.
Pathway Gene Accession # SNP ID nsSNP Freq. Wild Type Variant
DNA repair BRCA1 NM_007294.1 SNP000007492
rs799917 P871L 2 868 SKRQSFAPF 0.599 *S* -
BRCA1 NM_007294.1 rs4986852 S1040N 1 §1041 KEASSSNIN 0.557 *S* -
ERCC2 NM_000400.1 SNP000000054
rs1799792 H201Y 1 - 201 YSILYANVV 0.745 *Y*
ERCC4 NM_005236.1 SNP000000067 P379S 1 and 2 - 379 LESNSKWEA 0.507 *S*
ERCC5 NM_000123.1 SNP001026027
rs2307491 S311C 1 a) 311 SLPSSSKMH 0.990 *S*
b) §310 ESLPSSSKM 0.645 *S* -
IGHMBP2 NM_002180.1 SNP000012785
rs622082 T671A 2 §672 GPATSTRTG 0.634 *S* -
LIG4 NM_002312.2 rs3093765 P231S 1 - 231 QLHDSSVGL 0.562 *S*
OGG1 NM_002542.4 SNP000064679 S326C 2 326 DLRQSRHAQ 0.990 *S* -
WRN NM_000553.2 SNP001026663
rs3087414 S1079L 1 a) §1083 SKTVSSGTK 0.790 *S*
b) 1084 KTVSSGTKE 0.829 *S* -
XRCC1 NM_006297.1 SNP000064196
rs25491 P309S 1 - 309 EPRRSRAGP 0.996 *S*
XRCC3 NM_005432.2 SNP000000060 T241M 2 §241 SLGATLREL 0.849 *T* -
Cell Cycle CCND3 NM_001760.2 rs1051130 S259A 2 259 LREASQTSS 0.982 *S* -
CCNI NM_006835.2 rs4252903 V207I 1 §208 LAMVSLEME 0.664 *S* -
CDKN1A NM_000389.2 SNP000003435
rs1801270
GAI870831
GAI1503061 S31R 2 31 SEQLSRDCD 0.924 *S* -
NFKB1 NM_003998.2 rs4648099 H712Q 1 716 HVDSTTYDG 0.595 *T* -
Table 2 Distribution of the nsSNPs predicted to alter the phosphorylation sites.
nsSNP DNA repair Cell cycle Total
Abolished ≥1 putative phosphorylated residue (S, T or Y) at the nsSNP location 3 2 5
Abolished ≥1 putative phosphorylated residue by changing the kinase recognition motif 5 2 7
Created ≥1 putative phosphorylated residue (S, T or Y) at the nsSNP location 4 0 4
Created ≥1 putative phosphorylated residue by changing the kinase recognition motif 0 0 0
Table 3 Common nsSNPs with a possible role in cancer predisposition. Only the information derived from the studies on the protein function as well as the studies with a suggestion of disease-association have been included. 1 and 2 under the frequency column represents the nsSNP with minor allele frequencies <5% and ≥5%, respectively [20-21].
Pathway Gene nsSNP Possible effect on phosphorylation Frequency Functional analysis Cancer risk association
DNA repair BRCA1 P871L Abolishes at S869 2 - -
ERCC4 P379S Creates at S379 1 and 2 - -
IGHMBP2 T671A Abolishes at S672 2 - -
OGG1 S326C Abolishes at S326 2 Yamane et al. [55] Sugimura et al. [52];
Xing et al. [53];
Elahi et al. [54]
XRCC3 T241M Abolishes and T241 2 Yoshihara et al. [51] Winsey et al. [49];
Kuschel et al. [47];
Shen et al. [50];
Figueiredo et al. [48]
Cell cycle CCND3 S259A Abolishes at S359 2 - -
CDKN1A S31R Abolishes at S31 2 Chedid et al. [59] Wu et al. [58];
Roh et al. [56];
Tsai et al. [57]
==== Refs
Johnson LN O'Reilly M Control by phosphorylation Curr Opin Struct Biol 1996 6 762 769 8994876 10.1016/S0959-440X(96)80005-4
Cohen P The regulation of protein function by multisite phosphorylation – a 25 year update Trends Biochem Sci 2000 25 596 601 11116185 10.1016/S0968-0004(00)01712-6
Pawson T Specificity in signal transduction: from phosphotyrosine-SH2 domain interactions to complex cellular systems Cell 2004 116 191 203 14744431 10.1016/S0092-8674(03)01077-8
Zhang HG Wang J Yang X Hsu HC Mountz JD Regulation of apoptosis proteins in cancer cells by ubiquitin Oncogene 2004 23 2009 2015 15021888 10.1038/sj.onc.1207373
Szumiel I Monitoring and signaling of radiation-induced damage in mammalian cells Radiat Res 1998 150 S92 101 9806612
Graves JD Krebs EG Protein phosphorylation and signal transduction Pharmacol Ther 1999 82 111 121 10454190 10.1016/S0163-7258(98)00056-4
Hunter T Signaling – 2000 and beyond Cell 2000 100 113 127 10647936 10.1016/S0092-8674(00)81688-8
Henneke G Koundrioukoff S Hubscher U Multiple roles for kinases in DNA replication EMBO Rep 2003 4 252 256 12634841 10.1038/sj.embor.embor774
McGowan CH Regulation of the eukaryotic cell cycle Prog Cell Cycle Res 2003 5 1 4 14593695
Vermeulen K Van Bockstaele DR Berneman ZN The cell cycle: a review of regulation, deregulation and therapeutic targets in cancer Cell Prolif 2003 36 131 149 12814430 10.1046/j.1365-2184.2003.00266.x
Westermann S Weber K Post-translational modifications regulate the microtubule function Nat Rev Mol Cell Biol 2003 4 938 947 14685172 10.1038/nrm1260
Lester LB Scott JD Anchoring and scaffold proteins for kinases and phosphatases Recent Prog Horm Res 1997 52 409 429 9238861
Chasman D Adams RM Predicting the functional consequences of non-synonymous single nucleotide polymorphisms: structure-based assessment of amino acid variation J Mol Biol 2001 307 683 706 11254390 10.1006/jmbi.2001.4510
Ng PC Henikoff S Predicting deleterious amino acid substitutions Genome Res 2001 11 863 874 11337480 10.1101/gr.176601
Ng PC Henikoff S Accounting for human polymorphisms predicted to affect protein function Genome Res 2002 12 436 446 11875032 10.1101/gr.212802
Wang Z Moult J SNPs, protein structure, and disease Hum Mutat 2001 17 263 270 11295823 10.1002/humu.22
Ramensky V Bork P Sunyaev S Human non-synonymous SNPs: server and survey Nucleic Acids Res 2002 30 3894 3900 12202775 10.1093/nar/gkf493
Stitziel NO Tseng YY Pervouchine D Goddeau D Kasif S Liang J Structural location of disease-associated single-nucleotide polymorphisms J Mol Biol 2003 327 1021 1030 12662927 10.1016/S0022-2836(03)00240-7
Clifford RJ Edmonson MN Nguyen C Buetow KH Large-scale analysis of non-synonymous coding region single nucleotide polymorphisms Bioinformatics 2004 20 1006 1014 14751981 10.1093/bioinformatics/bth029
Savas S Kim DY Ahmad MF Shariff M Ozcelik H Identifying functional genetic variants in DNA repair pathways using protein conservation analysis Cancer Epidemiol Biomarkers Prev 2004 13 801 807 15159313
Savas S Ahmad MF Shariff M Kim DY Ozcelik H Candidate nsSNPs that can affect the functions and the interactions of cell cycle proteins Proteins 2005 58 697 705 15617026 10.1002/prot.20367
Blom N Gammeltoft S Brunak S Sequence- and structure-based prediction of eukaryotic protein phosphorylation sites J Mol Biol 1999 294 1351 1362 10600390 10.1006/jmbi.1999.3310
Norbury CJ Zhivotovsky B DNA damage-induced apoptosis Oncogene 2004 23 2797 2808 15077143 10.1038/sj.onc.1207532
Latif C Harvey SH O'Connell MJ Ensuring the stability of the genome: DNA damage checkpoints Scientific World Journal 2001 1 684 702 12805771
Malumbres M Barbacid M To cycle or not to cycle: a critical decision in cancer Nat Rev Cancer 2001 1 222 231 11902577 10.1038/35106065
Duker NJ Chromosome breakage syndromes and cancer Am J Med Genet 2002 115 125 129 12407692 10.1002/ajmg.10688
Thompson LH Schild D Recombinational DNA repair and human disease Mutat Res 2002 509 49 78 12427531
Sarasin A An overview of the mechanisms of mutagenesis and carcinogenesis Mutat Res 2003 544 99 106 14644312
Thacker J Zdzienicka MZ The mammalian XRCC genes: their roles in DNA repair and genetic stability DNA Repair (Amst) 2003 2 655 672 12767346
Motoyama N Naka K DNA damage tumor suppressor genes and genomic instability Curr Opin Genet Dev 2004 14 11 16 15108799 10.1016/j.gde.2003.12.003
Zhu X Wen Z Xu LZ Darnell JE Jr Stat1 serine phosphorylation occurs independently of tyrosine phosphorylation and requires an activated Jak2 kinase Mol Cell Biol 1997 6618 6623 9343425
Ouchi T Lee SW Ouchi M Aaronson SA Horvath CM Collaboration of signal transducer and activator of transcription 1 (STAT1) and BRCA1 in differential regulation of IFN-gamma target genes Proc Natl Acad Sci U S A 2000 97 5208 5213 10792030 10.1073/pnas.080469697
Mi J Qiao F Wilson JB High AA Schroeder MJ Stukenberg PT Moss A Shabanowitz J Hunt DF Jones NJ Kupfer GM FANCG is phosphorylated at serines 383 and 387 during mitosis Mol Cell Biol 2004 24 8576 8585 15367677 10.1128/MCB.24.19.8576-8585.2004
Ou YH Chung PH Sun TP Shieh SY p53 C-Terminal Phosphorylation by CHK1 and CHK2 Participates in the Regulation of DNA-Damage-induced C-Terminal Acetylation Mol Biol Cell 2005 16 1684 1695 15659650 10.1091/mbc.E04-08-0689
Goode EL Ulrich CM Potter JD Polymorphisms in DNA repair genes and associations with cancer risk Cancer Epidemiol Biomarkers Prev 2002 11 1513 1530 12496039
Koike H Suzuki K Satoh T Ohtake N Takei T Nakata S Yamanaka H Cyclin D1 gene polymorphism and familial prostate cancer: the AA genotype of A870G polymorphism is associated with prostate cancer risk in men aged 70 years or older and metastatic stage Anticancer Res 2003 23 4947 4951 14981950
Chang BL Zheng SL Isaacs SD Wiley KE Turner A Li G Walsh PC Meyers DA Isaacs WB Xu J A polymorphism in the CDKN1B gene is associated with increased risk of hereditary prostate cancer Cancer Res 2004 64 1997 1999 15026335
Wheeler DL Church DM Edgar R Federhen S Helmberg W Madden TL Pontius JU Schuler GD Schriml LM Sequeira E Suzek TO Tatusova TA Wagner L Database resources of the National Center for Biotechnology Information: update Nucleic Acids Res 2004 D35 40 14681353 10.1093/nar/gkh073
Chenna R Sugawara H Koike T Lopez R Gibson TJ Higgins DG Thompson JD Multiple sequence alignment with the Clustal series of programs Nucleic Acids Res 2003 31 3497 3500 12824352 10.1093/nar/gkg500
Songyang Z Blechner S Hoagland N Hoekstra MF Piwnica-Worms H Cantley LC Use of an oriented peptide library to determine the optimal substrates of protein kinases Curr Biol 1994 4 973 982 7874496 10.1016/S0960-9822(00)00221-9
Kreegipuu A Blom N Brunak S Jarv J Statistical analysis of protein kinase specificity determinants FEBS Lett 1998 430 45 50 9678592 10.1016/S0014-5793(98)00503-1
Blom N Sicheritz-Ponten T Gupta R Gammeltoft S Brunak S Prediction of post-translational glycosylation and phosphorylation of proteins from the amino acid sequence Proteomics 2004 4 1633 1649 15174133 10.1002/pmic.200300771
O'Donovan C Martin MJ Gattiker A Gasteiger E Bairoch A Apweiler R High-quality protein knowledge resource: SWISS-PROT and TrEMBL Brief Bioinform 2002 3 275 284 12230036
Peri S Navarro JD Amanchy R Kristiansen TZ Jonnalagadda CK Surendranath V Niranjan V Muthusamy B Gandhi T Gronborg M Ibarrola N Deshpande N Shanker K Shivashankar HN Rashmi BP Ramya MA Zhao Z Chandrika KN Padma N Harsha HC Yatish AJ Kavitha MP Menezes M Choudhury DR Suresh S Ghosh N Saravana R Chandran S Krishna S Joy M Anand SK Madavan V Joseph A Wong GW Schiemann WP Constantinescu SN Huang L Khosravi-Far R Steen H Tewari M Ghaffari S Blobe GC Dang CV Garcia JG Pevsner J Jensen ON Roepstorff P Deshpande KS Chinnaiyan AM Hamosh A Chakravarti A Pandey A Development of human protein reference database as an initial platform for approaching systems biology in humans Genome Res 2003 13 2363 2371 14525934 10.1101/gr.1680803
Kreegipuu A Blom N Brunak S PhosphoBase, a database of phosphorylation sites: release 2.0 Nucleic Acids Res 1999 27 237 239 9847189 10.1093/nar/27.1.237
Diella F Cameron S Gemund C Linding R Via A Kuster B Sicheritz-Ponten T Blom N Gibson TJ Phospho.ELM: a database of experimentally verified phosphorylation sites in eukaryotic proteins BMC Bioinformatics 2004 5 79 15212693 10.1186/1471-2105-5-79
Kuschel B Auranen A McBride S Novik KL Antoniou A Lipscombe JM Day NE Easton DF Ponder BA Pharoah PD Dunning A Variants in DNA double-strand break repair genes and breast cancer susceptibility Hum Mol Genet 2002 11 1399 1407 12023982 10.1093/hmg/11.12.1399
Figueiredo JC Knight JA Briollais L Andrulis IL Ozcelik H Polymorphisms XRCC1-R399Q and XRCC3-T241M and the risk of breast cancer at the Ontario site of the Breast Cancer Family Registry Cancer Epidemiol Biomarkers Prev 2004 13 583 591 15066923
Winsey SL Haldar NA Marsh HP Bunce M Marshall SE Harris AL Wojnarowska F Welsh KI A variant within the DNA repair gene XRCC3 is associated with the development of melanoma skin cancer Cancer Res 2000 60 5612 5616 11059748
Shen M Hung RJ Brennan P Malaveille C Donato F Placidi D Carta A Hautefeuille A Boffetta P Porru S Polymorphisms of the DNA repair genes XRCC1, XRCC3, XPD, interaction with environmental exposures, and bladder cancer risk in a case-control study in northern Italy Cancer Epidemiol Biomarkers Prev 2003 12 1234 1240 14652287
Yoshihara T Ishida M Kinomura A Katsura M Tsuruga T Tashiro S Asahara T Miyagawa K XRCC3 deficiency results in a defect in recombination and increased endoreduplication in human cells EMBO J 2004 23 670 680 14749735 10.1038/sj.emboj.7600087
Sugimura H Kohno T Wakai K Nagura K Genka K Igarashi H Morris BJ Baba S Ohno Y Gao C Li Z Wang J Takezaki T Tajima K Varga T Sawaguchi T Lum JK Martinson JJ Tsugane S Iwamasa T Shinmura K Yokota J hOGG1 Ser326Cys polymorphism and lung cancer susceptibility Cancer Epidemiol Biomarkers Prev 1999 8 669 674 10744126
Xing DY Tan W Song N Lin DX Ser326Cys polymorphism in hOGG1 gene and risk of esophageal cancer in a Chinese population Int J Cancer 2001 95 140 143 11307145 10.1002/1097-0215(20010520)95:3<140::AID-IJC1024>3.0.CO;2-2
Elahi A Zheng Z Park J Eyring K McCaffrey T Lazarus P The human OGG1 DNA repair enzyme and its association with orolaryngeal cancer risk Carcinogenesis 2002 23 1229 1234 12117782 10.1093/carcin/23.7.1229
Yamane A Kohno T Ito K Sunaga N Aoki K Yoshimura K Murakami H Nojima Y Yokota J Differential ability of polymorphic OGG1 proteins to suppress mutagenesis induced by 8-hydroxyguanine in human cell in vivo Carcinogenesis 2004 25 1689 1694 15073047 10.1093/carcin/bgh166
Roh JW Kim JW Park NH Song YS Park IA Park SY Kang SB Lee HP p53 and p21 genetic polymorphisms and susceptibility to endometrial cancer Gynecol Oncol 2004 93 499 505 15099969 10.1016/j.ygyno.2004.02.005
Tsai FJ Lin HJ Chen WC Tsai CH Tsai SW A codon 31ser-arg polymorphism of the WAF-1/CIP-1/p21/tumour suppressor gene in Chinese primary open-angle glaucoma Acta Ophthalmol Scand 2004 82 76 80 14738489 10.1111/j.1395-3907.2004.0180.x
Wu MT Wu DC Hsu HK Kao EL Yang CH Lee JM Association between p21 codon 31 polymorphism and esophageal cancer risk in a Taiwanese population Cancer Lett 2003 201 175 180 14607331 10.1016/S0304-3835(03)00469-5
Chedid M Michieli P Lengel C Huppi K Givol D A single nucleotide substitution at codon 31 (Ser/Arg) defines a polymorphism in a highly conserved region of the p53-inducible gene WAF1/CIP1 Oncogene 1994 9 3021 3024 8084608
Roninson IB Oncogenic functions of tumour suppressor p21(Waf1/Cip1/Sdi1): association with cell senescence and tumour-promoting activities of stromal fibroblasts Cancer Lett 2002 179 1 14 11880176 10.1016/S0304-3835(01)00847-3
Coqueret O New roles for p21 and p27 cell-cycle inhibitors: a function for each cell compartment? Trends Cell Biol 2003 13 65 70 12559756 10.1016/S0962-8924(02)00043-0
Menzel HJ Sarmanova J Soucek P Berberich R Grunewald K Haun M Kraft HG Association of NQO1 polymorphism with spontaneous breast cancer in two independent populations Br J Cancer 2004 90 1989 1994 15138483 10.1038/sj.bjc.6601779
Wenham RM Schildkraut JM McLean K Calingaert B Bentley RC Marks J Berchuck A Polymorphisms in BRCA1 and BRCA2 and risk of epithelial ovarian cancer Clin Cancer Res 2003 9 4396 4403 14555511
Fay JC Wu CI Sequence divergence, functional constraint, and selection in protein evolution Annu Rev Genomics Hum Genet 2003 4 213 235 14527302 10.1146/annurev.genom.4.020303.162528
|
16111488
|
PMC1208866
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Aug 19; 5:107
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-107
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-1091611782810.1186/1471-2407-5-109Case ReportHodgkin's disease as unusual presentation of post-transplant lymphoproliferative disorder after autologous hematopoietic cell transplantation for malignant glioma Zambelli Alberto [email protected] Daniele [email protected] Fausto [email protected] Mario [email protected] Laura [email protected] Prada Gian Antonio [email protected] U.O. di Oncologia Medica I, IRCCS, Fondazione Salvatore Maugeri, Pavia, Italy2 Servizio di Anatomia Patologica, IRCCS, Fondazione Salvatore Maugeri, Pavia, Italy3 Servizio di Virologia, IRCCS, Policlinico San Matteo, Pavia, Italy2005 23 8 2005 5 109 109 1 2 2005 23 8 2005 Copyright © 2005 Zambelli 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
Post-transplant lymphoproliferative disorder (PTLD) is a complication of solid organ and allogeneic hematopoietic stem cell transplantation (HSCT); following autologous HSCT only rare cases of PTLD have been reported.
Here, a case of Hodgkin's disease (HD), as unusual presentation of PTLD after autologous HSCT for malignant glioma is described.
Case presentation
60-years old man affected by cerebral anaplastic astrocytoma underwent subtotal neurosurgical excision and subsequent high-dose chemotherapy followed by autologous HSCT. During the post HSCT course, cranial irradiation and corticosteroids were administered as completion of therapeutic program.
At day +105 after HSCT, the patient developed HD, nodular sclerosis type, with polymorphic HD-like skin infiltration.
Conclusion
The clinical and pathological findings were consistent with the diagnosis of PTLD.
==== Body
Background
Post-transplant lymphoproliferative disorders (PTLDs) are severe complications that occur in solid organ and allogeneic hematopoietic cell transplantation (HSCT), and their incidence ranges from 1% to 20% [1].
Most PTLDs are classified as B-cell (rarely T-cell) PTLD, often related to Epstein-Barr virus (EBV) infection. Sometimes, after allogeneic HSCT, non Hodgkin lymphoma (NHL) or Hodgkin's disease (HD) have been described as PTLDs [2].
Herein, we report the first case of HD, as unusual presentation of PTLD, in a patient undergoing high dose sequential chemotherapy and autologous HSCT for malignant glioma.
Case presentation
A 60-year-old male was diagnosed with anaplastic astrocytoma involving the left parietal and temporal lobes. The patient underwent surgical resection with subsequent magnetic resonance imaging (MRI)-proven complete excision.
In the context of a clinical trial, exploring the role of high-dose chemo-radiotherapy in malignant glioma, the patient underwent high dose sequential chemotherapy consisting of cyclophosphamide 7 g/m2 plus rhG-CSF (5 μg/kbw) and hematopoietic cells harvest, followed by methotrexate 8 g/m2 and thiotepa 900 mg/m2. Autologous unfractionated hematopoietic stem cells were administered at doses of 7.5 × 106/kg CD34+ cells and the patient experienced an uncomplicated hematological recovery within 10 days.
One month later, the patient received 70-Gy of fractioned whole-brain irradiation, and steroids (dexamethasone 8 mg i.v. bid) were administered for 2 months and than tapered at the end of radiotherapy.
On day +105 he was hospitalized because of the occurrence of right hilum lymphoadenopathy associated with cutaneous lesions, as limited erythematous plaques, with well-demarcated margins and scalings.
Surgical nodes and skin excision showed the presence of HD, nodular sclerosis type, with atypical HD-like elements, infiltrating the skin. The staging evaluation, performed by CT scan of abdomen and thorax and bone marrow biopsy, was negative for distant involvement (stage IA) and no treatment was administered after radical surgery.
Eight months after HSCT, the malignant glioma relapsed and the patient received 2 cycles of carmustine (240 mg/m2 i.v. q.6 weeks) without clinical benefit, followed by palliative neurosurgical subtotal excision. The steroids treatment was resumed (dexamethasone 4 mg, daily) and chronically administered as supportive care.
Six months later, histological-confirmed HD relapse occurred as bilateral hilum lymphoadenopaty. Because of deteriorated neurological clinical conditions, the patient declined any further treatment and died 2 months later for progressive glioma.
Lymph node and skin biopsies were fixed in 10% buffered formalin and paraffin embedded. Micron sections were then prepared and stained with hematoxylin and eosin. Paraffin section immunohistochemical stains were performed using a biotin-streptavidin technique. The lymph node structure and the skin biopsy were diffusely effaced by a growth consisting of polymorphic features of small T-cell-rich B cell lymphoma/HD-like, and typical Hodgkin's and Reed-Sternberg cells (HRSCs), nodular sclerosis type with limited areas of lymphocyte depletion (REAL) (Fig. 1 and 2). At immunohistochemistry, HRSCs expressed the following molecules: CD30+, CD43+, CD15+/-, CD3+/-; and were negative for EMA, ALK and CD20.
EBV latent membrane protein (LMP-1) was weakly positive at HRSCs immunohistochemistry, but EBV-DNA determination by PCR [3] performed in multiple node biopsy fragments and in peripheral blood mononuclear cells (PBMC) yielded negative results.
According to the clinical, histological and immunohistochemical findings, HD consistent with non EBV-associated PTLD was diagnosed.
Conclusion
Usually reported after allogeneic transplantation, PTLD is a rare event after autologous HCT (Table 1). In most of cases, the onset is within the first in 3 months after HCT (vs 6 months in allogeneic setting), during the time of immunosuppession.
Although HD is not the commonest form of PTLD manifestation, several reports classified HD as a possible polymorphic aspect of the post-transplant malignancies, and HD-like lesions are included among the distinctive categories of PTLD [4].
The early time of onset after HCT and the presence of risk factors for immunological impairment are considered critical for distinguishing PTLD from classic form of HD occurring later in the post-transplant period [5].
EBV appears to be the major contributor to the pathogenesis of PTLD. Uncertain is the promoting role of EBV in EBV-negative PTLDs, while a possible involvement of EBV infection in HD or T-cell PTLD appears controversial. Beside the speculative interest on the pathogenetic mechanisms of the different presentations of PTLD, EBV-negative forms of the disease pose important questions on diagnosis and treatment of PTLD.
In this case, the patient presented a post-transplant EBV-negative HD; the treatment administered for malignat glioma, consisting of immuno-suppressive procedures (HDS with autologous HCT, radiotherapy and steroids), played a role in immunologic derailment and exposed the patient to an high risk of PTLD.
In our opinion, the rapid onset of HD after HSCT (+105 day), the clinical and the histopathological characteristics are consistent with a rare presentation of HD as EBV-negative PTLD. Surveillance for PTLD should be considered also in patients affected by solid malignancies receiving autologous HSCT.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AZ and DL equally contributed to the elaboration of the manuscript, FB performed the virological analyses, GADP performed the revision of the manuscript as responsible of the transplantation program, MS, LV performed the pathological analyses
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was partially supported by Ministero della Salute, Ricerca Finalizzata 2002 convenzione no. 08923502 and Ricerca Corrente 2002, grant no.80541
Figures and Tables
Figure 1 H&E stained sections of lymph node (fig.1) and skin (fig.2) excision in which the diagnostic Reed-Sternberg cell support the diagnosis of Hodgkin disease.
Figure 2 H&E stained sections of lymph node (fig.1) and skin (fig.2) excision in which the diagnostic Reed-Sternberg cell support the diagnosis of Hodgkin disease.
Table 1 PTLDs after autologous HCT in adults
Source (N° of cases) Disease HCT PTLD (days) PTLD (type)
Young 1989[6] (2) T cell LL T cell purged 48/31 NHL polymorphic B cell
Chao 1993[7] (1) FML B cell purged 52 NHL monomorphic B cell
Shephered 1995[8] (1) CML Unmanipulated 43 NHL polymorphic B cell
Briz 1997[9] (1) T cell ALL T cell purged 81 NHL monomorphic B cell
Hauke 1998[10] (2) HD/FLL Unmanipulated 87/38 NHL polymorphic B cell
Peniket 1998[11] (1) MM CD34+ selected 80 NHL monomorphic B cell
Lohrish 2000 [12] (1) BL Unmanipulated 39 NHL polymorphic B cell
Yufu 2000 [13] (1) HD Unmanipulated 900 T cell PTLD
Jenkins 2002 [14] (3) 2MM/1NHL CD34+ selected/Unmanipulated 31/NA NHL polymorphic B cell
ALL: acute lymphoblastic leukemia; CML: chronic myelogenous leukemia; FML: follicular mixed small cleaved and large lymphoma; FLL: follicular large cell lymphoma; LCL: large cell lymphoma; LL: lymphoblastic lymphoma; BL: Burkitt's lymphoma; HD: hodgkin disease; NHL: non hodgkin lymphoma; MM: multiple myeloma PTLD: post transplant lymphoproliferative disorder; HCT: hematopoietic cell transplantation; NA: not available.
==== Refs
Swinnen LJ Overview of posttransplant B-cell lymphoproliferative disorders Semin Oncol 1999 26 21 25 10561014
Deeg J Socie G Malignancies after hematopoietic stem cell transplantation: many question, some answers Blood 1998 91 1833 44 9490664
Baldanti F Grossi P Furione M Simoncini L Sarasini A Comoli P Maccario R Fiocchi R Gerna G High levels of Epstein-Barr virus DNA in blood of solid-organ transplant recipients and their value in predicting posttransplant lymphoproliferative disorders J Clin Microbiol 2000 38 613 9 10655355
Harris NL Ferry JA Swerdlow SH Posttransplant lymphoproliferative disorders: summury of Society for Hematopathology Workshop Semin Diagn Pathol 1997 14 8 14 9044505
Rowlings PA Curtis RE Passweg JR Deeg HJ Socie G Travis LB Kingma DW Jaffe ES Sobocinski KA Horowitz MM Increased incidence of Hodgkin's disease after allogeneic bone marrow transplantation J Clin Oncol 1999 17 3122 7 10506608
Young L Alfieri C Hennessy K Evans H O'Hara C Anderson KC Ritz J Shapiro RS Rickinson A Kieff E Expression of Epstein-Barr virus transformation-associated genes in tissues of patients with EBV lymphoproliferative disease N Engl J Med 1989 321 1080 5 2552313
Chao NJ Berry GJ Advani R Horning SJ Weiss LM Blume KG Epstein-Barr virus-associated lymphoproliferative disorder following autologous bone marrow transplantation for non-Hodgkin's lymphoma Transplantation 1993 55 1425 8 8390736
Shepherd JD Gascoyne RD Barnett MJ Coghlan JD Phillips GL Polyclonal Epstein-Barr virus-associated lymphoproliferative disorder following autografting for chronic myeloid leukemia Bone Marrow Transplant 1995 15 639 41 7655394
Briz M Fores R Regidor C Busto MJ Ramon y Cajal S Cabrera R Diez JL Sanjuan I Fernandez MN Epstein-Barr virus associated B-cell lymphoma after autologous bone marrow transplantation for T-cell acute lymphoblastic leukaemia Br J Haematol 1997 98 485 7 9266955 10.1046/j.1365-2141.1997.2153034.x
Hauke RJ Greiner TC Smir BN Vose JM Tarantolo SR Bashir RM Bierman PJ Epstein-Barr virus-associated lymphoproliferative disorder after autologous bone marrow transplantation: report of two cases Bone Marrow Transplant 1998 21 1271 4 9674863 10.1038/sj.bmt.1701258
Lohrisch CA Nevill TJ Barnett MJ Hogge DE Connors JM Keown PA Gascoyne RD Development of a biologically distinct EBV-related lymphoproliferative disorder following autologous bone marrow transplantation for an EBV-negative post-renal allograft Burkitt's lymphoma Leuk Lymphoma 2000 39 195 201 10975399
Peniket AJ Perry AR Williams CD MacMillan A Watts MJ Isaacson PG Goldstone AH Linch DC A case of EBV-associated lymphoproliferative disease following high-dose therapy and CD34-purified autologous peripheral blood progenitor cell transplantation Bone Marrow Transplant 1998 22 307 9 9720750 10.1038/sj.bmt.1701335
Yufu Y Kimura M Kawano R Noguchi Y Takatsuki H Uike N Ohshima K EBV associated T cell lymphoproliferative disorder following autologous blood stem cell transplantation for relapsed Hodgkin's disease Bone Marrow Transplant 2000 26 1339 41 11223975 10.1038/sj.bmt.1702721
Jenkins D Difrancesco L Chaudhry A Morris D Gluks S Jones A Woodman R Brown CB Russel J Stewart DA Successful traetment of post-transplantation lymphoproliferative disorder in autologous blood stem cell transplant recipients Bone Marrow Transplant 2002 30 321 6 12209355 10.1038/sj.bmt.1703603
|
16117828
|
PMC1208867
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Aug 23; 5:109
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-109
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-1121614304510.1186/1471-2407-5-112Research ArticleFamily history of colorectal cancer in Iran Mahdavinia Mahboobeh [email protected] Faraz [email protected] Reza [email protected] Nasim [email protected] Ahmad [email protected] Mahshid [email protected] Naser [email protected] Reza [email protected] Digestive Disease Research Centre (DDRC), Shariati Hospital, Tehran University of Medical sciences, Tehran, Iran2 General surgery ward, Mehr Hospital, Tehran, Iran3 Pathology Department, Mehr Hospital, Tehran, Iran2005 5 9 2005 5 112 112 2 5 2005 5 9 2005 Copyright © 2005 Mahdavinia 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
Previous reports show a high proportion of young CRC patients in Iran. In this study we aim to look for the clustering of colorectal cancer in families of a series of CRC patients from Iran.
Methods
The family history of cancer is traced in 449 CRC patients of which 112 were 45 yrs or younger and 337 were older than 45 yrs at time of diagnosis. The patients were admitted in two hospitals in Tehran, during a 4-year period.
Results
Clinical diagnosis of HNPCC was established in 21 (4.7%) probands. Family history of CRC was more frequently reported by early-onset than by late-onset patients (29.5% vs. 12.8%, p < 0.001).
Distribution of tumor site differed significantly between those with and without family history of CRC. Right colon cancer was the most frequent site (23/45, 35.4%) observed in patients with positive family history of colorectal cancer.
Conclusion
The relatively high frequency of CRC clustering along with HNPCC in our patients should be further confirmed with larger sample size population-based and genetic studies to establish a cost effective molecular screening for the future.
==== Body
Background
Colorectal carcinoma (CRC) is the third most common cause of cancer related deaths in the world [1] with some well known hereditary forms. Hereditary non-polyposis colorectal cancer (HNPCC) is the most common type of hereditary CRC. The frequency of HNPCC varies worldwide ranging from 1 to 6% [2]. This variability could be due to discrepancies in definitions (clinical vs. molecular criteria) and also regional heterogeneity [3-5].
Apart from defined genetic syndromes of CRC, both retrospective and prospective studies proved that familial history of colorectal cancer could increase a person's lifetime risk of colorectal cancer significantly [6,7].
CRC with age-adjusted rate of 6–7.9 per 100,000 person per year is the fourth most common cancer in Iran [8,9]. Early-onset colorectal cancer (less than 40 yrs of age at time of diagnosis) comprises almost one fifth of all CRC cases in the country [9,10]. This proportion is considerably lower in high-risk countries, with rates ranging from 2% to 8% [11,12]. The high proportions of young CRC cases seen in Iran, and probably many neighboring countries, are mainly due to the young age-structure of these countries and relatively low rates of CRC in older individuals [10].
Up to now there have been no reports addressing the familial aggregation of colorectal cancer in Iran. In the current study, we aimed to study the clustering of colorectal cancer in families of CRC patients and make a comparison of frequency of colorectal and other types of cancer in family members of early-onset with that of late-onset CRC patients.
Methods
We reviewed all patients' documents with a pathologically confirmed diagnosis of colorectal adenocarcinoma admitted in two hospitals in Tehran between January 2000 and February 2004. Patients with familial adenomatous polyposis or those with underlying inflammatory bowel disease were excluded. Of 662 patients registered in the database, 174 patients were lost due to change of address or contact phone number, 33 refused to be interviewed and 6 were excluded because of incomplete family history accordingly, 449 CRC cases (112 younger than 45 yrs and 337 older than 45 yrs at diagnosis) or their close family members participated in this study.
Demographic, clinical and tumor-related characteristics of patients were recorded based on their hospital documents. These parameters included gender, age at diagnosis, place and date of birth and tumor-related factors such as location, stage, degree of differentiation and mucus production.
All of these patients or their siblings/parents (in the case the patient was dead or not available) were interviewed to trace their family history of cancer including occurrence of malignancy in the family, type of cancer and the age at diagnosis of the affected family member. Pedigrees were drawn at least up to second-degree relatives. The obtained pedigrees were reconfirmed by interviewing another member of family. In addition, we tried to verify reported malignancies in relatives by asking for their medical records, if available. In Iran, family relations are very strong and people are usually aware of serious diseases such as cancer in their relatives.
Patients belonging to families fulfilling the AmsterdamII criteria, including at least three members with an HNPCC-associated cancer (colorectal, endometrial, small bowel, ureter, renal pelvis) in at least two successive generations, one being a first-degree relative of the other two and at least one diagnosed before the age of 50 years, were classified as HNPCC [13].
Patients from families which did not fulfill Amsterdam criteria but with at least two relatives with colorectal cancer in a first- or second-degree relationship were classified as Hereditary Colorectal Cancer (HCRC).
This study was approved by the Digestive Disease Research Center of Tehran University of Medical Sciences Institutional Review Board and informed consents were obtained from patients or their families participating in the study.
Statistical analysis
Qualitative variables were compared by chi-squared test with Yates correction when needed. A P value of less than 0.05 was considered to indicate a statistically significant difference. All calculations were performed using the 11.5 SPSS software package (SPSS Inc., Chicago, IL, USA).
Results
Four-hundred forty nine (449) CRC patients with pathologically confirmed colorectal adenocarcinoma were enrolled in the study.
Tumor sites of early-onset and late-onset group are separately shown in table 1. No significant difference in localization of tumor was observed between the two groups.
Family history of cancer up to second-degree relatives were observed in 60 (53.5%) of early-onset and 144 (43.5%) of late-onset patients respectively. The most common cancers affecting first and second degree relatives are shown in table 2. History of colorectal cancer in at least one relative was significantly more frequent in early-onset patients; 33 (29.5%) comparing to 43 (12.8%) cases in late-onset group. Frequency of other types of cancers reported in the family did not differ significantly between the two age groups.
The frequency of hereditary types of colorectal cancer in two age groups is shown in Table 3. Among 21 HNPCC families, nineteen fulfilled the Amsterdam criteria I having at least three members with CRC in two consecutive generations, one being a first-degree relative of the other two and at least one diagnosed before the age of 50 years [14]. The other 2 families fulfilled the Amsterdam criteria II with history of colorectal and endometrial cancer in the family.
Distribution of tumor site differed significantly between patients with family history of CRC and those without this history (Table 4,†); In total, right colon cancer was seen more frequently than other locations in probands with family history of CRC, in contrast to higher percentage of left colon carcinoma in patients without this history.
Rectal carcinoma was rarely seen in young patients with positive family history; 12.1% comparing to 38% in those without family history (p = 0.025) or cases belonging to the other age group.
Discussion
This report represents the first study to characterize the profile of familial CRC aggregation in Iranian patients. Although we tried to include all recorded patients in this study, familial profile of 213 (32%) cases could not be determined and analyzed here (see method). However, there were no significant differences in age, sex and tumor localization of participating and non participating patients.
The estimation of the frequency of HNPCC among CRC patients based on family history varies in different populations. In a multicenter study on Finnish population, the frequency of families meeting the Amsterdam criteria was 1.7% [15]. Ponz et al estimated the frequency of HNPCC among CRC cases to be 3.4–4.5% in Northern Italy [3], while in Spanish population; clinical diagnosis of HNPCC was established in 2.5% of CRC patients [16]. In a pilot study by Soliman a et al, 7.2% of Egyptian colorectal patients had family history suggestive of hereditary non polyposis colorectal cancer [17]. Discrepancies in the reported frequency of HNPCC between populations probably reflect population differences. Frequency of this syndrome in Iran has not yet been studied. In our study, a total of 21 cases (4.7%) met the Amsterdam criteria II. This relatively high frequency of HNPCC in our patients should be further confirmed with larger sample size, population-based, and genetic studies.
HNPCC clinical diagnosis was significantly more prevalent among younger patients in agreement with previous studies [18,19]. In some other studies, the mean age at diagnosis for HNPCC patients was reported to be between 52–60 yrs [20,21]. However, we found that almost 3% of patients over 45 yrs belonged to families fulfilling Amsterdam criteria. This indicates that taking detailed family history even in old cases is necessary and could identify CRC families with known genetic risk.
Also, we have found higher frequency of familial clustering of CRC apart from HNPCC cases in younger patients (Table 3).This in agreement with studies in other populations, suggests stronger genetic background in younger onset CRC patients [22].
We observed 15 families who did not fulfill HNPCC criteria, but with more than 2 CRC relatives in the family (Table 3). It was previously suggested that some familial cancers might be due to common environmental exposures of family members rather than genetic clustering of cancer [23], but more recently; it has been shown that familial risk for CRC are mainly due to heritable causes [24]. Therefore, these families could be added to HNPCC families to undergo analysis of MMR genes [25]. However, not all CRC clustering could be attributed to defects in genes involved in mismatch repair [26], and unknown loci may be responsible for much of the familial aggregation of CRC [27].
It is suggested that family history of CRC is related to the localization of tumor. Some studies proposed a stronger familial component for proximal than for distal colon cancer [6,28], while this association was not observed in other studies [29]. In our study, right-sided tumors occurred more frequently in patients with positive family history of CRC compared to those without this history (36.9% vs. 17.7%, p < 0.001). This pattern was almost the same in patients over and under 45 yrs of age (Table 4). Also, we have found rectal cancer to be less frequent among younger probands with positive CRC family history. This is in agrremnet with the previous report of Olsson L and Lindblom A, in which the frequency of sigmoid cancer was shown to be lower among familial CRC cases compared to sporadic patients [21]. Altogether, these findings could indicate a difference in carcinogenic pathways based on the tumor location with heritable causes mostly affecting the right colon and exogenous factors responsible for carcinogenesis in the distal part of the large bowel. It has been suggested that chromosomal instability affecting allelic losses of APC and P53 are more frequent in distal colonic and rectal tumors than in proximal lesions. On the other hand, microsatellite instability is more common in proximal tumors [30].
Appropriate screening strategies should be considered to decrease the burden of CRC in Iran.Our ongoing molecular and genetic studies including analysis of micro satellite instability and mismatch repair gene mutations would most likely help us to characterize the molecular basis for the observed clustering of colorectal cancer among families of younger onset patients and could lead to the adoption of simpler and more cost effective molecular screening in the future.
Conclusion
Clinical diagnosis of HNPCC was observed in 21 (4.7%) probands in our study. Familial clustering of CRC was more frequent in younger probands. The right side of the colon was more frequently affected in patients with positive family history of CRC. This indicates that a detailed family history is mandatory in colorectal cancer patients with particular attention to younger onset cases and those with right-sided tumors.
Authors' contributions
MM and FB designed and conducted the study, interviewed the patients, analyzed the data, and drafted the manuscript.
RA, NN and AK assisted in conducting the study and interviewing the patients.
MH and NR reviewed and approved the pathology reports of the patients.
RM supervised the study scientifically, has been involved in designing the study and preparing the manuscript and revising it for scientific content and has given final approval of the version to be published.
All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We are thankful to Mrs. Mohseni-far for contacting patients, Miss Noroozi in Mehr Hospital for her kind help in reviewing records and Dr. Annette Ratcliff for helping in language correction.
Figures and Tables
Table 1 CRC characteristics among young and old patients.
Variable Early-onset (n = 112) Late-onset (n = 337) P Value†
Sex (M/F) 43/69 138/199 NS††
Tumor Location
Right 30(26.8%) 62(18.4%) NS
Left 43(38.4%) 134(39.8%) NS
Rectum 36(32.1%) 141(41.8%) NS
NOS 3(2.6%) 0 ___
†Assigns the significance of the differences between two group
†† NS: Not significant
Table 2 Frequent cancer sites in relatives of early and late-onset CRC patients.
Early-onset (n = 112) Late-onset (n = 337) P Value†
Family History (FHx) of cancer 60(53.6%) 144(42.7%) NS††
FHx of CRC 33(29.4%) 43(12.8%) <0.001
Breast 11(9.8%) 17(5%) NS
Stomach 9(8%) 17(5%) NS
Lung 5(4.5%) 25(7.5%) NS
Leukemia & Lymphoma 4(3.6%) 14(4.2%) NS
Brain 4(3.6%) 9(2.7%) NS
Endometrial 2(1.8%) 12(3.6%) NS
†Assigns the significance of the differences between two group
†† NS: Not significant
Table 3 Hereditary forms of CRC in two age groups.
Variable Total
(n = 449) Early-onset
(n = 112) Late-onset
(n = 337) P Value†
HNPCC 21(4.7%) 12(10.7%) 9(2.9%) <0.001
HCRC 15(3.4%) 9(8%) 6(1.7%) <0.001
One first-degree relative with CRC 26(5.8%) 7(6%) 19(5.6%) NS††
†Assigns the significance of the differences between two group
†† NS: Not significant
Table 4 distribution of tumor sites by family history of colorectal cancer according to age of the proband.
Age group Positive FH† of CRC Negative FH of CRC
CRC localization
Total† N = 76 N = 373
Righta 28(36.9%) 66(17.7%)
Left 24(31.6%) 167(44.8%)
Rectum 22(29%) 139(37.3%)
NOS 2(2.5%) 1(0.03%)
Over 45 N = 43 N = 294
Rightb 14(32.6%) 49(16.7%)
Left 11(25.6%) 136(46.3%)
Rectum 18(41.8%) 109(37.1%)
Less than 45 N = 33 N = 79
Rightc 14(42.4%) 17(21.5%)
Left 13(39.4%) 31(39.2%)
Rectum 4(12.1%) 30(38%)
NOS 2(6.1%) 1(1.3%)
†: difference of tumor site distribution between two groups (p < 0.001)
a,b,c: p < 0.05
==== Refs
Parkin DM Global cancer statistics in the year 2000 Lancet Oncol 2001 2 533 43 11905707 10.1016/S1470-2045(01)00486-7
Lynch HT de la Chapelle A Hereditary colorectal cancer N Engl J Med 2003 348 919 32 12621137
Ponz de Leon M Sassatelli R Benatti P Roncucci L Identification of hereditary nonpolyposis colorectal cancer in the general population: the 6-year experience of a population-based registry Cancer 1993 71 3493 3501 8387880
Rustgi AK Hereditary gastrointestinal polyposis and nonpolyposis syndromes N Engl J Med 1994 331 1694 1702 7969362 10.1056/NEJM199412223312507
Aaltonen LA Salovaara R Kristo P Canzian F Hemminki A Peltomaki P Chadwick RB Kaariainen H Eskelinen M Jarvinen H Mecklin JP de la Chapelle A Incidence of hereditary nonpolyposis colorectal cancer and the feasibility of molecular screening for the disease N Engl J Med 1998 338 1481 1487 9593786 10.1056/NEJM199805213382101
Fuchs CS Giovannucci EL Colditz GA Hunter DJ Speizer FE Willett WC A prospective study of family history and the risk of colorectal cancer N Engl J Med 1994 331 1669 74 7969357
St John DJ McDermott FT Hopper JL Debney EA Johnson WR Hughes ES Cancer risk in relatives of patients with common colorectal cancer Ann Intern Med 1993 118 785 790 8470852
Sadjadi A Malekzadeh R Derakhshan MH Sepehr A Nouraie M Sotoudeh M Yazdanbod A Shokoohi B Mashayekhi A Arshi S Majidpour A Babaei M Mosavi A Mohagheghi MA Alimohammadian M Cancer occurrence in Ardabil: results of a population-based cancer registry from Iran Int J Cancer 2003 107 113 8 12925965
Hosseini SV Izadpanah A Yarmohammadi H Epidemiological changes in colorectal cancer in Shiraz, Iran: 1980–2000 ANZ J Surg 2004 74 547 9 15230787 10.1111/j.1445-2197.2004.03064.x
Ansari R Mahdavinia M Sadjadi A Nouraie M Kamangar F Bishehsari F Fakheri H Semnani S Arshi S Zahedi MJ Darvish-Moghadam S Mansour-Ghanei F Mosavi A Malekzadeh R Incidence and age distribution of colorectal cancer in Iran, results of a population-based cancer registry Can Lett
Griffin PM Liff JM Greenberg RS Clark WS Adenocarcinomas of the colon and rectum in persons under 40 years old. A population-based study Gastroenterology 1991 100 1033 40 2001800
Mitry E Benhamiche AM Jouve JL Clinard F Finn-Faivre C Faivre J Colorectal adenocarcinoma in patients under 45 years of age: comparison with older patients in a well-defined French population Dis Colon Rectum 2001 44 380 7 11289284
Vasen HF Watson P Mecklin JP Lynch HT New clinical criteria for hereditary nonpolyposis colorectal cancer (HNPCC, Lynch syndrome) proposed by the International Collaborative group on HNPCC Gastroenterology 1999 116 1453 6 10348829
Vasen HF Mecklin JP Khan PM Lynch HT The International Collaborative Group on Hereditary Non-Polyposis Colorectal Cancer (ICG-HNPCC) Dis Colon Rectum 1991 34 424 5 2022152 10.1007/BF02053699
Mecklin JP Jarvinen HJ Hakkiluoto A Hallikas H Hiltunen KM Harkonen N Kellokumpu I Laitinen S Ovaska J Tulikoura J Frequency of hereditary nonpolyposis colorectal cancer. A prospective multicenter study in Finland Dis Colon Rectum 1995 38 588 93 7774468 10.1007/BF02054116
Pinol V Andreu M Castells A Paya A Bessa X Rodrigo J Gastrointestinal Oncology Group of the Spanish Gastroenterological Association Frequency of hereditary non-polyposis colorectal cancer and other colorectal cancer familial forms in Spain: a multicentre, prospective, nationwide study Eur J Gastroenterol Hepatol 2004 16 39 45 15095851
Soliman AS Bondy ML Levin B El-Badawy S Khaled H Hablas A Ismail S Adly M Mahgoub KG McPherson RS Beasley RP Familial aggregation of colorectal cancer in Egypt Int J Cancer 1998 77 811 816 9714045
Mecklin JP Järvinen HJ Clinical features of colorectal carcinoma in cancer family syndrome Dis Colon Rectum 1986 29 160 4 3943429
Lynch HT Smyrk TC Watson P Lanspa SJ Lynch JF Lynch PM Cavalieri RJ Boland CR Genetics, natural history, tumor spectrum, and pathology of hereditary nonpolyposis colorectal cancer: an updated review Gastroenterology 1993 104 1535 49 8482467
Peel DJ Ziogas A Fox EA Gildea M Laham B Clements E Kolodner RD Anton-Culver H Characterization of hereditary nonpolyposis colorectal cancer families from a population-based series of cases J Natl Cancer Inst 2000 92 1517 22 10995807 10.1093/jnci/92.18.1517
Olsson L Lindblom A Family history of colorectal cancer in a Sweden county Fam Cancer 2003 2 87 93 14574157 10.1023/A:1025734200635
Paraf F Jothy S Colorectal cancer before the age of 40: a case-control study Dis Colon Rectum 2000 43 1222 6 11005487
Khoury MJ Beaty TH Liang K-Y Can familial aggregation of disease be explained by familial aggregation of environmental risk factors? Am J Epidemiol 1988 127 674 683 3341366
Hemminki K Chen B Familial risk for colorectal cancers are mainly due to heritable causes Cancer Epidemiol Biomarkers Prev 2004 13 1253 6 15247139
Vasen HFA What is hereditary nonpolyposis colorectal cancer (HNPCC) Anticancer Res 1994 14 1613 1616 7979195
Liu B Parsons R Papadopoulos N Nicolaides NC Lynch HT Watson P Jass JR Dunlop M Wyllie A Peltomaki P de la Chapelle A Hamilton SR Vogelstein B Kinzler KW Analysis of mismatch repair genes in hereditary non-polyposis colorectal cancer patients Nat Med 1996 2 169 74 8574961 10.1038/nm0296-169
Lewis CM Neuhausen SL Daley D Black FJ Swensen J Burt RW Cannon-Albright LA Skolnick MH Genetic heterogeneity and unmapped genes for colorectal cancer Cancer Res 1996 56 1382 8 8640829
Andrieu N Launoy G Guillois R Ory-Paoletti C Gignoux M Estimation of the familial relative risk of cancer by site from a French population based family study on colorectal cancer (CCREF study) Gut 2004 53 1322 8 15306593 10.1136/gut.2003.036376
Kee F Collins BJ Families at risk of colorectal cancer: who are they? Gut 1992 33 787 90 1624161
Lynch HT de la Chapelle A Genetic susceptibility to non-polyposis colorectal cancer J Med Genet 1999 36 801 18 Review 10544223
|
16143045
|
PMC1208868
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Sep 5; 5:112
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-112
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-981608684010.1186/1471-2407-5-98Research ArticleExpression of transforming growth factor-beta-1 and p27Kip1 in pancreatic adenocarcinomas: relation with cell-cycle-associated proteins and clinicopathologic characteristics Culhaci Nil [email protected] Ozgul [email protected] Sedat [email protected] Huseyin [email protected] Ibrahim [email protected] Mujde [email protected] Ilhan [email protected] Funda [email protected] Department Of Pathology, Adnan Menderes University, Faculty of Medicine, Aydin, Turkey2 Hepatobiliary Study Group, Dokuz Eylül University, Faculty of Medicine, Izmir, Turkey2005 8 8 2005 5 98 98 30 11 2004 8 8 2005 Copyright © 2005 Culhaci 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 purpose of our study was to investigate the immunohistochemical expression of TGF-β1 and p27 in pancreatic adenocarcinomas and to compare the findings with the clinicopathological features and survival. We also aimed to evaluate the expression of TGF-β1 and p27 in the context of other cell cycle and proliferation markers such as cyclin D1 and Ki-67.
Methods
We examined TGF-β1 and p27 expression immunohistochemically in 63 cases of invasive ductal adenocarcinoma of the pancreas. Standard streptavidin-biotin immunperoxidase method was used for immunostaining and the stained slides were examined microscopically using semiquantitative criteria.
Results
TGF-β1 stained the cytoplasms of the tumor cells in 43 cases [68.3%]. There was a statistically significant difference among TGF-β1 staining scores in terms of clinicopathologic factors such as blood vessel invasion, stage and distant metastasis [p < 0.05]. Of the 63 tumors evaluated 23 [36.5%] were positive for p27 within the nucleus. An inverse correlation was found between p27 immunoreactivity and grade [p < 0.05]. But no significant correlation was found between p27 and other parameters. Among the patients with survival data 27 patients had RO resections and these cases were considered in survival analysis. In the univariate analysis, neither TGF-β1 nor p27 expression was related with patient survival.
Conclusion
Our findings suggest that in pancreatic carcinoma, TGF-β1 expression is related to tumor growth and metastasis. But it is not associated with cell cycle proteins. p27 expression is reduced in pancreatic adenocarcinomas and decreased protein levels of p27 may play a role in the differentiation of pancreatic cancer.
==== Body
Background
Pancreatic cancer is a malignant tumor with an extremely poor prognosis. This tumor is highly aggressive and patients with this form of cancer have a short survival after diagnosis. Even when the tumor is localized, the mean survival time after radical resection varies from 10 to 20 months [1]. The mechanisms of the aggressive growth and metastasis are not yet extensively understood.
Several studies indicated that proliferative activity of tumor cells, as well as tumor angiogenesis, inactivation of tumor supressor genes, overexpression of growth factors may play role in pancreatic carcinogenesis and may help to predict patient outcome [2-8]. Recent studies denoted that alterations in growth factors and growth factor receptors seem to influence the biologic behaviour of pancreatic cancer cells [2]. Growth factors are involved in carcinogenesis, where they influence a variety of functions including cell proliferation, invasion, metastasis, angiogenesis, local immune system functions, and extracellular matrix formation [2]. Growth factors do not only stimulate cell proliferation, but they may also act as growth inhibitors, depending on the cell type and the stimulatory pathway that is involved. Transforming growth factor-β [TGF-β] is such an example, being a growth stimulator in fibroblastic cells with TGF-β receptors, but a negative regulator in epithelial cells.
TGF-β belongs to a family of homologous polypeptides that includes three major isoforms [TGF-β1, TGF-β2, TGF-β3]. It has been reported that TGF-β influence different cell functions, including growth, proliferation and differentiation. It can influence cancer growth in various ways, such as by stimulating angiogenesis, suppressing cancer-directed immune functions, increasing the expression of adhesion molecules and extracellular matrix components [9]. Human pancreatic cancer cells may exhibit loss of responsiveness to TGF-β-mediated growth inhibition as a consequence of altered TGF-β expression as well as a result of postreceptor alterations [10]. It has also been demonstrated that TGF-β induced cell cycle arrest can be partially attributed to the regulatory effects of TGF-β on both the expression and activity of cyclin-dependent kinase inhibitors [CDKI] such as p21 and p27. The binding of these inhibitors to spesific cyclin-dependent kinase [CDK] complexes blocks their activity and causes cell cycle arrest [11,12].
Alterations in cell cycle regulatory mechanisms play an important role in the tumor development. Cell cycle progression is regulated by a series of cyclins, CDKs and CDKIs. p27, a member of the Cip/Kip family is a low molecular weight CDKI, which is able to arrest cell cycle progression by complexing CDKs and their activity [13]. Low p27 expression has been reported to be a poor prognostic factor in a variety of human cancers including prostate, lung, squamous cell carcinomas [13-18].
In this study, we investigated the immunohistochemical expressions of TGF-β1 and p27 in pancreatic adenocarcinomas and the results were correlated with the clinicopathologic characteristics of the cases and the patients' survival to find out if these factors could be used as an additional predictor of the disease extent and patient outcome. Additionally, we evaluated the expression of TGF-β1 and p27 in the context of other cell cycle and proliferation markers; cyclin D1 and Ki-67.
Methods
Pancreatic cancer tissues were obtained from 63 patients [36 female, 27 male] undergoing pancreatic surgery for primary pancreatic adenocarcinoma in Dokuz Eylül University Hospital between the years 1996 and 2003. The mean age of the patients was 62 years, with a range of 42–82. The patients were not subjected to any sort of chemotherapy or radiation therapy prior to resection. Surgical procedures consisted of either distal pancreatectomy or Whipple procedure. Some of the patients underwent palliative surgical treatment. The histological diagnosis of each specimen was provided by standard light microscopic evaluation of the routinely processed and parafin embedded tissues. The H&E stained slides of each case were taken from the pathology archieves and reviewed by two pathologists without knowledge of the patients' outcome. The tumors were typed as proposed by WHO and graded as well, moderate or poorly differentiated. International Union Against Cancer [UICC] TNM classification system [sixth edition] was used for staging. Staging was based on the information at the time of diagnosis. Lymphatic vessel, blood vessel and perineural invasions as well as lymph node, resection margin and adjacent organ involvements were reevaluated in each tumor.
Immunohistochemistry
The most representative tumor tissue block was chosen from each case and 5 μm sections were taken to poly-L lysin coated slides for immunuhistochemical staining. Standard streptavidin biotin immunperoxidase method was used for immunostaining with TGF-β1 [Santa Cruz, sc-146, California, USA], p27Kip1 [DAKO, Glostrup, Denmark, M7203, dilution: 1/50], cyclin D1 [NeoMarkers, Fremont CA, Clone DCS-6, MS210-P, dilution: 1/50] and Ki-67 [NeoMarkers, Fremont CA, Clone MB67, MS1006-A, dilution: 1/50] antibodies. This indirect immunoperoxidase staining procedure was performed at room temperature. Appropriate tissue sections known to react with both antibodies as positive controls for each primary antibody were also stained simultaneously.
Analysis of immunohistochemical data
TGF-β1 scoring
The stained slides were examined microscopically using the following parameters and semiquantitative criteria. Cytoplasmic immunostaining in tumor cells were accepted as positive. For evaluation of the immunohistochemical results, the intensity of the immunoreaction and the number of the immunoreactive cancer cells were scored on the tissue slides. The intensity of the immunoreaction was classified into 4 levels: 0: no staining, 1: weak staining, 2: moderate staining, 3: intense staining. In addition, the percentage of the immunoreactive cancer cells was recorded; 0, negative; 1, less than 33% positive staining; 2, 33% to 66% positive staining, 3, more than 66% positive staining. A combined score of 0 to 6 was assigned. Tumors given a score 0 to 1 were classified as negative; those given a score of 2 were classified as weakly positive; a score of 3 to 4 was considered moderately positive; and a score of 5 to 6 was considered strongly positive [19].
p27 scoring
Approximately 500 cells from the selected area were counted, and the number and percentage of cells with nuclear p27 staining were noted. Cells that had either weak or strong nuclear staining were considered to be positive for p27 protein expression. Cells without nuclear staining were considered negative. The percentage of cells expressing p27 was recorded as the ratio of positive cells-to-the total number of cells counted. p27 expression was grouped as high or low. Patients with low p27 expression had less than 30% of the nuclei in the specimen staining positive, while those with high levels of p27 expression had more than 30% of the nuclei staining for the bound antibody [15].
