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
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1608950110.1371/journal.pmed.0020233Research ArticlePhysiologyDiabetes/Endocrinology/MetabolismMedical ImagingDiabetesGeneticsMedical ImagingNutrition and MetabolismDecreased Insulin-Stimulated ATP Synthesis and Phosphate Transport in Muscle of Insulin-Resistant Offspring of Type 2 Diabetic Parents Insulin Resistance-Mitochondrial DysfunctionPetersen Kitt F
1
Dufour Sylvie
1
3
Shulman Gerald I
1
2
3
*1Department of Internal Medicine, Yale University School of Medicine, New Haven, Connecticut, United States of America,2Department of Cellular and Molecular Physiology, Yale University School of Medicine, New Haven, Connecticut, United States of America,3 Howard Hughes Medical Institute, Yale University School of Medicine, New Haven, Connecticut, United States of AmericaKumar Sudesh Academic EditorUniversity of Warwick Medical SchoolUnited Kingdom*To whom correspondence should be addressed. E-mail: gerald. [email protected]
Competing Interests: The authors have declared that no competing interests exist. GIS is a member of the editorial board of PLOS Medicine.
Author Contributions: KFP, SD, and GIS designed the study and analyzed the data. KFP enrolled patients. KFP, SD, and GIS contributed to writing the paper.
9 2005 16 8 2005 2 9 e2334 1 2005 3 6 2005 Copyright: © 2005 Petersen 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.
Insulin Resistance in the Offspring of Parents with Type 2 Diabetes
Type 2 Diabetes: Insulin Resistance May Be the Result of Mitochondrial Dysfunction
Background
Insulin resistance is the best predictor for the development of type 2 diabetes. Recent studies have shown that young, lean, insulin-resistant (IR) offspring of parents with type 2 diabetes have reduced basal rates of muscle mitochondrial phosphorylation activity associated with increased intramyocellular lipid (IMCL) content, which in turn blocks insulin signaling and insulin action in muscle. In order to further characterize mitochondrial activity in these individuals, we examined insulin-stimulated rates of adenosine triphosphate (ATP) synthesis and phosphate transport in skeletal muscle in a similar cohort of participants.
Methods and Findings
Rates of insulin-stimulated muscle mitochondrial ATP synthase flux and insulin-stimulated increases in concentrations of intramyocellular inorganic phosphate (Pi) were assessed by 31P magnetic resonance spectroscopy (MRS) in healthy, lean, IR offspring of parents with type 2 diabetes and healthy, lean control participants with normal insulin sensitivity. IMCL content in the soleus muscle of all participants was assessed by 1H MRS. During a hyperinsulinemic-euglycemic clamp, rates of insulin-stimulated glucose uptake were decreased by approximately 50% in the IR offspring compared to the control participants (p = 0.007 versus controls) and were associated with an approximately 2-fold increase in IMCL content (p < 0.006 versus controls). In the control participants rates of ATP synthesis increased by approximately 90% during the hyperinsulinemic-euglycemic clamp. In contrast, insulin-stimulated rates of muscle mitochondrial ATP synthesis increased by only 5% in the IR offspring (p = 0.001 versus controls) and was associated with a severe reduction of insulin-stimulated increases in the intramyocellular Pi concentrations (IR offspring: 4.7% ± 1.9% versus controls: 19.3% ± 5.7%; p = 0.03). Insulin-induced increases in intramyocellular Pi concentrations correlated well with insulin-stimulated increases in rates of ATP synthesis (r = 0.67; p = 0.008).
Conclusions
These data demonstrate that insulin-stimulated rates of mitochondrial ATP synthesis are reduced in IR offspring of parents with type 2 diabetes. Furthermore, these IR offspring also have impaired insulin-stimulated phosphate transport in muscle, which may contribute to their defects in insulin-stimulated rates of mitochondrial ATP synthesis.
==== Body
Introduction
Type 2 diabetes affects about 171 million people worldwide, and the number of people likely to be affected by diabetes is expected to double by 2030 [1,2]. Diabetes develops when resistance to insulin action combines with impaired insulin secretion, resulting in hyperglycemia. Cross-sectional studies have demonstrated the presence of insulin resistance in virtually all patients with type 2 diabetes, and prospective studies have demonstrated the presence of insulin resistance one to two decades prior to the onset of the disease [3,4]. In addition, insulin resistance in the insulin-resistant (IR) offspring of parents with type 2 diabetes has been shown to be the best predictor for the later development of the disease [5]. Although the mechanism for their insulin resistance is unknown, previous studies have demonstrated a strong relationship between dysregulated intracellular fatty acid metabolism and insulin resistance in skeletal muscle [6–8]. Increases in certain intramyocellular lipid (IMCL) metabolites such as acyl CoAs and diacylglycerol have in turn been shown to block insulin signaling by activating a serine kinase cascade [9–12], resulting in reduced insulin-stimulated muscle glucose transport activity [10] and decreased muscle glycogen synthesis [13,14]. Recent combined 1H and 31P magnetic resonance spectroscopy (MRS) studies have demonstrated an approximately 2-fold increase in the IMCL content in a group of healthy, young, lean, IR offspring of parents with type 2 diabetes associated with a 30% reduction in basal rates of muscle mitochondrial adenosine triphosphate (ATP) production. These data suggest that defects in mitochondrial activity, due to reduction in mitochondrial content and/or function, may be an important contributing factor to their increased IMCL content and insulin resistance [15]. Consistent with these functional MRS data two recent gene microarray studies have found a coordinated reduction in peroxisome proliferator-activated receptor γ coactivator 1 responsive genes that are involved in mitochondrial oxidative phosphorylation activity in muscle biopsy samples obtained from obese participants with type 2 diabetes [16,17] and their overweight first-degree relatives [17].
In order to further characterize mitochondrial function in these individuals we measured insulin-stimulated rates of muscle mitochondrial ATP synthesis and changes in intramyocellular inorganic phosphate (Pi) concentrations using 31P MRS in a similar group of young, lean, IR offspring of parents with type 2 diabetes and body mass index (BMI)–age–activity-matched control participants. These individuals are ideal for examining the earliest defects responsible for the pathogenesis of insulin resistance since, in contrast to patients with diabetes, they are young, lean, healthy, and they have none of the other confounding factors that are likely to be present in patients with type 2 diabetes.
Methods
Participants
All volunteers were prescreened to be healthy, taking no medications, lean, nonsmoking, of normal birth weight (>2.25 kg), and sedentary, as defined by an activity index questionnaire [18]. IR offspring (four male and three female) had at least one parent or grandparent with type 2 diabetes and at least one other family member with type 2 diabetes. All qualifying participants underwent a complete medical history and physical examination along with blood tests to verify normal: blood and platelet count, electrolytes, aspartate amino transferase, alanine amino transferase, blood urea nitrogen, creatinine, cholesterol, and triglyceride. IR offspring were defined by an insulin (240 pmol/m2/min)–stimulated glucose infusion rate <4.85 mg/kg/min, whereas the control participants were defined by a rate of insulin-stimulated glucose uptake >6 mg/kg/min.
Written consent was obtained from each participant after the purpose, nature, and potential complications of the studies were explained. The protocol was approved by the Yale University Human Investigation Committee.
Diet and Study Preparation
Participants were instructed to abstain from any exercise and to eat a regular, weight maintenance diet containing at least 150 g of carbohydrate per day for the 3 d prior to the studies. To minimize changes in ovarian hormonal effects on glucose metabolism, the female participants were studied during the follicular phase (days 0–12) of the menstrual cycle [19].
Hyperinsulinemic-Euglycemic Clamp
Participants were admitted to the Yale University–New Haven Hospital General Clinical Research Center the evening before the studies and remained fasting from 10 p.m. until the end of the study the next day. At 6 a.m. antecubital IV catheters were inserted, and at 7 a.m. the participants were placed inside the 2.1T magnetic resonance spectrometer (Biospec Spectrometer, Bruker Instruments, Billerica, Massachusetts, United States) with the right calf positioned over the 31P receiver coil. After baseline 31P MRS measurements of ATP synthesis flux and muscle intramyocellular Pi concentrations (described below), a primed-continuous insulin infusion was begun (240 pmol/m2/min) to raise plasma insulin concentrations to approximately 480 pmol/l and to maintain them at this level throughout the remaining 150 min of the clamp. Blood was collected every 5 min for determination of plasma glucose concentrations, and a variable intravenous infusion of dextrose (200 g/l) was administered to raise plasma glucose concentrations to 100 mg/dl and maintained at this concentration until the end of the clamp.
Metabolites and Hormones
Plasma glucose concentrations were measured using a Glucose Analyzer II (Beckman Instruments, Fullerton, California, United States). Plasma concentrations of insulin were measured using double-antibody radioimmunoassay kits (Linco, St. Louis, Missouri, United States). Plasma fatty acid concentrations were determined using a microfluorimetric method [20]. Urine nitrogen content was measured at the Mayo Medical Laboratories (Rochester, Minnesota, United States).
31P MRS
Rates of mitochondrial phosphorylation activity were assessed by 31P MRS saturation transfer performed at 36.31 MHz using a flat, concentric probe made of a 9-cm-diameter inner coil (for 31P) and a 13-cm outer coil tuned to proton frequency for scout imaging and shimming as previously described [15,21]. Unidirectional rates of ATP synthesis were measured at baseline and during the insulin clamp, using the saturation transfer method applied to the exchange between intracellular phosphate and ATP [21]. The steady-state intramyocellular Pi magnetization was measured in the presence of a selective irradiation of the γ resonance of ATP and compared to the equilibrium intracellular phosphate magnetization in a control spectrum (without irradiation of γ ATP) [21]. Total acquisition time for 31P spectra was about 120 min (from t = 30 to 150 min of the clamp).
1H MRS of IMCL Content
On a separate day, after a 12-h fast, all participants were transported by wheelchair to the Yale Magnetic Resonance Center and localized 1H MRS of the soleus muscle to assess IMCL content were acquired on a 2.1T Biospec as previously described [22].
Indirect Calorimetry
Rates of whole-body energy expenditure and glucose and fat oxidation were assessed by indirect calorimetry (Deltratrack Metabolic Monitor, Sensormedics, Anaheim, California, United States) during the last 20 min of the baseline period and during the last 20 min of the clamp [23].
Calculations
Hyperinsulinemic-euglycemic clamp
The rates of glucose infusion were calculated in 15-min blocks from 60–150 min during the insulin clamp. The data were corrected for urinary glucose loss and glucose space and were expressed as milligrams glucose metabolized per kilogram body weight per minute.
Indirect calorimetry
Basal and insulin-stimulated glucose and lipid oxidation rates were measured by the ventilated hood technique, using a Deltatrack Metabolic Monitor (as previously described [24]). The nonprotein respiratory quotients for 100% oxidation of fat and for oxidation of carbohydrates were 0.707 and 1.00, respectively [24]. Nonoxidative glucose metabolism was calculated by subtracting the amount of glucose oxidized from the total amount of glucose infused.
Statistical analyses
Statistical analyses were performed using the unpaired t-test to detect statistical differences between control and IR offspring, and the paired student t-tests were performed to test for interparticipant differences. A p-value (two-sided) of <0.05 was considered significant. All data are expressed as mean ± SEM.
Results
Participant Characteristics
The control participants and IR offspring were group matched for age, weight, height, BMI, and activity (Table 1). All participants were lean, nonsmoking, and had a sedentary lifestyle without regular participation in sports activities.
Table 1 Characteristics of the Two Groups of Participants
Hyperinsulinemic-Euglycemic Clamp
Rates of insulin-stimulated glucose metabolism were approximately 50% lower in the IR offspring compared to the control participants (control: 8.56 ± 0.88 mg/[kg/min] and IR offspring: 4.54 ± 0.11 mg/[kg/min]; p < 0.001) (Figure 1A). This reduction in insulin-stimulated rate of whole-body glucose metabolism was associated with an approximately 90% increase in the IMCL content in the IR offspring (1.68% ± 0.21% versus controls: 0.86 ± 0.09%, p = 0.006) and could mostly be attributed to an approximately 60% reduction (p < 0.001) in nonoxidative glucose disposal (Table 2).
Figure 1 Effect of Insulin Stimulation
(A) Rates of whole-body glucose metabolism, (B) rates of muscle mitochondrial ATP synthesis, and (C) intramyocellular Pi concentrations.
Table 2 Rates of Energy Expenditure, Glucose and Lipid Oxidation, and Respiratory Quotient Values in the Basal State and during the Hyperinsulinemic-Euglycemic Clamp in Control Individuals and IR Offspring
31P MRS
During insulin stimulation, rates of mitochondrial phosphorylation activity increased by approximately 90% in the control participants (Figure 1B). In contrast, insulin-stimulated ATP synthase flux was severely blunted in the IR offspring (ΔATP flux: 5% ± 2%; p = 0.001) (Figure 1B). During insulin stimulation concentrations of intramyocellular Pi increased by 19.3% ± 5.7% in the control participants, whereas insulin-stimulated increases in intramyocellular Pi concentrations were also severely blunted in the IR offspring (ΔPi: 4.7% ± 1.9%; p = 0.03 versus controls) (Figure 1C). There was a strong correlation between insulin-induced increases in intramyocellular Pi concentrations, and the insulin-stimulated increases in rates of ATP synthesis (r = 0.67; p = 0.008). In order to determine whether or not these changes in intramyocellular Pi might be attributable to insulin-induced alterations in phosphocreatinine (PCr) metabolism, we also assessed the PCr/ATP ratio before and during the clamp in both groups of participants. We found no changes in the PCr/ATP ratio during the hyperinsulinemic-euglycemic clamp compared to the basal period in either the control participants (basal: 4.36 ± 0.06 versus clamp: 4.30 ± 0.31) or IR offspring (basal: 3.82 ± 0.17 versus clamp: 3.91 ± 0.33).
Discussion
Recent studies have demonstrated an approximately 30% reduction in basal rates of mitochondrial ATP synthesis in healthy, young, lean, IR offspring of parents with type 2 diabetes, which were associated with an approximately 2-fold increase in IMCL content [15]. We postulated that reductions in mitochondrial activity might predispose these individuals to increased IMCL accumulation, which in turn would lead to defects in insulin signaling and insulin action in skeletal muscle. In order to further characterize the regulation of mitochondrial activity in these individuals, we assessed insulin-stimulated rates of ATP synthesis in a similar group of young, lean, IR offspring as well as in a group of age–BMI–activity-matched control group participants using 31P MRS. We found that insulin was a potent stimulator of muscle ATP synthesis in that it increased this flux by 90% over basal rates in the control participants. In contrast, insulin-stimulated rates of ATP synthase flux were severely blunted in the IR offspring where a similar increase in plasma insulin concentrations increased this flux by only 5% over basal rates. These data demonstrate that the defects in basal rates of mitochondrial phosphorylation activity, previously found in the IR offspring, are even more pronounced under insulin-stimulated conditions and are consistent with a recent in vitro study that found reduced insulin-stimulated rates of oxidative phosphorylation activity in isolated mitochondria obtained from muscle biopsy samples of participants with type 2 diabetes [25]. Taken together these data suggest that the observed defect in insulin-stimulated rates of ATP synthesis is a very early defect in the pathogenesis of insulin resistance and type 2 diabetes.
Insulin also stimulates intramyocellular Pi transport into skeletal muscle, which accounts for much of its hypophosphatemic effects in vivo. Previous in vitro studies by Polgreen et al. [26] have demonstrated that insulin stimulates Na-dependent intramyocellular Pi influx in a mouse myoblast cell line in a relatively rapid manner that is independent of protein synthesis. These authors speculated that insulin regulates intramyocellular Pi transport into skeletal muscle by recruitment of intramyocellular Pi transporters to the cell surface in a fashion similar to insulin regulation of glucose transport activity in skeletal muscle. Using 31P MRS to assess intramyocellular Pi content before and during the hyperinsulinemic-euglycemic clamp, we found a marked reduction of insulin-stimulated intramyocellular Pi transport into skeletal muscle of the IR offspring. The parallel reductions in insulin-stimulated rates of muscle glucose metabolism, which can mostly be attributed to reduced insulin-stimulated glucose transport activity [27], and reduced insulin-stimulated phosphate transport activity suggests that a common insulin-signaling defect may be responsible for both of these abnormalities [28]. We also observed a strong correlation between insulin-stimulated increases in intramyocellular phosphate concentrations and insulin-stimulated increases in rates of ATP synthesis in control participants and IR offspring. Inorganic phosphate is the substrate for the phosphorylation of adenosine diphosphate to ATP during oxidative phosphorylation, and it is also a putative cytosolic signaling molecule in the regulation of this process [29,30]. Recent in vitro studies in isolated heart mitochondria have demonstrated that inorganic phosphate regulates oxidative phosphorylation by influencing both the proton motive force and the rate of ATP production [31]. Taken together these data suggest that insulin regulation of phosphate transport may be an important mechanism by which insulin regulates ATP synthesis in human skeletal muscle. If this is true, then it is possible that the observed reductions in insulin-stimulated rates of muscle ATP synthesis, which were observed in the IR offspring, may be secondary to the increased IMCL content that results in impaired insulin signaling and decreased insulin-stimulated phosphate transport [9–12]. While it is also possible that reduced insulin-stimulated muscle capillary recruitment [32–34] might be responsible for the observed reductions in insulin stimulation of ATP synthesis and intramyocellular Pi transport in the IR offspring, recent studies in humans [27,35–37] have shown that the majority of insulin resistance can be accounted for by a decrement in insulin's stimulatory effects on cellular glucose uptake and that stimulation of blood flow and capillary recruitment plays only a minor role in mediating insulin effects on glucose metabolism in human skeletal muscle. Therefore it appears unlikely that reductions in insulin-stimulated muscle capillary recruitment have a major role in explaining the observed reductions in insulin-stimulated ATP synthesis and intramyocellular Pi transport in the IR offspring.
In summary, these data provide further evidence of mitochondrial dysfunction in skeletal muscle of IR offspring of parents with type 2 diabetes. These individuals also manifest severe defects in insulin-stimulated phosphate transport in skeletal muscle, which may contribute to their observed defects in insulin-stimulated rates of mitochondrial ATP synthesis. Given the potentially important role that intracellular phosphate might have on regulating mitochondrial ATP production, we propose that regulation of intramyocellular Pi transport into the muscle cell by insulin may be an important mechanism by which insulin regulates mitochondrial ATP synthesis in skeletal muscle.
Patient Summary
Background
Type 2 diabetes is one of the fastest growing chronic diseases. We know quite a bit about the risk factors that predispose people to diabetes (such as being overweight), but we do not understand well how the disease starts (usually in adults but increasingly in teenagers and even children).
Why Was This Study Done?
Children whose parents are diabetic have a higher risk of developing type 2 diabetes themselves. Some of them have the first signs of the disease (a condition called insulin resistance, which is a strong predictor for subsequent development of diabetes) while they are still young and lean. The group of Gerald Shulman (who is the senior author of this study) has studied such individuals to understand the causes and effects of their insulin resistance. Having previously found some abnormalities in the mitochondria (the parts of cells involved in generating energy), they studied mitochondrial function in more detail here.
What Did the Researchers Do and Find?
They compared seven young lean insulin-resistant adults with a family history of diabetes with seven lean young adults without diabetic family members and with normal insulin sensitivity. Specifically, they looked at energy generated in the muscles in response to insulin stimulation. They found that while insulin increases energy production in the muscles of the control individuals by approximately 90%, it had very little effect in the insulin-resistant individuals. They also studied the amount of inorganic phosphate (an essential trace element) in muscle cells of both groups, and how it was affected by insulin. They found that in control individuals, insulin results in an increase of phosphate transport into the muscle cells and that this is also much reduced in insulin-resistant individuals.
What Does This Mean?
Intracellular phosphate is a key regulator of energy generation, and this study provides more support for the notion that insulin resistance compromises proper functioning of energy generation in the mitochondria of muscle cells. This and similar studies should eventually result in a better understanding of the early events in diabetes. The hope is that this will then allow the development of therapies that might prevent the disease or treat its symptoms more effectively.
Where Can I Find More Information Online?
The following Web sites contain information on diabetes and insulin resistance.
US National Diabetes Clearinghouse (English and Spanish):
http://diabetes.niddk.nih.gov/dm/pubs/insulinresistance/
http://diabetes.niddk.nih.gov/spanish/indexsp.asp
Stanford University Web site on syndrome X, diabetes, and insulin resistance:
http://syndromex.stanford.edu/
Diabetes UK:
http://www.diabetes.org.uk/
American Diabetes Association (English and Spanish):
http://www.diabetes.org/
http://www.diabetes.org/espanol/default.jsp
Information on Dr. Shulman's research:
http://www.hhmi.org/research/investigators/shulman.html
The authors would like to thank Yanna Kosover, Mikhail Smolgovsky, Amy Dennean, Pritpal Rhandawa, M.D., and the staff of the Yale University–New Haven Hospital General Clinical Research Center for expert technical assistance with the studies and the volunteers for participating in this study. These studies were supported by grants from the United States Public Health Service: R01 AG-23686 (KFP), P01 DK-068229 (GIS), P30 DK-45735 (GIS), and M01 RR-00125 (Yale University–New Haven General Clinical Research Center). GIS is the recipient of a Distinguished Clinical Scientist Award from the American Diabetes Association and an investigator of the Howard Hughes Medical Institute. The funding agencies had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Citation: Petersen KF, Dufour S, Shulman GI (2005) Decreased insulin-stimulated ATP synthesis and phosphate transport in muscle of insulin-resistant offspring of type 2 diabetic parents. PLoS Med 2(9): e233.
Abbreviations
ATPadenosine triphosphate
BMIbody mass index
IMCLintramyocellular lipid
IRinsulin-resistant
MRSmagnetic resonance spectroscopy
PCrphosphocreatinine
Piinorganic phosphate
==== Refs
References
Zimmet P Alberti KG Shaw J Global and societal implications of the diabetes epidemic Nature 2001 414 782 787 11742409
Wild S Roglic G Green A Sicree R King H Global prevalence of diabetes: Estimates for the year 2000 and projections for 2030 Diabetes Care 2004 27 1047 1053 15111519
Lillioja S Mott DM Howard BV Bennett PH Yki-Jarvinen H Impaired glucose tolerance as a disorder of insulin action. Longitudinal and cross-sectional studies in Pima Indians N Engl J Med 1988 318 1217 1225 3283552
Lillioja S Mott DM Spraul M Ferraro R Foley JE Insulin resistance and insulin secretory dysfunction as precursors of non-insulin-dependent diabetes mellitus. Prospective studies of Pima Indians N Engl J Med 1993 329 1988 1992 8247074
Warram JH Martin BC Krolewski AS Soeldner JS Kahn CR Slow glucose removal rate and hyperinsulinemia precede the development of type II diabetes in the offspring of diabetic parents Ann Intern Med 1990 113 909 915 2240915
Perseghin G Ghosh S Gerow K Shulman GI Metabolic defects in lean nondiabetic offspring of NIDDM parents: A cross-sectional study Diabetes 1997 46 1001 1009 9166672
Perseghin G Scifo P De Cobelli F Pagliato E Battezzati A Intramyocellular triglyceride content is a determinant of in vivo insulin resistance in humans: A 1H-13C nuclear magnetic resonance spectroscopy assessment in offspring of type 2 diabetic parents Diabetes 1999 48 1600 1606 10426379
Pan DA Lillioja S Kriketos AD Milner MR Baur LA Skeletal muscle triglyceride levels are inversely related to insulin action Diabetes 1997 46 983 988 9166669
Griffin ME Marcucci MJ Cline GW Bell K Barucci N Free fatty acid-induced insulin resistance is associated with activation of protein kinase C theta and alterations in the insulin signaling cascade Diabetes 1999 48 1270 1274 10342815
Dresner A Laurent D Marcucci M Griffin ME Dufour S Effects of free fatty acids on glucose transport and IRS-1-associated phosphatidylinositol 3-kinase activity J Clin Invest 1999 103 253 259 9916137
Yu C Chen Y Cline GW Zhang D Zong H Mechanism by which fatty acids inhibit insulin activation of IRS-1 associated phosphatidylinositol 3-kinase activity in muscle J Biol Chem 2002 277 50230 50236 12006582
Itani SI Ruderman NB Schmieder F Boden G Lipid-induced insulin resistance in human muscle is associated with changes in diacylglycerol, protein kinase C, and IkappaB-alpha Diabetes 2002 51 2005 2011 12086926
Boden G Chen X Ruiz J White JV Rossetti L Mechanisms of fatty acid-induced inhibition of glucose uptake J Clin Invest 1994 93 2438 2446 8200979
Roden M Price TB Perseghin G Petersen KF Rothman DL Mechanism of free fatty acid-induced insulin resistance in humans J Clin Invest 1996 97 2859 2865 8675698
Petersen KF Dufour S Befroy D Garcia R Shulman GI Impaired mitochondrial activity in the insulin-resistant offspring of patients with type 2 diabetes N Engl J Med 2004 350 664 671 14960743
Mootha VK Lindgren CM Eriksson KF Subramanian A Sihag S PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes Nat Genet 2003 34 267 273 12808457
Patti ME Butte AJ Crunkhorn S Cusi K Berria R Coordinated reduction of genes of oxidative metabolism in humans with insulin resistance and diabetes: Potential role of PGC1 and NRF1 Proc Natl Acad Sci U S A 2003 100 8466 8471 12832613
Baecke JA Burema J Frijters JE A short questionnaire for the measurement of habitual physical activity in epidemiological studies Am J Clin Nutr 1982 36 936 942 7137077
Diamond MP Jacob R Connolly-Diamond M DeFronzo RA Glucose metabolism during the menstrual cycle. Assessment with the euglycemic, hyperinsulinemic clamp J Reprod Med 1993 38 417 421 8331618
Miles J Glasscock R Aikens J Gerich J Haymond M A microfluorometric method for the determination of free fatty acids in plasma J Lipid Res 1983 24 96 99 6833886
Lebon V Dufour S Petersen KF Ren J Jucker BM Effect of triiodothyronine on mitochondrial energy coupling in human skeletal muscle J Clin Invest 2001 108 733 737 11544279
Mayerson AB Hundal RS Dufour S Lebon V Befroy D The effects of rosiglitazone on insulin sensitivity, lipolysis, and hepatic and skeletal muscle triglyceride content in patients with type 2 diabetes Diabetes 2002 51 797 802 11872682
Petersen KF Hendler R Price T Perseghin G Rothman DL 13C/31P NMR studies on the mechanism of insulin resistance in obesity Diabetes 1998 47 381 386 9519743
Lusk G Animal calorimetry: Analysis of the oxidation of mixtures of carbohydrates and fat: A correction J Biol Chem 1924 59 41 42
Stump CS Short KR Bigelow ML Schimke JM Nair KS Effect of insulin on human skeletal muscle mitochondrial ATP production, protein synthesis, and mRNA transcripts Proc Natl Acad Sci U S A 2003 100 7996 8001 12808136
Polgreen KE Kemp GJ Leighton B Radda GK Modulation of Pi transport in skeletal muscle by insulin and IGF-1 Biochim Biophys Acta 1994 1223 279 284 8086500
Cline GW Petersen KF Krssak M Shen J Hundal RS Impaired glucose transport as a cause of decreased insulin-stimulated muscle glycogen synthesis in type 2 diabetes N Engl J Med 1999 341 240 246 10413736
Shulman GI Cellular mechanisms of insulin resistance J Clin Invest 2000 106 171 176 10903330
Chance B Williams GR The respiratory chain and oxidative phosphorylation Adv Enzymol Relat Subj Biochem 1956 17 65 134 13313307
Lardy HA Wellman H Oxidative phosphorylations: Role of inorganic phosphate and acceptor systems in control of metabolic rates J Biol Chem 1952 195 215 224 14938372
Bose S French S Evans FJ Joubert F Balaban RS Metabolic network control of oxidative phosphorylation: Multiple roles of inorganic phosphate J Biol Chem 2003 278 39155 39165 12871940
Coggins M Lindner J Rattigan S Jahn L Fasy E Physiologic hyperinsulinemia enhances human skeletal muscle perfusion by capillary recruitment Diabetes 2001 50 2682 2690 11723050
Gudbjornsdottir S Sjostrand M Strindberg L Lonnroth P Decreased muscle capillary permeability surface area in type 2 diabetic subjects J Clin Endocrinol Metab 2005 90 1078 1082 15536160
Sjostrand M Gudbjornsdottir S Strindberg L Lonnroth P Delayed transcapillary delivery of insulin to muscle interstitial fluid after oral glucose load in obese subjects Diabetes 2005 54 152 157 15616023
Parsonage W Hetmanski D Cowley A Differentiation of the metabolic and vascular effects of insulin in insulin resistance in patients with chronic heart failure Am J Cardiol 2002 89 696 703 11897212
Laine H Yki-Jarvinen H Kirvela O Tolvanen T Raitakari M Insulin resistance of glucose uptake in skeletal muscle cannot be ameliorated by enhancing endothelium-dependent blood flow in obesity J Clin Invest 1998 101 1156 1162 9486987
Utriainen T Nuutila P Takala T Vicini P Ruotsalainen U Intact insulin stimulation of skeletal muscle blood flow, its heterogeneity and redistribution, but not of glucose uptake in non-insulin-dependent diabetes mellitus J Clin Invest 1997 100 777 785 9259575
|
16089501
|
PMC1184227
|
CC BY
|
2021-01-05 10:40:31
|
no
|
PLoS Med. 2005 Sep 16; 2(9):e233
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020233
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 10.1371/journal.pmed.0020292SynopsisPhysiologyDiabetes/Endocrinology/MetabolismMedical ImagingDiabetesNutrition and MetabolismGeneticsMedical ImagingType 2 Diabetes: Insulin Resistance May Be the Result of Mitochondrial Dysfunction Synopsis9 2005 16 8 2005 2 9 e292Copyright: © 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.
Decreased Insulin-Stimulated ATP Synthesis and Phosphate Transport in Muscle of Insulin-Resistant Offspring of Type 2 Diabetic Parents
Insulin Resistance in the Offspring of Parents with Type 2 Diabetes
==== Body
The role of insulin resistance (IR) in type 2 diabetes, the most frequently encountered metabolic disorder in the world, has attracted much attention in recent years. Virtually all patients with type 2 diabetes have IR, which usually appears some 10–20 years before the disease itself. Although the existence of the relationship between IR and type 2 diabetes is well recognized, the underlying mechanisms are poorly understood. A study by Kitt Falk Petersen and colleagues provides important new information on the underlying pathogenic mechanisms that lead to the development of IR.
Recent findings have suggested that inherited defects in mitochondrial oxidative phosphorylation activity might play a key role in the development of IR. Studies have demonstrated a relationship between dysregulated fatty acid metabolism, fat accumulation in muscle cells, and IR in skeletal muscle. This fat accumulation appears to interfere with insulin signaling, resulting in reduced insulin-stimulated muscle glucose transport activity and decreased muscle glycogen synthesis. Magnetic resonance spectroscopy (MRS) studies have found reduced basal rates of muscle mitochondrial ATP production, associated with increased intramyocellular lipid content. Consistent with these results, microarray studies have found a coordinated reduction in PGC-1á-responsive genes in patients with obesity and type 2 diabetes and their overweight first-degree relatives.
Building on these findings, Petersen and colleagues examined insulin-stimulated rates of muscle ATP synthesis and phosphate transport, to investigate whether mitochondrial function is affected not only in the fasting state but also during insulin stimulation and to determine whether inherited defects in mitochondrial oxidative phosphorylation activity might be responsible for IR.
Participants in the study were in their late 20s, lean, nonsmoking individuals with IR who were offspring of parents with type 2 diabetes. The participants were thus selected to be free of other risk factors for IR such as obesity and smoking. Such individuals are ideal for examining the earliest metabolic defects responsible for the pathogenesis of IR since they have none of the confounding factors that are likely to be present in patients with type 2 diabetes. A metabolic defect in this group is likely to be an early event of genetic origin and, therefore, is potentially a primary cause of the later development of diabetes. The individuals in the control group were healthy, nonsmoking, and matched for age and weight to the individuals in the IR group. Insulin was administered using a hyperinsulinemic-euglycemic clamp, and rates of mitochondrial phosphorylation activity (muscle ATP synthesis) were assessed by MRS.
Mitochondria in human muscle cell
Rates of insulin-stimulated glucose uptake were decreased by approximately 50% in the individuals with IR compared to the controls and were associated with an approximately 2-fold increase in intramyocellular lipid content. In the control individuals, rates of ATP synthesis increased by approximately 90% during the hyperinsulinemic-euglycemic clamp. In contrast, insulin-stimulated rates of muscle mitochondrial ATP synthesis increased by only 5% in the individuals with IR. This small increase in muscle mitochondrial ATP synthesis in the individuals with IR was associated with a severe reduction of insulin-stimulated increases in intramyocellular phosphorus concentrations.
The authors say their study provides further evidence that IR in skeletal muscle of individuals with IR who are offspring of parents with type 2 diabetes may be related to defects in mitochondrial dysfunction. Furthermore, there were also severe defects in insulin-stimulated phosphate transport into skeletal muscle in the individuals with IR, which may be part of the defect leading to impaired ATP synthesis in the muscle of these individuals.
The implications of the study are also discussed in a Perspective by Anton Wagenmakers (DOI: 10.1371/journal.pmed.0020289). Although he suggests that the defects underlying IR are likely also to involve organs other than muscle, he notes the clinical relevance of the main finding, which might explain the weight maintenance problems that individuals with obesity and IR have.
|
0
|
PMC1184228
|
CC0
|
2021-01-05 10:40:31
|
no
|
PLoS Med. 2005 Sep 16; 2(9):e292
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020292
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1609289110.1371/journal.pbio.0030277Research ArticleMolecular Biology/Structural BiologyBiochemistryIn VitroFunctional Identification of Catalytic Metal Ion Binding Sites within RNA Catalytic Metal Ion Sites within RNAHougland James L
1
Kravchuk Alexander V
2
Herschlag Daniel [email protected]
2
Piccirilli Joseph A [email protected]
1
3
4
1Department of Chemistry, University of Chicago, Illinois, United States of America,2Department of Biochemistry, Stanford University, California, United States of America,3Department of Biochemistry and Molecular Biology, University of Chicago, Illinois, United States of America,4Howard Hughes Medical Institute, University of Chicago, Illinois, United States of AmericaJoyce Gerald Academic EditorDepartment of Molecular Biology, Scripps Research InstituteUnited States of America9 2005 16 8 2005 16 8 2005 3 9 e2771 2 2005 9 6 2005 Copyright: © 2005 Hougland 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 Ol' Switcheroo Shows How an RNA Enzyme Splices Itself
The viability of living systems depends inextricably on enzymes that catalyze phosphoryl transfer reactions. For many enzymes in this class, including several ribozymes, divalent metal ions serve as obligate cofactors. Understanding how metal ions mediate catalysis requires elucidation of metal ion interactions with both the enzyme and the substrate(s). In the Tetrahymena group I intron, previous work using atomic mutagenesis and quantitative analysis of metal ion rescue behavior identified three metal ions (MA, MB, and MC) that make five interactions with the ribozyme substrates in the reaction's transition state. Here, we combine substrate atomic mutagenesis with site-specific phosphorothioate substitutions in the ribozyme backbone to develop a powerful, general strategy for defining the ligands of catalytic metal ions within RNA. In applying this strategy to the Tetrahymena group I intron, we have identified the pro-S
P phosphoryl oxygen at nucleotide C262 as a ribozyme ligand for MC. Our findings establish a direct connection between the ribozyme core and the functionally defined model of the chemical transition state, thereby extending the known set of transition-state interactions and providing information critical for the application of the recent group I intron crystallographic structures to the understanding of catalysis.
A combination of substrate atomic mutagenesis with site-specific substitutions in the ribozyme backbone allow the ligands of catalytic metal ions to be identified.
==== Body
Introduction
Phosphoryl transfer reactions occur ubiquitously in biology, playing roles in gene replication, recombination, and expression. To orchestrate these central biological processes, biomacromolecules have harnessed the catalytic power of divalent metal ions [1–10]. The chemistry underlying these cellular events hinges critically on the metal ion interactions that occur during function. However, there exists little functional data defining the coordination environment of individual metal ions during catalysis and correspondingly few methods for obtaining such information.
Structural analyses can serve as a powerful starting point for investigation of catalytic metal ions, but these approaches provide no direct information about transition-state interactions. Reflecting this uncertainty, for a given enzyme or enzyme family the number of active-site metal ions and their proposed interactions during catalysis can vary with both enzyme and observer ([2–10] and references therein). For example, restriction enzymes have been crystallized with zero, one, two, or three metal ions in their active sites [7,8,10], and different RNase H crystal structures support either a one or two metal ion mechanism [3]. This variability in metal ion binding also occurs for RNA. In structures of the hammerhead ribozyme and tRNA, Mg2+ and Mn2+ occupy different sites [11,12], and recent group I intron structures exhibit differences in the number, location, and identity of active-site metal ions [13–15]. Thus, only the combination of structural and functional studies can establish metalloenzyme mechanisms unambiguously.
To understand how metalloenzymes utilize the catalytic power of metal ions, we must identify individual catalytic metal ions functionally, determine their relationships to each other, define their coordination environment, and establish the network of interactions that position the coordinating groups. Metal ion rescue experiments using substrate mutations offer a strategy to identify specific catalytic metal ion interactions within enzyme active sites, a particularly formidable challenge in ribozymes due to the sea of metal ions that interact electrostatically with the anionic phosphodiester backbone. Functionally deleterious sulfur or nitrogen perturbations that exhibit rescue upon increasing cation softness (e.g., replacing Mg2+ with Cd2+ or Mn2+) suggest direct metal ion coordination during catalysis [16,17]. This approach has revealed catalytic metal ion interactions with enzyme substrates in the Tetrahymena group I ribozyme [18–21], the ai5γ group II intron [22,23], RNase P [24], the hammerhead ribozyme [25,26], the human spliceosome [27,28], and many protein enzymes (e.g., [16,29–32] and references therein).
The Tetrahymena group I ribozyme catalyzes nucleotidyl transfer from an oligonucleotide substrate that mimics the natural 5′-splice site to an exogenous guanosine (G) that serves as the nucleophile in a reaction analogous to the first step of group I intron self-splicing (Equation 1) [33,34].
Metal ion rescue experiments have identified four atoms within the oligonucleotide substrate and G nucleophile that interact with metal ions in the chemical transition state [18–21]. To determine whether one or several distinct metal ions mediate these interactions, Shan et al. developed thermodynamic fingerprint analysis, quantitatively analyzing the reactivity of modified substrates relative to unmodified substrates over a range of rescuing metal ion concentrations [35]. In this approach, the reactions for both modified and native substrates start from the same ground state and monitor the same elementary reaction steps. The resulting rescue profiles serve as distinctive “fingerprints” for the rescuing metal ion(s), revealing by comparison whether the same or distinct metal ions interact with the identified substrate ligands. Thermodynamic fingerprint analysis and related analyses [36] using a series of substrates bearing single or multiple atomic perturbations have provided functional evidence for a network of three distinct metal ions within the Tetrahymena ribozyme active site (Figure 1), making a total of five interactions with the reaction's transition state. Metal ions coordinate to the 3′-oxygen leaving group (MA), the 3′-oxygen on the G nucleophile (MB), and the 2′-hydroxyl of the G nucleophile (MC). Two of these metal ions (MA and MC) also contact the pro-S
P oxygen of the scissile phosphate [35,36]. However, the ligands within the ribozyme core architecture that bind and position these metal ion cofactors remain largely unknown, leaving the catalytic metal ion coordination environments undefined.
Figure 1 Model of the Tetrahymena Ribozyme Transition State during the First Step of Splicing
The three identified catalytic metal ions (MA, MB, and Mc) and their transition-state interactions with the U(-1) 3′-oxygen, G nucleophile 3′-oxygen, and G 2′-OH, respectively, along with the MA and MC interactions with the scissile phosphate pro-S
P oxygen, are shown by red dots [35,36]. The 2′-OH of U(-1) participates in a hydrogen-bonding network with the 2′-OH of A207 and the exocyclic amine of the G·U wobble pair [58,72] and donates a hydrogen bond to the adjacent 3′-oxygen in the transition state [71]; the hydrogen bonds are shown as hashed lines.
The non-bridging phosphate oxygens of the RNA backbone commonly serve as ligands for divalent metal ions. For the Tetrahymena group I ribozyme and other RNA enzymes, phosphorothioate interference studies have generated a plethora of ligand candidates for metal ions [17,26,37–52]. However, there have been few attempts to link these putative ligands to metal ions directly involved in catalysis [42,53,54]. Using the Tetrahymena group I ribozyme as a model system, we have combined thermodynamic fingerprint analysis with an array of atomically perturbed substrates and ribozyme site- and stereo-specific phosphorothioate mutations to develop a general functional approach for identifying ligands for the catalytic metal ions. Our findings establish a direct connection between the ribozyme core and the functionally defined model of the chemical transition state, thereby providing information critical for the application of the recent group I intron crystallographic structures to the understanding of catalysis.
Results
Choosing Sites for Phosphorothioate Substitution within the Ribozyme Core
Backbone mutation sites were chosen prior to the release of the recently reported group I intron structures [13–15]. To guide our choice of substitution sites, we focused on previously reported interferences arising from random R
P-phosphorothioate incorporation into the phylogenetically conserved core regions of the Tetrahymena group I intron. As Mg2+ coordinates poorly to sulfur, the R
P-phosphorothioate interferences could reflect direct disruption of a metal ion interaction with the pro-R
P phosphate oxygen, indirect disruption of a metal interaction with the geminal pro-S
P phosphate oxygen, or other effects. A literature survey identified 14 sites of R
P-phosphorothioate interference within the ribozyme's conserved core [17,47,52,55]. Herein we analyze ribozymes containing site-specific phosphorothioate incorporation at six sites within the J6/7 and P7 regions of the ribozyme (Figure 2A). We constructed these mutant ribozymes semi-synthetically with both R
P- and S
P-phosphorothioate mutations at these six sites (Figure 2B), resulting in 12 variant ribozymes.
Figure 2 Construction of Ribozymes Containing Site-Specific Phosphorothioate Substitutions
(A) Secondary structure of the Tetrahymena group I ribozyme. The ribozyme conserved core is highlighted in blue, and the six positions of phosphorothioate substitution within the J6/7 and P7 regions are labeled in green. The internal guide sequence at the 5′-terminus of the ribozyme is labeled “IGS.” (B) Ligation strategy for constructing mutant ribozymes. Following ion exchange HPLC purification of the phosphorothioate diastereomers (C), the phosphorothioate-substituted mutation oligonucleotides were ligated into full-length ribozymes by two successive-splint mediated ligations using T4 DNA ligase [77], as described in Materials and Methods
.
The variant ribozymes were characterized kinetically within the known framework of the Tetrahymena ribozyme reaction (Figure 3; [33,56,57] and references therein). The oligonucleotide substrate (S; Table 1) binds to the ribozyme (E) in two steps. First, S forms Watson–Crick base pairs with the ribozyme's internal guide sequence (see Figure 2A) to give the open complex (E·S)O. The resulting P1 helix then “docks” into the ribozyme core via tertiary interactions, forming the closed complex (E·S)C ([33,57–59] and references therein). G binds to give the ternary (E·S·G)C complex, and the reaction proceeds through the phosphoryl transfer step (k
chem), resulting in cleavage of the oligonucleotide substrate.
Figure 3 The Tetrahymena Ribozyme Reaction Pathway
The ribozyme binds the oligonucleotide substrate (in two steps) and the exogenous G that serves as the nucleophile in the ribozyme reaction as described in the text ([33,56,57] and references therein). K
G
d(E), K
G
d(E·S)O, and K
G
d(E·S)c are G dissociation constants from free E, (E·S)O, and (E·S)C, respectively. K
IGS
d is the dissociation constant for the oligonucleotide substrate from the internal guide sequence, and K
dock and K′dock are the docking equilibria for the E·S and E·S·G complexes, respectively. K
S
d(E) and K
S
d(E·G) are the observed dissociation constants for the oligonucleotide substrate from free E and the E·G complex, respectively, and k
chem is the observed rate of oligonucleotide substrate cleavage.
Table 1 Oligonucleotide Substrates Used Herein
We first tested whether Cd2+, a thiophilic metal ion that can adopt octahedral coordination geometry like Mg2+ [60–62], stimulates the ability of the phosphorothioate containing ribozymes to catalyze oligonucleotide substrate cleavage (Figure 4). Under conditions of saturating ribozyme and G (10 mM MgCl2), several of the phosphorothioates affected catalysis significantly (data not shown, and see Table 2 below), but upon addition of 0.1–1.0 mM Cd2+, only one of the variant ribozymes, the C262-S
P variant, experienced significant stimulation (Figure 4 and data not shown). This stimulation suggested that a functionally important Cd2+ phosphorothioate interaction may occur at the C262-S
P position, leading us to focus predominantly on this variant ribozyme. The other positions tested, in principle, remain viable ligand candidates as the absence of rescue cannot be taken as evidence that the modification site does not serve as a metal ion ligand. However, metal ion rescue experiments analogous to those described below for the C262-S
P ribozyme suggest that none of the other phosphorothioate substitutions have large effects on binding of the known catalytic metal ions (see Figure S1).
Figure 4 The Effect of Cd2+ on Activity of Phosphorothioate-Containing Ribozymes
Reactions contained saturating G and ribozyme, and monitored cleavage of −1d,rSA5 (see Table 1) in the presence or absence of 0.1 mM Cd2+ (10 mM Mg2+ background). The largest Cd2+ stimulation occurred with the C262-S
P variant ribozyme. Significant Cd2+ stimulation remained unique to this ribozyme at all Cd2+ concentrations tested (0.1–10 mM; data not shown). Concentrations of Cd2+ above 1 mM inhibit the ribozyme reaction with a steep concentration dependence (see, e.g., Figures S4 and S5A). The small Cd2+ stimulation observed for the U258 variant ribozymes is consistent with coordination of a metal ion important for structural stability, as proposed by Lindqvist et al. [78].
Table 2 G Binding and Reactivity for Ribozymes with Site-Specific Phosphorothioate Substitutions within J6/7 and P7
Cd2+ Accelerates a Non-Chemical Step in the C262-S
P Ribozyme Reaction
Metal ion rescue experiments provide definitive information about transition-state metal ion-ligand interactions only when conducted under conditions in which the same reaction steps are monitored and the chemical step limits the reaction rate [35]. To learn more about the apparent Cd2+ stimulation and to establish appropriate reaction conditions under which the chemical step could be monitored, we undertook basic characterization of the C262-S
P ribozyme. We first present data concerning the rate-limiting step of the C262-S
P ribozyme reaction, and then assess the effects of this mutation on individual reaction steps.
Previous work established that, in the wild-type (WT) ribozyme, the chemical step follows a pre-equilibrium loss of a proton, presumably from the 3′-hydroxyl of the attacking G (see Figure 1). This leads to a log-linear pH dependence with a slope of one under conditions of rate-limiting chemistry [33,63–65]. However, in the absence of Cd2+, the pH dependence for cleavage of the −1d,rSA5 substrate (Table 1) by the C262-S
P variant ribozyme is essentially flat above pH 5 (Figure 5A, closed circles), in contrast to the slope of one for the WT ribozyme [63,65]. This difference suggests that, under the reaction conditions, a conformational change limits the rate for the variant ribozyme. In the presence of low concentrations of Cd2+, the variant behaves more like the WT ribozyme, exhibiting a log-linear pH dependence with slope one (Figure 5A, open circles). This Cd2+-induced change in pH dependence suggests that Cd2+ accelerates the non-chemical step sufficiently to render the chemical step rate-limiting throughout the entire pH range.
Figure 5 The pH Dependence of C262-S
P Ribozyme Activity Reveals a Non-Chemical Rate-Limiting Step
(A) Cleavage of −1d,rSA5 under saturating conditions, as given in Figure 3. In the absence of Cd2+ (·), the pH dependence flattens above pH ∼4.7; the data are fit to k
max = 1/(1+10n(pKa-pH)). Addition of 1 mM Cd2+ (s○) leads to log-linear pH dependence (slope = 1.1 ± 0.1) over the pH range of 4–7, consistent with the chemical step being rate-limiting [33,63–65]. The Cd2+ rescue observed in Figure 3 results from acceleration of the non-chemical step, as indicated by the vertical arrow, such that this non-chemical step is no longer rate-limiting. The increased slope observed at low pH for reactions both with and without Cd2+ is consistent with ribozyme inhibition due to multiple independent protonations [65].
(B) Cleavage of −1r,dSA5 under saturating ribozyme and G conditions. Cleavage of −1r,dSA5 is log-linearly dependent on pH with and without added Cd2+ (0 mM Cd2+, • ; 1 mM Cd2+, s○) (best fit slopes of 0.8 in both cases), consistent with the chemical step being rate-limiting in both cases.
Although the molecular basis for the rate-limiting, non-chemical step and its stimulation by Cd2+ in the C262-S
P ribozyme reaction remains undefined, comparison to the reaction of −1r,dSA5 offers some insight. The −1r,dSA5 substrate reacts from the ternary complex (E·S·G) with a log-linear pH dependence whose slope approximates one, even in the absence of Cd2+ (Figure 5B, closed circles). Addition of Cd2+ has no effect on the reaction rate or the pH-dependence (Figure 5B, open circles). These results suggest that the chemical step limits the reaction of −1r,dSA5 both in the presence and absence of Cd2+. As −1r,dSA5 binds primarily in the open complex with the WT ribozyme (see Figure 3) [36,66] and −1d,rSA5 binds in the closed complex [57,58], the rate-limiting non-chemical step observed with −1d,rSA5 may reflect phosphorothioate-induced formation of an altered closed complex that must rearrange prior to the chemical step. Acceleration of this non-chemical step depends on Cd2+; increasing the Mg2+ concentration does not accelerate this step, nor does Mn2+ or Zn2+ (data not shown). The Cd2+ dependence suggests that this non-chemical step may require metal ion coordination to the backbone phosphorothioate. Future study of this metal-ion-dependent non-chemical step revealed by phosphorothioate incorporation may provide an opportunity for deeper understanding of the role of metal ion coordination in conformational rearrangements within RNA.
This initial characterization of the C262-S
P ribozyme allowed us to choose appropriate reaction conditions for further characterization. For analyzing the binding of G and oligonucleotide substrates to the variant ribozyme, we included Cd2+ in sufficient but subsaturating concentrations to render the chemical step rate-limiting without influencing the measured binding constants (i.e., subsaturating Cd2+ in order to have no significant effect on substrate affinities; [Cd2+] ≤ 1 mM) [35]. For metal ion rescue reactions in which a constant Cd2+ concentration was not feasible, we chose oligonucleotide substrates (Table 1) or reaction pHs that render the chemical step rate-limiting.
Phosphorothioate Substitution Effects on Reactivity and Binding
We analyzed the 12 phosphorothioate-substituted variants within the J6/7 and P7 regions of the ribozyme (see Figure 2A) under a variety of reaction conditions. As noted above, reactions with the C262-S
P variant contained Cd2+ to maintain rate-limiting chemistry. Cd2+ had no significant effect on the behavior of the other ribozymes (data not shown). Representative data from this survey—phosphorothioate substitution effects on the reaction E·S·G → products (k
ternary) and G binding to the E·S complex —are given in Table 2. This survey revealed one position at which phosphorothioate substitution strongly perturbs (greater than 10-fold) G binding: the S
P phosphorothioate at nucleotide C262, the same variant that exhibited Cd2+ stimulation.
Based on the G-binding results, the Cd2+ rescue of a non-chemical step, and screens for the effects of J6/7 and P7 phosphorothioates on rescue by the catalytic metal ions (see Figure S1), we decided to characterize the C262-S
P ribozyme further, focusing on how the phosphorothioate affects the individual reaction steps (see Figure 3). We found that the C262-S
P substitution had no significant effect on oligonucleotide substrate association or dissociation (Table S1). We then further investigated the observed effect of the C262-S
P phosphorothioate substitution on G binding. In reactions with the −(1–3)d,rSA5 substrate, which binds to the WT ribozyme in the open complex (E·S)O (based on observed G-binding affinity [56,67] and Mn2+ stimulation of cleavage activity [66] [data not shown]), G binds to the C262-S
P variant and the WT ribozyme with the same affinity. Thus, in the absence of docked oligonucleotide substrate, the C262-S
P phosphorothioate has no effect on G binding (Table 3). The C262-S
P phosphorothioate affects subsequent reaction steps, however. The ternary open complex, (E·S·G)O, for the variant reacts about 40-fold slower than that for the WT ribozyme, suggesting possible effects on oligonucleotide substrate docking into the G-bound active site or on the chemical step.
Table 3 G Binding and Reactivity with WT and C262-S
P Ribozymes
In the presence of WT ribozyme, G and S bind cooperatively to give the ternary closed complex, (E·S·G)C. G binds to (E·S)C five- to ten-fold tighter than to (E·S)O or free enzyme [56,67]. Using −1d,rSA5 to form (E·S)C and −(1–3)d,rSA5 to form (E·S)O, we reproduced this coupling for the WT ribozyme (Table 3). In contrast, the C262-S
P ribozyme exhibited no coupling. Indeed, the presence of docked oligonucleotide substrate with the C262-S
P ribozyme weakens G binding. Overall, the C262-S
P ribozyme closed complex has an approximately 40-fold weaker affinity for G than does the WT ribozyme closed complex (Table 3). Reaction chemistry from the ternary closed complex remains unaffected, as (E·S·G)C formed with −1d,rSA5 has the same rate for the WT and C262-S
P ribozymes (Table 3).
Previous work has linked the coupled binding of G and the oligonucleotide substrate to MC, which mediates contacts to the G 2′-hydroxyl and the scissile phosphate pro-S
P oxygen (see Figure 1) [68,69]. The loss of coupled binding between the oligonucleotide substrate and G induced upon C262-S
P phosphorothioate substitution therefore raised the possibility that the C262 pro-S
P phosphoryl oxygen resides near the MC binding site in the ribozyme tertiary structure.
A Linkage between C262-S
P and One of the Catalytic Metal Ions
The apparent functional connection between the C262 pro-S
P phosphoryl oxygen atom and the metal ion bound at site C (MC) could occur through direct interaction of this oxygen with MC, or indirectly through a chain of interactions. We used the C262-S
P ribozyme to ascertain whether direct coordination occurs, analyzing the metal ion rescue behavior for each of the previously identified metal ion sites. If the C262-S
P phosphorothioate directly coordinates to a Cd2+ ion bound at one of the known catalytic metal ion binding sites, we expect the stronger Cd2+–sulfur interaction to shift the metal ion rescue profile toward lower Cd2+ concentration relative to the WT ribozyme.
We determined the MA, MB, and MC profiles for the WT and C262-S
P ribozymes according to approaches described previously [35,36,68]. Table 4 lists the substrates and kinetic regimes for the reactions used to obtain each rescue profile. For each rescue profile, WT and mutant ribozyme reactions with both modified and unmodified substrates must start from the same ground state and monitor the chemical step. Unless indicated otherwise, we chose substrates and conditions so that the substrate bearing the modification is not bound within the active site in the starting ground state. This ensures that substrate modifications have no effect on ground-state binding of the rescuing metal ion to the WT and mutant ribozymes. The supporting information provides full details of the methods, models, and equations used to fit the Cd2+ rescue dependencies.
Table 4 Metal Rescue Assays Used to Probe Catalytic Metal Ions in the Tetrahymena Ribozyme
To monitor Cd2+ binding at the metal ion site A, we followed the reactivity of an oligonucleotide substrate containing a 3′-thiophosphoryl linkage at the cleavage site, Sm3′S (Figure 6A) [21,35]; i.e., Cd2+ specifically rescues the cleavage rate of Sm3′S relative to the unmodified 3′-oxygen oligonucleotide substrate (k
rel, k
3′S/k
3′O). The C262-S
P and WT ribozymes exhibit nearly identical profiles for rescue at metal ion site A (Figure 6A). The small deviation at high Cd2+ is beyond experimental error (as indicated by error bars in Figure 6A) and may reflect an indirect effect from Cd2+ occupancy at a different site (see Protocol S1 ). The lack of significant change in Cd2+ rescue at metal site A provides no evidence for direct contact between the S
P-phosphorothioate at residue 262 and a Cd2+ ion binding at metal site A. Therefore, we conclude that the C262 pro-S
P phosphoryl oxygen in the WT ribozyme is not a ligand for MA.
Figure 6 The C262-S
P Phosphorothioate Perturbs the Rescue Profile of MC
Cd2+ rescue profiles for MA, MB, and MC with WT (s○) and C262-S
P (▪) ribozymes are displayed. In each case, the specific metal ion-substrate contact being probed is indicated by closed red dots.
(A) MA rescue of Sm3′S cleavage. In the Tetrahymena ribozyme transition-state model, the oligonucleotide substrate 3′-thiophosphoryl modification at U(-1) is shown in green. MA rescue reactions monitoring cleavage of −(1–3d)rSA5 and Sm3′S were carried out under E·S·G → products (k
ternary) conditions as described in Materials and Methods
. The Cd2+ profiles for MA rescue are fit to a Hill equation, facilitating comparison between the WT and C262-S
P ribozyme profiles (see Protocol S1). The fits give Hill constants of 1 and 1.2 for the WT and C262-S
P ribozymes, respectively. Error bars not seen are obscured by the data symbols.
(B) MB rescue of CUCG3′SA cleavage. In the transition-state model, the G 3′-thio modification is shown in green. MB rescue reactions monitoring cleavage of CUCGA and CUCG3′SA were performed under E·P + CUCG3′XA → products [(k
c/K
m)CUCG(3′X)A] conditions, as described in Materials and Methods. MB rescue is fit to a model in which one Cd2+ ion binds to E·P and stimulates reaction of CUCG3′SA.
(C) MC rescue of −1r,dSA5 cleavage by subsaturating GN ((k
c/K
m)G(or GN
) conditions). In the transition-state model, the 2′-amino group of the G nucleophile is shown in blue. MC rescue of −1r,dSA5 cleavage under E·S + G(or GN) → products ((k
c/K
m)G(or GN
) conditions) was performed as described in Materials and Methods. k
rel data were fit to a model in which one Cd2+ ion binds and rescues reaction with GN.
(D) MC rescue of −1r,dSA5 cleavage by saturating GN. In the transition-state model, the 2′-amino group of the G nucleophile is again shown in blue. MC rescue of −1r,dSA5 cleavage under E·S·G/GN→ products (k
ternary) conditions was measured as described in Materials and Methods. k
rel data were fit to a model in which one Cd2+ ion binds and rescues reaction with GN; with the C262-S
P ribozyme, the fit to this model gives = 0.22 ± 0.08 mM. Cd2+ dependencies of observed cleavage rates in the MA, MB, and MC rescue reactions are displayed in Figures S2, S3, and S4, respectively.
To follow Cd2+ binding to the MB site, we monitored Cd2+ rescue of the reactivity of CUCG3′SA relative to CUCGA in the reverse reaction with E·P (Equation 2) [20,35]. CUCG3′SA contains a 3′-thiophosphoryl linkage at the cleavage site.
In reactions catalyzed by both the WT and C262-S
P ribozymes, Cd2+ specifically stimulates the reactivity of CUCG3′SA more than 500-fold. This rescue fits well to a model in which a single Cd2+ ion binds to the MB site and rescues the reaction (Figure 6B). As C262-S
P phosphorothioate incorporation exhibits no effect on the MB Cd2+ rescue profile, we conclude that the C262 pro-S
P phosphoryl oxygen in the WT ribozyme is not a ligand for MB.
To follow Cd2+ binding to the MC site, we monitored Cd2+ rescue of oligonucleotide substrate cleavage by 2′-aminoguanosine (GN) relative to G [19,35,68]. We first probed Cd2+ binding to the MC site in the E·S complex, conducting reactions at subsaturating G or GN concentrations ((k
c/K
m)G or (k
c/K
m
conditions). Specific Cd2+ stimulation of cleavage with GN occurs with both the WT and C262-S
P ribozymes, giving linear rescue curves throughout the Cd2+ concentration range tested with slopes that reflect a single Cd2+ stimulating cleavage by GN (Figure 6C). In contrast to the MA and MB rescue profiles discussed above, the MC rescue curve for the C262-S
P variant shifts to the left, with 16-fold lower Cd2+ concentrations required to achieve the same level of rescue as the WT. This shift suggests that the S
P-phosphorothioate at nucleotide C262 interacts directly with the Cd2+ ion binding at the MC site, a model we test further below.
The MC rescue profiles for both the WT and C262-S
P ribozymes increase linearly up to the highest experimentally accessible Cd2+ concentrations (Figure 6C), indicating that the rescuing Cd2+ ion binds to the E·S complexes of these ribozymes under these conditions with a dissociation constant that exceeds 10 mM. Without saturation behavior, we cannot definitively ascertain whether the C262 phosphorothioate-induced shift in the MC rescue profile emanates from tighter Cd2+ binding. Consequently, we probed Cd2+ binding to the MC site in the presence of saturating GN. Bound GN provides MC with an additional ligand, the nitrogen of the 2′-amino group, which should interact strongly with Cd2+ [70].
We conducted the rescue experiment as described above but included GN at saturating concentration to form the E·S·GN ternary complex. As with subsaturating GN, both the WT and C262-S
P ribozymes experienced significant specific Cd2+ rescue (Figure 6D). The rescue profile for the WT E·S·GN complex remains linear throughout the Cd2+ concentration range tested, indicating that even in the presence of bound GN, Cd2+ binds to the MC site with an apparent affinity of greater than 10 mM (50 mM Mg2+ background). In striking contrast, the Cd2+ rescue profile for reaction of the C262-S
P E·S·GN complex exhibits saturation behavior, with an apparent dissociation constant of ∼ 0.2 mM. The C262-S
P E·S·GN ternary complex therefore binds the rescuing Cd2+ ion more than 50-fold tighter than does C262-S
P ribozyme E·S binary complex, suggesting that the 2′-amino group of GN interacts with MC in the ground state. Moreover, the C262-S
P E·S·GN complex binds the rescuing Cd2+ ion more than 50-fold tighter than does the corresponding WT ternary complex, showing that the phosphorothioate substitution at C262 enhances Cd2+ binding to metal ion site C. Taken together, these data further support a model in which the C262-S
P phosphorothioate interacts directly with the Cd2+ ion in the MC site and strongly implicate the C262 pro-S
P oxygen as a ligand for MC in the WT ribozyme.
An Independent Test of the C262 pro-S
P Phosphoryl Oxygen as a MC Ligand
The reaction of an oligonucleotide substrate bearing a S
P-phosphorothioate at the cleavage site (SSp, Table 1) also experiences Cd2+ stimulation with the WT ribozyme. The Cd2+ rescue profile for this modified oligonucleotide substrate exhibits a slope of two, suggesting that two Cd2+ ions contact the non-bridging sulfur atom in the transition state [36]. Thermodynamic fingerprint analysis established that these rescuing metal ions bind at sites A and C (Figure 7A) [36]. The SSp oligonucleotide substrate therefore offers another strategy by which to test whether the C262-S
P phosphorothioate enhances Cd2+ binding to the MC site. If the MC site in the C262-S
P ribozyme becomes saturable in the presence of bound GN, as described in the previous section, then the rescue profile for SSp cleavage should transition from an apparent dependence on two Cd2+ ions to a dependence on one Cd2+ ion. The results described below meet this prediction, thereby providing additional quantitative support for the assignment of the pro-S
P phosphoryl atom of residue C262 as a ligand for MC.
Figure 7 Cd2+ Rescue of SSp Cleavage by the C262-S
P Ribozyme
Reactions were performed under E·S·G/GN → products (k
ternary) conditions, as described in Materials and Methods, with −(1–3)d,rSA5 serving as the unmodified control oligonucleotide substrate. Cd2+ dependencies for oligonucleotide substrate cleavage by G and GN are displayed in Figure S5.
(A) Model of the Tetrahymena ribozyme transition state for rescue of SSp reaction by Cd2+ ions bound at MA and Mc. Closed red circles indicate metal ion coordinations considered important for rescue, and the S
P phosphorothioate modification at the scissile phosphate is denoted in green.
(B) k
rel versus Cd2+ for cleavage of SSp with G. The data are fit to a model in which two Cd2+ ions rescue reaction, with neither Cd2+ binding site saturating (see Protocol S1).
(C) Model of the Tetrahymena ribozyme transition state for rescue of SSp cleavage by two Cd2+ ions bound at MA and MC in the presence of saturating GN. Metal ion coordinations important for rescue are denoted by closed red circles, and the GN amino group is shown in blue.
(D) k
rel versus Cd2+ for cleavage of SSp with GN. The dotted line is a fit to the model used in (B), for two Cd2+ ions rescuing reaction and neither Cd2+ binding site saturating. The solid line is a fit to a model in which two Cd2+ ions rescue reaction, with one site saturating with a K
Mc
d,app = 0.2 mM (see Protocol S1).
The Cd2+ rescue profile for SSp cleavage by C262-S
P ribozyme with saturating G fits well to a model with two Cd2+ ions rescuing SSp cleavage with neither metal ion saturating, analogous to that for the WT ribozyme (Figure 7B) [36]. This profile matches expected rescue behavior, given that neither the MA nor MC (subsaturating GN) individual rescue profiles exhibit saturation with the C262-S
P ribozyme (see Figures 6A and 6C, respectively). In contrast, with saturating GN, Cd2+ rescue of SSp cleavage by the C262-S
P ribozyme no longer fits well to a dependence on two nonsaturating Cd2+ ions (see Figure 7C and the dashed line in Figure 7D). Rather, the best fit gives a dependence on only one Cd2+ ion, suggesting that one of the two metal ion sites involved in rescuing the SSp reaction saturates. Assuming that one Cd2+ ion binds to the C262-S
P ribozyme with = 0.2 mM, the value determined for Cd2+ binding to metal site C in the presence of saturating GN (see Figure 6D), we obtain the solid line in Figure 7D. In contrast, for the WT ribozyme, Cd2+ rescue of the SSp reaction in the presence of saturating GN exhibits a dependence on two Cd2+ ions up to the highest accessible Cd2+ concentration [36]. Therefore, as predicted, the C262-S
P phosphorothioate substitution enhances binding of one of the rescuing Cd2+ ions in the reaction of SSp, strongly supporting the conclusion that MC interacts directly with the pro-S
P atom of the C262 phosphate.
Discussion
Metalloenzymes that catalyze phosphoryl transfer play multiple roles throughout biology, serving as kinases, phosphatases, polymerases, and nucleases, among other functions [1–10]. We have established a powerful new experimental paradigm with which to identify ligands that coordinate to catalytic metal ions. Our analysis of simultaneous atomic perturbations within the Tetrahymena ribozyme core and its substrates, under conditions that allow valid thermodynamic comparisons, provides strong evidence that the pro-S
P oxygen at residue C262 serves as a ligand for metal ion C (MC, Figure 8A). A single phosphorothioate substitution at this site, alone and in combination with GN, changes the metal ion affinity and specificity at the MC site while having little or no effect on the MA and MB sites. Although we cannot determine unambiguously whether the interaction between the pro-S
P oxygen at C262 and MC occurs via outer sphere or inner sphere coordination in the natural ribozyme, the presence of a direct, inner sphere metal ion-ligand interaction provides the simplest model to account for our observations. More extensive rearrangement by the ribozyme active site to accommodate the backbone mutation appears less likely, as the metal ion that contacts C262-S
P satisfies all the transition-state contacts proposed for MC. Further demonstrating the efficacy of this approach, detailed analysis of phosphorothioate mutations in the P4 and J5/4 regions of the ribozyme core identified the pro-S
P oxygen at position C208 as a ligand for MA (AVK, JLH, JAP, and DH, unpublished data), consistent with the previous proposal of Szewczak et al. [53].
Figure 8 Functional and Structural Models of Group I Intron Active Sites
(A) Model of the Tetrahymena ribozyme transition state from functional data with the C262 pro-S
P phosphoryl oxygen and C208 pro-S
P oxygen coordinating to MC and MA, respectively ([35,36,53,58,71,72, 74], data herein, and AVK, JLH, JAP, and DH, unpublished results).
(B) Model of Mg2+ binding in the crystal structure of a thermostable variant of the Tetrahymena group I ribozyme (derived from PDB file 1X8W) [14]. The model shown is derived from molecule C, although the position of the metal ion appears to vary among the four molecules observed in the asymmetric unit. The putative Mg2+ ion (dashed green circle) is 2.4 Å from the pro-S
P oxygen of C262 and 2.1 Å from the terminal G (ωG) 2′-OH. U(-1) is not shown, as the crystallized form of the thermostable Tetrahymena variant ribozyme lacks this nucleotide.
(C) Model of K+ binding from the crystal structure of the Azoarcus group I ribozyme (derived from PDB file 1T42) [15]. The putative K+ ion (dashed green circle) is 2.4 Å from the pro-S
P oxygen of G128 (C262 homologue) and 2.8 Å from the modeled position for the terminal G (ωG) 2′-OH; the Azoarcus intron construct that was crystallized contained 2′-deoxyguanosine at ωG, and electron density for the K+ ion is observed only in the presence of this 2′-deoxyguanosine modification at ωG.
(D) Model of proposed Mg2+ binding site derived from the crystal structure of the Twort group I ribozyme [13]. The crystallographic data lack clear density for a metal ion in the region expected for the MC binding site but show the pro-S
P phosphoryl oxygen of A120 (C262 homologue) and the 2′-OH of ωG appropriately juxtaposed to coordinate a single Mg2+ ion (purple circle).
Under some conditions, the C262-S
P ribozyme E·S·G ternary complex must undergo a rate-limiting conformational change en route to the chemical transition state. Cd2+ accelerates this non-chemical step in this C262-S
P ribozyme, suggesting that MC, which interacts with the phosphorothioate during the reaction, mediates this conformational change. Currently, we lack sufficient information to speculate further on the molecular basis of this conformational change. The ability to induce a new or existing conformational change may provide a future opportunity to investigate the relationship between RNA function and dynamics.
Crystal structures of three different group I introns have emerged in the past several months, providing a structural context for these functionally identified catalytic metal ligands [13–15]. The three structures converge beautifully with respect to the global architecture and reveal the highly electrostatic character of the active site, consistent with functional studies. These structures also underscore the difficulty of defining catalytic metal ion interactions, exhibiting limited agreement regarding the active-site conformation and the number, charge, and location of bound metal ions. This variability in metal ion number and location notwithstanding, the C262 pro-S
P oxygen (or its equivalent) lies in proximity to the ωG 2′-hydroxyl in all three structures (Figure 8). The Azoarcus and Tetrahymena crystals contain electron density for a metal ion within coordination distance of this phosphoryl oxygen [14,15], whereas the Twort ribozyme structure lacks such electron density [13]. The data presented herein establish the importance of this proximity for function and suggest that the active-site configuration in the crystals bears at least some relevance to the active-site configuration in the transition state. This agreement between structural and functional data for metal ion C supports the assignment of a conserved catalytic metal ion binding site at the top of the P7 helix in group I introns (see Figure 2A).
Extensive group I intron biochemical investigations, conducted for two decades in the absence of atomic resolution structures, have revealed an intricate network of interactions surrounding the reaction center in the transition state interconnected by hydrogen bonds, metal ion coordinations, and the atomic configuration of the reactants (see Figure 1). The results suggest a constellation of three metal ions at the active site [35,36], making five atomic interactions to the transition state. In addition to these catalytic metal ion coordinations, the 2′-hydroxyl group of U(-1) donates a hydrogen bond to the adjacent 3′-oxygen leaving group in the transition state [71]. The 2′-hydroxyl group of A207 donates a hydrogen bond to the 2′-OH of U(-1) and accepts a hydrogen bond from the exocyclic amine of G22, thereby bridging the cleavage site 2′-OH and the G·U wobble pair [58,72]. The recent group I intron crystal structures, though unable to define the number, location, or catalytic interactions of metal ions unambiguously, provide an opportunity to visualize how these functionally defined networks extend deeper into the ribozyme core.
Based on the recent crystal structure of the Azoarcus group I intron, an alternative model for the transition state was proposed in which only two metal ions interact with the reaction center [15]. In this model, one metal ion coordinates to the 3′-oxygen leaving group of the cleavage site uridine and the pro-S
P oxygen of the scissile phosphate, the same interactions proposed for MA in the functional model (Figure 8A). The other metal ion coordinates directly with the 2′-hydroxyl group of G as does MC in the functional model, but coordinates indirectly to the pro-Sp oxygen of the scissile phosphate, in contrast to the direct interaction implicated by the functional data. A metal ion interaction with the 3′-oxygen of G, mediated by MB in the functional model, is absent from the crystallographic model. No direct evidence for a MB-binding site was obtained from the X-ray structure. The resting structure of the RNA in the crystal could adopt a conformation that differs from the active structure and thereby exclude MB from the crystals. Alternatively, the MB rescue profile could reflect the recruitment of a thiophilic metal ion to the active site during the reaction of the sulfur-containing G analogue. As the 3′-oxygen of the G undergoes a large charge rearrangement in the transition state, we suggest that a direct metal ion interaction with the 3′-oxygen of G is likely, such that MB is present in the normal reaction or metal ion C makes both 2′- and 3′-interactions with G [73]. These different proposals for specific metal ion coordination configurations during the ribozyme reaction highlight the need for further structural and functional tests and refinements of the catalytic models.
To achieve a unified description of catalytic function that integrates the biochemically defined interaction networks and the crystallographically defined three-dimensional structure, we must establish “anchor points”—functionally verifiable linkages between transition-state interactions and the enzyme's core. Our analysis establishes the C262 pro-S
P oxygen as such an anchor point. Anchor points provide critical information about the spatial arrangement of catalytic groups within the global architecture of the enzyme. Together, the transition state model, anchor points, and the recent X-ray structures establish a powerful foundation to build toward an in-depth understanding of how cooperative structure adopted by this RNA and other enzymes engenders enormous catalytic power and exquisite specificity.
Materials and Methods
Materials
WT Tetrahymena ribozyme was prepared as described previously [34]. All oligonucleotide substrates (see Table 1) were prepared and 5′-end-labeled using standard methods [34,35,54,74]. Oligonucleotides with thio substitutions were prepared by published procedures [75]. Oligonucleotides containing phosphorothioate diastereomers were separated by anion exchange HPLC [42,54]. Reverse phase HPLC of purified diastereomers, under conditions in which the R
P diastereomer elutes before the S
P, allowed assignment of each diastereomer's configuration [76].
Ribozyme preparation
Variant ribozymes were constructed semisynthetically using successive splint-mediated ligations [77]. Synthetic oligonucleotides corresponding to nucleotides 255–274 of the ribozyme, each containing a single phosphorothioate modification at the desired mutation site, were purchased from Dharmacon (Lafayette, Colorado, United States). Following phosphorothioate separation, synthetic oligonucleotides were 5′-phosphorylated with T4 polynucleotide kinase and cold ATP. Constructs corresponding to nucleotides 22–254 and 274–409 of the Tetrahymena ribozyme were transcribed using DNA templates produced by PCR truncation of the plasmid-encoded ribozyme sequence, with excess GMP present in the transcription of the 3′-construct to yield a 5′-monophosphate. The transcripts were ligated to the synthetic oligonucleotide via two successive splint-mediated ligations with T4 DNA ligase to yield full-length ribozyme containing a single R
P- or S
P-phosphorothioate mutation at the desired site.
Ligated ribozymes appear to contain approximately 40% inactive enzyme fraction, as indicated by biphasic kinetics under conditions in which oligonucleotide substrate cleavage occurs faster than oligonucleotide substrate dissociation (data not shown). Evidence suggests that the inactive ribozyme fraction in these semisynthetic ribozymes may be due to errors at the ligation junctions resulting from ligation of a subset of transcribed RNA constructs with incorrect termini (K. Travers, V. Diankov, and DH, unpublished data). Under conditions used in this work to characterize variant ribozyme reactivity, the inactive fraction did not contribute to the cleavage activity monitored in our assays of ribozyme activity. The inactive fraction did not affect association and dissociation rate constants measured by pulse chase experiments, as evidenced by monophasic binding behavior (data not shown). The activity of WT ribozyme constructed by ligation varied less than 2-fold from that of transcribed WT ribozyme (Table S2) after accounting for the presence of the inactive ribozyme fraction. The inactive ribozyme fraction does not affect the conclusions in this work.
General kinetic methods
All cleavage reactions were single turnover, with ribozyme in excess of radiolabeled S (S*), and were carried out at 30 °C in 50 mM buffer and 50 mM MgCl2 unless noted otherwise. All reactions without Cd2+ contained 0.1 mM EDTA. The buffers used were NaOAC (pH 4.0–5.2), NaMES (pH 5.6–6.7), and NaMOPS (pH 7.0–7.5). Reaction mixtures containing all components except Cd2+, EDTA, and radiolabeled oligonucleotide substrate were pre-incubated at 50 °C for 30 min to renature the ribozyme. Reactions were followed and analyzed as described previously [33,36,74].
Determination of rate and equilibrium constants
For oligonucleotide substrates that bind to the WT ribozyme in the closed complex, e.g. −1d,rSA5 (see Table 1), we define k
c as the first-order rate constant for the reaction of the ternary complex (E·S·G)C → products. Values of k
c were determined at pH 7.0, with ribozyme saturating with respect to oligonucleotide substrate (20–100 nM E, < 1 nM) and with saturating G (2 mM, as reported in Table 2).
For cleavage of oligonucleotide substrates that are known to bind in the open complex to the WT ribozyme, e.g., −1r,dSA5 and SSp [36,66], we define k
ternary as the first-order rate constant for the reaction E·S·G/GN → products. We also used k
ternary to describe cleavage reactions catalyzed by variant ribozymes for which the oligonucleotide substrate binding mode has not been assigned definitively. Values of k
ternary were determined at pH 7.0, with ribozyme saturating with respect to oligonucleotide substrate (20–50 nM E, ribozyme concentration at least 4-fold above , data not shown) and with saturating G or GN (2 mM, ∼ 500 μM; ∼ 100–150 μM) ([68], Tables 2 and 3, and data not shown).
(k
c/K
m)G and (k
c/K
m
are the second-order rate constants for the reaction E·S + G (or GN) → products. In these experiments, we used the −1r,dSA5 oligonucleotide substrate that binds to the WT and variant ribozymes in the open complex [36,66]. Values of (k
c/K
m)G and (k
c/K
m
were determined at pH 7.0, with ribozyme saturating with respect to oligonucleotide substrate (20–50 nM E, ribozyme concentration at least 4-fold above , data not shown) and with subsaturating G or GN (30 μM G or GN, ∼ 500 μM; ∼ 100–150 μM) ([68], Table 3, and data not shown).
(k
c/K
m)CUCG(3′X)A (where X = O or S) is the second-order rate constant for the reaction E·P + CUCGA → products. Values of (k
c/K
m)CUCG(3′X)A were determined at pH 6.5 with trace amounts of radiolabeled CUCG3′XA and E·P subsaturating with respect to CUCG3′XA (X = O, 10–20 nM E·P, > 100 nM; X = S, 100–200 nM E·P, > 400 nM) (data not shown). To maintain the chemical step as rate-limiting in reactions catalyzed by the WT ribozyme, an oligonucleotide product with a 2′-deoxyribothymidine at the 3′-terminus was used (CCCUCdT); 2′-deoxyribose incorporation at this site slows the chemical step approximately 103-fold [74]. Reactions in the presence of the C262-S
P variant ribozyme used an all-ribose oligonucleotide product (CCCUCU).
Association (k
on) and dissociation (k
off) rate constants were measured by a gel mobility shift assay using pulse-chase methods [33,56]. Specific conditions and experimental details for these experiments are described in Protocol S2.
Experimental errors
All titrations, binding constants, and rate determinations were repeated at least three times. Error bars, when present, indicate standard deviation from at least three determinations. Error bars may be obscured by data symbols. Plots without error bars display representative data. Reported errors in all tables, both in the text and Supporting Information, are the standard deviation of at least three independent measurements.
Data analysis
The Cd2+ concentration dependencies for rescue of reaction of modified substrates were analyzed according to previously described methods ([35,36] and Protocol S1).
Supporting Information
Figure S1 Phosphorothioate Effects on MA, MB, and MC Rescue of Modified Substrates by Cd2+
Stimulation of modified ribozyme substrates was assessed by comparing cleavage activity in 50 mM Mg2+ alone to observed activity in the presence of 1 mM Cd2+ and 50 mM Mg2+. The cleavage rate in the presence of Cd2+ was divided by the observed cleavage rate in Mg2+ alone, resulting in a relative rate (k
rel). These k
rel values were then normalized to the WT ribozyme; i.e.,
for the WT ribozyme equals one.
(A) Normalized stimulation of cleavage of an oligonucleotide substrate containing a 3′-thiophosphoryl linkage by 1 mM Cd2+; reactions performed as described in Figures 6 and S2 and Table 4.
(B) Normalized stimulation of cleavage of a 3′-splice site analogue containing a 3′-thiophosphoryl linkage by 1 mM Cd2+; reactions performed as described in Figures 6 and S3 and Table 4.
(C) Normalized stimulation of oligonucleotide substrate cleavage by subsaturating GN by 1 mM Cd2+; reactions performed as described in Figures 6 and S4 and Table 4.
(2.4 MB PDF).
Click here for additional data file.
Figure S2 Cd2+ Specifically Stimulates Cleavage of an Oligonucleotide Substrate Containing a 3′-Thiophosphoryl Linkage
(A) [Cd2+] dependencies for the reaction E·S·G → products (k
ternary) for −(1–3)d,rSA5 (s○) and Sm3′S (•) with the WT ribozyme (see Materials and Methods).
(B) [Cd2+] dependencies of the rate of reaction E·S·G → products (k
ternary) for −(1–3)d,rSA5 (□) and Sm3′S (▪) with the C262-S
P ribozyme (see Materials and Methods). Error bars not seen are obscured by data symbols.
(2.8 MB PDF).
Click here for additional data file.
Figure S3 Cd2+ Specifically Stimulates Cleavage of a 3′-Splice Site Analogue Containing a 3′-Thiophosphoryl Linkage
(A) [Cd2+] dependencies of the rate of the reaction E·P + CUCG3′XA → products ((k
c/K
m)CUCG(3′X)A) for CUCGA (s○) and CUCG3′SA (•) with the WT ribozyme (see Materials and Methods)
(B) [Cd2+] dependencies of the rate of the reaction E·P + CUCG3′XA → products ((k
c/K
m)CUCG(3′X)A) for CUCGA (□) and CUCG3′SA (▪) with the C262-S
P ribozyme (see Materials and Methods). Error bars not seen are obscured by data symbols.
(3.1 MB PDF).
Click here for additional data file.
Figure S4 Cd2+ Specifically Stimulates Oligonucleotide Substrate Cleavage by GN
(A) [Cd2+] dependencies of the rate of the reaction E·S + G (or GN) → products ((k
c/K
m)G(or GN)) for G(s○) and GN(•) cleavage of −1r,dSA5 with the WT ribozyme (see Materials and Methods).
(B) [Cd2+] dependencies of the rate of the reaction E·S + G (or GN) → products ((k
c/K
m)G(or GN)) for G (□) and GN(▪) cleavage of −1r,dSA5 with the C262-S
P ribozyme (see Materials and Methods).
(C) [Cd2+] dependencies of the rate of the reaction E·S·G/GN → products (k
ternary) for G(s○) and GN(•) cleavage of −1r,dSA5 with the WT ribozyme (see Materials and Methods).
D) [Cd2+] dependencies of the rate of the reaction E·S·G/GN → products (k
ternary) for G(□) and GN(▪) cleavage of −1r,dSA5 with the C262-S
P ribozyme (see Materials and Methods).
(2.9 MB PDF).
Click here for additional data file.
Figure S5 Cd2+ Specifically Stimulates Cleavage of an Oligonucleotide Substrate Containing a S
P-Phosphorothioate Substitution at the Cleavage Site
(A) [Cd2+] dependencies of the rate of the reaction E·S·G → products [k
ternary] for SSp (▪) and −(1–3)d,rSA5 (□) cleavage with G with the C262-S
P ribozyme (see Materials and Methods).
(B) [Cd2+] dependencies of the rate of the reaction E·S·G/GN → products [k
ternary] for SSp cleavage by GN (▪) and −(1–3)d,rSA5 cleavage by G (□) with the C262-S
P ribozyme (see Materials and Methods).
(2.8 MB PDF).
Click here for additional data file.
Protocol S1 Analysis of Cd2+ Rescue Profiles
(126 KB DOC).
Click here for additional data file.
Protocol S2 Measuring Oligonucleotide Substrate Association and Dissociation Rate Constants
(26 KB DOC).
Click here for additional data file.
Table S1 Oligonucleotide Substrate Binding to WT and C262-S
P Ribozymes
(25 KB DOC).
Click here for additional data file.
Table S2 G Binding and Reactivity of WT Ribozymes Constructed by Transcription and Ligation
(19 KB DOC).
Click here for additional data file.
Accession Numbers
The Protein Data Bank (http://www.rcsb.org/pdb/) accession numbers for the group I intron crystal structures discussed in this paper are Azoarcus (PDB ID 1U6B), Tetrahymena (PDB ID 1X8W), and Twort ribozyme (PDB ID 1Y0Q).
We thank members of the Piccirilli and Herschlag labs for helpful discussion and comments on the manuscript, J. Olvera for preparation of T4 DNA ligase, and K. Hougland for assistance with figures. We also thank B. Golden for discussion and sharing unpublished data. JLH was supported in part by the Predoctoral Training Program at the Interface of Chemistry and Biology (2 T32 GM008720–06) at the University of Chicago. This work was supported by NIH Grant GM49243 to DH and a grant from the Howard Hughes Medical Institute to JAP. JAP is an investigator at the Howard Hughes Medical Institute.
Conflicts of interest. The authors have declared that no conflicts of interest exist.
Author contributions. JLH, AVK, DH, and JAP conceived and designed the experiments. JLH performed the experiments and analyzed the data. JLH and AVK contributed reagents/materials/analysis tools. JLH, DH, and JAP wrote the paper.
Citation: Hougland JL, Kravchuk AV, Herschlag D, Piccirilli JA (2005) Functional identification of catalytic metal ion binding sites within RNA. PLoS Biol 3(9): e277.
Abbreviations
Eribozyme
Gguanosine
GN2′-aminoguanosine
Soligonucleotide substrate
WTwild-type
==== Refs
References
Dismukes GC Manganese enzymes with binuclear active sites Chem Rev 1996 96 2909 2926 11848845
Strater N Lipscomb WN Klabunde T Krebs B Two-metal ion catalysis in enzymatic acyl- and phosphoryl-transfer reactions Ang Chem Int Ed 1996 35 2024 2055
Cowan JA Metal activation of enzymes in nucleic acid biochemistry Chem Rev 1998 98 1067 1087 11848925
Wilcox DE Binuclear metallohydrolases Chem Rev 1996 96 2435 2458 11848832
Heikinheimo P Lehtonen J Baykov A Lahti R Cooperman BS The structural basis for pyrophosphatase catalysis Structure 1996 4 1491 1508 8994974
Steitz TA Steitz JA A general two-metal-ion mechanism for catalytic RNA Proc Natl Acad Sci U S A 1993 90 6498 6502 8341661
Galburt EA Stoddard BL Catalytic mechanisms of restriction and homing endonucleases Biochemistry 2002 41 13851 13860 12437341
Pingoud A Jeltsch A Structure and function of type II restriction endonucleases Nucleic Acids Res 2001 29 3705 3727 11557805
Stec B Holtz KM Kantrowitz ER A revised mechanism for the alkaline phosphatase reaction involving three metal ions J Mol Biol 2000 299 1303 1311 10873454
Kovall RA Matthews BW Type II restriction endonucleases: Structural, functional and evolutionary relationships Curr Opin Chem Biol 1999 3 578 583 10508668
Shi HJ Moore PB The crystal structure of yeast phenylalanine tRNA at 1.93 angstrom resolution: A classic structure revisited RNA 2000 6 1091 1105 10943889
Scott WG Murray JB Arnold JRP Stoddard BL Klug A Capturing the structure of a catalytic RNA intermediate: The hammerhead ribozyme Science 1996 274 2065 2069 8953035
Golden BL Kim H Chase E Crystal structure of a phage Twort group I ribozyme-product complex Nat Struct Mol Biol 2005 12 82 89 15580277
Guo F Gooding AR Cech TR Structure of the Tetrahymena ribozyme: Base triple sandwich and metal ion at the active site Mol Cell 2004 16 351 362 15525509
Adams PL Stahley MR Kosek AB Wang JM Strobel SA Crystal structure of a self-splicing group I intron with both exons Nature 2004 430 45 50 15175762
Jaffe EK Cohn M Divalent cation-dependent stereospecificity of adenosine 5'-O-(2-thiotriphosphate) in hexokinase and pyruvate-kinase reactions-Absolute stereochemistry of diastereoisomers of adenosine 5'-O-(2-thiotriphosphate) J Biol Chem 1978 253 4823 4825 670166
Christian EL Yarus M Metal coordination sites that contribute to structure and catalysis in the group I intron from Tetrahymena
Biochemistry 1993 32 4475 4480 7683490
Yoshida A Sun SG Piccirilli JA A new metal ion interaction in the Tetrahymena ribozyme reaction revealed by double sulfur substitution Nat Struct Biol 1999 6 318 321 10201397
Sjogren AS Pettersson E Sjoberg BM Stromberg R Metal ion interaction with cosubstrate in self-splicing of group I introns Nucleic Acids Res 1997 25 648 653 9016608
Weinstein LB Jones B Cosstick R Cech TR A second catalytic metal ion in a group I ribozyme Nature 1997 388 805 808 9285596
Piccirilli JA Vyle JS Caruthers MH Cech TR Metal-ion catalysis in the Tetrahymena ribozyme reaction Nature 1993 361 85 88 8421499
Gordon PM Sontheimer EJ Piccirilli JA Kinetic characterization of the second step of group II intron splicing: Role of metal ions and the cleavage site 2'-OH in catalysis Biochemistry 2000 39 12939 12952 11041859
Sontheimer EJ Gordon PM Piccirilli JA Metal ion catalysis during group II intron self-splicing: Parallels with the spliceosome Gen Dev 1999 13 1729 1741
Warnecke JM Furste JP Hardt WD Erdmann VA Hartmann RK Ribonuclease P (RNase P) RNA is converted to a Cd(2+)ribozyme by a single Rp-phosphorothioate modification in the precursor tRNA at the RNase P cleavage site Proc Natl Acad Sci U S A 1996 93 8924 8928 8799129
Scott EC Uhlenbeck OC A re-investigation of the thio effect at the hammerhead cleavage site Nucleic Acids Res 1999 27 479 484 9862968
Peracchi A Beigelman L Scott EC Uhlenbeck OC Herschlag D Involvement of a specific metal ion in the transition of the hammerhead ribozyme to its catalytic conformation J Biol Chem 1997 272 26822 26826 9341112
Sontheimer EJ Sun SG Piccirilli JA Metal ion catalysis during splicing of premessenger RNA Nature 1997 388 801 805 9285595
Gordon PM Sontheimer EJ Piccirilli JA Metal ion catalysis during the exon-ligation step of nuclear pre-mRNA splicing: Extending the parallels between the spliceosome and group II introns RNA 2000 6 199 205 10688359
Cohn M Some properties of the phosphorothioate analogs of adenosinetriphosphate as substrates of enzymic reactions Acc Chem Res 1982 15 326 332
Eckstein F Nucleoside phosphorothioates Ann Rev Biochem 1985 54 367 402 2411211
Curley JF Joyce CM Piccirilli JA Functional evidence that the 3'-5′ exonuclease domain of Escherichia coli DNA polymerase I employs a divalent metal ion in leaving group stabilization J Am Chem Soc 1997 119 12691 12692
Aubert SD Li YC Raushel FM Mechanism for the hydrolysis of organophosphates by the bacterial phosphotriesterase Biochemistry 2004 43 5707 5715 15134445
Herschlag D Cech TR Catalysis of RNA cleavage by the Tetrahymena thermophila ribozyme. 1. Kinetic description of the reaction of an RNA substrate complementary to the active site Biochemistry 1990 29 10159 10171 2271645
Zaug AJ Grosshans CA Cech TR Sequence-specific endoribonuclease activity of the Tetrahymena ribozyme—Enhanced cleavage of certain oligonucleotide substrates that form mismatched ribozyme substrate complexes Biochemistry 1988 27 8924 8931 3069131
Shan S Yoshida A Sun SG Piccirilli JA Herschlag D Three metal ions at the active site of the Tetrahymena group I ribozyme Proc Natl Acad Sci U S A 1999 96 12299 12304 10535916
Shan S Kravchuk AV Piccirilli JA Herschlag D Defining the catalytic metal ion interactions in the Tetrahymena ribozyme reaction Biochemistry 2001 40 5161 5171 11318638
Basu S Strobel SA Thiophilic metal ion rescue of phosphorothioate interference within the Tetrahymena ribozyme P4-P6 domain RNA 1999 5 1399 1407 10580468
Boudvillain M Pyle AM Defining functional groups, core structural features and inter-domain tertiary contacts essential for group II intron selfsplicing: A NAIM analysis EMBO J 1998 17 7091 7104 9843513
Cate JH Hanna RL Doudna JA A magnesium ion core at the heart of a ribozyme domain Nat Struct Biol 1997 4 553 558 9228948
Christian EL Yarus M Analysis of the role of phosphate oxygens in the group-I intron from Tetrahymena
J Mol Biol 1992 228 743 758 1469712
Crary SM Kurz JC Fierke CA Specific phosphorothioate substitutions probe the active site of Bacillus subtilis ribonuclease P RNA 2002 8 933 947 12166648
Gordon PM Piccirilli JA Metal ion coordination by the AGC triad in domain 5 contributes to group II intron catalysis Nature Struct Biol 2001 8 893 898 11573097
Harris ME Pace NR Identification of phosphates involved in catalysis by the ribozyme RNase-P RNA RNA 1995 1 210 218 7585250
Christian EL Kaye NM Harris ME Helix P4 is a divalent metal ion binding site in the conserved core of the ribonuclease P ribozyme RNA 2000 6 511 519 10786842
Christian EL Kaye NM Harris ME Evidence for a polynuclear metal ion binding site in the catalytic domain of ribonuclease P RNA EMBO 2002 21 2253 2262
Jones FD Strobel SA Ionization of a critical adenosine residue in the Neurospora Varkud satellite ribozyme active site Biochemistry 2003 42 4265 4276 12680781
Ortoleva-Donnelly L Szewczak AA Gutell RR Strobel SA The chemical basis of adenosine conservation throughout the Tetrahymena ribozyme RNA 1998 4 498 519 9582093
Oyelere AK Kardon JR Strobel SA pKa perturbation in genomic hepatitis delta virus ribozyme catalysis evidenced by nucleotide analogue interference mapping Biochemistry 2002 41 3667 3675 11888283
Ryder SP Oyelere AK Padilla JL Klostermeier D Millar DP Investigation of adenosine base ionization in the hairpin ribozyme by nucleotide analog interference mapping RNA 2001 7 1454 1463 11680850
Ryder SP Strobel SA Nucleotide analog interference mapping of the hairpin ribozyme: Implications for secondary and tertiary structure formation J Mol Biol 1999 291 295 311 10438622
Yean SL Wuenschell G Termini J Lin RJ Metal-ion coordination by U6 small nuclear RNA contributes to catalysis in the spliceosome Nature 2000 408 881 884 11130730
Strauss-Soukup JK Strobel SA A chemical phylogeny of group I introns based upon interference mapping of a bacterial ribozyme J Mol Biol 2000 302 339 358 10970738
Szewczak AA Kosek AB Piccirilli JA Strobel SA Identification of an active site ligand for a group I ribozyme catalytic metal ion Biochemistry 2002 41 2516 2525 11851398
Wang SL Karbstein K Peracchi A Beigelman L Herschlag D Identification of the hammerhead ribozyme metal ion binding site responsible for rescue of the deleterious effect of a cleavage site phosphorothioate Biochemistry 1999 38 14363 14378 10572011
Michel F Westhof E Modeling of the 3-dimensional architecture of group-I catalytic introns based on comparative sequence-analysis J Mol Biol 1990 216 585 610 2258934
Karbstein K Carroll KS Herschlag D Probing the Tetrahymena group I ribozyme reaction in both directions Biochemistry 2002 41 11171 11183 12220182
Narlikar GJ Khosla M Usman N Herschlag D Quantitating tertiary binding energies of 2'OH groups on the P1 duplex of the Tetrahymena ribozyme: Intrinsic binding energy in an RNA enzyme Biochemistry 1997 36 2465 2477 9054551
Knitt DS Narlikar GJ Herschlag D Dissection of the role of the conserved G*U pair in group I RNA self-splicing Biochemistry 1994 33 13864 13879 7947795
Bevilacqua PC Kierzek R Johnson KA Turner DH Dynamics of ribozyme binding of substrate revealed by fluorescence-detected stopped-flow methods Science 1992 258 1355 1357 1455230
Venkataraman D Du YH Wilson SR Hirsch KA Zhang P A coordination geometry table of the d-block elements and their ions J Chem Educ 1997 74 915 918
Shriver DF Atkins P Langford CH Inorganic chemistry 1994 New York W.H. Freeman and Co 913
Feig AL Uhlenbeck OC Gesteland RF Cech TR Atkins JF The role of metal ions in RNA biochemistry The RNA world. 2nd ed 1998 Cold Spring Harbor (New York) Cold Spring Harbor Press 287 320
Herschlag D Khosla M Comparison of pH dependencies of the Tetrahymena ribozyme reactions with RNA 2'-substituted and phosphorothioate substrates reveals a rate-limiting conformational step Biochemistry 1994 33 5291 5297 8172903
Herschlag D Piccirilli JA Cech TR Ribozyme-catalyzed and nonenzymatic reactions of phosphate diesters—Rate effects upon substitution of sulfur for a nonbridging phosphoryl oxygen atom Biochemistry 1991 30 4844 4854 2036355
Knitt DS Herschlag D pH dependencies of the Tetrahymena ribozyme reveal an unconventional origin of an apparent pK(a) Biochemistry 1996 35 1560 1570 8634287
Shan SO Herschlag D An unconventional origin of metal-ion rescue and inhibition in the Tetrahymena group I ribozyme reaction RNA 2000 6 795 813 10864040
McConnell TS Cech TR Herschlag D Guanosine binding to the Tetrahymena ribozyme—Thermodynamic coupling with oligonucleotide binding Proc Natl Acad Sci U S A 1993 90 8362 8366 8378306
Shan SO Herschlag D Probing the role of metal ions in RNA catalysis: Kinetic and thermodynamic characterization of a metal ion interaction with the 2'-moiety of the guanosine nucleophile in the Tetrahymena group I ribozyme Biochemistry 1999 38 10958 10975 10460151
Profenno LA Kierzek R Testa SM Turner DH Guanosine binds to the Tetrahymena ribozyme in more than one step, and its 2'-OH and the nonbridging pro-Sp phosphoryl oxygen at the cleavage site are required for productive docking Biochemistry 1997 36 12477 12485 9376352
Martell AE Smith RM Critical stability constants 1976 New York Plenum Press
Yoshida A Shan S Herschlag D Piccirilli JA The role of the cleavage site 2'-hydroxyl in the Tetrahymena group I ribozyme reaction Chem Biol 2000 7 85 96 10662698
Strobel SA Ortoleva-Donnelly L A hydrogen-bonding triad stabilizes the chemical transition state of a group I ribozyme Chem Biol 1999 6 153 165 10074469
Hougland JL Piccirilli JA Forconi M Lee J Herschlag D Gesteland RF Atkins JF Cech TR How the group I intron works: A case study of RNA structure and function The RNA world, 3rd edition 2005 Cold Spring Harbor (New York) Cold Spring Harbor Press
Herschlag D Eckstein F Cech TR The importance of being ribose at the cleavage site in the Tetrahymena ribozyme reaction Biochemistry 1993 32 8312 8321 7688573
Sun SG Yoshida A Piccirilli JA Synthesis of 3'-thioribonucleosides and their incorporation into oligoribonucleotides via phosphoramidite chemistry RNA 1997 3 1352 1363 9409625
Slim G Gait MJ Configurationally defined phosphorothioate-containing oligoribonucleotides in the study of the mechanism of cleavage of hammerhead ribozymes Nucleic Acids Res 1991 19 1183 1188 1709484
Moore MJ Sharp PA Site-specific modification of pre-messenger-RNA—The 2'-hydroxyl groups at the splice sites Science 1992 256 992 997 1589782
Lindqvist M Sandstrom K Liepins V Stromberg R Graslund A Specific metal-ion binding P4-P6 triple-helical domain sites in a model of a group I intron RNA 2001 7 1115 1125 11497430
Narlikar GJ Bartley LE Khosla M Herschlag D Characterization of a local folding event of the Tetrahymena group I ribozyme: Effects of oligonucleotide substrate length pH, and temperature on the two substrate binding steps Biochemistry 1999 38 14192 14204 10571993
Karbstein K Herschlag D Extraordinarily slow binding of guanosine to the Tetrahymena group I ribozyme: Implications for RNA preorganization and function Proc Natl Acad Sci U S A 2003 100 2300 2305 12591943
|
16092891
|
PMC1184590
|
CC BY
|
2021-01-05 08:21:26
|
no
|
PLoS Biol. 2005 Sep 16; 3(9):e277
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030277
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1608663310.1371/journal.pbio.0030283Research ArticleCell BiologyDevelopmentNeuroscienceIn VitroMus (mouse)Niche-Independent Symmetrical Self-Renewal of a Mammalian Tissue Stem Cell Homogenous Neural Stem CellsConti Luciano
1
2
Pollard Steven M
1
Gorba Thorsten
1
¤Reitano Erika
2
Toselli Mauro
3
Biella Gerardo
3
Sun Yirui
1
Sanzone Sveva
2
Ying Qi-Long
1
Cattaneo Elena
2
Smith Austin [email protected]
1
1Institute for Stem Cell Research, University of Edinburgh, Edinburgh, United Kingdom,2Department of Pharmacological Sciences and Center of Excellence on Neurodegenerative Diseases, University of Milan, Milan, Italy,3Institute of Physiological and Pharmacological Sciences, University of Pavia, Pavia, ItalyLovell-Badge Robin Academic EditorNational Institute for Medical ResearchUnited Kingdom9 2005 16 8 2005 16 8 2005 3 9 e28329 11 2004 14 6 2005 Copyright: © 2005 Conti 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.
Bake a Batch of Stem Cells
Pluripotent mouse embryonic stem (ES) cells multiply in simple monoculture by symmetrical divisions. In vivo, however, stem cells are generally thought to depend on specialised cellular microenvironments and to undergo predominantly asymmetric divisions. Ex vivo expansion of pure populations of tissue stem cells has proven elusive. Neural progenitor cells are propagated in combination with differentiating progeny in floating clusters called neurospheres. The proportion of stem cells in neurospheres is low, however, and they cannot be directly observed or interrogated. Here we demonstrate that the complex neurosphere environment is dispensable for stem cell maintenance, and that the combination of fibroblast growth factor 2 (FGF-2) and epidermal growth factor (EGF) is sufficient for derivation and continuous expansion by symmetrical division of pure cultures of neural stem (NS) cells. NS cells were derived first from mouse ES cells. Neural lineage induction was followed by growth factor addition in basal culture media. In the presence of only EGF and FGF-2, resulting NS cells proliferate continuously, are diploid, and clonogenic. After prolonged expansion, they remain able to differentiate efficiently into neurons and astrocytes in vitro and upon transplantation into the adult brain. Colonies generated from single NS cells all produce neurons upon growth factor withdrawal. NS cells uniformly express morphological, cell biological, and molecular features of radial glia, developmental precursors of neurons and glia. Consistent with this profile, adherent NS cell lines can readily be established from foetal mouse brain. Similar NS cells can be generated from human ES cells and human foetal brain. The extrinsic factors EGF plus FGF-2 are sufficient to sustain pure symmetrical self-renewing divisions of NS cells. The resultant cultures constitute the first known example of tissue-specific stem cells that can be propagated without accompanying differentiation. These homogenous cultures will enable delineation of molecular mechanisms that define a tissue-specific stem cell and allow direct comparison with pluripotent ES cells.
Austin Smith and colleagues derive neural stem cells from mouse embryonic stem cells and demonstrate their long-term propagation in vitro.
==== Body
Introduction
Stem cells are capable of generating identical progeny through unlimited numbers of cell divisions whilst retaining the ability to respond to physiological demands by producing daughters committed to differentiate. In vivo, stem cells are thought to reside in specific cellular microenvironments, or niches, that constitute privileged settings for support of self-renewal [1–4]. In tissues that utilise stem cells to sustain cell turnover, the stem cell compartment must be renewed in balance with the production of transit-amplifying progenitors [5]. This requires either equivalence between symmetrical self-renewal and commitment divisions, or an asymmetric mode of stem cell division. Expansion of stem cells, in vivo or in vitro, unambiguously requires symmetrical self-renewal. However, with the notable exception of embryonic stem (ES) cells, it has proven extremely problematic to propagate homogenous cultures of stem cells ex vivo. Epidermal stem cells [6] and neural stem cells [7] can be expanded in vitro, although accompanied by differentiation. It is unclear whether this reflects a dependence of tissue stem cells on a cellular niche, an intrinsic bias of tissue stem cells towards asymmetric division, or a failure to develop appropriate culture conditions to suppress commitment and sustain symmetrical self-renewal, as has been achieved for ES cells [8].
Neural stem cells appear to be sustained in a complex niche in the mammalian brain [9–11]. In 1992, Weiss and Reynolds made the landmark discovery that neural stem cells could be maintained in culture via propagation of floating cell clusters termed “neurospheres” [7]. Neurospheres consist predominantly of committed progenitors mixed with differentiated astrocytes and neurons. This mixed cellular environment likely provides a niche that sustains relatively few stem cells [12]. The neurosphere assay has proven invaluable in demonstrating the potential to give rise to stem cells in the developing and adult central nervous system (CNS) of rodents and primates [13–15]. However, neurospheres have significant limitations. The stem cells maintained within neurospheres are not directly identifiable, have not been purified, and have an uncertain relationship to CNS precursor cells in vivo [16]. Cellular complexity is a barrier to molecular and biochemical dissection of self-renewal and commitment mechanisms [17]. Heterogeneity also undermines comparative analytical approaches such as global expression profiling [16]. Furthermore, there is variation between as well as within cultures, which can give rise to contradictory data from different laboratories [18]. Finally, neurospheres differentiate much more readily into astrocytes than neurons in vitro [18] and in vivo [19], providing little enthusiasm for pharmacological screening or therapeutic applications [20]. Neural progenitor cells are also propagated in adherent cultures supported by fibroblast growth factor 2 (FGF-2) [21,22], but without genetic transformation [23,24], neuronal differentiation potential is usually progressively lost in these conditions [25,26]. As in Drosophila, mammalian neural progenitor cells may undergo asymmetric divisions in vivo [27,28] and in vitro [29]. However, the incidence of asymmetric versus symmetric division in true stem cells, either in vivo or in neurospheres, is unknown. Here we have investigated the potential for symmetrical self-renewal of neural stem cells and maintenance of neuronal differentiation capacity in fully defined adherent cultures.
Results
Derivation of Self-Renewing Adherent Neural Stem Cells from ES Cells
Mouse ES cells differentiate efficiently into neural precursor cells upon withdrawal of serum in adherent monolayer culture [8] or via treatment of embryoid bodies with retinoic acid [30,31]. These precursors have previously been expanded in FGF-2 with transient retention of neuronal differentiation potential [31,32], but continuous propagation is accompanied by restriction to glial fates ([33] and unpublished data). A similar switch from neurogenic to gliogenic differentiation is consistently observed in adherent cultures of primary foetal progenitors and is suggested to recapitulate the developmental progression during formation of the nervous system [25,34].
We induced neural precursor differentiation from ES cells in serum-free adherent monoculture [35,36]. After 7 d, cells were re-plated in basal medium (NS-A plus N2) in the presence of either FGF-2 alone or FGF-2 plus epidermal growth factor (EGF). Importantly, NS-A media does not support propagation of residual undifferentiated ES cells and they are thereby eliminated from the cultures. Neural precursors initially associate into floating clusters in this media. After 3–5 d, these aggregates were harvested, separating them from any adherent differentiated cells, and re-plated in fresh medium. They attached within 2–3 d and outgrew a population of bipolar cells. Upon passaging, these cells did not persist in FGF-2 alone, but in FGF-2 plus EGF they proliferated in the absence of other cell types. These bipolar cells, named LC-1, can be continuously and rapidly propagated with a doubling time of approximately 24 h.
LC1 cells express the immature neural marker nestin and are immunoreactive with the RC2 antibody, which recognises neural precursors, but expression of the astrocyte differentiation marker glial fibrillary acidic protein (GFAP) or of neuronal antigens is negligible (Parts b, c, and d in Figure 1A). Upon exposure to serum or BMP 4, LC1 cells adopt astrocyte morphology within 48 h and subsequently uniformly express GFAP (Part e in Figure 1A). In contrast, cells with fine extended processes appear after re-plating on laminin without EGF for 5–7 d and then withdrawing FGF-2. These cells express neuronal markers type III β-tubulin (Part f in Figure 1A), microtubule associated protein-2 (MAP2) (Figure 1B), and neuN (not shown). High numbers of such neuronal cells, between 30% and 40% of total surviving cells, are generated with no significant decline even after 115 passages (Part g in Figure 1B). Together with the observation that LC1 cells retain diploid chromosome content at late passages (unpublished data), these data suggest the presence of self-renewing neural stem (NS) cell cultures.
Figure 1 Generation of NS Cells from ES Cells
(A) The adherent NS cell culture (LC1) propagated in EGF and FGF-2 (a), shows no expression of neuronal (b) or astrocyte (c) antigens, and uniform expression of the precursor marker RC2 (d) and nestin (not shown). LC1 cells differentiate into GFAP immunopositive astrocytes (e) upon addition of serum and generate TuJ1 immunopositive neurons (f) upon growth factor withdrawal.
(B) The proportion of neurons obtained remains greater than 35% of total cells after 115 passages (g). Immunostaining for MAP2 of LC1 differentiation shown at passage 16 (h) and passage 158 (i).
NS cells competent for both glial and neuronal differentiation have subsequently been obtained using the procedure described for LC1 from more than ten ES cell lines, originating from three independent derivations: E14, CGR8, and R1. For all NS lines examined more than 95% of cells express nestin and are immunoreactive with RC2 in the presence of FGF-2 plus EGF. These NS cells are transfectable by electroporation (unpublished data) and can reliably be recovered from standard cryopreservation.
To assess whether the serum-free adherent neural induction protocol is a prerequisite for NS cell generation, ES cells were induced to differentiate by embryoid body formation and exposure to retinoic acid in serum-containing medium [30]. Aggregates were subjected to Sox2-βgeo lineage selection with G418 [31,37] for 48 h to enrich for neural precursors, then dissociated and cultured in the presence of FGF-2 and EGF without serum. Floating clusters formed that subsequently attached and outgrew Sox2-positive, nestin-positive, proliferative cells, which displayed the bipolar morphology and lattice growth typical of NS cells, as well as the capacity for astrocyte and neuronal differentiation after multiple passages (unpublished data).
Clonogenic NS Cells Derive from Sox1-Positive Pan-Neural Precursors
To investigate the origin of NS cells and to determine whether the floating cluster phase is essential for their generation, we induced neural commitment in monolayer and then maintained the neural precursors in N2B27 medium, in which condition they remain adherent [38]. To eliminate undifferentiated ES cells and non-neural differentiation products under these conditions, we again exploited lineage selection [31]. In this case we used 46C ES cells in which the green fluorescent protein (GFP)irespac reporter/selection cassette is integrated into the Sox1 gene, an early marker of neural specification [39]. Transient puromycin selection after differentiation induction yields a purified population of neural precursors with minimal residual ES cells [40] (Parts a and b in Figure 2A). Either FGF-2 alone or FGF-2 plus EGF were then applied to the Sox1-expressing neural precursors maintained in N2B27 medium. Appreciable numbers of bipolar cells of NS morphology appeared only in the presence of both factors (Part c in Figure 2A). Initial heterogeneity of the population reduced with two to three passages. as astrocytes and other cell types decreased in number. Notably, expression of Sox1-GFP (and endogenous Sox1) is lost during this process (Parts c and d in Figure 2A), but the cells remain positive for Sox2 and nestin. As they began to dominate the cultures, the bipolar cells formed extensive lattices. To establish the presence of clonogenic NS cells, single cells were isolated in microwells and expanded as adherent cultures (Parts e and f in Figure 2A). Five clonal lines were derived with morphology and growth characteristics similar to the bulk population. Initially one clone, NS-5, was characterised in detail, but subsequently all essential features described below were confirmed for other clones. These cells lack detectable expression of the pluripotency factors Oct4 and Nanog, and also of the early neural marker Sox1, but retain the pan-neuroepithelial marker Sox2 (Part r in Figure 2C). They are competent for astrocyte and neuronal differentiation (Figure 2B). Like LC1 cells, NS-5 cells uniformly express nestin, RC2, and other neural precursor markers (Figure 2C), and lack detectable GFAP expression (Part p in Figure 2C). We conclude that NS cells can be generated through a transient Sox1-positive neuroectodermal precursor via continuous adherent culture.
Figure 2 Clonal NS Cells Generated through Sox1 Neural Lineage Selection
(A) Phase image of neural precursors at passage 1 (a) and 5 (c), with (b) and (d) showing corresponding Sox1-GFP fluorescence. Image (e) shows a single cell, 1 h after plating in Terasaki well, and (f) shows a phase-contrast image of clonal cell line at passage 20.
(B) Differentiation of NS-5 cells into astrocytes (g,h) and neurons (j,k) with loss of nestin immunoreactivity (i,l).
(C) These NS-5 cells are immunoreactive for neural precursor cell/radial glia markers (m–o,q,r) and negative for GFAP (p).
(D) Clones of NS-5 cells exhibit homogenous expression of BLBP with no immunoreactivity for GFAP in the presence of EGF/FGF (s), and generate neurons upon growth factor withdrawal (t).
(E) Metaphase spread of NS-5 (passage 31).
To establish that the NS-5 clone can generate sub-clones of the same phenotype, cells were plated at clonal density in EGF plus FGF-2. Multiple colonies were generated, all consisting of bipolar cells. Every colony shows immunoreactivity with RC2 (not shown) and expression of brain lipid binding protein (BLBP) (Figure 2D) in virtually all cells, with no detectable GFAP (Part s in Figure 2D). Several colonies were picked and expanded over ten or more passages with retention of these characteristics. Every colony can also be induced to differentiate into neurons (Part t in Figure 2D). Finally, we prepared metaphase spreads from NS-5 and determined a modal chromosome count of 40 (Figure 2E). The NS-5 clone therefore represents a clonogenic, genetically stable, NS cell line that self-renews continuously without requirement for a specialised cellular niche.
Contributions of EGF and FGF to NS Cell Self-Renewal and Lineage Commitment
As outlined above, NS cells were derived by culture only in the combination of FGF-2 plus EGF. This contrasts with previous studies, including our own, which used FGF-2 alone, resulting, after several passages, in populations of glial restricted progenitors [31–33]. We examined whether NS cells remained continuously dependent on EGF. When EGF is withdrawn from the cultures, massive cell death ensues after 20 h (Figures 3A and 3B), and the few cells that survive adopt differentiated morphology. This cell death is associated with presence of activated caspase 3, indicative of apoptosis (Figures 3C and 3D). Serum or BMP override the cell death response and drive the NS cells into astrocytic differentiation. We conclude that EGF firstly allows the maintenance of stem cells with competence for neurogenesis in contrast to FGF-2 alone, and secondly supports self-renewal of NS cells, acting in part via suppression of apoptosis.
Figure 3 NS Cells Die or Begin to Differentiate in the Absence of EGF
Unlike proliferating cultures in FGF plus EGF (A,C), NS cells on gelatin die by caspase-3-mediated programmed cell death 20 h after removal of EGF (B,D). This death can be overcome if cells are cultured on a laminin substrate in FGF-2 only (F). Under these conditions, cells become slow-dividing and extend longer processes (G,H). Most cells retain RC2 immunoreactivity (H), but a minority begin neuronal differentiation marked by TuJ1 expression (J).
We found that laminin could preserve cell viability in the absence of EGF. Simultaneous removal of EGF and FGF-2 on laminin results in astrocyte differentiation (unpublished data). However, NS cells cultured on laminin with FGF-2 alone do not express GFAP. Instead, they develop more extended processes (Figures 3E and 3F), and slow their rate of cell division. To date, we have been unable to maintain proliferation upon dissociation and passaging using FGF-2, and the cells die out when this is attempted. The majority of cells remain immunoreactive with RC2 in the presence of FGF-2, but TuJ1-immunostained immature neurons, which are never apparent in FGF-2 plus EGF, can be detected at low frequency (Figures 3G–3J). Upon subsequent withdrawal of FGF-2, the neuroblast marker doublecortin is expressed by a sub-population of cells (unpublished data). Many cells of neuronal morphology that express a diagnostic neuronal immunophenotype then appear (see Results), especially in presence of the neuronal viability supplement B27 [41]. These data suggest that upon release from EGF stimulation, the combination of laminin plus FGF-2 primes NS cells for neuroblast commitment with differentiation ensuing on mitogen withdrawal.
Neuronal Differentiation of NS Cells
To assess the frequency of cells within NS cultures that are capable of neuronal differentiation, we plated NS-5 cells at clonal density on laminin, expanded for 12 d in EGF/FGF-2 followed by FGF-2 alone for 5 d, then a further 7 d in B27-supplemented media without growth factor. Every colony (126/126) produced TuJ1-positive cells (Part t in Figure 2D). These data indicate that all colony forming cells in NS cultures are competent for neuronal differentiation. Most neurons are immunopositive for GAD67 (glutamic acid decarboxylase) (Figure 4A) and gamma-aminobutyric acid (unpublished data) and by 7 d a sub-population shows expression of the mature marker synaptophysin (Figure 4B).
Figure 4 Phenotype and Electrical Activity of NS Cell–Derived Neurons
(A) This LC1 NS cell–derived neuron at 27 d of differentiation displays mature morphology and expresses GAD67.
(B) Expression of MAP2/synaptophysin after 7 d of differentiation.
(C) Superimposed inward and outward current tracings obtained at different membrane potentials (between −70 and +40 mV from a holding potential of −90 mV), from NS cell–derived neurons after differentiation for 6 (i), 20 (ii), and 30 d (iii).
(D) Superimposed voltage responses obtained following injection of depolarising rectangular current pulses in the same three cells (i–iii) by switching from voltage- to current-clamp immediately after current recordings shown in (c) were obtained. The dashed line represents a voltage level of −60 mV.
(E)Average Na+ currents elicited at −20 mV from cells cultured in differentiating medium for increasing times as indicated by labels. Bars indicate SE.
(F) Superimposed inward currents elicited at −40 mV and 0 mV in 10 mM Ba2+ and in the presence of TTX; the holding potential was −90 mV.
(G) Current/voltage relationship from the same cell as in (F).
For unambiguous assignment of neuronal identity, we investigated the electrophysiological properties of differentiated NS cells. Figure 4C shows current recordings obtained during whole-cell voltage-clamp steps to depolarising test potentials. A sizeable outward voltage-gated current, with features of a delayed-rectifier K+ current, is present by 6 d of differentiation (trace i). At later stages of differentiation (20 and 30 d, traces ii and iii), the current amplitude increases only slightly. By contrast, the amplitude of the inward current increases dramatically. Figure 4D shows the voltage responses elicited in the same cells after switching from voltage-clamp to current-clamp mode. The excitability properties of the cells correlate with the magnitude of the inward voltage-gated conductance. Thus, an overshooting action potential with a relatively fast depolarisation rate was elicited in the cell differentiated for 30 d (trace iii). The fast inactivating inward current was completely blocked by the selective Na+ channel blocker tetrodotoxin (1 μM) and peaked at a test potential between −20 and −10 mV (unpublished data), typical features of voltage-gated Na+ currents in neurons. The Na+ current amplitude at −20 mV develops during differentiation (Figure 4E). The regenerative potential (ΔV measured between the threshold and the peak) elicited by the Na+ current under current clamp conditions ranged between 0 and +20 mV during the first 15 d (n = 6), but after 25 d reached values between +30 and +70 mV (n = 6). Voltage-gated Ca2+ channel conductances were also detected (Figure 4F). The fast activating and relatively fast inactivating (τh = 21 ms) current component elicited at −40 mV is reminiscent of the neuronal low-voltage activated (LVA) Ca2+ channel current [42]. By contrast, the Ba2+ current elicited at 0 mV, displaying a slow (τh = 73 ms) and incomplete inactivation, has the typical features of the neuronal high-voltage activated (HVA) Ca2+ channel current. The presence in this cell of two distinct, LVA and HVA, Ca2+ channel conductances is confirmed by the current-voltage relationship (Figure 4G). On average, the LVA current peaked at −40 mV, while the I/V relationship for the HVA current peaked at 0 mV. A HVA Ba2+ current was detectable in 19 out of 27 cells, while a LVA current component was measured in 60% of the cells already expressing a HVA Ca2+ current (n = 13). In summary, NS cell–derived neurons are electrophysiologically active, exhibiting excitability properties and underlying voltage-gated Na+ and Ca2+ conductances typical of maturing nerve cells.
NS Cells Exhibit Phenotypic Similarities to Radial Glia
Undifferentiated NS cells were then examined in more detail to gauge their developmental identity. By RT-PCR and Genechip analyses, they were found to lack pluripotency marker genes such as Oct-4,
nanog, and Eras and markers of mesoderm or endoderm (Figure 5A and unpublished data). They express Pax6, Glast, and BLBP mRNAs (Figure 5A), and are immunopositive for nestin, RC2, vimentin, 3CB2, SSEA1/Lex1, Pax6, and prominin (Figures 5C, 5D, and S1). This set of markers is considered diagnostic for neurogenic radial glia, precursors of both neurons and astrocytes during development of the nervous system [43–45]. Retinoic acid treatment of ES cells has recently been shown to induce radial glia-like cells as transient intermediates during neuronal differentiation [46,47]. Propagation of these radial glia cells was not described in these reports, however.
Figure 5 ES Cell–Derived or Forebrain-Derived NS Cells are Similar to Radial Glia
NS cells were derived from independent ES cell lines (CGR8, E14Tg2a) or primary cortical (Cor-1) and striatal (Str-1) tissue.
(A) RT-PCR of stem cell/radial glia markers.
(B) RT-PCR for pan-neural and region-specific transcriptional regulators.
(C) Double immunostaining for Pax6 and Pax6/RC2 (a,b), Olig2 and Olig2/RC2 (c,d) and Olig2/Pax6 (f). DAPI only for Olig2/Pax6 (e).
(D) The ES cell–derived line (CGR8-NS) and foetal cell–derived line (Cor-1) are indistinguishable from LC1 by morphology and NS cell/radial glial marker immunoreactivity (g,h,k,l), and can each differentiate into neurons (i,m) and astrocytes (j,n).
(E) The ability of Cor-1 to generate neurons (TuJ1+) is retained after 16 passages, more than 30 generations (p,o).