Ki-67 scoring
Fifty-two cases were stained with Ki-67. Nuclear immunostaining in tumor cells were accepted as positive. At least 10 fields containing approximately 500 tumor cells were counted each per case at 400 × magnification. The number of Ki-67-staining reactive cells in each field was determined as a percentage of the total number of tumor cells counted [20].
cyclin D1 scoring
Fifty-two cases were stained with cyclin D1. Nuclear immunostaining in tumor cells were evaluated according to Holland et al [21]. Cyclin D1 expression in tumors was graded as negative: less than 10% of cells stained; 1+: 10–50% of cells stained; 2+: 50–75% of cells stained and 3+: more than 75% of cells stained.
Statistical analysis
Data were analyzed by a computer software SPSS for Windows 10.0. A p value <0.05 was considered statistically significant. Immunohistochemical scores were compared among prognostic parameter subgroups by the chi-square and Fishers exact tests. Correlations between non-parametric and parametric values were tested with Kendall's tau-b and Spearman correlation tests respectively. The effect of prognostic and immunohistochemical scores on the patient survival was tested in potentially curatively operated patients who had R0 resections, with Cox regression analysis.
Results
Clinicopathologic features
Clinicopathologic features of the patients is shown in Table 1. Thirty-six were women, 27 were men. The median age of the patients was 62 years, with a range of 42–82. The size of the resected tumors ranged between 1 and 13 cm [median 4.5 cm]. Sixteen tumors were well differentiated, 33 were moderately differentiated, and 14 were poorly differentiated. Four, 23, 3 and 33 patients had stage I, II, III and IV disease. Thirty-nine [61.9%] patients had one and more than one regional lymph node metastasis and 33 [52.4%] patients had distant metastasis. Among resected tumors, 44 [69.8%] showed perineural invasion, 26 [41.3%] showed lymphatic vessel and 21 [33.3%] showed blood vessel invasion. Survival data were available for 54 patients. At follow-up, 38 patients [60.3%] had died of disease. The overall median survival of these 38 patients was 13 months. Sixteen patients were still alive with no evidence of disease. The remaining patients were either lost to follow-up or their death status was unknown. Among the patients with survival data 27 patients had RO resections and these cases were considered in survival analysis.
Immunohistochemical analysis
TGF-β1 stained the cytoplasm of the tumor cells in 43 cases [68.3%] [Figure 1]. Strongly positive staining was observed in 14 tissues [22.2%] while 19 tumors [30.1%] were moderately and 10 [15.9%] were weakly positive. There was a statistically significant positive correlation among TGF-β1 staining scores in terms of clinicopathologic factors such as blood vessel invasion [p = 0.01], stage [p = 0.05] and distant metastasis [p < 0.01] [Table 2]. But the other prognostic parameters did not relate significantly with TGF-β1 staining scores. Additionally, we could not find an association with cyclin D1 [Table 3] and Ki-67.
p27 immunoreactivity was concentrated within the nucleus, but concomitant faint cytoplasmic staining was also observed [Figure 2]. Of the 67 tumors evaluated, 40 [63.5%] were negative, 17 [27%] were assessed to have low grade and 6 [9.5%] were assessed to have as high grade p27 staining. Positive staining for p27 was found dominantly in low-grade tumors [p = 0.02] [Table 2]. Although the number was so small high grade p27 staining tended to be in low stage tumors [stage grouping was performed as: I and II as low, III and IV as high], but the difference was not statistically significant. On the other hand, there was no association between p27 expression and other clinicopathologic parameters. In statistical analysis, there was also no difference between p27 and cyclin D1 expression and Ki-67.
Forty-nine [77.8%] tumors showed positive nuclear staining with cyclin D1 while 3 [4.8%] cases showed no staining. There was a statistically significant difference among cyclin D1 nuclear staining scores in terms of tumor size [p = 0.03]. The other prognostic parameters did not relate significantly with nuclear or cytoplasmic staining with cyclin D1.
In the whole group, the mean proliferative index of the tumors were 41.5%, ranging from 2% to 97%. In statistical analysis, no significant relation was found between mean proliferative index of the tumors and prognostic parameters.
TGF-β1, p27 and clinical outcome
The median survival time was 13.4 months [range, 0–59 months] for low-expressors and the median survival time was 13.7 months [range, 1–38 months] for high-expressors for TGF-β1. The median survival time was 13.6 months [range, 0–59 months] for low-expressors and the median survival time was 13.4 months [range, 6–26 months] for high-expressors for p27. In the univariate analysis, neither TGF-β1 nor p27 expression was related with patient survival.
Discussion
Several tumor markers are studied to explain the mechanism of pancreatic carcinogenesis. Dysregulation of the normal cell cycle machinery is integrated to the neoplastic process and there is evidence implicating loss of cell cycle control in the development and progression of most human cancers, including pancreatic cancer. Regulation of cell cycle depends on the complex association between cyclin, CDK and CDKIs. According to their effects on cell cycle progression, these regulators qualify as proto-oncogenes [cyclin D1, D2, E] or tumor supressors [p21, p27]. Recently, studies have been performed in demonstration of p21, p27, cyclin D in tumors [17,22-25].
p27 is a CDKI that negatively regulate cyclin and its low expression results in proliferation of DNA damaged cells. Recent studies indicate that reduced expression of p27 is an independent predictor of poor outcome in cancers such as oral tongue squamous cell carcinoma, prostate cancer, gastrointestinal tract cancers, and even laryngeal precancerous lesions [14-18,26,27]. Lu et al demonstrated that loss of p27 expression independently predicts a poor prognosis for patients with resectable pancreatic adenocarcinomas [24]. Also anomalous overexpression of p27 in human pancreatic endocrine tumors were reported [25]. Inverse correlation between p27 expression and tumor grade, lymph node metastasis and stage of tumor was reported [14]. In the present study, we found positive staining for p27 in only minority of pancreatic adenocarcinoma cases and the staining was weak indicating that p27 is down-regulated during neoplastic process in pancreas. Additionally, although the difference was not statistically significant, p27 was dominantly expressed in low stage tumors. These results confirmed the previous reports. In a few reports, p27 and p21 staining did not correlate significantly with any clinicopathologic parameters [28,29], but there was an association between p27 and cyclin D1 expression indicating that the balance between the two opposing regulators was important for the end result of cell cycle progression [22]. Del Pizzo and Armengol found a positive correlation between p27 and cyclin E [30,31]. In the present study we could not find any relation between p27 and cyclin D1 expression.
It is well established that cell proliferation kinetics are important in predicting the prognosis of various tumors. Recenty, the highly proliferative activity as measured by Ki-67 antigen has been shown to correlate with a reduced p27 expression in various tumors suggesting that decreased expression of p27 may play important role in the increased cell turnover [32]. But the lack of correlation between cell cycle regulators and proliferation index has been also reported in some studies [22,27,31]. In the present study, we did not find any significant association between p27, cyclin D1 and proliferative activity. Cell cycle regulators might be more closely related with differentiation than proliferative activity, so induction of them may perhaps be necessary, but probably not sufficient for inhibition of cell cycle.
Growth factors, such as TGF-β and growth factor receptors have been shown to be overexpressed in a variety of cancers [10,33]. TGF-β can act as both a tumor suppresor and as a stimulator of tumor progression, invasion and metastasis. TGF-β inhibits normal cell proliferation and this effect is induced by binding specific cell receptors, type I [TβR-IALK5] and type II [TβR-II]. Both receptors were present in most cancer tissues and presence of them was associated with advanced tumor stage [10]. In pancreatic cancers, all three TGF-β isoforms are expressed at high levels and the presence of TGF-β isoforms is associated with shorter postoperative survival [34]. TGF-β isoforms were also markedly increased in hepatocellular carcinoma and prostate carcinoma [35-37]. Overproduction of TGF-β1 and loss of TGF-β-RII expression were found to be associated with poor clinical outcome [10]. Nevertheless several reports have indicated that TGF-β1 expression showed a significant correlation with low grade tumors and with longer survival time [19,29]. In the present study there was a statistically significant difference between TGF-β1 staining scores in terms of clinicopathologic factors such as blood vessel invasion, distant metastasis and stage. Our analysis revealed that overproduction of TGF-β1 correlates with tumor progression and metastasis. In the present study we could not find a correlation between TGF-β1 and p27 expression. TGF-β1 might influence cancer growth in other ways rather than cell cycle proteins. Hashimoto demonstrated a significant correlation between TGF-β1 and p21 expression and stated that the evaluation of TGF-β1 and p21 expression might be an effective tool in the prediction of the prognosis of patients with pancreatic cancer [29]. Grau and Ito also reported that TGF-β1 induces p21 expression in pancreatic cancer cells [11,38]. p21 may play more important role in the biology of pancreatic cancer than p27. We found no correlation between TGF-β1 and other proliferation markers studied. TGF-β1 induction of other CDKIs, such as p15 may partially participate in the inhibition of mitosis.
There is conflicting data in the literature about the correlation of TGF-β1 expression with survival in pancreatic cancer. In most series, the presence of TGF-β1 is associated with shorter postoperative survival [34], although in some other studies TGF-β1 expression showed a significant correlation with longer survival time [19,29]. On the other hand, there are reports indicating that TGF-β expression has no association with survival [39]. In the present study, to have the same treatment protocol only the patients which had RO resections were considered in survival analysis and TGF-β1 and p27 expression did not show a significant influence on the survival of patients. The small number of the patients may be the reason not to have definitive conclusions. TGF-β's role in controlling cellular growth and differentiation is very complex. In this study, findings may suggest that TGF-β1 may be a factor in the extention of the disease in pancreatic adenocarcinomas. But, progress may depend on the participation of other regulatory signals. Additional parameters related to prognosis and standard prognostic parameters are needed in the prediction of the disease outcome. Further studies that involve screening a larger number of patients with invasive disease will be required to establish the potential clinical usefulness of cell cyle regulators as prognostic markers in pancreatic cancer.
Conclusion
In the present study we examined the expression paterns of TGF-β1 and p27, which involved in cell cycle regulation, in pancreatic adenocarcinoma. Our findings suggest that in pancreatic carcinoma, TGF-β1 overexpression is more pronounced in metastatic and progressive ones. p27 expression is reduced in pancreatic adenocarcinomas and it may play a role in the differentiation of pancreatic cancer. Therefore, evaluation of these proteins might be useful in determining the agressive capacity of these tumors at an earlier stage.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
NC carried out the pathological studies, drafted and wrote the manuscript
ÖS participited in the pathological studies and performed the statistical analysis
SK, HA, İA performed the surgery and follow-up of the patients as surgeons
MS, İÖ made the endoscopic studies and performed the diagnosis as gastroenterologists
FO made the radiological analysis and participited in the diagnosis as a radiologist
All authors read and approved the final manuscript
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Figure 1 Intense [score 6] cytoplasmic TGF-β1 immunoreactivity of pancreatic ductal carcinoma cells [anti-TGF-β1, original magnification, × 400].
Figure 2 High level nuclear p27 immunoreactivity of pancreatic carcinoma cells [anti-p27, original magnification, × 400].
Table 1 Clinicopathological features
Number of patients [%]
Sex
Female 36 [57.1]
Male 27 [42.9]
Histologic grade
Well 16 [25.4]
Moderate 33 [52.4]
Poor 14 [22.2]
Stage
I 4 [6.4]
II 23 [36.5]
III 3 [4.8]
IV 33 [52.4]
Lymph node metastasis
Negative 24 [38.1]
Positive 39 [61.9]
Perineural invasion
Negative 19 [30.2]
Positive 44 [69.8]
Distant metastasis
Negative 30 [47.6]
Positive 33 [52.4]
Total 63 [100]
Table 2 Tissue marker expressions and tumor characteristics
TGF-β1 p 27 Cyclin D1
Variables
0 1 2 3 p 0 1 2 p 0 1 p
Sex
Female 13 7 8 8 0.68 26 6 4 0.10 0 31 0.06
Male 7 3 11 6 14 11 2 3 18
Grade
Well 8 2 5 1 0.58 12 3 1 0.02 2 13 0.42
Moderate 7 6 10 10 23 9 1 1 25
Poor 5 2 4 3 5 5 4 0 11
Stage
I 4 0 0 0 0.05 3 1 0 0.97 1 3 0.30
II 11 4 8 0 14 6 3 0 18
III 0 1 1 1 3 0 0 0 3
IV 5 5 10 13 20 10 3 2 25
Lymph node m.
Negative 10 6 7 4 0.34 19 7 1 0.34 3 19 0.23
Positive 8 3 6 9 15 7 4 0 22
Perineural inv.
Negative 3 1 3 2 0.95 6 1 2 0.21 1 7 0.47
Positive 16 8 11 9 28 13 3 2 32
Distant met.
Negative 15 5 9 1 <0.01 9 4 0 0.51 2 8 0.09
Positive 5 5 10 13 31 13 6 1 41
Blood vessel in.
Negative 14 6 10 4 0.01 23 9 2 0.62 3 25 1
Positive 4 3 5 9 13 5 3 1 17
Table 3 Relationships between TGF-β1, p 27 and cyclin D1
TGF-β1 immunoreactivity
0 1 2 3 p value
Cyclin D1
0 1 1 0 1 0.6
1 19 6 12 12
p 27
0 14 6 13 7 0.7
1 5 3 4 5
2 1 1 2 2
==== Refs
Solcia E Capella C Klöppel G Tumors of the exocrine pancreas Atlas of Tumor Pathology Tumors of the Pancreas Third series, fascicle 20 1997 Washington, D.C: AFIP 64 88
Ozawa F Friess H Tempia-Caliera A Kleeff J Büchler MW Growth factors and their receptors in pancreatic cancer Teratog Carcinog Mutagen 2001 21 27 44 11135319 10.1002/1520-6866(2001)21:1<27::AID-TCM4>3.0.CO;2-9
Itakura J Ishiwata T Shen B Kornmann M Korc M Concomitant over-expression of vascular endothelial growth factor and its receptors in pancreatic cancer Int J Cancer 2000 85 27 34 10585578 10.1002/(SICI)1097-0215(20000101)85:1<27::AID-IJC5>3.0.CO;2-8
Karademir S Sökmen S Terzi C Sağol Ö Özer E Astarcıoğlu H Çoker A Astarcıoğlu İ Tumor angiogenesis as a prognostic predictor in pancreatic cancer J Hepatobiliary Pancreat Surg 2000 7 489 495 11180876 10.1007/s005340070020
Apple SK Hecht JR Lewin DN Jahromi SA Grody WW Nieberg RK Immunohistochemical evaluation of K-ras, p53, and HER-2/neu expression in hyperplastic, dysplastic, and carcinomatous lesions of the pancreas: evidence for multistep carcinogenesis Hum Pathol 1999 30 123 129 10029438 10.1016/S0046-8177(99)90265-4
Eskelinen MJ Haglund UH Prognosis of human pancreatic adenocarcinoma: review of clinical and histopathological variables and possible uses of new molecular methods Eur J Surg 1999 165 292 306 10365829 10.1080/110241599750006811
Hruban RH Offerhaus GJA Kern SE Goggins M Wilentz RE Yeo CJ Tumor-suppresor genes in pancreatic cancer J Hepatobiliary Pancreat Surg 1998 5 383 391 9931387 10.1007/s005340050062
Wei D Le X Zheng L Wang L Frey JA GAo AC Peng Z Huang S Xiong HQ Abbruzzese JL Xie K Stat3 activation regulates the expression of vascular endothelial growth factor and human pancreatic cancer angiogenesis and metastasis Oncogene 2003 22 319 329 12545153 10.1038/sj.onc.1206122
Akhurst R Derynck R TGF-beta signaling in cancer- a double-edged sword Trends Cell Biol 2001 11 S44 S51 11684442
Lu Z Friess H Graber HU Gou X Schilling M Zimmermann A Korc M Büchler MW Presence of two signaling TGF-β receptors in human pancreatic cancer cor TGF-β relates with advanced tumor stage Dig Dis Sci 1997 42 2054 2063 9365135 10.1023/A:1018814416903
Grau AM Zhang L Wang W Ruan S Evans DB Abbruzzese JL Zhang W Chiao PJ Induction of p21waf1 expression and growth inhibition by transforming growth factor β involve the tumor suppresor gene DPC4 in human pancreatic adenocarcinoma cells Cancer Res 1997 57 3929 3934 9307274
Govinden R Bhoola KH Genealogy, expression, and cellular function of transforming growth factor-β Pharmacol Ther 2003 98 257 265 12725873 10.1016/S0163-7258(03)00035-4
Michalides RJAM Cell cycle regulators: mechanisms and their role in aetiology, prognosis, and treatment of cancer J Clin Pathol 1999 52 555 568 10645224
Mineta H Miura K Suzuki I Takebayashi S Amano H Araki K Harada H Ichimura K Wennerberg JP Dictor MR Low p27 expression correlates with poor prognosis for patients with oral tongue squamous cell carcinoma Cancer 1999 85 1011 1017 10091782 10.1002/(SICI)1097-0142(19990301)85:5<1011::AID-CNCR1>3.0.CO;2-0
Yang RM Naitoh J Murphy M Wang H Phillipson J Dekernion JB Loda M Reiter RE Low p27 expression predicts poor disease-free survival in patients with prostate cancer J Urol 1998 159 941 945 9474188 10.1097/00005392-199803000-00088
Thomas GV Schrage MI Rosenfelt L Kim JH Salur G Dekernion JB Dorey F Said J Reiter RE Preoperative prostate needle biopsy p27 correlates with subsequent radical prostatectomy p27, Gleason grade and pathological stage J Urol 2000 164 1987 1991 11061897 10.1097/00005392-200012000-00028
Vis AN Noordzij MA Fitoz K Wildhagen MF Schröder FH VanDerKwast TH Prognostic value of cell cycle proteins p27Kip1 and MIB-1, and the cell adhesion protein CD44s in surgically treated patients with prostate cancer J Urol 2000 164 2156 2161 11061947 10.1097/00005392-200012000-00083
Claudio PP Zamparelli A Garcia FU Claudio L Ammirati G Farina A Bovicelli A Russo G Giordano GG McGinnis DE Giordano A Cardi G Expression of cell-cycle-regulated proteins pRb2/p130, p107, p27Kip1, p53, and Ki-67 (MIB-1) in prostatic gland adenocarcinoma Clin Cancer Res 2002 8 1808 1815 12060621
Coppola D Lu L Fruehauf JP Kyshtoobayeva A Karl RC Nicosia SV Yeatman TJ Analysis of p53, p21WAF1, and TGF-β1 in human ductal adenocarcinoma of the pancreas Am J Clin Pathol 1998 110 16 23 9661918
De Manzoni G Verlato G Tomezzoli A Study on Ki-67 immunoreactivity as a prognostic indicator in patients with advanced gastric cancer Jpn J Clin Oncol 1998 28 534 537 9793024 10.1093/jjco/28.9.534
Holland TA Elder J McCloud JM Hall C Deakin M Fryer AA Elder JA Hoban PR Subcellular localisation of cyclin D1 protein in colorectal tumors is associated with p21 WAF1/CIP1 expression and correlates with patient survival Int J Cancer 2001 95 302 306 11494229 10.1002/1097-0215(20010920)95:5<302::AID-IJC1052>3.0.CO;2-#
Oh YL Choi JS Song SY Ko YH Han BK Nam SJ Yang JH Expression of p21Waf1, p27Kip1 and cyclin D1 proteins in breast ductal carcinoma in situ: relation with clinicopathologic characteristics and with p53 expression and estrogen receptor status Pathol Int 2001 51 94 99 11169147 10.1046/j.1440-1827.2001.01173.x
Tallini G Garcia-Rostan G Herrero A Zelterman D Viale G Bosari S Carcangiu ML Downregulation of p27kıp1 and Ki67/Mib1 labeling index support the classification of thyroid carcinoma into prognostically relevant categories Am J Surg Pathol 1999 23 678 685 10366150 10.1097/00000478-199906000-00007
Lu CD Morita S Ishibashi T Hara H Isozaki H Tanigawa N Loss of p27Kip1 expression independently predicts poor prognosis for patients with resectable pancreatic adenocarcinoma Cancer 1999 85 1250 1260 10189129 10.1002/(SICI)1097-0142(19990315)85:6<1250::AID-CNCR6>3.0.CO;2-#
Guo SS Wu X Shimoide AT Wong J Sawicki MP Anomalous overexpression of p27Kip1 in sporadic pancreatic endocrine tumors J Surg Res 2001 96 284 288 11266285 10.1006/jsre.2001.6085
Chetty R p27 protein and cancers of the gastrointestinal tract and liver: an Overview J Clin Gastroenterolog 2003 37 23 27 10.1097/00004836-200307000-00008
Hirai T Hayashi K Takumida M Ueda T Hirakawa K Yajin K Reduced expression of p27 is correlated with progression in precancerous lesions of the larynx Auris Nasus Larynx 2003 30 163 168 12753988 10.1016/S0385-8146(03)00007-5
Zamparelli A Mascıullo V Bovicelli A Santini D Ferrandina G Minimo C Terzano P Costa S Cinti C Ceccarelli C Mancuso S Scambia G Bovicelli L Giordano S Expression of cell-cycle-associated proteins pRb2/p130 and p27Kip1 in vulvar squamous cell carcinomas Hum Pathol 2001 32 4 9 11172288 10.1053/hupa.2001.20371
Hashimoto K Nio Y Sumi S Toga T Omori H Itakura M Yano S Correlation between TGF-beta1 and p21 (WAF1/CIP1) expression and prognosis in resectable invasive ductal carcinoma of the pancreas Pancreas 2001 22 341 347 11345133 10.1097/00006676-200105000-00002
Del Pizzo JJ Borkowski A Jacobs SC Kyprianou N Loss of cell cycle regulators p27Kip1 and cyclin E in transitional cell carcinoma of the bladder correlates with tumor grade and patient survival Am J Pathol 1999 155 1129 1136 10514396
Armengol C Boix L Bachs O Sole M Fuster J Sala M Llovet JM Rodes J Bruix J p27Kip1 is an independent predictor of recurrence after surgical resection in patients with small hepatocellular carcinoma J Hepatol 2003 38 591 597 12713869 10.1016/S0168-8278(03)00025-4
Gelen T Elpek GO Aksoy NH Ogus M Keles N p27 labeling index and proliferation in gastrointestinal stromal tumors: correlations with clinicopathologic factors and recurrence Jpn J Clin Oncol 2003 33 346 352 12949061 10.1093/jjco/hyg071
Bellone G Carbone A Tibaudi D Mauri F Ferrero I Smirne C Suman F Rivetti C Migliaretti G Camandona M Palestro G Emanuelli G Rodeck U Differential expression of transforming growth factors-β1, -β2 and -β3 in human colon carcinoma Eur J Cancer 2001 37 224 233 11166150 10.1016/S0959-8049(00)00391-9
Friess H Yamanaka Y Büchler M Ebert M Beger HG Gold LI Korc M Enhanced expression of transforming growth factor β isoforms in pancreatic cancer correlates with decreased survival Gastroenterology 1993 105 1846 1856 8253361
Abou-Shady M Baer HU Friess H Berberat P Zimmermann A Graber H Gold LI Korc M Büchler MW Transforming growth factor betas and their signaling receptors in human hepatocellular carcinoma Am J Surg 1999 177 209 215 10219856 10.1016/S0002-9610(99)00012-4
Wikstrom P Stattin P Franck-Lissbrant I Damber JE Bergh A Transforming growth factor beta1 is associated with angiogenesis, metastasis, and poor clinical outcome in prostate cancer Prostate 1998 37 19 29 9721065 10.1002/(SICI)1097-0045(19980915)37:1<19::AID-PROS4>3.0.CO;2-3
Logullo AF Nonogaki S Miguel RE Kowalski LP Nishimoto IN Pasini FS Federico MHH Brentani RR Brentani MM Transforming growth factor β1 (TGF β1) expression in head and neck squamous cell carcinoma patients as related to prognosis J Oral Pathol Med 2003 32 139 145 12581383 10.1034/j.1600-0714.2003.00012.x
Ito D Fujimoto K Doi R Koizumi M Toyoda E Mori T Kami K Kawagucgi Y Whitehead R Imamura M Chronic exposure of transforming growth factor beta 1 confers a more aggressive tumor phenotype through downregulation of p21Waf1/Cip1 in conditionally immortalized pancreatic epithelial cells Surgery 2004 136 364 374 15300203 10.1016/j.surg.2004.05.012
Sears D Erickson RA Sayage-Rabie L Escobar MC TGF-β1 and p53 staining in CT-guided and endoscopic ultrasound fine-needle aspirates of pancreatic adenocarcinoma Dig Dis Sci 2004 49 828 832 15259505 10.1023/B:DDAS.0000030095.71501.04
|
16086840
|
PMC1208869
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Aug 8; 5:98
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-98
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-991609115010.1186/1471-2407-5-99Research ArticleCharacterization of the linkage disequilibrium structure and identification of tagging-SNPs in five DNA repair genes Allen-Brady Kristina [email protected] Nicola J [email protected] Genetic Epidemiology, Department of Medical Informatics; University of Utah School of Medicine; 391 Chipeta Way, Suite D; Salt Lake City, Utah, 84108, USA2005 9 8 2005 5 99 99 22 4 2005 9 8 2005 Copyright © 2005 Allen-Brady and Camp; 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
Characterization of the linkage disequilibrium (LD) structure of candidate genes is the basis for an effective association study of complex diseases such as cancer. In this study, we report the LD and haplotype architecture and tagging-single nucleotide polymorphisms (tSNPs) for five DNA repair genes: ATM, MRE11A, XRCC4, NBS1 and RAD50.
Methods
The genes ATM, MRE11A, and XRCC4 were characterized using a panel of 94 unrelated female subjects (47 breast cancer cases, 47 controls) obtained from high-risk breast cancer families. A similar LD structure and tSNP analysis was performed for NBS1 and RAD50, using publicly available genotyping data. We studied a total of 61 SNPs at an average marker density of 10 kb. Using a matrix decomposition algorithm, based on principal component analysis, we captured >90% of the intragenetic variation for each gene.
Results
Our results revealed that three of the five genes did not conform to a haplotype block structure (MRE11A, RAD50 and XRCC4). Instead, the data fit a more flexible LD group paradigm, where SNPs in high LD are not required to be contiguous. Traditional haplotype blocks assume recombination is the only dynamic at work. For ATM, MRE11A and XRCC4 we repeated the analysis in cases and controls separately to determine whether LD structure was consistent across breast cancer cases and controls. No substantial difference in LD structures was found.
Conclusion
This study suggests that appropriate SNP selection for an association study involving candidate genes should allow for both mutation and recombination, which shape the population-level genomic structure. Furthermore, LD structure characterization in either breast cancer cases or controls appears to be sufficient for future cancer studies utilizing these genes.
==== Body
Background
Candidate gene association studies are a powerful study design for complex diseases such as cancer. Advances in association studies have been furthered by the recent discovery of single nucleotide polymorphisms (SNPs); their vast density throughout the genome, ease of genotyping and moderate cost contribute greatly to their utility. Association testing is efficient when the SNPs being analyzed represent the entire genetic variation of the gene. It has been suggested that nearby SNPs are organized into regions of high linkage disequilibrium (LD) separated by short segments of very low LD [1-6]. In Caucasians, high LD regions may vary in length from a few kb to >300 kb[2,6,7]. Regions of high LD contain redundant information and can be reduced to smaller subsets of tagging-SNPs (tSNPs)[8], such that tSNPs identify all common haplotypes within the region of high LD. A number of algorithms have been proposed to define regions of high LD and tSNPs[4,8-14]. Thus far, no consensus of which algorithm is best has been achieved. Several studies have suggested the utility of matrix decomposition algorithms.[12,13,15-17]. One advantage of these algorithms is that SNPs in high LD are not required to be contiguous nor mutually exclusive, a flexibility that is necessary for analyzing small genomic regions and rare variants. Further, these methods are stable with regards to marker density, minor allele frequency, analysis window, and possible analysis window length[18].
Growing evidence appears to suggest that tumorigenesis is a multi-step process of genetic alterations that transform a normal human cell into a malignant derivative[19]. The ability of a cell to maintain genomic stability through DNA repair mechanisms is essential to prevent tumor initiation and progression. A number of different types of cancer have been attributed to defective DNA repair including xeroderma pigmentosum[20], hereditary nonpolyposis colorectal cancer[21], and breast cancer due to mutations in BRCA1 and BRCA2 as well as other DNA repair genes (e.g., ATM, TP53 and CHK2)[22]. Many published candidate gene association studies involving DNA repair genes and cancer risk have assessed risk by examining a single SNP per gene or a single locus at a time analysis approach. Unfortunately, the former approach is often inadequate in comprehensively accounting for the genetic variation of a gene, and the latter incurs multiple testing corrections, which usually eliminate all or most of the association evidence found. It has been suggested that use of haplotypes in association studies may have increased power over single-allele studies[8]. Descriptions of haplotype diversity and LD structure as well as identification of potential tSNPs will be key for success in candidate gene association studies.
Here we describe haplotypes, LD structure and potential tSNPs in five DNA repair breast cancer susceptibility genes: ATM, MRE11A, NBS1, RAD50, and XRCC4. We used a matrix decomposition algorithm based on a method of principal components analysis[13]; this method does not require SNPs to be in contiguous block structure. Characterization of the LD structure and tSNPs are necessary for the design of future effective association studies.
Methods
Subjects
This study is part of a larger study involving 139 high-risk Caucasian breast cancer families, defined as high risk because cancer rates in these families were significantly higher than the general population rate determined using the Utah Population Database (UPDB) [23-25]. All breast cancer cases in the larger cohort met at least one of the following criteria: 1.) their family tested negative for a BRCA1 or BRCA2 mutation, 2.) the case themselves tested negative for the same BRCA1/2 mutation that was present in their family, or 3.) their family had a low probability of carrying a BRCA1/2 mutation based on the number of breast cancer cases present in the family and/or ages at diagnosis of breast cancer within the family. Therefore, all breast cancer cases in the larger study had a low residual probability of their cancer being due to mutations in BRCA1/2. Breast cancer diagnosis information was obtained from medical records for the subject or the Utah Cancer Registry.
For this LD characterization study, we selected a panel of 94 individuals (47 female breast cancer cases and 47 female controls), chosen randomly from separate kindreds to ensure independence. Both cases and controls were chosen such that comparisons of LD structure could be made between the groups. The sample size of 188 chromosomes is larger than generally used for this type of study [26-29], but inadequate for an association analysis. This current study is not a case-control study and associations with disease were not assessed.
Blood samples were collected on all subjects and all individuals signed consent to participate this study. This study was approved by the University of Utah Institutional Review Board.
Genes and SNP selection
For each gene of interest (i.e., ATM, MRE11A, NBS1, RAD50 and XRCC4), all SNPs available from Applied Biosystems[30], within each gene and the flanking 10 kb on either side, that had been validated to have a minor allele frequency greater than 0.01 in Caucasians were selected. For ATM (on chromosome 11q22-q23), which spans approximately 143 kb and contains 64 exons, 14 SNPs were studied with a SNP resolution of 1 SNP/10,489 bp. For MRE11A (11q21), which spans approximately 76 kb and contains 20 exons, 11 SNPs were studied with a SNP resolution of 1 SNP/8539 bp. For NBS1 (8q21), which contains 16 exons and spans about 51 kb, 5 SNPs were studied with a SNP resolution of 1 SNP/8256 bp. The RAD50 gene (5q31) spans approximately 87 kb contains 25 exons, and we studied 10 SNPs at a resolution of 1 SNP/10,533 bp. Finally, for XRCC4 (5q13-q14) with 8 exons and approximately 276 kb in length, we studied 21 SNPs at a resolution of 1 SNP/13,198 bp. The vast majority of the SNPs studied were intronic (see Table 1).
Genotyping
For the ATM and XRCC4 all SNPs that met the above criteria were genotyped on our panel of 94 subjects. For MRE11A, one SNP repeatedly failed to amplify (rs10831224) and was removed from the study.
Genomic DNA was isolated and purified using standard phenol/chloroform DNA extraction. SNP genotyping was performed using the fluorogenic 5' nuclease TaqMan Assay[31] (Applied Biosystems). The TaqMan Assay requires TaqMan PCR Master Mix (Applied Biosystems), which we used according to manufacturer's instructions, yielding a final volume of 5 μl per well. PCR amplification was also performed according to the Applied Biosystems protocol. The 7900HT Sequence Detection System (Applied Biosystems) was used to measure each fluorescent dye-labeled probe specific for each allele studied and results were analyzed with the Sequence Detection Software (Applied Biosystems).
Haplotype structure and tSNP selection
Haplotypes and haplotype frequencies were estimated from unphased genotype data using an expectation-maximization algorithm, SNPHAP[32]. SNPHAP uses a maximum-likelihood program to predict multilocus haplotypes. Haplotypes with a frequency of at least 0.01 were analyzed using a two-step PCA method[13]. This method does not require that groups of SNPs be contiguous along a DNA fragment and also allows SNPs to be present in more than one group. In step I, LD groups are determined. In brief, the PCA method extracts factors (LD groups) to capture ≥ 90% of the genetic diversity. An LD group is defined as those SNPs that load onto the same factor. In step II, tSNPs are selected for each LD group. Each LD group is considered separately and the PCA method again extracts factors; tSNPs are chosen as the SNPs with the highest factor loading. When a number of SNPs load equally well on an LD group, these can all be considered potential tSNPs. Under such circumstances, we selected the single SNP that performed best in the genotyping assay. This was done in order to minimize errors in allele calls.
We compared our genotype data for ATM, MRE11A, and XRCC4 with genotyping data for these same genes obtained from Applied Biosystems (ABI)[30] on 45 Caucasians. We found good concordance in allele frequencies between the data sets. Further, we applied the same LD characterization to both data sets and found excellent concordance in the LD groups and potential tSNPs (see Results). We therefore characterized LD groups and tSNPs for NBS1 and RAD50 using the genotyping data available online.
We also examined whether differences existed between LD group structure and tSNP selection when cases and controls were considered separately. This analysis could only be performed for ATM, MRE11A, and XRCC4.
Results
Characteristics of the SNPs studied are listed in Table 1. Minor allele frequencies from our 94 subjects compared well with those listed by Applied Biosystems[30]. Despite the very low minor allele frequencies in some of the SNPs studied, we observed heterozygosity for all SNPs genotyped.
Table 2 lists the haplotypes with a frequency > 0.01 obtained from SNPHAP, and the LD group designation and the tSNPs that were selected using the PCA method, for ATM, MRE11A, and XRCC4. Haplotypes are reported using the standard convention of designating the major allele as '1' and the minor allele as '2', in order to more easily spot occurrences of the minor allele. Please see Table 1 for the corresponding base pair change. For ATM, 7 haplotypes overall were observed and 5 had a frequency > 0.01. Using the PCA method, a single LD group was identified, encompassing the entire gene and accounting for 98.8% of the genetic variance across the gene. From this single LD group, a single tSNP (A13) was selected.
For MRE11A, we observed 9 haplotypes in total and 6 with frequency > 0.01. From the PCA analysis, four LD groups were identified based on these 6 haplotypes with a frequency > 0.01, and accounted for 99.1% of the genetic variance. The LD groups did not conform to haplotype blocks. SNP M4 separated LD group 1 into two parts and M8 separated LD group 2. Each LD group was represented by a single tSNP, such that the tSNP set contained 4 tSNPs (M6, M10, M11, and M14).
For XRCC4, we observed 26 haplotypes overall; 13 of which had a frequency >0.01. From the PCA method, four LD groups were observed which accounted for 97.2% of the variance. Similarly to MRE11A, the LD groups were not contiguous blocks. LD group 1 was divided by X9 and LD group 2 was divided by X15. Each of the LD groups could be represented by a single SNP resulting in the tSNP set (X2, X9, X14, and X21).
Table 3 shows the LD groups and tSNPs for ATM, MRE11A and XRCC4 using our panel of 94 subjects and using the 45 Caucasian subjects from Applied Biosystems[30]. For these three genes, we observed the same number of LD groups containing precisely the same SNPs for both data sets. The difference between the results was in the number of potential tSNPs for each LD group. For the majority of LD groups, the potential tSNPs using Applied Biosystems data were a subset of those from our data. This is perhaps expected, because our sample size was more than double their size and is therefore capable of better resolution.
Table 4 lists the haplotypes, LD group designation, potential tSNPs, and tSNP selected per group for NBS1 and RAD50 using the Applied Biosystems' data. For NBS1, 6 haplotypes overall were observed and all 6 haplotypes had a frequency > 0.01. Using the PCA method, two LD groups were identified and accounted for 93.8% of the variance. Two tSNPs were sufficient to tag these groups (N1, N2). However, N5 could replace N2 with no reduction in the variance explained. For the RAD50 gene, in order to include two available rare SNPs in the analysis, we lowered the haplotype acceptance threshold to 0.009. We observed a total of 14 haplotypes, 10 with a frequency greater than 0.01. Using the PCA method, we identified three LD groups, which accounted for 91.5% of the variance. Similarly to MRE11A and XRCC4, the LD groups for RAD50 were not contiguous blocks. Three tSNPs were sufficient to tag the groups (R1, R3, and R10), although R5 could replace R1 and R6 could replace R3 with no loss of variance explained.
For ATM, MRE11A, and XRCC4, we compared haplotypes and LD structure between the breast cancer cases and controls. For ATM and XRCC4 no difference in the LD structure was observed when cases and controls were analyzed separately. For the MRE11A gene differences in LD structure were noted, however, these were minor and likely attributable to small sample size since the differences were driven by 3 rare haplotypes (frequency = 0.02).
Discussion
Identification of the most informative markers to use in a large-scale association analysis for studies of complex disease, such as breast cancer, is critical to the success of the study. The key to this process is to select SNPs that are most informative about the underlying haplotype structure in a population of interest. As haplotype based designs have been suggested as being more powerful than the single-allele approach for association studies[8], a haplotype-based approach should result in more accurate and definitive findings. In this study, we have described haplotypes and characterized the LD structure of the ATM, MRE11A, and XRCC4 genes using a panel of 94 subjects, including breast cancer cases from high-risk breast cancer families as well as controls. Further, we identified tSNPs that can be used in future haplotype-based association studies. A similar analysis was performed for NBS1 and RAD50 using publicly available genotype data. We identified, using Principal Components Analysis[13], a single LD group for ATM, four noncontiguous LD groups for MRE11A, two LD groups for NBS1, three noncontiguous LD groups for RAD50, and four noncontiguous LD groups for XRCC4. In each case, the LD groups captured greater than 90% of the variance of the total SNPs available from Applied Biosystems across the gene. Furthermore for each gene, we present tSNPs that could be selected to represent the gene.
It is of interest that the LD structure for three of these five DNA repair genes did not conform to the haplotype block model, that is, that the LD groups did not contain contiguous SNPs. This was true whether the genotyping data came from our own study or from Applied Biosystems. Although we did not directly sequence these genes to identify all possible variants, the discontinuity we observed illustrates that the underlying LD structure cannot conform to contiguous haplotype blocks. A more flexible LD group representation (as supported under principle components analysis) fit the data better and appears to be stable to differences in minor allele frequency. Similar findings of a complex pattern of LD structure were recently reported in a high-resolution study of the ELAC2 gene[15]. Our results suggest that when studying small genomic regions and low frequency variants (<0.2), mutation is an important dynamic in LD structure, and the simple recombination-only model used in classical haplotype block methods does not fit the data well and hence will lead to a poor selection of tSNPs.
Due to the stability of the results for ATM, MRE11A and XRCC4, we pursued two additional DNA repair genes of interest (i.e., NBS1 and RAD50). Applied Biosystems provides freely-available genotyping data for four ethnically diverse populations of 45 subjects in each, therefore, even with limited funds, the haplotype structure and selection of tSNPs can be estimated for a study prior to any genotyping costs. However, caution must be used if this option is exercised as one's population must be one of Applied Biosystems' ethnic cohorts (i.e., Caucasian, African American, Chinese, or Japanese) and our experience is that occasionally errors exist in the data.
Of the genes studied here, only ATM has previously been studied in any depth for LD structure. The reason that ATM has received so much attention is that patients with the recessive disease ataxia-telangiectasia, due to a mutation in the ATM gene, have a 100-fold increased risk of cancer[33,34] and obligate heterozygous carriers of ATM mutations may have an increased risk of cancer, particularly breast cancer [35-39], although this finding is controversial[40,41]. Extensive LD across the ATM gene has previously been reported [42-44], and sequence analysis reveals that ATM polymorphisms are relatively rare resulting in low overall sequence diversity[44]. Thus, it follows that only a small number of haplotypes have been found, particularly in Caucasian populations of European descent. Thorstenson et al [44] predicted seven haplotypes in populations throughout the world, only three of which were found in Europeans or the Americans. Bonnen et al [43] identified 22 unique haplotypes, seven of which occurred in Caucasians, and only five of these occurred at a frequency of greater than 5% among Caucasians. We observed five haplotypes for the ATM gene, but only two of these could be considered common haplotypes (>0.01) and together accounted for 96% of all chromosomes. A recently published study using those haplotypes defined by Thorstenson et al[44] and Bonnen et al[43] identified five haplotype tagging-SNPs that were necessary to capture all of these haplotypes with a frequency >1%[45]. In our study, which is limited to Applied Biosystems' validated SNPs, we found that one tSNP was sufficient to represent 98.8% of the total genetic variance for all the SNPs available. The results of our study differed from these other studies due most likely to differences in the minor allele frequency range of the SNPs utilized. Our minor allele frequency for the 14 SNPs studied in the ATM gene varied minimally from 0.43 – 0.45. Thorstenson et al[44] and Bonnen et al[43] included 2 and 3 SNPs, respectively, that had minor allele frequencies <0.25. Population structure exists in SNP-allele frequencies[43] and as observed by the results of this study, exclusion of rarer SNPs has an impact on the frequency of haplotypes that are observed.
Comparison of haplotype and LD structure between cases and controls for ATM, MRE11A, and XRCC4 indicated that LD structure for these genes were similar in both groups. Results for ATM and XRCC4 were identical and only minor differences in LD structure were noted for MRE11A due to three rare haplotypes. A recent study has reported that rare haplotypes may be important for disease susceptibility and in their study these rare haplotypes had significant effects on their phenotype of interest[46]. Therefore, if rare haplotypes are of interest to an investigator, it may be prudent to characterize LD in both cases and controls and select tSNPs that comprehensively cover the diversity of both groups. However, most studies to date have empirically found that LD structure is similar across phenotype[1,47]. If major differences in LD structure were to exist, this would have a profound effect on guidelines for tSNP selection and for application of projects such as the HapMap[48,49].
Some limitations are inherent in this study and must be pointed out. First, we did not sequence our genes of interest and thus all of the genetic diversity within these genetic regions may not be captured. Our results must be interpreted in light of this. The gold standard is to identify all variants within a gene and select a subset of tSNPs from this set. It would be interesting to evaluate the robustness of our findings using sequence data. However, the SNPs examined were relatively evenly spaced, on the order of 1 SNP every 10 kb, and our results are important as they illustrate how smaller budget studies can best select tSNPs. Second, our sample size was modest (188 chromosomes), although larger than other previous studies examining LD and tSNPs [26-29]. Finally, haplotype block and haplotype-tagging SNP analyses have been suggested to only be reliable when markers are dense, otherwise marker sets have considerable loss of information[50]. This result may extend to PCA methods, however, the matrix decomposition algorithm used has been suggested to be stable with regards to varying levels of marker density[18].
Conclusion
In conclusion, we have described haplotypes, linkage disequilibrium structure, and identified tSNPs from all available Applied Biosystems' validated SNPs in ATM, MRE11A, NBS1, RAD50, and XRCC4 genes in a Caucasian population. As has been found for other genes, we identified LD structures that did not conform to contiguous haplotype block structures. This illustrates the importance of using flexible methods, such as matrix decomposition, that allow for multiple population dynamics such as recombination, mutation and selection. Although the gold standard for SNP characterization across a candidate gene is sequencing to identify all variants, we describe a low-budget means to characterize the LD structure and select tSNPs using publicly available data. Comprehensive characterization of the LD structure at genes of interest will be essential for future, effective association studies.
Electronic database information
The data from the 94 breast cancer case and control subjects for these tables is publicly available at under Supplemental Materials to Publication. On request from Dr. Nicola Camp a username and password to access the data will be given.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
KAB assisted in the study design, performed the genotyping, and drafted the manuscript. NJC conceived of the study and its design and helped to draft the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Kristina Allen-Brady is an NLM fellow, supported by NLM grant T15 LM0724. This research was supported by a dissertation research grant from the Susan G. Komen Breast Cancer Foundation for Kristina Allen-Brady (DISS0201521, to NJC) and an NIH NCI grant CA 098364 (to NJC). We appreciate the assistance of Kim Nguyen (Genetic Epidemiology) and Michael Hoffman (Family and Preventive Medicine) for their help in the laboratory. We also thank Helaman Escobar (Director of Sequencing and Genomics) and Michael Klein (Genomics) from the Core Resource Facilities, University of Utah, for use of their equipment and assistance on this project. Data collected for this publication was assisted by the Utah Cancer Registry supported by National Institutes of Health, Contract NO1-PC-35141, Surveillance, Epidemiology and End Results (SEER) Program, with additional support from the Utah Department of Health and the University of Utah. Partial support for all datasets within the Utah Population Database (UPDB) was provided by the University of Utah Huntsman Cancer Institute.
Figures and Tables
Table 1 Characteristics of SNPs analyzed
Gene SNP Code SNP ID Base change* Position† MAF‡ ABI reported MAF§ # bp from the most 5' SNP
ATM A1 rs228589 T/A Flanking 0.45 0.33 0
ATM A2 rs228591 G/A mRNA-utr 0.45 0.33 4125
ATM A3 rs641605 T/C Intron 0.45 0.33 8,711
ATM A4 rs228599 A/G Intron 0.44 0.31 14,452
ATM A5 rs600931 T/C Intron 0.45 0.35 24,127
ATM A6 rs228592 A/C Intron 0.45 0.33 29,981
ATM A7 rs664677 T/C Intron 0.43 0.33 49,974
ATM A8 rs1003623 T/C Intron 0.45 0.33 59,374
ATM A9 rs609261 C/T mRNA-utr, intron 0.45 0.32 64,926
ATM A10 rs645485 G/A Intron 0.45 0.32 75,655
ATM A11 rs673281 A/G Intron 0.45 0.31 88,861
ATM A12 rs227061 G/A mRNA-utr, intron 0.45 0.34 112,121
ATM A13 rs227062 A/G mRNA-utr, intron 0.45 0.33 112,175
ATM A14 rs652311 A/G Flanking 0.45 0.36 146,861
MRE11 M1 rs646130 T/C Flanking 0.3 0.39 0
MRE11 M2 rs491404 G/C Flanking 0.3 0.4 9192
MRE11 M3 rs10831227 G/A Intron 0.3 0.4 16,336
MRE11 M4 rs601341 G/A Intron 0.38 0.36 28,536
MRE11 M5 rs554715 T/C Intron 0.3 0.4 32,986
MRE11 M6 rs556477 A/G Intron 0.3 0.4 40,565
MRE11 M7 rs1805365 A/G Intron 0.02 0.02 61,721
MRE11 M8 rs680695 A/G Intron 0.34 0.36 72,913
MRE11 M9 rs1009455 C/G Intron 0.02 0.01|| 85,033
MRE11 M10 rs1009456 C/A locus-region, mRNA-utr 0.01 0.02 87,401
MRE11 M11 rs10831234 C/T Flanking 0.09 0.06 93,946
NBS1 N1 rs12680687 G/T Intron - ** 0.28 0
NBS1 N2 rs709816 A/G Coding-synon - 0.45 16,323
NBS1 N3 rs1805790 C/T Intron - 0.39 23,313
NBS1 N4 rs741778 C/G Intron - 0.36 33,415
NBS1 N5 rs1805841 C/G Intron - 0.45 41,282
RAD50 R1 rs2522406 G/A Flanking - 0.01 0
RAD50 R2 rs2244012 C/T Intron - 0.19 12,116
RAD50 R3 rs2299015 T/G Intron - 0.19 12,388
RAD50 R4 rs2299014 G/T Intron - 0.41 14,290
RAD50 R5 rs2706377 A/G Intron - 0.01 50,388
RAD50 R6 rs2301713 C/T intron - 0.19 62,887
RAD50 R7 rs2040703 C/G Intron - 0.22 83,149
RAD50 R8 rs2240032 C/T Intron - 0.18 88,018
RAD50 R9 rs1800925 C/T Flanking - 0.19 103,700
RAD50 R10 rs2066960 C/A Flanking - 0.17 105,326
XRCC4 X1 rs1993948 T/A Flanking 0.46 0.47 0
XRCC4 X2 rs1478485 G/A mRNA-utr 0.47 0.45 8247
XRCC4 X3 rs11951257 T/C Intron 0.47 0.45 31,031
XRCC4 X4 rs10045104 C/T Intron 0.43 0.42 40,082
XRCC4 X5 rs6452526 C/T Intron 0.47 0.43 64,531
XRCC4 X6 rs1382369 G/A Intron 0.47 0.43 69,149
XRCC4 X7 rs1382368 C/T Intron 0.47 0.41 78,795
XRCC4 X8 rs1382363 C/T Intron 0.47 0.42 80,292
XRCC4 X9 rs13180316 G/A Intron 0.23 0.26 87,173
XRCC4 X10 rs11741420 A/T Intron 0.47 0.44 98,452
XRCC4 X11 rs2731861 T/C Intron 0.47 0.45 112,984
XRCC4 X12 rs2662238 G/A Intron 0.46 0.45 127,027
XRCC4 X13 rs1039786 C/T Intron 0.46 0.45 127,761
XRCC4 X14 rs963248 T/C Intron 0.19 0.16 161,614
XRCC4 X15 rs301276 G/A Intron 0.23 0.23 175,451
XRCC4 X16 rs35268 T/C Intron 0.16 0.13 216,216
XRCC4 X17 rs301286 T/C Intron 0.16 0.18 230,675
XRCC4 X18 rs301289 C/T Intron 0.17 0.17 233,955
XRCC4 X19 rs2386275 G/A Intron 0.09 0.12 270,260
XRCC4 X20 rs2891980 T/C Intron 0.09 0.13 270,383
XRCC4 X21 rs1056503 T/G Coding-synon 0.09 0.12 276,697
* Base change listed as Major allele / Minor allele
† Position obtained from the University of California, Santa Cruz Genome Browser ; Flanking = within 10 kb of either side of gene; Locus region = variation in region of gene, but not in transcript; mRNA-utr = variation in transcript, but not in coding region interval
‡ MAF = minor allele frequency using our panel of 94 breast cancer case and control subjects
§Applied Biosystems reported minor allele frequency in Caucasians
|| Corrected value. Applied Biosystems acknowledged error in reported minor allele frequency of 0.49 on their web site, but it has not been updated.
** NBS1 and RAD50 were not genotyped in the current study. All analyses for these two genes were performed using the raw genotype data freely available online from Applied Biosystems. Base change obtained from University of California, Santa Cruz Genome Browser.
Table 2 Haplotypes with frequency>0.01, LD group characterization and tSNPs selected using Utah genotyping data*
a. ATM
A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 A11 A12 A13 A14 Freq
1 1 1 1 1 1 1 1 1 1 1 1 1 1 0.54
2 2 2 2 2 2 2 2 2 2 2 2 2 2 0.42
2 2 2 2 2 2 1 2 2 2 2 2 2 2 0.01
2 2 2 1 2 2 1 2 2 2 1 2 2 2 0.01
1 1 1 1 1 1 1 1 1 1 2 1 1 1 0.01
LD Group and tSNP Designation
1 1 1 1 1 1 1 1 1 1 1 1 1† 1
b. MRE11A
M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 Freq
2 2 2 1 2 2 1 1 1 1 1 0.30
1 1 1 2 1 1 1 2 1 1 1 0.28
1 1 1 1 1 1 1 1 1 1 1 0.25
1 1 1 2 1 1 1 1 1 1 2 0.09
1 1 1 1 1 1 1 2 1 1 1 0.06
1 1 1 2 1 1 2 1 2 2 1 0.01
LD Group and tSNP Designation
1 1 1 4† 1 1† 2 4 2 2† 3†
c. XRCC4
X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 Freq
2 2 2 2 2 2 2 2 1 2 2 2 2 1 1 1 1 1 1 1 1 0.35
1 1 1 1 1 1 1 1 2 1 1 1 1 1 2 1 1 1 1 1 1 0.19
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0.11
1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 2 2 2 1 1 1 0.10
1 2 2 2 2 2 2 2 1 2 2 2 2 1 1 1 1 1 1 1 1 0.05
1 1 1 1 1 1 1 1 2 1 1 1 1 1 2 1 1 1 2 2 2 0.03
1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 2 1 1 1 1 1 0.02
2 2 2 2 2 2 2 2 1 2 2 2 2 1 1 1 1 1 2 2 2 0.02
2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0.02
2 2 2 1 2 2 2 2 1 2 2 2 2 2 1 2 2 2 1 1 1 0.02
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 0.02
1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 2 2 1 1 1 0.01
1 1 1 1 1 1 2 1 1 1 1 1 2 1 1 2 2 1 1 1 0.01
LD Group and tSNP Designation
1 1† 1 1 1 1 1 1 4† 1 1 1 1 2† 4 2 2 2 3 3 3†
* Analysis considers the total panel of 94 individuals together
† tSNP selected / group
Table 3 Comparison of LD groups for the Utah breast cancer cases and controls with Applied Biosystems (ABI) data*
Gene Group Utah breast cancer case/control SNPs Utah potential tSNPs Utah % variance captured/group ABI SNPs ABI potential tSNPs ABI % variance captured/group
ATM 1 A1-A14 A1-A3, A5, A6, A8-A10, A12-A14 98.8% A1-A14 A1-A3, A5, A8, A13, A14 98.2%
MRE11 1 M1, M2, M3, M5, M6 M1, M2, M3, M5, M6 100% M1, M2, M3, M5, M6 M1, M2, M3, M5, M6 100%
2 M7, M9, M10 M10 84.3% M7, M9, M10 M7, M9, M10 100%
3 M11 M11 100% M11 M11 100%
4 M4, M8 M4, M8 82.2% M4, M8 M4, M8 83.9%
XRCC4 1 X1-X8, X10-X13 X2-X3, X5-X8, X10-X11 95.3% X1-X8, X10-X13 X2, X3, X10, X11, X13 96.0%
2 X14, X16-X18 X14 91.6% X14, X16-X18 X14, X17, X18 93.5%
3 X19-X21 X19-X21 100% X19-X21 X19-X21 100%
4 X9, X15 X9, X15 97.4% X9, X15 X9, X15 96.8%
*We used Applied Biosystems' validated SNP genotype data for 45 Caucasian subjects.
Table 4 Haplotypes with frequency>0.01, LD group characterization and tSNP selected using data from Applied Biosystems*
a. NBS1
N1 N2 N3† N4† N5† Frequency
1 1 1 1 1 0.55
2 2 2 2 2 0.26
1 2 2 2 2 0.10
1 2 2 1 2 0.03
2 2 1 1 2 0.03
1 2 1 1 2 0.03
LD Group and tSNP Designation
2‡ 1‡ 1 1 1
b. RAD50
R1† R2 R3† R4† R5 R6† R7 R8 R9 R10 Frequency
1 1 1 1 1 1 1 1 1 1 0.50
1 1 1 2 1 1 1 1 1 1 0.21
1 2 2 2 1 2 2 2 2 1 0.11
1 1 1 1 1 1 1 1 1 2 0.08
1 2 2 2 1 2 2 2 2 2 0.05
1 2 2 2 1 2 2 1 2 2 0.01
1 2 2 2 1 2 2 2 1 2 0.01
1 1 1 2 1 2 2 1 2 1 0.01
1 1 1 1 1 2 2 2 2 1 0.01
2 2 1 2 2 1 2 1 1 2 0.009§
LD Group and tSNP Designation
2‡ 1 1‡ 1 2 1 1 1 1 3‡
*We used Applied Biosystems' validated SNP genotype data for 45 Caucasian subjects.
† Allele designations have been changed from that listed by ABI to conform to the convention 1 = common allele, 2 = rare allele.
‡ tSNP selected / group
§ The haplotype with a frequency 0.009 was also analyzed to allow inclusion of rare variants at R1 and R5.
==== Refs
Daly MJ Rioux JD Schaffner SF Hudson TJ Lander ES High-resolution haplotype structure in the human genome Nat Genet 2001 29 229 232 11586305 10.1038/ng1001-229
Gabriel SB Schaffner SF Nguyen H Moore JM Roy J Blumenstiel B Higgins J DeFelice M Lochner A Faggart M Liu-Cordero SN Rotimi C Adeyemo A Cooper R Ward R Lander ES Daly MJ Altshuler D The structure of haplotype blocks in the human genome Science 2002 296 2225 2229 12029063 10.1126/science.1069424
Reich DE Cargill M Bolk S Ireland J Sabeti PC Richter DJ Lavery T Kouyoumjian R Farhadian SF Ward R Lander ES Linkage disequilibrium in the human genome Nature 2001 411 199 204 11346797 10.1038/35075590
Patil N Berno AJ Hinds DA Barrett WA Doshi JM Hacker CR Kautzer CR Lee DH Marjoribanks C McDonough DP Nguyen BT Norris MC Sheehan JB Shen N Stern D Stokowski RP Thomas DJ Trulson MO Vyas KR Frazer KA Fodor SP Cox DR Blocks of limited haplotype diversity revealed by high-resolution scanning of human chromosome 21 Science 2001 294 1719 1723 11721056 10.1126/science.1065573
Dawson E Abecasis GR Bumpstead S Chen Y Hunt S Beare DM Pabial J Dibling T Tinsley E Kirby S Carter D Papaspyridonos M Livingstone S Ganske R Lohmussaar E Zernant J Tonisson N Remm M Magi R Puurand T Vilo J Kurg A Rice K Deloukas P Mott R Metspalu A Bentley DR Cardon LR Dunham I A first-generation linkage disequilibrium map of human chromosome 22 Nature 2002 418 544 548 12110843 10.1038/nature00864
Phillips MS Lawrence R Sachidanandam R Morris AP Balding DJ Donaldson MA Studebaker JF Ankener WM Alfisi SV Kuo FS Camisa AL Pazorov V Scott KE Carey BJ Faith J Katari G Bhatti HA Cyr JM Derohannessian V Elosua C Forman AM Grecco NM Hock CR Kuebler JM Lathrop JA Mockler MA Nachtman EP Restine SL Varde SA Hozza MJ Gelfand CA Broxholme J Abecasis GR Boyce-Jacino MT Cardon LR Chromosome-wide distribution of haplotype blocks and the role of recombination hot spots Nat Genet 2003 33 382 387 12590262 10.1038/ng1100
Ke X Hunt S Tapper W Lawrence R Stavrides G Ghori J Whittaker P Collins A Morris AP Bentley D Cardon LR Deloukas P The impact of SNP density on fine-scale patterns of linkage disequilibrium Hum Mol Genet 2004 13 577 588 14734624 10.1093/hmg/ddh060
Johnson GC Esposito L Barratt BJ Smith AN Heward J Di Genova G Ueda H Cordell HJ Eaves IA Dudbridge F Twells RC Payne F Hughes W Nutland S Stevens H Carr P Tuomilehto-Wolf E Tuomilehto J Gough SC Clayton DG Todd JA Haplotype tagging for the identification of common disease genes Nat Genet 2001 29 233 237 11586306 10.1038/ng1001-233
Zhang K Deng M Chen T Waterman MS Sun F A dynamic programming algorithm for haplotype block partitioning Proc Natl Acad Sci U S A 2002 99 7335 7339 12032283 10.1073/pnas.102186799
Weale ME Depondt C Macdonald SJ Smith A Lai PS Shorvon SD Wood NW Goldstein DB Selection and evaluation of tagging SNPs in the neuronal-sodium-channel gene SCN1A: implications for linkage-disequilibrium gene mapping Am J Hum Genet 2003 73 551 565 12900796 10.1086/378098
Ke X Cardon LR Efficient selective screening of haplotype tag SNPs Bioinformatics 2003 19 287 288 12538253 10.1093/bioinformatics/19.2.287
Meng Z Zaykin DV Xu CF Wagner M Ehm MG Selection of genetic markers for association analyses, using linkage disequilibrium and haplotypes Am J Hum Genet 2003 73 115 130 12796855 10.1086/376561
Horne BD Camp NJ Principal component analysis for selection of optimal SNP-sets that capture intragenic genetic variation Genet Epidemiol 2004 26 11 21 14691953 10.1002/gepi.10292
Carlson CS Eberle MA Rieder MJ Yi Q Kruglyak L Nickerson DA Selecting a maximally informative set of single-nucleotide polymorphisms for association analyses using linkage disequilibrium Am J Hum Genet 2004 74 106 120 14681826 10.1086/381000
Camp NJ Swensen J Horne BD Farnham JM Thomas A Cannon-Albright LA Tavtigian SV Characterization of linkage disequilibrium structure, mutation history, and tagging SNPs, and their use in association analyses: ELAC2 and familial early-onset prostate cancer Genet Epidemiol 2004 28 232 243 10.1002/gepi.20054
Lin Z Altman RB Finding haplotype tagging SNPs by use of principal components analysis Am J Hum Genet 2004 75 850 861 15389393 10.1086/425587
Nyholt DR A simple correction for multiple testing for single-nucleotide polymorphisms in linkage disequilibrium with each other Am J Hum Genet 2004 74 765 769 14997420 10.1086/383251
Belmont J Yu F Hardenbol P Lu X Moorhead M Scott G Ghose S Pasternak S Willis T Faham M Leal SM Taylor J Morris R Kaplan N Gibbs RA High Density SNP Map Reveals Interrupted and Interlaced Organziation of Linkage Disequilibrium Among Markers: ; Toronto, Ontario, Canada. 2004
Hanahan D Weinberg RA The hallmarks of cancer Cell 2000 100 57 70 10647931 10.1016/S0092-8674(00)81683-9
Magnaldo T Sarasin A Xeroderma pigmentosum: from symptoms and genetics to gene-based skin therapy Cells Tissues Organs 2004 177 189 198 15388993 10.1159/000079993
Chung DC Rustgi AK The hereditary nonpolyposis colorectal cancer syndrome: genetics and clinical implications Ann Intern Med 2003 138 560 570 12667026
Khanna KK Jackson SP DNA double-strand breaks: signaling, repair and the cancer connection Nat Genet 2001 27 247 254 11242102 10.1038/85798
Skolnick M Bean L Dintelman SM Mineau G A computerized family history data base system Sociol Soc Res 1979 63 506
Skolnick M Skolnick M The Utah genealogical database: A resource for genetic epidemiology Banbury Report No 4: Cancer Incidence in Defined Populations 1980 Cold Spring Harbor, NY, Cold Spring Harbor Laboratory Press
Pedigree and Population Resource: Utah Population Database
Park BL Kim LH Shin HD Park YW Uhm WS Bae SC Association analyses of DNA methyltransferase-1 (DNMT1) polymorphisms with systemic lupus erythematosus J Hum Genet 2004 49 642 646 15375672 10.1007/s10038-004-0192-x
Fullerton SM Buchanan AV Sonpar VA Taylor SL Smith JD Carlson CS Salomaa V Stengard JH Boerwinkle E Clark AG Nickerson DA Weiss KM The effects of scale: variation in the APOA1/C3/A4/A5 gene cluster Hum Genet 2004 115 36 56 15108119 10.1007/s00439-004-1106-x
Setiawan VW Hankinson SE Colditz GA Hunter DJ De Vivo I HSD17B1 gene polymorphisms and risk of endometrial and breast cancer Cancer Epidemiol Biomarkers Prev 2004 13 213 219 14973105
Kosoy R Yokoi N Seino S Concannon P Polymorphic variation in the CBLB gene in human type 1 diabetes Genes Immun 2004 5 232 235 14961073 10.1038/sj.gene.6364057
Applied Biosystems
Lee LG Connell CR Bloch W Allelic discrimination by nick-translation PCR with fluorogenic probes Nucleic Acids Res 1993 21 3761 3766 8367293
Clayton DG SNPHAP
Morrell D Cromartie E Swift M Mortality and cancer incidence in 263 patients with ataxia-telangiectasia J Natl Cancer Inst 1986 77 89 92 3459930
Swift M Morrell D Massey RB Chase CL Incidence of cancer in 161 families affected by ataxia-telangiectasia N Engl J Med 1991 325 1831 1836 1961222
Morrell D Chase CL Swift M Cancers in 44 families with ataxia-telangiectasia Cancer Genet Cytogenet 1990 50 119 123 2253179 10.1016/0165-4608(90)90245-6
Broeks A Urbanus JH Floore AN Dahler EC Klijn JG Rutgers EJ Devilee P Russell NS van Leeuwen FE van 't Veer LJ ATM-heterozygous germline mutations contribute to breast cancer-susceptibility Am J Hum Genet 2000 66 494 500 10677309 10.1086/302746
Chenevix-Trench G Spurdle AB Gatei M Kelly H Marsh A Chen X Donn K Cummings M Nyholt D Jenkins MA Scott C Pupo GM Dork T Bendix R Kirk J Tucker K McCredie MR Hopper JL Sambrook J Mann GJ Khanna KK Dominant negative ATM mutations in breast cancer families J Natl Cancer Inst 2002 94 205 215 11830610
Thorstenson YR Roxas A Kroiss R Jenkins MA Yu KM Bachrich T Muhr D Wayne TL Chu G Davis RW Wagner TM Oefner PJ Contributions of ATM mutations to familial breast and ovarian cancer Cancer Res 2003 63 3325 3333 12810666
Izatt L Greenman J Hodgson S Ellis D Watts S Scott G Jacobs C Liebmann R Zvelebil MJ Mathew C Solomon E Identification of germline missense mutations and rare allelic variants in the ATM gene in early-onset breast cancer Genes Chromosomes Cancer 1999 26 286 294 10534763 10.1002/(SICI)1098-2264(199912)26:4<286::AID-GCC2>3.0.CO;2-X
FitzGerald MG Bean JM Hegde SR Unsal H MacDonald DJ Harkin DP Finkelstein DM Isselbacher KJ Haber DA Heterozygous ATM mutations do not contribute to early onset of breast cancer Nat Genet 1997 15 307 310 9054948 10.1038/ng0397-307
Olsen JH Hahnemann JM Borresen-Dale AL Brondum-Nielsen K Hammarstrom L Kleinerman R Kaariainen H Lonnqvist T Sankila R Seersholm N Tretli S Yuen J Boice JDJ Tucker M Cancer in patients with ataxia-telangiectasia and in their relatives in the nordic countries J Natl Cancer Inst 2001 93 121 127 11208881 10.1093/jnci/93.2.121
Li A Huang Y Swift M Neutral sequence variants and haplotypes at the 150 Kb ataxia-telangiectasia locus Am J Med Genet 1999 86 140 144 10449649 10.1002/(SICI)1096-8628(19990910)86:2<140::AID-AJMG10>3.0.CO;2-X
Bonnen PE Story MD Ashorn CL Buchholz TA Weil MM Nelson DL Haplotypes at ATM identify coding-sequence variation and indicate a region of extensive linkage disequilibrium Am J Hum Genet 2000 67 1437 1451 11078475 10.1086/316908
Thorstenson YR Shen P Tusher VG Wayne TL Davis RW Chu G Oefner PJ Global analysis of ATM polymorphism reveals significant functional constraint Am J Hum Genet 2001 69 396 412 11443540 10.1086/321296
Tamimi RM Hankinson SE Spiegelman D Kraft P Colditz GA Hunter DJ Common ataxia telangiectasia mutated haplotypes and risk of breast cancer: a nested case-control study Breast Cancer Res 2004 6 R416 22 15217510 10.1186/bcr809
Liu PY Zhang YY Lu Y Long JR Shen H Zhao LJ Xu FH Xiao P Xiong DH Liu YJ Recker RR Deng HW A survey of haplotype variants at several disease candidate genes: the importance of rare variants for complex diseases J Med Genet 2005 42 221 227 15744035 10.1136/jmg.2004.024752
Thompson D Stram D Goldgar D Witte JS Haplotype tagging single nucleotide polymorphisms and association studies Hum Hered 2003 56 48 55 14614238 10.1159/000073732
The International HapMap Project Nature 2003 426 789 796 14685227 10.1038/nature02168
HapMap
Iles MM The effect of SNP marker density on the efficacy of haplotype tagging SNPs--a warning Ann Hum Genet 2005 69 209 215 15720301 10.1046/j.1529-8817.2004.00141.x
|
16091150
|
PMC1208870
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Aug 9; 5:99
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-99
|
oa_comm
|
==== Front
BMC Cardiovasc DisordBMC Cardiovascular Disorders1471-2261BioMed Central London 1471-2261-5-241610918110.1186/1471-2261-5-24Study ProtocolPAIS: paracetamol (acetaminophen) in stroke; protocol for a randomized, double blind clinical trial. [ISCRTN 74418480] van Breda Eric J [email protected] der Worp H Bart [email protected] Gemert H Maarten A [email protected] Ale [email protected] L Jaap [email protected] Gijn Jan [email protected] Peter J [email protected] Diederik WJ [email protected] PAIS investigators 1 Erasmus Medical Center Rotterdam, the Netherlands2 University Medical Center Utrecht, the Netherlands3 Meander Medical Center, the Netherlands2005 19 8 2005 5 24 24 28 4 2005 19 8 2005 Copyright © 2005 van Breda et al; licensee BioMed Central Ltd.2005van Breda 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 patients with acute stroke, increased body temperature is associated with large lesion volumes, high case fatality, and poor functional outcome. A 1°C increase in body temperature may double the odds of poor outcome. Two randomized double-blind clinical trials in patients with acute ischemic stroke have shown that treatment with a daily dose of 6 g acetaminophen (paracetamol) results in a small but rapid and potentially worthwhile reduction of 0.3°C (95% CI: 0.1–0.5) in body temperature. We set out to test the hypothesis that early antipyretic therapy reduces the risk of death or dependency in patients with acute stroke, even if they are normothermic.
Methods/design
Paracetamol (Acetaminophen) In Stroke (PAIS) is a randomized, double-blind clinical trial, comparing high-dose acetaminophen with placebo in 2500 patients. Inclusion criteria are a clinical diagnosis of hemorrhagic or ischemic stroke and the possibility to start treatment within 12 hours from onset of symptoms. The study will have a power of 86% to detect an absolute difference of 6% in the risk of death or dependency at three months, and a power of 72% to detect an absolute difference of 5%, at a 5% significance level.
Discussion
This is a simple trial, with a drug that only has a small effect on body temperature in normothermic patients. However, when lowering body temperature with acetaminophen does have the expected effectiveness, 20 patients will have to be treated to prevent dependency or death in one.
==== Body
Background
In patients with acute stroke, increased body temperature is associated with high case fatality and poor functional outcome [1]. In the observational Copenhagen Stroke study, a 1°C increase in body temperature measured within 12 hours after stroke onset doubled the odds of poor outcome. The relation between body temperature and outcome was not affected by severity of symptoms at admission, lesion volume, or stroke type [1-6]. This suggests that even a small reduction in body temperature in acute stroke could improve outcome.
Phase II studies
We conducted two randomized, placebo-controlled clinical trials to study whether early treatment with acetaminophen reduces body temperature in patients with acute ischemic stroke confined to the carotid territory. In the first trial, seventy-six patients were randomized to daily treatment with either 3000 or 6000 mg acetaminophen, or placebo. In the high-dose group this resulted in a 0.4°C (95% CI: 0.1 to 0.7°C) lower body temperature than placebo treatment at 24 hours after onset [7]. The second trial was conducted to study the effect of ibuprofen and to confirm the effect of high-dose acetaminophen on body temperature. Seventy-five patients with acute ischemic stroke of the anterior circulation were randomized to daily treatment with either 2400 mg ibuprofen, 6000 mg acetaminophen, or placebo during 5 days [8]. Treatment with high-dose acetaminophen resulted in a 0.3°C reduction (95% CI: 0.0 to 0.6°C) in body temperature at 24 hours compared to treatment with placebo [9]. A pooled analysis of the data from both studies showed that a significant decrease of body temperature occurred within 4 hours after start of treatment with high-dose acetaminophen [10]. These studies suggest that high-dose acetaminophen can induce a small, but early reduction in body temperature in patients with acute stroke.
Pharmacological properties and toxicity of acetaminophen
Acetaminophen is a potent inhibitor of prostaglandin production within the central nervous system. This presumably accounts for its analgesic and antipyretic properties. It is rapidly absorbed through the gastrointestinal tract and uniformly distributed through most body fluids. After oral administration peak plasma levels are reached after 30 minutes to 1 hour [11]. Rectal administration may lead to a slightly slower absorption, but effective plasma levels are reached after approximately one hour [12]. Acetaminophen is mainly conjugated in the liver, and then excreted in the urine. Plasma half-life is 1.5 to 3 hours. Following a dose of 140 mg/kg bodyweight or more, the glucuronide pathway may become saturated, and (hepato-) toxicity may result. In the above-mentioned pilot trials, acetaminophen in a dose of 6 g/day was not associated with hepatotoxicity or other side effects [7,9].
Aim and purpose of the study
In guidelines for the treatment of acute ischemic or hemorrhagic stroke, antipyretic therapy is recommended in patients who develop fever [13,14]. This recommendation is based on expert opinion and observational studies but not on randomized clinical trials. Because observational studies suggest that every decrease in body temperature, regardless of the initial body temperature, is potentially benificial, antipyretic therapy may even improve outcome in patients without fever. The aim of PAIS is to test whether early antipyretic treatment with acetaminophen in a daily dose of 6000 mg for a period of 3 days will improve functional outcome in patients with acute stroke, even if they are normothermic.
Methods
Study design
PAIS is a multicenter, double-blind, placebo-controlled clinical trial that aims to include 2500 patients with acute stroke. The statistical analysis will be performed on an intention-to-treat basis [15]. Publication of the study results will be on behalf of the PAIS investigators.
Study medication
Patients will be treated for 3 days with acetaminophen in a daily dose of 6 times 1000 mg, or matching placebo. The study medication will be administered through identical tablets. The study medication will be provided in white paper boxes, numbered consecutively with a medication number. Each box contains 40 identical tablets of acetaminophen 500 mg, or placebo, and one suppository of acetaminophen 1000 mg, or placebo. Every gift of medication consists of 2 tablets except the first gift, which can be given as a suppository. The suppository can be administered at the discretion of the treating physician, for example when a patient has swallowing difficulties but has not yet been given a nasogastric tube.
Inclusion- and exclusion criteria
Inclusion criteria are a clinical diagnosis of ischemic stroke or primary intracerebral hemorrhage, the possibility to confirm the diagnosis with CT or MRI within 24 hours after inclusion in the study, the possibility to start treatment within 12 hours from onset of symptoms (for patients who noticed symptoms after waking from sleep, the time last seen well is taken as the time of onset of symptoms), age over 18 years, and signed informed consent.
Exclusion criteria are body temperature of less than 36.0°C or more than 39.0°C, a history of liver disease or elevated liver enzymes (ASAT, ALAT, AP, or gamma-GT) to more than twice the upper limit of normal, a history of alcohol abuse, hypersensitivity to acetaminophen, death appearing imminent at the time of inclusion, and any pre-stroke impairment that has led to dependency (modified Rankin scale (mRS) > 2) and therefore interferes with the assessment of functional outcome.
Randomization and treatment schedule
Patients will be assigned to treatment with acetaminophen 1000 mg, 6 times daily, or to matching placebo, for three days. Treatment assignment will be random. The study medication will be packed in separate boxes with a unique number. The study number is printed on an adhesive label, to be put on the inclusion form.
The study numbers correspond to those on a computer-generated list with the assigned treatment. An independent statistician, who is not involved in the study, will provide the list. The pharmacist of each center will receive a list that indicates the treatment allocation for each randomized patient.
Safety concerns and adverse events
Local investigators are advised to be careful not to overlook infections in patients who are treated according to the study protocol and might be treated with acetaminophen. Therefore, treating physicians are advised to lower their threshold for clinical suspicion of infection, and to start diagnostic studies and antibiotic treatment earlier than usual.
Serious adverse events will be reported by the local investigator to the PAIS trial office and include any infection that leads to prolonged hospital stay or is life threatening, death from any cause, liver failure, and gastro-intestinal hemorrhage that occurs during hospitalization and within the first 14 days after randomization.
Outcomes
The primary outcome measure will be functional outcome, as determined by the score on the modified Rankin scale (mRS) at 3 months [16]. Outcome will be dichotomized and defined as good (mRS 0 to 2) or poor (mRS 3 to 6). Other outcome measures will include an alternative dichotomization of the mRS (0 to 3 versus 4 to 6), the score on the Barthel index (BI) at 3 months [17], and body temperature at 24 hours from start of treatment. In addition, quality of life will be measured at three months with the EuroQol-5D [18,19].
Design
Time path
The study will run for four years. Patient inclusion will continue during the first 3.5 years. Three months will be needed to conclude the follow-up, and three months to run the final analysis and prepare the final report. In order to include 2500 patients in the study within the given timeframe, an annual recruitment rate of more than 700 patients will have to be realized.
Study activities
Day 0 is defined as the time period between onset of stroke and inclusion in the study. All baseline investigations are therefore carried out on day 0, except for the CT or MRI scans, which may be done within 24 hours after inclusion into the study. Day 1 commences directly after inclusion into the study. All time periods in the study are measured relative to the time of the start of treatment.
Baseline data
At baseline the medical history, including previous strokes or TIAs, will be assessed and a general and neurological examination will be carried out. The National Institutes of Health stroke scale (NIHSS) is used to assess stroke severity at inclusion in the study [20]. Laboratory investigations will include a full blood count, glucose, electrolytes, creatinine, and liver enzymes: ALAT, ASAT, AP, and gamma-GT. A brain CT or MRI will be done within 24 hours after inclusion in the study. Body temperature (tympanic or rectal) will be measured at inclusion, and 24 hours later. In each individual patient, the mode of thermometry at 24 hours will be similar to that at baseline.
After completing the one-page inclusion form, the local investigator will send it by fax to the trial office. The data will be automatically added to the study database by means of optical character recognition soft- and hardware (Teleform, Verity Inc. Sunnyvale, U.S.A.). All data entries will be verified by a study assistant. The trial coordinator will compare the data provided by the local investigators with those in the source documents in a random sample of at least 10% of the patients.
Day 14 or discharge
After 14 days or at discharge, if earlier, the discharge destination, the number of remaining tablets and suppositories, and functional status (BI) will be assessed. Based on the neurological examination and the results of CT, ECG, duplex, and other studies, etiological stroke type will be assessed according to the TOAST classification [21]. Serious adverse events that occurred during 14 days after start of treatment will be recorded.
Three month follow-up
Functional outcome at three months (mRS, BI) and quality of life (EuroQol-5D) will be assessed by telephone interview with the patients themselves or their caregivers. The telephone interview will be conducted by the central trial office.
Data analysis
The data will be analyzed on an intention-to-treat basis. In the analysis the occurrence of the primary outcome (mRS>2) will be compared between the two treatment arms by computing the relative risk, expressed as a risk ratio with a 95% confidence interval. In order to increase the power of the study, adjustments will be made with multiple logistic regression for any imbalance in the following prognostic variables: time since onset, baseline temperature, stroke severity, stroke type (hemorrhagic versus ischemic), ischemic stroke subtype (lacunar versus non-lacunar), and thrombolytic therapy, as suggested by Hernandez [22].
The treatment effect will be evaluated in specific subgroups, i.e. in patients treated early (i.e. within 9 hours from onset), in patients with ischemic stroke (as opposed to hemorrhagic stroke), and in patients with non-lacunar ischemic stroke.
Interim analysis of safety and effectiveness
During the period of recruitment into the study, every year interim analyses of in-hospital mortality and of any other information that is available on major outcome events including serious adverse events believed to be due to treatment will be supplied, in strict confidence, to the chairman of the data monitoring committee, along with any other analyses that the committee may request. In the light of these analyses, the data monitoring committee will advise the chairman of the steering committee if, in their view, the randomized comparisons in PAIS have provided both: 1) proof beyond reasonable doubt that for all, or for some, specific types of patients, one particular treatment is clearly indicated or clearly contra-indicated in terms of a net difference in mortality, and 2) evidence that might reasonably be expected to influence materially the patient management of the many clinicians who are already aware of the results of other main trials. The steering committee can then decide whether to modify intake to the study (or to seek extra data). Unless this happens, however, the steering committee, the collaborators, and the central administrative staff (except those who produce the confidential analyses) will remain ignorant of the interim analyses. Appropriate criteria of "proof beyond reasonable doubt" cannot be specified precisely, but some members of the committee have expressed sympathy with the view that a difference of at least 3 standard deviations in an interim analysis of a major outcome event may be needed to justify halting, or modifying such a study prematurely. If this criterion were to be adopted, it would have the practical advantage that the exact number of interim analyses would be of little importance, and so no fixed schedule is proposed.
Collaborators, and all others associated with the study, may write through the PAIS trial-office in Rotterdam to the chairman of the data monitoring committee, drawing attention to any worries that they may have about the possibility of particular side-effects, or about particular categories of patients requiring special consideration, or about any other matters that may be relevant [23].
Sample size
In the Copenhagen Stroke Study, the odds ratio of poor outcome increased by a factor of 2.2 (95% CI: 1.4 to 3.5), with each degree Celsius increase in body temperature [5,24]. We expect that 50% of the patients assigned to placebo will have a poor outcome at three months [7,9,10]. A relative odds reduction with a factor of 2.2 per degree Celsius of body temperature would imply an absolute reduction in poor outcome of 18.75%. A 0.3°C reduction in body temperature would then theoretically lead to a 6% absolute or 12% relative reduction of poor outcome. Two thousand five hundred patients (1250 in each arm) will provide a power of 86% to detect at least such an effect, and a power of 72% to detect an absolute reduction of 5%.
Discussion
Several studies have demonstrated a strong relationship between an increased body temperature in the first hours after stroke and poor functional outcome [3,5,24-26]. To date, it is still unclear whether this relationship is causal. Some authors suggest that increased body temperatures after stroke are just an epiphenomenon of extensive cerebral damage and, thereby, of poor outcome. In a recent study of 725 consecutive patients admitted within 6 hours from the onset of acute ischemic stroke, no relation was found between initial body temperature and outcome [27]. The authors used a rather insensitive method of statistical analysis (Spearman correlation and comparison of median modified Rankin scale scores with a nonparametric test). All patients with a body temperature of 37.0°C were treated with acetaminophen. Consequently, the results of this study should be interpreted with care.
In the Copenhagen Stroke study, the relationship between body temperature and outcome remained present after adjustment for initial stroke severity [5]. Furthermore, early body temperature measurements (within the first 6 hours) seemed to be more strongly related to outcome than later measurements [24]. This suggests that the relationship with poor outcome is not confounded by the occurrence of secondary infections, such as pneumonia or urinary tract infection, because these usually appear later in the course of the disease. Arguments for a causal relationship stem from the observation that the effect of body temperature on outcome is independent of the size of the brain lesion and from the beneficial effect of temperature-lowering treatment on infarct volume in animal models. Animal studies have demonstrated that higher body temperatures may worsen ischemic damage through an increase of blood-brain-barrier permeability and increased metabolic demands, resulting in acidosis and higher levels of deleterious excitatory amino-acids [28]. In a recent meta-analysis of controlled animal studies on the effect of hypo- and hyperthermia in focal cerebral ischemia, Miyazawa showed that hyperthermia increases infarct volume whereas hypothermia reduces infarct volume [29]. These reproducible observations from observational clinical studies and animal experiments strongly suggest that hypothermia may be a potent neuroprotective intervention, but this has never been studied in adequately powered clinical trials.
Randomized clinical trials of hypothermia in brain injury and hypoxic brain damage have provided conflicting results. In patients with coma after closed head injury, treatment with hypothermia, with body temperature reaching 33°C within eight hours after injury, was not effective in improving outcome [30]. In patients who had been successfully resuscitated after cardiac arrest, hypothermia with body temperatures between 32°C and 34°C increased the chances of favorable outcome and reduced mortality [31,32]. Perhaps the lack of effect in traumatic brain injury can be explained by the abundant presence of direct, i.e. non-ischemic damage.
Other methods of body temperature reduction
The feasibility of different methods of reducing body temperature in patients with acute stroke has been studied in several pilot studies.
A case-control study was conducted in 74 patients (17 cases, 56 controls) to assess the feasibility and safety of reducing body temperature to approximately 35.5°C with cooling blankets in combination with pethidine to prevent shivering [33].
Another study on the feasibility of several methods of lowering body temperature was conducted in eight patients. Two patients were treated with 1 gram acetaminophen at 4-hour intervals. Two patients were cooled with cooling blankets, in two patients sponging with 70% alcohol was applied, and 2 patients served as a control group and were only monitored. Target temperature reductions of 1°C were reached within 6 hours [34].
A study in 50 patients with a severe middle cerebral artery infarct demonstrated that body temperature can be reduced to 32 to 33°C, with the use of cooling blankets, alcohol and ice bags, under complete anesthesia [35]. The procedure had many side effects; the most frequent complications were thrombocytopenia (70%), bradycardia (62%), and pneumonia (48%).
The many complications that occurred in the study of Schwab can be explained by the fact that these patients all had severe middle cerebral artery infarcts. In the remaining studies safety concerns were less prominent.
The COOL-AID study was a randomized controlled study of endovascular cooling to 33°C compared to standard medical treatment in 40 patients with acute ischemic stroke. Shivering was suppressed by warming blankets and sedatives [36].
Most studies showed that cooling using the different methods was feasible, although it was more labor intense in the more invasive methods that induced larger body temperature decreases. However, in the COOL-AID study the feasibility was poor: 5 of the 18 treated patients did not reach the target temperature.
Mild reductions in body temperature – around 1°C – can be reached with external cooling blankets and general measures, within six hours, or more rapidly with an endovascular temperature management system. This approach requires mild sedation or morfine to reduce shivering, may be uncomfortable to patients, and is labor intensive. More aggressive approaches require anesthesia, and may induce increased risk of pulmonary- and other complications. This suggests that there is a need for a simple, medical intervention that may reduce body temperature to a lesser extent, but is cheap and safe.
Ethical considerations
Observational studies have shown an association between an increased body temperature and a poor outcome. Several national and international guidelines suggest that raised body temperature in stroke patients should be treated [13]. However, the efficacy of temperature-lowering treatment to improve functional outcome has not yet been demonstrated in randomized trials [37]. This dilemma is the rationale for PAIS.
PAIS: a simple trial
PAIS is a large, randomized, multi-center clinical trial. To keep the threshold for including patients low, the study design has been kept simple. The amount of data to be gathered is therefore limited; it consists of two one-page forms, one to be filled out at inclusion and one to be filled out at discharge. The local investigator will keep a log of randomized patients. The 3-month follow-up will be conducted by telephone from the central trial office. New centers are welcome to join the trial.
Conclusion
The PAIS trial will test the efficacy of a modest temperature reduction by high-dose acetaminophen to improve functional outcome after stroke. The treatment strategy tested is cheap and safe. When lowering body temperature with acetaminophen will have the expected effectiveness, 20 patients will have to be treated to prevent functional dependence or death in one
Abbreviations
ASAT – aspartate aminotransferase
ALAT – alanine aminotransferase
AP – alkaline phosphatase
BI – Barthel Index
GT – glutamyl transpeptidase
CRP – C-reactive protein
CT – computed tomography
mRS – modified Rankin Scale
MRI – magnetic resonance imaging
NIHSS – National Institutes of Health Stroke Scale
PAIS – Paracetamol (Acetaminophen) In Stroke
TOAST – Trial of ORG 10172 in Acute Stroke Treatment
Competing interests
The author(s) declare that they have no competing interests.
Financial support
This study is financially supported by the Netherlands Heart Foundation; Grant no 2002 B148.
Authors' contributions
All authors contributed equally to this article.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Azzimondi G Bassein L Nonino F Fiorani L Vignatelli L Re G D'Alessandro R Fever in acute stroke worsens prognosis. A prospective study Stroke 1995 26 2040 2043 7482646
Castillo J Martinez F Leira R Prieto JM Lema M Noya M Mortality and morbidity of acute cerebral infarction related to temperature and basal analytic parameters Cerebrovasc Dis 1994 4 66 71
Castillo J Davalos A Marrugat J Noya M Timing for fever-related brain damage in acute ischemic stroke Stroke 1998 29 2455 2460 9836750
Fukuda H Kitani M Takahashi K Body temperature correlates with functional outcome and the lesion size of cerebral infarction Acta Neurol Scand 1999 100 385 390 10589798
Reith J Jorgensen HS Pedersen PM Nakayama H Raaschou HO Jeppesen LL Olsen TS Body temperature in acute stroke: relation to stroke severity, infarct size, mortality, and outcome Lancet 1996 347 422 425 8618482 10.1016/S0140-6736(96)90008-2
Wang Y Lim LL Levi C Heller RF Fisher J Influence of admission body temperature on stroke mortality Stroke 2000 31 404 409 10657413
Dippel DWJ van Breda EJ Van Gemert HMA van der Worp HB Meijer RJ Kappelle LJ Koudstaal PJ Effect of paracetamol (acetaminophen) on body temperature in acute ischemic stroke: a double-blind, randomized phase II clinical trial Stroke 2001 32 1607 1612 11441208
van Breda EJ van der Worp HB Van Gemert HMA Meijer RJ Kappelle LJ Koudstaal PJ Dippel DWJ PISA. The effect of paracetamol (acetaminophen) and ibuprofen on body temperature in acute stroke: protocol for a phase II double-blind randomised placebo-controlled trial BMC Cardiovascular Disorders 2002 2 7 11918829 10.1186/1471-2261-2-7
Dippel DWJ van Breda EJ van der Worp HB van Gemert HM Meijer RJ Kappelle LJ Koudstaal PJ Effect of paracetamol (acetaminophen) and ibuprofen on body temperature in acute ischemic stroke PISA, a phase II double-blind, randomized, placebo-controlled trial [ISRCTN98608690] BMC Cardiovascular Disorders 2003 3 2 12657165 10.1186/1471-2261-3-2
Dippel DWJ van Breda EJ van der Worp HB van Gemert HM Kappelle LJ Algra A Koudstaal PJ Timing of the effect of acetaminophen on body temperature in patients with acute ischemic stroke Neurology 2003 61 677 679 12963761
Clissold SP Paracetamol and phenacetin Drugs 1986 32 46 59 3552585
Blume H Ali SL Elze M Kramer J Scholz ME [The relative bioavailability of paracetamol in suppositories in comparison to tablets] Arzneimittelforschung 1996 46 975 980 8992964
Adams HP JrAdams RJ Brott T del Zoppo GJ Furlan A Goldstein LB Grubb RL Higashida R Kidwell C Kwiatkowski TG Marler JR Hademenos GJ Guidelines for the early management of patients with ischemic stroke: A scientific statement from the Stroke Council of the American Stroke Association Stroke 2003 34 1056 1083 12677087 10.1161/01.STR.0000064841.47697.22
Broderick JP Adams HP JrBarsan W Feinberg W Feldmann E Grotta J Kase C Krieger D Mayberg M Tilley B Zabramski JM Zuccarello M Guidelines for the management of spontaneous intracerebral hemorrhage: A statement for healthcare professionals from a special writing group of the Stroke Council, American Heart Association Stroke 1999 30 905 915 10187901
Moher D Schulz KF Altman D The CONSORT statement: revised recommendations for improving the quality of reports of parallel-group randomized trials JAMA 2001 285 1987 1991 11308435 10.1001/jama.285.15.1987
van Swieten JC Koudstaal PJ Visser MC Schouten HJ van Gijn J Interobserver agreement for the assessment of handicap in stroke patients Stroke 1988 19 604 607 3363593
Mahoney FI Barthel DW Functional evaluation: the Barthel index MD State Med J 1965 14 61 65 14258950
The EuroQol Group EuroQol: a new facility for the measurement of health related quality of life Health Policy 1990 16 199 208 10109801 10.1016/0168-8510(90)90421-9
Dolan P Modeling valuations for EuroQol health states Medical Care 1997 35 1095 1108 9366889 10.1097/00005650-199711000-00002
Brott T Adams HP JrOlinger CP Marler JR Barsan WG Biller J Spilker J Holleran R Eberle R Hertzberg V Measurements of acute cerebral infarction: a clinical examination scale Stroke 1989 20 864 870 2749846
Adams HPJ Bendixen BH Kappelle LJ Biller J Love BB Gordon DL Marsh EE Classification of subtype of acute ischemic stroke. Definitions for use in a multicenter clinical trial. TOAST. Trial of Org 10172 in Acute Stroke Treatment Stroke 1993 24 35 41 7678184
Hernandez AV Steyerberg EW Habbema JD Covariate adjustment in randomized controlled trials with dichotomous outcomes increases statistical power and reduces sample size requirements J Clin Epidemiol 2004 57 454 460 15196615 10.1016/j.jclinepi.2003.09.014
International Stroke Trial Collaborative Group The International Stroke Trial (IST): a randomised trial of aspirin, subcutaneous heparin, both, or neither among 19435 patients with acute ischaemic stroke Lancet 1997 349 1569 1581 9174558 10.1016/S0140-6736(97)04011-7
Jorgensen HS Reith J Pedersen PM Nakayama H Olsen TS Body temperature and outcome in stroke patients Lancet 1996 348 193 8684167 10.1016/S0140-6736(05)66135-1
Hajat C Hajat S Sharma P Effects of poststroke pyrexia on stroke outcome : a meta-analysis of studies in patients Stroke 2000 31 410 414 10657414
Schwarz S Hafner K Aschoff A Schwab S Incidence and prognostic significance of fever following intracerebral hemorrhage Neurology 2000 54 354 361 10668696
Boysen G Christensen H Stroke Severity Determines Body Temperature in Acute Stroke Stroke 2001 32 413 417 11157175
Schaller B Graf R Hypothermia and stroke: the pathophysiological background Pathophysiology 2003 10 7 35 14643901 10.1016/j.pathophys.2003.09.001
Miyazawa T Tamura A Fukui S Hossmann KA Effect of mild hypothermia on focal cerebral ischemia. Review of experimental studies Neurol Res 2003 25 457 464 12866192 10.1179/016164103101201850
Clifton GL Miller ER Choi SC Levin HS McCauley S Smith KR Muizelaar JP Wagner FC Marion DW Luerssen TG Chesnut RM Schwartz M Lack of effect of induction of hypothermia after acute brain injury N Engl J Med 2001 344 556 563 11207351 10.1056/NEJM200102223440803
Bernard SA Gray TW Buist MD Jones BM Silvester W Gutteridge G Smith K Treatment of comatose survivors of out-of-hospital cardiac arrest with induced hypothermia N Engl J Med 2002 346 557 563 11856794 10.1056/NEJMoa003289
The Hypothermia after Cardiac Arrest Study Group Mild therapeutic hypothermia to improve the neurologic outcome after cardiac arrest N Engl J Med 2002 346 549 556 11856793 10.1056/NEJMoa012689
Kammersgaard LP Rasmussen BH Jorgensen HS Reith J Weber U Olsen TS Feasibility and safety of inducing modest hypothermia in awake patients with acute stroke through surface cooling: A case-control study: the Copenhagen Stroke Study Stroke 2000 31 2251 2256 10978060
Meijer RJ Visser H Koudstaal PJ Dippel DWJ Lowering body temperature in acute ischemic stroke without artificial ventilation and heavy sedation: a feasibility study J Stroke Cerebrovasc Dis 2001 10 157 160 10.1053/jscd.2001.26869
Schwab S Georgiadis D Berrouschot J Schellinger PD Graffagnino C Mayer SA Feasibility and safety of moderate hypothermia after massive hemispheric infarction Stroke 2001 32 2033 2035 11546893
De Georgia MA Krieger DW Abou-Chebl A Devlin TG Jauss M Davis SM Koroshetz WJ Rordorf G Warach S Cooling for Acute Ischemic Brain Damage (COOL AID): a feasibility trial of endovascular cooling Neurology 2004 63 312 317 15277626
Hacke W Kaste M Bogousslavsky J Brainin M Chamorro A Lees K Leys D Kwiecinski H Toni P Langhorne P Diener C Hennerici M Ferro J Sivenius J Gunnar N Bath P Olsen TS Gugging M European Stroke Initiative Recommendations for Stroke Management-update 2003 Cerebrovasc Dis 2003 16 311 337 14584488 10.1159/000072554
|
16109181
|
PMC1208871
|
CC BY
|
2021-01-04 16:30:07
|
no
|
BMC Cardiovasc Disord. 2005 Aug 19; 5:24
|
utf-8
|
BMC Cardiovasc Disord
| 2,005 |
10.1186/1471-2261-5-24
|
oa_comm
|
==== Front
BMC Cell BiolBMC Cell Biology1471-2121BioMed Central London 1471-2121-6-301609554110.1186/1471-2121-6-30Research ArticleInhibition of dynamin-dependent endocytosis increases shedding of the amyloid precursor protein ectodomain and reduces generation of amyloid β protein Carey Robyn M [email protected] Brigitte A [email protected] Ignacio [email protected] Barbara E [email protected] Department of Pathology and Laboratory Medicine, Boston University School of Medicine, 715 Albany Street, Rm. L808, Boston MA 02118, USA2 Gemeinnützige Salzburger Landeskliniken Betriebsgesellschaft mbH, Universitätsklinik für Innere Medizin III, Paracelsus Medizinische Privatuniversität, Müllner Hauptstrasse 48, A-5020 Salzburg, Austria2005 11 8 2005 6 30 30 18 3 2005 11 8 2005 Copyright © 2005 Carey et al; licensee BioMed Central Ltd.2005Carey 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 amyloid precursor protein (APP) is transported via the secretory pathway to the cell surface, where it may be cleaved within its ectodomain by α-secretase, or internalized within clathrin-coated vesicles. An alternative proteolytic pathway occurs within the endocytic compartment, where the sequential action of β- and γ-secretases generates the amyloid β protein (Aβ). In this study, we investigated the effects of modulators of endocytosis on APP processing.
Results
Human embryonic kidney cells were transfected with a dominant negative mutant of dynamin I, an important mediator of clathrin-dependent endocytosis, and APP proteolysis was analyzed. Overexpression of the mutant dynamin (dyn I K44A) resulted in increased shedding of the APP ectodomain (sAPPα), accumulation of the C-terminal α-secretase product C83, and a reduction in the release of Aβ. Levels of mature APP on the cell surface were increased in cells expressing dyn I K44A, and internalization of surface-immunolabeled APP, assessed by fluorescence microscopy, was inhibited. Dynamin is a substrate for protein kinase C (PKC), and it was hypothesized that activators of PKC, which are known to stimulate α-secretase-mediated cleavage of APP, might exert their effects by inhibiting dynamin-dependent endocytosis. However, the internalization of surface-biotinylated APP was unaffected by treatment of cells with phorbol 12-myristate 13-acetate in the presence of the α-secretase inhibitor TAPI-1.
Conclusion
The results indicate that APP is internalized by a dynamin-dependent process, and suggest that alterations in the activity of proteins that mediate endocytosis might lead to significant changes in Aβ production.
==== Body
Background
The amyloid precursor protein (APP) is a single-pass transmembrane protein that gives rise to the small peptides (known as Aβ) that form amyloid deposits in the brains of patients with Alzheimer's disease (AD) [1,2]. Aβ peptides are generated by the successive cleavage of APP by proteases known respectively as β- and γ-secretases. Alternatively, APP may be cleaved within the Aβ domain by α-secretases, now believed to be members of the disintegrin and metalloprotease (ADAM) family [3-5]. This latter process precludes the formation of Aβ, and results in the shedding of a large soluble N-terminal fragment of APP (sAPPα) into the extracellular or intra-luminal space. Cleavage of APP by α-secretases may occur in a late compartment of the secretory pathway, or at the cell surface [6].
APP ectodomain shedding occurs in both a constitutive and a regulated fashion. A key mediator of regulated shedding is protein kinase C (PKC), whether it is stimulated directly by phorbol esters, or as a consequence of the activation of receptors coupled to phosphoinositide turnover. Although the stimulation of APP shedding by PKC activators has been extensively documented [7], the mechanism is still unclear. Direct phosphorylation of the APP intracellular domain is not required, since phorbol esters are still able to increase shedding of a C-terminally truncated form of APP, or of APP constructs in which serine or threonine residues in the cytoplasmic domain have been replaced with alanine [8-10]. Likewise, C-terminal truncation of the putative α-secretase ADAM17/TACE (tumor necrosis factor-α converting enzyme) did not prevent up-regulation of its activity toward its substrate tumor necrosis factor-α (TNFα) by phorbol 12-myristate 13-acetate (PMA) [11]. On the other hand, phorbol ester-regulated cleavage of TrkA by TACE was found to be dependent, in part, on phosphorylation of threonine 735 within the TACE cytoplasmic domain [12]. Thus, phosphorylation of ADAM proteases may modulate their activity, at least toward certain substrates.
PKC-mediated effects on vesicular trafficking might also affect APP processing. A study showing that PKC activation increases the formation of APP-containing secretory vesicles from the trans-Golgi network [13], suggested that accelerated trafficking of APP to the cell surface might underlie the increase in sAPPα release induced by PKC. Alternatively, inhibition of endocytosis could increase sAPPα release by prolonging the interaction of APP with secretases on the cell surface. APP is found within clathrin-coated vesicles [14,15], which mediate the internalization of many cell surface proteins. Clathrin-dependent endocytosis is regulated by the high-molecular weight GTPase dynamin, which forms oligomeric rings around the neck of the forming vesicle, and severs it from the plasma membrane [16]. Dynamin activity, in turn, is reportedly governed by PKC [17-19], raising the possibility that PKC might modulate internalization, and therefore secretory cleavage, of APP, via an effect on endocytosis.
The aims of the present study were two-fold: to examine the effects of an inhibitor of dynamin function on APP processing, and to determine if PKC activation stimulates APP shedding via inhibition of endocytosis. Overexpression of a dominant negative dynamin mutant in HEK cells co-transfected with APP695 increased surface expression of APP and release of sAPPα, while inhibiting the internalization of full-length APP. The dynamin mutant also increased formation of C83, the C-terminal stub generated by α-secretase-mediated cleavage of APP, and reduced the release of Aβ peptides. These results contrast with a recent study, in which induction of dominant negative dynamin (dyn I K44A) increased both sAPPα release and Aβ formation [20]. Although activation of PKC by treatment with the phorbol ester PMA stimulates shedding of the APP ectodomain, PMA had no effect on the internalization of surface-biotinylated APP. Our observations provide direct evidence that APP internalization is a dynamin-dependent process. Moreover, the results indicate that activators of PKC do not promote sAPPα release via inhibition of endocytosis.
Results
Ectodomain shedding of APP is increased in cells transfected with dyn I K44A
The GTPase dynamin is an important mediator of clathrin-dependent endocytosis and synaptic vesicle recycling, and is required for the internalization of many cell surface proteins, including growth factor and G-protein coupled receptors [21]. To determine if dynamin regulates the internalization of APP, HEK-M3 cells were transiently transfected with APP695 and either an empty vector or a plasmid encoding the dominant-negative dynamin mutant dyn I K44A, which is deficient in GTP binding and GTPase activity [22]. The APP695 isoform was used for these transfection studies, since it is not expressed by HEK cells; and can be distinguished on western blots from the longer endogenous APP isoforms. Cells transfected with empty vectors alone were used as additional controls. Levels of sAPPα in the medium, and cellular full-length APP, were detected by western blotting using 6E10 antibodies, and antibodies to the APP C-terminus (APP-CT), respectively. The expression of endogenous dynamin, assessed using an antibody that recognizes both dynamin I and dynamin II isoforms, was low in cells transfected with APP695 or empty vector, and robust overexpression of the mutant protein was observed in cells transfected with the plasmid encoding dyn I K44A (Fig. 1A, lower panel). Dynamin I is neuron-specific, whereas dynamin II is widely expressed. The immunoreactive band present in cells that were not transfected with the dynamin plasmid presumably represents dynamin II, since no signal was detectable when lysates from these cells were immunoblotted with antibodies specific for dynamin I (not shown). An increase of approximately five-fold in the release of endogenous and transfected sAPPα from cells transfected with dyn I K44A was observed (Fig. 1A, upper panel, and Fig. 1B). Levels of cellular full-length APP were also increased in cells expressing the dynamin mutant (Fig. 1A, middle panel). Quantitation of the band comprising mature cellular APP695 and immature endogenous APP (APPendo) showed that levels in the presence of the mutant dynamin increased to 2.86 ± 0.97-fold control levels (mean ± SEM, n = 3). Overnight treatment of APP695 transfectants with the lysosomal protease inhibitor chloroquine (50 μM) increased levels of this band to a similar extent (to 2.30 ± 0.02-fold control, mean ± SEM from 3 experiments).
Figure 1 Release of sAPPα is increased by dyn I K44A overexpression. HEK-M3 cells were transiently transfected with APP695 and dyn I K44A or empty vector. After 48 hours, the growth medium was replaced with serum-free medium, and collected after 2 hours. Proteins in media extracts and cell lysates were size-fractionated on SDS gels and analyzed by immunoblotting. A. Expression of dyn I K44A increased release of endogenous and co-transfected sAPPα (upper panel; bands were detected with 6E10 antibodies) and levels of cellular APP (middle panel; bands were detected with anti-APP-CT). Arrows indicate mature (m) and immature (im) isoforms of endogenous and transfected APP. The mutant dynamin was expressed at high levels in transfected cells (lower panel). The lanes depicted in each panel were derived from the same blot. B. Levels of sAPPα in the media extracts were quantitated by densitometry and values were expressed as means ± SEM from 3 experiments. *, significantly different from the other two groups, by analysis of variance and Fisher's Least Significant Difference test.
Inhibition of dynamin function increases surface expression of APP
The effect of the dynamin mutant on surface expression of APP was next examined by surface biotinylation of transiently transfected HEK-M3 cells. In cells transfected with empty vector alone, one diffuse band of biotinylated APP representing mature endogenous APP was visible on western blots (Fig. 2A, lanes 1 and 2). Cells transfected with APP695 expressed an additional, more rapidly migrating band, representing mature cell surface APP695 (Fig. 2A, lanes 3 and 4). Surface expression of both endogenous and transfected APP695 was greatly increased in cells overexpressing dyn I K44A (Fig. 2A, lanes 5 and 6). As shown in Fig. 1, both mature isoforms of APP were released into the medium of APP695 transfectants, and release of both isoforms was markedly increased in cells overexpressing dyn I K44A.
Figure 2 Surface expression of APP is increased by dyn I K44A overexpression. A. HEK-M3 cells were transiently transfected with APP695 and dyn I K44A or empty vector. After 48 hours, cells were surface biotinylated, lysed, and incubated with streptavidin-agarose beads. Biotinylated proteins were immunoblotted with antibodies to the APP C-terminal. Full-length, biotinylated, endogenous and co-transfected APP are indicated by arrows. B. HEK-695 cells were transiently transfected with dyn I K44A or empty vector. Immunoblot analysis of cell lysates showed increased formation of the APP C-terminal fragment (C83) generated by α-secretase cleavage in cells transfected with the dynamin mutant. C. Results from B were quantitated, normalized and expressed as means ± SEM from 3 experiments. *, p < 0.05 by paired t-test.
Expression of dyn I K44A increases formation of the APP C-terminal fragment C83
Cleavage of APP by α-secretase results in the release of the soluble ectodomain fragment sAPPα, and leaves a C-terminal stub, known as C83, in the cell membrane. In HEK cells stably overexpressing APP695 (HEK-695 cells) a protein corresponding in size to C83 was detected by western blotting of cell lysates with antibodies to the APP C-terminus (Fig. 2B). The corresponding β-secretase product C99 was not detectable under these conditions. Levels of C83 were significantly increased in cells transfected with dyn I K44A, relative to levels in cells transfected with empty vector alone (Fig. 2B and 2C), consistent with the increase in ectodomain shedding observed in cells expressing the dynamin mutant.
Internalization of APP is inhibited in cells transfected with dyn I K44A
Our results can be interpreted to suggest that overexpression of dyn I K44A inhibits endocytosis of APP. The increase in cellular levels of APP observed in cells expressing the mutant dynamin, accordingly, could be the result of decreased internalization and degradation of full-length APP (Fig. 1A), and indeed, the lysosomal protease inhibitor chloroquine exerted a similar effect. However, it could also be argued that the increases in APP surface expression and shedding caused by the dynamin mutant were secondary to elevations in APP expression. Therefore, in order to directly examine the effect of dynamin on APP endocytosis, APP internalization in living cells was assessed using an immuno-labeling assay. Live HEK cells stably overexpressing APP695 were incubated at 4°C with 6E10 antibodies, in order to label cell surface APP. The cells were washed and warmed to 37°C for various intervals, and then fixed, permeabilized, and stained with Alexa 488-conjugated secondary antibodies. In cells that were labeled on ice and then fixed and stained prior to warming, APP was largely confined to the plasma membrane (Fig. 3A, 0 min). After 10 minutes at 37°C, most of the immunofluorescence was contained within intracellular punctuate structures distributed throughout the cytoplasm, indicating that surface APP had moved into an endosomal compartment. By 60 minutes, the immunofluorescent signal representing internalized APP had for the most part coalesced at a perinuclear site (Fig. 3A). Only background fluorescence was observed in cells that were incubated with antibodies to paxillin, an intracellular protein (not shown). APP internalization was next determined in these cells 48 hours after the cultures were transiently transfected with dyn I K44A. The cells were surface-labeled with 6E10 antibodies, and incubated at 37°C for 10 minutes. They were then fixed, permeabilized, and stained with Alexa 488-conjugated anti-mouse IgG to detect APP. Dynamin I expression was assessed by co-staining cells with goat anti-dynamin I antibodies followed by Alexa 594-conjugated anti-goat IgG. Immunoblot analysis of cell lysates confirmed that this antibody, which is specific for dynamin I, does not detect endogenous dynamin in non-transfected cells (not shown). In cells expressing the mutant dynamin I (Fig. 3B, right panel), APP immunoreactivity was restricted to the cell surface (Fig. 3B, left panel), whereas untransfected cells within the same culture exhibited a punctate pattern of APP immunofluorescence indicative of internalization (Fig. 3B, arrows). These results provide direct evidence that dyn I K44A inhibits endocytosis of APP.
Figure 3 APP internalization is inhibited in cells transfected with dyn I K44A. A. HEK-695 cells were surface-immunolabeled with 6E10 antibodies for 45 min while on ice. The cells were then incubated at 37°C for varying periods of time, to allow internalization to occur. Prior to warming (0 min), APP immunofluorescence was confined to the cell membrane. Within 5 minutes APP immunoreactivity was located within punctate structures near the cell membrane. By 60 minutes, most of the immunoreactivity had coalesced at a perinuclear site. B. HEK-695 cells were transiently transfected with dyn I K44A. After 48 hours, cells were surface-labeled with 6E10 antibodies, then incubated at 37° for 10 min. In untransfected cells (arrows) APP (green) was internalized within intracellular vesicles within 10 minutes. In dyn I K44A-transfected cells (red), APP immunoreactivity was still largely membrane-associated at this time-point, indicating that internalization was impaired in cells expressing the dynamin mutant. Bar, 10 μm.
Overexpression of dyn I K44A inhibits the formation of Aβ peptides
In order to determine if dynamin inhibition affects the formation of Aβ peptides, levels of Aβ1–40 in the medium of HEK-695 cells transiently transfected with either an empty vector or dyn I K44A were measured by ELISA. Overexpression of the dynamin I mutant caused a significant reduction in Aβ release (Fig. 4), suggesting that APP internalization is necessary for generation of Aβ in HEK cells.
Figure 4 Dyn I K44A inhibits Aβ formation. HEK-695 cells were transiently transfected with empty vector or dyn I K44A. Medium was collected for 24 hours and analyzed for Aβ1–40 levels by sandwich ELISA. Levels of Aβ1–40 were significantly lower in the medium of cells transfected with dyn I K44A. *, p < 0.05 by paired t-test.
Activation of PKC does not affect internalization of APP
Activation of PKC by administration of phorbol esters, or via stimulation of receptors coupled to PKC, increases APP ectodomain shedding, but the mechanism remains unclear [7]. It is known that inhibiting APP endocytosis via mutation of internalization motifs, or truncation of the cytoplasmic domain, also increases ectodomain shedding [23-25], raising the possibility that physiological mechanisms that regulate shedding might act by targeting the endocytic machinery. To address this question, the effect of the PKC activator PMA on APP internalization in HEK cells stably overexpressing APP695 was determined using reversible surface biotinylation. Shedding was inhibited by incubating cells with the α-secretase inhibitor TAPI-1 for 1 hour prior to biotinylation. Cells were surface-biotinylated while on ice, then incubated at 37°C in DMEM containing TAPI-1 and either PMA (1 μM) or the vehicle dimethylsulfoxide (DMSO) for varying periods of time. Media and cell lysates were collected and processed as described (see "Methods"). As a control for stripping efficiency, some cultures were biotinylated and stripped while remaining on ice (Fig. 5A). Despite the presence of TAPI-1, there was a slight, but detectable, time-dependent increase of biotinylated sAPPα695 in the medium. This was not affected by PMA, indicating that TAPI-1 effectively blocked the signal-dependent release of APP (Fig. 5B, upper panel). Levels of internalized biotinylated APP declined over the 60 minute incubation period, consistent with the degradation of internalized APP, or the removal of the biotin label in an endocytic compartment. APP internalization was not significantly altered by the presence of PMA (Fig. 5B, middle panel, and 5C). In the absence of TAPI-1, PMA caused a marked increase in secretion of endogenous and transfected sAPPα from HEK-695 cells, as expected (Fig. 5D).
Figure 5 Activation of PKC does not affect APP internalization. HEK-695 cells were pre-treated with TAPI-1 in serum free DMEM for 1 hour prior to biotinylation. After biotinylation and quenching, cells were incubated at 37°C for various time periods in the presence of TAPI-1 and either DMSO or PMA (1 μM). The medium was collected and biotinylated sAPPα was isolated and analyzed by immunoblot using 6E10 antibodies. The cells were stripped and lysed, and biotinylated APP was isolated and assessed by immunoblot analysis with antibodies to the APP C-terminal. A. Internalization of biotinylated APP was nearly absent in cells that were biotinylated and stripped (B/S) while still on ice. NB, non-biotinylated; B, biotinylated, not stripped. B. PMA, in the presence of TAPI-1, did not affect release of biotinylated sAPPα(upper panel), or internalization of full-length biotinylated APP (middle panel). C. Bands depicting biotinylated and internalized APP were quantitated by densitometry, normalized, and expressed as means ± SEM from 3 experiments. D. In the absence of TAPI-1, PMA caused a marked increase in release of endogenous and transfected sAPPα from HEK-695 cells.
Discussion
Cleavage of APP within the Aβ domain by α-secretases is of great physiological interest, not only because it precludes the formation of Aβ, but also because it generates a soluble N-terminal fragment, sAPPα, that exhibits neuroprotective properties [26,27]. Moreover, shedding of the ectodomain is a prerequisite for cleavage of the intracellular domain by γ-secretases; a process that liberates a C-terminal fragment with transcriptional activity [28-30]. Although the up-regulation of APP shedding by activation of PKC-dependent signaling pathways has been well-documented [7], the mechanism mediating this response is still obscure.
The present study was undertaken to determine if inhibitors of dynamin function would affect ectodomain shedding of APP. We first showed that APP internalization is dependent on the activity of dynamin, a large molecular weight GTPase that mediates both clathrin-dependent endocytosis, and internalization of caveolae, by promoting the separation of endocytic vesicles from the plasma membrane [22,31]. In confirmation of a recent study [20], we found that overexpression of a dominant negative dynamin mutant protein in HEK cells increased surface expression of full-length APP, and release of sAPPα. Thus, although cleavage of APP by α-secretases occurs largely in an intracellular compartment in many cell types (reviewed in [6]), our results suggest that inhibition of dynamin function, by preventing internalization of APP, increases its dwell-time on the cell surface, and prolongs its interaction with α-secretases at the plasma membrane. Similar elevations in APP secretion are induced by mutations of the APP cytoplasmic domain that inhibit internalization [23-25]. Consistent with the observed increase in α-secretase mediated cleavage, expression of the dynamin mutant increased cellular levels of C83, the C-terminal stub remaining after α-secretase-mediated cleavage of APP (Fig. 2B and 2C).
The increase in sAPPα release in HEK cells overexpressing dyn I K44A was associated with a reduction in the release of Aβ1–40, (Fig. 4), a result in keeping with reports that Aβ is generated in an endocytic compartment [24,25,32-34]. Our results are also in agreement with a study by Ehehalt et al. [35] who found that overexpression of a dyn K44A mutant protein reduced formation of the Aβ peptide in mouse neuroblastoma N2a cells. In contrast, Chyung and Selkoe reported that Aβ generation was increased in HeLa cells following induction of dyn K44A expression [20]. The increased Aβ formation observed in the latter study occurred in the absence of any alteration in the synthesis or maturation of APP, and suggested that, in HeLa cells, processing of APP by β- and γ-secretases occurs at the plasma membrane [20]. Indeed, an active γ-secretase complex was subsequently isolated from the plasma membrane of HeLa cells [36]. As a possible explanation for the reduction in Aβ observed by Ehehalt et al. [35] in cells overexpressing dyn K44A, Chyung and Selkoe pointed out that those workers measured formation of radiolabeled Aβ in cells labeled for 1 hour with [35S]methionine, and surmised that the mutant dynamin reduced generation of labeled Aβ by increasing the amount of unlabeled APP at the cell surface, and diluting the concentration of labeled precursor available for cleavage by β- and γ-secretases. In support of the notion that Aβ can be generated at the cell surface, Ehehalt et al [35] showed that when endocytosis was blocked by transfection with dyn K44A, the reduction in Aβ could be partially rescued by antibody cross-linking of APP and the β-secretase, β-site APP-cleaving enzyme (BACE). The decrease in total Aβ1–40 generation in HEK cells overexpressing dyn I K44A described in the present report might simply reflect reductions in the precursor pool due to increased cleavage of APP by α-secretase. This result is consistent with earlier studies showing that upregulation of α-secretase cleavage by PKC activation in HEK cells [37], or via mutations of the APP cytoplasmic domain in stably transfected HEK or Chinese hamster ovary (CHO) cells [23-25], is associated with decreased Aβ formation. The discrepancies among these studies might be due at least in part to cell-specific differences in the compartments where APP comes into contact with α- and β/γ-secretases, or in the relative capacities of the different secretases to cleave APP within a specific compartment.