NS cells generally have elongated bipolar morphology, lamellate extensions, end-feet, and oval nuclei anticipated for radial glia [48]. Some more flattened cells and rounded cells with short extensions are also present. Immunostaining for the metaphase marker phosphorylated histone H3 indicates that the highly compacted cells are mitotic (unpublished data). Time-lapse videomicroscopy demonstrates a dynamic interconversion of morphology (Videos S1 and S2). In addition, time lapse reveals that NS cell nuclei undergo pronounced migration up and down the entire length of the cell process (Video S3). Such interkinetic nuclear migration is a well-characterised feature of neuroepithelial and radial glia cells in vivo [49]. It is striking to observe this occur in isolated cells, indicating that nuclear movement is a cell-autonomous property independent of cell–cell contacts or epithelial architecture.
All NS cells examined express the same panel of radial glia markers (Figures 5A and S2), plus the neural precursor markers Sox2, Sox3, and Emx2, and the bHLH (basic helix–loop helix) transcription factors Olig2 and Mash1 (Figure 5B). Although suggestive overall of telencephalic character, this set of markers does not neatly correlate with a specific regional identity. The presence of Olig2 and Mash1 is not characteristic of dorsal forebrain, and may reflect the ex vivo environment and a response to FGF-2. Indeed, Olig2 has recently been found to be induced in spinal cord precursor cells and in neurospheres cultured in FGF-2, indicative of a relaxation of developmental specification [50,51]. Absence of Sox1, but maintenance of Sox2, is a noteworthy feature of NS cells, in view of the postulated determinative function of these transcription factors [52]. Whilst Sox1 marks all early neuroectodermal precursors, our data show that it is not retained in stem cells, where Sox2 may play the key role. NS cells also express Emx2, which is implicated in expansion of neural precursor cells [53,54] (Figures 5B and S1). Also, Dlx2, expressed in transit amplifying neuroblasts, but not stem cells in the sub-ventricular zone [55], is not detected in NS cells.
Most importantly, NS cells appear highly homogenous, staining uniformly for the various antigens examined. Double immunohistochemistry shows co-expression of Pax6/RC2, Olig2/RC2, Olig2/Pax6 (Figure 5C), and Pax6/Emx2 (Figure S1) in virtually every cell. Together with the absence of GFAP and TuJ1, this is indicative of pure symmetrical self-renewal divisions.
Derivation of NS Cells from Foetal Brain and from Neurospheres
ES cells are adapted for in vitro self-renewal [56], and this could in turn predispose for the propagation of derivative tissue stem cells. However, multiple phenotypic characteristics suggest that NS cells may be culture analogues of neurogenic radial glia. We therefore examined whether NS cell derivation depended on an epigenetic configuration carried over from ES cells or if they could be isolated from foetal neural tissue. Primary foetal CNS cells were harvested from embryonic day (E)16.5 mouse foetal forebrain and cultured in NS-A plus EGF/FGF-2. Initially, the cells adhered poorly to plastic and spontaneously formed floating clusters. After 6–7 d, these clusters were transferred to fresh dishes where they settled onto gelatin-coated plastic. Fourteen days later, outgrowths were trypsinised and re-plated. In three separate experiments, cells morphologically identifiable as NS cells proliferated and were subsequently expanded into continuous cell lines. These foetal brain derivatives express the same radial glia and neurogenic markers as the ES cell–derived NS cells (Parts k and l in Figure 5D, and Figure S2) and show consistent mRNA profiles by RT-PCR (Figures 5A and 5B) and by GeneChip analyses (unpublished data). They are likewise competent for neuronal and astrocyte differentiation (Parts m and n in Figure 5D). Cortex-derived Cor-1 cells were plated as single cells, and then colonies were subjected to sequential growth factor withdrawal as described for NS-5. Every colony produced TuJ1-positive neurons. This indicates that all clonogenic cells in the Cor-1 culture are neurogenic. These Cor-1 cells were also readily sub-cloned and continuously expanded from individual cells with retention of phenotypic markers of radial glia, as well as neuronal and astrocyte differentiation potential (Figure S2), indicative of self-renewal. Thus, NS cells derived from foetal brain share the key properties of ES cell–derived NS cells.
It has been reported that cells expressing radial glia markers persist in neurospheres [44] and that neurospheres can “differentiate” into radial glia [57]. It has also been shown that neurospheres can be obtained, albeit inefficiently, by continuous suspension culture during neural differentiation of ES cells [58]. We reasoned that NS cells may in fact be the resident stem cells within the neurosphere. Frozen/thawed, passage 40, mouse neurospheres derived from foetal forebrain were allowed to attach to gelatin-coated plastic in the presence of EGF and FGF-2. Bipolar cells outgrew that are indistinguishable from NS cells. These cells can be serially propagated as uniformly RC2-positive, GFAP-negative populations and then induced to differentiate into astrocytes or neurons (Figure S3). We conclude that radial glia-like cells present in neurospheres give rise to NS cells in adherent culture in the presence of FGF-2 plus EGF. Conversely, we observe that NS cells of either ES cell or foetal brain origin will readily form neurospheres, if detached from the substratum either mechanically or due to overgrowth. This suggests that NS cells/radial glia cells are likely the neurosphere forming stem cells. However, in contrast to adherent cultures, in neurospheres, stem cells constitute only a fraction of the cell population. This is presumably because aggregation induces differentiation, analogous to embryoid body differentiation by ES cell aggregates [59].
Survival and Differentiation of NS Cells Transplanted into the Rodent Brain
We investigated the behaviour of NS cells upon transplantation into mouse brain. ES cell–derived LC1 cells, transduced with a lentiviral enhanced GFP expression vector, were introduced into the developing brain by intra-uterine injection at E14.5 [60]. Animals were sacrificed after birth and the presence of GFP-positive cells was examined in brain sections. Mainly, GFP-labelled NS cell progeny migrated into striatum and cortex, with a few cells in the ventral telencephalon and olfactory bulbs. Immunohistochemical analyses revealed co-expression of enhanced GFP with the precursor marker nestin, neuronal markers TuJ, NeuN, and MAP2, and in lesser numbers with GFAP (Figure S4). NS cells were also injected into the adult mouse hippocampus and striatum. In this case, GFP-positive cells remained localised to the vicinity of the injection site. Four weeks after grafting, 44.4 ± 5.7% of GFP-expressing cells had neuronal morphology and were immunopositive for MAP2, 37.4 ± 6.1% expressed GFAP, and 4.2 ± 1.9% retained expression of nestin (Figure 6). A fraction of cells showed weak expression of either GFAP or neuronal markers that did not justify definitive classification but may be indicative of early stages of differentiation. The proliferative marker Ki67 was detected in only 1.0 ± 0.6% of GFP-positive cells, indicating that NS cells withdraw from the cell cycle in vivo. Well-differentiated GFP-labelled cells are also readily detected by GFP immunostaining in animals sacrificed 15 wk (n = 4) and 6 mo (n = 4) after grafting. We observed no histological evidence of unregulated proliferation or tumour formation in a total of 43 brains examined from 1–6 mo after transplantation, all with donor contributions. Furthermore, NS cells grafted to mouse kidney capsules did not proliferate or give rise to teratomas. These data indicate that NS cells can survive and differentiate in both foetal and adult brain environments and, unlike ES cells [61], they do not give rise to teratomas. Moreover, the relatively high frequency of neuronal differentiation is in contrast to grafts of passaged neurospheres, which appear predisposed in favour of glial differentiation [19,62]. The latter may reflect the incidence of committed glial progenitors present in neurospheres and seemingly absent in NS cell cultures
Figure 6 NS Cells Incorporate and Differentiate within the Adult Brain
(A–H) Confocal images of LC1 NS cells, lentivirally transduced with enhanced GFP, 4 wk post-grafting into hippocampus (A,B) or striatum (C–H). (B) and (D) show higher magnification of the insets in panels (A) and (C), respectively. Examples of enhanced GFP grafted NS cells (green) showing co-expression (yellow) of the neuronal markers TuJ (E, red) or MAP-2 (F, red), astroglial marker GFAP (G, red), neural progenitor marker nestin (H, red).
(I) Quantitative analysis of graft-derived neuronal (MAP2), astroglial (GFAP), progenitor (Nestin), and proliferating (Ki67) cells, 4 wk after transplantation into adult mouse striatum. Data are means (± standard deviation) of at least 500 enhanced GFP+ cells from five independent animals. DG, Dentate Gyrus; ST, Striatum. Scale bars: A,C, 100 μm; B, D, E, 40 μm.; F–H, 20 μm.
Derivation of NS-Like Cells from Human ES Cells and Foetal Cortex
Finally, we investigated whether similar NS cells could be isolated from human sources. In the process of attempting to derive human ES cells from donated supernumerary embryos, we observed, after 5–6 wk of culture, extensive spontaneous differentiation into rosette structures typical of neuroepithelial cells (Parts a and b in Figure 7A). These colonies were manually transferred into feeder-free culture in NS expansion medium. After a further 3–4 wk, bipolar cells similar to NS cells emerged from these cultures (parts c and d in Figure 7A) and have been continuously cultured for 5 mo. We also sourced Carnegie stage 19–20 human foetal cortex from elective terminations. Following tissue dissociation, cells initially formed floating aggregates that after 7 d were re-plated and allowed to attach to gelatin-coated plastic as for derivation of NS cells from mouse foetal brain (Parts i and j in Figure 7B). Proliferating cultures were established (Part k in Figure 7B). Human cultures derived from either ES cells or foetal forebrain were characterized by the presence of flattened cells associated with the bipolar cells. However, all cells express immature precursor markers nestin, vimentin, and 3CB2 (Figures 7A and 7B). Time-lapse monitoring confirmed that the two cell morphologies are plastic and interconvertible (Video S4). These human cells exhibit moderate levels of GFAP unlike mouse NS cells, but consistent with the known activity of the human GFAP promoter in radial glia [48,63]. They proliferate more slowly than the mouse cells, with doubling times of 5–10 d. After sequential withdrawal of EGF and FGF-2, they generate mixed populations of TuJ1-positive neuron-like cells and GFAP-positive cells (Part q in Figure 7C). Near-pure populations of cells with typical astrocyte morphology and intense GFAP immunoreactivity are readily produced after exposure to serum (Part r in Figure 7C). These data suggest that NS cells may be obtained from human as for mouse, although species-specific refinements may be required for optimal propagation and differentiation.
Figure 7 Human ES Cell or Foetal-Derived NS Cells
(A) Derivation from human ES cells: human ES cell primary culture (a), differentiation of human ES cells into neural-rosette structures (b), passage 9 in NS expansion medium (c), individual cells exhibit radial glial morphology (d), and immunostaining for NS cell/radial glia markers (e–h).
(B) Derivation from human foetal forebrain: floating clusters (i) generated from cortex, attachment and outgrowth (j), passage 5 in NS expansion medium (k), radial glia morphology (l), and NS cell/radial glial markers (m–p).
(C) Differentiation of human foetal NS cells: TuJ1 positive neuronal cells generated by sequential growth factor withdrawal (q), and GFAP positive astrocytes induced by exposure to serum (r).
Discussion
The findings reported here establish that the growth factors EGF and FGF-2, plus insulin provided in N2, are sufficient to sustain robust expansion of neural stem cells in defined monoculture, liberated from any requirement for a specific cellular niche. Previous propagation of mammalian CNS precursor cells [64] has predominantly relied on short-term progenitor cell cultures [22,25], genetic immortalization of progenitors [23], or maintenance of stem cells within neurospheres [7]. Progenitor cells from the adult rat hippocampus have been propagated in adherent culture with FGF-2 whilst retaining ability to form neurons [65], albeit at low efficiency without inductive astrocytes [66]. These hippocampal-derived cells are heterogeneous by immunohistochemical staining and “undergo asymmetric cell divisions to continually replenish the supply of multipotent progenitors” [65]. In contrast, the uniformity and stability of marker expression in NS cells, combined with the long-term retention of neuronal differentiation efficiency, indicate that differentiation is completely suppressed by FGF-2 plus EGF in adherent culture. Consequently, the NS cells undergo continuous symmetrical self-renewal divisions.
The critical difference between the adherent culture system we describe and those used by others appears to be the use of EGF in addition to FGF-2. Although often applied to neurosphere cultures, EGF is typically omitted for attached cells. We find that continuous provision of EGF is essential for the derivation and propagation of NS cells, whether sourced from mouse ES cells or foetal brain. This may be related to the role of ErbB signalling in establishment of radial glia in vivo [67] and crosstalk with the Notch pathway [68]. Also, EGF functions to suppress apoptosis in NS cells (Figure 4B and 4D).
The evidence suggests that NS cells are closely related to a radial glia lineage. Whether any radial glia actually function as self-renewing stem cells in vivo is uncertain, since they exist only transiently during foetal development. However, recent fate mapping analysis [69] indicates that a subset of radial glia give rise to the sub-ventricular zone astrocytes that function as adult neural stem cells [70,71]. A possible relationship has also been proposed between hippocampal progenitors and radial glia [72]. The derivation of NS cells may reflect this intrinsic potential of radial glia to convert into stem cells. EGF has been reported to down-regulate expression of Dlx2 in transit-amplifying cells of the sub-ventricular zone and promote their conversion into neurosphere forming cells [55]. Consistent with this, we find that Dlx2 is not expressed in NS cells. Intriguingly, evidence has recently been presented that the EGF receptor undergoes asymmetric segregation in a proportion of RC2-staining neural progenitor cell divisions in vivo [28]. The retention or loss of the EGF receptor is suggested to facilitate alternative fate choices. Retention by both daughters of full responsiveness to EGF may be a central aspect of NS cell propagation. This could represent a crucial divergence from the circumstances of radial progenitor cells in vivo that underpins self-renewal in vitro.
NS cells do not express pluripotent cell-specific transcription factors Oct-4 and Nanog, but show appropriate expression of neural genes and absence of mesoderm and endoderm-specific genes. Their close relationship to a defined endogenous neural precursor cell, radial glia, adds further interest. NS cells can be cryopreserved, and may be transiently or stably transfected by electroporation or lipofection (unpublished data), or transduced with lentiviral vectors. They may be derived from previously engineered ES cells or transgenic mice or be genetically modified after derivation, opening new windows for genetic intervention into self-renewal and lineage commitment decisions in the nervous system and for investigation of neurodegenerative processes and oncogenic transformation. The potential of NS cells to generate different neuronal sub-types has yet to be determined, but their engraftment after transplantation into the adult brain suggests the potential for delivery of cell replacement and gene therapies. Whilst long-term stability and functional integration in vivo will have to be established in future studies, the preliminary data of human analogues to mouse NS cells provides encouragement for this approach.
In the context of fundamental stem cell biology, homogenous expansion of any stem cell in defined conditions has hitherto been the exclusive preserve of the ES cell. Ex vivo propagation of tissue stem cells has invariably been accompanied by differentiation, often interpreted as reflecting an intrinsic bias towards asymmetric division. The findings reported here show that this is definitively not the case for at least one class of neural stem cell. NS cells undergo sustained symmetrical self-renewal divisions with complete suppression of differentiation in response to FGF-2 and EGF. They thus provide a directly accessible system for molecular characterization and experimental manipulation of the stem cell state. Therefore, NS cells offer the first known tissue stem cell resource for direct comparison with ES cells in order to delineate common and distinct mechanistic features of lineage-restricted and pluripotent stem cells.
Materials and Methods
Mouse cell culture and differentiation
ES cells and neural differentiation are detailed elsewhere [38]. The LC1 and other ES cell–derived NS cells were routinely generated by re-plating d 7 adherent neural differentiation cultures (typically 2–3 × 106 cells into a T75 flask) on uncoated plastic in NS-A medium (Euroclone, Milan, Italy) supplemented with modified N2 [36] and 10 ng/ml of both EGF and FGF-2 (NS expansion medium). Over 3–5 d, cells formed aggregates that, after harvesting and sedimentation to remove debris, subsequently attached to fresh plastic and outgrew NS cells. After addition of 0.5 μg/ml of puromycin to differentiating adherent cultures at d 7, 46C-NS cells were generated. Cells were re-plated 3 d later into an uncoated T75 flask in N2B27 media with 10 ng/ml of both EGF and FGF-2 (Peprotech, Rocky Hill, New Jersey, United States) in the absence of puromycin. To derive clonal lines, including NS-5, single cells were plated into 96-well microwell plates (Nalge Nunc International, Rochester, New York, United States) by limiting dilution. and the presence of one cell per well was scored 1 h after plating.
For derivation directly from foetal CNS, primary cultures were generated using standard protocols from cortex or striatum of E16.5 mouse embryos and subsequently allowed to attach on flasks treated with 0.1% gelatin. Outgrowing cells were then expanded on gelatin using NS expansion medium. Clonal derivatives of the cortical line Cor-1 were established by plating at very low density (1,000 cells per 9-cm plate) and expanding individual colonies.
For derivation from established neurospheres, derived from E14 foetal brain and maintained for 40 passages in EGF plus FGF-2, cultures were dissociated to single cells using Accutase (Sigma, St. Louis, Missouri, United States) and plated at 104 cells/ml on gelatin-coated culture flasks in NS expansion medium.
For passaging established NS cell lines, we routinely used trypsin/EDTA or PBS and split cells 1:3 to 1:5 every 2–3 d. For astrocyte differentiations, NS cells were re-plated onto 4-well plates at 1 × 105 cells/well in NS-A medium supplemented with 1% fetal calf serum or 10 ng/ml BMP4 (R&D Systems, Minneapolis, Minnesota, United States). For neuronal differentiation, 5 × 104 NS cells were plated into poly-ornithine/laminin treated wells in NS-A supplemented with FGF-2 alone. After 7 d, the media was switched to NS-A supplemented with B27 (GIBCO, San Diego, California, United States) without growth factor. Half of the medium was exchanged every 2–3 d during the differentiation. For clonal differentiation, 1,000 cells from NS-5 or Cor-1, cultures were plated in 10-cm plates pre-treated with laminin, expanded for 12 d in EGF/FGF-2, and differentiated in situ as above. For electrophysiological studies, 1.5 × 105 NS cells were plated into poly-L-ornithine-treated 35-mm dishes in NS-A medium supplemented with N2 and B27 (both at 0.5%) and FGF-2 (5 ng/ml). After 7 d, the medium was switched to the mix NS-A:Neurobasal (1:1), supplemented with B27 (GIBCO) without growth factors. To sustain neuronal maturation, after a further 7 d, the medium was switched to the mix NS-A:Neurobasal (1:3) supplemented with B27 (GIBCO) and brain derived neurotrophic factor (20 ng/ml) and nerve growth factor (R&D Systems; 50 ng/ml). Throughout neuronal differentiation, half of the medium was replaced every 2–3 d. Further details of NS cell derivation, propagation, and differentiation are provided in Protocol S1.
Characterisation of NS cells
Immunocytochemistry was performed using appropriate TRITC or FITC secondary conjugates and nuclear counterstaining with DAPI. Primary antibodies were used at the following dilutions: Nestin (1:10), Vimentin (1:50), Pax6 (1:5), 3CB2 (1:20), RC2 (1:50) (DSHB, Iowa City, Iowa, United States); TuJ (1:200) (Covance, Berkeley, California, United States); GFAP (1:300) (poly and mono, Sigma); MAP2 (1:200) (Chemicon, Temecula, California, United States; and Becton Dickinson, Palo Alto, California, United States); NeuN (1:200), gamma-aminobutyric acid (1:200), Gad65/67 (1:200) (Chemicon); Synaptophysin (1:200) (Sigma); Doublecortin (1:200) (Santa Cruz Biotechnology, Santa Cruz, California, United States), caspase-3 active (1:300) (R&D Systems), Olig2 (1:5000) (H. Takebayashi); Emx2 (1:2000) (A. Corte); BLBP (1:500) (N. Heins); prominin/mAb13A4 (1:200) (W. Huttner). Negative controls were ES cells, differentiated NS cells, or secondary alone. For RT-PCR, total RNA was extracted using RNeasy kit (Qiagen, Valencia, California, United States), and cDNA was generated using Superscript II (Invitrogen, Carlsbad, California, United States). PCR was performed for 30 cycles for all markers except β-actin (25 cycles). Details of primers and amplicon size are provided in Table S1. For metaphase spreads, cells were treated with 5 ml of 0.56% KCl for 20 min, fixed in methanol:acetic acid (3:1) on ice for 15 min, spread onto glass slides, and stained with TOPRO-3 (Molecular Probes, Eugene, Oregon, United States).
Solutions for electrophysiological and patch clamp recording
Recordings were made from LC1 cells differentiated between passages 20–25. Seals between electrodes and cells were established in a bath solution consisting of (in mmoles/l): 155 NaCl, 1.0 CaCl2, 1 MgCl2, 3.0 KCl, 10 glucose, and 10 HEPES/NaOH (pH 7.4). After establishing the whole-cell configuration, for current-clamp recording and for total current recording in voltage-clamp, the pipette filling solution contained (in mmoles/l): 128 KCl, 10 NaCl, 11 EGTA, 4 Mg-ATP, and 10 HEPES/KOH (pH 7.4). For the study of voltage-gated Na+ channels under voltage clamp conditions, the patch pipette was filled with (in mmoles/l): 130 CsCl, 10 NaCl, 20 TEA-Cl, 10 EGTA, 2 MgCl2, 4 Mg-ATP, and 10 HEPES/CsOH (pH 7.4), and the extracellular solution contained (in mmoles/l): 130 NaCl, 2 CaCl2, 2 MgCl2, 10 glucose, 5 tetrethylammonium-Cl, CdCl2 0.2, and 10 HEPES/NaOH (pH 7.4). For the study of voltage-gated Ca2+ channels, the patch pipette was filled with (in mmoles/l): 120 CsCl, 20 TEA-Cl, 10 EGTA, 2 MgCl2, 4 Mg-ATP, and 10 HEPES/CsOH (pH 7.4), and the extracellular solution contained (in mmoles/l): 130 NaCl, 10 BaCl2, 10 glucose, 5 tetrethylammonium-Cl, 10 4-AP 1, TTX 10−3, and HEPES/NaOH (pH 7.4). Ionic currents were recorded under voltage-clamp conditions using the patch-clamp whole-cell configuration at room temperature (20–24 °C) with an Axopatch 200B patch-clamp amplifier (Axon Instruments, Burlingame, California, United States) and digitized at sampling intervals of 26–100 μsec using a Digidata 1322A A/D converter (Axon Instruments) interfaced with an IBM-compatible PC. Stimulation, acquisition, and data analysis were carried out using the software packages: pClamp 9 (Axon Instruments) and ORIGIN 6 (Microcal Software, Northampton, Massachusetts, United States). For voltage-clamp experiments, linear components of leak and capacitive currents were first reduced by analogue circuitry and then almost completely cancelled with the P/N method. Patch pipettes were made from borosilicate glass tubing and fire polished. Pipettes had a final resistance of 3–4 MΩ when filled with internal solution. Currents were filtered at 5 KHz.
Human embryo and foetal cultures
Research on human tissue with informed consent was approved by the Research Ethics Committee of Lothian Health Board. Frozen supernumerary human embryos were donated for research under licence R0132 issued by the Human Fertilisation and Embryology Authority. Inner cell masses were isolated by immunosurgery and cultured on human foreskin fibroblasts in medium supplemented with 15% serum replacement (Invitrogen) plus human leukaemia inhibitory factor and FGF-2 [73]. After three to four passages, cells with ES cell morphology differentiated into rosettes of neuroepithelial-like cells. These colonies were passaged into NS expansion medium without feeders or serum replacement. Human foetal tissue was obtained following elective termination with consent for research according to the Polkinghorne guidelines [74]. Cortex was dissected from Carnegie stage 19–20 foetuses and processed as described for mouse foetal tissue. In some cases, LIF (100 U/ml) was added to the expansion medium [75].
To induce neuronal differentiation, a similar protocol was followed as for mouse NS cells but with addition of brain derived neurotrophic factor (R&D Systems; 10 ng/ml) after the first 7 d without EGF, and retention of FGF-2 (5 ng/ml) until 14 d.
Transplants
To provide a convenient reporter, LC1 cells at passage 12 were transduced with lentiviral GFP. Transplants were performed after expansion of these cells for a further 9–28 passages. Foetal surgery was performed as described previously [60]. Using a glass capillary, 5 × 104 cells in a volume of 1 μl of HBSS were injected into the telencephalic vesicles of E14.5 Sprague-Dawley rat foetuses exposed under transillumination. Injected foetuses were replaced into the abdominal cavity for development to term. After delivery, animals were sacrificed at 7 d (postnatal day [P] 1, n = 16) and 5 wk (P30, n = 8) post-transplantation. For adult transplantations, 129 or CD1 mice were placed in a Kopf stereotaxic frame and received an injection of 2 × 105 NS cells suspended in 5 μl of HBSS into the striatum (n = 22) or hippocampus (n = 21). Transplanted mice were sacrificed after 2 (n = 16) and 4 wk (n = 10) and perfused transcardially with 4% paraformaldehyde. Cryosections (16 μm) were stained with the following antibodies: (mouse), NeuN (1:100) and Ki67 (1:10) (Chemicon), MAP2 (1:200; Becton Dickinson), Nestin (1:5; Ron McKay); (rabbit), βIII tubulin (1:500; Covance); GFAP (1:200; Dako, Glostrup, Denmark); secondary antibodies, Texas Red (Vector Laboratories, Burlingame, California, United States) (Jackson ImmunoResearch, West Grove, Pennsylvania, United States) and AlexaFluor 488 (Molecular Probes). Sections were preserved in antifading solution and analysed on Nikon TE2000-S ECLIPSE (Nikon, Tokyo, Japan) and Bio-Rad Radiance 2100 (Bio-Rad, Hercules, California, United States) confocal microscopes. Further cohorts of animals were sacrificed after 15 wk (n = 4) and 6 mo (n = 4) and analysed by antibody staining for GFP because the fluorescence signal was very low.
Supporting Information
Figure S1 Immunohistochemical Staining of NS Cells
(A) Radial glia/neural precursor markers: GLAST (Slc1a3) (a), prominin (b), and LeX/SSEA-1 (c), in NS-5 cells (equivalent results with LC1 not shown).
(B) Co-expression of transcription factors Emx2 and Pax6: DAPI (d), Emx2 (e), and Pax6 (f) double immunostaining (g, overlay) of LC1 NS cells (equivalent results with NS-5 not shown).
(292 KB PDF).
Click here for additional data file.
Figure S2 Mouse ES Cell–Derived and Foetal Cortex–Derived NS Cells Uniformly Express Radial Glia Markers
NS cells derived from CGR8 ES cells or from E16 foetal cortex (Cor-1 and clonal derivative Cor-1.3) were analysed for expression of the indicated markers by immunochemistry. Examination at high power shows that the radial glia markers are each expressed in almost all cells whilst they are uniformly negative for GFAP.
(641 KB PDF).
Click here for additional data file.
Figure S3 NS Cells Can Be Derived from Expanded Foetal Forebrain Neurospheres
NS line derived from a long-term foetal neurosphere culture (40 passages) exhibits identical morphology (a) to ES-derived NS lines, expresses neural precursor cell/radial glial marker immunoreactivity (b–d), and can differentiate into neurons (e) and astrocytes (f).
(284 KB PDF).
Click here for additional data file.
Figure S4 NS Cells Migrate and Differentiate after Transplantation in Foetal Rat Brain
Confocal image of NS cells, lentivirally transduced with enhanced GFP, 1 wk after transplantation into the ventricle of E14.5 rats (a). Donor cells migrate from the ventricle into the parenchyma in clusters and as single cells. Grafted cells show co-localization (yellow) of enhanced GFP (green) and; the neuronal marker MAP2 (b, red); astroglia marker GFAP (c, red) or; progenitor cell marker, nestin (d, red). Quantitative analysis (e) of graft-derived neuronal (NeuN and TuJ-1), astroglial (GFAP), progenitor (Nestin), and proliferating (Ki67) cells, one week after transplantation. Data are means (± standard deviation) of at least 500 enhanced GFP+ cells from five independent animals. LV, lateral ventricle. Scale bars: (a) 200 μm; (b–d), 20 μm.
(154 KB PDF).
Click here for additional data file.
Protocol S1 Derivation and Manipulation of NS Cell Lines
(59 KB DOC).
Click here for additional data file.
Table S1 Primers Used for RT-PCR
(74 KB DOC).
Click here for additional data file.
Video S1 Mouse ES Cell–Derived (E14T-NS) NS Cells Show Dynamic Morphology (Low Magnification)
3 frames/second; 22 s running time.
(7.5 MB AVI).
Click here for additional data file.
Video S2 Mouse ES Cell–Derived (E14T-NS) NS Cells Show Dynamic Morphology (Higher Magnification)
3 frames/second; 72 s running time.
(4 MB MOV).
Click here for additional data file.
Video S3 Mouse Foetal Cortex–Derived NS Cell Exhibits Interkinetic Nuclear Migration
5 frames/second; 43 s running time.
(1.3 MB AVI).
Click here for additional data file.
Video S4 Human Foetal Cortex–Derived NS Cells Show Dynamic Morphology, Alternating between Bipolar and Spread-Out States
10 frames/second; 20 s running time.
(2.1 MB AVI).
Click here for additional data file.
Accession Numbers
The Swiss-Prot (http://www.ebi.ac.uk/swissprot) accession numbers for the genes and gene products discussed in this paper are Ascl1 (Mash1) (Q02067), Dlx2 (P40764), Egf (P01132), Emx2 (Q04744), Fabp7 (BLBP, B-FABP), (P51880), Fgf2 (bFGF) (P15655), Gad1 (Gad67) (P48318), Gfap (P03995), Nanog (Q7TN85), Nestin (Q6P5H2), Olig2 (Q9EQW6), Pax6 (P63015), Pax7 (P47239), Pou5f1 (Oct4, Oct3/4) (P20263), Slc1a3 (Glast) (P56564), Sox1 (P53783), Sox2 (P48432), Sox3 (P53784), and Vim (P20152).
We are indebted to Helen Mardon, Barbara Evans, and Janet Carver at the University of Oxford and Oxford Fertility Unit; to Richard Anderson, Anne Johnstone, and Joan Creiger at the Edinburgh Royal Infirmary; and to all tissue donors. We thank colleagues who generously provided antibodies. Richard Wallbank contributed to mouse NS cell analysis; Louise Taylor, and Jenny Nichols contributed to human cell cultures; Giulio Simonutti and David Kelly contributed to imaging and time lapse, and Lorenzo Magrassi contributed to transplantation. We are grateful to Sally Lowell for comments on the manuscript. This research was supported by the Biotechnology and Biological Sciences Research Council UK, the Medical Research Council UK, the Wellcome Trust, Telethon, Fondo Incentivazione Ricerca di Base (MIUR), and the European Commission Framework VI Programme, EuroStemCell.
Competing interests. The University of Edinburgh has filed a patent application on methods of deriving and culturing neural stem cells relating to this study. This patent has been licensed to Stem Cell Sciences Ltd. AS holds non-voting equity (ca 5%) in Stem Cell Sciences Ltd.
Author contributions. LC, SMP,TG, EC, and AS conceived and designed the experiments. LC, SMP, TG, ER, MT, GB, YS, SS, QLY, and AS performed the experiments. LC, SMP, TG, ER, MT, GB, and YS analyzed the data. LC, SMP and AS wrote the paper.
¤ Current address: Liggins Institute, University of Auckland, Auckland, New Zealand
Citation: Conti L, Pollard SM, Gorba T, Reitano E, Toselli M, et al. (2005) Niche-independent symmetrical self-renewal of a mammalian tissue stem cell. PLoS Biol 3(9): e283.
Abbreviations
BLBPbrain lipid binding protein
CNScentral nervous system
Eembryonic day
EGFepidermal growth factor
ESembryonic stem
FGF-2fibroblast growth factor 2
GFAPglial fibrillary acidic protein
GFPgreen fluorescent protein
HVAhigh-voltage activated
LVAlow-voltage activated
MAP2microtubule associated protein-2
NSneural stem
Ppostnatal day
==== Refs
References
Schofield R The relationship between the spleen colony-forming cell and the hemopoietic stem cell Blood Cells 1978 4 7 25 747780
Watt FM Hogan BL Out of Eden: Stem cells and their niches Science 2000 287 1427 1430 10688781
Spradling A Drummond-Barbosa D Kai T Stem cells find their niche Nature 2001 414 98 104 11689954
Doetsch F A niche for adult neural stem cells Curr Opin Genet Dev 2003 13 543 550 14550422
Potten CS Loeffler M Stem cells: Attributes, cycles, spirals, pitfalls and uncertainties. Lessons for and from the crypt Development 1990 110 1001 1020 2100251
Barrandon Y Green H Three clonal types of keratinocyte with different capacities for multiplication Proc Natl Acad Sci U S A 1987 84 2302 2306 2436229
Reynolds BA Weiss S Generation of neurons and astrocytes from isolated cells of the adult mammalian central nervous system Science 1992 255 1707 1710 1553558
Ying QL Nichols J Chambers I Smith A BMP induction of Id proteins suppresses differentiation and sustains embryonic stem cell self-renewal in collaboration with STAT3 Cell 2003 115 281 292 14636556
Lim DA Tramontin AD Trevejo JM Herrera DG Garcia-Verdugo JM Noggin antagonizes BMP signaling to create a niche for adult neurogenesis Neuron 2000 28 713 726 11163261
Shen Q Goderie SK Jin L Karanth N Sun Y Endothelial cells stimulate self-renewal and expand neurogenesis of neural stem cells Science 2004 304 1338 1340 15060285
Doetsch F Caille I Lim DA Garcia-Verdugo JM Alvarez-Buylla A Subventricular zone astrocytes are neural stem cells in the adult mammalian brain Cell 1999 97 703 716 10380923
Garcion E Halilagic A Faissner A ffrench-Constant C Generation of an environmental niche for neural stem cell development by the extracellular matrix molecule tenascin C Development 2004 131 3423 3432 15226258
Weiss S Dunne C Hewson J Wohl C Wheatley M Multipotent CNS stem cells are present in the adult mammalian spinal cord and ventricular neuroaxis J Neurosci 1996 16 7599 7609 8922416
Sanai N Tramontin AD Quinones-Hinojosa A Barbaro NM Gupta N Unique astrocyte ribbon in adult human brain contains neural stem cells but lacks chain migration Nature 2004 427 740 744 14973487
Morshead CM Reynolds BA Craig CG McBurney MW Staines WA Neural stem cells in the adult mammalian forebrain: A relatively quiescent subpopulation of subependymal cells Neuron 1994 13 1071 1082 7946346
Suslov ON Kukekov VG Ignatova TN Steindler DA Neural stem cell heterogeneity demonstrated by molecular phenotyping of clonal neurospheres Proc Natl Acad Sci U S A 2002 99 14506 14511 12381788
Sinor AD Lillien L Akt-1 expression level regulates CNS precursors J Neurosci 2004 24 8531 8541 15456827
Morshead CM Benveniste P Iscove NN van der Kooy D Hematopoietic competence is a rare property of neural stem cells that may depend on genetic and epigenetic alterations Nat Med 2002 8 268 273 11875498
Winkler C Fricker RA Gates MA Olsson M Hammang JP Incorporation and glial differentiation of mouse EGF-responsive neural progenitor cells after transplantation into the embryonic rat brain Mol Cell Neurosci 1998 11 99 116 9647689
Rossi F Cattaneo E Opinion: Neural stem cell therapy for neurological diseases: Dreams and reality Nat Rev Neurosci 2002 3 401 409 11988779
Cattaneo E McKay R Proliferation and differentiation of neuronal stem cells regulated by nerve growth factor Nature 1990 347 762 765 2172829
Johe KK Hazel TG Muller T Dugich-Djordjevic MM McKay RD Single factors direct the differentiation of stem cells from the fetal and adult central nervous system Genes Dev 1996 10 3129 3140 8985182
Frederiksen K Jat PS Valtz N Levy D McKay R Immortalization of precursor cells from the mammalian CNS Neuron 1988 1 439 448 2856096
Ryder EF Snyder EY Cepko CL Establishment and characterization of multipotent neural cell lines using retrovirus vector-mediated oncogene transfer J Neurobiol 1990 21 356 375 2307979
Temple S Division and differentiation of isolated CNS blast cells in microculture Nature 1989 340 471 473 2755510
Qian X Shen Q Goderie SK He W Capela A Timing of CNS cell generation: A programmed sequence of neuron and glial cell production from isolated murine cortical stem cells Neuron 2000 28 69 80 11086984
Chenn A McConnell SK Cleavage orientation and the asymmetric inheritance of Notch1 immunoreactivity in mammalian neurogenesis Cell 1995 82 631 641 7664342
Sun Y Goderie SK Temple S Asymmetric distribution of EGFR receptor during mitosis generates diverse CNS progenitor cells Neuron 2005 45 873 886 15797549
Shen Q Zhong W Jan YN Temple S Asymmetric Numb distribution is critical for asymmetric cell division of mouse cerebral cortical stem cells and neuroblasts Development 2002 129 4843 4853 12361975
Bain G Kitchens D Yao M Huettner JE Gottlieb DI Embryonic stem cells express neuronal properties in vitro Dev Biol 1995 168 342 357 7729574
Li M Pevny L Lovell-Badge R Smith A Generation of purified neural precursors from embryonic stem cells by lineage selection Curr Biol 1998 8 971 974 9742400
Okabe S Forsberg-Nilsson K Spiro AC Segal M McKay RDG Development of neuronal precursor cells and functional postmitotic neurons from embryonic stem cells in vitro Mech Dev 1996 59 89 102 8892235
Brustle O Jones KN Learish RD Karram K Choudhary K Embryonic stem cell-derived glial precursors: A source of myelinating transplants Science 1999 285 754 756 10427001
Temple S The development of neural stem cells Nature 2001 414 112 117 11689956
Ying QL Stavridis M Griffiths D Li M Smith A Conversion of embryonic stem cells into neuroectodermal precursors in adherent monoculture Nat Biotechnol 2003 21 183 186 12524553
Ying QL Smith AG Defined conditions for neural commitment and differentiation Methods Enzymol 2003 365 327 341 14696356
Billon N Jolicoeur C Ying QL Smith A Raff M Normal timing of oligodendrocyte development from genetically engineered, lineage-selectable mouse ES cells J Cell Sci 2002 115 3657 3665 12186951
Ying QL Smith AG Defined conditions for neural commitment and differentiation Methods Enzymol 2003 365 327 341 14696356
Aubert J Stavridis MP Tweedie S O'Reilly M Vierlinger K Screening for mammalian neural genes via fluorescence-activated cell sorter purification of neural precursors from Sox1-gfp knock-in mice Proc Natl Acad Sci U S A 2003 100 11836 11841 12923295
Stavridis MP Smith AG Neural differentiation of mouse embryonic stem cells Biochem Soc Trans 2003 31 45 49 12546651
Brewer GJ Torricelli JR Evege EK Price PJ Optimized survival of hippocampal neurons in B27-supplemented Neurobasal, a new serum-free medium combination J Neurosci Res 1993 35 567 576 8377226
Carbone E Lux HD Kinetics and selectivity of a low-voltage–activated calcium current in chick and rat sensory neurones J Physiol 1987 386 547 570 2445968
Campbell K Gotz M Radial glia: Multi-purpose cells for vertebrate brain development Trends Neurosci 2002 25 235 238 11972958
Hartfuss E Galli R Heins N Gotz M Characterization of CNS precursor subtypes and radial glia Dev Biol 2001 229 15 30 11133151
Noctor SC Flint AC Weissman TA Dammerman RS Kriegstein AR Neurons derived from radial glial cells establish radial units in neocortex Nature 2001 409 714 720 11217860
Liour SS Yu RK Differentiation of radial glia-like cells from embryonic stem cells Glia 2003 42 109 117 12655595
Bibel M Richter J Schrenk K Tucker KL Staiger V Differentiation of mouse embryonic stem cells into a defined neuronal lineage Nat Neurosci 2004 7 1003 1009 15332090
Rakic P Elusive radial glial cells: Historical and evolutionary perspective Glia 2003 43 19 32 12761862
Noctor SC Martinez-Cerdeno V Ivic L Kriegstein AR Cortical neurons arise in symmetric and asymmetric division zones and migrate through specific phases Nat Neurosci 2004 7 136 144 14703572
Gabay L Lowell S Rubin LL Anderson DJ Deregulation of dorsoventral patterning by FGF confers trilineage differentiation capacity on CNS stem cells in vitro Neuron 2003 40 485 499 14642274
Hack MA Sugimori M Lundberg C Nakafuku M Gotz M Regionalization and fate specification in neurospheres: The role of Olig2 and Pax6 Mol Cell Neurosci 2004 25 664 678 15080895
Pevny LH Sockanathan S Placzek M Lovell-Badge R A role for SOX1 in neural determination Development 1998 125 1967 1978 9550729
Heins N Cremisi F Malatesta P Gangemi RM Corte G Emx2 promotes symmetric cell divisions and a multipotential fate in precursors from the cerebral cortex Mol Cell Neurosci 2001 18 485 502 11922140
Galli R Fiocco R De Filippis L Muzio L Gritti A Emx2 regulates the proliferation of stem cells of the adult mammalian central nervous system Development 2002 129 1633 1644 11923200
Doetsch F Petreanu L Caille I Garcia-Verdugo JM Alvarez-Buylla A EGF converts transit-amplifying neurogenic precursors in the adult brain into multipotent stem cells Neuron 2002 36 1021 1034 12495619
Buehr M Smith A Genesis of embryonic stem cells Philos Trans R Soc Lond B Biol Sci 2003 358 1397 1402 14511487
Gregg C Weiss S Generation of functional radial glial cells by embryonic and adult forebrain neural stem cells J Neurosci 2003 23 11587 11601 14684861
Tropepe V Hitoshi S Sirard C Mak TW Rossant J Direct neural fate specification from embryonic stem cells: A primitive mammalian neural stem cell stage acquired through a default mechanism Neuron 2001 30 65 78 11343645
Doetschman TC Eistetter H Katz M Schmidt W Kemler R The in vitro development of blastocyst-derived embryonic stem cell lines: Formation of visceral yolk sac, blood islands and myocardium J Embryol Exp Morphol 1985 87 27 45 3897439
Magrassi L Ehrlich ME Butti G Pezzotta S Govoni S Basal ganglia precursors found in aggregates following embryonic transplantation adopt a striatal phenotype in heterotopic locations Development 1998 125 2847 2855 9655807
Brustle O Spiro AC Karram K Choudhary K Okabe S In vitro-generated neural precursors participate in mammalian brain development Proc Natl Acad Sci U S A 1997 94 14809 14814 9405695
Englund U Bjorklund A Wictorin K Migration patterns and phenotypic differentiation of long-term expanded human neural progenitor cells after transplantation into the adult rat brain Brain Res Dev Brain Res 2002 134 123 141 11947943
Malatesta P Hartfuss E Gotz M Isolation of radial glial cells by fluorescent-activated cell sorting reveals a neuronal lineage Development 2000 127 5253 5263 11076748
Gottlieb DI Large-scale sources of neural stem cells Annu Rev Neurosci 2002 25 381 407 12052914
Palmer TD Takahashi J Gage FH The adult rat hippocampus contains primordial stem cells Mol Cell Neurosci 1997 8 389 404 9143557
Song H Stevens CF Gage FH Astroglia induce neurogenesis from adult neural stem cells Nature 2002 417 39 44 11986659
Schmid RS McGrath B Berechid BE Boyles B Marchionni M Neuregulin 1-erbB2 signaling is required for the establishment of radial glia and their transformation into astrocytes in cerebral cortex Proc Natl Acad Sci U S A 2003 100 4251 4256 12649319
Ever L Gaiano N Radial ‘glial' progenitors: Neurogenesis and signaling Curr Opin Neurobiol 2005 15 29 33 15721741
Merkle FT Tramontin AD Garcia-Verdugo JM Alvarez-Buylla A Radial glia give rise to adult neural stem cells in the subventricular zone Proc Natl Acad Sci U S A 2004 101 17528 17532 15574494
Alvarez-Buylla A Garcia-Verdugo JM Tramontin AD A unified hypothesis on the lineage of neural stem cells Nat Rev Neurosci 2001 2 287 293 11283751
Doetsch F The glial identity of neural stem cells Nat Neurosci 2003 6 1127 1134 14583753
Alvarez-Buylla A Doetsch F Seril B Garcia-Verdugo JM Gage FH Bjorklund A Prochiantz A Christen Y Astrocytic nature of adult neural stem cells in vivo Stem cells in the nervous system: Functional and clinical implications 2004 Berlin Springer-Verlag 43 56
Hovatta O Mikkola M Gertow K Stromberg AM Inzunza J A culture system using human foreskin fibroblasts as feeder cells allows production of human embryonic stem cells Hum Reprod 2003 18 1404 1409 12832363
The Polkinghorne Report Review of the guidance on the research use of fetuses and fetal material 1989 London HMSO
Guyette TW Carpenter MA Accuracy of pressure-flow estimates of velopharyngeal orifice size in an analog model and human subjects J Speech Hear Res 1988 31 537 548 3230884
|
16086633
|
PMC1184591
|
CC BY
|
2021-01-05 08:21:27
|
no
|
PLoS Biol. 2005 Sep 16; 3(9):e283
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030283
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1608660810.1371/journal.pbio.0030289Research ArticleDevelopmentEvolutionPhysiologyNutritionDrosophilaThe Temporal Requirements for Insulin Signaling During Development in Drosophila
Temporal Requirements for InsulinShingleton Alexander W [email protected]
1
Das Jayatri
1
Vinicius Lucio
2
Stern David L
1
1Ecology and Evolutionary Biology, Princeton University, Princeton, New Jersey, United States of America,2Leverhulme Centre for Human Evolutionary Studies, University of Cambridge, Cambridge, United KingdomJohnston Laura Academic EditorColumbia UniversityUnited States of America9 2005 16 8 2005 16 8 2005 3 9 e2891 10 2004 17 6 2005 Copyright: © 2005 Shingleton 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.
For Insulin Signaling Pathways in Flies, Size Matters
Recent studies have indicated that the insulin-signaling pathway controls body and organ size in Drosophila, and most metazoans, by signaling nutritional conditions to the growing organs. The temporal requirements for insulin signaling during development are, however, unknown. Using a temperature-sensitive insulin receptor (Inr) mutation in Drosophila, we show that the developmental requirements for Inr activity are organ specific and vary in time. Early in development, before larvae reach the “critical size” (the size at which they commit to metamorphosis and can complete development without further feeding), Inr activity influences total development time but not final body and organ size. After critical size, Inr activity no longer affects total development time but does influence final body and organ size. Final body size is affected by Inr activity from critical size until pupariation, whereas final organ size is sensitive to Inr activity from critical size until early pupal development. In addition, different organs show different sensitivities to changes in Inr activity for different periods of development, implicating the insulin pathway in the control of organ allometry. The reduction in Inr activity is accompanied by a two-fold increase in free-sugar levels, similar to the effect of reduced insulin signaling in mammals. Finally, we find that varying the magnitude of Inr activity has different effects on cell size and cell number in the fly wing, providing a potential linkage between the mode of action of insulin signaling and the distinct downstream controls of cell size and number. We present a model that incorporates the effects of the insulin-signaling pathway into the Drosophila life cycle. We hypothesize that the insulin-signaling pathway controls such diverse effects as total developmental time, total body size and organ size through its effects on the rate of cell growth, and proliferation in different organs.
Studies using a temperature-sensitive insulin receptor elucidate the temporal requirements for insulin signaling in Drosophila; insulin signaling at different times during development affects many characters, such as total developmental time, total body size and organ size.
==== Body
Introduction
Development in multicellular animals is a process that involves both tight control and flexibility in the regulation of cell size and cell number. Tight control is necessary to produce an animal in which each organ is of an appropriate size relative to the whole body. Flexibility is necessary to produce an animal in which the whole body is of an appropriate size relative to environmental conditions. One key environmental factor shown to influence size is nutrition. In animals as diverse as humans and flies, malnutrition delays development and reduces adult body and organ size. Recent studies indicate that the insulin-signaling pathway (Gene Ontology ID GO:0008286) coordinates growth with nutritional condition in most metazoans, and is remarkably conserved. In Drosophila, expression of insulin-like peptides is nutritionally regulated [1], as is expression of IGF1 (insulin-like growth factor 1) in developing rats [2]. Flies carrying mutations of the insulin receptor (Inr) and mice carrying mutations of the IGF1 receptor show delayed development and growth deficiency with a reduction in body and organ size [3–6]. These and similar “knock out” experiments have demonstrated the gross effects of the insulin-signaling pathway on adult phenotype. Nevertheless, little is known of how the pathway acts during development to affect changes in the adult. In mice, such elucidation is hampered by the inaccessibility of the developing fetus enclosed in the uterus. In flies, however, developing larvae can be easily studied and manipulated.
Here we explore the temporal requirement for insulin signaling in developing Drosophila
melanogaster. In Drosophila, development proceeds through three larval instars to pupariation, pupation, and finally adult eclosion. Feeding is restricted to the first (L1), second (L2), and most of the third (L3) instar. Early in the third instar, the larvae reach a “critical size” at which point they have acquired sufficient nutrients to complete development in the absence of food [7–10]. After critical size is attained, larvae stop feeding, leave the food, and search for a pupariation site. There is, however, a delay between the time critical size is reached and the time larvae stop feeding and reach their maximum body size. Both critical size and the time between reaching critical size and the cessation of feeding are unaffected by nutrition [8,10]. Consequently, nutrition ostensibly affects final adult size through the amount of feeding in the fixed period between critical size and pupariation [10].
Because feeding affects final size only in the period between critical size and pupariation [8,10], one may predict that insulin signaling affects size only during the same period. However, organ growth continues well after pupariation [11,12] and is autonomously modulated by the insulin-signaling pathway [5,13]. Insulin signaling may therefore continue to influence size after pupariation. Details of the temporal requirements for insulin signaling are unknown in Drosophila, but a recent study on the Drosophila forkhead transcription factor (dFOXO) suggests that they vary during development [14]. Unphosphorylated dFOXO negatively regulates growth, but activation of the insulin receptor signaling pathway induces dFOXO phosphorylation, excluding it from the nucleus and inhibiting its activity [15,16]. Constitutive activation of dFOXO in the first and second instars causes developmental arrest, whereas dFOXO activation in the third instar causes a reduction in adult size [14]. Work on Caenorhabditis elegans also indicates that the timing requirements for insulin signaling might vary during the lifecycle. In C. elegans, the pathway acts during early development to regulate diapause, and during adulthood to influence ageing [17]. Resolving the links between larval size, nutrition, and developmental progression on the one hand, and the time-dependence of insulin-pathway gene effects on the other, requires a more precise understanding of the temporal requirements for insulin signaling during development.
We have used a temperature-sensitive mutation of the insulin receptor to investigate the role of the insulin-signaling pathway during different stages of development in Drosophila. By reducing insulin receptor activity at different points in development, we identify the periods of development during which insulin signaling affects adult phenotype. We show that total developmental time is affected by changes in insulin signaling only if those changes occur early in development, before a larva has reached critical size. In contrast, final body and organ size are affected by changes in insulin signaling only if those changes occur late in development, once a larva has passed critical size. Insulin signaling continues to influence organ size, but not body size, after larvae have stopped feeding and throughout much of pupal development. Not all organs respond equally to changes in insulin signaling, however. We find that the genitals show a limited response to suppression of insulin signaling in InrE19/InrGC25 flies, implicating insulin signaling in the control of allometries.
Results
Temperature-Sensitive Suppression of the Insulin Receptor
During a study of the interaction of environmental factors and the insulin-signaling pathway, we discovered that flies trans-heteroallelic for insulin receptor mutations InrE19 and InrGC25 show a temperature-sensitive suppression of the insulin pathway. We initially reared InrE19/InrGC25 flies, wild-type Oregon-R flies, and flies homozygous for chico1, at 18 °C, 25 °C, and 29 °C, and measured the wing areas of the adults. Chico is the fly ortholog of the insulin receptor substrate. chico1 is a null mutation, and homozygotes show a partial loss of insulin receptor function with a more than 50% reduction in body size relative to wild-type and a much reduced viability [5]. Both Oregon-R and chico1 homozygote flies showed a smaller wing area with higher rearing temperature (Figure 1). The response to temperature is the same in both genotypes (the lines relating wing size to temperature are parallel in Figure 1), indicating that temperature influences size independent of the insulin-signaling pathway. However, InrE19/InrGC25 flies have a wing size similar to Oregon-R flies when reared at 18 °C but a wing size similar to chico1 mutants when reared at 25 °C (Figure 1). This suggests that InrE19/InrGC25 flies show temperature-sensitive suppression of Inr activity. Changing the rearing temperature does not, however, simply switch Inr “on” or “off” in InrE19/InrGC25 flies. At 18 °C, InrE19/InrGC25 flies have smaller wings than wild-type flies, suggesting that the insulin receptor is still partially suppressed at this temperature. Additionally, very few InrE19/InrGC25 flies survived to adulthood when raised at 25 °C. and none survived at 29 °C. The difference between InrE19/InrGC25 flies reared at 18 °C and 25 °C appears, therefore, to be a consequence of the degree to which Inr is suppressed.
Figure 1 Temperature Sensitivity in InrE19/InrGC25 Flies
Increasing the rearing temperature of Inr
E19/InrGC25 females from 18 °C to 25 °C causes a reduction in wing area from approximately wild-type (Oregon-R [Ore-R]) to that of an insulin pathway mutant (chico). Wing size is expressed as percentage area of Oregon-R female wing at 25 °C. No InrE19/InrGC25 flies survived rearing at 29 °C. The standard errors are smaller than the markers.
To confirm that these temperature-dependent effects were due to changes in Inr activity, we assessed the activity of the insulin-signaling pathway in InrE19/InrGC25 flies reared at different temperatures. First, we looked at the cellular localization of dFOXO in InrE19/InrGC25 late third instar larvae reared at 17 °C and at 24 °C. Down-regulation of the insulin-signaling pathway dephosphorylates dFOXO and causes it to move from the cytoplasm to the nucleus. Antibody stains with anti-dFOXO show an increase in nuclear versus cytoplasmic localization of dFOXO in InrE19/InrGC25 larvae when reared at 24 °C relative to 17 °C (Figure 2A and 2B). We did not observe change in nuclear versus cytoplasmic staining for InrE19 or InrGC25/TM6B-Tb controls reared at 17 °C versus 25 °C (Figure 2C and 2D). Second, we assessed Inr activity in InrE19/InrGC25 second instar larvae using the tGPH reporter gene [18]. tGPH is under the control of the Drosophila ß-tubulin promotor and produces a GFP (green fluorescent protein) fused to the pleckstrin homology domain of the receptor for phosphoinositides-1. Under the action of phosophoinositide 3-kinase (PI3K), this fusion protein, GPH, becomes localized to the cell membrane. PI3K is itself activated by the insulin receptor, and so increased localization of GPH to the cell membrane indicates increased Inr activity. In InrE19/InrGC25 larvae reared at 15 °C, GPH is localized to the membrane, but this localization is lost when the larvae are moved to 25 °C for 12 h (Figure 2E and 2F). Localization does not appear to be lost at 25 °C in InrE19 or InrGC25/TM6B-Tb controls (Figure 2G and 2H). The phenotypic effects of temperature on InrE19/InrGC25 flies do, therefore, appear to be a consequence of temperature-sensitive suppression of the insulin-signaling pathway. However, we do not know the molecular basis for the temperature sensitivity of the Inr
E19
/Inr
GC25 transheterozygotes. It may arise from temperature-sensitive expression of one or both of the mutant alleles or from decreased function of the mutant receptor at higher temperatures.
Figure 2 Increasing the Temperature of InrE19/InrGC25 Flies Suppresses the Insulin-Signaling Pathway
The dFOXO panel shows localization of dFOXO protein in the fat body, the propidium iodide panel shows the position of the nuclei, and the merge panel clarifies the degree of dFOXO localization to the nuclei.
(A) Endogenous dFOXO in the fat body of InrE19/InrvGC25 third instar larvae has weak nuclear localization at 17 °C.
(B) Increase in rearing temperature causes a decrease in cytoplasmic distribution and an increase in nuclear localization of dFOXO, consistent with a decrease in the level of insulin signaling
(C and D) Temperature has no detectable effect on dFOXO localization in InrE19/TM3 control flies.
(E) GPH membrane localization reveals high levels of insulin signaling in the fat body of InrE19/InrGC25 second instar larvae reared at 15 °C. GPH is in green, DNA is stained blue. (F) This localization is lost when the larvae are moved to 25 °C for 12 h, consistent with a decrease in the level of insulin signaling.
(G and H) Temperature has no detectable effect on GPH membrane localization in InrE19/TM3 control flies.
Due to the low viability of InrE19/InrGC25 flies reared at 25 °C, all subsequent experiments involved comparisons of flies reared at 17 °C with flies reared at 24 °C. Further, we used the InrE19/TM3 siblings reared under identical conditions as controls. InrE19/TM3 flies show only a moderate reduction in Inr activity and have a slight reduction in body size relative to wild-type [4]. Because temperature affects overall body-size in wild-type flies [19], it is necessary to distinguish between changes in InrE19/InrGC25 phenotype that are a consequence of changes in the level of Inr expression from those resulting from changes in rearing temperature. To do this we report the phenotype of InrE19/InrGC25 flies as a percentage of the phenotype of InrE19/TM3 control flies that have undergone temperature shifts at the same developmental time.
The Changing Role of Insulin Signaling during Development
We used the temperature-sensitive Inr mutants to investigate the role of the insulin pathway during Drosophila development. We reduced Inr activity during different periods of development by transferring InrE19/InrGC25 flies from a permissive 17 °C to a restrictive 24 °C at different points in development. After the switch, the flies were left to complete development at 24 °C. We were able to identify temperature-sensitive periods (TSPs) during which increasingly earlier switches from 17 °C to 24 °C resulted in increasingly abnormal phenotypes.
Total developmental time is sensitive to Inr activity only before the middle of the third larval instar (Figure 3A). Switching InrE19/InrGC25 flies from 17 °C to 24 °C changes the time to adult eclosion during the first 9 d of development; the earlier the switch, the greater the delay in eclosion. After the ninth day of development at 17 °C, when the flies are approximately 40% through the third instar (Figure 3B), a shift to the restrictive temperature does not delay adult eclosion. At 17 °C, InrE19/InrGC25 flies show a slight delay in eclosion relative to InrE19/TM3 flies, suggesting that Inr activity is still a little impaired at this temperature. The delay in InrE19/InrGC25 flies reared at both 17 °C and 24 °C relative to the InrE19/TM3 controls occurs predominantly through extension of the third instar (Figure 4).
Figure 3 Suppression of Inr Expression in InrE19/InrGC25 Flies Affects Developmental Time and Adult Size
(A) Developmental time and adult wing size of InrE19/InrGC25 females switched from 17 °C to 24 °C increasingly late in development, expressed as percentage of developmental time and adult wing size of InrE19/TM3 females maintained under identical thermal conditions. Temperature-control flies were maintained at 17 °C throughout development. TSPs of female InrE19/InrGC25 for wing area and delayed eclosion can be seen as regions of the chart where switching from 17 °C to 24 °C increasingly early in development results in increasingly abnormal phenotypes (that is, where the gradient of the relationship between switch day and phenotype is non-zero). For delayed adult eclosion, the TSP of female InrE19/InrGC25 is before the ninth day of development at 17 °C. For reduced wing size, the TSP of female InrE19/InrGC25 is between the ninth and approximately the 20th day of development at 17 °C.
(B) The stages of development of InrE19/InrGC25 flies at 17 °C (A, adult; E, embryo; L1, first instar; L2, second instar; L3, third instar; P, pupae). The point at which suppression of the insulin pathway changes from delaying adult development to reducing adult wing size occurs approximately 40% into the third instar (vertical gray bar)
(C) Dry mass of InrE19/InrGC25 males switched from 17 °C to 24 °C at different points in development, expressed as percentage of dry mass of InrInrE19/TM3 males maintained under identical thermal conditions. The TSP of male InrE19/InrGC25 for reduced adult mass is after the ninth day of development but before pupariation.
(D) Proportion of 17 °C InrE19/InrGC25 larvae pupariating when completely starved at different points in development. The point at which 50% of larvae pupariate in the absence of food marks the critical size. The critical size is reached approximately 40% through the third instar and coincides with the end of the TSP for delayed eclosion and the beginning of the TSP for reduced wing size and adult dry mass (vertical gray bar). All pupariating larvae successfully completed metamorphosis and eclosed as adults.
Figure 4 Developmental Delay in 24 °C InrE19/InrGC25 Flies Occurs primarily through Elongation of the Third Larval Instar
Area shows percentage of individuals (n = 20) in each developmental stage at different times in development. Time is shown in DDs to control for the effect of temperature on developmental rate.
In contrast, adult wing area is sensitive to Inr activity only during late third instar and early pupation (see Figure 3A). Switching InrE19/InrGC25 flies from 17 °C to 24 °C changes adult wing size between day 9 and 20; the earlier the shift, the smaller the wings The precise TSP of female InrE19/InrGC25 flies for wing area may be smaller than implied by Figure 3A because the data are based on population rather than individual measures. The wings may be insensitive to changes in Inr activity as early as day 17. Before day 9, however, shifting the flies to the restrictive temperature earlier in development has no additional effect on adult wing size. At 17 °C, the wings of InrE19/InrGC25 flies are slightly smaller than in InrE19/TM3 control flies, again suggesting that Inr activity is marginally reduced at this temperature.
We tested whether total body mass was sensitive to Inr activity over the same period as wing size. We compared the dry mass of adult InrE19/InrGC25 males reared under several thermal conditions: 24 °C, 17 °C, and a series of samples switched from 17 °C to 24 °C on days 9 through 16 (pupariation is at approximately day 13 at 17 °C). Adult body mass is sensitive to reduction in Inr activity between day 9 and day 13 at 17 °C (see Figure 3C). Shifting InrE19/InrGC25 flies to the restrictive temperature after pupariation, and therefore after larvae have stopped feeding, has no influence on adult mass. Shifting the flies earlier than day 9 has no additional effect on adult body mass.
Inr activity therefore influences total development time, adult wing size, and adult mass for different periods of development. We tested whether the time at which down-regulation of the insulin pathway switches from delaying development to reducing the size of the resulting flies coincides with attainment of critical size. In practice, the critical size is determined as the size at which 50% of larvae proceed to pupariation in the absence of food [8]. We measured the timing of critical size in InrE19/InrGC25 flies reared at 17 °C under our experimental conditions and found that it is attained at approximately day 9 (see Figure 3D), coinciding with the time at which Inr activity switches from affecting developmental time to body size. Therefore, larval critical size coincides with the shift in Inr function.
A Reduction in Inr Activity Changes Body Chemistry
To better understand why reduction of Inr activity after pupariation influenced adult wing size but not overall body mass, we measured the protein, lipid, glycogen, and sugar content of InrE19/InrGC25 flies reared at 17 °C up to pupariation, then either switched to 24 °C or maintained at 17 °C.
Reducing Inr activity after pupariation results in an approximate doubling of free-sugar concentration in InrE19/InrGC25 flies (Table 1). There is no similar response in protein, glycogen, or lipid levels. Lipid levels are, however, elevated in both InrE19/InrGC25 flies maintained at 17 °C and those switched to 24 °C at pupariation relative to InrE19/TM3 flies (Table 1). InrE19/InrGC25 flies grown at a restrictive 24 °C for their entire development also have elevated lipid levels (138% ± 13.3 of InrE19/TM3 flies). This suggests that the slight deficiency in insulin signaling at 17 °C may account for the entire effect on lipid levels.
Table 1 Protein, Sugar, Glycogen, and Lipid Content of InrE19/InrGC25 and InrE19/TM3 Flies Reared at 17 °C until Pupariation and then Switched to 24 °C or Maintained at 17 °C
A Reduction in Inr Activity Affects Different Organs Differently
Different organs typically grow at different rates in developing animals, a phenomenon called “allometry.” Patterns of allometry in populations can also result from organs growing at the same rate but starting and stopping growth at different times in development [20]. We investigated whether the insulin-signaling pathway might be involved in the regulation of allometry by examining whether different organs respond differently to reductions in Inr activity induced by a temperature shift of InrE19/InrGC25 flies from 17 °C to 24 °C. We examined organs located at the anterior, median, and posterior of males: the maxillary palp, wing, and genital arch. We chose the maxillary palp rather than another anterior organ because it is similar in size to the genital arch, allowing us to control for absolute size in this comparison.
Figure 5 shows the relative areas of wings, genital arch, and maxillary palps of male InrE19/InrGC25 and male InrE19/TM3 flies reared at either 17 °C or 24 °C. At 17 °C, all three organs are smaller in InrE19/InrGC25 males than InrE19/TM3 males. At 24 °C, the wing and maxillary palps are further reduced in InrE19/InrGC25 males, consistent with a further reduction in the level of Inr activity. In contrast, there is no difference in the size of the genital arches, relative to InrE19/TM3 males, in flies reared at 17 °C and 24 °C.
Figure 5 Different Organs Respond Differently to Suppression of Inr Activity
Bars show organ area in InrE19/InrGC25 males as a percentage of area in InrE19/TM3 males, to control for temperature effects. Bars with different letters indicate organs that differ: A, B, and C are significantly different at α = 0.05 (Tukey-Kramer pairwise comparison). Mean areas of all organs given in Table S1. s.e., standard error.
This suggests that the size of the genital arches may not be regulated by Inr activity. However, the arches are smaller in InrE19/InrGC25 males than in InrE19/TM3 males at both rearing temperatures. It is possible that this size difference is a consequence of genetic background, unrelated to differences in Inr activity in the two genotypes. Alternatively, the genital arches may show a limited response to changes in Inr activity. They may be sensitive to a mild reduction in Inr activity experienced by InrE19/InrGC25 flies reared at 17 °C, but be insensitive to a further reduction experienced by InrE19/InrGC25 flies reared at 24 °C.
To distinguish between these two hypotheses, we generated large Minute clones homozygous for chico1. Homozygous chico1 clones produce phenotypes identical to mutant Inr clones [13] and autonomously cause a dramatic size reduction in fly wings and eyes, whereas the heterozygous chico1 cells behave as wild-type [5]. We identified males that were homozygous for chico1 throughout one side of the genitals and heterozygous for chico1 on the other. If genital size is independent of the insulin-signaling pathway, then the genitals from either compartment should be identical in size. Each comparison was made within a single male, automatically controlling for total body size. We found that genital arches consisting of mutant chico1 clones were 16% smaller than paired genital arches on the same male (genital arch area: chico1 mutant = 2,840 ± 60 μm2, chico1 wild-type = 3,230 ± 110 μm2, paired-sample t test: n = 6, t = 1.67, p = 0.0214). This reduction of 16% is consistent with the 16% reduction observed in InrE19/InrGC25 males relative to controls (Figure 5). In contrast, maxillary palps consisting of mutant chico1 clones were 45% smaller than paired palps on the same male (maxillary palp area: chico1 mutant = 4,820 ± 100 μm2, chico wild-type = 8,640 ± 170 μm2, n = 7). The genital arches do, therefore, show a limited response to changes in insulin signaling. They are sensitive to a mild reduction in Inr activity, such as observed in InrE19/InrGC25 flies reared at 17 °C. Further reduction in Inr activity, such as observed in InrE19/InrGC25 flies reared at 24 °C, has no additional effect on genital arch size but does have an effect on maxillary palp and wing size.
A Reduction in Inr Activity Affects Cell Size and Cell Number Independently
To investigate the cellular basis for the effect of reduction of Inr activity on size, we compared the size and number of epidermal wing cells in Inr
E19
/Inr
GC25 flies with InrE19/TM3 flies, reared at 17 °C and 24 °C. The wing-size difference between InrE19/InrGC25 and InrE19/TM3 flies reared at 17 °C is due to smaller cells but not fewer cells in the wings of InrE19/InrGC25 males (compare relative sizes at 17 °C, Figure 6, Table S2). In InrE19/InrGC25 females reared at 17 °C, cell size is also smaller than in InrE19/TM3females, but there are also slightly fewer cells. In both males and females, the additional difference in wing-size observed in InrE19/InrGC25 relative to InrE19/TM3 flies reared at 24 °C is due to fewer, but not smaller, cells: Cell size in InrE19/InrGC25 wings relative to InrE19/TM3 wings is the same at both 17 °C and 24 °C (Figure 6, Table S2). Reduction in wing area through suppression of the insulin pathway in InrE19/InrGC25 flies is therefore predominantly a consequence of smaller cells at 17 °C but fewer cells at 24 °C.
Figure 6 A Reduction in Inr Activity Affects Cell Size and Cell Number Independently
(A) At 17 °C, the difference in wing area between InrE19/InrGC25 and InrE19/TM3 flies is due to a difference in cell size, whereas at 24 °C the difference is due to an additional difference in cell number. Bars show wing area, cell area, and cell number in Inr
E19/Inr
GC25 flies as a percentage of area or number in InrE19/TM3 flies.
(B) At 17 °C the reduced Inr activity in InrE19/InrGC25 flies reduces cell area to approximately 85% the area in InrE19/TM3 flies, whereas at 24 °C there is no further reduction in cell area, but there is a reduction in cell number to approximately 75% of the number in InrE19/TM3 flies. Mean wing and cell area, and cell number are given in Table S2.
Discussion
Suppressing the insulin-signaling pathway extends developmental time and reduces final adult size in Drosophila. By varying the activity of Inr we have demonstrated that these effects depend on when in development the suppression occurs; Inr suppression affects total developmental time early in development, and body and organ size late in development. The transition from affecting developmental time to affecting body and organ size occurs when the fly reaches the critical size, the point at which development can be completed in the absence of food. The effect of reduced Inr activity on size varies from organ to organ, implicating the insulin pathway in the control of allometry in Drosophila. In addition, varying the magnitude of Inr activity has different effects on cell size and cell number.
Insulin Signaling, Critical Size, and Developmental Delay
We found that a reduction in Inr activity after critical size does not delay development (see Figure 3A). Therefore reducing Inr activity affects total developmental time by delaying the time at which larvae reach critical size. Critical size has been identified as the key stage in insect maturation that determines the point at which, in holometabolous insects, a larva becomes committed to metamorphosis [21]. In Drosophila it is also the minimal viable weight necessary to survive pupation. Consequently, because larvae can complete development without any additional feeding after reaching its critical size, the critical size sets the lower limit of final adult size.
How insects measure critical size is largely unknown. Our results indicate that a reduction in Inr activity delays the point at which Drosophila larvae reach critical size. We hypothesize that critical size measurement involves a specific organ or organs, and that it is the slow growth of this organ or organs that delays development in Inr mutants. Possible candidates include the imaginal discs and the fat body. Regeneration of damaged imaginal discs delays pupariation in Drosophila [22,23], indicating that some or all of the imaginal discs need to grow to a particular size before a larva can pupariate. Complete removal of the discs does not, however, delay pupariation [23]. Another organ must therefore measure critical size and initiate pupariation, with the immature imaginal discs inhibiting pupariation. Slow growth of either the imaginal discs or a “critical-size organ” could delay development in Inr mutants. Recent studies suggest this “critical-size organ” could be the fat body, which functions as a nutrient sensor and is involved in the coordination of organismal growth [24]. Suppressing Inr/PI3K signaling in the fat body alone is sufficient to inhibit larval growth and mimics the effects of starvation [18].
Insulin Signaling and Final Body and Organ Size
A reduction in Inr activity after critical size reduces final adult size and organ size (see Figure 3A and 3C). Inr activity influences adult body size and wing size for different periods of development. As expected, adult body size is influenced by Inr activity only between critical size and pupariation, after which the larva does not feed and becomes a “closed system,” neither gaining nor losing mass. However, insulin signaling continues to influence the final size of adult organs well into the pupal stage. InrE19/InrGC25 flies switched from 17 °C to 24 °C at pupariation have the same mass as InrE19/InrGC25 flies maintained at 17 °C, but have reduced wings. At the same time, the temperature-shifted flies have a much higher free-sugar concentration as adults. These two findings appear to be linked. Considerable cell proliferation in the imaginal discs occurs after the larva has stopped feeding [11,12,25], and this proliferation relies entirely on stored nutrients as an energy source. Nutrient storage occurs predominantly in the fat body cells, which accumulate reserves of proteins, lipids, and glycogen (the major carbohydrate storage compound) during the third larval instar [18]. Both starvation and suppression of the insulin pathway cause these nutrients to be mobilized for use by growing cells [18]. The finding that InrE19/InrGC25 mutants reared at 24 °C have elevated free-sugar levels, but do not have elevated glycogen levels, indicates that they are able to mobilize their carbohydrate reserves, but that the free-sugars are not taken up by growing organs, and remain in the haemolymph.
Insulin Signaling and Cell Size and Cell Number
Any mechanism that influences organ size does so by changing cell size, cell number, or both. Although mutations of Inr, PI3K, and chico reduce both cell size and cell number [4,5,13,26], downstream components of the insulin-signaling pathway can affect cell size and cell number independently. The RPS6-p70-protein kinase (S6K) branch of the pathway appears to influence cell size but not cell number [27], whereas the dFOXO branch appears to influence cell number but not cell size [15,16]. Because these signaling branches diverge downstream of the insulin receptor, it has not been clear how insulin signaling could affect cell size and number differentially. These changes are ultimately a consequence of changes in the relative rates of cell growth and division [28]. For example, the reduction in cell size, but not cell number, in S6K mutants implies a reduction in the rate of cell growth but not of cell division. Conversely, the reduction in cell number, but not cell size, when dFOXO is over-expressed implies that the rates of cell growth and division are reduced equally. Finally, a reduction in both cell size and cell number will result if both the rates of cell growth and division are reduced, if the former is reduced to a greater extent than the latter.
Our results support the hypothesis that insulin signaling differentially affects cell size and cell number via different levels of insulin receptor activity. Slightly reduced levels of activity, as in InrE19/InrGC25 flies raised at 17 °C, reduce cell size, possibly through a reduction in the rate of cell growth but not of cell division. Further reductions in the levels of activity, as in InrE19/InrGC25 flies raised at 24 °C, reduces cell number only, possibly through a subsequent balanced reduction in both the rate of cell growth and cell division. These data are consistent with a result from Bohni et al. [5]. They showed that the wings of chico2 homozygotes are smaller due to a reduction of both cell size (17%) and cell number (27%). However, a further reduction in insulin signaling through the removal of a single copy of Inr enhances the small-size phenotype exclusively through a reduction in cell number but not cell size. Although we cannot exclude the possibility that the differences in cell size between InrE19/InrGC25 and InrE19/TM3 flies are a consequence of genetic background unrelated to differences in Inr activity, the reduction in cell number alone is clearly due to reduction in insulin signaling (see Figure 5).
Insulin Signaling and Allometry
Variation in insulin signaling appears to affect the allometric relationship between organs. For example, at 17 °C, male InrE19/InrGC25 flies have body size, wings, maxillary palps, and genital arches approximately 85% of wild-type. Increasing the rearing temperature of InrE19/InrGC25 flies from 17 °C to 24 °C, however, causes a further reduction in wing, maxillary palp, and overall body size but does not affect the size of the genital arches (Figure 5). Consequently, flies reared at 24 °C have larger genitals relative to their body and wings compared to flies reared at 17 °C. The apparently restricted response of the genitals to changes in insulin signaling is not a consequence of the particular alleles used in this study: chico-mutant clones also have much less of an effect on size when they are in the genital arches than when they are in the maxillary palps.
The mechanism by which different organs respond differently to changes in insulin signaling is unclear. Organs may vary in their expression of the insulin receptor gene or may limit the activity of certain downstream components of the insulin signaling pathway. In 17 °C InrE19/InrGC25 males, the genital arches, wings, and maxillary palps are all reduced by approximately the same amount relative to 17 °C InrE19/TM3males. In the wing, this reduction in area is a consequence of a reduction in cell size. A further decrease in Inr activity (through an increase in rearing temperature to 24 °C) reduces wing area through a reduction in cell number alone. If the genital arches are like the wing, then their response to insulin signaling may be restricted to changes in cell size and not cell number. The cells of the genital arches may therefore be deficient in components of the insulin-signaling pathway that regulate cell number but not components that regulate cell size.
A Model of the Insulin-Signaling Regulation of Growth and Development
The insulin-signaling pathway appears to play a different role after critical size than before. Similarly, a temperature-sensitive lethal mutation, l(1)ts-1126, which reduces the rate of cell proliferation in Drosophila, delays pupariation when larvae are moved to a restrictive temperature before the third instar, but reduces adult size when larvae are moved to a restrictive temperature late in the third instar [29]. The two effects of reduced Inr activity may therefore result from the same process: a reduction in the rate of cell growth and proliferation. We have developed a model of the insulin-signaling regulation of growth and development in Drosophila (Figure 7). (A similar model has recently been developed by Davidowitz and Nijhout [30] to explain variation in body size in response to temperature in the tobacco hornworm, Manduca sexta.) A reduction in Inr activity prior to critical size slows cell growth and proliferation and delays the time at which the larvae reaches critical size. Critical size is not substantially influenced by nutritional conditions [8] or insulin signaling in Drosophila, although this does not seem to be the case for all insects, for example, M. sexta [21]. Once critical size is reached, the time to pupariation and adult eclosion is fixed, as are the remaining periods of growth prior to adult differentiation of individual imaginal discs. The duration of these intervals are uninfluenced by nutritional conditions or insulin signaling. A reduction in Inr activity during these periods also slows cell growth and proliferation, but now reduces the amount of growth attained before differentiation, resulting in smaller organs and a smaller fly. Because different structures grow for different periods, they are sensitive to Inr activity at different times. For example, wing size begins to be insensitive to changes in Inr activity at approximately the same time as cell proliferation ceases, around 25% into the pupal stage. Adult body mass becomes insensitive to changes in Inr activity just before pupariation, when the larvae stops feeding and final body size is fixed.
Figure 7 A Model of the Insulin-Signaling Regulation of Growth and Development
(A) Under normal conditions, imaginal discs grow to a critical size, which initiates an increase in the ecdysteroid titer. When ecdysteroid levels rise above a maximum threshold, the discs cease cell proliferation and undergo differentiation, fixing their final size. A, adult; E, embryo; L1–L3, first to third larval instar; P, pupa.
(B) In Inr mutants, growth of imaginal discs to critical size is slowed, retarding development. When critical size is reached, the ecdysteroid titer again increases, rising above the maximum threshold for cell proliferation in the imaginal discs. Temporal changes in the ecdysteroid titer are unaffected by insulin signaling. Because the rate of cell proliferation is slowed, the imaginal discs are smaller when they begin to differentiate, reducing final organ size. Different discs have different thresholds of sensitivity to ecdysteroid and so cease cell proliferation at different times. Hormones other than ecdysteroids may also be involved.
A key component of this model is that after critical size, the remaining periods of growth of individual imaginal discs and of the body as a whole are fixed and uninfluenced by insulin signaling. Our data show that the length of the period between critical size and pupariation, the remaining period of growth of the body as a whole, is not substantially affected by a reduction in Inr activity. In Drosophila and other insects, this interval is controlled by endocrine events. In Drosophila, a small peak in the ecdysteroid titer coincides with attainment of critical size [31], followed by a second peak 12 h later, just before the larvae leave the food [32]. In M.
sexta this second peak acts directly on the nervous system to initiate wandering behavior [33], and the same is likely true for Drosophila [34]. The period in which Inr activity can influence final body size is therefore terminated by hormones. Importantly, hormones other than insulin also control the period of cell proliferation in the imaginal discs. For example, in M.
sexta, ecdysteroids govern the phases of eye development during metamorphosis [35,36]. When the ecdysteroid titer rises above a minimum threshold just before pupariation, it stimulates a wave of cell proliferation to pass across the eye primordium. This proliferation is sustained until the ecdysteroid titer rises above a maximum level in the middle of pupal development, whereupon cell proliferation stops and maturation of the ommatidia begin. These and similar data in other insects [37–39], including Drosophila [40–43], suggest that cell proliferation in imaginal discs may be temporally regulated by thresholds of sensitivities to fluctuating levels of ecdysteroids and juvenile hormone (JH) [44]. Different organs have different thresholds of sensitivity and hence grow for different periods of time. Insulin signaling may therefore control body and organ size by regulating the amount of growth attained during these periods of cell proliferation.
This model requires that changes in ecdysteroid and JH levels are unaffected by the insulin-signaling pathway. It is known that adult Inr and chico mutant flies have reduced levels of JH and impaired ovarian ecdysone synthesis [45,46]. However, the same hormone fluctuations that putatively control the period of cell proliferation also initiate pupariation [34] and, because the timing of pupariation is unaffected by Inr activity, it seems likely that the temporal dynamics of the hormonal cascade in the larvae are also unaffected by Inr activity. We predict, therefore, that the insulin-signaling pathway regulates cell proliferation in imaginal discs but that the duration of proliferative phases are controlled by other endocrine cues, such as JH and ecdysteroids, that are themselves unaffected by the insulin-signaling pathway.
This model demonstrates how the pleiotropic effects of insulin signaling on developmental time and final body and organ size can be separated, and may be available for independent evolutionary modification. For example, changes in the relative size of an organ may occur by organ-specific modifications in its growth response to insulin signaling, through organ-specific changes in the expression of Inr, adjustments in Inr activity, or adjustments in the expression and activity of downstream components of the insulin-signaling pathway. Alternatively, changes in the period of an organ's growth, through alterations in its sensitivity to other endocrine cues, may have a similar effect [44]. In metazoans in general, each organ has a unique timetable for cellular events in tissue development. Our model, and the data upon which it is based, indicate that in order to understand the effects of insulin signaling on adult phenotype it is necessary to understand how temporal changes in insulin signaling interact with this timetable.
Materials and Methods
Mutant stocks
InrGC25 is a chromosomal inversion with a breakpoint upstream of the Inr. InrE19 is an uncharacterized mutation induced by ethyl methanosulfonate. Both were described in Chen et al. [4] and obtained from the Bloomington Stock Center. The chico
1 allele is a P-element insertion allele whose phenotype is similar to the null chico
2 allele [5] and was kindly provided by Ernst Hafen. The tub-GFP-PH flies were kindly supplied by Bruce Edgar. The flies were maintained as chico1/CyO [5], InrE19, and InrGC25 [4] balanced over TM6B-Tb, TM3, TM3-pAct-GFP, and tub-GFP-PH; Inr
E19/TM3 [18] at 17 °C on standard yeast cornmeal agar medium.
Immunolocalization
We crossed InrE19/TM6-Tb with InrGC25/TM6-Tb flies and reared them at either 17 °C or 24 °C. When the larvae reached third instar, we genotyped them as either insulin receptor mutants (InrE19
/InrGC25) or wild-type (InrE19 or InrGC25/TM6B-Tb). The fat bodies were dissected out in PBS, fixed in 4% paraformaldehyde for 10 min, and stored in absolute methanol at −20 °C. They were permeabilized with PBT (0.3% Triton X in PBS) for 30 min, washed in BBT (0.3% bovine serum albumin in PBT) for 30 min, then blocked in NGS/BBT (3% normal goat serum in BBT) for 30 min. They were stained with anti-dFOXO 2095 [16] (kindly provided by O. Puig) (1:1,000 in PBT) and fluorescein anti-rabbit (1:500 in PBT) (Vector Labs, Burlingame, California, United States). DNA was stained with propidium iodide (1:1,000 in PBS with 1 μg of RNase A). We mounted the fat bodies in Vectashield (Vector Labs) for observation under a confocal microscope (Perkin Elmer UltraVIEW RS3; PerkinElmer Life and Analytical Sciences, Boston, Massachusetts, United States).
GPH localization
We crossed tub-GFP-PH;InrE19/TM6B-Tb with Inr
GC25/TM6B-Tb flies and reared them at 15 °C. When the larvae reached second instar, they were removed from their food, washed, and genotyped as insulin receptor mutants (tub-GFP-PH;InrE19/InrGC25) or wild-type (tub-GFP-PH;InrE19 or InrGC25/TM6B-Tb). They were then transferred to fresh food and maintained at 15 °C or 25 °C for 24 h. We then dissected the larvae in 100% methanol kept at 15 °C or 25 °C depending on the temperature treatment of the larvae, and stored their fat bodies without additional fixing in 100% methanol at −20 °C. We quickly washed the fat bodies in 50% methanol in PBS, and then 100% PBS before mounting them in Vectashield with Hoechst 33342 (2:1,000) to stain DNA. We observed the fat bodies under a confocal microscope (Zeiss LSM 510; Zeiss, Gena, Germany).
Temperature shift
We reared flies at a number of temperature regimes. They were either maintained at 17 °C, 18 °C, 24 °C, or 25 °C or switched from 17 °C to 24 °C on day 1 to day 27 of development. We set up four bottles for each temperature regime, each containing standard yeast cornmeal agar medium. Two were the product of male InrE19/TM6B-Tb and female InrGC25/TM3 and two were the product of female InrE19/TM6B-Tb and male InrGC25/TM3. Each bottle contained between 100 and 200 eggs laid over a 4-h period. Temperature shifts were performed according to the median age of flies in each bottle. Larvae were reared on standard yeast cornmeal agar medium. Total developmental time was not significantly different between InrE19/InrGC25 flies from the two different crosses, so the data from all bottles maintained under the same temperature regime were pooled (data not shown).
Measurements and data handling
Bottles were inspected daily at 4:00 PM and any flies that had eclosed in the previous 24 h were collected and preserved in 80% ethanol. Flies were then genotyped and the eclosion times to the nearest day of each InrE19/InrGC25 and InrE19/TM3 control fly was recorded. We used InrE19/TM3 rather than InrGC25/TM6B-Tb flies as controls because TM6-Tb has Tubby as a marker, which affects developmental timing and adult shape [47]. Ten to 15 InrE19/InrGC25 and InrE19/TM3 female wings from each temperature regime were dissected and mounted in lactic acid:ethanol (4:5). Ten InrE19/InrGC25 and InrE19/TM3 males reared at 24 °C and 15 InrE19/InrGC25 and InrE19/TM3 males reared at 17 °C were also dissected and their wings, maxillary palps, and genital arches mounted in either lactic acid:ethanol (wings) or Hoyer's Solution (maxillary palps and genital arches). Digital images of the wings, maxillary palps, and genital arches were captured and measured using IPLab 3.9 (Scanylitics, Fairfax, Virginia, United States). Wing cell size was estimated using the number of trichomes in a 0.01 mm2 square between the veins IV and V of the dorsal wing blade. An index of the total number of wing cells was estimated by multiplying the number of trichomes in the area by the total wing size and dividing by 0.01. Dry masses of individual male flies were measured with an analytical balance (Mettler Toledo AX26 DeltaRange; Mettler-Toledo, Columbus, Ohio, United States).
The total developmental time for each fly was converted into degree-days (DDs) to control for the effects of temperature on growth rates. DD is a measure of the heat accumulation above a minimal temperature (To), and is calculated as:
where T
r is the experimental temperature, and n
r is the number of days maintained at T
r, summed across all experimental temperatures j to k. In this case, we used only two experimental temperatures, 17 °C and 24 °C. We calculated the value of T
0 such that the developmental time of control InrE19/TM3 flies was constant, irrespective of temperature regime. To do this we regressed total developmental time in DDs of InrE19/TM3 flies against the time of their temperature shift from 17 °C to 24 °C, using different values of T
0. The value of T
0 that minimized the slope of this regression line was found to be 9.978 °C. When T
0 is 9.978 °C, the average developmental time of InrE19/TM3 flies is 156.04 ± 0.338 DD. We then converted the total developmental time for each InrE19/InrGC25 fly into DDs using T
0 = 9.978, and expressed it as the average percentage of developmental time of InrE19/TM3 flies (156.04 DD), along with the standard error of the percentages.
Rearing temperature also affects body, organ, and cell size and cell number [19]. Consequently, all these measurements in InrE19/InrGC25 flies were expressed as a percentage of the value in InrE19/TM3 flies maintained under the same temperature regime. Wing area of InrE19/InrGC25 females and dry mass of InrE19/InrGC25 males switched from 17 °C to 24 °C increasingly late in development, was expressed as the percentage of wing area and dry mass of InrE19/TM3 flies switched from 17 °C to 24 °C at the same percent of total development at 17 °C. See Figure S1 for details.
Clonal analysis
We crossed males carrying the alleles chico1/CyO with females carrying the alleles f36a;M(2)Z f+37C/CyO . Larvae were subjected to X-rays (1,000 rad) at 24–72 h after egg-laying. Genitals and maxillary palps of male offspring without balancer chromosomes were studied for f36a bristles.
Developmental staging of InrE19/InrGC25 and InrInrE19/TM3 at 17 °C and 24 °C
We crossed InrE19/TM3-pAct-GFP flies with InrGC25/TM3-pAct-GFP flies and collected the eggs over a period of 4 h. After 24 h, we selected InrE19/InrGC25 and InrE19 or InrGC25
/TM3-pAct-GFP eggs using a fluorescence microscope to detect GFP activity. The eggs were transferred to 12 mm ø Petri dishes containing standard yeast cornmeal agar medium. Each dish contained 20 eggs of only one genotype and was maintained at either 17 °C or 24 °C. We prepared three dishes: InrE19/InrGC25 maintained at 17 °C; InrE19/InrGC25 maintained at 24 °C; and InrE19 or InrGC25
/TM3-pAct-GFP maintained at 17 °C. From day 2, ten individuals were randomly selected from each Petri dish and their developmental stage was recorded using mouthpart development, before being returned to the Petri dish.
Body chemistry assays
White prepupae of InrE19/InrGC25 and InrE19/TM3 flies raised at 17 °C were transferred to moistened Kimwipes (Kimberly-Clark, Neenah, Wisconsin, United States) in Petri dishes. Prepupae of each genotype were split between further development at 17 °C and 24 °C. We collected adults within 8 h of eclosion, after their cuticle had hardened, and immediately froze them in liquid nitrogen for later analysis. Wet and dry masses of individual flies were measured with an analytical balance (Mettler Toledo AX26 DeltaRange). All metabolic assays were performed on individual dried males (n = 10 for each assay). Glycogen and sugar content were measured using a protocol of van Handel [48]. Lipid content was quantified using another protocol of van Handel [49]. To determine protein concentration, flies were homogenized in 100 μl 0.1 M Na2HPO4. Ten μl of each sample was combined with the Bio-Rad protein assay dye reagent (Bio-Rad Laboratories, Hercules, California, United States) following the manufacturer's instructions and spectrophotometrically assayed at 595 nm.
Supporting Information
Figure S1 Fitted Values for Wing Aarea of InrE19/TM3 Control Flies
Wing area of InrE19/InrGC25 flies in Figure 2 is expressed as percentage of wing area of InrE19/TM3 flies kept under the same temperature regime, using the fitted values shown on the plot. Temperature affects wing area differently before and after critical size. Consequently, fitted values were calculated by regressing wing area of InrE19/TM3 flies against transfer age before critical size (35% development) and after critical size. A similar method was used to determine the dry mass of InrE19/TM3 males, except a single regression analysis was used to calculate the fitted values.
(59 KB PDF).
Click here for additional data file.
Table S1 Wing, Genital Arch, and Maxillary Palp Area of InrE19/InrGC25 and InrE19/TM3 Males Reared at 17 °C and 24 °C
(37 KB DOC).
Click here for additional data file.
Table S2 Total Area, Cell Area, and Cell Number of the Wings of InrE19/InrGC25 and InrE19/TM3 Flies Reared at 17 °C and 24 °C
(45 KB DOC).
Click here for additional data file.
Accession Numbers
The FlyBase (http://flybase.bio.indiana.edu/search/) accession numbers for the genes and gene products discussed in this paper are Chico (Fbgn0024248), chico1 (FBal0031303), dFOXO (FBgn0038197), Inr (FBgn0013984), InrE19 (FBal0094021), InrGC25 (FBal0010755), PI3K (Fbgn0015279), and S6K (FBgn0015806). The National Center for Biotechnology Information (NCBI) (http://www.ncbi.nlm.nih.gov/) accession number for IGF1 receptor is NM_010513.
We thank Fred Nijhout and three anonymous referees for critical and helpful comments on early versions of this manuscript. We thank Miguel Gaspar for his assistance with imaging. AWS was supported by Princeton University Council on Science and Technology, JD was supported by a Howard Hughes Medical Institute Predoctoral Fellowship. DLS acknowledges the National Institutes of Health, the David and Lucile Packard Foundation, and Princeton University for financial support. LV thanks Michael Akam and his lab for guidance, and St John's College (Cambridge), the Cambridge Overseas Trust, and the Overseas Award Scheme for funding.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. JD, LV, DLS, and AWS conceived, designed, and performed the experiments. AWS analyzed the data and wrote the paper.
Citation: Shingleton AW, Das J, Vinicius L, Stern DL, (2005) The temporal requirements for insulin signaling during development in Drosophila. PLoS Biol 3(9): e289.
Abbreviations
DDdegree day
JHjuvenile hormone
PI3Kphosophoinositide 3-kinase
TSPtemperature sensitive period
==== Refs
References
Ikeya T Galic M Belawat P Nairz K Hafen E Nutrient-dependent expression of insulin-like peptides from neuroendocrine cells in the CNS contributes to growth regulation in Drosophila
Curr Biol 2002 12 1293 1300 12176357
Shambaugh GE Radosevich JA Glick RP Gu DS Metzger BE Insulin-like growth factors and binding proteins in the fetal rat: Alterations during maternal starvation and effects in fetal brain cell culture Neurochem Res 1993 18 695 703 8510796
Baker J Liu JP Robertson EJ Efstratiadis A Role of insulin-like growth factors in embryonic and postnatal growth Cell 1993 75 73 82 8402902
Chen C Jack J Garofalo RS The Drosophila insulin receptor is required for normal growth Endocrinology 1996 137 846 856 8603594
Bohni R Riesgo-Escovar J Oldham S Brogiolo W Stocker H Autonomous control of cell and organ size by CHICO, a Drosophila homolog of vertebrate IRS1–4 Cell 1999 97 865 875 10399915
Liu JP Baker J Perkins AS Robertson EJ Efstratiadis A Mice carrying null mutations of the genes encoding insulin-like growth factor I (Igf-1) and type 1 IGF receptor (Igf1r) Cell 1993 75 59 72 8402901
Beadle G Tatum E Clancy C Food level in relation to rate of development and eye pigmentation in Drosophila melanogaster
Biol Bull 1938 75 447 462
De Moed GH Kruitwagen C De Jong G Scharloo W Critical weight for the induction of pupariation in Drosophila melanogaster : Genetic and environmental variation J Evol Biol 1999 12 852 858
Bakker K An analysis of factors which determine success in competition for food among larvae of Drosophila melanogaster
Arch Neerl Zool 1961 14 200 281
Robertson FW Ecological genetics of growth in Drosophila .6. Genetic correlation between duration of larval period and body size in relation to larval diet Genet Res 1963 4 74 96
Garcia-Bellido A Merriam JR Parameters of the wing imaginal disc development of Drosophila melanogaster
Dev Biol 1971 24 61 87 5001010
Postlethwait JH Schneiderman HA A clonal analysis of development in Drosophila melanogaster : morphogenesis, determination, and growth in the wild-type antenna Dev Biol 1971 24 477 519 5578888
Brogiolo W Stocker H Ikeya T Rintelen F Fernandez R An evolutionarily conserved function of the Drosophila insulin receptor and insulin-like peptides in growth control Curr Biol 2001 11 213 221 11250149
Kramer JM Davidge JT Lockyer JM Staveley BE Expression of Drosophila FOXO regulates growth and can phenocopy starvation BMC Dev Biol 2003 3 5 12844367
Junger MA Rintelen F Stocker H Wasserman JD Vegh M The Drosophila Forkhead transcription factor FOXO mediates the reduction in cell number associated with reduced insulin signaling J Biol 2003 2 20 12908874
Puig O Marr MT Ruhf ML Tjian R Control of cell number by Drosophila FOXO: Downstream and feedback regulation of the insulin receptor pathway Genes Dev 2003 17 2006 2020 12893776
Dillin A Crawford DK Kenyon C Timing requirements for insulin/IGF-1 signaling in C. elegans
Science 2002 298 830 834 12399591
Britton JS Lockwood WK Li L Cohen SM Edgar BA
Drosophila' s insulin/PI3-kinase pathway coordinates cellular metabolism with nutritional conditions Dev Cell 2002 2 239 249 11832249
French V Feast M Partridge L Body size and cell size in Drosophila : The developmental response to temperature J Insect Physiol 1998 44 1081 1089 12770407
Stern DL Emlen DJ The developmental basis for allometry in insects Development 1999 126 1091 1101 10021329
Davidowitz G D'Amico LJ Nijhout HF Critical weight in the development of insect body size Evol Dev 2003 5 188 197 12622736
Ursprung H Hadorn E [Further research on model growth in combination with partly dissociated wing imaginal disks of Drosophila melanogaster
Dev Biol 1962 4 40 66 13923941
Simpson P Berreur P Berreur-Bonnenfant J The initiation of pupariation in Drosophila : dependence on growth of the imaginal discs J Embryol Exp Morph 1980 57 155 165 7430927
Colombani J Raisin S Pantalacci S Radimerski T Montagne J A nutrient sensor mechanism controls Drosophila growth Cell 2003 114 739 749 14505573
Garcia-Bellido A Merriam JR Clonal parameters of tergite development in Drosophila Dev Biol 1971 26 264 276 5158535
Weinkove D Neufeld TP Twardzik T Waterfield MD Leevers SJ Regulation of imaginal disc cell size, cell number and organ size by Drosophila class I(A) phosphoinositide 3-kinase and its adaptor Curr Biol 1999 9 1019 1029 10508611
Montagne J Stewart MJ Stocker H Hafen E Kozma SC
Drosophila S6 kinase: A regulator of cell size Science 1999 285 2126 2129 10497130
Neufeld TP Shrinkage control: Regulation of insulin-mediated growth by FOXO transcription factors J Biol 2003 2 18 12974982
Simpson P Schneiderman HA A temperature sensitive mutation that reduces mitotic rate in Drosophila melanogaster
Rouxs Arch Dev Biol 1976 179 215 236
Davidowitz G Nijhout HF The physiological basis of reaction norms: The interaction among growth rate, the duration of growth and body size Integr Comp Bio 2004 44 443 449 21676730
Berreur P Porcheron P Berreur-Bonnenfant J Simpson P Ecdysteroid levels and pupariation in Drosophila melanogaster
J Exp Zool 1979 210 347 352
Schwartz MB Imberski RB Kelly TJ Analysis of metamorphosis in Drosophila melanogaster : Characterization of giant, an ecdysteroid-deficient mutant Dev Biol 1984 103 85 95 6425099
Dominick OS Truman JW The physiology of wandering behaviour in Manduca sexta II. The endocrine control of wandering behaviour J Exp Biol 1985 117 45 68 4067505
Riddiford LM Bate M Martinez Arias A Hormones and Drosophila development The development of Drosophila melangogaster 1993 New York Cold Spring Harbor Laboratory Press 899 939
Champlin DT Truman JW Ecdysteroids govern two phases of eye development during metamorphosis of the moth, Manduca sexta
Development 1998 125 2009 2018 9570766
Champlin DT Truman JW Ecdysteroid control of cell proliferation during optic lobe neurogenesis in the moth Manduca sexta
Development 1998 125 269 277 9486800
Berreur P Bougues R Effects of ecdysone on in vivo growth of wing disks of Calliphora erythrocephala
J Insect Physiol 1975 21 915 919
Oberlander H Leach CE Shaaya E Juvenile hormone and juvenile hormone mimics inhibit proliferation in a lepidopteran imaginal disc cell line J Insect Physiol 2000 46 259 265 12770230
Kawasaki H Kawasaki H (1995) Ecdysteroid concentration inducing cell proliferation brings about the imaginal differentiation in the wing disc of Bombyx mori in vitro Dev Growth Differ 1995 37 575 580
Peel DJ Milner MJ The response of Drosophila imaginal disk cell-lines to ecdysteroids Rouxs Arch Dev Biol 1992 202 23 35
Currie DA Milner MJ Evans CW The growth and differentiation in vitro of leg and wing imaginal disc cells from Drosophila melanogaster
Development 1988 102 805 814
Chihara CJ Fristrom JW Petri WH King DS The assay of ecdysones and juvenile hormones on Drosophila imaginal disks in vitro J Insect Physio 1972 18 1115 1123
Cullen CF Milner MJ Parameters of growth in primary cultures and cell lines established from Drosophila imaginal discs Tissue Cell 1991 23 29 39 1905427
Emlen DJ Allen CE Genotype to phenotype: Physiological control of trait size and scaling in insects Integr Comp Biol 2003 43 617 634 21680471
Tu MP Yin CM Tatar M Impaired ovarian ecdysone synthesis of Drosophila melanogaster insulin receptor mutants Aging Cell 2002 1 158 160 12882346
Tatar M Kopelman A Epstein D Tu MP Yin CM A mutant Drosophila insulin receptor homolog that extends life-span and impairs neuroendocrine function Science 2001 292 107 110 11292875
Craymer L [New mutants report] Drosoph Inf Serv 1980 55 197 200
Van Handel E Rapid determination of glycogen and sugars in mosquitoes J Am Mosq Control Assoc 1985 1 299 301 2906671
Van Handel E Rapid determination of total lipids in mosquitoes J Am Mosq Control Assoc 1985 1 302 304 2906672
|
16086608
|
PMC1184592
|
CC BY
|
2021-01-05 08:21:25
|
no
|
PLoS Biol. 2005 Sep 16; 3(9):e289
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030289
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1608660810.1371/journal.pbio.0030289Research ArticleDevelopmentEvolutionPhysiologyNutritionDrosophilaThe Temporal Requirements for Insulin Signaling During Development in Drosophila
Temporal Requirements for InsulinShingleton Alexander W [email protected]
1
Das Jayatri
1
Vinicius Lucio
2
Stern David L
1
1Ecology and Evolutionary Biology, Princeton University, Princeton, New Jersey, United States of America,2Leverhulme Centre for Human Evolutionary Studies, University of Cambridge, Cambridge, United KingdomJohnston Laura Academic EditorColumbia UniversityUnited States of America9 2005 16 8 2005 16 8 2005 3 9 e2891 10 2004 17 6 2005 Copyright: © 2005 Shingleton 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.
For Insulin Signaling Pathways in Flies, Size Matters
Recent studies have indicated that the insulin-signaling pathway controls body and organ size in Drosophila, and most metazoans, by signaling nutritional conditions to the growing organs. The temporal requirements for insulin signaling during development are, however, unknown. Using a temperature-sensitive insulin receptor (Inr) mutation in Drosophila, we show that the developmental requirements for Inr activity are organ specific and vary in time. Early in development, before larvae reach the “critical size” (the size at which they commit to metamorphosis and can complete development without further feeding), Inr activity influences total development time but not final body and organ size. After critical size, Inr activity no longer affects total development time but does influence final body and organ size. Final body size is affected by Inr activity from critical size until pupariation, whereas final organ size is sensitive to Inr activity from critical size until early pupal development. In addition, different organs show different sensitivities to changes in Inr activity for different periods of development, implicating the insulin pathway in the control of organ allometry. The reduction in Inr activity is accompanied by a two-fold increase in free-sugar levels, similar to the effect of reduced insulin signaling in mammals. Finally, we find that varying the magnitude of Inr activity has different effects on cell size and cell number in the fly wing, providing a potential linkage between the mode of action of insulin signaling and the distinct downstream controls of cell size and number. We present a model that incorporates the effects of the insulin-signaling pathway into the Drosophila life cycle. We hypothesize that the insulin-signaling pathway controls such diverse effects as total developmental time, total body size and organ size through its effects on the rate of cell growth, and proliferation in different organs.
Studies using a temperature-sensitive insulin receptor elucidate the temporal requirements for insulin signaling in Drosophila; insulin signaling at different times during development affects many characters, such as total developmental time, total body size and organ size.
==== Body
Introduction
Development in multicellular animals is a process that involves both tight control and flexibility in the regulation of cell size and cell number. Tight control is necessary to produce an animal in which each organ is of an appropriate size relative to the whole body. Flexibility is necessary to produce an animal in which the whole body is of an appropriate size relative to environmental conditions. One key environmental factor shown to influence size is nutrition. In animals as diverse as humans and flies, malnutrition delays development and reduces adult body and organ size. Recent studies indicate that the insulin-signaling pathway (Gene Ontology ID GO:0008286) coordinates growth with nutritional condition in most metazoans, and is remarkably conserved. In Drosophila, expression of insulin-like peptides is nutritionally regulated [1], as is expression of IGF1 (insulin-like growth factor 1) in developing rats [2]. Flies carrying mutations of the insulin receptor (Inr) and mice carrying mutations of the IGF1 receptor show delayed development and growth deficiency with a reduction in body and organ size [3–6]. These and similar “knock out” experiments have demonstrated the gross effects of the insulin-signaling pathway on adult phenotype. Nevertheless, little is known of how the pathway acts during development to affect changes in the adult. In mice, such elucidation is hampered by the inaccessibility of the developing fetus enclosed in the uterus. In flies, however, developing larvae can be easily studied and manipulated.
Here we explore the temporal requirement for insulin signaling in developing Drosophila
melanogaster. In Drosophila, development proceeds through three larval instars to pupariation, pupation, and finally adult eclosion. Feeding is restricted to the first (L1), second (L2), and most of the third (L3) instar. Early in the third instar, the larvae reach a “critical size” at which point they have acquired sufficient nutrients to complete development in the absence of food [7–10]. After critical size is attained, larvae stop feeding, leave the food, and search for a pupariation site. There is, however, a delay between the time critical size is reached and the time larvae stop feeding and reach their maximum body size. Both critical size and the time between reaching critical size and the cessation of feeding are unaffected by nutrition [8,10]. Consequently, nutrition ostensibly affects final adult size through the amount of feeding in the fixed period between critical size and pupariation [10].
Because feeding affects final size only in the period between critical size and pupariation [8,10], one may predict that insulin signaling affects size only during the same period. However, organ growth continues well after pupariation [11,12] and is autonomously modulated by the insulin-signaling pathway [5,13]. Insulin signaling may therefore continue to influence size after pupariation. Details of the temporal requirements for insulin signaling are unknown in Drosophila, but a recent study on the Drosophila forkhead transcription factor (dFOXO) suggests that they vary during development [14]. Unphosphorylated dFOXO negatively regulates growth, but activation of the insulin receptor signaling pathway induces dFOXO phosphorylation, excluding it from the nucleus and inhibiting its activity [15,16]. Constitutive activation of dFOXO in the first and second instars causes developmental arrest, whereas dFOXO activation in the third instar causes a reduction in adult size [14]. Work on Caenorhabditis elegans also indicates that the timing requirements for insulin signaling might vary during the lifecycle. In C. elegans, the pathway acts during early development to regulate diapause, and during adulthood to influence ageing [17]. Resolving the links between larval size, nutrition, and developmental progression on the one hand, and the time-dependence of insulin-pathway gene effects on the other, requires a more precise understanding of the temporal requirements for insulin signaling during development.
We have used a temperature-sensitive mutation of the insulin receptor to investigate the role of the insulin-signaling pathway during different stages of development in Drosophila. By reducing insulin receptor activity at different points in development, we identify the periods of development during which insulin signaling affects adult phenotype. We show that total developmental time is affected by changes in insulin signaling only if those changes occur early in development, before a larva has reached critical size. In contrast, final body and organ size are affected by changes in insulin signaling only if those changes occur late in development, once a larva has passed critical size. Insulin signaling continues to influence organ size, but not body size, after larvae have stopped feeding and throughout much of pupal development. Not all organs respond equally to changes in insulin signaling, however. We find that the genitals show a limited response to suppression of insulin signaling in InrE19/InrGC25 flies, implicating insulin signaling in the control of allometries.
Results
Temperature-Sensitive Suppression of the Insulin Receptor
During a study of the interaction of environmental factors and the insulin-signaling pathway, we discovered that flies trans-heteroallelic for insulin receptor mutations InrE19 and InrGC25 show a temperature-sensitive suppression of the insulin pathway. We initially reared InrE19/InrGC25 flies, wild-type Oregon-R flies, and flies homozygous for chico1, at 18 °C, 25 °C, and 29 °C, and measured the wing areas of the adults. Chico is the fly ortholog of the insulin receptor substrate. chico1 is a null mutation, and homozygotes show a partial loss of insulin receptor function with a more than 50% reduction in body size relative to wild-type and a much reduced viability [5]. Both Oregon-R and chico1 homozygote flies showed a smaller wing area with higher rearing temperature (Figure 1). The response to temperature is the same in both genotypes (the lines relating wing size to temperature are parallel in Figure 1), indicating that temperature influences size independent of the insulin-signaling pathway. However, InrE19/InrGC25 flies have a wing size similar to Oregon-R flies when reared at 18 °C but a wing size similar to chico1 mutants when reared at 25 °C (Figure 1). This suggests that InrE19/InrGC25 flies show temperature-sensitive suppression of Inr activity. Changing the rearing temperature does not, however, simply switch Inr “on” or “off” in InrE19/InrGC25 flies. At 18 °C, InrE19/InrGC25 flies have smaller wings than wild-type flies, suggesting that the insulin receptor is still partially suppressed at this temperature. Additionally, very few InrE19/InrGC25 flies survived to adulthood when raised at 25 °C. and none survived at 29 °C. The difference between InrE19/InrGC25 flies reared at 18 °C and 25 °C appears, therefore, to be a consequence of the degree to which Inr is suppressed.
Figure 1 Temperature Sensitivity in InrE19/InrGC25 Flies
Increasing the rearing temperature of Inr
E19/InrGC25 females from 18 °C to 25 °C causes a reduction in wing area from approximately wild-type (Oregon-R [Ore-R]) to that of an insulin pathway mutant (chico). Wing size is expressed as percentage area of Oregon-R female wing at 25 °C. No InrE19/InrGC25 flies survived rearing at 29 °C. The standard errors are smaller than the markers.
To confirm that these temperature-dependent effects were due to changes in Inr activity, we assessed the activity of the insulin-signaling pathway in InrE19/InrGC25 flies reared at different temperatures. First, we looked at the cellular localization of dFOXO in InrE19/InrGC25 late third instar larvae reared at 17 °C and at 24 °C. Down-regulation of the insulin-signaling pathway dephosphorylates dFOXO and causes it to move from the cytoplasm to the nucleus. Antibody stains with anti-dFOXO show an increase in nuclear versus cytoplasmic localization of dFOXO in InrE19/InrGC25 larvae when reared at 24 °C relative to 17 °C (Figure 2A and 2B). We did not observe change in nuclear versus cytoplasmic staining for InrE19 or InrGC25/TM6B-Tb controls reared at 17 °C versus 25 °C (Figure 2C and 2D). Second, we assessed Inr activity in InrE19/InrGC25 second instar larvae using the tGPH reporter gene [18]. tGPH is under the control of the Drosophila ß-tubulin promotor and produces a GFP (green fluorescent protein) fused to the pleckstrin homology domain of the receptor for phosphoinositides-1. Under the action of phosophoinositide 3-kinase (PI3K), this fusion protein, GPH, becomes localized to the cell membrane. PI3K is itself activated by the insulin receptor, and so increased localization of GPH to the cell membrane indicates increased Inr activity. In InrE19/InrGC25 larvae reared at 15 °C, GPH is localized to the membrane, but this localization is lost when the larvae are moved to 25 °C for 12 h (Figure 2E and 2F). Localization does not appear to be lost at 25 °C in InrE19 or InrGC25/TM6B-Tb controls (Figure 2G and 2H). The phenotypic effects of temperature on InrE19/InrGC25 flies do, therefore, appear to be a consequence of temperature-sensitive suppression of the insulin-signaling pathway. However, we do not know the molecular basis for the temperature sensitivity of the Inr
E19
/Inr
GC25 transheterozygotes. It may arise from temperature-sensitive expression of one or both of the mutant alleles or from decreased function of the mutant receptor at higher temperatures.
Figure 2 Increasing the Temperature of InrE19/InrGC25 Flies Suppresses the Insulin-Signaling Pathway
The dFOXO panel shows localization of dFOXO protein in the fat body, the propidium iodide panel shows the position of the nuclei, and the merge panel clarifies the degree of dFOXO localization to the nuclei.
(A) Endogenous dFOXO in the fat body of InrE19/InrvGC25 third instar larvae has weak nuclear localization at 17 °C.
(B) Increase in rearing temperature causes a decrease in cytoplasmic distribution and an increase in nuclear localization of dFOXO, consistent with a decrease in the level of insulin signaling
(C and D) Temperature has no detectable effect on dFOXO localization in InrE19/TM3 control flies.
(E) GPH membrane localization reveals high levels of insulin signaling in the fat body of InrE19/InrGC25 second instar larvae reared at 15 °C. GPH is in green, DNA is stained blue. (F) This localization is lost when the larvae are moved to 25 °C for 12 h, consistent with a decrease in the level of insulin signaling.
(G and H) Temperature has no detectable effect on GPH membrane localization in InrE19/TM3 control flies.
Due to the low viability of InrE19/InrGC25 flies reared at 25 °C, all subsequent experiments involved comparisons of flies reared at 17 °C with flies reared at 24 °C. Further, we used the InrE19/TM3 siblings reared under identical conditions as controls. InrE19/TM3 flies show only a moderate reduction in Inr activity and have a slight reduction in body size relative to wild-type [4]. Because temperature affects overall body-size in wild-type flies [19], it is necessary to distinguish between changes in InrE19/InrGC25 phenotype that are a consequence of changes in the level of Inr expression from those resulting from changes in rearing temperature. To do this we report the phenotype of InrE19/InrGC25 flies as a percentage of the phenotype of InrE19/TM3 control flies that have undergone temperature shifts at the same developmental time.
The Changing Role of Insulin Signaling during Development
We used the temperature-sensitive Inr mutants to investigate the role of the insulin pathway during Drosophila development. We reduced Inr activity during different periods of development by transferring InrE19/InrGC25 flies from a permissive 17 °C to a restrictive 24 °C at different points in development. After the switch, the flies were left to complete development at 24 °C. We were able to identify temperature-sensitive periods (TSPs) during which increasingly earlier switches from 17 °C to 24 °C resulted in increasingly abnormal phenotypes.
Total developmental time is sensitive to Inr activity only before the middle of the third larval instar (Figure 3A). Switching InrE19/InrGC25 flies from 17 °C to 24 °C changes the time to adult eclosion during the first 9 d of development; the earlier the switch, the greater the delay in eclosion. After the ninth day of development at 17 °C, when the flies are approximately 40% through the third instar (Figure 3B), a shift to the restrictive temperature does not delay adult eclosion. At 17 °C, InrE19/InrGC25 flies show a slight delay in eclosion relative to InrE19/TM3 flies, suggesting that Inr activity is still a little impaired at this temperature. The delay in InrE19/InrGC25 flies reared at both 17 °C and 24 °C relative to the InrE19/TM3 controls occurs predominantly through extension of the third instar (Figure 4).
Figure 3 Suppression of Inr Expression in InrE19/InrGC25 Flies Affects Developmental Time and Adult Size
(A) Developmental time and adult wing size of InrE19/InrGC25 females switched from 17 °C to 24 °C increasingly late in development, expressed as percentage of developmental time and adult wing size of InrE19/TM3 females maintained under identical thermal conditions. Temperature-control flies were maintained at 17 °C throughout development. TSPs of female InrE19/InrGC25 for wing area and delayed eclosion can be seen as regions of the chart where switching from 17 °C to 24 °C increasingly early in development results in increasingly abnormal phenotypes (that is, where the gradient of the relationship between switch day and phenotype is non-zero). For delayed adult eclosion, the TSP of female InrE19/InrGC25 is before the ninth day of development at 17 °C. For reduced wing size, the TSP of female InrE19/InrGC25 is between the ninth and approximately the 20th day of development at 17 °C.
(B) The stages of development of InrE19/InrGC25 flies at 17 °C (A, adult; E, embryo; L1, first instar; L2, second instar; L3, third instar; P, pupae). The point at which suppression of the insulin pathway changes from delaying adult development to reducing adult wing size occurs approximately 40% into the third instar (vertical gray bar)
(C) Dry mass of InrE19/InrGC25 males switched from 17 °C to 24 °C at different points in development, expressed as percentage of dry mass of InrInrE19/TM3 males maintained under identical thermal conditions. The TSP of male InrE19/InrGC25 for reduced adult mass is after the ninth day of development but before pupariation.
(D) Proportion of 17 °C InrE19/InrGC25 larvae pupariating when completely starved at different points in development. The point at which 50% of larvae pupariate in the absence of food marks the critical size. The critical size is reached approximately 40% through the third instar and coincides with the end of the TSP for delayed eclosion and the beginning of the TSP for reduced wing size and adult dry mass (vertical gray bar). All pupariating larvae successfully completed metamorphosis and eclosed as adults.
Figure 4 Developmental Delay in 24 °C InrE19/InrGC25 Flies Occurs primarily through Elongation of the Third Larval Instar
Area shows percentage of individuals (n = 20) in each developmental stage at different times in development. Time is shown in DDs to control for the effect of temperature on developmental rate.
In contrast, adult wing area is sensitive to Inr activity only during late third instar and early pupation (see Figure 3A). Switching InrE19/InrGC25 flies from 17 °C to 24 °C changes adult wing size between day 9 and 20; the earlier the shift, the smaller the wings The precise TSP of female InrE19/InrGC25 flies for wing area may be smaller than implied by Figure 3A because the data are based on population rather than individual measures. The wings may be insensitive to changes in Inr activity as early as day 17. Before day 9, however, shifting the flies to the restrictive temperature earlier in development has no additional effect on adult wing size. At 17 °C, the wings of InrE19/InrGC25 flies are slightly smaller than in InrE19/TM3 control flies, again suggesting that Inr activity is marginally reduced at this temperature.
We tested whether total body mass was sensitive to Inr activity over the same period as wing size. We compared the dry mass of adult InrE19/InrGC25 males reared under several thermal conditions: 24 °C, 17 °C, and a series of samples switched from 17 °C to 24 °C on days 9 through 16 (pupariation is at approximately day 13 at 17 °C). Adult body mass is sensitive to reduction in Inr activity between day 9 and day 13 at 17 °C (see Figure 3C). Shifting InrE19/InrGC25 flies to the restrictive temperature after pupariation, and therefore after larvae have stopped feeding, has no influence on adult mass. Shifting the flies earlier than day 9 has no additional effect on adult body mass.
Inr activity therefore influences total development time, adult wing size, and adult mass for different periods of development. We tested whether the time at which down-regulation of the insulin pathway switches from delaying development to reducing the size of the resulting flies coincides with attainment of critical size. In practice, the critical size is determined as the size at which 50% of larvae proceed to pupariation in the absence of food [8]. We measured the timing of critical size in InrE19/InrGC25 flies reared at 17 °C under our experimental conditions and found that it is attained at approximately day 9 (see Figure 3D), coinciding with the time at which Inr activity switches from affecting developmental time to body size. Therefore, larval critical size coincides with the shift in Inr function.
A Reduction in Inr Activity Changes Body Chemistry
To better understand why reduction of Inr activity after pupariation influenced adult wing size but not overall body mass, we measured the protein, lipid, glycogen, and sugar content of InrE19/InrGC25 flies reared at 17 °C up to pupariation, then either switched to 24 °C or maintained at 17 °C.
Reducing Inr activity after pupariation results in an approximate doubling of free-sugar concentration in InrE19/InrGC25 flies (Table 1). There is no similar response in protein, glycogen, or lipid levels. Lipid levels are, however, elevated in both InrE19/InrGC25 flies maintained at 17 °C and those switched to 24 °C at pupariation relative to InrE19/TM3 flies (Table 1). InrE19/InrGC25 flies grown at a restrictive 24 °C for their entire development also have elevated lipid levels (138% ± 13.3 of InrE19/TM3 flies). This suggests that the slight deficiency in insulin signaling at 17 °C may account for the entire effect on lipid levels.
Table 1 Protein, Sugar, Glycogen, and Lipid Content of InrE19/InrGC25 and InrE19/TM3 Flies Reared at 17 °C until Pupariation and then Switched to 24 °C or Maintained at 17 °C
A Reduction in Inr Activity Affects Different Organs Differently
Different organs typically grow at different rates in developing animals, a phenomenon called “allometry.” Patterns of allometry in populations can also result from organs growing at the same rate but starting and stopping growth at different times in development [20]. We investigated whether the insulin-signaling pathway might be involved in the regulation of allometry by examining whether different organs respond differently to reductions in Inr activity induced by a temperature shift of InrE19/InrGC25 flies from 17 °C to 24 °C. We examined organs located at the anterior, median, and posterior of males: the maxillary palp, wing, and genital arch. We chose the maxillary palp rather than another anterior organ because it is similar in size to the genital arch, allowing us to control for absolute size in this comparison.
Figure 5 shows the relative areas of wings, genital arch, and maxillary palps of male InrE19/InrGC25 and male InrE19/TM3 flies reared at either 17 °C or 24 °C. At 17 °C, all three organs are smaller in InrE19/InrGC25 males than InrE19/TM3 males. At 24 °C, the wing and maxillary palps are further reduced in InrE19/InrGC25 males, consistent with a further reduction in the level of Inr activity. In contrast, there is no difference in the size of the genital arches, relative to InrE19/TM3 males, in flies reared at 17 °C and 24 °C.
Figure 5 Different Organs Respond Differently to Suppression of Inr Activity
Bars show organ area in InrE19/InrGC25 males as a percentage of area in InrE19/TM3 males, to control for temperature effects. Bars with different letters indicate organs that differ: A, B, and C are significantly different at α = 0.05 (Tukey-Kramer pairwise comparison). Mean areas of all organs given in Table S1. s.e., standard error.
This suggests that the size of the genital arches may not be regulated by Inr activity. However, the arches are smaller in InrE19/InrGC25 males than in InrE19/TM3 males at both rearing temperatures. It is possible that this size difference is a consequence of genetic background, unrelated to differences in Inr activity in the two genotypes. Alternatively, the genital arches may show a limited response to changes in Inr activity. They may be sensitive to a mild reduction in Inr activity experienced by InrE19/InrGC25 flies reared at 17 °C, but be insensitive to a further reduction experienced by InrE19/InrGC25 flies reared at 24 °C.
To distinguish between these two hypotheses, we generated large Minute clones homozygous for chico1. Homozygous chico1 clones produce phenotypes identical to mutant Inr clones [13] and autonomously cause a dramatic size reduction in fly wings and eyes, whereas the heterozygous chico1 cells behave as wild-type [5]. We identified males that were homozygous for chico1 throughout one side of the genitals and heterozygous for chico1 on the other. If genital size is independent of the insulin-signaling pathway, then the genitals from either compartment should be identical in size. Each comparison was made within a single male, automatically controlling for total body size. We found that genital arches consisting of mutant chico1 clones were 16% smaller than paired genital arches on the same male (genital arch area: chico1 mutant = 2,840 ± 60 μm2, chico1 wild-type = 3,230 ± 110 μm2, paired-sample t test: n = 6, t = 1.67, p = 0.0214). This reduction of 16% is consistent with the 16% reduction observed in InrE19/InrGC25 males relative to controls (Figure 5). In contrast, maxillary palps consisting of mutant chico1 clones were 45% smaller than paired palps on the same male (maxillary palp area: chico1 mutant = 4,820 ± 100 μm2, chico wild-type = 8,640 ± 170 μm2, n = 7). The genital arches do, therefore, show a limited response to changes in insulin signaling. They are sensitive to a mild reduction in Inr activity, such as observed in InrE19/InrGC25 flies reared at 17 °C. Further reduction in Inr activity, such as observed in InrE19/InrGC25 flies reared at 24 °C, has no additional effect on genital arch size but does have an effect on maxillary palp and wing size.