Modulation of endocytosis might represent a mechanism for physiological regulation of APP processing by PKC-dependent signaling pathways. PKC phosphorylates dynamin, thereby activating its GTPase activity [17], and inhibiting its association with phospholipids in vitro [18]. In nerve terminals, dynamin must be dephosphorylated in order to promote retrieval of synaptic vesicles following exocytosis, and re-phosphorylation is required for the next round of endocytosis that follows a second stimulus [19]. Persistent phosphorylation of dynamin might therefore be predicted to interfere with endocytosis. Contrary to expectation, the PKC activator PMA did not affect the rate of APP internalization, as determined by reversible biotinylation in the presence of the α-secretase inhibitor TAPI-1 (Fig. 5). Thus, although PKC activation can modulate endocytosis of a variety of transmembrane proteins, either positively, in the case of β1 integrin, GABA receptors, and the dopamine transporter [38-41], or negatively, as is the case with μ-opioid receptors [42], we could not find evidence for a modulatory effect of phorbol esters on APP internalization. Others have shown that PKC activation increases APP ectodomain shedding in PC12 cells by stimulating trafficking of APP through the secretory pathway [13]. In contrast, surface expression of APP was reduced in CHO cells that were surface biotinylated following treatment with PMA and TAPI, suggesting that in these cells, PMA did not increase trafficking of APP to the plasma membrane, but possibly stimulated α-secretase-mediated cleavage within an intracellular compartment that was partially resistant to TAPI [43]. Interestingly, the motor neuron-derived trophic factor neuregulin-1, a ligand for the tyrosine kinase receptors ErbB3 and ErbB4, was found to increase the rate of internalization and degradation of APP in cultured myotubes, while decreasing release of the ectodomain [44]. This report lends credence to the hypothesis that modulation of APP internalization may represent a physiological mechanism for regulation of sAPPα release.
Conclusion
Our results show that experimental manipulations that interfere with the function of the endocytic machinery can inhibit APP internalization, and shift APP proteolysis to a non-amyloidogenic pathway, in HEK cells. In HeLa cells, in contrast, an interfering dynamin mutant increased both α-secretase cleavage of APP and Aβ formation [20], suggesting that cell-specific differences in APP metabolism may influence the consequences of altered endocytosis. The levels of a number of proteins important for clathrin-mediated recycling of synaptic vesicles, including dynamin, and the clathrin assembly-mediating adapter proteins AP2 and AP180, are reduced in the brains of AD patients [45]. Moreover, exposure of neurons to Aβ in vitro was recently reported to reduce dynamin levels [46]. It is therefore possible that alterations in clathrin-mediated endocytosis play a role in the abnormal metabolism of APP that is characteristic of AD. Finally, given the putative role of APP as a cell surface signaling molecule in the brain [47], it is important to consider the possibility that alterations in APP endocytosis may contribute to the pathologic process by disrupting the normal signaling function of APP.
Methods
Materials
Antibodies and other reagents were obtained from the following sources: 6E10 antibodies to sAPPα from Signet Laboratories (Dedham, MA), antibodies to the C-terminus of APP (APP-CT) from Zymed Labs (San Francisco, CA), anti-dynamin monoclonal antibodies from BD Biosciences (San Diego, CA), goat polyclonal antibodies specific for dynamin I from Santa Cruz Biotechnology (Santa Cruz, CA), and goat anti-mouse IgG and goat anti-rabbit IgG peroxidase-conjugated secondary antibodies from BioRad (Hercules CA). Immunofluorescence-conjugated secondary antibodies including Alexa Fluor 488-conjugated goat or donkey anti-mouse IgG, and Alexa Fluor 594-conjugated rabbit anti-goat IgG, and ProLong Anti-fade mounting medium were obtained from Molecular Probes (Eugene, OR). Mini-gels and reagents for electrophoresis were obtained from BioRad (Hercules CA), and polyvinylidene difluoride (PVDF) membranes were purchased from Perkin-Elmer (Boston, MA). The metalloproteinase inhibitor, tumor necrosis factor-α protease inhibitor (TAPI-1), was obtained from Peptides International (Louisville, KY). 2-Mercaptoethanesulfonic acid sodium salt, iodoacetamide, and phorbol 12-myristate 13-acetate (PMA) were obtained from Sigma-Aldrich (St. Louis MO). Sulfo-NHS-SS-Biotin was purchased from Pierce (Rockford, IL), Other reagents and materials were acquired from Fisher Scientific (Pittsburgh PA).
Cell culture
HEK-M3 cells (HEK cells stably transfected with M3 muscarinic receptors) and HEK-695 cells (HEK cells stably overexpressing APP695; a gift from Dr. Dennis Selkoe) were grown in Dulbecco's Modified Eagle Medium (DMEM)/F-12 supplemented with 10% Fetal Bovine Serum (Invitrogen Life Technologies, Carlsbad, CA) and maintained at 37°C in an atmosphere of 95% air, 5% CO2. HEK-M3 cells were used in some of these studies because the regulation of constitutive and receptor-coupled sAPPα release has been well characterized in this line [5,48,49].
Transient transfections
Cells were transiently transfected with plasmids encoding APP695 (a gift from Dr. Carmela Abraham) and dyn I K44A (a gift from Dr. Marc Caron), or with an empty pcDNA3 vector, using Lipofectamine Plus™ reagent (Invitrogen Life Technologies, Carlsbad CA) according to the manufacturer's specifications. Experiments were carried out 48 hours later.
Cell surface biotinylation
Confluent HEK cells were pre-incubated in serum-free DMEM for 2 hours, then washed in phosphate buffered saline (PBS), pH 7.9, supplemented with 1 mM Ca++ and 2 mM Mg++. Surface biotinylation was carried out by incubating the cells for 30 min on ice with Sulfo-NHS-SS-Biotin (0.5 mg/ml in PBS). Culture dishes were kept on ice in the dark and gently rocked during the incubation period. The biotin reagent was quenched by treating the cells with two 15 min washes of 50 mM glycine in PBS. Cells were rinsed again with PBS and lysed in a buffer containing 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 2 mM 4-(2-aminoethyl)benzenesulfonyl fluoride, 1 μg/ml leupeptin, 1% (v/v) Nonidet P-40, 0.05% (w/v) sodium dodecyl sulfate, 0.5% (w/v) deoxycholate. Lysates were incubated overnight with streptavidin-coated agarose beads (Pierce, Rockford, IL) at 4°C in a rotary mixer to isolate biotin-labeled proteins. Isolates were size-fractionated on SDS gels, and analyzed for APP content by immunoblotting.
Reversible biotinylation
HEK cells were pre-treated for 1 hour at 37°C in serum-free DMEM containing TAPI-1 (50 μM), then surface-biotinylated as described above. Cells were then incubated at 37° for various time periods in the presence of TAPI-1 and either PMA (1 μM) or DMSO (vehicle control). The cells were placed on ice and the remaining surface biotin was removed by applying two 20 minute washes of a stripping buffer (50 mM 2-mercaptoethanesulfonic acid (sodium salt); 150 mM NaCl; 1 mM EDTA and 0.2% BSA in 20 mM Tris, pH 8.6). The 2-mercaptoethanesulfonic acid was quenched with a buffer containing iodoacetamide (50 mM iodoacetamide, 1% BSA in PBS, pH 7.4) for 30 minutes, and cells were rinsed with PBS. To assess the efficiency of the stripping procedure, some cultures were biotinylated and then stripped while remaining on ice.
Western blotting
The protein content of cell lysates was measured using the bicinchoninic acid reagent (Sigma, St Louis MO). Medium was collected, cleared by centrifugation, desalted, lyophilized, and resuspended in SDS-PAGE loading buffer, as previously described [5]. Lysates were centrifuged to remove insoluble material, and diluted in 2X loading buffer. Samples were normalized for protein content and size-fractionated on 7.5% or 10–20% Tris-HCl mini-gels. Proteins were transferred to PVDF membranes, which were then blocked in 5%-powdered milk in Tris-buffered saline with 0.15% Tween-20 for 2 hours, and probed overnight with primary antibodies. The next day, membranes were washed, and incubated with goat anti-mouse IgG or goat anti-rabbit IgG peroxidase-conjugated secondary antibodies and bands were detected using an enhanced chemiluminescence reagent (Western Lightning, Pierce). Membranes were imaged on a Kodak 440CF Image Station and quantitated using Kodak 1D Image Analysis software.
Immunofluorescence microscopy
Cells were plated on nitric acid-washed coverslips coated with poly-D-lysine and placed in 30-mm tissue culture dishes. After 48 hours in growth medium, live cells were washed with PBS, and incubated on ice with 6E10 antibodies (at a dilution of 1:200 in PBS) to label surface APP. Cells were then transferred to an incubator and maintained at 37°C for various time periods. Cells were fixed in 3.0% paraformaldehyde in PBS for 10 minutes at room temperature, permeabilized in 0.1% Triton X-100, and blocked in 1% bovine serum albumin in PBS. APP was detected by incubating the cells with goat anti-mouse antibodies conjugated with Alexa Fluor 488. When double-labeling of APP and dynamin was required, cell preparations were incubated with goat anti-dynamin I primary antibodies (1:400), washed, then incubated with donkey Alexa Fluor 488-conjugated anti-mouse IgG, and rabbit Alexa Fluor 594-conjugated anti-goat IgG (1:200). After washing in PBS, cells were mounted with ProLong Anti-Fade mounting medium and left overnight to dry. Specimens were examined using a conventional fluorescence microscope equipped with appropriate band-pass filters, and images were captured with a Spot RT-KE camera (Diagnostics Instruments, Sterling Heights MI).
Aβ measurement
HEK-695 cells were plated and transiently transfected with dyn I K44A or with empty vector, as described above, and allowed to grow for 48 hours. The growth medium was removed, and the cells were rinsed with serum-free DMEM. Fresh DMEM was then placed on the cells and they were incubated overnight. The next day, the medium was collected and a 1 ml aliquot was analyzed by enzyme-linked immunosorbent assay (ELISA) using a kit from Signet Laboratories (Dedham MA). Standards and samples were prepared and incubated in the plate overnight at 4°C. The ELISA was performed the next day according to the manufacturer's instructions.
Authors' contributions
RMC carried out the immunoblotting, biotinylation and immunofluorescence studies, and helped to draft the manuscript. BAB participated in the transfection and immunoblotting experiments. IL-C contributed to the design and execution of the immunofluorescence studies. BES conceived of the study, participated in its design, and assisted in drafting and editing the manuscript. All authors read and approved the final manuscript.
Acknowledgements
We thank Dr. Dennis Selkoe for the gift of APP695-transfected HEK cells, Dr. Marc Caron for the dynamin K44A construct, and Dr. Carmela Abraham for the APP695 plasmid. This work was supported by NIH grants NS30791 and MH59775 (to BES). RC was supported in part by a training grant (NIH-AG00115).
==== Refs
Selkoe DJ Alzheimer's disease: Genes, proteins, and therapy Physiol Rev 2001 81 741 766 11274343
Suh YH Checler F Amyloid precursor protein, presenilins, and α-synuclein: Molecular pathogenesis and pharmacological applications in Alzheimer's disease Pharmacol Rev 2002 54 469 525 12223532 10.1124/pr.54.3.469
Buxbaum JD Liu K-N Luo Y Slack JL Stocking KL Peschon JJ Johnson RS Castner BJ Cerretti DP Black RA Evidence that tumor necrosis factor α converting enzyme is involved in regulated α-secretase cleavage of the Alzheimer protein precursor J Biol Chem 1998 273 27765 27767 9774383 10.1074/jbc.273.43.27765
Lammich S Kojro E Postina R Gilbert S Pfeiffer R Jasionowski M Haass C Fahrenholz F Constitutive and regulated α-secretase cleavage of Alzheimer's amyloid precursor protein by a disintegrin metalloprotease Proc Natl Acad Sci USA 1999 96 3922 3927 10097139 10.1073/pnas.96.7.3922
Slack BE Ma LK Seah CC Constitutive shedding of the amyloid precursor protein ectodomain is up-regulated by tumour necrosis factor-α converting enzyme Biochem J 2001 357 787 794 11463349 10.1042/0264-6021:3570787
Checler F Processing of the β-amyloid precursor protein and its regulation in Alzheimer's disease J Neurochem 1995 65 1431 1444 7561836
Mills J Reiner PB Regulation of amyloid precursor protein cleavage J Neurochem 1999 72 443 460 9930716 10.1046/j.1471-4159.1999.0720443.x
Da Cruz e Silva OAB Iverfeldt K Oltersdorf T Sinha S Lieberburg I Ramabhadran TV Suzuki T Sisodia SS Gandy S Greengard P Regulated cleavage of Alzheimer β-amyloid precursor protein in the absence of the cytoplasmic tail Neuroscience 1993 57 873 877 8309547 10.1016/0306-4522(93)90031-A
Hung AY Selkoe DJ Selective ectodomain phosphorylation and regulated cleavage of β-amyloid precursor protein EMBO J 1994 13 534 542 8313898
Efthimiopoulos S Felsenstein KM Sambamurti K Robakis NK Refolo LM Study of the phorbol ester effect on Alzheimer amyloid precursor processing: Sequence requirements and involvement of a cholera toxin sensitive protein J Neurosci Res 1994 38 81 90 8057394 10.1002/jnr.490380111
Reddy P Slack JL Davis R Cerretti DP Kozlosky CJ Blanton RA Shows D Peschon JJ Black RA Functional analysis of the domain structure of tumor necrosis factor-α converting enzyme J Biol Chem 2000 275 14608 14614 10799547 10.1074/jbc.275.19.14608
Diaz-Rodriguez E Montera JC Esparis-Ogando A Yuste L Pandiella A Extracellular signal-regulated kinase phosphorylates tumor necrosis factor α-converting enzyme at threonine 735: a potential role in regulated shedding Mol Biol Cell 2002 13 2031 2044 12058067 10.1091/mbc.01-11-0561
Xu HX Greengard P Gandy S Regulated formation of Golgi secretory vesicles containing Alzheimer β-amyloid precursor protein J Biol Chem 1995 270 23243 23245 7559474 10.1074/jbc.270.40.23243
Nordstedt C Caporaso GL Thyberg J Gandy SE Greengard P Identification of the Alzheimer β/A4 amyloid precursor protein in clathrin-coated vesicles purified from PC12 cells J Biol Chem 1993 268 608 612 8416966
Marquez-Sterling NR Lo ACY Sisodia SS Koo EH Trafficking of cell-surface β-amyloid precursor protein: Evidence that a sorting intermediate participates in synaptic vesicle recycling J Neurosci 1997 17 140 151 8987743
Mousavi SA Malerod L Berg T Kjeken R Clathrin-dependent endocytosis Biochem J 2004 377 1 16 14505490 10.1042/BJ20031000
Robinson PJ Sontag J-M Liu J-P Fykse EM Slaughter C McMahon H Südhof TC Dynamin GTPase regulated by protein kinase C phosphorylation in nerve terminals Nature 1993 365 163 166 8371759 10.1038/365163a0
Powell KA Valova VA Malladi CS Jensen ON Larsen MR Robinson PJ Phosphorylation of dynamin I on Ser-795 by protein kinase C blocks its association with phospholipids J Biol Chem 2000 275 11610 11617 10766777 10.1074/jbc.275.16.11610
Cousin MA Tan TC Robinson PJ Protein phosphorylation is required for endocytosis in nerve terminals: potential role for the dephosphins dynamin I and synaptojanin, but not AP180 or amphiphysin J Neurochem 2001 76 105 116 11145983 10.1046/j.1471-4159.2001.00049.x
Chyung JH Selkoe DJ Inhibition of receptor-mediated endocytosis demonstrates generation of amyloid β-protein at the cell surface J Biol Chem 2003 278 51035 51043 14525989 10.1074/jbc.M304989200
Schmid SL McNiven MA De Camilli P Dynamin and its partners: a progress report Curr Opin Cell Biol 1998 10 504 512 9719872 10.1016/S0955-0674(98)80066-5
Damke H Baba T Warnock DE Schmid SL Induction of mutant dynamin specifically blocks endocytic coated vesicle formation J Cell Biol 1994 127 915 934 7962076 10.1083/jcb.127.4.915
Haass C Hung AY Schlossmacher MG Teplow DB Selkoe DJ β-amyloid peptide and a 3-kDa fragment are derived by distinct cellular mechanisms J Biol Chem 1993 268 3021 3024 8428976
Koo EH Squazzo SL Evidence that production and release of amyloid β-protein involves the endocytic pathway J Biol Chem 1994 269 17386 17389 8021238
Perez RG Soriano S Hayes JD Ostaszewski B Xia W Selkoe DJ Chen X Stokin GB Koo EH Mutagenesis identifies new signals for β-amyloid precursor protein endocytosis, turnover, and the generation of secreted fragments, including Aβ42 J Biol Chem 1999 274 18851 18856 10383380 10.1074/jbc.274.27.18851
De Strooper B Annaert W Proteolytic processing and cell biological functions of the amyloid precursor protein J Cell Sci 2000 113 1857 1870 10806097
Mattson MP Cellular actions of β-amyloid precursor protein and its soluble and fibrillogenic derivatives Physiol Rev 1997 77 1081 1132 9354812
Cao X Südhof TC A transcriptively active complex of APP with Fe65 and histone acetyltransferase Tip60 Science 2001 293 115 120 11441186 10.1126/science.1058783
Gao Y Pimplikar SW The gamma-secretase-cleaved C-terminal fragment of amyloid precursor protein mediates signaling to the nucleus Proc Natl Acad Sci USA 2001 98 14979 14984 11742091 10.1073/pnas.261463298
Leissring MA Murphy MP Mead TR Akbari Y Sugarman MC Jannatipour M Anliker B Müller U Saftig P De Strooper B Wolfe MS Golde TE LaFerla FM A physiologic signaling role for the gamma-secretase-derived intracellular fragment of APP Proc Natl Acad Sci USA 2002 99 4697 4702 11917117 10.1073/pnas.072033799
Henley JR Krueger EWA Oswald BJ McNiven MA Dynamin-mediated internalization of caveolae J Cell Biol 1998 141 85 99 9531550 10.1083/jcb.141.1.85
Soriano S Chyung ASC Chen X Stokin GB Lee VMY Koo EH Expression of β-amyloid precursor protein-CD3gamma chimeras to demonstrate the selective generation of amyloid β1–40 and amyloid β1–42 peptides within secretory and endocytic compartments J Biol Chem 1999 274 32295 32300 10542269 10.1074/jbc.274.45.32295
Daugherty BL Green SA Endosomal sorting of amyloid precursor protein-P-selectin chimeras influences secretase processing Traffic 2001 2 908 916 11737828 10.1034/j.1600-0854.2001.21206.x
Grbovic OM Mathews PM Jiang Y Schmidt SD Dinakar R Summers-Terio NB Ceresa BP Nixon RA Cataldo AM Rab5-stimulated up-regulation of the endocytic pathway increases intracellular β-cleaved amyloid precursor protein carboxyl-terminal fragment levels and Aβ production J Biol Chem 2003 278 31261 31268 12761223 10.1074/jbc.M304122200
Ehehalt R Keller P Haass C Thiele C Simons K Amyloidogenic processing of the Alzheimer β-amyloid precursor protein depends on lipid rafts J Cell Biol 2003 160 113 123 12515826 10.1083/jcb.200207113
Chyung JH Raper DM Selkoe DJ γ-Secretase exists on the plasma membrane as an intact complex that accepts substrates and effects intramembrane cleavage J Biol Chem 2005 280 4383 4392 15569674 10.1074/jbc.M409272200
Hung AY Haass C Nitsch RM Qiu WQ Citron M Wurtman RJ Growdon JH Selkoe DJ Activation of protein kinase C inhibits cellular production of the amyloid β-protein J Biol Chem 1993 268 22959 22962 8226807
Daniels GM Amara SG Regulated trafficking of the human dopamine transporter. Clathrin-mediated internalization and lysosomal degradation in response to phorbol esters J Biol Chem 1999 274 35794 35801 10585462 10.1074/jbc.274.50.35794
Ng T Shima D Squire A Bastiaens PIH Gschmeissner S Humphries MJ Parker PJ PKCα regulates β1 integrin-dependent cell motility through association and control of integrin traffic EMBO J 1999 18 3909 3923 10406796 10.1093/emboj/18.14.3909
Filippova N Sedelnikova A Zong Y Fortinberry H Weiss DS Regulation of recombinant gamma-aminobutyric acid (GABA)(A) and GABA(C) receptors by protein kinase C Mol Pharmacol 2000 57 847 856 10779366
Loder MK Melikian HE The dopamine transporter constitutively internalizes and recycles in a protein kinase C-regulated manner in stably transfected PC12 cell lines J Biol Chem 2003 278 22168 22174 12682063 10.1074/jbc.M301845200
Ueda H Inoue M Matsumoto T Protein kinase C-mediated inhibition of μ-opioid recetor internalization and its involvement in the development of acute tolerance to peripheral μ-agonist analgesia J Neurosci 2001 21 2967 2973 11312280
Jolly-Tornetta C Wolf BA Protein kinase C regulation of intracellular and cell surface amyloid precursor protein (APP) cleavage in CHO695 cells Biochemistry 2000 39 15282 15290 11106509 10.1021/bi001723y
Rosen KM Ford BD Querfurth HW Downregulation and increased turnover of β-amyloid precursor protein in skeletal muscle cultures by neuregulin-1 Exp Neurol 2003 181 170 180 12781990 10.1016/S0014-4886(03)00031-1
Yao PJ Synaptic frailty and clathrin-mediated synaptic vesicle trafficking in Alzheimer's disease Trends Neurosci 2004 27 24 29 14698607 10.1016/j.tins.2003.10.012
Kelly BL Vassar R Ferreira A Beta-amyloid-induced dynamin 1 depletion in hippocampal neurons: A potential mechanism for early cognitive decline in Alzheimer's disease J Biol Chem 2005 JBC Papers in Press published on July 15, 2005 as doi:10.1074/jbc.M503259200. 16002400
Neve RL McPhie DL Chen Y Alzheimer's disease: a dysfunction of the amyloid precursor protein Brain Res 2000 886 54 66 11119687 10.1016/S0006-8993(00)02869-9
Nitsch RM Slack BE Wurtman RJ Growdon JH Release of Alzheimer amyloid precursor derivatives stimulated by activation of muscarinic acetylcholine receptors Science 1992 258 304 307 1411529
Petryniak MA Wurtman RJ Slack BE Elevated intracellular calcium concentration increases secretory processing of the amyloid precursor protein by a tyrosine phosphorylation-dependent mechanism Biochem J 1996 320 957 963 9003386
|
16095541
|
PMC1208872
|
CC BY
|
2021-01-04 16:31:30
|
no
|
BMC Cell Biol. 2005 Aug 11; 6:30
|
utf-8
|
BMC Cell Biol
| 2,005 |
10.1186/1471-2121-6-30
|
oa_comm
|
==== Front
BMC Dev BiolBMC Developmental Biology1471-213XBioMed Central London 1471-213X-5-171610916910.1186/1471-213X-5-17Research ArticleInactivation of the Huntington's disease gene (Hdh) impairs anterior streak formation and early patterning of the mouse embryo Woda Juliana M [email protected] Teresa [email protected] Paige [email protected] Mabel P [email protected] Ronald A [email protected] Marcy E [email protected] Molecular Neurogenetics Unit, Center for Human Genetic Research, Massachusetts General Hospital, CNY-149, 13th Street, Charlestown MA 02129, USA2 Department of Genetics, Case Western Reserve University, 10900 Euclid Avenue, Cleveland, OH 44106, USA3 University of Queensland, Faculty of Health Sciences, St Lucia QLD 4072, Australia4 Department of Pathology, Harvard Medical School, 77 Avenue Louis Pasteur, NRB-850A, Boston MA 02115, USA2005 18 8 2005 5 17 17 21 5 2005 18 8 2005 Copyright © 2005 Woda et al; licensee BioMed Central Ltd.2005Woda 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
Huntingtin, the HD gene encoded protein mutated by polyglutamine expansion in Huntington's disease, is required in extraembryonic tissues for proper gastrulation, implicating its activities in nutrition or patterning of the developing embryo. To test these possibilities, we have used whole mount in situ hybridization to examine embryonic patterning and morphogenesis in homozygous Hdhex4/5 huntingtin deficient embryos.
Results
In the absence of huntingtin, expression of nutritive genes appears normal but E7.0–7.5 embryos exhibit a unique combination of patterning defects. Notable are a shortened primitive streak, absence of a proper node and diminished production of anterior streak derivatives. Reduced Wnt3a, Tbx6 and Dll1 expression signify decreased paraxial mesoderm and reduced Otx2 expression and lack of headfolds denote a failure of head development. In addition, genes initially broadly expressed are not properly restricted to the posterior, as evidenced by the ectopic expression of Nodal, Fgf8 and Gsc in the epiblast and T (Brachyury) and Evx1 in proximal mesoderm derivatives. Despite impaired posterior restriction and anterior streak deficits, overall anterior/posterior polarity is established. A single primitive streak forms and marker expression shows that the anterior epiblast and anterior visceral endoderm (AVE) are specified.
Conclusion
Huntingtin is essential in the early patterning of the embryo for formation of the anterior region of the primitive streak, and for down-regulation of a subset of dynamic growth and transcription factor genes. These findings provide fundamental starting points for identifying the novel cellular and molecular activities of huntingtin in the extraembryonic tissues that govern normal anterior streak development. This knowledge may prove to be important for understanding the mechanism by which the dominant polyglutamine expansion in huntingtin determines the loss of neurons in Huntington's disease.
==== Body
Background
Huntington's disease (HD) is a dominantly inherited neurodegenerative disorder that is caused by CAG repeats in the HD locus that extend a polyglutamine tract in a ubiquitous HEAT domain protein called huntingtin [1]. The molecular mechanism by which the new property that is conferred on huntingtin by the polyglutamine expansion leads to the hallmark loss of striatal neurons in HD is not known. However, polyglutamine expansions in unrelated proteins that target distinct neuronal cell populations cause distinct 'polyglutamine' neurodegenerative disorders. This observation strongly suggests that the striatal cell specificity of the polyglutamine expansion in the context of huntingtin must be determined by some aspect of huntingtin's structure, subcellular location or activities [2].
Huntingtin is postulated to function as a flexible ~350 kDa HEAT domain scaffold that may facilitate the assembly and possibly the subcellular location of large protein complexes [3-7]. Huntingtin's large number of diverse cytoplasmic and nuclear protein binding partners strongly suggest that huntingtin may participate in a variety of cellular processes that range from trafficking of growth factor complexes to gene transcription (reviewed in [5,8,9]. However, despite the potential importance of huntingtin's normal function to our understanding of how the dominant polyglutamine mutation causes HD pathology, huntingtin's precise molecular and cellular activities have not been defined.
Therefore, we, and others, set out to discover huntingtin's essential activities by studying the effects of huntingtin deficiency in the mouse. Inactivation of the mouse HD gene (Hdh) has shown that huntingtin is not required for cell viability, as evidenced by the survival of mouse embryonic stem cells and neurons that lack huntingtin [10-12]. However, huntingtin is needed at the level of the organism for proper mammalian embryonic development [10,13,14]. Complete lack of huntingtin results in developmental arrest during gastrulation, while severe reduction of huntingtin levels results in abnormal neurogenesis and perinatal lethality [15].
Analysis of huntingtin deficient Hdhex4/5/Hdhex4/5 embryos reveals that homozygous inactivation of the mouse HD gene does not overtly affect development until E7.0. By E7.5, mutant embryos exhibit a shortened primitive streak, reduced size and, by morphology, lack a node and head folds. Mutants are rapidly resorbed by E8.0 [10]. Importantly, the expression of huntingtin only in extraembryonic tissues in chimeras rescues this gastrulation phenotype, suggesting that huntingtin is required only in cells of the extraembryonic lineage and acts in a cell non-autonomous manner at this stage [16].
Extraembryonic tissues are essential for supplying nutrients and signals that direct anterior/posterior axis formation and patterning in the developing embryo (reviewed in [17]), implicating huntingtin in either or both of these processes. Of these possibilities, the nutritive role has been more extensively investigated. However, huntingtin deficient embryos do not display obvious visceral endoderm defects, with the notable exception of compromised iron transport in later stage mutants, although iron uptake is undisturbed [16] and endocytosis is not impaired in huntingtin deficient embryos or embryonic stem cells [16,18].
By the same token, huntingtin shuttles through the nucleus, where it is required for proper nuclear localization of its transcription factor partners, suggesting that huntingtin may play a role in transcription cascades in extraembryonic tissues that pattern the embryo [18]. Therefore, we have examined this hypothesis, by monitoring the expression of genes that determine normal embryonic patterning and morphogenesis in Hdhex4/5/Hdhex4/5 huntingtin deficient embryos. Our results support and refine the hypothesis, indicating that huntingtin is required for proper mesoderm patterning and for normal regional restriction of the expression of a subset of growth and transcription factors.
Results
Huntingtin-deficient embryos exhibit abnormal streak progression and paraxial mesoderm production
Since extraembryonic tissues supply nutrients to the developing embryo, we tested the possibility that huntingtin deficiency may perturb this function by performing RT-PCR analysis to examine the expression of a panel of 'nutritive' genes in E7.5 wild-type and Hdhex4/5/Hdhex4/5 huntingtin deficient embryos. Consistent with a previous report [16], no obvious differences were found in the expression of "nutritive" genes (Hnf4, Afp, Tfn, ApoAI, Apo-AIV, and ApoB) or genes involved in yolk sac hematopoiesis or vasculogenesis (Ttr, Rbp, Flt1, Flk1, Tal1, Rbtn2, GATA1) (data not shown), suggesting that huntingtin is not essential for the proper expression of genes required for the nutritive function of the extraembryonic tissues.
To investigate huntingtin's developmental activities, we then analyzed the expression of genes which pattern the early embryo or mark morphogenic landmarks in wild-type and Hdhex4/5/Hdhex4/5 embryos by whole mount and section in situ hybridization. The dissections confirmed previous morphologic data at E7.0–7.5 that all Hdhex4/5/Hdhex4/5 homozygotes exhibit abnormal morphology, including shortened primitive streak and a lack of morphological head folds or node [10,13]. The results of in situ hybridization analysis also confirmed that all three germ layers and extraembryonic tissue are formed in huntingtin deficient embryos.
Otx2, normally expressed in the anterior neuroectoderm and anterior visceral endoderm [19], is expressed in mutant embryos at E7.5 (Fig. 1A,B) although the expression domain appears reduced. Similarly, Hesx1 expression is grossly normal in mutant embryos, with expression localized to the AVE and neuroectoderm (Fig. 1C–F, [20]), although the expression domain also appears reduced. These results indicate appropriate specification and movement of anterior visceral endoderm (AVE) cells from the distal tip and suggest that neuroectoderm is induced in the mutant embryos.
Figure 1 AVE displacement and anterior neurectoderm induction occur normally in the absence of huntingtin. Whole mount in situ hybridization analysis of Otx2 (A,B) and Hesx (C-F) in E7.5 normal (A,C,E) and mutant (B,D,F) embryos reveals that neuroectoderm and anterior visceral endoderm (AVE) develop normally in huntingtin deficient embryos, although the neuroectoderm expression domain is reduced. Asymmetrical expression of Hesx in mutant embryos (F) suggests that left-right transcriptional control is maintained. Hnf3β expression in the definitive endoderm extends around the distal tip and is reduced in the AVE (*) in both normal (G,I) and mutant embryos (H,J). Taken together, these results suggest normal ectoderm and endoderm induction and localization in Hdhex4/5/Hdhex4/5 embryos. Embryos are shown in lateral views, with anterior to the left in all pictures with the exception of E and F. Embryos are viewed from the anterior aspect in E and F.
To examine definitive endoderm formation, the expression of Hnf3β (FoxA2) in mutant and wild-type embryos was analyzed. In wild-type embryos, Hnf3β expression is confined to the node and anterior definitive endoderm (Fig. 1G,I[21]). Mutant embryos exhibit Hnf3β-reactive definitive endoderm over the disorganized anterior streak region and proceeding rostrally around the distal tip (Fig. 1H,J). In both normal and mutant embryos, the AVE exhibits little Hnf3β expression. Therefore, huntingtin deficiency does not greatly affect Hnf3β regulation or the reorganization of the visceral endoderm.
The lack of a morphological node and presence of a shortened streak, together with reduced neuroectoderm and lack of headfolds, suggest that anterior streak formation may be impaired in huntingtin deficient embryos. To investigate this possibility, we examined mesoderm formation in mutant embryos. Mesoderm is specified in the mutant embryos, as marked by the expression of T (Brachyury) and Evx1 (Fig. 2A–F). However, close inspection of the data reveals abnormal patterning within this tissue and its derivatives.T, normally expressed in the primitive streak, node and axial head process/notochord mesoderm [22], is detected in the shortened streak and axial mesoderm in Hdhex4/5/Hdhex4/5 embryos, extending rostrally from a region of weakly positive cells (Fig. 2A,B). T expression appears weaker, however, in the anterior streak, corresponding to cells that will give rise to axial mesoderm (Fig. 2D). T is also ectopically expressed in mutant extraembryonic mesoderm at the anterior embryonic junction and along the chorion (Fig. 2B,D). Similarly, Evx1, normally expressed in primitive streak mesoderm at E7.5 with highest levels in proximal cells [23], is expressed in the proximal shortened streak but is also aberrantly expressed throughout the extraembryonic mesoderm, allantois and chorion (Fig. 2E,F). Extraembryonic mesoderm, derived from the proximal streak, does not normally express T or Evx1 in wild-type embryos [22]. Therefore, the inappropriate expression of T and Evx1, the shortened primitive streak, and the absence of a morphological node, all suggest that the anterior primitive streak is deficient in the mutant embryos.
Figure 2 Huntingtin is required for formation of anterior primitive streak and paraxial mesoderm. Whole mount and section insitu hybridizations of E7.5 embryos shows T (Brachyury) (A-D) is expressed in the primitive streak, node, axial mesoderm and Evx1 (E-F) is expressed in the primitive streak, most strongly in the proximal streak wild-type embryos. However, in mutant embryos, both T (B, D) and Evx1 (F) are ectopically expressed in the extraembryonic region. Wnt3A expression is reduced in mutant embryos (H), although the localization of its expression to the proximal streak is the same as in wild-type embryos (G). Analysis of paraxial mesoderm markers Tbx6 (I,J) and Dll1 (K,L), reveals that these markers are reduced in mutant embryos (J,L), suggesting impaired paraxial mesoderm production in the absence of huntingtin. Embryos in A-H are shown in a lateral view with anterior oriented to the left. Embryos in I-L are shown in a posterior view (I,K) or near posterior (J,L) view with proximal oriented toward the top. In (C,D), al = allantois, a = amnion, ch = chorion, ee = embryonic node(N), em = extraembryonic mesoderm, n = node, ps = primitive streak. Rather than a node, mutant embryos exhibit a region of disorganized cells (*) at the distal extent of the short primitive streak.
The anterior streak generates paraxial mesoderm. Therefore we examined paraxial mesoderm formation in wild-type and mutant embryos, revealing deficits in mesoderm patterning. Starting at E.7.5, Wnt3A is expressed in the primitive streak in cells fated to become paraxial mesoderm. In huntingtin deficient mutants, Wnt3a is induced in the proximal streak (Fig. 2G,H), confirming stage appropriate posterior development, in contrast to the absence of anterior head folds. However, expression of Wnt3a is noticeably reduced in Hdhex4/5/Hdhex4/5 embryos, suggesting a defect in paraxial mesoderm development (Fig. 2H). Reduced expression of Tbx6 in the mesoderm lateral to the primitive streak in mutant embryos confirms this interpretation (Fig. 2I,J). Furthermore, in mutant embryos at E7.5, the expression of Dll1 in the distal streak region and in only a narrow swath of cells located laterally confirms the paucity of paraxial mesoderm (Fig. 2K,L, [24]). These results strongly suggest that anterior primitive streak formation is impaired, resulting in reduced axial and paraxial mesoderm formation and impaired neural development.
Impaired regional restriction of growth factor expression in the absence of huntingtin
To elucidate the apparent patterning deficits, we next analyzed signaling molecules that are required for early patterning. Nodal, a member of the Tgfβ family of secreted molecules is required for the formation and maintenance of the primitive streak and induction of the AVE [25-27]. Nodal is normally expressed throughout the epiblast and overlying visceral endoderm at early post implantation stages [28], but later becomes restricted to the posterior of the embryo to the site of primitive streak with asymmetrical visceral endoderm expression marking the left-right axis. By E7.5, Nodal expression is restricted to the node. Nodal expression was assessed in Hdhex4/5/Hdhex4/5 embryos heterozygous for the Ndl lac Z allele [28,29]. Notably, heterozygous loss of nodal does not alter the Hdhex4/5/Hdhex4/5 phenotype, as determined by morphology of Hdhex4/5/Hdhex4/5:Ndllacz/Ndl+ embryos compared with Hdhex4/5/Hdhex4/5 embryos (data not shown). In contrast to wild-type embryos, which exhibit tight restriction of Nodal.LacZ expression to the node, Hdhex4/5/Hdhex4/5:Ndllacz/Ndl+ embryos express Nodal.LacZ throughout the endoderm overlying the epiblast, with higher levels in the posterior in an asymmetric pattern (Fig. 3A–D). The lack of tight restriction of nodal signal is consistent with a failure to form an organized node structure.
Figure 3 Impaired regional restriction of gene expression in huntingtin deficient embryos. X-gal staining of Nodal-LacZ embryos shows staining in endoderm near the node of normal embryos (A,C) but broad staining in mutant embryos (B, D), although expression is higher in the posterior. The tight node expression of Nodal in normal embryos (C) is lost in mutant embryos (D), consistent with the loss of a morphological node in the absence of huntingtin. Whole mount and in situ hybridization of E7.5 day embryos reveals that Fgf8 is detected in the proximal streak and is downregulated in cells migrating out of the streak in normal embryos (E,G). In contrast, Fgf8 remains highly expressed in mutant embryos (F,H). Transient expression of Gsc in the definitive endoderm overlying the prospective head region in normal embryos (I,K) is distinguished in other cell layers in normal embryos but remains unrestricted in mutant embryos (J,L). Earlier posterior expression of Gsc is also maintained in mutant embryos (J) while it is down-regulated in normal embryos (I). Embryos (A,B,E,F,G,H,I,K,L) are shown in a lateral view with anterior oriented to the left. Embryos (C,D) are in a posterior view.
Fgf8 signaling is also essential for normal gastrulation in the mouse embryo. Fgf8 is required for cell migration away from the primitive streak [30]. Expressed just prior to streak formation in the posterior epiblast and visceral endoderm, Fgf8 is restricted to the streak mesoderm at E7.5 in a decreasing proximal-distal gradient and is downregulated in cells shortly after they exit the streak (Fig. 3E,G). In Hdhex4/5/Hdhex4/5 embryos, Fgf8 expression is strongly expressed in the posterior region in the primitive streak and ectopically in the endoderm overlying the entire epiblast (Fig. 3F,H). However, streak derivatives appear to migrate normally as evidenced by the proper anterior expression of markers such as Otx2, Hnf3β and Hesx1 anteriorly (Fig. 1). Therefore, mutant embryos exhibit normal migration of streak derivatives but display impaired Fgf8 repression in mutant endoderm.
Hdhex4/5/Hdhex4/5 embryos also fail to restrict the expression of goosecoid (Gsc). Normally, Gsc is initially expressed in the visceral endoderm and proximal, posterior streak where the primitive streak will form prior to gastrulation. As the primitive streak forms and extends, Gsc is expressed in the distal streak, the node, and the axial mesoderm extending anteriorly from the node (Fig. 3I,K, [31,32]). However, in the mutant Hdhex4/5/Hdhex4/5 embryos, high levels of Gsc expression remain unrestricted in the endoderm overlying the entire embryo and ectopically in cells adjacent to the ectoplacental cone (Fig. 3J,L). These results suggest that, in contrast to proper Hnf3β regulation, Gsc remains inappropriately activated in mutant visceral and definitive endoderm, implicating huntingtin in the proper restriction of this homeodomain transcription factor.
Huntingtin is not required for expression of extraembryonic signaling molecules
Previous studies of chimeric embryos suggest that huntingtin is required only in the extraembryonic tissue for proper development [16]. Signals from the extraembryonic tissue are critical for the induction of embryonic signals and for patterning the epiblast. Consequently, we examined extraembryonic development in huntingtin deficient embryos. Hnf4 is a transcription factor expressed in the primitive endoderm as soon as this tissue becomes distinct and is a key regulator of visceral endoderm secreted factors such as alphafetoprotein, apolipoproteins, and transferrin. Inactivation of Hnf4 results in impaired gastrulation [33,34]. At E7.5, Hnf4 is expressed in the columnar visceral endoderm cells at the extraembryonic-ectoderm junction (Fig. 4A, [33]). In Hdhex4/5/Hdhex4/5 embryos, consistent with normal primitive and visceral endoderm differentiation, Hnf4 expression appears normal, although the signal is stronger in mutant embryos compared to wild-type embryos (Fig. 4B). Similarly, Pem, a transcription factor expressed in proximal visceral endoderm and ectoplacental cone in wild-type embryos at E7.5, also is expressed in these tissues in the mutant embryos (Fig. 4C,D[35]). However, Pem expressing visceral endoderm hangs over the anterior of the mutant embryos, revealing abnormal location despite grossly normal differentiation.
Figure 4 Normal expression of extraembryonic markers in huntingtin deficient embryos. Whole mount in situ hybridization analysis at E7.5 of markers of the extraembryonic tissues reveals grossly normal expression in the absence of huntingtin. Hnf4, expressed in the visceral endoderm at the junction of embryonic-ectoderm junction (A), is normal in mutant embryos, although the signal is slightly higher (B). Similarly, the expression of Pem transcripts is maintained in mutant embryos (D) similar to normal embryos (C), although Pem is expressed in the abnormal lopsided overhang of visceral endoderm over the anterior of the mutant embryos. Expression of extraembryonic signaling molecules is unaffected by the loss of huntingtin, as evidenced by the expression of Bmp4 (E,F) in the extraembryonic ectoderm, and Lefty1 and Dkk1 (I-L) in the AVE in mutant embryos. Bmp4 is not localized, however, to a ring of extraembryonic ectoderm in mutant embryos (F) as in normal embryos (E). Primitive germ cells (PCGs) are induced normally in both wild-type (G) and mutant embryos (H), suggesting the Bmp4 signaling from the extraembryonic ectoderm to the epiblast is normal. Lefty1 expression appears disorganized in mutant embryos (I) compared to wild-type embryos (J). In contrast, the anterior expression of Dkk1 in the AVE in mutant embryos (L) matches the wild-type expression pattern (K). Despite normal AVE formation, head folds fail to form in mutant embryos, even when cultured in nutrient rich media for 24 hours. Wild-type E7.5 embryos, when cultured in 75% rat serum, develop somites (M), heart (white arrow, N) and head folds (blue arrow head, N) in culture. In contrast, huntingtin deficient embryos continue to live in culture but do not form headfolds, heart or somites (O). Embryos are shown in a lateral view (A-F, I-J) with anterior oriented to the left. Embryos in (G,H,K,L) are shown in an anterior view with proximal oriented up.
Signals from the extraembryonic tissues, including the anterior visceral endoderm and extraembryonic ectoderm are required for proper formation and patterning of the epiblast [17]. Bmp4 is a signaling molecule that is first expressed uniformly throughout the extraembryonic ectoderm and subsequently is localized to a ring of extraembryonic ectoderm adjacent to the epiblast (Fig. 4E, [36]). A key factor in regulating the formation of the node and primitive streak, Bmp4 is required for patterning the embryo along the proximodistal axis [37-40]. In the absence of huntingtin, Bmp4 expression is properly maintained in the Hdhex4/5/Hdhex4/5 extraembryonic ectoderm but is also expressed throughout the extraembryonic ectoderm (Fig. 4F) in a pattern that is similar to early Bmp4 expression rather than being restricted to a ring of extraembryonic ectoderm as seen in the wild-type embryos To assess Bmp4 signaling from the extraembryonic ectoderm, we evaluated primordial germ cells (PGCs), which require Bmp4 for their induction [37]. PGCs can first be detected at E7.0 and subsequently underlie the posterior portion of the primitive streak. Whole mount staining of E7.5 mutant and wild-type embryos for alkaline phosphatase activity reveals that PCGs form in Hdhex4/5/Hdhex4/5 embryos, suggesting that Bmp4 signaling is functional in the absence of huntingtin (Fig. 4G,H).
The anterior visceral endoderm (AVE) is also an extraembryonic source of signals that are critical for early patterning. Wnt and nodal antagonists, Dkk1 (mdkk-1) and Lefty1 respectively, are expressed in the AVE and are important in limiting the posteriorization of the anterior embryo by restricting Nodal and Wnt signaling [41-43]. In Hdhex4/5/Hdhex4/5 embryos, both Dkk1 (Fig. 4I,J) and Lefty1 (Fig. 4K,L) are expressed normally in the AVE as compared with wild-type embryos. However, Dkk-1 levels appear to be slightly increased in Hdhex4/5/Hdhex4/5 embryos, although the pattern of Dkk-1 expression remains unchanged and this increase may just reflect the same amount of expression in a smaller area. Therefore, the ectopic expression of Nodal (Fig. 3A–D) and the decreased Wnt3a expression (Fig. 2H) in mutant embryos do not appear to be result of changes in the expression pattern of Lefty1 or Dkk1.
Despite normal AVE formation and neuroectoderm induction, head folds do not form in Hdhex4/5/Hdhex4/5 embryos. Therefore, to determine whether mutant embryos are inherently capable of forming head folds, embryos harvested at stage E7.5 were allowed to progress in rich culture medium in vitro for 24 hours. Wild-type embryos continued to develop head folds, somites and hearts (Fig. 4M,N). In contrast, mutant stage 7.5 embryos did not develop headfolds, hearts or somites, although these embryos continued to live (Fig. 4O). These results strongly suggest that in the absence of huntingtin, embryos are unable to undergo organogenesis, even if they continue to live past E7.5 in a nutrient rich environment.
Discussion
We have investigated the embryonic processes that require huntingtin in order to more precisely delineate huntingtin's essential molecular and cellular activities and to provide clues to the mechanism by which the dominant polyglutamine expansion mutation in huntingtin leads to HD pathogenesis. In pursuing the finding that huntingtin is needed only in extraembryonic tissues for normal gastrulation, our data fail to provide evidence of abnormal nutritive gene expression in Hdhex4/5/Hdhex4/5 embryos. Instead, our results reveal that huntingtin is required for normal anterior streak formation and the consequent production of paraxial mesoderm, with a previously unrecognized role for huntingtin in the proper extinction of transiently and/or dynamically expressed genes.
Indeed, the hallmark of the huntingtin deficient molecular phenotype is the impaired down-regulation of a subset of dynamically expressed genes, after the proper onset of expression. This phenomenon does not reflect a lack of anterior/posterior axis formation, as evidenced by the formation of the AVE anteriorly and the primitive streak posteriorly. Nor can it be simply explained by delayed development, as stage-specific markers, such as Wnt3a and primordial germ cells, which are detectable at E7.0 in wild-type embryos, are induced appropriately. Furthermore, the expression of T and Evx1 in the extraembryonic mesoderm of mutant embryos is not a feature of wild-type embryos, even at earlier stages. This strongly suggests that in huntingtin deficient embryos, the migration of the distal streak derivatives to the extraembryonic mesoderm occurs normally but that the down-regulation of these genes is impaired. This impairment may also explain the failure of huntingtin deficient embryos to properly restrict the expression of Fgf8, Nodal and Gsc. Thus, huntingtin may play a direct role in the transcriptional regulation, or mRNA stability of these genes or it may act indirectly by intersecting with other pathways that regulate the expression of these genes.
The requirement for huntingtin in the extraembryonic tissues had prompted us to test whether impaired extraembryonic signals might be responsible for the dysregulation of gene expression within the epiblast that is observed in Hdhex4/5/Hdhex4/5 embryos. Extraembryonic development in Hdhex4/5/Hdhex4/5 embryos is associated with mildly elevated levels expression of Hnf4 in the primitive endoderm and Pem in the lopsided anterior chorion but the expression of other known signals, such as Bmp4 from the extraembryonic ectoderm, and Dkk1 and Lefty1 from the AVE, appear to be normal, although the slight increase in Dkk-1 expression in Hdhex4/5/Hdhex4/5 embryos suggests that further investigation into Wnt signaling is warranted. Moreover, extraembryonic Bmp4 signaling is not impaired in the absence of huntingtin, as the induction of PCGs in mutant embryos is normal, implying proper transport and secretion of the appropriate extraembryonic signals. However, Nodal, Fgf8 and Gsc are expressed ectopically in the visceral endoderm of Hdhex4/5/Hdhex4/5 embryos. Both Nodal and Fgf8, important growth factors required for normal development of the epiblast, are tightly regulated during gastrulation. Therefore, misexpression of either or both of these factors, or of goosecoid, in the visceral endoderm could contribute to the Hdhex4/5/Hdhex4/5 mutant phenotype. In addition, it is possible that other extraembryonic signal(s) that we have not analyzed may also be affected by the lack of huntingtin activity in extraembryonic cells in mutant embryos.
Huntingtin deficient embryos also fail to form headfolds, and to undergo organogenesis, even after culturing in nutrient rich media. The absence of headfold formation in these embryos does not appear to be a result of a failure to induce neurectoderm or a failure to form the AVE, since mutant embryos express markers such as Otx2, Ddk1, Lefty1 and Hesx1. In addition, since node formation is not required for neural induction [44-46], the failure to form a node in huntingtin deficient embryos is also unlikely to explain the lack of headfolds. The apparent reduction of paraxial mesoderm in Hdhex4/5/Hdhex4/5 embryos could explain the lack of headfolds since paraxial mesoderm is important for the full development of neuroectoderm, and consequently, headfolds. Alternatively, the inability to manifest headfolds could suggest that huntingtin is required at a very early stage for normal CNS development. This conclusion is consistent with the finding that severely reduced levels of huntingtin, from a hypomorphic Hdh allele, lead to abnormal brains later in embryonic development [15].
The cardinal features of complete Hdh inactivation that we observe are similar to the phenotypes that stem from the complete inactivation of the Polycomb group gene (Pc-g) Eed (embryonic ectoderm development). Indeed, complete deficiency for either huntingtin or the eed protein leads to abnormal streak development, lack of headfold formation, ectopic T, Evx1 and Nodal expression and disruption of anterior primitive streak mesoderm production [47]. Interestingly, Eed protein is also required for proper trophoblast development and normal maintenance of imprinted X-inactivation and genomic imprinting [47-49], suggesting that these activities warrant investigation in huntingtin deficient embryos.
Thus, our observations provide unexpected starting-points in the search for huntingtin's precise molecular activities, which began with the discovery that this HEAT domain protein hosts the dominant polyglutamine property that is the fundamental basis of HD pathogenesis. In HD patients and in accurate genetic replicas, HD CAG knock-in mice, the dominant mutation specifically affects the major population of neurons in the striatum, without impairing huntingtin's essential activities in embryonic development [50-53]. Indeed, homozygous HD patients make no wild-type huntingtin, and, in the mouse, a single mutant Hdh allele's worth of mutant huntingtin can fully rescue huntingtin deficiency embryonic phenotypes [15,51]. The quest to understand the HD mechanism, therefore, is aimed at delineating the huntingtin activity that may explain the striatal cell specificity of the polyglutamine mutant version of huntingtin. One hypothesis is that huntingtin is normally involved in gene transcription, as proposed for NRSF/REST mediated BDNF expression [54]. Now, our finding that huntingtin can be absolutely necessary for the appropriate regulation of genes with dynamic expression patterns in vivo, provides a compelling reason to elucidate the cellular machinery that is necessary for huntingtin mediated gene regulation.
Conclusion
Our findings indicate that huntingtin is required for proper patterning of the epiblast during early embryogenesis, for proper anterior streak and node formation, primitive streak progression, paraxial mesoderm and head fold formation, as well as for the proper restriction of transiently expressed growth and transcription factor genes. Knowledge of the molecular basis of these changes in huntingtin deficient embryos should facilitate the identification of the cellular pathways that are dependent on huntingtin activities. These will be important for implicating candidates to be assessed in the extraembryonic signals that determine anterior streak progression in the developing embryo and in delineating the dominant activity of the polyglutamine tract in huntingtin that determines the striatal specificity of HD.
Methods
Mice and genotyping
The Hdhex4/5 mice carrying a pGKneo insertion/replacement inactivating mutation in the mouse HD gene homologue have been described previously [10]. The experiments were conducted in accordance with an IACUC approved protocol, through the MGH Subcommittee on Animal Research. Mutant Hdhex4/5/Hdhex4/5 and normal littermates were obtained in timed pregnancies from mating of Hdhex4/5/Hdh+ heterozygotes, genotyped by PCR assay, as described [10]. The day of plug was taken to be E0.5. Embryos that were morphologically normal were pooled separately from morphologically mutant embryos for analysis. Nodal expression was determined in embryos from matings of Hdhex4/5/Hdh+; NdllacZ/Ndl+ compound heterozygotes genotyped by PCR assay as described in [29].
Whole mount and section in situ hybridization and β-gal staining
After dissection in PBS, embryos were fixed overnight in 4% paraformaldehyde at 4°C. For sections, decidua fixed in 4% paraformaldehyde, were embedded in paraffin and sectioned at 7 microns. RNA in situ hybridizations were performed as described previously [55]. Nodal.lacZ expression was assessed by β-galactosidase staining as reported [29], on embryos post fixed in 4% paraformaldehyde. Embryos were mounted in 80% glycerol before being photographed.
The huntingtin deficient phenotype is fully penetrant at each of the stages that were assessed [10]. Three to six embryos were evaluated for each marker, with every embryo exhibiting the same mutant phenotype in each case.
Alkaline phosphatase staining of Primordial Germ Cells (PCGs)
After dissections, embryos were fixed in 4% paraformaldehyde briefly and washed and stored in 1 × PBS/0.1% TX-100 at 4°C. Embryos were washed once with Tris-Maleate Buffer (25 mM Tris-Maleate, pH = 9.0, 0.8 mM MgCl2) and were subsequently incubated in alkaline phosphatase staining solution (25 mM Tris-Maleate, pH = 9.0, 0.8 mM MgCl2, 0.4 mg/ml alpha-naphthyl phosphate, 1 mg/ml Fast-Red). Stained embryos were washed in 1 × PBS/0.1% TX-100.
Whole embryo culture
Embryos were dissected at E7.5 and washed in DMEM. Embryos were then cultured individually in 1 ml of culture media (75% immediately centrifuged rat serum and 25% DMEM [56]) for 24 hours while rotating in a 37°C incubator in 5% CO2. Embryos were then fixed in 4% paraformaldehyde for analysis.
Abbreviations
AVE, anterior visceral endoderm; HD, Huntington's disease gene; HD, Huntington's disease; Hdh, mouse HD gene homologue; PCGs, primordial germ cells
Authors' contributions
JMW, TC, PH-M and MD performed whole mount and in situ hybridization assays. MEM and RC contributed to the conception of this study. JMW, TC, PH-M and MEM drafted the manuscript and RC contributed to its finalization. All authors read and approved the final manuscript.
Acknowledgements
We are grateful to Drs. A. Gossler, J. Darnell, Jr., J Rossant, G. Keller, S. Orkin, G. Martin, T. Yamaguchi, A. McMahon, R. Maas, K., Muneoka, A. Simeone, Hamada H. and C. Niehrs for the generous gifts of clones and antibody reagents and Dr. E. Robertson for NdllacZ mice. We would like to thank Kathy Molyneaux for her helpful suggestions and technical assistance. We also thank Vladimir Vrbanac, Janice Espinola and Edith Toral Lopez for assistance with animal husbandry. We also thank the members of the MacDonald lab for helpful discussions during the completion of this work. This work was supported by the NINDS grants NS32765 and NS16367, and grants from the Foundation for the Care and Cure of Huntington's disease and with the support of the Huntington's Disease Society of America Coalition for the Cure and the Hereditary Disease Foundation. Juliana M. Woda is the recipient of the Milton Wexler Postdoctoral Fellowship from the Hereditary Disease Foundation.
==== Refs
Huntington's Disease Collaborative Research Group A novel gene containing a trinucleotide repeat that is expanded and unstable on Huntington's disease chromosomes Cell 1993 72 971 83 8458085 10.1016/0092-8674(93)90585-E
Gusella JF MacDonald ME Molecular genetics: unmasking polyglutamine triggers in neurodegenerative disease Nat Rev Neurosci 2000 1 109 15 11252773 10.1038/35039051
Takano H Gusella JF The predominantly HEAT-like motif structure of huntingtin and its association and coincident nuclear entry with dorsal, an NF-κB/REL/dorsal family transcription factor BMC Neuroscience 2002 3 15 12379151 10.1186/1471-2202-3-15
Andrade MA Bork P HEAT repeats in the Huntington's disease protein Nat Genet 1995 11 115 6 7550332 10.1038/ng1095-115
MacDonald ME Huntingtin: alive and well and working in middle management Sci STKE 2003 2003 pe48 14600292
Marcora E Gowan K Lee JE Stimulation of NeuroD activity by huntingtin and huntingtin-associated proteins HAP1 and MLK2 Proc Natl Acad Sci U S A 2003 100 9578 83 12881483 10.1073/pnas.1133382100
Gauthier LR Charrin BC Borrell-Pages M Dompierre JP Rangone H Cordelieres FP De Mey J MacDonald ME Lessmann V Humbert S Huntingtin controls neurotrophic support and survival of neurons by enhancing BDNF vesicular transport along microtubules Cell 2004 118 127 38 15242649 10.1016/j.cell.2004.06.018
Harjes P Wanker EE The hunt for huntingtin function: interaction partners tell many different stories Trends Biochem Sci 2003 28 425 33 12932731 10.1016/S0968-0004(03)00168-3
Gusella JF MacDonald ME Huntingtin: a single bait hooks many species Current Opinion in Neurobiology 1998 8 425 30 9687360 10.1016/S0959-4388(98)80071-8
Duyao MP Auerbach AB Ryan A Persichetti F Barnes GT McNeil SM Ge P Vonsattel JP Gusella JF Joyner AL Inactivation of the mouse Huntington's disease gene homolog Hdh Science 1995 269 407 10 7618107
Metzler M Helgason CD Dragatsis I Zhang T Gan L Pineault N Zeitlin SO Humphries RK Hayden MR Huntingtin is required for normal hematopoiesis Human Molecular Genetics 2000 9 387 94 10655548 10.1093/hmg/9.3.387
Metzler M Chen N Helgason CD Graham RK Nichol K McCutcheon K Nasir J Humphries RK Raymond LA Hayden MR Life without huntingtin: normal differentiation into functional neurons Journal of Neurochemistry 1999 72 1009 18 10037472 10.1046/j.1471-4159.1999.0721009.x
Zeitlin S Liu JP Chapman DL Papaioannou VE Efstratiadis A Increased apoptosis and early embryonic lethality in mice nullizygous for the Huntington's disease gene homologue Nature Genetics 1995 11 155 63 7550343 10.1038/ng1095-155
Nasir J Floresco SB O'Kusky JR Diewert VM Ricmhan JM Zeisler J Borowski A Marth JD Phillips AG Hayden MR Targeted disruption of the Huntington's disease gene results in embryonic lethality and behavioral and morphological changes in heterozygotes Cell 1995 81 811 23 7774020 10.1016/0092-8674(95)90542-1
White JK Auerbach W Duyao MP Vonsattel JP Gusella JF Joyner AL MacDonald ME Huntingtin is required for neurogenesis and is not impaired by the Huntington's disease CAG expansion Nature Genetics 1997 17 404 10 9398841 10.1038/ng1297-404
Dragatsis I Efstratiadis A Zeitlin S Mouse mutant embryos lacking huntingtin are rescued from lethality by wild-type extraembryonic tissues Development 1998 125 1529 39 9502734
Beddington RS Robertson EJ Anterior patterning in mouse Trends Genet 1998 14 277 84 9676530 10.1016/S0168-9525(98)01499-1
Hilditch-Maguire P Trettel F Passani LA Auerbach A Persichetti F MacDonald ME Huntingtin: an iron-regulated protein essential for normal nuclear and perinuclear organelles Human Molecular Genetics 2000 9 2789 97 11092755 10.1093/hmg/9.19.2789
Simeone A Acampora D Mallamaci A Stornaiuolo A D'Apice MR Nigro V Boncinelli E A vertebrate gene related to orthodenticle contains a homeodomain of the bicoid class and demarcates anterior neuroectoderm in the gastrulating mouse embryo Embo J 1993 12 2735 47 8101484
Thomas P Beddington R Anterior primitive endoderm may be responsible for patterning the anterior neural plate in the mouse embryo Curr Biol 1996 6 1487 96 8939602 10.1016/S0960-9822(96)00753-1
Ang SL Wierda A Wong D Stevens KA Cascio S Rossant J Zaret KS The formation and maintenance of the definitive endoderm lineage in the mouse: involvement of HNF3/forkhead proteins Development 1993 119 1301 15 8306889
Wilkinson DG Bhatt S Herrmann BG Expression pattern of the mouse T gene and its role in mesoderm formation Nature 1990 343 657 9 1689462 10.1038/343657a0
Dush MK Martin GR Analysis of mouse Evx genes: Evx-1 displays graded expression in the primitive streak Dev Biol 1992 151 273 87 1349539 10.1016/0012-1606(92)90232-6
Bettenhausen B Hrabe de Angelis M Simon D Guenet JL Gossler A Transient and restricted expression during mouse embryogenesis of Dll1, a murine gene closely related to Drosophila Delta Development 1995 121 2407 18 7671806
Zhou X Sasaki H Lowe L Hogan BL Kuehn MR Nodal is a novel TGF-beta-like gene expressed in the mouse node during gastrulation Nature 1993 361 543 7 8429908 10.1038/361543a0
Conlon FL Lyons KM Takaesu N Barth KS Kispert A Herrmann B Robertson EJ A primary requirement for nodal in the formation and maintenance of the primitive streak in the mouse Development 1994 120 1919 28 7924997
Schier AF Shen MM Nodal signalling in vertebrate development Nature 2000 403 385 9 10667782 10.1038/35000126
Varlet I Collignon J Norris DP Robertson EJ Nodal signaling and axis formation in the mouse Cold Spring Harb Symp Quant Biol 1997 62 105 13 9598342
Collignon J Varlet I Robertson EJ Relationship between asymmetric nodal expression and the direction of embryonic turning Nature 1996 381 155 8 8610012 10.1038/381155a0
Sun X Meyers EN Lewandoski M Martin GR Targeted disruption of Fgf8 causes failure of cell migration in the gastrulating mouse embryo Genes Dev 1999 13 1834 46 10421635
Filosa S Rivera-Perez JA Gomez AP Gansmuller A Sasaki H Behringer RR Ang SL Goosecoid and HNF-3beta genetically interact to regulate neural tube patterning during mouse embryogenesis Development 1997 124 2843 54 9226455
Blum M Gaunt SJ Cho KW Steinbeisser H Blumberg B Bittner D De Robertis EM Gastrulation in the mouse: the role of the homeobox gene goosecoid Cell 1992 69 1097 106 1352187 10.1016/0092-8674(92)90632-M
Chen WS Manova K Weinstein DC Duncan SA Plump AS Prezioso VR Bachvarova RF Darnell JE Jr Disruption of the HNF-4 gene, expressed in visceral endoderm, leads to cell death in embryonic ectoderm and impaired gastrulation of mouse embryos Genes Dev 1994 8 2466 77 7958910
Duncan SA Nagy A Chan W Murine gastrulation requires HNF-4 regulated gene expression in the visceral endoderm: tetraploid rescue of Hnf-4(-/-) embryos Development 1997 124 279 87 9053305
Lin TP Labosky PA Grabel LB Kozak CA Pitman JL Kleeman J MacLeod CL The Pem homeobox gene is X-linked and exclusively expressed in extraembryonic tissues during early murine development Dev Biol 1994 166 170 9 7958444 10.1006/dbio.1994.1305
Waldrip WR Bikoff EK Hoodless PA Wrana JL Robertson EJ Smad2 signaling in extraembryonic tissues determines anterior-posterior polarity of the early mouse embryo Cell 1998 92 797 808 9529255 10.1016/S0092-8674(00)81407-5
Lawson KA Dunn NR Roelen BA Zeinstra LM Davis AM Wright CV Korving JP Hogan BL Bmp4 is required for the generation of primordial germ cells in the mouse embryo. [comment] Genes & Development 1999 13 424 36 10049358
Fujiwara T Dehart DB Sulik KK Hogan BL Distinct requirements for extra-embryonic and embryonic bone morphogenetic protein 4 in the formation of the node and primitive streak and coordination of left-right asymmetry in the mouse Development 2002 129 4685 96 12361961
Fujiwara T Dunn NR Hogan BL Bone morphogenetic protein 4 in the extraembryonic mesoderm is required for allantois development and the localization and survival of primordial germ cells in the mouse Proceedings of the National Academy of Sciences of the United States of America 2001 98 13739 44 11707591 10.1073/pnas.241508898
Winnier G Blessing M Labosky PA Hogan BL Bone morphogenetic protein-4 is required for mesoderm formation and patterning in the mouse Genes & Development 1995 9 2105 16 7657163
Sakuma R Ohnishi Yi Y Meno C Fujii H Juan H Takeuchi J Ogura T Li E Miyazono K Hamada H Inhibition of Nodal signalling by Lefty mediated through interaction with common receptors and efficient diffusion Genes Cells 2002 7 401 12 11952836 10.1046/j.1365-2443.2002.00528.x
Perea-Gomez A Lawson KA Rhinn M Zakin L Brulet P Mazan S Ang SL Otx2 is required for visceral endoderm movement and for the restriction of posterior signals in the epiblast of the mouse embryo Development 2001 128 753 65 11171400
Perea-Gomez A Rhinn M Ang SL Role of the anterior visceral endoderm in restricting posterior signals in the mouse embryo Int J Dev Biol 2001 45 311 20 11291861
Klingensmith J Ang SL Bachiller D Rossant J Neural induction and patterning in the mouse in the absence of the node and its derivatives Dev Biol 1999 216 535 49 10642791 10.1006/dbio.1999.9525
Episkopou V Arkell R Timmons PM Walsh JJ Andrew RL Swan D Induction of the mammalian node requires Arkadia function in the extraembryonic lineages Nature 2001 410 825 30 11298452 10.1038/35071095
Ang SL Rossant J HNF-3 beta is essential for node and notochord formation in mouse development Cell 1994 78 561 74 8069909 10.1016/0092-8674(94)90522-3
Faust C Schumacher A Holdener B Magnuson T The eed mutation disrupts anterior mesoderm production in mice Development 1995 121 273 85 7768172
Morin-Kensicki EM Faust C LaMantia C Magnuson T Cell and tissue requirements for the gene eed during mouse gastrulation and organogenesis Genesis 2001 31 142 6 11783004 10.1002/gene.10017
Wang J Mager J Chen Y Schneider E Cross JC Nagy A Magnuson T Imprinted X inactivation maintained by a mouse Polycomb group gene Nat Genet 2001 28 371 5 11479595 10.1038/ng574
Folstein SE Huntington's Disease: A Disorder of Families 1989 Baltimore, MD: Johns Hopkins Press
Lin CH Tallaksen-Greene S Chien WM Cearley JA Jackson WS Crouse AB Ren S Li XJ Albin RL Detloff PJ Neurological abnormalities in a knock-in mouse model of Huntington's disease Human Molecular Genetics 2001 10 137 44 11152661 10.1093/hmg/10.2.137
Vonsattel JP DiFiglia M Huntington disease J Neuropathol Exp Neurol 1998 57 369 84 9596408
Wheeler VC Gutekunst CA Vrbanac V Lebel LA Schilling G Hersch S Friedlander RM Gusella JF Vonsattel JP Borchelt DR Early phenotypes that presage late-onset neurodegenerative disease allow testing of modifiers in Hdh CAG knock-in mice Human Molecular Genetics 2002 11 633 40 11912178 10.1093/hmg/11.6.633
Zuccato C Tartari M Crotti A Goffredo D Valenza M Conti L Cataudella T Leavitt BR Hayden MR Timmusk T Huntingtin interacts with REST/NRSF to modulate the transcription of NRSE-controlled neuronal genes Nat Genet 2003 35 76 83 12881722 10.1038/ng1219
Conlon RA Herrmann BG Detection of messenger RNA by in situ hybridization to postimplantation embryo whole mounts Methods Enzymol 1993 225 373 83 8231865
Sturm K Tam PP Isolation and culture of whole postimplantation embryos and germ layer derivatives Methods Enzymol 1993 225 164 90 8231855
|
16109169
|
PMC1208873
|
CC BY
|
2021-01-04 16:30:31
|
no
|
BMC Dev Biol. 2005 Aug 18; 5:17
|
utf-8
|
BMC Dev Biol
| 2,005 |
10.1186/1471-213X-5-17
|
oa_comm
|
==== Front
BMC Dev BiolBMC Developmental Biology1471-213XBioMed Central London 1471-213X-5-181612021210.1186/1471-213X-5-18Research ArticleRestricted mobility of Dnmt1 in preimplantation embryos: implications for epigenetic reprogramming Grohmann Maik [email protected] Fabio [email protected] Lothar [email protected] Natalia [email protected] Michael [email protected] M Cristina [email protected] Heinrich [email protected] Department of Biology II, Ludwig Maximilians University Munich, Grosshadernerstr. 2, 82152 Planegg-Martinsried, Germany2 Max Delbruck Center for Molecular Medicine, Berlin, Germany3 Max-Planck-Institute for Molecular Genetics, Berlin, Germany2005 24 8 2005 5 18 18 4 7 2005 24 8 2005 Copyright © 2005 Grohmann et al; licensee BioMed Central Ltd.2005Grohmann 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
Mouse preimplantation development is characterized by both active and passive genomic demethylation. A short isoform of the prevalent maintenance DNA methyltransferase (Dnmt1S) is found in the cytoplasm of preimplantation embryos and transiently enters the nucleus only at the 8-cell stage.
Results
Using GFP fusions we show that both the long and short isoforms of Dnmt1 localize to the nucleus of somatic cells and the cytoplasm of preimplantation embryos and that these subcellular localization properties are independent of phosphorylation. Importantly, photobleaching techniques and salt extraction revealed that Dnmt1S has a very restricted mobility in the cytoplasm, while it is highly mobile in the nucleus of preimplantation embryos.
Conclusion
The restricted mobility of Dnmt1S limits its access to DNA and likely contributes to passive demethylation and epigenetic reprogramming during preimplantationdevelopment.
==== Body
Background
In mammals establishment and maintenance of DNA methylation patterns are crucial for embryonic development, cell differentiation, silencing of transposable elements, X inactivation and allele-specific expression of imprinted genes [1]. DNA methyltransferases (Dnmts) are responsible for establishment and maintenance of methylation patterns. In contrast to Dnmt3a and 3b, which catalyze de novo methylation of unmethylated DNA, Dnmt1 shows a preference for hemi-methylated DNA and is targeted to replication foci by binding to PCNA during S-phase [2-4]. Thus, Dnmt1 is thought to maintain genomic methylation through DNA replication by reproducing the cytosine methylation pattern of the parental DNA strand onto the newly synthesized strand.
Genomic methylation patterns undergo drastic changes during gametogenesis and early embryonic development. In the germ line, methylation patterns are erased early in development and gamete-specific ones are established during gametogenesis [5]. In the mouse zygote there is a drastic decrease of DNA methylation in the paternal genome within a few hours after fertilization (active demethylation) and both the maternal and paternal genomes undergo progressive demethylation during segmentation stages [6-9]. This is followed by establishment of new, tissue specific methylation patterns beginning around the time of implantation [9,10].
Different isoforms of Dnmt1 are encoded by the mouse dnmt1 locus. A longer isoform (Dnmt1L) is expressed in somatic and embryonic stem cells where it is strictly nuclear, except in post-mitotic neurons where it is also found in the cytoplasm [2,11,12]. A shorter, maternally contributed isoform lacking 118 amino acids at the N-terminus (Dnmt1S) is found in the cytoplasm of maturing oocytes and preimplantation embryos and enters the nucleus only transiently at the 8-cell stage [13-16]. The methylation maintenance function of Dnmt1 is shared by the long and short isoforms as the latter can rescue methylation patterns and differentiation potential in ES cells and mice lacking the former [11,17]. It is believed that retention of Dnmt1S in the cytoplasm of preimplantation embryos may prevent maintenance of gamete-specific methylation patterns, determining their erasure by passive demethylation and thus contributing to epigenetic reprogramming of the embryo. However, it is far from clear how methylation patterns at imprinted loci and transposable elements are maintained throughout preimplantation development and how Dnmt1S is prevented from entering the nucleus. Interestingly, during Xenopus early embryonic development a Dnmt1 isoform equivalent to the mouse long isoform is present in the nuclei and only limited demethylation occurs [18,19].
Here we investigated the localization of GFP fusions of the long and short Dnmt1 isoforms in mouse preimplantation embryos and directly compared their mobility in the nucleus and cytoplasm of living embryos.
Results and discussion
To directly compare the subcellular localization of the two Dnmt1 isoforms in cycling somatic cells and preimplantation embryos we expressed GFP-fusions of Dnmt1S and L (Fig. 1A and [4]) in both systems. After microinjection of the expression constructs in 1-cell embryos both fusion proteins were localized in the cytoplasm of preimplantation embryos (Fig. 1C), while they were exclusively nuclear in transfected mouse myoblasts (Fig. 1B). These results confirm earlier immunolocalization studies and indicate that the differential localization of the two Dnmt1 isoforms in somatic cells and embryos does not depend on the additional N-terminal 118 amino acids in Dnmt1L [13-16,20]. Nuclear localization of both isoforms in somatic cells is likely due to the fact that all active nuclear localization sequences are found within the region shared by the two Dnmt1 isoforms [15]. Indeed, overexpression of both isoforms by injection of 2–4 fold more plasmid DNA resulted in nuclear localisation of a fraction of the fusion proteins also in preimplantation embryos (Fig. 1C, 3A and 3C), suggesting a saturable cytoplasmic retention mechanism.
Figure 1 Subcellular localisation of Dnmt1 isoforms in mouse somatic cells and preimplantation embryos. A) Schematic representation of GFP-Dnmt1 fusion proteins. The start codons of the long (ATGL) and the short (ATGS) isoforms are indicated. The catalytic domain of Dnmt1 is in black. Subcellular localisation of GFP-Dnmt1 fusions in somatic cells (B) and 2-cell embryos (C). In B mouse C2C12 myoblasts were transfected with either the GFP-Dnmt1S (left pair of panels) or the GFP-Dnmt1L expression constructs (right pair of panels) and imaged by confocal microscopy. The left panel in each pair shows the phase contrast image, while the right panel shows GFP fluorescence (scale bars = 5 μm). In C the same expression constructs were microinjected in pronuclei at the 1-cell stage and embryos were further cultured until the 2-cell stage (scale bars = 20 μm). Both the short and the long isoforms of Dnmt1 are localised in the nucleus of myoblasts and in the cytoplasm of 2-cell embryos. Small amounts of fusion proteins in embryonic nuclei (arrowheads) are due to overexpression of Dnmt1 and consequent saturation of the cytoplasmic retention mechanism.
Figure 3 Restricted mobility of GFP-Dnmt1S in the cytoplasm of mouse preimplantation embryos. A) The GFP-Dnmt1S expression construct was microinjected in 1-cell stage embryos and a portion of the cytoplasm of one blastomere was bleached at the 2-cell stage. The upper row shows bleaching of a living embryo while the lower row shows a fixed control. Note that only regions immediately adjacent to the bleached area show decreased fluorescence. A very sharp bleaching boundary in fixed controls shows that such decrease is not due to poor sharpness of the bleaching beam, but to diffusion of GFP-Dnmt1S from adjacent sites. B) Salt extraction of endogenous Dnmt1S from 2-cell embryos. Soluble (S) and insoluble (P) fractions were analysed by immunoblotting with an anti-Dnmt1 antibody. C) Localisation dependent mobility of Dnmt1S in 1-cell embryos. A square bleached area (indicated) including a small fraction of the male pronucleus (outlined) was produced in a 1-cell embryo microinjected with the GFP-Dnmt1S construct. After bleaching no fluorescence remained in the entire pronucleus, while in the cytoplasm fluorescence was depleted only within and in proximity of the bleached area, indicating that the mobility of GFP-Dnmt1S is specifically restricted in the cytoplasm. Insets on the right show magnifications of the bleached area at the indicated time points. A) and C) show optical sections obtained by confocal microscopy (scale bars = 20 μm).
Posttranslational modification, in particular phosphorylation, is a well documented mechanism controlling the nuclear-cytoplasmic localization of a large number of proteins. The only posttranslational modification reported so far for Dnmt1 is phosphorylation of serine 514 (396 in Dnmt1S) [21]. To test whether the phosphorylation state of this residue determines the differential localization of GFP-Dnmt1S we generated mutations either mimicking the phosphorylated state (S396D) or preventing it (S396A; Fig. 2A). As shown in figure 2C and 2D neither mutation changed the nuclear localization of GFP-Dnmt1S in somatic cells or the cytoplasmic localization in preimplantation embryos.
Figure 2 Dnmt1S localisation is independent of phosphorylation. A) Schematic representation of GFP-Dnmt1S phosphorylation mutants. B) Western blot of transfected Cos-7 cells probed with an antibody against the N-terminal domain of Dnmt1 [25] showing expression of GFP-Dnmt1s wild type (wt) and phosphorylation mutants S396A and S396D with the expected molecular mass. Weak bands at 190 kDa representing endogenous Dnmt1L are indicated. C) and D) localisation of GFP-Dnmt1S phosphorylation mutants in mouse somatic cells and preimplantation embryos, respectively. Both mutant proteins localise to the nucleus of somatic cells and to the cytoplasm of preimplantation embryos like wild type GFP-Dnmt1S (Fig. 1B and C). In D embryos were fixed and immunostained with an anti-GFP antibody detected with a red fluorescent secondary antibody. The upper row shows GFP fluorescence and the lower row shows the immunofluorescent signal (IF). The IF gradient from the periphery to the centre of some embryos is an optical artefact depending on signal intensity [15].
To further investigate whether GFP-Dnmt1S is soluble or tightly bound to cytoplasmic structures in preimplantation embryos we tested fluorescence recovery after photobleaching of a defined region in the cytoplasm of microinjected embryos (Fig. 3A). Fluorescence recovery was very slow in living embryos with very little decrease of fluorescence in the non-bleached regions apart for those just next to the bleached area. This slow recovery could be explained by repeated association with fixed binding sites restricting its diffusion in the cytoplasm. We further assayed the stability of such binding of the endogenous Dnmt1S by salt extraction experiments (Fig. 3B). Essentially all Dnmt1S was found in the insoluble fraction with up to 250 mM NaCl, but was solubilized at 400 mM. This result provides biochemical evidence for strong interactions of Dnmt1S in early embryos. Since Dnmt1 is highly mobile in nuclei of somatic cells (data not shown) we directly compared the mobility of the short Dnmt1 isoform in the cytoplasm to that in the nucleus of the same embryo. For this we extensively bleached a region spanning across the nuclear-cytoplasmic boundary (Fig. 3C). Although only about one tenth of the nuclear volume was illuminated, the nuclear fluorescence was entirely depleted, indicating a high mobility of GFP-Dnmt1S molecules in the nucleus. In contrast, only regions within and close to the targeted area in the cytoplasm showed decreased fluorescence. As above, the rate of fluorescence recovery was slow, which is consistent with Dnmt1S binding to fixed structures in the cytoplasm. These results clearly show that Dnmt1S has a much higher mobility in the nucleus than in the cytoplasm of preimplantation embryos.
In summary, the subcellular localization of Dnmt1 does not depend on either the additional 118 amino acids at the N-terminus of Dnmt1L or the phosphorylation state at serine 396 of Dnmt1S. Previously, we showed that Dnmt1S fused to a SV40 NLS was still retained in the cytoplasm, ruling out that masking of the NLS of Dnmt1S could be the mechanism preventing nuclear localization like in the case of IκB preventing nuclear import of NF-κB. The biochemical extraction data indicate a strong binding of Dnmt1S, the overexpression experiments show that these binding sites are saturable and the fluorescence bleaching results demonstrate that GFP-Dnmt1S is specifically immobilized in the cytoplasm of preimplantation embryos.
Conclusion
Taken together the data presented in this study argue for a strong binding of Dnmt1S to immobile structures in the cytoplasm of early embryos. Sequestration of Dnmt1 in the cytoplasm of embryos is likely to prevent full maintenance of methylation patterns during preimplantation development. Since mice expressing reduced levels of Dnmt1 protein already show severe genomic hypomethylation [22], even partial cytoplasmic retention of Dnmt1 could account for global passive demethylation and epigenetic reprogramming. At the same time, preferential targeting of remaining nuclear Dnmt1 could allow maintenance of parental methylation patterns at imprinted loci and silencing of endogenous retroviral elements.
Methods
DNA constructs
The expression construct for GFP-Dnmt1S (pEGMT1S) was derived by subcloning the cDNA insert from pEMT [23] into the Kpn I / Xma I site of the pEGFP-C1 vector (Clontech). The mutations S396D and S396A were introduced in pEMT by site directed mutagenesis using the QuikChange system (Stratagene) and the corresponding cDNA inserts were subcloned into the pEGFP-C1 vector as above. The expression construct for GFP-Dnmt1L (pEGMT1L) was obtained in two steps. First a 3.3 kb Xma I fragment from a PCR product spanning the whole N-terminal regulatory domain of Dnmt1L was cloned in the pEGFP-C1 vector to obtain pEGNMT. Next a Hind III / Psh AI fragment of pEGMT1S encompassing part of the EGFP coding sequence and the 5' part of the partial Dnmt1 cDNA till shortly downstream of ATG4 [11] was replaced with the corresponding fragment from pEGNMT, which contains the complete 5' sequence of Dnmt1L cDNA, including the part encoding the additional 118 amino acids. In all vectors transcription is under the control of the CMV promoter.
Embryo collection, culture, microinjection and cell transfection
Mouse embryos were obtained from FVB/N superovulated females mated with FVB/N males. For superovulation females (3–4 weeks old) were injected intraperitoneally with 5 IU of pregnant mare serum (PMS) followed by intraperitoneal injection of 5 IU of human chorionic gonadotropin (hCG) 46–48 hours later. Matings were set up right after hCG injection. Zygotes were collected from the oviduct of 0.5 days post-coitum (d.p.c.) donors into M2 medium containing 300 μg/ml hyaluronidase to remove the cumulus cells. After washing in M2 medium the zygotes were transferred to M16 medium in microdrop cultures for further development up to the blastocyst stage in a humidified 5% CO2 incubator at 37°C. Culture conditions and media compositions were as described [24]. For injection, plasmid DNA was linearized, purified with standard glassmilk absorption techniques (Qiagen) and resuspended in microinjection buffer (8 mM Tris, pH 7.4 and 0.15 mM EDTA) at a final concentration of 10 μg/ml. DNA was injected into male pronuclei. C2C12 mouse myoblasts and Cos-7 cells were cultured and transfected as described [4,15].
Immunofluorescence
Embryos were fixed with 3.7% formaldehyde in PBS for 10 min, permeabilised with 0.25% Triton X-100 in PBS and blocked with 0.2% fish skin gelatine in PBS for 30 min. Both primary and secondary antibodies were diluted in blocking solution and applied for 1 hr. The rabbit polyclonal antibody to GFP was from Abcam and was detected with an anti-rabbit secondary antibody conjugated with Alexa Fluor 568 (Molecular Probes). Embryos were washed with 0.1% NP-40 in PBS and mounted in Mowiol.
Confocal Microscopy and Fluorescence Recovery after Photobleaching (FRAP)
Confocal images (2 μm thick optical slices) were taken with a confocal laser scanning microscope LSM510 (Zeiss) using a 488 nm and a 543 nm laser lines for GFP and Alexa Fluor 568 excitation and BP 500–530 nm and LP 570 nm filters for detection, respectively. For live cell microscopy, microinjected embryos were transferred to 8-well LabTek chambers (Nunc) containing M2 medium. FRAP experiments were performed at RT on a LSM510 using a 63 × Plan-Apochromat oil immersion objective (N.A. 1.4; Zeiss) with a 488 nm argon laser line. After acquisition of an initial (prebleach) optical section, a selected area was bleached for up to 30 s at 100 % laser power. Fluorescence recovery was then monitored by time lapse imaging 2 s and 5, 10 and 15 min after bleaching at low laser power.
Salt extraction of mouse embryos
2-cell stage embryos were washed in PBS and transferred to a fresh Eppendorf tube in a 10 μl volume. 10 μl of 2 × EBC lysis buffer (100 mM Tris-HCl pH 8,0; 1% NP-40, pepstatin, aprotinin, leupeptin 2 μg/ml each, 200 μg/ml PMSF and either 200 mM, 500 mM or 800 mM NaCl) were added to each sample, followed by 5 min incubation on ice. 45 embryos were processed in each case. After centrifugation at 13,000 rpm and 4°C for 5 min soluble (supernatant) and insoluble (pellet) fractions were analysed by SDS-PAGE followed by immunoblotting and incubation with an antibody against the N-terminal domain of Dnmt1 [25].
Abbreviations
GFP: green fluorescent protein; Dnmt: DNA methyltransferase.
Authors' contributions
MG performed all the experimental work on mouse embryos. FS contributed fig. 2C and to writing of the manuscript. LS contributed fig. 1B. NA and MB provided crucial assistance to setting up experiments on mouse embryos. MCC and HL were involved in the conception and coordination of this project and the writing of the manuscript. All authors approved the final manuscript.
Acknowledgements
We are grateful to Petra Domaing (Max Delbruck Center, Berlin) for technical assistance. This work was funded by the Deutsche Forschungsgemeinschaft.
==== Refs
Bird A DNA methylation patterns and epigenetic memory Genes Dev 2002 16 6 21 11782440 10.1101/gad.947102
Leonhardt H Page AW Weier HU Bestor TH A targeting sequence directs DNA methyltransferase to sites of DNA replication in mammalian nuclei Cell 1992 71 865 873 1423634 10.1016/0092-8674(92)90561-P
Chuang LSH Ian HI Koh TW Ng HH Xu G Li BFL Human DNA-(Cytosine-5) Methyltransferase-PCNA Complex as a Target for p21WAF1 Science 1997 277 1996 2000 9302295 10.1126/science.277.5334.1996
Easwaran HP Schermelleh L Leonhardt H Cardoso MC Replication-independent chromatin loading of Dnmt1 during G2 and M phases EMBO Rep 2004 5 1181 1186 15550930 10.1038/sj.embor.7400295
Hajkova P Erhardt S Lane N Haaf T El-Maarri O Reik W Walter J Surani MA Epigenetic reprogramming in mouse primordial germ cells Mechanisms of Development 2002 117 15 23 12204247 10.1016/S0925-4773(02)00181-8
Mayer W Niveleau A Walter J Fundele R Haaf T Demethylation of the zygotic paternal genome Nature 2000 403 501 502 10676950
Oswald J Engemann S Lane N Mayer W Olek A Fundele R Dean W Reik W Walter J Active demethylation of the paternal genome in the mouse zygote Curr Biol 2000 10 475 478 10801417 10.1016/S0960-9822(00)00448-6
Rougier N Bourc'his D Gomes DM Niveleau A Plachot M Paldi A Viegas-Pequignot E Chromosome methylation patterns during mammalian preimplantation development Genes Dev 1998 12 2108 2113 9679055
Santos F Hendrich B Reik W Dean W Dynamic reprogramming of DNA methylation in the early mouse embryo Dev Biol 2002 241 172 182 11784103 10.1006/dbio.2001.0501
Monk M Boubelik M Lehnert S Temporal and regional changes in DNA methylation in the embryonic, extraembryonic and germ cell lineages during mouse embryo development Development 1987 99 371 382 3653008
Gaudet F Talbot D Leonhardt H Jaenisch R A short DNA methyltransferase isoform restores methylation in vivo J Biol Chem 1998 273 32725 32729 9830015 10.1074/jbc.273.49.32725
Inano K Suetake I Ueda T Miyake Y Nakamura M Okada M Tajima S Maintenance-type DNA methyltransferase is highly expressed in post-mitotic neurons and localized in the cytoplasmic compartment J Biochem (Tokyo) 2000 128 315 321 10920268
Carlson LL Page AW Bestor TH Properties and localization of DNA methyltransferase in preimplantation mouse embryos: implications for genomic imprinting Genes Dev 1992 6 2536 2541 1340468
Mertineit C Yoder JA Taketo T Laird DW Trasler JM Bestor TH Sex-specific exons control DNA methyltransferase in mammalian germ cells Development 1998 125 889 897 9449671
Cardoso MC Leonhardt H DNA methyltransferase is actively retained in the cytoplasm during early development J Cell Biol 1999 147 25 32 10508852 10.1083/jcb.147.1.25
Howell CY Bestor TH Ding F Latham KE Mertineit C Trasler JM Chaillet JR Genomic Imprinting Disrupted by a Maternal Effect Mutation in the Dnmt1 Gene Cell 2001 104 829 838 11290321 10.1016/S0092-8674(01)00280-X
Ding F Chaillet JR In vivo stabilization of the Dnmt1 (cytosine-5)- methyltransferase protein Proc Natl Acad Sci U S A 2002 99 14861 14866 12397175 10.1073/pnas.232565599
Stancheva I El-Maarri O Walter J Niveleau A Meehan RR DNA Methylation at Promoter Regions Regulates the Timing of Gene Activation in Xenopus laevis Embryos Developmental Biology 2002 243 155 165 11846484 10.1006/dbio.2001.0560
Shi L Suetake I Kawakami T Aimoto S Tajima S Xenopus eggs express an identical DNA methyltransferase, Dnmt1, to somatic cells J Biochem (Tokyo) 2001 130 359 366 11530011
Ratnam S Mertineit C Ding F Howell CY Clarke HJ Bestor TH Chaillet JR Trasler JM Dynamics of Dnmt1 methyltransferase expression and intracellular localization during oogenesis and preimplantation development Dev Biol 2002 245 304 314 11977983 10.1006/dbio.2002.0628
Glickman JF Pavlovich JG Reich NO Peptide Mapping of the Murine DNA Methyltransferase Reveals a Major Phosphorylation Site and the Start of Translation J Biol Chem 1997 272 17851 17857 9211941 10.1074/jbc.272.28.17851
Gaudet F Hodgson JG Eden A Jackson-Grusby L Dausman J Gray JW Leonhardt H Jaenisch R Induction of tumors in mice by genomic hypomethylation Science 2003 300 489 492 12702876 10.1126/science.1083558
Czank A Hauselmann R Page AW Leonhardt H Bestor TH Schaffner W Hergersberg M Expression in mammalian cells of a cloned gene encoding murine DNA methyltransferase Gene 1991 109 259 263 1662657 10.1016/0378-1119(91)90618-L
Nagy A Gertsenstein M Vintersten K Behringer R Manipulating the mouse embryo 2003 Cold Spring Harbor, Cold Spring Harbor Laboratory Press
Li E Bestor TH Jaenisch R Targeted mutation of the DNA methyltransferase gene results in embryonic lethality Cell 1992 69 915 926 1606615 10.1016/0092-8674(92)90611-F
|
16120212
|
PMC1208874
|
CC BY
|
2021-01-04 16:30:32
|
no
|
BMC Dev Biol. 2005 Aug 24; 5:18
|
utf-8
|
BMC Dev Biol
| 2,005 |
10.1186/1471-213X-5-18
|
oa_comm
|
==== Front
BMC DermatolBMC Dermatology1471-5945BioMed Central London 1471-5945-5-91610917310.1186/1471-5945-5-9Case ReportPsoriasis vulgaris flare during efalizumab therapy does not preclude future use: a case series Lowes Michelle A [email protected] James A [email protected] James G [email protected] Ross StC [email protected] Laboratory for Investigative Dermatology, The Rockefeller University, New York, NY, USA2 Department of Dermatology, Royal Prince Alfred Hospital, New South Wales, Australia2005 18 8 2005 5 9 9 2 5 2005 18 8 2005 Copyright © 2005 Lowes et al; licensee BioMed Central Ltd.2005Lowes 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
Severe psoriasis vulgaris can be extremely difficult to treat in some patients, even with the newer biological therapies available today.
Case presentations
We present two patients with severe chronic plaque psoriasis who received numerous systemic anti-psoriatic therapies with varied results. Both responded well to initial treatment with efalizumab (anti-CD11a), but then experienced a flare of their disease after missing a dose. However, after disease stablization, both patients responded well to re-introduction of efalizumab, one patient requiring concurrent treatment with infliximab (anti-TNF-α).
Conclusion
These cases are presented to characterize this "flare" reaction, and to inform health care providers that efalizumab can still be administered after disease flare, and again may be a successful therapy.
==== Body
Background
Psoriasis may be a long-lasting disease resulting in great morbidity in affected patients. Newer biological therapies may offer a real alternative to those with severe disease, and they are associated with a different toxicity profile than traditional systemic therapies [1]. The agents currently approved by the US FDA are alefacept (anti-CD2, Amevive, Biogen), efalizumab (anti-CD11a, Raptiva, Genentec Inc) and etanercept (anti-TNF receptor, Enbrel, Amgen). Infliximab (anti-TNF-α, Remicade, Centocor) has not yet been approved for psoriasis vulgaris, although it has recently been approved for psoriatic arthritis.
Efalizumab is a humanized monoclonal antibody targeting the α chain of the T cell adhesion integrin lymphocyte function-associated antigen (LFA)-1. The LFA-1- intracellular adhesion molecule (ICAM)-1 interaction plays a crucial role in T cell adhesion at several key points in immune activation pathways. By binding to ICAM-1 on dendritic cell (DCs), endothelial cells and keratinocytes, T cells may be activated, migrate, and interact with keratinocytes respectively. The mechanism of action is not yet completely understood, however during therapy peripheral lymphocytosis is observed, which is most likely due to inhibition of T cell trafficking and blockade of memory T cells entering inflamed skin [2]. Efalizumab is associated with a rebound flare reaction in approximately 5% of patients when therapy is ceased [3]. However, we were not able to find reports of exacerbations of psoriasis while on therapy, as in these cases.
Infliximab is a chimeric anti-TNF-α monoclonal antibody which gives excellent results in the majority of patients at a dose of 5 mg/kg per infusion [4]. Case reports of combination therapies with two biological agents have not yet been reported for psoriasis. The main concern with this therapeutic strategy is the risk of opportunistic infection and malignancy, which should be constantly considered. We present these cases to document the clinical and histological appearance of the flare reaction occurring during previously-effective efalizumab therapy, and demonstrate that this agent can be reintroduced with good clinical effect.
Case presentations
Patient 1 is a 51 year old man from Ecuador, with severe large plaque psoriasis for 15 years, and a strong family history of psoriasis. His medical background included recent-onset hypertension and diabetes, and renal calculi. He takes lisinopril and glyburide, as well as doxepin and atarax when required. His past psoriasis treatments include topical steroids, methotrexate (not tolerated due to nausea), and UVB with minimal effect.
He was first seen at The Rockefeller University, NY, USA, in December 2000 and received numerous courses of biological therapies in the context of our clinical trials program. He initially received efalizumab (100 mg [1 mg/kg] sc weekly for 12 weeks) with good effect. His re-treatment with efalizumab was required in May 2001 because of a sunburn-induced flare, and was permitted under our clinical trial protocol. Another psoaisis flare in Sept 2001 was treated with alefacept (7.5 mg IV for 12 weekly doses) with good effect. Subsequent disease exacerbations were managed well with a course of daclizumab (anti-CD25) therapy, NB-UVB, and cyclosporine.
Due to previous success with efalizumab and recent USA FDA approval, a disease flare in March 2004 was managed with efalizumab at the standard dose (1 mg/kg/wk sc) at a private clinic. However, he missed a dose in June 2004 and his skin flared again, so he re-attended our clinic (Fig 1). Despite missing a dose, there was still leukocyte CD11a saturation by efalizumab (Fig. 2B, solid line identical to isotype, shaded). At this time, his psoriasis was complicated by Staphlococcal skin infection. To gain control of his skin disease and while waiting for his skin infection to respond to antibiotic treatment (dicloxicillin), he was given low-dose NB-UVB. Efalizumab was re-commenced in September 2004 with good result (110 mg/wk, 1 mg/kg). This has been continued and the patient is currently in remission.
Figure 1 Clinical photos of patient 1 during the flare reaction in June 2004 (upper panels), with follow-up photos after standard-dose treatment with efalizumab for three months (lower panels).
Figure 2 FACS analysis of peripheral blood lymphocytes (gated) of patient 1, pre-treatment (A) and during flare (B). Isotype (Becton Dickenson) is shaded, solid line shows T cell binding of efalizumab-FITC (custom design, Genetech), and dotted line shows binding of an anti-CD11a antibody which binds to a different epitope (clone 25.3, Immunotech). The pre-treatment sample (A) shows high-level binding of both anti-CD11a antibodies. The sample acquired during flare (B) shows the ex-vivo level of binding of efalizumab-FITC is the same as isotype, as there is already saturation of this CD11a epitope by therapeutic administration of efalizumab. There is also reduced binding of the 25.3 epitope indicating significant down-regulation of CD11a surface expression.
Immunohistochemistry during the flare reaction (Fig. 3), shows there are relatively less lesional CD3+ cells compared to untreated psoriasis, and they are predominantly all CD8+. CD103 is also expressed on epidermal T cells. In addition, there are abundant CD11c+ and inducible nitric oxide synthase (iNOS)+ inflammatory cells infiltrating the dermis and epidermis. CD14+ cells are relatively rare (not shown).
Figure 3 Immunohistochemistry of lesional skin biopsies during psoriatic flare of patient 1 and 2, showing similar features (magnification × 10). Haematoxylin and eosin (H&E) stain showing epidermal hyperplasia, elongation of rete ridges, dilatation of dermal papilla blood vessels, and mononuclear and neutrophil leukocyte infiltration. Some CD3 (BD) T cells infiltrate the dermis and epidermis, and these are mostly CD8+ (BD Pharmingen), although there are less T cells than during untreated psoriasis. Most of the epidermal T cells are CD103+ (Biodesign, Kennebunnk, ME). CD11c+ (BD Pharmingen) and iNOS+ (R&D Systems) cells are dramatically increased compared to non-lesional skin, and stain in a dendritic pattern.
Patient 2 is a 33 year-old Australian-born Caucasian female with a 14 year history of severe plaque psoriasis, and psoriatic arthritis for 12 years. Her first presentation of psoriasis was with erythroderma at 19 years old. Her past medical history included appendicectomy, infectious mononucleosis, and Chloroquine-resistant P. falciparum malaria at age 15 years following travel to Papua New Guinea. She was not taking any medications at the time of initial referral to Royal Prince Alfred Hospital Department of Dermatology, New South Wales, Australia, in 1994.
Past rotational treatments for her psoriasis and psoriatic arthritis included cyclosporine (1990 and 1995), methotrexate (1993 and 1996), acitretin (1994 and 1997), as well as periodic courses of NB-UVB. She was hospitalized in 1999 for erythrodermic psoriasis. In 2001 her psoriasis flared and other therapies tried without success were mycophenolate mofetil, hydroxyurea, tacrolimus, and thioguanine. From July 2001 to November 2002 she received periodic treatment with cyclosporine and acitretin therapy, with poor disease control.
Over the next 2.5 years, this patient was treated with a number of newer biological therapies with standard dosing, sometimes requiring cyclosporine cover and subsequent withdrawal as the patient responded. Etanercept (anti-TNF receptor) (Dec 2002–Feb 2003, 25 mg sc twice week) was initially successful, but was ceased on relapse to erythroderma. Alefacept (anti-CD2) (15 mg IM weekly for 3 months) had no beneficial effect.
Efalizumab (anti-CD11a) (80 mg [1 mg/kg] sc weekly, Dec 2003) induced a dramatic clinical response, and after 6 weeks, cyclosporine was ceased. The patient missed the 18th dose of efalizumab, but the regular dose was given at the subsequent visit. Two days after the "catchup" dose, the patient developed a psoriatic flare involving extensive plaques on all body surfaces. Immunohistochemistry at this time (Fig. 3) shows similar results to patient 1: there are relatively reduced CD3+ and CD8+ cells, epidermal T cells are CD103+, and there are abundant CD11c+ and iNOS+ inflammatory cells infiltrating the dermis and epidermis.
During this flare, NB-UVB was administered concomitantly with efalizumab, however a UVB-induced burn was associated with recurrence of her psoriasis. However, by June 2004 there was almost confluent severe plaque psoriasis again with features of erythroderma. The efalizumab dose was increased to 125 mg (1.5 mg/kg) sc weekly, but was ceased in Sept 2004 as there was no further improvement.
Infliximab (anti-TNF-α) (400 mg [5 mg/kg] at 0 and 2 weeks, followed by 400 mg every 6 weeks) was commenced with excellent effect for 4 months, when the patient then experienced another flare of her psoriasis. It was decided to carefully combine infliximab and efalizumab therapy. In February 2005, 62.5 mg (0.7 mg/kg) efalizumab was given sc, and a second dose of 125 mg (1.4 mg/kg) a week later, and she started to respond clinically. She is currently well controlled on this therapy.
Discussion
We present two cases of severe large plaque psoriasis, with both patients clearing with initial efalizumab treatment, experiencing a flare of their disease after missing a dose of therapy, and responding well to reintroduction of efalizumab after disease stablization. These case reports illustrate that this therapy can be safely reintroduced with good clinical effect in those with limited therapeutic options (Fig. 1), and support our view that this flare reaction is not an allergic hypersensitivity event.
During the flare reaction, flow cytometric analysis of circulating lymphocytes demonstrated that there was persisting saturation of the efalizumab-CD11a epitope, and there was also down-regulation of CD11a expression (Fig. 2). The possible mechanisms of psoriasis flare while on a saturating dose of efalizumab include (1) allergic hypersensitivity, (2) the development of anti-human neutralizing antibodies, (3) lowering of tissue concentration of the drug, or (4) an external trigger inducing other types of leukocytes to enter the skin via an LFA-1-independent mechanism causing inflammation. Safe and effective reintroduction of efalizumab, and the lack of eosinophils on biopsy argue for a non-hypersensitivity mechanism of the flare. If neutralizing antibodies had developed, there would not be efalizumab-CD11a epitope saturation, and further administration would not lead to skin clearing. It is difficult to correlate circulating levels and tissue levels of efalizumab, but it is possible that missing a dose might decrease tissue levels. Finally, we cannot rule out the possibility that entry of non-lymphocyte leukocytes into the skin due to an external trigger such as skin infection might lead to skin inflammation, especially in patient 1. However, the frequency of flare reactions is much less than external events, suggesting that it is an uncommon consequence.
We have also characterized this flare reaction histologically (Fig. 3), demonstrating two main points. In these cases, we found increased CD8+ dominant T cell infiltrates. However, overall there were fewer T cells compared to other psoriasis patients before efalizumab treatment. The CD8+ T cells are located mostly in the epidermis and they express the integrin αE (CD103). The αE subunit combines with the β7 subunit and binds to E-cadherin on keratinocytes, which mediates epidermal T cell trafficking. Thus CD8+ T cell trafficking during LFA-1 (a β2 integrin) blockade by efalizumab may be regulated by αEβ7 integrin, permitting epidermal entry of T cells during disease flare.
Secondly, there are abundant CD11c+ and iNOS+ cells in the dermis and epidermis during flare reactions. These markers represent a dendritic cell subset, and may be important in the pathogenesis of psoriasis. These cells may be able to enter the skin via an LFA-1 independent mechanism, and may be playing a direct role in the induction of or maintenance of psoriasis, causing the phenotypic features of erythema and hyperplasia.
Although we would always prefer the use of a single therapeutic agent where possible, the concurrent use of two biologicals of different classes can be considered in those patients with difficult-to-treat severe psoriasis vulgaris with limited therapeutic options. The main reason for caution with combination therapy is that the safety of such combinations is not yet established and so there is limited safety data. One of the main potential risks of such a combination is a decreased response to infection. Patients should be educated about this risk and seek medical attention early if they develop any new symptoms of infection. The main long-term risk of any of the biologicals includes malignancy, and this may be increased when more than one agent is used concurrently. Again, careful regular examination and screening where appropriate is warranted. The length of time a person is on two agents should be tailored to the patient.
Conclusion
In conclusion, we present two patients who experienced a flare of their severe psoriasis while on initially effective efalizumab therapy. Patient 1 still had saturation of available efalizumab-CD11a sites on circulating lymphocytes, even though he had missed a dose, and was presumably in a therapeutic range. Subsequently, both patients were able to restart efalizumab with effect, one requiring concurrent administration with infliximab. This is reassuring, given the limited therapeutic options for certain psoriasis patients. We have also characterized the histological appearance of this flare reaction, with abundant CD11c+ and iNOS+ DCs, and less CD3, CD8+ and CD103+ cells than are normally seen in psoriasis. These DCs may be playing a critical role in the psoriasis flare and possibly psoriasis.
List of abbreviations
lymphocyte function-associated antigen-1 (LFA-1)
inducible nitric oxide synthase (iNOS)
intracellular adhesion molecule-1 (ICAM-1)
dendritic cell (DCs)
Competing interests
JGK has worked as a consultant for Genentech Inc. and Serono. The other authors do not have any financial interest related to this work.
Authors' contributions
ML drafted the case report of patient 1, analyzed the flow cytometry and immunohistochemistry, and drafted the manuscript. JK supervised treatment of patient 1, analyzed the flow cytometry and immunohistochemistry, and contributed significantly to the manuscript. RB supervised treatment of patient 2, and initiated the effort to characterize the flare reaction and publish the case reports. JT treated patient 2 and drafted the case report of patient 2. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The work presented in this paper was primarily supported by National Institutes of Health (NIH) grants; a General Clinical Research Center grant (M01-RR00102) from the National Center for Research Resources at the NIH, NIH grants R01 AI-49572 and AI-49832. Written consent was obtained from the patients for publication of this study. In the Laboratory for Investigative Dermatology, The Rockefeller University, we would like to acknowledge T Kikuchi who performed the FACS studies, A Khatcherian who performed the immunohistochemistry, P Gilleaudeau and M Whalen who treated patient 1 and provided the clinical photographs.
==== Refs
Kormeili T Lowe NJ Yamauchi PS Psoriasis: immunopathogenesis and evolving immunomodulators and systemic therapies; U.S. experiences Br J Dermatol 2004 151 3 15 15270867 10.1111/j.1365-2133.2004.06009.x
Vugmeyster Y Kikuchi T Lowes MA Howell K Chamian F Kagen MH Gilleaudeau P Lee E Dummer W Pippig S Hunte B Bodary S Krueger JG Efalizumab (anti-CD11a)-induced increase in leukocyte numbers in psoriasis patients is preferentially mediated by blocked entry of memory CD8+ T cells into the skin Clin Immunol 2004 113 38 46 15380528 10.1016/j.clim.2004.06.001
Cather JC Menter A Modulating T cell responses for the treatment of psoriasis: a focus on efalizumab Expert Opin Biol Ther 2003 3 361 370 12662148 10.1517/14712598.3.2.361
Winterfield LS Menter A Gordon K Gottlieb A Psoriasis treatment: current and emerging directed therapies Ann Rheum Dis 2005 64 Suppl 2 ii87 90; discussion ii91-2 15708946 10.1136/ard.2004.032276
|
16109173
|
PMC1208875
|
CC BY
|
2021-01-04 16:29:46
|
no
|
BMC Dermatol. 2005 Aug 18; 5:9
|
utf-8
|
BMC Dermatol
| 2,005 |
10.1186/1471-5945-5-9
|
oa_comm
|
==== Front
BMC Evol BiolBMC Evolutionary Biology1471-2148BioMed Central London 1471-2148-5-441610721410.1186/1471-2148-5-44CorrespondenceQuasispecies theory in the context of population genetics Wilke Claus O [email protected] Keck Graduate Institute of Applied Life Sciences, 535 WatsonDrive, Claremont, California 91711, USA2 Digital Life Laboratory, California Institute of Technology, Mail Code 136-93, Pasadena, California 91125, USA2005 17 8 2005 5 44 44 16 2 2005 17 8 2005 Copyright © 2005 Wilke; licensee BioMed Central Ltd.2005Wilke; 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 number of recent papers have cast doubt on the applicability of the quasispecies concept to virus evolution, and have argued that population genetics is a more appropriate framework to describe virus evolution than quasispecies theory.
Results
I review the pertinent literature, and demonstrate for a number of cases that the quasispecies concept is equivalent to the concept of mutation-selection balance developed in population genetics, and that there is no disagreement between the population genetics of haploid, asexually-replicating organisms and quasispecies theory.
Conclusion
Since quasispecies theory and mutation-selection balance are two sides of the same medal, the discussion about which is more appropriate to describe virus evolution is moot. In future work on virus evolution, we would do good to focus on the important questions, such as whether we can develop accurate, quantitative models of virus evolution, and to leave aside discussions about the relative merits of perfectly equivalent concepts.
==== Body
Background
Quasispecies theory describes the evolution of an infinite population of asexual replicators at high mutation rate [1,2]. Quasispecies theory is often cited as the theory to describe the evolution of RNA viruses [3], but in recent years several authors have questioned whether quasispecies theory has any relevance for virus evolution [4-7]. Esteban Domingo has responded to this criticism from an experimentalist's point of view [8]. However, the fundamental issue in this discussion is of theoretical nature, and has not yet been addressed in detail. The fundamental issue is whether quasispecies theory and population genetics are two competing theories, and whether virology ulitmately has to decide for or against one or the other. Some quasispecies opponents have argued that quasispecies theory contradicts population genetics (e.g. "This model contrasts sharply with conventional population genetics models ..." in [7]), and that there is no evidence that favors quasispecies theory over classical population genetics [6]. On the other hand, some quasispecies proponents have also voiced the position that quasispecies theory goes beyond population genetics, and virologists in general have frequently used the term quasispecies inappropriately (see e.g. the discussion on this topic by Eigen, Ref. [9]). I find this discussion somewhat frustrating, because quasispecies theory is simply a subset of theoretical population genetics, and it is mathematically equivalent to the theory of mutation-selection balance. The only real difference between quasispecies theory and mutation-selection balance is that they have been developed largely independently by two separate schools of research, and that these schools of research have often focused on somewhat different questions and special cases. Quasispecies theory treats multiple loci, whereas early work on mutation-selection balance has focused on one- or two-locus models. On the other hand, most work on population genetics considers finite populations and includes stochastic effects, whereas quasispecies theory is first and foremost a deterministic description of infinite populations.
Quasispecies theory has its origin in a seminal paper written by Eigen in 1971 [10], in which he studied the error-prone self-replication of biological macromolecules, primarily with the goal of understanding the origin of life. However, Eigen did not yet use the term quasispecies in this 1971 paper; he coined this term in a later paper coauthored with Peter Schuster [11]. These early papers by Eigen, Schuster, and coworkers (reviewed in [1,2]) were some of the first to study theoretically the extreme nucleotide heterogeneity caused by highly error-prone replication. As a consequence, they generated interest among researchers working on RNA viruses, as these viruses were found to replicate at high mutation rates and have extremely polymorphic populations [3,12-14].
Eigen's papers also generated substantial interest among theoreticians (mostly physicists), who found the description of highly error-prone replication an interesting theoretical challenge. Unfortunately, much of the theoretical follow-up work [15-21] has focused on a particular fitness landscape, the single-peak (or master sequence) fitness landscape, in which a single sequence (the master sequence) has superior fitness 1 + s, and all other sequences have inferior fitness 1. As a result, much of the generality of Eigen's original work, as well as its connection to population genetics, have been obscured, and the conclusions of these special-case studies are frequently taken to be general predictions of quasispecies theory.
Because of the development of quasispecies theory independently from population genetics, and because of the widespread emphasis on a single fitness landscape in quasispecies theory, many authors now hold a set of beliefs about quasispecies theory that do not correspond to the actual predictions of the theory. These beliefs are:
1. Quasispecies theory is at odds with population genetics.
2. Quasispecies theory is inapplicable if populations are finite and there is neutral drift.
3. Quasispecies theory predicts an error threshold.
In the next three sections, I will address each of these points in detail. However, first I have to define what exactly I mean by quasispecies theory.
Throughout this paper, by quasispecies theory I mean specifically Eq. (6) in Ref. [11],
where xi(t) is the concentration of sequence i, Wij = AjQij is the product of the replication rate (fitness) Aj of sequence j and the mutation probability Qij from sequence j to i, and E(t) is the total production of new sequences,
In my definition of quasispecies theory, I also include straightforward generalizations of the above equation that have been used in the quasispecies literature, such as the discrete-time quasispecies equation, which can be written as [22,23]
and leads to the same steady-state solution as Eq. (1). Both Eqs. (1) and (3) can be mapped onto linear equations, and then solved by diagonalizing the matrix Wij. In both cases, the steady-state solution is given by the dominant eigenvector of Wij.
The mapping onto a linear system assumes that the Wij, which consist of the fitness landscape (as given by the Aj) and the mutation landscape (as given by the Qij), are constants. In the most general case, fitness will depend on the mutant frequencies xi, as different mutants may make use of different resources, and the relative resource concentrations change as the mutant frequencies change. It turns out that the mapping onto a linear system is still valid if resource abundances change due to external factors [24], but not if resources change in response to increasing or decreasing mutant frequencies xi. In this latter case, which corresponds to frequency-dependent selection, the conclusions drawn from quasispecies theory do not apply.
Is quasispecies theory at odds with population genetics?
Several recent papers present quasispecies theory as a theory that is alternative to (and maybe even contradictory to) standard population genetics [6,7]. Is there any merit to this position? Is quasispecies theory somehow at odds with standard population genetics?
Let us investigate what form the quasispecies equations take in a simple example. Consider a single locus with two alleles a and A, and assume that the A allele has a selective advantage s over the a allele. Further, assume that allele a mutates into allele A, and likewise allele A into allele a, with probability μ. Then, in Eq. (1), we have WAA = (1 + s)(1 - μ), WaA = (1 + s)μ, WAa = μ, Waa = 1 - μ, and hence (note that xa(t) = 1 - xA(t))
If we set the mutation rate μ to zero, then this equation turns into
that is, into the standard logistic equation that describes the rise of a beneficial allele in an otherwise homogeneous population. Thus, we can recover standard population dynamics from the quasispecies equations. Now, let us calculate the steady state solution of Eq. (4) for an arbitrary mutation rate. We set , and find
and of course xa = 1 - xA. For μ = 0, this expression becomes xA = 1, which simply means that the A allele will reach fixation in the absence of any mutation pressure. As μ increases, xA decreases, and xa increases. For a positive μ, even though the a allele is removed from the population by selection, it is constantly regenerated from the A allele by mutation pressure, and thus reaches a positive equilibrium frequency. If the mutation rate is sufficiently high, then the equilibrium frequency of the a allele, maintained by the balance of selection and mutation pressure, can be substantial. In summary, we find that for the case of a single locus with two alleles, the quasispecies model predicts logistic growth of the beneficial allele in the absence of mutations, and mutation-selection balance in the presence of mutations.
Now consider the multi-locus case. A classic paper on mutation-selection balance is the one by Kimura and Maruyama, written in 1966 [25]. In this paper, Kimura and Maruyama study the mutational load of a haploid, asexually reproducing population. I will now show that this model is also a special case of the quasispecies equations. Kimura and Maruyama assume that the frequency xi of a sequence with i mutations changes from one generation to the next according to (Eq. (3.1) in Ref. [25]):
where wi is the fitness of a sequence with i mutations, , and μ is the mutation rate (note that Kimura and Maruyama use fi instead of xi and 2M instead of μ). Now, define the mutation matrix Qij as
and write the matrix Wij in Eq. (3) as Wij = wjQij. Then, we see that E(t) as defined in Eq. (2) becomes . Furthermore, the sum ∑jWijxj in Eq. (3) runs from j = 0 to j = i, since Qij = 0 for i <j. After introducing a new index k = i - j, we can rewrite the sum as
which demonstrates that Eq. (7) follows directly from the quasispecies equation Eq. (3). As a consequence, the quasispecies model is in agreement with the Haldane-Muller principle [26], which means that the mutational load L of a population described by the quasispecies model is in many (but not all) cases approximately given by L= 1 - e-μ. (Deviations from this principle arise for example from the presence of neutral mutations [27,28].)
Now that we have seen that quasispecies theory and mutation-selection balance are equivalent, the question remains whether Eigen just reinvented parts of population genetics, or actually contributed to the development of the field. While Eigen was not the first to consider mutation-selection balance (this concept goes back to Wright and Fisher in the early 20th century), by studying multi-locus mutation-selection equations at arbitrary mutation rate he was certainly at the forefront of theoretical population genetics in the late 1970s and early 1980s. The first analytic solutions to equations of the form Eq. (1) were found by Thompson and McBride in 1974 [29] and independently by Jones et al. in 1976 [30]. These works were directly influenced by Eigen's seminal paper of 1971 [10]. On the population genetics side, Moran was the first to solve Eq. (3) [31], also in 1976, but was unaware of the work by Eigen, Thompson, McBride, Jones, and coworkers.
One of the reasons why the quasispecies model is sometimes perceived to be at odds with standard population genetics is that it predicts (under certain conditions, I should add) that the equilibrium state of the population, which is given by the dominant eigenvector of the matrix Wij, is a stable mixture of closely related mutants. This mixture of mutants, also called a mutant cloud or quasispecies, does not necessarily have to contain the fastest-replicating individual sequence that exists in the fitness landscape. In other words, sequences with high fitness can lose out against sequences with lower fitness that have better support from their mutational neighbors [32,33], an effect which has been termed survival of the flattest (Figure 1).
Figure 1 Schematic drawing of the survival-of-the-flattest effect. At low mutation rate μ, all individuals accumulate close to the top of the local fitness peak, and hence the individuals on peak A outcompete the individuals on peak B. At high mutation rate, most individuals on the steep peak A are located at low fitness values, while the individuals on the flat peak B remain close to the local optimum. As a consequence, the mean fitness of the individuals on peak B exceeds that of the individuals on peak A, and thus the former outcompete the latter.
It is important to understand that the emergence of a quasispecies is not something that has been put into the model ad hoc, but is a necessary consequence of the mutation-selection equations. We see in Eqs. (1) and (3) that the model is built on reproduction of individual sequences, but that mutations provide coupling between the different sequence types. When the mutation rate is low, then the quasispecies model predicts that the fastest-replicating sequence takes over the population, as we witness from the emergence of the logistic equation Eq. (5) at zero mutation rate. However, when the mutation rate is high, then the coupling between sequences caused by mutations can become stronger than the individual selection coefficients, and a quasispecies forms. Note that this effect will arise in any model of mutation-selection balance that correctly takes into account the coupling of different mutants at high mutation rate.
Does quasispecies theory apply to finite populations?