A Reduction in Inr Activity Affects Cell Size and Cell Number Independently
To investigate the cellular basis for the effect of reduction of Inr activity on size, we compared the size and number of epidermal wing cells in Inr
E19
/Inr
GC25 flies with InrE19/TM3 flies, reared at 17 °C and 24 °C. The wing-size difference between InrE19/InrGC25 and InrE19/TM3 flies reared at 17 °C is due to smaller cells but not fewer cells in the wings of InrE19/InrGC25 males (compare relative sizes at 17 °C, Figure 6, Table S2). In InrE19/InrGC25 females reared at 17 °C, cell size is also smaller than in InrE19/TM3females, but there are also slightly fewer cells. In both males and females, the additional difference in wing-size observed in InrE19/InrGC25 relative to InrE19/TM3 flies reared at 24 °C is due to fewer, but not smaller, cells: Cell size in InrE19/InrGC25 wings relative to InrE19/TM3 wings is the same at both 17 °C and 24 °C (Figure 6, Table S2). Reduction in wing area through suppression of the insulin pathway in InrE19/InrGC25 flies is therefore predominantly a consequence of smaller cells at 17 °C but fewer cells at 24 °C.
Figure 6 A Reduction in Inr Activity Affects Cell Size and Cell Number Independently
(A) At 17 °C, the difference in wing area between InrE19/InrGC25 and InrE19/TM3 flies is due to a difference in cell size, whereas at 24 °C the difference is due to an additional difference in cell number. Bars show wing area, cell area, and cell number in Inr
E19/Inr
GC25 flies as a percentage of area or number in InrE19/TM3 flies.
(B) At 17 °C the reduced Inr activity in InrE19/InrGC25 flies reduces cell area to approximately 85% the area in InrE19/TM3 flies, whereas at 24 °C there is no further reduction in cell area, but there is a reduction in cell number to approximately 75% of the number in InrE19/TM3 flies. Mean wing and cell area, and cell number are given in Table S2.
Discussion
Suppressing the insulin-signaling pathway extends developmental time and reduces final adult size in Drosophila. By varying the activity of Inr we have demonstrated that these effects depend on when in development the suppression occurs; Inr suppression affects total developmental time early in development, and body and organ size late in development. The transition from affecting developmental time to affecting body and organ size occurs when the fly reaches the critical size, the point at which development can be completed in the absence of food. The effect of reduced Inr activity on size varies from organ to organ, implicating the insulin pathway in the control of allometry in Drosophila. In addition, varying the magnitude of Inr activity has different effects on cell size and cell number.
Insulin Signaling, Critical Size, and Developmental Delay
We found that a reduction in Inr activity after critical size does not delay development (see Figure 3A). Therefore reducing Inr activity affects total developmental time by delaying the time at which larvae reach critical size. Critical size has been identified as the key stage in insect maturation that determines the point at which, in holometabolous insects, a larva becomes committed to metamorphosis [21]. In Drosophila it is also the minimal viable weight necessary to survive pupation. Consequently, because larvae can complete development without any additional feeding after reaching its critical size, the critical size sets the lower limit of final adult size.
How insects measure critical size is largely unknown. Our results indicate that a reduction in Inr activity delays the point at which Drosophila larvae reach critical size. We hypothesize that critical size measurement involves a specific organ or organs, and that it is the slow growth of this organ or organs that delays development in Inr mutants. Possible candidates include the imaginal discs and the fat body. Regeneration of damaged imaginal discs delays pupariation in Drosophila [22,23], indicating that some or all of the imaginal discs need to grow to a particular size before a larva can pupariate. Complete removal of the discs does not, however, delay pupariation [23]. Another organ must therefore measure critical size and initiate pupariation, with the immature imaginal discs inhibiting pupariation. Slow growth of either the imaginal discs or a “critical-size organ” could delay development in Inr mutants. Recent studies suggest this “critical-size organ” could be the fat body, which functions as a nutrient sensor and is involved in the coordination of organismal growth [24]. Suppressing Inr/PI3K signaling in the fat body alone is sufficient to inhibit larval growth and mimics the effects of starvation [18].
Insulin Signaling and Final Body and Organ Size
A reduction in Inr activity after critical size reduces final adult size and organ size (see Figure 3A and 3C). Inr activity influences adult body size and wing size for different periods of development. As expected, adult body size is influenced by Inr activity only between critical size and pupariation, after which the larva does not feed and becomes a “closed system,” neither gaining nor losing mass. However, insulin signaling continues to influence the final size of adult organs well into the pupal stage. InrE19/InrGC25 flies switched from 17 °C to 24 °C at pupariation have the same mass as InrE19/InrGC25 flies maintained at 17 °C, but have reduced wings. At the same time, the temperature-shifted flies have a much higher free-sugar concentration as adults. These two findings appear to be linked. Considerable cell proliferation in the imaginal discs occurs after the larva has stopped feeding [11,12,25], and this proliferation relies entirely on stored nutrients as an energy source. Nutrient storage occurs predominantly in the fat body cells, which accumulate reserves of proteins, lipids, and glycogen (the major carbohydrate storage compound) during the third larval instar [18]. Both starvation and suppression of the insulin pathway cause these nutrients to be mobilized for use by growing cells [18]. The finding that InrE19/InrGC25 mutants reared at 24 °C have elevated free-sugar levels, but do not have elevated glycogen levels, indicates that they are able to mobilize their carbohydrate reserves, but that the free-sugars are not taken up by growing organs, and remain in the haemolymph.
Insulin Signaling and Cell Size and Cell Number
Any mechanism that influences organ size does so by changing cell size, cell number, or both. Although mutations of Inr, PI3K, and chico reduce both cell size and cell number [4,5,13,26], downstream components of the insulin-signaling pathway can affect cell size and cell number independently. The RPS6-p70-protein kinase (S6K) branch of the pathway appears to influence cell size but not cell number [27], whereas the dFOXO branch appears to influence cell number but not cell size [15,16]. Because these signaling branches diverge downstream of the insulin receptor, it has not been clear how insulin signaling could affect cell size and number differentially. These changes are ultimately a consequence of changes in the relative rates of cell growth and division [28]. For example, the reduction in cell size, but not cell number, in S6K mutants implies a reduction in the rate of cell growth but not of cell division. Conversely, the reduction in cell number, but not cell size, when dFOXO is over-expressed implies that the rates of cell growth and division are reduced equally. Finally, a reduction in both cell size and cell number will result if both the rates of cell growth and division are reduced, if the former is reduced to a greater extent than the latter.
Our results support the hypothesis that insulin signaling differentially affects cell size and cell number via different levels of insulin receptor activity. Slightly reduced levels of activity, as in InrE19/InrGC25 flies raised at 17 °C, reduce cell size, possibly through a reduction in the rate of cell growth but not of cell division. Further reductions in the levels of activity, as in InrE19/InrGC25 flies raised at 24 °C, reduces cell number only, possibly through a subsequent balanced reduction in both the rate of cell growth and cell division. These data are consistent with a result from Bohni et al. [5]. They showed that the wings of chico2 homozygotes are smaller due to a reduction of both cell size (17%) and cell number (27%). However, a further reduction in insulin signaling through the removal of a single copy of Inr enhances the small-size phenotype exclusively through a reduction in cell number but not cell size. Although we cannot exclude the possibility that the differences in cell size between InrE19/InrGC25 and InrE19/TM3 flies are a consequence of genetic background unrelated to differences in Inr activity, the reduction in cell number alone is clearly due to reduction in insulin signaling (see Figure 5).
Insulin Signaling and Allometry
Variation in insulin signaling appears to affect the allometric relationship between organs. For example, at 17 °C, male InrE19/InrGC25 flies have body size, wings, maxillary palps, and genital arches approximately 85% of wild-type. Increasing the rearing temperature of InrE19/InrGC25 flies from 17 °C to 24 °C, however, causes a further reduction in wing, maxillary palp, and overall body size but does not affect the size of the genital arches (Figure 5). Consequently, flies reared at 24 °C have larger genitals relative to their body and wings compared to flies reared at 17 °C. The apparently restricted response of the genitals to changes in insulin signaling is not a consequence of the particular alleles used in this study: chico-mutant clones also have much less of an effect on size when they are in the genital arches than when they are in the maxillary palps.
The mechanism by which different organs respond differently to changes in insulin signaling is unclear. Organs may vary in their expression of the insulin receptor gene or may limit the activity of certain downstream components of the insulin signaling pathway. In 17 °C InrE19/InrGC25 males, the genital arches, wings, and maxillary palps are all reduced by approximately the same amount relative to 17 °C InrE19/TM3males. In the wing, this reduction in area is a consequence of a reduction in cell size. A further decrease in Inr activity (through an increase in rearing temperature to 24 °C) reduces wing area through a reduction in cell number alone. If the genital arches are like the wing, then their response to insulin signaling may be restricted to changes in cell size and not cell number. The cells of the genital arches may therefore be deficient in components of the insulin-signaling pathway that regulate cell number but not components that regulate cell size.
A Model of the Insulin-Signaling Regulation of Growth and Development
The insulin-signaling pathway appears to play a different role after critical size than before. Similarly, a temperature-sensitive lethal mutation, l(1)ts-1126, which reduces the rate of cell proliferation in Drosophila, delays pupariation when larvae are moved to a restrictive temperature before the third instar, but reduces adult size when larvae are moved to a restrictive temperature late in the third instar [29]. The two effects of reduced Inr activity may therefore result from the same process: a reduction in the rate of cell growth and proliferation. We have developed a model of the insulin-signaling regulation of growth and development in Drosophila (Figure 7). (A similar model has recently been developed by Davidowitz and Nijhout [30] to explain variation in body size in response to temperature in the tobacco hornworm, Manduca sexta.) A reduction in Inr activity prior to critical size slows cell growth and proliferation and delays the time at which the larvae reaches critical size. Critical size is not substantially influenced by nutritional conditions [8] or insulin signaling in Drosophila, although this does not seem to be the case for all insects, for example, M. sexta [21]. Once critical size is reached, the time to pupariation and adult eclosion is fixed, as are the remaining periods of growth prior to adult differentiation of individual imaginal discs. The duration of these intervals are uninfluenced by nutritional conditions or insulin signaling. A reduction in Inr activity during these periods also slows cell growth and proliferation, but now reduces the amount of growth attained before differentiation, resulting in smaller organs and a smaller fly. Because different structures grow for different periods, they are sensitive to Inr activity at different times. For example, wing size begins to be insensitive to changes in Inr activity at approximately the same time as cell proliferation ceases, around 25% into the pupal stage. Adult body mass becomes insensitive to changes in Inr activity just before pupariation, when the larvae stops feeding and final body size is fixed.
Figure 7 A Model of the Insulin-Signaling Regulation of Growth and Development
(A) Under normal conditions, imaginal discs grow to a critical size, which initiates an increase in the ecdysteroid titer. When ecdysteroid levels rise above a maximum threshold, the discs cease cell proliferation and undergo differentiation, fixing their final size. A, adult; E, embryo; L1–L3, first to third larval instar; P, pupa.
(B) In Inr mutants, growth of imaginal discs to critical size is slowed, retarding development. When critical size is reached, the ecdysteroid titer again increases, rising above the maximum threshold for cell proliferation in the imaginal discs. Temporal changes in the ecdysteroid titer are unaffected by insulin signaling. Because the rate of cell proliferation is slowed, the imaginal discs are smaller when they begin to differentiate, reducing final organ size. Different discs have different thresholds of sensitivity to ecdysteroid and so cease cell proliferation at different times. Hormones other than ecdysteroids may also be involved.
A key component of this model is that after critical size, the remaining periods of growth of individual imaginal discs and of the body as a whole are fixed and uninfluenced by insulin signaling. Our data show that the length of the period between critical size and pupariation, the remaining period of growth of the body as a whole, is not substantially affected by a reduction in Inr activity. In Drosophila and other insects, this interval is controlled by endocrine events. In Drosophila, a small peak in the ecdysteroid titer coincides with attainment of critical size [31], followed by a second peak 12 h later, just before the larvae leave the food [32]. In M.
sexta this second peak acts directly on the nervous system to initiate wandering behavior [33], and the same is likely true for Drosophila [34]. The period in which Inr activity can influence final body size is therefore terminated by hormones. Importantly, hormones other than insulin also control the period of cell proliferation in the imaginal discs. For example, in M.
sexta, ecdysteroids govern the phases of eye development during metamorphosis [35,36]. When the ecdysteroid titer rises above a minimum threshold just before pupariation, it stimulates a wave of cell proliferation to pass across the eye primordium. This proliferation is sustained until the ecdysteroid titer rises above a maximum level in the middle of pupal development, whereupon cell proliferation stops and maturation of the ommatidia begin. These and similar data in other insects [37–39], including Drosophila [40–43], suggest that cell proliferation in imaginal discs may be temporally regulated by thresholds of sensitivities to fluctuating levels of ecdysteroids and juvenile hormone (JH) [44]. Different organs have different thresholds of sensitivity and hence grow for different periods of time. Insulin signaling may therefore control body and organ size by regulating the amount of growth attained during these periods of cell proliferation.
This model requires that changes in ecdysteroid and JH levels are unaffected by the insulin-signaling pathway. It is known that adult Inr and chico mutant flies have reduced levels of JH and impaired ovarian ecdysone synthesis [45,46]. However, the same hormone fluctuations that putatively control the period of cell proliferation also initiate pupariation [34] and, because the timing of pupariation is unaffected by Inr activity, it seems likely that the temporal dynamics of the hormonal cascade in the larvae are also unaffected by Inr activity. We predict, therefore, that the insulin-signaling pathway regulates cell proliferation in imaginal discs but that the duration of proliferative phases are controlled by other endocrine cues, such as JH and ecdysteroids, that are themselves unaffected by the insulin-signaling pathway.
This model demonstrates how the pleiotropic effects of insulin signaling on developmental time and final body and organ size can be separated, and may be available for independent evolutionary modification. For example, changes in the relative size of an organ may occur by organ-specific modifications in its growth response to insulin signaling, through organ-specific changes in the expression of Inr, adjustments in Inr activity, or adjustments in the expression and activity of downstream components of the insulin-signaling pathway. Alternatively, changes in the period of an organ's growth, through alterations in its sensitivity to other endocrine cues, may have a similar effect [44]. In metazoans in general, each organ has a unique timetable for cellular events in tissue development. Our model, and the data upon which it is based, indicate that in order to understand the effects of insulin signaling on adult phenotype it is necessary to understand how temporal changes in insulin signaling interact with this timetable.
Materials and Methods
Mutant stocks
InrGC25 is a chromosomal inversion with a breakpoint upstream of the Inr. InrE19 is an uncharacterized mutation induced by ethyl methanosulfonate. Both were described in Chen et al. [4] and obtained from the Bloomington Stock Center. The chico
1 allele is a P-element insertion allele whose phenotype is similar to the null chico
2 allele [5] and was kindly provided by Ernst Hafen. The tub-GFP-PH flies were kindly supplied by Bruce Edgar. The flies were maintained as chico1/CyO [5], InrE19, and InrGC25 [4] balanced over TM6B-Tb, TM3, TM3-pAct-GFP, and tub-GFP-PH; Inr
E19/TM3 [18] at 17 °C on standard yeast cornmeal agar medium.
Immunolocalization
We crossed InrE19/TM6-Tb with InrGC25/TM6-Tb flies and reared them at either 17 °C or 24 °C. When the larvae reached third instar, we genotyped them as either insulin receptor mutants (InrE19
/InrGC25) or wild-type (InrE19 or InrGC25/TM6B-Tb). The fat bodies were dissected out in PBS, fixed in 4% paraformaldehyde for 10 min, and stored in absolute methanol at −20 °C. They were permeabilized with PBT (0.3% Triton X in PBS) for 30 min, washed in BBT (0.3% bovine serum albumin in PBT) for 30 min, then blocked in NGS/BBT (3% normal goat serum in BBT) for 30 min. They were stained with anti-dFOXO 2095 [16] (kindly provided by O. Puig) (1:1,000 in PBT) and fluorescein anti-rabbit (1:500 in PBT) (Vector Labs, Burlingame, California, United States). DNA was stained with propidium iodide (1:1,000 in PBS with 1 μg of RNase A). We mounted the fat bodies in Vectashield (Vector Labs) for observation under a confocal microscope (Perkin Elmer UltraVIEW RS3; PerkinElmer Life and Analytical Sciences, Boston, Massachusetts, United States).
GPH localization
We crossed tub-GFP-PH;InrE19/TM6B-Tb with Inr
GC25/TM6B-Tb flies and reared them at 15 °C. When the larvae reached second instar, they were removed from their food, washed, and genotyped as insulin receptor mutants (tub-GFP-PH;InrE19/InrGC25) or wild-type (tub-GFP-PH;InrE19 or InrGC25/TM6B-Tb). They were then transferred to fresh food and maintained at 15 °C or 25 °C for 24 h. We then dissected the larvae in 100% methanol kept at 15 °C or 25 °C depending on the temperature treatment of the larvae, and stored their fat bodies without additional fixing in 100% methanol at −20 °C. We quickly washed the fat bodies in 50% methanol in PBS, and then 100% PBS before mounting them in Vectashield with Hoechst 33342 (2:1,000) to stain DNA. We observed the fat bodies under a confocal microscope (Zeiss LSM 510; Zeiss, Gena, Germany).
Temperature shift
We reared flies at a number of temperature regimes. They were either maintained at 17 °C, 18 °C, 24 °C, or 25 °C or switched from 17 °C to 24 °C on day 1 to day 27 of development. We set up four bottles for each temperature regime, each containing standard yeast cornmeal agar medium. Two were the product of male InrE19/TM6B-Tb and female InrGC25/TM3 and two were the product of female InrE19/TM6B-Tb and male InrGC25/TM3. Each bottle contained between 100 and 200 eggs laid over a 4-h period. Temperature shifts were performed according to the median age of flies in each bottle. Larvae were reared on standard yeast cornmeal agar medium. Total developmental time was not significantly different between InrE19/InrGC25 flies from the two different crosses, so the data from all bottles maintained under the same temperature regime were pooled (data not shown).
Measurements and data handling
Bottles were inspected daily at 4:00 PM and any flies that had eclosed in the previous 24 h were collected and preserved in 80% ethanol. Flies were then genotyped and the eclosion times to the nearest day of each InrE19/InrGC25 and InrE19/TM3 control fly was recorded. We used InrE19/TM3 rather than InrGC25/TM6B-Tb flies as controls because TM6-Tb has Tubby as a marker, which affects developmental timing and adult shape [47]. Ten to 15 InrE19/InrGC25 and InrE19/TM3 female wings from each temperature regime were dissected and mounted in lactic acid:ethanol (4:5). Ten InrE19/InrGC25 and InrE19/TM3 males reared at 24 °C and 15 InrE19/InrGC25 and InrE19/TM3 males reared at 17 °C were also dissected and their wings, maxillary palps, and genital arches mounted in either lactic acid:ethanol (wings) or Hoyer's Solution (maxillary palps and genital arches). Digital images of the wings, maxillary palps, and genital arches were captured and measured using IPLab 3.9 (Scanylitics, Fairfax, Virginia, United States). Wing cell size was estimated using the number of trichomes in a 0.01 mm2 square between the veins IV and V of the dorsal wing blade. An index of the total number of wing cells was estimated by multiplying the number of trichomes in the area by the total wing size and dividing by 0.01. Dry masses of individual male flies were measured with an analytical balance (Mettler Toledo AX26 DeltaRange; Mettler-Toledo, Columbus, Ohio, United States).
The total developmental time for each fly was converted into degree-days (DDs) to control for the effects of temperature on growth rates. DD is a measure of the heat accumulation above a minimal temperature (To), and is calculated as:
where T
r is the experimental temperature, and n
r is the number of days maintained at T
r, summed across all experimental temperatures j to k. In this case, we used only two experimental temperatures, 17 °C and 24 °C. We calculated the value of T
0 such that the developmental time of control InrE19/TM3 flies was constant, irrespective of temperature regime. To do this we regressed total developmental time in DDs of InrE19/TM3 flies against the time of their temperature shift from 17 °C to 24 °C, using different values of T
0. The value of T
0 that minimized the slope of this regression line was found to be 9.978 °C. When T
0 is 9.978 °C, the average developmental time of InrE19/TM3 flies is 156.04 ± 0.338 DD. We then converted the total developmental time for each InrE19/InrGC25 fly into DDs using T
0 = 9.978, and expressed it as the average percentage of developmental time of InrE19/TM3 flies (156.04 DD), along with the standard error of the percentages.
Rearing temperature also affects body, organ, and cell size and cell number [19]. Consequently, all these measurements in InrE19/InrGC25 flies were expressed as a percentage of the value in InrE19/TM3 flies maintained under the same temperature regime. Wing area of InrE19/InrGC25 females and dry mass of InrE19/InrGC25 males switched from 17 °C to 24 °C increasingly late in development, was expressed as the percentage of wing area and dry mass of InrE19/TM3 flies switched from 17 °C to 24 °C at the same percent of total development at 17 °C. See Figure S1 for details.
Clonal analysis
We crossed males carrying the alleles chico1/CyO with females carrying the alleles f36a;M(2)Z f+37C/CyO . Larvae were subjected to X-rays (1,000 rad) at 24–72 h after egg-laying. Genitals and maxillary palps of male offspring without balancer chromosomes were studied for f36a bristles.
Developmental staging of InrE19/InrGC25 and InrInrE19/TM3 at 17 °C and 24 °C
We crossed InrE19/TM3-pAct-GFP flies with InrGC25/TM3-pAct-GFP flies and collected the eggs over a period of 4 h. After 24 h, we selected InrE19/InrGC25 and InrE19 or InrGC25
/TM3-pAct-GFP eggs using a fluorescence microscope to detect GFP activity. The eggs were transferred to 12 mm ø Petri dishes containing standard yeast cornmeal agar medium. Each dish contained 20 eggs of only one genotype and was maintained at either 17 °C or 24 °C. We prepared three dishes: InrE19/InrGC25 maintained at 17 °C; InrE19/InrGC25 maintained at 24 °C; and InrE19 or InrGC25
/TM3-pAct-GFP maintained at 17 °C. From day 2, ten individuals were randomly selected from each Petri dish and their developmental stage was recorded using mouthpart development, before being returned to the Petri dish.
Body chemistry assays
White prepupae of InrE19/InrGC25 and InrE19/TM3 flies raised at 17 °C were transferred to moistened Kimwipes (Kimberly-Clark, Neenah, Wisconsin, United States) in Petri dishes. Prepupae of each genotype were split between further development at 17 °C and 24 °C. We collected adults within 8 h of eclosion, after their cuticle had hardened, and immediately froze them in liquid nitrogen for later analysis. Wet and dry masses of individual flies were measured with an analytical balance (Mettler Toledo AX26 DeltaRange). All metabolic assays were performed on individual dried males (n = 10 for each assay). Glycogen and sugar content were measured using a protocol of van Handel [48]. Lipid content was quantified using another protocol of van Handel [49]. To determine protein concentration, flies were homogenized in 100 μl 0.1 M Na2HPO4. Ten μl of each sample was combined with the Bio-Rad protein assay dye reagent (Bio-Rad Laboratories, Hercules, California, United States) following the manufacturer's instructions and spectrophotometrically assayed at 595 nm.
Supporting Information
Figure S1 Fitted Values for Wing Aarea of InrE19/TM3 Control Flies
Wing area of InrE19/InrGC25 flies in Figure 2 is expressed as percentage of wing area of InrE19/TM3 flies kept under the same temperature regime, using the fitted values shown on the plot. Temperature affects wing area differently before and after critical size. Consequently, fitted values were calculated by regressing wing area of InrE19/TM3 flies against transfer age before critical size (35% development) and after critical size. A similar method was used to determine the dry mass of InrE19/TM3 males, except a single regression analysis was used to calculate the fitted values.
(59 KB PDF).
Click here for additional data file.
Table S1 Wing, Genital Arch, and Maxillary Palp Area of InrE19/InrGC25 and InrE19/TM3 Males Reared at 17 °C and 24 °C
(37 KB DOC).
Click here for additional data file.
Table S2 Total Area, Cell Area, and Cell Number of the Wings of InrE19/InrGC25 and InrE19/TM3 Flies Reared at 17 °C and 24 °C
(45 KB DOC).
Click here for additional data file.
Accession Numbers
The FlyBase (http://flybase.bio.indiana.edu/search/) accession numbers for the genes and gene products discussed in this paper are Chico (Fbgn0024248), chico1 (FBal0031303), dFOXO (FBgn0038197), Inr (FBgn0013984), InrE19 (FBal0094021), InrGC25 (FBal0010755), PI3K (Fbgn0015279), and S6K (FBgn0015806). The National Center for Biotechnology Information (NCBI) (http://www.ncbi.nlm.nih.gov/) accession number for IGF1 receptor is NM_010513.
We thank Fred Nijhout and three anonymous referees for critical and helpful comments on early versions of this manuscript. We thank Miguel Gaspar for his assistance with imaging. AWS was supported by Princeton University Council on Science and Technology, JD was supported by a Howard Hughes Medical Institute Predoctoral Fellowship. DLS acknowledges the National Institutes of Health, the David and Lucile Packard Foundation, and Princeton University for financial support. LV thanks Michael Akam and his lab for guidance, and St John's College (Cambridge), the Cambridge Overseas Trust, and the Overseas Award Scheme for funding.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. JD, LV, DLS, and AWS conceived, designed, and performed the experiments. AWS analyzed the data and wrote the paper.
Citation: Shingleton AW, Das J, Vinicius L, Stern DL, (2005) The temporal requirements for insulin signaling during development in Drosophila. PLoS Biol 3(9): e289.
Abbreviations
DDdegree day
JHjuvenile hormone
PI3Kphosophoinositide 3-kinase
TSPtemperature sensitive period
==== Refs
References
Ikeya T Galic M Belawat P Nairz K Hafen E Nutrient-dependent expression of insulin-like peptides from neuroendocrine cells in the CNS contributes to growth regulation in Drosophila
Curr Biol 2002 12 1293 1300 12176357
Shambaugh GE Radosevich JA Glick RP Gu DS Metzger BE Insulin-like growth factors and binding proteins in the fetal rat: Alterations during maternal starvation and effects in fetal brain cell culture Neurochem Res 1993 18 695 703 8510796
Baker J Liu JP Robertson EJ Efstratiadis A Role of insulin-like growth factors in embryonic and postnatal growth Cell 1993 75 73 82 8402902
Chen C Jack J Garofalo RS The Drosophila insulin receptor is required for normal growth Endocrinology 1996 137 846 856 8603594
Bohni R Riesgo-Escovar J Oldham S Brogiolo W Stocker H Autonomous control of cell and organ size by CHICO, a Drosophila homolog of vertebrate IRS1–4 Cell 1999 97 865 875 10399915
Liu JP Baker J Perkins AS Robertson EJ Efstratiadis A Mice carrying null mutations of the genes encoding insulin-like growth factor I (Igf-1) and type 1 IGF receptor (Igf1r) Cell 1993 75 59 72 8402901
Beadle G Tatum E Clancy C Food level in relation to rate of development and eye pigmentation in Drosophila melanogaster
Biol Bull 1938 75 447 462
De Moed GH Kruitwagen C De Jong G Scharloo W Critical weight for the induction of pupariation in Drosophila melanogaster : Genetic and environmental variation J Evol Biol 1999 12 852 858
Bakker K An analysis of factors which determine success in competition for food among larvae of Drosophila melanogaster
Arch Neerl Zool 1961 14 200 281
Robertson FW Ecological genetics of growth in Drosophila .6. Genetic correlation between duration of larval period and body size in relation to larval diet Genet Res 1963 4 74 96
Garcia-Bellido A Merriam JR Parameters of the wing imaginal disc development of Drosophila melanogaster
Dev Biol 1971 24 61 87 5001010
Postlethwait JH Schneiderman HA A clonal analysis of development in Drosophila melanogaster : morphogenesis, determination, and growth in the wild-type antenna Dev Biol 1971 24 477 519 5578888
Brogiolo W Stocker H Ikeya T Rintelen F Fernandez R An evolutionarily conserved function of the Drosophila insulin receptor and insulin-like peptides in growth control Curr Biol 2001 11 213 221 11250149
Kramer JM Davidge JT Lockyer JM Staveley BE Expression of Drosophila FOXO regulates growth and can phenocopy starvation BMC Dev Biol 2003 3 5 12844367
Junger MA Rintelen F Stocker H Wasserman JD Vegh M The Drosophila Forkhead transcription factor FOXO mediates the reduction in cell number associated with reduced insulin signaling J Biol 2003 2 20 12908874
Puig O Marr MT Ruhf ML Tjian R Control of cell number by Drosophila FOXO: Downstream and feedback regulation of the insulin receptor pathway Genes Dev 2003 17 2006 2020 12893776
Dillin A Crawford DK Kenyon C Timing requirements for insulin/IGF-1 signaling in C. elegans
Science 2002 298 830 834 12399591
Britton JS Lockwood WK Li L Cohen SM Edgar BA
Drosophila' s insulin/PI3-kinase pathway coordinates cellular metabolism with nutritional conditions Dev Cell 2002 2 239 249 11832249
French V Feast M Partridge L Body size and cell size in Drosophila : The developmental response to temperature J Insect Physiol 1998 44 1081 1089 12770407
Stern DL Emlen DJ The developmental basis for allometry in insects Development 1999 126 1091 1101 10021329
Davidowitz G D'Amico LJ Nijhout HF Critical weight in the development of insect body size Evol Dev 2003 5 188 197 12622736
Ursprung H Hadorn E [Further research on model growth in combination with partly dissociated wing imaginal disks of Drosophila melanogaster
Dev Biol 1962 4 40 66 13923941
Simpson P Berreur P Berreur-Bonnenfant J The initiation of pupariation in Drosophila : dependence on growth of the imaginal discs J Embryol Exp Morph 1980 57 155 165 7430927
Colombani J Raisin S Pantalacci S Radimerski T Montagne J A nutrient sensor mechanism controls Drosophila growth Cell 2003 114 739 749 14505573
Garcia-Bellido A Merriam JR Clonal parameters of tergite development in Drosophila Dev Biol 1971 26 264 276 5158535
Weinkove D Neufeld TP Twardzik T Waterfield MD Leevers SJ Regulation of imaginal disc cell size, cell number and organ size by Drosophila class I(A) phosphoinositide 3-kinase and its adaptor Curr Biol 1999 9 1019 1029 10508611
Montagne J Stewart MJ Stocker H Hafen E Kozma SC
Drosophila S6 kinase: A regulator of cell size Science 1999 285 2126 2129 10497130
Neufeld TP Shrinkage control: Regulation of insulin-mediated growth by FOXO transcription factors J Biol 2003 2 18 12974982
Simpson P Schneiderman HA A temperature sensitive mutation that reduces mitotic rate in Drosophila melanogaster
Rouxs Arch Dev Biol 1976 179 215 236
Davidowitz G Nijhout HF The physiological basis of reaction norms: The interaction among growth rate, the duration of growth and body size Integr Comp Bio 2004 44 443 449 21676730
Berreur P Porcheron P Berreur-Bonnenfant J Simpson P Ecdysteroid levels and pupariation in Drosophila melanogaster
J Exp Zool 1979 210 347 352
Schwartz MB Imberski RB Kelly TJ Analysis of metamorphosis in Drosophila melanogaster : Characterization of giant, an ecdysteroid-deficient mutant Dev Biol 1984 103 85 95 6425099
Dominick OS Truman JW The physiology of wandering behaviour in Manduca sexta II. The endocrine control of wandering behaviour J Exp Biol 1985 117 45 68 4067505
Riddiford LM Bate M Martinez Arias A Hormones and Drosophila development The development of Drosophila melangogaster 1993 New York Cold Spring Harbor Laboratory Press 899 939
Champlin DT Truman JW Ecdysteroids govern two phases of eye development during metamorphosis of the moth, Manduca sexta
Development 1998 125 2009 2018 9570766
Champlin DT Truman JW Ecdysteroid control of cell proliferation during optic lobe neurogenesis in the moth Manduca sexta
Development 1998 125 269 277 9486800
Berreur P Bougues R Effects of ecdysone on in vivo growth of wing disks of Calliphora erythrocephala
J Insect Physiol 1975 21 915 919
Oberlander H Leach CE Shaaya E Juvenile hormone and juvenile hormone mimics inhibit proliferation in a lepidopteran imaginal disc cell line J Insect Physiol 2000 46 259 265 12770230
Kawasaki H Kawasaki H (1995) Ecdysteroid concentration inducing cell proliferation brings about the imaginal differentiation in the wing disc of Bombyx mori in vitro Dev Growth Differ 1995 37 575 580
Peel DJ Milner MJ The response of Drosophila imaginal disk cell-lines to ecdysteroids Rouxs Arch Dev Biol 1992 202 23 35
Currie DA Milner MJ Evans CW The growth and differentiation in vitro of leg and wing imaginal disc cells from Drosophila melanogaster
Development 1988 102 805 814
Chihara CJ Fristrom JW Petri WH King DS The assay of ecdysones and juvenile hormones on Drosophila imaginal disks in vitro J Insect Physio 1972 18 1115 1123
Cullen CF Milner MJ Parameters of growth in primary cultures and cell lines established from Drosophila imaginal discs Tissue Cell 1991 23 29 39 1905427
Emlen DJ Allen CE Genotype to phenotype: Physiological control of trait size and scaling in insects Integr Comp Biol 2003 43 617 634 21680471
Tu MP Yin CM Tatar M Impaired ovarian ecdysone synthesis of Drosophila melanogaster insulin receptor mutants Aging Cell 2002 1 158 160 12882346
Tatar M Kopelman A Epstein D Tu MP Yin CM A mutant Drosophila insulin receptor homolog that extends life-span and impairs neuroendocrine function Science 2001 292 107 110 11292875
Craymer L [New mutants report] Drosoph Inf Serv 1980 55 197 200
Van Handel E Rapid determination of glycogen and sugars in mosquitoes J Am Mosq Control Assoc 1985 1 299 301 2906671
Van Handel E Rapid determination of total lipids in mosquitoes J Am Mosq Control Assoc 1985 1 302 304 2906672
|
0
|
PMC1184593
|
CC BY
|
2021-01-05 08:21:26
|
no
|
PLoS Biol. 2005 Sep 16; 3(9):e307
|
latin-1
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030307
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030310SynopsisMolecular Biology/Structural BiologyBiochemistryIn VitroThe Ol' Switcheroo Shows How an RNA Enzyme Splices Itself Synopsis9 2005 16 8 2005 16 8 2005 3 9 e310Copyright: © 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.
Functional Identification of Catalytic Metal Ion Binding Sites within RNA
==== Body
Ribonucleic acid (RNA) is a dogma breaker. The “central dogma” of cellular biochemistry mandates that deoxyribonucleic acid (DNA) stores information, and RNA copies this information and uses it to direct the assembly of amino acid building blocks into proteins, such as enzymes. Enzymes catalyze important chemical reactions in the cell, such as the breakdown of glucose or the synthesis of urea.
When biochemists discovered catalytic RNA, they had to ditch the dogma. Because of its structure, it turns out, RNA can act as an enzyme and catalyze reactions. While two strands of DNA tend to zip up into the famous double helix, RNA usually goes solo. The single RNA strand folds back on itself to create myriad tangled arrangements. Some of these arrangements create an active center, the place on the RNA where the enzymatic magic happens. The many RNA enzymes and protein enzymes that use metal atoms to do their job are called metalloenzymes. One example of an important structural motif in RNA metalloenzymes is the group I intron, which can snip itself out of an RNA segment. Understanding exactly how the RNA and the metals interact will help to provide precise answers about how the enzyme really works.
Through X-ray crystallography, researchers have revealed many structural features of group I introns. But X-ray crystallography creates images of the enzyme frozen in time; it does not catch an enzyme in action. In a new study, Joseph Piccirilli, Daniel Herschlag, and colleagues discovered that a particular oxygen atom on a particular nucleotide in a group I RNA must bind to a particular magnesium ion in order for the reaction under study to proceed normally. The oxygen atom is known as the pro-SP phosphoryl oxygen at nucleotide C262 in the intron from the unicellular Tetrahymena thermophila protozoan.
Since there's no way to watch the oxygen and metal hook up during the reaction, how do the researchers know they do? The researchers used the powerful techniques of metal ion rescue and atomic mutagenesis. Here's how it worked. They figured out how well the group I intron reaction works with a normal enzyme. Then, they replaced the oxygen in question with a sulfur atom. The reaction didn't work as well because, by the rules of chemistry, sulfur doesn't like to bind to magnesium. But sulfur does like manganese and cadmium ions. So they replaced the magnesium with one of these other metal ions and measured the reaction. The researchers saw that these other metal ions restored (or “rescued”) enzymatic activity. In short, the enzyme needs a bond where the oxygen and the magnesium are, but the bond doesn't have to be between oxygen and magnesium.
As complicated as that is, plucking out one atom and trading it for another is itself a tricky business. Because most enzymes are made of stubborn amino acids and not nucleotides, atomic mutagenesis can be difficult. And usually when researchers have tried atomic mutagenesis, they've mutated the substrate (the molecule that the reaction acts upon) instead of the enzyme (the molecule that acts). Here, Piccirilli, Herschlag, and colleagues directed the applications of atomic mutagenesis to the molecule that does the work.
RNA enzymes called ribozymes require metal atoms to function. The site of metal-ribozyme interaction was studied by changing components of both the catalytic center of the ribozyme (the backbone) and its target substrate
To test that a specific oxygen in the intron binds to the magnesium ion, the researchers first had to compile a short list of potential atoms to which the magnesium might bind. By combining literature data from structural models and functional studies with a random sprinkling of sulfur atoms in the intron to find critical oxygen contacts, Piccirilli, Herschlag, and colleagues established a group of specific oxygen atoms to watch. They tried the metal rescue experiment with each of these oxygens, and the only enzyme rescued by the metal switch was the one in which they changed the C262 oxygen to a sulfur. Therefore, they concluded that this specific oxygen atom makes a critical contact with the magnesium ion. The strategy of atomic mutagenesis combined with metal ion rescue can be used to help understand the mechanism of other RNA and protein metalloenzymes.
|
0
|
PMC1184594
|
CC BY
|
2021-01-05 08:21:25
|
no
|
PLoS Biol. 2005 Sep 16; 3(9):e310
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030310
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030310SynopsisMolecular Biology/Structural BiologyBiochemistryIn VitroThe Ol' Switcheroo Shows How an RNA Enzyme Splices Itself Synopsis9 2005 16 8 2005 16 8 2005 3 9 e310Copyright: © 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.
Functional Identification of Catalytic Metal Ion Binding Sites within RNA
==== Body
Ribonucleic acid (RNA) is a dogma breaker. The “central dogma” of cellular biochemistry mandates that deoxyribonucleic acid (DNA) stores information, and RNA copies this information and uses it to direct the assembly of amino acid building blocks into proteins, such as enzymes. Enzymes catalyze important chemical reactions in the cell, such as the breakdown of glucose or the synthesis of urea.
When biochemists discovered catalytic RNA, they had to ditch the dogma. Because of its structure, it turns out, RNA can act as an enzyme and catalyze reactions. While two strands of DNA tend to zip up into the famous double helix, RNA usually goes solo. The single RNA strand folds back on itself to create myriad tangled arrangements. Some of these arrangements create an active center, the place on the RNA where the enzymatic magic happens. The many RNA enzymes and protein enzymes that use metal atoms to do their job are called metalloenzymes. One example of an important structural motif in RNA metalloenzymes is the group I intron, which can snip itself out of an RNA segment. Understanding exactly how the RNA and the metals interact will help to provide precise answers about how the enzyme really works.
Through X-ray crystallography, researchers have revealed many structural features of group I introns. But X-ray crystallography creates images of the enzyme frozen in time; it does not catch an enzyme in action. In a new study, Joseph Piccirilli, Daniel Herschlag, and colleagues discovered that a particular oxygen atom on a particular nucleotide in a group I RNA must bind to a particular magnesium ion in order for the reaction under study to proceed normally. The oxygen atom is known as the pro-SP phosphoryl oxygen at nucleotide C262 in the intron from the unicellular Tetrahymena thermophila protozoan.
Since there's no way to watch the oxygen and metal hook up during the reaction, how do the researchers know they do? The researchers used the powerful techniques of metal ion rescue and atomic mutagenesis. Here's how it worked. They figured out how well the group I intron reaction works with a normal enzyme. Then, they replaced the oxygen in question with a sulfur atom. The reaction didn't work as well because, by the rules of chemistry, sulfur doesn't like to bind to magnesium. But sulfur does like manganese and cadmium ions. So they replaced the magnesium with one of these other metal ions and measured the reaction. The researchers saw that these other metal ions restored (or “rescued”) enzymatic activity. In short, the enzyme needs a bond where the oxygen and the magnesium are, but the bond doesn't have to be between oxygen and magnesium.
As complicated as that is, plucking out one atom and trading it for another is itself a tricky business. Because most enzymes are made of stubborn amino acids and not nucleotides, atomic mutagenesis can be difficult. And usually when researchers have tried atomic mutagenesis, they've mutated the substrate (the molecule that the reaction acts upon) instead of the enzyme (the molecule that acts). Here, Piccirilli, Herschlag, and colleagues directed the applications of atomic mutagenesis to the molecule that does the work.
RNA enzymes called ribozymes require metal atoms to function. The site of metal-ribozyme interaction was studied by changing components of both the catalytic center of the ribozyme (the backbone) and its target substrate
To test that a specific oxygen in the intron binds to the magnesium ion, the researchers first had to compile a short list of potential atoms to which the magnesium might bind. By combining literature data from structural models and functional studies with a random sprinkling of sulfur atoms in the intron to find critical oxygen contacts, Piccirilli, Herschlag, and colleagues established a group of specific oxygen atoms to watch. They tried the metal rescue experiment with each of these oxygens, and the only enzyme rescued by the metal switch was the one in which they changed the C262 oxygen to a sulfur. Therefore, they concluded that this specific oxygen atom makes a critical contact with the magnesium ion. The strategy of atomic mutagenesis combined with metal ion rescue can be used to help understand the mechanism of other RNA and protein metalloenzymes.
|
0
|
PMC1184595
|
CC BY
|
2021-01-05 08:21:25
|
no
|
PLoS Biol. 2005 Sep 16; 3(9):e320
|
latin-1
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030320
|
oa_comm
|
==== Front
Ann Clin Microbiol AntimicrobAnnals of Clinical Microbiology and Antimicrobials1476-0711BioMed Central London 1476-0711-4-101598753310.1186/1476-0711-4-10ResearchDevelopment of a Multiple-Locus Variable number of tandem repeat Analysis (MLVA) for Leptospira interrogans and its application to Leptospira interrogans serovar Australis isolates from Far North Queensland, Australia Slack Andrew T [email protected] Michael F [email protected] Meegan L [email protected] Lee D [email protected] WHO/FAO/OIE Collaborating Centre for Reference & Research on Leptospirosis, Centre for Public Health Sciences, Queensland Health Scientific Services, Brisbane, Australia2005 30 6 2005 4 10 10 4 5 2005 30 6 2005 Copyright © 2005 Slack et al; licensee BioMed Central Ltd.2005Slack 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
Leptospirosis is a zoonotic disease caused by the genus, Leptospira. Leptospira interrogans is the most common genomospecies implicated in the disease. Epidemiological investigations are needed to distinguish outbreak situations or to trace reservoirs of the organisms. Current methodologies used for typing Leptospira have significant drawbacks. The development of an easy to perform yet high resolution method is needed for this organism.
Methods
In this study we have searched the available genomic sequence of L. interrogans serovar Copenhageni strain Fiocruz L1-130 for the presence of tandem repeats [1]. These repeats were evaluated against reference strains for diversity. Six loci were selected to create a Multiple Locus Variable Number of Tandem Repeats (VNTR) Analysis (MLVA) to explore the genetic diversity within L. interrogans serovar Australis clinical isolates from Far North Queensland.
Results
The 39 reference strains used for the development of the method displayed 39 distinct patterns. Diversity Indexes for the loci varied between 0.80 and 0.93 and the number of repeat units at each locus varied between less than one to 52 repeats. When the MLVA was applied to serovar Australis isolates three large clusters were distinguishable, each comprising various hosts including Rattus species, human and canines.
Conclusion
The MLVA described in this report, was easy to perform, analyse and was reproducible. The loci selected had high diversity allowing discrimination between serovars and also between strains within a serovar. This method provides a starting point on which improvements to the method and comparisons to other techniques can be made.
==== Body
Background
Leptospirosis, is the zoonotic disease caused by the spirochete Leptospira. Leptospirosis is characterised either as a febrile illness with sudden onset, or a 'flu like' illness. Patients present with chills, headaches, myalgia, and abdominal pain. In addition, patients may present with renal and pulmonary complications. It is considered an emerging infectious disease with large documented outbreaks occurring world-wide [2]. The majority of isolates detected in Queensland, Australia belong to the genomospecies, L. interrogans, with the dominant serovars being Zanoni or Australis. Typically Leptospirosis infections in Queensland are a result of occupational exposure with the majority of cases occurring in farming or animal based industries [3].
The genus Leptospira contains 17 genomospecies as shown by DNA-DNA hybridisation [4]. Under the current genotypic classification system, pathogenic and non pathogenic serovars may reside within the same genomospecies [2]. Many animals both domesticated and native serve as the reservoir for this bacterium. Leptospires are shed into the environment via the urine of these animals, where they survive in the soil or freshwater. Human infections result from contact with the contaminated soil/water or from direct contact with animals or their infectious body fluids [5].
Epidemiological investigation of this organism is important for the ability to distinguish individual cases from related cases such as in an outbreak situation. Also of importance is determining which animal is the likely common source of the infection, as containing or preventing the spread of the vector is paramount to controlling the disease.
Molecular typing methods have been described for L. interrogans including Randomly Amplified Polymorphic DNA (RAPD) [6-9], Pulsed Field Gel Electrophoresis (PFGE) [10], Arbitrarily Primed PCR (AP-PCR) [11,12] and most recently Fluorescent Amplified Fragment Length Polymorphism (FAFLP) [13]. Each of these methods have their disadvantages such as insufficient discriminatory power, poor inter-lab and intra-lab reproducibility, difficulties with database and dissemination of data [14]. They may also require specialised equipment such as DNA sequencers or contour-clamped homogeneous electric field electrophoresis systems. In addition leptospires have their own particular problems when using the above methods, the fastidious nature of the organism does not easily allow for the large volume of culture material required for PFGE and cultures can be quite prone to contamination with other bacteria, which may influence the accuracy of low stringency PCR methods or AFLP. As an alternative to the above methods, investigation of Variable Number of Tandem Repeats (VNTR) has been described for various organisms. These include Salmonella enterica [14,15], Staphylococcus aureus [16], Yersinia pestis [17], Mycobacterium tuberculosis [18], Francisella tularensis [19], Legionella pneumophila [20], Brucella spp. [6,21], Escherichia coli O157:H7 [22]and Borrelia spp[23]. VNTR are repeated DNA sequences of varying copy number. They are caused by slipped strand mispairing during DNA replication [24,25]. VNTRs can provide information relating to both the evolutionary and functional areas of bacterial diversity[25]. The ability to detect VNTRs in micro-organisms has been greatly enhanced by the availability of whole genomic sequences and software that can search for VNTR loci from these sequences. 1, [26,27] Once these polymorphisms are located, flanking primers can then be designed to amplify these variable length regions thus allowing differentiation of copy numbers using the size of the resultant amplicon. This can be done using standard agarose gel electrophoresis and if a higher resolution is required, fluorescent labelling and fragment sizing via a DNA sequencer can be used. VNTR is therefore applicable to a wide range of laboratories, including those which may have simple equipment such as thermal cyclers and agarose gel electrophoresis but do not have access to sophisticated equipment such as DNA sequencers. Furthermore when VNTR is applied to multiple loci as a typing scheme such as in Multiple Locus VNTR Analysis (MLVA) greater discriminatory power and more accurate determination of genetic relatedness is achieved [17,19,28,29]. Recently Majed et al [30]described a MLVA typing scheme for L. interrogans Sensu Stricto, this research highlighted the value of using VNTR as a typing scheme but was limited to the identification of serovar by comparing the size of the repeat to that of known reference strains. It should be noted that in that study several reference strains had identical VNTR patterns including serovars Australis and Bratislava, serovars Copenhageni and Icterohaemorragiae and serovars Romanica and Wolffi. It was subsequently validated against a small number of clinical isolates. In this paper, we report on the development of a MLVA scheme using novel VNTR loci selected from the sequence of a published L. interrogans genome [1] and evaluate its usefulness as a phylogenetic typing method using reference strains and clinical isolates from Far North Queensland, Australia.
Methods
Bacterial Strains
Thirty-nine reference strains of L. interrogans were obtained from the reference culture collection maintained by the WHO/FAO/OIE Collaborating Centre for Reference & Research on Leptospirosis, Brisbane, Australia (Table 1). In addition to the reference strains, ninety-eight isolates of L. interrogans serovar Australis were analysed [Additional File One]. These isolates were recovered from human and animal specimens. Human isolates were cultured using 0.2–0.5 mL of whole blood inoculated into 3 mL of semi solid Ellinghausen McCollugh Johnson Harris broth supplemented with 0.15% agar (EMJH, Difco lab, USA). These were then sub-cultured into EMJH broth within one week of receipt at the laboratory. Cultures were incubated for a further six weeks at 30°C and inspected weekly using dark ground microscopy. Positive cultures were identified using hyperimmune antisera and the Cross Agglutination Absorption Test (CAAT). 3 mm cubes of kidney or 100 μL of urine from rodents were inoculated into 3 mL semi solid EMJH media, these were incubated at 30°C for six weeks and inspected weekly using dark field microscopy. Positive cultures were identified as above. Once identified, isolates were stored in liquid nitrogen using EMJH media containing 2.5% dimethyl sulfoxide (DMSO).
Table 1 Leptospira interrogans reference strains used for VNTR loci selection.
Serogroup Serovar Strain Area of isolation Source
Australis Australis Ballico Australia Human
Pomona Pomona Pomona Australia Human
Sejroe Medanesis Hond HC Indonesia Dog
Icterohaemorrhagie Copenhageni M20 Denmark Human
Mini Swaijak Swaijak Australia Human
Sejroe Hardjo Hardoprajitno Indonesia Human
Icterohaemorrhagie Icterohaemorrhagie Ictero 1 Japan Human
Autumnalis Autumnalis Akiyami A Japan Human
Canicola Canicola Hond Utrecht IV Netherlands Dog
Australis Muenchen Munchen C90 Germany Human
Australis Fugis Fudge Malaysia Human
Australis Lora Lora Italy Human
Autumnalis Weerasinghe Weerasinghe Sri Lanka Human
Bataviae Bataviae Swart - -
Bataviae Paidjan Paidjan Indonesia Human
Canicola Benjamini Benjamin Indonesia Human
Canicola Binjei Binjei Indonesia Human
Canicola Broomi Patane Australia Human
Djasiman Djasiman Djasiman - -
Icterohaemorrhagie Gem Simon Sri Lanka Human
Pyrogenes Abramis Abraham Malaysia Human
Pyrogenes Biggis Biggs Malaysia Human
Pyrogenes Camlo Lt64-67 Vietnam Human
Sejroe Geyaweera Geyaweera Sri Lanka Human
Pyrogenes Zanoni Zanoni Australia Human
Pyrogenes Robinsoni Robinson Australia Human
Autumnalis Bangkinang Bangkinang 1 Indonesia Human
Autumnalis Carlos C3 Phillippines Toad
Autumnalis Mooris Moores Malaysia Human
Bataviae Losbanos LT101-69 Phillippines Rat
Canicola Sumneri Sumner Malaysia Human
Canicola Jonsis Jones Malaysia Human
Djasman Sentot Sentot Indonesia Human
Djasiman Gurungi Gurung Malaysia Human
Pyrogenes Guaratuba An7705 Brazil Opossum
Icterohaemorrhagie Smithi Smith Malaysia Human
Sejroe Wolffi 3705 Indonesia Human
Sejroe Ricardi Richardson Malaysia Human
Sejroe Haemolytica Marsh Malaysia Human
DNA Extractions
Once the cultures had reached a density equivalent to a 0.5 McFarland standard (1.5 × 108 cells/mL), cells were harvested by aspirating 500 μL of culture, centrifuged at 12,000 g for 4 minutes and resuspended in 200 μL of Phosphate Buffered Saline (PBS). DNA was extracted using the Roche High Pure Template kit as per manufacturer's instruction.
VNTR primer design
The two chromosomes of L. interrogans Copenhageni strain Fiocruz L1-130 deposited in GenBank under accession numbers, NC005823 and NC005824 were used to detect the VNTR Loci. Analysis using the Tandem Repeat Finder (TRF) program [1,26] was used to identify potential VNTR loci. Primer Premier 5.0 (Premier Biosoft) was used to design PCR primers for amplifying the loci. Primers were designed within the flanking regions, with a theoretical melting temperature of 57°C to 60°C (Table 2).
Table 2 PCR primers used in Study
Primer Name Direction Sequence (5'-3') Theoretical Melting Temperature (°C)
V8 Forward CAA GTG TTC GAC AAG AAT GAG 57.4
Reverse CTC ACC GGT AGA ACG CTC TTT T 58.4
V27 Forward TCG TCG GGT GAG CTA AAA TAT 57.0
Reverse TTC TTT CGG TGG CAA GG TTT 59.8
V29 Forward ATC GTT TTG GCA GTT TTT GCT 57.7
Reverse CTA GAA AAT TCC GCG TAG GG 57.2
V30 Forward AAG TAA GAT AGG TTC GGC GTT TA 57.9
Reverse ACT TGG GTG TTA ATC GCA AAA 57.7
V36 Forward TGG TTC TTG GGG TAA TTC TGT T 58.2
Reverse CTA CCA GGA GAT TAT CAA AAC GAA 57.9
V50 Forward CTT GTT GGA TCA CAA TAC GAA CTA TA 58.4
Reverse GGTAAGGGACAAAGTAAGTGAAGC 58.9
VNTR PCR amplification
PCR amplification of the VNTR loci was performed in a total volume of 50 μL containing 1X PCR Buffer II (Applied Biosystems, Foster City, Calif.), 2 mM MgCl2, 200 μM dNTP mix (Amersham Pharmacia Biotech, Piscataway, N.J), 10 pmol each of forward and reverse primer (Table 2), 1 unit of Amplitaq Gold (Applied Biosystems, Foster City, Calif.), 2 μL of the DNA preparation and double distilled water (ddH2O) making up the volume to 50 μL. The PCRs were run on a GeneAmp 9700 thermal cycler (Applied Biosystems, Foster City, Calif.). An initial denaturation at 95°C for 9 minutes, was followed by 35 cycles of a three step cycle protocol: 94°C for 30 seconds, 58°C for 60 seconds and 72°C for 60 seconds and a final extension of 72°C for 7 minutes. Each PCR product (15 μL) was resolved by electrophoresis (2 hours at 80 V) through a 2% agarose gel containing 0.5 ug ethidium bromide and buffered with 1X TBE (90 mM Tris-borate, 1 mM EDTA, pH 8). Allelic sizes were estimated using a 100 bp DNA plus Ladder (MBI Fermentas, Vilnius, Lithuania) as a size marker. Gels were visualised using UV transillumination and the images captured using the ChemiDoc XRS System (BioRad, Hercules, Calif.) (Figure 2).
Sequencing
The PCR products from eight selected references strains were sequenced using the same primers used to amplify the products. PCR product clean-up was performed using an enzyme digestion containing 1 μl of 10X Antarctic Phosphatase buffer (New England Biolabs, Beverly, Mass), 2 units of Exonuclease I (MBI Fermentas, Vilnius, Lithuania), 2 units of Antarctic Phosphatase (New England Biolabs, Beverly, Mass), 7 μL of PCR product and ddH2O to the final volume of 10 μL. This mix was incubated for 20 min at 37°C follow by 5 min at 80°C to inactivate the enzymes. Sequencing was performed using 1 μL of Big Dye Terminator V3.1 ready reaction mix (Applied Biosystems, Foster City, Calif.) with 7 μL of 2.5x Applied Biosystem sequencing dilution buffer, 3.2 pmol of primer, 3 μL of PCR product and ddH2O to the final volume of 20 μL. The thermal cycling was perform according to the manufacturer's instructions with the exception of increasing the cycles to 45. The sequencing products were cleaned up using sodium acetate-ethanol precipitation before being run on an ABI-373 sequencer. The sequences were aligned and analysed using Vector NTI Suite 9 (Invitrogen, Carlsbad, Calif.).
Data Analysis
Using the Quantity One 1D Analysis software package (BioRad, Hercules, Calif.), the agarose gel images were analysed and allelic sizes estimated. Allelic sizes were then converted into repeat copy numbers using Microsoft Excel software package [Additional File One], using the formula: Number of Repeats (bp) = [Fragment size (bp) – Flanking regions (bp)] / Repeat size (bp). The repeat copy numbers were then rounded down to form whole numbers. When repeat numbers were less than one, they were rounded down to zero, whilst no amplification was represented by the number ninety-nine. This created a numerical profile which was analysed as a character dataset using Bionumerics software package version 3.5 (Applied-Maths, Sint-Martens-Latern, Belgium). Clustering analysis was done using the categorical parameter and the Ward coefficient. Nei's Diversity Index of the VNTR loci was calculated from the range of alleles generated from the reference strains utilising the formula; D = 1-Σ(allele frequency)2 [31].
Results
Identification of VNTR markers
The Tandem Repeat Finder program identified 189 repeat motifs within the genome of L. interrogans serovar Copenhageni strain Fiocruz L1-130. 186 of the repeats were identified from chromosome 1 and only three were found in chromosome 2. 53 repeats were identified as being suitable for further analysis based upon the size of the repeat, number of repeat units present and also whether the sequence was conserved within the repeats. Preliminary testing against the reference strains identified 25 loci out of the 53 that were polymorphic between different serovars. The remaining 25 loci either failed to amplify any DNA or were amplified but were monomorphic. A subset of the L. interrogans serovar Australis clinical isolates that were considered geographically unrelated was used to determine whether the 25 selected loci were also polymorphic within a serovar. Six loci were found to contain variable repeat copy numbers within a serovar and were then re-applied to the 39 reference strains. Amplification of the six loci was possible from the 39 reference strains tested with the exceptions of locus V8 from serovar Djasiman, locus V27 from serovar Swaijak and Robinsoni, locus V29 from serovars Swaijak, Canicola, Broomi, Robinsoni and Jonsis. Amplification was also not possible for locus V36 from serovars Munchen and Fugis also for Locus V50 in serovars Lora and Geyaweera. PCR amplification was attempted three times for these serovars, no amplicons were detected at each attempt. The different allele sizes were caused by the loss or addition of repeat units confirmed by the sequencing of the PCR products. Sequence data was entered into GenBank [GenBank: DQ023538 – DQ023553].
For the 39 reference strains the number of repeats in the six loci varied between no repeats in all loci up to 52 repeats in the V36 locus. The number of alleles per locus varied between six in V27 and nineteen in V36. The diversity index ranged from the lowest of 0.80 in V27 to 0.93 in V36. (Table 3)
Table 3 Characteristic of the Six VNTR loci
Loci Repeat Motif Repeat size (bp)a Total Flanking Regions (bp)a Repeat range (min-max) No. of alleles Diversity (D)
8 GGAAAACTCAACACAA
CGCTCTTTATGAATCG
CGTT 36 124 0–16 14 0.88
27 TTGTGGGAACTCTTAC
AATTTGAGATTTTACA
GTAAAACTTGGAAGTT
GTGGGAACTCTTACAA
TTTGAGATTTTACAGTA
AAACTTGGAAATTGTG
GGAACTCTTACAACTT
GAGATTTTACAGTGGG
ACTTTGAAG 138 183 0–4 6 0.80
29 GATTTTACAGTTAGAC
TTTGAAATTGTGGGAA
CTCCCACGGATTTGG 47 90 0–17 14 0.92
30 TCCCACATATTCAAGA
TTAAACTGTAAAATTGT
GATTTGTGGTAGT 46 228 0–12 12 0.88
36 CTTAGACTTTGTGTGA
GTTCCCACATTTTAAA
GTAAAA 38 161 0–52 19 0.93
50 AAAATGTAGGAACTAC
CACAAACACTGACTTT
ACAGATAAATTCTC 46 106 012 9 0.83
L. interrogans reference strains clustering analysis
Clustering analysis (Figure 1) positioned the reference strains into three large clusters. These clusters each contained a diverse selection of serovars with no bias towards the grouping of serogroups together, with the exception of serovar Iceterohaemorrhagie strain Ictero 1, Copenhageni strain M20, serovar Hardjo strain Hardjoprajitno and Haemolytica strain Marsh. Both of these pairs of serovars belong to the same serogroups: Icterohaemorrhagie and Sejroe respectively.
Figure 1 Leptospira interrogans reference strains: clustering analysis using MLVA Data. Clustering analysis was done using the categorical and ward options using Bionumerics software package version 3.5 (Applied-Maths, Sint-Martens-Latern, Belgium).
L. interrogans serovar Australis clustering analysis
To further evaluate the VNTR loci selected for the MLVA, the typing scheme was applied to 98 isolates of L. interrogans serovar Australis. All six VNTR loci were amplified from all of these clinical isolates, which varied in geography of isolation and host. Clustering analysis [Additional File Two and Additional File Three] revealed three major clusters, each containing several sub-groups. Three clusters contained a mix of Rattus species and human whilst two contained canine isolates. Two of the major clusters show significant geographical distributions towards the two main townships in the area: Tully and Innisfail, whilst the remaining major cluster was more diverse in geographical distributions. As the isolates were taken over a limited timeframe, not surprisingly there was no discernable pattern in regards to the introduction or extinction of strains over time in the area.
Discussion
The MLVA assay presented was easy to perform and analyse, as it consisted of six individual PCR reactions and agarose gel electrophoresis. Selected reference strain isolates were run in tandem during the initial evaluation by different individuals to assess reproducibility, each time they displayed identical fragment sizes as determined by sequencing and agarose gel electrophoresis (data not shown). Dilutions of the template DNA were used to evaluate whether the fragment sizes were template concentration dependant. Dilutions of 1:10 and 1:100 did not effect fragment size but did result in reduced PCR product yield (data not shown). The MLVA assay was proven to be reproducible under varying laboratory conditions. The two limitations of this assay are firstly the use of agarose gel electrophoresis to separate fragments for allelic sizing, due to inherent inaccuracies of this method to size bands of close molecular weights given that the resolution is dependant on agarose composition and concentration and secondly in rounding partial repeat copy numbers to the nearest whole number to make the data analysis easier, isolates that had partial repeats were treated as if they contained whole repeats. As illustrated in Additional file 1, by simplifying the repeat copy numbers we have artificially reduced the resolution of the method and its ability to distinguish between closely related strains that may only vary by a partial repeat at one or more of the loci.
The diversity index calculated for each locus suggests that the loci selected are of highly polymorphic nature and therefore have greater discriminatory power between similar strains than loci with a lower diversity indexes would have. Whilst it has been previously reported in organisms such as Francisella tularensis [19] and Yersinia pestis [29] that higher copy repeat numbers may confer higher allelic variability, it was not demonstrated with this study. This may be due to the loci having similar repeat copy sizes (36–47 bp) with the exception of locus V27 with a repeat size of 138 bp. The lack of amplification from loci V8, V27, V29, V36 and V50 from certain reference strains may be due to sequence diversity in the flanking regions up or downstream from the repeat regions or the lack of the VNTR loci all together. Further investigations using isolates of these serovars is needed to determine whether there is sufficient diversity in the remaining loci for it to be valid as a typing method. In the study by Majed et al. [30], they noted from the dendrogram that the isolates had clustered into distinct global geographical regions. Interestingly in this study, we found that the dendrogram showed no bias towards this global geographical clustering of reference strains.
The different genomospecies of Leptospira were not used in the selection and development of the VNTR loci for this typing scheme. The basis for this decision was that the other genomospecies are considered to be significantly different from L. interrogans based upon DNA-DNA Hybridisation [32,33] and may not possess the same primer binding sites or indeed the same VNTR loci. In addition due to the use of a serologically based Leptospiral taxonomy system, several serovars belong to more than one genomospecies [32,33], thus it would be doubtful that the MLVA typing scheme described in this article would be useful in determining the genomospecies of these related serovars due to the significant genetic differences[34]
When the MLVA assay was applied to L. interrogans serovar Australis isolates collected from 1995 to 2004, the six selected loci appeared to show less diversity. This could be due to the fact that the isolates were taken from within a limited geographical area of Far North Queensland. Despite this apparent limited diversity, the phylogenetic analysis revealed several large albeit weakly linked clusters. All of these clusters contained a mixture of hosts, and it would be possible to speculate that the transmission of serovar Australis to humans in that area is via the native rodent population indigenous to that area. Another possible risk whilst not proven in Australia, is that transmission of the organism may occur via canines to humans [2]. Indeed in this study the two strains isolated from canines are genetically similar to two strains isolated from humans. Also of interest, the reference culture for serovar Australis; strain Ballico which was isolated from a patient in North Queensland in 1934, shows homology with LT958 and QHR371A a human and Rattus sordidus isolate respectively, both from the Tully region. Further investigations using MLVA of isolates, could add further detail to the depth of knowledge into the population of serovar Australis and to determine other possible transmission sources to humans.
Further improvements to this method are possible to increase both practicality and discriminatory power for typing of L. interrogans isolates. These improvements could include the use of fluorescently labelled primers and fragment analysis using a DNA sequencer to accurately assign repeat sizes. Multiplexing of the six targets would rationalise the number of PCR reactions needed to complete the MLVA and would also decrease costs in terms of reagents and labour. Improvements that could be introduced to the analysis of the data may include using an allele designation system as described by Lindstedt et al [34]. In addition to improving the method, comparisons between other molecular typing methods such as FAFLP or a sequence based typing scheme, would ultimately determine the validity of the MLVA assay as molecular epidemiology tool for L. interrogans.
The method has potential application in furthering the understanding of Leptospiral molecular epidemiology. As this method can be performed without specialised equipment, a broader range of laboratories including those in developing countries could potentially use this scheme as part of their isolate typing. This method was also easily standardised within our laboratory, with multiple users and different thermal cyclers employed to achieve the same results. This level of standisation at an inter laboratory level would allow the transfer of the method into another laboratory more effectively than that of a method that was operator or equipment specific. The simplification of MLVA data into a concise and portable numerical format as suggested in this article makes it easier to be comprehended by non-technical staff such as public health authorities. In addition, the format of allele data is similar to the allele string that is used for Multiple Locus Sequence Typing (MLST). As an alternative to using Bionumerics, software freely available on the internet such as Sequence Type Analysis and Recombinational Tests (START) [35] could be used for the phylogenetic analysis. Bionumerics was ultimately selected over START due to the advanced features of Bionumerics including evolutionary and population modelling.
Further assessment of L. interrogans isolates globally is required to confirm that the selected VNTR loci possess sufficient diversity to be used a typing scheme on an international level.
Conclusion
We have developed a novel MLVA typing scheme which is simple, robust reproducible and cost effective. The six VNTR loci chosen for this assay showed a high level of diversity between reference strains. When this method was applied to a collection of clinical isolates, it was possible to observe distant relationships between suspected reservoirs and humans. This method provides a starting point for further investigations into the molecular epidemiology of L. interrogans infections.
Competing interests
The author(s) declare that they have no competing interests
Authors' contributions
AS had primary responsibility for study design, conducting the typing work and preparation of the manuscript. MD provided laboratory support by culturing and maintaining culture collections. MS provided laboratory support by performing serological identification of isolates. LS had intellectual contributions. All authors have read and approved the final manuscript.
Figure 2 PCR products from the six selected VNTR loci of various L. interrogans reference strains. PCR products electrophoresised through a 2% agarose gel. M: 100 bp DNA Ladder plus (MBI Fermentas, Vilnius, Lithuania); 1, Serovar Zanoni strain Zanoni; 2, Serovar Autmnalis strain Akiyami A; 3, Serovar Canicola strain Hond Utrecht IV; 4, Serovar Pomona strain Pomona; 5, Serovar Hardjo strain Hardjoprajitno; 6, Serovar Muenchen strain Muenchen C90; 7, Serovar Weerasinghe Strain Weerasinghe; 8, Serovar Paidjan strain Paidjan; 8, Serovar Biggis strain Biggs; 9, Serovar Bangkinang strain Bangkinang 1; 10, Serovar Jonsis strain Jones.
Supplementary Material
Additional File 1
Characteristics of Leptospira interrogans reference strains and serovar australis strains and allelic profiles for all strains tested.
Click here for file
Additional File 2
Leptospira interrogans serovar Australis: clustering analysis of MLVA data part one.
Click here for file
Additional File 3
Leptospira interrogans serovar Australis: clustering analysis of MLVA data part two.
Click here for file
Acknowledgements
The authors wish to acknowledge Mr Shane Byrne for reviewing the manuscript.
==== Refs
Nascimento AL Ko AI Martins EA Monteiro-Vitorello CB Ho PL Haake DA Verjovski-Almeida S Hartskeerl RA Marques MV Oliveira MC Menck CF Leite LC Carrer H Coutinho LL Degrave WM Dellagostin OA El-Dorry H Ferro ES Ferro MI Furlan LR Gamberini M Giglioti EA Goes-Neto A Goldman GH Goldman MH Harakava R Jeronimo SM Junqueira-de-Azevedo IL Kimura ET Kuramae EE Lemos EG Lemos MV Marino CL Nunes LR de Oliveira RC Pereira GG Reis MS Schriefer A Siqueira WJ Sommer P Tsai SM Simpson AJ Ferro JA Camargo LE Kitajima JP Setubal JC Van Sluys MA Comparative genomics of two Leptospira interrogans serovars reveals novel insights into physiology and pathogenesis J Bacteriol 2004 186 2164 2172 15028702 10.1128/JB.186.7.2164-2172.2004
Levett PN Leptospirosis Clin Microbiol Rev 2001 14 296 326 11292640 10.1128/CMR.14.2.296-326.2001
Smythe LBLSMDMHC National Leptospirosis surveillance report number 12 January-December 2003 2003 , WHO/FAO/OIE Collaborating Centre for Reference & Research on Leptospirosis, Queensland Health Scientific Services (QHSS), Brisbane, Australia.
Bharti AR Nally JE Ricaldi JN Matthias MA Diaz MM Lovett MA Levett PN Gilman RH Willig MR Gotuzzo E Vinetz JM Leptospirosis: a zoonotic disease of global importance Lancet Infect Dis 2003 3 757 771 14652202 10.1016/S1473-3099(03)00830-2
Woo TH Patel BK Smythe LD Symonds ML Norris MA Dohnt MF Identification of pathogenic Leptospira genospecies by continuous monitoring of fluorogenic hybridization probes during rapid-cycle PCR J Clin Microbiol 1997 35 3140 3146 9399509
Corney BG Colley J Djordjevic SP Whittington R Graham GC Rapid identification of some Leptospira isolates from cattle by random amplified polymorphic DNA fingerprinting J Clin Microbiol 1993 31 2927 2932 8263177
Corney BG Colley J Graham GC Simplified analysis of pathogenic leptospiral serovars by random amplified polymorphic DNA fingerprinting J Med Microbiol 1997 46 927 932 9368533
Gerritsen MA Smits MA Olyhoek T Random amplified polymorphic DNA fingerprinting for rapid identification of leptospiras of serogroup Sejroe J Med Microbiol 1995 42 336 339 7752212
Ramadass P Meerarani S Venkatesha MD Senthilkumar A Nachimuthu K Characterization of leptospiral serovars by randomly amplified polymorphic DNA fingerprinting Int J Syst Bacteriol 1997 47 575 576 9103653
Herrmann JL Bellenger E Perolat P Baranton G Saint Girons I Pulsed-field gel electrophoresis of NotI digests of leptospiral DNA: a new rapid method of serovar identification J Clin Microbiol 1992 30 1696 1702 1629323
Roy S Biswas D Vijayachari P Sugunan AP Sehgal SC A 22-mer primer enhances discriminatory power of AP-PCR fingerprinting technique in characterization of leptospires Trop Med Int Health 2004 9 1203 1209 15548317 10.1111/j.1365-3156.2004.01322.x
Perolat P Merien F Ellis WA Baranton G Characterization of Leptospira isolates from serovar hardjo by ribotyping, arbitrarily primed PCR, and mapped restriction site polymorphisms J Clin Microbiol 1994 32 1949 1957 7989548
Vijayachari P Ahmed N Sugunan AP Ghousunnissa S Rao KR Hasnain SE Sehgal SC Use of fluorescent amplified fragment length polymorphism for molecular epidemiology of leptospirosis in India J Clin Microbiol 2004 42 3575 3580 15297500 10.1128/JCM.42.8.3575-3580.2004
Ramisse V Houssu P Hernandez E Denoeud F Hilaire V Lisanti O Ramisse F Cavallo JD Vergnaud G Variable Number of Tandem Repeats in Salmonella enterica subsp. enterica for Typing Purposes J Clin Microbiol 2004 42 5722 5730 15583305 10.1128/JCM.42.12.5722-5730.2004
Liu Y Lee MA Ooi EE Mavis Y Tan AL Quek HH Molecular typing of Salmonella enterica serovar typhi isolates from various countries in Asia by a multiplex PCR assay on variable-number tandem repeats J Clin Microbiol 2003 41 4388 4394 12958274 10.1128/JCM.41.9.4388-4394.2003
Sabat A Krzyszton-Russjan J Strzalka W Filipek R Kosowska K Hryniewicz W Travis J Potempa J New method for typing Staphylococcus aureus strains: multiple-locus variable-number tandem repeat analysis of polymorphism and genetic relationships of clinical isolates J Clin Microbiol 2003 41 1801 1804 12682193 10.1128/JCM.41.4.1801-1804.2003
Adair DM Worsham PL Hill KK Klevytska AM Jackson PJ Friedlander AM Keim P Diversity in a variable-number tandem repeat from Yersinia pestis J Clin Microbiol 2000 38 1516 1519 10747136
Frothingham R Meeker-O'Connell WA Genetic diversity in the Mycobacterium tuberculosis complex based on variable numbers of tandem DNA repeats Microbiology 1998 144 ( Pt 5) 1189 1196 9611793
Farlow J Smith KL Wong J Abrams M Lytle M Keim P Francisella tularensis strain typing using multiple-locus, variable-number tandem repeat analysis J Clin Microbiol 2001 39 3186 3192 11526148 10.1128/JCM.39.9.3186-3192.2001
Pourcel C Vidgop Y Ramisse F Vergnaud G Tram C Characterization of a tandem repeat polymorphism in Legionella pneumophila and its use for genotyping J Clin Microbiol 2003 41 1819 1826 12734211 10.1128/JCM.41.5.1819-1826.2003
Bricker BJ Ewalt DR Halling SM Brucella 'HOOF-Prints': strain typing by multi-locus analysis of variable number tandem repeats (VNTRs) BMC Microbiol 2003 3 15 12857351 10.1186/1471-2180-3-15
Noller AC McEllistrem MC Pacheco AG Boxrud DJ Harrison LH Multilocus variable-number tandem repeat analysis distinguishes outbreak and sporadic Escherichia coli O157:H7 isolates J Clin Microbiol 2003 41 5389 5397 14662916 10.1128/JCM.41.12.5389-5397.2003
Farlow J Postic D Smith KL Jay Z Baranton G Keim P Strain typing of Borrelia burgdorferi, Borrelia afzelii, and Borrelia garinii by using multiple-locus variable-number tandem repeat analysis J Clin Microbiol 2002 40 4612 4618 12454161 10.1128/JCM.40.12.4612-4618.2002
Bzymek M Lovett ST Instability of repetitive DNA sequences: the role of replication in multiple mechanisms Proc Natl Acad Sci U S A 2001 98 8319 8325 11459970 10.1073/pnas.111008398
van Belkum A Scherer S van Alphen L Verbrugh H Short-sequence DNA repeats in prokaryotic genomes Microbiol Mol Biol Rev 1998 62 275 293 9618442
Benson G Tandem repeats finder: a program to analyze DNA sequences Nucleic Acids Res 1999 27 573 580 9862982 10.1093/nar/27.2.573
Keim P Price LB Klevytska AM Smith KL Schupp JM Okinaka R Jackson PJ Hugh-Jones ME Multiple-locus variable-number tandem repeat analysis reveals genetic relationships within Bacillus anthracis J Bacteriol 2000 182 2928 2936 10781564 10.1128/JB.182.10.2928-2936.2000
Klevytska AM Price LB Schupp JM Worsham PL Wong J Keim P Identification and characterization of variable-number tandem repeats in the Yersinia pestis genome J Clin Microbiol 2001 39 3179 3185 11526147 10.1128/JCM.39.9.3179-3185.2001
Majed Z Bellenger E Postic D Pourcel C Baranton G Picardeau M Identification of variable-number tandem-repeat loci in Leptospira interrogans sensu stricto J Clin Microbiol 2005 43 539 545 15695642 10.1128/JCM.43.2.539-545.2005
Weir BS Genetic data analysis: methods for discrete population data analysis. 1990 Sunderland, Mass., Sinauer Associates, Inc. 150 156
Feresu SB Steigerwalt AG Brenner DJ DNA relatedness of Leptospira strains isolated from beef cattle in Zimbabwe Int J Syst Bacteriol 1999 49 Pt 3 1111 1117 10425768
Brenner DJ Kaufmann AF Sulzer KR Steigerwalt AG Rogers FC Weyant RS Further determination of DNA relatedness between serogroups and serovars in the family Leptospiraceae with a proposal for Leptospira alexanderi sp. nov. and four new Leptospira genomospecies Int J Syst Bacteriol 1999 49 Pt 2 839 858 10319510
Lindstedt BA Vardund T Aas L Kapperud G Multiple-locus variable-number tandem-repeats analysis of Salmonella enterica subsp. enterica serovar Typhimurium using PCR multiplexing and multicolor capillary electrophoresis J Microbiol Methods 2004 59 163 172 15369852 10.1016/j.mimet.2004.06.014
Jolley KA Feil EJ Chan MS Maiden MC Sequence type analysis and recombinational tests (START) Bioinformatics 2001 17 1230 1231 11751234 10.1093/bioinformatics/17.12.1230
|
15987533
|
PMC1185519
|
CC BY
|
2021-01-04 16:38:21
|
no
|
Ann Clin Microbiol Antimicrob. 2005 Jun 30; 4:10
|
utf-8
|
Ann Clin Microbiol Antimicrob
| 2,005 |
10.1186/1476-0711-4-10
|
oa_comm
|
==== Front
BMC BiochemBMC Biochemistry1471-2091BioMed Central London 1471-2091-6-121601880310.1186/1471-2091-6-12Research ArticleCharacterization of 17α-hydroxysteroid dehydrogenase activity (17α-HSD) and its involvement in the biosynthesis of epitestosterone Bellemare Véronique [email protected] Frédérick [email protected] Rock [email protected] Van [email protected] Oncology and Molecular Endocrinology Research Center Laval University Medical Center (CHUL) 2705 Laurier Boulevard Quebec, (Quebec) G1V 4G2, Canada2005 14 7 2005 6 12 12 2 2 2005 14 7 2005 Copyright © 2005 Bellemare et al; licensee BioMed Central Ltd.2005Bellemare 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
Epi-testosterone (epiT) is the 17α-epimer of testosterone. It has been found at similar level as testosterone in human biological fluids. This steroid has thus been used as a natural internal standard for assessing testosterone abuse in sports. EpiT has been also shown to accumulate in mammary cyst fluid and in human prostate. It was found to possess antiandrogenic activity as well as neuroprotective effects. So far, the exact pathway leading to the formation of epiT has not been elucidated.
Results
In this report, we describe the isolation and characterization of the enzyme 17α-hydroxysteroid dehydrogenase. The name is given according to its most potent activity. Using cells stably expressing the enzyme, we show that 17α-HSD catalyzes efficienty the transformation of 4-androstenedione (4-dione), dehydroepiandrosterone (DHEA), 5α-androstane-3,17-dione (5α-dione) and androsterone (ADT) into their corresponding 17α-hydroxy-steroids : epiT, 5-androstene-3β,17α-diol (epi5diol), 5α-androstane-17α-ol-3-one (epiDHT) and 5α-androstane-3α,17α-diol (epi3α-diol), respectively. Similar to other members of the aldo-keto reductase family that possess the ability to reduce the keto-group into hydroxyl-group at different position on the steroid nucleus, 17α-HSD could also catalyze the transformation of DHT, 5α-dione, and 5α-pregnane-3,20-dione (DHP) into 3α-diol, ADT and 5α-pregnane-3α-ol-20-one (allopregnanolone) through its less potent 3α-HSD activity. We also have over-expressed the 17α-HSD in Escherichia coli and have purified it by affinity chromatography. The purified enzyme exhibits the same catalytic properties that have been observed with cultured HEK-293 stably transfected cells. Using quantitative Realtime-PCR to study tissue distribution of this enzyme in the mouse, we observed that it is expressed at very high levels in the kidney.
Conclusion
The present study permits to clarify the biosynthesis pathway of epiT. It also offers the opportunity to study gene regulation and function of this enzyme. Further study in human will allow a better comprehension about the use of epiT in drug abuse testing; it will also help to clarify the importance of its accumulation in breast cyst fluid and prostate, as well as its potential role as natural antiandrogen.
==== Body
Background
Epitestosterone (17α-hydroxy-4-androstene-3-one) is an epimer of testosterone (T). Its concentration in the urine is used as reference substance in the control of T abuse [1]. EpiT was identified for the first time as an androgen metabolite produced by rabbit liver slices [2]. It has been also observed that slices of rabbit, guinea pig and dog liver, mare's ovary, ox and sheep blood as well as guinea pig kidney, ovary and testis possess the ability to produce epiT from T and 4-dione [3]. In the mouse, the kidney is a major site of epiT formation, while the production in the liver is negligible [4]. In castrated male bovine, it has been observed that blood and liver both possess a high ability to convert T into epiT [5]. No interconversion of T and epiT has been observed in testes of bulls, rabbits or rats, even though it has been found that testes is a source of endogenous epiT in these species [3].
In young boys, the concentration of epiT is higher than T, however it declines in adulthood to a epiT/T ratio of approximately 1 [6]. In hyperplasic prostate, epiT concentration is comparable to that of 4-dione, which represents about twice the amount of T and half the concentration of dihydrotestosterone (DHT) [7]. The excretion of epiT in urine is slightly lower than that of T [8-11]. Plasma concentrations of epiT decline with age and were established at approximately 2.5 nmol/l in adult men and 1.2 nmol/l in women. Since epiT does not originate from endogenous T, and because the ratio of urinary T to epiT in adults is almost constant, this ratio has been used as a basis for the detection of exogenously administered T : the median ratio in normal healthy men is about 1, while being significantly elevated in case of testosterone abuse. Urinary T/epiT ratio from 1 to 6 is considered normal by the International Olympic Committee, while one displaying a ratio greater than 6 is suspected of anabolic steroid use [12]. In order to apply this assay, it is important to assume that the biosynthesis of epitestosterone does not originate from testosterone, and that both epimers undergo similar clearance. It is also important to assume that racial or individual variations do not affect this ratio. Because of these important assumptions, parameters that may influence this ratio and possibly lead to false positive results have been intensively debated since the introduction of the T/epiT ratio in doping analysis [13]. The lack of knowledge about the gene encoding the enzyme responsible for the formation of epiT has made the identification of appropriate parameters difficult.
In humans, the interconversion of T and epiT is negligible [14]. Using labeled-T, it has been shown that epiT does not originate from T. Studying the 16-ene-synthetase reaction in human testicular homogenates, Weusten et al. [15] hypothesized that 5-androstene-3β,17α-diol (epi5-diol) and 5,16-androstadien-3β-ol are synthesized from pregnenolone in a single step through a 16-ene-synthase, and then that epi5-diol is further converted to epiT by 3β-hydroxysteroid dehydrogenase (3β-HSD). On the other hand, a recent study showed that administration of 4-dione to healthy male subjects increases the urinary excretion rates of epiT, thus suggesting that epiT could be biosynthesized from 4-dione [16]. The authors suggested that a likely candidate to convert 4-dione to epiT would be 17α-hydroxysteroid dehydrogenase (17α-HSD). This enzyme has been purified from liver and kidney tissue from rabbits and hamsters [17-19]. Accordingly, in this report, we describe the isolation and characterization of a cDNA encoding an enzyme exhibiting a strong 17α-HSD acivity, efficiently catalyzing the conversion of 4-dione into epiT.
Results
Characterization of mouse 17α-HSD cDNA and gene
The mouse genome project has made available the sequence of a cluster containing eight members of the aldo-keto reductase family located in the chromosome 13 [20]. Using specific primers, we have cloned a cDNA fragment containing the entire coding region of the gene identified as AKR1C21 without notification of any activity. Sequence analysis of the cDNA fragment shows that it encodes a putative protein of 323 amino acids having a calculated MW of 36745 Daltons. We have deposited the sequence in GenBank under the accession number AY742217. Comparison of the deduced amino acid sequence with that of other aldo-keto reductase members (Figure 1) shows that mouse 17α-HSD shares 77, 70 and 72 % amino acid sequence identity with mouse type 5 17β-HSD, 3α-HSD and 20α-HSD respectively. It also shares 70, 69, 71, 72, 72, 73, 74 % with rat 3α-HSD and 20α-HSD, rabbit 20α-HSD as well as human 3α-HSD1, 17β-HSD5, 3α-HSD3 and 20α-HSD. The genomic structure as derived from public data bank indicates that mouse 17α-HSD gene is 12.5 kb long covering 9 exons and is transcribed into a 1.2 kb mRNA.
Figure 1 Alignment of the amino acid sequence of mouse 17α-HSD with those of related enzymes. Amino acid sequence of mouse 17α-HSD was aligned with sequences of mouse (m), rat (r), rabbit (rb) and human (h) enzyme members of the AKR1c subfamily. The corresponding name according to the nomenclature for aldo-keto reductase (AKR) family members are: m17α-HSD, AKR1C21; m3α-HSD, AKR1C14; m20α-HSD, AKR1C18; m17β-HSD5, AKR1C6; r3α-HSD, AKR1C9; r20α-HSD, AKR1C8; h3α-HSD1, AKR1C4; h3α-HSD3, AKR1C2; h17β-HSD5 or 3α-HSD2, AKR1C3; and h20α-HSD, AKR1C1. Amino acids are given in conventional single letter code and numbered on the right. Dashes and dots, respectively, represent identical and missing amino acids.
Identification of epiT by HPLC
Preliminary data using [14C]4-dione as substrate showed that the product resulting from the transformation of 4-dione by the enzyme overexpressed in HEK-293 cells is different from T, the metabolite produced by a 17β-HSD activity. To verify the identity of this metabolite, we used HPLC to analyze extracts of culture medium of HEK-293 cells over expressing 17α-HSD in presence of [14C]4-dione (Figure 2B). We also analyzed metabolites obtained after [14C]4-dione incubation with the pure enzyme (Figure 2C). Comparison of the HPLC elution and TLC migration profile of the product with a commercial epiT standard clearly indicates that the product is epitestosterone, the 17α-epimer of testosterone. Using TLC separation, we were also able to verify the identity of epi3α-diol, which is produced from ADT, in addition of epitestosterone In order to confirm the 17α-HSD nature of the activity, comparison of the products obtained from incubation of 4-dione (Figure 3A) and ADT (Figure 3B) with 17β-HSD type 5 and 17α-HSD was accomplished.
Figure 2 Identification by HPLC of epiT produced from HEK-293 cells stably expressing 17α-HSD. (A) Elution profile of non labeled steroids; 4-dione (1st peak), testosterone (2nd peak) and epitestosterone (3rd peak). (B) Products extracted from enzymatic assay done with cells stably expressing 17α-HSD, substrate is 4-dione. (C) Products extracted from enzymatic assay done with purified enzyme; substrate is 4-dione. Separation and identification of metabolites were performed as described in Materials and Methods.
Figure 3 Identification of the 17α-HSD activity by TLC. A- Incubation of [14C]-4-dione with cells stably expressing mouse 17β-HSD type 5 (1) and 17α-HSD (2) activity. Standards of 4-dione (3), T (4) and epiT (5) have been deposited and co-migrated. B- Incubation of [14C]-ADT with cells cells stably expressing mouse 17β-HSD type 5 (1) and 17α-HSD (2) activity. Standards of ADT (3), 3α-diol (4) and epi3α-diol (5) have been deposited and co-migrated.
Substrate specificity of the mouse 17α-HSD activity
Using 17α-HSD stably expressed in HEK-293 cells, we have characterized the substrate specificity of this enzyme in cultured cells by comparison with non transfected cells. As illustrated in Figure 4 and Table 1, mouse 17α-HSD possesses two activities. The strongest activity is 17α-reductase activity responsible for the transformation of 4-dione into epiT as well as the transformation of ADT, 5α-dione and DHEA into epi3α-diol, epiDHT and epi5-diol, respectively. The second activity is 3α-reductase, responsible for the transformation of DHT to 3α-diol, 5α-dione to ADT and DHP to allopregnanolone. Moreover, we found that purified enzyme displays the same catalytic activity as the overexpressed enzyme in intact cells.
Figure 4 Substrate specificity of mouse 17α-HSD activity of HEK-293 cells stably transfected with pCMVneo-m17α-HSD. The experiments were performed using HEK-293 cells stably expressing 17α-HSD in culture. 0,1 μM of the indicated [14C]-and [3H]-labeled steroid was added to culture medium for one hour. Testo, conversion of testosterone to 4-dione and vice-versa; E1, conversion of estrone to estradiol and vice-versa; DHEA, conversion of dehydroepiandrosterone to 5-diol; 5α-dione, conversion of androstanedione to androsterone (ADT) and vice-versa; DHT, conversion of dihydrotestosterone to 3α-diol and vice-versa; DHP, conversion of dihydroprogesterone to allopregnanolone and vice-versa; Preg and Prog, conversion of pregnenolone and progesterone to 20α-OHPreg and 20α-OHProg. The error bar indicates mean ± SEM of triplicate assays. Incubation, extraction, separation and quantification were performed as described in Materials and Methods.
Table 1 Kinetic constants of 17α-HSD for various substrates
Km (μM) Vmax (μmol/min) Vmax / Km
4-dione 0,4 0,3 0,7
ADT 0,9 0,3 0,3
DHEA 3,5 0,4 0,1
DHT 2,0 0,7 0,3
Tissue distribution of mouse 17α-HSD
Using Q_RT-PCR to quantify mRNA expression levels of 17α-HSD in male and female (Table 2) mouse tissues, we found that the expression of this enzyme is highly specific for the mouse kidney, while type 5 17β-HSD, the enzyme catalyzing the formation of T, is markedly expressed in the liver. Both enzymes are expressed at very high levels, more than 15 millions copies/μg total RNA. Since the estimated amount of total RNA in a single liver cell is about 50 pg [21], we consider that an expression level of 20,000 copies/μg total RNA corresponds to approximately 1 copy/cell in an homogenous cell population. Therefore, 15 millions copies/μg total RNA should correspond to 750 copies/cell.
Table 2 Quantification of mRNA expression levels of 17α-HSD and 17β-HSD5 in various mouse tissues using Realtime PCR
17α-HSD (number of copies/μg RNA ± SEM) 17β-HSD5 (number of copies/μg RNA ± SEM)
male female male female
Pituitary gland 130 ± 92 692 ± 251 572 ± 992 2986 ± 2591
Adrenal 0 ± 0 12036 ± 356 53445 ± 11410 3714 ± 1960
Liver 2344 ± 6 332 ± 73 15369159 ± 2701597 14995082 ± 987906
Kidney 22226386 ± 1068282 17083578 ± 928740 45536 ± 9194 102586 ± 35774
Spleen 21610 ± 1369 7466 ± 438 22791 ± 4358 8777 ± 5673
Lung 0 ± 0 0 ± 0 0 ± 0 6097 ± 2642
Thymus 2029 ± 40 3790 ± 134 407 ± 705 10943 ± 7200
Stomach 0 ± 0 0 ± 0 11858 ± 4062 8539 ± 4606
Colon 13860 ± 2184 6299 ± 1049 256686 ± 37628 17438 ± 21120
Heart 0 ± 0 0 ± 0 0 ± 0 6821 ± 4402
Testis 6333 ± 490 - 0 ± 0 -
Prostate 0 ± 0 - 1095 ± 1896 -
Preputial gland 2681 ± 261 - 14551 ± 12605 -
Ovary - 1980 ± 120 - 7208 ± 6397
Uterus - 0 ± 0 - 8328 ± 3819
Clitoral gland - 6516 ± 337 - 14419 ± 12591
Mammary gland - 0 ± 0 - 5666 ± 3188
Discussion
The present report describes the isolation and characterization of a cDNA sequence encoding the enzyme 17α-HSD. We have shown that this enzyme converts mainly 17-ketosteroids into 17α-hydroxysteroids : 4-dione to epiT, ADT to epi3α-diol, 5α-dione to epiDHT and DHEA to epi5-diol. Since the enzyme is able to catalyze 4 distincts 17-ketosteroids into its 17α-hydroxy-compounds, and because all the four products have shown identitical profile on HPLC and TLC with commercial products, the results, thus, represent a strong evidence that the activity catalyzed by this specific enzyme is a 17-ketoreductase activity producing 17α-hydroxysteroids.
17α-HSD belongs to the aldo-keto reductase family and shares 77 % amino acid with the mouse type 5 17β-HSD, an enzyme catalyzing the transformation of 4-dione into T [22]. To our knowledge, this is the first example of two highly homologous enzymes belonging to the same gene family, and catalyzing the formation of two distinct epimers from a same substrate. This will represent an interesting model to study the mechanism of the 17α- and 17β- stereospecificity. In addition to the difference in substrate specificity, mouse 17α-HSD and 17β-HSD5 show distinct and specific mRNA tissue distribution : 17α-HSD is highly expressed in kidney while 17β-HSD5 is abundant in the liver. This is in agreement with previous reports showing that mouse liver does not produce epiT [3].
We have previously shown that mouse type 5 17β-HSD is specifically expressed in the liver [22], while human type 5 17β-HSD is more widely expressed [23]. It is noteworthy that 17α-HSD and type 5 17β-HSD belongs to aldo-keto reductase family. Members of this family share very high homology although they catalyze different activities and are expressed in different tissues; for example, mouse type 5 17β-HSD shares 77, 88 and 86 % identity with mouse 17α-HSD, 3α-HSD and 20α-HSD respectively. Therefore, previous studies based on hybridation experiments such as northern blot analysis could not distinguish different members of the aldo-keto-reductase family. Using Realtime-PCR with specific primers, as described in the present manuscript, mouse type 5 17β-HSD could be specifically identified, distinctly from 17α-HSD.
Our results show that 17α-HSD is able to convert 4-dione to epiT as well as DHEA to epi5α-diol, and thus suggest that epiT could be produced through two different pathways involving 17α-HSD (Figure 5). However, the catalytic efficiency of 3β-HSD and 17α-HSD for DHEA and 4-dione respectively, strongly suggest that the main pathway leading to the formation of epiT is the conversion of DHEA by 3β-HSD and the 4-dione product being further converted into epiT. In a previous study described by Weusten et al. [15], it was shown that a non-negligible quantity of epi5-diol is produced by 16-ene-synthase activity, the reaction catalyzing the transformation of pregnenolone into 5,16-androstadien-3β-ol. The authors thus suggested that epi5-diol could be produced directly from pregnenolone, and epiT being produced by the subsequent conversion of epi5-diol by 3β-HSD. In contrast, our data strongly suggests that the epi5-diol originates from the transformation of pregnenolone to DHEA by P450c17, followed by the conversion of DHEA into epi5-diol by the enzyme 17α-HSD. Previously, we have shown that P450c17 possesses two activities, 17α-hydroxylase/17-20 lyase and 16-ene synthase, that are able to convert pregnenolone into DHEA and 5,16-androstadien-3β-ol, respectively [24]; however, no epi5-diol has been detected.
Figure 5 Diagram illustrating the 2 putative pathways for the conversion of DHEA to epiT via 17α-HSD and 3β-HSD. The thickness of the arrows indicates the relative importance of each pathway
The present first report of 17α-HSD will help to better understand the physiological role of epiT. It offers a tool to further study at molecular level the role of 17α-HSD in human, espectially for T abuse testing in sports using epiT as a control. It also permits to investigate the importance of epitestosterone in some previously reported interesting phenomenons such as its accumulation in mammary cyst fluid [25] and in human prostate [7,26], as well as its potential neuroprotective effects [27], and natural anti-androgenic effects [28].
Methods
Isolation of mouse 17α-HSD
A cDNA fragment of a coding region of mouse 17α-HSD (AKR1C21) was amplified by polymerase chain reaction (PCR) from a mouse spleen cDNA and the oligoprimer pair (5'-ggg-gtc-gac-ttt-gaa-gag-gga-cac-ata-atg-a-3' and 5'-ggg-ggt-acc-acc-cat-agg-ctt-ttc-agg-aga-3') derived from the DNA sequence NM_029901 from GenBank database. Mouse spleen cDNA was obtained by reverse transcription of 20 μg of mouse spleen total RNA using 400 U SuperScript II reverse transcriptase (Invitrogen, Burlington, Ontario, Canada) and oligo-d(T)24 as primer in a reaction buffer containing 50 mM Tris-HCl pH 8.3, 75 mM KCl, 3 mM MgCl2, 10 mM DTT and 0.5 mM dNTPs. The resulting cDNA fragments were subcloned into a pCMVneo expression vector (pCMVneo-m17α-HSD) which was subsequently used to produce a stably transfected HEK-293 cell line. Plasmid DNA was prepared using the Qiagen Mega Kit (Qiagen, Chatsworth, CA, USA). Sequence of the pCMVneo-m17α-HSD was determined using an ABI 3730/XL automatic sequencer, to verify the identity of the amplified sequence.
Stable expression in HEK-293 cells
Stable transfection of pCMVneo-m17α-HSD into HEK-293 cells was performed as described previously [29]. Briefly, HEK-293 cells were cultured in 6-well falcon flasks to approximately 3 × 105 cells/well in Minimum Essential Medium (MEM) (Invitrogen) supplemented with 10% (vol/vol) FCS (Wisent, Saint-Bruno, Québec, Canada) at 37°C under a 95% air- 5% CO2 humidified atmosphere. Five μg of pCMVneo-m17α-HSD was transfected using Exgen 500 reagent (Fermentas, Burlington, Ontario, Canada). After 6 h incubation at 37°C, the transfection medium was removed and 2 ml of MEM were added. Cells were further cultured for 48 h, then transferred into 10 cm Petri dishes and cultured in MEM containing 700 μg/ml of G-418 (Invitrogen) in order to inhibit the growth of non-transfected cells. The medium containing G-418 was changed every two days until resistant colonies were observed.
Overexpression and purification of mouse 17α-HSD
The cDNA encoding mouse 17α-HSD was subcloned into a pGEX vector (Amhersham Biosciences, Baie d'Urfé, Québec, Canada) and expressed in Escherichia coli BL21(DE3) pLysS as a fusion protein with glutathione-S-transferase (GST). The fusion protein was isolated using a Glutathione-Sepharose 4B column, as described by the manufacturer. The purified 17α-HSD enzyme was separated from GST by digestion with thrombin. 17α-HSD is not adsorbed on the DEAE column and is recuperated in the flow-through fraction, while GST and fusion protein remain on the column. With this method we obtain about 10 mg of a purified enzyme preparation per 100 ml of cell culture. Analysis of samples obtained during the purification process was performed using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), as described before [30]. The broad range molecular weight standards was purchased from Bio-Rad (Missisauga, Ontario, Canada).
Assay of enzymatic activity
Enzymatic activities of 17α-HSD were determined using both purified enzyme and the cultured HEK-293 cells stably transfected with pCMVneo-m17α-HSD, as previously described [29]. Briefly, 3 μg of purified 17α-HSD were incubated with 10 mM NADPH, 0.1 μM of 4-dione in phosphate saline buffer, 50 mM, pH 7.3, for 20 minutes. For the intact cells, 0.1 μM of the [14C]-labeled steroid (PerkinElmer, Boston, Massachussetts, USA) was added to freshly changed culture medium in a 6-well culture plate. Non-transfected HEK-293 cells were used as control of the background. After incubation, the steroids were extracted with 2 ml of ether. The organic phases were pooled and evaporated to dryness. The steroids were then solubilized in 50 μl of dichloromethane, applied to Silica gel 60 thin layer chromatography (TLC) plates (Merck, Darmstad, Germany). To obtain a better separation and identification of metabolites, different solvent systems were used. Metabolites of substrates 4-dione, DHEA and 5α-dione were separated in chloroform : ether (9:1), while DHP, DHT and ADT products were separated in the toluene : acetone (4:1) solvent system. Substrates and metabolites were identified by comparison with reference steroids, revealed by autoradiography and quantified using the PhosphorImager System (Molecular Dynamics, Sunnyvale, California, USA). The enzymatic reaction was carried out using the condition in which the activity varies linearly with the enzyme concentration and incubation time, indicating that the cofactor concentration produced by the cells is in excess; the reverse reaction was consequently prevented. In our conditions, this linearity was observed at even more than 60% transformation. Determination of the kinetic parameters was done by Lineweaver-Burk graph analysis using Enzfitter software.
Identification of epitestosterone by High Performance Liquid Chromatography (HPLC)
14C-Labeled steroids were analyzed using Waters NovaPak reverse-phase C18 HPLC column (3.9 × 150 mm, 4 μm). The mobile phase was MeOH/H2O/THF (26 : 56 : 18 v/v), with a flow rate of 0.7 ml.min-1. Radioactivity was monitored in the eluent using Beckman 171 HPLC Radioactivity Monitoring System. Unlabelled steroids (4-dione, DHEA, 5α-dione, ADT, T, 5-diol, DHT, 3α-diol, epiT, epi5-diol, epiDHT and epi3α-diol) were obtained from Steraloids (Newport, Rhode Island, USA) and used as standards.
Tissue collection and RNA preparation
Total RNA of indicated tissues was isolated using Trizol Reagent (Invitrogen, Burlington, Ontario) as described by the manufacturer. Twenty μg of total RNA was converted to cDNA by incubation at 42°C for 1 h with 400 U SuperScript II reverse transcriptase (Invitrogen), using oligo-d(T)24 as primer in a reaction buffer containing 50 mM Tris-HCl pH 8.3, 75 mM KCl, 3 mM MgCl2, 10 mM DTT and 0.5 mM dNTPs. The tissues were collected in C57BL6 mice at 12-15 weeks of age obtained from Charles River, Inc. (Saint-Constant, Québec, Canada). The mice were housed individually in vinyl cages. The photoperiod was 12 h of light and 12 h of darkness (lights on at 07:15 h). Certified rodent food (Lab Rodent Diet) and tap water were provided ad libitum. The experiment was conducted in an animal facility approved by the Canadian Council on Animal Care (CCAC) and the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). The study was performed in accordance with the CCAC Guide for Care and Use of Experimental Animals. The collected organs were rapidly trimmed, snap-frozen in liquid nitrogen and stored at -80°C until RNA extraction.
Tissue distribution of 17α-HSD mRNA using RealTime PCR
Total RNA from pituitary gland, adrenal, liver, kidney, spleen, thymus, stomach, heart, lung, ovary, uterus, clitoral gland, mammary gland, testis, prostate and preputial gland, prepared as described above, were analyzed for the expression of 17α-HSD mRNA using quantitative RealTime RT-PCR (Q_RTPCR). cDNA corresponding to 20 pg of the initial total RNA was used to perform fluorescent-based Realtime PCR quantification using the LightCycler Realtime PCR apparatus (Roche Inc. Nutley, NJ). Reagents were obtained from the same company and were used as described by the manufacturer. The conditions for the PCR reactions were: denaturation at 95°C for 10 sec, annealing at 62°C for 5 sec and elongation at 72°C for 8 sec. Oligoprimer pairs (5'-ttg-att-gcc-ctt-cgc-tac-cag-3', 5'-aaa-tgg-cag-cag-gta-tgt-atc-gc-3') allowed the amplification of approximately 170 bp of the mouse 17α-HSD sequence. Data calculation and normalization was performed using second derivative and double correction method as previously described [31]. 17α-HSD mRNA expression levels are expressed as number of copies/μg total RNA using a standard curve of Cp versus logarithm of the quantity. The standard curve was established using known cDNA amounts of 0, 102, 103, 104, 105 and 106 copies of cDNA and a LightCycler 3.5 program provided by the manufacturer (Roche Inc).
Abbreviations
T testosterone
epiT epitestosterone
4-dione androstenedione
epi5-diol 5-androstene-3α,17α-diol
ADT androsterone
DHEA dehydroepiandrosterone
5α-dione androstanedione
DHP 5α-pregnane-3,20-dione
DHT 5α-dihydrotestosterone
epi3α-diol 5α-androstane-3α,17α-diol
5α-pregnane-3α-ol-20-one allopregnanolone
3α-diol 5α-androstane-3α,17α-diol
5-diol 5-androstene-3β,17α-diol
epiDHT 5α-androstane-3-one-17α-ol
HSD hydroxysteroid dehydrogenase
PCR polymerase chain reaction
Authors' contributions
VB has participated in the design of the study and in redaction of the manuscript; she has carried out the molecular biology manipulations, all enzymatic assays, and everything surrounding the culture of the cells. FF has taken care of the entire process of enzyme's purification. RB has participated in the conception of the study, especially the purified enzyme part. VLT conceived the study and was implicated in the redaction of the article. All authors read and approved final manuscript.
Figure 6 SDS-PAGE of fractions obtained during the purification process of 17α-HSD. MW, molecular weight standards; CE, cell extract of 100000 g; FP, fusion protein; DP, protein obtained after digestion with thrombin; 17α-HSD, purified.
Acknowledgements
We would like to thank Canadian Institutes of Health Research (CIHR) for financial support. Lucille Lacoste and Mélanie Robitaille are acknowledged for their skillful technical assistance.
==== Refs
Kuoppasalmi K Karjalainen U Tehunen R Doping analysis in Helsinki 1983 Clinical Chemistry Research Foundation Publications 1984 Helsinki, Painotalo Miktor 32 35
Clark LC Kochakian CD The in vitro metabolism of testosterone to 4-androstenedione-3,17 cis-testosterone and other steroids by rabbit liver slices Journal of biological chemistry 1947 170 22 23
Starka L Epitestosterone J Steroid Biochem Mol Biol 2003 87 27 34 14630088 10.1016/S0960-0760(03)00383-2
Arimasa N Kochakian CD Epitestosterone and 5alpha-androstane-3alpha,17beta-diol: the characteristic metabolites of androst-4-ene-3,17-dione produced by mouse kidney in vitro Endocrinology 1973 92 72 82 4734142
Martin RP Fecal metabolites of testosterone-4-14C in the bovine male castrate Endocrinology 1966 78 907 913 5935505
Lapcik O Hampl R Hill M Starka L Plasma levels of epitestosterone from prepuberty to adult life J Steroid Biochem Mol Biol 1995 55 405 408 8541237 10.1016/0960-0760(95)00181-6
Starka L Hampl R Hill M Lapcik O Bilek R Petrik R Epitestosterone in human blood and prostatic tissue Eur J Clin Chem Clin Biochem 1997 35 469 473 9228331
De Nicola AF Dorfman RI Forchielli E Urinary excretion of epitestosterone and testosterone in normal individuals and hirsute and virilized females Steroids 1966 7 351 366 5960067 10.1016/0039-128X(66)90106-1
Bilek R Hampl R Putz Z Starka L Radioimmunoassay of epitestosterone: methodology, thermodynamic aspects and applications J Steroid Biochem 1987 28 723 729 3695520 10.1016/0022-4731(87)90404-3
France JT Knox BS Urinary excretion of testosterone and epitestosterone in hirsutism Acta Endocrinol (Copenh) 1967 56 177 187 6072685
Longhino N Tajic M Vedris M Jankovic D Drobnjak P Urinary excretion of androstenedione, testosterone, epitestosterone and dehydroepiandrosterone during the normal menstrual cycle Acta Endocrinol (Copenh) 1968 59 644 651 4240704
Donike M Barwald KR Klostermann K Schanzer W Zimmermann J Heck H, Hollmann W and Liesen H Nachweis von exogenem testosteron in sport Leistung und Gesendheit 1983 Kohl, Germany, Deutscher Azte-Verlag 293 298
Dehennin L Peres G Plasma and urinary markers of oral testosterone misuse by healthy men in presence of masking epitestosterone administration Int J Sports Med 1996 17 315 319 8858400
Dray F Ledru MJ [Metabolism of epitestosterone. Absence of peripheral interconversion of epitestosterone and testosterone and existence of a production of epitestosterone sulfate in normal adult men] C R Acad Sci Hebd Seances Acad Sci D 1966 262 679 681 4955903
Weusten JJ Legemaat G van der Wouw MP Smals AG Kloppenborg PW Benraad T The mechanism of the synthesis of 16-androstenes in human testicular homogenates J Steroid Biochem 1989 32 689 694 2739409 10.1016/0022-4731(89)90513-X
Catlin DH Leder BZ Ahrens BD Hatton CK Finkelstein JS Effects of androstenedione administration on epitestosterone metabolism in men Steroids 2002 67 559 564 11996927 10.1016/S0039-128X(02)00005-3
Kochakian CD 17 alpha and 17 beta-oxidoreductases of adult female hamster liver and kidney J Steroid Biochem 1982 17 541 546 6960220 10.1016/0022-4731(82)90013-9
Lau PC Layne DS Williamson DG Comparison of the multiple forms of the soluble 3(17) alpha-hydroxysteroid dehydrogenases of female rabbit kidney and liver J Biol Chem 1982 257 9450 9456 6955303
Hasnain S Williamson DG Properties of the multiple forms of the soluble 17alpha-hydroxy steroid dehydrogenase of rabbit liver Biochem J 1977 161 279 283 849262
Vergnes L Phan J Stolz A Reue K A cluster of eight hydroxysteroid dehydrogenase genes belonging to the aldo-keto reductase supergene family on mouse chromosome 13 J Lipid Res 2003 44 503 511 12562828 10.1194/jlr.M200399-JLR200
Manual LC LightCycler presentation 2002 Quebec, Roche Molecular Biochemicals 32
Rheault P Charbonneau A Luu-The V Structure and activity of the murine type 5 17beta-hydroxysteroid dehydrogenase gene(1) Biochim Biophys Acta 1999 1447 17 24 10500239
Dufort I Rheault P Huang XF Soucy P Luu-The V Characteristics of a highly labile human type 5 17beta-hydroxysteroid dehydrogenase Endocrinology 1999 140 568 574 9927279 10.1210/en.140.2.568
Soucy P Lacoste L Luu-The V Assessment of porcine and human 16-ene-synthase, a third activity of P450c17 Eur J Biochem 2003 270 1349 1357 12631293 10.1046/j.1432-1033.2003.03508.x
Bicikova M Szamel I Hill M Tallova J Starka L Allopregnanolone, pregnenolone sulfate, and epitestosterone in breast cyst fluid Steroids 2001 66 55 57 11090659 10.1016/S0039-128X(00)00140-9
Hill M Bilek R Safarik L Starka L Analysis of relations between serum levels of epitestosterone, estradiol, testosterone, IGF-1 and prostatic specific antigen in men with benign prostatic hyperplasia and carcinoma of the prostate Physiol Res 2000 49 Suppl 1 S113 8 10984080
Hammond J Le Q Goodyer C Gelfand M Trifiro M LeBlanc A Testosterone-mediated neuroprotection through the androgen receptor in human primary neurons J Neurochem 2001 77 1319 1326 11389183 10.1046/j.1471-4159.2001.00345.x
Nuck BA Lucky AW Epitestosterone: a potential new antiandrogen J Invest Dermatol 1987 89 209 211 3598212 10.1111/1523-1747.ep12470564
Huang XF Luu-The V Molecular characterization of a first human 3(alpha-->beta)-hydroxysteroid epimerase J Biol Chem 2000 275 29452 29457 10896656 10.1074/jbc.M000562200
Laemmli UK Cleavage of structural proteins during the assembly of the head of bacteriophage T4 Nature 1970 227 680 685 5432063 10.1038/227680a0
Van LT Paquet N Calvo E Cumps J Improved real-time RT-PCR method for high-throughput measurements using second derivative calculation and double correction Biotechniques 2005 38 287 293 15727135
|
16018803
|
PMC1185520
|
CC BY
|
2021-01-04 16:26:24
|
no
|
BMC Biochem. 2005 Jul 14; 6:12
|
utf-8
|
BMC Biochem
| 2,005 |
10.1186/1471-2091-6-12
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-621596703310.1186/1471-2407-5-62Research ArticlePrognostic impact of epidermal growth factor receptor (EGFR) expression on loco-regional recurrence after preoperative radiotherapy in rectal cancer Azria David [email protected] Frederic [email protected] Nicolas [email protected] Abderrahim [email protected] Claire [email protected] Philippe [email protected] Marc [email protected] Pierre [email protected] Mahmut [email protected]èlegrin André [email protected] Jean-Bernard [email protected]èzenas Simon [email protected] Department of Radiation Oncology, Val d'Aurelle Cancer Institute, Montpellier, France2 INSERM, EMI 0227, Val d'Aurelle Cancer Institute, Montpellier, France3 Department of Pathology, Val d'Aurelle Cancer Institute, Montpellier, France4 Department of Radiation Oncology, Centre Hospitalier Universitaire Vaudois, Lausanne, Switzerland5 Department of Surgical Oncology, Val d'Aurelle Cancer Institute, Montpellier, France6 Department of Medical and Digestive Oncology, Val d'Aurelle Cancer Institute, Montpellier, France7 Biostatistics Unit, Val d'Aurelle Cancer Institute, Montpellier, France2005 20 6 2005 5 62 62 17 12 2004 20 6 2005 Copyright © 2005 Azria 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
Epidermal growth factor receptor (EGFR) represents a major target for current radiosensitizing strategies. We wished to ascertain whether a correlation exists between the expression of EGFR and treatment outcome in a group of patients with rectal adenocarcinoma who had undergone preoperative radiotherapy (RT).
Methods
Within a six-year period, 138 patients underwent preoperative radiotherapy and curative surgery for rectal cancer (UICC stages II-III) at our institute. Among them, 77 pretherapeutic tumor biopsies were available for semi-quantitative immunohistochemical investigation evaluating the intensity and the number (extent) of tumor stained cells. Statistical analyses included Cox regression for calculating risk ratios of survival endpoints and logistic regression for determining odds ratios for the development of loco-regional recurrences.
Results
Median age was 64 years (range: 30–88). Initial staging showed 75% and 25% stage II and III tumors, respectively. RT consisted of 44-Gy pelvic irradiation in 2-Gy fractions using 18-MV photons. In 25 very low-rectal-cancer patients the primary tumor received a boost dose of up to 16 Gy for a sphincter-preservation approach. Concomitant chemotherapy was used in 17% of the cases. All patients underwent complete total mesorectal resection. Positive staining (EGFR+) was observed in 43 patients (56%). Median follow-up was 36 months (range: 6–86). Locoregional recurrence rates were 7 and 20% for EGFR extent inferior and superior to 25%, respectively. The corresponding locoregional recurrence-free survival rate at two years was 94% (95% confidence interval, CI, 92–98%) and 84% (CI 95%, 58–95%), respectively (P = 0.06). Multivariate analyses showed a significant correlation between the rate of loco-regional recurrence and three parameters: EGFR extent superior to 25% (hazard ratio = 7.18, CI 95%, 1.17–46, P = 0.037), rectal resection with microscopic residue (hazard ratio = 6.92, CI 95%, 1.18–40.41, P = 0.032), and a total dose of 44 Gy (hazard ratio = 5.78, CI 95%, 1.04–32.05, P = 0.045).
Conclusion
EGFR expression impacts on loco-regional recurrence. Knowledge of expression of EGFR in rectal cancer could contribute to the identification of patients with an increased risk of recurrences, and to the prediction of prognosis.
==== Body
Background
In patients with rectal carcinoma, pelvic recurrence is a major source of morbidity and mortality. Despite improvements in surgical approaches, local recurrence may occur in up to 30% of patients treated with surgery including total mesorectal excision [1]. Since 2001, the Dutch Colorectal Cancer Group Trial [2] has confirmed that a short course of radiotherapy (RT) reduced the rate of pelvic recurrence at 3 years, from 10.1% to 3.4%. In addition, a meta-analysis of 19 randomized trials including preoperative RT tends to show that it provides a gain of three percent at 5 years in overall survival [3]. However despite these recent intensive clinical investigations, there is still a need to develop novel strategies in the management of patients with locally advanced rectal cancer.
Advances in the understanding of the molecular biology of rectal cancer have opened many new research directions. Increasing effort has been directed towards developing molecular targeted therapies or searching for molecular markers that are useful either in predicting treatment outcome or in selecting patients for specific molecular targeted therapies, based on particular tumor characteristics. None of the recent studies has identified convincing data to warrant routine clinical application of any marker such as p53 [4,5], or apoptosis regulators [6].
To date, no data have become available that shed light on the impact of EGFR expression on local and distant relapse in patients treated with preoperative RT and extensive local surgery i.e. abdominoperineal excision or low anterior resection with total mesorectal excision. We present here the prognostic impact of EGFR expression on locoregional recurrence in 77 patients treated with preoperative RT at our institute.
Methods
Patient selection and pretreatment evaluation
Within a six-year period (April 1996 and September 2002), 138 patients underwent preoperative radiotherapy and curative surgery for rectal cancer (UICC stages II-III) at the Val d'Aurelle Cancer Institute of Montpellier, France. A carcinoma was considered a primary rectal carcinoma if it was located in the lower third (<6 cm from the anal verge), middle third (6–12 cm), and upper third of the rectum (above 12 cm). Pretherapy biopsies were available for analysis in 77 patients and were evaluable for the statistical results. Diagnostic and distant disease extension studies consisted of colorectal endoscopy with biopsies, rectal ultrasonography (uT), presurgical carcinoembryonic antigen (CEA) value, abdominal and pelvic computed tomography (CT) scans, chest X-ray or CT-scan and routine laboratory studies. All patients were metastasis-free at diagnosis.
EGFR immunohistochemical assay (IHC)
IHC of the tumor biopsies was performed by using the Dako autostainer (DakoCytomation, Glostrup, Denmark) and the EGFR Pharm Dx kit® K 1494 (Dako Cytomation, Glostrup, Denmark), according to the manufacturer's instructions with the reagents supplied with the kit. Briefly, sections of 3 μm were mounted on silanized slides and allowed to dry overnight at 37°C. After deparaffinization and rehydratation, slides were incubated with 3% hydrogen peroxide solution for 5 min. After a washing procedure with the supplied buffer, tissue sections were covered for 5 min with protein K solution. The slides were then incubated for 30 min with the primary mouse anti-EGFR MAb (clone 2-18C9), which binds to a formalin-resistant epitope near the ligand-binding site on the extra cellular domain of the EGFR and recognizes both wild type and mutant type (vIII). After two rinses in buffer, the slides were incubated with the detection system for 30 min (labeled polymer-HRP). Tissue staining was visualized with a DAB substrate chromogen solution. Slides were counterstained with hematoxylin, dehydrated, and mounted. Negative control sections were processed without the primary antibody but with an irrelevant murine IgG1 supplied with the kit. Negative and positive control cell slides provided with the kit EGFR Pharm Dx® were also used, to ensure that each assay run was performed appropriately and according to protocol specifications. Furthermore, perineurium was considered as a positive internal control on tumor slides. EGFR assessment was realized according to the EGFR Pharm Dx® scoring guidelines. Results were reported as positive when a complete or incomplete circumferential membrane staining was observed in at least 1% of the tumor cells. Staining was defined as any IHC staining of tumor cell membranes above background level, i.e., weak, moderate, or strong. Absence of or cytoplasm staining was reported as negative. In addition to these standardized criteria, the pathologist performed a semi-quantitative evaluation reporting both intensity and percentage (extent) of tumor cells staining blinded to clinical data (Figure 1).
Preoperative radiation therapy (RT) and surgical modalities
Patients were treated in supine position with a 3-field (posterior and two opposed laterals) isocentric technique using 18-MV photon beams daily, five times a week. The daily dose at the isocenter was 2 Gy; the total dose to the entire pelvis was 44 Gy. In 25 very low-rectal-cancer patients, primary tumor received a boost dose of up to 16 Gy for a sphincter-preservation approach. Clinical target volume (CTV) included the tumor and the entire rectum, the anterior wall of the sacrum and the posterior wall of the prostate or vagina, and the following lymph nodes: perirectal, presacral, hypogastric, obturator, and low common iliac nodes. The planning target volume (PTV) included the clinical target volume plus a 1–1.5-cm margin. The superior margin was the L5-S1 interspace in most patients; in some patients with tumors very close to the anus, however, the cranial margin was placed somewhat lower, but always at least 5 cm above the tumor area. The lateral margins were 1 cm outside the bony margins of the true pelvis. The posterior margin was placed just posterior to the sacrum. The anterior margin was dependent on the anterior extension (gross tumor volume, GTV) of the primary tumor. Individually shaped blocks were used to shield normal tissues. The boost volume covered the primary tumor plus a 1.5-cm margin using a 3-field (posterior and oblique) technique. Standard or CT-scan simulation was used. With the CT-scan simulator (Picker PQ 2000 + ACQSIM), GTV, CTV, and PTV were determined as defined above, the treated volume and the irradiated volume according to the ICRU report 50 [7]. Fields were marked during initial CT-scan simulation after the ICRU reference was calculated.
Thirteen patients (17%) received concomitant chemotherapy. In eight patients, the chemotherapy regimen consisted of continuous infusion of 200–250 mg/m2/day of 5-fluorouracil (5-FU) alone beginning on the first day of radiation therapy, five days a week for 5 weeks. Oxaliplatin (40 mg/m2/day at days 1, 8, 15, 22, and 29) and leucovorin (100 mg/m2/day at days 1–2, 15–16, and 29–30) were added at the same protocol of 5-FU for two and three patients, respectively.
Median time between the last day of radiotherapy and surgery was 41 days (range: 13–97). The choice of the surgical procedure was at the surgeon's discretion. In all cases, the entire mesorectum was removed. Specimens were inked for lateral margin determination. R1 resection was defined as lateral clearance less than one mm.
Clinical, operative, and histopathological data were recorded prospectively in a computerized registry database including patient age, gender, tumor site, tumor stage according to UICC stage [8], histological differentiation, gross morphology, tumor size, local invasion, nodal status and type of surgery.
Follow-up
All patients were seen on regular follow-up including clinical history, physical examination, laboratory investigations, abdominal ultrasonography, chest X-ray, and endoscopy (sigmoidoscopy after 6 months, total colonoscopy after one year). They were followed semi-annually during the period of 2–5 years postoperatively until death or the closing date of the study (July 2004). Any regrowth of tumor within the pelvis was considered as a local recurrence. The diagnosis of a pelvic recurrence was preferably proven by histology and/or cytology; however, in the majority of cases, the diagnosis was made on clinical or radiological grounds. Data collected were entered prospectively into the registry database. Median follow-up of all patients was 36 months (range: 6–86 months).
Statistical methods
The characteristics of EGFR staining were examined for correlation with tumor- and patient-related prognostic factors. The cut-off of 25% of EGFR staining corresponded to the third quartile of EGFR extent and was then selected for all statistical correlations. Categorical variables were reported by means of contingency tables. Furthermore, for continuous variables the median and range were computed.
To investigate the association between trial features, univariate statistical analyses were performed using Pearson's Chi-2 test or Fisher's exact test when applicable.