In the previous section, I have established that the quasispecies model is equivalent to the theory of mutation-selection balance in an infinite, haploid, asexual population. However, this equivalence does not necessarily imply that the quasispecies model applies to populations of RNA viruses, because these populations are finite. Jenkins et al. [5] argue that the total sequence space of an RNA virus is much larger than the sequence space a finite population of realistic size can cover, and that therefore the deterministic equations of the quasispecies model are inapplicable, because virus evolution is dominated by random genetic drift. A priori, this is a reasonable objection, and we have to test whether the quasispecies equations are indeed useless in any realistic setting, or whether maybe complete coverage of the sequence space is not necessary to observe quasispecies effects. (By quasispecies effects, I mean that the population behaves in a way that can only be explained through strong mutational coupling between genotypes. An example would be the observation of the survival of the flattest effect.)
First, let us have a look at some theoretical studies of finite populations that have been carried out within the context of the quasispecies model [19,34-38]. In general, in these studies the deterministic equations are taken as the starting point, and then the authors derive corrections to these equations that take into account deviations from the deterministic behavior caused by the finite population size. Thus, at least in these model systems, the deterministic equations provide a reasonable starting point to understand the population dynamics. Van Nimwegen et al. make this point particularly clear by showing that in certain cases, we can understand the behavior of a finite population from a deterministic description of a population that occupies the sequence space around a local optimum [37]. In this model, information about other local optima (which would be available to an infinite population) is not necessary to accurately describe the behavior of the finite population on the given local optimum.
However, one could still object that these models may be describing idealized and simplified situations that differ too much from the reality of RNA viruses to be of any relevance. To counter this argument, we have to ask whether there is a more general way to determine the relevance of quasispecies theory to finite populations of RNA viruses. The hallmark of quasispecies theory (and of course of any theory of mutation-selection balance) is that for a sufficiently high mutation rate, we must take into account the formation of a quasispecies to obtain a faithful description of the population dynamics. Therefore, the question is under what conditions does mutational coupling become so strong that we can observe quasispecies effects in a finite population. Can theory help us to address this question?
In the formation of a quasispecies, the population minimizes the mutational load by accumulating sequences that have a reduced probability to suffer from deleterious mutations [27,28,39,40]. This effect has been termed evolution of mutational robustness [27]. Van Nimwegen et al. studied this effect for RNA evolution, and found that mutational robustness evolved as long as the product of mutation rate μ and effective population size Ne was significantly larger than one, μNe ≫ 1 [27]. I have recently made similar observations in simulations of protein evolution [41]. What is interesting about the latter simulations is that in the regime of mutational robustness, the proteins continued to accumulate mutations, and in fact accumulated mutations faster than in the regime in which mutational robustness did not evolve. This observation demonstrates that the existence of a stable master sequence is not a necessary consequence of quasispecies evolution, in contrast to the key assumption of the study by Jenkins et al. [5]. These results can be understood with the theory of quasispecies fixation, which shows that an individual invading sequence has a positive fixation probability precisely when the mutant cloud that this sequence will spawn has higher fitness than the currently established mutant cloud, regardless of the individual fitness of the invading sequence [42,43]. Note that, in line with the previous section, this theory is again equivalent to the general theory of fixation in a haploid, asexually replicating population [44,45].
Finally, the recent paper by Comas et al. [7] also provides evidence that quasispecies effects can be observed in surprisingly small populations, populations far too small to cover the relevant sequence space. Comas et al. studied to what extent the survival-of-the-flattest effect would be affected by population size, and found that population size played only a minor role in determining the position of the critical mutation rate at which the flatter strain began to outcompete the fitter strain. (I had previously found similar results in simulated RNA evolution [46].) Even in populations of size 250, Comas et al. consistently observed outcompetition of the fitter strain by the flatter strain at high mutation rate. Note that the digital organisms in these experiments had sizes of between 54 and 272 instructions, chosen from an alphabet of 28 different instructions, so that the complete sequence space of these organisms was between 1047 and 1068 sequences large. Clearly, a population of size 250 (or even several thousand, for that matter) cannot even come close to complete coverage of such a huge sequence space.
The previous paragraphs show that on purely theoretical grounds, there is no reason to assume that quasispecies effects cannot play a role in finite populations of RNA viruses. Nevertheless, to date we have no experimental evidence that unequivocally demonstrates such effects in a specific experimental system. On the other hand, selection for specific, individual mutants is common (see e.g. Ref. [47]). What does this experimental evidence imply for quasispecies theory? First, quasispecies theory covers both cases, those in which mutational coupling can be neglected, and those in which it can't. The latter is a second-order effect that becomes relevant only when there is no strong selection for individual sequences [40]. Thus, it is not surprising that in cases where selection is strong, such as in the case of resistance or escape mutants, we don't see quasispecies effects. Second, because quasispecies effects are second-order, it may be difficult to detect them experimentally, and experiments more sensitive than the ones carried out to date may be necessary to demonstrate their presence unequivocally.
In summary, we currently have no evidence (theoretical or experimental) that contradicts the existence of quasispecies effects in finite populations of RNA viruses, but we also have no experimental evidence in favor of it. Theoretical studies and computer simulations indicate that quasispecies effects should become important when the product of effective population size and genomic mutation rate is significantly larger than one. Since for many RNA viruses the genomic mutation rate is already on the order of one [48,49], even moderately large populations of RNA viruses, or populations that undergo regular bottlenecks, are candidates for quasispecies behavior.
Does quasispecies theory predict an error threshold?
The error threshold is probably one of the most misunderstood concepts of quasispecies theory. Eigen described the error threshold in his 1971 paper as a limit to the amount of information a genome can store at a given mutation rate [10]. If the mutation rate is increased beyond this limit, then the population structure breaks down, and the population disperses over sequence space.
The first important point to understand about the error threshold is that it is a deterministic effect. This means that the position of the error threshold depends only weakly on the population size, and that even in an infinite population the error threshold occurs at a finite mutation rate [19,35]. In this way, the error threshold differs markedly from Muller's ratchet [50], which occurs only at finite population sizes and disappears in the deterministic limit. Second, the error threshold's existence and position depend on the choice of the fitness landscape [51-56]. Even though the error threshold is usually perceived as a general prediction of quasispecies theory, most of the literature that studies the error threshold considers only the single-peak fitness landscape, and disregards all other possible fitness functions [10,15-21]. The single-peak fitness landscape has the unrealistic property that all sequences have a positive replication rate, that is, there are no lethal mutants. As a result, at high mutation rates these sequences compete with the master sequence (the single sequence that has higher fitness than all other sequences), and can win this competition at sufficiently high mutation rate by sheer abundance. Can the error threshold occur in a more realistic fitness landscape that contains lethal genotypes? No. Wagner and Krall have proven mathematically that the condition for the existence of an error threshold is precisely the complete absense of lethal genotypes [53]. An intuitive explanation for this result is provided in Fig. 2.
Figure 2 Schematic drawing of the error threshold. If a fitness landscape has a positive minimum fitness (case A), then at a sufficiently high mutation rate all individuals are pushed to this minimum level. The selective strength on the narrow peak is not sufficient to counteract the mutation pressure. If a fitness landscape has no minimum fitness (case B), then the mutation pressure pushes a large fraction of the population to zero fitness. The individuals with zero fitness (shown in gray) are inviable, and thus do not compete with the individuals on the fitness peak. Therefore, a few individuals will always remain on the top of the fitness peak. Note that this conclusion holds only when two assumptions are met: (i) The population is infinite. (Otherwise, stochastic effects push the population away from the peak, and we observe Muller's ratchet.) (ii) Selection is soft, that is, only relative fitness differences matter, and the overall population size is held constant at all times. (If selection is hard, then the population size will decline as the mutation rate is increased, and eventually the population can go extinct. This case is mutational meltdown.)
There is certainly no lack of lethal mutants in viruses [57,58], and as a consequence, viruses cannot suffer from an error threshold as defined by Eigen. If this is the case, then how can we understand the concept of lethal mutagenesis, which has recently proven successful in a variety of viruses [59-62], and for which the error threshold is generally cited as the underlying theory? The truth of the matter is that the two concepts are mostly unrelated. To understand the difference between the two, we have to understand the difference between soft and hard selection. Soft selection means that the population size is held constant, regardless of the mean fitness of the population. Under soft selection, populations cannot go extinct by definition. Since the quasispecies model is usually studied in the context of soft selection (even though it can be generalized to hard selection), the error threshold per se makes no statements about population extinction. The alternative model is hard selection, where the population size will decline if the mean fitness of the population is too low. Extinction due to mutation pressure can occur under hard selection, and is usually called mutational meltdown [63-65]. Mutational meltdown will operate in any fitness landscape, as long as the population size is sufficiently small, the mutation rate sufficiently large, or the hard selection pressure sufficiently strong. While lethal mutagenesis is probably a valid antiviral therapy, referring to it as an error-threshold related effect is at best a misnomer, and can at worst lead to poor treatment decisions brought about by a misunderstanding of the actual evolutionary dynamics that unfold under lethal mutagenesis.
Finally, it is interesting to note that in certain fitness landscapes, several error-threshold-like transitions can occur one after the other as the mutation rate increases [66]. At each transition, the population loses the ability to take advantage of a particular region of sequence space, and delocalizes over this region, while remaining localized in other regions. Tannenbaum and Shakhnovich have termed this effect the error cascade [66], and in fact, the survival-of-the-flattest effect [32,33,46] can be considered as a special case of this error cascade.
Conclusion
To summarize, I have provided arguments for the following conclusions: Quasispecies theory is in perfect agreement with population genetics, it can make usefull predictions for finite populations if the product of population size and mutation rate is large, and it predicts an error threshold only for fitness landscapes that lack lethal mutants, and which therefore have little relevance for virus evolution.
However, these arguments do not imply that quasispecies theory is the final answer to all questions of virus evolution. Quasispecies theory has its short-comings that need to be addressed in future modeling work. Ironically, the biggest shortcoming of quasispecies theory, as far as I can see, does not have its origin in quasispecies theory being at odds with population genetics, but rather in quasispecies theory being too similar to the population genetics theory of asexual, haploid organisms. Viruses differ from other forms of life in that they don't have a well-defined ploidy. When a single virus particle infects a cell, the virus can be considered a haploid organism, and indeed the quasispecies model makes this assumption. However, frequently several virus particles coinfect the same cell, in which case the ploidy is given by the number of coinfecting particles. Standard population genetics has no model for such a variable-ploidy organism, and only a handful of authors have considered the theoretical implications of viral coinfection in detail [67-81].
Authors' contributions
COW carried out all aspects of this work.
Acknowledgements
I would like to thank Eddie Holmes, Santiago Elena, Igor Rouzine, and John Coffin for helpful comments on earlier versions of this manuscript. This work was in part supported by NIH grant AI 065960.
==== Refs
Eigen M McCaskill J Schuster P Molecular Quasi-Species J Phys Chem 1988 92 6881 6891 10.1021/j100335a010
Eigen M McCaskill J Schuster P The Molecular Quasi-Species Adv Chem Phys 1989 75 149 263
Domingo E Biebricher CK Eigen M Holland JJ Quasispecies and RNA Virus Evolution: Principles and Consequences 2001 Georgetown, TX: Landes Bioscience
Moya A Elena SF Bracho A Miralles R Barrio E The evolution of RNA viruses: A population genetics view Proc Natl Acad Sci USA 2000 97 6967 6973 10860958 10.1073/pnas.97.13.6967
Jenkins GM Worobey M Woelk CH Holmes EC Evidence for the non-quasispecies evolution of RNA viruses Mol Biol Evol 2001 18 987 994 11371587
Holmes EC Moya A Is the Quasispecies Concept Relevant to RNA Viruses? J Virol 2002 76 460 462 11739715 10.1128/JVI.76.1.460-462.2002
Comas I Moya A Gonzalez-Candelas F Validating viral quasispecies with digital organisms: A re-examination of the critical mutation rate BMC Evol Biol 2005 5 5 15651995 10.1186/1471-2148-5-5
Domingo E Quasispecies Theory in Virology J Virol 2002 76 463 465 10.1128/JVI.76.1.463-465.2002
Eigen M On the nature of virus quasispecies Trends Microbiol 1996 4 216 218 8795155 10.1016/0966-842X(96)20011-3
Eigen M Selforganization of matter and the evolution of biological macromolecules Die Naturwissenschaften 1971 58 456 523 10.1007/BF00623322
Eigen M Schuster P The hypercycle, a principle of natural self-organization. Part A: Emergence of the hypercycle Die Naturwissenschaften 1977 64 541 565 593400 10.1007/BF00450633
Domingo E Sabo D Taniguchi T Weissmann C Nucleotide sequence heterogeneity of an RNA phage population Cell 1978 13 735 744 657273 10.1016/0092-8674(78)90223-4
Holland J Spindler K Horodyski F Grabau E Nichol S VandePol S Rapid evolution of RNA genomes Science 1982 215 1577 1585 7041255
Domingo E Holland JJ RNA virus mutations and fitness for survival Annu Rev Microbiol 1997 51 151 178 9343347 10.1146/annurev.micro.51.1.151
Swetina J Schuster P Self-replication with errors: a model for polynucleotide replication Biophys Chem 1982 16 329 345 7159681 10.1016/0301-4622(82)87037-3
Galluccio S Exact solution of the quasispecies model in a sharply peaked fitness landscape Phys Rev E 1997 56 4526 4539 10.1103/PhysRevE.56.4526
Alves D Fontanari JF Error threshold in finite populations Phys Rev E 1998 57 7008 7013 10.1103/PhysRevE.57.7008
Campos PRA Fontanari JF Finite-size scaling of the quasispecies model Phys Rev E 1998 58 2664 2667 10.1103/PhysRevE.58.2664
Campos PRA Fontanari JF Finite-size scaling of the error threshold transition in finite populations J Phys A 1999 32 L1 L7
Tannenbaum E Deeds EJ Shakhnovich EI Equilibrium distribution of mutators in the single fitness peak model Phys Rev Lett 2003 91 138105 14525341 10.1103/PhysRevLett.91.138105
Tannenbaum E Deeds EJ Shakhnovich EI Semiconservative replication in the quasispecies model Phys Rev E 2004 69 061916 10.1103/PhysRevE.69.061916
Demetrius L Selection and evolution in macromolecular systems J Theor Biol 1983 103 619 643 6195487 10.1016/0022-5193(83)90286-2
Demetrius L Schuster P Sigmund K Polynucleotide evolution and branching processes Bull Math Biol 1985 47 239 262 4027436 10.1016/S0092-8240(85)90051-5
Wilke CO Ronnewinkel C Martinetz T Dynamic Fitness Landscapes in Molecular Evolution Phys Rep 2001 349 395 446 10.1016/S0370-1573(00)00118-6
Kimura M Maruyama T The mutational load with epistatic gene interactions in fitness Genetics 1966 54 1337 1351 17248359
Burger R The Mathematical Theory of Selection, Recombination, and Mutation 2000 Chichester: Wiley
van Nimwegen E Crutchfield JP Huynen M Neutral Evolution of Mutational Robustness Proc Natl Acad Sci USA 1999 96 9716 9720 10449760 10.1073/pnas.96.17.9716
Wilke CO Adami C Evolution of mutational robustness Mut Res 2003 522 3 11 12517406
Thompson CJ McBride JL On Eigen's Theory of Self-Organization of Matter and the Evolution of Biological Macromolecules Math Biosci 1974 21 127 142 10.1016/0025-5564(74)90110-2
Jones BL Enns RH Rangnekar SS On the theory of selection of coupled macromolecular systems Bull Math Biol 1976 38 15 28 10.1016/S0092-8240(76)80040-7
Moran PAP Global stability of genetic systems governed by mutation and selection Math Proc Camb Phil Soc 1976 80 331 336
Schuster P Swetina J Stationary mutant distributions and evolutionary optimization Bull Math Biol 1988 50 635 660 3219448 10.1016/S0092-8240(88)80059-4
Wilke CO Wang JL Ofria C Lenski RE Adami C Evolution of digital organisms at high mutation rate leads to survival of the flattest Nature 2001 412 331 333 11460163 10.1038/35085569
McCaskill JS A stochastic theory of macromolecular evolution Biological Cybernetics 1984 50 63 73 10.1007/BF00317940
Nowak M Schuster P Error Thresholds of Replication in Finite Populations–Mutation Frequencies and the Onset of Muller's Ratchet J Theor Biol 1989 137 375 395 2626057
Woodcock G Higgs PG Population evolution on a multiplicative single-peak fitness landscape J Theor Biol 1996 179 61 73 8733432 10.1006/jtbi.1996.0049
van Nimwegen E Crutchfield JP Mitchell M Statistical Dynamics of the Royal Road Genetic Algorithm Theoretical Computer Science 1999 229 41 102 10.1016/S0304-3975(99)00119-X
Rouzine IM Wakeley J Coffin JM The solitary wave of asexual evolution Proc Natl Acad Sci USA 2003 100 587 592 12525686 10.1073/pnas.242719299
Bornberg-Bauer E Chan HS Modeling evolutionary landscapes: Mutational stability, topology, and superfunnels in sequence space Proc Natl Acad Sci USA 1999 96 10689 10694 10485887 10.1073/pnas.96.19.10689
Wilke CO Adaptive Evolution on Neutral Networks Bull Math Biol 2001 63 715 730 11497165 10.1006/bulm.2001.0244
Wilke CO Molecular clock in neutral protein evolution BMC Genetics 2004 5 25 15333142 10.1186/1471-2156-5-25
Iwasa Y Michor F Nowak MA Evolutionary dynamics of invasion and escape J Theor Biol 2003 226 205 214 14643190 10.1016/j.jtbi.2003.08.014
Wilke CO Probability of Fixation of an Advantageous Mutant in a Viral Quasispecies Genetics 2003 163 467 474 12618386
Barton NH Linkage and the limits to natural selection Genetics 1995 140 821 841 7498757
Johnson T Barton NH The Effect of Deleterious Alleles on Adaptation in Asexual Populations Genetics 2002 162 395 411 12242249
Wilke CO Selection for fitness versus selection for robustness in RNA secondary structure folding Evolution 2001 55 2412 2420 11831657
Price DA Goulder PJR Klenerman P Sewell AK Easterbrook PJ Troop M Bangham CRM Phillips RE Positive selection of HIV-1 cytotoxic T lymphocyte escape variants during primary infection Proc Natl Acad Sci USA 1997 94 1890 1895 9050875 10.1073/pnas.94.5.1890
Drake JW Rates of spontaneous mutation among RNA viruses Proc Natl Acad Sci USA 1993 90 4171 4175 8387212
Drake JW Holland JJ Mutation rates among RNA viruses Proc Natl Acad Sci USA 1999 96 13910 13913 10570172 10.1073/pnas.96.24.13910
Haigh J Accumulation of deleterious mutations: Muller's ratchet Theor Popul Biol 1978 14 251 267 746491 10.1016/0040-5809(78)90027-8
Maynard Smith J Models of evolution Proc R Soc London B 1983 219 315 325
Charlesworth B Mutation-selection balance and the evolutionary advantage of sex and recombination Genet Res 1990 55 199 221 2394378
Wagner GP Krall P What is the difference between models of error thresholds and Muller's ratchet? J Math Biol 1993 32 33 44 10.1007/BF00160372
Higgs PG Error thresholds and stationary mutant distributions in multi-locus diploid genetics models Genet Res 1994 63 63 78
Wiehe T Model dependency of error thresholds: the role of fitness functions and contrasts between the finite and infinite sites models Genet Res 1997 69 127 136 10.1017/S0016672397002619
Baake E Wagner H Mutation-selection models solved exactly with methods of statistical mechanic Genet Res 2001 78 93 117 11556140 10.1017/S0016672301005110
Malpica JM Fraile A Moreno I Obies C Drake JW García-Arenal F The Rate and Character of Spontaneous Mutation in an RNA Virus Genetics 2002 162 1505 1511 12524327
Sanjuán R Moya A Elena SF The distribution of fitness effects caused by single-nucleotide substitutions in an RNA virus Proc Natl Acad Sci USA 2004 101 8396 8401 15159545 10.1073/pnas.0400146101
Pariente N Sierra S Lowenstein PR Domingo E Efficient virus extinction by combinations of a mutagen and antiviral inhibitors J Virol 2001 75 9723 9730 11559805 10.1128/JVI.75.20.9723-9730.2001
Crotty S Cameron CE Andino R RNA virus error catastrophe: Direct molecular test by using ribavirin Proc Natl Acad Sci USA 2001 98 6895 6900 11371613 10.1073/pnas.111085598
Grande-Perez A Sierra S Castro MG Domingo E Lowenstein PR Molecular indetermination in the transition to error catastrophe: Systematic elimination of lymphocytic choriomeningitis virus through mutagenesis does not correlate linearly with large increases in mutant spectrum complexity Proc Natl Acad Sci USA 2002 99 12938 12943 12215495 10.1073/pnas.182426999
Anderson JP Daifuku R Loeb LA Viral error catastrophe by mutagenic nucleosides Annu Rev Microbiol 2004 58 183 205 15487935 10.1146/annurev.micro.58.030603.123649
Gabriel W Lynch M Burger R Muller's ratchet and mutational meltdowns Evolution 1993 47 1744 1757
Lynch M Burger R Butcher D Gabriel W The mutational meltdown in asexual populations J Hered 1993 84 339 344 8409355
Lynch M Conery J Burger R Mutation accumulation and the extinction of small populations Am Nat 1995 146 489 518 10.1086/285812
Tannenbaum E Shakhnovich EI Solution of the quasispecies model for an arbitrary gene network Phys Rev E 2004 70 021903 10.1103/PhysRevE.70.021903
Nee S Maynard Smith J The evolutionary biology of molecular parasites Parasitology 1990 100 S5 S18 2172893
Bangham CRM Kirkwood TBL Defective interfering particles: effects in modulating virus growth and persistence Virology 1990 179 821 826 2238471 10.1016/0042-6822(90)90150-P
Kirkwood TBL Bangham CRM Cycles, chaos, and evolution in virus cultures: A model of defective interfering particles Proc Natl Acad Sci USA 1994 91 8685 8689 8078942
Szathmáry E Natural selection and dynamical coexistence of defective and complementing virus segments J theor Biol 1992 157 383 406 1465021
Szathmáry E Co-operation and defection: playing the field in virus dynamics J theor Biol 1993 165 341 356 8114502 10.1006/jtbi.1993.1193
Turner PE Burch CL Hanley KA Chao L Hybrid frequencies confirm limit to coinfection in the RNA bacteriophage Φ6 J Virol 1999 73 2420 2424 9971826
Turner PE Chao L Prisoner's dilemma in an RNA virus Nature 1999 398 441 443 10201376 10.1038/18913
Brown SP Collective action in an RNA virus J Evol Biol 2001 14 821 828 10.1046/j.1420-9101.2001.00317.x
Frank SA Within-host spatial dynamics of viruses and defective interfering particles J theor Biol 2000 206 279 290 10966764 10.1006/jtbi.2000.2120
Frank SA Multiplicity of infection and the evolution of hybrid incompatibility in segmented viruses Heredity 2001 87 522 529 11869342 10.1046/j.1365-2540.2001.00911.x
Wilke CO Novella IS Phenotypic mixing and hiding may contribute to memory in viral quasispecies BMC Microbiology 2003 3 11 12795816 10.1186/1471-2180-3-11
Novella IS Reissig DD Wilke CO Density-dependent selection in vesicular stomatitis virus J Virol 2004 78 5799 5804 15140977 10.1128/JVI.78.11.5799-5804.2004
Wilke CO Reissig DD Novella IS Replication at periodically changing multiplicity of infection promotes stable coexistence of competing viral populations Evolution 2004 58 900 905 15154565
Froissart R Wilke CO Montville R Remold SK Chao L Turner PE Co-infection weakens selection against epistatic mutations in RNA viruses Genetics 2004 168 9 19 15454523 10.1534/genetics.104.030205
Bretscher MT Althaus CL Müller V Bonhoeffer S Recombination in HIV and the evolution of drug resistance: for better or for worse? BioEssays 2004 26 180 188 14745836 10.1002/bies.10386
|
16107214
|
PMC1208876
|
CC BY
|
2021-01-04 16:29:16
|
no
|
BMC Evol Biol. 2005 Aug 17; 5:44
|
utf-8
|
BMC Evol Biol
| 2,005 |
10.1186/1471-2148-5-44
|
oa_comm
|
==== Front
BMC Fam PractBMC Family Practice1471-2296BioMed Central London 1471-2296-6-351611148710.1186/1471-2296-6-35Research ArticleA randomized trial of mail vs. telephone invitation to a community-based cardiovascular health awareness program for older family practice patients [ISRCTN61739603] Karwalajtys Tina [email protected] Janusz [email protected] Larry W [email protected] Cheryl [email protected] Lisa [email protected] Bea [email protected] Christopher [email protected] James E [email protected] Department of Family Medicine, McMaster University, Hamilton, Canada2 Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Canada3 Social and Public Health Services Department, City of Hamilton, Canada4 Élisabeth Bruyère Research Institute, a University of Ottawa and SCO Health Service Partnership, Ottawa, Canada5 Centre for Evaluation of Medicines, Father Sean O'Sullivan Research Centre, St. Joseph's Healthcare Hamilton, and McMaster University, Faculty of Health Sciences, Hamilton, Canada6 Division of Geriatric Medicine, McMaster University, Hamilton, Canada2005 19 8 2005 6 35 35 11 3 2005 19 8 2005 Copyright © 2005 Karwalajtys et al; licensee BioMed Central Ltd.2005Karwalajtys 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
Family physicians can play an important role in encouraging patients to participate in community-based health promotion initiatives designed to supplement and enhance their in-office care. Our objectives were to determine effective approaches to invite older family practice patients to attend cardiovascular health awareness sessions in community pharmacies, and to assess the feasibility and acceptability of a program incorporating invitation by physicians and feedback to physicians.
Methods
We conducted a prospective randomized trial with 1 family physician practice and 5 community pharmacies in Dundas, Ontario. Regular patients 65 years or older (n = 235) were randomly allocated to invitation by mail or telephone to attend pharmacy cardiovascular health awareness sessions led by volunteer peer health educators. A health record review captured blood pressure status, monitoring and control. At the sessions, volunteers helped patients to measure blood pressure using in-store machines and a validated portable device (BPM-100), and recorded blood pressure readings and self-reported cardiovascular risk factors. We compared attendance rates in the mail and telephone invitation groups and explored factors potentially associated with attendance.
Results
The 119 patients invited by mail and 116 patients contacted by telephone had a mean age of 75.7 (SD, 6.4) years and 46.8% were male. Overall, 58.3% (137/235) of invitees attended a pharmacy cardiovascular health awareness session. Patients invited by telephone were more likely to attend than those invited by mail (72.3% vs. 44.0%, OR 3.3; 95%CI 1.9–5.7; p < 0.001).
Conclusion
While the attendance in response to a telephone invitation was higher, response to a single letter was substantial. Attendance rates indicated considerable interest in community-based cardiovascular health promotion activities. A large-scale trial of a pharmacy cardiovascular health awareness program for older primary care patients is feasible.
==== Body
Background
Hypertension is a major contributor to cardiovascular disease and associated morbidity and mortality among older adults in Canada. The prevalence of hypertension increases with age and more than half of men and women 65 to 74 years of age have a systolic or diastolic blood pressure above the target level of 140/90 mmHg (or 130/80 mmHg if diabetes or target organ damage is present) recommended by the Canadian guidelines [1,2]. High blood pressure is an important and modifiable risk factor for heart disease, stroke, renal failure, peripheral vascular disease [3], and Alzheimer's disease [4]. Since elevated blood pressure is generally asymptomatic, it is important to find effective strategies to facilitate diagnosis and control of hypertension and promotion of the cardiovascular health in communities across Canada.
The diagnosis of hypertension is primarily made in the offices of family physicians [3]. Although provincial survey data indicate that nearly all older adults in Ontario visit their physician annually [5], there is evidence that one half to one third of patients have elevated blood pressures for which no diagnosis is recorded or treatment prescribed [6].
Effective diagnosis and control of hypertension depends on physicians' awareness of a patient's blood pressure status and cardiovascular risk profile over time [7]. Diagnosing hypertension can be complicated and frequently requires multiple assessments over time [8]. In-office monitoring may be limited by infrequent patient visits, co-morbid conditions, time or space constraints, remuneration and the presence of 'white coat', or 'masked' hypertension.
Since multiple, accurate blood pressure readings are needed to diagnose hypertension and regular monitoring is necessary for effective, on-going control of high blood pressure, accurate supplementary readings, a cardiovascular health profile, and more active involvement of patients in their care may help physicians to better manage hypertension in their practices. Blood pressure monitoring in a familiar, convenient community setting such as community pharmacies, with support from trained peer volunteers and delivery of patient-specific blood pressure readings and cardiovascular risk factor information to family physicians, pharmacists and the patients, could help overcome barriers to effective management [9].
Family physicians can play an important role in encouraging patients to participate in health promotion initiatives designed to supplement and enhance their in-office care. In a study by McAuley et al, a letter of invitation from the family physician was found to be an effective strategy to recruit women for routine mammography, with nearly 70% success [10]. Given that public opinion polls demonstrate that physicians are among the most trusted professionals [11], patient acceptance of new approaches to routine care may be largely dependent on their physician's endorsement and involvement. Older patients with a long relationship with their family physician may be particularly receptive to the physician's advice. A recent study recruiting older adults in family practice achieved a 70% success rate and found that the median time of patient association with the current physician was 10.2 years in the intervention group and 11.5 years in the control group, with a standard deviation of 9 years [12].
The program described here seeks to: 1) maximize use of existing resources such as community pharmacies and volunteers in cardiovascular health promotion; 2) ensure that community-based cardiovascular risk assessment is linked to family physicians who can provide follow up. The program is unique because it incorporates collection of patient-reported risk factors in addition to blood pressure readings, and 'closes the loop' by delivering results to the physician. However, there are many community programs offering blood pressure monitoring, and the idea of inviting older adults via family practices to attend a new program for cardiovascular health assessment prompted some scepticism – would they come? A first step was to determine the effectiveness of different approaches for inviting older adults to participate in pharmacy cardiovascular health awareness sessions, and the appeal of a program incorporating invitation by physicians and feedback of patient-specific results to physicians.
We conducted a randomized trial to determine the best method (mail or telephone) for family physicians' offices to invite patients to attend a session, and to identify factors that predicted attendance. In preparation for a larger scale randomized controlled trial, we also performed a cost analysis of invitation method and success rate, investigated the operational and methodological aspects of the proposed intervention, and queried patients' willingness to continue to attend community pharmacy blood pressure sessions and their preferences for the time and frequency of the sessions.
Methods
The study design was a prospective randomized trial (see Figure 1). The primary end-point was the overall attendance among patients invited by mail compared to those invited by telephone. Potential predictors of attendance, including cardiovascular risk factors recorded in patient health records, were assessed. The study was approved by the Hamilton Health Sciences / McMaster University Faculty of Health Sciences Research Ethics Board.
Figure 1 Study design and patient flow.
Eligible patients were identified from the practice roster of one family physician. Patients were eligible to participate if they were 65 years of age or older, considered by the physician to be regular patients, community dwelling, and able to attend a cardiovascular health promotion session in a local pharmacy. Patients were excluded if they suffered from dementia or a serious, non-cardiovascular disease or condition, or were non-English speaking and could not attend with an English-speaking companion. An electronic list of patients 65 years or older (n = 490) was generated using billing data. Of these, 235 met the eligibility criteria and were included in the trial. The list was reviewed and eligibility verified by the physician. The targeting of regular patients reflects the increasing move to rostered practices in Ontario and the potential importance of an established relationship with the provider, where mail/telephone contact related to health promotion would not be unusual.
Computer-generated random numbers were used to allocate eligible patients to invitation by mail (n = 119) or by telephone (n = 116) to attend one of five blood pressure monitoring sessions scheduled in five local pharmacies. To prevent potential contamination of the study arms, patients sharing the same address and/or surname were assigned identical identification numbers and thus randomly allocated to the same group. Mailed letters used an electronic signature from the physician. Patients assigned to the telephone group received a telephone call from a member of the physician's office staff who followed a protocol that limited repeat calls to three and included a structured script. Patients who were unreachable by either contact method (letters returned to sender or calls busy/unanswered) were included in the intention-to-treat analysis of attendance by invitation method in order to establish the feasibility of the program in real practice.
Research nurses performed a baseline health record review to confirm patient eligibility, collect demographic and health status factors that might influence session attendance, and document blood pressure monitoring and control. Data collected included age, gender, formal diagnosis or suspicion of hypertension, related physician comments, cardiovascular risk factors (diabetes, high cholesterol, smoking, family history), and all blood pressure readings recorded in the chart in the previous six months.
The five pharmacy blood pressure sessions were held during a 10-day period in April of 2001. Attending patients were asked to provide signed informed consent to participate. Patients were able to 'opt-out' of the feedback to the physician, however no participants in this small trial objected to having their results forwarded to the physician. At the sessions, volunteer Peer Health Educators trained by a public health nurse helped participants to measure their blood pressure using both a portable, automated device (BPM-100, VSM MedTech Ltd., Coquitlam, Canada) [13] and the in-store device, and recorded results and additional patient-reported cardiovascular risk factors. The portable device had been validated and met international standards for accuracy [13,14].
The session recording form captured, in triplicate, blood pressure readings and a checklist of 14 cardiovascular risk factors for distribution to the patient, family physician and regular pharmacist. A patient questionnaire collected demographic and general health information, current blood pressure status and history, history of related health problems, and preferences for place, time and frequency of blood pressure monitoring.
Statistical analysis
To achieve a power of 80% to detect a difference between groups of 20% or more, the significance level (alpha) was set at 0.05 and at least 103 patients were required for each study arm. The comparability of groups was established maintaining the denominator of all patients randomized (n = 235), regardless of whether or not telephone or mail contact was successful. The probability of a Type I error (alpha) was chosen to be 0.05 (two-tailed) in all analyses.
Patients allocated to the mail or telephone contact group were compared on data collected in the health record review: median age, gender, mean blood pressure, level of blood pressure, previous diagnosis of hypertension, smoking status, family history of cardiovascular disease, high cholesterol, and diabetes. The mean of the blood pressure readings recorded in the last six months, collected in the health record review, were categorized in three levels, based on whether the systolic or diastolic pressure exceeded upper boundaries.
To determine whether attending patients systematically responded differently to the invitation method, attenders and non-attenders were compared on median age, gender, mean blood pressure, and previous diagnosis of hypertension using t-tests or chi-square tests as appropriate. We examined the potential association of factors captured in the health record review (formal diagnosis or suspicion of hypertension, related physician comments, cardiovascular risk factors, blood pressure control) with attendance at the cardiovascular health promotion sessions, using chi-square tests.
To determine what combination of variables best predicted the likelihood of patient attendance at community-based cardiovascular health promotion sessions, logistic regression was performed, controlling for allocation group. Variables entered into the stepwise model included contact method, age (65–74 yr or 75+), gender, diagnosis of hypertension, and the four cardiovascular risk factors (diabetes, high cholesterol, smoking, family history) collected in the health record review.
A retrospective cost-analysis compared the cost per one patient attending in each invitation arm.
Results
Patients allocated to each of the experimental groups (mail and telephone invitation) were found to be comparable on median age, gender, mean blood pressure, level of blood pressure, previous diagnosis of hypertension, and four additional cardiovascular risk factors (see Table 1). Patients who attended a session and those who did not were also found to be comparable on median age, gender, previous diagnosis of hypertension, and four additional cardiovascular risk factors (see Table 2).
Table 1 Characteristics of 235 patients randomly allocated to mail or telephone invitation to attend a cardiovascular health promotion session.
Patient Characteristics Mail invitation n = 116(%) Telephone invitation n = 119(%)
Median Age (min 65 – max 96) Male 76 77
Female 76 75
Gender Male 55 (47.4) 55 (46.2)
Mean blood pressure in health record (mm Hg)* 136 / 71 136 / 72
Formal diagnosis of hypertension* 50 (44.2) 56 (50.0)
Health record blood pressure readings (mm Hg)* ≤ 140 and ≤90 62 (54.4) 65 (57.5)
≥140 or ≥90 to ≤160 and ≤100 38 (33.3) 36 (31.9)
≥160 or ≥100 or no readings 14 (12.3) 12 (10.6)
Cardiovascular risk factors* Smoking 47 (41.2) 37 (32.7)
Family history 33 (29.2) 36 (31.9)
High cholesterol 40 (35.4) 45 (39.8)
Diabetes 17 (15.0) 22 (19.5)
* During the health record review, 8 patients were discovered to be ineligible and 1 audit was incomplete. Therefore, the denominator for health record review data ranges from 112 to 114.
Table 2 Characteristics of the 98 non-attending patients compared to 137 patients who attended a cardiovascular health promotion session.
Patient Characteristics Non-attenders n = 98(%) Attenders n = 137(%)
Median Age (yr; 65–96) Male 76 75
Female 76 77
Gender Male 45 (45.9) 65 (47.4)
Mean blood pressure in health record (mm Hg) 135 / 71 137 / 72
Formal diagnosis of hypertension* 46 (51.7) 60 / 136 (44.1)
Family physician-reported cardiovascular risk factors* Smoking 34 (38.2) 50 (36.5)
Family history† 20 (22.5) 49 (35.8)
High cholesterol 37 (41.6) 48 (35.0)
Diabetes 14 (15.7) 25 (18.2)
* During the health record review, 8 patients were discovered to be ineligible and 1 audit was incomplete. Therefore the denominator for health record review data is 89 for the non-attenders and 136-7 for attenders
† p < 0.05
In the mail contact group, 8 letters were returned to sender. In the telephone contact group, all patients were reached in three tries or called back, and 4 were discovered to be ineligible on contact.
Overall, 58.3% (137/235) of invited patients attended a session. In the group invited by mail, 44.0% (51/116) patients attended, compared to 72.3% (86/119) among those invited by telephone. In univariate analyses, patients invited by telephone were significantly more likely to attend than were those invited by mail (OR 3.3; 95%CI 1.9–5.7; p < 0.001), and patients with a family history of cardiovascular disease noted in their chart were also significantly more likely to attend (OR 1.9; 95%CI 1.1–3.5; p = 0.049). In multivariate logistic regression modelling (n = 226), factors remaining significantly associated with attendance were invitation method (telephone) (OR 3.9, 95%CI 2.2–7.0; p < 0.001) and family history of cardiovascular disease recorded in the health record (OR 2.0, 95%CI 1.0–3.7; p = 0.38) (see Table 3).
Table 3 Univariate and multivariate analyses to detect an association between specific variables and attendance at a session.
Variable Attending (%) Univariate analyses OR (95% CI) Multivariate analyses OR (95% CI)*
Invitation Method Mail 51 (44.0) 1.0† 1.0†
Telephone 86 (72.3) 3.3 (1.9–5.7) ‡ 3.9 (2.2–7.0) ‡
Age (min 65 – max 96 yrs) 65–74 58 (58.0) 1.0†
75+ 79 (58.5) 1.0 (0.6–1.7)
Gender Female 72 (59.6) 1.0†
Male 65 (59.0) 1.1 (0.6–1.8)
Diagnosis of Hypertension§ No 76 (63.9) 1.0†
Yes 60 (56.6) 0.7 (0.4–1.3)
Family physician-reported cardiovascular risk factors|| Smoking 87 (61.3) 1.0†
50 (59.5) 0.9 (0.5–1.6)
Family history 88 (56.1) 1.0† 1.0†
49 (71.0) 1.9 (1.0–3.5) ‡ 2.0 (1.0–3.7) ‡
High cholesterol 89 (63.1) 1.0†
48 (56.5) 0.8 (0.4–1.3)
Diabetes 112 (59.9) 1.0†
25 (64.1) 1.2 (0.6–2.4)
* model based on 226 cases with data from health record review
† reference category
‡ p < 0.05
§ based on 225 cases with data from health record review
|| based on 226 cases with data from health record review
Of the patients who attended a session and completed the questionnaire (n = 130), 80.0% (95/119) indicated an interest in attending again and 70.2% (85/121) preferred a session in the morning. The preference expressed for different invitation methods was close to evenly divided (49.6% mail and 50.4% telephone) and there was a tendency for patients to state a preference for the type of invitation they had actually received, mail (72.1%; 31/43) or telephone (64.3%; 45/70), with more patients contacted by telephone attending.
Mailing costs included postage ($0.47 × 116 patients = $54.52), stationary/printing ($8.50) and research assistant time to prepare the letters and correct addressing problems (4 hrs × $12/hr = $48.00). The telephone cost was for research assistant time to prepare a patient list and calling log (2 hrs × $12/hr = $24.00), and practice staff time (6 hrs at $25/hr = $150.00). Since 51 patients invited by mail attended, the cost per successful recruitment of one patient was $111.02 / 51 = $2.18. Of patients invited by telephone, 86 attended, for a per-patient cost of $174.00 / 86 = $2.02, indicating a minimal difference in efficiency.
Discussion
The pharmacy sessions were well attended, indicating that organized community-based blood pressure monitoring with feedback to the family physician is feasible, and could enhance the diagnosis and management of hypertension by increasing the number of accurate readings available to the physician, and raising awareness among older adults of cardiovascular risk factors. The pilot demonstrated the advantage of incorporating invitation by telephone rather than mail, however the attendance rates for both groups and preferences reported by attendees demonstrated the success and acceptance of either method. Although attendance was higher in the telephone group, results also support use of mailed invitations, which are more easily implemented on a larger scale. The cost analysis demonstrated a minimal difference, and patient preference among attenders reflected the acceptability of either invitation method.
These findings are in keeping with similar studies of reminder strategies for preventive services in family practice. One study found an equal procedure completion rate among patients 15 years of age or older (n = 5883) of five preventive procedures with use of letter and telephone reminders, which was higher than that achieved through computerized physician reminders [15]. Procedure completion rates in both the letter and telephone groups were 42.0%, and letters were found to be more cost-effective.
Attendance at the Dundas sessions was higher, at 58.3% of invited patients on average, than might be expected from a single contact, particularly since participation involved attending one of a limited number of sessions held on specific days at five community pharmacies. The excellent attendance in response to a single letter or telephone call from the physician's office demonstrates that engaging older adults to participate in community-based initiatives, in cooperation with family physicians, is an effective and low-cost approach to cardiovascular health promotion.
Only two factors (invitation method and family history of cardiovascular disease) entered into the logistic regression model accounted for a significant portion of the variance in attendance, however these results are reassuring in the context of this population-health program. It was expected that male patients or very elderly patients might be less likely to attend community-based blood pressure monitoring, however, logistic regression analysis reveals that the program achieved good coverage of all patients, across the key variables of age, gender, hypertensive status, and cardiovascular risk factors.
A number of limitations of this pilot could be remedied in a larger study. For instance, the results of this study in one family physician's practice may not be generalizable to other practices, the success of one nurse in inviting patients may not be representative, and the limited time period in which the sessions were offered may have precluded attendance by some patients.
Conclusion
While the attendance in response to a telephone invitation was higher, response to a single letter was substantial. Contact by the family physicians office was effective in encouraging older patients to participate in community-based assessment of cardiovascular risk factors, with feedback of results to the family physician. The pilot provided important perspectives toward expanding the program to a larger number of family physician practices, pharmacies, and older adults.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
All authors reviewed the manuscript and contributed to its critical revision for important intellectual content. TK contributed to the study design and was responsible for the analysis and interpretation of data and preparation of the manuscript. JK contributed to the study design and supervised the analyses. LC contributed to the study design. CL was a co-investigator contributing to the study design. LD contributed to the study design and recruited pharmacists. BM contributed to the study design and was responsible for training volunteers, coordinating the sessions, and data collection. CP is a co-investigator contributing to the study design. JEW contributed to the study design and invited patients from his practice.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors would like to acknowledge the contributions of The CHAT Working Group:
Chantal Belley; Margaret Black; Kim Canary; Larry W. Chambers; Lisa Dolovich; Tom Elmslie; Barbara Farrell; Manal Guirguis-Younger; Heather L. Hall; Maureen Harmer; Alexandra Hendriks; William Hogg; Janusz Kaczorowski; Tina Karwalajtys; Crystal LaRose; Cheryl Levitt; Pamela Logan; Beatrice McDonough; Robert S. McKelvie; Rolf J. Sebaldt; Constance Sellors; Brenda Szabo; Lehana Thabane; Jennifer Thompson and Claire Zanetti.
This research was supported by grants provided by Biovail Pharmaceuticals Canada, formerly Crystaal Corporation.
==== Refs
Joffres MR Ghadirian P Fodor JG Petrasovits A Chockalingam A Hamet P Awareness, treatment, and control of hypertension in Canada Am J Hypertens 1997 10 1097 1102 9370379 10.1016/S0895-7061(97)00224-0
Zarnke KB McAlister FA Campbell NR Levine M Schiffrin EL Grover S McKay DW Myers MG Wilson TW Rabkin SW Feldman RD Burgess E Bolli P Honos G Lebel M Mann K Abbott C Tobe S Petrella R Touyz RM Canadian Hypertension Recommendations Working Group The 2001 Canadian recommendations for the management of hypertension: Part one – Assessment for diagnosis, cardiovascular risk, causes and lifestyle modification Can J Cardiol 2002 18 604 624 12107419
Feldman RD Campbell N Larochelle P Bolli P Burgess ED Carruthers SG Floras JS Haynes RB Honos G Leenen FH Leiter LA Logan AG Myers MG Spence JD Zarnke KB Task force for the development of the 1999 Canadian recommendations for the management of hypertension. 1999 Canadian recommendations for the management of hypertension CMAJ 1999 161 S1 S17 10624417
Kivipelto M Helkala EL Laakso MP Hanninen T Hallikainen M Alhainen K Soininen H Tuomilehto J Nissinen A Midlife vascular risk factors and Alzheimer's disease in later life: longitudinal, population based study BMJ 2001 322 1447 1451 11408299 10.1136/bmj.322.7300.1447
Ontario Ministry of Health and Long Term Care Ontario Health Survey 1996–1997 1999
Hooker RC Cowap N Newson R Freeman GK Better by half: hypertension in the elderly and the 'rule of halves': a primary care audit of the clinical computer record as a springboard to improving care Fam Pract 1999 16 123 128 10381016 10.1093/fampra/16.2.123
Pocock SJ McCormack V Gueyffier F Boutitie F Fagard RH Boissel JP A score for predicting risk of death from cardiovascular disease in adults with raised blood pressure, based on individual patient data from randomized controlled trials BMJ 2001 323 75 81 11451781 10.1136/bmj.323.7304.75
Hemmelgarn BR Zarnke KB Campbell NR Feldman RD McKay DW McAlister FA Khan N Schiffrin EL Myers MG Bolli P Honos G Lebel M Levine M Padwal R Canadian Hypertension Education Program Evidence-Based Recommendations Task Force The 2004 Canadian Hypertension Education Program recommendations for the management of hypertension: Part I – Blood pressure measurement, diagnosis and assessment of risk Can J Cardiol 2004 20 31 40 14968141
Chambers LW Kaczorowski J Levitt C Karwalajtys T McDonough B Lewis J Blood pressure self-monitoring in pharmacies. Building on existing resources Can Fam Physician 2002 48 1594 12449541
McAuley RG Rand C Levine M Recruiting women for breast screening Can Fam Physician 1997 43 883 888 9154360
POLLARA Public Trust Index 1998
Sellors J Kaczorowski J Sellors C Dolovich L Woodward C Willan A Goeree R Cosby R Trim K Sebaldt R Howard M Hardcastle L Poston J A randomized controlled trial of a pharmacist consultation program for family physicians and their elderly patients CMAJ 2003 169 17 22 12847034
Wright JM Mattu GS Perry TL JrGelferc ME Strange KD Zorn A Chen Y Validation of a new algorithm for the BPM-100 electronic oscillometric office blood pressure monitor Blood Press Monit 2001 6 161 165 11518840 10.1097/00126097-200106000-00008
Mattu GS Perry TL JrWright JM Comparison of the oscillometric blood pressure monitor (BPM-100 (Beta)) with the auscultatory mercury sphygmomanometer Blood Press Monit 2001 6 153 159 11518839 10.1097/00126097-200106000-00007
Rosser WW McDowell I Newell C Use of reminders for preventive procedures in family medicine CMAJ 1991 145 807 814 1913409
|
16111487
|
PMC1208877
|
CC BY
|
2021-01-04 16:29:12
|
no
|
BMC Fam Pract. 2005 Aug 19; 6:35
|
utf-8
|
BMC Fam Pract
| 2,005 |
10.1186/1471-2296-6-35
|
oa_comm
|
==== Front
BMC GastroenterolBMC Gastroenterology1471-230XBioMed Central London 1471-230X-5-261608683310.1186/1471-230X-5-26Research ArticleEpidemiologic study of chronic hepatitis B virus infection in male volunteer blood donors in Karachi, Pakistan Akhtar Saeed [email protected] Muhammad [email protected] Salman [email protected] Farrukh [email protected] Sarffraz Hussain [email protected] Department of Community Medicine and Behavioral Sciences, Faculty of Medicine, Kuwait University PO Box 24923, Safat 13110, Kuwait2 Department of Community Health Sciences, Aga Khan University, Karachi 74800, Pakistan3 Department of Pathology and Microbiology, Aga Khan University, Karachi 74800, Pakistan4 Husaini Blood Bank, Karachi, Pakistan2005 8 8 2005 5 26 26 22 1 2005 8 8 2005 Copyright © 2005 Akhtar 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 magnitude of chronic infection with hepatitis B virus (HBV) varies substantially between the countries. A better understanding of incidence and/ or prevalence of HBV infection and associated risk factors provides insight into the transmission of this infection in the community. The purpose of this investigation was to estimate the prevalence of and to identify the risk factors associated with chronic infection with HBV, as assessed by HBV surface antigen (HBsAg) positivity, in asymptomatic volunteer male blood donors in Karachi, Pakistan.
Methods
Consecutive blood donations made at the two large blood banks between January 1, 1998 and December 31, 2002 were assessed to estimate the prevalence of HBsAg positivity. To evaluate the potential risk factors, a case-control study design was implemented; cases (HBsAg positives) and controls (HBsAg negatives), were recruited between October 15, 2001 and March 15, 2002. A pre-tested structured questionnaire was administered through trained interviewers to collect the data on hypothesized risk factors for HBV infection. Sera were tested for HBsAg using commercially available kits for enzyme linked Immunosorbant assay-III.
Results
HBsAg prevalence in the male volunteer blood donors was 2.0 % (7048/351309). Multivariate logistic regression analysis showed that after adjusting for age and ethnicity, cases were significantly more likely than controls to have received dental treatment from un-qualified dental care provider (adjusted odds ratio (OR) = 9.8; 95% confidence interval (CI): 2.1, 46.1), have received 1–5 injections (adjusted OR = 3.3; 95% CI: 1.1, 9.6), more than 5 injections (adjusted OR = 1.4; 95% CI: 1.4, 12.7) during the last five years or have received injection through a glass syringe (adjusted OR = 9.4; 95% CI: 2.6, 34.3). Injury resulted in bleeding during shaving from barbers (adjusted OR = 2.3; 95% CI: 1.1, 4.8) was also significant predictor of HBsAg positivity.
Conclusion
Prevalence of HBsAg positivity in the male volunteer blood donors in Karachi was 2%. Infection control measures in health-care settings including safe injection practices and proper sterilization techniques of medical instruments and education of barbers about the significance of sterilization of their instruments may reduce the burden of HBV infection in this and similar settings. There is also an urgent need of developing locally relevant guidelines for counseling and management of HBsAg positive blood donors.
hepatitis B virushepatitis B virus surface antigen, prevalencerisk factorsvolunteer blood donorsPakistan
==== Body
Background
In the absence of effective screening programs, hepatitis B virus (HBV) is responsible for a substantial proportion of cases of post-transfusion hepatitis, liver cirrhosis and hepatocellular carcinoma [1]. An estimated 2 billion people are infected with HBV worldwide, among them 350 millions are chronic carriers: hepatitis B surface antigen (HBsAg) positive [2]. HBsAg positivity in developed countries varies from 0.6 percent in Wales, England, to 1.2 percent in Texas, USA. However, higher prevalences of infection with HBV have been reported from various parts of the developing world including 3.5% in Gaza, Palestine [3], 1.6%–7.7 % in Brazil [4,5], 19.6 % in Egypt [6], and 2%–10 % from various parts of India [7].
Intravenous drug use, needle stick injuries, haemodialysis, tattooing and multiple sexual partners have been identified as common modes of HBV transmission in the developed world [8]. In many developing countries however, the relative contributions of various routes of HBV infection have not been defined in population-based studies. Due to a lack of universal and appropriate blood screening in these countries, the risk of post-transfusion HBV infection is still unknown. Parenteral routes implicated as the most likely factors for HBV transmission include un-sterilized needles and syringes in health-care settings [9-11], Moreover, in low socio-economic settings, horizontal transmissions of HBV through contact with infected family member have also been reported [12], but these findings are yet to be verified.
The national estimates for prevalence and/or incidence of HBV infection in Pakistan are unknown. However, studies in selected groups have shown variable prevalence of chronic infection with HBV as assessed by HBsAg positivity: 7% in health professionals [13], 2%–14% in blood donors [14-17]. Pre-employment screening revealed 2.6% HBsAg positivity among the healthy individuals in northern Pakistan [18]. Moreover, some hospital-based studies have revealed that 30% – 42% of the cases of chronic liver disease [19,20] and 78% of the cases of hepatocellular carcinoma [19] were positive for HBsAg.
Developed countries have been successful in reducing the risk of HBV spread by interrupting some of the known routes of HBV transmission and through mass HBV vaccinations. The vaccine against HBV infection is available in most of the developing world including Pakistan, but its high cost limits the widespread use. Recently, Pakistan initiated universal HBV vaccination for neonates through its expanded program of immunization with the assistance of Global Alliance for Vaccines and Immunization [2]. However, public health benefits of this initiative require some time to accrue as the program focuses on neonates only [10]. Therefore, a multi-prong approach needs to be undertaken to curtail the spread of HBV infection in Pakistan and perhaps other developing countries in the region.
Volunteer blood donors are generally regarded as a healthier segment of any community, as blood banks usually have strict selection criteria that helps identify, and consequently bleed healthy donors only [21]. The proportion of HBsAg positive donors and risk factor(s) associated with HBsAg positive status among these healthy individuals may reflect the magnitude of chronic HBV infection in the general population. Therefore, we conducted this study to estimate the prevalence of and to identify the risk factors associated with HBsAg positivity in asymptomatic male volunteer blood donors in Karachi.
Methods
Study design and setting
The study design, setting and data collection procedures have been described elsewhere [22], and we briefly outlined here. This study was conducted at two large blood banks (I and II) in Karachi. Blood bank I is located in a tertiary care teaching hospital in private sector and caters for the need of inpatients. Blood bank II belongs to a non-government organization (NGO) and caters the need of all those who need blood transfusion including patients of leukaemia, haemophilia, thalassaemia and other blood related diseases. Both the blood banks receive blood donations as replacements from volunteers and admit for screening apparently healthy individuals. Preliminary screening includes a personal interview with the donor and exclusion of those who admit to known risk factors for hepatitis B and C and human immunodeficiency virus infections.
HBsAg test results of consecutive blood donations made between January 1, 1998 and December 31, 2002 at blood bank I, and from January 1, 1999 to December 31, 2002 at blood bank II were available to assess the proportions of HBsAg positive donors. To evaluate the potential risk factors for HBsAg positivity, a case-control study design was implemented.
Recruitment of cases and controls
Between October 15, 2001 and March 15, 2002, eligible blood donors aged 18–64 years, and HBsAg positive were defined as cases and those who were HBsAg negative were taken as controls. Study subjects were contacted after obtaining their addresses and phone numbers from records of respective blood banks. After explaining the objectives and potential risks/benefits of the study, subjects were invited to participate in the study.
Data collection and serology
A pre-tested structured questionnaire was used to collect data from both cases and controls regarding demographic, and socioeconomic attributes, various potential parenteral exposures to blood or blood products and non-parentral modes of HBV transmission. Both the blood banks use commercially available enzyme-linked immunosorbant assay-III (ELISA) kits to detect HBsAg as a marker for chronic HBV infection. ELISA results are interpreted essentially following the instructions of the manufacturer. This study complied with the human subjects' protection requirement of the institutional ethics committee.
Data analysis
Data were managed and analyzed using Epi-Info (version 6.04: Centres for Disease Control and Prevention, Atlanta, GA, USA) and SPSS (version 10.0: SPSS Inc., Chicago, IL, USA) respectively. Overall HBsAg prevalence and HBsAg prevalence by year and by blood bank was computed. Chi-square analysis for trend was carried out to assess the significance of the change in proportions of HBsAg positive donors over the entire study period. Descriptive statistics were computed for demographic variables for both cases and controls. To asses univariate associations between HBsAg positivity and hypothesized risk factors, odds ratios (ORs) and their corresponding 95% confidence intervals (CIs) were calculated using simple logistic regression analysis. A final set of independent risk factors for HBsAg positivity among the donors was derived by a backward stepwise logistic regression model. The risk factors significantly (P = 0.2) related with outcome on univariate analyses were considered for inclusion in the final model. After arriving at main effects model, plausible interaction terms were also evaluated for inclusion in the model. The adjusted ORs and their 95% CIs were obtained from final model and used for substantive interpretation of the model. The fit of the final model was assessed by the Hosmer-Lemeshow's goodness-of-fit test [23].
Results
HBsAg prevalence
The overall HBsAg prevalence in this study was 2.0% (7048/351309). HBsAg positivity among the donors at blood bank I was 1.0 % (782/75752), whereas, corresponding figure for the donors at blood bank II was 2.3% (6266/275557). Blood donations and proportions of HBsAg positive donors by blood bank and by year are given in Table 1. The proportions of HBsAg positive donors were consistently higher at blood bank II than blood bank I across the study period. Trend analysis revealed an overall significant (P < 0.001) increase in the proportions of HBsAg positive donors from 1998 to 2002 for combined data and for the data from blood bank I. However, for blood bank II there was significant (P < 0.001) downward trend in the proportions of HBsAg positive donors.
Characteristics of cases and controls
In this study 64 cases and 260 controls were enrolled. About 81% of cases and 86% of controls were 35 years of age or less. Forty seven percent of cases and 61% of controls belonged to Mohajir ethnicity, almost representing the composition of general population of Karachi. Fifty three percent of cases and 61% of controls had 10 or more years of schooling and 45% of cases were ever-married compared to 56% of controls. The distributions of profession, household income, number of blood donations in the past, parenteral and non-parenteral factors considered are given in Table 2.
Risk factor analysis
On univariate analyses, death of a family member due to liver disease, dental treatment received from a un-qualified provider, therapeutic injections received in the past, type of syringe used for therapeutic injections, intravenous infusions received in the past, bleeding during shaving from barbers and sexual intercourse with multiple partners were significantly (P ≤ 0.05) associated with HBsAg positivity (Table 2).
Final multivariate logistic regression model revealed that cases were significantly more likely than controls to have received dental treatment from un-qualified provider (adjusted OR = 9.8; 5% CI: 2.1, 46.1). Also, cases compared to controls were significantly more likely to have received 1–5 injections (adjusted OR = 3.3; 5% CI: 1.1, 9.6) or >5 injections (adjusted OR = 1.4; 5% CI: 1.4, 12.7) during the last five years. Furthermore, cases were significantly more likely to have been injected medicine with a glass syringe by health-care providers (adjusted OR = 9.4; 5% CI: 2.6, 34.3). Bleeding during shaving from barbers was independently and significantly associated with HBsAg positivity (adjusted OR = 2.3; 5% CI: 1.1, 4.8) in this study. The Hosmer-Lemeshow goodness-of-fit test statistic showed a good fit for the final model (χ2 = 7.82, P = 0.349) (Table 3).