Survival times to all events were measured from the day of surgery to the time of the event or to last news if no event occurred. Relapse-free survival (RFS; event was all relapse), locoregional recurrence-free survival (LRFS; event was locoregional recurrence), and distant metastasis-free survival (MFS; event was distant metastasis relapse) rates were estimated according to the Kaplan-Meier method. Patients not presenting the event of interest were considered censored at the last known follow up of time. Survival curves were drawn, and the logrank test was performed to assess differences between the groups.
Cox's proportional hazards regression using a stepwise selection procedure was used to investigate prognostic factors. Hazards ratios with 95% confidence interval, CI, are presented.
All P values reported were two-sided, and differences were considered as significant at the 5% level. Data were analyzed with software STATA 7.0 (Stata Corporation, College Station, TX, USA).
Results
EGFR expression
The semi-quantitative analysis of EGFR expression is summarized in Table 1. Fifty-six percent of the cases demonstrated EGFR expression, and 44% had negative staining. EGFR staining extent superior to 25% was observed in 26% of the cases, and the staining intensity was graded as strong in 8%. Strong staining intensity occurred statistically more frequently in those cases with EGFR extent ≥25% than in those with <25% (P = 0.018).
EGFR and clinical characteristics of the study population
A total of 77 patients were evaluable for EGFR expression. Median age was 64 years (range: 30–88). Twenty-six (34%) were female and 19 (25%) were staged as stage III patients. A majority of patients presented T3/T4 (72%) rectal tumor. Initial tumor was located in the lower third (n = 52, 68%), middle third (n = 17, 22%), and upper third (n = 8, 10%) of the rectum. Twenty-five patients (32%) received a total radiation dose of 60 Gy for a sphincter-conserving approach. Thirteen patients (17%) and 25 patients (32%) received preoperative concomitant chemo-radiotherapy and adjuvant chemotherapy, respectively. Microscopic incomplete surgery (R1) was achieved in 7 patients (9%), and all corresponded to initially T4 tumors.
We compared the distribution of patients and tumor characteristics and treatment according to EGFR expression (staining intensity and extent) to assess the presence of potential imbalances in the known prognostic variables. Table 2 shows no significant differences between the groups in the distribution of known clinical prognostic indicators of loco-regional control and survival, i.e., age, gender, stage group, tumor location, preoperative total dose RT, concomitant chemo-radiotherapy, type of surgery, resection margins. Neither was any imbalance observed for patients who received adjuvant chemotherapy.
EGFR expression and relapse
Overall tumor progression, caused by local recurrence alone (n = 1, 1.3%), distant metastases alone (n = 8, 10.4%), and both of them (n = 7, 9.1%) occurred in 16 patients (20.8%).
Patients with EGFR extent ≥25% had a higher locoregional recurrence rate (20% vs 7%). The two-year LRFS rate was 94% (92–98%) in patients with EGFR extent <25% and 84% (58–95%) in patients with EGFR extent ≥25% with a borderline statistical difference (P = 0.06, Figure 2).
An EGFR extent of ≥25% was associated with poorer MFS (84% [59–95%] vs 95% [84–98%]) but the difference did not achieve statistical significance. Metastatic evolution corresponded to lung, liver, peritoneum, bone, and brain in 60%, 40%, 26%, 13% and 7%, respectively. No difference was observed in the pattern of metastatic failure according to EGFR status.
Patients with strong EGFR staining intensity had a higher loco-regional recurrence rate (17% vs 10%) and a poorer RFS than those with negative to moderate staining intensity but without significant statistical difference.
Univariate analysis did not show any significant association of tumor local recurrence with age (P = 0.48), gender (P = 0.81), UICC stage III (P = 0.08), tumor location (P = 0.60), preoperative chemotherapy (P = 0.73), preoperative RT (P = 0.26), type of surgery (P = 0.66), resection margins (P = 0.10), and adjuvant chemotherapy (P = 0.75). To adjust for prognostic factors, the clinical parameters described in Table 2 were included in the multivariate analysis using the Cox proportional hazards model, i.e., EGFR extent (<25% vs ≥25%), resection (complete [R0] vs R1), tumor stage (II vs III), preoperative total dose RT (44 vs 60 Gy), gender, tumor location (lower third vs other thirds), type of surgery, resection margins, age (≤64 vs >64 years old), preoperative concomitant chemo-radiotherapy (no vs yes), EGFR intensity (negative to moderate vs strong), delay from the last day of RT to the day of surgery (≤41 vs >41 days) and uT.
EGFR extent expression, R1 resection, and 44-Gy total dose radiation were the independent prognostic factors that predicted locoregional failure with P values of 0.037, 0.032, and 0.045, respectively (Table 3). For both RFS and MFS, stage III tumor was detected as an independent prognostic factor with P values of 0.024 (hazard ratio = 4.08, CI 95%, 1.21–13.82) and 0.023 (hazard ratio = 4, CI 95%, 1.22–13.13), respectively. Concomitant preoperative chemo-radiotherapy was detected as a potential prognostic factor for RFS and MFS but statistical analysis showed only a trend towards significance P = 0.057 and 0.070, respectively. Margin resection <1 mm was also detected as a significant prognostic factor for MFS (hazard ratio = 5.03, CI 95%, 1.02–24.78, P = 0.047). EGFR expression predicted neither RFS (hazard ratio = 1.11, CI 95%, 0.28–4.46, P = 0.88) nor OS (hazard ratio = 1.26, CI 95%, 0.30–5.41, P = 0.753).
Discussion
The identification of parameters that reflect biological behavior of individual cancer tissues correlating with tumor aggressiveness is a key determinant of prognosis and a fundamental issue for the improvement of cancer therapy. Despite recent progress in defining the molecular mechanisms of cancer development and tumor progression, only a few individual biomarkers providing prognostic information have been identified. Among them, the EGFR pathways attracted the most attention of cancer investigators.
EGFR (HER1), a transmembrane glycoprotein, is a member of the large receptor tyrosine family encoded by a gene located in human chromosome 7p12. EGFR exists in inactive monomer form or in active dimer form. Dimerization can take place between different receptors in order to develop homologue (homodimers) or heterologue (heterodimers) dimers [9]. In either normal or malignant cells, the activation of EGF receptor cascades may have multiple consequences such as cell growth, differentiation, and proliferation. EGF receptor cascades may also promote malignant transformation, angiogenesis, and/or metastatic dissemination [10,11].
In addition, the cell membrane has been known for some time to be a secondary target for ionizing radiation. This phenomenon may provoke the pathways of mitogen-activated protein kinase (MAPK), phosphatidyl inositol-3-phosphate kinase (PI3K), and MAPK8 activation [12], which can modulate cell proliferation or death. Preclinical and clinical studies associate EGFR expression with radioresistance [13-16]. Ionizing radiation produces several types of cellular response via activation of multiple transduction pathways resulting in cell death, differentiation, or proliferation. Following irradiation, the MAPK pathway was recently reported to be a cellular "SOS" signal initiator starting from EGF receptors [17]. MAPK pathway activation via EGFR receptors was reported in many malignant human cell lines [17-19]. This activation is similar to the one produced by physiological concentrations of EGF (0.1 nM), and seems to act as a radioprotector [16,17,19,20]. Moreover, it has been recently shown that EGF-receptor and MAPK signal pathway activation following ionizing radiation depends on the proteolytic clivage of TGFβ precursor and functional activation of autocrine TGFβ [21]. STAT-3 signal pathway activation by phosphorylation via EGF receptors can be initiated by ionizing radiation, and it results with a radioprotective effect by apoptosis inhibition [22-24]. An inverse relation between the number of EGF receptors and tumor radiocurability is reported in several murine cell lines. In these models, radiation-induced apoptosis was decreased when important levels of EGF receptor were expressed on the cells [25,26]. Clinical consequences of these findings would be tailoring treatment according to a simple predictive assay of radiosensitivity based on the EGF-receptor expression. Clinical data pertaining to the relationship between EGFR expression and the success of radiotherapy are sparse and equivocal. Nevertheless, with respect to squamous-cell cancer of the head and neck, EGFR is among the best-studied examples [27-33], and positive and negative correlations between EGFR levels and tumor recurrences were reported in laryngeal cancers after radiotherapy [34-36]. The relationship of EGFR levels to the prognosis in unresectable pharyngeal or nasopharyngeal cancer patients treated by chemo-radiotherapy was recently reported [37-39].
In colorectal cancer, EGFR expression was evaluated in resected tumors [40]. The authors found significantly higher EGFR levels in stage III cancers than in stages I and II. It was then concluded that high EGFR expression is associated with poor prognosis. Another group [41], found 72 cases of EGFR-positive expression in 82 resected colorectal adenocarcinomas (88%). The extent of EGFR expression (>50%) revealed significant differences in survival times. In our study, a significant correlation between the positive tumor cell percentage greater than 25% and the rate of locoregional recurrence was detected (P = 0.037).
We did not assess the predictive value of EGFR on tumor response after preoperative treatment. This question was recently tackled by Giralt et al [42]. The authors analyzed EGFR expression of 45 locally advanced-rectal-cancer patients treated with preoperative radiotherapy and total mesorectal resection. Immunochemistry for EGFR was determined at the preradiation diagnostic biopsy and in the resected specimens. EGFR positivity was observed in 29 of 45 tumors (64%) and was associated with neither clinical tumor stage nor clinical nodal stage. The overall response rate was 34% in EGFR positive patients vs. 62% in those who did not express EGFR (P = 0.07). Only one of the seven pathologic complete remission patients was EGFR positive (P = 0.003). The link between the positive EGFR expression and the microscopic response on surgical specimen seems to be logical, but we fail to assess it in our series. Such a relationship should be based on a large tumor sampling and needs a very strict procedure at the macroscopic level, to ensure that the whole tumor is analyzed after a neoadjuvant treatment. An exhaustive tissue material should allow a precise analysis of the entire spectrum of tumor regression, i.e., complete, partial or none, as it has been proposed by Dvorack et al [43]. It should then be of interest to correlate these well documented histopathological data with biological parameters such as EGFR.
In our study EGFR expression was not found to be an independent prognostic factor for overall survival in patients with rectal cancer. Other studies, described above [40,41], have reported variable results making it difficult to draw firm conclusions about a possible relationship between EGFR expression and overall survival. Probably, the variation in results is due to (i) the use of different laboratory tests, (ii) varying extent of follow-up, and (iii) heterogeneity in the population of colon- and rectum-cancer patients. (i) EGFR assessment in previous studies was obtained by using different antibodies, different methods of antigen retrieval, and different cut-off values. In this study, we used uniform reagents provided by a kit allowing minimized variations in results and a reproducible method. Therefore, our results detected a low percentage of EGFR immunopositivity (56%) as compared with other colorectal cancer trials [44] probably due to the numerous IHC techniques. (ii) In this study we did not have a sufficiently long follow-up to give definitive conclusions on the prognostic impact on EGFR expression and overall survival. In fact, our analysis was only based on a group of patients with rectal carcinoma, a disease with a natural history different from that of colon carcinoma especially with respect to the tendency to recur locally. (iii) Reasons for recurrence after curative resection for colorectal carcinoma are not completely clear. Several theories have been put forward including, amongst others, microscopic deposits in the lymphatics, inadequate distant and lateral resection margins, exfoliated tumor cells at time of surgery, presence of malignant cells in the anastomosis and, finally, tumor aggressiveness related to biological behaviour. It is known that reported recurrence after resection for rectal carcinoma is commonly higher than after colon carcinoma [45,46], and differences in prognosis have also been reported between high and low rectal carcinomas [47]. Risk factors that have previously been associated with increased recurrence rates include amongst others patient age, gender, tumor stage, site of lesion (colon vs rectum), infiltration of adjacent organs, histopathological criteria, tumor size, lymph-node involvement and radial resection margins [45,46,48-50]. In rectal cancer, in particular, the impact of surgery and adequate lymph-node dissection related to the risk of local recurrence have been highlighted [51]. Several studies have evaluated the prognostic significance of EGFR on survival in colorectal cancer but, to our knowledge, not specifically focusing on rectal cancer recurrence.
In our study, multivariate statistical model identified EGFR expression as a significant independent predictor of recurrence following preoperative and curative surgery for rectal cancer. Two possible explanations for this relationship are to be considered. Firstly, EGFR overexpressing tumors exhibit a more aggressive behavior leading to more pelvic recurrences and in a lesser extent to more distant metastases. A second possible explanation is that EGFR overexpressing tumors present decreased intrinsic radiosensitivity as explained at the first part of this discussion and lead to more pelvic recurrences. This explanation is supported by the fact that the majority of the observed recurrences in our study appeared in the irradiated areas.
Therapeutic approaches targeting EGFR signaling pathways either alone or in combination with radiation or cytotoxic agents are being intensively investigated [52]. Strategies that are in various stages of development include blockade of the extracellular receptor domain [44] and inhibition of the intracellular tyrosine kinase activity [53]. Data presented in Figure 2 suggest that tumor radiation-sensitization through the inhibition of EGFR signaling could yield a therapeutic gain by increasing the locoregional control rate in patients with EGFR-overexpressing rectal cancer.
Conclusion
Knowledge of expression of EGFR in rectal cancer can contribute to the identification of patients with an increased risk of recurrences. Our results have to be related to several confounding factors such as the small number of events and the retrospective approach. Nevertheless, the potential of introducing routine EGFR immunohistochemistry as a diagnostic tool into the clinical practice of rectal cancer management still has to be undertaken, and may allow clinicians to deliver targeted therapies even in patients with a poorer prognosis.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
DA conceived the study, participated in its design and coordination. FB performed all EGFR immunohistochemical analyses. NB made acquisition of the data. AZ, MO, MY, CL, PR, AP, and JBD participated in the design of the study, in its analysis and in the interpretation of the data. DA, FB, ST, and MO drafted the manuscript. ST performed all statistical analyses. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors thank Ms Lavaill for excellent technical assistance, and Ms. Frances Godson for excellent editorial assistance.
Figures and Tables
Figure 1 Patterns of EGFR expression. a) Case with an extent positivity ≥25% with strong membrane staining of tumor cells. Insert and arrow: note the membranous positivity pattern (immunoperoxydase × 100 and × 400). b) Case with an extent positivity <25% with weak membrane staining of some tumor cells (arrow) (immunoperoxydase × 400). c) Case with negative staining (immunoperoxydase × 200).
Figure 2 Locoregional recurrence-free survival curves according to epidermal growth factor receptor (EGFR) expression extent.
Table 1 EGFR immunohistochemical staining characteristics in rectal-cancer patients
EGFR staining Patientsa (%) EGFR extent Patientsa (%) EGFR intensity Patientsa (%)
Negative 34 (44) - - - -
Positive 43 (56) <25% 23 (30) Weak and moderate 21 (27)
Strong 2 (3)
≥25% 20 (26) Weak and moderate 16 (21)
Strong 4 (5)
aNumber of patients
Table 2 EGFR expression and clinical characteristics of the study population
EGFR extent P value EGFR intensity P value
Parameters <25% ≥25% Negative to moderate Strong
Age (y)
≤64 26 (45.6)a 13 (65) 36 (50.7) 3 (50)
>64 31 (54.4) 7 (35) 0.14 35 (49.3) 3 (50) 0.97
Gender
Male 38 (66.7) 13 (65) 47 (66.2) 4 (66.7)
Female 19 (33.3) 7 (35) 0.89 24 (33.8) 2 (33.3) 0.98
Stage groupb
II 43 (75.4) 15 (75) 54 (76.1) 4 (66.7)
III 14 (24.6) 5 (25) 0.97 17 (23.9) 2 (33.3) 0.61
Tumor location
Lower third 39 (68.4) 13 (65) 46 (64.8) 6 (100)
Middle third 12 (21.1) 5 (25) 17 (23.9) -
Upper third 6 (10.5) 2 (10) 0.96 8 (11.3) - 0.21
Preoperative
RT-CTc 8 (14) 5 (25) 0.26 11 (15.5) 2 (33.3) 0.26
Preoperative RT
44 Gy 40 (70.2) 12 (60) 48 (67.6) 4 (66.7)
60 Gy 17 (28.8) 8 (40) 0.40 23 (32.4) 2 (33.3) 0.96
Type of surgeryd
AP 46 (80.7) 17 (85) 58 (81.7) 5 (83.3)
LAR 11 (19.3) 3 (15) 0.67 13 (18.3) 1 (16.7) 0.92
Resection
Margins
- 52 (93) 17 (85) 63 (90) 6 (100)
+ 4 (7) 3 (15) 0.30 7 (10) - 0.41
Adjuvant
chemotherapy 16 (28.1) 9 (45) 0.16 24 (33.8) 1 (16.7) 0.39
aData are presented as number of patients, with the percentage in parentheses.
bAccording to AJCC (American Joint Committee on Cancer) 1997.
cPreoperative chemo-radiotherapy.
dAP, abdominoperineal excision; LAR, low anterior resection
Table 3 Cox multivariate regression analysis for loco-regional recurrence-free survival (LRFS)a
Prognostic variable Hazard ratio 95% CIb Pvalue
EGFR extent <25%
≥25% 1
7.18 1.12 to 46 0.037
Resection R0
R1 1
6.92 1.18 to 40.42 0.032
Preoperative RT 60 Gy
44 Gy 1
5.78 1.04 to 32.06 0.045
aResults were adjusted on tumor stage, gender, type of surgery, surgical margin, age, preoperative radio-chemotherapy, EGFR intensity, delay from radiotherapy to surgery, and uT
bCI, confidence interval
==== Refs
Kane JM Petrelli NJ Controversies in the surgical management of rectal cancer Semin Radiat Oncol 2003 13 403 418 14586830 10.1016/j.semradonc.2003.07.001
Kapiteijn E Marijnen CA Nagtegaal ID Putter H Steup WH Wiggers T Rutten HJ Pahlman L Glimelius B van Krieken JH Leer JW van de Velde CJ Preoperative radiotherapy combined with total mesorectal excision for resectable rectal cancer N Engl J Med 2001 345 638 646 11547717 10.1056/NEJMoa010580
Colorectal Cancer Collaborative Group Adjuvant radiotherapy for rectal cancer: a systematic overview of 8,507 patients from 22 randomised trials Lancet 2001 358 1291 1304 11684209 10.1016/S0140-6736(01)06409-1
Wiggenraad R Tamminga R Blok P Rouse R Hermans J The prognostic significance of p53 expression for survival and local control in rectal carcinoma treated with surgery and postoperative radiotherapy Int J Radiat Oncol Biol Phys 1998 41 29 35 9588914 10.1016/S0360-3016(98)00043-1
Lopez-Crapez E Ychou M Thèzenas S Simony-Lafontaine J Thirion A Azria D Bibeau F Grenier J Senesse P P53 status and response to radiotherapy in rectal cancer : a prospective multilevel analysis B J Cancer 2005 92 2114 2121 15956964 10.1038/sj.bjc.6602622
Tannapfel A Nusslein S Fietkau R Katalinic A Kockerling F Wittekind C Apoptosis, proliferation, bax, bcl-2 and p53 status prior to and after preoperative radiochemotherapy for locally advanced rectal cancer Int J Radiat Oncol Biol Phys 1998 41 585 591 9635706 10.1016/S0360-3016(98)00076-5
International Commission on Radiation Units and Measurements (ICRU) Prescribing, recording, and reporting photon beam therapy. ICRU report 62 2001 Bethesda, Maryland
Sobin L Wittekind C UICC eds TNM Classification of Malignant Tumours 1997 New York: Wiley-Liss
Alroy I Yarden Y The ErbB signaling network in embryogenesis and oncogenesis: signal diversification through combinatorial ligand-receptor interactions FEBS Lett 1997 410 83 86 9247128 10.1016/S0014-5793(97)00412-2
Hudziak RM Schlessinger J Ullrich A Increased expression of the putative growth factor receptor p185HER2 causes transformation and tumorigenesis of NIH 3T3 cells Proc Natl Acad Sci U S A 1987 84 7159 7163 2890160
Allred DC Clark GM Molina R Tandon AK Schnitt SJ Gilchrist KW Osborne CK Tormey DC McGuire WL Overexpression of HER-2/neu and its relationship with other prognostic factors change during the progression of in situ to invasive breast cancer Hum Pathol 1992 23 974 979 1355464 10.1016/0046-8177(92)90257-4
Schmidt-Ullrich RK Dent P Grant S Mikkelsen RB Valerie K Signal transduction and cellular radiation responses Radiat Res 2000 153 245 257 10669545
Wollman R Yahalom J Maxy R Pinto J Fuks Z Effect of epidermal growth factor on the growth and radiation sensitivity of human breast cancer cells in vitro Int J Radiat Oncol Biol Phys 1994 30 91 98 8083133
Gupta AK McKenna WG Weber CN Feldman MD Goldsmith JD Mick R Machtay M Rosenthal DI Bakanauskas VJ Cerniglia GJ Bernhard EJ Weber RS Muschel RJ Local recurrence in head and neck cancer: relationship to radiation resistance and signal transduction Clin Cancer Res 2002 8 885 892 11895923
Barker FG Simmons ML Chang SM Prados MD Larson DA Sneed PK Wara WM Berger MS Chen P Israel MA Aldape KD EGFR overexpression and radiation response in glioblastoma multiforme Int J Radiat Oncol Biol Phys 2001 51 410 418 11567815 10.1016/S0360-3016(01)01609-1
Balaban N Moni J Shannon M Dang L Murphy E Goldkorn T The effect of ionizing radiation on signal transduction: antibodies to EGF receptor sensitize A431 cells to radiation Biochim Biophys Acta 1996 1314 147 156 8972728 10.1016/S0167-4889(96)00068-7
Schmidt-Ullrich RK Mikkelsen RB Dent P Todd DG Valerie K Kavanagh BD Contessa JN Rorrer WK Chen PB Radiation-induced proliferation of the human A431 squamous carcinoma cells is dependent on EGFR tyrosine phosphorylation Oncogene 1997 15 1191 1197 9294612 10.1038/sj.onc.1201275
Carter S Auer KL Reardon DB Birrer M Fisher PB Valerie K Schmidt-Ullrich R Mikkelsen R Dent P Inhibition of the mitogen activated protein (MAP) kinase cascade potentiates cell killing by low dose ionizing radiation in A431 human squamous carcinoma cells Oncogene 1998 16 2787 2796 9652746 10.1038/sj.onc.1201802
Kavanagh BD Dent P Schmidt-Ullrich RK Chen P Mikkelsen RB Calcium-dependent stimulation of mitogen-activated protein kinase activity in A431 cells by low doses of ionizing radiation Radiat Res 1998 149 579 587 9611096
Goldkorn T Balaban N Shannon M Matsukuma K EGF receptor phosphorylation is affected by ionizing radiation Biochim Biophys Acta 1997 1358 289 299 9366260 10.1016/S0167-4889(97)00063-3
Dent P Reardon DB Park JS Bowers G Logsdon C Valerie K Schmidt-Ullrich R Radiation-induced release of transforming growth factor alpha activates the epidermal growth factor receptor and mitogen-activated protein kinase pathway in carcinoma cells, leading to increased proliferation and protection from radiation-induced cell death Mol Biol Cell 1999 10 2493 2506 10436007
David M Wong L Flavell R Thompson SA Wells A Larner AC Johnson GR STAT activation by epidermal growth factor (EGF) and amphiregulin. Requirement for the EGF receptor kinase but not for tyrosine phosphorylation sites or JAK1 J Biol Chem 1996 271 9185 9188 8621573 10.1074/jbc.271.8.4134
Grandis JR Drenning SD Zeng Q Watkins SC Melhem MF Endo S Johnson DE Huang L He Y Kim JD Constitutive activation of Stat3 signaling abrogates apoptosis in squamous cell carcinogenesis in vivo Proc Natl Acad Sci U S A 2000 97 4227 4232 10760290 10.1073/pnas.97.8.4227
Park OK Schaefer TS Nathans D In vitro activation of Stat3 by epidermal growth factor receptor kinase Proc Natl Acad Sci U S A 1996 93 13704 13708 8942998 10.1073/pnas.93.24.13704
Akimoto T Hunter NR Buchmiller L Mason K Kian Ang K Milas L Inverse relationship between epidermal growth factor receptor expression and radiocurability of murine carcinomas Clin Cancer Res 1999 5 2884 2890 10537357
Sheridan MT O'Dwyer T Seymour CB Mothersill CE Potential indicators of radiosensitivity in squamous cell carcinoma of the head and neck Radiat Oncol Investig 1997 5 180 186 9327497 10.1002/(SICI)1520-6823(1997)5:4<180::AID-ROI3>3.0.CO;2-U
Christensen ME Therkildsen MH Poulsen SS Bretlau P Immunoreactive transforming growth factor alpha and epidermal growth factor in oral squamous cell carcinomas J Pathol 1993 169 323 328 8492225 10.1002/path.1711690308
Dassonville O Formento JL Francoual M Ramaioli A Santini J Schneider M Demard F Milano G Expression of epidermal growth factor receptor and survival in upper aerodigestive tract cancer J Clin Oncol 1993 11 1873 1878 8410112
Grandis JR Tweardy DJ TGF-alpha and EGFR in head and neck cancer J Cell Biochem 1993 17F 188 191 10.1002/jcb.240531027
Lee CS Redshaw A Boag G Epidermal growth factor receptor immunoreactivity in human laryngeal squamous cell carcinoma Pathology 1997 29 251 254 9271009
Maiorano E Favia G Maisonneuve P Viale G Prognostic implications of epidermal growth factor receptor immunoreactivity in squamous cell carcinoma of the oral mucosa J Pathol 1998 185 167 174 9713343 10.1002/(SICI)1096-9896(199806)185:2<167::AID-PATH70>3.0.CO;2-E
Santini J Formento JL Francoual M Milano G Schneider M Dassonville O Demard F Characterization, quantification, and potential clinical value of the epidermal growth factor receptor in head and neck squamous cell carcinomas Head Neck 1991 13 132 139 2022478
Etienne MC Pivot X Formento JL Bensadoun RJ Formento P Dassonville O Francoual M Poissonnet G Fontana X Schneider M Demard F Milano G A multifactorial approach including tumoural epidermal growth factor receptor, p53, thymidylate synthase and dihydropyrimidine dehydrogenase to predict treatment outcome in head and neck cancer patients receiving 5-fluorouracil Br J Cancer 1999 79 1864 1869 10206306 10.1038/sj.bjc.6690297
Miyaguchi M Olofsson J Hellquist HB Expression of epidermal growth factor receptor in glottic carcinoma and its relation to recurrence after radiotherapy Clin Otolaryngol 1991 16 466 469 1742894
Wen QH Miwa T Yoshizaki T Nagayama I Furukawa M Nishijima H Prognostic value of EGFR and TGF-alpha in early laryngeal cancer treated with radiotherapy Laryngoscope 1996 106 884 888 8667988 10.1097/00005537-199607000-00019
Almadori G Cadoni G Galli J Ferrandina G Scambia G Exarchakos G Paludetti G Ottaviani F Epidermal growth factor receptor expression in primary laryngeal cancer: an independent prognostic factor of neck node relapse Int J Cancer 1999 84 188 191 10096253 10.1002/(SICI)1097-0215(19990420)84:2<188::AID-IJC16>3.0.CO;2-1
Ang KK Berkey BA Tu X Zhang HZ Katz R Hammond EH Fu KK Milas L Impact of epidermal growth factor receptor expression on survival and pattern of relapse in patients with advanced head and neck carcinoma Cancer Res 2002 62 7350 7356 12499279
Magne N Pivot X Bensadoun RJ Guardiola E Poissonnet G Dassonville O Francoual M Formento JL Demard F Schneider M Milano G The relationship of epidermal growth factor receptor levels to the prognosis of unresectable pharyngeal cancer patients treated by chemo- radiotherapy Eur J Cancer 2001 37 2169 2177 11677103 10.1016/S0959-8049(01)00280-5
Chua DT Nicholls JM Sham JS Au GK Prognostic value of epidermal growth factor receptor expression in patients with advanced stage nasopharyngeal carcinoma treated with induction chemotherapy and radiotherapy Int J Radiat Oncol Biol Phys 2004 59 11 20 15093894 10.1016/j.ijrobp.2003.10.038
Steele RJ Kelly P Ellul B Eremin O Epidermal growth factor receptor expression in colorectal cancer Br J Surg 1990 77 1352 1354 2276016
Mayer A Takimoto M Fritz E Schellander G Kofler K Ludwig H The prognostic significance of proliferating cell nuclear antigen, epidermal growth factor receptor, and mdr gene expression in colorectal cancer Cancer 1993 71 2454 2460 8095852
Giralt J Eraso A Armengol M Rossello J Majo J Ares C Espin E Benavente S de Torres I Epidermal growth factor receptor is a predictor of tumor response in locally advanced rectal cancer patients treated with preoperative radiotherapy Int J Radiat Oncol Biol Phys 2002 54 1460 1465 12459370 10.1016/S0360-3016(02)03752-5
Dworak O Keilholz L Hoffmann A Pathological features of rectal cancer after preoperative radiochemotherapy Int J Colorectal Dis 1997 12 19 23 9112145 10.1007/s003840050072
Cunningham D Humblet Y Siena S Khayat D Bleiberg H Santoro A Bets D Mueser M Harstrick A Verslype C Chau I van Custem E Cetuximab monotherapy and cetuximab plus irinotecan in irinotecan-refractory metastatic colorectal cancer N Engl J Med 2004 351 337 345 15269313 10.1056/NEJMoa033025
Galandiuk S Wieand HS Moertel CG Cha SS Fitzgibbons RJ Pemberton JH Wolff BG Patterns of recurrence after curative resection of carcinoma of the colon and rectum Surg Gynecol Obstet 1992 174 27 32 1729745
Obrand DI Gordon PH Incidence and patterns of recurrence following curative resection for colorectal carcinoma Dis Colon Rectum 1997 40 15 24 9102255
Konn M Morita T Hada R Yamanaka Y Sasaki M Munakata H Suzuki H Inoue S Endoh M Sugiyama Y Survival and recurrence after low anterior resection and abdominoperineal resection for rectal cancer: the results of a long-term study with a review of the literature Surg Today 1993 23 21 30 8461602 10.1007/BF00308995
Michelassi F Vannucci L Ayala JJ Chappel R Goldberg R Block GE Local recurrence after curative resection of colorectal adenocarcinoma Surgery 1990 108 787 792 2218892
Arbman G Nilsson E Hallbook O Sjodahl R Local recurrence following total mesorectal excision for rectal cancer Br J Surg 1996 83 375 379 8665198
de Haas-Kock DF Baeten CG Jager JJ Langendijk JA Schouten LJ Volovics A Arends JW Prognostic significance of radial margins of clearance in rectal cancer Br J Surg 1996 83 781 785 8696739
Bjerkeset T Edna TH Rectal cancer: the influence of type of operation on local recurrence and survival Eur J Surg 1996 162 643 648 8891623
Azria D Larbouret C Robert B Culine S Ychou M Verrelle P Dubois JB Pelegrin A Anti-EGF receptors and radiation therapy: current translational research and early clinical trials Bull Cancer 2003 90 202 212
Sartor CI Epidermal growth factor family receptors and inhibitors: radiation response modulators Semin Radiat Oncol 2003 13 22 30 12520461 10.1053/srao.2003.50003
|
15967033
|
PMC1185521
|
CC BY
|
2021-01-04 16:03:05
|
no
|
BMC Cancer. 2005 Jun 20; 5:62
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-62
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-631596704310.1186/1471-2407-5-63Research ArticleP-cadherin expression and survival rate in oral squamous cell carcinoma:an immunohistochemical study Lo Muzio Lorenzo [email protected] Giuseppina [email protected] Antonio [email protected] Corrado [email protected] Giuseppe [email protected] Rosario [email protected] Gregorio [email protected] Lillo Alfredo [email protected] Francesco [email protected] Department of Surgical Sciences, University of Foggia, Foggia, Italy2 Department of Dental Sciences "G. Messina", University of Palermo, Palermo, Italy3 Institute of Histology, University of Bologna, Bologna, Italy4 Institute of Pathology, University of Ancona, Ancona, Italy5 Department of Dental Sciences, Second University of Naples, Naples, Italy6 Institute of Maxillofacial Surgery, University of Ferrara, Ferrara, Italy2005 21 6 2005 5 63 63 2 1 2005 21 6 2005 Copyright © 2005 Lo Muzio 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
P-cadherin (P-cad) is a transmembrane molecule involved in the cell-cell adhesion and similar to E-cadherin (E-cad), but less investigated in oncology, especially in in vivo studies. Aims of the present study were to assess the prevalence of P-cad expression in oral squamous cell carcinoma (OSCC) and to verify whether P-cad can be considered a marker of prognosis in patients with OSCC.
Methods
In a retrospective study, a cohort of 67 OSCC patients was investigated for P-cad expression and its cellular localization by immunohistochemistry; some respective healthy margins of resection were similarly investigated as standard controls. After grouping for P-cad expression, OSCCs were statistically analyzed for the variables age, gender, histological grading (G), TNM, Staging, and overall survival rate. Univariate and multivariate analyses were performed.
Results
37 cases (55.2%) of OSCC showed membranous/cytoplasmic positivity for P-cad, whereas 30 (44.8 %) were negative. Although with some differences in membranous vs cytoplasmic localization of P-cad in OSCC with different G, no statistical association was found between P-cad expression and any variables considered at baseline. In terms of prognostic significance, P-cad non expression was found to have an independent association with poorer overall survival rate than P-cad expressing group (P = 0.056); moreover, among P-cad +ve patients the best prognosis was for those OSCC with membranous (P < 0.0001) than those with cytoplasmic P-cad expression.
Conclusion
On the basis of these results, it is possible to suggest P-cad as an early marker of poor prognosis. The abnormal or lack of P-cad expression could constitute an hallmark of aggressive biological behavior in OSCC
==== Body
Background
Invasive OSCC, in spite of improved therapeutic procedures, actually show a generally poor prognosis since its local aggressiveness and metastases. In particular, the incidence of lymph node metastases has been found to be significantly associated with several factors; among these, not only macroscopic parameters, such as clinical stage, localization and thickness of primary tumours [1-10], but also microscopic-molecular parameters from differentiation of tumoral cells up to their skill for adherence [11-14]. Recently, many investigations have been performed in this latter direction, until to know that intercellular adhesiveness is mediated by a family of glycoproteins named cadherins [15]. This family is composed of an extra-cellular domain, involved in Ca++-dependent homophilic binding to adjacent cells, a trans-membrane domain, and an intra-cellular domain which binds to proteins called catenins [16]. In epithelial cells, this adhesiveness is mediated by epithelial-cadherin (E-cad), a 120-kd transmenbrane glycoprotein, localized mainly in the zonula adherens junctions. The cadherin family includes also other members: neural-cadherin (N-cadherin) [17], placental cadherin (P-cad) [18] and liver cell adhesion molecule (L-CAM), and more than 20 cadherins have been described in the central nervous system, liver and vascular endothelial cells and in other tissues and organs [19,20]. In particular, P-cad is a protein homologous to E-cad. E-cadherin is involved in the adherens type of intercellular junctions of keratinocytes, while P-cad is detected on the cell-cell contact surface of basal keratinocytes in normal mouse and human epidermis and cells migrating into the suprabasal compartiment down-regulate P-cad expression. Both these molecules interact with cytoskeleton by alpha-catenin.
P-cad is expressed in mouse placenta [18], epithelia [21], the basal cell of the skin [22,23], playing an important role in the morphogenesis of epidermis and skin appendage [22,24,25]. The expression of P-cad in epithelial tissues appears to identify cell populations with proliferative activity, and its expression decreases as cells differentiate [23,26].
The possible role exerted by cadherins in human carcinogenesis has been suggested by a number of studies [27,28]. Down-regulation of E-cad was reported to be directly related to invasiveness and progression of many human epithelial tumours [28], including oral squamous cell carcinomas (OSCC) [29].
While E-cad expression has been extensively studied in many forms of human cancers, including OSCC [27,30-40], less is known about the expression levels of P-cad in human cancers [23,41-51] and, particularly, in OSCCs in vivo [35,39,40,52].
Aims of the present study were to assess the prevalence of P-cad expression in oral squamous cell carcinoma (OSCC) and to verify whether P-cad can be considered a marker of prognosis in patients with OSCC.
Methods
Sixty-seven patients affected by histologically proven OSCC were consecutively recruited among those surgically treated in a multicentric study between January 1992 and December 1997. The patients, never treated before for OSCC, included 45 males (67.2%) and 22 females (32.8%). They ranged in age from 18 to 87 years (median age 65 years) at the time of admission; 20 (29.8%) had neck nodes, and none had evidence of distant metastases. Tumours were classified according to U.I.C.C. 2000 classification (UICC 2000), reaching to the following Stage Grouping: Stage I for 30 OSCC, II for 15, III for 11 and IV for 11.
Although recruited in different centres, all of OSCCs were treated according to the common and current Guidelines dedicated[53]. In particular, sAlthough treated in deifferent centers, all odf OSCCs were Alkjkkhhhpurgery on T was the treatment of choice and always performed at the initial course of the protocol with curative intent (i.e. only tumour resections in safe margins were done). When radiotherapy was considered useful, it was usually done 3 weeks after surgery, with external beam and the dose was equivalent to 60 or 65 Gy in 6 or 7 weeks. Chemotherapy, when prescribed, consisted of cisplatin (100 mg/m2 body-surface area) given as intravenous infusion followed by continuous 24-hour intravenous infusion of fluorouracil (1,000 mg/m2 per day) for five days. globally up to 3 cycles. All patients were followed up and examined on a monthly basis for the first year after treatment, every 2 months for the second year, and every 3 months thereafter. At our baseline, an overall disease-specific survival was calculated at 72 months for all patients plus cases censored (for death).
Immunohistochemistry
5 μm serial sections from routinely formalin-fixed paraffin-embedded blocks were cut for each case, and one section stained with haematoxylin-eosin (H.E.) was used to confirm the histopathological diagnosis. Only sections showing sufficient epithelium to assess 1000 cells were considered for this study. In addition, some microscopically healthy specimens of oral mucosa from juxtaposed sites to OSCC went to similar investigation as standard controls.
Immunocytochemistry was then performed on the remaining sections mounted on poly-L-lysine-coated glass slides. Endogenous peroxidase was quenched by incubating the sections for 20 minutes with 0,3% hydrogen peroxide in methanol. To improve the staining pattern, the sections were boiled three times for 3 minutes in 10 mM citrate buffer as an antigen retrieval method. In order to prevent non-specific binding of antibodies, sections were then pre-incubated with non-immune mouse serum (1:20; Dakopatts, Hamburg, Germany) and diluted in PBS/BSA (1%) for 25 minutes at room temperature. After washing twice with Tris-HCl buffer, primary antibodies were applied. As positive controls, the immunoreactivity of 5 normal skin sections from leg was evaluated. A negative control was also performed in each run by substituting primary antibodies with non-immune serum (DAKO Antibody Diluent, Dakopatts, Hamburg, Germany). All the slides were washed twice in Tris-HCl buffer between each step. Commercially available mouse monoclonal IgG antibody against P-cad (Transduction Laboratories, Lexington, Kentucky, U.S.A.), packaged at 0.25 mg/ml, was used at a dilution of 1:300. Then two methods were applied: Labeled streptavidin-biotin-peroxidase technique (LSAB-HRP) and Labeled Streptavidin-biotin-alkaline phosphatase technique (LSAB-AP). In LSAB-HRP technique sequential 20-minute incubation with biotinylated linking antibody and horseradish peroxidase-labeled streptavidin (Dako LSAB + kit, HRP) were performed at room temperature. The peroxidase activity was developed by incubation with 3.3'-diaminobenzidine (DAB, Vector Laboratories, Burlinghame, USA) as a substrate chromogen solution. The slides were then counterstained with hematoxylin. In LSAB-AP technique sequential 20-minute incubations with biotinylated anti-mouse immunoglogulins and streptavidin conjugated to alkaline phosphatase were performed. Finally, a new fast red substrate system (K0597, Dako, Glostrup, Denmark) was applied as a chromogen solution. The specificity of this antibody has been described in the literature [42].
Evaluation of immunostaining
The number of P-cad-expressing tumour cells was estimated as a percentage of the final number of 500 neoplastic cells of each case, and scored in two categories: score 0 (≤ 5% of cells were positive), score 1 (P-cad expression in > 5 of cells %). The expansion of P-cad-positive cells in the spinous layer was defined as anomalous P-cad expression [47].
Statistical analysis
Univariate analysis
Differences between P-cad expression values and the variables considered were analysed by means of Student-Newmann-Keuls' test (simple or in multiple comparison) and by ANOVA. The difference was considered significant when p-value was ≥ 0.05. Disease-specific survival curves were calculated according to the product-limit method (Kaplan-Meier algorithm). Time zero was defined as the date of the patient's initial diagnosis. Patients who are still alive were included in the total number at risk of death only up to the time of their last follow-up. Therefore, the survival rate only changed when death occurred. Patients dead during the follow-up period (i.e. 72 months) were considered as censored. The calculated survival rate was the maximum estimate of the true survival curve. Log rank test was used to compare survival curves, generated by stratifications for a variable of interest.
Cox regression analysis
Afterwards, Cox regression analysis was applied to determine the single contribution of covariates on survival rate. Cox regression analysis compares survival data while taking into account the statistical value of independent variables, such as age and sex, on whether or not an event (i.e. death) is likely occur. If the associated probability was less then 5% (p < .05), the difference was considered statistically significant. In the process of doing the regression analysis, odds ratio (OR) and 95% confidence interval (CIs) were calculated. Stepwise Cox analysis allowed us to detect the variables most associated to survival.
Results
P-cad expression
First of all, the paradigmatic P-cad expression was obtained from the standard controls (Fig. 1a): in these, in fact, only a membranous staining was observed at the basal layer of histologically normal oral epithelium; predominantly on the membrane of only a thin line of cells basally located, with occasionally moderate parabasal staining. The intensity of staining for P-cadherin progressively reduced from basal to parabasal layers and stopped in the spinous layer. No staining for P-cadherin was observed in the upper layer.
37 cases (55.2%) OSCC showed membranous/cytoplasmic positivity for P-cad (Group P-cad +ve)(Table 1), whereas 30 (44.8 %) were found negative (Group P-cad -ve). When examined the cell staining pattern of positive cases, 25 cases showed a prevalent membranous pattern (Fig. 1b-c), while 12 had a prevalent cytoplasmic pattern (Fig. 1d). Worthy of note, within the Group P-cad +ve, dedifferentiated areas showed both membranous and cytoplasmic P-cad up-regulation: well-differentiated (G1) oral carcinomas showed P-cad up-regulation, while P-cad expression homogeneously reduced in scarcely differentiated oral squamous cell carcinomas (G3), and it shifted to membranous/cytoplasmic co-localization, predominantly cytoplasmic in distribution, or alternatively was absent in a large numbers of cells. Although with these differences in membranous vs cytoplasmic localization of P-cad in OSCC with different G, no statistical association was found between P-cad expression and any variables (age, sex, G, T, N and Stage grouping) considered at baseline.
The second part of the analysis planned the study of survival rates with respect to P-cad expression. Although the global disease-specific survival rate at 72 months was 51.0 %, irrespectively of the extent of the tumour or treatment (Fig. 2), the survival rates in the same cohort distributed according to P-cad expression (Group P-cad +ve vs. P-cad -ve) was 79.0% vs 40.0% respectively (p-value = 0.04 Log Rank Test). Survival curves stratified according to P-cad expression are illustrated in Fig 3. Still in terms of overall survival, within P-cad +ve group, OSCCs (n = 12) with a prevalent citoplasmic pattern of P-cad showed poorer survival rates than those (n = 25) with a prevalent membranous P-cad expression (P <0.0001). Besides the classic parameters related to survival rate and predictor of a poor outcome (e.g. G, T, N, stage and recurrence), a stepwise model introducing P-cad non-expression, without considering recurrence (parameter with the highest OR), showed that also P-cad non expression is significantly associated to survival, together with grading and stage. (Table 2, 3).
Discussion
Both E- and P-cad play a pivotal role in the maintenance of the epithelial structure, even if they are expressed in distinct regions of the epithelium. E-cad is expressed on all epithelial layers, while P-cad is predominantly expressed in the basal layer of stratified squamous epithelia, the proliferative compartment [54-59]. E- and P-cad expression is altered in premalignant and malignant skin tumors, as demonstrated by reduced E-cad and aberrant P-cad expression in human squamous cell carcinomas [59], indicating the importance of coordinated cadherin expression for maintaining normal epidermal structure [60]. Malignant keratinocytes probably acquire different mechanisms for regulating the expression of these two cadherins [61]. P-cad seems to play a role in the maintenance of the epithelial phenotype and may be involved, together with E-cad, in the final stage of tumor progression in epidermal carcinogenesis, being a marker of hyperproliferative activity [47]. Studies on epidermis [26], gastric epithelium [49], and mammary epithelium [62] showed that P-cad controls cell proliferation in these tissues. P-cad expression seems to be related to tumour progression in gastric [50] and gingival carcinomas [35], while its expression is higher in poorly differentiated than in well-differentiated lung carcinomas [23]. In particular, well-differentiated oral carcinomas showed P-cad expression similar to normal oral mucosa or up-regulated, while P-cad expression homogeneously reduced in low-differentiated oral squamous cell carcinomas or its localization shifted to the cytoplasm, in accordance with other studies on oral mucosa [39] or gastrointestinal mucosa [48,63]. Williams et al. (1988) reported a loss of membranous immunostaining at the periphery of the islands of carcinoma with a cytoplasmic immunostaining or a complete loss[39]. In contrast, towards the centre of the islands the more differentiated cells showed mild or moderate membranous staining in well- or moderately-differentiated carcinomas, reflecting the pattern seen in dysplasia [39]. Recently, also in oral premalignant lesions induced in rats it has been found that E-cad and P-cad have similar location of expression as in OSCC and that just P-cad aberrant expression could be a strong marker of carcinogen progression[64].
On the basis of the current knowledge on P-cad and of a previous research conducted by one of the center involved[52], in the present research our main goal was to conduct an in vivo study on the clinical outcome (e.g overall survival rate) of OSCC with respect of quality and quantity of P-cad expression. Hence, in terms of prognostic significance, the lack of P-cad expression (44.8%) was found to have an independent association with poorer overall survival rate than P-cad expressing group; moreover, the abnormal (cytoplasmic) expression of P-cad is also associated to a poorer prognosis when compared to that normally membranous. In these latter cases, expression of P-cad shifted to membranous/cytoplasmic co-localisation, predominantly cytoplasmic in distribution, or alternatively was absent from a large numbers of cells. These two main results are consistent with a very recent Italian research [52] in which also other tissues (i.e. lymph node and bone metastases) were investigated.
As regards the lack of P-cad expression, it can be interpreted as a late event prior to invasion, as shown by loss of its expression in dysplasia adjacent to infiltrating carcinomas [39]. The loss of P-cad expression probably comes after P-cad cytoplasmic relocalization. Loss of P-cad expression in OSCC is associated with tumour invasion, while P-cad membranous staining in OSCC is probably due to the up-regulation seen in tumour cell lines and dysplasias. Therefore, in the initial phase of tumour growth the high expression of P-cad may be crucial in the formation of a tumour mass which is ready to progress and metastasize [50]. Whereas anomalous P-cad expression in the spinous layer of epithelium overlying tumour can be a biological marker for keratinocyte atypia and/or premalignant changes[65]. In fact, the continued expression of P-cad in the invasive cells can contribute to the maintenance of the epithelioid phenotype of the carcinoma cells [65]. An experimental study on squamous cell carcinoma cell lines showed aberrant expression in cancer cells, whereas E-cad expression was reduced [61]. SCC cells probably acquire the ability to express P-cad and this molecule plays a role in tumour progression [61]. Elevated [Ca++] determined increased cell-surface P-cad expression in SCC cell lines by up-regulation of de novo P-cad synthesis, while in normal keratinocytes calcium-induced cell-surface P-cad expression is a result of the translocation of pre-formed P-cad from the cytosol without up-regulation of P-cad synthesis [60]. These results suggest the existence of a unique mechanism for regulating the P-cad expression gene in tumour cells.
Bagutti et al. (1998) showed no correlation between P-cad expression and differentiation of tumour cells[40], while Sakaki et al. (1994) showed a complete loss of P-cad expression in poorly-differentiated gingival SCC[35]. Cytoplasmic relocalization or loss of P-cad expression may be responsible, together with other known/unknown upregulated oncogenes and downregulated tumour suppressor genes, for the later stages of tumour progression, such as invasive growth and metastasis [50].
On the basis of our present results, no association was found between P-cad expression and any of parameters considered (age, gender, G, T, N, Stage grouping), datum not consistent with the Italian research cited above [52], probably due to the different sample size.
A limit of the present study could be the fact that assessment of P-cad expression was done according to two scores; a larger sample size is warranted in the future in order to score OSCC in several groups according to P-cad expression with respect to prognosis.
The main present findings, based on overall survival rate, emphasise the role of P-cad expression as an early marker of OSCC prognosis, earlier than recurrence, as being the best marker, unfortunately, detected only after its onset.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
LLM conceived of the study, participated in its design- coordination and helped to draft the manuscript. GC participated in the design of the study, helped to analyse data and to draft the manuscript. AF participated in the design of the study and performed the statistical analysis. CS performed immunohistochemical study. RS revised the article critically. GL made substantial contributions to acquisition, analysis and interpretation of data. GP performed immunohistochemical study. ADL performed immunohistochemical study. FC has been involved in drafting the article. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was supported by a Grant from The Italian MIUR, year 2004.
Figures and Tables
Figure 1 a) Strong basal-parabasal membranous expression of P-cadherin in oral hyperplastic epithelium (LSAB-AP, nuclear counterstaining with haematoxylin, ×106); b) Membranous P-cadherin expression in oral moderately-differentiated SCC (LASB-AP, ×400); c) Membranous expression of P-cadherin in an area of stromal infiltration from moderately-differentiated OSCC (LASB-AP, nuclear counterstaining with haematoxylin, ×160); d) Cytoplasmic expression of P-cadherin in a case of lowly differentiated OSCC infiltrating stroma (LSAB-HRP, ×250)
Figure 2 Overall disease-specific survival rate at 72-months (+ censored cases).
Figure 3 Disease-specific survival rate calculated according to P-cad categories. Log Rank 3.38, 1 df p-value = 0.0661 (+ censored cases)
Table 1 OSCC grouped by P-cad expression and their features.
Variables No. P-cad-ve (%) P-cad+ve n. (%) Mean Standard deviation P < 0.05
Cases 67 30(44.8) 37 (55.2)
Age
</= 65 years 32 16 16 1.40 1.10 P = 0.363°
65 years 35 14 21 1.65 1.13
Sex
Male 45 20 25 1.55 1.07 P = 0.253°
Female 22 10 12 1.50 1.22
Grading
G1 21 6 15 1.85 1.10 P = 0.143*
G2 30 15 15 1.53 1.07
G3 16 9 7 1.12 1.14
Staging
I 30 10 20 1,667 0,4795 P = 0,993*
II 15 6 9 1,600 0,5071
III 11 7 4 1,364 0,5045
IV 11 4 7 1,636 0,5045
°Student-Newmann-Keuls' test.
*One-way Analysis of Variance (ANOVA) and Student-Newman-Keuls Multiple Comparisons Test
Table 2 Cox regression for survival rate with all variables known, apart from P-cad.
Variables OR 95% CI p-value
Lower Upper
Grading 2.341 1.292 4.241 .005
T .391 .162 .940 .036
N .194 .062 .607 .005
Stage 6.553 2.327 18.456 .000
Recurrence 23.170 7.697 69.745 .000
Table 3 Cox regression analyses forced for the presence of P-cad and without considering recurrence.
Varziables OR 95.0% CI p-value
Lower Upper
P-cad 4.068 .964 17.172 .056
Grading 1.942 1.164 3.240 .011
Stage 1.461 1.114 1.915 .006
==== Refs
Berenson JR Yang J Mickel RA Frequent amplification of the bcl-1 locus in head and neck squamous cell carcinomas Oncogene 1989 4 1111 1116 2476705
Po Wing Yuen A Lam KY Lam LK Ho CM Wong A Chow TL Yuen WF Wei WI Prognostic factors of clinically stage I and II oral tongue carcinoma-A comparative study of stage, thickness, shape, growth pattern, invasive front malignancy grading, Martinez-Gimeno score, and pathologic features Head Neck 2002 24 513 520 12112547 10.1002/hed.10094
P O Pillai G Patel S Fisher C Archer D Eccles S Rhys-Evans P Tumour thickness predicts cervical nodal metastases and survival in early oral tongue cancer Oral Oncol 2003 39 386 390 12676259 10.1016/S1368-8375(02)00142-2
O'Brien CJ Lauer CS Fredricks S Clifford AR McNeil EB Bagia JS Koulmandas C Tumor thickness influences prognosis of T1 and T2 oral cavity cancer--but what thickness? Head Neck 2003 25 937 945 14603454 10.1002/hed.10324
Lam P Au-Yeung KM Cheng PW Wei WI Yuen AP Trendell-Smith N Li JH Li R Correlating MRI and histologic tumor thickness in the assessment of oral tongue cancer AJR Am J Roentgenol 2004 182 803 808 14975989
Lim SC Zhang S Ishii G Endoh Y Kodama K Miyamoto S Hayashi R Ebihara S Cho JS Ochiai A Predictive markers for late cervical metastasis in stage I and II invasive squamous cell carcinoma of the oral tongue Clin Cancer Res 2004 10 166 172 14734465
Ross GL Soutar DS MacDonald DG Shoaib T Camilleri IG Robertson AG Improved staging of cervical metastases in clinically node-negative patients with head and neck squamous cell carcinoma Ann Surg Oncol 2004 11 213 218 14761927 10.1245/ASO.2004.03.057
Russolo M Giacomarra V Papanikolla L Tirelli G Prognostic indicators of occult metastases in oral cancer Laryngoscope 2002 112 1320 1323 12169922 10.1097/00005537-200207000-00035
Sparano A Weinstein G Chalian A Yodul M Weber R Multivariate predictors of occult neck metastasis in early oral tongue cancer Otolaryngol Head Neck Surg 2004 131 472 476 15467620 10.1016/j.otohns.2004.04.008
Sheahan P O'Keane C Sheahan JN O'Dwyer TP Effect of tumour thickness and other factors on the risk of regional disease and treatment of the N0 neck in early oral squamous carcinoma Clin Otolaryngol 2003 28 461 471 12969352 10.1046/j.1365-2273.2003.00748.x
Shear M Hawkins DM Farr HW The prediction of lymph node metastases from oral squamous carcinoma Cancer 1976 37 1901 1907 1260692
Frierson HFJ Cooper PH Prognostic factors in squamous cell carcinoma of the lower lip Hum Pathol 1986 17 346 354 3957335
Umeda M Yokoo S Take Y Omori A Nakanishi K Shimada K Lymph node metastasis in squamous cell carcinoma of the oral cavity: correlation between histologic features and the prevalence of metastasis Head Neck 1992 14 263 272 1517076
Yamamoto E Miyakawa A Kohama G Mode of invasion and lymph node metastasis in squamous cell carcinoma of the oral cavity Head Neck Surg 1984 6 938 947 6724960
Takeichi M Hatta K Nose A Nagafuchi A Identification of a gene family of cadherin cell adhesion molecules Cell Differ Dev 1988 25 Suppl 91 94 3061598 10.1016/0922-3371(88)90104-9
Gumbiner BM McCrea PD Catenins as mediators of the cytoplasmic functions of cadherins J Cell Sci Suppl 1993 17 155 158 8144692
Hatta K Takeichi M Expression of N-cadherin adhesion molecules associated with early morphogenetic events in chick development Nature 1986 320 447 449 3515198 10.1038/320447a0
Nose A Takeichi M A novel cadherin cell adhesion molecule: its expression patterns associated with implantation and organogenesis of mouse embryos J Cell Biol 1986 103 2649 2658 3539943 10.1083/jcb.103.6.2649
Suzuki S Sano K Tanihara H Diversity of the cadherin family: evidence for eight new cadherins in nervous tissue Cell Regul 1991 2 261 270 2059658
Buxton RS Cowin P Franke WW Garrod DR Green KJ King IA Koch PJ Magee AI Rees DA Stanley JR Nomenclature of the desmosomal cadherins J Cell Biol 1993 121 481 483 8486729 10.1083/jcb.121.3.481
Hirai Y Nose A Kobayashi S Takeichi M Expression and role of E- and P-cadherin adhesion molecules in embryonic histogenesis. I. Lung epithelial morphogenesis Development 1989 105 263 270 2806125
Hirai Y Nose A Kobayashi S Takeichi M Expression and role of E- and P-cadherin adhesion molecules in embryonic histogenesis. II. Skin morphogenesis Development 1989 105 271 277 2806126
Shimoyama Y Hirohashi S Hirano S Noguchi M Shimosato Y Takeichi M Abe O Cadherin cell-adhesion molecules in human epithelial tissues and carcinomas Cancer Res 1989 49 2128 2133 2702654
Wheelock MJ Jensen PJ Regulation of keratinocyte intercellular junction organization and epidermal morphogenesis by E-cadherin J Cell Biol 1992 117 415 425 1373144 10.1083/jcb.117.2.415
Lewis JE Jensen PJ Wheelock MJ Cadherin function is required for human keratinocytes to assemble desmosomes and stratify in response to calcium J Invest Dermatol 1994 102 870 877 8006450 10.1111/1523-1747.ep12382690
Hodivala KJ Watt FM Evidence that cadherins play a role in the downregulation of integrin expression that occurs during keratinocyte terminal differentiation J Cell Biol 1994 124 589 600 8106556 10.1083/jcb.124.4.589
Birchmeier W Behrens J Cadherin expression in carcinomas: role in the formation of cell junctions and the prevention of invasiveness Biochim Biophys Acta 1994 1198 11 26 8199193
Birchmeier W E-cadherin as a tumor (invasion) suppressor gene Bioessays 1995 17 97 99 7748170 10.1002/bies.950170203
Downer CS Speight PM E-cadherin expresson in normal, hyperplastic and malignant oral epithelium. Oral Oncol, Eur J Cancer 1993 29B 303 305 10.1016/0964-1955(93)90053-H
Bowie GL Caslin AW Roland NJ Field JK Jones AS Kinsella AR Expression of the cell-cell adhesion molecule E-cadherin in squamous cell carcinoma of the head and neck Clin Otolaryngol 1993 18 196 201 8365008
Fuller LC Allen MH Montesu M Barker JN Macdonald DM Expression of E-cadherin in human epidermal non-melanoma cutaneous tumours Br J Dermatol 1996 134 28 32 8745882 10.1046/j.1365-2133.1996.d01-739.x
Kinsella AR Bowie GL Field JK Jones AS Expression of the cell-cell adhesion molecule E-cadherin in tongue carcinoma cell lines J Laryngol Otol 1994 108 957 961 7829949
Mattijssen V Peters HM Schalkwijk L Manni JJ van 't Hof-Grootenboer B de Mulder PH Ruiter DJ E-cadherin expression in head and neck squamous-cell carcinoma is associated with clinical outcome Int J Cancer 1993 55 580 585 8406985
Sakaki T Wato M Otake S Shirasu R Tanaka A Localization of E-cadherin adhesion molecules in human gingiva and gingival carcinoma Acta Pathol Jpn 1993 43 99 106 8484337
Sakaki T Wato M Kaji R Mushimoto K Shirasu R Tanaka A Correlation of E- and P-cadherin expression with differentiation grade and mode of invasion in gingival carcinoma Pathol Int 1994 44 280 286 8044295
Schipper JH Frixen UH Behrens J Unger A Jahnke K Birchmeier W E-cadherin expression in squamous cell carcinomas of head and neck: inverse correlation with tumor dedifferentiation and lymph node metastasis Cancer Res 1991 51 6328 6337 1933895
Schipper JH Unger A Jahnke K E-cadherin as a functional marker of the differentiation and invasiveness of squamous cell carcinoma of the head and neck Clin Otolaryngol 1994 19 381 384 7834876
Andrews NA Jones AS Helliwell TR Kinsella AR Expression of the E-cadherin-catenin cell adhesion complex in primary squamous cell carcinomas of the head and neck and their nodal metastases Br J Cancer 1997 75 1474 1480 9166940
Williams HK Sanders DS Jankowski JA Landini G Brown AM Expression of cadherins and catenins in oral epithelial dysplasia and squamous cell carcinoma J Oral Pathol Med 1998 27 308 317 9725568
Bagutti C Speight PM Watt FM Comparison of integrin, cadherin, and catenin expression in squamous cell carcinomas of the oral cavity J Pathol 1998 186 8 16 9875134 10.1002/(SICI)1096-9896(199809)186:1<8::AID-PATH156>3.0.CO;2-H
Paul R Ewing CM Jarrard DF Isaacs WB The cadherin cell-cell adhesion pathway in prostate cancer progression Br J Urol 1997 79 Suppl 1 37 43 9088271
Soler AP Harner GD Knudsen KA McBrearty FX Grujic E Salazar H Han AC Keshgegian AA Expression of P-cadherin identifies prostate-specific-antigen-negative cells in epithelial tissues of male sexual accessory organs and in prostatic carcinomas. Implications for prostate cancer biology Am J Pathol 1997 151 471 478 9250159
Shimoyama Y Gotoh M Terasaki T Kitajima M Hirohashi S Isolation and sequence analysis of human cadherin-6 complementary DNA for the full coding sequence and its expression in human carcinoma cells Cancer Res 1995 55 2206 2211 7743525
Foty RA Steinberg MS Measurement of tumor cell cohesion and suppression of invasion by E- or P-cadherin Cancer Res 1997 57 5033 5036 9371498
Matsuyoshi N Tanaka T Toda K Imamura S Identification of novel cadherins expressed in human melanoma cells J Invest Dermatol 1997 108 908 913 9182820 10.1111/1523-1747.ep12292703
Palacios J Benito N Pizarro A Suarez A Espada J Cano A Gamallo C Anomalous expression of P-cadherin in breast carcinoma. Correlation with E-cadherin expression and pathological features Am J Pathol 1995 146 605 612 7534041
Pizarro A Gamallo C Benito N Palacios J Quintanilla M Cano A Contreras F Differential patterns of placental and epithelial cadherin expression in basal cell carcinoma and in the epidermis overlying tumours Br J Cancer 1995 72 327 332 7640213
Sanders DS Bruton R Darnton SJ Casson AG Hanson I Williams HK Jankowski J Sequential changes in cadherin-catenin expression associated with the progression and heterogeneity of primary oesophageal squamous carcinoma Int J Cancer 1998 79 573 579 9842964 10.1002/(SICI)1097-0215(19981218)79:6<573::AID-IJC4>3.0.CO;2-H
Shimoyama Y Hirohashi S Expression of E- and P-cadherin in gastric carcinomas Cancer Res 1991 51 2185 2192 2009537
Yasui W Sano T Nishimura K Kitadai Y Ji ZQ Yokozaki H Ito H Tahara E Expression of P-cadherin in gastric carcinomas and its reduction in tumor progression Int J Cancer 1993 54 49 52 8478147
Rasbridge SA Gillett CE Sampson SA Walsh FS Millis RR Epithelial (E-) and placental (P-) cadherin cell adhesion molecule expression in breast carcinoma J Pathol 1993 169 245 250 8383197 10.1002/path.1711690211
Lo Muzio L Pannone G Mignogna MD Staibano S Mariggio MA Rubini C Procaccini M Dolci M Bufo P De Rosa G Piattelli A P-cadherin expression predicts clinical outcome in oral squamous cell carcinomas Histol Histopathol 2004 19 1089 1099 15375751
Ord RA Blanchaert RHJ Current management of oral cancer. A multidisciplinary approach J Am Dent Assoc 2001 132 Suppl 19S 23S 11803648
Takeichi M Morphogenetic roles of classic cadherins Curr Opin Cell Biol 1995 7 619 627 8573335 10.1016/0955-0674(95)80102-2
Borradori L Sonnenberg A Hemidesmosomes: roles in adhesion, signaling and human diseases Curr Opin Cell Biol 1996 8 647 656 8939649 10.1016/S0955-0674(96)80106-2
Moles JP Watt FM The epidermal stem cell compartment: variation in expression levels of E-cadherin and catenins within the basal layer of human epidermis J Histochem Cytochem 1997 45 867 874 9199672
Nicholson LJ Pei XF Watt FM Expression of E-cadherin, P-cadherin and involucrin by normal and neoplastic keratinocytes in culture Carcinogenesis 1991 12 1345 1349 2070502
Fujita M Furukawa F Fujii K Horiguchi Y Takeichi M Imamura S Expression of cadherin cell adhesion molecules during human skin development: morphogenesis of epidermis, hair follicles and eccrine sweat ducts Arch Dermatol Res 1992 284 159 166 1503501 10.1007/BF00372710
Shirahama S Furukawa F Wakita H Takigawa M E- and P-cadherin expression in tumor tissues and soluble E-cadherin levels in sera of patients with skin cancer J Dermatol Sci 1996 13 30 36 8902651 10.1016/0923-1811(95)00493-9
Wakita H Furukawa F Baba S Takigawa M Human squamous-cell-carcinoma cell line (DJM-1) cells synthesize P- cadherin molecules via an elevation of extracellular calcium: calcium regulates P-cadherin-gene expression at the translational level via protein tyrosine phosphorylation Int J Cancer 1997 73 432 439 9359492 10.1002/(SICI)1097-0215(19971104)73:3<432::AID-IJC19>3.0.CO;2-E
Wakita H Shirahama S Furukawa F Distinct P-cadherin expression in cultured normal human keratinocytes and squamous cell carcinoma cell lines Microsc Res Tech 1998 43 218 223 9840799 10.1002/(SICI)1097-0029(19981101)43:3<218::AID-JEMT3>3.0.CO;2-S
Daniel CW Strickland P Friedmann Y Expression and functional role of E- and P-cadherins in mouse mammary ductal morphogenesis and growth Dev Biol 1995 169 511 519 7781895 10.1006/dbio.1995.1165
Sanders DS Perry I Hardy R Jankowski J Aberrant P-cadherin expression is a feature of clonal expansion in the gastrointestinal tract associated with repair and neoplasia J Pathol 2000 190 526 530 10727977 10.1002/(SICI)1096-9896(200004)190:5<526::AID-PATH564>3.0.CO;2-9
Sakaki T Tamura I Kadota H Kakudo K Changing expression of E- and P-cadherin during rat tongue carcinogenesis induced by 4-nitroquinoline 1-oxide J Oral Pathol Med 2003 32 530 537 12969227 10.1034/j.1600-0714.2003.00174.x
Cano A Gamallo C Kemp CJ Benito N Palacios J Quintanilla M Balmain A Expression pattern of the cell adhesion molecules. E-cadherin, P- cadherin and alpha 6 beta 4 intergrin is altered in pre-malignant skin tumors of p53-deficient mice Int J Cancer 1996 65 254 262 8567126 10.1002/(SICI)1097-0215(19960117)65:2<254::AID-IJC21>3.0.CO;2-C
|
15967043
|
PMC1185522
|
CC BY
|
2021-01-04 16:03:05
|
no
|
BMC Cancer. 2005 Jun 21; 5:63
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-63
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-811602950310.1186/1471-2407-5-81Research ArticleCommon variation in EMSY and risk of breast and ovarian cancer: a case-control study using HapMap tagging SNPs Benusiglio Patrick R [email protected] Fabienne [email protected] Craig [email protected] Joan [email protected] Robert N [email protected] Paula [email protected] Alison [email protected] Douglas F [email protected] Bruce AJ [email protected] Paul D [email protected] Department of Oncology, University of Cambridge, Strangeways Research Laboratory, Worts Causeway, Cambridge CB1 8RN, UK2 Centre National de Genotypage, 2 rue Gaston Cremieux, CP 5721, 91057 Evry Cedex, France3 EPIC, University of Cambridge, Strangeways Research Laboratory, Worts Causeway, Cambridge CB1 8RN, UK4 Department of Genetic Epidemiology, University of Cambridge, Strangeways Research Laboratory, Worts Causeway, Cambridge CB1 8RN, UK2005 19 7 2005 5 81 81 10 3 2005 19 7 2005 Copyright © 2005 Benusiglio 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
EMSY could be involved in low-level susceptibility to breast and ovarian cancer. Gene amplification is seen in a proportion of breast and ovarian tumours and correlates with poor prognosis in breast cancer patients. Furthermore, the EMSY protein silences a transcription activation domain in BRCA2 exon 3.
Methods
We used a genetic association study design to determine if common genetic variation (frequency ≥ 5%) in EMSY was associated with breast or ovarian cancer risk in the British population. Haplotype tagging single-nucleotide polymorphisms (htSNPs) were selected from the HapMap database and genotyped using Taqman® in two large study sets of white British women (n [breast set] = 2343 cases and 2284 controls, n [ovarian set] = 864 cases and 864 controls). HapMap data might be insufficient to tag genetic variation in EMSY comprehensively. We therefore screened the gene promoter and coding sequences with denaturing high performance liquid chromatography in order to identify additional SNPs that are most likely to be functional.
Results
HapMap data on 22 SNPs show that 4 htSNPs tag 4 common haplotypes: rs2282611 (5'up t>g), rs4245443 (IVS7 g>a), rs2513511 (IVS16 a>g), rs2155220 (3'down c>t). We observed no association between any of the genotypes or associated haplotypes and breast or ovarian cancer risk. Seventeen out of the 18 remaining HapMap polymorphisms (94%) were well tagged by the 4 selected htSNPs (r2s > 0.8). Genotype frequencies for two further SNPs identified by screening and located near exon-intron boundaries, rs2508740 (IVS9 a>g) and rs11600501 (IVS10 c>t), were also similar in cases and controls. In order to simulate unidentified SNPs, we performed the leave-one-out cross-validation procedure on the HapMap data; over 95% of the common genetic variation was well represented by tagging polymorphisms. We are therefore likely to have tagged any common, functional variants present in our population.
Conclusion
We found no association between common genetic variation in EMSY and risk of breast or ovarian cancer in two large study sets of white British women.
==== Body
Background
Breast and ovarian cancer are two of the most common causes of cancer in women in the United Kingdom (Office for National Statistics). Together, they account for about a third of all new cancer cases and a quarter of cancer deaths. Positive family history is a well-established risk factor for both diseases: the risk to first-degree relatives of a case is about 2 times the population risk [1-3]. Most of the excess familial risk associated with breast and ovarian cancer is likely to be genetic in origin [3,4]. However, only a small proportion of this risk is accounted for by known highly predisposing genes, BRCA1 and BRCA2, while the remainder might be explained by a combination of weakly predisposing alleles [5-10]. EMSY (C11orf30) is a novel gene that could be involved in low-level predisposition to breast and ovarian cancer [11]. Its recent discovery generated widespread interest [12,13]. The gene maps to chromosome 11q13, spans 103.3 kilobases and comprises 20 coding exons. EMSY is amplified in 12% of breast cancers and 17% of high-grade ovarian cancers and its amplification has been associated with an increased risk of relapse as well as decreased survival in breast cancer patients [11,14]. Furthermore, the EMSY protein silences the transcriptional activation potential of BRCA2 exon 3, a region deleted in a Swedish breast and ovarian cancer family [11,15].
The case-control study design is well suited to the identification of small-effect genes that are likely to underlie common, complex diseases such as breast or ovarian cancer: a difference in allele frequency is sought between affected individuals and unrelated controls [16]. Two approaches have been proposed. The traditional, hypothesis-driven approach is to investigate single-nucleotide polymorphisms (SNPs) in coding regions, since they are more likely to have a functional role and to influence directly the traits under study [17]. The alternative, indirect approach is to select a set of haplotype-tagging SNPs (htSNPs); htSNPs are informative polymorphisms that best characterize haplotype diversity and therefore genetic variation within the gene [18,19]. They serve as markers to detect associations between a particular region and diseases, whether or not the SNPs themselves have a functional effect [20,21]. It is not necessary to genotype all possible polymorphisms because the alleles of SNPs that are physically close to each other tend to be correlated with each other: they are in linkage disequilibrium (LD) [22]. The HapMap online database allows the indirect approach to be applied readily to many genes or regions [23]. By December 2004, the database held the genotypes of 90 individuals with northern and western European ancestry for over 850'000 SNPs.
We used a genetic association study design to determine if variation in EMSY was associated with breast or ovarian cancer risk. In order to have good power to detect small relative risks, we restricted our attention to common SNPs and haplotypes (frequency ≥ 5%). We first selected htSNPs using HapMap data. We also screened the gene promoter and coding regions in order to identify additional polymorphisms that are likely to be functional, as HapMap might be insufficient to tag genetic variation comprehensively. The selected SNPs were then genotyped in two large case-control sets (one breast cancer set and one ovarian cancer set) of white British women.
Methods
Patients and controls
Cases were drawn from the breast and ovarian arms of the SEARCH Cancer Study, an ongoing population-based study with cases ascertained through the East Anglia and West Midlands cancer registries in the United Kingdom [5,24]. All women diagnosed after 1990 with invasive breast cancer under the age of 70 years, or epithelial ovarian cancer under the age of 75 years, were eligible for inclusion. Approximately 65% of eligible breast cancer patients and 60% of ovarian cancer patients have enrolled in the study. Women taking part were asked to provide a 20-ml blood sample for DNA analysis and to complete a comprehensive epidemiological questionnaire. We carried out genotyping on sub-sets consisting of the first 2343 (breast cancer) and 864 (ovarian cancer) cases. Median age at diagnosis was 51 years for breast cancer cases (age range 25 to 69) and 55 years for ovarian cancer cases (age range 16 to 74). Two thousand two hundred and eighty-four and 864 controls were randomly drawn from the Norfolk component of the European Prospective Investigation of Cancer (EPIC), for the breast and the ovarian studies, respectively [25]. The EPIC-Norfolk cohort comprises 25,000 individuals resident in Norfolk (East Anglia), ages 45–74 years. The ethnic background of both cases and controls is similar, with over 98% being white Europeans. Ethical approval was obtained from the Anglia and Oxford Multicentre Research Committee and the Norwich Local Research Ethics Committee and informed consent was obtained from each patient.
SNP identification and selection
We selected htSNPs from the HapMap database (, public releases up to September 2004) with the TagSNPs program [26], including 5 kilobases upstream and downstream the gene and aiming for a minimum r2h of 0.8. r2h is a measure of correlation between haplotypes defined by all SNPs and haplotypes defined by the selected htSNPs. At the time of selection, genotypes were only available for the Centre d'Etude du Polymorphisme Humain (CEPH) samples: these were collected in 1980 from people living in Utah with ancestry from northern and western Europe.
In order to screen the gene promoter and coding regions for polymorphisms, we performed denaturing high performance liquid chromatography (DHPLC) using the Wavemaker detection system (version 4.1, Transgenomics, Crewe, United Kingdom), followed by sequencing (3100 Genetic Analyser, Applied Biosystems, Warrington, United Kingdom) on genomic DNA from 48 random breast cancer cases. A 600-base pair putative promoter starting 500 base pairs upstream the gene was identified with gene2promoter , a program that allows for automated extraction of groups of promoters from a list of accession numbers or gene IDs.
Genotyping
Genotyping was carried out using Taqman® (Applied Biosystems) according to manufacturer's instructions. Primers and probes were supplied directly by Applied Biosystems except those for IVS9 a>g that were designed using Primer Express Oligo Design Software v2.0 (Applied Biosystems). Sequences are available on request. Reactions were carried out at 60°C in 384-well plates with cases and controls plated together. Each plate included 2 negative controls with no DNA and 12 samples duplicated on a separate quality control plate. Plates were read on the ABI Prism 7900 using the Sequence Detection Software (Applied Biosystems). Failed genotypes were not repeated.
Statistical methods
For each SNP, deviation of genotype frequencies in controls from the Hardy-Weinberg equilibrium was assessed by a χ2 test with one degree of freedom (df). Genotype frequencies in cases and controls were compared by a χ2 test for heterogeneity (2df). Genotype-specific risks were estimated as odds ratios (OR) using standard cross-product ratio and confidence intervals were calculated using the variance of the log (OR), which was estimated by the standard Taylor expansion. A comparison of haplotype frequencies between cases and controls was carried out using the haplo.score routine implemented in S-plus [27]. Haplotypes with an estimated frequency of less than 5% were pooled. Haplo.score uses a likelihood that depends on estimated haplotype frequencies to test the statistical association between haplotypes and phenotype. It is based on score statistics, which provide both global tests and haplotype-specific tests [27]. The LDA program [28] was used to calculate pairwise LD for SNP pairs in the breast cancer study set. LDA is a Java-based program implementing the EM algorithm for pairwise LD analysis [28].
Power was determined using standard statistical methods [29]. We have over 90% power at the 1% significance level to detect a dominant allele with a frequency of 0.2 that confers a relative risk of breast cancer of 1.3 or a relative risk of ovarian cancer of 1.6. Power to detect recessive alleles at the 1% significance level is more limited: 59% for an allele with a frequency of 0.2 that confers a relative risk of breast cancer of 1.5 or 51% for an allele with a frequency of 0.3 that confers a relative risk of ovarian cancer of 1.5.
Results
Genotypes for 22 common EMSY SNPs were available in HapMap, none of the SNPs were in coding regions. The working density was therefore of one SNP per 5 kilobases. The gene consisted of only one LD block [21]. There were 5 common haplotypes which constituted 92% of all the observed haplotypes and were tagged by 5 htSNPs: rs2282611 (5'up t>g), rs4245443 (IVS7 g>a), rs2513511 (IVS16 a>g), rs2155220 (3'down c>t) and rs7106446 (table 1). Taqman® assays were successfully designed for the first four, but an assay could not be designed for rs7106446. There were no alternative SNPs with similar tagging properties. We were thus left with 4 htSNPs tagging 4 common haplotypes.
Genotyping success rate was over 92%. None of the genotype distributions in the controls differed significantly from those expected under Hardy-Weinberg equilibrium. There was no evidence that any of the SNPs is associated with breast (table 2) or ovarian cancer (table 3); genotype-specific OR were all close to unity with narrow confidence intervals. There was no association of genotype with age in controls and, as expected, age adjusted risks were close to the unadjusted risks (data not shown). The 4 htSNPs generated 5 common haplotypes in our population; the global test of association was not significant for breast cancer (P = 0.27) or for ovarian cancer (P = 0.93), nor were there any differences between cases and controls for the individual haplotype frequencies (Additional file: 1). The number of common haplotypes tagged by the 4 selected htSNPs differed between HapMap (n = 4) and our study (n = 5) because two rare HapMap haplotypes tagged by SNP rs1939468 were grouped into our fifth common haplotype (Additional file: 1).
Screening of the promoter and coding regions revealed two further SNPs located near exon-intron boundaries, rs2508740 (IVS9 a>g) located 4 base pairs upstream exon 10 and rs11600501 (IVS10 c>t) located 14 base pairs upstream exon 11; neither of these were associated with breast or ovarian cancer risk (tables 2 and 3). At the time of study, there were four putative, non-validated coding SNPs mentioned in the dbSNP database : rs1954782, rs11822571, rs3753051 and rs1047196. We did not detect any of them. LD was strong (D' > 0.7) across pairs involving all SNPs except IVS10 c>t while IVS7 g>a and IVS9 a>g were in nearly perfect LD (r2 = 0.94) (figure 1).
Discussion
This is the first association study reporting results on EMSY, a gene of importance through its interaction with BRCA2 and its amplification status in tumours. We found no association between any of the EMSY genotypes or associated haplotypes and risk of breast or ovarian cancer in a white British population. We could have failed to observe a true association because of a Type II statistical error, but the large size of our study gives us high statistical power and strongly reduces the likelihood that our results are influenced by chance fluctuations in the case or control genotype frequency [30].
An alternative reason for failure to observe a true association could be that our set of tagging SNPs are poor markers of a true causal variant, which would either be one of the known SNPs in the gene or an as yet unidentified variant. In HapMap, common EMSY haplotypes were tagged by 5 SNPs. However, an assay for rs7106446 could not be designed and thus our htSNP set was suboptimal. Where a tagging SNP is used as a marker for a true disease-predisposing SNP the effective sample size is proportional to the bivariate correlation coefficient (r2) between the marker and causal SNPs [31]. r2s is the squared correlation coefficient between multi-locus haplotypes and individuals SNPs and is analogous to r2. In order to establish how well we had tagged the known SNPs with our set of tagging SNPs, we calculated r2s [26] between the 4 selected htSNPs and the 18 remaining HapMap polymorphisms. Seventeen (94%) SNPs were tagged with r2s > 0.8 but 1 SNP, rs7106446, was tagged with r2s = 0.46. Loss of power was therefore marginal for all HapMap SNPs except rs7601446; for a SNP tagged with r2s = 0.85, we had 89% power at the 1% significance level to detect a dominant allele with a frequency of 0.3 that confers a relative risk of ovarian cancer of 1.5.
It is also possible that we have not adequately tagged an unidentified, disease-predisposing SNP. Whole-gene resequencing across a sample population would be required to identify all existing polymorphisms and allow investigators to select htSNPs that tag all genetic variants. The HapMap project does not re-sequence the genome; it validates SNPs from the dbSNP public database, aiming for a density of polymorphisms that cover the whole of genetic variation across the human genome. Comprehensive tagging requires a genotyping density of around 1 SNP per 2.5 kb [32]. The 1 SNP per 5 kb density available for EMSY in HapMap might therefore be insufficient. In order to identify additional SNPs that are most likely to be functional [17], we screened the gene promoter and coding sequence with DHPLC, a technique with an estimated sensitivity of 94% [33]. Two SNPs near exon-intron boundaries were identified but neither was associated with breast or ovarian cancer.
We also assessed how well a set of htSNPs would have tagged any unidentified SNPs using a leave-one-out cross validation procedure on the HapMap data: each of the 22 known SNPs were dropped in turn and htSNPs selected from the remaining 21, thus simulating unidentified polymorphisms [32]. The ability of htSNPs to tag the dropped SNP was then evaluated by calculating r2s [26]. Mean r2s was 0.94. Twenty-one (95.4%) out of 22 dropped SNPs were tagged with r2s > 0.8, and only 1 (4.6%) was tagged with 0.4 < r2s < 0.8. Over 95% of the common genetic variation in EMSY should therefore be well represented by tagging polymorphisms. We are therefore likely to have tagged any common, functional variants present in our population. After this study was completed and the first version of the manuscript submitted, genotyping data in a white American population for rs3753051, a synonymous coding SNP in exon 19, were released in dbSNP. We were able to assess how this polymorphism was tagged by our set of SNPs as genotypes from the same individuals were also available for 5'up t>g, IVS9 a>g, IVS16 a>g and 3'down c>t; SNP rs3753051 was perfectly tagged (r2 = 1) by 5'up t>g.
This study design can not exclude the involvement of a rare allele in predisposition to breast or ovarian cancer; for example, CHEK2*1100delC, has a frequency of around 1% and was recently shown to confer a two-fold increased risk of breast cancer [34]. Our study set would be too small to detect the effect of such an allele if it doubled the risk of ovarian cancer. Some authors have advocated the use of histopathologic or demographic data that subclassify individuals in order to identify homogeneous subsets for analysis [35]. In the absence of any main effect or strong biological rationale, we have not carried out subgroup analyses as much larger sample sizes would be required to obtain reliable results. The number of possible post-hoc, subgroup analyses is large and there is a strong possibility that one or more tests will be statistically significant simply by chance [36].
We are reporting results for a set of htSNPs selected from HapMap. We used genotypes for the CEPH samples to choose htSNPs. According to the HapMap website, it is unclear how accurately the CEPH samples reflect the patterns of genetic variation in people with northern and western European ancestry. Our results suggest that they correctly predict genetic variation in our white British population: allele frequencies in the breast study controls were similar to those obtained from HapMap (P = 0.57, 0.99, 0.88 and 0.85 for 5'up t>g, IVS7 g>a, IVS16 a>g and 3'down c>t, respectively), thus strengthening the argument for a widespread use of the database for htSNPs selection. A predisposing SNP might have a differential effect in another ethnic group via gene-gene or gene-environment interactions, although in a recent study of the genetic effects for 43 validated gene-disease associations across 697 study populations of various descents, Ioannidis et al. concluded that, even if frequencies of polymorphisms varied among populations, their biological impact on the risk for common diseases should be consistent across traditional ethnic boundaries [37,38].
Conclusion
We saw no association between common SNPs in EMSY or their associated haplotypes with risk of breast or ovarian cancer in two large study sets of white British women.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
PRB carried out the experiments, performed the analyses and wrote the manuscript under the supervision of FL, BAJP and PDP. CL managed the genotyping process. JM and RL collected DNA from cancer cases and EPIC controls, respectively. PS contributed to the haplotype analyses. AD was the laboratory manager. DFE was the statistical advisor. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional file 1
Click here for file
Acknowledgements
PRB is supported by the Ligue Genevoise contre le Cancer (N/Ref 0208). BAJP is a Gibb Fellow, DFE is a Principal Fellow and PDP is a Senior Clinical Research Fellow of Cancer Research United Kingdom. This work was funded by Cancer Research United Kingdom. We are grateful to Shahana Ahmed, Don Conroy and Oluseun Ajai for their technical help.
Figures and Tables
Figure 1 Linkage disequilibrium (LD). Pairwise (LD) measures of D' (left bottom half) and r2 (right top half) for the six single-nucleotide polymorphisms (SNPs) genotyped in the breast and ovarian cancer study sets.
Table 1 Haplotype-tagging single-nucleotide polymorphisms (htSNPs) selected from the HapMap database
dbSNP id SNP name Frequency Location
rs2282611 5'up t>g 0.35 START-3136
rs4245443 IVS7 g>a 0.40 Intron 7 +40
rs2513511 IVS16 a>g 0.14 Intron 16 +1947
rs2155220 3'down c>t 0.45 END +3584
Table 2 Single-nucleotide polymorphisms (SNPs) in the breast cancer study set. Allele frequencies, genotype frequencies and genotype-specific risks in 2343 women with breast cancer and 2284 controls. OR, odds ratio; CI, confidence intervals; RAF, rare allele frequency; M/M, common homozygotes; M/m, heterozygotes; m/m, rare homozygotes; df, degrees of freedom.
SNP Series RAF M/M n (%) M/m n (%) m/m n (%) Number genotyped P 2 df
5'up t>g Cases 0.33 989 (45) 961 (44) 231 (11) 2181
Controls 0.32 1069 (47) 959 (42) 249 (11) 2277 0.42
OR
(95% CI) 1
(ref) 1.08
(0.96–1.23) 1.00
(0.82–1.22)
IVS7 g>a Cases 0.40 727 (36) 980 (48) 317 (16) 2024
Controls 0.40 803 (37) 1025 (47) 364 (16) 2192 0.51
OR
(95% CI) 1
(ref) 1.06
(0.92–1.21) 0.96
(0.80–1.15)
IVS16 a>g Cases 0.13 1629 (75) 502 (23) 36 (2) 2167
Controls 0.14 1690 (75) 526 (23) 42 (2) 2258 0.87
OR
(95% CI) 1
(ref) 0.99
(0.86–1.14) 0.89
(0.57–1.40)
3'down c>t Cases 0.44 638 (32) 986 (49) 399 (20) 2023
Controls 0.44 691 (31) 1073 (49) 440 (20) 2204 0.98
OR
(95% CI) 1
(ref) 1.00
(0.87–1.14) 0.98
(0.83–1.17)
IVS9 a>g Cases 0.39 760 (37) 988 (48) 299 (15) 2047
Controls 0.39 847 (39) 1009 (46) 343 (16) 2199 0.29
OR
(95% CI) 1
(ref) 1.09
(0.96–1.24) 0.97
(0.81–1.17)
IVS10 c>t Cases 0.05 1824 (91) 177 (9) 3 (0) 2004
Controls 0.04 1956 (91) 187 (9) 3 (0) 2146 0.99
OR
(95% CI) 1
(ref) 1.02
(0.82–1.26) 1.07
(0.22–5.32)
Table 3 Single-nucleotide polymorphisms (SNPs) in the ovarian cancer study set. Allele frequencies, genotype frequencies and genotype-specific risks in 864 women with ovarian cancer and 864 controls. OR, odds ratio; CI, confidence intervals; RAF, rare allele frequency; M/M, common homozygotes; M/m, heterozygotes; m/m, rare homozygotes; df, degrees of freedom.
SNP Series RAF M/M n (%) M/m n (%) m/m n (%) Number genotyped P 2 df
5'up t>g Cases 0.31 346 (47) 315 (43) 69 (9) 730
Controls 0.32 392 (46) 369 (43) 92 (11) 853 0.65
OR
(95% CI) 1
(ref) 0.97
(0.79–1.19) 0.85
(0.60–1.20)
IVS7 g>a Cases 0.40 222 (37) 289 (48) 95 (16) 606
Controls 0.40 304 (36) 405 (48) 140 (16) 849 0.90
OR
(95% CI) 1
(ref) 0.98
(0.78–1.23) 0.93
(0.68–1.27)
IVS16 a>g Cases 0.14 479 (74) 164 (25) 6 (1) 649
Controls 0.13 652 (76) 188 (22) 13 (2) 853 0.22
OR
(95% CI) 1
(ref) 1.19
(0.93–1.51) 0.63
(0.24–1.66)
3'down c>t Cases 0.44 219 (30) 370 (51) 141 (19) 730
Controls 0.43 283 (33) 412 (48) 159 (19) 854 0.41
OR
(95% CI) 1
(ref) 1.16
(0.93–1.45) 1.15
(0.86–1.53)
IVS9 a>g Cases 0.39 319 (37) 398 (47) 134 (16) 851
Controls 0.39 327 (38) 402 (47) 133 (15) 862 0.97
OR
(95% CI) 1
(ref) 1.01
(0.82–1.25) 1.03
(0.78–1.37)
IVS10 c>t Cases 0.04 778 (92) 68 (8) 0 (0) 846
Controls 0.05 776 (91) 77 (9) 2 (0) 855 0.29
OR
(95% CI) 1
(ref) 0.88
(0.63–1.24) -
==== Refs
Edmondson RJ Monaghan JM The epidemiology of ovarian cancer Int J Gynecol Cancer 2001 11 423 429 11906544 10.1046/j.1525-1438.2001.01053.x
Familial breast cancer: collaborative reanalysis of individual data from 52 epidemiological studies including 58,209 women with breast cancer and 101,986 women without the disease Lancet 2001 358 1389 1399 11705483 10.1016/S0140-6736(01)06524-2
Pharoah PD Ponder BA The genetics of ovarian cancer Best Pract Res Clin Obstet Gynaecol 2002 16 449 468 12413928 10.1053/beog.2002.0296
Ponder BA Cancer genetics Nature 2001 411 336 341 11357140 10.1038/35077207
Prevalence and penetrance of BRCA1 and BRCA2 mutations in a population-based series of breast cancer cases. Anglian Breast Cancer Study Group Br J Cancer 2000 83 1301 1308 11044354 10.1054/bjoc.2000.1407
Dite GS Jenkins MA Southey MC Hocking JS Giles GG McCredie MR Venter DJ Hopper JL Familial risks, early-onset breast cancer, and BRCA1 and BRCA2 germline mutations J Natl Cancer Inst 2003 95 448 457 12644538
Peto J Collins N Barfoot R Seal S Warren W Rahman N Easton DF Evans C Deacon J Stratton MR Prevalence of BRCA1 and BRCA2 gene mutations in patients with early-onset breast cancer J Natl Cancer Inst 1999 91 943 949 10359546 10.1093/jnci/91.11.943
Easton DF How many more breast cancer predisposition genes are there? Breast Cancer Res 1999 1 14 17 11250676 10.1186/bcr6
Pharoah PD Antoniou A Bobrow M Zimmern RL Easton DF Ponder BA Polygenic susceptibility to breast cancer and implications for prevention Nat Genet 2002 31 33 36 11984562 10.1038/ng853
Gayther SA Russell P Harrington P Antoniou AC Easton DF Ponder BA The contribution of germline BRCA1 and BRCA2 mutations to familial ovarian cancer: no evidence for other ovarian cancer-susceptibility genes Am J Hum Genet 1999 65 1021 1029 10486320 10.1086/302583
Hughes-Davies L Huntsman D Ruas M Fuks F Bye J Chin SF Milner J Brown LA Hsu F Gilks B Nielsen T Schulzer M Chia S Ragaz J Cahn A Linger L Ozdag H Cattaneo E Jordanova ES Schuuring E Yu DS Venkitaraman A Ponder B Doherty A Aparicio S Bentley D Theillet C Ponting CP Caldas C Kouzarides T EMSY links the BRCA2 pathway to sporadic breast and ovarian cancer Cell 2003 115 523 535 14651845 10.1016/S0092-8674(03)00930-9
King MC A novel BRCA2-binding protein and breast and ovarian tumorigenesis N Engl J Med 2004 350 1252 1253 15028830 10.1056/NEJMcibr033528
Livingston DM EMSY, a BRCA-2 partner in crime Nat Med 2004 10 127 128 14760417 10.1038/nm0204-127
Rodriguez C Hughes-Davies L Valles H Orsetti B Cuny M Ursule L Kouzarides T Theillet C Amplification of the BRCA2 pathway gene EMSY in sporadic breast cancer is related to negative outcome Clin Cancer Res 2004 10 5785 5791 15355907
Nordling M Karlsson P Wahlstrom J Engwall Y Wallgren A Martinsson T A large deletion disrupts the exon 3 transcription activation domain of the BRCA2 gene in a breast/ovarian cancer family Cancer Res 1998 58 1372 1375 9537232
Risch NJ Searching for genetic determinants in the new millennium Nature 2000 405 847 856 10866211 10.1038/35015718
Tabor HK Risch NJ Myers RM Opinion: Candidate-gene approaches for studying complex genetic traits: practical considerations Nat Rev Genet 2002 3 391 397 11988764 10.1038/nrg796
Rebbeck TR Ambrosone CB Bell DA Chanock SJ Hayes RB Kadlubar FF Thomas DC SNPs, haplotypes, and cancer: applications in molecular epidemiology Cancer Epidemiol Biomarkers Prev 2004 13 681 687 15159297
Freedman ML Penney KL Stram DO Le Marchand L Hirschhorn JN Kolonel LN Altshuler D Henderson BE Haiman CA Common variation in BRCA2 and breast cancer risk: a haplotype-based analysis in the Multiethnic Cohort Hum Mol Genet 2004 13 2431 2441 15317758 10.1093/hmg/ddh270
Cardon LR Abecasis GR Using haplotype blocks to map human complex trait loci Trends Genet 2003 19 135 140 12615007 10.1016/S0168-9525(03)00022-2
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
Pharoah PD Dunning AM Ponder BA Easton DF Association studies for finding cancer-susceptibility genetic variants Nat Rev Cancer 2004 4 850 860 15516958 10.1038/nrc1476
Gibbs RA Belmont JW Hardenbol P Willis TD Yu F Yang H Ch'ang LY Huang W Liu B Shen Y Tam PK Tsui LC Waye MM Wong JT Zeng C Zhang Q Chee MS Galver LM Kruglyak S Murray SS Oliphant AR Montpetit A Hudson TJ Chagnon F Ferretti V Leboeuf M Phillips MS Verner A Kwok PY Duan S Lind DL Miller RD Rice JP Saccone NL Taillon-Miller P Xiao M Nakamura Y Sekine A Sorimachi K Tanaka T Tanaka Y Tsunoda T Yoshino E Bentley DR Deloukas P Hunt S Powell D Altshuler D Gabriel SB Zhang H Matsuda I Fukushima Y Macer DR Suda E Rotimi CN Adebamowo CA Aniagwu T Marshall PA Matthew O Nkwodimmah C Royal CD Leppert MF Dixon M Stein LD Cunningham F Kanani A Thorisson GA Chakravarti A Chen PE Cutler DJ Kashuk CS Donnelly P Marchini J McVean GA Myers SR Cardon LR Abecasis GR Morris A Weir BS Mullikin JC Sherry ST Feolo M Altshuler D Daly MJ Schaffner SF Qiu R Kent A Dunston GM Kato K Niikawa N Knoppers BM Foster MW Clayton EW Wang VO Watkin J Gibbs RA Belmont JW Sodergren E Weinstock GM Wilson RK Fulton LL Rogers J Birren BW Han H Wang H Godbout M Wallenburg JC L'Archeveque P Bellemare G Todani K Fujita T Tanaka S Holden AL Lai EH Collins FS Brooks LD McEwen JE Guyer MS Jordan E Peterson JL Spiegel J Sung LM Zacharia LF Kennedy K Dunn MG Seabrook R Shillito M Skene B Stewart JG Valle DL Jorde LB Belmont JW Chakravarti A Cho MK Duster T Foster MW Jasperse M Knoppers BM Kwok PY Licinio J Long JC Marshall PA Ossorio PN Wang VO Rotimi CN Royal CD Spallone P Terry SF Lander ES Lai EH Nickerson DA Altshuler D Bentley DR Boehnke M Cardon LR Daly MJ Deloukas P Douglas JA Gabriel SB Hudson RR Hudson TJ Kruglyak L Kwok PY Nakamura Y Nussbaum RL Royal CD Schaffner SF Sherry ST Stein LD Tanaka T The International HapMap Project Nature 2003 426 789 796 14685227 10.1038/nature02168
Dicioccio RA Song H Waterfall C Kimura MT Nagase H McGuire V Hogdall E Shah MN Luben RN Easton DF Jacobs IJ Ponder BA Whittemore AS Gayther SA Pharoah PD Kruger-Kjaer S STK15 polymorphisms and association with risk of invasive ovarian cancer Cancer Epidemiol Biomarkers Prev 2004 13 1589 1594 15466974
Day N Oakes S Luben R Khaw KT Bingham S Welch A Wareham N EPIC-Norfolk: study design and characteristics of the cohort. European Prospective Investigation of Cancer Br J Cancer 1999 80 Suppl 1 95 103 10466767
Stram DO Haiman CA Hirschhorn JN Altshuler D Kolonel LN Henderson BE Pike MC Choosing haplotype-tagging SNPS based on unphased genotype data using a preliminary sample of unrelated subjects with an example from the Multiethnic Cohort Study Hum Hered 2003 55 27 36 12890923 10.1159/000071807
Schaid DJ Rowland CM Tines DE Jacobson RM Poland GA Score tests for association between traits and haplotypes when linkage phase is ambiguous Am J Hum Genet 2002 70 425 434 11791212 10.1086/338688
Ding K Zhou K He F Shen Y LDA--a java-based linkage disequilibrium analyzer Bioinformatics 2003 19 2147 2148 14594722 10.1093/bioinformatics/btg276
Armitage P Berry G The size of a statistical investigation Statistical Methods in Medical Research 1994 6.6 3rd edition Oxford, Blackwell Scientific Publications 195 206
Cox DG Hankinson SE Kraft P Hunter DJ No association between GPX1 Pro198Leu and breast cancer risk Cancer Epidemiol Biomarkers Prev 2004 13 1821 1822 15533915
Zondervan KT Cardon LR The complex interplay among factors that influence allelic association Nat Rev Genet 2004 5 89 100 14735120 10.1038/nrg1270
Ahmadi KR Weale ME Xue ZY Soranzo N Yarnall DP Briley JD Maruyama Y Kobayashi M Wood NW Spurr NK Burns DK Roses AD Saunders AM Goldstein DB A single-nucleotide polymorphism tagging set for human drug metabolism and transport Nat Genet 2005 37 84 89 15608640
Klein B Weirich G Brauch H DHPLC-based germline mutation screening in the analysis of the VHL tumor suppressor gene: usefulness and limitations Hum Genet 2001 108 376 384 11409863 10.1007/s004390100500
CHEK2*1100delC and Susceptibility to Breast Cancer: A Collaborative Analysis Involving 10,860 Breast Cancer Cases and 9,065 Controls from 10 Studies Am J Hum Genet 2004 74 1175 1182 15122511 10.1086/421251
Rebbeck TR Martinez ME Sellers TA Shields PG Wild CP Potter JD Genetic variation and cancer: improving the environment for publication of association studies Cancer Epidemiol Biomarkers Prev 2004 13 1985 1986 15598750
Colhoun HM McKeigue PM Davey SG Problems of reporting genetic associations with complex outcomes Lancet 2003 361 865 872 12642066 10.1016/S0140-6736(03)12715-8
Cui J Zhou X Chazaro I DeStefano AL Manolis AJ Baldwin CT Gavras H Association of polymorphisms in the promoter region of the PNMT gene with essential hypertension in African Americans but not in whites Am J Hypertens 2003 16 859 863 14553966 10.1016/S0895-7061(03)01026-4
Ioannidis JP Ntzani EE Trikalinos TA 'Racial' differences in genetic effects for complex diseases Nat Genet 2004 36 1312 1318 15543147 10.1038/ng1474
|
16029503
|
PMC1185523
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Jul 19; 5:81
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-81
|
oa_comm
|
==== Front
BMC Cardiovasc DisordBMC Cardiovascular Disorders1471-2261BioMed Central London 1471-2261-5-191599241210.1186/1471-2261-5-19Research ArticleAre antifibrinolytic drugs equivalent in reducing blood loss and transfusion in cardiac surgery? A meta-analysis of randomized head-to-head trials Carless Paul A [email protected] Annette J [email protected] Barrie J [email protected] David A [email protected] Discipline of Clinical Pharmacology, School of Medical Practice and Population Health, Faculty of Health, University of Newcastle, New South Wales, Australia2005 4 7 2005 5 19 19 15 11 2004 4 7 2005 Copyright © 2005 Carless et al; licensee BioMed Central Ltd.2005Carless 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
Aprotinin has been shown to be effective in reducing peri-operative blood loss and the need for re-operation due to continued bleeding in cardiac surgery. The lysine analogues tranexamic acid (TXA) and epsilon aminocaproic acid (EACA) are cheaper, but it is not known if they are as effective as aprotinin.
Methods
Studies were identified by searching electronic databases and bibliographies of published articles. Data from head-to-head trials were pooled using a conventional (Cochrane) meta-analytic approach and a Bayesian approach which estimated the posterior probability of TXA and EACA being equivalent to aprotinin; we used as a non-inferiority boundary a 20% increase in the rates of transfusion or re-operation because of bleeding.
Results
Peri-operative blood loss was significantly greater with TXA and EACA than with aprotinin: weighted mean differences were 106 mls (95% CI 37 to 227 mls) and 185 mls (95% CI 134 to 235 mls) respectively. The pooled relative risks (RR) of receiving an allogeneic red blood cell (RBC) transfusion with TXA and EACA, compared with aprotinin, were 1.08 (95% CI 0.88 to 1.32) and 1.14 (95% CI 0.84 to 1.55) respectively. The equivalent Bayesian posterior mean relative risks were 1.15 (95% Bayesian Credible Interval [BCI] 0.90 to 1.68) and 1.21 (95% BCI 0.79 to 1.82) respectively. For transfusion, using a 20% non-inferiority boundary, the posterior probabilities of TXA and EACA being non-inferior to aprotinin were 0.82 and 0.76 respectively. For re-operation the Cochrane RR for TXA vs. aprotinin was 0.98 (95% CI 0.51 to 1.88), compared with a posterior mean Bayesian RR of 0.63 (95% BCI 0.16 to 1.46). The posterior probability of TXA being non-inferior to aprotinin was 0.92, but this was sensitive to the inclusion of one small trial.
Conclusion
The available data are conflicting regarding the equivalence of lysine analogues and aprotinin in reducing peri-operative bleeding, transfusion and the need for re-operation. Decisions are sensitive to the choice of clinical outcome and non-inferiority boundary. The data are an uncertain basis for replacing aprotinin with the cheaper lysine analogues in clinical practice. Progress has been hampered by small trials and failure to study clinically relevant outcomes.
==== Body
Background
Excessive peri-operative bleeding during cardiac surgery involving cardiopulmonary bypass contributes to overall morbidity and mortality [1-6]. Blood loss frequently leads to transfusion of allogeneic blood products, which expose patients to the risk of transfusion-related adverse effects such as febrile non-hemolytic transfusion reactions, transfusion errors and blood-borne infections [2,7]. Concerns about blood safety, continual blood shortages and rising costs of blood bank operations have generated interest in the reduction of transfusion requirements during and after surgery. A popular approach is to minimize peri-operative bleeding through the prophylactic use of the antifibrinolytic agents aprotinin, tranexamic acid (TXA), and epsilon aminocaproic acid (EACA) [8].
Aprotinin, the benchmark compound, is the most widely used and best established antifibrinolytic medication. It is a non-specific broad-spectrum serine protease inhibitor mainly derived from bovine lungs [9]. TXA and EACA are synthetic lysine analogues, which act principally by blocking lysine binding sites on plasminogen molecules, inhibiting plasmin formation and thereby fibrinolysis [10].
Several published systematic reviews have shown aprotinin to be efficacious in reducing peri-operative blood loss, patient exposure to allogeneic blood transfusion and the need for re-operation due to continued or recurrent bleeding [1,2,11-13]. TXA and EACA also have demonstrated efficacy in placebo-controlled trials [1,2,12,13], but the available literature does not allow conclusions to be drawn about the comparative clinical performance of these agents. It is important to establish the relative performance of these agents as aprotinin is substantially more expensive than either TXA or EACA.
In synthesizing the available literature we were interested in whether TXA and EACA are as effective (i.e. no worse than) as the more expensive drug, aprotinin. To achieve this aim we performed a meta-analysis of data obtained from head-to-head randomized controlled trials of aprotinin, TXA, and EACA and performed tests of equivalence (non-inferiority) using a Bayesian approach.
Methods
Search strategy
This systematic review was undertaken using the methods established by the Cochrane Collaboration [14]. Databases searched were: Medline (1966–September 2003), EMBASE (1980–September 2003), Current Contents (1993–Week 34 2003) and the Cochrane Central Register of Controlled Trials (CENTRAL – The Cochrane Library, Issue 2, 2003). Initially we used unrestricted search strategies, employing exploded MeSH (Medical Subject Headings) terms and specific text-word terms for aprotinin, tranexamic acid, and epsilon aminocaproic acid. The specific text-word terms included: 'aprotinin', 'antilysin', 'contrical', 'kallikrein-trypsin', 'kallikrein inhibitor$', 'kallikrein inactivator$', 'tranexamic acid', 'cyklokapron', 't-amcha', 'amca', 'amcha', urugol', 'transamin', 'kabi', 'exacyl', 'anvitoff', 'epsilon aminocaproic acid', 'amicar', and 'lederle'. The truncation character "$" was used in Medline and EMBASE to retrieve all possible suffix variations of the root word or phrase. In Medline, EMBASE (Excerpta Medica Database), and Current Contents two search filters were used to restrict and improve the specificity of the electronic database searches. Firstly, the ISPOT (International Study of Peri-operative Transfusion) filter [11] which identifies blood transfusion trials, and secondly, a modified version of the Cochrane Collaboration filter [15], which identifies randomized controlled trials. These search filters were combined with the MeSH and relevant text-word terms for aprotinin, TXA, and EACA. Experts in the field were contacted to identify relevant reports or projects in progress relevant to the review. The bibliographies of identified trials, review articles, and reports were searched for potentially relevant studies. Studies were retrieved regardless of language.
Study selection criteria
Two reviewers (PAC and AJM) independently evaluated identified articles for eligibility. Studies were eligible for inclusion if they were randomized parallel-group trials, evaluated the drugs as prophylactic interventions in the context of adult elective cardiac surgery, involved the intravenous administration of the trial agents during the pre and/or intra-operative period, and included in their study outcomes the numbers of individuals who received allogeneic RBC transfusions, or the volume of allogeneic RBCs received by subjects in the intervention groups. Duplicate publications, studies involving only children (less than 18 years), and trials that only administered the study drugs during the post-operative period were not considered for review.
Data extraction
The outcomes measured included: the numbers of patients exposed to allogeneic red blood cell transfusion, and/or the amounts of allogeneic RBC transfused (expressed as units), blood loss (expressed as milliliters), the rates of re-operation for bleeding (re-exploration), non-fatal myocardial infarction, stroke, thrombosis, and mortality. Data were extracted from each trial by two reviewers (PAC and AJM), checked for consistency and accuracy, and then entered into a computer database for analysis.
Data analysis
Dichotomous data (e.g. required re-operation for bleeding or numbers of patients who were transfused) and continuous data (e.g. mean volume of blood loss and mean units of allogeneic RBC transfused) were analyzed using Cochrane Review Manager 4.1 (MetaView 4.1) [16]. Trials were excluded from analysis if they did not report conventional measures of dispersion (standard deviations or standard errors) along with means for continuous data (or if we were unable to calculate these from the raw data). Data expressed in milliliters (mls) of blood transfused were converted to units by dividing by 300. Outcomes are expressed as pooled relative risks (RR) or weighted mean differences (WMD) (for continuous variables) using a random effects model [17]. The Q statistic was used to assess heterogeneity of treatment effect [17]. We also used a Bayesian approach (utilizing WinBUGS software) to model the results of the individual trials as a binomial experiment. We employed a random effects model to calculate the pooled risk ratio, using the methods described by Warn et al.[18]. We used a Uniform (0,1) prior for the risk of allogeneic RBC transfusion with aprotinin treatment (consistent with the reported 50% transfusion rate in cardiac surgery) and estimated a prior for re-operation rates with aprotinin from the results of a published systematic review [12]. We integrated the posterior distribution curve for the RR between various pairs of limits to summarize the probabilities of interest. In doing this we were indifferent to the probability of superiority of lysine analogues over aprotinin, but included those areas of the curves in the calculation of the probabilities of non-inferiority. We selected a non-inferiority boundary of 20% (delta value) for re-operation data and the rate of transfusion with allogeneic blood (i.e. TXA & EACA were considered non-inferior to aprotinin if the upper limit of the 95% CI for the pooled RR was ≤1.2). The delta value was varied during sensitivity analysis (i.e. 5% to 40%).
Assessment of study methodological quality
Studies were assessed for methodological quality by two independent raters (PAC and AJM), using criteria proposed by Schulz et al.[19]. These specify four items of assessment: double-blinding, allocation concealment, participant inclusion/exclusion and methods used to achieve randomization. Disagreements were resolved by consensus. Inter-rater agreement for each item of methodological quality was assessed by comparing the observed agreement with that expected by chance. A kappa statistic (which expresses the agreement beyond chance as a proportion of the maximum possible agreement) was calculated for each item assessed. Kappa is equal to one when there is perfect agreement between raters.
Results
We identified twenty randomized, head-to-head trials involving comparisons of aprotinin TXA and EACA in elective adult cardiac surgery, which reported information on the main outcomes of interest [6,20-38]. One trial [39] was excluded from the analysis due to a lack of usable data (i.e. for continuous data the results were reported as medians [25th–75th percentiles]; the number of patients transfused ≥1 unit of allogeneic RBC transfusion was not reported).
Characteristics of included studies
The 20 included trials (Tables 1 &2) randomized a total of 2430 subjects to receive either aprotinin, TXA, or EACA. The majority (n = 11) compared aprotinin to TXA. There were only three head-to-head trials of aprotinin versus EACA, three trials of TXA versus EACA, and three trials that compared the three antifibrinolytic drugs with each other. The median size of trial arms was 25 participants (range; 14–522). For each of the three intervention groups the mean age of study participants ranged from 60.5–62.4 years. Most study participants were male (77–79%). The publication period of the trials spanned nine years (1993 to 2001). Only one trial was published in a language other than English and was translated before being included in the analysis [23]. The trials were heterogeneous in terms of drug dose and treatment regimen (Table 3).
Table 1 Characteristics of Included Studies
Study Year Country Type of cardiac surgery Interventions
Isetta et al. [25] 1993 France NR HD APR (n = 70) vs. LD APR (n = 70) vs. TXA (n = 70) vs. Control (n = 70)
Blauhut et al. [27] 1994 Switzerland CABG HD APR (n = 14) vs. TXA (n = 14) vs. Control (n = 14)
Penta de Peppo et al. [20] 1995 Italy CABG & Valve Sx. HD APR (n = 15) vs. TXA (n = 15) vs. EACA (n = 15) vs. Control (n = 15)
Corbeau et al. [23] 1995 France CABG & Valve Sx. HD APR (n = 43) vs. TXA (n = 41) vs. Control (n = 20)
Pugh et al. [22] 1995 UK Primary CABG LD APR (n = 21) vs. TXA (n = 22) vs. Control (n = 23)
Speekenbrink et al. [21] 1995 The Netherlands Primary CABG PP APR (n = 15) vs. TXA (n = 15) vs. DIP (n = 12) vs. Control (n = 15)
Menichetti et al. [24] 1996 Italy Primary CABG HD APR (n = 24) vs. TXA (n = 24) vs. EACA (n = 24) vs. Control (n = 24)
Pinosky et al. [33] 1997 USA Primary CABG TXA (n = 20) vs. EACA (n = 20) vs. Placebo (n = 19)
Mongan et al. [31] 1998 USA Primary CABG HD APR (n = 75) vs. TXA (n = 75)
Hardy et al. [26] 1998 Canada Primary CABG TXA (n = 42) vs. EACA (n = 46) vs. Placebo (n = 44)
Eberle et al. [29] 1998 Germany Primary CABG HD APR (n = 20) vs. EACA (n = 20)
Misfeld et al. [30] 1998 Germany Primary CABG LD APR (n = 14) vs. TXA (n = 14) vs. Control (n = 14)
Casati et al. [28] 1999 Italy Primary CABG & Valve Sx. HD APR (n = 67) vs. TXA (n = 70) vs. EACA (n = 66)
Bernet et al. [34] 1999 Switzerland Primary CABG HD APR (n = 28) vs. TXA (n = 28)
Nuttall et al. [32] 2000 USA Re-do CABG & Valve Sx. HD APR (n = 40) vs. TXA (n = 45) vs. TXA+ANH (n = 32) vs Placebo (n = 43)
Maineri et al. [38] 2000 Italy Primary CABG TXA (n = 24) vs. EACA (n = 24)
Wong et al. [37] 2000 Canada Re-do CABG & Valve Sx. HD APR (n = 39) vs. TXA (n = 38)
Casati et al. [35] 2000 Italy Primary CABG & Valve & ASD Repair HD APR (n = 518) vs. TXA (n = 522)
Greilich et al. [36] 2001 USA Primary CABG HD APR (n = 24) vs. EACA (n = 23) vs. Placebo (n = 25)
Ray et al. [6] 2001 Australia CABG & Valve Sx. LD APR (n = 49) vs. EACA (n = 51)
ANH = acute normovolemic hemodilution, APR = aprotinin, ASD = atrial septal defect, CABG = coronary artery bypass graft, DIP = dipyridamole, EACA = epsilon aminocaproic acid, HD = high dose, LD = low dose, NR = not reported, PP = pump prime, Sx. = surgery, TXA = tranexamic acid
Table 2 Characteristics of Included Studies
Study Co-interventions Transfusion threshold Anti-platelet use
Isetta et al. [25] PO CS - re-transfusion of SMB Hct<20% during CPB
Hct<25% 4 hrs post CPB
Hct<27% post-op. NR
Blauhut et al. [27] NR Hct<30% post-op. Excluded pts. pre-operatively treated with ASA + NSAIDs
Penta de Peppo et al. [20] IO CS + IO & PO re-transfusion of SMB Post-op. non-monitored pts. Hb<7.0 g/dL
Monitored pts. Hb<8.5 g/dL Discontinued NSAIDs 24 hrs before Sx.
Corbeau et al. [23] NR Hct<20% during CPB
Hct<25% at the end of surgery
Hct<30 post extubation Anti-platelet aggregation drugs ceased 10 days pre-operatively
Pugh et al. [22] IO CS + ANH (1 unit of WB collected pre-CPB then re-transfused post CPB) Hct<20% during CPB
Hct<30% off CPB Aspirin use within 10 days of the operation: LD APR = 67%, TXA = 91%, Control = 78%
Speekenbrink et al. [21] NR NR Aspirin discontinued 2–4 days before Sx.
Menichetti et al. [24] NR Hct<30% post-operatively Excluded pts. who had taken ASA or DIP until 2 weeks pre-op.
Pinosky et al. [33] NR Hct<20% + surgeon preference Pre-operative aspirin use: TXA = 25%, EACA = 40%, Placebo = 42%
Mongan et al. [31] NR Hb<6.0 g/dL during CPB
Hb<8.0 g/dL off CPB Pre-operative aspirin use: HD APR = 44%, TXA = 53%
Hardy et al. [26] IO CS & Re-infusion of SMB were not used Hb<7.0 g/dL during CPB
Hb<8.0 g/dL off CPB NR
Eberle et al. [29] IO & PO CS used Hct<27% - post-operative + accompanied by signs & symptoms of hypovolemia Intra-operative IV ASA: HD APR = 5.0%, EACA = 15%
Misfeld et al. [30] NR Hb<8.0 g/dL Excluded pts. receiving ASA treatment within 5 days of Sx.
Casati et al. [28] IO CS used + PAD Hb<6.0 g/dL during CPB
Hb<8.0 g/dL off CPB + clinical condition Pts. receiving ASA treatment within 5 days of Sx.: HD APR = 37.8%, TXA = 40.9%, EACA = 35.3%
Bernet et al. [34] PO CS Hct<25% PO All pts. were treated with 100 mg ASA daily until Sx.
Nuttall et al. [32] PAD not used Hb<7.0 g/dL during CPB Excluded pts. taking ASA daily (≥325 mg) before Sx.
Maineri et al. [38] IO CS + PO re-infusion of SMB Hct<30% IO
Hct<28% PO NR
Wong et al. [37] IO CS + PO re-infusion of SMB Hb<7.0 g/dL IO
Hb<8.0 g/dL PO Excluded pts. receiving ASA treatment within 5 days of Sx.
Casati et al. [35] IO CS used Hb<6.0 g/dL during CPB
Hb<8.0 g/dL PO Pts. receiving ASA treatment before Sx.: HD APR = 17.8%, TXA = 18.8%
Greilich et al. [36] IO CS used PO SMB was not used Hb<8.0 g/dL Pts. receiving ASA treatment before Sx.: HD APR = 88%, EACA = 90%, Placebo = 79%
Ray et al. [6] NR NR ASA within 10 days before Sx.: LD APR = 22.4%, EACA = 33.3%
ANH = acute normovolemic hemodilution, APR = aprotinin, ASA = acetylsalicylic acid, CABG = coronary artery bypass graft, CPB = cardiopulmonary bypass, CS = cell salvage, DIP = dipyridamole, EACA = epsilon aminocaproic acid, Hb = hemoglobin, Hct = hematocrit, HD = high dose, LD = low dose, NR = not reported, NSAIDs = non-steroidal anti-inflammatory drugs, PP = pump prime, IO = intra-operative, PO = post-operative, SMB = shed mediastinal blood, Sx. = surgery, TXA = tranexamic acid, WB = whole blood
Table 3 Summary of drug dose and treatment regimens
Study Aprotinin TXA EACA
Isetta et al. [25] L = 2.0 × 106 KIU
M = 0.5 × 106 KIU/h
P = 2.0 × 106 KIU L = 15 mg/kg NS
L = 0.5 × 106
M = 0.5 × 106
Blauhut et al. [27] L = 2.0 × 106 KIU
M = 0.5 × 106 KIU/h
P = 1.0 × 106 KIU L = 10 mg/kg
M = 1.0 mg/kg/h NS
Penta de Peppo et al. [20] L = 2.0 × 106 KIU
M = 0.5 × 106 KIU/h
P = 2.0 × 106 KIU L = 10 mg/kg M = 1.0 mg/kg/h L = 10 g
M = 2.0 g/h for 5 h
Corbeau et al. [23] L = 2.0 × 106 KIU
M = 0.5 × 106 KIU/h
P = 2.0 × 106 KIU L = 15 mg/kg
E = 15 mg/kg NS
Pugh et al. [22] L = 1.0 × 106 KIU
P = 1.0 × 106 KIU L = 2.5 g
P = 2.5 g NS
Speekenbrink et al. [21] P = 2.0 × 106 KIU L = 10 mg/kg
M = 1.0 mg/kg/h NS
Menichetti et al. [24] L = 2.0 × 106 KIU
M = 0.5 × 106 KIU/h
P = 2.0 × 106 KIU L = 10 mg/kg
M = 3.0 mg/kg/h
P = 10 mg/kg L = 80 mg
M= 30 mg/kg/h
P = 80 mg/kg
Pinosky et al. [33] NS L = 15 mg/kg
M = 1.0 mg/kg/h for 6 h L = 150 mg/kg
M = 10 mg/kg/h for 6 h
Mongan et al. [31] L = 2.0 x 106 KIU
M = 0.5 × 106 KIU/h
P = 2.0 × 106 KIU L = 15 mg/kg
M = 2.0 mg/kg/h for 6 h NS
Hardy et al. [26] NS L = 10 g L = 15 g
M = 1.0 g/h
Eberl et al. [29] L = 2.0 × 106 KIU
M = 0.5 × 106 KIU/h
P = 2.0 × 106 KIU NS L = 10 g
M = 2.5 g/h
P = 10 g
Misfeld et al. [30] P = 1.0 × 106 KIU
E = 0.2 × 106 KIU/h for 5 h L = 10 mg/kg
M = 1.0 mg/kg/h NS
Casati et al. [28] L = 2.0 × 106 KIU
M = 0.5 × 106 KIU/h
P = 2.0 × 106 KIU L = 1.0 g
M = 400 mg/h
P = 500 mg L = 5.0 g
M = 2.0 g/h
P = 2.5 g
Bernet et al. [34] L = 2.0 × 106 KIU
M = 0.5 × 106 KIU/h
P = 2.0 × 106 KIU L = 10 g NS
Nuttall et al. [32] L = 2.0 × 106 KIU
M = 0.5 × 106 KIU/h
P = 2.0 × 106 KIU L = 10 mg/kg
M = 1.0 mg/kg/h NS
Maineri et al. [38] NS L = 20 mg/kg
M = 2.0 mg/kg/h L = 10 g
M = 2.0 g/h
Wong et al. [37] L = 2.0 × 106 KIU
M = 0.5 × 106 KIU/h
P = 2.0 × 106 KIU L = 10 g NS
Casati et al. [35] L = 2.0 × 106 KIU
M = 0.5 × 106 KIU/h
P = 2.0 × 106 KIU L = 1.0 g
M = 400 mg/h
P = 500 mg NS
Greilich et al. [36] L = 2.0 × 106 KIU
M = 0.5 × 106 KIU/h
P = 2.0 × 106 KIU NS L = 100 mg/kg
M = 2.5 mg/kg/h
P = 5.0 g
Ray et al. [6] L = 1.0 × 106 KIU
P = 1.0 × 106 KIU NS L = 5.0 g
M = 1.25 g/h
P = 5.0 g
L = loading dose, M = maintenance dose/continuous infusion, P = pump prime dose, E = after protamine administration, KIU = kallikrein inhibitor units, NS = not studied, mg = milligram, g = gram, kg = kilogram, h = hour
Methodological quality of the studies
Nineteen of the 20 trials were assessed for methodological quality by the two raters (PAC and AJM). As the non-English language study [23] could not be adequately assessed by the two raters, it was excluded from the analysis of the reliability of quality assessment procedure. For the four items of the Schulz criteria [19] used to assess trial quality, the observed agreement was good with kappa scores ranging from 0.92 to 1.0. Generally, the methodological quality of the trials reviewed was poor. Double-blinding was reported in eight trials (42%), concealment of treatment allocation was judged to be adequate in four trials (21%), and only four trials (21%) described the method used to generate allocation sequences (i.e. randomization procedure). Follow-up was judged to have been complete in five trials (26%). For seven trials there was incomplete follow-up; however for these trials only a small number of exclusions were reported making differential withdrawal an unlikely source of bias. For the seven remaining trials a rationale for the withdrawal of study subjects was not provided. As the majority of trials were of poor methodological quality stratification of the data by methodological quality and subgroup analyses were uninformative. We were therefore unable to determine whether treatment effect estimates varied due to study methodological quality.