Discussion
The objectives of this study were to estimate the prevalence of and to identify the risk factors for chronic infection with HBV as assessed by the HBsAg positivity through ELISA-III among volunteer blood donors in Karachi. An overall 2% HBsAg prevalence among volunteer blood donors in this study was recorded, which is lower than the figures reported from some other high HBV prevalence developing countries including 3.4% in Georgia [24], 4.5% in Taiwan [25], and 5.8% in Indonesia [26]. With HBsAg prevalence of 2%, Pakistan falls into intermediate range of HBV infection suggesting a need for screening of pregnant women at minimum and universal childhood vaccination against HBV. However, a much lower prevalence of infection with HBV has been reported from some developed countries, for instance 0.8% in Australia [27] and 1.1% in Denmark [28], which reflect the efficacy of blood donor selection policies, effective screening programs and very low HBV prevalence in general population in these countries. Similar strategies may play a significant role in reducing post-transfusion HBV infection in Pakistan.
The proportions of HBsAg positive blood donors were consistently higher at blood bank II (2.3%) than at blood bank I (1.0%) across the study period. As mentioned earlier that blood bank I receives blood donations from the relatives/friends of the hospital in-patients only, and usually patients from middle and higher socioeconomic class utilize the services of this hospital. Being a subsidiary of an NGO, blood bank II caters the need of all patients requiring blood transfusion in the city including patients from public funded hospitals, which are generally attended by the lower or middle class patients. It has been well documented that HBV infection is more prevalent in low socioeconomic settings in Indonesia [29], or perhaps similar setting in Pakistan. Therefore, the difference in the HBsAg positivity may be attributed to the apparent difference in populations attending these two blood banks. Although trend analysis revealed significant changes over time (1998–2002) in proportions of donors positive for HBsAg at blood bank I, blood bank II and for combined data but these changes do not seem to be substantial. Therefore, given the large sample size and the observational design of the study, this statistically significant change in the proportions over time may be discounted. Nonetheless, further study is indicated to discern this pattern using detailed time series data.
We found that subjects with a history of dental treatment received from unqualified provider had an increased risk for HBsAg positivity; consistent with previous research [30]. Dental practice by unregistered practitioners is common in developing countries. These non-medical personals, besides their technical incompetence, do not properly sterilize their equipments and thereby transmit blood borne infections to their patients. Strict enforcement of legislation to ban such illegal practices may result in substantial reduction in HBV transmission in this and similar settings.
Therapeutic injections received during the past five years were significantly associated with HBsAg status in this study. HBV transmission through contaminated needles is well established [31]. In Pakistan, the proportion of injections per prescription is one of the highest compared to some other countries [32-34]. In developing countries, a large proportion of patients preferred injected medicines and considered them more efficacious than other routes of drug administration [32,34]. In addition to patient preference for injections, physicians' prescribing practices, their belief in better efficacy of injected drugs, direct observation of the prescribed therapy, patients demand and financial incentive have been reported as the reasons for increased frequency of injections in developing countries [35]. Therefore, interventions to improve injection safety and reduce injection overuse would have a substantial impact on the incidence of infection with HBV.
This study showed a strong association between use of glass syringe for therapeutic injection and HBsAg status. The use of glass syringes in administering therapeutic injections has been common in the past, and is still in practice in most of the low socioeconomic settings in developing countries. The 'boiling method' employed to sterilize glass syringes in most of the health facilities is in-adequate for complete sterilization. Therefore, these glass syringes act as a source for transmitting HBV and other blood borne pathogens [7,36-38]. Taking into account the hazards of glass syringes, their use should be banned by public health authorities in Pakistan and perhaps other developing countries in the region.
Bleeding during facial shaving from barbers was significantly associated with the outcome in this study. Facial shaving from barbers has been repeatedly documented as a risk factor for transmission of hepatitis B and C viruses in various countries [39-41]. Barbers in this part of the world are mostly un-aware of the transmission of blood borne pathogens through shaving tools [41]. The repeated use of potentially contaminated razors and other non-sterilized shaving tools most likely infect their clients with HBV and other blood-borne infections. Therefore, health education programs focusing on barbers' community may contribute to the reduction in HBV transmission in this and other similar settings.
Some limitations of this study need to be considered in interpreting the present findings. Volunteer blood donors are a low-risk healthier segment of any community [42], who are further screened for symptoms of various medical conditions at the blood banks. Hence, we expect a considerable 'healthy donor effect – relatively weaker associations of the risk factors with HBsAg positivity in this population. Furthermore, subjects with long-standing HBsAg positive status could be sicker with hepatocelluar carcinoma or cirrhosis, perhaps too sick to come to donate blood and resulting into underestimation of prevalence of HBsAg positivity. Recall bias – an inherent limitation of a case-control design, might have been introduced in measurement of some of the variables, especially when past histories of injections and intravenous infusions were explored [43]. However, to minimize recall bias, we asked injection and infusion history of different eras: six months to ten years. Any bias that might have occurred must be non-differential, thus yielded conservative estimates of observed relationships. For economic reasons we were unable to test for some other markers of active HBV infection such as HBV DNA and/or hepatitis B e antigen (HBeAg), and IgM anti-HBc – an indicator of early acute HBV infection,. Therefore, some donors in this category may have been missed since some patients with acute self-limited primary HBV infection never have detectable HBsAg in the blood [44]. However, considering the sample size in this study, we hope such misclassification to be minimal. In the case-control arm of the study, we did not test controls for anti-HBs (antibodies to HBsAg) to exclude those who might have resolved past infection, became immune and lacked potential to become case as defined in this study. Ascertainment of controls' serostatus may have non-differentially led to controls who are not susceptible to HBV infection due to previous infection or immunization, thereby biasing results toward the null. Future studies should take this aspect in to account. The proportion of female donors in the selected blood banks was quite low; therefore, excluded from the present analysis. Study in female blood donors might show a different set of risk factors.
Conclusion
The prevalence of HBsAg positivity was 2% in male volunteer blood donors in this study. A history of dental treatment received from un-qualified provider, the number of therapeutic injection received during the past five years and use of glass syringe for therapeutic injections, bleeding during facial shaving from a barber were independent risk factors for HBsAg positivity in this study population. Therefore, educational intervention targeted on health-care professionals about the importance of infection control measures may include safe injection practices and proper sterilization of medical and dental instruments. Education of barbers about the significance of sterilization of their instruments may help in reducing the burden of community-acquired infection with HBV and other blood-borne pathogens in this and similar settings. Strict enforcement of legislation to ban un-qualified dental practitioners may further help curb the HBV spread. There is also an urgent need to develop locally relevant guidelines for further management of HBsAg positive donors. A cohort study of donors preferably on those who routinely volunteer should focus both on acute and well as chronic HBV infection to seek better estimate of the magnitude of the problem by incorporating relevant diagnostic tests.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SA conceived and designed the study. SA and MY designed the study questionnaire. MY, SD, FH SHJ participated in the implementation of the study. SA and MY were responsible for data analysis and manuscript writing. All the authors read and approved the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We sincerely thank the study participants. We highly appreciate the cooperation and assistance we received from the administration and staff of the blood banks. We also gratefully acknowledge the help of Dr. Mohammad Mustafa Qureshi and Dr. Hafiz Irfan in the conduct of this study. We wish to thank the reviewers of the journal for their thoughtful comments, which helped in bringing about substantive clarity in this manuscript.
Figures and Tables
Table 1 Total blood donations and proportions of hepatitis B virus (HBV) infected donors at two large blood banks in Karachi, 1998–2002
Blood bank 1 Blood bank 2 Blood banks combined
Year Total HBV+ (%)a Total HBV+ (%)b Total HBV+ (%)c
1998** 15068 117 (0.8) -- -- 15068 117 (0.8)
1999 14856 72 (0.5) 62566 1530 (2.4) 77422 1602 (2.1)
2000 15525 95 (0.6) 66486 1566 (2.4) 82011 1661 (2.0)
2001 15084 230 (1.5) 74154 1654 (2.2) 89238 1884 (2.1)
2002 15219 268 (1.8) 72351 1516 (2.1) 87570 1784 (2.0)
Total 75752 782 (1.0) 75557 6266 (2.3) 351309 7048 (2.0)
**Data for 1998 were not available for blood bank 2;aχ2trend = 170.81 (P < 0.001); bχ2trend = 21.28 (P < 0.001); cχ2trend = 123.05 (P < 0.001)
Table 2 Univariate analysis of putative risk factors for hepatitis B virus infection in a case-control study of volunteer blood donors from two large blood banks, Karachi, Pakistan
Variables Cases Controls OR† 95% CI‡
n = 64 % n = 260 %
Blood bank
Blood bank 1 38 (59.4) 118 (45.4) 1.0 -
Blood bank II 26 (40.6) 142 (54.6) 0.6 0.3–1.0
Age (completed years)
15–25 23 (35.9) 139 (53.5) 1.0 -
26–35 29 (45.3) 86 (33.1) 2.0 1.1–3.8
> 36 12 (18.8) 35 (13.5) 2.1 0.9–5.6
Ethnicity
Mohajir 30 (46.9) 158 (60.8) 1.0 -
Punjabi 8 (12.5) 45 (17.3) 0.9 0.4–2.2
Pathan 11 (17.2) 17 (6.5) 3.4 1.5–8.0
Sindhi 5 (7.8) 15 (5.8) 1.8 0.6–5.2
Baloch and others 10 (15.6) 25 (9.6) 2.1 0.9–4.8
Education (completed school years)
>10 34 (53.1) 159 (61.2) 1.0 -
≤ 10 30 (46.9) 101 (38.8) 1.4 0.8–2.5
Marital status
Never married 29 (45.3) 146 (56.2) 1.0 -
Ever married 35 (54.7) 114 (43.8) 1.5 0.9–2.7
Profession
Un-employed 8 (12.5) 50 (19.2) 1.0 -
Self-employed 19 (29.7) 39 (15.0) 3.0 1.1 – 8.6
Service provider 37 (57.8) 171 (65.8) 1.4 0.6 – 3.4
*Household income (Pak Rs./mo)
> 15000 11 (18.3) 33 (15.3) 1.0 -
8001–15000 16 (26.7) 71 (33.0) 0.7 0.3 – 1.8
5001–8000 9 (15.0) 50 (23.3) 0.5 0.2 – 1.6
≤ 5000 24 (40.0) 61 (28.4) 1.2 0.5 – 2.9
Number of times donated blood
1 21 (32.8) 93 (35.8) 1.0 -
2 13 (20.3) 52 (20.0) 1.1 0.5–2.6
>2 30 (46.9) 115 (44.2) 1.2 0.6–2.3
History of hospital admission
0 50 (78.1) 196 (75.4) 1.0 -
1 13 (20.3) 53 (20.4) 0.9 0.5–2.0
>1 1 (1.6) 11 (4.2) 0.4 0.1–2.8
History of dental treatment
Never had dental treatment 31 (48.4) 183 (70.4) 1.0 -
Treated by dentist 24 (24.0) 73 (28.1) 1.9 1.1–3.5
Treated by others 9 (9.0) 4 (1.5) 13.2 3.9–45.8
Last injection received (completed months)
Never received injection 5 (7.8) 63 (24.2) 1.0 -
≥ 6 34 (53.1) 123 (47.3) 3.5 1.3–9.3
< 6 25 (39.1) 74 (28.5) 4.3 1.5–11.8
Injections received in last one year
0 27 (42.2) 167 (64.2) 1.0 -
1 8 (12.5) 19 (7.3) 2.6 1.0–6.5
2 – 5 19 (29.7) 49 (18.8) 2.4 1.2–4.7
≥ 6 10 (15.6) 25 (9.6) 2.5 1.1–5.7
Injections received in five years
0 15 (23.4) 119 (45.8) 1.0 -
1–5 23 (35.9) 84 (32.3) 2.2 1.1–4.4
> 5 26 (40.6) 57 (21.9) 3.6 1.8–7.4
Injections received in ten years
0 40 (62.5) 167 (64.2) 1.0 -
1–10 16 (16.0) 75 (28.8) 0.9 0.5–1.7
>10 8 (8.0) 18 (6.9) 1.9 0.7–4.6
Type of syringe used for the last injection
Never had injection 8 (12.5) 92 (35.4) 1.0 -
Plastic syringe 31 (48.4) 131 (50.4) 2.7 1.1–6.2
Glass syringe 25 (39.1) 37 (14.2) 7.8 3.2–18.8
IV infusion received from
Never had drip 36 (56.3) 156 (60.0) 1.0 -
Hospital 6 (9.4) 49 (18.8) 0.5 0.2–1.3
GP's and others 22 (34.4) 55 (21.2) 1.7 0.9–3.2
IV infusion received in last one year
0 53 (82.8) 240 (92.3) 1.0 -
1 5 (7.8) 16 (6.2) 1.4 0.5–4.0
≥ 2 6 (9.4) 4 (1.5) 6.8 1.9–25.0
IV infusion received in last five years
0 47 (73.4) 202 (77.7) 1.0 -
1–4 12 (18.8) 54 (20.8) 1.0 0.5–1.9
≥ 5 5 (7.8) 4 (1.5) 5.3 1.4–20.8
IV infusions received in last ten years
0 42 (65.6) 171 (65.8) 1.0 -
1–4 15 (23.4) 79 (30.4) 0.8 0.5–1.5
≥ 5 7 (10.9) 10 (3.8) 2.9 1.0–8.0
Ear pierced
No 62 (96.9) 248 (95.4) 1.0 -
Yes 2 (3.1) 12 (4.6) 0.7 0.2–3.1
Tattooing
No 61 (95.3) 249 (95.8) 1.0 -
Yes 3 (4.7) 11 (4.2) 1.1 0.3–4.1
Shaving from barber
No 17 (26.6) 100 (38.5) 1.0 -
Yes 47 (73.4) 160 (61.5) 1.7 0.9–3.2
Frequency of going to barber
Never 17 (26.6) 100 (38.5) 1.0 -
Daily 17 (26.6) 65 (25.0) 1.5 0.7–3.2
Once a week 19 (29.7) 59 (22.7) 1.9 0.9–3.9
More than once a week 11 (17.2) 36 (13.8) 1.8 0.8–4.2
Death of a family member due to liver disease
No 55 (85.9) 24.8 (95.4) 1.0 -
Yes 9 (14.1) 12 (4.6) 3.4 1.4–8.4
Lived with jaundiced family member
No 38 (59.4) 190 (73.1) 1.0 -
Yes 26 (40.6) 70 (26.9) 1.9 1.0–3.4
History of sexual contact
No history of sexual contact 24 (37.5) 130 (50.0) 1.0 -
Sexual contact with ≥ 1 partners 40 (62.5) 130 (50.0) 1.6 0.9–3.0
• Commercial Sex worker †Odds ratio ‡Confidence Interval
•χ Who provides services like laborer, driver, policeman, factory worker etc.
Table 3 Multivariate logistic regression model of risk factors associated with hepatitis B virus infection in asymptomatic volunteer blood donors in a case-control study, Karachi, Pakistan.
Adjusted odds ratio 95 % confidence interval
History of dental treatment
Never had dental treatment 1.0 -
Treated by dentist 1.6 0.7 – 3.6
Treated by un-qualified provider 9.8 2.1 – 46.1
Number of therapeutic injection received in last five years
0 1.0 -
1–5 3.3 1.1 – 9.6
> 5 4.2 1.4 – 12.7
Type of syringe used for the last injection
Never had injection 1.0 -
Plastic 1.8 0.5 – 6.4
Glass 9.4 2.6 – 34.3
Injury resulted in bleeding during facial shaving from barber
No 1.0 -
Yes 2.3 1.1 – 4.8
==== Refs
Torbenson M Thomas DL Occult hepatitis B Lancet Infec Dis 2002 2 479 486 12150847 10.1016/S1473-3099(02)00345-6
World Health Organization Hepatitis B Fact sheet 2000 WHO/204
Yassin K Awad R Tebi AJ Queder A Laaser U Prevalence and risk factors of HBsAg in Gaza: implications for prevention and control J Infect 2002 44 252 256 12099733 10.1053/jinf.2001.0998
Arboleda M Castilho MC Fonseca JC Albuquerque BC Saboia RC Yoshida CF Epidemiological aspects of hepatitis B and D virus infection in thenorthern region of Amazonas, Brazil Trans Roy Soc Trop Med Hyg 1995 89 481 483 8560515 10.1016/0035-9203(95)90074-8
Passos AD Gomes UA Figueiredo JF do Nascimento MM de Oliveira JM Gaspar AM Yoshida CF Influence of migration on the prevalence of serologic hepatitis B markers in a rural community. 1. Analysis of prevalence by birthplace Revista De Saude Publica 1993 27 30 35 8310266
el-Sayed HF Abaza SM Mehanna S Winch PJ The prevalence of hepatitis B and C infections among immigrants to a newly reclaimed area endemic for Schistosoma mansoni in Sinai, Egypt Acta Tropica 1997 68 229 237 9386797 10.1016/S0001-706X(97)00097-1
Chowdhury A Santra A Chaudhuri S Ghosh A Banerjee P Mazumder DN Prevalence of hepatitis B infection in the general population: a rural community based study Trop Gastroenterol 1999 20 75 77 10484893
Custer B Sullivan SD Hazlet TK Iloeje U Veenstra DL Kowdley KV Global epidemiology of hepatitis B virus J Clin Gastroenterol 38 S158 S168 15602165
Aylward B Lloyd J Zaffran M McNair-Scott R Evans P Reducing the risk of unsafe injections in immunization programmes: financial and operational implications of various injection technologies Bull World Health Organ 1995 73 531 540 7554027
Usman HR Akhtar S Rahbar MH Hamid S Moattar T Luby SP Injections in health care settings: a risk factor for acute hepatitis B virus infection in Karachi, Pakistan Epidemiol Infect 2003 130 293 300 12729198 10.1017/S0950268802008178
Khan AJ Luby SP Fikree F Karim A Obaid S Dellawala S Mirza S Malik T Fisher-Hoch S McCormick JB Unsafe injections and the transmission of hepatitis B and C in a periurban community in Pakistan Bull World Health Organ 2000 78 956 963 10994278
Doganci T Uysal G Kir T Bakirtas A Kuyucu N Doganci L Horizontal transmission of hepatitis B virus in children with chronic hepatitis B World J Gastroenterol 2005 11 418 420 15637758
Shaikh MH Prevalence of HBV markers in health care personals and control J Coll Physicians Surg Pak 1995 5 19 21
Chaudhry NT Khan SJ Khan TA Saeed M Syed M Iqbal J Ahmed S Saleem M Yaqub Prevalence of hepatitis B carriers and blood group frequency in voluntary blood donors J Ayub Med Coll Abbottabad 1996 8 29 32
Rehman K Khan AA Haider Z Shahzad A Iqbal J Khan RU Ahmad S Siddiqui A Syed SH Prevalence of seromarkers of HBV and HCV in health care personnel and apparently healthy blood donors J Pak Med Assoc 1996 46 152 154 8993043
Kakepoto GN Bhally HS Khaliq G Kayani N Burney IA Siddiqui T Khurshid M Epidemiology of blood-borne viruses: a study of healthy blood donors in Southern Pakistan Southeast Asian J Trop Med Public Health 1996 27 703 706 9253870
Tareen S Eslick GD Kam EP Byles JE Durrani AB Maree SM High prevalence of hepatitis B virus (HBV) among male blood donors in a developing country: urgent need for systematic screening Scand J Infect Dis 2002 34 712 713 12374375 10.1080/00365540210147804
Khokhar N Gill ML Malik GJ General seroprevalence of hepatitis C and hepatitis B virus infections in population J Coll Physicians Surg Pak 2004 14 534 536 15353136
Tong CY Khan R Beeching NJ Tariq WU Hart CA Ahmad N Malik IA The occurance of hepatitis B virus and HCV in Pakistani patients with chronic liver disease and hepatocellular carcinoma Epidemiol Infect 1996 117 327 332 8870630
Khan TS Rizvi FJ Hepatitis B seropositivity among chronic liver disease patients in Hazara division Pakistan J Ayub Med Coll Abbottabad 2003 15 54 55
Pillonel J Saura C Courouce AM Prevalence of HIV, HTLV, and hepatitis B and C viruses in blood donors in France, 1992–1996 Transfus Clin Biol 1998 5 305 312 9836391 10.1016/S1246-7820(98)85001-5
Akhtar S Younus M Adil S Jafri SH Hassan F Hepatitis C virus infection in asymptomatic male volunteer blood donors in Karachi, Pakistan J Viral Hepat 2004 11 527 535 15500553 10.1111/j.1365-2893.2004.00518.x
Hosmer DW JrLemeshow S Applied Logistic Regression 1989 New York: John Wiley & Sons 307
Butsashvili M Tsertsvadze T McNutt LA Kamkamidze G Gvetadze R Badridze N Prevalence of hepatitis B, hepatitis C, syphilis and HIV in Georgian blood donors Eur J Epidemiol 2001 17 693 695 12086085 10.1023/A:1015566132757
Tsai SJ Chen CJ Hsieh YS Yang CS Seroprevalence of anti-HCV among voluntary blood donors in Taiwan Zhonghua Min Guo Wei Sheng Wu Ji Mian Yi Xue Za Zhi 1991 4 61 70
Budihusodo U Sulaiman HA Akbar HN Lesmana LA Waspodo AS Noer HM Akahane Y Suzuki H Seroepidemiology of HBV and HCV infection in Jakarta, Indonesia Gastroenterol Jpn 1991 26 196 201 1909267
Correll PK Law MG Seed CR Gust A Buring M Dax EM Keller AJ Kaldor JM Variant Creutzfeldt-Jakob disease in Australian blood donors: estimation of risk and the impact of deferral strategies Vox Sang 2001 81 6 11 11520409 10.1046/j.1423-0410.2001.00056.x
Christensen PB Titlestad IL Homburg KM Georgsen J Kristensen T Hepatitis B core antibodies in Danish blood donors: a surrogate marker of risk behaviour Vox Sang 2001 81 222 227 11903997 10.1046/j.0042-9007.2001.00113.x
Akbar N Basuki B Mulyanto M Garabrant DH Sulaiman A Noer HM Ethnicity, socioeconomic status, transfusions and risk of hepatitis B and hepatitis C infection J Gastroenterol Hepatol 1997 12 752 757 9430042
Cisneros-Castolo M Hernandez-Ruiz L Ibarra-Robles IE Fernandez-Garate IH La Pena JE Prevalence of hepatitis B virus infection and related risk factors in a rural community of Mexico Am J Trop Med Hyg 2001 65 759 763 11791971
Kane A Lloyd J Zaffran M Simonsen L Kane M Transmission of hepatitis B, hepatitis C and human immunodeficiency viruses through unsafe injections in the developing world: model-based regional estimates Bull World Health Organ 1999 77 801 807 10593027
Simonsen L Kane A Lloyd J Zaffran M Kane M Unsafe injections in the developing world and transmission of blood-borne pathogens: a review Bull World Health Organ 1999 77 789 800 10593026
Thaver IH Prescribing patterns of primary care providers in squatter areas of Karachi J Pak Med Assoc 1995 45 301 302 8920612
Janjua NZ Akhtar S Hutin YJF Injection use in two districts of Pakistan: implications for disease prevention Int J Qual Health Care 2005
Reeler AV Anthropological perspectives on injections: a review Bull Wrld Hlth Org 2000 78 135 143
Ko YC Li SC Yen YY Yeh SM Hsieh CC Horizontal transmission of hepatitis B virus from siblings and intramuscular injection among preschool children in a familial cohort Am J Epidemiol 1991 133 1015 1023 2035502
Singh J Bhatia R Gandhi JC Kaswekar AP Khare S Patel SB Oza VB Jain DC Sokhey J Outbreak of viral hepatitis B in a rural community in India linked to inadequately sterilized needles and syringes Bull World Health Organ 1998 76 93 98 9615501
Voelker R Eradication efforts need needle-free delivery JAMA 1999 281 1879 1881 10349879 10.1001/jama.281.20.1879
Mele A Corona R Tosti ME Palumbo F Moiraghi A Novaco F Galanti C Bernacchia R Ferraro P Beauty treatments and risk of parenterally transmitted hepatitis: results from the hepatitis surveillance system in Italy Scand J Infect Dis 1995 27 441 444 8588131
Yee LJ Weiss HL Langner RG Herrera J Kaslow RA Leeuwen DJV Risk factors for acquisition of hepatitis C virus infection: a case series and potential implications for disease surveillance BMC Infect Dis 2001 1 8 11518542 10.1186/1471-2334-1-8
Bari A Akhtar S Rahbar MH Luby SP Risk factors for hepatitis C virus infection in male adults in Rawalpindi-Islamabad, Pakistan Trop Med Int Health 2001 6 732 738 11555441 10.1046/j.1365-3156.2001.00779.x
Crook PD Jones ME Hall AJ Mortality of hepatitis B surface antigen-positive blood donors in England and Wales Int J Epidemiol 2003 32 118 124 12690022 10.1093/ije/dyg039
Schlesselman JJ Stolley PD Case-control studies: Design, Conduct, Analysis 1982 New York: Oxford University Press
Robinson WS Mandell GL, Nemmett JE, Dolin R Hepatitis B virus and Hepatitis D virus Principles and Practice of Infectious Diseases 1995 New York: Churchill Livingstone 1406 1439
|
16086833
|
PMC1208878
|
CC BY
|
2021-01-04 16:03:27
|
no
|
BMC Gastroenterol. 2005 Aug 8; 5:26
|
utf-8
|
BMC Gastroenterol
| 2,005 |
10.1186/1471-230X-5-26
|
oa_comm
|
==== Front
BMC GenetBMC Genetics1471-2156BioMed Central London 1471-2156-6-431609822610.1186/1471-2156-6-43Research ArticleA screen for proteins that interact with PAX6: C-terminal mutations disrupt interaction with HOMER3, DNCL1 and TRIM11 Cooper Simon T [email protected] Isabel M [email protected] University of Edinburgh, School of Molecular and Clinical Medicine, Medical Sciences (Medical Genetics), Molecular Medicine Centre, Western General Hospital, Crewe Road Edinburgh EH4 2XU2005 12 8 2005 6 43 43 24 1 2005 12 8 2005 Copyright © 2005 Cooper and Hanson; licensee BioMed Central Ltd.2005Cooper and Hanson; 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 PAX6 protein is a transcriptional regulator with a key role in ocular and neurological development. Individuals with heterozygous loss-of-function mutations in the PAX6 gene have malformations of the eye and brain. Little is known about the interactions of PAX6 with other proteins, so we carried out a systematic screen for proteins that interact with PAX6.
Results
We used bioinformatics techniques to characterise a highly conserved peptide at the C-terminus of the PAX6 protein. Yeast two-hybrid library screens were then carried out to identify brain-expressed proteins that interact with the C-terminal peptide and with the entire PAX6 proline-serine-threonine-rich domain. Three novel PAX6-interacting proteins were identified: the post-synaptic density (PSD) protein HOMER3, the dynein subunit DNCL1, and the tripartite motif protein TRIM11. Three C-terminal PAX6 mutations, previously identified in patients with eye malformations, all reduced or abolished the interactions.
Conclusion
Our preliminary data suggest that PAX6 interacts with HOMER3, DNCL1 and TRIM11. We propose that the interaction of PAX6 with HOMER3 and DNCL1 is a mechanism by which synaptic activation could lead to changes in neuronal transcriptional activity, and that some of the neural anomalies in patients with PAX6 mutations could be explained by impaired protein-protein interactions.
==== Body
Background
The PAX6 protein is a member of the PAX (paired-box) family of transcriptional regulators and is essential for normal ocular and neural development [1]. Heterozygous mutations of the human PAX6 gene cause aniridia (absence of the iris) and a range of other congenital eye malformations [2]. Neural defects such as foveal hypoplasia and optic nerve hypoplasia are common in PAX6-associated eye disease [3-5]. Homozygous mutations in man and mouse are lethal and result in severe developmental abnormalities including anophthalmia, severe reduction of the olfactory structures and gross brain malformations [2,6]. The roles of PAX6 in brain development have mainly been studied in homozygous mutant mice or rats and include arealisation of the cerebral cortex [7], formation of the prosencephalon-mesencephalon boundary [8], axon guidance [8], differentiation of neurons from glia [9] and neuronal migration in the cerebellum [10].
The discovery of multiple and diverse roles for PAX6 in brain development prompted MRI analyses of aniridia patients, and a range of distinctive brain anomalies were uncovered. The most common and striking of these was absence or hypoplasia of the anterior commissure [11]. Other defects included absence or hypoplasia of the pineal gland, cortical polymicrogyria, white matter changes in the corpus callosum and grey matter changes in the cerebellum [11-13]. Functional changes included hyposmia and abnormal inter-hemispheric auditory transfer [11,14].
The defining feature of all PAX proteins is the presence of a 128 amino acid DNA-binding paired domain that has been highly conserved over evolution [1]. In addition to the paired domain, PAX6 also contains a DNA-binding homeodomain and a proline, serine and threonine-rich (PST) domain at the C-terminus [1,6]. The PST domain, which encompasses the C-terminal 145 amino acids of PAX6, has been shown to act as a transcriptional activator [6]. The PAX6 protein directly regulates a wide range of target genes [1,2] including Pax2 [15], Ngn2 [16] and glucagon [17].
The Pax6 gene has a spatially and temporally complex expression pattern in the eye, brain, nasal structures, spinal cord and pancreas [1]. Although PAX6 is clearly involved in multiple developmental processes, common themes are now emerging concerning the role of PAX6 in neural tissues. Gradients of Pax6 expression are important for determining positional characteristics in the retina [18] and the neocortex [7]. PAX6 plays a role in development of specific axonal connections between the retina and the brain [18] and within the forebrain [8,19]. It is also involved in the differentiation of neural cell types from multipotent precursors in the retina [16] and the cerebral cortex [9] through activation of bHLH genes such as Ngn2 and Mash1. These studies provide a clear link between PAX6 function in the retina and the brain, and are of particular relevance to the neurological phenotypes of individuals with PAX6 mutations.
It is becoming apparent that transcription factors do not act in isolation but are dependent on interactions with other proteins to carry out their function [20,21]. These interactions introduce more specificity into the regulatory function of a given transcription factor. To date only three PAX6 protein-protein interactions have been described: with SOX2 on the lens-specific enhancer element of the δ-crystallin gene [22]; with MDIA, which modulates PAX6 activity in early neuronal development [23], and with MAF proteins on the glucagon promoter, which causes increased expression of this pancreatic hormone gene [17].
Here we report the preliminary results of the first systematic screen for proteins that interact with PAX6. We used sequence alignment algorithms and secondary structure prediction programs to define a new domain of 32 amino acids at the C-terminal end of the PAX6 protein. We then screened a brain library with this peptide using the yeast two-hybrid technique and identified three novel interacting proteins, HOMER3, DNCL1 and TRIM11. The interaction between PAX6 and these proteins was disrupted by naturally occurring C-terminal PAX6 mutations.
Results
A highly conserved C-terminal PAX6 peptide
We and others [31-33] noted that there is significant sequence conservation at the C-terminal end of the PAX6 protein. BLAST analysis of the amino acid sequence of the PAX6 PST domain (aa 278–422) revealed a highly conserved motif within the last 28 amino acids (beginning at the 'GLISP' motif, Figure 1a). Strong conservation was seen in distantly related species such as axolotl (Ambystoma mexicanum) and sea urchin (Paracentrotus lividus) (Figure 1a).
Figure 1 Characterisation of the PAX6 C-terminal peptide. (a) CLUSTAL alignment of the terminal 42 amino acids of PAX6 from diverse species. (*) indicates invariant residues, (:) indicates highly similar substitutions and (.) indicates moderately similar substitutions. (b) Secondary structure analysis of the highly conserved terminal 28 amino acids predicts a single beta sheet (arrow) in the SVPVQ peptide. (c) When 4 residues are added in the N-terminal direction, 2 beta sheets are now predicted.
We subjected the whole PAX6 PST domain to secondary structure analysis using JPRED, a program that uses a number of different protein structure prediction algorithms to generate a consensus secondary structure (Figure 1) [25,26]. The PST domain was largely devoid of predicted secondary structure except for the C-terminal region, which contained two predicted beta sheets within the highly conserved domain, one in the 'GLISP' motif and one in the 'SPVPQ' motif (identical to the pattern shown in Figure 1c). Initially we defined the C-terminal domain as running from the 'GLISP' motif up to the stop codon, since this region was most highly conserved and contained strongly predicted secondary structure elements. However when we performed secondary structure prediction analysis on this 28 amino acid peptide, the first beta sheet was lost (Figure 1b). Addition of another 4 amino acids (TTST) immediately before 'GLISP' caused recovery of the first beta sheet (Figure 1c). Although these 4 residues are not highly conserved (Figure 1a), they appear to be important for seeding the first beta sheet and therefore for secondary structure. Thus we define the C-terminal peptide as being the final 32 amino acids of PAX6, running from threonine 391 to the stop codon (top line of Figure 1c).
Yeast two-hybrid screening
We hypothesised that the C-terminal peptide might be involved in protein-protein interactions, and we tested this by screening a cDNA library using the yeast two-hybrid system with a construct (PAX6CTP) in which the 32 amino acid C-terminal peptide was fused to the yeast GAL4 DNA binding domain. We chose to screen a mouse brain cDNA library as no human libraries were available. Given the fact that the amino acid sequence of the PAX6 protein is identical in man and mouse, we reasoned that a mouse brain library would yield relevant interactors. For comparison we also carried out the screen using a construct containing the whole PST domain (PAX6PST).
The C-terminal peptide screen gave 15 colonies that were positive with all three reporters and the PST domain screen gave 62 colonies. The interacting plasmids were isolated and the cDNA inserts sequenced. Three cDNAs were identified 3 or more times, Homer3 (NM_011984), Dncl1 (Dynein cytoplasmic light chain 1, NM_019682, also known as Pin or Dlc8) and Trim11 (Tripartite motif protein family member 11, NM_053168). Homer3 (6 clones) and Dncl1 (2 clones) were identified in the C-terminal peptide screen. Homer3 (7 clones), Dncl1 (2 clones) and Trim11 (6 clones) were identified in the PST domain screen. All cDNA inserts were in-frame with the coding region of the pPC86 GAL4 activation domain. None of the cDNAs was present in a list of known false positives [34].
HOMER3 is a member of the HOMER family of neuronal post-synaptic density (PSD) proteins [35]. DNCL1 is a subunit of two motor protein complexes, dynein and myosin-Va, both of which are involved in intracellular trafficking of proteins and organelles in neurons [36,37]. TRIM11 is a member of the tripartite motif protein family and contains a RING finger, a B-box zinc finger, a coiled coil domain and a B30.2 domain [38]. The possible significance of the interactions between these proteins and PAX6 is discussed below.
Semi-quantitative PCR
To check that the Homer3, Dncl1 and Trim11 clones were not identified multiple times solely because they are highly abundant in the library, we performed a semi-quantitative PCR assay. We compared the relative abundance of Homer3, Dncl1, Trim11 and Pax6 with Gapdh and Atp5a1. Gapdh and Atp5a1 both show strong constitutive expression in a variety of tissues including the brain [29,30]. Homer3, Dncl1, Trim11 and Pax6 were only amplified strongly after 35 cycles of PCR (Figure 2) and were therefore present at relatively low levels compared to Gapdh (amplified strongly after 25 cycles) and Atp5a1 (amplified strongly after 30 cycles; Figure 2).
Figure 2 Semi-quantitative PCR analysis of Homer3, Dncl1, Trim11 and Pax6 in the mouse brain cDNA library. Library cDNA was amplified with primers specific for Pax6 (600 bp), Homer3 (485 bp band), Dncl1 (485 bp) and Trim11 (609 bp) for 20, 25, 30 or 35 cycles. Atp5a1 (415 bp) and Gapdh (450 bp), which are highly expressed in the brain, are included for comparison. M indicates the Φ×174 HaeIII DNA size marker; the positions of the 603 bp and 310 bp marker bands are indicated.
We concluded that Homer3, Dncl1 and Trim11 clones were not highly abundant in the library. This is consistent with the idea that they were pulled out because the encoded proteins interact specifically with the C-terminal peptide or PST domain of PAX6.
Yeast two-hybrid pairwise interactions
By library screening we identified two proteins (HOMER3 and DNCL1) that interact with the C-terminal peptide and three proteins (HOMER3, DNCL1 and TRIM11) that interact with the whole PST domain. This suggests that HOMER3 and DNCL1 interact specifically with the C-terminal peptide while TRIM11 interacts with a more N-terminal part of the PST domain. We conducted pairwise tests between specific constructs to confirm the interactions identified in the library screen and to further investigate the interaction between PAX6 and HOMER3, DNCL1 and TRIM11. The Dncl1 and Trim11 clones that were pulled out of the library were full-length, but the Homer3 cDNAs lacked the N-terminal 70 amino acids. The missing coding region was inserted into the truncated Homer3 cDNA to give a full-length expression construct (see Methods). Pairwise interactions were carried out with both the full-length and truncated Homer3 clones.
We confirmed that the whole PAX6 PST domain interacts with HOMER3 (full-length and truncated constructs), TRIM11 and DNCL1, as all three reporter genes were strongly activated in pairwise tests (Figure 3; Table 1). In contrast the interaction between the C-terminal peptide and HOMER3 or DNCL1 could not be confirmed with pairwise tests (Table 1). Occasionally, partial suppression of growth on plates containing 5-fluoro-orotic acid was observed, indicating low-level activation of the URA3 reporter; however HIS3 and LacZ activation were not observed. The reasons for this are not clear, although it may be that the pairwise tests were of sub-optimal sensitivity compared to the library screen. However we were able to confirm that the C-terminal peptide is important for the interaction with HOMER3 and DNCL1 because interaction with the PAX6PST-CT construct, which lacks the final 32 amino acids, was completely abolished (Figure 3, Table 1). TRIM11 interacted equally well with PAX6PST and PAX6PST-CT (Figure 3, Table 1). This is consistent with the library screens in which TRIM11 was isolated with the PST domain but not with the C-terminal peptide and supports the idea that the C-terminal peptide is not important for the interaction between PAX6 and TRIM11.
Figure 3 LacZ reporter gene activation in pairwise tests. pPC86 constructs are shown across the top, and pDBLeu constructs are shown down the right hand side. PAX6PST, PAX6 PST domain; PAX6PST/Q422R, PAX6 PST domain with the Q422R mutation; PAX6PST/X423L, PAX6 PST domain with the X423L mutation, PAX6PST/1615del10, PAX6 PST domain with the 1615 del10 mutation; PAX6PST-CT, PAX6 PST domain minus the C-terminal peptide. 'Truncated HOMER3' is HOMER3 lacking the N-terminal 70 amino acids. Five control strains are shown for comparison (left). These range from non-interactor (A) to strong interactor (E).
Table 1 Pairwise interaction tests between normal and mutant PAX6 constructs and HOMER3, DNCL1 and TRIM11. +++: strong interaction; ++: moderate interaction; +: weak interaction; (+) borderline interaction with one or two reporters activated at very low levels; 0: no interaction. PAX6PST, PAX6 PST domain; PAX6CTP, PAX6 C-terminal peptide; PAX6PST-CT, PAX6 PST domain minus the C-terminal peptide; PAX6PST/Q422R, PAX6 PST domain with the Q422R mutation; PAX6PST/X423L, PAX6 PST domain with the X423L mutation, PAX6PST/1615del, PAX6 PST domain with the 1615 del10 mutation. HOMER3-FL, HOMER3 full-length clone; HOMER3-Tr, truncated HOMER3 clone lacking the N-terminal 70 amino acids.
pDBLeu constructs pPC86 constructs
HOMER3-FL HOMER3-Tr DNCL1 TRIM11
PAX6PST +++ ++ ++ ++
PAX6CTP 0 0 0 0
PAX6PST-CT (+) 0 0 ++
PAX6PST/Q422R ++ + + ++
PAX6PST/X423L + 0 (+) ++
PAX6PST/1615del10 (+) 0 0 ++
Having confirmed that HOMER3, DNCL1 and TRIM11 interact with the PAX6 PST domain, we next investigated how the interactions were affected by three C-terminal PAX6 mutations that have been previously described in patients with ocular anomalies. The first mutation is a single nucleotide substitution 1627A>G that causes a glutamine to arginine amino acid substitution in the last codon of PAX6. This missense mutation (Q422R) has been reported in two patients, one affected by anterior segment dysgenesis with uveal ectropion and one with typical aniridia and foveal hypoplasia [27,33].
The second mutation (1615del10) was found in an aniridia family [28]. This frame-shifting deletion occurs just before the PAX6 stop codon and is predicted to cause translational read-through into the 3' untranslated region, generating a protein in which the last 5 amino acids of the C-terminal peptide are replaced by a 103 amino acid-extension. Affected individuals in this family showed unusual neurobehavioural traits including impaired social cognition and poor verbal inhibition [28]. MRI analysis revealed grey matter abnormalities in the frontal lobe, temporal lobe and cerebellum, and white matter deficits in the corpus callosum [13].
The third mutation 1629insT (X423L) has been reported in several aniridia patients [11,12,27,33]. Insertion of a single T nucleotide at position 1629 changes the stop codon (TAA) to a leucine codon (TTA) and generates a full length PAX6 protein with a C-terminal extension that extends for a further 35 amino acids into the 3' untranslated region. MRI analysis of six patients with this mutation revealed variable brain defects including absence or hypoplasia of the anterior commissure, pineal gland and olfactory bulbs [12]. Two patients had temporal polymicrogyria, one in association with epilepsy [12].
The three mutations were introduced into the PAX6PST construct, and pairwise tests were carried out to investigate the interaction of each mutant protein with HOMER3, DNCL1 and TRIM11. All three mutations had a clear effect on the interactions. The most subtle mutation (Q422R) caused a reduction in the interaction with HOMER3 and DNCL1 (Figure 3, Table 1). The C-terminal extension mutations X423L and1615del10 mutations both dramatically reduced or abolished the interaction with HOMER3 and DNCL1 (Figure 3, Table 1).
None of the three mutations affected the interaction with TRIM11 which again is consistent with the hypothesis that TRIM11 interacts with a more N-terminal part of the PST domain.
Discussion
On the basis of secondary structure predictions and amino acid sequence conservation, we defined a novel PAX6 protein domain, which we have called the C-terminal peptide. We performed yeast two-hybrid library screens with the C-terminal peptide and the whole PST domain and we identified three novel interacting proteins, HOMER3, DNCL1 and TRIM11. In library screens, HOMER3 and DNCL1 interacted with the C-terminal peptide and the PST domain while TRIM11 interacted only with the PST domain, suggesting that HOMER3 and DNCL1 specifically interact with the C-terminal peptide while TRIM11 interacts with a more N-terminal part of the PST domain. The interactions between the PST domain and HOMER3, DNCL1 and TRIM11 were confirmed in pairwise tests. We were not able to confirm the interaction between HOMER3 or DNCL1 with the C-terminal peptide construct in pairwise tests, but we showed that the C-terminal peptide was important for PAX6/HOMER3 or PAX6/DNCL1 interaction because HOMER3 and DNCL1 did not interact with a PST domain construct lacking the C-terminal peptide.
HOMER3 is found in the PSD of neurons and directly binds to type I metabotropic glutamate receptors, which act via phospholipase C to stimulate IP3-mediated release of Ca2+ from intracellular vesicles [35,39]. HOMER3 is a member of the HOMER family of proteins that are constitutively expressed in the brain and play a role in post-synaptic signalling and receptor trafficking by forming multivalent links with various receptors and PSD scaffolding proteins [35,39-41]. HOMER proteins have also been implicated in axon guidance during brain development [42].
DNCL1 is a subunit of two intracellular transport protein complexes, dynein and myosin Va [36]. Dynein and myosin Va are involved in the microtubule-based and actin-based movement respectively of proteins, organelles and vesicles in neurons [36,37]. Myosin Va is enriched in the PSD [43], and DNCL1 binds to a variety of PSD proteins including guanylate kinase domain-associated protein [44] and neuronal nitric oxide synthase [45].
TRIM11 is a member of the mouse tripartite motif protein family (also known as the RBCC family), and contains the three characteristic structural motifs of this protein family, a RING finger, a B-box zinc finger, and a coiled coil domain, as well as a B30.2 domain that is found in many but not all TRIM proteins [38]. TRIM11 interacts with Humanin, a protein that suppresses the neurotoxicity associated with Alzheimer's disease [46]. TRIM11 lowers Humanin levels by a mechanism that appears to involve ubiqutin-mediated proteasomal degradation [46].
At present our data must be considered preliminary because the interactions have not been confirmed by any other approach. However it is interesting to speculate that the interaction of PAX6 with HOMER3 and DNCL1 may be the basis of a mechanism by which synaptic signalling causes changes in gene expression. We propose that PAX6 is sequestered in the PSD by binding to HOMER3. Since receptor activation causes dissociation of HOMER proteins [35], PAX6 may be released as a result of synaptic activity, allowing it to interact with DNCL1. The PAX6/DNCL1 complex could then participate in myosin Va-mediated transport along the PSD-associated actin cytoskeleton followed by dynein-mediated transport along the microtubule network, eventually reaching the nucleus [36]. Since TRIM11 is implicated in protein degradation [46], it may play a role in PAX6 protein turnover.
Precedents for association of transcription factors with the post-synaptic density include STAT3 and CREB, which act as messengers between the synapse and the nucleus [47,48]. Although the full-length PAX6 protein is predominantly nuclear, there is good evidence in mouse, quail and nematode for an isoform that lacks the paired domain, is both nuclear and cytoplasmic, and binds DNA through the homeodomain alone [49-51]. In mice the paired-less isoform is relatively abundant in brain [49]; however its subcellular localisation in neurones, and the possibility of an association with the PSD, remains to be investigated.
Regarding the subcellular localisation of the putative interacting proteins, HOMER3 is predominantly found at the interface between the PSD and the cytoplasm [39], DNCL1 is chiefly cytoplasmic, although nuclear localisation has been reported [51], and TRIM11 is both nuclear and cytoplasmic [38]. Therefore cytoplasmic PAX6 could potentially interact with all three proteins, while nuclear PAX6 could interact with TRIM11 and DNCL1. At the tissue level, HOMER3 expression has been detected in thymus and lung but it has mainly been studied in brain where it is found in the forebrain, hippocampus and cerebellum [39]. DNCL1 and TRIM11 both have wide expression domains that include the brain [44,38]. Thus the expression of all three interactors overlaps with PAX6 at the tissue level [2,7,8,10]. We detected co-expression of PAX6, HOMER3, DNCL1 and TRIM11 by RT-PCR in human adult brain RNA (IH, L Harrison and A Brown, data not shown).
We demonstrated that the interaction between the PST domain and HOMER3 or DNCL1 was impaired by three naturally occurring mutations that are located in the PAX6 C-terminal peptide. The Q422R mutation, which involves a glutamine to arginine substitution at the last amino acid position of PAX6, caused a reduction in the interaction with HOMER3 and DNCL1. The X423L and 1615del10 mutations severely reduced or completely abolished the interaction with HOMER3 and DNCL1. The predicted effect of the X423L and 1615del10 mutations is to cause translation into the 3' untranslated region, thus generating proteins with abnormal extensions that might be expected to disrupt the conformation of the C-terminal end of the protein. The 1615del10 mutation also removes the last 5 amino acids of the C-terminal peptide [28].
None of the mutations affected the interaction with TRIM11, suggesting that they do not alter the conformation of the more N-terminal part of the PST domain. All our data are consistent with the hypothesis that TRIM11 does not interact with the C-terminal peptide, but interacts with the PST domain between the homeodomain and the C-terminal peptide.
Most aniridia patients are heterozygous for mutations that introduce a premature termination codon into the PAX6 open reading frame [2,52]. These alleles would be expected to encode truncated proteins or no protein at all if the mutant RNA is degraded by nonsense-mediated decay [52]. We propose that the brain anomalies that have been observed in aniridia patients may be partly explained by impaired interaction between PAX6 and HOMER3, DNCL1 and TRIM11. The neurobehavioural phenotype associated with 1615del10 and the polymicrogyria associated with X423L may result from a specific effect of these unusual C-terminal extension mutations. There is evidence that signalling and transport mechanisms that were initially characterized in the brain may also be conserved in the retina, suggesting that impaired PAX6 protein-protein interactions may also have implications for the retinal defects observed in individuals with PAX6 mutations [53,54].
Conclusion
We have presented preliminary evidence that the neurodevelopmental transcriptional regulator PAX6 interacts with HOMER3, DNCL1 and TRIM11. We suggest that the interaction of PAX6 with HOMER3 and DNCL1 is a mechanism by which synaptic signalling could lead to regulated changes in gene expression in neurons. We also propose that some of the neural anomalies in patients with PAX6 mutations may be explained by impaired protein-protein interactions.
Methods
Bioinformatics techniques
Sequence database searches were carried out using the BLAST program available through the Bioinformatics Applications at the Rosalind Franklin Centre for Genomics Research [24]. Protein sequences that were highly homologous to the C-terminus of human PAX6 were aligned using CLUSTAL [24]. Secondary structure prediction was performed using the JPRED consensus method [25,26].
Yeast two-hybrid constructs
All cDNA and amino acid numbering is based on the human PAX6 cDNA and protein reference sequences available from the Human PAX6 Allelic Variant Database web site [27]. Standard PCR and subcloning techniques were used to make three PAX6 cDNA constructs in the pDBLeu expression vector (ProQuest Two-Hybrid System, Invitrogen), which generates a protein fused to the yeast GAL4 DNA binding domain. PAX6PST contains the whole PST domain (amino acids 278–422 of the full-length PAX6 protein). Primers were ST001 (forward) 5'-AAA AGT TCG ACT GCC AGC AAC ACA CCT AGT C-3' and ST005 (R) 5'-TTT TGC GGC TTT TTA CTG TAA TCT TGG CCA GTA TTG-3'. PAX6CTP contains the newly defined C-terminal peptide alone (391–422). Primers were ST004 (F) 5'-AAA AGT CGA CTA CCA CTT CAA CAG GAC TCA TT-3' and ST005 (R). PAX6PST-CT contains the PST domain minus the C-terminal peptide (278–390). This was made by cutting PAX6PST with NdeI and NotI to drop out the C-terminal peptide, and inserting a synthetic linker between the two restriction sites. All fragments generated by PCR or with linkers were sequenced to check that no errors had been introduced.
A PAX6 PST domain construct containing the mutation 1627A>G (Q422R) was generated in the same way as the PAX6PST construct, but using the reverse PCR primer ST006 5'-TTT TGC GGC CGC TTT TTA CCG TAA TCT TGG CCA GTA TTG AG-3', which contains the mutant nucleotide substitution (underlined).
cDNA sequences containing the mutations 1615del10 [28] and 1629insT (X423L) [12] were generated by PCR from reverse transcribed RNA (a gift from Dr K Williamson and Prof V van Heyningen). Primers were ST015 (F) 5'-CCC ACA TAT GCA GAC ACA C-3' and ST031 (R) 5'-TTG CGG CCG CAT CCA TCC AGT CTA CAT TGT TC-3'. The PAX6PST construct was cut with NdeI and NotI to release the normal C-terminal peptide, and the mutant sequence was inserted.
Yeast two-hybrid library screens
A mouse brain cDNA library (ProQuest, Invitrogen) was screened with the PAX6PST and PAX6CTP pDBLeu constructs. The library was constructed in the pPC86 vector, which produces proteins fused to the yeast GAL4 activation domain. The system uses three GAL4-activated reporter genes, HIS3, URA3 and lacZ, to identify positive interactions. Reporters are activated when the bait protein fused to the GAL4 DNA binding domain (pDBLeu) interacts with the prey protein fused to the GAL4 activation domain (pPC86), thus reconstituting GAL4 function. HIS3 activation allows growth on plates lacking histidine. Weak URA3 activation suppresses growth on plates containing 5-fluro-orotic acid while strong URA3 activation permits growth on plates lacking uracil. LacZ activation causes X-gal to turn blue in a beta-galactosidase assay. All assays were carried out in parallel with the five ProQuest control yeast strains A-E, which range from non-interactor (A) to strong interactor (E).
All procedures were carried out according to the supplier's protocols. Briefly, chemically competent MaV203 yeast cells were co-transformed with the cDNA library and the bait plasmid pDBLeu PAX6CTP or pDBLeu PAX6PST. Transformants (5 × 107) were plated on medium lacking histidine to check for HIS3 activation. HIS3 positives were then assayed for all 3 reporters. The pPC86 prey plasmid was isolated from all HIS3/URA3/LacZ positives and the cDNA insert sequenced. BLAST searches were performed to identify the cDNA insert [24].
Pairwise interactions
Specific interactions were tested by transforming competent MaV203 yeast cells with one bait construct (in pDBLeu) and one prey construct (in pPC86) and testing the resulting colonies for activation of the HIS3, URA3 and LacZ reporters as before.
To create a full-length Homer3 clone for pairwise tests, a cDNA clone (IMAGE clone 3602414, accession number BE569374) containing the missing N-terminal 70 amino acids was identified by a BLAST search of the EST nucleotide sequence database and obtained from the Rosalind Franklin Centre for Genomics Research. The missing fragment was amplified from the IMAGE clone by PCR and inserted into the pPC86-Homer3 plasmid to create a full-length expression construct.
Semi-quantitative PCR
To check the relative representation of clones in the cDNA library, semi-quantitative PCR was performed on Pax6, Homer3, Dncl1, Trim11 and the constitutively expressed genes Gapdh and Atp5a1 [29,30]. Primers were designed to cross at least one intron, so that only correctly spliced clones were amplified. Primer sequences were: Pax6-F CAG CCA AAA TAG ATC TAC CTG; Pax6-R CGA TCA CAT GCT CTC TCC TT; Homer3-F CCC AGG TGG CTG TAG AGC; Homer3-R CTC TAC ACA GTG CAA AGC TCA G; Trim11-F GTG CAG GAT GTG AAG CTG; Trim11-R GCC TGC AGA TAG TCA TAG GG; Dncl1-F CAA AAA TGC AGA CAT GTC G; Dncl1-R CTA AGG GAG AAA AAA ATG GGG; Gapdh-F: CAT CAC CAT CTT CCA GGA GC; Gapdh-R: ATG ACC TTG CCC ACA GCC TT; Atp5a1-F: CAC ACG TGA GAT GTC CTC CA; Atp5a1-R: CAC AGA GAT TCG GGG ATA A. 10 ng library cDNA were amplified in a reaction containing 1xAmpliTaq polymerase buffer (Perkin Elmer), 1.5 mM MgCl2, 200μM each primer and 2.5 units of AmpliTaq polymerase (Perkin Elmer). PCR conditions were (95°C for 30 sec) × 1, (94°C for 30 sec, 55°C for 30 sec, 72°C for 30 sec) × 32 and (72°C for 2 min) × 1. Products were resolved on a 2.5% agarose gel with ΦX174/HaeIII size markers (Promega).
Abbreviations
PST domain, proline-, serine- and threonine-rich domain; PSD, post-synaptic density; PCR, polymerase chain reaction.
Authors' contributions
STC carried out the bioinformatic analyses and all experimental work. IMH conceived, designed and supervised the study, and obtained funding. The manuscript was prepared jointly by STC and IMH, who have both read and approved the final version.
Acknowledgements
We gratefully acknowledge Dr A Brown for technical advice on the yeast two-hybrid system and the Rosalind Franklin Centre for Genomics Research for supplying the Homer3 IMAGE clone. STC was supported by Fight for Sight and IMH was supported by a Career Development Award from the UK Medical Research Council.
==== Refs
Simpson TI Price DJ Pax6, a pleiotropic player in development Bioessays 2002 24 1041 1051 12386935 10.1002/bies.10174
van Heyningen V Williamson KA PAX6 in sensory development Hum Mol Genet 2002 11 1161 1167 12015275 10.1093/hmg/11.10.1161
Nelson LB Spaeth GL Nowinski TS Margo CE Jackson L Aniridia. A review Surv Ophthalmol 1984 28 621 642 6330922 10.1016/0039-6257(84)90184-X
Azuma N Nishina S Yanagisawa H Okuyama T Yamada M PAX6 missense mutation in isolated foveal hypoplasia Nat Genet 1996 13 141 142 8640214 10.1038/ng0696-141
Azuma N Yamaguchi Y Handa H Tadokoro K Asaka A Kawase E Yamada M Mutations of the PAX6 gene detected in patients with a variety of optic-nerve malformations Am J Hum Genet 2003 72 1565 1570 12721955 10.1086/375555
Glaser T Jepeal L Edwards JG Young SR Favor J Maas RL PAX6 gene dosage effect in a family with congenital cataracts, aniridia, anophthalmia and central nervous system defects Nat Genet 1994 7 463 471 7951315 10.1038/ng0894-463
Bishop KM Rubenstein JL O'Leary DD Distinct actions of Emx1, Emx2, and Pax6 in regulating the specification of areas in the developing neocortex J Neurosci 2002 22 7627 7638 12196586
Mastick GS Davis NM Andrew GL Easter SS Pax-6 functions in boundary formation and axon guidance in the embryonic mouse forebrain Development 1997 124 1985 1997 9169845
Heins N Malatesta P Cecconi F Nakafuku M Tucker KL Hack MA Chapouton P Barde YA Gotz M Glial cells generate neurons: the role of the transcription factor Pax6 Nature Neurosci 2002 5 308 315 11896398 10.1038/nn828
Engelkamp D Rashbass P Seawright A van Heyningen V Role of Pax6 in development of the cerebellar system Development 1999 126 3585 3596 10409504
Sisodiya SM Free SL Williamson KA Mitchell TN Willis C Stevens JM Kendall BE Shorvon SD Hanson IM Moore AT van Heyningen V PAX6 haploinsufficiency causes cerebral malformation and olfactory dysfunction in humans Nat Genet 2001 28 214 216 11431688 10.1038/90042
Mitchell TN Free SL Williamson KA Stevens JM Churchill AJ Hanson IM Shorvon SD Moore AT van Heyningen V Sisodiya SM Polymicrogyria and absence of pineal gland due to PAX6 mutation Ann Neurol 2003 53 658 663 12731001 10.1002/ana.10576
Ellison-Wright Z Heyman I Frampton I Rubia K Chitnis X Ellison-Wright I Williams SC Suckling J Simmons A Bullmore E Heterozygous PAX6 mutation, adult brain structure and fronto-striato-thalamic function in a human family Eur J Neurosci 2004 19 1505 1512 15066147 10.1111/j.1460-9568.2004.03236.x
Bamiou DE Musiek FE Sisodiya SM Free SL Davies RA Moore A van Heyningen V Luxon LM Deficient auditory interhemispheric transfer in patients with PAX6 mutations Ann Neurol 2004 56 503 509 15389894 10.1002/ana.20227
Schwarz M Cecconi F Bernier G Andrejewski N Kammandel B Wagner M Gruss P Spatial specification of mammalian eye territories by reciprocal transcriptional repression of Pax2 and Pax6 Development 2000 127 4325 4334 11003833
Marquardt T Ashery-Padan R Andrejewski N Scardigli R Guillemot F Gruss P Pax6 is required for the multipotent state of retinal progenitor cells Cell 2001 105 43 55 11301001 10.1016/S0092-8674(01)00295-1
Planque N Leconte L Coquelle FM Benkhelifa S Martin P Felder-Schmittbuhl MP Saule S Interaction of Maf transcription factors with Pax-6 results in synergistic activation of the glucagon promoter J Biol Chem 2001 276 35751 35760 11457839 10.1074/jbc.M104523200
Baumer N Marquardt T Stoykova A Ashery-Padan R Chowdhury K Gruss P Pax6 is required for establishing the naso-temporal and dorsal characteristics of the optic vesicle Development 2002 129 4535 4545 12223410
Jones L Lopez-Bendito G Gruss P Stoykova A Molnar Z Pax6 is required for the normal development of the forebrain axonal connections Development 2002 129 5041 5052 12397112
Oikawa T Yamada T Molecular biology of the Ets family of transcription factors Gene 2003 303 11 34 12559563 10.1016/S0378-1119(02)01156-3
Westerman BA Murre C Oudejans CB The cellular Pax-Hox-helix connection Biochim Biophys Acta 2003 1629 1 7 14522074
Kamachi Y Uchikawa M Tanouchi A Sekido R Kondoh H Pax6 and SOX2 form a co-DNA-binding partner complex that regulates initiation of lens development Genes Dev 2001 15 1272 1286 11358870 10.1101/gad.887101
Tominaga T Meng W Togashi K Urano H Alberts AS Tominaga M The Rho GTPase effector protein, mDia, inhibits the DNA binding ability of the transcription factor Pax6 and changes the pattern of neurite extension in cerebellar granule cells through its binding to Pax6 J Biol Chem 2002 277 47686 47691 12324464 10.1074/jbc.M207539200
Rosalind Franklin Centre for Genomics Research BioinformaticsApplications
JPRED, a Consensus Method for Protein Secondary Structure Prediction
Cuff JA Barton GJ Application of multiple sequence alignment profiles to improve protein secondary structure prediction Proteins 2000 40 502 511 10861942 10.1002/1097-0134(20000815)40:3<502::AID-PROT170>3.0.CO;2-Q
The Human PAX6 Allelic Variant Database
Heyman I Frampton I van Heyningen V Hanson I Teague P Taylor A Simonoff E Psychiatric disorder and cognitive function in a family with an inherited novel mutation of the developmental control gene PAX6 Psychiatr Genet 1999 9 85 90 10412187
Patel P Boyd CA Johnston DG Williamson C Analysis of GAPDH as a standard for gene expression quantification in human placenta Placenta 2002 23 697 698 12361689 10.1053/plac.2002.0859
Pickard BS Malloy MP Clarke L LeHellard S Ewald H Mors O Porteous DJ Blackwood DHR Muir WJ Candidate psychiatric illness genes identified in patients with pericentric inversions of chromosome 18 Psy Gen
Glardon S Callaerts P Halder G Gehring WJ Conservation of Pax-6 in a lower chordate, the ascidian Phallusia mammillata Development 1997 124 817 825 9043063
Mikkola I Bruun JA Bjorkoy G Holm T Johansen T Phosphorylation of the transactivation domain of Pax6 by extracellular signal-regulated kinase and p38 mitogen-activated protein kinase J Biol Chem 1999 274 15115 15126 10329718 10.1074/jbc.274.21.15115
Singh S Chao LY Mishra R Davies J Saunders GF Missense mutation at the C-terminus of PAX6 negatively modulates homeodomain function Hum Mol Genet 2001 10 911 918 11309364 10.1093/hmg/10.9.911
Table of False Positives
Xiao B Tu JC Worley PF Homer: a link between neural activity and glutamate receptor function Curr Opin Neurobiol 2000 10 370 374 10851183 10.1016/S0959-4388(00)00087-8
Vale RD The molecular motor toolbox for intracellular transport Cell 2003 112 467 480 12600311 10.1016/S0092-8674(03)00111-9
Espindola FS Suter DM Partata LB Cao T Wolenski JS Cheney RE King SM Mooseker MS The light chain composition of chicken brain myosin-Va: calmodulin, myosin-II essential light chains, and 8-kDa dynein light chain/PIN Cell Motil Cytoskeleton 2000 47 269 281 11093248 10.1002/1097-0169(200012)47:4<269::AID-CM2>3.0.CO;2-G
Reymond A Meroni G Fantozzi A Merla G Cairo S Luzi L Riganelli D Zanaria E Messali S Cainarca S Guffanti A Minucci S Pelicci PG Ballabio A The tripartite motif family identifies cell compartments EMBO J 2001 20 2140 2151 11331580 10.1093/emboj/20.9.2140
Xiao B Tu JC Petralia RS Yuan JP Doan A Breder CD Ruggiero A Lanahan AA Wenthold RJ Worley PF Homer regulates the association of group 1 metabotropic glutamate receptors with multivalent complexes of homer-related, synaptic proteins Neuron 1998 21 707 716 9808458 10.1016/S0896-6273(00)80588-7
Tu JC Xiao B Naisbitt S Yuan JP Petralia RS Brakeman P Doan A Aakalu VK Lanahan AA Sheng M Worley PF Coupling of mGluR/Homer and PSD-95 complexes by the Shank family of postsynaptic density proteins Neuron 1999 23 583 592 10433269 10.1016/S0896-6273(00)80810-7
Tu JC Xiao B Yuan JP Lanahan AA Leoffert K Li M Linden DJ Worley PF Homer binds a novel proline-rich motif and links group 1 metabotropic glutamate receptors with IP3 receptors Neuron 1998 21 717 726 9808459 10.1016/S0896-6273(00)80589-9
Foa L Rajan I Haas K Wu GY Brakeman P Worley P Cline H The scaffold protein, Homer1b/c, regulates axon pathfinding in the central nervous system in vivo Nat Neurosci 2001 4 499 506 11319558
Walikonis RS Jensen ON Mann M Provance DW JrMercer JA Kennedy MB Identification of proteins in the postsynaptic density fraction by mass spectrometry J Neurosci 2000 20 4069 4080 10818142
Naisbitt S Valtschanoff J Allison DW Sala C Kim E Craig AM Weinberg RJ Sheng M Interaction of the postsynaptic density-95/guanylate kinase domain-associated protein complex with a light chain of myosin-V and dynein J Neurosci 2000 20 4524 4534 10844022
Rodriguez-Crespo I Straub W Gavilanes F Ortiz de Montellano PR Binding of dynein light chain (PIN) to neuronal nitric oxide synthase in the absence of inhibition Arch Biochem Biophys 1998 359 297 304 9808772 10.1006/abbi.1998.0928
Niikura T Hashimoto Y Tajima H Ishizaka M Yamagishi Y Kawasumi M Nawa M Terashita K Aiso S Nishimoto I A tripartite motif protein Trim11 binds and destabilizes Humanin, a neuroprotective peptide against Alzheimer's disease-relevant insults Eur J Neurosci 2003 17 1150 1158 12670303 10.1046/j.1460-9568.2003.02553.x
Murata S Usuda N Okano A Kobayashi S Suzuki T Occurrence of a transcription factor, signal transducer and activators of transcription 3 (Stat3), in the postsynaptic density of the rat brain Brain Res Mol Brain Res 2000 78 80 90 10891587 10.1016/S0169-328X(00)00077-2
Suzuki T Usuda N Ishiguro H Mitake S Nagatsu T Okumaura-Noji K Occurrence of a transcription factor, cAMP response element-binding protein (CREB), in the post-synaptic sites of the brain Brain Res Mol Brain Res 1998 30 69 77 9795144 10.1016/S0169-328X(98)00199-5
Mishra R Gorlov IP Chao LY Singh S Saunders GF PAX6, paired domain influences sequence recognition by the homeodomain J Biol Chem 2002 277 49488 49494 12388550 10.1074/jbc.M206478200
Carriere C Plaza S Caboche J Dozier C Bailly M Martin P Saule S Nuclear localization signals, DNA binding, and transactivation properties of quail Pax-6 (Pax-QNR) isoforms Cell Growth Differ 1995 6 1531 1540 9019158
Zhang Y Ferreira HB Greenstein D Chisholm A Emmons SW Regulated nuclear entry of the C. elegans Pax-6 transcription factor Mech Dev 1998 78 179 187 9858725 10.1016/S0925-4773(98)00171-3
Kaiser FJ Tavassoli K Van den Bemd GJ Chang GT Horsthemke B Moroy T Ludecke HJ Nuclear interaction of the dynein light chain LC8a with the TRPS1 transcription factor suppresses the transcriptional repression activity of TRPS1 Hum Mol Genet 2003 12 1349 1358 12761050 10.1093/hmg/ddg145
Vincent MC Pujo AL Olivier D Calvas P Screening for PAX6 gene mutations is consistent with haploinsufficiency as the main mechanism leading to various ocular defects Eur J Hum Genet 2003 11 163 169 12634864 10.1038/sj.ejhg.5200940
Brandstatter JH Dick O Boeckers TM The postsynaptic scaffold proteins ProSAP1/Shank2 and Homer1 are associated with glutamate receptor complexes at rat retinal synapses J Comp Neurol 2004 475 551 563 15236236 10.1002/cne.20194
Libby RT Lillo C Kitamoto J Williams DS Steel KP Myosin Va is required for normal photoreceptor synaptic activity J Cell Sci 2004 117 4509 4515 15316067 10.1242/jcs.01316
|
16098226
|
PMC1208879
|
CC BY
|
2021-01-04 16:30:39
|
no
|
BMC Genet. 2005 Aug 12; 6:43
|
utf-8
|
BMC Genet
| 2,005 |
10.1186/1471-2156-6-43
|
oa_comm
|
==== Front
BMC GenetBMC Genetics1471-2156BioMed Central London 1471-2156-6-441610217610.1186/1471-2156-6-44Research ArticleTests for the replication of an association between Egfr and natural variation in Drosophila melanogaster wing morphology Palsson Arnar [email protected] James [email protected] Ian [email protected] Greg [email protected] Department of Genetics' North Carolina State University, Raleigh, NC 27695, USA2 Department of Ecology and Evolution, University of Chicago, Chicago, IL 60637, USA3 The Department of Biochemistry, University of Sussex, Brighton, BN1 9QG, UK2005 15 8 2005 6 44 44 17 3 2005 15 8 2005 Copyright © 2005 Palsson et al; licensee BioMed Central Ltd.2005Palsson 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
Quantitative differences between individuals stem from a combination of genetic and environmental factors, with the heritable variation being shaped by evolutionary forces. Drosophila wing shape has emerged as an attractive system for genetic dissection of multi-dimensional traits. We utilize several experimental genetic methods to validation of the contribution of several polymorphisms in the Epidermal growth factor receptor (Egfr) gene to wing shape and size, that were previously mapped in populations of Drosophila melanogaster from North Carolina (NC) and California (CA). This re-evaluation utilized different genetic testcrosses to generate heterozygous individuals with a variety of genetic backgrounds as well as sampling of new alleles from Kenyan stocks.
Results
Only one variant, in the Egfr promoter, had replicable effects in all new experiments. However, expanded genotyping of the initial sample of inbred lines rendered the association non-significant in the CA population, while it persisted in the NC sample, suggesting population specific modification of the quantitative trait nucleotide QTN effect.
Conclusion
Dissection of quantitative trait variation to the nucleotide level can identify sites with replicable effects as small as one percent of the segregating genetic variation. However, the testcross approach to validate QTNs is both labor intensive and time-consuming, and is probably less useful than resampling of large independent sets of outbred individuals.
==== Body
Background
Elucidation of the specific genetic variants that underlie natural phenotypic variation constitutes a major challenge for evolutionary geneticists. Our understanding of evolution will remain incomplete until the relative proportions of deleterious, (nearly) neutral and adaptive factors are documented, in terms of number of loci, their individual and joint effects as well as mode of expression [1]. Several practical issues complicate this endeavor. First, assessment of the contribution of loci and nucleotide variants can be confounded by chance effects, leading to inflated estimates [2]. Second, precise assessment of the effects of segregating polymorphisms on phenotypes depends critically on accurate mapping of the variants, down to individual quantitative trait nucleotides (QTN). Third, environmental interaction, epistasis and pleiotropy, all add complexity to the architecture of genetic variation[1,3].
Most common implementations of quantitative trait locus (QTL) mapping have low bias with respect to genomic coverage, but only identify allelic variation between two strains. In model organisms, these approaches allow assessment of marginal and epistatic effects, since the experiments are conducted with a large number of offspring, often in laboratory settings that reduce environmental variance. In practice, QTL are rarely resolved to individual loci or exact causal genetic variants [3-5], although several studies on plants offer exceptions [6,7] (reviewed in [8]). In D. melanogaster, QTL loci have also been dissected with quantitative complementation tests [9,10] and/or by linkage disequilibrium (LD) mapping involving a candidate region or locus. These approaches have the resolution to establish a significant contribution of allelic variation at single genes [9,11-20] and even specific nucleotides [21-23].
Successful implication of allelic and nucleotide variation in candidate genes in the production of phenotypic variation is aided by low amounts of LD, due to substantial historical recombination, in the fly genome. LD mapping in D. melanogaster can be implemented with varying degrees of control over genetic and environmental variance from wild caught individuals, laboratory reared iso-female lines, inbred strains, chromosome extraction lines and strains with introgressed chromosome regions. It is now clear that the power and resolution of association studies varies among organisms according to the extent of haplotype structure, and that different experimental approaches must be taken to verify associations in each organism. Despite the lesson from LD mapping in humans that extensive repetition, across cohorts and populations, is crucial to verify allelic contributions [24,25], replication of associations in model organisms is almost non-existent. More research into genetic approaches to validation of QTN effects is needed.
Drosophila wing shape has been used extensively as a model for the study of integration of developmental and quantitative genetics [26,27] and for analysis of the evolution of clinal variation in morphology [28-30]. More specifically, wing shape has proven to be an amenable system for studies on developmental modularity and integration [31], developmental stability [32], selection responses [33-35], laboratory adaptation [36] and more recently for the quantitative genetic dissection of patterning [23,37-41]. Wing shape is commonly described by geometric morphometric tools [42] that capture variation in the locations of landmarks at junctions of veins, cross-veins and the wing margin. The veins have a stereotypical configuration in the Sophophoran family of Drosophilids, with only minor differences documented between species [43], but diversity of shape is considerable [44,45]. Wing shape is highly polygenic [26,33,34,46] and we proposed that the spacing and length of veins is a major source of this variation [47].
QTL mapping and quantitative complementation tests support the involvement of venation loci, including components of the EGFR/Ras pathway, in naturally occurring wing shape variation [38,41]. These observations led us to test association between allelic variation in the Egfr locus and shape, by sequencing ~11 kb of the locus in 210 inbred lines from two North American localities, NC and CA [23,48]. Significant association of six polymorphisms in Egfr with aspects of wing shape and size, either as main effects or by interaction with population or sex, were reported. A follow-up with wild caught flies confirmed one of the associations, suggesting that QTN effects responsible for less than one percent of the variation for a complex trait can be isolated [49].
The aim of the current study was to assess the capacity of a series of controlled cross designs to validate the contribution of Egfr polymorphisms to naturally occurring variation for wing shape and size. Three schemes were employed, two involving crosses among a subset of the NC lines (a round robin in which 71 nearly isogenic lines were each tested in six random crosses to each other; and a backcross of each of 79 of the lines to two of the most phenotypically extreme lines), and a third involving test crosses between an independent set of Kenyan second chromosomes and the Samarkand wild-type and EgfrE1 and blistered1 mutant alleles (Figure 1). Only one of the six previously reported associations replicated in all datasets, the variant in the Egfr promoter that showed the most significant main effect in the original study and that also replicated in the wild caught flies [49]. However, when we increased the genotyping in the inbred lines, an interesting dichotomy appeared: the association persisted in the North Carolinian sample but vanished in the Californian population. These results argue for the need of large samples, direct contrast of genetic designs, and most importantly increased replication across populations to fully explore the utility of LD mapping to ascertain nucleotide differences affecting continuous variation of evolutionary importance. They also have implications for the fundamental question of whether quantitative genetic variants have variable effects in different populations [50,51].
Figure 1 Schema of the three experimental crosses. (A) In the round robin (RR), each of 71 inbred lines from NC was crossed to six other lines to produce heterozygous offspring. Six loops of the type shown were used. (B) In the Backcross (BC), each of 76 NC lines were crossed to two phenotypically divergent backgrounds, NC025 and NC144, again resulting in heterozygous offspring. (C) Each of 26 Kenyan second chromosomes extracted into the Samarkand background were crossed to regular Samarkand or Samarkand lines carrying blistered1 or EgfrEllipse mutations.
Results
Similarity of shape variation between datasets
Comparison of genotype-phenotype associations between datasets requires that the phenotypic measurements be comparable. We have adopted principal component (PC) descriptors of shape, and although these are modified subtly by inclusion of more wing data [23] overall the shape metrics extracted from each dataset individually are remarkably similar as depicted for consensus configurations of standardized principal component deviations in Figure 2B–I. This is true both for major (for example C1) and minor (W7) principal components, suggesting that shape variation in North American and African populations of D. melanogaster wings reduces to few shared dimensions (see also reference [52]). Furthermore, the eigenvalue decomposition for principal components derived from the individual experiments is qualitatively similar, as shown in Figure 3A. The only exception is the Backcross dataset, where the first PC's for the central region and the whole wing have unusually extreme values. This commonality of the axes of shape variation justified re-extraction of PC's for all datasets jointly, and these joint values were used for all subsequent tests of association. Note that the use of "jointly" or "separately" derived PC's has negligible effect on the test statistics for genetic terms and estimated effects (Table 1 and Additional table 1).
Figure 2 Shape of the D. melanogaster wing was captured from 9 landmarks (A) and analyzed with TPS software [68]. Veins are labeled both by the Comstock and Needham [69] nomenclature and using developmental genetic terminology (L1–L5). Individual inter-vein regions are designated by the letters A-E. (B – I): Shape differences derived from the four different datasets, namely the inbred (B, F), backcross (C, G), round robin (D, H), and Kenyan (E, I) panels. B-E shows the first principal component for landmarks of the central region of the wing (C1). The region is specified by longitudinal veins L3 and L4 and the placement of the cross veins and the margin. F-I show the shape differences for the seventh principal component for the whole wing data (W7). The whole wing and the central portion are not drawn to scale. Dark lines represent negative values, and gray positive. For whole wing PC 7 (W7) the extremes are at +/- 0.01 units and for central region PC 1 (C1) the values are +/- 0.2.
Figure 3 Eigenvalue distribution for the PC's, derived from each of the five datasets individually. (A) Values for the 9 PC's capturing variation in the whole wing, and (B) for the 3 PC's capturing variation for the five landmarks that define the central region of the wing. The decomposition of eigenvalues is comparable for all datasets, with the exception of the Backcross (BC) dataset which deviates qualitatively for the first component in both panels. To illustrate the consistency of shape capture, the principal components for the NC and CA populations were extracted individually for this Figure. All other analyses were conducted on PC's estimated for the CA and NC populations jointly.
Table 1 Retesting the effects of Egfr SNP's on wing shape and size
Separately derived a Jointly derived a
SNP Termb Trait Typec INB RR BC KI INB RR BC KI
C31656T Gtyp × Sex Area N **** . . . **** . . .
C30200Td Gtyp C1 N **** ** **** *** **** ** **** ***
T31634C Gtyp × Sex × Pop C2 N **** . . . **** . . .
T39389C Gtyp D1 S **** . . . **** . . .
T40722C Gtyp × Pop Area S **** . . na **** . . na
C30505A Gtyp W7 N *** *** * ** *** *** ** *
a. As described in the Materials and Methods, PC's were calculated for the datasets individually (separately) or for all the data concatenated (jointly).
b. Term denotes the genetic term most significant in the original study. The same term is reported for the repeats, except for the RR where we could only test genotype effects and sites T31634C and T40722C where the genotype terms are reported as population terms are not available.
c. Type indicates the nature of the polymorphism, N: non-coding and S: synonymous.
d. The significance of the T30200C to C1 association is here reported for the data from Palsson and Gibson 2004. After re-genotyping the p-values reduce to 0.062 and 0.061 for the separately and jointly derived data respectively, when analyzed over the NC and CA populations. Note however that the results for RR and BC are for the re-genotyped data. Significance of terms: "." non-significant, "*": 0.05 > p > 0.01, "**":p > 0.001, "***":p > 0.0001, "****":p > 0.00001. P-values are not adjusted to correct for the seventeen new independent tests conducted.
Absence of support for effects of Egfr on wing size
In order to re-evaluate our previously published associations between wing size and Egfr polymorphisms, recrossing of inbred lines used earlier and testcrosses of additional African chromosomes was carried out. Neither of the two variants affecting size of the wing (C31656T and T40722C) in the initial study gave a significant association in any of the three new datasets (Table 1: RR, round robin; BC, backcross; KI, Kenyan introgressions). In the initial study, polymorphism C31656T had the strongest association, a Genotype by Sex interaction (p = 0.000002) that also exhibited a possible three way interaction of Population, Sex and Genotype (p = 0.001). As the three-way interaction was primarily caused by larger difference in the CA than the NC sample [23], the lack of signal in the crossed NC lines is not surprising. Similarly, while T40722C had previously opposite effects on size depending on population, its contribution in the NC population was neither replicated in the BC and RR recrossing experiments nor in the Kenyan sample. These results indicate that the previously reported association of Egfr with wing size was likely a false positive even though it was significant after adjustment for the number of multiple comparisons experiment-wide.
Replicable effects of one Egfr variant on wing shape
The two crossing schemes and the Kenyan introgressions were used to re-evaluate the contribution of four Egfr variants to aspects of wing shape. Only one polymorphism T30200C, was significant and had consistent effects in all of these experiments. This variant resides in the second alternate promoter in a putative GAGA factor binding site, and contributes to the first principal component of the central region of the wing (C1: Table 1 and Figure 2B–E). One other polymorphism, C30505A in the same promoter, was also significant in all experiments, but had opposite effects on shape metric W7 in the Kenyan sample compared to the Inbred, BC and RR experiments. The inconsistency of the effects casts serious doubt on this association.
Neither of two other previously reported putative associations [23], the sex and population dependent contribution of site T31634C to the width of the central region (C2) nor the contribution of T39389C to the posterior region (D1) were supported by the new data. The lack of association of T39389C prompted us to re-examine epistatic effects which included this particular site and associated with variance in the posterior region (D1) of the wing [23]. The three site Egfr haplotype (G6065T, T39389C, and T40110C) and also each of the two site haplotypes had given highly significant association in the original panel of inbred lines. Due to smaller sample size in our recrossing datasets, testing of this pattern could only be conducted with the BC dataset, but the previous epistatic interactions were not confirmed (data not shown). In summary, only one of the Egfr polymorphisms previously implicated to impact wing shape was corroborated by the new data.
Breakdown of the T30200C association in the Californian population
Previously, due to incomplete genotyping around exon 2, the contribution of T30200C to the central region of the wing was only evaluated with 79 NC and 43 CA lines [23]. Analyses by population found highly significant association in the North Carolinian sample (p = 0.00002) but only marginal association in the west coast sample (p = 0.04) (see Additional Table 1). In order to obtain a better estimate of the magnitude of the effect of T30200C on cross-vein placement, and to investigate the apparent difference in effect between populations, extra genotyping was conducted. The sampling of this polymorphism was increased by re-genotyping the surviving lines from the two populations. Repeating the analysis of variance with 121 NC lines reduced the significance of the association of the T30200C polymorphism (p = 0.002). More dramatically, the addition of 30 more alleles to the CA lines (N = 76) rendered the originally marginal association non-significant (p = 0.9) (Additional Table 1). Inspection of estimated genotypic effects demonstrates this clearly (Figure 4 and Additional Table 2), as the homozygous classes have nearly identical values for the CA population. Evaluation of the effect of this site in the full dataset without population as a term in the model also renders the association non-significant (p > 0.05).
Figure 4 Effects of the T30200C polymorphism on C1 in females across experiments and genetic configurations (designated on the X-axis). INB refers to the inbred populations CA and NC, while INB_RR and INB_BC denote the subsets of inbred lines corresponding to the lines used for the recrossing experiments. Likewise, C144 and C25 indicate the estimated effects of the site in backcrosses to line NC144 and NC025 respectively. The last three points show the effects estimates for the three test crosses involving the Kenyan introgressions (KI), namely wildtype Samarkand chromosomes, the blistered1 mutant, and the EgfrEllipse allele. Each point represents the least square estimate plus or minus one standard error unit for the indicated homozygous genotypes. See Additional Table 2 for corresponding ANOVA's and values for tests with the older genotype data. The PC's were extracted from all datasets jointly to ensure that the axis of variation and units are comparable.
Magnitude of the Egfr allelic contribution
Estimates of the genotypic effects of T30200C on wing shape are comparable across all of the datasets. There was a slight reduction in observed contribution after the extra genotyping (Additional Table 2), and the estimated difference between homozygote classes was smaller in the RR data than in the NC lines, with the CC and TC heterozygotes being indistinguishable. This suggestion of dominance is opposite that observed in a large sample of outbred flies [49] in which heterozygotes resembled TT homozygotes (dominance was non-zero in this study), but it should be noted that CC homozygotes are very infrequent in the current study. In the BC experiments, only TT and TC genotypes were available but the magnitude of the difference between genotypic classes was nearly identical in both backcrosses (to NC025 and NC144) and in the RR experiment (Figure 4 and Additional Table 2). The general differences were again of the same magnitude and direction in the testcrosses involving the Kenyan chromosomes, and they scaled additively with the genetic background (Samarkand, E1 or bs1 carrying chromosomes).
Experimental designs and potential sites with weak effects
In order to compare the gene-wide patterns of association for each design, the association statistic for the Genotype effect of each site along the Egfr locus is plotted for the three experiments in Figure 5. In each plot, higher significance is toward the top, with thresholds drawn at p = 0.05 and p = 0.0001 as before [23]. The analysis focuses on the effects on trait C1, on basis of the assumption that the T30200C association implicates this shape metric as being affected by variation in Egfr.
Figure 5 Association plots for tests of association between Egfr and shape parameter C1. Each plot shows the negative logarithm of the p-value for the test statistic at each polymorphic SNP from 5' to 3 along the Egfr locus. (A) Association profile from the whole NC inbred panel (N = 121, solid line) and the Kenyan chromosomes (N = 26, broken line). (B) The association profile for the RR experiment (outcrossed, N = 71, broken line) and the subset of NC lines (in inbred condition, N = 71, solid line) used for the outcrossing scheme. (C) Similarly the profiles for the BC panel (backcrossed to NC025 and NC144, N = 79, broken line) and the corresponding NC set (N = 79, solid line). The X scale is broken to indicate gaps between contigs with the gene structure represented below (exons as boxes in three contigs of non-coding sequence). Site T30200C (indicated) is located at the 5' most end of contig 2. Lines corresponding to single site significance levels α = 0.05 (negative log p: 1.30) and more conservative gene wide α = 0.0001 (negative log p: 4.0). Trait values for all lines and crosses come from data that were processed jointly in TPSrelw [68].
The first general result is that the small sample of Kenyan introgressions provides more highly significant sites than the total NC sample (with the exception of T30200C there are no significant associations in common between these two populations). Similarly the RR design yielded more significant test statistics (three sites in the first exon) then the BC or inbred panels and had 55 sites exceeding the test-wise significance threshold of p = 0.05. The observed jaggedness of the association profiles likely reflects stochastic fluctuations in the p-values in experiments with relatively small sample size. One interpretation of the data is that the inbred and backcross designs provide better dampening of this stochastic fluctuation then do studies with round robin crossed inbred lines.
The second result is that, in both the RR and BC experiments, the shape of the association profile tracks quite closely with that of the corresponding profile for the set of nearly isogenic lines used to set up the testcrosses. This was not anticipated, since NC025 and NC144 lines have very different wing shapes and each contribute 25% of the genetic variation in the BC, while the RR combines the genetic variation of the 71 inbred lines in equal proportions. Evidently genetic correlations between the different testcrosses are sufficient to produce similar association profiles, whether or not these accurately report QTN effects.
Finally, in order to test whether other sites in Egfr affect the cross-vein placement we performed a combined Mixed model ANOVA on the three NC datasets (NC, RR and BC). Eleven independent polymorphisms summarized in Table 2 were observed to be significant at the experiment-wide significance level of p < 0.0001, including site T30200C. Most of these sites are not significant in the CA and Kenyan datasets, but the direction of the genotypic effects generally correspond with the NC panels (only 2/14 are non-concordant, one tailed Fisher exact test yields p = 0.052). Only one of these new candidate variants, C6085G in the less conserved of the two alternate first N-terminal exon, alters the protein, while the remaining are non-coding or silent. Interestingly, one of these silent polymorphisms is C40620T, which also associates with cryptic variation for eye-roughness in inbred lines and wild flies [21]. Note however, if the Egfr variants are tested against other principal component measures of wing shape, similar number of sites emerge at the level of p < 0.0001 (data not shown) suggesting the caveat that this approach may be inherently noisy.
Table 2 Significance of Egfr polymorphisms on central region shape in NC, CA and Kenyan samples
Effect and significancec Frequency of rare allele
Site Significance in NC Locationa Typeb CA KI NC CA K
C6085G 7.98 × 107 E1 R (OK) ns (Rev) ns 15/125 13/75 4/27
T30200C 6.11 × 1010 P2 N (Rev) ns (OK) *** 26/121 21/76 8/17
A31442T 5.49 × 108 I2 N (OK) ns (OK) ns 13/92 3/24 2/20
A36644T 9.60 × 107 I2 N (Rev) ns (OK) * 26/116 3/20 9/24
A36761C 5.17 × 106 I2 N na na 35/84 na na
Del37192d 3.44 × 105 I2 N (OK) ns (OK) ns 8/106 10/77 7/24
A37282G 9.95 × 106 I2 N nd nd 13/106 1/77 0/23
T39160C 2.10 × 105 E4 S (OK) ns (OK) # 41/128 26/76 16/35
In39534d 8.81 × 105 I5 N (OK) ns (OK) ns 27/110 6/27 3/33
C40620Te 7.37 × 107 E6 S (OK) ns (OK) # 46/123 19/79 7/36
G42242A 8.88 × 105 3' UTR N nd nd 5/105 1/12 1/25
a. Location within the Egfr locus, where P2 refers to the second promoter and the other indicators to the respective introns (I), exons (E) and the 3'UTR.
b. Type indicates the nature of the polymorphism; R: replacement, N: non-coding and S: synonymous.
c. Effects of the polymorphisms in the same direction as in NC are indicated by "OK" and those in the reverse direction are designated by "Rev", with the significance of the genotypic term indicated. "ns": not significant, "#:" 0.1 > p > 0.05. "*": p > 0.01, "**": p > 0.001, "***": p > 0.0001. P-values are not adjusted to correct for the number of tests conducted. "na": genotypes not available, "nd": not computed because of allele rarity.
d. Del37192 is a one base pair deletion and In39534 a four base insert (AACC repeated).
e. Site C40620T is the same as site 8697 described by Dworkin et al. [21].
Discussion
Previously, fine mapping of the association between polymorphisms in the candidate locus Egfr and wing shape and size in D. melanogaster in 210 inbred lines from two North American populations [23] implicated six Egfr variants or linked polymorphisms as causal variants. In this study we aimed to re-evaluate their involvement through further genetic analysis by generating heterozygous lines derived from crosses of a subset of the original lines and by test crosses with a small sample of African chromosomes. Only one of the retested variants was significant in all datasets and gave consistent effects: the T30200C polymorphism that affects a principal component capturing variation in relative distance between the two cross-veins. However, even the estimated absolute magnitude of this effect is dependent on the survey population and crossing scheme. These results highlight the difficulties in validating weak quantitative effects through experimental genetic approaches and suggest that resampling of outbred populations may be the more conclusive approach to dissection of QTL to the nucleotide level.
The T30200C association persists
There are at least three possible explanations for the observed restriction of statistical support for the association of T30200C with wing shape to just two of the three populations sampled. The first is that the observed associations in NC and Kenyan samples are false positives, namely that T30200C or linked variants in Egfr do not contribute to shape of the central region of the wing. This seems unlikely, since significant association was also observed in a large sample of outbred NC flies [49] and the association was also replicated in both of the testcross experiments described here.
Two alternative explanations are consistent with the statistical significance being indicative of a true contribution of Egfr polymorphisms to wing shape in NC. One is that the effect of T30200C is masked by genetic variation that is unique to the CA population. Another possibility is that T30200C is not the real causative variant, but exhibits high LD with the causative site in the NC and Kenyan populations but weak LD in the CA population. Since LD in the Egfr decays to background levels over several hundred bases and no differences were observed between NC and CA in their patterns of LD or allele frequencies, while both North American populations diverge considerably from the Kenyan sample [48], this latter explanation is also unlikely. T30200C does not differ in frequency between NC and CA (Fisher's exact test, p = 0.88), but it does lie adjacent to a 23 kb intron that has not been sequenced in the population sample and could conceivably harbor the true causative variant. However, we favor the hypothesis that one or more modifier loci that differentiate the two North American populations mask the expression of variation due to the Egfr in the CA sample.
Two developmental genetic arguments also lend support to the hypothesis that the T30200C variant is the causal site. First, our prediction that this site affects a GAGA factor binding element in the Egfr promoter, is supported by genetic interaction between the two loci [53,54]. Second, the association between Egfr and cross vein placement is in accord with developmental genetic evidence. Specifically, flies heterozygous for different Egfr alleles lack the majority of the L4 vein and the entire proximal cross-vein [53,55,56]. Recall that shape changes corresponding to principle component C1 for the central region of the wing (Figure 2B–E) represent variation in the distance between the cross-veins, both of which connect with vein L4.
Detection of natural alleles with subtle effects
Quantitative traits in D. melanogaster are now being dissected with QTL mapping, quantitative complementation tests and by testing specific allelic variants by LD mapping. While several studies have found significant association between markers in candidate gene regions and continuous phenotypes [9,11-20] direct re-evaluations of these relationships remain rare. Mackay and Langley [18] found that large insertions around the achaete-scute locus influence bristle number, and this inference was corroborated in a second sample [16]. Geiger-Thornsberry and Mackay [57] confirmed the involvement of two previously identified Delta polymorphisms [15] on bristle number when the same flies were reared under different environmental conditions. Also, we found that three tightly linked silent Egfr polymorphisms affect cryptic variation in eye roughness in inbred lines, and then confirmed the finding in an independent sample of wild caught flies [21]. These studies corroborate the involvement of allelic variation in specific genes with quantitative traits. On the other hand, MacDonald and Long [58] failed to confirm the involvement of a large indel in the 5' region of hairy on bristle number that was previously observed [20]. Moreover, even though both Lai et al. [12] and Lyman et al. [17] implicated scabrous in variation for bristle number, these two studies differed in which markers were typed and by criteria for evaluation of significance (Lai et al. [12] reported an excess of associations with p-value below 0.05 while Lyman et al. [17] found three individual significant sites after permutation testing). Finally Genissel et al. [59] asked if the reported Delta bristle association [16] was caused by common replacement polymorphisms in the gene but were not able to identify the hypothesized causal variant.
In summary, several studies have aimed to validate the contribution of allelic to phenotypic variation, but interpretation is complicated by numerous differences between the studies, including: which population is sampled, the genetic designs, the types of genetic markers employed, and control over environmental variation. Additionally, while negative or only weakly suggestive results are sometimes reported [58-61], bias towards publication of positive results may prevent honest evaluation of the nature of the genetic basis of quantitative traits. In theory, once particular polymorphisms have been associated with an evolutionarily important trait, experimental genetic approaches can be used to confirm the functional differences between alleles [62-66]. However, due to technical complexity such methods have yet to be deployed to systematically gauge the effect of segregating variation in Drosophila. In the case of the Egfr, the proposed regulatory regions are too extensive to evaluate the dynamic contribution of allelic variants to vein and intervein determination, so extensive replication is the only viable approach to dissection of QTN effects.
Mapping resolution and experimental designs
Successful fine mapping of QTL depends on multiple factors such as the magnitude of effect, pattern of LD in the region, available genetic resources, appropriateness of the selection of candidate genes/regions/molecular markers, and the dependence of expression of genetic variation on the experimental settings. The experiments reported here were designed to evaluate the potential for defined crosses to further dissect the role of QTN in subtle quantitative variation, but no obvious recommendations (apart from the need for deep sampling) are forthcoming since the different approaches only produce broadly comparable results.
The round robin and backcross approaches were designed to evaluate the degree to which effects observed in inbred lines are also seen in mixed genetic backgrounds. If the effects of the SNP are additive and there is no epistasis, then they should be just as strong in the testcrosses as in the nearly isogenic lines, with the caveat that there are three genotypes at each SNP to compare instead of just two. The BC design differs in two distinct ways from the RR design, namely the reduced genetic variation (two genomes contribute 50% of the alleles) and the capacity to detect epistatic effects. This latter could occur by interaction between the QTN and other loci, either due to de-canalization as these other loci perturb the phenotype away from the population mean, or simply because QTN effects may generally be so modified by the background that they are only observed in certain backgrounds. The similarity of the estimated genotypic mean differences over the two BC backgrounds and the close tracking of means in the KI experiment (Figure 4), suggests that the reduced genotypic variance is responsible for higher significance of the T30200C association in the BC cross. While this argues for the additivity of the genotypic effects in this case, it is not clear that similar effects will be observed for other traits or loci.
While the ten new highly significant sites in the combined model may be false negatives in the initial lines, more data would be required to confirm that they are true positives. These results indicate that recrossing and deeper population sampling has at best low power to detect novel candidate sites with subtle effects on the phenotype. Consequently, the testcrosses do not obviously outperform the inbred line analysis or bring us any closer to resolving true positive QTN from false positives. Even with a relatively large experiment such as this, the amount of labor and time spent on setting up several hundred crosses and phenotyping several thousand wings does not overcome sampling biases. Even if our analyses suggest that other sites in Egfr may affect cross-vein placement, a considerably larger sample than explored here would be required to validate these sites. The testcross results strongly suggest that we can eliminate highly significant results from the first experiment as false positives, but can not conclusively resolve the question of whether the Egfr QTL resolves to a single or several QTN.
Conclusion
The Egfr contribution to shape variation in D. melanogaster wings reported in Palsson and Gibson [23], and replicated here and in Dworkin et al. [49], represent the best validated example of allelic contribution to continuous morphological variation in flies. While we can not assert that the polymorphism implicated is the causative variant, the evidence and literature cited provide hypotheses testable with experimental genetics. The practical lesson from the observation that five of the six retested Egfr variants failed to validate in testcrosses is that stochastic factors have a substantial impact on analysis of the genetic basis of continuous phenotypes in studies involving fewer than 200 inbred lines. Apparent conditional polymorphisms may be especially sensitive to these effects of chance, and all unreplicated association studies in Drosophila should be considered with this caveat in mind. We suggest that measurement of a very large number of offspring is essential for replication and validation in association studies, and that these are better sampled in outbred wild individuals than in laboratory lines. The declining cost of genotyping will facilitate this transition to large scale mapping of quantitative traits to single nucleotides in ecological settings.
Methods
Stocks and crossing schemes
Three separate experiments were conducted to re-evaluate the contribution of Egfr on wing shape (Figure 1). Two involved recrossing, by round-robin (RR) and modified backcross (BC) designs, (71 and 76 NC lines respectively, with 70 being shared). The RR crossing scheme is a partial diallele cross, with the 73 lines being crossed three times as sire and three times as dam. The mating scheme was derived by permutation. In the BC design, males from 76 NC lines were crossed independently with females from two strains NC025 and NC144. These inbred lines have extreme PC1 values for the anterior and posterior regions of the wing. The third experiment (KI) involved an independent set of Kenyan alleles from Ron Woodruff [48]. Second chromosomes were substituted into the Samarkand background by a 4 generation crossing scheme, utilizing stocks kindly provided by Trudy MacKay. Similarly an Egfr allele, Ellipse (E1), and a blistered allele (bs1) were substituted into Sam. The wild-type chromosomes were tested over these two mutations and the wild-type Samarkand second chromosomes in three replicate crosses arranged in random blocks. All crosses involved three males crossed to three females, and where conducted in two (RR and BC) or three replicates (KI).
Fly rearing and scoring of wings
Flies were reared at 25°C in standard cornmeal medium with a constant light/dark cycle. Density was controlled by placing two virgin females and two males in a vial and discarding parents on the 2nd or 4th day depending on visual assessment of egg density. The right wing of eight to ten randomly selected individuals per sex (except only RR females) from each vial was scored. In the Kenyan introgression experiment the visible marker Cy on the balancer chromosome distinguished genotypes of lines inviable as 2nd chromosome homozygotes. Handling of specimens and data processing was identical to previous experiments [47]. In short, wings were dissected at the hinge and arranged on glass slides and held in place with a cover slip. Within 48 hours, wings were digitally photographed at 4× magnification with a Spot camera, mounted on a Nikon microscope. Images were processed with Adobe Photoshop version 5, and landmarks captured in Scion Image (freeware available [67]). The nine landmarks at the junctions of veins and wing margin are depicted in Figure 2A. One author, JD, digitized the back-cross and ~65% of the round-robin while the remaining specimens (35% of RR, Inbred and Kenyan) were scored by AP. No significant "investigator" effects were found in an analysis of 1000 RR wings scored by both authors (not shown).
Extracting common axes of shape variation
Shape variation was summarized with the TPSrelw software version 1.39 (freeware available [68]) by calculating relative warps for a set of landmarks, for the whole wing or individual regions (Figure 2A). The procedure involves "partial Procrustes" superimposition, by iterated rotation and alignment of specimens, rescaling to unit size, prior to extraction of the relative warps. The relative warps are essentially principal components (PC's), and will be referred to as such henceforth.
Egfr genotype matrix
Genotypes used for the association tests were derived from our earlier sequence data [48]. The BC and RR recrossing was not designed to test particular polymorphisms, and therefore generated heterozygotes and sometimes both homozygotes at particular nucleotide positions. For instance in the BC design, of the six sites retested, T31634C and C30505A were not typed in NC144. Furthermore, of the remaining four polymorphisms, the lines differed only at T40722C. Note this does not mean that their Egfr haplotypes are highly similar, as 167 out the 232 common Egfr sites genotyped in both lines differ, with several recombination events evident. F1 lines that were missing a genotype of one parent where omitted from the analysis for that particular genotype. In the Kenyan sample, only the variant T40722C was not tested, as it was only available in one Kenyan line. The Egfr alleles were not sequenced in the three tester chromosomes, leading to tests on haploid data.
Re-genotyping of T30200C
The T30200C polymorphism in the non-coding region upstream of alternative exon one [48] was re-genotyped in the NC and CA lines in 2004. The previous sample was incomplete due to high level of PCR failure that we attributed to repetitive elements in the region [48]. Therefore an alternative strategy for genotyping was deployed, utilizing the observation that this polymorphism affects a Restriction Length Fragment Polymorphism (RFLP) for the DraIII restriction endonuclease. As before, a single male from each line was genotyped [48]. For PCR, the following new primers were utilized as described in [49]: 5'-GTGGCTCGTAATGTGAAACT-3' and 5'-GCGTTACTGGTGGGATGAATCAAG-3'. Of the 210 original lines characterized in 2001–2002, 198 were still surviving in 2004 and were regenotyped. Three discrepancies were found, all in the NC panel (NC065, NC075, NC116). In the case of NC065 heterozygosity for the 3'end of the locus was noted in the original study and it is consequently quite possible that two alleles were segregating when the line was initially genotyped. Contamination of either DNA samples or stocks maintained over this period are also formal possibilities, particularly for the other two lines. These three lines were dropped from the re-analyses.
Analysis of phenotypic variation
All statistical analysis used SAS version 8.2 (SAS Institute, Cary, NC). The estimation of line effects and extraction of line means was implemented with the LSMEANS option in Proc GLM. The model for the RR dataset was:
Y = μ + Line + Rep(Line) + ε
where Line represents each of the F1 lines generated by the round-robin crosses, and Rep the replicate vial. For the Back-cross and the Kenyan introgression, a more complicated model was used, accounting for the effects of Cross (to NC025 and NC144 or to Sam, E1 and bs), Sex or Line.
Y = μ + Cross + Sex + C × S + Line + S × L + C × L + C × S × L + Rep(C × L) + S × R(C × L) + ε
In both models terms including Line and Rep are considered random. We also performed the analysis without Rep as a term, with the same results.
Tests of quantitative nucleotide effects
The main aim of these experiments was to re-evaluate the six sites which gave significant signals for wing size and shape in [23]. The RR experiment focused on females from a single population (NC) and a simple model was implemented in Proc Mixed:
Y = μ + Gtyp + Rep(Gtyp) + ε
Gtyp is the fixed effect of Genotype, and Rep is a random term, again the replicate vials. For the back-cross and the Kenyan test cross, the model accounted for the contribution of sex and cross:
Y = μ + Gtyp + Sex + Cross + G × S + G × C + G × S × C + Line(G × C) + ε
The mean effects of polymorphisms were estimated by the LSMEANS option. Reduced models, by crosses, and extended, by including replicates were also studied and were in accord.
In order to gauge the effects of additional sites in Egfr on the C1 we utilized a related model, substituting the Cross term with a fixed experiment (Exp) term to demarcate the NC, BC and RR datasets, and restricting the analysis to females as the RR panel had no males. The sire and dam are random effects nested within the fixed effects:
Y = μ + Gtyp + Exp + G × E + dam × sire(G × C) + Rep(dam × sire × G × C) + ε
Sites with probability of genotype term below 0.0001, where then investigated for consistency in genotypic effects and their significance in the CA and KI dataset.
Authors' contributions
AP, GG and JD designed experiments. JD crossed and scored RR and BC experiments, AP crossed and scored KI, Inbreds and parts of RR dataset. AP conducted statistical analysis. ID regenotyped the T30200C variant. AP, ID and GG wrote the manuscript and all authors approved the final version.
Supplementary Material
Additional table 1
ANOVA tables for site T30200C and wing shape. The file shows the results of Analysis of Variance for the T30200C variant in the Egfr promoter and the first principle component for the shape of the central region of the wing.
Click here for file
Additional Table 2
Genotypic effects of T30200C on the first PC of the central region of the wing. This table illustrates the genotypic effects (and standard errors) for the T30200C association to wing shape of the central region.
Click here for file
Acknowledgements
Trudy MacKay kindly provided the Samarkand stocks for the introgression and advice on analysis. Thanks to Marcos Antezana, Lisa Goering, Todd Martin and Jean-Claude Walser for comments on the manuscript.
==== Refs
Barton NH Keightley PD Understanding quantitative genetic variation Nat Rev Genet 2002 3 11 21 11823787 10.1038/nrg700
Beavis WD Washington DC: American Seed Trade Association The power and deceit of QTL experiments: lessons from comparative QTL studies 1994 49th Annual Corn and Sorghum Research Conference pp 252 268
Mackay TFC The genetic architecture of quantitative traits Annual Review of Genetics 2001 35 303 339 11700286 10.1146/annurev.genet.35.102401.090633
Bolivar VJ Cook MN Flaherty L Mapping of quantitative trait loci with knockout/congenic strains Genome Research 2001 11 1549 1552 11544198 10.1101/gr.194001
Moore KJ Nagle DL Complex trait analysis in the mouse: The strengths, the limitations and the promise yet to come Annu Rev Genet 2000 34 653 686 11092842 10.1146/annurev.genet.34.1.653
Frary A Nesbitt TC Grandillo S Knaap E Cong B Liu J Meller J Elber R Alpert KB Tanksley SD fw2.2: a quantitative trait locus key to the evolution of tomato fruit size Science 2000 289 85 88 10884229 10.1126/science.289.5476.85
Wilson LM Whitt SR Ibanez AM Rocheford TR Goodman MM Buckler 4th ES Dissection of maize kernel composition and starch production by candidate gene association Plant Cell 2004 16 2719 2733 15377761 10.1105/tpc.104.025700
Remington DL Ungerer MC Purugganan MD Map-based cloning of quantitative trait loci: progress and prospects Genetical Research 2001 78 213 218 11865710
De Luca M Roshina NV Geiger-Thornsberry GL Lyman RF Pasyukova EG Mackay TFC Dopa decarboxylase (Ddc) affects variation in Drosophila longevity Nature Genetics 2003 34 429 433 12881721 10.1038/ng1218
Pasyukova EG Vieira C Mackay TF Deficiency mapping of quantitative trait loci affecting longevity in drosophila melanogaster Genetics 2000 156 1129 1146 11063689
Game AY Oakeshott JG Associations between restriction site polymorphism and enzyme activity variation for esterase 6 in Drosophila melanogaster Genetics 1990 126 1021 1031 1981760
Lai C Lyman RF Long AD Langley CH MacKay TFC Naturally occurring variation in bristle number and DNA polymorphisms at the scabrous locus of Drosophila melanogaster Science 1994 266 1697 1702 7992053
Laurie CC Bridgeham JT Choudhary M Association between DNA sequence variation and variation in expression of the Adh gene in natural populations of Drosophila melanogaster Genetics 1991 129 489 499 1683848
Lazzaro BP Sceurman BK Clark AG Genetic basis of natural variation in D. melanogaster antibacterial immunity Science 2004 303 1873 1876 15031506 10.1126/science.1092447
Long AD Lyman RF Langley CH Mackay TFC Two sites in the Delta gene region contribute to naturally occurring variation in bristle number in Drosophila melanogaster Genetics 1998 149 999 1017 9611209
Long AD Lyman RF Morgan AH Langley CH Mackay TFC Both naturally occurring insertions of transposable elements and intermediate frequency polymorphisms at the achaete-scute complex are associated with variation in bristle number in Drosophila melanogaster Genetics 2000 154 1255 1269 10757767
Lyman RF Lai CQ Mackay TFC Linkage disequilibrium mapping of molecular polymorphisms at the scabrous locus associated with naturally occurring variation in bristle number in Drosophila melanogaster Genetical Research 1999 74 303 311 10689806 10.1017/S001667239900419X
Mackay TFC Langley CH Molecular and phenotypic variation in the achaete-scute region of Drosophila melanogaster Nature 1990 348 64 66 2122256 10.1038/348064a0
Odgers WA Healy MJ Oakeshott JG Nucleotide polymorphism in the 5' promoter region of esterase 6 in Drosophila melanogaster and its relationship to enzyme activity variation Genetics 1995 141 215 222 8536969
Robin C Lyman RF Long AD Langley CH Mackay TF hairy: a quantitative trait locus for Drosophila sensory bristle number Genetics 2002 162 155 164 12242230
Dworkin I Palsson A Birdsall K Gibson G Evidence that Egfr contributes to cryptic genetic variation for photoreceptor determination in natural populations of Drosophila melanogaster Current Biology 2003 13 1888 1893 14588245 10.1016/j.cub.2003.10.001
Inomata N Goto H Itoh M Isono K A Single-Amino-Acid Change of the Gustatory Receptor Gene, Gr5a, Has a Major Effect on Trehalose Sensitivity in a Natural Population of Drosophila melanogaster Genetics 2004 167 1749 1758 15342513 10.1534/genetics.104.027045
Palsson A Gibson G Association between nucleotide variation in Egfr and wing shape in Drosophila melanogaster Genetics 2004 167 1187 1198 15280234 10.1534/genetics.103.021766
Colhoun HM McKeigue PM Smith GD Problems of reporting genetic associations with complex outcomes Lancet 2003 361 865 872 12642066 10.1016/S0140-6736(03)12715-8
Ioannidis JPA Ntzani EE Trikalinos TA Contopoulos-Ioannidis DG Replication validity of genetic association studies Nature Genetics 2001 29 306 309 11600885 10.1038/ng749
Cowley DE Atchley WR Quantitative genetics of Drosophila melanogaster. II. Heritabilities and genetic correlations between sexes for head and thorax traits Genetics 1988 119 421 433 17246429
Whitlock MC Fowler K The changes in genetic and environmental variance with inbreeding in Drosophila melanogaster Genetics 1999 152 345 353 10224265
Calboli FCF Gilchrist GW Partridge L Different cell size and cell number contribution in two newly established and one ancient body size cline of Drosophila subobscura Evolution 2003 57 566 573 12703946
Gilchrist S Partridge L A comparison of the genetic basis of wing size divergence in three parallel body size clines of Drosophila melanogaster Genetics 1999 153 1775 1787 10581284
Huey RB Gilchrist GW Carlson ML Berrigan D Serra L Rapid evolution of a geographic cline in size in an introduced fly Science 2000 287 308 309 10634786 10.1126/science.287.5451.308
Klingenberg CP Zaklan SD Morphological intergration between development compartments in the Drosophila wing Evolution Int J Org Evolution 2000 54 1273 1285 11005294
Klingenberg CP McIntyre GS Zaklan SD Left-right asymmetry of fly wings and the evolution of body axes Proc Biol Sci 1998 265 1255 1259 9699316 10.1098/rspb.1998.0427
Weber KE Selection on wing allometry in Drosophila melanogaster Genetics 1990 126 975 989 2127580
Weber KE Increased selection response in larger populations. I. Selection for wing-tip height in Drosophila melanogaster at three population sizes Genetics 1990 125 579 584 2116358
Weber KE How small are the smallest selectable domains of form? Genetics 1992 130 345 353 1541394
Santos M Iriarte PF Cespedes W Balanya J Fontdevila A Serra L Swift laboratory thermal evolution of wing shape (but not size) in Drosophila subobscura and its relationship with chromosomal inversion polymorphism Journal of Evolutionary Biology 2004 17 841 558 15271084 10.1111/j.1420-9101.2004.00721.x
Mezey JG Houle D Nuzhdin SV Naturally Segregating Quantitative Trait Loci Affecting Wing Shape of Drosophila melanogaster Genetics 2005 169 2101 2113 15520257 10.1534/genetics.104.036988
Palsson A Gibson G Quantitative developmental genetic analysis reveals that the ancestral dipteran wing vein prepattern is conserved in Drosophila melanogaster Dev Genes Evol 2000 210 617 622 11151298 10.1007/s004270000107
Weber K Eisman R Higgins S Morey L Patty A Tausek M Zeng ZB An analysis of polygenes affecting wing shape on chromosome 2 in Drosophila melanogaster Genetics 2001 159 1045 1057 11729152
Weber KE Eisman R Morey L Patty A Sparks J Tausek M Zeng ZB An analysis of polygenes affecting wing shape on chromosome 3 in Drosophila melanogaster Genetics 1999 153 773 786 10511557
Zimmerman E Palsson A Gibson G Quantitative trait loci affecting components of wing shape in Drosophila melanogaster Genetics 2000 155 671 683 10835390
Bookstein FL Morphometric tools for landmark data: geometry and biology 1991 Cambridge, Massachusetts , Cambridge University Press 435
Stark J Bonacum J Remsen J DeSalle R The evolution and development of dipteran wing veins: a systematic approach Annu Rev Entomol 1999 44 97 129 9990717 10.1146/annurev.ento.44.1.97
Galpern P The use of common principle component analysis in studies of phenotypic evolution: an example from the Drosophilidae Zoology 2000 Toronto , University of Toronto 131
Houle D Mezey JG Galpern P Carter A Automated measurement of Drosophila wings BMC Evolutionary Biology 2003 3 25 14670094 10.1186/1471-2148-3-25
Weber K Johnson N Champlin D Patty A Many P-Element Insertions Affect Wing Shape in Drosophila melanogaster Genetics 2005 169 1461 1475 15545659 10.1534/genetics.104.027748
Birdsall K Zimmerman E Teeter K Gibson G Genetic variation for the positioning of wing veins in Drosophila melanogaster Evol Dev 2000 2 16 24 11256413 10.1046/j.1525-142x.2000.00034.x
Palsson A Rouse A Riley-Berger R Dworkin I Gibson G Nucleotide variation in the Egfr locus of Drosophila melanogaster Genetics 2004 167 1199 1212 15280235 10.1534/genetics.104.026252
Dworkin I Palsson A Gibson G Replication of an Egfr-Wing Shape Association in a Wild-Caught Cohort of Drosophila melanogaster Genetics 2005 169 2115 2125 15687273 10.1534/genetics.104.035766
Goldstein DB Hirschhorn JN In genetic control of disease, does 'race' matter? Nat Genet 2004 36 1243 1244 15565101 10.1038/ng1204-1243
Ioannidis JP Ntzani EE Trikalinos TA 'Racial' differences in genetic effects for complex diseases Nat Genet 2004 36 1312 1318 15543147 10.1038/ng1474
Mezey JG Houle D The dimensionality of genetic variation for wing shape in Drosophila melanogaster Evolution 2005 in press
Angulo M Corominas M Serras F Activation and repression activities of ash2 in Drosophila wing imaginal discs Development 2004 131 4943 4953 15371308 10.1242/dev.01380
Bejarano F Busturia A Function of the Trithorax-like gene during Drosophila development Developmental Biology 2004 268 327 341 15063171 10.1016/j.ydbio.2004.01.006
Diaz-Benjumea FJ Garcia-Bellido A Behaviour of cells mutant for an EGF receptor homologue of Drosophila in genetic mosaics Proc Biol Sci 1990 242 36 44 1980740
Sturtevant MA Roark M Bier E The Drosophila rhomboid gene mediates the localized formation of wing veins and interacts genetically with components of the EGF-R signaling pathway Genes Dev 1993 7 961 973 8504935
Geiger-Thornsberry GL MacKay TFC Association of single-nucleotide polymorphisms at the Delta locus with genotype by environment interaction for sensory bristle number in Drosophila melanogaster Genetical Research 2002 79 211 218 12220128 10.1017/S0016672302005621
Macdonald SJ Long AD A potential regulatory polymorphism upstream of hairy is not associated with bristle number variation in wild-caught Drosophila Genetics 2004 167 2127 2131 15342546 10.1534/genetics.104.026732
Genissel A Pastinen T Dowell A Mackay TF Long AD No evidence for an association between common nonsynonymous polymorphisms in delta and bristle number variation in natural and laboratory populations of Drosophila melanogaster Genetics 2004 166 291 306 15020426 10.1534/genetics.166.1.291
Goering LM Gibson G Genetic variation for dorsal-ventral patterning of the Drosophila melanogaster eggshell Evol dev 2005 7 81 88 15733305 10.1111/j.1525-142X.2005.05009.x
Nikoh N Duty A Gibson G Effects of population structure and sex on association between serotonin receptors and Drosophila heart rate Genetics 2004 168 1963 1974 15611167 10.1534/genetics.104.028712
Choudhary M Laurie CC Use of in vitro mutagenesis to analyze the molecular basis of the difference in Adh expression associated with the allozyme polymorphism in Drosophila melanogaster Genetics 1991 129 481 488 1743488
Greenberg AJ Moran JR Coyne JA Wu CI Ecological adaptation during incipient speciation revealed by precise gene replacement Science 2003 302 1754 1757 14657496 10.1126/science.1090432
Laurie CC Stam LF The Effect of an Intronic Polymorphism on Alcohol Dehydrogenase Expression in Drosophila melanogaster Genetics 1994 138 379 385 7828821
Odgers WA Aquadro CF Coppin CW Healy MJ Oakeshott JG Nucleotide polymorphism in the Est6 promoter, which is widespread in derived populations of Drosophila melanogaster, changes the level of Esterase 6 expressed in the male ejaculatory duct Genetics 2002 162 785 797 12399389
Stam LF Laurie CC Molecular dissection of a major gene effect on a quantitative trait: The level of alcohol dehydrogenase expression in Drosophila melanogaster Genetics 1996 144 1559 1564 8978044
Scion Image
Rohlf, F. J. TPS relative warp analysis software Windows
Comstock JH Needham JG The wings of insects American Naturalist 1898 32 43 903 10.1086/276766
|
16102176
|
PMC1208880
|
CC BY
|
2021-01-04 16:30:38
|
no
|
BMC Genet. 2005 Aug 15; 6:44
|
utf-8
|
BMC Genet
| 2,005 |
10.1186/1471-2156-6-44
|
oa_comm
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.