Meta-analyses
TXA vs. Aprotinin (10 trials, 1707 subjects)
On average, TXA was inferior to aprotinin in reducing 24 hour blood loss (WMD 106 mls, 95% CI 37 to 176 mls; Fig 1). This apparent disadvantage of TXA was not reflected in the transfusion data. For the five trials (N = 357 subjects) that reported on the amount of blood transfused, the mean numbers of red cell units did not differ between the two drugs; WMD 0.06 units (95% CI -0.18 to 0.31 units). The rate of red cell transfusion in patients treated with TXA was 37.2% compared with 36.5% with aprotinin (Cochrane RR 1.08, 95% CI 0.88 to 1.32; Fig 2). The equivalent Bayesian posterior mean relative risk was 1.11 (95% BCI 0.92 to 1.45). Data on re-operation rates were sparse (Fig 3). The Cochrane estimate of the pooled RR for re-operation with TXA compared to aprotinin was close to one (RR 0.98, 95% CI 0.51 to 1.88). In contrast, the Bayesian posterior mean risk ratio was 0.63 (95% BCI 0.16 to 1.46). Most of the difference between TXA and aprotinin seemed to be contributed by one study (Nuttall et al.[32]). This study reported re-operation rates of 0/45 with tranexamic acid and 6/45 with aprotinin, equating to an absolute risk reduction of 13% [risk difference (RD) -0.13, 95% CI -0.24 to -0.03]. In comparison, none of the remaining trials reached statistical significance for this outcome with the risk differences ranging from -0.03 to 0.07 and the 95% confidence intervals including unity (RD = 0). Excluding the data from this one trial [32] changed the mean posterior RR to 0.93 (95% BCI 0.30 to 1.96).
Figure 1 Forest plot of 10 comparative trials of TXA and aprotinin – weighted mean difference in blood loss.
Figure 2 Forest plot of 10 comparative trials of TXA and aprotinin – pooled relative risk of requiring an allogeneic red cell transfusion.
Figure 3 Forest plot of 9 comparative trials of TXA and aprotinin – pooled relative risk of needing re-operation for bleeding.
For RBC transfusion the estimated posterior probability of non-inferiority TXA to aprotinin (with a pooled RR threshold of 1.2) was 0.82. If the threshold was set to 1.1 the posterior probability of non-inferiority was 0.57 (Fig 4). The probabilities of non-inferiority of TXA for re-operation were higher than for transfusion, being 0.92 and 0.90 for the delta values of 20% and 10% respectively, but fell to 0.69 and 0.64 when the data from Nuttall et al.[32] were excluded from the calculations.
Figure 4 Posterior probability of TXA being considered non-inferior to aprotinin at different delta values (transfusion data).
EACA vs. Aprotinin (6 trials; 399 subjects)
EACA was inferior to aprotinin in controlling blood loss over 24 hours (WMD 184 mls, 95% CI 134 to 235 mls; Fig 5). However, the mean number of units of allogeneic RBC transfused did not differ between the drugs (WMD -0.22 units, 95% CI -0.52 to 0.09 units). Transfusion rates were similar for EACA and aprotinin: Cochrane RR 1.14 (95% CI 0.84 to 1.55); Bayesian posterior mean risk ratio 1.08 (95% BCI 0.73 to 1.52). Using a non-inferiority threshold value of 1.2 for the pooled RR, the probability of EACA being non-inferior to aprotinin was 0.76. With the threshold set at 1.1 the posterior probability of non-inferiority dropped to 0.54 (Fig 6). There were insufficient data to analyze the effects of treatment on re-operation rates.
Figure 5 Forest plot of 4 comparative trials of EACA and aprotinin – weighted mean difference in blood loss.
Figure 6 Posterior probability of EACA being considered non-inferior to aprotinin at different delta values (transfusion data).
Other outcomes
Analyses of other clinical outcomes such as all cause mortality, myocardial infarction and stroke were generally uninformative because of the sparse data, but we saw no trends favoring any of the drugs studied here, compared with the others (data not displayed).
Direct comparisons between TXA and EACA revealed no clinically meaningful or significant differences therefore we did not perform non-inferiority tests for these two agents.
Discussion
Aprotinin has become a widely used adjunct in cardiac surgery [2], a practice that is supported by the results of a large number of placebo-controlled trials [1,11,12]. These trials have demonstrated reductions in allogeneic red cell transfusion, and the need for re-operation due to bleeding. Placebo-controlled trials of tranexamic acid (TXA) and epsilon aminocaproic acid (EACA) have also demonstrated efficacy, but the data are sparse and it is unclear from the published indirect comparisons whether they are as effective as aprotinin [11,12]. This is not an academic question as both agents are substantially cheaper than aprotinin. For example, an average course of treatment with aprotinin in Canada costs CAN$1000, compared with CAN$100-275 for TXA and approximately CAN$50 for EACA [37].
So there are financial pressures to switch from aprotinin to the synthetic lysine analogues. But this should only be contemplated if there is a high degree of confidence that the treatments are clinically equivalent. Conventional meta-analysis provides pooled estimates of differences between treatments (with uncertainty reflected in the width of the confidence intervals). But to demonstrate an acceptable level of 'equivalence' we need to estimate and interpret the probability of a drug's efficacy lying within a 'non-inferiority' boundary [40]. We have to make a judgment about what level of non-inferiority is acceptable, and agree on a tolerable probability of breaching this threshold. These are difficult judgments and we accept that our approach is somewhat arbitrary.
In this paper we used the rates of blood loss, transfusion with allogeneic red cells and re-operation due to continued or recurrent bleeding as the outcome variables. Adequate mortality and morbidity data were not available from the trials. Both lysine analogues seemed inferior to aprotinin in controlling peri-operative blood loss, but the increments were small (between 100 and 200 mls), and of uncertain clinical significance. In the case of red cell transfusions we set the non-inferiority boundaries at 1.2 (a relative 20% increase) in the base case analyses. The rate of transfusion for aprotinin-treated patients in these trials was around 35%, therefore a non-inferiority threshold of 1.2 translates into an absolute increase of around 6.9% in transfusion frequency in this population. In the case of TXA the probability of non-inferiority with this threshold was 0.82, but was slightly lower in the case of EACA (0.76) because of sparse data. To achieve a higher level of confidence in the 'equivalence' of TXA, for example 90%, it is necessary to tolerate a non-inferiority boundary of 1.4 – an absolute increase in the transfusion rate of around 12%. It is difficult to know how this will be viewed by clinicians, but some may consider it as an unsatisfactory basis for switching from a drug of proven efficacy.
As blood transfusion is a practice variable, as opposed to a clinical end-point variable, it requires a degree of subjectivity on the part of clinicians. The decision to transfuse is complex and sometimes arbitrary. It will be influenced by local transfusion protocols, the patient's pre-operative hemoglobin (Hb), the estimated degree of blood loss and the presence of co-morbidity (particularly coronary disease). We do not think that such decisions are likely to be sensitive to the modest differences in blood loss reviewed here, in fact that is what the data indicate.
Our analyses encouraged us to have greater confidence in the equivalence of TXA to aprotinin in preventing the need for re-operation than the need for transfusion. But we remain uncertain about these data. For re-operation, with the threshold for the pooled RR set at 1.2, the probability of TXA being non-inferior to aprotinin was 0.92. This is moderately higher than the probability of 0.82 for RBC transfusion. This is because the Bayesian estimate for the posterior mean RR was 0.63, with a high proportion of the posterior probability distribution below a value of 1.0. Consequently, the integrated area below the non-inferiority boundary of 1.2 was high. Re-operation was uncommon in this population, being required by only 2.5% of aprotinin recipients. Although the point estimates of the RR suggested a trend in favor of TXA (not seen for other outcomes), the confidence intervals were wide and the results changed (unfavorably for TXA) when a single small trial (Nuttall et al.[32]), which contributed disproportionately to the difference between the drugs, was excluded from calculation. In addition, these trends are not paralleled by improvements in blood loss (which was worse with TXA than with aprotinin) or transfusion requirements. For these reasons we think that the findings should be interpreted cautiously.
Heterogeneity in trial outcomes was not particularly prominent in our analyses. For the main study outcome (i.e. number of patients transfused allogeneic blood) heterogeneity was not statistically significant (TXA vs. aprotinin, p = 0.13; EACA vs. aprotinin p = 0.55). Although the results for blood loss indicated statistically significant heterogeneity (TXA vs. aprotinin, p = 0.0005) it appears that the data from one trial contributed to this result (Menichetti et al., 1996). When the data from this trial were removed from the analysis heterogeneity was no longer significant (p = 0.29).
We were unable to formally assess the impact that the use of anti-platelet agents had on treatment effect estimates as the majority of trials either excluded patients that had been treated with acetylsalicyclic acid (ASA) or dipyridamole (DIP) within 5–10 days of surgery or discontinued treatment with these agents pre-operatively to avoid excessive bleeding. However, in those trials that included ASA or DIP treated patients generally treatment with these agents was evenly distributed across trial arms.
Stratification of trial data by the use of cell salvage proved only marginally informative. Subgroup analysis indicated that for the six trials that used cell salvage the pooled relative risk of receiving an allogeneic RBC transfusion in those patients treated with TXA was 0.97 (95%CI 0.84 to 1.12) compared to 1.54 (95%CI 0.82 to 2.91) for the four studies that did not report the use of cell salvage. Although there appeared to be a trend toward a reduced risk of transfusion in those trials that used cell salvage both results failed to reach statistical significance with the 95% confidence intervals crossing unity. For EACA subgroup analysis was uninformative due to the small number of trials.
Conclusion
The conclusions that can be drawn from these data are limited for a number of other reasons. The studies were of generally poor quality. This is regrettable as trials of drugs are generally easier to conduct well than trials of different transfusion thresholds or surgical techniques. We have not examined the data for publication bias and are uncertain what effect this might have as the trial comparisons involved active treatments. We have not explored heterogeneity in detail, but it was not particularly prominent in these analyses. The main limitation was the small size of the trials and the reliance on transfusion rates rather than more clinically meaningful endpoints. Doubts about the clinical performance of a treatment are tolerable when the clinical consequences are slight. However, when the result of treatment failure is an unplanned visit to the operating theatre and a further sternotomy or thoracotomy to deal with the source of continued bleeding we need assurance about the equivalence of our treatment choices. In our view the data reviewed here do not provide this reassurance and larger comparative studies using clinically important endpoints are necessary.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
PAC conducted the literature search, screened articles for eligibility, assessed methodological quality of included studies, extracted data, analyzed data, interpreted results, and wrote manuscript. AJM screened articles for eligibility, assisted with data extraction and methodological assessment of included studies. BJS performed Bayesian analysis of data and provided statistical consultancy for this project. DAH conceived study project and provided critique of successive drafts of the manuscript. All listed authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This research was supported by a grant from the National Health and Medical Research Council of Australia.
==== Refs
Munoz JJ Birkmeyer NJ Birkmeyer JD O'Connor GT Dacey LJ Is epsilon-aminocaproic acid as effective as aprotinin in reducing bleeding with cardiac surgery?: a meta-analysis Circulation 1999 99 81 89 9884383
Levi M Cromheecke ME de Jonge E Prins MH de Mol BJM Briet E Buller HR Pharmacological strategies to decrease excessive blood loss in cardiac surgery: a meta-analysis of clinically relevant endpoints [Review] Lancet 1999 354 1940 1947 10622296 10.1016/S0140-6736(99)01264-7
Smith CR Management of bleeding complications in redo cardiac operations. [Review] [29 refs] Ann Thorac Surg 1998 65 S2 8; discussion S27-8 9563396 10.1016/S0003-4975(98)00070-8
Rich JB The efficacy and safety of aprotinin use in cardiac surgery. [Review] [20 refs] Ann Thorac Surg 1998 66 S6 11 9869434 10.1016/S0003-4975(98)00964-3
Bennett-Guerrero E Spillane WF White WD Muhlbaier LH Gall SAJ Smith PK Newman MF Epsilon-aminocaproic acid administration and stroke following coronary artery bypass graft surgery Ann Thorac Surg 1999 67 1283 1287 10355397 10.1016/S0003-4975(99)00116-2
Ray MJ O'Brien MF Comparison of epsilon aminocaproic acid and low-dose aprotinin in cardiopulmonary bypass: efficiency, safety and cost Ann Thorac Surg 2001 71 838 843 11269462 10.1016/S0003-4975(00)02229-3
Cohen G Ivanov J Weisel RD Rao V Mohabeer MK Mickle DA Aprotinin and dipyridamole for the safe reduction of postoperative blood loss Ann Thorac Surg 1998 65 674 683 9527194 10.1016/S0003-4975(97)01428-8
Fergusson D van Walraven C Coyle D Laupacis A Economic evaluations of technologies to minimize perioperative transfusion: a systematic review of published studies. International Study of Peri-operative Transfusion (ISPOT) investigators Transfus Med Rev 1999 13 106 117 10218234
Fritz H Wunderer G Biochemistry and applications of aprotinin, the kallikrein inhibitor from bovine organs. [Review] [250 refs] Arzneimittel-Forschung 1983 33 479 494 6191764
Dunn CJ Goa KL Tranexamic acid: a review of its use in surgery and other indications. [Review] [132 refs] Drugs 1999 57 1005 1032 10400410
Laupacis A Fergusson D Drugs to minimize perioperative blood loss in cardiac surgery: meta-analyses using perioperative blood transfusion as the outcome. The International Study of Peri-operative Transfusion (ISPOT) Investigators Anesth Analg 1997 85 1258 1267 9390590 10.1097/00000539-199712000-00014
Henry DA Moxey AJ Carless PA O'Connell D McClelland B Henderson KM Sly K Laupacis A Fergusson D Anti-fibrinolytic use for minimising perioperative allogeneic blood transfusion. [Review] [179 refs] Cochrane Database Syst Rev 2003 CD001886
Fremes SE Wong BI Lee E Mai R Christakis GT McLean RF Goldman BS Naylor CD Metaanalysis of prophylactic drug treatment in the prevention of postoperative bleeding Ann Thorac Surg 1994 58 1580 1588 7526811
Clarke M Oxman AD Cochrane Reviews' Handbook 4.0 [updated July 1999] 2000 2000 Oxford.UK., The Cochrane Collaboration.
Dickersin K Larson K Establishing and maintaining an international register of RCTs. 1996 Oxford.United Kingdom.(UK), Cochrane Collaboration
The Cochrane Collaboration Software Development Group Review Manager Software - MetaView 4.1 2000 Oxford.UK., The Cochrane Collaboration
DerSimonian R Laird N Meta-analysis in clinical trials Control Clin Trials 1986 7 177 188 3802833 10.1016/0197-2456(86)90046-2
Warn DE Thompson SG Spiegelhalter DJ Bayesian random effects meta-analysis of trials with binary outcomes: methods for the absolute risk difference and relative risk scales Stat Med 2002 21 1601 1623 12111922 10.1002/sim.1189
Schulz KF Chalmers I Hayes RJ Altman DG Empirical evidence of bias. Dimensions of methodological quality associated with estimates of treatment effects in controlled trials JAMA 1995 273 408 412 7823387 10.1001/jama.273.5.408
Penta de Peppo A Pierri MD Scafuri A De Paulis R Colantuono G Caprara E Tomai F Chiariello L Intraoperative antifibrinolysis and blood-saving techniques in cardiac surgery. Prospective trial of 3 antifibrinolytic drugs Texas Heart Institute Journal 1995 22 231 236 7580360
Speekenbrink RG Vonk AB Wildevuur CR Eijsman L Hemostatic efficacy of dipyridamole, tranexamic acid, and aprotinin in coronary bypass grafting Annals of Thoracic Surgery 1995 59 438 442 7531423 10.1016/0003-4975(94)00865-5
Pugh SC Wielogorski AK A comparison of the effects of tranexamic acid and low-dose aprotinin on blood loss and homologous blood usage in patients undergoing cardiac surgery Journal of Cardiothoracic and Vascular Anesthesia 1995 9 240 244 7545448
Corbeau JJ Monrigal JP Jacob JP Cottineau C Moreau X Bukowski JG Subayi JB Delhumeau A COMPARAISON DES EFFETS DE L'APROTININE ET DE L'ACIDE TRANEXAMIQUE SUR LE SAIGNEMENT EN CHIRURGIE CARDIAQUE Comparative effects of aprotinin and tranexamic acid on blood loss in cardiac surgery Annales Francaises d'Anesthesie et de Reanimation 1995 14 154 161 7486272
Menichetti A Tritapepe L Ruvolo G Speziale G Cogliati A Di Giovann C Pacilli M Criniti A Changes in coagulation patterns, blood loss and blood use after cardiopulmonary bypass: aprotinin vs tranexamic acid vs epsilon aminocaproic acid J Cardiovasc Surg (Torino) 1996 37 401 407 8698787
Isetta C Gunness TK Samat C Paolini G Lugrin D Sanchez B Jourdan J Antifibrinolytic Treatment and Homologeous Transfusion in Cardiac Surgery. European Heart Journal 1993 14 424
Hardy JF Belisle S Dupont C Harel F Robitaille D Roy M Gagnon L Prophylactic tranexamic acid and epsilon-aminocaproic acid for primary myocardial revascularization Ann Thorac Surg 1998 65 371 376 9485231 10.1016/S0003-4975(97)01016-3
Blauhut B Harringer W Bettelheim P Doran JE Spath P Lundsgaard-Hansen P Comparison of the effects of aprotinin and tranexamic acid on blood loss and related variables after cardiopulmonary bypass Journal of Thoracic & Cardiovascular Surgery 1994 108 1083 1091 7527112
Casati V Guzzon D Oppizzi M Cossolini M Torri G Calori G Alfieri O. Hemostatic effects of aprotinin, tranexamic acid and epsilon- aminocaproic acid in primary cardiac surgery Ann Thorac Surg 1999 68 2252 2256 10617012 10.1016/S0003-4975(99)00866-8
Eberle B Mayer E Hafner G Heinermann J Dahm M Prellwitz W Dick W Oelert H High-dose epsilon-aminocaproic acid versus aprotinin: antifibrinolytic efficacy in first-time coronary operations Ann Thorac Surg 1998 65 667 673 9527193 10.1016/S0003-4975(97)01424-0
Misfeld M Dubbert S Eleftheriadis S Siemens HJ Wagner T Sievers HH Fibrinolysis-adjusted perioperative low-dose aprotinin reduces blood loss in bypass operations Ann Thorac Surg 1998 66 792 799 9768932 10.1016/S0003-4975(98)00646-8
Mongan PD Brown RS Thwaites BK Tranexamic acid and aprotinin reduce postoperative bleeding and transfusions during primary coronary revascularization Anesth Analg 1998 87 258 265 9706913 10.1097/00000539-199808000-00005
Nuttall GA Oliver WC Ereth MH Santrach PJ Bryant SC Orszulak TA Schaff HV Comparison of blood-conservation strategies in cardiac surgery patients at high risk for bleeding Anesthesiology 2000 92 674 682 10719945 10.1097/00000542-200003000-00010
Pinosky ML Kennedy DJ Fishman RL Reeves ST Alpert CC Ecklund J Kribbs S Spinale FG Kratz JM Crawford R Gravlee GP Dorman BH Tranexamic acid reduces bleeding after cardiopulmonary bypass when compared to epsilon aminocaproic acid and placebo J Card Surg 1997 12 330 338 9635271
Bernet F Carrel T Marbet G Skarvan K Stulz P Reduction of blood loss and transfusion requirements after coronary artery bypass grafting: similar efficacy of tranexamic acid and aprotinin in aspirin-treated patients J Card Surg 1999 14 92 97 10709819
Casati V Guzzon D Oppizzi M Bellotti F Franco A Gerli C Cossolini M Torri G Calori G Benussi S Alfieri O Tranexamic acid compared with high-dose aprotinin in primary elective heart operations: effects on perioperative bleeding and allogeneic transfusions J Thorac Cardiovasc Surg 2000 120 520 527 10962414 10.1067/mtc.2000.108016
Greilich PE Okada K Latham P Kumar RR Jessen ME Aprotinin but not epsilon-aminocaproic acid decreases interleukin-10 after cardiac surgery with extracorporeal circulation: randomized, double-blind, placebo-controlled study in patients receiving aprotinin and epsilon-aminocaproic acid Circulation 2001 104 I265 I269 11568067
Wong BI McLean RF Fremes SE Deemar KA Harrington EM Christakis GT Goldman BS Aprotinin and tranexamic acid for high transfusion risk cardiac surgery Ann Thorac Surg 2000 69 808 816 10750765 10.1016/S0003-4975(99)01419-8
Maineri P Covaia G Realini M Caccia G Ucussich E Luraschi M Crosta A Foresti B Chiaranda M Postoperative bleeding after coronary revascularization. Comparison between tranexamic acid and epsilon-aminocaproic acid Minerva Cardioangiologica 2000 48 155 160 11048468
Bennett-Guerrero E Sorohan JG Gurevich ML Kazanjian PE Levy RR Barbera AV White WD Slaughter TF Sladen RN Smith PK Newman MF Cost-benefit and efficacy of aprotinin compared with epsilon- aminocaproic acid in patients having repeated cardiac operations: a randomized, blinded clinical trial Anesthesiology 1997 87 1373 1380 9416723 10.1097/00000542-199712000-00017
Wiens BL Choosing an equivalence limit for noninferiority or equivalence studies.[see comment][erratum appears in Control Clin Trials 2002 Dec;23(6):774] Control Clin Trials 2002 23 2 14 11852160 10.1016/S0197-2456(01)00196-9
|
15992412
|
PMC1185524
|
CC BY
|
2021-01-04 16:30:06
|
no
|
BMC Cardiovasc Disord. 2005 Jul 4; 5:19
|
utf-8
|
BMC Cardiovasc Disord
| 2,005 |
10.1186/1471-2261-5-19
|
oa_comm
|
==== Front
BMC Cardiovasc DisordBMC Cardiovascular Disorders1471-2261BioMed Central London 1471-2261-5-201600883210.1186/1471-2261-5-20Research ArticlePredictors and prognosis of paroxysmal atrial fibrillation in general practice in the UK Ruigómez Ana [email protected] Saga [email protected] Mari-Ann [email protected]ía Rodríguez Luis Alberto [email protected] Centro Español de Investigación Farmacoepidemiológica (CEIFE), Madrid, Spain2 AstraZeneca R&D Mölndal, Sweden3 Section of Preventive Cardiology, Göteborg University, Sweden4 Department of Public Health and Caring Science, Uppsala University, Sweden2005 11 7 2005 5 20 20 18 1 2005 11 7 2005 Copyright © 2005 Ruigómez et al; licensee BioMed Central Ltd.2005Ruigómez 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
Natural history of paroxysmal atrial fibrillation (AF) is not very well documented. Clinical experience suggests that paroxysmal AF could progress to chronic AF with estimates ranging between 15 and 30% over a period of 1–3 years. We performed an epidemiologic study to elucidate the natural history of paroxysmal AF, this study estimated its incidence in a general practice setting, identified associated factors and analyzed the progression into chronic AF as well as the mortality rate.
Methods
Using the UK General Practice Research Database (GPRD), we identified patients aged 40–89 years with a first-recorded episode of paroxysmal AF during 1996. Risk factors were assessed using 525 incident paroxysmal AF cases confirmed by the general practitioner (GP) and a random sample of controls. We follow-up paroxysmal AF patients and estimated their mortality rate and progression to chronic AF.
Results
The incidence of paroxysmal AF was 1.0 per 1,000 person-years. Major risk factors for paroxysmal AF were age and prior valvular heart disease, ischaemic heart disease, heart failure and hyperthyroidism. During a mean follow-up of 2.7 years, 70 of 418 paroxysmal AF patients with complete information progressed to chronic AF. Risk factors associated with progression were valvular heart disease (OR 2.7, 95% CI 1.2–6.0) and moderate to high alcohol consumption (OR 3.0, 95% CI 1.1–8.0). Paroxysmal AF patients did not carry an increased risk of mortality, compared to an age and sex matched sample of the general population. There was a suggestion of a small increased risk among patients progressing to chronic AF (RR 1.5, 96% CI 0.8–2.9).
Conclusion
Paroxysmal AF is a common arrhythmia in the general practice setting, increasing with age and commonly associated with other heart diseases. It sometimes is the initial presentation and then progress to chronic AF. A history of valvular heart disease and alcohol consumption are associated with this progression.
==== Body
Background
The incidence and natural history of paroxysmal atrial fibrillation (AF) have not been well studied[1]. Paroxysmal AF is usually defined temporally as intermittent periods of AF intertwined with episodes of normal sinus rhythm, normally lasting less than a week [2,3]. It could be a unique self-terminating episode of AF or recurrent ones [4]. When the arrhythmia is sustained in time, AF is designated persistent or chronic AF. However, the differentiation of paroxysmal AF from more chronic forms of AF is often based on the history given by symptomatic patients. This could be misleading, as asymptomatic paroxysmal AF is quite common [5,6]. Paroxysmal AF may be the initial presentation of the arrhythmia or can continue with recurrent episodes that may eventually become chronic as the end result AF[4,7,8]. Some reports have shown that about 25% of paroxysmal AF patients will develop a more persistent form of AF [3]. The rate of transition from paroxysmal to chronic varies considerably with the underlying aetiology, being more frequent among patients with rheumatic heart disease and ischaemic heart disease. Progression from paroxysmal to chronic form does not depend on age or gender but on longer duration of the initial paroxysmal AF episode [8,9].
Most of the studies published on atrial fibrillation patients are based on hospital patients, or cohorts of patients undergoing specific clinical diagnostic tests and procedures. We performed an epidemiologic study in the general practice setting using information recorded by the general practitioners during their daily consultations.
The objectives of this study were to estimate the incidence rate of paroxysmal AF as compared with chronic AF in a general practice setting [10], to describe comorbidity and health care utilization and to identify factors associated with paroxysmal occurrence. We also estimated the rate of progression to chronic AF, as well as the mortality rate among patients with first-detected paroxysmal AF.
Methods
We performed a cohort study using data from the General Practice Research Database (GPRD). The general practitioners (GPs) who anonymously provide the GPRD with their data systematically record information on demographics, medical diagnoses, referrals to consultants and hospitals, and written prescriptions for their patients. The database contains computerized information for more than two million residents in the UK registered with GPs and is administered by the Medicines and Healthcare products Regulatory Agency (MHRA), which organizes this information to be used for research projects. Numerous epidemiological studies support its validity and data completeness, including a recent study on AF [11,12].
Study population
We identified a total of 703,777 patients in the GPRD meeting the following conditions during 1996. Patients were 40–89 years old, enrolled with a GP for more than 2 years and with at least one health contact before January 1996. These eligibility criteria were applied to ensure that all study members had incurrred in some recent contact with the GP prior to entering the study period and were then at risk of being cared "actively" by their GP. Patients with a code for heart rhythm disorders (ICD 8th edition: 4160–4169) or cancer (ICD 8th: 1400–2099) before January 1996 were omitted.
Case definition and validation
We identified 2,098 patients attending to the general practitioner with a first-ever recorded diagnosis of AF/flutter (ICD 8th: 4163). After reviewing the computerized records for all of them, 1,972 patients were considered potential cases of AF. For all of them a questionnaire was sent to the GPs requesting additional information. Additionally, 623 patients with unspecific codes of supra-ventricular and sinus arrhythmias other than AF were identified. Since these codes could mask a diagnosis of AF, we sent a questionnaire to an approximately 10% random sample (n = 68). The GP was asked to confirm the diagnosis of AF, and to classify it as chronic or paroxysmal based on the following criteria: Chronic AF was defined as persistence of the arrhythmia, with the episode of AF not converting to sinus rhythm within 1 week (this definition include persistent and permanent AF cases on terminology proposed by the ACC/AHA/ESC guidelines) [4]. We considered paroxysmal AF when the arrhythmia did revert spontaneously or following treatment to sinus rhythm within a week. We also requested the GPs to confirm whether this diagnosis was the first ever and to provide information on diagnostic tests, procedures and aetiology of the AF. Patient confidentiality was always preserved.
We received 1,888 valid questionnaires (95% response rate). Five hundred twenty-five patients were confirmed as paroxysmal AF cases and 1,109 were patients with chronic AF. The remaining patients were not confirmed as having AF (n = 125) or had an episode of AF before entering the study period (n = 129). The confirmation rate was 98% among patients originally identified with AF codes and 30% among those with unspecific codes of arrhythmia. Patients with chronic forms of AF were analysed separately, and the main results have been published recently [10,12].
Incidence analysis
We estimated the incidence rate of paroxysmal AF stratified by age groups and sex. We used confirmed incident cases of paroxysmal AF as the numerator and the sum of person-years in the study population as the denominator within age and sex strata.
Nested case-control analysis to ascertain risk factor for paroxysmal AF
We used all confirmed cases of paroxysmal AF (as in the cohort analysis), and their index date was the date of the paroxysmal AF diagnosis. We then assigned a random date between 1 January and 31 December 1996 to all members of the study population. If this random date was included in her/his contribution of person-time to the study, that person became an eligible control and we used that date as index date. Finally, 5,000 controls were randomly sampled from the pool of eligible controls. This selection process warrants that the likelihood of being selected as a control is proportional to the person-time at risk.
We used the information recorded by the GP in the computerized database for both cases and controls to ascertain demographic data, as well as the prevalence of medical history before index date for the following conditions: ischaemic heart disease (IHD), valvular disease, heart failure (HF), hypertension, cerebrovascular disease (CVD), diabetes and hyperthyroidism. We also ascertained the role of smoking status, body mass index (BMI) and alcohol consumption. We computed estimates of the odds ratio (OR) and 95% confidence interval (CI) of AF associated with risk factors, using unconditional logistic regression adjusted by age and sex.
Mortality follow-up analysis
Using the study population in which AF patients were identified and applying the same eligibility criteria as for the AF cohort, we randomly sampled a cohort of 5,000 individuals free of AF, matched to the cohort of paroxysmal AF patients by age and sex. The paroxysmal AF cohort was followed up from date of AF diagnosis and the general population cohort from a random date during 1996 until the earlier of either death or end of follow-up (April 2001). Survival probability was computed in both cohorts, and we estimated the relative risk (RR) of mortality associated with AF, using Cox proportional hazard regression. We retrieved death information from computerized files and from the questionnaire filled by the GP, and ascertained the cause of death using the two sources of information.
Follow-up analysis for progression to chronic AF
In order to assess which paroxysmal AF patients progressed to chronic AF, we sent a second questionnaire to the GPs requesting depersonalised copies of medical records for all paroxysmal AF patients. At the time of this second request, we received valid information on 418 cases; the remaining patients (n = 107) were no longer reachable. The GPs confirmed that 70 patients (17%) developed chonic AF before April 2001, 192 continued with recurrent episodes of paroxysmal AF and the remaining (n = 156) did not present any more episodes after the first one.
We used a life table analysis to show the proportion of patients with initial paroxysmal AF, that progressed to chronic AF during the follow-up period.
A nested case-control analysis was performed among the 418 paroxysmal AF patients with valid information to assess risk factors for progression to chronic forms. In this analysis, we used as cases the 70 patients who progressed to chronic AF, and we used their date of diagnosis of chronic AF as index date. All remaining AF patients (n = 348) were used as controls. Estimates of progression risk and 95% CI were computed using unconditional logistic regression. We collected recorded information on the following risk factors: smoking status, BMI and alcohol consumption, as well as prior history of HF, valvular disease, IHD, CVD, hypertension and diabetes. We assessed the association between drug treatment and the development of persistent AF. We also estimated the mortality rate in both groups: those who had paroxysmal AF and those who developed chronic AF.
Results
The incidence rate of paroxysmal AF was 1.0 per 1000 person-years (95% CI 0.9–1.1). In a previous reported study on the same source population we found that the incidence of chronic AFwas higher (1.7, 95% CI 1.6–1.8) [10]. The incidence increased with age in both AF groups, but this was more pronounced among persons aged 70 years and older in the cohort of chronic AF, compared with the paroxysmal AF cohort (Fig. 1). The incidence of paroxysmal AF was similar in males and females.
Figure 1 Incidence of paroxysmal atrial fibrillation in comparison to chronic atrial fibrillation in UK General Practice [10].
Characteristics of the 525 patients with incident paroxysmal AF are shown in Table 1. The mean age at presentation was significantly higher in the female population (73 years; SD = 10) than in the male population (67 years; SD = 11). The most frequent aetiology among initially detected paroxysmal AF as assigned by the GP was IHD in both sexes (43%), while no specific cause was given in 32% of the cases. Less than a third of patients with an initial paroxysmal AF episode had a cardioversion attempt close to the time of diagnosis. Digoxin and beta-blockers were the most frequently prescribed drugs during the 3 months after the paroxysmal AF episode. Close to half of the paroxysmal AF patients (48%) did not receive warfarin or aspirin during the first 3 months after diagnosis.
Table 1 Distribution of age, aetiology, diagnostic tests and pattern of treatment among paroxysmal atrial fibrillation patients by sex
Female Male
n = 268 (%) n = 257 (%)
Age
40–59 30 (11.2) 63 (24.5)
60–69 57 (21.3) 74 (28.8)
70–79 107 (39.9) 81 (31.5)
80+ 74 (27.6) 39 (15.2)
AF aetiology assigned by the GP*
IHD 117 (43.7) 110 (42.8)
Valvular 24 (9.0) 13 (5.1)
Other cardiac diseases 19 (7.1) 20 (7.8)
Non-cardiac diseases 28 (10.4) 16 (10.1)
Unknown 80 (29.9) 88 (34.2)
Tests done to confirm diagnosis
ECG alone 158 (59.0) 170 (66.1)
Other test (with/without ECG) 46 (17.1) 33 (12.8)
Unknown 64 (23.9) 54 (21.0)
Cardioversion attempts
No or unknown 178 (66.4) 170 (66.2)
Pharmacological only 79 (29.5) 69 (26.8)
Electrical only 8 (3.0) 13 (5.1)
Both 3 (1.1) 5 (1.9)
Use of AF treatment drugs †
Amiodarone 27 (10.1) 40 (15.6)
Verapamil 8 (3.0) 7 (2.7)
Diltiazem 13 (4.9) 13 (5.1)
Beta-blockers 82 (30.6) 65 (25.3)
Digoxin 100 (37.3) 80 (31.1)
Use of antithrombotics/anticoagulants†
No use 136 (50.7) 115 (44.7)
Warfarin only 28 (10.4) 36 (14.0)
Aspirin only 92 (34.3) 91 (35.4)
Both 12 (4.5) 15 (5.8)
*17 cases (3.2%) had the episode after a cardiovascular surgery.
†In the first 3 months after initial diagnosis.
Baseline characteristics of paroxysmal AF patients and controls are shown in Table 2. Age was the most important risk factor for paroxysmal AF. Individuals 70 years old or more presented a relative risk greater than eight-fold of developing paroxysmal AF after adjustment for cardiovascular comorbidity and other risk factors. Male patients and those with a moderate to high alcohol consumption carried an increased risk of paroxysmal AF.
Table 2 Risk of paroxysmal atrial fibrillation associated with age, sex, and other factors
Paroxysmal AF cases Controls
n = 525 (%) n = 5000 (%) OR* (95% CI)
Age
40–49 31 (5.9) 1565 (31.3) 1
50–59 62 (11.8) 1343 (26.9) 2.1 (1.3–3.2)
60–69 131 (25.0) 997 (19.9) 5.0 (3.3–7.5)
70–79 188 (35.8) 769 (15.4) 8.3 (5.5–12.5)
80–89 113 (21.5) 326 (6.5) 10.9 (7–17.1)
Sex
Female 268 (51.0) 2647 (52.9) 1
Male 257 (49.0) 2353 (47.1) 1.3 (1.0–1.6)
Smoking†
Non-smoker 313 (59.6) 2736 (54.7) 1
Smoker 94 (17.9) 1226 (24.5) 0.8 (0.6–1.0)
Ex-smoker 42 (8.0) 280 (5.6) 1.0 (0.7–1.5)
Body mass index (BMI) †
<20 25 (4.8) 201 (4.0) 1.4 (0.9–2.4)
20–24 135 (25.7) 1466 (29.3) 1
25–29 162 (30.9) 1420 (28.4) 1.1 (0.8–1.4)
30+ 62 (11.8) 601 (12.0) 1.1 (0.8–1.5)
Alcohol consumption (units per week) †#
None 185 (35.2) 1597 (31.9) 1
1–7 units 114 (21.7) 1112 (22.2) 1.2 (0.9–1.6)
8–21 units 69 (13.1) 718 (14.4) 1.4 (1.0–1.9)
>21 units 31 (5.9) 292 (5.8) 1.7 (1.1–2.6)
Comorbidity
IHD 156 (29.7) 415 (8.3) 2.1 (1.6–2.6)
Valvular disease 30 (5.7) 41 (0.8) 4.2 (2.4–7.3)
Heart failure 81 (15.4) 117 (2.3) 2.5 (1.8–3.5)
Hypertension 197 (37.5) 859 (17.2) 1.4 (1.2–1.8)
Cerebrovascular disease 68 (13.0) 195 (3.9) 1.5 (1.1–2.1)
Diabetes 36 (6.9) 194 (3.9) 0.9 (0.6–1.4)
Hyperthyroidism 19 (3.6) 44 (0.9) 3.6 (2.0–6.5)
*OR: Odds ratio adjusted by age, sex and all the variables in the table, using unconditional logistic regression.
†Missing data in smoking (15.1%), BMI (26.3%) and alcohol use (25.5%).
# 1 unit = 10 mL of pure ethanol.
Close to 70% of paroxysmal AF patients presented some cardiovascular morbidity compared with 29% of controls. Hypertension and IHD were the most prevalent diagnoses in both groups (Table 2). Patients with underlying valvulopathies had a four-fold increased risk of having paroxysmal AF. Other conditions independently associated with the development of paroxysmal AF were heart failure (OR 2.5; 95% CI 1.8–3.5) and hyperthyroidism (OR 3.6; 95% CI 2.0–6.5). We did not find any major association with diabetes.
During a mean follow-up period of 2.7 years, 67 paroxysmal AF patients died. The mortality rate among paroxysmal AF patients was 4.2 per 100 person-years. The relative risk of mortality among paroxysmal AF patients was similar to that in an age and sex matched sample of the general population free of AF after adjustement for other co-comorbidity (table 3). The most frequent cause of death was heart disease including IHD (46.3%), followed by cancer (13.4%). In nine cases, no specific cause of death was assigned.
Table 3 Mortality rate and relative risk of death associated with paroxysmal atrial fibrillation
Age and sex matched Cohort free of AF n = 5000 Paroxysmal AF cohort n = 525
Person-years 14298 1606
Deaths 483 67
Mortality rate/100 person-years (95% CI) 3.38 (3.09–3.69) 4.17 (3.30–5.26)
Relative risk (95% CI) 1 1.2 (1.0–1.6)
Adjusted relative risk*(95% CI) 1 1.0 (0.75–1.3)
*Relative risk estimated by Cox regression model, including age, sex, smoking, heart failure, ischaemic heart disease, hypertension, cerebrovascular disease and diabetes.
Seventy patients with paroxysmal AF eventually progressed into chronic AF during the follow-up period. The progression rate was 6.2 per 100 person-years, with a slightly higher rate among men (6.8 per 100 person-years) than women (5.6 per 100 person-years). Fig. 2 shows the proportion of patients developing chronic AF over time, using life table analysis. It also present the number of patients with paroxysmal AF at risk and the number of patients progressing to chronic AF in each time interval. Patients dying or leaving the practice were censored from follow-up. More than half of the cases of progression (n = 48) occurred during the first year after the initially paroxysmal AF onset, resulting in a progression rate of 13.6 per 100 person-years during the first year of follow-up. Among those not progressing to chronic AF (n = 348), 156 had only the initial single episode recorded and 192 presented recurrent episodes of paroxysmal AF during follow-up.
Figure 2 Proportion of patients with paroxysmal atrial fibrillation progressing to chronic AF. Number of patients at risk and number of new chronic AF cases in each time period, in UK General Practice.
Figure 3 presents the main characteristics studied as potential risk factors for the progression to chronic AF. Age and sex were not associated with progression to persistent AF. Only cardiac morbidity, mainly valvular disease and heart failure were associated with significantly greater risk of developing chronic AF. Moderate to high consumption of alcohol (more than 21 units per week) was associated with a three-fold increased risk of progression (fig. 3). We did not find any association between use of antiarrhytmic drugs after initial diagnosis of paroxysmal AF and progression or not to chronic forms of AF (data not shown), but we observed that patients progressing to persistent AF were more likely to have been treated with warfarin after their initial PAF diagnosis (OR: 2.9; 95%CI: 1.6–1.9) compared to those not progressing.
Figure 3 Risk of progression to chronic atrial fibrillation among paroxysmal atrial fibrillation patients (Odds ratio estimates adjusted by age and sex, using logistic regression).
Eleven patients died among the patients progressing to chronic AF and 39 in the subgroup not progressing. The mortality rate was higher among patients progressing to chronic AF (5.2 per 100 person-years) than in the group who did not (3.6 per 100 person-years). The age- and sex-adjusted relative risk of mortality was 1.5 (95% CI 0.8–2.9) among patients who had progressed to sustained forms of AF, compared with those not progressing.
Discussion
This is the first study, to our knowledge presenting follow-up data in patients initially diagnosed with paroxysmal atrial fibrillation from the perspective of a general practice setting, and we found the annual incidence of paroxysmal AF to be 1.0 per 1000 person-years. A slightly lower incidence rate (0.6 per 1000) has been reported in a study where paroxysmal AF patients attending hospital were identified [13]. The incidence of paroxysmal AF was almost half the one we reported for chronic AF in the same source population [10]. This is well in line with what has been reported previously – that paroxysmal episodes represent between 35% and 66% of all cases of AF, depending on the study population and the definitions used [1,3,14]. It has been reported that the load of patients with atrial fibrillation is likely to increase substantially in the next years, explained only in part by the aging of the population [15]. Our results show, as previously reported, that patients with paroxysmal AF are younger and have less comorbidity than patients with chronic AF [10,13,16]. This age difference could reflect the progressive nature of AF. The differences in morbidity and mortality among the two forms of AF suggest the possibility that paroxysmal and persistent AF may be different diseases with different risk factors and different pathogenic substrates, although clearly overlapping in part [17-19], paroxysmal being a more benign disease than chronic.
It is difficult to assess an underlying cause in all AF patients [3]. In our study, the specific cause was not reported in 32% of paroxysmal AF patients. We observed that the most frequent underlying associated conditions in paroxysmal AF were coronary heart disease, rheumatic heart disease and hypertension. These causes have also been found for chronic AF [12], although there appears to be a great variability between studies.
It has been recommended that pharmacological management of patients with newly discovered AF requires knowledge of its pattern of presentation (paroxysmal, persistent, or permanent) [4]. Anticoagulation with warfarin has been proposed for patients with paroxysmal forms when there is underlying heart disease [9,20-22], as no major differences in risk of stroke between chronic and paroxysmal AF have been reported. However it is not clear whether patients with limited episodes of paroxysmal AF require anticoagulation, and the decision must be individualized for each patient [4]. We found that close to half of the patients attending general practice with initial paroxysmal AF were not given warfarin or aspirin in the three months after the first-detected episode – a higher proportion than the one we observed among patients with chronic forms of AF [10]. This could be due in part to the fact that 37% of patients with paroxysmal AF presented a single episode of paroxysmal AF without any recurrence during the follow-up, and consequently anticoagulant or antiplatelet therapies may not be recommended in this subgroup of patients [19].
Clinical experience suggests that paroxysmal AF is often perpetuated and frequently it progresses to chronic AF, with estimates ranging between 20% and 30% within a period of 1–3 years [1,23,24]. A ten year follow-up study of paroxysmal patients in Trieste reported that 34% developed chronic AF [18]. The rate of progression varied according to aetiology, with patients with rheumatic valve disease carrying the highest rate of progression (66%) [4]. A recent 14 years follow up study on Japanese patients with initial paroxysmal AF, reported that 77% of them developed into its chronic form (5,5% of patients per year) [25]. In our study, we found that 17% of paroxysmal AF patients progressed to chronic AF during an average follow-up of close to 3 years, and history of valvular heart disease and moderate to high alcohol consumption were identified as the major independent predictors of progression to chronic AF.
Only a few small studies have examined factors predicting progression from paroxysmal to persistent AF [23,24]. The study by Abe et al. of 122 consecutive patients with paroxysmal AF reported a progression of 11% to chronic AF during a period of 2 years. They did not find any significant differences in age, sex or presence of organic heart diseases in the patients who developed chronic AF, compared with paroxysmal AF patients who did not progress[23]. Another study reported an annual progression rate of 22% [24]. Chronic and persistent forms of AF have been reported to carry a greater mortality than paroxysmal AF [17,23,24]. Our study suggests that patients progressing to chronic AF had a slightly higher increased risk of mortality than those not progressing.
Some limitations need to be taken into account when interpreting our results. Our findings are the reflection of detection and management of patients with newly diagnosed AF in general practice. Furthermore, our data are based on physicians assessments of patients symptoms and when available results from diagnostic tests to determine if the atrial fibrillation was paroxysmal or chronic (we could not specifically distinguish between permanent or persistent), and consequently will not be as accurate as prospective studies based on events detected after Holter monitoring, pacemaker insertion or ablation surgery. Our study was observational based on the information recorded by general practitioners during their daily practice. This shares the limitations of insufficient information at times but has the advantage to study the occurrence and determinants of AF patients in the real world of general practice. Therefore, it is likely that we may not have ascertained all cases of paroxysmal atrial fibrillation occurring in the study population, resulting in some underestimation of paroxymal AF. We found that a large proportion of patients had a unique episode of PAF, but we could not verify whether this was due to effective treatment with antiarrhythmic drugs or was incomplete reporting of subsequent paroxysmal episodes.
As we rely on the judgment of the GPs as well as their records and tests, some information was missing, for example GP could not assign a specific aetiology in 32% of paroxysmal AF patients, and we could not distinguish between those who would truly be "lone AF" and those with an underlying etiology not recorded by the GP. Also the rate of progression from paroxysmal to chronic AF could have been underestimated as we did not obtain complete response for all followed-up patients.
Conclusion
In summary, we observed that paroxysmal AF is a common arrhythmia in a general practice setting, increases with age, and is commonly associated with other cardiac morbidity. Paroxysmal AF is sometimes the initial presentation of an arrhythmia progressing to chronic AF. Valvular heart disease and alcohol consumption are associated with this progression. Patients with paroxysmal AF do not present an increased risk of mortality compared to the general population after adjustment for other comorbidity, except among the subgroup progressing to chronic AF where a small mortality excess risk was observed.
Competing interests
AR and LAGR work at CEIFE, which received a research grant from AstraZeneca for this study. MAW and SJ are employees and shareholders of AstraZeneca. The authors do not expect any companies to which they have provided services to gain or lose financially from the materials in this study.
Authors' contributions
AR and LAGR conceived the study, participated in its design, performed the analysis and wrote the manuscript. MAW and SJ contributed to the review of patients profiles, data analysis and participated in the writing of the mansucript. 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 thank the participating GPs for their collaboration. This study was supported by a research grant from AstraZeneca R&D.
==== Refs
Aboaf AP Wolf PS Paroxysmal atrial fibrillation. A common but neglected entity Arch Intern Med 1996 156 362 7 8607721 10.1001/archinte.156.4.362
Fuster V Rydén LE Asinger RW Cannom DS Crijns HJ Frye RL Halperin JL Kay GN Klein WW Levy S McNamara RL Prystowsky EN Wann LS Wyse DG Gibbons RJ Antman EM Alpert JS Faxon DP Fuster V Gregoratos G Hiratzka LF Jacobs AK Russell RO Smith SC Klein WW Alonso-Garcia A Blomstrom-Lundqvist C De Backer G Flather M Hradec J Oto A Parkhomenko A Silber S Torbicki A American College of Cardiology/American Heart Association/European Society of Cardiology Board. ACC/AHA/ESC guidelines for the management of patients with atrial fibrillation: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines and the European Society of Cardiology Committee for Practice Guidelines and Policy Conferences (Committee to Develop Guidelines for the Management of Patients With Atrial Fibrillation) J Am Coll Cardiol 2001 38 1231 66 11583910 10.1016/S0735-1097(01)01587-X
Lévy S Classification system of atrial fibrillation Curr Opin Cardiol 2000 15 54 7 10666661 10.1097/00001573-200001000-00007
Lip GYH Hee FL Paroxysmal atrial fibrillation QJM 2001 94 665 78 11744787 10.1093/qjmed/94.12.665
Page RL Wilkinson WE Clair WK McCarthy EA Pritchett EL Asymptomatic arrhythmias in patients with symptomatic paroxysmal atrial fibrillation and paroxysmal supraventricular tachycardia Circulation 1994 89 224 7 8281651
Kaufman ES Waldo AL The impact of asymptomatic atrial fibrillation J Am Coll Cardiol 2004 43 47 52 14715182 10.1016/j.jacc.2003.10.013
Lévy S Breithardt G Campbell RWF Camm AJ Daubert JC Allessie M Aliot E Capucci A Cosio F Crijns H Jordaens L Hauer RN Lombardi F Luderitz B Atrial fibrillation: current knowledge and recommendations for management Eur Heart J 1998 19 1294 320 9792255 10.1053/euhj.1998.1050
Lévy S Epidemiology and classification of atrial fibrillation J Cardiovasc Electrophysiol 1998 9 S78 82 9727680
Petersen P Thromboembolic complications of atrial fibrillation and their prevention: a review Am J Cardiol 1990 65 24C 8C 2137283 10.1016/0002-9149(90)90111-D
Ruigómez A Johansson S Wallander MA García Rodríguez LA The incidence of chronic atrial fibrillation in general practice and its treatment pattern J Clin Epidemiol 2002 55 358 63 11927203 10.1016/S0895-4356(01)00478-4
García Rodríguez LA Pérez Gutthann S Use of the UK General Practice Research Database for pharmacoepidemiology Br J Clin Pharmacol 1998 45 419 26 9643612 10.1046/j.1365-2125.1998.00701.x
Ruigómez A Johansson S Wallander MA García Rodríguez LA Risk of mortality in a cohort of patients newly diagnosed with chronic atrial fibrillation BMC Cardiovasc Disord 2002 2 5 11897013 10.1186/1471-2261-2-5
Goudevenos JA Vakalis JN Giogiakas V Lathridou P Katsouras C Michalis LK Sideris DA An epidemiological study of symptomatic paroxysmal atrial fibrillation in northwest Greece Europace 1999 1 226 33 11220559 10.1053/eupc.1999.0059
Sudlow M Rodgers H Kenny RA Thomson R Population based study of use of anticoagulants among patients with atrial fibrillation in the community BMJ 1997 314 1529 30 9183202
Go AS Hylek EM Phillips KA Chang Y Henault LE Selby JV Singer DE Prevalence of diagnosed atrial fibrillation in adults: national implications for rhythm management and stroke prevention: the AnTicoagulation and Risk Factors in Atrial Fibrillation (ATRIA) Study JAMA 2001 285 2370 5 11343485 10.1001/jama.285.18.2370
Lévy S Maarek M Coumel P Guize L Lekieffre J Medvedowsky JL Sebaoun A for the College of French Cardiologists Characterization of different subsets of atrial fibrillation in general practice in France: the ALFA study Circulation 1999 99 3028 35 10368121
Johansson S Wallander MA García Rodríguez LA Ruigómez A Permanent and paroxysmal atrial fibrillation in general practice Pharmacoepidemiol Drug Saf 2003 12 S 109
Scardi S Mazzone C Pandullo C Goldstein D Poletti A Humar F Lone atrial fibrillation: prognostic differences between paroxysmal and chronic forms after 10 years of follow-up Am Heart J 1999 137 686 91 10097231
Gersh BJ Solomon A Lone atrial fibrillation: Epidemiology and natural history Am Heart J 1999 137 592 5 10097218
Domanski MJ The epidemiology of atrial fibrillation Coron Artery Dis 1995 6 95 100 7780625
Lip GYH Hart RG Conway DSG Antithrombotic therapy for atrial fibrillation BMJ 2002 325 1022 5 12411366 10.1136/bmj.325.7371.1022
Inoue H Atarashi H for the Research Group for Antiarrhythmic Drug Therapy Risk factors for thromboembolism in patients with paroxysmal atrial fibrillation Am J Cardiol 2000 86 852 5 11024400 10.1016/S0002-9149(00)01105-X
Abe Y Fukunami M Yamada T Ohmori M Shimonagata T Kumagai K Kim J Sanada S Hori M Hoki N Prediction of transition to chronic atrial fibrillation in patients with paroxysmal atrial fibrillation by signal-averaged electrocardiography: a prospective study Circulation 1997 96 2612 6 9355901
Sakamoto H Kurabayashi M Nagai R Fujii J Prediction of transition to chronic atrial fibrillation in patients with paroxysmal atrial fibrillation Circulation 1998 98 1045 6 9737530
Kato T Yamashita T Sagara K Iinuma H Fu LT Progressive nature of paroxysmal atrial fibrilaltion-Observations from a 14-year follow-up study Circ J 2004 68 568 572 15170094 10.1253/circj.68.568
|
16008832
|
PMC1185525
|
CC BY
|
2021-01-04 16:30:06
|
no
|
BMC Cardiovasc Disord. 2005 Jul 11; 5:20
|
utf-8
|
BMC Cardiovasc Disord
| 2,005 |
10.1186/1471-2261-5-20
|
oa_comm
|
==== Front
BMC Chem BiolBMC Chemical Biology1472-6769BioMed Central London 1472-6769-5-21607899510.1186/1472-6769-5-2Research ArticleElectronic properties of amino acid side chains: quantum mechanics calculation of substituent effects Dwyer Donard S [email protected] LSU Health Sciences Center, School of Medicine, Shreveport, LA 71130 USA2005 3 8 2005 5 2 2 23 3 2005 3 8 2005 Copyright © 2005 Dwyer; licensee BioMed Central Ltd.2005Dwyer; 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
Electronic properties of amino acid side chains such as inductive and field effects have not been characterized in any detail. Quantum mechanics (QM) calculations and fundamental equations that account for substituent effects may provide insight into these important properties. PM3 analysis of electron distribution and polarizability was used to derive quantitative scales that describe steric factors, inductive effects, resonance effects, and field effects of amino acid side chains.
Results
These studies revealed that: (1) different semiempirical QM methods yield similar results for the electronic effects of side chain groups, (2) polarizability, which reflects molecular deformability, represents steric factors in electronic terms, and (3) inductive effects contribute to the propensity of an amino acid for α-helices.
Conclusion
The data provide initial characterization of the substituent effects of amino acid side chains and suggest that these properties affect electron density along the peptide backbone.
==== Body
Background
How the amino acid sequence of a protein determines its native tertiary structure is one of the most perplexing questions in biology. The formation of secondary structure (α-helices, β-strands and coils/turns) is an intermediate step in this process, although, in some cases, this may occur very late in folding just prior to consolidation of the final 3-D structure [1-4]. Hydrophobicity and steric effects are two major factors that govern protein folding [5-7]. In addition, I [8] have recently suggested that electronic properties of amino acids, including inductive effects, may contribute to the propensity for secondary structure. This possibility merits further investigation especially in view of several recent findings. First, although the hydrophobicity of an amino acid correlates with preference for β-strand and coil conformations, it does not predict tendency to form α-helices [8]. This suggests that adoption of β-strands vs. α-helices may be driven by different molecular forces. Second, electronic effects have provided important insights into structural preferences and have dramatically revised our thinking about the factors that impact rotation about a single bond. For example, the fact that ethane prefers the staggered conformation over the eclipsed conformation has long been ascribed to steric factors [9]. However, Pophristic and Goodman [10] demonstrated that hyperconjugative effects (electron delocalization into antibonding orbitals) rather than steric effects explain the conformational preference of ethane in support of earlier suggestions [11,12]. Finally, recent studies suggest that inductive effects are involved in helix formation/stabilization. Thus, inductive effects have been invoked to explain the enhanced stability of helical structures in collagen that contain fluoroproline substitutions [13,14] and to account for the preference of amino acids for α-helical structures [8]. Despite the emerging significance of electronic effects for conformational preference, little is known about the electronic properties of amino acid side chains. In order to address this shortcoming, I have applied computational chemistry, i.e., quantum mechanics (QM) calculations, to the characterization of the electronic effects of amino acids.
Electronic (substituent) effects of various chemical groups have been characterized in some detail and related to basic chemical properties including rotational flexibility and pKa [15-19]. Previously, I [8] presented theoretical arguments for considering amino acid side chains as substituents of the peptide backbone that affect electron densities and bond angles as a function of their electronic properties. Electronic effects were initially quantified in terms of the pKa at the amino group and localized electronic effects (eσ) estimated from the work of Charton [15]. However, as shown by, Taft [19,20], Chalvet et al. [16], Charton [15], and Topsom [18], the substituent effects that determine the pKa of a chemical group can be partitioned into more fundamental factors, which include inductive (through-bond) and field (through-space) effects, polarizability, and resonance effects. QM methods have been successfully applied to the derivation of substituent effects of certain chemical groups in substituted phenols [21,22], bicyclooctane carboxylic acids [15,18], and other substrates [23,24]. Until now, there has not been a detailed characterization of substituent constants for amino acid side chains.
Theoretical considerations
The structure and properties of a molecule are determined by its electronic configuration or charge distribution [25,26]. Moreover, the electronic properties of substituent groups affect the structure, reactivity, and rotational flexibility of the substituted host molecule. Electron delocalization, including hyperconjugation in saturated molecules such as ethane, contributes to rotational freedom in molecules [10]. Rotation about the main chain bonds of proteins ultimately determines the secondary and tertiary structure of a protein, as observed by Ramachandran and colleagues [27]. It is worth noting that there is electron delocalization along the main chain, which modulates the chemical properties of proteins [28-30]. Elsewhere, I [8] have suggested that amino acid side chains can be considered substituent groups along the peptide backbone that affect the local electron distribution and rotational flexibility. However, the substituent effects of amino acids have not been systematically characterized. Therefore, a major goal of this work is to provide an initial characterization of the substituent effects of amino acids determined from QM calculations and equations that describe proton dissociation. Hammett [31] sought to account for substituent effects of chemical groups with two terms: (1) the substituent constant, σ, (reflecting intrinsic physicochemical properties of a group), and (2) a reaction constant, ρ, which specifies the nature of the reaction, the medium and temperature. Considerable evidence supports the notion that substituent effects reflect the intrinsic electronic properties of a chemical group and its environment, including temperature and solvent [15-19]. A similar concept might be applied to protein folding, i.e., the native structure is determined by inherent physicochemical properties of amino acids in concert with temperature and solvent effects. This paper lays out a general strategy for determining the inherent electronic properties of amino acid side chains, and presents an initial quantitative analysis of substituent effects that include inductive, resonance, field, and steric effects. The possible relationship of these properties to secondary structural preferences has also been explored.
Results and discussion
Calculation of electronic effects
Previous analysis of Hammett constants has revealed that substituent effects represent an amalgam of electronic effects. The work of Taft [20], Charton [15], Chalvet et al. [16], and Topsom [18] provided the theoretical framework for partitioning the substituent effects of amino acid side chains into fundamental electronic properties. The collective contribution of various factors to the electronic properties of molecules, including proton dissociation at the amino group of amino acids, can be written as:
pKa (amino group) = σF + [σI + σR] + σα (1)
where field effects (σF), together with inductive (σI) and resonance effects (σR), constitute the localized electronic effect of Charton (σ*), and σα represents polarizability (steric effects). The general inductive term, [σI+R], consists of both inductive (σ) effects and resonance (π) effects. Inductive versus resonance effects can be distinguished by examination of substituent effects in saturated versus non-saturated ring systems such as substituted bicyclooctane carboxylic acids [15,18]. This general strategy was applied here to the characterization of a series of cyclohexanols and phenols with amino acid side chains substituted in the 4-position. The Mulliken population from QM calculations for the hydroxyl group was used as a potential indicator or reporter of electronic effects of the attached side chain groups. Changes in the electron distribution at the hydroxyl moiety mainly reflect inductive effects (sometimes equated with the electronegativity of a substituent) of the side chains in cyclohexanol and inductive plus resonance effects in phenol. Detailed derivation of inductive, resonance, and polarizability (steric) effects is described below. With this information and knowledge of amino acid pKa's, it was possible to calculate field effects with equation (1). The pKa at the amino group was used in this analysis because previous work showed a close association between electron density at this group (as measured in NMR studies) and secondary structure [32,33].
Several predictions follow from this theoretical background. First, electronic features derived from the QM calculations should be consistent across methodologies, at least for different semiempirical QM methods. Second, the electronic properties obtained from QM calculations should correlate with empirically derived substituent constants (e.g., from Charton's work). Finally, if a particular electronic effect contributes to protein folding, there should be an association between that effect and folding preference.
Evaluation of QM methods
For the QM calculations, semiempirical methods, PM3, AM1, and MNDO, were used to characterize amino acid side chains. Ab initio methods (both Hartree-Fock and DFT) could theoretically be employed for this analysis and may ultimately offer a more accurate picture of electronic properties of amino acids. Nevertheless, semiempirical methods are still commonly used and perform comparably to ab initio methods in many cases [34-37]. Thus, PM3 and MNDO QM methods were evaluated for their ability to accurately represent the electronic properties of a series of substituted phenol molecules. There was generally a good correspondence between these two semiempirical methods in the Mulliken population data calculated for the hydroxyl atoms with r values from linear regression analysis > 0.9 (Pearson coefficient). The PM3 method was somewhat superior overall and was chosen as the main approach for this analysis. The reliability of the PM3 calculations was established by assessing the ability of this method to predict the pKa's of a series of substituted phenols in relation to experimental data. Electron populations and bond lengths were computed for a series of phenols with substitutions (mainly at the 4-position) that included chlorine atoms, and nitro, amine and ethyl groups. Linear regression analysis revealed that there was a highly significant correlation (correlation coefficient, r = -0.9) between the O-H bond lengths of the substituted phenols and their experimentally determined pKa (Figure 1A). The Mulliken population data at the hydrogen atom also showed a similar high degree of correlation (r = 0.9) with the pKa (Figure 1B). Thus, PM3 QM values faithfully predict the dissociation behavior of the alcohol moiety in this model system.
Figure 1 Linear regression analysis was performed to determine the correlation between experimentally-derived pKa's for a series of substituted phenols and (A) O-H bond lengths, and (B) Mulliken populations derived from PM3 calculations. The pKa values were determined by Hanai et al. [57].
Quantification of the electronic properties of amino acids
PM3 calculations were then performed on each of the 20 amino acids for an initial characterization of their electronic properties. Mulliken population analysis of the heavy chain atoms and the polarizability of each residue are summarized in Table 1. There were sizeable differences among the amino acids in the Mulliken population data especially at the nitrogen and Cα atoms. In addition, there was a significant correlation between the Mulliken population data at the nitrogen atom and the pKa at the amino group (r = 0.6, p < 0.01). The correlation was quite striking when cysteine was omitted from the analysis due to the anomalous pKa for its amino group. In this case, the correlation coefficient between the pKa at the amino group and the Mulliken population at the nitrogen atom was 0.8 (p < 0.005). These observations were consistent with the success of PM3 in predicting the pKa of substituted phenols on the basis of the Mulliken population data at the hydroxyl group. Mulliken values at the nitrogen atom also showed a highly significant correlation with the localized electronic effect scale of Charton (eσ) as compiled previously [8] (r = -0.7, p < 0.002). The localized electronic effect includes field, inductive, and resonance effects [8]. Therefore, the QM values calculated for the amino acids reflect complex electronic factors (pKa and eσ) that may be further partitioned into more fundamental components.
Table 1 PM3 data for the 20 amino acids.
Mulliken populationa Polarizability
Amino Acid Cyclohexanol Phenol (A3)
N Cα C O O H O H
P 5.1340 4.2388 3.5314 6.6336 6.3268 0.7883 6.2674 0.7784 4.3
C 4.1918 4.6375 3.4483 6.5367 6.3264 0.7882 6.2595 0.7773 2.7
A 4.2048 4.5978 3.4552 6.5453 6.3277 0.7888 6.2615 0.7781 1.1
I 4.4251 4.4995 3.5018 6.5272 6.3281 0.7889 6.2614 0.7782 4.3
E 4.4723 4.4447 3.5051 6.6012 6.3454 0.7951 6.2811 0.7899 4.1
V 4.2068 4.6039 3.4648 6.5416 6.3267 0.7884 6.2673 0.7784 3.2
L 4.2043 4.5929 3.4581 6.5472 6.3269 0.7885 6.2609 0.7784 4.2
D 4.5761 4.3934 3.5257 6.5880 6.3482 0.7934 6.2775 0.7954 3.0
G 4.1766 4.7053 3.4629 6.5511 6.3271 0.7883 6.2614 0.7774 0.03
W 4.2111 4.5755 3.4645 6.5537 6.3272 0.7889 6.2685 0.7789 12.1
M 4.1951 4.6201 3.4468 6.5369 6.3271 0.7891 6.2620 0.7770 5.1
H 4.2906 4.5323 3.4741 6.5614 6.3285 0.7882 6.2617 0.7795 6.3
S 4.1828 4.6620 3.4635 6.5441 6.3271 0.7880 6.2617 0.7776 1.6
F 4.2128 4.5783 3.4638 6.5490 6.3272 0.7887 6.2618 0.7780 8.0
Q 4.2029 4.6050 3.4588 6.5447 6.3263 0.7873 6.2596 0.7767 4.8
Y 4.2091 4.5836 3.4611 6.5514 6.3268 0.7888 6.2620 0.7776 8.8
T 4.1873 4.6438 3.4584 6.5433 6.3265 0.7878 6.2660 0.7771 2.7
R 4.2113 4.5381 3.4861 6.5801 6.3127 0.7857 6.2469 0.7699 8.5
K 4.2248 4.5119 3.5041 6.5645 6.3137 0.7867 6.2541 0.7653 5.2
N 4.3075 4.5431 3.4674 6.5528 6.3267 0.7869 6.2608 0.7754 3.7
aThese data refer to results of PM3 calculations for the main chain atoms of the amino acids and the hydroxyl group of cyclohexanol or phenol.
Inductive and resonance effects of side chains
QM calculations were performed on the substituted cyclohexanol and phenol reporter molecules. The H atom (side chain) of glycine represented the zero point for the derivation of inductive and resonance effects. Values below that of glycine were assigned negative numbers to reflect the fact that decreased electron density at these atoms would encourage proton dissociation, thus tending to lower the pKa of the hydroxyl group. Specifically, the Mulliken population data for the H atom of the hydroxyl moiety were used to determine inductive effects because this value showed a high degree of correlation with the pKa of substituted phenols in the test panel. Thus, inductive effects (σI) were calculated with equation (2) (see the Methods section) and reflected the difference from the glycine (cyclohexanol) reference data (Table 2). The general trends seemed reasonable because the acidic side chains of aspartic and glutamic acid produced opposite effects from the positive side chains of arginine and lysine, and alkyl groups were weak electron donors in this system as expected. There was an excellent correspondence between the σI scale derived from QM calculations and the localized electronic effects of Charton (eσ) determined from experimental data (r = -0.9) (Table 3). As expected, the Mulliken population data were highly correlated (r = 0.99) with the electron densities calculated with PM3. Moreover, the values derived from PM3 calculations showed excellent correlation with those obtained with other semiempirical methods including MNDO (r = 0.98; Figure 2A) and AM1 (r = 0.99; Figure 2B). The fact that three separate QM methods yielded similar overall results lends support for the trends reported here even if the calculated values include a measure of uncertainty. These observations suggest that Mulliken population data can reveal fundamental behavior of a molecule in terms of electronic effects, despite potential limitations of this measure.
Table 2 Electronic properties of amino acid side chains.
pKaa σI HMΔPH σR σα σF AI HNNMRb
P 10.60 0 0.10 0.10 -0.04 0.02c 0 -
C 10.28 -0.01 -0.01 0.01 -0.03 0.06c 0.01 8.18
A 9.69 0.05 0.05 0 -0.01 0.05 0.05 8.12
I 9.68 0.06 0.08 0.02 -0.04 0.04 0.06 7.99
E 9.67 0.68 1.25 0.57 -0.04 -1.14 0.68 8.40
V 9.62 0.01 0.09 0.08 -0.03 -0.04 0.01 8.08
L 9.60 0.02 0.07 0.05 -0.04 -0.03 0.02 7.99
D 9.60 0.51 1.80 1.29 -0.03 -1.77 0.51 8.38
G 9.60 0 0 0 0 0 0 8.36
W 9.39 0.06 0.15 0.09 -0.12 -0.24 0.06 8.03
M 9.21 0.08 -0.04 -0.12 -0.05 -0.30 0.08 8.12
H 9.17 -0.01 0.21 0.22 -0.06 -0.58 0.01 8.36
S 9.15 -0.03 -0.05 -0.02 -0.02 -0.38 0.03 8.30
F 9.13 0.04 0.06 0.02 -0.08 -0.45 0.04 7.93
Q 9.13 -0.10 -0.07 0.03 -0.05 -0.35 0.10 8.19
Y 9.11 0.05 0.02 -0.03 -0.09 -0.42 0.05 8.10
T 9.10 -0.05 -0.03 0.02 -0.03 -0.44 0.05 8.17
R 9.04 -0.26 -0.75 -0.49 -0.08 0.27 0.26 8.23
K 8.95 -0.16 -1.11 -0.95 -0.05 0.51 0.16 8.29
N 8.80 -0.14 -0.20 -0.06 -0.04 -0.56 0.24 8.33
aThe pKa data were taken from Edsall [58]. bThe HNNMR data were derived from the NMR work of Wishart et al. [32] and represent chemical shift values in p.p.m. for the amide proton of residues in the coil conformation. The other terms have been defined in the text. cEstimated on the basis of chemically similar groups due to anomalous pKa.
Table 3 Correlations between electronic properties and folding preferencesa.
Index HNNMR σI σR σα σF AI CαMULL
Kyte-Doolittle -0.8** 0 0 -0.2 0.3 -0.5* 0.4
Water vapor 0.8** 0.2 0.2 0.1 -0.6* 0.5* -0.4
Bulk -0.3 -0.1 -0.2 0.9** 0 0.2 -0.4
Gyration -0.2 0 -0.2 0.9** 0 0.3 -0.5*
α-helix -0.1 0.3 0.1 0.2 -0.2 0.6** -0.7**
β-strand -0.8** -0.2 -0.1 0.3 0.3 -0.4 0.2
coil 0.7** 0.1 0.4 -0.3 -0.4 0.3 -0.3
eσ -0.1 -0.9** -0.9** 0.3 0.7** -0.6** 0.4
pKa -0.1 0.3 0.3 -0.3 0.1 0 0
σI 0.2
σR 0.2 0.8**
σα -0.4 -0.1 -0.2
σF -0.4 -0.8** -0.9** 0
AI 0.5* 0.7** 0.5* 0 -0.6**
CαMULL -0.2 -0.6* -0.4 -0.2 0.5* -0.8**
aLinear regression analysis was performed to determine possible correlations between the various electronic scales presented here. The first 2 indices of hydrophobicity were taken from Kyte and Doolittle [46], the bulk scale was from Kidera et al. [38], and the gyration scale is that of Levitt [39]. The secondary structure preferences and eσ scales were described previously [8]. The r values from the analysis are presented here and directions of the slope are indicated by the signs. Statistical analysis of the data revealed significant correlations: *p < 0.05, **p < 0.01.
Figure 2 Results of linear regression analysis comparing Mulliken population data for the hydroxyl hydrogen atom in substituted cyclohexanol calculated with PM3 vs. (A) MNDO and (B) AM1 methods.
An additional scale (AI) is presented in Table 2 derived from the absolute values of the σI index. This scale is presented to emphasize the fact that, from the perspective of the main chain atoms, there may be little difference between strong electron donation by a side chain group (e.g. acidic moieties of aspartic acid and glutamic acid) to the amino group and strong electron withdrawal from the carboxyl group by an electron acceptor (e.g. charged side chains of lysine and arginine). This suggestion is supported by the high degree of correlation (r = -0.8) between the AI scale and the Mulliken population data at the Cα carbon (CαMULL) (see Table 3).
The resonance effects scale (σR) was derived according to equation (3) (see the Methods section). These values showed a high degree of correlation (r = 0.9) with the independently-derived resonance scale of Hansch et al. [19], which included 7 chemical groups that correspond to amino acid side chains. Furthermore, the σR scale correlated with the eσ constants of Charton (r = -0.9), which reflect a combination of inductive, field, and resonance effects.
Polarizability index
PM3 calculations of polarizability were obtained for each of the amino acid side chains (Table 1) and a normalized polarizability index (σα) was derived (Table 2). This measure reflects both the deformability and size of a substituent. Linear regression analysis revealed that the polarizability index was very similar to scales that represent steric or bulk factors of amino acids. Thus, there was a highly significant correlation with both the composite bulk scale of Kidera et al. [38] (r = 0.9) and interestingly, the side chain gyration scale of Levitt [39] (r = 0.9) (Table 3), which includes implicit vibrational contributions. It is known that vibrational (Raman) spectra of peptides display consistent shifts in relation to the size of the amino acid side chain [40]. Information about the correct sign to apply to the σα scale in equation (1) derives from two observations. First, others have assigned a negative value to polarizability effects on protonation [23]. Second, steric effects are known to encourage proton dissociation and lower the pKa [25].
Field effects of amino acids
The next step was to calculate the field effects of the amino acid side chains by substituting into equation (1). Field effects include electrostatic interactions between charged side chains and main chain groups and polarization effects from H-bonding between OH and NH groups of the side chains and the peptide backbone. To solve equation (1), the various indices (σF, σI, etc.) were weighted equally, which represented a first approximation of the relative contributions to the pKa. Charton and others [15,16] employed weighting factors in the range of 0.5–2 for similar analyses of substituent constants, so the basic assumption in the present work was consistent with these values. Normalized indices were established for polarizability and inductive/resonance effects by multiplying raw calculations by 0.01 or 100 so that the individual components of equation (1) were on the same scale (equal weighting). The work described here was focused on the relative electronic properties of amino acid side chains and not the absolute value for field effects, inductive effects, etc. In order to simplify the calculations for this analysis, the pKa values at the amino group were referenced to glycine (0), e.g., asparagine was -0.80 (rather than 8.80) and proline was 1.0 (rather than 10.60), and an arithmetic scale was used. The field effects index (σF) derived from these calculations is summarized in Table 2. The high correlation between σF and the independently-derived localized electronic effect scale of Charton (eσ) supported the overall validity of this measure of field effects (r = 0.7; Table 3).
To summarize the findings thus far, it was shown that the Mulliken population data for amino acid side chains revealed similar trends when several different semiempirical methods were used for the QM calculations. Second, the Mulliken population at the nitrogen atom and the polarizability scale were highly correlated with empirical data concerning the pKa and steric effects, respectively, of amino acids. Finally, the scales for inductive, resonance, and field effects showed strong correlation with the localized electronic effect scale of Charton (eσ), which was derived from experimental observations.
Relationship to secondary structure
The next objective was to determine whether any of the electronic scales correlated with the folding preferences of the amino acids. A clear relationship between a particular electronic property and secondary structure preference might provide fundamental insights into the forces that drive protein folding. Moreover, although the hydrophobicity of an amino acid is a good predictor of its preference for β-strand and coil conformations, this measure is a poor predictor of helix propensity [8]. The secondary structural preferences used for this analysis were derived previously [8] from an analysis of over 24,000 residues. Our scales show good (0.72–0.8) [41-43] to excellent (0.83–0.93) correlation [44,45] with structural preferences reported by other groups. Of the indices presented here, the empirical HNNMR index is the best predictor of secondary structure at least for β-strand and coil conformations (Table 3). Given the close correlation between the HNNMR scale and various hydrophobicity scales, this relationship is not surprising. Furthermore, the correlation between hydrophobicity and β-strand and coil preference is confirmed here for both the Kyte-Doolittle scale [46] (r values: coil, -0.6; β, 0.7; α, -0.1), and the partition coefficient in water vapor (coil, -0.7; β, -0.7; α, 0.06) (Table 3). However, these scales are completely inadequate for predicting the propensity of amino acids for α-helical conformations. The simple electronic property that best predicts preference for α-helices is the Mulliken population at the Cα atom (CαMULL) derived from the PM3 calculations (r = -0.7). Previous work from this laboratory suggested that electronic effects along the peptide backbone contribute to α-helical preference [8]. Although the inductive scale (σI) in Table 2 does not predict the propensity of amino acids for α-helices, the absolute value of this index (AI) shows a significant correlation with helix preference (r = 0.6; Table 3). The AI scale is highly correlated with the CαMULL values (r = -0.8).
One possible interpretation of these findings would be that opposite processes related to electron delocalization along the peptide backbone produce similar conditions that favor formation of α-helices. More specifically, electron donation by a side chain (e.g., glutamic acid) to the amino group and electron withdrawal by a side chain (e.g., lysine) from the carboxyl group may exert similar overall effects on the electron distribution along the main chain. In both cases, the inductive effects of the side chains disrupt the normal electron flow from the carboxyl to the amide group. The net result would be a decrease in π-character along the backbone (an increase in electron density), an increase in bond length, and enhanced rotational flexibility. This flexibility may be required for adoption of α-helices.
The hydrophobicity of amino acids reasonably predicts strand and coil conformations, but is a poor predictor of α-helices. Nevertheless, solvent effects clearly help to drive protein folding. In contrast to hydrophobicity, electronic scales that predict α-helices (CαMULL and AI) tend to fare poorly in the prediction of other secondary structures. These observations suggest that folding into α-helices versus coils and β-strands may be driven by different forces. Inductive effects appear to play a significant role in helix formation, whereas polarity and solvent effects are the major determinants of other secondary structures. Thus, helix formation is opposed by high polarity near the main chain and by disruption of inductive effects. Amino acids that prefer α-helices have a higher average electron density at the main chain atoms (from PM3 calculations), which would mean longer bond lengths and greater rotational freedom. By contrast, amino acids with a propensity for β-strands tend to have a lower electron density at the main chain atoms, which would produce the opposite effects. These predictions received initial support from an analysis of bond lengths in β-strands vs. α-helices in a panel of 7 proteins with high resolution (< 0.93 Å) crystal structures. As seen in Fig. 3, bonds involving the nitrogen atom along the main chain of α-helices are slightly, but significantly, longer than those of the β-strands, which is consistent with the increased electron density at this atom determined from our QM calculations and NMR data [32]. The longer bonds imply greater rotational freedom and less π character in α-helices compared to β-strands. The proposal that electron densities at the main chain atoms ultimately determine α-helix propensity is consistent with the observations of Wishart et al. [32] and of Creamer and Rose [7] who concluded that "general factors that drive helix formation must originate in the backbone." Furthermore, this notion is consistent with the role of inductive effects in the formation of helical structures as suggested by earlier studies [8,13,14]. Here, we have independently arrived at the critical role of inductive effects in helix formation and have for the first time provided quantitative estimates of the inductive effects of the 20 natural amino acids.
Figure 3 Bond lengths in high resolution structures of β-strands and α-helices. The average bond lengths (in Å) are shown for main chain bonds: (A) C-N, (B) N-Cα, and (C) Cα-C. In each case, the length of bonds in the α-helices is significantly longer than that in β-strands (Student's t-test, p < 0.01).
Additional determinants of protein folding
The main goal of these studies was to provide a more precise description of the electronic properties of amino acids in order to relate these features to protein folding. Few studies have explored this topic despite the fact that a better understanding of folding hinges on a detailed analysis of electron distributions and molecular orbitals of the main chain atoms. Towards this end, electronic properties of amino acid side chains have been derived from two major sources: quantum mechanical (PM3) calculations of Mulliken populations and the solution of equations that relate substituent effects to inductive effects, field effects, and polarizability. Semiempirical QM methods have been used with success to predict electronic effects such as charge transfer [47], proton affinities [48,49], rotational states related to protonation [50], and heat of formation [34,35]. Although more recent ab initio methods (including the application of density functional theory) may prove superior, in some cases the results with semiempirical approaches have been comparable to those obtained with more demanding ab initio calculations [35-37]. QM calculations have previously been used to define electronic effects of substituents in terms of surrogate measures that include electron densities and bond lengths at proton donor groups [18,22-24,51]. For example, bond lengths in pentaoxyphosphoranes calculated from ab initio methods showed a highly significant correlation with experimentally measured pKa's [51]. However, the derivation of electronic properties is potentially limited by certain factors such as the relative weighting of the various scales in the solution of equation (1) and the accuracy of the QM calculations. Nevertheless, Topsom [18] concluded that absolute measures from QM calculations are not necessary for most studies of substituent effects. Hopefully, this preliminary analysis will stimulate further development of the conceptual framework needed to precisely define the electronic features of amino acid side chains.
Notwithstanding these potential limitations, the work presented here reveals a potential role of electronic factors, in particular inductive effects, in determining preference for secondary structure. The significance of these effects should not be underestimated because studies have shown that inductive effects extend across non-conjugated bonds in proteins [30] and may even affect electron density over a distance of several residues [52]. Despite progress in characterizing factors that affect protein folding, hydrophobic effects and electronic effects do not fully account for the structural preferences of amino acids. Most likely, the remaining forces that contribute to folding result from two types of context effect: nearest neighbor and tertiary stabilization effects [4]. Tertiary stabilization refers to the observations of Kabsch and Sander [53] that the same 5 amino acids could be found in both α-helical and β-strand conformations in different proteins. Presumably in attaining the energy minimum of the whole protein, smaller modules may assume secondary structures that do not represent the energy minimum of that particular module owing to contact-assisted structural consolidation during condensation of folding [4]. Of course, tertiary stabilization ultimately involves various electronic effects: electrostatic interactions, dispersion forces, dipole alignment, and rotational flexibility (including hyperconjugation).
Conclusion
This paper presents a thorough description of the electronic properties of amino acid side chains. Quantitative scales were derived for representing inductive, resonance, and field effects, and polarizability (steric) factors. Regression analysis revealed that Mulliken population values at the Cα atom and inductive effects were the best predictors of helix preference. Thus, preference for secondary conformation appears to be influenced by the electronic properties of amino acid side chains. With further refinement of these properties, it may be possible to describe protein folding purely in electronic terms, including electron densities, inductive effects, field effects, and polarizability. The correlation data presented here suggest that such a strategy may yield important new insights into factors that promote the folding of proteins.
Methods
QM calculations
Computational analysis was performed with a Silicon Graphics Indigo2 workstation outfitted with the Insight II software package (Accelrys; San Diego, CA). PM3 [54] calculations of Mulliken populations and polarizability were performed using the MOPAC program with restricted Hartree-Fock methods. For comparison, MNDO [55] and AM1 [56] methods were also used to calculate properties in initial studies. The electronic features of amino acids were analyzed in one of several contexts. With the exception of proline, individual amino acids were evaluated in their zwitterion form in order to gain insight into their electronic properties independent of the context of a protein. Analysis of the various molecules was performed in the absence of solvent to simplify the system and to focus on inherent tendencies of amino acid side chain groups. The side chains of aspartic acid and glutamic acid carried a net charge of -1 e.u., whereas the side chains of arginine and lysine were +1 e.u. All other side chains were neutral.
In order to distinguish inductive vs. resonance effects, in some QM calculations the amino acid side chains (from Cβ outward) were attached at the 4-position to the reporter molecules cyclohexanol and phenol, which in their unsubstituted forms represented glycine (i.e., a hydrogen atom side chain). The geometries of the amino acids and substituted rings were optimized a priori through energy minimization and data were averaged from the two lowest energy structures. Both the absolute and relative values for the Mulliken populations for these two conformations were very consistent with a correlation coefficient of 0.99. The electronic structure of amino acids in other conformations will be somewhat different; however, analysis of a myriad of possible higher energy structures is not possible. Consequently, we have focused on the lowest energy conformations to derive intrinsic properties of amino acid side chains. Small deviations from the lowest energy conformation have little effect on the overall QM calculations (r = 0.99), whereas large deviations from this structure are uncommon and therefore less reflective of inherent properties. The behavior of the hydroxyl moiety (bond lengths and Mulliken populations) in substituted cyclohexanol and phenol was evaluated as an indicator of the substituent effects of the side chains. Various groups have used a similar approach to study the effects of other types of chemical substituents [18,22,23,51]. Mulliken population values derived from the zwitterion data are summarized in Table 1 for the heavy chain atoms of the 20 amino acids. In addition, polarizability values (see next section) derived from these calculations are presented.
Derivation of the polarizability (σα) scale
Charton [15] and Chalvet et al. [16] included steric factors in their derivation of substituent effects, whereas Topsom [18] and Graton et al. [23] included a polarizability term in their equations. It appears that both terms refer to the same effect, namely the overall size and deformability of a chemical group. Thus, we considered these terms to be roughly equivalent. Because polarizability can be evaluated directly from QM calculations, this is the convention that has been adopted for the present work. The average polarizability (α component) was determined with PM3 calculations as described above. The original values for the 20 amino acid side chains ranged from 1–13 Å3. In order to normalize the various substituent scales, these original values were multiplied by a factor of 10-2 to obtain the data presented in Table 2.
Derivation of inductive (σI) and resonance (σR) scales
In order to tease apart inductive versus resonance effects of substituents, various groups have characterized the effect of a substituent in the context of π interactions (i.e., attached to a phenol ring) and compared this with effects produced in molecules that lack significant resonance, such as bicyclooctanes or cyclohexane [15,18]. A similar approach was used here to characterize the amino acids. Side chain atoms from Cβ outward were bonded to cyclohexanol or phenol at the 4-position with the Biopolymer module of the software package. Cyclohexanol and phenol served as the standards for comparison and represented the glycine side chain. The structures were subjected to extensive energy minimization prior to QM calculations with the PM3 semiempirical method. Key values for the hydroxyl atoms provided the basis for derivation of the electronic properties of amino acid side chains. Inductive effects (σI) of side chains were derived from Mulliken population analysis of the hydroxyl hydrogen atom (HMULL) in cyclohexanol according to equation (2), where aa represents any amino acid and gly represents the glycine reference data (cyclohexanol).
(HMULLaa – HMULLgly) X 100 = HMΔCY (2)
The values were multiplied by 100 in order to normalize them in relation to the pKa values. These normalized Mulliken population data are referred to in this section as HMΔCY. These values also represent the inductive effects (σI) of the amino acid side chains. Similar normalized Mulliken population data for the amino acid side chains in the context of phenol (HMΔPH) were calculated from the PM3 results. The resonance effect (σR) scale was then derived according to equation (3).
HMΔPH – HMΔCY = σR (3)
Bond length analysis
A panel of 7 proteins was selected from the Protein Data Bank on the basis of their high resolution (< 0.93 Å) crystal structures and inclusion of both α-helices and β-strands. The panel included: crambin (1ejg; 0.54 Å resolution), aldose reductase (1us0; 0.66 Å), syntenin (1r6j; 0.73 Å), subtilisin (1gci; 0.78 Å), α-lytic protease (1ssx; 0.83 Å), ribonuclease (1dy5; 0.87 Å), and cholesterol oxidase (1n4w; 0.92 Å). Bond lengths along the main chain of randomly selected secondary structures were measured automatically. A total of 450 bonds were examined in α-helical conformations and 343 in β-strands.
Acknowledgements
The author thanks Dr. Stephan Witt and Dr. Ronald Bradley for reviewing the manuscript and for helpful discussions.
==== Refs
Karplus M Weaver DL Protein-folding dynamics Nature 1976 260 404 406 1256583 10.1038/260404a0
Kim PS Baldwin RL Intermediates in the folding reactions of small proteins Annu Rev Biochem 1990 59 631 660 2197986 10.1146/annurev.bi.59.070190.003215
Fersht AR Optimization of rates of protein folding: The nucleation-condensation mechanism and its implication Proc Natl Acad Sci USA 1995 92 10869 10873 7479900
Daggett V Fersht AR Is there a unifying mechanism for protein folding? Trends Biochem Sci 2003 28 18 25 12517448 10.1016/S0968-0004(02)00012-9
Kauzmann W Some factors in interpretation of protein denaturation Adv Protein Chem 1959 14 1 63 14404936
Dill KA Dominant forces in protein folding Biochemistry 1990 29 7133 7155 2207096 10.1021/bi00483a001
Creamer TP Rose GD Side-chain entropy opposes α-helix formation but rationalizes experimentally determined helix-forming propensities Proc Natl Acad Sci USA 1992 89 5937 5941 1631077
Dwyer DS Electronic properties of the amino acid side chains contribute to the structural preferences in protein folding J Biomol Struct Dyn 2001 18 881 92 11444376
Lowe JP The barrier to internal rotation in ethane Science 1973 179 527 532
Pophristic V Goodman L Hyperconjugation not steric repulsion leads to the staggered structure of ethane Nature 2001 411 565 568 11385566 10.1038/35079036
Brunck TK Weinhold F Quantum-mechanical studies on the origin of barriers to internal rotation about single bonds J Am Chem Soc 1979 101 1700 1709 10.1021/ja00501a009
Weinhold F A new twist on molecular shape Nature 2001 411 539 541 11385553 10.1038/35079225
Eberhardt ES Panasik N JrRaines RT Inductive effects on the energetics of prolyl peptide bond isomerization: implications for collagen folding and stability J Am Chem Soc 1996 118 12261 12266 10.1021/ja9623119
DeRider ML Wilkens SJ Waddell MJ Bretscher LE Weinhold F Raines RT Markley JL Collagen stability: insights from NMR spectroscopic and hybrid density functional computational investigations of the effect of electronegative substituents on prolyl ring conformations J Am Chem Soc 2002 124 2497 2505 11890798 10.1021/ja0166904
Charton M Taft RW Electrical effect substituent constants for correlation analysis Physical Organic Chemistry 1981 13 New York: Wiley 120 252
Chalvet O Daudel R Peradejordi F Lowdin PO, Pullman B Application of the molecular orbitals to the study of base strength Molecular Orbitals in Chemistry, Physics and Biology 1964 New York: Academic Press 475 484
Pross A Radom L Taft RW A theoretical approach to substituent interactions in substituted benzenes Physical Organic Chemistry 1981 13 New York: Wiley 1 60
Topsom RD Some theoretical studies of electronic substituent effects in organic chemistry Prog Phys Org Chem 1987 16 125 191
Hansch C Leo A Taft RW A survey of Hammett substituent constants and resonance and field parameters Chem Rev 1991 91 165 195 10.1021/cr00002a004
Taft RW Polar and steric substituent constants for aliphatic and o-benzoate groups from rates of esterification and hydrolysis of esters J Am Chem Soc 1952 74 3120 3128 10.1021/ja01132a049
Brinck T Haeberlein M Jonsson M A computational analysis of substituent effects on the O-H bond dissociation energy in phenols: polar versus radical effects J Am Chem Soc 1997 119 4239 4244 10.1021/ja962931+
Zhang HY Sun YM Wang XL Electronic effects on O-H proton dissociation energies of phenolic cation radicals: a DFT study J Org Chem 2002 67 2709 2712 11950325 10.1021/jo016234y
Graton J Berthelot M Gal JF Girard S Laurence C Lebreton J Le Questel JY Maria PC Naus P Site of protonation of nicotine and nornicotine in the gas phase: pyridine or pyrrolidine nitrogen? J Am Chem Soc 2002 124 10552 10562 12197757 10.1021/ja017770a
Alhaider AA Selassie CD Chua SO Lien EJ Measurements of ionization constants and partition coefficients of guanazole prodrugs J Pharmaceut Sci 1982 71 89 93
Bader RF Atoms in Molecules A Quantum Theory 1990 Oxford: Clarendon Press
Wiberg KB Hadad CM Breneman CM Laidig KE Murcko MA LePage TJ The response of electrons to structural changes Science 1991 252 1266 1272
Ramachandran GD Sasisekharan V Conformation of polypeptides and proteins Adv Prot Chem 1968 23 283 437
Eley DD Spivey DI Semiconductivity in hydrated hemoglobin Nature 1960 188 725 13726329
Patten F Gordy W Temperature effects on free radical formation and electron migration in irradiated proteins Proc Natl Acad Sci USA 1960 46 1137 1144 16590725
Pruetz WA Land EJ Charge transfer in peptides. Pulse radiolysis investigation of one-electron reactions in dipeptides of tryptophan and tyrosine Int J Radiat Biol 1979 36 513 520
Hammett LP The effect of structure upon the reactions of organic compounds. Benzene derivatives J Am Chem Soc 1937 59 96 103 10.1021/ja01280a022
Wishart DS Sykes BD Richards FM Relationship between nuclear magnetic resonance chemical shift and protein secondary structure J Mol Biol 1991 222 311 333 1960729 10.1016/0022-2836(91)90214-Q
Osapay K Case DA Analysis of proton chemical shifts in regular secondary structure of proteins J Biomol NMR 1994 4 215 30 8019135
Tubert-Brohman I Guimaraes CRW Repasky MP Jorgensen WL Extension of the PDDG/PM3 and PDDG/MNDO semiempirical molecular orbital methods to the halogens J Comput Chem 2004 25 138 150 14635001 10.1002/jcc.10356
Stewart JJ Comparison of the accuracy of semiempirical and some DFT methods for predicting heats of formation J Mol Model 2004 10 6 12 14655037 10.1007/s00894-003-0157-6
Casadesus R Moreno M Gonzalez-Lafont A Lluch JM Repasky MP Testing electronic structure methods for describing intermolecular H. H interactions in supramolecular chemistry J Comput Chem 2004 25 99 105 14634997 10.1002/jcc.10371
McCormack AL Somogyi A Dongre AR Wysocki VH Fragmentation of protonated peptides: surface-induced dissociation in conjunction with a quantum mechanical approach Anal Chem 1993 65 2859 2872 8250266 10.1021/ac00068a024
Kidera A Konishi Y Oka M Ooi T Scheraga HA Statistical analysis of the physical properties of the 20 naturally occurring amino acids J Prot Chem 1985 4 23 55 10.1007/BF01025492
Levitt MA Simplified representation of protein conformations for rapid simulation of protein folding J Mol Biol 1976 104 59 107 957439 10.1016/0022-2836(76)90004-8
Weaver JL Williams RW Amide III frequencies for ala-X peptides depend on the X amino acid size Biopolymers 1990 30 593 597 2265230 10.1002/bip.360300511
Chou PY Fasman GD Empirical predictions of protein conformation Annu Rev Biochem 1978 47 251 276 354496 10.1146/annurev.bi.47.070178.001343
Wojcik J Altmann KH Scheraga HA Helix-coil stability constants for the naturally occurring amino acids in water. XXIV. Half cystine parameters from random poly(hydroxybutylglutamine-co-S-methythio-L-cysteine Biopolymers 1990 30 121 134 10.1002/bip.360300113
Chakrabartty A Baldwin RL Stability of α-helices Adv Prot Chem 1995 46 141 176
Chou PY Fasman GD Conformational parameters for amino acids in helical, β-sheet, and random coil regions calculated from proteins Biochemistry 1974 13 211 245 4358939 10.1021/bi00699a001
Williams RW Chang A Juretic D Loughran S Secondary structure predictions and medium range interactions Biochim Biophys Acta 1987 916 200 204 3676331
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
van der Vaart A Merz KM Jr The role of polarization and charge transfer in the solvation of biomolecules J Am Chem Soc 1999 121 9182 9190 10.1021/ja9912325
Berthelot M Decouzon M Gal JF Laurence C Le Questel JY Maria PC Tortajada J Gas-phase basicity and site of protonation of polyfunctional molecules of biological interest: FT-ICR experiments and AM1 calculations on nicotines, nicotinic acid derivatives, and related compounds J Am Chem Soc 1991 56 4490 4494
Rutherford TJ Wilkie J Vu CQ Schnackerz KD Jacobson MK Gani D NMR studies and semi-empirical energy calculations for cyclic ADP-ribose Nucleosides Nucleotides Nucl Acids 2001 20 1485 1495 10.1081/NCN-100105243
Elmore DE Dougherty DA A computational study of nicotine conformations in the gas phase and in water J Org Chem 2000 65 742 747 10814006 10.1021/jo991383q
Davies JE Doltsinis NL Kirby AJ Roussev CD Sprik M Estimating pKa values for pentaoxyphosphoranes J Am Chem Soc 2002 124 6594 6599 12047179 10.1021/ja025779m
Gmeiner WH Facelli JC Quantum mechanical calculations and experimental measurement of N-terminal charge effects on 1HN and 1HCα chemical shifts in peptides Biopolymers 1996 38 573 581 8722227 10.1002/(SICI)1097-0282(199605)38:5<573::AID-BIP3>3.0.CO;2-P
Kabsch W Sander C On the use of sequence homologies to predict protein structure: Identical pentapeptides can have completely different conformations Proc Natl Acad Sci USA 1984 81 1075 1078 6422466
Stewart JJP Optimization of parameters for semiempirical methods. I. Method J Comput Chem 1989 10 209 220 10.1002/jcc.540100208
Dewar MJS Thiel W Ground states of molecules. 38. The MNDO method. Approximations and parameters J Am Chem Soc 1977 99 4899 4907 10.1021/ja00457a004
Dewar MJS Zoebisch EG Healy EF Stewart JJP AM1: a new general purpose quantum mechanical molecular model J Am Chem Soc 1985 107 3902 3909 10.1021/ja00299a024
Hanai T Koizumi K Kinoshita T Arora R Ahmed F Prediction of pKa values of phenolic and nitrogen-containing compounds by computational chemical analysis compared to those measured by liquid chromatography J Chromotog A 1997 762 55 61 10.1016/S0021-9673(96)01009-6
Edsall JT Cohn EJ, Edsall JT Dipolar ions and acid-base equilibria Proteins, amino acids and peptides as dipolar ions 1965 New York: Hafner Publishing 75 115
|
16078995
|
PMC1185526
|
CC BY
|
2021-01-04 16:30:50
|
no
|
BMC Chem Biol. 2005 Aug 3; 5:2
|
utf-8
|
BMC Chem Biol
| 2,005 |
10.1186/1472-6769-5-2
|
oa_comm
|
==== Front
BMC Evol BiolBMC Evolutionary Biology1471-2148BioMed Central London 1471-2148-5-391598751010.1186/1471-2148-5-39Research ArticleConservation of pregnancy-specific glycoprotein (PSG) N domains following independent expansions of the gene families in rodents and primates McLellan Andrew S [email protected] Wolfgang [email protected] Tom [email protected] Department of Biochemistry, Biosciences Institute, University College Cork, College Road, Cork, Ireland2 Tumor Immunology Group, LIFE Center, University Clinic Grosshadern, Ludwig-Maximilians-University Muenchen, Marchioninistrasse 23, D-81377 Muenchen, Germany2005 29 6 2005 5 39 39 21 1 2005 29 6 2005 Copyright © 2005 McLellan et al; licensee BioMed Central Ltd.2005McLellan 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
Rodent and primate pregnancy-specific glycoprotein (PSG) gene families have expanded independently from a common ancestor and are expressed virtually exclusively in placental trophoblasts. However, within each species, it is unknown whether multiple paralogs have been selected for diversification of function, or for increased dosage of monofunctional PSG. We analysed the evolution of the mouse PSG sequences, and compared them to rat, human and baboon PSGs to attempt to understand the evolution of this complex gene family.
Results
Phylogenetic tree analyses indicate that the primate N domains and the rodent N1 domains exhibit a higher degree of conservation than that observed in a comparison of the mouse N1 and N2 domains, or mouse N1 and N3 domains. Compared to human and baboon PSG N domain exons, mouse and rat PSG N domain exons have undergone less sequence homogenisation. The high non-synonymous substitution rates observed in the CFG face of the mouse N1 domain, within a context of overall conservation, suggests divergence of function of mouse PSGs. The rat PSG family appears to have undergone less expansion than the mouse, exhibits lower divergence rates and increased sequence homogenisation in the CFG face of the N1 domain. In contrast to most primate PSG N domains, rodent PSG N1 domains do not contain an RGD tri-peptide motif, but do contain RGD-like sequences, which are not conserved in rodent N2 and N3 domains.
Conclusion
Relative conservation of primate N domains and rodent N1 domains suggests that, despite independent gene family expansions and structural diversification, mouse and human PSGs retain conserved functions. Human PSG gene family expansion and homogenisation suggests that evolution occurred in a concerted manner that maintains similar functions of PSGs, whilst increasing gene dosage of the family as a whole. In the mouse, gene family expansion, coupled with local diversification of the CFG face, suggests selection both for increased gene dosage and diversification of function. Partial conservation of RGD and RGD-like tri-peptides in primate and rodent N and N1 domains, respectively, supports a role for these motifs in PSG function.
==== Body
Background
In tandemly repeated gene families, in which all members share a common function, there is a tendency for concerted evolution that is characterised by homogenisation of gene sequences [1]. Classical examples include the histone and ribosomal RNA genes. In such cases the expansion of gene families is driven by selection for high expression [2]. Concerted evolution is generally maintained by unequal crossover, intergenic gene conversion or other illegitimate recombination mechanisms [1,2]. Conversely, there are multigene families whose members encode diverse functions e.g. genes encoding immunoglobulin (Ig), T cell receptor (TCR) and major histocompatibility complex (MHC) proteins [1]. Such diversity occurs when there is less homogenisation than mutation, due to the evolution of specific programmed mutational mechanisms [3]. In addition, more complex modes exist; for example, the immunoglobulin heavy-chain variable-region (VH) genes encode proteins with identical functions, but exhibit little concerted evolution [4]. Instead, their evolution is governed by divergence and a birth-and-death process of gene duplication and dysfunctioning mutations [2].
Similar to other families of highly expressed trophoblast-specific genes such as the pregnancy-associated glycoproteins (PAG) [5], the pregnancy-specific glycoproteins, which are the most abundant foetal proteins in the maternal bloodstream during human late pregnancy, are encoded by multiple tandemly arrayed genes [6,7]. The PSG family of glycoproteins, with the related CEA-related cell adhesion molecule (CEACAM) proteins, are part of the immunoglobulin superfamily [8]. The Ig domain structure of the human and mouse PSGs differs, as follows: Human PSGs contain one V-like Ig domain (N), C2-like Ig domains (A and B) and relatively hydrophilic tails (C), with domain arrangements classified as type I (N-A1-A2-B2-C), type IIa (N-A1-B2-C), type IIb (N-A2-B2-C), type III (N-B2-C) and type IV (A1-B2-C) [9]. In contrast, mouse PSGs typically have three or more N domains followed by a single A domain [7,10]. The common ancestor of rodent and primate PSGs and CEACAMs was probably similar to CEACAM1, which is the only CEA family member with an identical gene structure in the human, rat and mouse that encodes all types of extracellular domains present in CEACAM and PSG proteins. The time of initial gene duplication is estimated at 90 Myr [11], approximately the time of rodent-primate divergence. The independent expansion of human and mouse PSG gene families occurred through further gene duplication and exon shuffling events [7,12,13].
The independent expansion of PSG gene families in rodents and primates indicates convergent evolution, implying that PSG function is conserved. These events can be interpreted in the context of evolutionary theories of parent-offspring and inter-sibling conflicts that promote transcriptional 'arms races' leading to high expression of trophoblast-specific genes that influence maternal investment in offspring [14,15]. In one scenario, duplicated PSG genes are selected because they increase effective PSG dosage, thereby enhancing an effect on maternal investment in offspring. In this context, it is noteworthy that human PSG N domains contain putative integrin-binding 'RGD' motifs that are proposed to mediate cell interactions with the extracellular matrix [16,17] and immune cells [18]. Such PSG-mediated functions could potentially influence trophoblast invasion or maternal immune cell function. However, not all human, and none of the mouse, PSGs contain an RGD motif [7], suggesting that, if human RGD motifs are functionally significant, there has been diversification of function of some human, and all mouse, PSGs, relative to a putative RGD-containing ancestor. In the context of parent-offspring conflict, such divergence might reflect co-evolution of PSGs and their receptors, similar to the co-evolution of ligand / receptor pairs observed in host-pathogen interactions [19,20].
In this study, we sought to analyse PSG evolution to determine the extent and patterns of rodent and primate PSG sequence divergence by analysing intraspecific and interspecies DNA substitution rates in PSG coding regions. We also sought evidence in support of functionality of RGD and RGD-like tri-peptide motifs in PSG amino-terminal effector domains.
Results
Pairwise comparisons of all 4-domain mouse PSG with all 4-domain human PSG full-length amino acid sequences indicates conservation of the amino-terminal N domain
With the exception of mouse PSG24, PSG30 and PSG31 and human PSG2 and PSG5, all PSGs for which full length sequences are available have a structure based on four Ig-like domains and a leader sequence that is cleaved during post-translational processing. The only type of domain found in all rodent and primate PSGs is the N domain located at the amino terminus. Indeed, this domain is shared by all members of the extended CEA family, suggesting that it may contain important functional motifs. We sought to test this hypothesis with respect to PSG function, by analysing both full-length PSG sequences and selected domains of possible functional importance. Alignments of full-length 4-domain human and mouse PSG protein sequences were generated with ClustalX, followed by pairwise comparisons of all mouse sequences with all human sequences. Mean Dayhoff PAM250 log scores were calculated for each alignment position and grouped by domain. The scores within each of the four domains were then visualised using box and whisker plots (which show the median value, upper and lower quartiles plus range) (Fig. 1). The N domains exhibited significantly higher scores (p < 0.001) than the other three domains, with positive scores indicating conservation. There was no evidence of interspecies conservation of the other domains, which is unsurprising given the known lack of orthology between human A1 / mouse N2, human A2 / mouse N3, and human B2 / mouse A domain pairs.
Figure 1 Box and whisker plots for Dayhoff PAM 250 scores determined by ClustalX alignment of full-length mouse PSGs with full-length human PSGs. At each position in the alignment, the Dayhoff PAM250 log score was determined for pairwise comparisons of each sequence in the set of mouse PSGs against all sequences in the set of human PSGs. Mouse Psg24, Psg30 and Psg31 along with human PSG2 and PSG5 were omitted from the analysis due to expansions or contractions of total domain complement which would complicate generation of the initial clustalX alignment. The scores were split into five groups, according to domain structure, and used to generate a box and whisker plot. Domain name abbreviations shown on the X-axis correspond to the following domain comparisons: L/L, human L versus mouse L domain; N/N1, human N versus mouse N1 domain; A1/N2, human A1 versus mouse N2 domain; A2/N3, human N2 versus mouse N3 domain; B2/A, human B2 versus mouse A domain. Significant differences of p < 0.0001 were observed for the N/N1 domain comparison when tested against the A1/N2 and A2/N3 data, and p < 0.0005 when tested against the B2/A data.
Novel rat PSG N1 domains identified by database searches
Rat N1 domain exon sequences were identified in NCBI and Ensembl databases. Three novel rat PSG genes were identified and named PSG41, PSG42 and PSG43 in keeping with accepted nomenclature [21]. We also identified a novel PSG40 splice variant with alternative leader and N1 domain exons, situated between the N1 and N2 domain exons of the published PSG40 sequence (NM_021677). Both BLAST and pattern matching methods retrieved the same rat PSG genes from different databases; therefore we considered our search to be exhaustive. All rat PSG genes were found to reside on contig NW_047556 and this was used for the prediction of remaining exons for each PSG gene based on BLAST generated alignments with mouse Psg gene sequences (Table 1). The CDS sequences of the novel predicted rat PSG genes and PSG40 splice variant are listed in additional file 1. We used our predicted sequences in preference to the publicly available sequences in our analyses.
SplitsTree analysis reveals relatively high contradiction in rat PSG N1 domain alignments, compared to mouse
Following the preliminary identification of amino-terminal N domain conservation, we planned to use an evolutionary tree building approach to further examine inter-domain relationships in rodent and primate PSGs. However, using split decomposition analysis, McLenachan et al. [22], in their study of a subset of human PSGs, concluded that it is not possible to accurately determine branch points in an evolutionary tree of human PSGs. Split decomposition analysis identifies contradictory relationships within alignment data; for example, there may be a pattern grouping PSGX and PSGY together, and another pattern grouping PSGY and PSGZ together [23]. This information is normally approximated when drawing evolutionary trees, however split decomposition is a non-approximation method that permits the building of trees with support indicated for relationships based on all patterns in the data. Such analysis can therefore predict to a limited extent the occurrence of sequence homogenisation e.g. by gene conversion or positive selection.
We performed split decomposition analysis on nucleotide sequences using the SplitsTree4 program [24] on the individual domain exons of mouse Psg genes (Fig. 2). For a more complete analysis of N1 domains we also performed the analysis using rat N1 domain exons, all known human N1 domain exons and all known baboon N1 domain exons (Fig. 3). We detected no conflicting signals for mouse Psg N1 domain exons (Fig. 2A), in contrast to the human N domain exons (Fig. 2B). However, our results for human N1 domains (Fig. 3B) differ from those obtained by McLenachan et al. [22] because we observed only two contradictions: i. regarding the relationship of PSG4 and PSG9 to each other, and to their nearest neighbours PSG3 and the common ancestor of PSG6 and PSG10 and, ii. between 'the relationship of PSG2 to PSG1 and PSG11'. This discrepancy is probably due to our inclusion of four extra PSG N1 domain sequences, and the fact that the PSG11 sequence (GenBank: M69025) used by McLenachan et al. [22] has been updated.
Figure 2 Split decomposition graphs for all mouse Psg domain exons and rat PSG N1 domain exons for observed (Hamming) distances. Split decomposition analysis was performed using nucleotide sequences for individual groups of PSG domain exons. (A) N1 domain exons; (B) N2 domain exons; (C) N3 domain exons (the N4 domain exon of Psg24 is used instead of N3; see Fig. 3B in McLellan et al. [41] for explanation); (D) A domain exons. Numbers indicate respective PSG genes. Scale bars represent 0.01 (A) or 0.1 nucleotide substitutions per site (B, C, D).
Figure 3 Split decomposition graphs for the rat PSG N1 domain exons, human PSG N domain exons, and baboon PSG N domain exons for observed (Hamming) distances. Split decomposition analysis was performed using nucleotide sequences for individual groups of PSG domain exons. (A) rat N1 domain exons; (B) human N domain exons; (C) baboon N domain exons. Scale bars represent 0.01 nucleotide substitutions per site. In (A) the putative N1 domain exon splice variants of PSG40 are identified with the suffix 'v1' for the published variant (NM_021677) and 'v2' for our predicted variant.
Analysis of the mouse N2 domains indicates numerous contradictions in the alignments of the Psg24, Psg29, Psg30, Psg31 and Psg32 group (Fig. 2B). In contrast, the N3 domains exhibit no discernable conflicts (Fig. 2C). The A domain only showed contradiction within the Psg24, Psg29, Psg30, Psg31 and Psg32 group (Fig. 2D). Examination of the rat PSG N1 domain exon alignments demonstrated minor contradictions between the common ancestor of PSG36, PSG37 and PSG39 and that of PSG38 and PSG41 (Fig. 3A). In contrast to all the other PSG N1 domains thus compared, the baboon PSGs demonstrate considerable conflicting signals as demonstrated by the 'spider's web' appearance of the SplitsTree graph (Fig. 3C).
Phylogenetic analysis indicates interspecific amino-terminal N domain conservation and identifies potential mouse / rat orthologues
Few examples of orthologous relationships between PSG sequences have been identified. In order to compare the relationship between rodent and primate amino-terminal N domain exon coding sequences, an NJ tree was produced (Fig. 4). The tree was generated from ClustalX alignments of nucleotide sequences, with bootstrapping 1000 times to test the reliability of branches. The human and baboon N sequences formed one distinct cluster, the mouse and rat N1 sequences formed a second, the mouse N2 domains formed a third and the mouse N3 domains formed a fourth. Of particular interest was the split between the ancestral N-type domain and the common ancestor of the N2 and N3 domains. The confidence of this split was 93% and demonstrates that the mouse N1 domains are more closely related to primate N domains than to the mouse N2 and N3 domains. A similar comparison of the entire set of mouse and human PSG domains confirmed that the interspecific N domain clustering is unique because the human PSG A1 and A2 domains segregated into distinct branches (sharing a common ancestor with the mouse A domains) and the B2 domains cluster on a distinct branch (Fig. 5).
Figure 4 Phylogeny of the mouse N1, N2 and N3 domains, rat N1 domains and human N domains. NJ-tree of N domain nucleotide sequences on ClustalX alignments of corresponding amino acid sequences showing the evolution of mouse (Mmu) PSG N1, N2 and N3 domains in comparison with rat (Rho) N1 domains, human (Hsa) N domains and baboon (Pha) N domains. Alignments were bootstrapped 1000 times yielding the values shown for the main branches. The scale bar represents 0.1 nucleotide substitutions per site.
Figure 5 Phylogeny of all known mouse and human PSG N, A and B domains. NJ-tree of mouse (Mmu) and human (Hsa) N, A and B domain nucleotide sequences on ClustalX alignments of corresponding amino acid sequences showing the evolutionary relationships between domain types. Alignments were bootstrapped 1000 times yielding the values shown on the major branches. Scale bar represents 0.1 nucleotide substitutions per site.
Mouse and rat PSG gene coding sequences were analysed using an NJ plot which highlighted four putative orthologous relationships, as follows: rat PSG36 and mouse Psg24; rat PSG40 and mouse Psg29; rat PSG42 and mouse Psg32; rat PSG38 and mouse Psg16 (Fig. 6). There is also distinct branching of rat PSG43 with mouse Psg30 and Psg31. The orthologous relationship is also supported for PSG36 and Psg24 because both contain five N domains.
Figure 6 NJ tree of alignments of complete CDS of all known mouse and rat PSGs. Sequences of PSG40 – PSG43 are de novo predictions. Data were bootstrapped 1000 times and all major branches yielded values of 95–100%. The scale bar represents 0.1 nucleotide substitutions per site.
PSG N domain sequences are generally conserved but alignments reveal specific regions that may be diverging
The crystal structure of mouse CEACAM1 (soluble murine sCEACAM1a [1,4]) has been resolved [25]. Comparison of the mouse PSG N1 domains identifies the predicted β-sheet-forming CFG β-strands as the most variable regions of the N domains (Fig. 7A). The CFG face of CEACAM N domains has been shown to interact with pathogens and mammalian proteins (Fig. 7B). Within Box 1 and Box 2, there is considerable variation between mouse N1 domains, which is illustrated quantitatively using Dayhoff charts (Figs. 8 – 10). Positive Dayhoff scores and generally low standard deviations indicate good conservation of mouse PSG N1 domains (Fig. 8), and even stronger conservation of human PSG N domains (Fig. 9). The latter may be explained by homogenisation of human PSG gene sequences [22]. Dayhoff score analysis using comparisons of all mouse N1 domain versus all human N domain ClustalX aligned sequences gives an indication, at the amino acid level, of the general pattern of evolution of these domains since the rodent / primate divergence (Fig. 10). Again, the majority of residues exhibit good conservation, and relatively little variability is observed between pair-wise comparisons particularly with regard to residues that are involved in protein folding. The reduction in size of Box 2 in Fig. 8 and Fig. 10 is explained by deletions of mouse DNA sequences, requiring exclusion of the corresponding amino acids from the analysis.
Figure 7 High amino acid sequence variability is found in the CFG faces of mouse PSG N1 domains. Alignments of the mouse PSG N1 domain amino acid sequences were performed using ClustalW. The locations of the β-strands (A-G) were derived from the crystal structure of the mouse CEACAM1 N domain [25], and are indicated by blue arrows. The boxed amino acids sequences form the CFG face of the N domain (deduced by structural modelling). (A) Alignment of mouse PSG N1 domain amino acid sequences. The signal peptide (leader) cleavage site is shown as a dotted line and N domain amino acid numbering commences from the first amino acid of the mature N domain. (B) Alignment of CEACAM N domains (minus signal sequences) showing all N domain interactions with pathogens and known binding partners (referenced as follows: 1 [48]; 2 [49]; 3 [50]; 4 [51]; 5 [52]; 6 [53]; 7 [54]; 8 [55]).
Figure 8 Dayhoff PAM250 plot for ClustalX-aligned mouse N1 domain amino acid sequence comparisons. At each position in the alignment, the Dayhoff PAM250 log score was determined for pairwise comparisons of each sequence in the set against all the others in the set. Mean and standard deviation were calculated for scores at each residue position. Regions representing the CFG face are boxed (1–3) and an RGD-like motif is indicated. Other specified amino acids are denoted by the single letter code. Note that amino acid positions are numbered in vertical orientation.
Figure 9 Dayhoff PAM250 plot for ClustalX-aligned human N domain amino acid sequence comparisons. At each position in the alignment, the Dayhoff PAM250 log score was determined for pairwise comparisons of each sequence in the set against all the others in the set. Mean and standard deviation were calculated for scores at each residue position. Regions representing the CFG face are boxed (1–3) and an RGD-like motif is indicated. Other specified amino acids are denoted by the single letter code. Note that amino acid positions are numbered in vertical orientation.
Figure 10 Dayhoff PAM250 plots for ClustalX-aligned N1 (mouse) and N (human) domain amino acid sequence comparisons. At each position in the alignment, the Dayhoff PAM250 log score was determined for pairwise comparisons of each sequence in the mouse set against all sequences in the human set. Mean and standard deviation were calculated for scores at each residue position. Regions representing the CFG face are boxed (1–3) and an RGD-like motif is indicated. Other specified amino acids are denoted by the single letter code. Note that amino acid positions are numbered in vertical orientation.
To gain further insight into mouse Psg N domain exon evolution, the N1, N2 and N3 domain exons of mouse Psg genes (mN1, mN2 and mN3, respectively), the N1 domain exons of rat PSG genes (rN1) and the N domain exons of human PSG genes (hN) were analysed in the following comparisons: mN1 vs mN2; mN1 vs mN3; mN2 vs mN3; mN1 vs rN1; mN1 vs hN. Synonymous (ds) and non-synonymous (dn) substitutions per synonymous and non-synonymous site, respectively, were determined in each case for all combinations of PSG gene pairwise comparisons, and box and whisker plots were generated from the data (Fig. 11). The majority of data points derived from individual comparisons lie under the 45° line of equivalence where dn = ds, and most variation in the comparisons lies within the values of ds (Fig. 11A). When the data are presented as box and whisker plots, the values are indicative of conservation, with median values ranging from 0.48 – 0.70 (Fig. 11B). The higher values for median dn/ds in the mN1 vs rN1 comparison appear to be the result of a tighter ds distribution as observed in Fig. 11A, with values not exceeding one substitution per synonymous site in any pairwise comparison.
Figure 11 Nonsynonymous versus synonymous substitution rates for pairwise comparisons between N domains. The number of nonsynonymous substitutions per nonsynonymous site (dn) and the number of synonymous substitutions per synonymous site (ds) was calculated using the method of Yang and Neilson [46] for pairwise nucleotide comparisons. The N1, N2 and N3 domains of mouse PSGs (mN1, mN2 and mN3, respectively), the N1 domain of rat PSGs (rN1) and the N domain of human PSGs (hN) comprised individual data sets that were analysed in the following comparisons: mN1 vs mN2; mN1 vs mN3; mN2 vs mN3; mN1 vs rN1; mN1 vs hN. (A) Plot of dn against ds where each data point represents a pairwise comparison of a nucleotide sequence taken from each set under comparison. The 45° line of equivalence is drawn where dn = ds. (B) Box and whisker plot of dn/ds calculated from the pairwise comparisons of all sequences in one dataset against all sequences in the other dataset. Significant differences of p < 0.0001 (calculated by the Mann-Whitney method) were observed between all comparisons except intra-mouse comparisons.
In view of the sequence variations in the CFG face, which are visible in alignments (Fig. 7A), against a background of overall conservation, as estimated from dn/ds analysis, we sought to determine whether the dn/ds values were higher in the CFG face than the ABED face of the N1 domain. Nucleotide sequence alignments were generated using all mouse Psg N1 domain exons (based on protein alignments), and the nucleotides present in the three sections comprising the CFG face (Boxes 1, 2 & 3; Fig. 7A) were separated from those comprising the ABED face. The two new sets of data were analysed individually to determine mean dn and ds values from pairwise comparisons of all sequences within each dataset (Fig. 12). A plot of dn vs ds for the ABED face of the mouse N1, N2 and N3 domains (Fig. 12A) demonstrates a distribution of pairwise-alignment data points which overwhelmingly lie below the line of equivalence. However, a similar plot generated from analysis of the CFG face has data points distributed approximately equally on both sides of the line of equivalence (Fig. 12B). This is due predominately to a higher number of non-synonymous substitutions. The values of dn/ds obtained for the CFG face in the N1, N2 and N3 domains of the mouse and the N1 domain of the rat are all significantly greater than the values obtained for the ABED face (p < 0.0001, Fig. 12C). The dn/ds values obtained for the mouse N1, N2 and N3 domain CFG faces equal or exceed 1.0, with the highest median value of 1.1 observed in the N1 domain. The rat N1 domains are more conserved, with dn/ds values derived from both the CFG and ABED faces under 1.0 on average.
Figure 12 Nonsynonymous versus synonymous substitution rates in the mouse N domain CFG and ABED faces. The nucleotide sequences encoding the CFG and ABED faces of the N1 domain and equivalent regions of the N2 and N3 domains were separated and compared individually. The number of nonsynonymous substitutions per nonsynonymous site (dn) and the number of synonymous substitutions per synonymous site (ds) were calculated using the method of Yang and Neilson [46] for pairwise nucleotide comparisons. Plots are shown of dn versus ds for regions comprising (A) the ABED face and, (B) the CFG face, where each data point represents a pairwise comparison of two nucleotide sequences taken from the dataset being examined. The 45° line of equivalence is drawn where dn = ds. (C) Box and whisker plot of dn/ds calculated from pairwise comparisons of all sequences in one dataset against all others in the set. Data derived from sets of rat N1 CFG and ABED faces were also analysed for comparative purposes. Significant differences between CFG and ABED faces for each domain are shown, where '***' is p < 0.0001 (calculated by the Mann-Whitney method).
Evidence of conservation of RGD-like motifs in mouse N1 domains
Within Box 3 of the CFG face (Fig. 7A) there is evidence of conservation of putative integrin-interacting RGD-like motifs in the mouse N1 domain, which may have functional significance. To investigate this possibility further, a survey of all mouse, rat, baboon and human PSG RGD, and related, motifs was compiled (Fig. 13). Extant primate and rodent PSG RGD-like motifs are linked in sequence space by an RGD motif encoded by the sequence CGA GGA GAT which, incidentally, is not observed in any of the extant PSG coding sequences. The most commonly observed motif, RGD, is encoded by CGA GGT GAT, and the majority of variants are closely related to this sequence. In rodents, RGE and HGE are the most commonly observed motifs. However, the NGK motif, which is not an RGD-like motif as we have defined it, is well represented, and is separated in sequence space from HGE by a transition and a transversion.
Figure 13 Relationship between RGD-like motifs in human, baboon, mouse and rat PSG N domains. The font size used for each tri-peptide motif represents relative abundance among the PSG proteins, and the codon sequences are shown underneath. Arrows represent single or double (x2) transitions (ts) or transversions (tv) as indicated. Motifs and codon sequences in grey type are intermediates that have not been observed in vivo. Primate RGD-like motifs cluster naturally in the left-hand box, whereas those of the rodents cluster in the right-hand box. The baboon derived PAE motif is an outlier and is bracketed.
Of the seventeen aligned mouse PSG N1 domain exon sequences, 53% possess a tri-peptide at the site of the RGD-like motif belonging to the RGD-like 5-1-4 tri-group (as defined in the Methods section). For comparative purposes, tri-groups were determined for tri-peptide motifs at fifty random positions within the alignment. The number of most commonly represented tri-groups at each position was expressed as a percentage of the number of aligned sequences, and the mean and standard deviation was determined to indicate the mean maximal tri-group representation for the 50 random alignment positions. The control value obtained was 67.6 ± 22.9%; the value of 53% of 5-1-4 tri-groups at the RGD site therefore lies within the control range, albeit 14.6% below the mean value. However, a more revealing statistic is derived from aligning the mouse N1 domains with the mouse N2 and N3 domains (see additional file 2), compared to aligning the mouse N1 domains with the human N domains. In the former comparison (mouse N1 vs N2 and N3 domains) the most commonly represented tri-group is 4-2-5, with 27% representation. This tri-group is not RGD-like and its representation is lower than the mean maximal tri-group representation of 49.8 ± 22.7% determined for fifty random alignment positions. However, when the mouse N1 domain is aligned with the human N domain, the most commonly represented tri-group is the RGD-like 5-1-4 group which has 59% representation, comparable to the mean maximal tri-group representation of 60.7 ± 20.4%.
Discussion
We recently collated the full-length coding sequences of the entire mouse Psg gene family [7]. In the present study we aimed to identify evolutionary signals embedded in Psg gene and PSG protein sequences to determine whether PSG protein function has diverged between the rodent and primate lineages, and to attempt to understand the reasons for the independent expansions of rodent and primate PSG gene families.
Mouse and human PSG protein amino-terminal N domains exhibit different patterns of evolution. McLenachan et al. [22] analysed the evolution of a subset of human PSGs using split decomposition analysis and found, in individual comparisons of N, A1, B2 and C domain exons, strong contradictions in alignments, which they suggested was due to gene conversion and/or positive selection. Our similar analysis of an expanded set of human PSG sequences revealed a detectable, but less marked, degree of homogenisation. Analysis of mouse N and A domain exons showed that, in general, there is less evidence of purifying selection compared to the human, although there are examples of gene conversions as described previously for the closely related Psg21 and Psg23 genes [12]. Detailed analysis of alignments using plots of Dayhoff scores confirmed the difference between mouse and human N domain evolution.
Using dn/ds analysis for interspecies comparisons, we found that the PSG protein amino-terminal N and N1 domains are relatively conserved, consistent with conservation of function in rodents and primates. However, inspection of mouse PSG N1 domain alignments, and scrutiny of corresponding Dayhoff scores, revealed regions of apparently poor conservation. These regions correspond to the CFG face within the N1 domain of CEACAM1. In the CEACAM family, the CFG face interacts with pathogens and mammalian proteins. Comparisons of dn/ds values obtained from the CFG and ABED faces of mouse N1, N2 and N3 domains confirmed that the CFG face has evolved more rapidly than the ABED face in all three domains. The greatest effect was observed in the N1 domain exon with a doubling of the dn/ds ratio in the CFG face compared with the ABED face. The dn/ds ratio of 1.1 suggests weak positive selection on the CFG face of the N1 domain. The increase in the dn/ds ratio appears to be mainly due to an increase in the dn value, indicative of diversification. The high dn/ds values for the CFG face in the N2 and N3 domains, which are not known to interact with ligands, could be due to a low contribution of these sequences to the structural integrity of the IgV-like domain.
Interestingly, the rat N1 domain CFG face does not appear to have evolved as rapidly as the mouse N1 domain, with a dn/ds ratio of 0.9. This observation, combined with the relatively smaller number of PSG genes identified in the rat (eight to date, compared to seventeen in the mouse) and the higher level of gene homogenisation implied by split decomposition analysis suggests that the rat PSG gene family has not expanded or diversified as extensively as the mouse. However, we cannot exclude the possibility that further rat PSG genes may yet be identified because there may be under-representation in the WGS database [26]. Notwithstanding this possibility, there has clearly been ongoing turnover of the PSG gene family in all of the lineages analysed, as there are no known human orthologues of rat and mouse PSGs, and only four potential orthologous relationships between known rat and mouse PSGs.
These findings suggest partial conservation of PSG N domain function across rodent and primate lineages. However, the relaxed constraint on the CFG face of mouse PSGs suggests diversification of binding partners or modification of existing ligand-binding kinetics, analogous to the CEACAMs. This observation receives experimental support from the recent observation that treatment of mouse macrophages in vitro with recombinant mouse PSG17N, or human PSG1 or PSG11, induces cytokine expression; however, only in the case of mouse PSG17N does this depend on CD9 receptor expression [27]. Divergence of PSG function is also suggested by differences in the level and developmental timing of expression of different mouse PSGs [7,12], expansion of N domain number in PSG24, PSG30 and PSG31 [7], and loss of secretory signals in PSG32 and in the brain-specific splice variant of PSG16.
As noted above, the only PSG receptor identified to date is the integrin-associated tetraspanin, CD9, which binds the N1 domain of mouse PSG17 but not, apparently, to human PSGs [28]. However, a peptide containing the RGD motif from the human PSG9 N domain binds to a receptor on a promonocytic cell line suggesting that some human PSGs may effect their functions through an integrin-type receptor [18]. In this context, the high frequency of the RGD motif on an exposed loop in primate PSG N domains (seven of ten in human and five of fifteen in baboon) may be significant. Rodent PSG N1 domains do not have an RGD motif, but have a high frequency of the RGD-like motifs RGE, HGE and HAE on the CFG face. Under the null hypothesis that these motifs are unlikely to underpin structural integrity of the N1 domain and are therefore free of constraint, our analysis reveals evidence of unexpected conservation of RGD-like motifs in the N1 domain, which have been lost in the N2 and N3 domains. Given the high transition and transversion rates in the N1 domain and the fact that the mouse N1, N2 and N3 domains share a common ancestor after the divergence of the rodent / primate lineages, the conservation of RGD-like motifs exclusively in the N1 domain may have functional significance. We note that the RGE motif in the context of the POEM protein induced apoptosis of MC3T3-E1 cells in vitro [29]. We speculate that certain RGE or RGE-like motifs may elicit weak cell attachment, followed by apoptosis – a combination of properties, reminiscent of snake venom disintegrins [30,31], that could have important functional implications in the context of the extensive tissue remodelling that occurs during placentation [32].
In summary, our data are consistent with experimental evidence indicating functional convergence of rodent and primate PSGs, in spite of the independent expansions of the gene families in the two lineages. In the context of parent-offspring conflict, the homogenisation of human PSG sequences is consistent with the theory that placental hormones encoded by multigene families are monofunctional and selected for high expression, possibly due to coevolution with physiologically conflicting maternal mechanisms [15]. However, the evidence for positive selection on the CFG face of the N1 domain implies divergent evolution of rodent PSGs. Allied to the evidence for functionality of putative integrin-interacting RGD-like motifs in rodents, a scenario can be envisaged whereby the different RGD-like motifs observed in human and baboon PSGs also suggest some degree of functional divergence in these species.
Conclusion
Our analysis provides evidence for conservation of rodent and primate PSG amino-terminal N domains, with ongoing independent expansion of the gene families in the two lineages. There has been some diversification of the CFG face of mouse N1 domains, a region that includes putative integrin-interacting RGD-like motifs. Our analysis provides reassurance that the mouse Psg gene family is a suitable model system for the analysis of human PSG gene function.
Methods
Perl programs were written to perform most general sequence manipulations and iterative tasks and executed under ActivePerl v5.8.3 [33] on a Windows 2000 (Microsoft) platform.
Identification of novel rat PSG N1 domain exons
Blast searches of the NCBI [34] and Ensembl [35] RGSC3.1 rat genome databases were performed using coding sequences from known rat PSGs (PSG36-PSG40) and mouse PSGs. Additionally, a search pattern was developed and used to interrogate the Rattus_norvegicus.RGSC3.1.nov.dna_rm.contig.fa.gz archive obtained from the Ensembl FTP resource [36]. The search pattern was derived manually from alignments of amino acid sequences from the N domain exon of all known mouse and rat PSGs (mouse PSG16-PSG32 and rat PSG36-PSG40) generated using the ClustalX 1.81 windows interface [37]. In PROSITE format [38] the search pattern used was S-x-R-E-x(5)-G-x(3)-[IL]-x(3)-T-x(2)D-x(3)-Y-x(17,18)-L-x-V. Analysis was performed essentially as described [39], with the program modified to search for the selected pattern in peptides of fifty amino acids or greater derived from genomic DNA sequences translated in all six open reading frames. ClustalX alignments were produced using the complete open reading frames returned by the program combined with the N1 domains of rat PSG36-PSG40. The alignments were trimmed to include only N1 domain exon sequence and a Neighbour-Joining tree was generated using MEGA version 2.1 software [40] to aid the identification of the new sequences.
Phylogenic analysis
Mouse PSG sequences were obtained from McLellan et al. [41], rat PSG sequences were obtained as described above, human PSG sequences were obtained by name searches at the NCBI Entrez (nucleotide or protein options) database [42] and baboon N1 domain sequences were obtained as described [43]. To generate protein alignments for examination by eye, a Web based ClustalW utility was used [44], otherwise protein sequences were aligned with the ClustalX using the default parameters. Nucleotide alignments were generated based on ClustalX protein alignments, such that where a single dash was placed in the amino acid alignment, three dashes were placed in the equivalent codon position in the nucleotide alignment. The nucleotide alignments were then analysed using SplitsTree version 4b [24] and software and NJ trees were generated from the data (with bootstrapping 1000 times to test the reliability of branches). Individual domains of the mouse PSGs were also analysed by the split decomposition method using the same software. During NJ or Splitstree tree-building, the Jukes-Cantor [45] correction for multiple hits was applied and positions with gaps were ignored.
Comparisons of amino acids encoded at each site within alignments
Multiple alignments of either one set (e.g. all mouse PSG N1 domain exons only) or two sets (e.g. all mouse PSG N1 and N2 domain exons) of amino acid sequences were produced using ClustalX. A Perl program was written to perform the subsequent analysis. At each position of the alignment, the Dayhoff PAM250 log score was determined for pairwise comparisons of each sequence in the set against all the others in the set in one-set analyses, or of all set 1 sequences against all set 2 sequences in two-set analyses. The mean and standard deviation of scores obtained for the pairwise comparisons at each site were determined to give an indication of the general level of conservation and variability at the site. Sites where gaps were present in any of the sequences were not analysed. Where full-length mouse and human PSG amino acid sequences were compared, the scores were split into five groups at domain junctions and a box and whisker plot produced.
Evolutionary analysis
ClustalX was used to produce multiple alignments of either one set of amino acid sequences (e.g. all mouse PSG N1 domain exons only) or two sets combined (e.g. all mouse PSG N1 and N2 domain exons). These alignments were used to inform the alignment of corresponding nucleotide sequences as described above. Values of ds and dn were determined for pairwise comparisons of each sequence in a set against all the others in the set for one-set analysis, or of all set 1 sequences against all set 2 sequences for two-set analysis. The analysis was performed according the method of Yang and Neilsen [46] using the 'YN00' program in the PAML3.14 software package [47]. Before each pairwise comparison was executed, pairs of aligned sequences were extracted from the alignment file, placed in a Phylip format file and gapped positions were removed. Plots of dn vs ds, and box and whisker plots of dn/ds were produced in order to visualise the data. Where statistical significance was evaluated, the Mann-Whitney test was applied.
Analysis of tri-peptide amino acid property groupings
A perl program was written to analyse ClustalX alignments of mouse and human PSG N domain exons. These alignments were inspected and modified where necessary. For a tri-peptide at a given position within an alignment, a tri-group code was generated for tri-peptide motifs based on amino acid properties of the residues in the motif where group 1 contains G, A, S, T; group 2: V, L, I, M; group 3: F, Y, W; group 4: D, N, E, Q; group 5: H, K, R; group 6: P; group 7: C. For example, an RGD tri-peptide motif is represented by tri-group code 5-1-4 as arginine is in group 5, glycine is in group 1, and aspartate is in group 4. Conversely, tri-group 5-1-4 is 'RGD-like' in terms of the biochemical properties of the constituent amino acids. The number of sequences in the alignment containing each group code at a given position was determined. The most highly represented group code in the alignment at that position was used in the analysis. The program was designed to compare a user selected tri-peptide motif position with fifty randomly selected tri-peptide motif positions.
Authors' contributions
A. McLellan performed data collection and analysis and co-wrote the manuscript. W. Zimmermann and T. Moore co-conceived the project and co-wrote the manuscript.
Table 1 Rat PSG genes: nomenclature and references. Previously and newly identified rat PSG genes are listed with GenBank references. Where the GNOMON predicted sequence in GenBank differs from our prediction this is denoted by a single asterix beside the nucleotide accession number. A double asterix indicates the prediction of a putative splice variant with an alternative leader and N1-domain exon.
gene name alternative/ old name accession number (nucleotide) accession number (protein) NW_047566.1 contig (CDS start and end positions and orientation) Notes
PSG36 CGM1 NM_012702 XP_218391 947162 – 958811 (F)
PSG37 CGM3 NM_019126 NP_061999 1543892 – 1552811 (R)
PSG38 similar to brain CEA XM_214842* XP_214842 859024 – 870967 (F)
PSG39 CGM6 XM_218398 XP_218398 1562043 – 1571612 (R)
PSG40 CGM8 NM_021677** NP_067709 1009499 – 1015560 (F) putative novel splice variant
PSG41 XM_218390* XP_218390 888980 – 904807 (F)
PSG42 1108370 – 1118331 (F)
PSG43 1149404 – 1185428 (F)
Supplementary Material
Additional File 1
An ASCII text file containing the CDS sequences of novel predicted rat PSG41, PSG42 and PSG43 and a novel splice variant of PSG40.
Click here for file
Additional File 2
A rich text format file containing the Clustal W amino acid sequence multialignment of PSG N1, N2 and N3 domains. The RGD-like motif is boxed for comparison between domains.
Click here for file
Acknowledgements
We thank two anonymous referees for helpful comments. This work was supported by the Irish Higher Education Authority Program for Research in Third Level Institutions funded under the National Development Plan, and an Irish Health Research Board / Wellcome Trust 'New Blood' Research Fellowship to T. Moore.
==== Refs
Ohta T Evolution of gene families Gene 2000 259 45 52 11163960 10.1016/S0378-1119(00)00428-5
Ota T Nei M Divergent evolution and evolution by the birth-and-death process in the immunoglobulin VH gene family Mol Biol Evol 1994 11 469 482 8015440
Ohta T On the evolution of multigene families Theor Popul Biol 1983 23 216 240 6612633 10.1016/0040-5809(83)90015-1
Gojobori T Nei M Concerted evolution of the immunoglobulin VH gene family Mol Biol Evol 1984 1 195 212 6443795
Hughes AL Green JA Garbayo JM Roberts RM Adaptive diversification within a large family of recently duplicated, placentally expressed genes Proc Natl Acad Sci U S A 2000 97 3319 3323 10725351 10.1073/pnas.050002797
Lin TM Halbert SP Spellacy WN Measurement of pregnancy-associated plasma proteins during human gestation Journal of Clinical Investigation 1974 54 576 582 4853116
McLellan AS Fischer B Dveksler G Hori T Wynne F Ball M Okumura K Moore T Zimmermann W Structure and evolution of the mouse pregnancy-specific glycoprotein (Psg) gene locus BMC Genomics 2005 6
Brummendorf T Rathjen FG Cell adhesion molecules. 1: immunoglobulin superfamily Protein Profile 1994 1 951 1058 8528906
Teglund S Zhou GQ Hammarstrom S Characterization of cDNA encoding novel pregnancy-specific glycoprotein variants Biochemical & Biophysical Research Communications 1995 211 656 664 7794280 10.1006/bbrc.1995.1862
Rudert F Saunders AM Rebstock S Thompson JA Zimmermann W Characterization of murine carcinoembryonic antigen gene family members Mammalian Genome 1992 3 262 273 1638085 10.1007/BF00292154
Rudert F Zimmermann W Thompson JA Intra- and interspecies analyses of the carcinoembryonic antigen (CEA) gene family reveal independent evolution in primates and rodents Journal of Molecular Evolution 1989 29 126 134 2509715
Ball M McLellan A Collins B Coadwell J Stewart F Moore T An abundant placental transcript containing an IAP-LTR is allelic to mouse pregnancy-specific glycoprotein 23 (Psg23): cloning and genetic analysis Gene 2004 325 103 113 14697515 10.1016/j.gene.2003.10.001
Zimmermann W Stanners CP The nature and expression of the rodent CEA families: evolutionary considerations Cell adhesion and communication mediated by the CEA family 1998 Amsterdam, Harwood Academic Publishers 31 55
Mills W Moore T Polyandry, life-history trade-offs and the evolution of imprinting at mendelian Loci Genetics 2004 168 2317 2327 15611195 10.1534/genetics.104.030098
Haig D Genomic imprinting, human chorionic gonadotropin, and triploidy Prenat Diagn 1993 13 151 8464837
Rooney BC Horne CH Hardman N Molecular cloning of a cDNA for human pregnancy-specific beta 1-glycoprotein:homology with human carcinoembryonic antigen and related proteins Gene 1988 71 439 449 3265688 10.1016/0378-1119(88)90061-3
Ruoslahti E Pierschbacher MD New perspectives in cell adhesion: RGD and integrins Science 1987 238 491 497 2821619
Rutherfurd KJ Chou JY Mansfield BC A motif in PSG11s mediates binding to a receptor on the surface of the promonocyte cell line THP-1 Molecular Endocrinology 1995 9 1297 1305 8544838 10.1210/me.9.10.1297
Yang Z Nielsen R Goldman N Pedersen AM Codon-substitution models for heterogeneous selection pressure at amino acid sites Genetics 2000 155 431 449 10790415
Hurst LD Smith NG Do essential genes evolve slowly? Curr Biol 1999 9 747 750 10421576 10.1016/S0960-9822(99)80334-0
Beauchemin N Draber P Dveksler G Gold P Gray-Owen S Grunert F Hammarstrom S Holmes KV Karlsson A Kuroki M Lin SH Lucka L Najjar SM Neumaier M Obrink B Shively JE Skubitz KM Stanners CP Thomas P Thompson JA Virji M von Kleist S Wagener C Watt S Zimmermann W Redefined nomenclature for members of the carcinoembryonic antigen family Experimental Cell Research 1999 252 243 249 11501563 10.1006/excr.1999.4610
McLenachan PA Lockhart PJ Faber HR Mansfield BC Evolutionary analysis of the multigene pregnancy-specific beta 1-glycoprotein family: separation of historical and nonhistorical signals J Mol Evol 1996 42 273 280 8919879
Bandelt HJ Dress AW Split decomposition: a new and useful approach to phylogenetic analysis of distance data Mol Phylogenet Evol 1992 1 242 252 1342941 10.1016/1055-7903(92)90021-8
Huson DH SplitsTree: analyzing and visualizing evolutionary data Bioinformatics 1998 14 68 73 9520503 10.1093/bioinformatics/14.1.68
Tan K Zelus BD Meijers R Liu JH Bergelson JM Duke N Zhang R Joachimiak A Holmes KV Wang JH Crystal structure of murine sCEACAM1a[1,4]: a coronavirus receptor in the CEA family Embo J 2002 21 2076 2086 11980704 10.1093/emboj/21.9.2076
Gibbs RA Weinstock GM Metzker ML Muzny DM Sodergren EJ Scherer S Scott G Steffen D Worley KC Burch PE Okwuonu G Hines S Lewis L DeRamo C Delgado O Dugan-Rocha S Miner G Morgan M Hawes A Gill R Celera Holt RA Adams MD Amanatides PG Baden-Tillson H Barnstead M Chin S Evans CA Ferriera S Fosler C Glodek A Gu Z Jennings D Kraft CL Nguyen T Pfannkoch CM Sitter C Sutton GG Venter JC Woodage T Smith D Lee HM Gustafson E Cahill P Kana A Doucette-Stamm L Weinstock K Fechtel K Weiss RB Dunn DM Green ED Blakesley RW Bouffard GG De Jong PJ Osoegawa K Zhu B Marra M Schein J Bosdet I Fjell C Jones S Krzywinski M Mathewson C Siddiqui A Wye N McPherson J Zhao S Fraser CM Shetty J Shatsman S Geer K Chen Y Abramzon S Nierman WC Havlak PH Chen R Durbin KJ Egan A Ren Y Song XZ Li B Liu Y Qin X Cawley S Cooney AJ D'Souza LM Martin K Wu JQ Gonzalez-Garay ML Jackson AR Kalafus KJ McLeod MP Milosavljevic A Virk D Volkov A Wheeler DA Zhang Z Bailey JA Eichler EE Tuzun E Birney E Mongin E Ureta-Vidal A Woodwark C Zdobnov E Bork P Suyama M Torrents D Alexandersson M Trask BJ Young JM Huang H Wang H Xing H Daniels S Gietzen D Schmidt J Stevens K Vitt U Wingrove J Camara F Mar Alba M Abril JF Guigo R Smit A Dubchak I Rubin EM Couronne O Poliakov A Hubner N Ganten D Goesele C Hummel O Kreitler T Lee YA Monti J Schulz H Zimdahl H Himmelbauer H Lehrach H Jacob HJ Bromberg S Gullings-Handley J Jensen-Seaman MI Kwitek AE Lazar J Pasko D Tonellato PJ Twigger S Ponting CP Duarte JM Rice S Goodstadt L Beatson SA Emes RD Winter EE Webber C Brandt P Nyakatura G Adetobi M Chiaromonte F Elnitski L Eswara P Hardison RC Hou M Kolbe D Makova K Miller W Nekrutenko A Riemer C Schwartz S Taylor J Yang S Zhang Y Lindpaintner K Andrews TD Caccamo M Clamp M Clarke L Curwen V Durbin R Eyras E Searle SM Cooper GM Batzoglou S Brudno M Sidow A Stone EA Payseur BA Bourque G Lopez-Otin C Puente XS Chakrabarti K Chatterji S Dewey C Pachter L Bray N Yap VB Caspi A Tesler G Pevzner PA Haussler D Roskin KM Baertsch R Clawson H Furey TS Hinrichs AS Karolchik D Kent WJ Rosenbloom KR Trumbower H Weirauch M Cooper DN Stenson PD Ma B Brent M Arumugam M Shteynberg D Copley RR Taylor MS Riethman H Mudunuri U Peterson J Guyer M Felsenfeld A Old S Mockrin S Collins F Genome sequence of the Brown Norway rat yields insights into mammalian evolution Nature 2004 428 493 521 15057822 10.1038/nature02426
Ha CT Waterhouse R Wessells J Wu JA Dveksler GS Binding of pregnancy-specific glycoprotein 17 to CD9 on macrophages induces secretion of IL-10, IL-6, PGE2, and TGF-{beta}1 J Leukoc Biol 2005
Waterhouse R Ha C Dveksler GS Murine CD9 is the receptor for pregnancy-specific glycoprotein 17 J Exp Med 2002 195 277 282 11805154 10.1084/jem.20011741
Morimura N Tezuka Y Watanabe N Yasuda M Miyatani S Hozumi N Tezuka Ki K Molecular cloning of POEM: a novel adhesion molecule that interacts with alpha8beta1 integrin J Biol Chem 2001 276 42172 42181 11546798 10.1074/jbc.M103216200
Gould RJ Polokoff MA Friedman PA Huang TF Holt JC Cook JJ Niewiarowski S Disintegrins: a family of integrin inhibitory proteins from viper venoms Proc Soc Exp Biol Med 1990 195 168 171 2236100
McLane MA Marcinkiewicz C Vijay-Kumar S Wierzbicka-Patynowski I Niewiarowski S Viper venom disintegrins and related molecules Proc Soc Exp Biol Med 1998 219 109 119 9790167
Chan CC Lao TT Cheung AN Apoptotic and proliferative activities in first trimester placentae Placenta 1999 20 223 227 10195745 10.1053/plac.1998.0375
ActiveState - Dynamic Tools for Dynamic Languages
Blast The Rat Genome
Ensembl BlastSearch (BlastView)
Ensembl FTP rat FASTA data
Thompson JD Gibson TJ Plewniak F Jeanmougin F Higgins DG The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools Nucleic Acids Res 1997 25 4876 4882 9396791 10.1093/nar/25.24.4876
PROSITE Pattern Format Example
McLellan AS Langlands K Kealey T Exhaustive identification of human class II basic helix-loop-helix proteins by virtual library screening Mech Dev 2002 119 Suppl 1 S285 91 14516699 10.1016/S0925-4773(03)00130-8
Kumar S Tamura K Jakobsen IB Nei M MEGA2: molecular evolutionary genetics analysis software Bioinformatics 2001 17 1244 1245 11751241 10.1093/bioinformatics/17.12.1244
Entrez Nucleotide
Zhou GQ Hammarstrom S Pregnancy-specific glycoprotein (PSG) in baboon (Papio hamadryas): family size, domain structure, and prediction of a functional region in primate PSGs Biol Reprod 2001 64 90 99 11133662
NPS@ : CLUSTALW multiple alignment
Jukes TH Cantor CR Munro HN Evolution of protein molecules Mammalian Protein Metabolism 1969 New York, Academic Press 21 132
Yang Z Nielsen R Estimating synonymous and nonsynonymous substitution rates under realistic evolutionary models Mol Biol Evol 2000 17 32 43 10666704
Yang Z PAML: a program package for phylogenetic analysis by maximum likelihood Comput Appl Biosci 1997 13 555 556 9367129
Comegys MM Lin SH Rand D Britt D Flanagan D Callanan H Brilliant K Hixson DC Two variable regions in carcinoembryonic antigen-related cell adhesion molecule1 N-terminal domains located in or next to monoclonal antibody and adhesion epitopes show evidence of recombination in rat but not in human J Biol Chem 2004 279 35063 35078 15184366 10.1074/jbc.M404431200
Markel G Gruda R Achdout H Katz G Nechama M Blumberg RS Kammerer R Zimmermann W Mandelboim O The critical role of residues 43R and 44Q of carcinoembryonic antigen cell adhesion molecules-1 in the protection from killing by human NK cells J Immunol 2004 173 3732 3739 15356119
Watt SM Teixeira AM Zhou GQ Doyonnas R Zhang Y Grunert F Blumberg RS Kuroki M Skubitz KM Bates PA Homophilic adhesion of human CEACAM1 involves N-terminal domain interactions: structural analysis of the binding site Blood 2001 98 1469 1479 11520797 10.1182/blood.V98.5.1469
Bos MP Hogan D Belland RJ Homologue scanning mutagenesis reveals CD66 receptor residues required for neisserial Opa protein binding J Exp Med 1999 190 331 340 10430622 10.1084/jem.190.3.331
Virji M Evans D Hadfield A Grunert F Teixeira AM Watt SM Critical determinants of host receptor targeting by Neisseria meningitidis and Neisseria gonorrhoeae: identification of Opa adhesiotopes on the N-domain of CD66 molecules Mol Microbiol 1999 34 538 551 10564495 10.1046/j.1365-2958.1999.01620.x
Rao PV Kumari S Gallagher TM Identification of a contiguous 6-residue determinant in the MHV receptor that controls the level of virion binding to cells Virology 1997 229 336 348 9126247 10.1006/viro.1997.8446
Wessner DR Shick PC Lu JH Cardellichio CB Gagneten SE Beauchemin N Holmes KV Dveksler GS Mutational analysis of the virus and monoclonal antibody binding sites in MHVR, the cellular receptor of the murine coronavirus mouse hepatitis virus strain A59 J Virol 1998 72 1941 1948 9499047
Taheri M Saragovi U Fuks A Makkerh J Mort J Stanners CP Self recognition in the Ig superfamily. Identification of precise subdomains in carcinoembryonic antigen required for intercellular adhesion J Biol Chem 2000 275 26935 26943 10864933
|
15987510
|
PMC1185527
|
CC BY
|
2021-01-04 16:29:16
|
no
|
BMC Evol Biol. 2005 Jun 29; 5:39
|
utf-8
|
BMC Evol Biol
| 2,005 |
10.1186/1471-2148-5-39
|
oa_comm
|
==== Front
BMC EcolBMC Ecology1472-6785BioMed Central London 1472-6785-5-51598750710.1186/1472-6785-5-5Research ArticleGenetic diversity in populations of asexual and sexual bag worm moths (Lepidoptera: Psychidae) Grapputo Alessandro [email protected] Tomi [email protected] Johanna [email protected] Silja [email protected] Department of Biological and Environmental Science, University of Jyväskylä, P.O. Box 35, FI-40014 Jyväskylä, Finland2005 29 6 2005 5 5 5 17 9 2004 29 6 2005 Copyright © 2005 Grapputo et al; licensee BioMed Central Ltd.2005Grapputo 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 the two-fold cost of sex, most of the higher animals reproduce sexually. The advantage of sex has been suggested to be its ability, through recombination, to generate greater genetic diversity than asexuality, thus enhancing adaptation in a changing environment. We studied the genetic diversity and the population structure of three closely related species of bag worm moths: two strictly sexual (Dahlica charlottae and Siederia rupicolella) and one strictly asexual (D. fennicella). These species compete for the same resources and share the same parasitoids.
Results
Allelic richness was comparable between the sexual species but it was higher than in the asexual species. All species showed high heterozygote deficiency and a large variation was observed among FIS values across loci and populations. Large genetic differentiation was observed between populations confirming the poor dispersal ability of these species. The asexual species showed lower genotype diversity than the sexual species. Nevertheless, genotype diversity was high in all asexual populations.
Conclusion
The three different species show a similar population structure characterised by high genetic differentiation among populations and low dispersal. Most of the populations showed high heterozygote deficiency likely due to the presence of null alleles at most of the loci and/or to the Wahlund effect. Although the parthenogenetic D. fennicella shows reduced genetic diversity compared to the sexual species, it still shows surprisingly high genotype diversity. While we can not totally rule out the presence of cryptic sex, would explain this high genotype diversity, we never observed sex in the parthenogenetic D. fennicella, nor was there any other evidence of this. Alternatively, a non-clonal parthenogenetic reproduction, such as automictic thelytoky, could explain the high genotypic diversity observed in D. fennicella.
==== Body
Background
Parthenogenetic females have a two-fold advantage over sexual females because they produce only the fecund sex while sexual females produce also males. The elimination of sex is predicted whenever the two strategies compete unless there are factors that overcome this disadvantage. Nevertheless, most of the higher animals reproduce sexually[1]. This leads to a fundamental question which continues to puzzle evolutionary biologists: how is sex maintained? A large body of theories seek to explain the maintenance of sex [2-7].
Advantages of sexual reproduction arise from genetic recombination in cross-fertilisation, which purges deleterious mutation and increases genetic variability in the population [8-10], enhancing adaptation in a changing environment. The idea that sexual reproduction and recombination may be favoured in changing environments has been the subject of several papers [11-15]. If a trait is subjected to stabilising selection, genetic variability introduces a genetic load as a consequence of the produced phenotypes that deviate from the optimum [16]. However, in a varying environment that exerts directional selection on a trait, genetic variability is essential because the response to selection will be proportional to the additive genetic variance in the population [15]. Under the mutation accumulation theory, the persistence of asexual lineages is more problematic unless asexuals are able to minimise the competition with sexuals through high dispersal rates [17].
Under the Red Queen hypothesis for sex [18], we should expect that heavy directional selection exerted by parasites can favour greater genetic variability in host populations. Parasites are more likely to infect the most common genotypes while rare genotypes, produced by sexual females, may escape infection [19-23]. Asexual reproduction is expected to be an unstable long term strategy since asexual females can only generate offspring with new genotypes through mutation.
The parasite hypothesis relies on several critical assumptions: the all-else-equal assumption and assumptions about the population structure and the genetic diversity of sexual and asexual populations [24]. The all-else-equal assumption (e.g. production of equal number and viability of offspring) depends on: how the asexuals originate, the type of parthenogenesis, and the degree of polyploidy of the asexuals [24,25]. The difference in population genetic structure between competing sexuals and asexuals may determine difference in the parasite infection load. Asexual hosts can persist in the long term, even in the presence of parasites, if they out-disperse their parasites [26,27]. The parasite hypothesis also assumes that sexual populations harbour higher levels of genetic diversity than asexual populations. The parasite hypothesis does not select for sex per se, but for diversity [28]. Thus, high clonal diversity could erode any advantage of sex. Howard and Lively [29] theoretically showed that host-parasite coevolution could lead to the accumulation of clones with different resistance genotypes and, in turn, to the elimination of sexual populations.
Few systems with coexisting sexual and asexual competitors are known. So, comparisons of genetic diversity between coexisting sexual and asexual populations are scarce: e.g. the freshwater snails, Potamopyrgus antipodarum [7,20] and genus Campeloma [17,24] and the aphid Rhopalosiphum padi [30]. Additional comparisons are needed to further evaluate the parasite hypothesis for sex.
Bag worm moths (Lepidoptera: Psychidae) provide an attractive case for investigating the coexistence of sexual and asexual reproduction in the same locations. In Lepidoptera, parthenogenetic reproduction is very rare. However, in the family Psychidae and especially among Dahlica species, parthenogenesis seems to have evolved several times [31,32]. A parthenogenetic (Dahlica fennicella, Suomalainen 1980) and two sexual species (Siederia rupicolella, Sauter 1954 and D. charlottae, Meier 1957) are common in Finland and often coexist in the same habitat. In these small insects (3–6 mm), adult females are always wingless, sessile and incapable of dispersing. Males are always winged but their dispersing ability is very limited and they can only fly short distances (between 10 and 100 m). Life cycle from egg to adult takes from one to two years, but the adults only live 3–6 days [33]. S. rupicolella and D. fennicella are very difficult to separate from each other and the only distinctive characters are their reproductive mode and genetic markers [32,34]. In central Finland, these bag worm moth species occur patchily in wooded habitats. The proportion of sexually and parthenogenetically reproducing species varies between locales from the total absence of the sexual species to only their presence.
Psychid larvae are often infected by at least two common species of Hymenopteran parasitoids, e.g. Orthizema spp. [31,34]. Kumpulainen et al. [34] found a strong positive correlation between parasite prevalence and the occurrence of sexual reproduction in Finnish bag worm moth populations for three consecutive years. S. rupicolella (sexual) was more abundant where parasitoids were more common, whereas D. fennicella (asexual) was more common in localities where parasitoids were scarce or absent. This result could argue in favour of the parasite hypothesis for the maintenance of sex. In light of this previous result, we investigated the genetic variability and the population structure of three closely related species of bag worm moths, two strictly sexual (Siederia rupicolella and Dahlica charlottae) and one parthenogenetic (D. fennicella) using isozyme variation.
Results
Genetic variability
Thirteen loci from ten isozymes were detected (Table 1). All were polymorphic in the two sexual species whereas twelve were polymorphic in the asexual species, with only fumaric acid (FUM) being monomorphic. No more than two bands were observed at all loci in the asexual D. fennicella, thus it is possible that this species is not tetraploid as are its relatives D. lichenella and D. triquetrella (tetraploid race) [35]. Ewens-Watterson [36] and Chakraborty's [37] test of neutrality indicated that the polymorphism observed, at the scale investigated, can confidently be assumed to be neutral.
Table 1 Isoenzymatic loci scored for Siederia rupicolella, Dahlica charlottae and D. fennicella. Recipes for buffers used are found at
Enzyme E.C.* Locus Buffer
- Isocitrate Dehydrogenase 1.1.1.42 IDH Phosphate 0.02 M pH 7.0
- Diaphorase 1.1.1.40 DIA Phosphate 0.02 M pH 7.0
- Glucose-6-Phosphate Dehydrogenase 1.1.1.49 G6PDH Tris-Maleate-EDTA- MgCl2 0.2 M pH 7.8
- 6-Phosphogluconate Dehydrogenase 1.1.1.44 6PGDH Tris-Maleate-EDTA- MgCl2 0.2 M pH 7.8
- Aspartate Aminotransferase 2.6.1.1 GOT Citrate Phosphate 0.04 M pH 6.4
- Phosphoglucomutase 5.4.2.2 PGM Tris-Maleate-EDTA- MgCl2 0.2 M pH 7.8
- Malate Dehydrogenase NADP+ 1.1.1.40 ME1
ME2
ME3 Tris Maleate 0.1 M pH 5.3
- Malate Dehydrogenase 1.1.1.37 MDH MDH2 Tris-Maleate-EDTA- MgCl2 0.2 M pH 7.8
- Glucose-6-Phosphate Isomerase 5.3.1.9 GPI Tris-Maleate-EDTA- MgCl2 0.2 M pH 7.8
- Fumarate Hydratase 4.2.1.2 FUM Tris Maleate 0.1 M pH 5.3
* Enzyme Commission Number
The estimates of genetic variability are shown in Table 2. Sufficient sampling of all three species in each population for population genetic analyses was not possible. There were two reasons for this; 1) although all three species were present to some extent in each location, they were not all abundant, particularly D. fennicella, and 2) females of the sexual species S. rupicolella are difficult to separate from D. fennicella without observing their mating behaviour. While sexual females secrete pheromones to attract males and do not lay eggs before mating, parthenogenetic females lay eggs in their larval case immediately after hatching from pupa. Species determination for sexual females was performed by experimental mating with a male. Because the adults are very short lived, females can mated with males hatching only few days apart restricting the sample size.
Table 2 Sample sizes, average number of alleles per locus, allelic richness, proportion of different genotypes (k), Simpson's index (D) and Evenness (E) (D and E calculated for asexual D. fennicella only), observed heterozygosity (Ho), gene diversity (Hs), and FIS are presented for each population of each sexual species. In the last column are presented the averaged values per species and the FST values among populations.
Jyväskylä Orimattila
D. charlottae Kö Lv1 Lv2 Lv3 Lv4 Muu Hj Pih Isa Sip1 Sal Tuk Vilj Vill
N 18 11 9 14 10 12 16 14 7 7 14 9 13 11 11.786
N. alleles ± (S.E.) 3.846 (1.039) 3.077 (0.485) 2.692 (0.537) 2.923 (0.611) 3.000 (0.553) 3.154 (0.761) 3.692 (0.888) 3.077 (0.564) 2.154 (0.281) 2.692 (0.511) 3.462 (0.719) 3.000 (0.707) 3.231 (0.649) 3.538 (0.860) 5.6154 (1.4485)
Allelic richness 3.029 2.871 2.543 2.593 2.765 2.790 3.049 2.702 2.119 2.582 2.877 2.804 2.884 3.114 2.766
k 1.000 1.000 0.889 0.929 1.000 1.000 1.000 0.786 1.000 1.000 1.000 1.000 1.000 1.000 0.972
Ho ± (S.E.) 0.297 (0.077) 0.307 (0.093) 0.249 (0.087) 0.295 (0.095) 0.308 (0.081) 0.245 (0.075) 0.266 (0.070) 0.251 (0.070) 0.374 (0.109) 0.393 (0.112) 0.248 (0.063) 0.274 (0.065) 0.284 (0.087) 0.285 (0.076) 0.2955 (0.1046)
Hs ± (S.E.) 0.469 (0.107) 0.548 (0.086) 0.451 (0.102) 0.437 (0.114) 0.508 (0.110) 0.447 (0.107) 0.505 (0.127) 0.502 (0.093) 0.405 (0.087) 0.425 (0.102) 0.524 (0.080) 0.468 (0.121) 0.528 (0.099) 0.558 (0.093) 0.4851 (0.0841)
FST (P) 0.168 (<0.001) 0.154 (<0.001) 0.1530 (<0.001)
S. rupicolella Kv Lv2 Lv4 Lv5 Muu Hj Sip1 Vill Pen
N 10 14 10 9 14 17 26 11 11 13.556
N. alleles ± (S.E.) 2.846 (0.337) 3.000 (0.358) 3.000 (0.340) 2.385 (0.368) 3.231 (0.426) 3.385 (0.549) 3.692 (0.634) 2.538 (0.332) 2.308 (0.208) 2.932 (0.815)
Allelic richness 2.742 2.794 2.783 2.275 2.914 2.988 3.001 2.480 2.198 2.686
k 1.000 0.929 1.000 1.000 0.929 1.000 1.000 0.909 0.909 0.964
Ho ± (S.E.) 0.364 (0.084) 0.327 (0.086) 0.247 (0.063) 0.253 (0.071) 0.312 (0.084) 0.343 (0.092) 0.266 (0.068) 0.303 (0.096) 0.192 (0.073) 0.290 (0.072)
Hs ± (S.E.) 0.534 (0.057) 0.512 (0.072) 0.507 (0.057) 0.380 (0.073) 0.492 (0.062) 0.526 (0.069) 0.501 (0.066) 0.477 (0.058) 0.399 (0.066) 0.481 (0.049)
FST (P) 0.114 (<0.001) 0.116 (<0.001) 0.101 (>0.001)
D. fennicella Hn Pih Isa Sip1 Sip2 Sip3
N 9 13 17 18 21 8 14.333
N. alleles ± (S.E.) 1.923 (0.211) 2.692 (0.365) 2.615 (0.241) 2.385 (0.350) 2.769 (0.411) 1.692 (0.208) 3.615 (0.488)
Allelic richness 1.692 2.123 2.126 1.955 2.125 1.611 1.939
k 1 1 0.824 0.778 0.667 0.500 0.795
D 9 13 9.966 11.571 10.756 3.556 9.642
E 1 1 0.712 0.827 0.964 0.889 0.899
Ho ± (S.E.) 0.376 (0.117) 0.383 (0.115) 0.363 (0.122) 0.443 (0.122) 0.436 (0.133) 0.375 (0.130) 0.397 (0.116)
Hs ± (S.E.) 0.308 (0.069) 0.458 (0.067) 0.448 (0.058) 0.379 (0.0793) 0.466 (0.066) 0.279 (0.076) 0.390 (0.033)
FST (P) 0.213 (<0.001) 0.213 (<0.001)
Allelic richness ranged from 1.61 in D. fennicella to 3.11 in D. charlottae. The expected heterozygosity (Hs) ranged from 0.279 in D. fennicella to 0.558 in D. charlottae. Allele richness and gene diversity (Hs) were similar in the two sexual species (1000 permutation: P = 0.488 and P = 1.00, respectively). Both sexual species harbour significantly higher allele richness and Hs than the asexual species (permutation tests: D. charlottae vs. D. fennicella P = 0.002 and P = 0.007, respectively and S. rupicolella vs. D. fennicella P = 0.012 and P = 0.012, respectively). As expected, the proportion of different genotypes (k) was close to 1 in the sexual populations and no differences were observed between the two species (Mann-Whitney test U14,9 = 50.5, P = 0.439). The proportion of different clones (k) was also high in the asexual species, ranging from 0.5 to 1 in the different populations; however it was significantly lower than in D. charlottae and almost significantly lower than in S. rupicolella (Mann-Whitney test U14,6 = 18, P = 0.020 and U9,6 = 13, P = 0.082). Evenness was very similar among asexual populations and it was very close to 1 because of the high genotype diversity. Significant deviation from the Hardy-Weinberg equilibrium was observed in most of the loci in all three species. In the sexual species, this deviation was due to heterozygote deficiency. The FIS values, over all loci indicated a significant deficiency of heterozygotes in all populations of both sexual species with the exception of Isosaari (Isa) and Sippulanniemi 1 (Sip 1) populations of D. charlottae (Table 3). In D. fennicella, instead, only two populations showed heterozygosity deficiency (Table 3). A large variation in FIS values was observed across all loci and populations in all three species with the exception of MDH2 for which no heterozygote individuals were ever observed (Table 3). The exclusion of this locus, however, did not change any of our results.
Null alleles might cause deviation from H-W proportion. The presence of null alleles at many of the loci analysed is strongly suggested by the significant correlation of the FIS values between the two sexual species (rs = 0.93, n = 12, P < 0.001) and between D. charlottae and D. fennicella (rs = 0.774, n= 11, P = 0.005). Calculation of the frequency of null alleles with the methods of Brookfield [38] indicate that null alleles are present in high frequencies in most of the loci in all three species (Table 3).
Table 3 FIS values and frequency of null alleles for each population of the three species at each locus. Null alleles frequencies (a*) were calculated with the method of Brookfield [38].
DIA FUM G6PDH GOT GPI IDH MDH MDH2 ME1 ME2 ME3 6PGDH PGM All
D. charlottae
Jyväskylä Kö1 FIS -0.278 NA 0.141 0.721 0.619 0.115 0.256 1 0.368 1 1 0.211 0.011 0.367
a* 0 0 0.027 0.257 0.215 0.038 0.092 0.363 0.116 0.172 0.095 0.042 0
Lv1 FIS -0.653 NA -0.297 0.773 0.828 -0.071 0.877 1 0.429 1 1 0.100 0.279 0.441
a* 0 0 0 0.202 0.265 0 0.348 0.410 0.145 0.346 0.284 0.014 0.094
Lv2 FIS -0.556 1 0.276 0.000 0.636 -0.297 0.429 1 1 1 NA -0.143 0.689 0.448
a* 0 0.165 0.069 0 0.130 0 0.133 0.331 0.308 0.417 0 0 0.258
Lv3 FIS 0.034 NA -1 0.840 0.576 -0.243 0.425 1 NA 1 NA 0.407 0.278 0.326
a* 0 0 0 0.243 0.214 0 0.146 0.252 0 0.417 0 0.092 0.097
Lv4 FIS -0.292 NA -0.125 0.385 0.717 -0.021 0.217 1 1 0.509 0 1 0.223 0.393
a* 0 0 0 0.128 0.271 0 0.070 0.383 0.153 0.169 0 0.319 0.067
Muu FIS 0.165 NA -0.100 0.681 0.701 0.079 0.133 1 1 1 1 0.429 0.471 0.481
a* 0.044 0 0 0.215 0.231 0.014 0.038 0.379 0.133 0.217 0.133 0.089 0.186
Hj FIS 0.167 NA -0.189 0.605 0.683 0.025 0.534 1 0.659 1 0 0.556 0.544 0.474
a* 0.045 0 0 0.220 0.285 0 0.221 0.418 0.095 0.105 0 0.173 0.229
Pih FIS -0.241 NA 0.226 0.658 0.506 0.158 0.477 1 0.871 1 1 -0.294 0.577 0.501
a* 0 0 0.049 0.105 0.139 0.052 0.195 0.384 0.294 0.290 0.333 0 0.223
Isa FIS -1 NA 0 0 0.091 -1 0.143 1 -0.200 1 0.750 -1 -0.714 -0.113
a* 0 0 0 0 0.006 0 0.021 0.2899 0 0.1967 0.238 0 0
Sip1 FIS 0.615 NA -0.091 -0.333 -0.161 -0.333 0.300 1 0.647 NA 1 -0.667 -0.448 0.076
a* 0.152 0 0 0 0 0 0.059 0.3725 0.158 0 0.197 0 0
Orimattila Sal FIS 0.393 -0.040 0.189 0.589 0.482 -0.026 0.816 1 0.662 0.769 1 0.154 0.564 0.527
a* 0.121 0 0.047 0.223 0.179 0 0.338 0.351 0.191 0.178 0.332 0.030 0.133
Tuk FIS -0.333 NA -0.200 0.632 0.575 0.127 0.744 1 -0.067 1 NA -0.091 0.571 0.415
a* 0 0 0 0.211 0.223 0.029 0.315 0.331 0 0.165 0 0 0.220
Vilj FIS 0.208 NA 0.294 0.222 0.177 -0.200 0.866 1 1 1 1 0.353 0.392 0.463
a* 0.066 0 0.061 0.073 0.048 0 0.311 0.362 0.327 0.321 0.124 0.103 0.155
Vill FIS -0.135 1 0.438 0.615 0.870 0.067 0.512 1 0 1 1 0.043 0.681 0.489
a* 0 0.142 0.130 0.232 0.335 0.009 0.195 0.284 0 0.142 0.332 0 0.289
S. rupicolella
Jyväskylä Kv FIS -0.800 NA 0.250 0.554 0.419 -0.268 0.534 1 0.442 0.852 0.542 0.351 -0.184 0.318
a* 0 0 0.067 0.198 0.091 0 0.185 0.296 0.159 0.319 0.190 0.096 0
Lv2 FIS -0.307 NA 0.559 0.458 0.780 -0.279 -0.078 1 0.579 0.882 NA 0.100 0.242 0.361
a* 0 0 0.192 0.165 0.178 0 0 0.413 0.216 0.330 0 0.014 0.080
Lv4 FIS -0.091 0 0.617 0.664 1 0.338 0.820 1 0.471 1 0 -0.207 -0.154 0.513
a* 0 0 0.189 0.223 0.390 0.120 0.269 0.398 0.148 0.296 0 0 0
Lv5 FIS 0.067 NA 0.368 -0.091 0.573 -0.085 -0.067 1 0.818 NA NA 0.478 0.190 0.333
a* 0 0 0.087 0 0.172 0 0 0.257 0.283 0 0 0.137 0.045
Muu FIS 0.007 NA -0.132 0.480 0.623 -0.226 0.182 1 0.196 1 1 -0.176 0.507 0.367
a* 0 0 0 0.178 0.158 0 0.063 0.364 0.033 0.333 0.290 0 0.170
Hj FIS -0.333 1 0.364 0.202 0.615 -0.200 0.813 1 0.509 0.652 NA 0.336 -0.049 0.347
a* 0 0.172 0.111 0.072 0.251 0 0.299 0.386 0.177 0.088 0 0.096 0
Sip1 FIS -0.090 NA -0.154 0.301 0.788 0.001 0.708 1 0.091 1 0.928 -0.091 0.358 0.468
a* 0 0 0 0.085 0.321 0 0.303 0.393 0.018 0.316 0.316 0 0.132
Orimattila Vill FIS -0.800 NA -0.047 1 1 -0.125 0.612 1 -0.296 1 1 -0.2000 0.474 0.366
a* 0 0 0 0.339 0.316 0 0.240 0.316 0 0.153 0.316 0 0.142
Pen FIS 0.438 NA 0.217 0.091 1 -0.524 0.680 1 0 1 1 NA -0.026 0.518
a* 0.130 0 0.054 0.014 0.332 0 0.225 0.373 0 0.301 0.332 0 0
D. fennicella
Jyväskylä Hn FIS 0.304 NA -1 0 -0.600 -0.091 0.448 1 NA NA NA -1 -1 -0.222
a* 0.093 0 0 0 0 0 0.108 0.331 0 0 0 NA 0
Pih FIS 0.268 NA -1 1 0.200 0.050 0.553 1 0.351 1 1 -1 -0.581 0.145
a* 0.110 0 0 0.321 0.000 0.058 0.301 0.381 0.256 0.299 0 0 0
Isa FIS 0.657 NA -1 1 -0.333 0.185 0.780 1 0.840 1 NA -0.800 -1 0.208
a* 0.058 0 0 0.333 0.067 0.008 0.205 0.279 0.067 0.233 0.172 0 0
Sip1 FIS -0.732 NA -1 1 -0.308 -0.244 0.532 1 NA -0.097 NA -0.846 -1 -0.168
a* 0 0 0 0.095 0 0 0.207 0.436 0 0 0 0 0
Sip2 FIS 0.650 NA -1 0.840 -0.772 -0.159 0.162 1 1 1 NA -1 -1 0.057
a* 0.218 0 0 0.185 0 0 0.052 0.393 0.329 0.329 0 0 0
Sip3 FIS 0.391 NA -1 NA NA -1 -0.029 1 NA NA NA -1 -1 -0.345
a* 0.091 0 0 0 0 0 0 0.347 0 0 0 0 0
Population differentiation
The overall differentiation among populations, FST, was high and significantly different from zero in all three species (Table 2), which is an indication of strong population structure. Mean FST values across loci and the 95% confidence interval (bootstrap over loci) are shown in Figure 2. The overall differentiation was significantly different in the three species. While the mean FST value of D. fennicella was not different from that of D. charlottae, it was significantly higher than that of S. rupicolella. Hierarchical analyses of molecular variance (AMOVA) indicated that the two areas, Jyväskylä and Orimattila, did not differ from each other for both sexual species. In D. charlottae the percentage of variance among sites was 1.61, P = 0.12, while it was 16.07, P < 0.001 among populations within sites. In S. rupicolella the percentage of variance among sites was 0.85, P = 0.27, while it was 11.45, P < 0.001 among populations within sites. We did not observe isolation by distance (Mantel test between FST/(1-FST) and the natural logarithm of the geographical distance) in any species at either of the sites: D. charlottae (Jyväskylä R = -0.257, P = 0.089; Orimattila R = 0.353, P = 0.512), S. rupicolella (R = 0.222, P = 0.338) and D. fennicella (R = - 0.300, P = 0.277).
Figure 2 Mean FST of each species and its 95% confidence interval obtained with bootstrap over loci. Asterisk indicates 5% significant level.
Sixty five different genotypes were detected among the 86 samples of D. fennicella. Only two of them were shared among different populations. One genotype was shared between Sippulanniemi 1 (Sip1) and Huviniemi (Hn) while the other was shared between Sippulanniemi 3 (Sip3) and Pihta (Pih). In both cases the two populations are geographically distant, and three populations in the same area (Sip1, 2 and 3 and Hn, Pih and Isa, respectively) did not share any genotypes.
Discussion
Analysis of isozyme variation in three species of psychid moths revealed that the genetic diversity of the sexual species D. charlottae and S. rupicolella is higher than that of the asexual species D. fennicella. Allele richness, gene and genotypic diversity were also higher in the sexual species than in the asexual species. Higher genotype diversity in sexual than in asexual populations is most likely the logical result of recombination and was an expected result. The sexual populations also showed higher allele diversity, which is a more intriguing outcome [30]. One possible explanation for this difference is that the asexual lineage retained only a portion of the diversity of its sexual ancestor. Alternatively, a lower per locus diversity in the asexual species could reflect lower population sizes compared to the sexual species. Asexual D. fennicella is, in general, rarer than the sexual species, although it was the most abundant species in some locations. A lower population size is also suggested by the higher differentiation (FST) among D. fennicella populations than among sexual populations.
Surprisingly, parthenogenetic D. fennicella showed a considerable amount of genotype diversity, with 65 different genotypes detected among 86 individuals. This was in contrast with a previous analysis with allozyme markers which found limited diversity among samples of D. fennicella [39]. This amount of genotype diversity was higher than that recently observed in Potamopyrgus snails [20] and that reported in animals (reviewed in[40]) and in plants (reviewed in [41]) in the previous allozymes literature. Interestingly, clonal lineages of D. fennicella were mostly restricted to single populations. Only two genotypes were shared among distant populations. The lack of a common broadly adapted haplotype spread over different populations is in conflict with the hypothesis of the general-purpose-genotype [42]. Instead, adaptation to different microclimates or other specific environmental conditions of these locales could explain the presence of many different genotypes, as suggested by Vrijenhoek's [43] frozen niche variation hypothesis. However, we found no significant differences in morphology, size and life-history characters between two different D. fennicella populations that would reflect ecological specialisation [34]. Although several studies have reported allozymes as not neutral (reviewed in [44,45]), in our study there were no indications that they deviate from neutrality, thus these markers are expected to be subjected more to drift than to selection. High genotypic diversity could indicate the presence of cryptic sex in the parthenogenetic species. Although we cannot completely rule out this hypothesis, we never observed sex in the species. All parthenogenetic females lay eggs immediately after hatching from pupa and never show the characteristic behaviour of sexual females when they secrete pheromones to attract potential mates (Kumpulainen et al. 2004). Moreover, mitochondrial sequences from sexual and asexual females clearly indicate these are two different species (Grapputo et al. 2005). This high genotypic diversity could also be explained by alternative types of parthenogenesis involving recombination, such as the automictic thelytoky [46].
High clonal diversity and the observed distribution of different clones could be the result of a restricted dispersal capacity and the fragmentation of suitable habitats for these psychid moths. Large differentiation was also observed among populations of diploid parthenogenetic D. triquetrella in the Alps but not among tetraploid populations of the same species in Finland [47]. The same pattern, however, could be explained by an extinction-colonisation process associated with a long persistence of the populations, which would explain the high intrapopulation diversity. Large genetic differentiation among populations was also observed in both the sexual species, D. charlottae and S. rupicolella, which is consistent with their extremely low ability for active dispersal (see also [31]) and the patchy distribution of suitable habitats. Nevertheless, psychid moths sometimes colonise new areas as suggested by the absence of D. charlottae in the Isosaari population in 1999 and its presence in 2000 (T. Kumpulainen, personal observation). Most probably, dispersal between different populations is a relatively rare event taking place as passive aerial dispersal of very small larvae [31]. The large genetic differentiation among D. charlottae and S. rupicolella populations is in contrast with the data obtained for populations of sexual D. triquetrella in the Alps by Lokki et al[47], where allelic frequencies were described as homogeneous among populations, although rigorous tests of population differentiation were not carried out.
The observed proportion of heterozygotes was not different between the two sexual species D. charlottae and S. rupicolella (0.29) and was very similar to that previously observed in another sexual species D. triquetrella (0.23) [47]. The level of heterozygosity was also highly similar among populations in both sexual species. D. charlottae and S. rupicolella, in contrast to D. triquetrella, were not in HW equilibrium for most of the loci and populations. Heterozygote deficiency has been widely reported in allozyme surveys of natural populations of marine invertebrates (reviewed in [48,49]) and also in fishes (e.g. [50,51]), amphibians and reptiles (reviewed in [52]). Alternative hypotheses have been advanced to explain such heterozygote deficiencies [48,49,53]. The high heterozygosity deficiency in all three species of bag worm moths could be explained by null alleles. The high variation across loci in FIS values correlate among species and the methods of Brookfield [38] for the calculation of null alleles frequencies strongly suggest that most of the loci in the three species are affected by null alleles. Most populations of sexual psychid moths are small, consisting of just 30 to 100 individuals. Suitable forest patches are also small and isolated. Moreover, females are apterous and unable to disperse. When sexual females emerge from pupae they quickly start to secrete pheromones to attract males. Once emerged, males respond promptly to the female pheromones because they have a very short adult life span (about 10 hours). Therefore, copulation most likely occurs between emerging adults that are both spatially and temporally close. This could create substructured populations and a Wahlund effect, both spatial and temporal, which could maintain a high number of alleles in the population but increase the homozygosity [54].
Conclusion
In summary, the three different moth species show a similar population structure characterised by high genetic differentiation among populations and low dispersal. The parthenogenetic D. fennicella shows reduced genetic diversity compared to the sexual species but still shows high genotype diversity that could indicate the presence of cryptic sex. All species show a very high heterozygote deficiency due to the presence of null alleles at most of the loci or to the Wahlund effect. DNA markers certainly need to be investigated to determine the causes of such heterozygote deficiency shown by the allozymes.
Methods
Source populations
Two sexual species, Siederia rupicolella and Dahlica charlottae, and an asexual species, D. fennicella were sampled to study their genetic variability and population structure. Samples were collected in April 2000 from 20 different study areas of suitable forest type [34]. All areas were situated in central Finland, 15 of them around the city of Jyväskylä (62 °15 N', 25°43 E') and five close to the town of Orimattila (60 °49 N', 25 °40 E') (Figure 1). Study areas consisted of old forest patches, separated by meadows and fields and sometimes by human settlements. All study areas were dominated by mixed forests of Norwegian spruce (Picea abies) and silver birch (Betula pendula), many of them also contained Scotch pine (Pinus sylvestris). Final instar larvae of all three moth species climb on tree trunks to pupate and they can easily be caught by setting tape traps on tree trunks. Larvae remain stuck on the tape and they can later be collected. Each collected larva was taken to the laboratory and kept individually until hatching to adult, allowing us to determine the reproduction mode and identify the species [34]. Samples were subsequently frozen at -80°C until analysis.
Figure 1 Map of the sampling sites in central Finland.
Electrophoresis
Frozen samples were squashed in 20 μl of grinding buffer (Tris-HCl 0.1 M pH = 8.0) and then applied to Titan® III cellulose acetate plates (76 mm × 76 mm) using the Super Z-12 applicator Kit (Helena laboratories) following the method of Hebert and Beaton [55]. Electrophoresis was carried out at room temperature at 200 volts for 20–25 minutes in the appropriate buffer for each enzyme as indicated in Table 1. Of twenty-three enzymes tested, ten were polymorphic (listed in Table 1) from which a total of thirteen loci could be scored. Enzymes excluded from the analysis because they were monomorphic or unreadable were: ACON (EC 4.2.1.3), AK (EC 2.7.4.3), ADH (EC 1.1.1.1), ALP (EC 3.1.3.1), ATT (EC 2.6.1.1), EST (EC 3.1.1.1), HEX (EC 2.7.1.1), LDH (EC 1.1.1.27), LAP (EC 3.4.11.1), MPI (EC 5.3.1.8), SOD (EC 1.15.1.1), α, α-Trehalase (EC 3.2.1.28) and SKDH (EC 1.1.1.25).
Data analysis
Tests of neutrality for each locus, population and species were carried out using the Ewens-Watterson test [36] with the software package Popgene [56]. The genetic diversity and population structure of each species were analysed using Fstat [57]. We tested the Hardy-Weinberg equilibrium (HW) for each locus and population by randomisation of alleles among individuals within populations. Significance levels were adjusted using the sequential Bonferroni correction for multiple comparisons [58]. For each population we estimated the number of alleles per locus, allelic richness [59], gene diversity (Hs) [60], observed heterozygosity (Ho) and FIS value. Frequency of null alleles per locus and population was estimated with the method of Brookfield [38] as implemented in Micro-Checker v.2.2.3 [61], which does not require detecting null allele homozygotes. Genotypic diversity (or clonal diversity in asexuals) within populations was determined simply as the proportion of different genotypes in the population k = G/N, where G is number of genotypes and N is the number of individuals in the population. For the asexual species, we also measured clonal diversity using Simpson's diversity index D = 1/∑pi, where pi is the frequency of the i-th clone (Simpson, 1949). D varies from 1 (monoclonal population) to N if each individual carries a different genotype. This measure takes into account the frequency of clones, but it depends on the sample size, so we also calculated the evenness (E) of Simpson's index E = D/Dmax, which is constrained between 0 and 1. Population structure was assessed by calculating FST [62] between populations and tested by permuting genotypes among samples because most of the populations were not in HW (as suggested in Fstat). Hierarchical analysis of molecular variance (AMOVA, [63]) including all populations and populations within the two areas (Jyväskylä and Orimattila) was performed with Arlequin ver. 2.000 [64]. If the differentiation between populations is due to isolation by distance, a positive correlation between genetic distance and geographical distance is expected. Isolation by distance was tested as suggested by Rousset [65] and a Mantel test was performed between populations in each site using Fstat.
List of abbreviations
ACON = Aconitate Hydratase, AK = Adenylate Kinase, ADH = Alcohol Dehydrogenase, ALP = Alkaline Phosphatase, AAT = Amino Aspartate Transferase, EST = Carboxylesterase, HEX = Hexokinase, LDH = Lactate Dehydrogenase, LAP = Leucine Aminopeptidase, MPI = Mannose-6-Phosphate Isomerase, SOD = Superoxide Dismutase and SKDH = shikimate dehydrogenase.
Authors' contributions
TK collected samples and performed most of the laboratory procedure with SP. AG and JM performed the analysis of the data and wrote the manuscript. All the authors contributed to the study.
Acknowledgements
We want to acknowledge K. Kulmala for assisting in the field work and in the insect laboratory. We thank M. Hietanen and M. Myllylä for their assistance in the genetic laboratory, P. Halme, V. Heino, T. Hiltunen, I. Kananen and A. Kolehmainen, for assisting in the field work, K. E. Knott and A. Veijanen for help in scoring allozyme gels. Finally, we acknowledge K. E. Knott, M. Björklund, P. Mutikainen and four anonymous reviewers for their valuable comments on previous versions of the manuscript. This study was carried out at the University of Jyväskylä and was financially supported by the Academy of Finland, project number 779874, and by the Centre of Excellence in Evolutionary Ecology.
==== Refs
West-Eberhard MJ The maintenance of sex as a developmental trap due to sexual selection Q Rev Biol 2005 80 47 54 15884735 10.1086/431024
Williams GC Sex and evolution 1975 Princeton (NJ), Princeton Univ. Press
Maynard Smith J The evolution of sex 1978 Cambridge, Cambridge Univ. Press
Bell G The masterpiece of nature: the evolution and genetics of sexuality 1982 Berkeley, Univ. of California Press
Kondrashov AS Classification of hypotheses on the advantage of amphimixis J Hered 1993 84 372 387 8409359
Burt A Sex, recombination, and the efficacy of selection - Was Weismann right? Evolution 2000 54 337 351 10937212
Jokela J Lively CM Dybdahl MF Fox JA Genetic variation in sexual and clonal lineages of a freshwater snail Biol J Linn Soc 2003 79 165 181 10.1046/j.1095-8312.2003.00181.x
Muller HJ The relation of recombination to mutational advance Mutat Res 1964 1 2 9 14195748
Kondrashov AS Deleterious mutations and the evolution of sexual reproduction Nature 1988 336 435 440 3057385 10.1038/336435a0
Rice WR Experimental tests of the adaptive significance of sexual recombination Nat Rev Genet 2002 3 241 251 11967549 10.1038/nrg760
Maynard Smith J Selection for recombination in a polygenic model - the mechanism Genet Res 1988 51 59 63 3366381
Crow JF An advantage of sexual reproduction in a rapidly changing environment J Hered 1992 83 169 173 1624761
Charlesworth B The evolution of sex and recombination in a varying environment J Hered 1993 84 345 350 8409356
Kondrashov AS Yampolsky LY Evolution of amphimixis and recombination under fluctuating selection in one and many traits Genet Res 1996 68 165 173
Bürger R Evolution of genetic variability and the advantage of sex and recombination in changing environments Genetics 1999 153 1055 1069 10511578
Mather K Polygenic inheritance and natural selection Biol Rev 1943 18 32 64
Johnson SG Population structure, parasitism, and the survivorship of sexual and autoploid parthenogenetic Campeloma limum Evolution 2000 54 167 175 10937193
Van Valen L A new evolutionary law Evol Theor 1973 1 1 30
Hamilton WD Sex versus non-sex versus parasite Oikos 1980 35 282 290
Fox JA Dybdahl MF Jokela J Lively CM Genetic structure of coexisting sexual and clonal subpopulations in a freshwater snail (Potamopyrgus antipodarum) Evolution 1996 50 1541 1548
Howard RS Lively CM The maintenance of sex by parasitism and mutation accumulation under epistatic fitness functions Evolution 1998 52 604 610
Ooi K Yahara T Genetic variation of geminiviruses: comparison between sexual and asexual host plant populations Mol Ecol 1999 8 89 97 10.1046/j.1365-294X.1999.00537.x
Lively CM Dybdahl MF Parasite adaptation to locally common host genotypes Nature 2000 405 679 681 10864323 10.1038/35015069
Johnson SG Leefe WR Clonal diversity and polyphyletic origins of hybrid and spontaneous parthenogenetic Campeloma (Gastropoda: Viviparidae) from the south-eastern United States J Evol Biol 1999 12 1056 1068 10.1046/j.1420-9101.1999.00099.x
Vrijenhoek RC Factors affecting clonal diversity and coexistence Am Zool 1979 19 787 797
Ladle RJ Johnstone RA Judson OP Coevolutionary dynamics of sex in a metapopulation: escaping the Red Queen P Roy Soc Lond B Bio 1993 253 155 160
Judson OP Preserving genes: a model of the maintenance of genetic variation in a metapopulation under frequency-dependent selection Genet Res 1995 65 175 191
Lively CM Howard RS Selection by parasites for clonal diversity and mixed mating Philos T Roy Soc B 1994 346 271 281
Howard RS Lively CM Parasitism, mutation accumulation and the maintenance of sex Nature 1994 367 554 557 8107824 10.1038/367554a0
Delmotte F Leterme N Gauthier JP Rispe C Simon JC Genetic architecture of sexual and asexual populations of the aphid Rhopalosiphum padi based on allozyme and microsatellite markers Mol Ecol 2002 11 711 723 11972759 10.1046/j.1365-294X.2002.01478.x
Hättenschwiler P Die Sackträger der Schweiz (Lepidoptera, Psychidae) Schmetterlinge und ihre Lebensräume Arten - Gefährdung - Schutz Band 2 1997 Basel, Switzerland, Pro Natura 165 308
Kumpulainen T Särkkä J, Olsbo P and Tynkkynen ML The evolution and maintenance of reproductive strategies in bag worm moths (Lepidoptera: Psychidae) Jyväskylä Studies in Biological and Environmental Science 2004 Jyväskylä, University of Jyväskylä 0 42
Suomalainen E The Solenobiinae species of Finland (Lepidoptera: Psychidae), with a description of a new species Ent Scand 1980 11 458 466
Kumpulainen T Grapputo A Mappes J Parasites and sexual reproduction in psychid moths Evolution 2004 58 1511 1520 15341153
Narbel M La Cytologie de la Parthénogénèse chez Solenobia s p. (Lepidopteres Psychides) Chromosoma 1950 4 56 90 14812634 10.1007/BF00325767
Watterson G The homozygosity test of neutrality Genetics 1978 88 405 417 17248803
Chakraborty R Mitochondrial DNA polymorphism reveals hidden heterogeneity within some Asian populations Am J Hum Genet 1990 47 87 94 2349953
Brookfield JFY A simple new method for estimating null allele frequency from heterozygote deficiency Mol Ecol 1996 5 453 455 8688964 10.1046/j.1365-294X.1996.00098.x
Suomalainen E Lokki J Saura A Genetic polymorphism and evolution in parthenogenetic animals. X. Solenobia species (Lepidoptera: Psychidae) Hereditas 1981 95 31 35
Parker ED Ecological implications of clonal diversity in parthenogenetic morphospecies Am Zool 1979 19 753 762
Ellstrand NC Roose ML Patterns of genotypic diversity in clonal plant species Am J Bot 1987 74 123 131
Lynch M Destabilizing hybridization, general-purpose genotypes and geographic parthenogenesis Q Rev Biol 1984 59 257 290 10.1086/413902
Vrijenhoek RC Wöhrmann K and Loeschcke V Ecological differentiation among clones: the frozen niche variation model Population biology and evolution 1984 Berlin, Springer-Verlag 217 231
Gillespie JH The causes of molecular evolution 1991 New York, Oxford University Press
Mitton JB Selection in natural populations 1997 New York, Oxford University Press
Suomalainen E Significance of parthenogenesis in the evolution of insects Annu Rev Entomol 1962 7 349 365 10.1146/annurev.en.07.010162.002025
Lokki J Suomalainen E Saura A Lankinen P Genetic polymorphism and evolution in parthenogenetic animals. II. Diploid and polyploid Solenobia triquetrella (Lepidoptera: Psychidae) Genetics 1975 79 513 525 1126629
Zouros E Foltz DW Possible explanations of heterozygote deficiency in bivalve molluscs Malacologia 1984 25 583 591
Singh SM Green RH Excess of allozyme homozygosity in marine molluscs and its possible biological significance Malacologia 1984 25 569 581
Allegrucci F Fortunato C Sbordoni V Genetic structure and allozyme variation of sea bass (Dicentrarchus labrax and D. punctatus) in the Mediterranean Sea Marine Biology 1997 128 347 358 10.1007/s002270050100
Castric V Bonney F Bernatchez L Landscape structure and hierarchical genetic diversity in the brook charr Salvelinus fontinalis Evolution 2001 55 1016 1028 11430638
Waldman B McKinnon JS Thornhill NW Inbreeding and outbreeding in fishes, amphibians and reptiles The natural history of inbreeding and outbreeding: theoretical and empirical perspectives 1993 Chicago, University of Chicago Press 250 283
Castric V Bernatchez L Belkhir K Bonhomme F Heterozygote deficiencies in small lacustrine populations of brook charr Salvelinus fontinalis Mitchill (Pisces, Salmonidae): a test of alternative hypotheses Heredity 2002 89 27 35 12080367 10.1038/sj.hdy.6800089
Hartl DL Clark AG Principles of population genetics 1997 3 Sunderland, MA, Sinauer Associates
Hebert PDN Beaton MJ Methodologies for allozyme analysis using cellulose acetate electrophoresis 1993 Beaumont, Texas, Helena Laboratories
Yeh FC Yang RC Boyle TJB Ye ZH Mao JX POPGENE, the user-friendly shareware for population genetic analysis 1997 University of Alberta, Canada, Molecular Biology and Biotechnology Centre
Goudet J FSTAT (vers. 1.2): a computer program to calculate F-statistics J Hered 1995 86 485 486
Rice WR Analysing tables of statistical tests Evolution 1989 43 223 225
Petit RJ El Mousadik A Pons O Identifying populations for conservation on the basis of genetic markers Conser Biol 1998 12 844 855 10.1046/j.1523-1739.1998.96489.x
Nei M Molecular Evolutionary Genetics 1987 New York, Columbia University Press
Van Oosterhout C Hutchinson WF Wills DPM Shipley P MICRO-CHECKER: software for identifying and correcting genotyping errors in microsatellite data Mol Ecol Notes 2004 4 535 538 10.1111/j.1471-8286.2004.00684.x
Weir BS Cockerham CC Estimating F-statistics for the analysis of population structure Evolution 1984 38 1358 1370
Excoffier L Smouse PE Quattro JM Analysis of molecular variance inferred from metric distances among DNA haplotypes - Application to human mitochondrial-DNA restriction data Genetics 1992 131 479 491 1644282
Schneider S Roessli D Excoffier L Arlequin ver. 2.000: A software for population genetic data analysis. 2000 University of Geneva, Switzerland, Genetics and Biometry Laboratory
Rousset F Genetic differentiation and estimation of gene flow from F-statistics under isolation by distance Genetics 1997 145 1219 1228 9093870
|
15987507
|
PMC1185528
|
CC BY
|
2021-01-04 16:29:15
|
no
|
BMC Ecol. 2005 Jun 29; 5:5
|
utf-8
|
BMC Ecol
| 2,005 |
10.1186/1472-6785-5-5
|
oa_comm
|
==== Front
BMC Emerg MedBMC Emergency Medicine1471-227XBioMed Central London 1471-227X-5-51602949410.1186/1471-227X-5-5Case ReportQT interval prolongation after sertraline overdose: a case report de Boer Rudolf A [email protected] Dijk Tonnis H [email protected] Nicole D [email protected] Melle Joost P [email protected] Department of Internal Medicine, Intensive Care Unit, Martini Hospital, Groningen, The Netherlands2 Department of Cardiology, University Medical Center Groningen, Groningen, The Netherlands2005 19 7 2005 5 5 5 26 4 2005 19 7 2005 Copyright © 2005 de Boer et al; licensee BioMed Central Ltd.2005de Boer 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
Selective serotonin reuptake inhibitors (SSRIs) are the most common antidepressants used in first-world countries and are generally well tolerated. Specifically, less cardiovascular toxicity has been reported in comparison with tricyclic antidepressants. Here we report QT interval prolongation after an overdose of the SSRI sertraline.
Case presentation
A previously healthy female patient presented with an attempted suicide with overdoses sertraline (2250 mg), diazepam (200 mg), and temazepam (400 mg). Routine laboratory studies were normal and her ECG upon admission showed a normal QT interval. The next day, her ECG showed prolongation of the QTc interval up to 525 ms. After discontinuation of sertraline the QT interval normalized. Echocardiography and exercise electrocardiography were normal. After hospitalization, the patient resumed sertraline in the normally recommended dose and QT interval remained within normal ranges.
Conclusion
It seems that the SSRI sertraline in overdose may cause QT interval prolongation.
==== Body
Background
Since their introduction in 1987, the use of Selective Serotonin Reuptake Inhibitors (SSRIs) has increased dramatically [1]. They clearly have a more favorable safety profile compared to tricyclic antidepressants [2], although prolongation of the QT interval has been reported as a side effect [3]. This is an important side effect since prolongation of the QT interval is strongly associated with life-threatening arrhythmias, most notably torsades de pointes. Although sertraline belongs to the same class of antidepressants, controversy persists whether this holds true for the SSRI sertraline [4]. Here we here present a patient with prolonged QT interval after sertraline overdose.
Case presentation
A 40-year old female patient was referred to our emergency department because of an intended overdose with 200 mg diazepam, 400 mg temazepam, and 2250 mg sertraline.
Her main complaints were fatigue and drowsiness. Blood pressure, pulse rate, and auscultation of the heart and lungs were normal. The patient was treated with sodiumsulfate and charcoal and was admitted to the intensive care unit for continuous control of vital signs. Routine laboratory studies (hematology, chemistry) were normal. Plasma levels of diazepam and temazepam were elevated, 1155 ugr/l (normal: 125 – 750 ugr/l) and 1710 ugr/l (normal: 300–900 ugr/l, respectively). Plasma levels of sertraline and desmethylsertraline were 174 ug/l (normal 20–55 ug/l [5]) and 276 ng/l, respectively.
Her ECG upon admission (upper panel of the figure) shows a sinus rhythm (77 b.p.m.) without conduction disturbances. QT interval in lead V2 was 370 ms. We used the Bazett method (QT time divided by the square root of the RR interval) to calculate the corrected QT (QTc). QTc at admission was 420 ms and negative T-waves were found in leads V1–V3. A second ECG, taken one day after admission (lower panel of the figure), showed a markedly prolonged QT interval with deepened negative T waves in leads V1–V3. QT interval was 520 ms in V2, at a heart rate (HR) of 63 b.p.m (QTc 525 ms). An old ECG (august 2002) showed a sinus rhythm with a HR of 63 b.p.m and a QT interval in lead V2 of 370 ms (QTc 373 ms; ECG not shown).
After 4 days the patient was discharged to a psychiatric hospital because the risk for another suicide attempt was deemed high by the psychiatric consultant. After discharge, the patient underwent further out-patient cardiac evaluation. Echocardiography revealed no structural heart disease. On exercise electrocardiography, patient reached 88% of her maximum HR – no abnormal ST-segment changes were observed. Hereafter, the use of sertraline was resumed in a dose of 50 mg twice daily under guidance of her psychiatrist. Control ECG revealed a normal QT interval (not shown).
Discussion
We here present a patient with prolonged QT interval associated with sertraline overdose. An acquired cause of QT prolongation was suspected since QT intervals had been normal on admission, about 3 hours after ingestion of 2250 mg of sertraline (11 times the maximum maximum recommended dose of 200 mg/day), and were markedly prolonged after one day in hospital. The QT interval normalized after sertraline withdrawal. Therefore, a temporal relation existed between the overdose of sertraline and the development of QT prolongation. However, other causes for QT prolongation, both acquired and inherited, must be considered. For example, combinations of psychoactive drugs have been shown to cause prolongation of the QT interval [6], and our patient ingested temazepam as well as nitrazepam in overdose.
Whereas previous clinical studies [7-10] did not reveal any QT prolongation as a side-effect of sertraline, this case report suggests it may have this potential. We are aware of 1 additional report by Amin et al [11] who described 'a clinically significant' increase in QT interval after treatment with 200 mg of sertraline, however the magnitude of QT prolongation was not specified.
Naturally, implications of this finding are limited because it is only a single case. Two other limitations deserve comment. First, we did not perform a rechallenge with high dosage of sertraline, since this would be unethical. Second, only one blood sample was taken to assess plasma concentration of sertraline – the sertraline plasma level was found clearly increased according to other reports [5,12]. It was therefore not possible to investigate the relation between the course of QT interval prolongation and their paralleled serum levels of sertraline
Conclusion
Our observation suggests that the SSRI sertraline may have the potential to prolong QT interval in rare cases. This case underscores the need for continuous post marketing surveillance.
List of abbreviations
HR heart rate
LV left ventricular
QTc Corrected QT interval
SSRI selective serotonin reuptake inhibitor
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RADB, THVD, and NDH cared for the patient in the intensive care unit, conducted QT analyses, and arranged laboratory samples. RADB and JPVM noticed that QT interval prolongation had not been discussed previously in the case of sertraline overdose. RADB, THVD, NDH wrote the paper, whereas JPVM critically revised the discussion for important intellectual content. All authors read and approved the final manuscript.
Figure 1 ECGs of the patient. ECG of the patient upon admission (upper panel) shows a normal sinus rhythm with a QT interval in lead V2 of 370 ms (QTc 420 ms). There were negative T-waves in leads V1–V3. A second ECG was obtained one day after admission (lower panel) shows a markedly prolonged QT interval of 520 ms in V2 (QTc 525 ms).
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Written consent was obtained from the patient for publication of the patient's details.
==== Refs
Meijer WE Heerdink ER Leufkens HG Herings RMC Egberts ACG Nolen WA Incidence and determinants of long-term use of antidepressants Eur J Clin Pharmacol 2004 60 57 61 14985889 10.1007/s00228-004-0726-3
Kelsey JE Nemeroff CB Sadock BJ, Sadock VA Selective serotonin reuptake inhibitors: introduction and overview Kaplan and Sadock's Comprehensive Textbook of Psychiatry 2000 7 Philadelphia: Lippincott Williams & Wilkins 2432 2435
Al-Khatib SM LaPointe NM Kramer JM Califf RM What clinicians should know about the QT interval JAMA 2003 289 2120 2127 12709470 10.1001/jama.289.16.2120
Gillespie JA Clary CM Medications that prolong the QT interval JAMA 2003 290 1025 (letter) 12941667 10.1001/jama.290.8.1025-b
Sala M Vicentini A Brambilla P Montomoli C Jogia JR Caverzasi E Bonzano A Piccinelli M Barale F De Ferrari GM QT interval prolongation related to psychoactive drug treatment: a comparison of monotherapy versus polytherapy Ann Gen Psychiatry 2005 4 1 15845138 10.1186/1744-859X-4-1
Isbister GK Bowe SJ Dawson A Whyte IM Relative toxicity of selective serotonin reuptake inhibitors (SSRIs) in overdose J Toxicol Clin Toxicol 2004 42 277 285 15362595 10.1081/CLT-120037428
Fisch C Knoebel SB Electrocardiographic findings in sertraline depression trials Drug Invest 1992 4 305 312
Fabre LF Abuzzahab FS Amin M Claghorn JL Mendels J Petrie WM Dube S Small JG Sertraline safety and efficacy in major depression: a double-blind fixed-dose comparison with placebo Biol Psychiatry 1995 38 592 602 8573661 10.1016/0006-3223(95)00178-8
Glassman AH O'Connor CM Califf RM Swedberg K Schwartz P Bigger JT JrKrishnan KR van Zyl LT Swenson JR Finkel MS Landau C Shapiro PA Pepine CJ Mardekian J Harrison WM Barton D Mclvor M Sertraline Antidepressant Heart Attack Randomized Trial (SADHEART) Group Sertraline treatment of major depression in patients with acute MI or unstable angina JAMA 2002 288 701 709 12169073 10.1001/jama.288.6.701
Amin M Lehmann H Mirmiran J A double-blind, placebo-controlled dose-finding study with sertraline Psychopharmacol Bull 1989 25 164 167 2690162
Preskorn SH A tale of two patients J Pract Psych Behav Health 1999 160 164
|
16029494
|
PMC1185529
|
CC BY
|
2021-01-04 16:31:03
|
no
|
BMC Emerg Med. 2005 Jul 19; 5:5
|
utf-8
|
BMC Emerg Med
| 2,005 |
10.1186/1471-227X-5-5
|
oa_comm
|
==== Front
BMC Fam PractBMC Family Practice1471-2296BioMed Central London 1471-2296-6-251596085210.1186/1471-2296-6-25Study ProtocolThe DiGEM trial protocol – a randomised controlled trial to determine the effect on glycaemic control of different strategies of blood glucose self-monitoring in people with type 2 diabetes [ISRCTN47464659] Farmer Andrew [email protected] Alisha [email protected] David P [email protected] Elizabeth [email protected] Ann Louise [email protected] Andrew [email protected] Department of Primary Health Care, University of Oxford, Old Road Campus, Oxford, OX3 7LF, UK2 School of Sport and Exercise Sciences, University of Birmingham, Birmingham, B15 2TT, UK3 School of Health and Related Research, University of Sheffield, 30 Regent Street, Sheffield, S1 4DA, UK4 General Practice and Primary Care Research Unit, University of Cambridge, Robinson Way, Cambridge, CB2 2SR, UK5 Oxford Centre for Diabetes, Endocrinology and Metebolism Churchill Hospital, Headington, Oxford, OX3 7LJ2005 16 6 2005 6 25 25 12 5 2005 16 6 2005 Copyright © 2005 Farmer et al; licensee BioMed Central Ltd.2005Farmer et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
We do not yet know how to use blood glucose self-monitoring (BGSM) most effectively in the self-management of type 2 diabetes treated with oral medication. Training in monitoring may be most effective in improving glycaemic control and well being when results are linked to behavioural change.
Methods/design
DiGEM is a three arm randomised parallel group trial set in UK general practices. A total of 450 patients with type 2 diabetes managed with lifestyle or oral glucose lowering medication are included. The trial compares effectiveness of three strategies for monitoring glycaemic control over 12 months (1) a control group with three monthly HbA1c measurements; interpreted with nurse-practitioner; (2) A self-testing of blood glucose group; interpreted with nurse- practitioner to inform adjustment of medication in addition to 1; (3) A self-monitoring of blood glucose group with personal use of results to interpret results in relation to lifestyle changes in addition to 1 and 2.
The trial has an 80% power at a 5% level of significance to detect a difference in change in the primary outcome, HbA1c of 0.5% between groups, allowing for an attrition rate of 10%. Secondary outcome measures include health service costs, well-being, and the intervention effect in sub-groups defined by duration of diabetes, current management, health status at baseline and co-morbidity. A mediation analysis will explore the extent to which changes in beliefs about self-management of diabetes between experimental groups leads to changes in outcomes in accordance with the Common Sense Model of illness. The study is open and has recruited more than half the target sample. The trial is expected to report in 2007.
Discussion
The DiGEM intervention and trial design address weaknesses of previous research by use of a sample size with power to detect a clinically significant change in HbA1c, recruitment from a well-characterised primary care population, definition of feasible monitoring and behaviour change strategies based on psychological theory and evidence, and measures along the hypothesised causal path from cognitions to behaviours and disease and well being related outcomes. The trial will provide evidence to support, focus or discourage use of specific BGSM strategies.
==== Body
Background
Diabetes is now a major public health problem. The number of people with diabetes is estimated to reach 330 million by 2030 [1]. There is a high burden from the disease: people with diabetes have an increased two to four fold risk of stroke and heart disease compared to the general population, and increased incidence of retinopathy, peripheral nerve damage and renal problems.
There is now strong evidence for the effectiveness of tight glycaemic control in reducing complications among people with diabetes [2]. However, the evidence with which to translate these research findings into guidance for delivery of health care is lacking [3]. In particular, efforts to promote self management of diabetes have shown limited and transient success in improving HbA1c levels [4]. Blood glucose self-monitoring (BGSM) is a technology that is frequently incorporated into self-management interventions, but has only been separately evaluated in a limited number of trials. Despite the lack of evidence, guidance is given both supporting and discouraging the use of BGSM.
BGSM was used to underpin insulin dose adjustment in the Diabetes Control and Complications Trial among people with type 1 diabetes, which clearly demonstrated the efficacy of glycaemic control in reducing diabetic complications. However, neither the rationale for BGSM nor its efficacy or effectiveness among people with type 2 diabetes is clear. Yet BGSM is now widely accepted as part of management of people with type 2 diabetes [5,6], and the costs associated with its use are rising rapidly [7]. Further trials are required to evaluate the benefit and cost-effectiveness of this technology and its place in the self management of people with type 2 diabetes.
Target population
People with type 2 diabetes are at risk from a range of macrovascular and microvascular diabetic complications. Large trials have confirmed the effectiveness of intensive glycaemic control at reducing these complications [2]. Tight glycaemic control can be achieved through lifestyle change and medications. The target population will comprise the majority of patients on the average practice list with diabetes; we will focus on those patients within 5–10 yrs of diagnosis who are still comparatively healthy, of average age around 55–65 y, and managed on a range of medications and lifestyle advice. The exclusion of those with regular experience of BGSM will be required to avoid randomising them to a group not using meters.
Limitations of previous research
Limited evidence for effectiveness from non randomised studies
A recent qualitative study has suggested that self-monitoring may be an important factor in helping people achieve a better understanding of their condition [8]. However, one study carried out in outpatient and general practitioner clinics in Italy found that increased frequency of monitoring was only associated with improved metabolic control in people able to adjust insulin doses [9]. A population study in the United States also found little relationship between testing frequency and HbA1c value [10]. By contrast, another large study using a cohort design carried out in a group model health maintenance organisation in California has suggested better glycaemic control among type 2 diabetes patients when using a BGSM compared to no use. Although attempts were made to control for differences between groups, the possibility of confounding between attitudes to self-care and use of BGSM cannot be excluded [11].
Limitations of and lack of evidence of effectiveness from randomised trials
Three systematic reviews have provided no evidence that self monitoring is effective in improving glycaemic control for people with type 2 diabetes when compared to urine testing and measurement of glycosylated haemoglobin (HbA1c) [12-14]. The majority of trials identified in these reviews have been carried out in small groups of people. Participants were not recruited from representative populations in the community and the strategies for use of the results from BGSM were not clearly defined. Two more recent studies, set both in hospitals and a family practice setting have adopted a more structured approach to relating blood glucose measurements to subsequent management decisions, but both trials have only published an analysis of people adhering to use of BGSM [15,16].
Research on mediators of effect not investigated in trials
There are a small number of studies that offer some insight into how BGSM might lead to improved blood glucose control among people with type 2 diabetes. BGSM may be helpful in the titration of therapy by either patients or by practitioners or both, but whether regular monitoring is more effective than periodic measurement of HbA1c is unknown. Evidence from qualitative studies with patients suggests that awareness of fluctuations in blood glucose levels may promote adherence to self-care behaviours, including medication taking, diet and physical activity in selected patients [8,17].
There is increasing work in the area of diabetes self-management that uses psychological theory to guide intervention and measurement of the processes of behaviour change. One approach, the Common Sense Model (CSM) [18], proposes that how people understanding threats to their health in central in determining efforts to minimise these health threats. For instance, if people with type 2 diabetes do not believe that physical activity affects their blood glucose levels, they have little reason to be more active to control their condition. Beliefs about illness can be categorised in terms of whether they relate to symptoms/ identity, cause, consequences, time lines, and control/ cure [18]. In support of the CSM, previous work has shown that beliefs about the consequences and controllability of diabetes, and the perceived effectiveness of treatment [19-21], predict patient adherence to recommended lifestyle management. Further, an intervention with myocardial infarction patients based on the CSM successfully managed to alter unhelpful beliefs, and led to faster return to work and fewer symptoms in the intervention group [23]. Further work using this approach to guide intervention and measures with people with type 2 diabetes may inform understanding of the potential mechanisms through which BGSM may improve health.
Limitations of previous interventions
Technology
The majority of previous trials have used reflectance meters rather than biosensor technology. The older meters required larger quantities of blood and took longer to produce a reading than current systems. Although when used correctly the older meters provided reliable information, in practice their accuracy, usability and so potential impact was limited and may have formed a barrier to their effectiveness without high levels of motivation.
Strategy for use of meters
Only one identified randomised trial specified the approach used with patients to support their interpretation of test results [15]. Patients were told that using the meter and keeping records in a diary would provide information that would help them adjust their diet and lifestyle. A defined counselling algorithm was used to help ensure uniformity of delivery, but the extent to which the counselling helped patients relate the results to behaviour change is unclear. However, the impact of the intervention on self-perception, self-reflection and beliefs are not reported. The only published results of the study are from a per-protocol analysis rather than an intention to treat analysis [15]. Therefore, further work in which the intervention includes linking decision making to behaviour change is required.
Quality assurance
Previous trials have not specified the efforts made to monitor delivery of interventions. These include both the process of attempting to ensure that participants understand the techniques being used, and that efforts are made to ensure that the intervention is delivered as per protocol, and that there is adequate separation from the comparison groups in the intervention delivery.
DiGEM Objectives
Primary objective
Our primary objective is to determine whether HbA1c is significantly lower in patients with type 2 diabetes allocated to each of two intervention groups (both receiving training in the techniques of blood glucose self-monitoring, but with one additionally receiving training in the interpretation and application of the results to diet, physical activity and medication adherence) compared to patients allocated to a control group (receiving standardised usual care involving intermittent measurement of HbA1c by health professionals).
Secondary objectives
Secondary objectives are:
(i) To compare well-being, satisfaction, health service use and economic cost between allocated groups.
(ii) To conduct an exploratory analysis of changes in mean HbA1c between regimens among sub-groups defined by duration of diabetes, current management, self-reported health status and co-morbidity.
(iii) To test how self-monitoring influences beliefs and behaviour using measures chosen on the basis of theoretical models of behaviour change.
Methods/design
DiGEM is a four-year study with an open, randomised controlled pragmatic parallel group trial design (Figure 1) with sequential recruitment from two centres. The trial is managed from the Department of Primary Health Care, University of Oxford following NHS R&D Health Technology Assessment Programme guidelines. The study protocol was approved by the Oxfordshire B Research Ethics Committee.
Figure 1 Consort diagram showing planned study numbers.
The study design is shown in Figure 1. Participants are randomised to three groups consisting of
(i) a control group receiving standardised usual care and three monthly measurement of HbA1c,
(ii) a self-testing group who, in addition to the above, are carrying out blood glucose self testing with the results interpreted by the study nurse,
(iii) a self monitoring group who, in addition to both of the above, are given support in interpreting and applying the results of blood testing to enhance motivation and maintain adherence to diet physical activity and medication regimens.
Following randomisation, patients receive the allocated education and training appropriate for their group, with follow up to maintain the interventions at four, 13, 26, and 39 weeks with a final assessment at 52 weeks.
Setting and recruitment
Non-insulin using patients with type 2 diabetes have been recruited from general practices in Oxfordshire and are being recruited from South Yorkshire to take part in the study. The 48 recruited practices represent a geographical spread of rural/suburban centres and cover a wide socio-economic range of patients. The mean number of patients recruited in the 24 practices in Oxfordshire was 10.2, standard deviation (SD) 5.6.
Patients suitable for trial inclusion are identified from practice generated computer lists. Eligible patients are sent an invitation to participate signed by their general practitioner accompanied by an information sheet and a reply paid envelope to facilitate response. One further letter is sent if no response is received in one month.
Study population: inclusion and exclusion criteria
Inclusion criteria are type 2 diabetes, aged 25 years or more at diagnosis, managed with lifestyle or oral hypoglycaemic agents, independent for activities of daily living. Exclusion criteria are use of blood glucose monitor twice a week or more often over the previous three months, current use of insulin, co-morbidity or limited life expectancy that would make intensive glycaemic control inappropriate, last clinic HbA1c or HbA1c at the assessment visit less than 6.2%, or unable to follow trial procedures.
Randomisation
Participants are randomly allocated to one of the three groups using a partial minimisation procedure to adjust the randomisation probabilities between groups to balance important covariates including duration of diabetes, HbA1c, and prior medication using a computer programme (Minim, ).
Baseline measures and follow up
The primary outcome measure is change in HbA1c between the baseline measurement at the assessment visit and 12-month visit. Secondary outcome measures include change in systolic and diastolic blood pressure, weight, serum cholesterol and HDL, self-reported smoking status, dietary intake and physical activity (the Diabetes Self Care Activities Questionnaire) [22], medication adherence (The Medication Adherence Rating Scale) [23], and the scores in the Diabetes Treatment Satisfaction Questionnaire [24], and the Well-being Questionnaire (12 item) [25].
Beliefs about diabetes and its management are assessed using the Illness perceptions Questionnaire [26], Beliefs about Medicines Questionnaire [27], and a questionnaire developed for the study about the effectiveness of changes in eating and physical activity on the course of diabetes, and attitudes to blood glucose self-monitoring. Table 1 summarises the measures and their timing.
Table 1 Study measures
Measures Baseline 3 months 6 months 9 months 12 months
Questionnaire measures:
Illness perceptions questionnaire [26]
Well-being questionnaire 12-item [25]
Beliefs about medicines questionnaire
Medication Adherence Report scale
Beliefs about physical activity and eating +
Beliefs about blood glucose monitoring +
Beliefs about using a blood glucose monitor+
Occupation and social class
Physiological measures
HbA1c
Blood pressure
Weight, height
Total and HDL cholesterol
Costs
Use of medical services
Costs of medication
Costs of delivering intervention
Blood glucose monitoring resources are measured by counting recorded entries in diaries. Medication use, episodes of hospitalisation and their duration will be noted and non-hospital health care resource use will be recorded by a questionnaire administered to all patients at each visit.
Trial procedures
Assessment
Participant eligibility for the study and willingness to be randomised to a group in which they might be required to test their own blood glucose six times a week or more is confirmed at the assessment visit. Following consent, beliefs about diabetes, the role of eating, physical activity and medication are discussed with all participants. A goal setting approach to lifestyle change is introduced and continued in subsequent visits. Baseline blood tests and clinical measurements are taken and questionnaires completed at this visit.
Following the assessment visit and confirmation of eligibility on the basis of HbA1c measurement, patients are randomised to one of three groups: control group, self-testing and self-monitoring.
Post-randomisation
At a visit two weeks after the assessment visit participants receive training and education appropriate to their allocated study group. The control group receives 3-monthly HbA1c measurements and identifies behavioural goals to improve glycaemic control. The self-testing group, in addition, is asked to use a blood glucose meter to record three fasting, pre-meal or 2-hour post meal readings on two days during the week. Treatment targets of fasting and pre-meal levels of 4–6 mmol/l and post meal levels of 6 to 8 mmol/l and advice about using these readings to identify high (>15 mmol/l) and low (<4 mmol/l) blood glucose readings are given. The self-monitoring group, in addition, is provided with training and support to encourage interpretation of readings and application to goals for lifestyle change based on the CSM in order to reach treatment targets.
Follow up visits
Subsequent follow up includes a telephone call 2 weeks after randomisation (after the post-randomisation visit rather than randomisation), and further visits at 4, 13, 26 and 39 weeks. The follow up visits differ according to the allocated group. Those allocated to the control group have a HbA1c measure two weeks before their scheduled visit in order for their glycaemic control to be discussed. The two groups using a meter are managed on the results of their blood glucose self-monitoring. The GP is notified of all HbA1c results and asked to consider changes in medication in line with the National Institute for Clinical Excellence diabetes guidelines. The GP is also notified if blood glucose readings are consistently above 15 mmol/l.
Study measures (see Table 1 and Figure 1)
Baseline self-report measures and measures of belief are completed at the assessment visit. Baseline biochemical measures and clinical measurements are also made. Repeated measurements are made at the 52-week follow up visit. Data on adverse reactions or complications are collected at each study visit along with information about use of medication and health services.
Quality assurance and fidelity of interventions
Patients are supplied with a blood glucose meter calibrated to provide plasma equivalent results. Meters are checked at the beginning of the study and at 26 weeks by study nurses with a test aliquot.
A script outlining the topics to be covered, and explanation of the theoretical basis of the intervention are used to support the nurses in their adherence to study protocol. Study nurses attended a six day phased course in psychological theory, behaviour change techniques and skills training in the intervention. Additional measures to ensure fidelity include self-review of taped interventions by the study nurses and external review by a researcher using a checklist to ensure delivery of the intervention according to protocol. Prompts built into the patient diaries help participants adhere to their allocated group intervention.
Statistical aspects
In the absence of data relating to change in HbA1c we have calculated sample size conservatively based on the absolute level of HbA1c at follow up. We set out to detect a difference in HbA1c of 0.5% between any two groups. At the outset we estimated the SD of HbA1c as 1.5, based on data from the UKPDS. In practice, the SD of baseline HbA1c among the 245 patients recruited in Oxford was 0.9, but we assumed that the SD among additional patients recruited elsewhere could be as high as 1.5. With 205 further patients, the overall SD would be 1.2, and there would be 150 patients in each group (135 allowing for 10% attrition). These numbers would give 93% power to detect a difference of 0.5% in HbA1c between any two trial groups (2-tailed alpha = 0.05). Figure 1 gives estimates of likely numbers in each group and attrition.
We propose to conduct a single analysis of main trial endpoints at the end of the study. The proposed intention to treat analysis will compare mean levels of HbA1c at follow up between the three study groups, with baseline HbA1c as a covariate, using analysis of covariance. Post-hoc t-tests between groups will be conducted in the event of a statistically significant result. Subsequent analysis will include comparing the two intervention groups against the control group.
We will estimate the intervention effect in sub-groups defined by duration of diabetes (above or below median duration), current management (oral hypoglycaemic dugs or dietary management only), health status at baseline (above or below median EQ-5D score) and co-morbidity (presence or absence of diabetes related complications). We will also explore the extent to which the measures of beliefs included in the study can explain changes in behaviour; firstly by comparing mean levels of beliefs e.g. about controllability of diabetes between experimental groups and secondly by a more formal mediation analysis [28]. Within group analyses will be used to determine fidelity to protocol and conformity to the theoretical model, between group analyses will be used to assess impact of the intervention on key variables proposed by the theoretical model. Additional exploratory analysis will include changes in behaviour in relation to perceived threat and changes in perceived thereat from diabetes.
Economic evaluation
The economic evaluation will be on an intention-to-treat basis. A cost-effectiveness/cost-utility analysis will be performed in which the difference in effectiveness will be compared to the difference in total costs between each study intervention group, and the results will be expressed as incremental cost-effectiveness ratios. Effectiveness will be measured in terms of change in HbA1c, and modelled for life years gained and quality adjusted life years gained. Unit costs will be attached to the resource items collected from the healthcare resource utilisation assessment using published national average costs and tariff averages for procedures to calculate costs. Mean values and 95% confidence intervals will be reported for each component of resource use and cost and for total costs and effectiveness. Sensitivity analyses will be performed on all aspects of the economic evaluation that are subject to uncertainty.
Discussion
This study will make an important contribution to the evidence-base for the use of blood glucose self-monitoring in non-insulin using patients with type 2 diabetes. It will provide a robust estimate of overall effect of use of the meters for self-testing alone and with a more intensive programme to support people in using their test results actively in self management of health related behaviours.
This trial will address two main problems associated with previous trials. Firstly, the study is adequately powered to identify a reduction in HbA1c of 0.5%, which is associated with clinically important reductions in diabetic complications. Secondly, study participants allocated to the most intensive group are explicitly supported in interpretation of their study results in relation to lifestyle changes. This is the first study in this field to use standardised delivery of these interventions together with the other measures supporting fidelity to protocol.
A further strength of the trial design is the inclusion of measures that explore the extent to which the interventions being used influences beliefs and behaviour. It will be possible to test whether the interventions, especially that based on the CSM, are successful in altering beliefs about diabetes, and whether the interventions were delivered with fidelity to their theoretical basis. It will also be possible to test whether differences in behaviour and HbA1c between experimental groups are due to the interventions bringing about changes in beliefs, as proposed by the CSM. Depending on the pattern of findings from the mediation analysis, it will be possible to conclude there is no mediation (none of the causal effect of intervention on behaviour is transmitted by beliefs about diabetes), total mediation (all of the effect is transmitted by beliefs) or partial mediation (part of the effect is transmitted by beliefs and part is direct). This will provide information about the extent to which the CSM can usefully inform intervention content to change beliefs and behaviours.
The trial is mainly generalisable to that group of patients willing to be randomised to no self-testing. It will be limited in its ability to inform management of people who are enthusiastic about regular meter use. Not only does this group include people who have already been recommended to use a meter by their doctor or nurse, but also includes people who have obtained a meter in the absence of medical advice. In health service terms, these are people who are likely to obtain a meter whether or not they are prescribed or offered one. A trial to address management in this group would be more difficult because of difficulties in identifying individuals who are both enthusiasts, yet willing not to be exposed to the use of self-monitoring. However, the detailed information about beliefs and the pre-specified sub-group analyses will provide information to inform design of future studies in this area by allowing refinement of interventions and accurate estimates to inform sample size calculations The detailed information from blood glucose diaries and relation to outcomes will also inform the refinement of training programmes and data interpretation.
This trial has features of both a pragmatic and an explanatory trial [29]. The use of three parallel groups, delivery of interventions according to protocol and extensive measures along a causal pathway are characteristic of phase III or explanatory trials. The use of a primary care setting, comparison of three feasible health service strategies, wide range of recruitment a long duration, and estimation of costs are, however, more characteristic of a pragmatic trial. This hybrid design, whilst answering a health service question, will provide additional information to guide future research. The extent to which the trial is able to deliver data that informs both of these agendas will inform design of trials of emerging health technologies where the opportunity to understand underlying mechanisms may be lost in an effort to simplify protocol design.
The results of this trial will be available in 2007.
Competing interests
The authors hold no financial or non-financial competing interests. The views expressed in this paper do not necessarily reflect those of the NHS.
Authors' contributions
AF, A-LK and AN had the original idea for the study and wrote the study protocol. AF, AW, DF and ALK developed study measures and intervention. AF, AW and LG have set up the study and made it run. AF is the guarantor of this paper.
Investigators/TSC and DMEC
Members of the writing committee for this paper were A. Farmer, A. Wade, D French E Goyder HAW Neil, A-L Kinmonth.
The study investigators are Oxford: A Farmer, D Mant, R Holman, S Ziebland, R. Holman, A Gray, and P Yudkin (trial statistician); Cambridge A-L Kinmonth; Birmingham D French
Members of the trial steering committee are N Stott (Chair), A Farmer, HAW Neil, S Sutton, H Tewson, D Chapman, H Hearnshaw, E Goyder, and M Jiwa.
Members of the data monitoring committee are C. Baigent (Chair), J Levy and K Wheatley.
Coordinating Centres: (Oxford) A Wade (trial coordinator), A Craven (trial administrator), J Simon (health economist) and A Fuller (data manager); (Sheffield) Vivienne Walker
Study Nurses (Oxford) M Selwood, H Kirlow, M Chapman, and S Turner; (Sheffield) A Casbolt, K Dobson, and A Willert.
Funding
Health Technology Assessment Programme and NHS R&D NHS Support Funding. AF is supported by an NHS R&D Career Development Award
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Medisense provided blood glucose meters (Optium). We are grateful to the patients taking part in this study and to their general practitioners for support and help. M McKinnon, W Hardeman, I Hobbis and J Donnelly have helped with training the nurses.
==== Refs
Wild S Roglic G Green A Sicree R King H Global Prevalence of Diabetes: Estimates for the year 2000 and projections for 2030 Diabetes Care 2004 27 1047 1053 15111519
UK Prospective Diabetes Study (UKPDS) Group Intensive blood glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33) Lancet 1998 352 837 853 9742976 10.1016/S0140-6736(98)07019-6
Garfield SA Malozowski S Chin MH Venkat Narayan KM Glasgow RE Green LW Hiss RG Krumholz HM Considerations for Diabetes Translational Research in Real-World Settings Diabetes Care 2003 26 2670 2674 12941736
Norris SL Lau J Smith SJ Schmid CH Engelgau MM Self-Management Education for Adults With Type 2 Diabetes: A meta-analysis of the effect on glycemic control Diabetes Care 2002 25 1159 1171 12087014
Blonde L Ginsberg BH Horn S Frequency of blood glucose monitoring in relation to glycaemic control in patients with type 2 diabetes Diabetes Care 2002 25 245 246 11772926
European Diabetes Policy Group A desktop guide to type 2 diabetes mellitus Diabet Med 1999 16 716 730 10510947 10.1046/j.1464-5491.1999.00166.x
Farmer AJ Neil A Variations in glucose self-monitoring during oral hypoglycaemic therapy in primary care (letter) Diabet Med 2004
Peel E Parry O Douglas M Lawton J Blood glucose self-monitoring in non-insulin-treated type 2 diabetes: a qualitative study of patients' perspectives Br J Gen Pract 2004 54 183 188 15006123
Franciosi M Pellegrini F DeBeradis G Belfiglio M Cavaliere D Di Nardo B Greenfield S Kaplan SH Sacco M Tognoni G Valentini M Nicolucci A The impact of blood glucose self-monitoring on metabolic control and quality of life in type 2 diabetic patients: an urgent need for better educational strategies Diabetes Care 2001 24 1870 1877 11679449
Harris MI Frequency of blood glucose monitoring in relation to glycaemic control in patients with type 2 diabetes Diabetes Care 2001 24 982
Karter AJ Ackerson LM Darbinian JA D'Agostino RB Ferrara A Liu J Selby JV Self-monitoring of blood glucose levels and glycaemic control: the Northern Kaiser Permanente Registry Am J Med 2002 111 1 9 11448654 10.1016/S0002-9343(01)00742-2
Coster S Gulliford MC Seed PT Powrie JK Swaminatham R Self-monitoring in Type 2 diabetes mellitus: a meta-analysis Diabet Med 2000 17 755 761 11131099 10.1046/j.1464-5491.2000.00390.x
Halimi S Apports de "auto surveillance glycemique dans la prise en charge des diabetique insulino (DID) et non insulino dependents (DNID) Diabetes Metab (Paris) 1998 24 35 41
Faas A Schellevis FG van Eijk JT The efficacy of self monitoring of blood glucose in NIDDM subjects: a criteria-based literature review Diabetes Care 1997 20 1482 1486 9283802
Schwedes U Siebolds M Mertes G Meal-Related Structured Self-Monitoring of Blood Glucose: Effect on diabetes control in non-insulin-treated type 2 diabetic patients Diabetes Care 2002 25 1928 1932 12401734
Guerci B Drouin P Grange V Bougneres P Fontaine P Kerlan V Passa P Thivolet C Vialettes B Charbonnel B Self-monitoring of blood glucose significantly improves metabolic control in patients with type 2 diabetes mellitus: the Auto-Surveillance Intervention Active (ASIA) study Diabetes Metab 2003 29 587 594 14707887
Fox MA Cassmeyer V Eaks GA Hamera E O'Connell K Knapp T Blood glucose self-monitoring usage and its influence on patients' perceptions of diabetes Diabetes Educ 1984 10 27 31 6386403
Leventhal H Nerenz DR Steele DJ Baum A, Taylor SE, Singer JE Illness representations and coping with health threats Handbook of psychology and health 1984 Hillsdale, NJ: Erlbaum 219 252
Hampson SE Glasgow RE Toobert DJ Personal models of diabetes and their relations to self-care activities Health Psychol 1990 9 632 646 2226390 10.1037//0278-6133.9.5.632
Hampson SE Glasgow RE Foster LS Personal models of diabetes among older adults: Relationship to self-management and other variables Diabetes Educ 1995 21 300 307 7621732
Hampson SE Glasgow RE Strycker LA Beliefs versus feelings: A comparison of personal models and depression for predicting multiple outcomes in diabetes Br J Health Psychol 2000 5 27 40 10.1348/135910700168748
Toobert DJ Hampson SE Glasgow RE The summary of diabetes self-care activities measure: results from 7 studies and a revised scale Diabetes Care 2000 23 943 950 10895844
Horne R Weinman J Patients' beliefs about prescribed medicines and their role in adherence to treatment in chronic physical illness J Psychosom Res 2001 47 555 567 10661603 10.1016/S0022-3999(99)00057-4
Bradley C Lewis KS Measures of psychological well-being and treatment satisfaction developed from the responses of people with tablet-treated diabetes Diabet Med 1990 7 445 451 2142043
Bradley C Bradley C The Well-being Questionnaire Handbook of Psychology and Diabetes 1994 Switzerland: Harwood Academic Publishers 89 110
Moss-Morris R Weinman J Petrie KJ Horne R Cameron LD Buick D The revised illness perception questionnaire (IPQ-R) Psychol Health 2002 17 1 6 10.1080/08870440290001494
Horne R Weinman J Hankins M The beliefs about medicines questionnaire: the development and evaluation of a new method for assessing the cognitive representation of medication Psychol Health 1999 14 1 24
Kenny DA Kashy DA Bolger N Gilbert DT, Fiske ST, Lindzey G Data analysis in social psychology The handbook of social psychology 1998 New York: McGraw-Hill 233 265
Schwartz D Lellouch J Explanatory and Pragmatic Attitudes in Therapeutical Trials J Chronic Disease 1967 20 637 648 4860352 10.1016/0021-9681(67)90041-0
|
15960852
|
PMC1185530
|
CC BY
|
2021-01-04 16:29:12
|
no
|
BMC Fam Pract. 2005 Jun 16; 6:25
|
utf-8
|
BMC Fam Pract
| 2,005 |
10.1186/1471-2296-6-25
|
oa_comm
|
==== Front
BMC Fam PractBMC Family Practice1471-2296BioMed Central London 1471-2296-6-281601417010.1186/1471-2296-6-28Research ArticleCorrelation of same-visit HbA1c test with laboratory-based measurements: A MetroNet study Schwartz Kendra L [email protected] Joseph C [email protected] Monina G [email protected] Patricia A [email protected] Anne Victoria [email protected] Department of Family Medicine, Wayne State University School of Medicine, 101 E. Alexandrine, Detroit, MI, 48201 USA2 Department of Family Medicine, St. John Health System, 24911 Little Mack, St. Clair Shores, MI, 48080 USA2005 13 7 2005 6 28 28 19 2 2005 13 7 2005 Copyright © 2005 Schwartz et al; licensee BioMed Central Ltd.2005Schwartz 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
Glycated hemoglobin (HbA1c) results vary by analytical method. Use of same-visit HbA1c testing methodology holds the promise of more efficient patient care, and improved diabetes management. Our objective was to test the feasibility of introducing a same-visit HbA1c methodology into busy family practice centers (FPC) and to calculate the correlation between the same-visit HbA1c test and the laboratory method that the clinical site was currently using for HbA1c testing.
Methods
Consecutive diabetic patients 18 years of age and older having blood samples drawn for routine laboratory analysis of HbA1c were asked to provide a capillary blood sample for same-visit testing with the BIO-RAD Micromat II. We compared the results of the same-visit test to three different laboratory methods (one FPC used two different laboratories).
Results
147 paired samples were available for analysis (73 from one FPC; 74 from the other). The Pearson correlation of Micromat II and ion-exchange HPLC was 0.713 (p < 0.001). The Micromat II mean HbA1c was 6.91%, which was lower than the 7.23% from the ion-exchange HPLC analysis (p < 0.001). The correlation of Micromat II with boronate-affinity HPLC was 0.773 (p < 0.001); Micromat II mean HbA1c 6.44%, boronate-affinity HPLC mean 7.71% (p < 0.001). Correlation coefficient for Micromat II and immuno-turbidimetric analysis was 0.927 (p < 0.001); Micromat II mean HbA1c was 7.15% and mean HbA1c from the immuno-turbidimetric analysis was 7.99% (p = 0.002). Medical staff found the same-visit measurement difficult to perform due to the amount of dedicated time required for the test.
Conclusion
For each of the laboratory methods, the correlation coefficient was lower than the 0.96 reported by the manufacturer. This might be due to variability introduced by the multiple users of the Micromat II machine. The mean HbA1c results were also consistently lower than those obtained from laboratory analysis. Additionally, the amount of dedicated time required to perform the assay may limit its usefulness in a busy clinical practice. Before introducing a same-visit HbA1c methodology, clinicians should compare the rapid results to their current method of analysis.
==== Body
Background
The percent HbA1c of glycated hemoglobin provides an estimate of blood glucose levels over a 3–4 month period. The HbA1c level is used for patient education and counseling, for feedback about diabetic control, to improve patient motivation, and to monitor management; thus its measurement should be optimally accurate and precise [1]. However, to date, there is no international standard for determining HbA1c [2-4], and various methodologies are commercially available. Tran [1] determined the physiological (changes over time between measurements) and analytic variation of two widely used laboratory assays, one a high performance liquid chromatography (HPLC) method, and the other an immunoassay [1]. The coefficient of variation (CV) for the HPLC was 2.6%. The 5.1% CV of the immunoassay method exceeded physiologically established limits of 2–3%, and those of the National Glycohemoglobin Standardization Program (3–4%).
Hosseini et al. [5] reported a relative ranking of assays to result in a normal HbA1c level by using the same patient's blood tested with five assays, each of which used a different method. They found that glycated hemoglobin results vary widely, with some assays consistently more likely to result in a "normal glycated hemoglobin" level than other assays, consequently resulting in differing implications for an individual patient to achieve a HbA1c level within the normal range. Ogawa [6] reported a case series where HbA1c was underestimated in the measurement by HPLC which excluded glycated abnormal hemoglobin [6]. These findings illustrate the potential usefulness for clinical practitioners to evaluate the performance of their method for determining HbA1c, especially if using different methodologies for the same patient.
Recent developments in medical technology allow clinicians to determine HbA1c test results during a patient's office visit. Several manufacturers offer an assay that can be performed by trained medical personnel and yield HbA1c results in five to ten minutes. We found only a few reports of the performance of such rapid tests used at the point of care [7,8], and one study that was conducted by the test manufacturer [9].
The objective of this pilot study was to test the feasibility of introducing a same-visit HbA1c methodology into busy family practice centers (FPCs) and to compare the results obtained from a point-of-care test with a laboratory-based technique. Specifically, our purpose was to determine: 1) if a specific rapid HbA1c methodology was accepted by medical support staff in two busy FPCs; and 2) how rapid HbA1c results compared with the standard laboratory methodology.
Methods
Study design
Patients were recruited for this cross-sectional study from two FPCs that are members of MetroNet, a metropolitan Detroit practice-based research network. At both sites, HbA1c analysis is routinely performed at an outside laboratory on venipuncture samples. Physicians, medical assistants, and research assistants identified consecutive diabetic patients 18 years of age and older whose physicians ordered HbA1c analysis. The study was explained to these eligible patients and informed consent obtained from those who wished to participate.
After patients were enrolled, a finger-prick blood sample was collected for in-office HbA1c testing with the BIO-RAD Micromat II. Since the BIO-RAD Micromat II is compatible with capillary, venous, and EDTA anti-coagulated blood samples, aliquots of these types were also acceptable for analysis. Research and medical staff were instructed to use finger-prick capillary samples whenever possible, but venous samples from the blood draw apparatus, or a drop of blood from the EDTA tube was substituted when necessary. At one FPC only finger-prick samples were used, while at the other FPC, thirteen MicroMat II samples were venous and five were EDTA anti-coagulated; the remaining 56 tests were performed using capillary blood samples.
The data collected included patient name, study site, the person performing same-visit HbA1c analysis, the date, and the rapid HbA1c result. Physicians were blinded to rapid HbA1c results, and relied on the laboratory analysis to make treatment decisions during the study period. One FPC used one of two different laboratories based on the patient's health insurance carrier. At one laboratory, the Primus Model 386 was used for HbA1c testing, which is a boronate – affinity HPLC method. The other laboratory used the Roche Integra 800, which uses an immuno-turbidimetric methodology. The laboratory of the second FPC used the Tosoh A1c 2.2 Plus, an ion-exchange HPLC, for analysis.
All three methodologies are aligned to Diabetes Control and Complications Trial (DCCT) and National Glycohemoglobin Standardization Program (NGSP) standards. All have linear response from HbA1c level of 3–4% to 20% or higher. The intra- and inter-assay coefficients of variation are displayed in Table 1. These values were either obtained directly from the laboratory performing the assay (Primus 386) or from the manufacturer. All are within NGSP acceptable limits.
Table 1 Coefficients of variation (CV) for three laboratory analyzers
Instrument Method Intra-assay CV Inter-assay CV
Primus 386 Boronate affinity HPLC 0.9% 2.9%
Roche Integra 800 Immuno-turbidimetric 2.3% 2.4%
Tosoh A1c2.2 Plus Ion-exchange HPLC 1.3% < 4.0%
The BIO-RAD Micromat II, which provides results in approximately 5 minutes, incorporates an affinity chromatography method that measures the percent glycated hemoglobin in the sample. According to the manufacturer, the analyzer then uses a factory-set algorithm to deliver an HbA1c result which is calibrated to the recommendations of the DCCT and is traceable to the NGSP. The intra-assay coefficient of variation is reported to range from 2.93 – 4.65%; higher at lower values of HbA1c. The inter-assay coefficient of variation is estimated to be higher; however values are not given in the package insert. The sensitivity of the assay ranges from 4 – 15% HbA1c. BIO-RAD representatives provided an in-service to help familiarize staff in the use and operation of the Micromat II analyzer.
Each HbA1c analysis with the Micromat II requires a single test cartridge, which consists of several tubes with reagents that are mixed and decanted into a collection reservoir for measurement. After a test cartridge has been placed into the Micromat II, a 20 microliter blood sample is added to the first tube. This initiates a series of aliquot additions and incubation steps. In total there are four decanting steps followed by four incubations. These incubations require a total time of 230 seconds and range from 40 seconds to 80 seconds in length. Quality control procedures were carried out as outlined in the Micromat II instruction manual. Controls and standards were run per the manufacturer's recommendation; results were always acceptable.
Analytic strategy
Data were analyzed separately by type of laboratory methodology. To evaluate the performance of the BIO-RAD Micromat II, Pearson correlations were calculated using the laboratory results as the standard. Scatter plots and regression lines were also examined. The mean absolute difference between the sample groups was determined to test the hypothesis that group means are equal (α = 0.05), using a two-sided paired t-test.
Results
One hundred fifty-six patients were enrolled into the study (75 from one FPC, and 81 from the other FPC). Nine different medical staff performed the rapid HbA1c testing. Data from nine patients were omitted: eight had missing laboratory results, and one result was out of the precision range of the machine (HbA1c = 18.1%). Therefore, 147 paired samples were available for analysis, 73 from one site and 74 from the other.
Considering first the data from the site that used two different laboratories: the boronate-affinity HPLC (n = 63) and the immuno-turbidimetric (n = 11), we found a significant correlation with the Micromat II results for both (Pearson r = 0.773, p < 0.001 and r = 0.927, p < 0.001, respectively) (Figures 1 and 2). The range of values was from 2.3% to 12.70%. The laboratory method yielded a mean HbA1c value that was significantly higher than that from the Micromat II for both methodologies (7.71 ± 1.99 vs. 6.44 ± 1.99, p < 0.001 and 7.99 ± 1.76 vs. 7.15 ± 1.72, p = 0.002, respectively (Table 1). Similarly, the Micromat II correlated well with the ion-exchange HPLC (n = 73, Pearson r = 0.713, p < 0.001) (Figure 3). Again, the mean HbA1c result from the laboratory was significantly greater than the mean from the Micromat II (7.23 ± 1.51 vs. 6.91 ± 1.34, p = 0.014). The range of results from these two methods was 3.6% to 15.80%.
Figure 1 Scatterplot and regression line of HbA1c values produced from boronate affinity HPLC and Micromat II.
Figure 2 Scatterplot and regression line of HbA1c values produced from immuno-turbidimetric and Micromat II methods.
Figure 3 Scatterplot and regression line of HbA1c values produced from ion-exchange HPLC and Micromat II methods.
Regarding feasibility and acceptability of introducing the same-visit Micromat II test into the busy clinical practice setting, we found that medical assistants were able to collect and analyze samples and produce same-visit results. However, the five minute time dedication for each individual analysis was not well tolerated by staff because of numerous competing demands that made it difficult to perform all the test steps in the time intervals prescribed.
Discussion
Physicians in ambulatory settings routinely send blood samples to laboratories for HbA1c testing, and then wait several days for the HbA1c test results. Thus, patient counseling and treatment adjustments based on HbA1c levels are delayed, and at times follow-up can be lost completely.
Recent advancements in technology now make it possible for physicians to incorporate point-of-care HbA1c results to evaluate and adjust treatment of their diabetic patients. Studies of the effect of same-visit HbA1c measurement found significantly improved glycemic control through 12-month follow-up [10,11]. This technology is gaining acceptance, and is now offered by a number of manufacturers. The same-visit HbA1c test provides the opportunity to improve diabetes care by discussing the value and adjusting management as needed during the same-visit, rather than waiting until the patient can be telephoned and/or scheduled for a future visit. HbA1c testing has been studied for its effect on improved glycemic control in trials primarily conducted in specialty clinics. Yet, little is published regarding the validity of the same-visit test result, and the feasibility of using a same-visit methodology in a busy primary care setting.
The manufacturer reports a correlation coefficient of 0.96 between the BIO-RAD Micromat II and HPLC methodology. However, the correlation coefficients we obtained in this clinical situation (r = 0.713; and r = 0.773 for the two different HPLC methods) were less than reported by the company. The highest correlation was with the immuno-turbidimetric methodology (r = 0.927). The mean HbA1c level obtained from Micromat II was significantly lower than that yielded from the three types of laboratory analysis, and this difference spanned the treatment threshold level currently recommended by the American Diabetes Association (ADA) [12]. Thus, for some patients, the Micromat II rapid test yielded a test result that was below the ADA treatment threshold of 7%, while the laboratory analysis produced a test result above 7%, suggesting the need for more intensive therapy.
Limitations
There are likely limitations to the generalizability of the study findings. First, the number of medical staff (n = 9) that collected samples and performed the HbA1c testing may have increased the variability of the same-visit test results. Similarly, the correlation between the laboratory and the same-visit methodologies may be improved when conducted under ideal conditions where sources of variation in the operation of the Micromat II are minimized. Secondly, introducing a research study into a busy clinical practice setting is often met with varying degrees of resistance. Thus, evaluating the acceptance of what may have been viewed by staff as a research technique may have limitations when generalizing the acceptance of a clinical procedure. However, our purpose was to conduct a correlation study in the real world setting of the busy FPC. We trained all clinical staff in the calibration and specimen analysis of the point-of-care instrument. From discussions with the clinical staff and physicians, we learned that there was variability among staff members to faithfully adhere to the Micromat II timed steps as outlined in the test kit instructions.
Conclusion
Same-visit HbA1c testing offers potential benefits for diabetes care, as patient results are available in the same-visit. However, clinicians should be aware that the rapid HbA1c technology may produce results that are lower than the method that they have been utilizing, and that the same-visit test may suggest a different treatment strategy than a result from their usual laboratory testing source. To overcome this barrier, we suggest that clinicians determine how the results of a same-visit HbA1c test compare with the outside laboratory reports on which they routinely base their treatment plans before incorporating the same-visit HbA1c test into their practice.
Abbreviations
HbA1c – Hemoglobin A1c
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
KS, JM and AVN developed the idea. KS, PW, and AVN designed and oversaw data collection. KS and AVN supervised statistical analyses performed by JM and MB. All authors contributed to the interpretation of the data and the writing and editing of the manuscript.
Table 2 Comparison of mean percent HbA1c (SD) of paired samples using three laboratory methodologies and BIO-RAD Micromat II same-visit assay
Mean percent HbA1c (SD)
Laboratory Methodology HPLC Micromat II p-value
Boronate affinity HPLC (n = 63) 7.71 (1.99) 6.44 (1.99) <0.001
Immuno-tubidimetric (n = 11) 7.99 (1.76) 7.15 (1.72) 0.002
Ion-exchange HPLC (n = 73) 7.23 (1.51) 6.91 (1.34) 0.014
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was supported in part by funding from DHHS HRSA (D12 HP00175).
==== Refs
Tran DV Hofer TL Lee T Cembrowski GS Unique approach to derivation of random error in laboratory assays: application to glycohemoglobin testing demonstrates poor clinical performance for immunochemistry assay Diabetes Technol Ther 2003 5 975 978 14709199 10.1089/152091503322641015
Groche D Hoeno W Hoss G Vogt B Herrmann Z Witzigmann A Standardization of two immunological HbA1c routine assays according to the new IFCC reference method Clin Lab 2003 49 657 661 14651337
Miedema K Towards worldwide standardisation of HbA1c determination Diabetologia 2004 47 1143 1148 15249996 10.1007/s00125-004-1453-0
Home P Mbanya JC Horton E Standardisation of glycated haemoglobin Bmj 2004 329 1196 1197 15550404 10.1136/bmj.329.7476.1196
Hosseini SS Bibler I Charles MA The narrow therapeutic window of glycated hemoglobin and assay variability Metabolism 1999 48 1498 1502 10599979 10.1016/S0026-0495(99)90236-6
Ogawa K Bando T Ogawa M Miyazaki A Nakanishi T Shimizu A Hemoglobin variant HbG-coushatta (beta-22 Glu --> Ala) found by dissociation of blood glucose from values of HbA1C measured by HPLC Intern Med 2003 42 781 787 14518662
Jermendy G Nadas J Farkas K [Rapid hemoglobin A1c determination (a new possibility in diabetes care)] Orv Hetil 1999 140 1251 1254 10377737
Gebrekidan A Gill G Wile D Tesfaye S An accurate and portable system for glycated haemoglobin measurement in the tropics Trop Doct 2004 34 94 95 15117136
Stivers CR Baddam SR Clark AL Ammirati EB Irvin BR Blatt JM A miniaturized self-contained single-use disposable quantitatitve test for hemoglobin A1c in blood at the point of care Diabetes Technol Ther 2000 2 517 526 11469613 10.1089/15209150050501916
Cagliero E Levina EV Nathan DM Immediate feedback of HbA1c levels improves glycemic control in type 1 and insulin-treated type 2 diabetic patients Diabetes Care 1999 22 1785 1789 10546008
Thaler LM Ziemer DC Gallina DL Cook CB Dunbar VG Phillips LS El-Kebbi IM Diabetes in urban African-Americans. XVII. Availability of rapid HbA1c measurements enhances clinical decision-making Diabetes Care 1999 22 1415 1421 10480502
Standards of medical care in diabetes Diabetes Care 2004 27 Suppl 1 S15 35 14693923
|
16014170
|
PMC1185531
|
CC BY
|
2021-01-04 16:29:13
|
no
|
BMC Fam Pract. 2005 Jul 13; 6:28
|
utf-8
|
BMC Fam Pract
| 2,005 |
10.1186/1471-2296-6-28
|
oa_comm
|
==== Front
BMC GastroenterolBMC Gastroenterology1471-230XBioMed Central London 1471-230X-5-221598516810.1186/1471-230X-5-22Research ArticleAlterations of tumor suppressor gene p16INK4a in pancreatic ductal carcinoma Attri Jyotika [email protected] Radhika [email protected] Siddhartha [email protected] Bishan Dass [email protected] Jaidev [email protected] Department of General Surgery, Postgraduate Institute of Medical Education and Research, Chandigarh, India2 Department of Cytology and Gynaec Pathology, Postgraduate Institute of Medical Education and Research, Chandigarh, India3 Department of Experimental Medicine and Biotechnology, Postgraduate Institute of Medical Education and Research, Chandigarh, India4 Department of Histopathology, Postgraduate Institute of Medical Education and Research, Chandigarh, India2005 28 6 2005 5 22 22 15 12 2004 28 6 2005 Copyright © 2005 Attri 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
Cell cycle inhibitor and tumor suppressor gene p16 / MTS-1 has been reported to be altered in a variety of human tumors. The purpose of the study was to evaluate primary pancreatic ductal adenocarcinomas for potentially inactivating p16 alterations.
Methods
We investigated the status of p16 gene by polymerase chain reaction (PCR), nonradioisotopic single strand conformation polymorphism (SSCP), DNA sequencing and hypermethylation analysis in 25 primary resected ductal adenocarcinomas. In addition, we investigated p16 protein expression in these cases by immunohistochemistry (IHC) using a monoclonal antibody clone (MS-887-PO).
Results
Out of the 25 samples analyzed and compared to normal pancreatic control tissues, the overall frequency of p16 alterations was 80% (20/25). Aberrant promoter methylation was the most common mechanism of gene inactivation present in 52% (13/25) cases, followed by coding sequence mutations in 16% (4/25) cases and presumably homozygous deletion in 12% (3/25) cases. These genetic alterations correlated well with p16 protein expression as complete loss of p16 protein was found in 18 of 25 tumors (72%).
Conclusion
These findings confirm that loss of p16 function could be involved in pancreatic cancer and may explain at least in part the aggressive behaviour of this tumor type.
==== Body
Background
Pancreatic cancer is a malignant neoplasm in the digestive tract the etiology of which is not known fully as yet. Recent advances in molecular oncology have provided explanations at the DNA level that multiple genetic changes contribute to pancreatic cancer development in which the p16 locus of tumor tissue is nearly always altered [1]. The p16 tumor suppressor gene located on 9p21 enodes a 16 KDa protein that acts as a cyclin dependent kinase (CDK) 4/6 inhibitor [2].
p16 belongs to an important group of proteins that includes the p15INK4b, p21waf1 and p27KIP1 which negatively regulate the G1 phase of cell cycle [2]. The p16 gene product binds to CDK4 and CDK6 inhibiting their interaction with cyclin D1. The inhibition of cyclin D1-CDK4/6 complex activity prevents retinoblastoma protein phosphorylation and release of E2F, leading to the inhibition of cell cycle in G1/S transition [3]. Genetic abnormalities inactivating the p16 gene thus confer a growth advantage to the cell contributing to tumorigenesis.
Inactivating alterations of the gene have been commonly identified in a number of human malignancies [4,5]. In cancers, functional loss of p16INK4a occurs as a consequence of somatic mutations, homozygous and heterozygous deletions [6,7]. A high frequency of homozygous deletion and mutation of this gene have been detected in cell lines derived from different types of tumors (glioma, breast, lung, bladder and melanoma) [6,8] suggesting that p16 may play an important role in the regulation of cellular growth in the majority of cell types. However, homozygous deletions and somatic mutations are rarely observed in primary tumors with p16 genetic alterations [9,10]. On the other hand, denovo methylation has also been proposed to be an important alternative mechanism of p16 gene inactivation [11]. DNA hypermethylation can inhibit transcription of tumor suppressor genes and mismatch repair genes (p16, hMLH1 and VHL), providing an epigenetic mechanism of selection during tumorigenesis [11,12].
Abnormalities of tumor suppressor gene p16 have been reported in a variety of human tumors but less information is available regarding alterations of p16 in primary pancreatic ductal carcinoma than in pancreatic cancer derived cell lines and xenografts. There have been a few reports on the p16 alterations in tissue specimen of primary pancreatic ductal adenocarcinomas till date [13-17]. To provide further evidence that the cell cycle inhibitor p16 might be relevant for pancreatic tumorigenesis, we performed a comprehensive analysis of the mechanisms involved in p16 inactivation such as mutation, hypermethylation and homozygous deletion, in primary ductal adenocarcinomas.
Methods
Tissue samples
A written informed consent was obtained from each patient for inclusion in this study which was carried out after obtaining a formal approval from the Institute Ethics Committee. Pancreatic cancer tissues were obtained from 25 patients (15 males, 10 females) undergoing surgery for pancreatic cancer. Normal tissue away from the main tumor mass was taken as control. The age of these patients ranged from 27–78 years. According to the tumor, node, metastases classification of International Union against cancer [18], there were 2 patients with stage I, 8 patients with stage II, 14 patients with stage III and one patient with stage IV disease.
Freshly removed pancreatic tissue samples were immediately fixed in 10% formalin for 24 hours and paraffin embedded. A section of each specimen was stained with hematoxylin and eosin and microscopically examined to confirm the diagnosis. In addition, fresh tissues for molecular analysis were frozen and stored at -80°C until use. One frozen section from each tissue subjected to molecular analysis was assessed histologically to ensure the presence of tumor and only those samples which contained >90% of tumor were included in the final analysis.
Detection of homozygous deletion in p16INK4a
Genomic DNA was isolated from the stored frozen tissue using the parallel RNA/DNA extraction kit (Qiagen, GmbH, Germany). PCR was performed using 100 ng template DNA and 10 pmol of each primer (Biobasic INC, Canada) in a volume of 50 μl containing 10 mM Tris HCl (pH 8.3), 50 mM KCl, 1.5 mM MgCl2, 2.0 mM of each dNTPs and 1.5 units Taq polymerase (ROCHE, GmbH, Germany). The following primer pairs were used:
Exon 1: 5' cggagagggggagaacagac 3' and 5' ctggatcggcctccgaccgtaac 3' 189 bp
Exon 2: 5' tgagggaccttccgcggc 3' and 5' gtcatgatgatgggcagcgc 3' 307 bp
Exon 3: 5' cacatccccgattgaaagaac 3' and 5' cagtgaatgaatgaaaatta 3' 489 bp
These primers were designed by us based on their mRNA sequence available in the Genbank [Genbank: 4502748]. The PCR program was set for an initial denaturation at 95°C for 5 min, 35 cycles of denaturation at 95°C for 1 min, annealing at 55–56°C (depending on primer pair), for 1 min, extension at 72°C for 1 min and final extension at 72°C for 7 min. PCR reaction products were electrophoresed on a 1.5% agarose gel and were visualized under UV illuminator. Each time, a positive control (normal lymphocyte DNA) was included in the PCR reaction.
Detection of mutation in p16INK4a gene by single strand conformation polymorphism (SSCP) analysis
All samples were analysed for mutation in exon 1, 2 and 3 of p16 by SSCP [19] and DNA sequencing of the PCR products revealing a mobility shift.
For SSCP analysis, 1 μl of each PCR product formed using the above mentioned PCR program and primers, was mixed with 9 μl SSCP loading buffer (98% v/v, formamide, 10 mM EDTA+0.05% Bromo-Phenol Blue+0.5% Xylene Cyanol) and incubated at 90°C for 5', followed by rapid cooling on ice. 10 μl of this mixture was loaded on a 6% acrylamide gel containing 1XTrisborate EDTA buffer. Following electrophoresis at 50 V at 4°C for 15–20 hrs, the gel was analysed by silver staining [20].
The presence of bands with variant migration pattern was confirmed by repeating PCR-SSCP at least prior to extraction of band for DNA sequence analysis.
DNA sequencing
PCR products that revealed mobility shift on SSCP analysis were send for DNA sequencing. Automated sequencing reactions were carried out using Perkin Elmer Big Dye sequence Terminator Mix (ABI / PE, Foster, CA) as per manufacturer's instructions and sequenced on an ABI 377 sequencer.
Methylation analysis of p16INK4a gene
The methylation status of 5' CpG islands of p16 gene by bisulfite modification of DNA and Methylation Specific PCR (MS-PCR) was performed according to the method of Herman et al [21].
Briefly, DNA (1 μg) in a volume of 50 μl was denatured by 0.2 M NaOH for 10' at 37°C, then 30 μl of 10 mM hydroquinone (Sigma, St. Louis, USA) & 520 μl of 3 M Na bisulfite, pH 5.0 (Sigma, St. Louis, USA), both freshly prepared were added to each sample. These were then incubated at 50°C/16 h. Modified DNA was purified using the wizard DNA purification Kit (Promega, Madison, WI, USA). Modification was completed by 0.3 M NaOH treatment for 5 min at room temperature, followed by ethanol precipitation, DNA was resuspended in distilled water and stored at -20°C.
PCR amplification
Sequences of primer pair (Biobasic, INC, Canada) used for MSPCR on p16INK4a gene were: unmodified or wild type primers (W), 5' cagaggtggggcggaccgc 3' and 5' cgggccgcggccgtgg 3'; methylated specific primers (M), 5' ttattagagggtggggcggatcgc 3' and 5' gaccccgaaccgcgaccgtaa 3'; unmethylated specific primers (U) 5' ttattagagggtggggtggattgt 3' and 5' caaccccaaaccacaaccataa 3'. PCR products identified by W, M and U primers were 140 bp, 150 bp and 151 bp respectively.
The PCR mixture contained 1 × PCR buffer (16.6 mM ammonium sulphate / 67 mM Tris, pH 8.8 / 6.7 mM MgCl2 / 10 mM 2-mercaptoethanol) dNTPs (each at 2 mM), Primers (100 pmol) and bisulfite treated DNA (~50 ng) or unmodified DNA (50–100 ng) in a final volume of 50 μl. PCR specific for unmodified DNA also included 5% DMSO. Reactions were hot started 95°C for 5' before the addition of 1.5 units of Taq polymerase (ROCHE, GmbH, Germany).
Amplification was carried out in the following conditions, 35 cycles at 95°C (30 sec), 60–65°C (depending on the type of primer pair used) (30 sec), 72°C (30 sec) followed by a final 5 min extension at 72°C. Each PCR product was loaded onto a 2% agarose gel, stained with ethidium bromide and visualized under UV illuminator. DNA from Raji cell line was used as a positive control for methylated alleles of this gene. DNA from normal lymphocyte was used as the control for unmethylated alleles.
Immunohistochemistry
Immunohistochemical detection was performed according to the avidin biotin complex method using the ABC staining kit (Santa Cruz Biotech INC., CA, USA). In brief, 5 μm sections were cut from formalin fixed, paraffin embedded tissue specimens. After treatment with blocking solution (0.03% H2O2 in methanol) to block endogenous peroxidase activity, the antigenic sites were unmasked by means of 3 cycles of 5 minutes microwave irradiation in 10 mM citrate buffer (pH 6.0). Sections were then incubated with the primary antibodies (Neomarkers, Fremont, USA) against p16 (clone MS-887-PO) used at 1:20 dilution, for 2 hours at room temperature. Sections were further incubated with the secondary biotinylated antibody followed by treatment with ABC reagent. The slides were developed using 3-3' diaminobenzidine as the chromogen and counterstained with hematoxylin followed by mounting with DPX.
The IHC results were scored by taking percentage positivity and intensity of staining into account. An intensity score of 0 = No staining, 1 = weak positivity; 2 = moderate positivity & 3 = strong positivity was given.
Statistical analysis
The following statistical tests were applied to analyze the data:
Wilcoxon Signed Rank Test, Chi Square Test and Pearson Correlation Coefficient Test. A probability value of less than 0.05 was considered to be significant.
Results
Paired normal and tumor DNA from 25 patients with pancreatic cancer were examined for the occurrence of p16 genetic alterations.
Homozygous deletion
We separately amplified exon 1, 2 & 3 of p16 using specific primers for homozygous deletion and for SSCP analysis of PCR products. In 3 out of 25 (12%) tumors, we failed to amplify exon 1, 2 and 3 whereas these three exons could be amplified from the corresponding normal pancreatic tissue (Fig 1). Hence, in these three cases there was presumably a homozygous deletion of the p16 gene.
SSCP analysis
To screen for mutation in p16 gene, exon 1, 2 and 3 were analysed by PCR-SSCP. 5 out of 25 cases analysed showed evidence for p16 mutation (exon1, one case; exon 2, four cases) (Fig 2). No mobility shift of exon-3 was found in any sample. DNA samples showing electrophoretic band shift mobility were reamplified and the product purified and used directly for sequencing. DNA sequence analysis of these abnormally migrating SSCP fragment revealed point mutations in 4 out of 5 cases (3 transversions and 1 missense mutation) (Fig. 3). Table 1 summarizes the SSCP and DNA sequencing data for these.
Methylation analysis of p16
Methylation status of p16INK4a gene was evaluated in 25 tumors. A total of 13 (52%) samples showed evidence of promoter methylation. In 11 out of these 13 cases, the gene was completely methylated while in the other 2 cases the gene was partially methylated (Fig. 4).
p16 immunohistochemistry
Loss of p16 protein expression was noted in 18 out of 25 tumors as determined by immunohistochemistry. However, 7 out of 25 (28%) ductal adenocarcinomas stained positive for p16 with weak/focal p16 nuclear staining in 4 and moderate positivity in 3 cases (Fig. 5A). In normal pancreas, p16 nuclear positivity was noted in islets of Langerhans with scattered non-specific cytoplasmic positivity in ductal and acinar cells (Fig. 5B).
Comparison of p16 gene alterations to p16 protein expression
The p16 gene alterations were compared to the p16 protein expression and the results are tabulated in Table 2. In the 18 adenocarcinomas negative for p16 expression, 11 had methylation of the promoter region. The p16 gene was deleted in 3 cases with mutations detected in 2 additional cases. However, two tumors negative for p16 expression did not reveal any of the genetic alterations.
Statistical analysis
These genetic alterations showed a significant correlation with the p16 protein expression (Pearson Correlation Coefficient Test and Chi Square Test, p < 0.01). However, no correlation was found between p16 gene alterations (mutation, deletion and hypermethylation) and age, TNM staging and histological differentiation (Wilcoxon Signed Rank Test and Chi Square Test, p > 0.05).
Discussion
The Rb / p16 tumor suppressive pathway can be abrogated in tumors by inactivation of any of the several members of the pathway [22,23]. For pancreatic carcinoma, this disruption is caused exclusively by inactivation of p16INK4a gene and, only rarely, the Rb gene [24,25]. Inactivation of p16 gene occurs through intragenic mutation, homozygous deletion and methylation associated transcriptional silencing. In the present study, all known mechanisms of p16 inactivation were analysed and we found evidence of p16 inactivation in a high proportion (80%) of the pancreatic tumors examined.
Most of the previous studies regarding pancreatic cancer have reported an elevated frequency of p16 gene alterations in pancreatic cancer derived cell line and xenografts [7,13,26,27]. There have been a few reports on the p16 alterations in tissue specimen of primary pancreatic ductal adenocarcinomas. Our report is the sixth such study and the first from India. A comparison of the result of all the reports with the present study is shown in Table 3. Homozygous deletion could not be detected in two studies by Gerdes et al [14] and Ohtsubo et al [15], but were observed in 35% cases in the study by Zhonghua et al [16]. In our study, we were unable to amplify the p16 exons 1–3 in 12% cases presumably indicating homozygous deletion which is in good agreement with 10% deletion frequency reported by Huang et al [13]. We identified mutation in 4 out of 25 (16%) cases which is consistent with 15% mutation frequency reported by Ohtsubo et al [15]. However, this frequency is lower than 22.5% reported by Gerdes et al [14]. Also, the frequency of mutation was higher in exon 2 than exon 1 (3 cases in exon 2 compared with 1 case for exon 1). This result is consistent with the previous reports [13,14]. Of the 4 points mutations reported here, 3 were transversions and 1 was a missense mutation. Importantly, one out of these four mutations (glutamine → valine at codon 88) was located in the ankyrin repeat III in a highly conserved region believed to form the CDK binding cleft and may have resulted in loss of p16 function [28]. Out of the remaining three mutations, one was in ankyrin repeat I and two others were in ankyrin repeat IV. None of these mutations involved the conserved amino acid and hence may not be crucial to p16 function [28]. On the other hand, two of these mutations resulted in a loss of p16 protein expression and so, it may be postulated that they could have affected the stability and half life of the mRNA or protein. This postulation needs confirmation by further experimental approaches. In 2 out of 18 tumors negative for p16 protein expression, we were unable to detect any genetic alteration in the form of deletion, mutation or promoter hypermethylation. However, presence of a mutation is not completely excluded, as we used PCR-SSCP as a screening tool and it is known from the published studies that this technique detects only about 70% of mutations [19,29].
We have observed hypermethylation to be a major mechanism of p16 inactivation. This frequency is higher than the 27% frequency reported by Gerdes et al [14] and Fukushima et al [30]. Other studies report and even lower frequency of 15% and 3.3% [15,17]. The later study was however performed on microdissected tissue samples using restriction enzyme analysis in contrast to our study on fresh samples using MSPCR. Another explanation for the observed differences in the methylation frequencies could be differences in the ethnicity of the population studied and the role of unknown environmental factors. Such a difference in the methylation profile is reported in the context of hepatocellular carcinomas by Shen et al [31]. In pancreatic cancer cell lines also, the prevalence of p16 promoter methylation has ranged from 18%-38% [27,32]. p16 promoter methylation has also been detected in pancreatic intraductal neoplasia adjacent to pancreatic cancer [28,33,34]. Moreover, the detection of p16 promoter methylation in the pancreatic fluid of patients with pancreatic cancer but not in chronic pancreatitis patients has suggested its potential role as a diagnostic marker in the differentiation of benign and malignant pancreatic disease [35].
Most of the previous studies which have correlated p16 alterations to survival data are limited by the fact that a comprehensive analysis of the p16 genetic alterations is lacking. Bartsch et al [36] found reduced survival in patients with p16 mutations but this study did include the analysis of p16 inactivation by homozygous deletions. Hu et al [37], found longer survival in patients with immunohistochemically p16 positive tumors compared with p16-negative ones, but this difference was not significant. Naka et al [38], identified a significantly longer survival in IHC p16 positive tumors, but the analysis of gene mutations was not examined in these studies. The association of p16 alterations with worse prognosis has also been reported by Gerdes et al [14] and Ohtsubo et al [15]. In the present analysis, although correlation to survival was not possible, p16 alterations did not show any association with the stage or histological grade of the tumor.
Conclusion
Overall, our findings support the observations that p16 / MTS-1 gene alterations play a key role in dysregulated growth of pancreatic cancer and that methylation of the promoter is an important mechanism of its inactivation in Indian patients with pancreatic ductal adenocarcinoma.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SJ carried out the molecular genetic studies and drafted the manuscript. SR participated in its design, coordination and helped to draft the manuscript. MS did DNA sequencing part of the study and coordinated the study. DRB did immunohistochemical part of the study. WJ provided the tissue samples and all the clinical information regarding the patient. All authors have 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 PCR analysis of p16 exon 1, 2 & 3 in normal pancreas and pancreatic tumor. M: Molecular weight marker; N: Normal Pancreas; T: Pancreatic Cancer.
Figure 2 PCR-SSCP analysis of p16 mutations in pancreatic cancer samples. Arrows indicates bands with a mobility shift.
Figure 3 Automated DNA sequencing results to determine point mutations in p16 exon 1 and 2.
Figure 4 Methylation specific PCR of p16 gene in normal pancreas and in pancreatic cancer. M: PCR with primers for methylated p16. U: PCR with primers for unmethylated p16. NL : Normal Lymphocyte DNA used as a negative control for methylation. Raji: Cell line DNA used as a positive control for methylation. MW: Molecular weight marker.
Figure 5 Immunohistochemical staining of p16 in representative cases of (A) pancreatic cancer (B) Normal pancreas (original magnification ×200).
Table 1 p16 sequence changes in pancreatic cancer.
Sample Exon Codon Nucleotide Altered Predicted Product
S-4 1 39 AAC → AAG Asparagine → Lysine
S-10 2 132 CGC → CGG Arginine → Arginine
S-17 2 88 GAG → GTG Glutamine → Valine
S-22 2 139 AGA → ACA Arginine → Threonine
Table 2 Comparison of p16 protein expression to the p16 gene alterations.
Sr. No. p16 protein expression p16 gene alteration
Deletion Mutation p16 promoter methylation
1. - - - + / -
2. - - - + / -
3. - + - - / -
4. - - + - / +
5. + - - + / +
6. - - - + / -
7. - + - - / -
8. - - - + / -
9. - - - + / -
10. + - + - / +
11. - - - + / -
12. - + - - / -
13. - - - + / -
14. + - - + / +
15. - - - + / -
16. - - - - / +
17. - - + - / +
18. - - - + / -
19. + - - - / +
20. + - - - / +
21. - - + / -
22. + - + - / +
23. - - - + / -
24. + - - - / +
25. - - - - / +
+ / + : Completely methylated
- / - : Unmethylated
+ / - : Partially methylated
Table 3 Studies showing p16 alterations in tissue specimen of primary pancreatic ductal adenocarcinomas.
Study Number of cases p16 gene alterations
Homozygous deletion Mutation Methylation
Huang et al (1996) 30 10% 17% NE
Zhonghua et al (2000) 35 35% 20% NE
Moore et al (2002) 34 NE 23% 15%
Gerdes et al (2002) 40 ND 22.5% 27.5%
Ohtsubo et al (2003) 60 ND 15% 3.3%
Present study 25 12% 16% 52%
NE: Not Examined; ND: Not Detected
==== Refs
Torrisani J Buscail L Molecular pathways of pancreatic carcinogenesis Ann Pathol 2002 22 349 55 12483152
Serrano M Hannon GJ Beach D A new regulatory motif in cell cycle control causing specific inhibition of cycling D/CDK4 Nature 1993 366 704 707 8259215 10.1038/366704a0
Cobrinik D Dowdy SF Hinds PW Mittnacht S Weinberg RA The retinoblastoma protein and the regulation of cell cycling Trends Biochem Sci 1992 17 312 315 1412705 10.1016/0968-0004(92)90443-D
De Vos S Miller CW Takeuchi S Gombart AF Cho SK Koeffler HP Alterations of CDKN2 (p16) in Non-small cell lung cancer Genes Chromosomes Cancer 1995 14 164 170 8589032
Hatta Y Hirama T Takeuchi S Lee E Pham E Miller C Strohmeyer T Wilczynski S Melmed S Koeffler H Alteration of p16 gene in-testicular, ovarian and endometrial malignancies J Urol 1995 154 1954 1957 7563391 10.1097/00005392-199511000-00104
Kamb A Gruis NA Weaver-Feldhaous J Liu Q Harshman K Tavtigian SV Stockert B Day RS Johnson BE Skolnick MH A cell cycle regulator potentially involved in genesis of many tumor types Science 1994 264 436 440 8153634
Caldas C Hahn SA daCosta LT Redston MS Schutte M Seymour AB Weinstim CL Hruban RH Yeo CJ Kern SE Frequent somatic mutations and homozygous deletions of p16 (MTSI) gene in pancreatic adenocarcinoma Nat Genet 1994 8 27 31 7726912 10.1038/ng0994-27
Nobori T Miura K Wu DJ Lois A Takabayashi K Carson DA Deletion of cyclin dependent kinase-4 inhibitor gene in multiple human cancers Nature 1994 368 753 756 8152487 10.1038/368753a0
Okajima E Fukuda T Okita S Tsutsumia M Hirao Y Konishi Y In frequent somatic alteration of p16/MTS1 in human primary superficial bladder cancer Cancer Lett 1996 103 227 231 8635161 10.1016/0304-3835(96)04225-5
Pollock PM Pearson JU Hayward NK Compilation of somatic mutation of CDKN2 gene in human cancers: non random distribution of base substitutions Genes Chromosomes Cancer 1996 15 77 88 8834170 10.1002/(SICI)1098-2264(199602)15:2<77::AID-GCC1>3.0.CO;2-0
Merlo A Herman JG Mao L Lee DJ Gabrielson E Burger PC Baylin SB Sidranski D 5' CpG island methylation is associated with transcriptional silencing of tumor suppressor p16 / CDKN2 / MTS1 in human cancers Nat Med 1995 1 686 692 7585152 10.1038/nm0795-686
Baylin SB Herman JG Graff JR Vertino PM Issa JP Alterations in DNA methylation : a fundamental aspect of neoplasia Adv Cancer Res 1998 72 141 196 9338076
Huang L Goodrow TL Zhang SY Klein Szanto AJP Chang H Ruggeri BA Deletion and mutation analysis of the p16/MTS-1 tumor suppressor gene in human ductal pancreatic carcinoma reveals higher frequency of abnormalities in tumor derived cell lines than in primary ductal adenocarcinomas Cancer Res 1996 56 1137 1141 8640773
Gerdes B Ramaswamy A Ziegler A Lang SA Kersting M Baumann R Wild A Moll R Rothmund M Bartsch DK p16INK4a is a prognostic marker in resected ductal pancreatic cancer: an analysis of p16INK4a, p53, MDM2 and Rb Ann Surg 2002 235 51 59 11753042 10.1097/00000658-200201000-00007
Ohtsubo K Watanabe H Yamaguchi Y Hu YX Motto Y Okai T Sawabu N Abnormalities of tumor suppressor gene p16 in carcinoma: Immunohistochemical and genetic findings compared with clinicopathological parameters J Gastroenterol 2003 38 663 671 12898359 10.1007/s00535-003-1119-6
Zhonghua Yi Xue Yi Chuan Xue Yuan Xue Za Zhi Study on the relationship on alterations and expression of p16 gene to pancreatic carcinoma Article in Chinese 2000 17 399 403
Moore PS Orlandini S Zamboni G Capelli P Rigaud G Falconi M Bassi C Lemoine NR Scarpa A Pancreatic tumors: molecular pathways implicated in ductal cancer are involved in ampullary but not in exocrine non-ductal or endocrine tumorigenesis Br J Cancer 2001 84 253 262 11161385 10.1054/bjoc.2000.1567
Sobin LH Wittekind CH TNM classification of malignant tumors 1997 5 Wiley Liss:New York 87 90
Hayashi K PCR-SSCP: a simple & sensitive method for detection of mutation in the genomic DNA PCR Methods Appl 1991 1 34 38 1842918
Bassam BJ Caetano-Anolles Greschoff PM Fast and sensitive silver staining of DNA in polyacylamide gels Anal Biochem 1991 196 80 83 1716076 10.1016/0003-2697(91)90120-I
Herman JG Graff JR Myohanen S Nelkin BD Baylin SB Methylation specific PCR: a novel PCR assay for methylation status of CpG islands Proc Natl Acad Sci USA 1996 93 9821 9826 8790415 10.1073/pnas.93.18.9821
Jiang W Zhang YJ Kahn SM Hollstein MC Santella RM Lu SH Harris CC Montesano R Weinstein IB Altered expansion of cyclin D1 and retinoblastoma genes in human osophageal cancer Proc Natl Acad Sci USA 1993 90 9026 9030 8415648
He J Allen JR Collins VP Allalunis-Turner MJ Godbout R Day RS James CD CDK4 amplification is an alternative mechanism of p16 gene homozygous deletion in glioma cell lines Cancer Res 1994 54 5804 5807 7954404
Seymour AB Hruban RH Redston MS Caldas G Powell SM Kinzler KW Yeo CH Kern SE Allelotype of pancreatic adenocarcinoma Cancer Res 1994 54 9761 64
Barton CM Mckie AB Hogg A Bia B Elia G Phillips SMA Ding SF Lemoine NR Abnormalities of the RB1 and DCC tumor suppressor gene: uncommon in human pancreatic adenocarcinoma Mol Carcinog 1995 13 61 69 7605581
Rozenblum E Schutte M Goggins M Hahn SA Panzer S Zahurak M Goodman SN Sohn TA Hruban RH Yeo CJ Kern SE Tumor suppressive pathways in pancreatic carcinoma Cancer Res 1997 57 1731 34 9135016
Schutte M Hruban RH Geradts J Maynard R Hilgers W Rabindran SK Hahn SA Moskaluk CA Waldhoff IS Schmiegel W Baylin SB Kern SE Herman JG Abrogation of the Rb/p16 tumor suppressor pathway in virtually all pancreatic carcinomas Cancer Res 1997 57 3126 3130 9242437
Yarbrough WG Buckmire RA Bessho M Liu ET Biologic and biochemical analysis of p16INK4a mutations from primary tumors JNCI 1999 91 1569 1574 10491434 10.1093/jnci/91.18.1569
Ravnik-Glavac M Glavac D Dean M Sensitivity of single strand conformation polymorphisms and heteroduplex method for mutation detection in the cystic fibrosis gene Hum Mol Genet 1994 3 801 807 7521710
Fukushima N Sato N Ueki T Rosty C Walter KM Wilentz RE Yeo CJ Hruban RJ Goggins M Aberrant methylation of preproenkephalin and p16 genes in pancreatic intraepithelial neoplasia and pancreatic ductal adenocarcinoma Am J Pathol 2002 160 1573 1581 12000709
Shen L Ahuja N Shen Y Habib NA Toyota M Rashid A Issa JPJ DNA methylation and environmental exposures in human hepatocellular carcinoma JNCI 2002 94 755 761 12011226
Ueki T Toyota M Sohn T Yeo CJ Issa JPJ Hruban RH Goggins M Hypermethylation of multiple genes in pancreatic adenocarcinoma Cancer Res 2000 60 1835 1839 10766168
Wilentz RE Geradts J Maynard R Offerhaus GJA King M Goggins M Yeo CJ Kern SE Hruban RH Inactivation of p16INK4a tumor suppressor gene in duct lesions: loss of intranuclear expression Cancer Res 1998 58 4740 4744 9788631
Gerdes B Ramaswamy A Kersting M Ernst M Lang S Schuermann M Wild A Bartsch DK p16INK4a alteration in chronic pancreatitis – indicator for high risk lesions for pancreatic cancer Surgery 2001 129 490 497 11283541 10.1067/msy.2001.112071
Klump B Hsieh CJ Nehls O Dette S Holzmann K Kiesslich I Jung M Sinn U Ortner M Porschen R Gregor M Methylation status of p14ARF and p16INK4a as detected in pancreatic secretions Br J Cancer 2003 88 217 222 12610506 10.1038/sj.bjc.6600734
Bartsch D Shevlin DW Callery MP Norton JA Wells SA JrGoodfellow PJ Reduced survival in patients with ductal pancreatic adenocarcinoma associated with CDKN2 mutation J Natl Cancer Inst 1996 88 680 682 8627645
Hu YX Watanabe H Ohtsubo K Yamaguchi Y Ha A Okai T Sawasu N Frequent loss of p16 expression and its correlation with clinicopathological parameters in pancreatic carcinoma Clin Cancer Res 1997 3 1473 1477 9815833
Naka T Kobayashi M Ashida K Toyota N Kaneko T Kaibara N Aberrant p16INK4 expression related to clinical stage and prognosis in patients with pancreatic cancer Int J Oncol 1998 12 1111 1116 9538137
|
15985168
|
PMC1185532
|
CC BY
|
2021-01-04 16:03:27
|
no
|
BMC Gastroenterol. 2005 Jun 28; 5:22
|
utf-8
|
BMC Gastroenterol
| 2,005 |
10.1186/1471-230X-5-22
|
oa_comm
|
==== Front
BMC GenetBMC Genetics1471-2156BioMed Central London 1471-2156-6-351596084710.1186/1471-2156-6-35Research ArticleTree measures and the number of segregating sites in time-structured population samples Forsberg Roald [email protected] Alexei J [email protected] Jotun [email protected] Bioinformatics Research Center (BiRC) and Department of Genetics and Ecology, University of Aarhus, Arhus, Denmark2 Department of Zoology, University of Oxford, Oxford, England3 Department of Statistics, University of Oxford, Oxford, England2005 16 6 2005 6 35 35 19 10 2004 16 6 2005 Copyright © 2005 Forsberg et al; licensee BioMed Central Ltd.2005Forsberg 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
Time-structured genetic samples are a valuable source of information in population genetics because they provide several correlated observations of the underlying evolutionary processes. In this paper we study basic properties of the genetic variation in time-structured samples as reflected in the genealogies relating individuals and the number of segregating sites observed. Our emphasis is on "measurably evolving populations" i.e. populations from which it is possible to obtain time-structured samples that span a significant interval of evolutionary time.
Results
We use results from the coalescent process to derive properties of time-structured samples. In the first section we extend existing results to attain measures on coalescent trees relating time-structured samples. These include the expected time to a most recent common ancestor, the expected total branch length and the expected length of branches subtending only ancient individuals. The effect of different sampling schemes on the latter measure is studied. In the second section we study the special case where the full sample consists of a group of contemporary extant samples and a group of contemporary ancient samples. As regards this case, we present results and applications concerning the probability distribution of the number of segregating sites where a mutation is unique to the ancient individuals and the number of segregating sites where a mutation is shared between ancient and extant individuals.
Conclusion
The methodology and results presented here is of use to the design and interpretation of ancient DNA experiments. Furthermore, the results may be useful in further development of statistical tests of e.g. population dynamics and selection, which include temporal information.
==== Body
Background
Time-structured genetic samples
Genetic samples obtained over several points in time are a valuable source of information in population genetics because they provide several correlated observations of the underlying evolutionary processes.
These time-structured samples can be separated into two qualitatively different groups. Firstly, samples may be taken over such a short evolutionary time that the occurrence of mutations between sampling points can be ignored. Samples of this type have a long standing history in the study of the process of drift and selection via observations of allele frequencies (see e.g. [1]). Secondly, time-structured samples may be obtained over intervals of evolutionary time that are long enough for mutation to become a relevant force in shaping the diversity between samples from different time points. To reflect the fact that the latter type of samples are capable of showing new variation arising, Drummond et al. used the term "Measurably Evolving Populations" (MEP) to describe populations from which biological and technological constraints allow samples of this type to be obtained [2]. Such measurably evolving populations arise from two principal sources, namely rapidly evolving microorganisms e.g. [3] and well characterised vertebrate subfossil material from which ancient DNA can be reliably amplified e.g. [4-6].
Besides a report by Nordborg [7], population genetic studies of MEPs have mostly focused on the construction and use of models for population genetic inference that incorporate the time structure of the data [8-12]. However, our interest here is not that of inference, but rather to study basic properties of time-structured samples obtained from measurably evolving populations. To this end, we use a simple model of a constant sized panmictic population of haploids and base our results on the standard coalescent process [13,14].
The paper consists of two parts. In the first part we use recursions to derive results for various measures on coalescent trees connecting time-structured samples having any time-structure. In particular, we focus on the effect that different sampling schemes have on the expected length of the branches in the tree upon which mutations can arise that are unique to the ancient individuals of the sample. This is a measure of the expected amount of genetic variation unique to the ancestral material and as such of intrinsic interest to the design of many ancient DNA studies, for example when the objective is to discover unique ancient haplotypes. In the second part, we study the case where the sample has a time-structure of only two time-points, one consisting of a number of contemporary extant individuals and one consisting of a number of contemporary ancient individuals. For this case we obtain the probability distributions of the number of segregating sites where the mutation is observed only in ancient lineages and of the number of segregating sites where the mutation is shared between ancient and extant lineages. Using these results we study the number of ancient samples needed to observe at least one unique or one shared mutation as a function of population parameters.
The results presented here should be of particular interest to studies of ancient DNA and the design of ancient DNA sampling schemes.
Notation
Consider the evolution of a haploid population comprising N individuals. We assume selective neutrality, no recombination and a Wright-Fisher model of propagation where each individual chooses its parent independently and at random from the individuals of the previous generation [15,16]. Time is measured in units of N generations, and the population is assumed to remain at a constant size, which is sufficiently large that the diffusion approximation of the coalescent process applies [13]. The complete sample is produced by sampling the population serially over a sampling time-interval consisting of n sampling points (Figure 1) each contributing one new individual to the process. Each sampling point is associated with a sampling time (τi) and hence the temporal configuration of the sample is completely determined by the ordered vector of times τ = (τ1,..., τn) with τ1 ≤ τ2 ≤ ⋃ ≤ τn. Sampling points and times are enumerated from the present going backwards. We let τ1 = 0 by definition, and define, τi,J = τi - τJ, i > j, as the difference in sampling times between sample pair i, j. Note that groups of samples may be taken at the same time so that τk = τk + 1⋃ = τk+g-1, for a group of size g.
Figure 1 Time-structured samples. An illustration of a time-structured sample. Time proceeds backward as indicated by the arrow at the left. Individuals sampled at five sampling points constitutes the full sample. With each sampling point is associated a sampling time (τ) and the first sampling time is zero by convention. Three samples are ancient (open circles) and constitute a sub-sample Q. Two individuals are extant (full circles) i.e. they are both sampled at time zero (τ1 = τ2 = 0). Hence, interval one has a length of zero time units and is not depicted. The sampling intervals are numbered by i = 1 ... 5 and interval five is the terminal interval where after no new samples are included. At the end of the terminal interval a most recent common ancestor (MRCA) is found and this time point is denoted TMRCA. The total number of lineages at time t in interval i is a stochastic process with an associated variable Ai(t). On the figure is given the total number of lineages at the beginning of each interval (Ai(0)) and for interval four the number at a specific time point t. Likewise, the number of lineages subtending leaves exclusive to Q at time t in interval i is a stochastic process with an associated variable . On the figure is given the number of lineages subtending leaves exclusively in Q at the beginning of each interval () and for interval four the number at a specific time point t. Mutations occurring in ancestors which subtend leaves exclusive to Q (branches represented by dotted lines) will create segregating sites where the mutation is unique to the individuals in Q i.e. the ancient individuals.
The standard coalescent
By a continuous-time approximation, Kingman has shown that for a contemporary sample (τi = 0, ∀i) a process termed the coalescent can describe the genealogical relationships of the individuals in the sample [13]. Kingman's result applies to the Wright-Fisher model and a broad class of other reproductive models that share a common set of requirements [13]. Briefly, the standard coalescent is a process describing the death of lineages through coalescence events. Some fundamental results concerning the standard coalescent which we use in the following are given here:
The waiting times Wn, Wn-1,..., W2 between successive coalescent events are exponentially distributed as
when time is measured in units of N generations [17].
Let time increase into the past and let the number of distinct line ages at time t, {A(t), t ≥ 0} be given by the death process described by the coalescent. Following [18] we have that given a lineages were sampled at time 0, the conditional probability distribution of the number of remaining lineages at time t is
for 2 ≤ b ≥ a, and for the case b = 1
where a[j] = a(a - 1)...(a - j + 1), and a(j) = a(a + 1)...(a + j - 1).
In the following it will be of interest to study the number of ancestors of sub-samples which in this context are comprised of ancient lineages. Let Q be a sub-sample of the full sample consisting of all ancient individuals and let Qc be the complement of Q which contains the extant individuals. It is our interest to study the number of ancestral lineages at some time t that subtend leaves in the tree which are exclusive to Q, {A(Q)(t), t ≥ 0}, as mutations occurring on these branches will be unique to Q, i.e. be found solely in ancient samples (Figure 1). However, this is equivalent to recording the total number of ancestral lineages at time t, A(t), and the number of ancestral lineages that subtend one or more leaves (not necessarily exclusively) in Qc { (t), t ≥ 0}, as the two are related by A(t) = A(Q) (t) + (t). The bivariate process {A(t), (t)} for a sample of contemporary individuals has been studied extensively by Saunders et al. [19], and from this we extract results for the conditional distribution of the bivariate process {A(t), A(Q) (t)} using the above relation
where Pr{A(t) = b|A(0) = a} is given by (2), and
The coalescent process of time-structured data
When samples have time-structure (τi > 0 for some i) the simple death process of the coalescent is replaced by a series of death processes, interrupted at specified points in time by new lineages entering the process (Figure 1). We note that this would correspond to a birth-death process of lineages if sampling events were random rather than known. However, lacking knowledge about the properties of the sampling process we restrict ourselves to condition on known sampling times. Thus, this serial coalescent process can be modelled on the basis of the standard coalescent process, by the following algorithm: At the second sampling point (no coalescence in first interval) a contemporary coalescent is initiated with two individuals; this process is continued for τ3,2 time units; at τ3 time units another individual enters the process; again a coalescent process is continued for τ4,3 time units until τ4 where yet another individual is added and so forth until τn is reached and the last individual is included; from here the process continues as a standard coalescent process (as no new lineages will be added) until the most recent common ancestor (MRCA) of the sample is reached. Therefore, we refer to this last interval as the termination interval (Figure 1). The lineage number at each sample point is a stochastic variable with a distribution that must be tracked through the sampling interval. It is clear that the following results are all conditional on the temporal configuration of the sample (τ) and consequently this dependence is suppressed from this point.
The mutation process
We shall be concerned with the sampling properties of nucleotide sequences. Hence, we adopt the infinitely many sites model [17], assuming that a single site experiences at most one mutational event so that every mutation that arises is represented in the sample. Furthermore, we assume that mutations are generated by a Poisson process with parameter . The compound parameter θ is given by θ = 2N μ where μ is the mutation rate per sequence per generation.
As branch lengths are measured in units of N generations, the expected number of segregating sites generated over a tree (or sub-section of a tree) with total branch length l, is .
Results
Measures on coalescent trees of time-structured samples
In this section we use the theory presented above to derive recursions describing various measures on coalescent trees of time-structured data.
Number of lineages through the sampling intervals
Let, {Ai(t), 0 ≥ t ≥ τi+1,i} be a stochastic variable representing the number of lineages in interval i, t time units after τi (see Figure 1). Consecutive death processes are related by single birth events, so that
The probability distribution of Ai(t) is found by summing over all the possible lineage numbers at the start of the interval permitted by the sample configuration
Notice that the last term is given by (2), and that Pr{Ai(0) = a}, the probability of observing a lineages at the start of the interval, is given by (5,7) and:
for i = 1, 2
for i > 2
Pr{Ai(0) = a} = Pr{Ai-1 (τi,i-1) = a - 1}.
Lastly, let {Ai} be the marginal lineage number in the interval [τi, τi+1) with probability
Time to the most recent common ancestor
From (1) and the above it follows that the expected time to the MRCA of a time structured sample, (TMRCA), is given by
Total branch length of the genealogy
The above results can also be applied to produce the expected total branch length for time-structured data (B(tot)), given by
where Bi is the branch length added in each of the n - 1 time-intervals that comprise the total sampling interval,
and B(term)) is the branch length added over the last interval
where the latter term corresponds to the expected total branch length of the tree relating a sample of size a in a standard coalescent process [17].
Number of lineages subtending leaves exclusive to a sub-sample
Let the function δ(Q, i) be a membership function for the sub-set Q so that
Let, { (t, 0 ≤ t ≤ τi+1,i} be a death process representing the number of lineages subtending leaves exclusively in Q, t time units after sampling individual i. Consecutive death processes ( (t), (t)) are related by single birth events, so that
The joint probability distribution over Ai(t) and (t) is found by summing over all the possible lineage configurations at the start of the interval that are permitted by the sample configuration and the structure of Q
where, , and where Pr{Ai(0) = a, (0) = d}, is given by (13,14) and (15). The unconditional probability distribution over (x) is given by
Lastly, let {} be the marginal probability of the number of lineages subtending leaves exclusively to Q over the interval [τi, τi+1). Consequently, is given by integrating over the interval length
The expected length of branches subtending leaves exclusive to a sub-sample
Let B(Q) denote the expected total length of all branches subtending leaves only in Q. Similar to (10), we have that
where {} is the expected branch length added over interval i, given by
and the expected branch length over the last interval {} is given by
which corresponds to conditioning on the number of different lineage types present at the initiation of the termination interval and then weighing the expected branch length over the contemporary coalescent process by the probability of observing a given number of ancestors exclusive to Q in the individual coalescent intervals.
Effect of sampling scheme on the expected number of segregating sites unique to ancient samples
Particular constraints on studies involving ancient DNA are the maximum age at which genetic material can be obtained and the number of ancient DNA samples obtainable. The former constraint arises due to problems with dating techniques, and more importantly, with the time-dependent degradation of DNA, whereas the latter constraint arises because material containing DNA may be hard to obtain or due to financial constraints on e.g. carbon dating of fossil material. Therefore, the planning of an ancient DNA study may consist of the construction of a sampling strategy which distributes a given number of ancient samples within a fixed maximum time-interval. A possible objective of such a study may be to maximise the expected number of segregating sites where the mutation is unique to the ancient samples i.e. maximising the expected length of branches subtending leaves exclusive to the ancient samples (see equation 18). This would, for example, be the case in studies which aim at maximising the number of unique ancient haplotypes observed in order to infer past mixing events of populations such as humans and Neanderthals [6,7,20] or patterns of e.g. colonisation from an ancient population [21]. Assuming that a number of contemporary extant individuals are sampled, two strategies for sampling ancient individuals appear as natural candidates. A stepwise sampling strategy where ancient samples are distributed evenly over the available maximum time-interval and a "bouquet" strategy where all ancient samples are sampled at the maximum age (Figure 2).
Figure 2 Sample types. The two sample types considered in the paper. The sampling time of individual i is indicated by τi and a total of eight individuals constitute the full sample. Of these eight, four constitute a contemporary extant sub-sample (full circles) and four constitute the ancient sub-sample (open circles). In the stepwise sampling scheme the ancient samples are evenly distributed over the maximum length of the sampling interval (τmax) which corresponds to τ8. In contrast, the ancient samples in the bouquet sampling scheme are all taken at the maximum time attainable.
In Figure 3 we plot the expected total length of branches subtending only ancient leaves as a function of the maximum length of the total sampling interval (τmax) and the sampling strategy. When τmax is zero, lineages from the ancient sub-sample Q and Qc are interchangeable in the coalescent process and the major part of the tree will contain lineages which subtend leaves in both Q and Qc. Initially, the expected branch length for the bouquet strategy increases rapidly with increasing τmax since this leaves time for extant samples to coalesce within themselves before the inclusion of the ancient sample and thus interfere less with the part of the tree subtending only ancient lineages. This means that in this part of the parameter space, large gains in uniquely ancient genetic diversity can be achieved by a relatively small elongation of the sampling interval. At higher values of τmax the probability of observing more than one extant lineage at the time of the ancient sample is minimal and thus the expected branch length goes towards the asymptotic value of including a single extant lineage. For the step-wise strategy the initial increase in expected branch length is less marked because the ancient individuals in this case are younger and share more branches with the ancestors of the extant individuals for a given τmax. However, as τmax increases the number of surviving lineages after each sampling interval approaches one and the expected branch length asymptotes towards a linearly increasing function.
Figure 3 Effect of sampling type on branch length. The expected total length of branches subtending only ancient individuals as a function of the maximum attainable sampling time (τmax). The expected total length (E(B(Q))) is shown for the stepwise and the bouquet sample types depicted in Figure 2. The number of extant individuals is 30 and the number of ancient individuals is 10. The maximum attainable sampling time is varied causing a sample type specific pattern of increase in the expected branch length (see text).
When the objective is to maximise the expected length of branches subtending only ancient lineages the bouquet strategy may intuitively seem the obvious choice as this maximises the total age of ancient material in the sample. However, Figure 3 shows that the bouquet sampling strategy only outperforms the stepwise strategy until a certain value of τmax is reached. Above this value it is advisable to distribute ancient samples evenly over the maximum attainable time-interval rather than to sample all ancient individuals as old as possible. This may be of particular relevance to species with small effective population sizes (τmax large), like top-level predators (saber-toothed tigers, lions and bears).
The case of two sample points
In this section we focus on the simpler case with only two sampling points i.e. the bouquet sampling strategy presented above (Figure 2). Our interest is to study the distribution of segregating sites between different classes of lineages. Under the infinite sites model, the number of segregating sites in a contemporary sample of individuals has been extensively studied by [17]. Furthermore, the transient distribution of segregating sites between two time points has been studied by [22].
Suppose that a total of n individuals are sampled. Of these, b ancient individuals are sampled at the same time and constitute the sub-sample Q of interest. The remaining n - b individuals are all extant i.e. sampled at time 0. Assume that there are a ancestors left of the extant individuals at time τn-b, i.e. when the ancient individuals included in Q are reached. Since no time passes (τn-b+1 = τn-b+2 = ... = τn) during the sampling of the b lineages in Q, we have that (0) = b and (0) = a.
The coalescence process in the terminal interval now proceeds through a + b - 1 intervals of waiting. Let tk denote the beginning of the kth coalescent interval {k = 0,1...a + b - 2}.
The total number of mutations occurring on the lineages through the coalescent process in the terminal interval (U(tot)) will give rise to segregating sites that can be divided into three categories: mutations which occur on lineages subtending only in Q and which therefore create segregating sites where the mutant is unique to the ancient individuals, (U(Q)), mutations occurring on lineages subtending only in Qc causing segregating sites where the mutation is unique to the extant individuals (U(E)), and mutations occurring on lineages subtending in both Q and Qc giving rise to segregating sites where the mutation is shared between the ancient and the extant individuals (U(S)); U(tot) = U(E) + U(S) + U(Q). In the following we assume that the ancestral state of the genetic element under study can be accurately inferred. In the case where a segregating site represents a mutation that is found in all individuals of one group (ancient or extant) and absent from all individuals in the other group, this knowledge is required to determine whether the segregating site represents a mutation that is unique to all ancient (in U(Q)) or to all extant individuals (in U(E)).
Segregating sites found only in ancient individuals
Through the coalescent process, lineages which subtend leaves exclusively in Q (ancient leaves) may die in two events: as two of these lineages coalesce, or as one of these coalesce with a lineage subtending leaves in Qc. Let there be ( (tk) = c) ancestors left subtending exclusively ancient leaves at the beginning of the kth coalescent interval.
The probability of losing a lineage subtending leaves exclusively in Q over the kth coalescent interval is
and the probability of keeping a lineage subtending leaves exclusively in Q over the kth coalescent interval is
Let be the number of mutations unique to the lineages in Q that occur in the kth coalescent interval. Following [23] we then have the probability
The probability distribution of the number of mutations unique to the lineages in Q which occur from interval k until the MRCA {} is then given by the recursion
and the full unconditional probability of seeing m mutations is found by summing over all possible start values of a
Segregating sites arising in terminal interval found only in extant individuals
It is evident that the probability distribution of the number of segregating sites where the mutation is unique to extant individuals can be found by exchanging Q and Qc in the above.
Segregating sites shared between ancient and extant individuals
Segregating sites where the mutation is shared between ancient and extant samples occur on ancestral lineages that subtend leaves in both the sub-sample and the complement. Past the last sampling interval, let { (t), t ≥ 0} represent the number of ancestors present at time t which subtend leaves in both Q and Qc. In a coalescent event, the number of lineages in (t) may be reduced as two of these lineages coalesce, it may remain constant, or it may be increased by one as a lineage in (t) coalesces with a lineage in (t).
Given that interval k is initiated with ( (tk) = s) lineages subtending leaves in both Q and Qc and ( (tk) = c) ancestors left subtending exclusively ancient leaves, we have that over the kth interval:
The probability of losing a lineage subtending leaves in both Q and Qc is
The probability of the number of lineages subtending leaves in both Q and Qc remaining constant while a lineage subtending leaves exclusively in Q is lost is
The probability of the number of lineages subtending leaves in both Q and Qc remaining constant whilst the number of lineages subtending leaves exclusively in Q is kept constant is
and the probability of gaining a lineage subtending leaves in both Q and Qc is
Let be the number of mutations shared between the sub-sample and the complement that occur in the kth coalescent interval. We then have the probability
The number of shared mutations which occur from interval k until the MRCA {} is then given by the recursion
and the full unconditional probability of seeing m mutations is found by summing over all possible start values of a
Total number of segregating sites arising in terminal interval
The total number of segregating sites created in the terminal interval can be found by treating all lineages as one sub-sample in either of the recursions given above.
Applications
In Figures 4 and 5 we plot two measures as a function of θ and the maximum attainable sampling time, τmax. The first of these is the number of ancient individuals that must be included at the second sample point to reach a probability greater than 0.95 of there being one or more segregating sites where the mutation is unique to the ancient individuals. Likewise, the latter is the number of ancient individuals that must be included at the second sample point to reach a probability greater than 0.95 of there being one or more segregating sites where the mutation is shared between ancient and extant isolates. Both are decreasing functions of θ which scales branch lengths in terms of mutations. At low values of θ the probability of seeing any mutations in the sample is low and a large amount of ancient individuals are required for any unique ancient or shared mutations to occur. The expected number of extant ancestors at the inclusion of the ancient sample decreases with τmax and thus the distribution of coalescent trees generated in the terminal interval becomes less dominated by lineages subtending extant leaves. As a consequence, the number of ancient individuals needed to produce any segregating sites where the mutation is unique to the ancient individuals also decreases with τmax (Figure 4) and the number needed to see any shared segregating sites increases with τmax (Figure 5). For both functions, the effect of increasing the number of extant samples can only be seen at low values of τmax (results not shown). This is due to the nature of the coalescent process where the coalescent intensity increases as a function of lineage number causing the additional extant lineages to coalesce before they can affect the substitution process in the terminal interval.
Figure 4 Segregating sites unique to ancient individuals. The number of ancient lineages required to ensure a probability higher than 0.95 of seeing at least one segregating site where the mutation is unique to ancient individuals. Samples are taken by the bouquet sampling strategy (see Figure 2) and the number of extant samples is fixed at 30. For computational reasons there is an upper limit of 50 ancient individuals. The number of lineages required is shown as a function of θ = 2N μ and the maximum attainable sampling time τmax. Contour lines of the required number of individuals are shown at the base of the plot.
Figure 5 Segregating sites shared by ancient and extant individuals. The number of ancient lineages required to ensure a probability higher than 0.95 of seeing at least one segregating site where the mutation is shared by ancient and extant individuals. Samples are taken by the bouquet sampling strategy (see Figure 2) and the number of extant samples is fixed at 30. For computational reasons there is an upper limit of 50 ancient individuals. The number of lineages required is shown as a function of θ = 2N μ and the maximum attainable sampling time τmax. Contour lines of the required number of individuals are shown at the base of the plot.
Typical parameter values observed in ancient DNA studies are: fragment length ~ 500 nucleotides, substitution rate per nucleotide site ~ 1·10-7(10-6 – 10-8 range for mitochondrial DNA) e.g. [4,24]. Since the parameters θ and τmax are measured in units of generations over the effective population size (Ne), knowledge of Ne and the maximal number of generations spanning the sampling interval (gmax) are needed to relate these parameter values to the plots. In Table 1 we list several combinations of the parameters Ne and gmax and their associated values of θ and τmax along with values from Figure 4 and Figure 5. To interpret the entries in Table 1 consider that in ancient DNA studies the maximal sampling time is often in the range 30.000 ~ 50.000 years. Dependent on the generation time of the studied organism the maximal number of generations spanning the sampling interval is then given by: .
Table 1 The effect of sample number under different evolutionary scenarios. Result in this table are based on typical parameter values from the ancient DNA literature and a bouquet sampling scheme (see text). Fixed parameter values are: fragment length = 500 nucleotides, substitution rate per nucleotide site = 1·10-7, number of extant samples = 30. Given these values any combination of the population size parameter (Ne) and the maximal number of generations spanning the sampling interval (gmax ) can be translated to a value of the population parameter (θ) and the maximum length of the sampling interval (τmax ) which are measured in units of generations (see text for definition). For various such parameter combinations we here list the associated values of θ, τmax, and the calculated number of ancient samples needed to have a greater than 95% chance of seeing at least one segregating site where the mutation is unique to the ancient individuals (nu) and the number of ancient samples needed to have a greater than 95% chance of seeing at least one segregating site where the mutation is shared between ancient and extant individuals (ns).
Ne θ gmax τmax nu ns
5000 0.5 1000 0.2 > 50 > 50
- - 5000 1.0 > 50 > 50
- - 10000 2.0 > 50 > 50
- - 30000 6.0 > 50 > 50
- - 50000 10.0 > 50 > 50
10000 1.0 1000 0.1 > 50 > 50
- - 5000 0.5 > 50 > 50
- - 10000 1.0 > 50 > 50
- - 30000 3.0 42 > 50
- - 50000 5.0 39 > 50
30000 3.0 1000 0.03 44 3
- - 5000 0.17 23 5
- - 10000 0.33 15 7
- - 30000 1.0 9 > 50
- - 50000 1.67 7 > 50
50000 5.0 1000 0.02 26 2
- - 5000 0.1 16 2
- - 10000 0.2 11 3
- - 30000 0.6 7 6
- - 50000 1.0 5 9
For a specific example consider the recent effort to sequence ancient human mitochondrial sequences from ~ 20.000 years ago [25]. If we assume an effective human population of ~ 10.000 and a generation time of 20 years [20], we have that for humans gmax ~ 0.1 at present. Given the population and mutation parameters chosen here, we see from Table 1 that a large number of ancient human samples (> 50) would be needed to ensure the finding of segregating sites where the mutation is unique to ancient humans. It is thus not surprising that the two ancient human mitochondrial haplotypes inferred in a study by Caramelli et al. [25] are both found to be circulating in the present day human population, and it is questionable whether unique ancient haplotypes can be obtained without a substantial elongation of either the sampling interval or the region sequenced.
Discussion
The use of time-structured population samples has a long standing tradition in the study of rapidly evolving microorganisms. Here, the temporal component of data sets has allowed researchers to explore complex hypotheses concerning e.g. the action of selection, host-parasite co-evolution and the evolutionary response of parasites to drugs (see [2] for a review). With the advent of ancient DNA technologies it has recently become possible to obtain time-structured genetic samples that span a large number of generations from multi-cellular organisms also. As an example of the potential that such time-structured samples holds for resolving long standing questions in evolutionary biology, ancient DNA data have been used to clarify the genetic relationship between our own species and the Neanderthals [6,7], to estimate absolute rates of nucleotide substitution [4] and to infer patterns of population demography in relation to e.g. climatic changes [21].
Inspired by these technological advances, our motivation was to elaborate existing results from population genetics to the case where samples have a time-structure and to relate the results to problems faced by experimentalists.
In the first section of the paper, we have presented recursions for various tree measures and applied these to explore the effect of sampling scheme on the expected number of segregating sites where the mutation is found only in ancient individuals. Besides guiding the choice of sampling strategy, these results allow experimentalists to consider whether any genetic variation previously unseen is likely to be discovered by including ancient samples, and if so, to compare the observed number of segregating sites where the mutation is unique to ancient individuals to the expected number under a neutral model. Other applications may be relevant for the construction of time-structured data sets, depending on the objective. As an example, studies may have the objective of estimating the rate of neutral substitution. For such an estimate, substitutions occurring over the sampling interval constitute information and substitutions occurring over the terminal interval constitute noise. Thus, it is desirable to construct sampling schemes which maximise the ratio of the expected length of branches in the sampling interval over the expected length of branches in the terminal interval, and the effect of e.g. sampling schemes on this measure could also be explored using the results from this paper.
In the second part of the paper we studied the simpler case where samples are taken at only two time-points. This simplification allowed us to present results concerning the full probability distribution of the number of segregating sites where the mutation is unique to the ancient individuals and of the number of segregating sites where the mutation is shared between ancient and extant individuals. As an application of these we have shown the number of ancient lineages needed for a sample to have a high probability of showing at least one segregating site where the mutation is unique to the ancient material or at least one segregating site where the mutation is shared between ancient and extant material. These results should be of interest to experimentalists who wish to evaluate the amount of information likely to be obtained from a given time-structured sample.
In the applications presented we have focused on moderate sample sizes. For large samples (>> 50) computational problems arise due to the computational load of the recursions and problems with computer representation of very small probabilities. If larger sample sizes are to be considered these problems must be circumvented. This could be done by simplification of recursions, the use of approximative calculations and the derivation of probabilities through Monte Carlo procedures [7]. Smaller sample sizes may, however, be sufficient to extract general properties as the effect of increasing sample size on e.g. total branch length in the coalescent quickly diminishes.
The results concerning tree measures are applicable to all time-structured samples. However, the results on segregating sites are derived under the assumption of the infinite sites model which is invalid for data sets taken over long time-intervals from rapidly evolving genetic elements such as the mito-chondrial control region or genomic regions of RNA viruses.
For simplicity we have only considered the case of a constant population, but if the demographic function is known, the effect of varying population size could be accommodated into the tree measure recursions given here via the generalisation of (2) given in [26]. This is also true for the tree measures concerning sub-samples, since the result in (4) is purely combinatorial and thus independent of the demographic function. For the results concerning the number of segregating sites, an extension to varying population size is not possible with the present approach as it relies on the independence between consecutive coalescent intervals.
Conclusion
The focus of this paper has been on sampling issues and the production of results that may be of use in the design and interpretation of experiments including time-structured genetic samples, particularly ancient DNA experiments. However, given that the inclusion of time-structure increases the statistical power in evolutionary inference, it is our hope that the results presented may also be useful in the pursuit of statistical tests, in the flavour of Tajima's D [27], which are applicable to time-structured samples.
Authors' contributions
RF conceived the idea for this manuscript, provided the majority of the results, implemented results into computer code, performed the experiments and wrote the manuscript. AJD helped in the discussion of results, the implementation into software and the design of experiments. JH aided in conceiving the idea for the manuscript, discussing results and providing mathematical results. All authors read and approved the final manuscript.
Acknowledgements
This study acknowledges support from the following grants: The Danish Natural Science Research Council grants 21-02-0206 and 51-00-0392 (RF), The National Institute of Health, USA grant 1-R01-GM60729-01 (RF), and EPSRC grant HAMJW and MRC grant HAMKA. Carsten Wiuf is thanked for useful discussions.
==== Refs
Williamson EG Slatkin M Using Maximum Likelihood to Estimate Population Size From Temporal Changes in Allele Frequencies Genetics 1999 152 755 761 10353915
Drummond AJ Pybus OG Rambaut A Forsberg R Rodrigo AG Measurably evolving populations Trends in Ecology and Evolution 2003 18 481 488 10.1016/S0169-5347(03)00216-7
Forsberg R Oleksiewicz MB Petersen AM Hein J Botner A Storgaard T A molecular clock dates the common ancestor of European-type porcine re productive and respiratory syndrome virus at more than 10 years before the emergence of disease Virology 2001 289 174 179 11689039 10.1006/viro.2001.1102
Lambert DM Ritchie PA Millar CD Holland B Drummond AJ Baroni C Rates of evolution in ancient DNA from Adelie penguins Science 2002 295 2270 2273 11910113 10.1126/science.1068105
Leonard JA Wayne RK Cooper A Population genetics of ice age brown bears Proc Natl Acad Sci USA 2000 97 1651 1654 10677513 10.1073/pnas.040453097
Krings M Stone A Schmitz RW Krainitzki H Stoneking M Paabo S Neandertal DNA sequences and the origin of modern humans Cell 1997 90 19 30 9230299 10.1016/S0092-8674(00)80310-4
Nordborg M On the probability of Neanderthal ancestry Am J Hum Genet 1998 63 1237 1240 9758610 10.1086/302052
Drummond AJ Nicholls GK Rodrigo AG Solomon W Estimating mutation parameters, population history and genealogy simultaneously from temporally spaced sequence data Genetics 2002 161 1307 1320 12136032
Fu YX Estimating mutation rate and generation time from longitudinal samples of DNA sequences Mol Biol Evol 2001 18 620 626 11264414
Rodrigo AG Felsenstein J Crandall KA Coalescent approaches to HIV-1 population genetics The Evolution of HIV 1999 Baltimore: Johns Hopkins University Press 233 272
Rodrigo AG Shpaer EG Delwart EL Iversen AK Gallo MV Brojatsch J Hirsch MS Walker BD Mullins JI Coalescent estimates of HIV-1 generation time in vivo Proc Natl Acad Sci USA 1999 96 2187 2191 10051616 10.1073/pnas.96.5.2187
Seo TK Thorne JL Hasegawa M Kishino H Estimation of effective population size of HIV-1 within a host: a pseudomaximum-likelihood approach Genetics 2002 160 1283 1293 11973287
Kingman JFC On the genealogy of large populations Journal of Applied Probability 1982 19A 27 43
Kingman JFC The coalescent Stock Process Appl 1982 13 235 248 10.1016/0304-4149(82)90011-4
Fisher RA The genetical theory of natural selection 1930 Oxford: Clarendon Press
Wright S Evolution in Mendelian populations Genetics 1931 16 97 159 17246615
Watterson GA On the number of segregating sites in genetical models without recombination Theoretical Population Biology 1975 7 256 276 1145509 10.1016/0040-5809(75)90020-9
Tavaré S Line-of-descent and genealogical processes, and their applications in population genetics models Theoretical Population Biology 1984 26 119 164 6505980 10.1016/0040-5809(84)90027-3
Saunders IW Tavare S Watterson GA On the genealogy of nested subsamples from a haploid population Advances in Applied Probability 1984 14 471 491
Cooper A Drummond AJ Willerslev E Ancient DNA: would the real Neandertal please stand up? Curr Biol 2004 14 431 433 10.1016/j.cub.2004.05.037
Shapiro B Drummond AJ Rambaut A Wilson MC Matheus PE Sher AV Pybus OG Gilbert MTP Barnes I Binladen J Willerslev E Hansen AJ Baryshnikov GF Burns JA Davydov S Driver JC Froese DG Harington CR Keddie G Kosintsev P Kunz ML Martin LD Stephenson RO Storer J Tedford R Zimov S Cooper A Rise and Fall of the Beringian Steppe Bison Science 2004 306 1561 1565 15567864 10.1126/science.1101074
Griffiths RC Transient distribution of the number of segregating sites in a neutral infinite-sites model with no recombination Journal of Applied Probability 1981 18 42 51
Hudson RR Gene genealogies and the coalescent process Oxford Surveys in evolutionary biology 1991 1 44
Barnes I Matheus P Shapiro B Jensen D Cooper A Dynamics of Pleistocene population extinctions in Beringian brown bears Science 2002 295 2267 2270 11910112 10.1126/science.1067814
Caramelli D Lalueza-Fox C Vernesi C Lari M Casoli A Mallegni F Chiarelli B Dupanloup I Bertranpetit J Barbujani G Bertorelle G Evidence for a genetic discontinuity between Neandertals and 24,000-year-old anatomically modern Europeans PNAS 2003 100 6593 6597 12743370 10.1073/pnas.1130343100
Griffiths RC Tavaré S The age of a mutation in a general coalescent tree Stochastic Models 1998 14 273 295
Tajima F Statistical Method for Testing the Neutral Mutation Hypothesis by DNA Polymorphism Genetics 1989 123 585 595 2513255
|
15960847
|
PMC1185533
|
CC BY
|
2021-01-04 16:30:39
|
no
|
BMC Genet. 2005 Jun 16; 6:35
|
utf-8
|
BMC Genet
| 2,005 |
10.1186/1471-2156-6-35
|
oa_comm
|
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-971600016810.1186/1471-2164-6-97Research ArticlePositional and functional mapping of a neuroblastoma differentiation gene on chromosome 11 De Preter Katleen [email protected] Jo [email protected] Björn [email protected] Philippa [email protected] Heike [email protected]ö Anders [email protected] Nigel P [email protected] Nurten [email protected] Wim [email protected] Roy Nadine [email protected] Scott [email protected]åhlman Sven [email protected] Frank [email protected] Center for Medical Genetics, Ghent University Hospital MRB 2nd floor, De Pintelaan 185, B-9000 Ghent, Belgium2 The Wellcome Trust Sanger Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge CB10 1SA, United Kingdom3 Department of Laboratory Medicine, Molecular Medicine, Lund University, University Hospital MAS, S-20502 Malmö, Sweden4 Department of Pathological Anatomy, Ghent University Hospital BLOK A, De Pintelaan 185, B-9000 Ghent, Belgium5 Sir Alastair Currie Cancer Research U.K. Laboratories, Division of Pathology, Molecular Medicine Centre, University of Edinburgh, Crewe Road, Edinburgh EH4 2XU, United Kingdom2005 6 7 2005 6 97 97 4 3 2005 6 7 2005 Copyright © 2005 De Preter et al; licensee BioMed Central Ltd.2005De Preter 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
Loss of chromosome 11q defines a subset of high-stage aggressive neuroblastomas. Deletions are typically large and mapping efforts have thus far not lead to a well defined consensus region, which hampers the identification of positional candidate tumour suppressor genes. In a previous study, functional evidence for a neuroblastoma suppressor gene on chromosome 11 was obtained through microcell mediated chromosome transfer, indicated by differentiation of neuroblastoma cells with loss of distal 11q upon introduction of chromosome 11. Interestingly, some of these microcell hybrid clones were shown to harbour deletions in the transferred chromosome 11. We decided to further exploit this model system as a means to identify candidate tumour suppressor or differentiation genes located on chromosome 11.
Results
In a first step, we performed high-resolution arrayCGH DNA copy-number analysis in order to evaluate the chromosome 11 status in the hybrids. Several deletions in both parental and transferred chromosomes in the investigated microcell hybrids were observed. Subsequent correlation of these deletion events with the observed morphological changes lead to the delineation of three putative regions on chromosome 11: 11q25, 11p13->11p15.1 and 11p15.3, that may harbour the responsible differentiation gene.
Conclusion
Using an available model system, we were able to put forward some candidate regions that may be involved in neuroblastoma. Additional studies will be required to clarify the putative role of the genes located in these chromosomal segments in the observed differentiation phenotype specifically or in neuroblastoma pathogenesis in general.
==== Body
Background
In addition to the well known group of high stage neuroblastomas with MYCN amplification and 1p-deletion, a second genetic subgroup of aggressive neuroblastomas has been delineated. This subgroup is characterised by the presence of 11q-deletions, often in association with 3p-deletions [1-5]. Both subgroups typically present with 17q-gain or a normal chromosome 17 copy number, which are the strongest independent genetic indicators of poor prognosis [6]. Deletions of 11q mostly affect a large distal part of the long arm. Only a few small deletions have been identified which delineated a tentative SRO (shortest region of overlap) at 11q23 between markers D11S1340 and D11S1299, encompassing a region of approximately 3 Mb [7]. More recently however, a neuroblastoma patient was reported with a constitutional 11q14.1-11q23.3 deletion that did not overlap with the proposed SRO [8]. Consequently, the presumed localisation of the 11q neuroblastoma tumour suppressor gene (or genes) remains ill defined, thus hampering the selection of positional candidate genes. For the 11q23 region we proposed SDHD as a putative candidate neuroblastoma tumour suppressor, but only two bona fide mutations could be identified[9].
In addition to the observed losses of 11q in neuroblastoma, the existence of a tumour suppressor gene on 11q has also been supported by functional evidence obtained by microcell mediated chromosome 11 transfer (MMCT) experiments [10]. Although these studies were initially aimed at investigating the role of chromosome 1p in tumour suppression, the control chromosome 11 transfer experiment unexpectedly produced clones with morphological features of differentiation. Introduction of chromosome 11 induced a more flattened and adherent morphology, with short neuritic processes, similar to the changes seen after a few days of growth in the presence of retinoic acid. As these microcell hybrids could be powerful models for the identification of candidate neuroblastoma suppressor or differentiation genes, we decided first to determine the genetic status of the chromosome 11 in the hybrid subclones prior to further experiments. To this purpose, the parental NGP cell line and the microcell hybrids after chromosome 11 transfer were analysed using high-resolution arrayCGH (microarray based comparative genomic hybridisation), FISH (fluorescence in situ hybridisation) and microsatellite heterozygosity mapping. Following the identification of a region on chromosome 11 with altered copy number, we measured the mRNA expression levels of genes in these regions in an attempt to find altered gene expression related to neurite outgrowth and differentiation.
Results
Morphological characterisation
The chromosome 11 status of the different microcell hybrid subclones used in this study and the reported chromosome 11 changes [10] are listed in Table 1. The morphology of the cells was comparable to the phenotype described by Bader and colleagues [10]. Cells of the parental cell line NGP.1A.TR1 (a tumour reconstitute of mutagenised NGP cells [10]) were non-adherent, spheroid and growing in cell clusters (Figure 1A). Subclones with an apparently intact transferred chromosome 11 (MCH574c4, c11, c13), as well as the clone with reported loss of a region on 11q (MCH574c10) exhibited features of induced differentiation, with more flattened and adherent cells and some short neuritic processes (Figure 1C). Subclone MCH574c3 with reported loss of part of 11p showed the same non-adherent phenotype as the parental cell line NGP.1A.TR1 (Figure 1B).
Table 1 Chromosome 11 status and morphology of the microcell hybrids (MCH) obtained after chromosome 11 transfer in parental NGP.1A.TR1 cells as determined by Bader and colleagues [10] and in this study
microcell hybrid subclone (NGP.1A.TR1 + chr 11) chromosome 11 status (in addition to parental NGP.1A.TR1 11q-loss) morphology
Bader et al. [10] this study
MCH574c4,c11,c13 no additional changes del(11)(pterp15.1) more flattened, adherent cells, some short neuritic processes
MCH574c10 del(11)(q23.3) (MCT128.1, HBI 18P2) del(11)(pterp15.1) more flattened, adherent cells, some short neuritic processes
MCH574c3 del(11)(p15.5) (HRAS) del(11)(pterp15.1)
del(11)(pterp13)
del(11)(q25qter) non-adherent, spheroid cells, growing in cell clusters
Figure 1 Cell morphology of parental cell line NGP.1A.TR1 (A) and chromosome 11 transferred subclone MCH574c3 (B) with non-adherent, spheroid cells, and subclone MCH574c10 (C) showing signs of induced differentiation such as short neuritic processes
Assessment of the organisation of the actin fibres using phalloidin staining confirmed the presence of neurites (and excluded stress fibres) in subclones MCH574c10 and MCH574c11[11].
ArrayCGH based chromosome 11 copy number assessment
ArrayCGH was performed for NGP.1A.TR1, MCH574c3 and MCH574c10 cells. These hybridisations failed to provide evidence for the reported 11q-deletion in the transferred chromosome of microcell hybrid MCH574c10 (Figure 2). Unexpectedly, the distal region of the short arm of one of the chromosomes 11 (11pter->11p15.1) was deleted in both MCH574c3 and MCH574c10. Microcell hybrid MCH574c3 presented with an additional larger deletion of 11pter->11p13, as well as a third deletion involving the most distal band (11q25->11qter) in one of the chromosomes 11. Deletion of a single BAC clone RP11-51B23 on 11p15.3 was detected in the parental NGP.1A.TR1 cells (Figure 2B). Thus far, this clone has not been recognised as being involved in polymorphic genomic deletions for this particular chromosomal region (own observations and Ensembl clone list). Deletions observed by arrayCGH were confirmed by FISH analysis using one BAC clone selected in each observed deleted region (RP11-734D5 on 11p15.3, RP11-48O9 on 11p13, RP11-545G16 on 11q25). This FISH analysis demonstrated that the 11pter->11p15.1 deletion was present in all other subclones that were not analysed with arrayCGH, i.e. MCH574c4, c11 and c13.
Figure 2 ArrayCGH results (log2 scale) of parental cell line NGP1A.TR1 and microcell hybrids MCH574c3 and MCH574c10 compared to a normal female control, with reported (red) and newly detected (orange) chromosome 11 deletion events, (A) parental cell line (NGP.1A.TR1), (B) MCH574c10 in which regional 11q-loss of the transferred chromosome 11 was reported [10] and (C) MCH574c3 with reported regional 11p-loss of transferred chromosome 11. FISH was used to confirm the results obtained by arrayCGH (data not shown).
Microsatellite heterozygosity mapping
To determine which of the chromosomes 11 exhibited loss of the 11pter->11p15.1, 11pter->11p13 and 11q25->11qter regions, microsatellite heterozygosity mapping in conjunction with FISH analysis of metaphase spreads was performed. Microsatellite markers D11S861 (on 11p15.2) and D11S1324 (on 11p14.1) were tested on NGP.1A.TR1, MCH574c3 and MCH574c10. These tests show that one of the two parental chromosomes 11 had lost the 11pter->11p15.1 region, while the 11pter->11p13 segment was lost in the transferred chromosome. FISH on metaphase spreads (clone RP11-545G16 on 11q25 in combination with RP11-206C1 on 11p15.1; clone RP11-709M17 on 11q25 in combination with clone RP11-4B7 on 11p15.2) demonstrated that the 11q25->11qter deletion occurred in the transferred chromosome 11, whereas the 11pter->11p15.1 deletion occurred in the normal parental chromosome 11 (and not in the parental der(11)t(2;11)) (Figure 2).
Breakpoint delineation of chromosome 11 deletions
The position of the deletion breakpoints was confirmed or refined by FISH analysis. The breakpoint of the del(11)(q22.1qter) resulting from an unbalanced translocation between chromosomes 2 and 11 in parental cell line NGP.1A.TR1 mapped within a 2.285 Mb segment located between BAC clones RP11-379J13 and RP11-49M9 (map position between 97.328 Mb and 99.613 Mb, NCBI 35 May 2004 assembly (hg17)). The breakpoint of the 11pter->11p15.1 deletion of the normal parental chromosome 11 in all microcell hybrids was assigned to a 229 kb segment between RP11-452G18 and RP11-358H18 (17.337 Mb – 17.596 Mb). The breakpoint of the larger 11p-deletion (11pter->11p13) present in MCH574c3 was located within a 921 kb segment flanked by clones RP11-48O9 and RP11-202M19 (33.038 Mb – 33.959 Mb). The breakpoint of the distal ± 13 Mb 11q25->11qter deletion of the transferred chromosome in subclone MCH574c3 was mapped between BAC clones RP11-340L13 and RP11-697E14 (131.149 Mb – 131.230 Mb). Flanking clones of the 11p15.3 deletion in NGP.1A.TR1 were also tested using FISH (RP11-734D5, RP11-573E11, RP11-47J17) demonstrating that the deletion involves at least a 707 kb segment including BAC clones RP11-573E11 and RP11-47J17 (12.202 Mb – 12.909 Mb) (Figure 3).
Figure 3 FISH analysis of BAC clones RP11-573E11 (panel A), RP11-51B23 and RP11-47J17 (not shown) in parental cells NGP.1A.TR1 (with a derivative chromosome 11, due to an unbalanced translocation between chromosomes 11 and 2) confirms the existence of a small deletion in 11p15.3. The breakpoint of the distal 11q25->11qter deletion of the transferred chromosome in subclone MCH574c3 was mapped between BAC clones RP11-340L13 (not shown) and RP11-697E14 (panel B).
mRNA expression profiling
As loss of 11q is a recurrent chromosomal aberration in a subgroup of advanced stage neuroblastomas, the 11q25->11qter region that is deleted in the MCH574c3 microcell hybrid is of particular interest. In an attempt to relate the observed morphology of induced neuronal differentiation to expression differences of genes located in this distal 11q25 segment, the expression of 6 known genes, i.e. HNT, OPCML, JAM3, THY28, ACAD8 and B3GAT1 was tested. Of particular interest are HNT and B3GAT1, because of their reported involvement in neurite outgrowth and neural crest development. We quantified the mRNA expression of these 6 genes in the microcell hybrids, the parental cell line and in neuroblastoma cell lines (SH-SY5Y, LA-N-5 and NTRK1 transfected SH-SY5Y) that were treated with inducers of differentiation [12-14]. While the expression of the genes is not significantly altered in the microcell hybrids compared to the parental cell line, the expression of HNT is significantly higher in different cell lines that are induced to differentiate (between 5 to 120 fold induction) (Figure 4).
Figure 4 Fold induction of HNT mRNA expression (log scale, versus control cultures, C) in neuroblastoma cell lines SH-SY5Y, LA-N-5 and NTRK1-transfected SH-SY5Y after induction of differentiation with RA and BDNF (retinoic acid and brain-derived neurotrophic factor), NGF (nerve growth factor) and TPA (12-O-tetradecanoyl-phorbol-13-acetate).
Discussion
In a search for candidate neuroblastoma genes located on chromosome arm 11q, we investigated microcell hybrids obtained by transfer of a normal chromosome 11 into NGP neuroblastoma cells with loss of 11q. Although initially designed as a control experiment, this transfer resulted in morphological changes in the obtained hybrids (without loss of tumorigenicity) and also yielded revertants after further culture [10]. The induced differentiation that was observed in all but one microcell hybrid is consistent with the presence of a neuroblastoma differentiation gene on chromosome 11. We thus anticipated that these hybrids might be of interest for functional mapping of the regions on chromosome 11 critically involved in neuroblastoma pathogenesis. To investigate this, we performed arrayCGH copy number analysis of these microcell hybrids. This allowed us to assess the status of the introduced (and parental) chromosomes 11 and to validate these hybrids as model system for further functional assays. The obtained results were surprising and puzzling. One particular microcell hybrid that did not show the expected differentiation features upon chromosome 11 transfer was shown to carry an 11q25->11qter deletion in the transferred chromosome. In addition we found that all microcell hybrid subclones presented with an 11pter->11p15.1 deletion, and that the MCH574c3 hybrids presented with an additional 11pter->11p13 deletion.
In line with previous successful functional analyses of microcell hybrids [15], the responsible gene for the observed changes in cell morphology is assumed to be located in one of the chromosomal regions that show a different copy number in the microcell hybrid subclones with differentiation features (MCH574c4, c10, c11 and c13) compared to the non-adherent, spheroid cell phenotype of parental cell line NGP.1A.TR1 and microcell hybrid subclone MCH574c3. Based upon our findings three regions can be identified as candidate regions harbouring a putative differentiation gene: (1) the 11q25->11qter region (lost in MCH574c3), (2) the 11p13->11p15.1 region (lost in MCH574c3 but not in the other MCH574 subclones) and (3) a small region of at least 706 kb on 11p15.3 (lost in NGP.1A.TR1) (Figure 5).
Figure 5 Regional copy numbers (deletion events are indicated in red) in cells with non-adherent, spheroid (parental) cell phenotype compared to cells with induced differentiation, demonstrating the three regions on chromosome 11 that may be involved in the phenotypic difference, i.e. a small region on 11p15.3 encompassing BAC clone RP11-51B23 (lost in NGP.1A.TR1) (A region), the 11p13->11p15.1 region (lost in MCH574c3 but not in the other MCH574 microcell hybrids) (B region) and the 11q25->11qter region (lost in MCH574c3) (C region) (* indicates the putative presence of a mutated gene).
As loss of distal 11q is a recurrent chromosomal aberration in MYCN single copy advanced stage neuroblastomas [3], we propose the 11q25->11qter chromosomal segment as the most likely candidate region for the presence of a differentiation gene. Despite efforts to define a shortest region of overlap (SRO) for 11q-loss in neuroblastoma by microsatellite heterozygosity mapping [7] and delineation of constitutional 11q-deletions [8,16], a consensus region for loss of 11q in neuroblastoma has not been defined thus far. In the light of the uncertainty of the boundaries of the 11q SRO, the 11q25->11qter region must be considered as potentially harbouring a neuroblastoma suppressor or differentiation gene. This region is present in two copies in microcell hybrid subclones MCH574c4, c10, c11 and c13 with differentiated morphology, but only in one copy in the non-adherent, spheroid cells from NGP.1A.TR1 and MCH574c3. Six known genes, i.e. HNT, OPCML, JAM3, THY28, ACAD8 and B3GAT1 are located in this distal 11q segment, of which two genes are of particular interest. HNT (neurotrimin) is reported to promote neurite outgrowth and adhesion [17]. B3GAT1 encodes for a protein that functions as the key enzyme in a glucuronyl transfer reaction during the biosynthesis of the carbohydrate epitope HNK1 (CD57) [18,19], which is a carbohydrate expressed in developmentally immature neural crest cells [20]. Interestingly, the expression of HNT is significantly increased in neuroblastoma cell lines that are induced to differentiate using RA (retinoic acid), RA plus BDNF (brain-derived neurotrophic factor), NGF (nerve growth factor) and TPA (12-O-tetradecanoyl-phorbol-13-acetate). However, HNT expression is not significantly different between the differentiated microcell hybrids and the parental cells. It is conceivable that the observed phenotypic changes are caused by small changes in expression that can not be reliably detected by Q-PCR. An alternative explanation is that the normal parental chromosome 11 harbours a mutated allele that is normally expressed at the mRNA level (Figure 5). Reintroduction of a wild type allele by chromosome transfer could repair the defect, leading to differentiation. This is in keeping with reversal to the non-adherent, spheroid morphology of the microcell hybrids that have lost the 11q25->11qter region of the transferred chromosome. Additional mutation, promoter hypermethylation and gene directed functional assays are needed to clarify which of the genes located within the deleted 11q25->11qter region are responsible for the differentiated phenotype.
While the 11q25 region is the best candidate region to harbour a differentiation gene, the observed deletions on the short arm of chromosome 11 may also account for the differentiated morphology. The observation of two independent deletion events along the distal part of chromosome arm 11p is suggestive for the involvement of this region. In particular, it is striking that all microcell hybrids in which chromosome 11 is transferred are characterised by the presence of an 11pter->11p15.1 deletion in the (prior to transfer) normal parental chromosome 11. This may either be the result of an early coincidental event during the transfer process, or indicative for a selection process against the presence of three copies of a growth suppressive gene in this region. The last hypothesis may be further supported by the presence of unbalanced 11p-deletions in 4% of neuroblastomas (14/394) [21,22].
Apart from highlighting at candidate 11q regions involved in neuroblastoma pathogenesis, this study clearly shows that it is important to monitor the transfer of the desired chromosome, as well as the genetic background of the cell line before and after chromosome transfer experiments. Selective pressure processes may occur during or after transfer of a chromosome, e.g. by chromosomal loss in order to maintain the viability of the microcell hybrids. Hence, detailed information on the chromosome copy number status before and after transfer is required in order to correlate phenotypic changes with chromosomal alteration. ArrayCGH has been proven to be a valuable screening method for evaluation of the chromosome alterations and for delineation of possible deletion events, allowing fine-mapping of the candidate regions that harbour candidate suppressor genes.
Conclusion
Microsatellite marker heterozygosity analysis, FISH and (array)CGH based copy number in neuroblastoma tumour specimens and patients with constitutional deletions have thus far not identified a consensus SRO for 11q-deletion. Here, we present an alternative strategy to pinpoint chromosomal regions or genes that may be important in neuroblastoma pathogenesis. Chromosome 11 transfer, followed by phenotype scoring and high-resolution copy number analysis delineated putative regions on chromosome 11 involved in tumour differentiation. Further mutation and functional analyses are required to clarify the putative involvement of genes localised in these regions in neuroblastoma.
Methods
Cell lines
The parental cell line, NGP.1A.TR1, and the chromosome 11 microcell transfer derived subclones MCH574c3, c4, c10, c11 and c13 used in this study have been described previously [10].
Cell lines were cultured following standard procedures and were digitally photographed under an inverted (phase-contrast) microscope, pelleted, snap-frozen and stored at -80°C for further processing. DNA was isolated using the QIAamp DNA mini kit (Qiagen). RNA was isolated from the snap-frozen cell pellets using the RNeasy Mini kit (Qiagen) according to the manufacturer's guidelines, followed by RNase free DNase treatment on column (Qiagen).
Culture conditions and details regarding differentiation protocols are given in [12,13]. The NTRK1-transfected SH-SY5Y cells used, were SH-SY5Y/trkA, clone 6:2 described in [23].
Phalloidin staining
Cell lines were fixed for 10 min in 4% paraformaldehyde/HEPES on ice. The excess of aldehydes is quenched for 5 min in 50 mM NH4Cl. After washing twice for 5 min in 1x PBS, extraction is performed for 5 min in acetone (20°C). The cells are washed again twice for 5 min in 1x PBS, followed by blocking in 0.2% Fish Skin Gelatine (FSG, Sigma)/PBS. During 60 min cells are incubated with Alexa 594-phalloidin (1 unit per section), dissolved in 0.2%FSG/PBS at 37°C. Cells are washed twice for 5 min in 1x PBS; nuclei are stained for 1 min with DAPI; sections are washed with 1x PBS and mounted in Vectashield.
ArrayCGH
ArrayCGH using 1 Mb BAC arrays was performed once for NGP.1A.TR1, MCH574c3 and MCH574c10 cells with normal female DNA as control. In addition, subclones MCH574c3 and MCH574c10 were hybridised to the same arrays with NGP.1A.TR1 DNA as control. Hybridisation of cell line and control DNA to the array was performed as described [24]. Using our in-house developed analysis and visualisation software, arrayCGHbase, data were normalised to the median ratio, and replicate median ratio profiles visualised [29].
FISH and microsatellite marker analysis
BAC clones and microsatellite markers were selected based on their chromosomal position using the Ensembl genome browser , the UCSC human genome browser (July 2003 freeze, ) or the Genome Database . Labelling and FISH (fluorescence in situ hybridisation) was performed as described [25]. Experimental conditions for the fluorescent based microsatellite screening can be obtained from the authors upon request.
Real-time quantitative RT-PCR based mRNA expression profiling
Primers were designed using Primer Express v2.0 (Applied Biosystems). Primer sequences are available in the public RTPrimerDB database : HNT (1078), OPCML (1079), JAM3 (1080), THY28 (1084), ACAD8 (1081), B3GAT1 (1082), HPRT1 (5), UBC (8) and GAPD (3) [26]. Relative expression levels were determined using an optimized two-step SYBR Green I RT-PCR assay [27]. PCR reagents were obtained from Eurogentec as SYBR Green I core reagents, prepared as 2x mastermixes, stored at -20°C and used according to the manufacturer's instructions. Reactions were run on an ABI5700 (Applied Biosystems). The comparative CT method was used for quantification. Gene expression levels were normalized using the geometric mean of the 3 most stable internal control genes in neuroblastoma (i.e. UBC, HPRT1 and GAPD) as reported previously [28].
Abbreviations
arrayCGH = microarray based comparative genomic hybridisation
FISH = fluorescence in situ hybridisation
MMCT = microcell mediated chromosome transfer
SRO = shortest region of overlap
Authors' contributions
KDP supervised the culturing of the microcell hybrids that were produced by SB, carried out the microsatellite marker analysis and drafted the manuscript. PC performed the arrayCGH hybridisations, under the supervision of HF and NC. KDP and BM analysed the arrayCGH data. WW performed the phalloidin staining and NVR evaluated the FISH results. NY performed real-time quantitative PCR. Neuroblastoma cells were induced to differentiate by AE under the supervision of SP. JV and FS participated in the study design and coordination, and were the final editors of the manuscript.
Acknowledgements
We greatly acknowledge Peter Degrave and Geert De Vos for the cell cultures, Anouck Waeytens and Isabel Rottiers for phalloidin staining, and Helén Nilsson for providing us with RNA from NTRK1-SH-SY5Y cells that were treated with differentiation inducers.
This text presents research results of the Belgian program of Interuniversity Poles of Attraction initiated by the Belgian State, Prime Minister's Office, Science Policy Programming. The scientific responsibility is assumed by the authors. KDP, JV and WW are supported by a post-doctoral grant from the Institute for the Promotion of Innovation by Science and Technology in Flanders (IWT). NVR is a post-doctoral researcher with the FWO. This work was supported by FWO-grant G.0028.00, VEO-grant 011V1302, BOF-grant 011F1200 and 011B4300, and GOA-grant 12051203.
==== Refs
Vandesompele J Van Roy N Van Gele M Laureys G Ambros P Heimann P Devalck C Schuuring E Brock P Otten J Gyselinck J De Paepe A Speleman F Genetic heterogeneity of neuroblastoma studied by comparative genomic hybridization Genes Chromosomes Cancer 1998 23 141 152 9739017 10.1002/(SICI)1098-2264(199810)23:2<141::AID-GCC7>3.0.CO;2-2
Vandesompele J Speleman F Van Roy N Laureys G Brinskchmidt C Christiansen H Lampert F Lastowska M Bown N Pearson A Nicholson JC Ross F Combaret V Delattre O Feuerstein BG Plantaz D Multicentre analysis of patterns of DNA gains and losses in 204 neuroblastoma tumors: how many genetic subgroups are there? Med Pediatr Oncol 2001 36 5 10 11464905 10.1002/1096-911X(20010101)36:1<5::AID-MPO1003>3.0.CO;2-E
Plantaz D Vandesompele J Van Roy N Lastowska M Bown N Combaret V Favrot MC Delattre O Michon J Benard J Hartmann O Nicholson JC Ross FM Brinkschmidt C Laureys G Caron H Matthay KK Feuerstein BG Speleman F Comparative genomic hybridization (CGH) analysis of stage 4 neuroblastoma reveals high frequency of 11q deletion in tumors lacking MYCN amplification Int J Cancer 2001 91 680 686 11267980 10.1002/1097-0215(200002)9999:9999<::AID-IJC1114>3.0.CO;2-R
Luttikhuis ME Powell JE Rees SA Genus T Chughtai S Ramani P Mann JR McConville CM Neuroblastomas with chromosome 11q loss and single copy MYCN comprise a biologically distinct group of tumours with adverse prognosis Br J Cancer 2001 85 531 537 11506492 10.1054/bjoc.2001.1960
Breen CJ O'Meara A McDermott M Mullarkey M Stallings RL Coordinate deletion of chromosome 3p and 11q in neuroblastoma detected by comparative genomic hybridization Cancer Genet Cytogenet 2000 120 44 49 10913676 10.1016/S0165-4608(99)00252-6
Vandesompele J Baudis M De Preter K Van Roy N Ambros P Bown N Brinkschmidt C Christiansen H Combaret V Lastowska M Nicholson J O’Meara M Plantaz D Stallings R Brichard B Van den Broecke C De Paepe A Laureys G Speleman F Unequivocal delineation of clinico-genetic subgroups and development of a new model for outcome prediction in neuroblastoma J Clin Oncol 2005 23 2280 99 submitted 15800319 10.1200/JCO.2005.06.104
Guo C White PS Weiss MJ Hogarty MD Thompson PM Stram DO Gerbing R Matthay KK Seeger RC Brodeur GM Maris JM Allelic deletion at 11q23 is common in MYCN single copy neuroblastomas Oncogene 1999 18 4948 4957 10490829 10.1038/sj.onc.1202887
Mosse Y Greshock J King A Khazi D Weber BL Maris JM Identification and high-resolution mapping of a constitutional 11q deletion in an infant with multifocal neuroblastoma Lancet Oncol 2003 4 769 771 14662434 10.1016/S1470-2045(03)01283-X
De Preter K Vandesompele J Hoebeeck J Vandenbroecke C Smet J Nuyts A Laureys G Combaret V Van Roy N Roels F Van Coster R Praet M De Paepe A Speleman F No evidence for two-hit involvement of SDHD in neuroblastoma pathogenesis BMC Cancer 2004 4 55 15331017 10.1186/1471-2407-4-55
Bader SA Fasching C Brodeur GM Stanbridge EJ Dissociation of suppression of tumorigenicity and differentiation in vitro effected by transfer of single human chromosomes into human neuroblastoma cells Cell Growth Differ 1991 2 245 255 1679663
Ling M Troller U Zeidman R Lundberg C Larsson C Induction of neurites by the regulatory domains of PKCdelta and epsilon is counteracted by PKC catalytic activity and by the RhoA pathway Exp Cell Res 2004 292 135 150 14720513 10.1016/j.yexcr.2003.08.013
Edsjo A Hallberg B Fagerstrom S Larsson C Axelson H Pahlman S Differences in early and late responses between neurotrophin-stimulated trkA- and trkC-transfected SH-SY5Y neuroblastoma cells Cell Growth Differ 2001 12 39 50 11205744
Edsjo A Lavenius E Nilsson H Hoehner JC Simonsson P Culp LA Martinsson T Larsson C Pahlman S Expression of trkB in human neuroblastoma in relation to MYCN expression and retinoic acid treatment Lab Invest 2003 83 813 823 12808116
Sidell N Retinoic acid-induced growth inhibition and morphologic differentiation of human neuroblastoma cells in vitro J Natl Cancer Inst 1982 68 589 596 7040765
Doherty AM Fisher EM Microcell-mediated chromosome transfer (MMCT): small cells with huge potential Mamm Genome 2003 14 583 592 14629108 10.1007/s00335-003-4002-0
Satge D Moore SW Stiller CA Niggli FK Pritchard-Jones K Bown N Benard J Plantaz D Abnormal constitutional karyotypes in patients with neuroblastoma: a report of four new cases and review of 47 others in the literature Cancer Genet Cytogenet 2003 147 89 98 14623457 10.1016/S0165-4608(03)00203-6
Gil OD Zanazzi G Struyk AF Salzer JL Neurotrimin mediates bifunctional effects on neurite outgrowth via homophilic and heterophilic interactions J Neurosci 1998 18 9312 9325 9801370
Mitsumoto Y Oka S Sakuma H Inazawa J Kawasaki T Cloning and chromosomal mapping of human glucuronyltransferase involved in biosynthesis of the HNK-1 carbohydrate epitope Genomics 2000 65 166 173 10783264 10.1006/geno.2000.6152
Marcos I Galan JJ Borrego S Antinolo G Cloning, characterization, and chromosome mapping of the human GlcAT-S gene J Hum Genet 2002 47 677 680 12522689 10.1007/s100380200103
Bronner-Fraser M Analysis of the early stages of trunk neural crest migration in avian embryos using monoclonal antibody HNK-1 Dev Biol 1986 115 44 55 3516760 10.1016/0012-1606(86)90226-5
Guo C White PS Hogarty MD Brodeur GM Gerbing R Stram DO Maris JM Deletion of 11q23 is a frequent event in the evolution of MYCN single-copy high-risk neuroblastomas Med Pediatr Oncol 2000 35 544 546 11107113 10.1002/1096-911X(20001201)35:6<544::AID-MPO10>3.0.CO;2-2
Schleiermacher G Janoueix-Lerosey I Combaret V Derre J Couturier J Aurias A Delattre O Combined 24-color karyotyping and comparative genomic hybridization analysis indicates predominant rearrangements of early replicating chromosome regions in neuroblastoma Cancer Genet Cytogenet 2003 141 32 42 12581896 10.1016/S0165-4608(02)00644-1
Lavenius E Gestblom C Johansson I Nanberg E Pahlman S Transfection of TRK-A into human neuroblastoma cells restores their ability to differentiate in response to nerve growth factor Cell Growth Differ 1995 6 727 736 7669728
Fiegler H Carr P Douglas EJ Burford DC Hunt S Scott CE Smith J Vetrie D Gorman P Tomlinson IP Carter NP DNA microarrays for comparative genomic hybridization based on DOP-PCR amplification of BAC and PAC clones Genes Chromosomes Cancer 2003 36 361 374 12619160 10.1002/gcc.10155
Van Roy N Laureys G Cheng NC Willem P Opdenakker G Versteeg R Speleman F 1;17 translocations and other chromosome 17 rearrangements in human primary neuroblastoma tumors and cell lines Genes Chromosomes Cancer 1994 10 103 114 7520263
Pattyn F Speleman F De Paepe A Vandesompele J RTPrimerDB: the real-time PCR primer and probe database Nucleic Acids Res 2003 31 122 123 12519963 10.1093/nar/gkg011
Vandesompele J De Paepe A Speleman F Elimination of primer-dimer artifacts and genomic coamplification using a two-step SYBR green I real-time RT-PCR Anal Biochem 2002 303 95 98 11906156 10.1006/abio.2001.5564
Vandesompele J De Preter K Pattyn F Poppe B Van Roy N De Paepe A Speleman F Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes Genome Biol 2002 3 RESEARCH0034 12184808 10.1186/gb-2002-3-7-research0034
Menten B Pattyn K Robbrecht P Michels E Buysse K Speleman F Mortier G De Paepe A van Vooren S Vermeesch J Moreau Y De Moor B Vermeulen S Speleman F Vandesompele J arrayCGHbase: an analysis platform for comparative genomic hybridization microarrays BMC Bioinformatics 2005 6 124 15910681 10.1186/1471-2105-6-124
|
16000168
|
PMC1185534
|
CC BY
|
2021-01-04 16:32:49
|
no
|
BMC Genomics. 2005 Jul 6; 6:97
|
utf-8
|
BMC Genomics
| 2,005 |
10.1186/1471-2164-6-97
|
oa_comm
|
==== Front
BMC Health Serv ResBMC Health Services Research1472-6963BioMed Central London 1472-6963-5-481598752310.1186/1472-6963-5-48Research ArticleThe impact of a pharmacist-managed dosage form conversion service on ciprofloxacin usage at a major Canadian teaching hospital: a pre- and post-intervention study Ho Bradley P [email protected] Tim TY [email protected] Robert M [email protected] Terryn L [email protected] Peter J [email protected] Pharmaceutical Sciences Clinical Services Unit, Vancouver General Hospital, Vancouver Coastal Health, 855 West 12th. Avenue, Vancouver, BC, Canada, V5Z 1M92 Faculty of Pharmaceutical Sciences, University of British Columbia, 2146 East Mall, Vancouver, BC, Canada, V6T 1Z32005 29 6 2005 5 48 48 7 3 2005 29 6 2005 Copyright © 2005 Ho et al; licensee BioMed Central Ltd.2005Ho 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 cost containment efforts, parenteral (IV) ciprofloxacin appears to be overutilized at Vancouver General Hospital. In November 2003, the Pharmacist-managed intravenous to oral (IV-PO) Dosage Form Conversion Service was implemented, enabling autonomous pharmacist-initiated dosage form conversion for ciprofloxacin. This study evaluates characteristics of ciprofloxacin use prior to and following implementation of this conversion service.
Methods
This was a single-centre, two-phase (pre/post), unblinded study. Phase I occurred between November 12, 2002 and November 11, 2003 (365 days), and Phase II between November 12, 2003 and March 11, 2004 (120 days). All patients receiving ciprofloxacin IV during these periods were reviewed. The primary endpoint was IV:PO ciprofloxacin use ratio. Secondary endpoints were total number of ciprofloxacin doses, proportion of inappropriate IV ciprofloxacin doses, cost of therapy between phases, and estimated cost avoidance with the intervention.
Results
Two hundred ciprofloxacin IV treatment courses were evaluated (100 per phase). The IV:PO ciprofloxacin use ratio was 3.03 (Phase I) vs. 3.48 (Phase II). Total number of doses and ratio of IV to total doses across phases were similar (p = 0.2830). IV-PO ciprofloxacin conversion occurred in 27/100 (27%) of IV courses in Phase I and 23/100 (23%) in Phase II. Proportion of inappropriate ciprofloxacin IV doses decreased between Phases I and II (244/521 (47%) vs. 201/554 (36%) (p = 0.0005), respectively). Furthermore, the proportion of pharmacist-preventable inappropriate ciprofloxacin IV doses was reduced between Phases I and II (114/244 (47%) vs. 65/201 (32%) (p = 0.0026). Proportional cost avoidance associated with total inappropriate IV use was $7,172/$16,517 (43%) (in Canadian dollars) in Phase I vs. $6,012/$17,919 (34%) in Phase II (p = 0.001). Similarly, proportional cost avoidance associated with pharmacist-preventable inappropriate IV doses was reduced from $3,367/$16,517 (20%) in Phase I to $1,975/$17,919 (11%) in Phase II (p = 0.001).
Conclusion
While overall utilization of ciprofloxacin remained unchanged and the proportion of IV to total doses was stable during the study period, the proportion of inappropriate IV doses and its associated costs appear to have declined subsequent to implementation of a Pharmacist-managed IV-PO Dosage Form Conversion Service. Such a program may be a beneficial adjunct in facilitating appropriate and cost-effective usage of ciprofloxacin.
==== Body
Background
The annual drug expenditures at Vancouver Hospital are approximately $13 million (in Canadian dollars). At the Vancouver General Hospital (VGH) site, anti-infectives accounted for an expenditure of $3.39 million or 25% of the 2002–03 fiscal year total drug costs. Ciprofloxacin ranked third among all drugs by cost and was the second highest annual expenditure within the anti-infective drug class at $646,000. In addition, ciprofloxacin expenditures increased 5% from the previous year. Of the 39,147 ciprofloxacin doses administered in the 2002–03 fiscal year, 18,297 doses (47%) were given intravenously (IV).
Ciprofloxacin is a fluoroquinolone antibacterial that is primarily active against aerobic gram-negative bacterial infections [1]. It can be administered via the IV and oral (PO) routes. Concentrations similar to those achieved with the IV formulation are possible when administering ciprofloxacin orally, as it is highly bioavailable [2,3]. The daily drug cost (including material and labour costs for dispensing, preparation, and administration) of a typical ciprofloxacin 400 mg IV regimen administered twice daily is $72.62 [4]. Conversely, the daily drug cost of an equivalent ciprofloxacin 500 mg PO regimen given twice daily is $8.48, or 12% of the IV regimen [4].
In an effort to optimize use and minimize drug expenditures, ciprofloxacin has been designated a reserved antimicrobial drug (RAD) at our institution and has been included in the existing intravenous to oral (IV-PO) Step-down Program since 1992 [5]. Under this initiative, the use of the PO formulation of ciprofloxacin has been promoted at our institution through the means of newsletters [6], chart talkers and notes [7,8], and direct pharmacist-physician interactions. Other drugs included in this program are cefuroxime, cefixime, clindamycin, fluconazole, levofloxacin, and metronidazole.
Despite these cost containment efforts, there is some evidence to suggest that the IV formulation of these drugs may not be optimally utilized. Previous investigations by others and ourselves have shown that the IV formulation is often initiated when the PO formulation can be used [9-13]. Of equal importance, conversion to the PO formulation does not appear to be undertaken in a timely manner [6,10,11,13,14]. This results in unnecessary medication costs, IV drug administration expenses, and potential exposure to adverse events associated with IV therapy (e.g. pain at injection site, phlebitis, and line infections).
Various authors have described criteria-based medication dosage form conversion programs in the literature [8,10,11,13,15-24]. Some institutions have implemented modified IV-PO conversion programs in which pharmacists are given the authority and responsibility to change dosage forms in accordance with established criteria (i.e. clinical stability of the patient, ability to tolerate PO medications, and lack of drug interactions that may impair drug absorption from the gastrointestinal tract) [10,17,18,23,24]. The anticipated benefit of a pharmacist-managed conversion program is that delays in IV-PO conversion will be reduced if a pharmacist can avoid the requirement of conferring with the initiating physician before commencing a dosage form change. These services have demonstrated that cost savings can be achieved when pharmacists are directly responsible for changing the route of administration for selected medications [24].
In November 2003, the Pharmacist-managed IV-PO Dosage Form Conversion Service was approved at VGH by the Antibiotic Use Subcommittee (AUS), Drugs and Therapeutics Committee (D&TC), and the Medical Advisory Committee (MAC). Several antimicrobial agents were included in this service; namely, ciprofloxacin, clindamycin, co-trimoxazole, fluconazole, levofloxacin, metronidazole, acyclovir, ampicillin, cefazolin, cefuroxime, penicillin G, ceftriaxone, imipenem-cilastatin, cloxacillin, erythromycin, and ticarcillin-clavulanate [25].
Our hypothesis was that ciprofloxacin IV was overutilized at VGH and that a more cost-effective use of this dosage form was possible with the implementation of a Pharmacist-managed IV-PO Dosage Form Conversion Service. Accordingly, this study was conducted to assess the impact of this service on the relative utilization of the IV versus PO dosage form of ciprofloxacin. To our knowledge, there are no published reports involving an assessment of the impact of a Pharmacist-managed IV-PO Dosage Form Conversion Service on ciprofloxacin usage characteristics at a major Canadian teaching hospital.
Methods
Literature review
A literature search of the Medline, EMBASE and IPA databases, as well as a bibliographic review from the cited articles was performed to retrieve references pertaining to pharmacy-managed IV-PO conversion programs. Additional references were obtained through the St. Paul's Hospital and Lions' Gate Hospital pharmacy departments, as these local institutions had established pharmacist-managed conversion services [18,23]. The information collected was used to formulate a policy and procedure for the Pharmacist-managed IV-PO Conversion Service at our institution. This document was approved by the AUS, D&TC, and MAC, and a hospital-wide service was implemented.
Study design
This was a single-centre, 2-phase (pre/post), unblinded study to assess the impact of a hospital-approved intervention aimed at improving the utilization of ciprofloxacin dosage forms. Phase I (365 days; November 12, 2002 to November 11, 2003) was designed to characterize ciprofloxacin usage patterns under the existing IV-PO Step-down Program. Phase II (120 days; November 12, 2003 to March 11, 2004) was designed to characterize the impact of the new Pharmacist-managed Dosage Form Conversion Service (implemented on November 12, 2003) on the relative utilization of the ciprofloxacin IV and PO dosage forms.
The primary endpoint was the relative utilization of IV and PO ciprofloxacin by dose (IV:PO ciprofloxacin use ratio). The secondary endpoints were the overall utilization of ciprofloxacin (by dose), the proportion of total IV doses considered to be inappropriate, and relative IV and PO total and treatment course acquisition costs between the two phases. Potential cost avoidance associated with the intervention was estimated from the data.
Intervention
Prior to the implementation of the program, pharmacists were educated on the approved conversion service through in-house presentations. A newsletter was distributed to all medical staff detailing the program [25].
Decentralized clinical pharmacists on the medical wards were expected to conduct target drug report reviews 5 days per week to identify inpatients who had been prescribed ciprofloxacin IV. Health records were then reviewed, and patients assessed to determine if IV-PO conversion criteria were met. A patient was eligible for IV-PO dosage form conversion after 48 hours of IV therapy if he/she 1) continued to need an antibiotic; 2) was clinically stable; 3) was capable of tolerating the PO dosage form; and 4) had no factors present that would adversely affect PO bioavailability (e.g. gastrointestinal abnormalities or drug interactions). Pharmacists could consult the Infectious Diseases service or Infectious Diseases Pharmacist at anytime with any questions regarding IV-PO conversion eligibility.
For patients who met the conversion criteria, the pharmacist would write the order for the PO regimen in the Physician's Orders section of the health record. If the pharmacist wanted to convert the patient to PO ciprofloxacin prior to 48 hours of IV therapy, they would first confer with the physician. In collaboration with the healthcare team, the pharmacist would monitor the patient for clinical progress and medication tolerability, and could convert the patient back to IV therapy as required.
A randomly selected convenience sample of 200 ciprofloxacin IV treatment courses (100 treatment courses per phase) was considered to be adequate to determine the impact of this new intervention. These courses were identified using computerized pharmacy records and a random sample was undertaken using a computer-generated number list.
All patients who were ordered ciprofloxacin IV were included in the sample collection. Patients were excluded if they did not receive any doses of ciprofloxacin IV, if their ciprofloxacin IV treatment course did not occur within the pre-specified phase to which they were randomized, or if their charts were unavailable from the Health Records department at VGH as of July 14, 2004.
Charts were reviewed to gather demographic data, ciprofloxacin utilization information, inappropriate IV doses, and pharmacist-preventable IV doses of ciprofloxacin.
Data collection and analysis
Data was collected by one investigator and entered into statistical analysis software (SPSS© Version 11.0). Any treatment course that this investigator considered to have four or more inappropriately administered ciprofloxacin IV doses was reviewed in collaboration with the coordinating investigator to ensure accuracy of interpretation.
Inferential statistics were performed. A two-sample Student's t-test was used for parametric data, the Mann-Whitney test was used for non-parametric data, and the Fisher's Exact and Chi-square tests were used for proportional analyses.
Definitions
For the purposes of this study, a ciprofloxacin IV dose was considered "inappropriate" when the patient met the criteria for use of the PO dosage form. A ciprofloxacin IV dose was considered "pharmacist-preventable" if the dose was administered when the patient met the criteria for use of the PO dosage form and the decentralized clinical pharmacist was considered to have had the opportunity to intervene (i.e. Monday to Friday between 08:00 and 16:00 hours, excluding statutory holidays). The inappropriate IV-PO ciprofloxacin acquisition cost was the differential cost between the IV and the PO dosage form at current contract prices multiplied by the number of inappropriate IV doses administered.
Results
Two hundred and fifteen health records of patients who were prescribed ciprofloxacin IV during the study period were reviewed. Of these, seven patients were excluded, as the ciprofloxacin IV treatment courses were not completed within the pre-specified treatment phase. Six patients were excluded, as no IV ciprofloxacin doses were actually received. Health records were not accessible at the time of the study for the remaining two patients. Accordingly, 200 treatment courses for 200 patients (100 per phase) were included for analysis. This represented a 4% (100/2411 treatment courses) sampling rate for Phase I and a 10% (100/994 treatment courses) sampling rate for Phase II.
Patient demographics are presented in Table 1. Patients receiving ciprofloxacin IV were equally distributed by gender, typically in their sixth/seventh decade of life with an average duration of hospital stay of approximately two weeks. Treatment courses were initiated in both surgical and medical service areas for a wide variety of infectious indications. There were no significant differences between the two phases in terms of age, gender, renal function, length of stay, and medical service area to which the patients were assigned. Most patients (75% in Phase I, 78% in Phase II) received IV ciprofloxacin in combination with one or more antibiotics.
Table 1 Patient demographics
Phase I Nov. 12, 02 to Nov. 11, 03 (365 days) Phase II Nov. 12, 03 to Mar. 11, 04 (120 days)
No. of patients 100 100
No. of treatment courses 100 100
Age (yr), median (range) 57 (17–93) 63 (16–91)
Gender, N
Male 45 50
SCr1 (μmol/L), median (range) 84 (40–541) 89 (35–641)
Length of Stay (d), mean (range) 12 (1–84) 17 (1–165)
Service Area, N
General Surgery 31 30
Medicine 22 11
Emergency 15 11
Intensive Care Unit 7 8
Urology 3 12
Other 222 283
Indication, N
Off-label indications4 38 35
Intra-abdominal infection 18 15
Respiratory tract infection 15 16
Urinary tract infection 15 15
Other 145 196
1Serum creatinine closest to start of ciprofloxacin IV treatment.
2 Other service areas: Thoracic, Respiratory, Spine, Hematology/BMT, Transplant, Cardiothoracic Surgery, Neurosciences Intensive Care Unit, Cardiology, Day Bed Unit, Same Day Admit Unit, Pre-admission Clinic, Trauma Special Care Unit, Neurosciences, Family Practice, Orthopedics, Vascular, and Gynecology.
3 Other service areas: Thoracic, Spine, Hematology/BMT, Transplant, Cardiology, Day Bed Unit, Same Day Admit Unit, Neurosciences, Family Practice, Vascular, Gynecology, Palliative Care, and Trauma.
4Off-label indications refer to those not approved on the manufacturer's drug monograph.
5Other indications: Empiric therapy in febrile neutropenia, skin and soft tissue, and septicemia.
6 Other indications: Empiric therapy in febrile neutropenia, skin and soft tissue, septicemia, and infectious diarrhea.
Of the 200 ciprofloxacin IV courses reviewed, the total number of doses and the ratio of IV to total doses across phases were similar (p = 0.2830) (Figure 1). The IV:PO ciprofloxacin use ratio was 3.03 in Phase I vs. 3.48 in Phase II.
Figure 1 Total number of ciprofloxacin doses. p = 0.2830 for ratio of IV to total number of ciprofloxacin doses between phases.
Ciprofloxacin treatment characteristics are described in Table 2. No significant differences were observed between the initial ciprofloxacin dosing strengths (p = 1.00), the initial dosing frequencies (p = 0.55), and the number of IV-PO conversions per treatment course (p = 0.73). IV-PO ciprofloxacin conversion occurred in 27/100 (27%) of IV treatment courses in Phase I and 23/100 (23%) of courses in Phase II (Table 2). The number of IV-PO conversions that were subsequently reversed to IV was 2 cases in Phase I and 3 cases in Phase II. Chart documentation of a pharmacist-initiated IV-PO conversion was recorded in 3/27 (11%) episodes in Phase I and 4/23 (17%) episodes in Phase II. There was no difference between phases with respect to the median number of ciprofloxacin doses administered and the median costs associated with each treatment course (Table 2).
Table 2 Ciprofloxacin treatment course characteristics
Phase I Phase II P value
Treatment regimen characteristics
Dosing Strength 1.00
200 mg IV 6 6
400 mg IV 94 94
Dosing Frequency 0.55
Once 17 22
Once daily 5 3
Twice Daily 78 75
IV to PO Conversion Rate (% by treatment course) 27 23 0.73
Ciprofloxacin doses/treatment course, median (range) 5 (1–33) 5 (1–44) 0.55
IV, median (range) 3 (1–33) 4 (1–25) 0.29
Inappropriate IV, mean (range) 2.4 (0–26) 2.0 (0–9) 0.33
Inappropriate & pharmacist-preventable IV, mean (range) 1.1 (0–24) 0.6 (0–6) 0.14
PO, mean (range) 1.7 (0–26) 1.6 (0–26) 0.83
Treatment course acquisition costs ($)
Ciprofloxacin, median (range) 99 (17–1089) 132 (17–825) 0.32
IV, median (range) 99 (17–1089) 132 (17–825) 0.28
Inappropriate IV, mean (range) 72 (0–790) 60 (0–274) 0.39
Inappropriate & pharmacist-preventable IV, mean (range) 34 (0–729) 20 (0–182) 0.17
For those patients who met the criteria for the use of an oral dosage form, 59/100 (59%) received one or more inappropriate doses of ciprofloxacin IV in Phase I compared to 61/100 (61%) in Phase II. There was a significant decrease in the proportion of inappropriate ciprofloxacin IV doses between phases (244/521 (47%) in Phase I vs. 201/554 (36%) in Phase II (p = 0.0005) (Figure 2). Furthermore, there was a significant reduction in the proportion of pharmacist-preventable inappropriate ciprofloxacin IV doses between Phase I and Phase II (114/244 (47%) vs. 65/201 (32%) (p = 0.0026) (Figure 2).
Figure 2 Number of total, inappropriate, and pharmacist-preventable inappropriate IV ciprofloxacin doses. p = 0.0005 for difference in the proportions of inappropriate IV ciprofloxacin doses between phases. p = 0.0026 for difference in the proportions of pharmacist-preventable inappropriate IV ciprofloxacin doses between phases.
The total cost of IV and PO ciprofloxacin for the treatment courses reviewed was $16,993 in Phase I and $18,332 in Phase II. Ciprofloxacin IV accounted for $16,517 (97%) of total ciprofloxacin costs in Phase I and $17,919 (98%) of these costs in Phase II (Figure 3). The proportional cost avoidance associated with inappropriate use of IV ciprofloxacin was $7,172/$16,517 (43%) in Phase I compared to $6,012/$17,919 (34%) in Phase II (p = 0.001). The proportional pharmacist-preventable cost avoidance associated with inappropriate IV ciprofloxacin use was reduced from $3,367/$16,517 (20%) in Phase I to $1,975/$17,919 (11%) in Phase II (p = 0.001).
Figure 3 Costs associated with total, inappropriate, and pharmacist-preventable inappropriate IV ciprofloxacin doses. p = 0.001 for difference in potential cost avoidance of inappropriate IV ciprofloxacin doses between phases. p = 0.001 for difference in potential cost avoidance of pharmacist-preventable inappropriate IV ciprofloxacin doses between phases.
Discussion
The purpose of this study was to assess the relative utilization of the IV and PO dosage form of ciprofloxacin subsequent to the implementation of the Pharmacist-managed Dosage Form Conversion Service. We did not aim to assess the appropriateness of ciprofloxacin usage for specific indications.
Overall, the IV:PO ratio of ciprofloxacin usage remained similar between the two phases, and the total number of ciprofloxacin doses did not change significantly. Initially, it was anticipated that implementation of the conversion service would reduce the IV:PO ratio, however, there were numerous variables that may have affected this endpoint. We were also interested in evaluating whether the number of inappropriate IV doses and pharmacist-preventable inappropriate IV doses could be reduced. Our results showed a 23% relative reduction in the proportion of inappropriate ciprofloxacin IV doses and a 32% relative reduction in the incidence of pharmacist-preventable inappropriate ciprofloxacin IV doses subsequent to the intervention.
A possible explanation for the decline in the inappropriate and pharmacist-preventable inappropriate ciprofloxacin IV doses was the drive to reduce hospital expenditures at our institution at the time this program was implemented. There was an increased awareness and emphasis for cost-effective prescribing. The conversion service was an adjunct to the cost savings initiatives and was readily adapted to our established practice.
VGH has had a pre-existing IV-PO Step-Down Program since 1992. In a previous study at our institution in 1992, the rate of patients eligible for IV-PO conversion was 52%, which is similar to the 59% observed in Phase I of our present study [6]. The rate of IV-PO conversions in 1992 was also comparable to our baseline in Phase I (34% vs. 27%, respectively). This suggests that the interventions of our IV-PO Step-down Program have remained relatively constant since 1992. With an effective IV-PO Step-down Program in place, it is possible that the magnitude of change associated with the implementation of the conversion service may have been blunted.
Pharmacist-initiated IV-PO conversion was documented in the health record in 4 cases in Phase I and 3 cases in Phase II. This low incidence of documentation may be attributed to the activities of the decentralized clinical pharmacists who attend patient care rounds and interact directly with physicians, so that orders are written during rounds for IV-PO conversion. Greater awareness of other health care professionals on the bioavailability of PO ciprofloxacin may also have resulted in the earlier usage of the PO dosage form.
Total costs of ciprofloxacin therapy were similar between the two phases. This was expected, as the conversion service would not alter the indications for ciprofloxacin use. However, the costs associated with inappropriate ciprofloxacin IV therapy and pharmacist-preventable inappropriate ciprofloxacin IV therapy declined from Phase I to II, which may be attributed to the increased interventions of the clinical pharmacists and the improved awareness for PO therapy post intervention.
Following implementation of the conversion service, 12% (65/554) of ciprofloxacin doses deemed inappropriate and preventable by pharmacists were still administered. Ideally, all of these doses should have been avoided. One full working day (excluding weekends and statutory holidays) was allotted as the time required for clinical pharmacists to assess these patients. This delay in IV-PO conversion may be explained in part by having our data collection period immediately after the introduction of the new service, as pharmacists may not yet have been comfortable exercising a dosage form conversion autonomously. The retrospective assessment for appropriateness by the investigator may also differ from that of the clinical pharmacist. It would be beneficial to obtain an internal assessment to discover the barriers associated with the program. Of course, it would be preferable to educate the medical staff to initiate PO regimens where indicated and avoid the use of the IV formulation.
Several limitations exist with this study. The retrospective, pre/post, unblinded design precludes the formulation of any direct causal relationships between the implementation of the conversion service and the subsequent reduction in inappropriate ciprofloxacin IV doses. The sample size of convenience may not have achieved the power required to detect a difference. In addition, the sampling rate was relatively low at 4.1% (100/2411 ciprofloxacin IV courses) in Phase I and 10.1% (100/994) in Phase II, and thus may not truly represent the characteristics of our population. Time restrictions resulted in differences in sampling periods between Phase I (365 days) and Phase II (120 days), which may affect the representation of ciprofloxacin IV treatment courses throughout the year. However, this should not directly influence the proportion of inappropriate and pharmacist-preventable inappropriate ciprofloxacin IV treatment doses. The assessment of this service was over a relatively short period, and so may not truly reflect the long-term impact of this program.
As with any unblinded study of this type, the potential for investigator assessment bias existed. As the evaluation of dose appropriateness was undertaken sequentially across phases, potential bias may have been introduced as the investigator gained experience during the process. To minimize this bias, charts were reviewed by a senior investigator when greater than four inappropriate and pharmacist-preventable inappropriate ciprofloxacin IV doses were identified by the junior investigator. A 100% concordance existed between the assessments made by the junior and senior investigators.
In the context of drug use optimization and cost minimization, implementation of a Pharmacist-managed IV-PO Dosage Form Conversion Service may be used to facilitate appropriate, cost-effective therapy. This service can be used in conjunction with other established methods including newsletters [6], chart talkers, notes [7,8], and direct pharmacist-physician interactions. To further promote antimicrobial use appropriateness, strategies aimed at affecting prescribing behaviour may be employed. These include individual physician prescribing feedback, multidisciplinary inservices in collaboration with infectious diseases physicians, and prescriber education through academic detailing.
Conclusion
In summary, the overall utilization of ciprofloxacin seems to have remained unchanged and the proportion of IV to total doses appears stable. However, the proportion of inappropriate IV doses and its associated costs appear to have declined subsequent to the implementation of a Pharmacist-managed IV-PO Dosage Form Conversion Service. Such a program may be a beneficial adjunct in facilitating appropriate and cost-effective usage of ciprofloxacin.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
BPH participated in the design of the study, performed data collection and analyses, and drafted the manuscript. TTYL participated in the design and coordination of the study, performed data and statistical analyses, and drafted and revised the manuscript. RMB participated in the design of the study, developed the analytical database, and revised the manuscript. TLN participated in the design of the study and revision of the manuscript. PJJ conceived the study, participated in its design, performed data and statistical analyses, and drafted and revised the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Bayer Inc Repchinsky C Cipro® product monograph Compendium of pharmaceuticals and specialties 2004: the Canadian drug reference for health professionals 2004 Ottawa, ON Canadian Pharmacists Association 425 428
Echols RM Antimicrobial practice. The selection of appropriate dosages for intravenous ciprofloxacin J Antimicrob Chemother 1993 31 783 787 8335506
Lettieri JT Rogge MC Kaiser L Echols RM Heller AH Pharmacokinetic profiles of ciprofloxacin after single intravenous and oral doses Antimicrob Agents Chemother 1992 36 993 996 1510426
Drugs and Therapeutics Committee Shalansky K, Hill S. Vancouver, BC Formulary of Vancouver Hospital and Health Sciences Centre 2003 Vancouver Hospital and Health Sciences Centre 14
Frighetto L Martinusen SM Mamdani F Jewesson PJ Ciprofloxacin use under a reserved drug and stepdown promotion program Can J Hosp Pharm 1995 48 35 42 10141061
Frighetto L New drug and drug products: parenteral ciprofloxacin (Cipro IV®) Vancouver Hospital and Health Sciences Centre Drugs and Therapeutics Newsletter 1992 Vancouver, BC: Department of Pharmacy, Vancouver Hospital and Health Sciences Centre
Bunz DM Frighetto L Gupta S Jewesson PJ Simple ways to promote cost containment DICP 1990 24 546 2343599
Martinez MJ Freire A Castro I Inaraja MT Ortega A Del Campo V Rodriguez I Bardan B Morano LE Garcia JF Clinical and economic impact of a pharmacist-intervention to promote sequential intravenous to oral clindamycin conversion Pharm World Sci 2000 22 53 58 10849923 10.1023/A:1008769204178
Davis C Sequential intravenous/oral ciprofloxacin as an empiric antimicrobial therapy: results of a Canadian multicenter study Clin Ther 1994 16 505 521 7923317
Fox ER Beckwith MC Tyler LS Pharmacy-administered IV to oral therapeutic interchange program: development, implementation, and cost-assessment Hosp Pharm 2003 38 444 452 462
Malfair SC Frighetto L Nickoloff DM Martinusen SM Jewesson PJ Evaluation of the use of cefuroxime and cefuroxime axetil in an intravenous-oral stepdown program Ann Pharmacother 1996 30 337 342 8729884
Marra CA Frighetto L Quaia CB de Lemos ML Warkentin DI Marra F Jewesson PJ A new ciprofloxacin stepdown program in the treatment of high-risk febrile neutropenia: a clinical and economic analysis Pharmacother 2000 20 931 940 10.1592/phco.20.11.931.35258
Zamin MT Pitre MM Conly JM Development of an intravenous-to-oral route conversion program for antimicrobial therapy at a Canadian tertiary care health facility Ann Pharmacother 1997 31 564 570 9161649
Grasela TH JrPaladino JA Schentag JJ Huepenbecker D Rybacki J Purcell JB Fiedler JB Clinical and economic impact of oral ciprofloxacin as follow-up to parenteral antibiotics DICP 1991 25 857 862 1949945
Drew RH Programs promoting timely sequential antimicrobial therapy: an American perspective J Infect 1998 3 9 9756363 10.1016/S0163-4453(98)92677-9
Hunter KA Dormaier GK Pharmacist-managed intravenous to oral step-down program Clin Ther 1995 17 534 540 7585857 10.1016/0149-2918(95)80119-7
Kirking DM Svinte MK Berardi RR Cornish LA Chaffee BW Ryan ML Evaluation of direct pharmacist intervention on conversion from parenteral to oral histamine H2-receptor antagonist therapy DICP 1991 25 80 84 1672572
Anon Pharmacy drug monitoring service: sequential drug therapy Lions' Gate Hospital Policy and Procedure Manual 2003 North Vancouver, BC: Lions' Gate Hospital C2.3
Maidment ID Why use intravenous antibiotics when oral will do? The Pharmaceutical J 1998 261 630 632
Okpara AU Maswoswe JJ Stewart K Criteria-based antimicrobial IV to oral conversion program Formulary 1995 30 343 348 10144873
Ramirez JA Antibiotic streamlining: development and justification of an antibiotic streamlining program Pharm Pract Manag Q 1996 16 19 34 10166232
Roberts BL Jr Decentralizing an i.v.-to-oral conversion program Am J Health Syst Pharm 1997 54 524 525 9066858
Anon IV to oral step-down St Paul's Hospital Medication Manual 2003 Vancouver, BC: St. Paul's Hospital
Wong-Beringer A Nguyen KH Razeghi J Implementing a program for switching from i.v. to oral antimicrobial therapy Am J Health Syst Pharm 2001 58 1146 1149 11449860
Anon Drug Cost Containment Strategies Vancouver Hospital and Health Sciences Centre Drug and Therapeutics Newsletter 2003 10 Vancouver, BC: Pharmaceutical Sciences Clinical Services Unit, Vancouver Hospital and Health Sciences Centre 1 3
|
15987523
|
PMC1185535
|
CC BY
|
2021-01-04 16:31:52
|
no
|
BMC Health Serv Res. 2005 Jun 29; 5:48
|
utf-8
|
BMC Health Serv Res
| 2,005 |
10.1186/1472-6963-5-48
|
oa_comm
|
==== Front
BMC Health Serv ResBMC Health Services Research1472-6963BioMed Central London 1472-6963-5-491601180610.1186/1472-6963-5-49Research ArticleThe development and implementation of a regional network of physiotherapists for exercise therapy in patients with peripheral arterial disease, a preliminary report Willigendael EM [email protected] BLW [email protected] der Berg C [email protected] RJThJ [email protected] MH [email protected] de RA [email protected] JAW [email protected] Atrium Medical Centre, Department of Surgery, Division of Vascular Surgery, Heerlen, The Netherlands2 Atrium Medical Centre, Department of Cardiology, Heerlen, The Netherlands3 Department of Epidemiology, University of Maastricht / KEMTA, Maastricht, The Netherlands2005 12 7 2005 5 49 49 2 12 2004 12 7 2005 Copyright © 2005 Willigendael et al; licensee BioMed Central Ltd.2005Willigendael 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
Exercise therapy (ET) is the main conservative and proven effective treatment of patients with intermittent claudication. Currently, the most frequent exercise prescription is a single 'go home and walk' advise, without supervision or follow-up. There is no evidence to support the efficacy of this advise and compliance is known to be low. Therefore, a systematic approach was used to guarantee quality and standardisation of treatment, optimal guideline adherence and improved of inter-professional communication between vascular surgeons and physiotherapists. In this preliminary report we would like to outline the steps taken for the development and implementation of the Network Exercise Therapy Parkstad
Methods
In October 2003 all 59 regional physiotherapy practices were invited to attend a symposium regarding ET in a physiotherapeutic setting. Attending physiotherapists interested in providing ET and willing to follow a certified course on ET, were asked to register. Three tastkgroups were formed to accomplish the set targets: Exercise therapy education, Exercise therapy implementation and continuity, and Inter-professional communication in the Parkstad region.
Results
In total 27 physiotherapists, from 22 different practices followed the educational program and are now trained and accredited to provide ET according to the guideline of the Royal Dutch Society for Physiotherapy. A web-based database wasdesigned to contain information on disease specific items provided by the vascular surgery department, and aspects with respect to ET registered by the physiotherapist. The information is regularly updated and available online. Access tothe database is restricted to vascular surgeons and physiotherapists in the network. The secondary purpose of the database is to register essential benchmark data for future analysis of ET in a physiotherapeutic setting in the Netherlands and to enable physiotherapists continuous feedback on patient performance. A triage system was developed to detect patients with a compromised cardiac history. This group receives ET at the in-hospital department of revalidation with the possibility of immediate consultation of a cardiologist in case of cardiac complications or even CPR.
Conclusion
The Network Exercise Therapy Parkstad of supervised ET is the first initiative in the Netherlands to provide ET close to the patient's home environment. With the implementation of supervised ET in an outpatient physiotherapeutic setting for all eligible patients with symptomatic PAD, the access to care has been improved. A web-based communication system provides physiotherapists and vascular surgeons with all the necessary and continues updated patient information. Future research, currently in progress, will investigate the therapeutic benefits and cost-effectiveness of exercise therapy in a physiotherapeutic setting.
==== Body
Background
Patients with peripheral arterial disease (PAD), with intermittent claudicationas the predominant clinical symptom, experience muscle aching or cramp during walking, secondary to muscle ischemia in the calf, thigh or buttocks. PAD is a manifestation of systemic atherosclerosis. The treatment of patients with PAD stage II (according to Fontaine) consists of vascular risk factor management, smoking cessation and exercise therapy (ET).
ET is the main conservative treatment for patients with intermittent claudication and proven effective.[1] The psychological, metabolical, and mechanical alterations that occur during the periods of exercise stimulate an adaptive response that ultimately reduces the symptoms of intermittent claudication. Besides the improvement in maximal walking distance, ET contributes to an increase in quality of life and a decrease in the number of vascular interventions.[2] Furthermore, with adequate ET, hypertension, hypercholesterolaemia, overweight, and diabetes, if present, are better regulated. These positive results of exercise therapy have been observed during training programs in hospitals or rehabilitation clinics. [3,4] However, major disadvantages of these hospital-based programs are that they are costly, their availability is limited, and patients are taken out of their family and working environment.
Currently, the most frequent exercise prescription for patients with PAD in theNetherlands, and considered best practice according to the Dutch general practitioners and vascular surgeons claudication intermittens guidelines, is a single'go home and walk' advise, without supervision or follow-up. A leaflet on unsupervised ET provided by the Vascular Patients Society is available. There is no evidence to support the efficacy of this advise and compliance is known to be low. [5,6] In studies comparing the 'go home and walk' advise to supervised ET, a clear advantage for supervised ET was present. [6-9] The ineffectiveness of home based ET is to a large extent caused by the pain caused by ET and the subsequent reluctance to exercise again. Factors like fear of pain, inadequate knowledge and poor general condition, contribute to the difficulty to start, sustain and maintain ET. ET in a physiotherapeutic setting has the benefits of adequate coaching and provides the stimulation and supervision deemed necessary to provide an advantage over the 'go home and walk' advice. Furthermore, the close contact with patients during ET in a physiotherapeutic setting provides coaching in the necessary changes in life-style, like weight control and smoking cessation.
The first national physiotherapeutic guideline on ET was issued in December 2003 by the Royal Dutch Society for Physiotherapy.[10] This guideline enablesthe facilitation of professional ET in the future. However, a national cross-sectional survey performed among 265, randomly selected Dutch physiotherapists showed deficits in the physiotherapists theoretical and practical skills regarding ET.[11] This illustrates again the fact that the existence of a guideline does not warrant optimal care or automatic goal attainment.
Therefore, a systematic approach was used to guarantee quality and standardisation of treatment, optimal guideline adherence and improved inter-professional communication between vascular surgeons and physiotherapists.
To improve the current care for patients with intermittent claudication, we setthe following goals:
• Optimalisation of the current knowledge and skills of the physiotherapists
• Improve the screening of patients allegeable for ET and to provide a safety-net for patients with a high cardiovascular risk
• ET should be accessible close to the patients home address
• Implementation of a continues monitoring and exchange of inter-professional patient information system
To achieve these goals, the so-called Network Exercise Therapy Parkstad steering committee was founded in October 2003. This steering committee consisted of avascular surgeon, three nurse practitioners of the vascular surgery department, an epidemiologist, five physiotherapists and a research fellow of the vascular surgery department. To our knowledge this concept is innovative for the Netherlands and abroad. Before future research will commence, we would like to outlinethe steps taken for the development and implementation of the Network ExerciseTherapy Parkstad
Methods
Exercise therapy network development
In October 2003 all 59 regional physiotherapy practices were invited to attend a symposium regarding ET in a physiotherapeutic setting. The symposium covered topics on the theoretical background, therapeutic benefits, and an introductionto the new Royal Dutch Society for Physiotherapy guideline on ET. An important issue of the symposium was an open discussion on the formation of the Network Exercise Therapy Parkstad to facilitate supervised ET for patients referred by vascular surgeons from the Atrium Medical Centre. At the end of the symposium, attending physiotherapists interested in providing ET, and in the possession of, or the willingness to acquire, a treadmill, and willing to follow a certified course on ET, were asked to register. From this group of physiotherapists we planned to invite five physiotherapists to join the aforementioned steering committee. Our aim was to obtain both physiotherapists from the Department of Revalidation in our hospital as well as regional physiotherapists.
Three taskgroups within the steering committee were formed to accomplish the set targets:
• Exercise therapy education
• Exercise therapy implementation and continuity
• Inter-professional communication in the Parkstad region
Exercise therapy education
To overcome the noted knowledge deficiencies of ET, this taskgroup organised anaccredited two-day 'in-hospital' educational program on ET provided by the Royal Dutch Society for Physiotherapy and the Dutch Paramedic Institute. This course had already been developed, and provides the necessary knowledge and skills to give professional supervised ET. The course covered the necessary theoreticalbackground, and practical skills to provide ET. Three months later, this coursewas followed by a one day follow-up training. This training took place to exchange practical experiences, and to pay extra attention to the use of clinimetrics, standardisation and optimisation of the protocol, as well as getting an introductory session on the use of the electronic patient file in the web-based database.
Exercise therapy implementation and continuity
In some patients with PAD, ET in an outpatient physiotherapeutic setting interferes with the present high cardiovascular risk. To provide ET for patients witha compromised cardiac history in a relatively safe environment, the second taskgroup developed a triage system to filter these patients prior to the ET prescription. In close collaboration with the department of cardiology, a decision tree has been developed.(Figure 1) This enabled the nurse practitioners at the department of vascular surgery to decide if ET can be performed in an outpatient setting or at the in-hospital department of revalidation with the possibility of immediate consultation of a cardiologist in case of cardiac complications or even CPR.
Figure 1 Decision tree to triage patients with a compromised cardiac history.
The aforementioned leaflet on unsupervised ET of the Vascular Patients Society, which has been used throughout the Netherlands for over 10 years, needed revision. The Vascular Patients Society gave us the opportunity to suggest some alterations in this leaflet to make it compliant with the possibility of supervised ET. In this revised version attention was given to the new insights on PAD, risk factor management and supervised as well as unsupervised ET.
ET requires a life-long adaptation of the patient's life-style, in which a continuation of daily walking plays an essential role. ET in a physiotherapeutic setting is only covered by medical insurance in the Netherlands for the durationof one year. To support the patient with the necessary life-style changes, the 'Regional Exercise Therapy Walking group Parkstad' has been formed for patientsto enrol after the termination of the supervised ET after a year.
Inter-professional communication in the Parkstad region
The last taskgroup tried to solve the so far absent or at least insufficient communication between the department of vascular surgery and physiotherapy practices. In close collaboration with the University of Maastricht, department of epidemiology, a web-based database was developed. The purpose of this database was twofold. The primary purpose was to facilitate adequate information exchange between vascular surgeons/nurse-practitioners and physiotherapists. The database was designed to contain information on disease specific items like the extentof PAD, present risk factor and co-morbidity provided by the vascular surgery department. The physiotherapist registered aspects with respect to the ET, like therapy progress and difficulties as well as patient compliance. The information is regularly updated and online available. The secondary purpose of the database is to register essential data for future analysis of ET in a physiotherapeutic setting and to enable physiotherapists to receive continuous feedback (benchmark data) on their performance. Furthermore, this database has been designed to be implemented nationwide, and when accomplished, a substantial and important level of evidence on the value of physiotherapy will be available. Similar databases are currently being developed for other therapies in a physiotherapeutic setting.
Results
Exercise therapy network implementation
Exercise therapy education
64 physiotherapists attended the symposium on the future of ET in the Parkstad region from 45 regional physiotherapy practices. Five physiotherapists, two from the Department of Rehabilitation, Atrium medical centre, and three regional physiotherapists joined the steering committee. In total 27 physiotherapists, from 22 different practices followed the educational program and are now trained and accredited to provide ET according to the guideline.
Exercise therapy implementation and continuity
Patients with a serious cardiac medical history receive ET in a hospital setting, but the vast majority is referred to a local physiotherapist. (Figure 2) Therevised ET leaflet of the Vascular Patients Society has been nationally released in the autumn of 2004. Students of the regional physiotherapy school, under supervision of a steering committee physiotherapist are coaching the 'Regional Exercise Therapy Walking group Parkstad'. The primary goal of this walking group is to stimulate patientsto remain physically active after the termination of the supervised ET. The walking group organises regular walks every 14 days, in which life style changes, like smoking cessation, weight control, and contacts with other patients play a central role.
Figure 2 Regional physiotherapy practice participation and distribution.
Inter-professional communication in the Parkstad region
The web based database provides vascular surgeons and physiotherapists with essential information. Before the patients are scheduled for a repetitive appointment at the hospital the nurse practitioner has access to all therapy related information. The same principle applies to the physiotherapist who receives all the medical updates. With this information the physiotherapist is better equipped to make an adequate judgement on the exercise capacity of the patient and thevascular surgeon can, based on therapy progress and compliance, make an adequate judgement on the necessity of a vascular intervention. Access to the databaseis restricted to vascular surgeons and physiotherapists in the network.
In June 2004 the Network Exercise Therapy Parkstad has started. It is the firstnational network to provide supervised ET in a physiotherapeutic setting close to the patients home. All patients with symptomatic PAD (Fontaine stage II) presented at the vascular surgery outpatient clinic who are considered for maximalnon-invasive treatment, are now primarily referred to one of the physiotherapists in the Network Exercise Therapy Parkstad for supervised ET.
Discussion
In the 1998 report of the Dutch Heart Foundation on obstacles in vascular care, the inadequate ET facilities and infrastructure was noted.[12] Recommendations were made to stimulate the use of ET and to develop an ET infrastructure. Six years later the Network Exercise Therapy Parkstad of supervised ET is the first initiative to provide ET close to the patient's home. Supervised ET has been provided on a small scale in the Netherlands, primarily hospital based. Withthe implementation of supervised ET in an outpatient physiotherapeutic setting for patients with symptomatic PAD, an attempt has been made to improve the quality of care, and the communication between the different care providers.
Less than half (49%) of the physiotherapists that attended the symposium, eventually jointed the network. This seems disappointing, but taken into account that only 30% of the Dutch practices is in the possession of the required treadmill, the results were better than expected. Furthermore, the majority of practices are already specialised in for instance, cardiac rehabilitation, pulmonary diseases, sport-physiotherapy or neurological disorders, leaving little room for an additional (and time consuming) speciality. The 22 practices that were capable to meet the set requirements, cover the region well and are highly motivated.
Network Exercise Therapy Parkstad
ET in a physiotherapeutic setting is more expensive than the 'go home and walk' advise. However, besides the increase in maximal walking distance, the prognosis of general health (diabetes, high cholesterol, hypertension, bodyweight, blood pressure, quality of life, life-style changes like smoking, and an inactive life-style) may improve. Potential differences in costs are likely to result from less morbidity and medical consumption related to PAD complications, vascular surgery, and associated cardiovascular diseases. It is expected that these savings will exceed the costs of ET in a physiotherapeutic setting. However, a true cost-effectiveness analysis should answer this question. On the other hand, ET in hospitals or rehabilitation clinics, as have been practiced abroad, is more costly. The availability of ET close to the patient's home address is one ofthe advantages of the Network Exercise Therapy Parkstad and is most likely to substantially improve therapy participation and compliance. The group that has not the benefit of supervised ET 'around the corner', is the 'safety-net group' for patients with a high cardiovascular risk. In the previous situation, these patients would not be engaged in ET or suffer an increased risk during unsupervised ET. Providing ET in a hospitalised setting enables patients to slowly reclaim their health in a protective environment. The developed decision tree (which has been incorporated in the web-based database), enables also vascular nursepractitioners to identify these high-risk patients.
Inter-professional communication in the Parkstad region
The Network Exercise Therapy Parkstad requires a different way of working and an intensified manner of communication between the department of vascular surgery and the participating physiotherapists. Working with the web-based database is, compared to the old situation, a new way of working and relatively time consuming. Future research will show the impact and consequences of this database.
Conclusion
This preliminary report describes the development and implementation of the Network Exercise Therapy Parkstad. The main goal of this regional network was to improve the access to care, and the communication between the different care providers. Future research, currently in progress, will investigate the therapeutic benefits and the cost-effectiveness of exercise therapy in a physiotherapeutic setting. The Network Exercise Therapy Parkstad is now operational and general practitioners have been invited to join and refer patients from the primary care setting. To guide other interested regions in similar projects an extensive implementation guideline is available.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
EW concept development and implementation of network and writing of the manuscript. BB management and implementation of network, preparation of manuscript. CB development and implementation of cardiac decision tree. RW concept implementation and critical review of manuscript. MP participated in the design of the concept and draft of manuscript. RB development of database, concept implementation and critical review of manuscript. JT primary concept development and implementation of network and writing of 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 would like to thank all participating physiotherapists from the Network Exercise Therapy Parkstad for their efforts and enthusiasm.
==== Refs
Leng GC Fowler B Ernst E Exercise for intermittent claudication Cochrane database Syst Rev 2000 2 CD000990 10796572
Stewart KJ Hiatt WR Regensteiner JG Hirsch AT Exercise training for claudication; a review N Engl J Med 2002 347 1941 51 12477945 10.1056/NEJMra021135
Franco A Legrand E Guidicelli H Quesda C Sarrazin R Gaultier R Results of physiotherapy of arterial disease at the stage of intermittent claudication by programed efforts training J Mal Vasc 1980 5 185 9 7462850
Schoop W Methods and results of physiotherapy in stage II arteriopathies J Mal Vasc 1980 5 181 4 7462849
Bartelink MEL Stoffers HEJH Biesheuvel C Hoes AW Walking exercise in patients with intermittent claudication; experience in routine clinical practice Br J Gen Pract 2004 54 196 200 15006125
Patterson R B Pinto B Marcus B Value of a supervised exercise program for the therapy of arterial claudication J Vasc Surg 1997 25 312 9 9052565
Cheetham DR Burgess L Ellis M Williams A Greenhalgh RM Davies AH Does supervised exercise offer adjuvant benefit over exercise advice alone for the treatment of intermittent claudication? A randomised trial Eur J Vasc Endovasc Surg 2004 27 17 23 14652832 10.1016/j.ejvs.2003.09.012
Degischer S Labs KH Hochstrasser J Aschwanden M Tschoepl M Jaeger KA Physical training for intermittent claudication: a comparison of structured rehabilitation versus home-based training Vasc Med 2002 7 109 15 12402991 10.1191/1358863x02vm432oa
Regensteiner JG Meyer TJ Krupski WC Cranford LS Hiatt WR Hospital vs home-based exercise rehabilitation for patients with peripheral arterial occlusive disease Angiology 1997 48 291 300 9112877
Jongert MWA Hendriks HJM Hoek van J Klaasboer Kogelman K Robeer GG Simens B KNGF Richtlijn claudicatio intermittens Ned Tijd Fysiother 2003 6 3 50
Willigendael EM Teijink JAW Bartelink ML Boiten J Büller HR Prins MH Physiotherapists for exercise therapy in patients with intermittent claudication Submitted
Dutch Heart Foundation Vaatpatiënten in beeld; knelpunten in de zorg en aanbevelingen Dutch Heart Foundation 1998 The Hague, The Netherlands
|
16011806
|
PMC1185536
|
CC BY
|
2021-01-04 16:31:53
|
no
|
BMC Health Serv Res. 2005 Jul 12; 5:49
|
utf-8
|
BMC Health Serv Res
| 2,005 |
10.1186/1472-6963-5-49
|
oa_comm
|
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-411592151110.1186/1471-2334-5-41Research ArticleSalivary antibodies induced by the seven-valent PncOMPC conjugate vaccine in the Finnish Otitis Media Vaccine Trial Nurkka Anu [email protected] Mika [email protected] Arto AI [email protected]äyhty Helena [email protected] FinOM Study Group [email protected] Department of Vaccines, National Public Health Institute (KTL), Helsinki, Finland2005 27 5 2005 5 41 41 8 2 2005 27 5 2005 Copyright © 2005 Nurkka et al; licensee BioMed Central Ltd.2005Nurkka 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
Mucosal antibodies have been suggested to have a role in defence against pneumococcal infections. We investigated here the ability of a seven-valent pneumococcal conjugate vaccine, PncOMPC, to induce mucosal immune response.
Methods
Healthy Finnish children (n = 111), a subcohort of the Finnish Otitis Media Vaccine Trial, were recruited and 56 of them were immunised with the PncOMPC at the age of 2, 4, and 6 months. At 12 months of age, 49 of them received the PncOMPC and 7 were vaccinated with the pneumococcal polysaccharide vaccine (PncPS) as a booster. The control group of 55 children received a hepatitis B vaccine at the same ages. Salivary anti-Pnc IgG, IgA, IgA1, and IgA2 antibodies to serotypes 6B, 14, 19F, and 23F were measured in both groups at the age of 7 and 13 months.
Results
Salivary anti-Pnc IgG and IgA were detected more often in the PncOMPC than in the control group. However, the difference between groups was significant only for 19F and 23F IgA concentrations at the age of 7 months. At the age of 13 months, antibody concentrations did not differ between PncOMPC and control groups. The rises in IgA concentrations between 7 and 13 months of age were mainly of subclass IgA1. Further, there is a clear trend that PncPS booster induces higher salivary anti-Pnc PS antibody concentrations than the PncOMPC.
Conclusion
We found that PncOMPC can induce a mucosal IgA response. However, the actual impact of mucosal antibodies in protection against pneumococcal infections is not clear.
==== Body
Background
Streptococcus pneumoniae (Pnc) frequently colonises mucosal epithelium at nasopharynx without causing any symptoms [1]. The carriage rate varies depending on the age, being highest in children under two years of age [2]. Further, the prevalence of pneumococcal carriage is higher in developing than developed countries [3]. Although the pneumococcal carriage is often harmless, it may lead to a local disease, e.g. to acute otitis media (AOM), sinusitis or to an invasive disease like pneumonia, meningitis or sepsis [3].
The main mechanism of defence against pneumococcus are anticapsular antibodies, which help in the phagocytosis or which can counteract acquisition of pneumococcus probably by preventing adhesion to the mucosal surface [4]. In serum the predominant immunoglobulin class is IgG. The salivary IgG is mainly derived from serum by leakage across capillaries and entering saliva through gingival crevices. However, some local production of IgG may take place [5,6]. At mucosal membranes IgA is the main immunoglobulin class and it is found most often in the secretory form (sIgA). The role of serum IgG in the defence against pneumococcus is obvious; IgG can activate complement efficiently and further lead to phagocytosis of bacteria. The function of mucosal antibodies in humans is less clear. However, there are several pieces of evidence, which suggest that they do have a role in the defence. The presence of pneumococcus in nasopharynx induces salivary antibodies against pneumococcal proteins and polysaccharides already in infants [7,8], and pneumococcal conjugate vaccines evoke mucosal immune response [5,9-15]. In addition to invasive disease and pneumonia, pneumococcal conjugate vaccines prevent also local infections like AOM and carriage [16-22]. Further, in an animal model mucosal antibodies prevented the acquisition of pneumococcus [23].
At the moment there is only one licensed pneumococcal conjugate vaccine, PncCRM. The vaccine contains seven pneumococcal capsular polysaccharides (PS) conjugated to a non-toxic variant form of diphtheria toxin (CRM197). The Kaiser Permanente Efficacy Trial in the USA showed that PncCRM is highly protective, 97.4% (95% CI 88.7 to 99.9%), against invasive pneumococcal disease caused by vaccine serotypes [16]. Among American Indian children, which are a high risk population for pneumococcal infection, the intention-to-treat total primary efficacy of PncCRM against invasive disease was 82.6% (95% CI 21.4 to 96.1%) [24]. The efficacy of a 9-valent PncCRM in HIV-infected and uninfected children has been studied in Soweto, South-Africa [25]. The vaccine prevented there 83% (95% CI 39–97%) of invasive pneumococcal infections due to vaccine serotypes in HIV-uninfected children. The respective number was 65% (95% CI 24–86%) in HIV-infected children.
In the Finnish Otitis Media (FinOM) Vaccine Trial, two pneumococcal conjugate vaccines, PncCRM and PncOMPC, were investigated in parallel regarding the efficacy against AOM using hepatitis B vaccine (HBV) as a control for both arms. The efficacy of the PncCRM vaccine was 57% (95% CI 44 to 67%) [19] and of the PncOMPC vaccine 56% (95% CI 44 to 66%) [22] against AOM caused by the vaccine serotypes. In the setting of the FinOM Vaccine Trial we have had a possibility to evaluate both the serum [19,22] and mucosal immunity [10] as surrogates of vaccine efficacy [26].
The aim of this study was to measure the mucosal antibody responses after vaccination with 3 or 4 doses of either the PncOMPC or the control vaccine. Measurement of anti-Pnc antibodies against serotypes 6B, 14, 19F and 23F were selected because these types were most commonly associated with AOM in the FinOM studies [10,19,27]. We compared the prevalence and concentrations of salivary antibodies at 7 and 13 months of age in the PncOMPC and control group. We also explored the effect of pneumococcal polysaccharide vaccine (PncPS) as a booster. The data on salivary antibody in the PncCRM group has been reported separately [10].
Methods
Vaccines, vaccinees and vaccinations
PncOMPC (Merck & Co., Inc., West Point, PA) is a heptavalent pneumococcal conjugate vaccine containing 5 μg of capsular PS of the serotype 6B, 3 μg of type 23F, 2 μg of types 18C and 19F, 1.5 μg of type 9V, and 1 μg of types 4 and 14 each conjugated individually to an outer membrane protein complex (OMPC) of Neisseria meningitidis serogroup B. Aluminium was used as an adjuvant and thiomersal as a preservative. Pneumovax® (Merck & Co., Inc.) is a 23-valent pneumococcal PS vaccine (PncPS), containing 25 μg of each capsular PS. HBV (Recombivax ™ Merck & Co., Inc.) was used as a control vaccine. One dose of HBV vaccine contains 5 μg of hepatitis B surface antigen.
The design of the FinOM Vaccine Trial has been described previously [19,22]. In this study we had a subcohort of 111 infants. 56 children were randomised to receive the PncOMPC vaccine at the age of 2, 4, and 6 months. At 12 months of age, 49 of them were vaccinated with the PncOMPC and 7 with the PncPS as a booster vaccine. The control group of 55 children received HBV at the same ages.
Additionally, the children received a combined diphtheria-tetanus-whole cell pertussis and Haemophilus influenzae type b conjugate vaccine (Tetramune®, Wyeth Pharmaceuticals) at 2, 4, 6, and 24 months, inactivated polio vaccine (Imovax®, Aventis Pasteur) at 7, 12, and 24 months and measles-mumps-rubella (MMR®II, Merck & Co., Inc.) at 18 months of age.
Sampling
Unstimulated saliva samples were taken at the age of 7 (n = 111) and 13 (n = 107) months. Mothers were advised not to breast feed infants during one hour before saliva sampling. Up to 2 ml of saliva was collected with a gentle aspiration using electronic suction device and samples were immediately frozen at -70°C. Before analysis the samples were centrifuged for 10 minutes with 15 000 rpm and the supernatant was used for analyses of anti-Pnc PS antibodies. The saliva samples were thawed only once. The volume of the sample was not always sufficient for all analyses.
Measurement of antibodies
Enzyme immunoassay (EIA) for saliva samples
IgA, IgA1, IgA2, and IgG antibodies to serotypes 6B, 14, 19F and 23F were measured by EIA as described previously [5,10,11,14].
Before calculation of concentrations, optical density (OD) readings from the PBS wells were subtracted from the antigen plates. For IgA and IgG results OD values of ≥ 0.05 (≥ 2 standard deviations of the blank) were regarded as positive. The concentrations of IgA and IgG were calculated in nanograms per millilitre (ng/ml) of saliva by using the 89-SF serum as a reference. The detection limit was 5 ng/ml for both IgA and IgG assays and for all serotypes. Samples with undetectable IgA and IgG were assigned a value 1.7 ng/ml; half a log less than the detection limit. The IgA1 and IgA2 results are presented as EIA units/ml calculated using 89-SF as a reference with a given calibration factor [28]. The detection limit of IgA1 and IgA2 assays for all the serotypes was 1.3 units/ml; and samples with undetectable IgA1 and IgA2 concentrations were given a value 0.65 units/ml.
Statistical methods
Results are given as geometric mean antibody concentrations (GMC) with 95% confidence intervals (CI). Differences in antibody concentrations in saliva samples between different ages were analysed with Wilcoxon Signed Ranks Test. Kruskal-Wallis test was used to compare differences between PncOPMC and HBV groups. Proportions of children having anti-Pnc antibodies in saliva were compared with Yates-corrected chi square test (χ2-test) or with Fisher's two-tailed exact test. Differences were considered significant when p-value was <0.05. The sample size was not specifically determined for this study.
Results
Anti-pneumococcal IgG in saliva
Anti-Pnc PS IgG was detected at the age of both 7 and 13 months very rarely, and there were no differences either in the proportion of positive samples or concentration between the PncOMPC and control groups except for serotype 14 at 13 months; there were more anti-Pnc PS IgG positive samples in the PncOMPC than in the control group (Table 1).
Table 1 Salivary anti-Pnc PS IgG antibodies in the PncOMPC, PncOMPC+PncPS booster and control groups; number and the proportion of positive samples at the age of 7 and 13 months.
Pnc serotype Age IgG
PncOMPC group Control group
N, samples Number positive (%) N, samples Number positive (%)
6B 7 mo 56 1 (2) 55 0
13 mo (PncOMPC booster) 44 3 (7) 55 0
13 mo (PncPS booster) 6 1 (17) NA
14 7 mo 56 2 (4) 55 2 (4)
13 mo (PncOMPC booster) 44 6 (14)a 55 1 (2)
13 mo (PncPS booster) 6 3 (50) NA
19F 7 mo 56 2 (4) 54 5 (9)
13 mo (PncOMPC booster) 44 6 (14) 55 2 (4)
13 mo (PncPS booster) 5 3 (60) NA
23F 7 mo 56 1 (2) 54 1 (2)
13 mo (PncOMPC booster) 44 0 55 1 (2)
13 mo (PncPS booster) 6 0 NA
a statistical difference in the number of positive samples between PncOMPC and control groups, p = 0.04.
A small group of 7 children received PncPS vaccine as a booster instead of PncOMPC at the age of 12 months. Further, the volume of saliva was not sufficient for all the EIA analyses. However, there was a clear trend that the PncPS boosting induced more often IgG to the saliva than the PncOMPC boosting (Table 1).
Anti-pneumococcal IgA in saliva
At the age of 7 months, anti-Pnc PS IgA was detected more often and the concentrations were higher in the PncOMPC than in the control group for all four serotypes (Table 2, Fig 1). However, the difference was statistically significant only for serotypes 19F (both the number of positive samples and concentration) and 23F (concentration). The geometric mean concentrations (GMCs) varied between 2.3 and 5.7 ng/ml, and 1.9 and 3.1 ng/ml in the PncOMPC and control groups, respectively (Fig 1).
Table 2 Salivary anti-Pnc PS IgA antibodies in the PncOMPC, PncOMPC+PncPS booster and control groups; number and the percentage of positive samples at the age of 7 and 13 months.
Pnc serotype Age IgA
PncOMPC group Control group
N, samples Number positive (%) N, samples Number positive (%)
6B 7 mo 56 17 (30) 55 9 (16)
13 mo (PncOMPC booster) 45 23 (51) 55 26 (47)
13 mo (PncPS booster) 7 5 (71) NA
14 7 mo 56 23 (41) 55 14 (25)
13 mo (PncOMPC booster) 45 19 (42) 55 16 (29)
13 mo (PncPS booster) 7 5 (71) NA
19F 7 mo 56 32 (57)a 55 17 (31)
13 mo (PncOMPC booster) 45 35 (78) 55 35 (64)
13 mo (PncPS booster) 7 7 (100) NA
23F 7 mo 56 11 (20) 55 3 (5)
13 mo (PncOMPC booster) 45 18 (40) 55 17 (31)
13 mo (PncPS booster) 7 5 (71) NA
a statistical difference in the number of positive samples between PncOMPC and control groups, p = 0.01.
Figure 1 Salivary anti-Pnc PS IgA concentrations (ng/ml) with 95% confidence intervals in the PncOMPC, PncOMPC+PncPS booster and control groups at the age of 7 and 13 months.
At the age of 13 months, there were no statistical differences between the PncOMPC and the control group either in the proportion of positive samples or in the anti-Pnc PS IgA concentrations (Table 2, Fig 1). After the booster, the GMCs ranged between 3.7 and 11.2 ng/ml, and 3.0 and 8.0 ng/ml in the PncOMPC and control group, respectively. However, there was a statistical difference between 7 and 13 months of age in the anti-Pnc PS IgA concentrations for serotype 23F in the PncCRM group and for serotypes 6B, 19F and 23F in the HBV group.
PncPS booster induced higher anti-Pnc PS IgA concentrations than the PncOMPC (Table 2, Fig 1). However, the small number of samples in the PncPS group did not allow any statistical analyses.
IgA1 and IgA2 subclasses
At the age of 7 months, IgA1 concentrations were mirroring the total anti-Pnc serotype specific IgA (Fig 1) and were significantly higher in the PncOMPC than in the control group for serotypes 6B, 19F, and 23F. However, at the age of 13 months, there were no statistical differences between the groups. The IgA2 concentrations did not differ between the groups either at the age of 7 or 13 months. Thus, rises in the antibody concentrations were seen only for IgA1 (Fig 2).
Figure 2 Salivary anti-Pnc PS IgA1 and IgA2 concentrations (EIA units/ml) with 95% confidence intervals in the PncOMPC, PncOMPC+PncPS booster and control groups at the age of 7 and 13 months.
Discussion
Salivary anti-Pnc polysaccharide antibodies, both IgG and IgA, were detected only slightly more often in the PncOMPC than in the control group. However, the proportion of positive samples and the antibody concentrations rose between 7 and 13 months of age in both groups and the differences between the vaccine groups diminished. This suggests that children in both groups had pneumococcal contacts which induced development of mucosal antibodies. This is in accordance with the findings that pneumococcal carriage induces salivary antibodies in children [7], also regardless of the previous pneumococcal vaccination status [29].
In addition to IgA, we measured also IgA subclass specific antibody development. IgA1 is the predominant subclass both in serum and saliva, IgA1:IgA2 ratios being roughly 9:1 in serum and 6:4 in saliva [30]. To improve their virulence, some bacteria including pneumococcus have developed proteases, which degrade IgA1 antibodies [31]. Recently, IgA1-proteases have been found to be important in the adherence of pneumococcus [32]. Due to methodological reasons the IgA1 and IgA2 concentrations detected in this study can not be compared as the data are given in EIA units that are not comparable. However, we found that the changes and the difference in the IgA concentrations were mainly due to changes in the IgA1 concentrations. This is in accordance with our previous findings [11,33]. This suggests that both IgA1 and IgA2 are present, but the vaccine induces IgA1, which, unfortunately, is susceptible to IgA proteases.
We have also reported the mucosal anti-Pnc PS antibodies in the PncCRM arm of the FinOM Vaccine Trial [10]. Though the PncCRM induced more often and higher concentrations of salivary anti-Pnc PS antibodies than the PncOMPC vaccine (all the samples were analysed together at the same time), the basic message from the two studies is the same. The data of this study and of the PncCRM arm [10] suggest that parenteral immunization with pneumococcal conjugate vaccines induce local IgA1 antibody formation, but the antibodies persist only for a short period. The concentrations of anti-Pnc PS IgA at 13 months and in the PncCRM arm also at 4 to 5 years of age were very similar in the pneumococcal vaccine and in the control vaccine groups indicating that no remarkable natural boosting of the vaccine-induced antibodies occurred [10]. IgG anti-Pnc PS antibodies were detected only rarely. The highest prevalence was found in the small group of children whose 4th pneumococcal vaccine dose at 12 months was PncPS vaccine. This is in accordance with the serum antibody concentrations that were significantly higher in the PncPS-boosted than in the PncOMPC-boosted children for all other serotypes except 23F [34].
Pneumococcal PS vaccine has been suggested to be used as a booster after a primary series with a pneumococcal conjugate vaccine. It is cheaper than the conjugate vaccine available and induces higher serum antibody concentrations than a conjugate booster [14,22,34]. However, a conjugate vaccine as a booster can be favourable for the persistence and quality of antibodies, because it stimulates the formation of high-affinity clones of memory B-cells [35]. There is also a possibility that PS booster may cause depletion of memory cells [36]. Further, no significant differences in the efficacy against AOM and recurrent AOM have been found between PS and conjugate booster [22,37]. We found in this study, that there is a clear trend that the PS vaccine as a booster induces mucosal anti-pneumococcal antibodies more often and in higher concentration than conjugate booster. The same effect of a polysaccharide booster in salivary antibody concentrations have been seen also previously [9,14]. However, there may be differences in the quality of salivary antibodies induced by the PS or conjugate vaccines.
Pneumococcal conjugate vaccines have been found to induce both humoral and mucosal immunity. Further, the vaccines have been found effective against invasive and local pneumococcal infections. However, the antibody concentrations or quality of antibodies needed for protection against disease and carriage are not known, even if there has been a lot of discussion about the protective levels [38,39]. Recently, a consensus about the estimate of serological surrogate of protection against invasive pneumococcal infections has been achieved. It is suggested that the 0.35 μg/ml concentration of anti-Pnc PS IgG in serum would predict the protection against pneumococcal infection at the population level . However, the value is a rough estimate and further, the function of antibodies, both opsonophagocytic activity and avidity, and immunological memory are also essential issues in serological immunity. In addition to the concentration and function of anti-pneumococcal antibodies in serum, the importance and function of mucosal anti-pneumococcal antibodies should be clarified. In the FinOM Vaccine Trial the PncCRM induced more salivary antibodies than the PncOMPC, but the efficacy of these vaccines against AOM was similar. This raises a question, which is the actual mechanism of protection? Anyway, we know that pneumococcal carriage, infections and conjugate vaccines all induce mucosal antibodies. But we do not know how important mucosal antibodies really are in the defence against pneumococcal infections compared with the antibodies in serum, and if mucosal immunisation would provide better local immune response.
Conclusion
We found that PncOMPC can induce a salivary IgA response. However, the actual impact of mucosal antibodies in protection against pneumococcal infections is not clear.
Competing interests
A Palmu has had travel paid for by Wyeth-Lederle and GlaxoSmithKline as an invited speaker at symposia and received an honorarium from Wyeth-Lederle. H Käyhty has provided consultancies on advisory boards for Aventis Pasteur, GlaxoSmithKline and ID Biomedical Corporation; has had travel paid for by Aventis Pasteur, GlaxoSmithKline, Spectrum Medical Sciences and Wyeth Lederle Vaccines as an invited speaker or expert at symposia; and has received honoraria from Aventis Pasteur, GlaxoSmithKline, and Wyeth Lederle Vaccines. The other authors declare no competing interests.
Authors' contributions
AN conducted the laboratory analyses and analysed the immunogenicity data. ML was responsible for the statistical analyses. AP was the study coordinator responsible for enrolment, clinical evaluation and saliva sampling of the study subjects and participated in the planning of study design together with the FinOM Study Group. HK supervised the immunogenicity analyses and participated in the planning of study design together with the FinOM Study Group. All authors contributed to the writing of the manuscript and approved the final version.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We thank all the children and families who participated in the study. Special thanks to Helena Jokinen, RN, for collecting the saliva samples. The FinOM studies were supported by Aventis Pasteur, Merck & Co. Inc., and Wyeth.
==== Refs
Gray BM Turner ME Dillon HC Jr Epidemiologic studies of Streptococcus pneumoniae in infants. The effects of season and age on pneumococcal acquisition and carriage in the first 24 months of life Am J Epidemiol 1982 116 692 703 7137156
Stenfors LE Räisänen S Occurrence of middle ear pathogens in the nasopharynx of young individuals. A quantitative study in four age groups Acta Otolaryngol 1990 109 142 148 2106762
Obaro S Adegbola R The pneumococcus: carriage, disease and conjugate vaccines J Med Microbiol 2002 51 98 104 11863272
Musher DM Chapman AJ Goree A Jonsson S Briles D Baughn RE Natural and vaccine-related immunity to Streptococcus pneumoniae J Infect Dis 1986 154 245 256 3722865
Nieminen T Käyhty H Virolainen A Eskola J Circulating antibody secreting cell response to parenteral pneumococcal vaccines as an indicator of a salivary IgA antibody response Vaccine 1998 16 313 319 9607048 10.1016/S0264-410X(97)00162-X
Berneman A Belec L Fischetti VA Bouvet JP The specificity patterns of human immunoglobulin G antibodies in serum differ from those in autologous secretions Infect Immun 1998 66 4163 4168 9712763
Simell B Kilpi TM Käyhty H Pneumococcal carriage and otitis media induce salivary antibodies to pneumococcal capsular polysaccharides in children J Infect Dis 2002 186 1106 1114 12355361 10.1086/344235
Simell B Korkeila M Pursiainen H Kilpi TM Käyhty H Pneumococcal carriage and otitis media induce salivary antibodies to pneumococcal surface adhesin a, pneumolysin, and pneumococcal surface protein a in children J Infect Dis 2001 183 887 896 11237805 10.1086/319246
Choo S Zhang Q Seymour L Akhtar S Finn A Primary and booster salivary antibody responses to a 7-valent pneumococcal conjugate vaccine in infants J Infect Dis 2000 182 1260 1263 10979930 10.1086/315834
Nurkka A Lahdenkari M Palmu A Käyhty H the FinOM Study Group Salivary antibodies induced by the seven-valent PncCRM conjugate vaccine in Finnish Otitis Media Vaccine Trial Vaccine 2004 23 298 304 15530671 10.1016/j.vaccine.2004.06.009
Korkeila M Lehtonen H Åhman H Leroy O Eskola J Käyhty H Salivary anti-capsular antibodies in infants and children immunised with Streptococcus pneumoniae capsular polysaccharides conjugated to diphtheria or tetanus toxoid Vaccine 2000 18 1218 1226 10649623 10.1016/S0264-410X(99)00393-X
Nieminen T Eskola J Käyhty H Pneumococcal conjugate vaccination in adults: circulating antibody secreting cell response and humoral antibody responses in saliva and in serum Vaccine 1998 16 630 636 9569475 10.1016/S0264-410X(97)00235-1
Nieminen T Käyhty H Leroy O Eskola J Pneumococcal conjugate vaccination in toddlers: mucosal antibody response measured as circulating antibody-secreting cells and as salivary antibodies Pediat Infect Dis J 1999 18 764 772 10493335 10.1097/00006454-199909000-00005
Nurkka A Åhman H Korkeila M Jäntti V Käyhty H Eskola J Serum and salivary anti-capsular antibodies in infants and children immunized with the heptavalent pneumococcal conjugate vaccine Pediatr Infect Dis J 2001 20 25 33 11176563
Nurkka A Åhman H Yaich M Eskola J Käyhty H Serum and salivary anti-capsular antibodies in infants and children vaccinated with octavalent pneumococcal conjugate vaccines, PncD and PncT Vaccine 2001 20 194 201 11567764 10.1016/S0264-410X(01)00250-X
Black S Shinefield H Fireman B Lewis E Ray P Hansen JR Efficacy, safety and immunogenicity of heptavalent pneumococcal conjugate vaccine in children Pediatr Infect Dis J 2000 19 187 195 10749457 10.1097/00006454-200003000-00003
Black SB Shinefield HR Ling S Hansen J Fireman B Spring D Effectiveness of heptavalent pneumococcal conjugate vaccine in children younger than five years of age for prevention of pneumonia Pediatr Infect Dis J 2002 21 810 815 12352800 10.1097/00006454-200209000-00005
Dagan R Givon-Lavi N Zamir O Sikuler-Cohen M Guy L Janco J Reduction of nasopharyngeal carriage of Streptococcus pneumoniae after administration of a 9-valent pneumococcal conjugate vaccine to toddlers attending day care centers J Infect Dis 2002 185 927 936 11920317 10.1086/339525
Eskola J Kilpi T Palmu A Jokinen J Haapakoski J Herva E Efficacy of a pneumococcal conjugate vaccine against acute otitis media N Engl J Med 2001 344 403 409 11172176 10.1056/NEJM200102083440602
Fireman B Black SB Shinefield HR Lee J Lewis E Ray P Impact of the pneumococcal conjugate vaccine on otitis media Pediatr Infect Dis J 2003 22 10 16 12544402 10.1097/00006454-200301000-00006
Mbelle N Huebner RE Wasas AD Kimura A Chang I Klugman KP Immunogenicity and impact on nasopharyngeal carriage of a nonavalent pneumococcal conjugate vaccine J Infect Dis 1999 180 1171 1176 10479145 10.1086/315009
Kilpi T Åhman H Jokinen J Lankinen KS Palmu A Savolainen H Protective Efficacy of a Second Pneumococcal Conjugate Vaccine against Pneumococcal Acute Otitis Media in Infants and Children: Randomized, Controlled Trial of a 7-Valent Pneumococcal Polysaccharide-Meningococcal Outer Membrane Protein Complex Conjugate Vaccine in 1666 Children Clin Infect Dis 2003 37 1155 1164 14557958 10.1086/378744
Malley R Stack AM Ferretti ML Thompson CM Saladino RA Anticapsular polysaccharide antibodies and nasopharyngeal colonization with Streptococcus pneumoniae in infant rats J Infect Dis 1998 178 878 882 9728564 10.1086/314523
O'Brien KL Moulton LH Reid R Weatherholtz R Oski J Brown L Efficacy and safety of seven-valent conjugate pneumococcal vaccine in American Indian children: group randomised trial Lancet 2003 362 355 361 12907008 10.1016/S0140-6736(03)14022-6
Klugman K Shabir M Huebner R Kohberger R Mbelle N Pierce N A trial of a 9-valent pneumococcal conjugate vaccine in children with and those without HIV infection N Engl J Med 2003 349 1341 8 14523142 10.1056/NEJMoa035060
Jokinen J Åhman H Kilpi T Mäkelä PH Käyhty H The concentration of anti-pneumococcal antibodies as a serological correlate or protection: An application to acute otitis media J Infect Dis 2004 190 545 50 15243930 10.1086/422531
Kilpi T Herva E Kaijalainen T Syrjänen R Takala AK Bacteriology of acute otitis media in a cohort of Finnish children followed for the first two years of life Pediatr Infect Dis J 2001 20 654 662 11465836 10.1097/00006454-200107000-00004
Quataert SA Kirch CS Wiedl LJ Phipps DC Strohmeyer S Cimino CO Assignment of weight-based antibody units to a human antipneumococcal standard reference serum, lot 89-S Clin Diagn Lab Immunol 1995 2 590 597 8548539
Zhang Q Arnaoutakis K Murdoch C Lakshman R Race G Burkinshaw R Finn A Mucosal immune responses to capsular pneumococcal polysaccharides in immunized preschool children and controls with similar nasal pneumococcal colonization rates Pediatr Infect Dis J 2004 23 307 313 15071283
Delacroix DL Dive C Rambaud JC Vaerman JP IgA subclasses in various secretions and in serum Immunology 1982 47 383 385 7118169
Kilian M Reinholdt J Lomholt H Poulsen K Frandsen EV Biological significance of IgA1 proteases in bacterial colonization and pathogenesis: critical evaluation of experimental evidence Apmis 1996 104 321 338 8703438
Weiser JN Bae D Fasching C Scamurra RW Ratner AJ Janoff EN Antibody-enhanced pneumococcal adherence requires IgA1 protease Proc Natl Acad Sci U S A 2003 100 4215 4220 12642661 10.1073/pnas.0637469100
Kauppi-Korkeila M Saarinen L Eskola J Käyhty H Subclass distribution of IgA antibodies in saliva and serum after immunization with Haemophilus influenzae type b conjugate vaccines Clin Exp Immunol 1998 111 237 242 9486387 10.1046/j.1365-2249.1998.00501.x
Blum MD Dagan R Mendelman PM Pinsk V Giordani M Li S A comparison of multiple regimens of pneumococcal polysaccharide-meningococcal outer membrane protein complex conjugate vaccine and pneumococcal polysaccharide vaccine in toddlers Vaccine 2000 18 2359 2367 10738092 10.1016/S0264-410X(00)00021-9
Anttila M Eskola J Åhman H Käyhty H Differences in the avidity of antibodies evoked by four different pneumococcal conjugate vaccines in early childhood Vaccine 1999 17 1970 1977 10217596 10.1016/S0264-410X(98)00458-7
MacLennan J Obaro S Deeks J Lake D Elie C Carlone G Immunologic memory 5 years after meningococcal A/C conjugate vaccination in infancy J Infect Dis 2001 183 97 104 11087205 10.1086/317667
Veenhoven RH Bogaert D Schilder AG Rijkers GT Uiterwaal CS Kiezebrink HH Nasopharyngeal pneumococcal carriage after combined pneumococcal conjugate and polysaccharide vaccination in children with a history of recurrent acute otitis media Clin Infect Dis 2004 39 911 9 15472839 10.1086/422651
Lee LH Frasch CE Falk LA Klein DL Deal CD Correlates of immunity for pneumococcal conjugate vaccines Vaccine 2003 21 2190 2206 12706710 10.1016/S0264-410X(03)00025-2
Jodar L Butler J Carlone G Dagan R Goldblatt D Käyhty H Serological criteria for evaluation and licensure of new pneumococcal conjugate vaccine formulations for use in infants Vaccine 2003 21 3265 3272 12804857 10.1016/S0264-410X(03)00230-5
|
15921511
|
PMC1185537
|
CC BY
|
2021-01-04 16:28:13
|
no
|
BMC Infect Dis. 2005 May 27; 5:41
|
utf-8
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-41
|
oa_comm
|
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-511597813510.1186/1471-2334-5-51Research ArticleImmunomodulatory intervention in sepsis by multidrug-resistant Pseudomonas aeruginosa with thalidomide: an experimental study Giamarellos-Bourboulis Evangelos J [email protected] Nikolaos [email protected] George [email protected] Vassilios [email protected] Vassilios [email protected] Despina [email protected] Panayotis E [email protected] Helen [email protected] 4th Department of Internal Medicine, University of Athens, Medical School, Athens, Greece2 Laboratory of Experimental Surgery and Surgical Research, University of Athens, Medical School, Athens, Greece3 Center for Biomedical Research, Academy of Athens, Athens, Greece2005 26 6 2005 5 51 51 4 11 2004 26 6 2005 Copyright © 2005 Giamarellos-Bourboulis et al; licensee BioMed Central Ltd.2005Giamarellos-Bourboulis 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
Thalidomide is an inhibitor of tumour necrosis factor-alpha (TNFα) that has been proven effective for the treatment of experimental sepsis by Escherichia coli. It was tested whether it might behave as an effective immunomodulator in experimental sepsis by multidrug-resistant (MDR) Pseudomonas aeruginosa.
Methods
Sepsis was induced by the intraperitoneal injection of 1 × 108 cfu/kg inoculum of the test isolate in a total of 109 Wistar rats divided in three groups as follows: group A controls; group B administered seed oil 30 minutes before bacterial challenge; and group C administered 50 mg/kg of thalidomide diluted in seed oil 30 minutes before bacterial challenge. Blood was sampled for estimation of endotoxins (LPS), TNFα, interferon-gamma (IFNγ), nitric oxide (NO) and malondialdehyde (MDA). LPS was measured by the QCL-1000 LAL assay, TNFα and IFNγ by ELISA, NO by a colorimetric assay and MDA by the thiobarbiturate assay.
Results
Mean (± SE) survival of groups A, B and C were 18.60 ± 1.84, 12.60 ± 0.60 and 30.50 ± 6.62 hours (p of comparisons A to C equal to 0.043 and B to C equal to 0.002). Decreased TNFα and NO levels were found in sera of animals of group C compared to group A. Plasma levels of LPS, MDA and IFNγ did not differ between groups.
Conclusion
Intake of thalidomide considerably prolonged survival in experimental sepsis by MDR P.aeruginosa an effect probably attributed to decrease of serum TNFα.
==== Body
Background
Nosocomial infections are commonly caused by multidrug-resistant Gram-negative pathogens. Management of these infections is difficult due to the lack of potent antimicrobial agents; thus a target for immunomodulatory intervention is created [1]. Thalidomide is an old regimen that has been proved potent in reducing the half-life of mRNA of the gene of tumour necrosis factor-alpha (TNFα) in human monocytes [2]. Its anti-angiogenic and anti-TNFα properties have led to its application for the treatment of erythema nodosum leprosum, of cutaneous lupus erythematosus, of Behçet's syndrome, of multiple myeloma and of HIV-related aphthous ulcers and wasting syndrome [3,4].
In a model of experimental sepsis by Escherichia coli, thalidomide was proved very effective in reducing serum levels of TNFα, a phenomenon that was associated with refraining of evolution to sepsis [5]. However, its effect on survival was not assessed. The immunomodulatory benefit of thalidomide would be of considerable importance for sepsis induced by multidrug-resistant isolates. The present study was designed to evaluate thalidomide in experimental sepsis by multidrug-resistant Pseudomonas aeruginosa. Interest was focused on the effect of thalidomide on a) survival after bacterial challenge, and b) serum levels of pro-inflammatory mediators.
Methods
Animals
A total of 109 male Wistar rats were enrolled in the study. Their mean (± SD) weight were 257.2 ± 40.2 g. The study received permit from the Veterinary Directorate of the Perfecture of Athens according to the Greek legislation in conformance to the Council Directive of the European Community. Rats were housed in metal cages and had access to tap water and standard balanced chow ad libitum. Temperature ranged between 18 and 22°C, relative humidity between 55 and 65% and the light/dark cycle was 6 am/6 pm.
Bacterial isolate
One multidrug-resistant blood isolate of P. aeruginosa derived from a patient with nosocomial sepsis was applied. Minimal inhibitory concentrations (MICs) of ticarcillin/clavulanate, piperacillin, ceftazidime, imipenem, meropenem, ciprofloxacin and amikacin were determined by a microdilution technique of a 0.1 ml final volume. MIC was considered as the lowest concentrations of the tested antimicrobial limiting visible bacterial growth after 18 hours of incubation at 35°C. The isolate was stored as multiple aliquots in skim milk (Oxoid Ltd, London, UK) under -70°C. One aliquot was removed from the fridge before each experiment. Single colonies were incubated at 37°C in 10 ml of Mueller-Hinton broth (Oxoid Ltd) for eight hours to yield a log-phase inoculum that was applied for bacterial challenge.
Study design
Animals were divided into three groups of treatment, as follows:
• Group A (n = 40), controls; in 20 of these animals survival was recorded after bacterial challenge and 20 were sacrificed five hours after bacterial challenge.
• Group B (n = 20), animals pre-treated with linseed oil 30 minutes before bacterial challenge; survival was then recorded.
• Group C (n = 49), animals pre-treated with thalidomide 30 minutes before bacterial challenge; survival was recorded for 24 and another 25 were sacrificed five hours after bacterial challenge.
Thalidomide was supplied as a white amorphous powder (ICN Biomedicals GmbH, Thüringer, Germany) insoluble in water. It was diluted in commercial linseed oil and it was administered via a gastric tube in animals of group C at a dose of 50 mg/kg, based on previous results [5]. The respective dose of linseed oil administered was 2 ml/kg. The same dose was administered in animals of group B. A 1 × 108 cfu/kg inoculum of the test isolate was injected intraperitoneally in all animals.
Survival was recorded at 12-hour time intervals. Five hours after bacterial challenge, a midline abdominal incision was performed. Intestines were displaced to the left and the inferior vena cava was recognized and punctured with a 19-gauge needle. Ten ml of blood was collected into pyrogen-free syringes and applied for culture and for the estimation of endotoxins (LPS), of tumour necrosis factor-alpha (TNFα), of interferon-gamma (IFNγ), of nitric oxide (NO) and of malondialdehyde (MDA). Animals were then sacrificed by an intramuscular injection of pentothal. The 5-hour time interval was selected as the most appropriate for blood sampling after estimation of pro-inflammatory mediators in groups of animals inoculated by the test pathogen with hourly time differences between each sampling (data not shown).
Blood culture and estimation of LPS, cytokines, NO and MDA
One ml of blood was added into flasks with growth medium (Becton Dickinson, Cockeysville, Md) and incubated at 35°C for a total of seven days. Identification of Gram-negative isolates from blood cultures was made by the API20NE test (bioMérieux, Paris, France).
Sampled blood was added into pyrogen-free tubes (Vacutainer, Becton Dickinson), and centrifuged; serum was kept refrigerated as multiple aliquots at -70°C until assayed. For the determination of LPS, serum was diluted 1:10 with pyrogen-free water (BioWhitaler, Maryland, USA) and incubated for five minutes at 70°C. LPS were then measured by the QCL-1000 LAL assay (BioWhitaker, Maryland, USA, lower detection limit 1 EU/ml). TNFα and IFNγ were estimated in serum by an enzymoimmunoassay (Diaclone, Paris, France). Lower detection limits for TNFα and IFNγ were 10 and 10 pg/ml respectively. All determinations were performed in duplicate.
Nitric oxide (NO) was estimated in serum samples by a colorimetric assay based on the production of velvet color after sequential addition of NADH and nitrate reductase (Assay Designs Inc., Ann Arbor Minnesota, USA). Optical density was read at 570 nm (Hitachi Spectrophotometer).
Lipid peroxidation in serum was assessed by the estimation of the concentration of malondialdehyde according to the thiobarbiturate assay, as already described [6,7]. Briefly, a 0.1 ml aliquot of each sample was mixed to 0.9 ml of trichloroacetic acid 20% (Merck, Darmstadt, Germany) and centrifuged at 12,000 g and 4°C for 10 minutes. The supernatant was incubated with 1 ml of PBS (pH: 7.2) and 1 ml of thiobarbituric acid 0.6% (Merck) for 20 minutes at 90°C. Optical density was then read at 535 nm (Hitachi Spectophotometer). MDA was determined in mM by a standard curve created with 1, 1, 3, 3-tetramethoxy-propane (Merck). A water sample treated in the same way was applied as a blank. All determinations were performed in duplicate.
Animal sacrifice was not performed for the estimation of pro-inflammatory mediators in animals of group B. That was based on the similar survival rate (see below) of groups A and B.
Statistical analysis
Survival of each group was estimated by Kaplan-Meier analysis; comparisons between different groups were performed by the log-rank test.
Concentrations of LPS and of serum pro-inflammatory parameters were expressed by their mean (± SE). Comparisons between groups A and C were performed by the Mann-Whitney U test. Any value of P equal to or below 0.05 was considered as significant.
Results
MICs of ticarcillin/clavulanate, piperacillin, ceftazidime, imipenem, meropenem, ciprofloxacin and amikacin for the test isolate were >256/4, >256, 16, 16, 16, >256 and >512 μg/ml respectively.
Mean (± SE) survival of group A was 18.60 ± 1.84 hours, of group B 12.60 ± 0.60 hours (P: 0.036 compared to group A) and of group C 30.50 ± 6.62 hours (P equal to 0.043 when compared to group A and equal to 0.002 when compared to group B). Survival curves of each group of treatment are given in Figure 1. Survival of one animal of group C was prolonged to 168 hours. After subtraction of that animal, the mean (± SE) survival of group C was 24.52 ± 2.97 hours (P equal to 0.049 when compared to group A and equal to 0.004 when compared to group B).
Figure 1 Comparative survival of Wistar rats after bacterial challenge with multidrug-resistant Pseudomonas aeruginosa. Group A: controls; group B: rats pre-treated with seed oil; and group C: rats pre-treated with thalidomide.
No diarrhea was noted over follow-up of groups B and C.
All animals had positive blood cultures. Comparisons of serum pro-inflammatory mediators of groups A and C estimated five hours after bacterial challenge are given in Table 1. Mean LPS levels of groups A and C were 13.12 and 13.52 EU/ml respectively. Respective values of TNFα were 153.9 and 74.7 pg/ml, of IFNγ 782.4 and 564.4 pg/ml, of NO 1,694.4 and 618.9 μM, and of MDA 4.30 and 4.564 mM respectively.
Table 1 Comparison of serum concentrations of endotoxins (LPS), nitric oxide (NO), malondialdehyde (MDA), tumour necrosis factor-alpha (TNFα) and interferon-gamma (IFNγ) of animals-controls (group A) and of animals pre-treated with thalidomide (group C) sacrificed five hours after bacterial challenge.
Group A Group C p
Mean ± SE
LPS (EU/ml) 13.12 ± 0.17 13.52 ± 0.44 NS
NO (μM) 1,694.4 ± 400.9 618.9 ± 350.3 0.009
MDA (mM) 4.30 ± 1.51 4.56 ± 0.54 NS
TNFα (pg/ml) 153.9 ± 53.8 74.7 ± 22.5 0.012
IFNγ (pg/ml) 782.4 ± 103.2 564.4 ± 116.5 NS
Discussion
The perspective of an immunomodulatory intervention in sepsis, as evolved from human studies with the application of monoclonal antibodies, was very promising. However the application of these antibodies in clinical practice failed to disclose particular benefit [8,9]. The need of such immunotherapies seems to be increasing for the field of nosocomial infections where the availability of antimicrobial agents is limited. Thalidomide has been proved effective in reducing serum TNFα in experimental sepsis induced by endotoxins [10,11] and E.coli [5] by a mechanism differing from monoclonal antibodies. The present study documented the effect of thalidomide in an experimental model simulating a nosocomial infection ie sepsis by multidrug-resistant P.aeruginosa. This is the first time, to our knowledge, that an agent with anti-TNFα properties is applied in an experimental model of sepsis triggered by a multidrug-resistant pathogen.
The applied model of sepsis was lethal as documented by the absolute mortality rate of animal-controls. Results revealed a considerable benefit of thalidomide intake on survival after bacterial challenge (Figure 1). Animals administered only the vehicle of thalidomide died earlier than controls a phenomenon aggravating the potency of thalidomide. Bacteremia and endotoxemia were equally present in all groups of treatment (Table 1) so that all animals had the same chance of evolution to sepsis. Thalidomide prolonged survival even in the presence of considerable endotoxaemia.
The action of thalidomide seems to be mediated by degradation of the mRNA of TNFα and probably of IFNγ [12]. In the present study, intake of thalidomide was accompanied by significant decrease of serum TNFα and of NO whereas serum levels of MDA and of IFNγ remained unaffected (Table 1). Its specific anti-TNFα effect has also been considered as the mode of action in experimental sepsis by E.coli [5]. Biosynthesis of NO is triggered by TNFα [13], so that decrease of NO following that of TNFα might be expected. Thalidomide intake did not influence serum levels of MDA and of IFNγ. MDA is the end product of the peroxidation of cell membrane lipids occurring during the septic cascade; lipid peroxidation is triggered by a variety of pro-inflammatory mediators [14,15]. The specific blockade of TNFα and not of any other cytokine by thalidomide might be consistent with its failure to affect lipid peroxidation. The latter observation is also consistent with the observation that thalidomide did prolong survival but that all animals eventually died (Figure 1). However, it should be mentioned that the activity of thalidomide might be mediated by other pathways different than inhibition of TNFα.
Conclusion
The present study revealed that intake of thalidomide considerably prolonged survival in experimental sepsis by multidrug-resistant P.aeruginosa an effect probably attributed to decrease of serum TNFα. These findings probably merit further research in order to elucidate their clinical relevance.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
EJGB participated in the design of the study, in the performance of estimation of inflammatory parameters and drafted the manuscript.
NB participated in the administration of the drugs and in the conduct of the animals' experiments
GL participated in the administration of the drugs and in the conduct of the animals' experiments
VP participated in the conduct of the animals' experiments
VK participated in the estimation of malondialdehyde
DP participated in study design
PEK participated in study design, in the conduct of the animals' experiments and drafted the manuscript
HG participated in study design and drafted the manuscript
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Rangel Frausto MS The epidemiology of bacterial sepsis Infect Dis Clin Nor Amer 1999 13 299 311 10.1016/S0891-5520(05)70076-3
Moreira A Sampaio E Zmuidzinas A Frindt P Smith K Kaplan G Thalidomide exerts its inhibitory action on TNFa by enhancing mRNA degradation J Exp Med 1993 177 1675 1680 8496685 10.1084/jem.177.6.1675
Calabrese L Fleitscher AB Jr Thalidomide: current and potential clinical applications Am J Med 2000 108 487 495 10781782 10.1016/S0002-9343(99)00408-8
Tsenova L Sokol K Freedman VJ Kaplan G A combination of thalidomide plus antibiotics protects rabbits from mycobacterial meningitis-associated death J Infect Dis 1998 177 1563 1572 9607834
Giamarellos-Bourboulis EJ Poulaki H Kostomitsopoulos N Dontas I Perrea D Karayannacos PE Giamarellou H Effective immunomodulatory treatment of Escherichia coli experimental sepsis with thalidomide Antimicrob Agents Chemother 2003 47 2445 2449 12878503 10.1128/AAC.47.8.2445-2449.2003
Giamarellos-Bourboulis EJ Plachouras D Tzivra A Koussoulas V Bolanos N Raftogiannis M Galani I Dontas I Dionyssiou-Asteriou A Giamarellou H Stimulation of innate immunity by susceptible and multidrug-resistant Pseudomonas aeruginosa: an in vitro and in vivo study Clin Exp Immunol 2004 135 240 246 14738451 10.1111/j.1365-2249.2003.02365.x
Muzio G Salvo RA Trombetta A Autelli R Maggiora M Terreno M Dianzani MU Canuto RA Dose-dependent inhibition of cell proliferation induced by lipid peroxidation products in rat hepatoma cells after enrichment with arachidonic acid Lipids 1999 34 705 711 10478928
Arndt P Abraham E Immunological therapy for sepsis Intensive Care Med 2001 27 S104 S115 11307366 10.1007/s001340000574
Vincent JL Sun Q Dubois MJ Clinical trials of immunomodulatory therapies in severe sepsis and septic shock Clin Infect Dis 2003 34 1084 1093 11914997 10.1086/339549
Arrieta O Ortiz-Reyes A Rembao D Calvillo M Rivera E Sotelo J Protective effect of pentoxifylline plus thalidomide against septic shock in mice Int J Exp Pathol 1999 80 11 16 10365082 10.1046/j.1365-2613.1999.00085.x
Schmidt H Rush B Simonian G Murphy T Hsieh J Condon M Thalidomide inhibits TNF response and increases survival following endotoxin injection in rats J Surg Res 1996 63 143 146 8661187 10.1006/jsre.1996.0237
Matthews SJ McCoy C Thalidomide: a review of approved and investigational uses Clin Ther 2003 25 342 395 12749503 10.1016/S0149-2918(03)80085-1
Annane D Sanquer V Sebille V Faye A Djuranovic D Raphael JC Gajdos B Bellissant E Compartmentalised inducible nitric-oxide synthase activity in septic shock Lancet 2000 355 1143 1148 10791377 10.1016/S0140-6736(00)02063-8
Cowley H Bacon P Goode H Webster N Jones JG Menon D Plasma antioxidant potential in severe sepsis: A comparison of survivors and nonsurvivors Crit Care Med 1996 24 1179 1183 8674332 10.1097/00003246-199607000-00019
Ekmekcioglu C Schweiger B Strauss-Blasche G Mundigler G Siostrzonek P Marktl W Urinary excretion of 8-iso-PGF2a in three patients during sepsis, recovery and state of health Prostaglandins Leukot Essent Fatty Acids 2002 66 441 442 12054915 10.1054/plef.2002.0371
|
15978135
|
PMC1185538
|
CC BY
|
2021-01-04 16:28:13
|
no
|
BMC Infect Dis. 2005 Jun 26; 5:51
|
utf-8
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-51
|
oa_comm
|
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-611603364110.1186/1471-2334-5-61Research ArticleLack of association between serological evidence of past Coxiella burnetii infection and incident ischaemic heart disease: nested case-control study McCaughey Conall [email protected] Liam J [email protected] James P [email protected] Peter V [email protected]'Neill Hugh J [email protected] Dorothy E [email protected] Jayne V [email protected] John WG [email protected] Pierre [email protected] Annie [email protected] Philippe [email protected] Michele [email protected] Dominique [email protected] Bernadette [email protected] Jean [email protected] Jean-Bernard [email protected] Regional Virus Laboratory, Royal Hospitals, Grosvenor Road, Belfast BT12 6BA, UK2 Department of Epidemiology and Public Health, Mulhouse Building, Queen's University of Belfast, BT12 6BJ, UK3 Department of Medicine, Mulhouse Building, Queen's University of Belfast BT12 6BJ, UK4 INSERM U258, Epidemiologie cardio-vasculaire et metabolique, Hopital Paul Brousse, 94807 Villejuif Cedex, France5 INSERM U508, Institut Pasteur de Lille, Lille, France6 Laboratoire d'Epidémiologie et de Santé Publique, Strasbourg, France7 INSERM U558, Faculté de Médicine Purpan, Toulouse, France2005 20 7 2005 5 61 61 14 3 2005 20 7 2005 Copyright © 2005 McCaughey et al; licensee BioMed Central Ltd.2005McCaughey 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
Coxiella burnetii causes the common worldwide zoonotic infection, Q fever. It has been previously suggested that patients who had recovered from acute Q fever (whether symptomatic or otherwise) may be at increased risk of ischaemic heart disease. We undertook this study to determine if past infection with Coxiella burnetii, the aetiological agent of Q fever, is a risk factor for the subsequent development of ischaemic heart disease.
Methods
A nested case-control study within the Prospective Epidemiological Study of Myocardial Infarction (PRIME). The PRIME study is a cohort study of 10,593 middle-aged men undertaken in France and Northern Ireland in the 1990s. A total of 335 incident cases of ischaemic heart disease (IHD) were identified and each case was matched to 2 IHD free controls. Q fever seropositivity was determined using a commercial IgG ELISA method.
Results
Seroprevalence of Q fever in the controls from Northern Ireland and France were 7.8% and 9.0% respectively. No association was seen between seropositivity and age, smoking, lipid levels, or inflammatory markers. The unadjusted odds ratio (95% CI) for Q fever seropositivity in cases compared to controls was 0.95 (0.59, 1.57). The relationship was substantially unaltered following adjustment for cardiovascular risk factors and potential confounders.
Conclusion
Serological evidence of past infection with C. burnetii was not found to be associated with an increased risk of IHD.
==== Body
Background
Q fever is a globally distributed, common, zoonotic infection caused by the bacteria Coxiella burnetii. A large proportion of cases of C. burnetii infection are asymptomatic. Where symptomatic infection occurs, typical signs and associated symptoms are headache, pyrexia, and respiratory tract infection including atypical pneumonia. Hepatitis may also occur. Chronic infection is well recognised, usually in the form of Q fever endocarditis.
Various seroepidemiological and molecular biology approaches have suggested a potential role of various viral and bacterial infections in the development of atherosclerosis. In this context it has been previously suggested that patients who recover from acute Q fever (whether symptomatic or otherwise) may be at increased risk of ischaemic heart disease(IHD)[1,2]. The first of these studies was a retrospective case-control study, a study design that is subject to several important biases including difficulty in ascertaining the temporality of relationships, and the second has been criticised for failing to adjust for important confounders[3]. Until now no prospective studies have examined this issue. We present a prospective investigation, examining the relationship between C. burnetii seropositivity and incident cardiovascular disease in a large cohort study of middle aged men.
Methods
Study design
The study was a nested case-control study within the Prospective Epidemiological Study of Myocardial Infarction (PRIME) study, which is a cohort study of middle-aged men in France and Northern Ireland (Belfast). The original purpose of this study was to investigate the relative roles of various risk factors on the development of ischaemic heart disease. Recruitment and examination methods have been fully described previously [4,5] but are briefly summarised here. A total of 10,593 men aged between 50–59 years were recruited from industry, various employment groups and general practices in Lille, Strasbourg, Toulouse and Belfast between 1991 and 1993. The sample was recruited to broadly match the social class structure of the background population. Each subject completed self-administered questionnaires on demographic, socio-economic factors and dietary habits after informed consent was obtained. Their responses were checked by medical staff and additional data collected during clinic attendance on educational level, occupational activity, personal and family history, tobacco and alcohol consumption, and physical activity. The London School of Hygiene and Tropical Medicine Cardiovascular (Rose) Questionnaire for Chest Pain on Effort and Possible Infarction [6] was also administered.
Clinical examination
Baseline investigations included a standard 12-lead electrocardiogram and standardised blood pressure measurements (measured on 2 occasions in the sitting position) using an automatic sphygmomanometer (Spengler SP9). Anthropometric measurements included height and weight without shoes and waist and hip circumferences. Subjects were considered to have a history of IHD at entry if they had one of the following: myocardial infarction (MI) and/or angina pectoris diagnosed by a physician, electrocardiographic evidence of MI, or a positive answer to the Rose questionnaire. There were 9,758 subjects without a history of IHD at entry into the study.
Case-control selection and follow-up
Subjects were contacted annually by letter and asked to complete a clinical event questionnaire. Phone contact was established with non-responders or their general practitioner. Coronary cases were defined as the presence of at least one of the following: non-fatal MI, death from IHD, or angina pectoris. Five-year follow-up has been completed (98% follow-up was achieved). The total number of incident IHD cases identified was 335. Each case was matched to 2 controls, who were study participants of the same age (± 3 years) recruited in the same centre on the same day (± 2 days) as the corresponding case and were free of IHD on the date of the ischaemic event of the case.
Laboratory methods
Baseline venous blood samples were collected at the start of the study after a 12 hour fast and centrifuged within 4 hours. All assays were performed blind to the investigating centre and by consecutive number so that samples from all 4 centres were included in each analytical run. All procedures were standardised between centres. Among other measurements, total and HDL cholesterol, triglycerides, and fibrinogen were assayed at recruitment on all subjects. C-reactive protein (CRP) and interleukin 6 (IL-6) were assayed on stored samples from cases and controls. Enzymatic methods were employed to measure total cholesterol and triglycerides using a commercial kit in an automatic analyser (Boehringer Mannheim, Germany). HDL cholesterol was measured by an enzymatic method (Boehringer Mannheim, Germany) after precipitation using phosphotungstate magnesium chloride. LDL cholesterol was calculated according to Friedewald and colleagues [7] Fibrinogen was measured according to the method of Clauss[8] High sensitivity CRP was measured by immunonephelometry (hs-CRP: Dade Behring, Reuil-Malmaison, France). IL-6 was measured by ELISA (R&D, Europe).
Q fever ELISA testing
The presence of IgG antibodies to Phase II C. burnetii in the baseline serum specimen was performed using a commercially available indirect ELISA test kit (Vircell, Granada, Spain. Cat. No. G1001). The standard manufacturer's protocol was followed for testing, use of controls, and interpretation The antigen used in the Vircell IgG assay is C. burnetii phase II Nine Mile, grown in cell culture. The cut off for a positive test result was based on the manufacturers stipulated guidelines. Provided that Optical Densities (OD) for positive, negative and cut off controls fell within the manufacturers stated guidelines the test was considered valid. The mean OD for cut off serum was determined. An antibody index was then determined for each sample tested Antibody Index = (sample OD/cut off serum mean OD) × 10. The results were then interpreted in line with the manufacturer's guidance as follows: <9 Negative; 9–11 Equivocal; >11 Positive.
Statistical methods
Smoking status was categorised as never smoked, ex-smoker, or current smoker of <20 or = 20 cigarettes a day. Alcohol intake was calculated in grams per week and categorised into none, <225 g/week and = 225 g/week. The following skewed variables were log transformed before inclusion in analyses: triglycerides, CRP, and IL-6. Equivocal Q fever results were treated as seropositive (n = 3). In the controls, Chi-square tests and t-tests were used to examine the differences in prevalence of Q fever seropositivity between Northern Irish and French subjects, the relationship between seropositivity and age and smoking status, and between seropositivity and lipid levels and markers of inflammation (fibrinogen, CRP and IL-6). Conditional (matched) logistic regression analysis was used to examine the relationship between Q fever seropositivity and case-control status. The relationship was adjusted for age, smoking, body mass index, systolic and diastolic blood pressure, social class, physical activity, and alcohol intake at baseline. Further adjustment was undertaken for lipids (total cholesterol, HDL cholesterol and triglycerides) and then for markers of inflammation. Finally, a country/Q fever seropositivity interaction term was entered into the model. All analyses were performed using Stata version 8.
Ethical approval was obtained from the Queen's University of Belfast Ethical Committee
Results
Serum was available for Q fever testing from 320 (95.5%) of the 335 cases and 622 (92.8%) of 670 controls. Q fever results were available for the case and at least one control in 317 cases (94.6%). Among the controls, there was no difference in the seroprevalence of Q fever antibodies between Northern Ireland and France; 7.8% and 9.0% seropositive respectively. Within France, 12.4% of controls from Toulouse were seropositive, while seropositivity in the other 2 centres was 7.0% (χ2 = 2.99, df 1, p = 0.08). There was no difference in age between seropositives and seronegatives; mean age (SD) was 54.9 years (3.11) and 55.2 years (2.69) respectively. No association was seen between seropositivity and age, smoking, lipid levels, or inflammatory markers (Table 1).
Table 1 Relationships between seropositivity for Q fever and characteristics of control subjects
Mean (SD) p-value
Seronegative Seropositive
Age (years) 55.2 (2.69) 54.9 (3.11) 0.45
Total cholesterol 2.23 (0.42) 2.15 (0.31) 0.20
HDL 0.48 (0.12) 0.48 (0.15) 0.69
Triglycerides* 1.37 (1.72) 1.35 (1.68) 0.81
Fibrinogen 3.37 (0.98) 3.53 (1.10) 0.16
CRP* 1.39 (2.73) 1.32 (2.01) 0.76
IL-6* 1.53 (1.97) 1.39 (2.73) 0.14
Number (%) Number (%)
Smoking never smoked 223 (25.8) 14 (18.2)
ex-smoker 360 (41.6) 38 (49.4)
current smoker < 20/day 164 (18.9) 11 (14.3)
current smoker ≥ 20/day 118 (13.6) 14 (18.2) 0.21
* geometric mean
The unadjusted and adjusted relationships between Q fever seropositivity and case-control status are shown in Table 2. The odds of Q fever seropositivity in cases compared to controls was not elevated, either before or after adjustment for the baseline variables age, smoking, body mass index, systolic and diastolic blood pressure, social class, physical activity, and alcohol intake. Further adjustment for lipids also had little effect on observed odds ratios. The relationship strengthened slightly on adjusting for the inflammatory markers fibrinogen, CRP and IL-6, but was not statistically significant. A country/Q fever seropositivity interaction term was not significant when added to the fully adjusted model.
Table 2 Unadjusted and adjusted relationships between Q fever seropositivity and case-control status.
Unadjusted odds ratio (95% CI) Adjusted odds ratio1 (95% CI) Adjusted odds ratio2 (95% CI) Adjusted odds ratio3 (95% CI)
Q fever status
Seronegative 1.00 (reference category) 1.00 (reference category) 1.00 (reference category) 1.00 (reference category)
Seropositive 0.95 (0.57 1.57) 0.96 (0.56, 1.64) 1.04 (0.59, 1.82) 1.20 (0.66, 2.18)
1 adjusted for age, smoking, body mass index, systolic and diastolic blood pressure, social class, physical activity, and alcohol intake
2 further adjustment for total cholesterol, HDL cholesterol, and triglycerides
3 further adjustment for fibrinogen, CRP, and IL-6
Discussion
Atheroma is an inflammatory condition which has been putatively associated with infections including a range of bacterial infections. Against this background it has been suggested that the long-term sequelae of C. burnetii infection may include an elevated risk of cardiovascular disease[1,2]. Original tentative suggestions of a link were in the form of a case report of acute MI in a patient with Q fever infection[9]. More recently Enders et al. have suggested that C. burnetii infection may have a modest association with coronary artery disease. This is based on the evaluation of 155 consecutive patients undergoing coronary angiography using C. burnetii serology[1] Further positive associations have been reported in a follow-up study of people infected in 1983 in a large outbreak of Q fever in Switzerland[10] However this study was criticised for not controlling for current and past cigarette smoking as a potential explanation for the excess cardiovascular morbidity and mortality observed[3]. In this nested case-control study we were able to adjust for smoking and a range of other potential confounders, but this adjustment had little effect on observed odds ratios. We did not find any association between serological evidence of prior Q-fever infection and incident IHD. However, it is possible that selection/survivor bias may have occurred to some extent in the study. People with more severe Q fever infections, who may be more likely to develop some arterial disease if an association exists, may not have been enrolled in the study because of illness or death. In the Swiss study acute Q fever was found to be associated more closely with stroke than with myocardial infartion, however the stroke outcome could not be tested in our study. The reported chronic fatigue syndrome associated with Q fever appears not to be associated with cardiac disease. [11]
Smoking
Previous literature has suggested an association between smoking status and Q fever. In a large outbreak of Q fever pneumonia in Birmingham, UK in 1989, 60 (55%) of 110 patients for whom smoking data were available were current smokers, 28 (25%) were ex-smokers, and only 22 (20%) had never smoked[3]. A subsequent case-control study in the same cohort demonstrated smoking to be a risk factor for Q fever[12] Smoking was also shown to be an independent risk factor for acquisition of Q fever infection in goat workers in rural Newfoundland[13] Several possible explanations exist for these findings. Smoking may reflect the extent of hand to mouth contact which may be important in transmission of the infection, especially in an occupational exposure setting. Alternatively, smokers may be more likely to get symptomatic rather than asymptomatic disease. Our finding of no association between smoking status and seroprevalence would tend to favour the second explanation, assuming that the reported association with smoking and symptomatic Q fever is valid. Alternatively, an explanation of our finding is that that there is no increased risk of infection, either silent or clinical, in smokers.
Geographic differences
Differences in seroprevalence between Northern Ireland and France might have been expected, as they are very different in terms of topography and farming practice. However, the cohort recruitment methods employed in both countries were likely to exclude farmers. There was some evidence of higher prevalence in the southern French centre compared to the other centres, although conventional statistical significance was not achieved. An overall seroprevalence figure of almost 10% in populations not known to be occupationally exposed to Q fever testifies to the ubiquity of infection with this zoonotic agent in human populations in Europe.
Conclusion
This prospective study did not find any association between serological evidence of past infection with C. burnetii and IHD.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
CMcC and LJM jointly conceived the study, secured funding, participated in its design and coordination and co-drafted the manuscript. LJM also performed the main data analysis. JPMcK performed the Q fever serology testing and participated in the coordination of the study. HJON and PVC participated in the design and coordination of the study. DEW assisted in the coordination of the study and assisted in the editing of the manuscript. JWGY, JVW, PD, AB, PA, MM, DA, BH, JF and J-BR were responsible for the conception, design and coordination of the PRIME study in France and Northern Ireland. All authors helped revise and subsequently approved the final manuscript. CMcC and LJM are joint guarantors.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors wish to acknowledge the help and advice on statistical methodology given by Dr CC Patterson, Department of Epidemiology and Public Health, Queen's University of Belfast. The authors also wish to thank all the men who volunteered to be participants in the PRIME study.
This study was funded by a grant from the Research & Development Office, a directorate of the Northern Ireland Health and Social Services Central Services Agency (RRG project 9.8). The PRIME study was funded by Merck Sharpe & Dohme / Chibret (France).
==== Refs
Ender PT Phares J Gerson G Taylor SE Regnery R Challener RC Dolan MJ Association of Bartonella species and Coxiella burnetii infection with coronary artery disease J Infect Dis 2001 183 831 834 11181164 10.1086/318831
Lovey PY Morabia A Bleed D Peter O Dupuis G Petite J Long term vascular complications of Coxiella burnetii infection in Switzerland: cohort study BMJ 1999 319 284 286 10426735
Wildman M Ayres JG Long term vascular complications of Coxiella burnetii infection. Cardiovascular risk factors cannot be ignored BMJ 2000 320 58 59 10617545 10.1136/bmj.320.7226.58
Ducimetiere P Ruidavets JB Montaye M Haas B Yarnell J Five-year incidence of angina pectoris and other forms of coronary heart disease in healthy men aged 50-59 in France and Northern Ireland: the Prospective Epidemiological Study of Myocardial Infarction (PRIME) Study Int J Epidemiol 2001 30 1057 1062 11689522 10.1093/ije/30.5.1057
Yarnell JW The PRIME study: classical risk factors do not explain the severalfold differences in risk of coronary heart disease between France and Northern Ireland. Prospective Epidemiological Study of Myocardial Infarction QJM 1998 91 667 676 10024924 10.1093/qjmed/91.10.667
Rose GA Blackburn H Gillum RR Priveas RJ Cardiovascular survey methods Monograph Series No. 56 1982 2nd Geneva, World Health Organisation
Friedewald WT Levy RI Fredrickson DS Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge Clin Chem 1972 18 499 502 4337382
Clauss A [Rapid physiological coagulation method in determination of fibrinogen] Acta Haematol 1957 17 237 246 13434757
Pierangeli L Nuzzolo L Calabro R Villanti P [Myocardial infarct in the presence of serum antibodies against Coxiella burneti] Boll Soc Ital Cardiol 1967 12 205 209 5606567
Dupuis G Petite J Peter O Vouilloz M An important outbreak of human Q fever in a Swiss Alpine valley Int J Epidemiol 1987 16 282 287 3301708
Ayres JG Wildman M Groves J Ment J Smith EG Beattie JM Long-term follow-up of patients from the 1989 Q fever outbreak: no evidence of excess cardiac disease in those with fatigue QJM 2002 95 539 546 12145393 10.1093/qjmed/95.8.539
Ayres JG Flint N Smith EG Tunnicliffe WS Fletcher TJ Hammond K Ward D Marmion BP Post-infection fatigue syndrome following Q fever QJM 1998 91 105 123 9578893 10.1093/qjmed/91.2.105
Hatchette TF Hudson RC Schlech WF Campbell NA Hatchette JE Ratnam S Raoult D Donovan C Marrie TJ Goat-associated Q fever: a new disease in Newfoundland Emerg Infect Dis 2001 7 413 419 11384518
|
16033641
|
PMC1185539
|
CC BY
|
2021-01-04 16:28:15
|
no
|
BMC Infect Dis. 2005 Jul 20; 5:61
|
utf-8
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-61
|
oa_comm
|
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-621605095910.1186/1471-2334-5-62Research ArticleA commercial line probe assay for the rapid detection of rifampicin resistance in Mycobacterium tuberculosis: a systematic review and meta-analysis Morgan Maureen [email protected] Shriprakash [email protected] Laura [email protected] Madhukar [email protected] Division of Epidemiology, School of Public Health, University of California, Berkeley, U.S.A2 Department of Medicine, Mahatma Gandhi Institute of Medical Sciences, Sevagram, India3 Departamento de Biomedicina Molecular, CINVESTAV-IPN, Mexico4 Division of Pulmonary & Critical Care Medicine, San Francisco General Hospital, University of California, San Francisco, U.S.A2005 28 7 2005 5 62 62 14 4 2005 28 7 2005 Copyright © 2005 Morgan et al; licensee BioMed Central Ltd.2005Morgan 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
Mycobacterium tuberculosis is a leading cause of death worldwide. In multi-drug resistant tuberculosis (MDR-TB) infectiousness is frequently prolonged, jeopardizing efforts to control TB. The conventional tuberculosis drug susceptibility tests are sensitive and specific, but they are not rapid. The INNO-LiPA Rif. TB ® (LiPA) is a commercial line probe assay designed to rapidly detect rifampicin resistance, a marker of MDR-TB. Although LiPA has shown promising results, its overall accuracy has not been systematically evaluated.
Methods
We did a systematic review and meta-analysis to evaluate the accuracy of LiPA for the detection of rifampicin-resistant tuberculosis among culture isolates and clinical specimens. We searched Medline, Embase, Web of Science, BIOSIS, and Google Scholar, and contacted authors, experts and the manufacturer. Fifteen studies met our inclusion criteria. Of these, 11 studies used culture isolates, one used clinical specimens, and three used both. We used a summary receiver operating characteristic (SROC) curve and Q* index to perform meta-analysis and summarize diagnostic accuracy.
Results
Twelve of 14 studies that applied LiPA to isolates had sensitivity greater than 95%, and 12 of 14 had specificity of 100%. The four studies that applied LiPA directly to clinical specimens had 100% specificity, and sensitivity that ranged between 80% and 100%. The SROC curve had an area of 0.99 and Q* of 0.97.
Conclusion
LiPA is a highly sensitive and specific test for the detection of rifampicin resistance in culture isolates. The test appears to have relatively lower sensitivity when used directly on clinical specimens. More evidence is needed before LiPA can be used to detect MDR-TB among populations at risk in clinical practice.
==== Body
Background
Tuberculosis (TB) continues to be a major public health problem, particularly in developing countries. The WHO estimates that one third of the world's population is infected with Mycobacterium tuberculosis, the causative agent of TB. There were an estimated 8.3 million new active cases and 1.8 million deaths from TB in the year 2000, making it the second greatest killer among infectious diseases worldwide [1].
The prevalence of multidrug-resistant TB (MDR-TB), defined as resistance to at least rifampicin (RIF) and isoniazid (INH), is rising in a number of geographic regions. According to a recent WHO report [1], the median prevalence of MDR-TB is 1% (range 0%–14.1%) among new cases and 9.3% (range 0%–48%) among previously treated cases. Rapid identification is essential for effective treatment and control of MDR-TB. Conventional methods of drug susceptibility testing (DST) include solid media-based methods such as the proportion, absolute concentration, and resistance ratio methods. These can take up to 12 weeks to produce definitive results, leading to prolonged infectiousness [2]. Liquid media-based tests, such as the BACTEC®, MB/BacT®, ESP® and MGIT® systems, are more rapid, but also more costly and require sophisticated laboratories and trained personnel [2].
Rifampicin works by binding to the beta-subunit of the RNA polymerase (coded for by the rpoB gene), inhibiting protein transcription [3]. DNA sequencing studies have shown that greater than 95% of the RIF-resistant strains have mutations within an 81 base pair hot-spot region (codons 507–533) of the rpoB gene [4]. Though more than 50 mutations within this region have been characterized by automated DNA sequencing, the majority involve point mutations at codons 516, 526, or 531 [5]. It is estimated that more than 90% of RIF-resistant TB is also resistant to INH, making RIF-resistance a good surrogate marker for MDR-TB [4,6]. The above observations have lead to the recent development of several genotypic methods for rapidly detecting RIF-resistance conferring mutations, including DNA sequencing, line probe assay, single-strand conformation polymorphism, DNA microarrays, RNA/RNA mismatch, and molecular beacons [2].
The commercially available INNO-LiPA Rif. TB kit (Innogenetics, Zwijndrecht, Belgium) is a line probe assay (LiPA) able to identify the M. tuberculosis complex and simultaneously detect genetic mutations in the rpoB gene region related to RIF-resistance [7]. The LiPA kit contains 10 oligonucleotide probes (one specific for the M. tuberculosis complex, five overlapping wild-type S probes, and four R probes for detecting specific mutations of resistant genotypes) immobilized on nitrocellulose paper strips [7].
LiPA is performed by extracting DNA from cultures or directly from clinical samples and amplifying the RIF- resistance-determining region of the rpoB gene using PCR. Biotinylated PCR products are then hybridized with the immobilized probes, and results are determined by colorimetric development. The M. tuberculosis isolate is considered RIF susceptible if all of the wild-type S probes give a positive signal and all of the R probes react negatively. RIF resistance is indicated by absence of one or more wild-type S probes. When RIF resistance is due to one of the four most frequently observed mutations, a positive reaction is obtained with one of the four R probes [7].
A number of studies [3,5,7-19] have evaluated the diagnostic accuracy of LiPA for detecting RIF resistance in diverse geographic settings. We conducted a systematic review and meta-analysis to evaluate the overall accuracy of line probe assay in the detection of RIF-resistant TB.
Methods
Search strategy
We searched the following databases for retrieving articles and abstracts based on primary studies: Pubmed, Embase, Biosis, Web of Science (all 1990–2004), and Google Scholar (December 2004) using the keywords and search terms "Tuberculosis", "Mycobacterium tuberculosis", "Tuberculosis, Multidrug-Resistant", "Drug Resistance", "Drug Resistance, Bacterial", "rifampicin", "Rifampin", "mutation", "mutant", "rpob", "rpob gene", "line probe", "line probe assay", "LiPA", and "INNO-LiPA". We also contacted authors and experts, including the manufacturer of the commercial INNO-LiPA Rif.Tb kit, for lists of references and unpublished data, and reviewed citations of relevant primary studies and review articles.
Study selection
We identified results from all primary studies evaluating the accuracy (sensitivity and specificity) of line probe assay (specifically, the commercial INNO-LiPA Rif. TB kit) for rapid detection of RIF-resistant TB in clinical specimens or isolates. Titles and/or abstracts of all citations were screened independently by two reviewers (MM and SK), with 85% agreement on articles warranting full text review. Differences between reviewers were reconciled by consensus, and the full text of all relevant studies was evaluated.
We included studies that met the following pre-determined criteria: (i) comparison of INNO-LiPA with a reference standard (including proportion method, radiometric BACTEC 460 method, and minimum inhibitory concentration method), (ii) evaluation of a minimum of ten RIF-sensitive and ten RIF-resistant samples.
Although our initial search had no language restrictions, studies not available in either English or Spanish language were excluded from the data extraction process.
Data extraction and assessment of study quality
All included articles were assessed by one reviewer (MM), who extracted data using a piloted data extraction form. A second reviewer (LF) independently extracted data from a subset (five out of fifteen) of the included studies, with an inter-rater agreement between the two reviewers of 80% for sensitivity and specificity data. Discrepancies between reviewers were reconciled by consensus. Extracted data included the reference standard used, type of sample (clinical specimen vs. isolate), outcome data (sensitivity and specificity as determined by comparison with the reference standard), and proportion of RIF-resistant samples that were determined to be MDR-TB.
We assessed study quality using the following criteria, based on the QUADAS criteria [20] for assessment of quality of diagnostic studies: (i) prospective enrolment of consecutive patients, (ii) comparison with an appropriate reference standard, (iii.) blind and independent comparison of the index test (LiPA) with a reference standard, and (iv) verification (partial or complete) of LiPA results by reference standards.
Data synthesis and meta-analysis
We used standard methods for diagnostic meta-analysis [23,24], and performed data analysis using the Meta-Disc (version 1.1.1) software [23].
We focused on sensitivity and specificity as measures of diagnostic accuracy of LiPA. These were computed by creating a two by two table of LiPA RIF-susceptibility results against reference standard RIF-susceptibility results for each study and cross-tabulating. Sensitivity (true positive rate [TPR]) in this case is defined as the proportion of samples determined to be RIF-resistant by a reference standard correctly identified as RIF-resistant by LiPA. Specificity (true negative rate or 1-false positive rate [FPR]) is defined as the proportion of samples determined to be RIF-sensitive by a reference standard correctly identified as RIF-sensitive by LiPA. We created forest plots to display estimates of accuracy and examine the heterogeneity (between-study variability) of the summary measures of sensitivity and specificity.
We summarized the joint distribution of TPR and FPR with a summary receiver operating characteristic (SROC) curve. SROC curves used in analyses of diagnostic accuracy are intended to represent the relationship between TPR and FPR across studies when test performance is evaluated at varying diagnostic thresholds [24]. Each study is a separate unit of analysis and contributes an estimate of TPR and FPR. Overall diagnostic performance of a test can be judged by the position and appearance of the SROC curve, which is fitted by using a regression model proposed by Moses et al [25]. The area under the curve (AUC) represents an overall summary measure of the curve and the test's overall ability to accurately distinguish cases from non-cases. The Q* index, the highest point on the SROC curve that intersects the anti-diagonal, represents a summarization of test performance where sensitivity and specificity are equal (so the probability of an incorrect test result is the same for cases and non-cases). An AUC of one represents perfect discriminatory ability, while a Q* index of one represents perfect accuracy [24].
Results
Description of included studies
Figure 1 illustrates the study selection process. Fifteen articles [3,5,7-19], all reporting results of primary studies, met eligibility criteria and are included in this review.
Figure 1 Study selection process and reasons for exclusion of studies.
Table 1 describes the characteristics and outcomes of the 15 included studies. Three studies [10,11,19] are listed twice in order to describe the outcome of a subgroup analysis of LiPA applied directly to clinical specimens. All studies were published between 1995 and 2004 and used the commercial INNO-LiPA Rif. TB kit according to the manufacturer's instructions. Eleven studies [3,5,7,12-18] tested LiPA exclusively on culture isolates, one study [8] tested LiPA directly on clinical specimens, and three studies [10,11,19] tested LiPA on both isolates and clinical specimens. Clinical specimens included sputum, bronchial aspirate, urine, tissue biopsy, cerebrospinal fluid, feces, skin exudates, and gastric juice aspirate.
Table 1 Description of studies included in meta-analysis.
Author (year) Country Reference Test Blinded to reference test? Sample Sample size (# resistant / # sensitive) Sensitivity (95% CI) Specificity (95% CI)
Ahmad (2002) Kuwait BACTEC 460 Not Specified Isolate 29/12 0.97 (.82–1.0) 1.0 (.74–1.0)
De Oliveira (1998) Brazil Proportion Not Specified Isolate 113/15 0.97 (.92–.99) 1.0 (.78–1.0)
Gamboa (1998) Spain BACTEC 460 Not Specified Isolate 46/13 1.0 (.92–1.0) 1.0 (.75–1.0)
Hirano (1999) Japan Proportion Not Specified Isolate 90/26 0.92 (.85–.97) 1.0 (.87–1.0)
Johansen (2003) Denmark BACTEC 460 Not Specified Isolate 35/24 0.97 (.85–1.0) 1.0 (.86–1.0)
Jureen (2004) Sweden BACTEC 460 Not Specified Isolate 27/26 1.0 (.87–1.0) 0.92 (.75–.99)
Lemus (2004) Belgium BACTEC 460, Proportion Yes Isolate 10/10 1.0 (.69–1.0) 1.0 (.69–1.0)
Rossau (1997) Belgium Proportion Not Specified Isolate 203/61 0.98 (.95–.1.0) 1.0 (.94–1.0)
Sintchenko (1999) Australia BACTEC 460 Not Specified Isolate 22/11 0.96 (.77–1.0) 1.0 (.72–1.0)
Somoskovi (2003) USA Proportion Not Specified Isolate 64/37 0.95 (.87–.99) 1.0 (.91–1.0)
Srivastava (2004) India MIC Not Specified Isolate 45/10 0.82 (.68–.92) 1.0 (.69–1.0)
Tracevska (2002) Latvia BACTEC 460 Not Specified Isolate 34/19 1.0 (.90–1.0) 1.0 (.82–1.0)
Traore (2000) Belgium Proportion Not Specified Isolate 266/145 0.99 (.96–1.0) 1.0 (.98–1.0)
Watterson (1998) England BACTEC 460, Proportion Not Specified Isolate 16/16 1.0 (.80–1.0) 0.94 (.70–1.0)
De Beenhouwer (1995) Belgium Proportion Not Specified Clinical Specimen 21/46 0.91 (.70–1.0) 1.0 (.92–1.0)
Gamboa (1998) Spain BACTEC 460 Not Specified Clinical Specimen 46/13 0.98 (.89–1.0) 1.0 (.75–1.0)
Johansen (2003) Denmark BACTEC 460 Not Specified Clinical Specimen 26/21 1.0 (.87–1.0) 1.0 (.84–1.0)
Watterson (1998) England BACTEC 460, proportion Yes Clinical Specimen 10/24 0.80 (.44–.98) 1.0 (.86–1.0)
The 15 studies evaluated 1738 specimens (mean 91; range 20 to 411), 1164 (67%) of which were RIF-resistant. Twelve of the 15 studies include a greater number of RIF-resistant than RIF-sensitive strains (mean 87 and 36 respectively). Six studies [3,7-9,15,18] used proportion method, six studies [5,10-12,14,17] used BACTEC 460, two studies [13,19] used both proportion method and BACTEC 460, and one study [16] used minimum inhibitory concentration (MIC) method as the reference test. Only two studies [13,19] explicitly reported blinding researchers to the results of the reference standard and/or LiPA. None of the studies prospectively enrolled consecutive patients, and all had complete verification of LiPA with a reference standard.
Accuracy of LiPA in isolates
Figure 2 illustrates a forest plot of estimates of sensitivity and specificity based on results of the 15 included studies and stratified by type of sample (isolate vs. clinical specimen). Figure 3 is a SROC curve of the same data. As seen in figure 2, of the 14 studies that applied LiPA to isolates, sensitivity ranged from 82% to 100%, and specificity ranged from 92% to 100%. Twelve studies [5,7,9-15,17-19] reported sensitivity >= 95%, and five of these studies [10,12,13,17,19] reported sensitivity of 100%. With the exception of two [12,19], all studies reported specificity of 100%.
Figure 2 Forrest plots of sensitivity and specificity. The point estimates of sensitivity and specificity from each study are shown as solid circles (culture isolates) and open rectangles (clinical specimens). Error bars are 95% confidence intervals (CI).
Figure 3 Summary Receiver Operator Curve (SROC) plot for line probe assay. Each solid circle (culture isolate) and open rectangle (clinical specimen) represents each study in the meta-analysis. The curve is the regression line that summarizes the overall diagnostic accuracy. SROC: summary receiver operating characteristic; AUC: area under the curve; SE(AUC): standard error of AUC; Q*: an index defined by the point on the SROC curve where the sensitivity and specificity are equal, which is the point closest to the top-left corner of the ROC space; SE(Q*): standard error of Q* index.
The SROC curve, figure 3, shows an area of 0.99 and Q* of 0.97, indicating a high level of overall accuracy.
Subgroup analysis of accuracy of LiPA in clinical specimens
As illustrated in figure 2, of the four studies that tested LiPA directly on clinical specimens [8,10,11,19], sensitivity estimates, although more variable than specificity, are consistently high (80% to 100%) with one study [11] reporting a sensitivity of 100%. The specificity estimates for all four studies are 100%.
Although still consistently high, sensitivity appears to be lower overall in clinical specimens than isolates. Additionally, one study [11] explicitly stated that 13 of the 60 samples tested were indeterminate due to failure at the PCR stage, making it impossible to perform LiPA. These are excluded from measures of sensitivity and specificity, indicating that the overall accuracy of LiPA applied to clinical specimens may be inflated in this study (and possibly others if they experienced similar indeterminate results that went unreported) when compared with performance in an actual clinical setting.
Rifampicin-resistance as a marker of MDR-TB
Four studies [5,8,12,18] determined the number of RIF-resistant samples that were also INH-resistant, thereby meeting the criteria for MDR-TB. On average, 91% of RIF-resistant samples were also INH-resistant.
Discussion
Principle findings
This meta-analysis suggests that the LiPA assay is highly sensitive and specific for detecting rifampicin-resistant TB both in culture isolates and, to a slightly lesser degree, clinical specimens. The majority of studies had sensitivity of 95% or greater, and nearly all were 100% specific.
Despite variations in patient populations, all 15 studies yielded consistently high estimates of sensitivity and specificity, so heterogeneity was not a concern in this meta-analysis [26].
Clinical implications
The currently employed DST methods typically delay the diagnosis of MDR-TB by at least one to two months. A more rapid method is needed to allow timely diagnosis and initiation of effective treatment. This meta-analysis demonstrates that LiPA yielded high overall sensitivity and specificity with a maximum joint sensitivity and specificity of 97% (based on the Q* index). The test may thus have a potential role in ruling in and ruling out the diagnosis of RIF-resistance. For example, assuming that 5% of TB patients in a clinical setting have RIF-resistant TB, a positive LiPA result (inferring RIF-resistance) would yield a positive predictive value of 83%, while a negative LiPA result would yield a negative predictive value of 99%. These test results would lead to a clinically meaningful increase in the probability of RIF-resistance from 5% to 83% if a test is positive, while a negative test would virtually rule out RIF-resistance. Because the test has a high sensitivity, a negative result would effectively rule out the probability of drug resistance. Similarly, because the test has a high specificity, a positive result would rule in drug resistance. However, the diagnostic accuracy of LiPA needs to be interpreted cautiously in low prevalence areas. For example, if the baseline prevalence of rifampicin resistance is 1%, a positive test would translate into a positive predictive value of only 66%, i.e. one false positive test for every two true positives. As with any diagnostic test, if used judiciously (ie in patients suspected of having MDR-TB, thereby raising the pretest probability) the accuracy of LiPA could be maintained even in low prevalence regions.
Because patients with MDR-TB are more likely to be put on an effective drug therapy regimen if the drug resistance is quickly detected, and thus are less likely to transmit MDR-TB to the community, the benefits of early detection of drug resistance can be substantial. A positive test in a high prevalence setting can lead to a highly meaningful shift from pre-test to post-test probability and thus may facilitate better outcomes.
LiPA has shown a high degree of accuracy when used on culture isolates, but this requires 2–6 weeks for primary isolation. Only four studies applied LiPA directly to clinical specimens, resulting in slightly more variation in the degree of accuracy than those studies using isolates. Additional research is needed to establish the accuracy of LiPA applied to clinical specimens, but the preliminary studies suggest that LiPA may help diagnose RIF-resistant TB within 24–48 hours of sample collection.
The cost of the commercial LiPA kit is $45 per sample tested. When additional costs for import and transport are taken into account, the actual cost per sample is as high as $116 [27]. Though this may be prohibitively expensive for routine use in the regions of the world with the highest prevalence and incidence of TB and MDR-TB, judicious use of LiPA for patients with a high likelihood of MDR-TB (for example, smear-positive patients with treatment failure or relapse from high incidence areas and/or previously treated patients) may be possible, particularly when weighed against the costs of undetected drug resistant TB. An additional challenge to widespread implementation of LiPA is the requirement of a lab with technical expertise in performing PCR.
Strengths and weaknesses of the review
This review has several strengths. We performed a comprehensive search for literature by exploring five electronic databases and by contacting authors, experts, and the manufacturer of the reviewed index test. Study selection was conducted independently by two reviewers, as was data extraction and quality review for a subset of included studies, and disagreements were resolved with discussion. We performed meta-analyses in accordance with published guidelines [21,22].
This review has some limitations. We excluded studies not available in English or Spanish language, which could introduce publication bias. However, a review of the abstracts of these papers suggests that the overall results are similar to the results in the included English and Spanish language studies. Publication bias may also be introduced by inflation of diagnostic accuracy estimates since studies that report positive results are more likely to be accepted for publication. The studies included in this meta-analysis apply LiPA to a total of 1738 MTB positive samples, of which 1164 are RIF-resistant. This prevalence of 67% differs significantly from the prevalence of MDR-TB seen in routine clinical practice settings, even in high prevalence regions such as Estonia (14.1%), Henan Province in China (10.8%), Latvia (9%), and the Russian oblasts of Ivanovo (9%) and Tomsk(6.5%) [1]. Because the specimens analyzed in the studies are not a true representation of specimens that a TB laboratory would actually receive, estimates of sensitivity and specificity may be inflated. Finally, estimates of sensitivity and specificity may be inflated in these studies due to exclusion of indeterminate results from measures of accuracy if failure occurred at the PCR stage, which precludes performance of LiPA on the specimen or isolate.
Implications for research
Additional studies are needed to establish the accuracy of LiPA used directly on clinical specimens. Study design should include selection of sputum samples from patients suspected of having MDR-TB (ie patients with treatment failure or relapse from high incidence areas and/or previously treated patients). Indeterminate results, the proportion of RIF-resistant specimens that meet MDR-TB criteria, patients' sputum smear status, and turnaround time for diagnosis should be reported.
Studies are also needed to establish clinical usefulness of rapid diagnosis of RIF-resistant TB in terms of the effect on clinical outcomes and TB transmission rates. Finally, studies are needed to establish the cost benefit advantages of LiPA over conventional DST.
Conclusion
Line probe assay has been shown to be highly sensitive and specific in the detection of rifampicin-resistant TB when used on culture isolates. There is a paucity of data on application of this test directly to clinical specimens, although based on a small number of studies, the test appears to be less promising. The cost of the kit may render the test impractical for widespread use in those regions of the world most affected by MDR-TB and most in need of a method for its rapid diagnosis. However if further studies indicate that line probe assay consistently and accurately detects RIF-resistant TB when applied directly to clinical specimens, it could be a useful test in select patient populations in which MDR-TB is strongly suspected.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MM designed the study, searched the databases, extracted the data, analyzed the results and wrote the manuscript. SK helped with study design, searching the databases, writing and revising the manuscript, and served as a second reviewer in screening articles for inclusion. MP formulated the research question, helped with study design, database searches, analysis, and in revising the manuscript. LF helped design the data abstraction form, provided critical input in laboratory associated issues and served as a second reviewer in extracting data. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
SK and MP are supported by the National Institutes of Health, Fogarty AIDS International Training Program (1-D43-TW00003-15), and NIH/NIAID (R01 AI34238).
We are grateful to the following authors who sent additional information on their primary studies: Wouter Mijs, Max Salfinger, Tatjana Tracevska, Cosme Alvarado Esquivel, Cengiz Cavusoglu, and Philip Suffys, and Isik Johansen. We thank Donna Adams (Innogenetics) for providing useful information on the INNO-LiPA Rif Tb kit, as well as a list of published references.
We thank Gloria Won for assisting us in the retrieval of references from Embase, and Dr. Ed Desmond, California Department of Health Services, for providing useful feedback on an earlier version of this manuscript.
==== Refs
The WHO/IUATLD Global Project on Anti-Tuberculosis Drug Resistance Surveillance Anti-Tuberculosis drug resistance in the world, report number 2 2004 Geneva
Heifets LB Cangelosi GA Drug susceptibility testing of Mycobacterium tuberculosis: a neglected problem at the turn of the century Int J Tuberc Lung Dis 1999 3 564 581 10423219
Hirano K Abe C Takahashi M Mutations in the rpo B gene of rifampin-resistant Mycobacterium tuberculosis strains isolated mostly in Asian countries and their rapid detection by line probe assay J Clin Microbiol 1999 37 2663 2666 10405418
Cavusoglu C Hilmioglu S Guneri S Bilgic A Characterization of rpo B mutations in Rifampin-resistant clinical isolates of Mycobacterium tuberculosis from Turkey by DNA sequencing and line probe assay J Clin Microbiol 2002 40 4435 4438 12454132 10.1128/JCM.40.12.4435-4438.2002
Ahmad S Mokaddas E Fares E Characterization of rpo B mutations in rifampin-resistant clinical Mycobacterium tuberculosis isolates from Kuwait and Dubai Diagn Microbiol Infect Dis 2002 44 245 252 12493171 10.1016/S0732-8893(02)00457-1
Drobniewski FA Wilson SM The rapid diagnosis of isoniazid and rifampin resistance in Mycobacterium tuberculosis – a molecular story J Med Microbiol 1998 47 189 196 9511823
Rossau R Traore H De Beenhouwer H Mijs W Jannes G De Rijk P Portaels F Evaluation of the INNO-LiPA Rif. TB assay, a reverse hybridization assay for the simultaneous detection of Mycobacterium tuberculosis complex and its resistance to rifampin Antimicrob Agents Chemother 1997 41 2093 2098 9333031
De Beenhouwer H Lhiang Z Jannes G Mijs W Machtelinckx L Rossau R Traore H Portaels F Rapid detection of Rifampicin resistance in sputum and biopsy specimens from tuberculosis patients by PCR and line probe assay Tuber Lung Dis 1995 76 425 430 7496004 10.1016/0962-8479(95)90009-8
De Oliveira M Rocha A Oelemann M Gomes H Fonseca L Werneck-Barreto A Valim A Rossetti M Rossau R Mijs W Vanderborght B Suffys P Rapid detection of resistance against Rifampicin in isolates of Mycobacterium tuberculosis from Brazilian patients using a reverse-phase hybridization assay J Microbiol Methods 2003 53 335 342 12689711 10.1016/S0167-7012(02)00253-1
Gamboa F Cardona PJ Manterola JM Lonca J Matas L Padilla E Manzano JR Ausina V Evaluation of a commercial probe assay for detection of rifampin resistance in Mycobacterium tuberculosis directly from respiratory and nonrespiratory clinical samples Eur J Clin Microbiol Infect Dis 1998 17 189 192 9665301
Johansen I Lundgren B Sosnovskaja A Thomsen V Direct detection of multidrug-resistant Mycobacterium tuberculosis in clinical specimens in low- and high-incidence countries by line probe assay J Clin Microbiol 2003 41 4454 4456 12958292 10.1128/JCM.41.9.4454-4456.2003
Jureen P Werngren J Hoffner S Evaluation of the line probe assay (LiPA) for rapid detection of Rifampicin resistance in Mycobacterium tuberculosis Tuberculosis 2004 84 311 316 15207806 10.1016/j.tube.2003.12.001
Lemus D Martin A Montoro E Portaels F Palomino J Rapid alternative methods for detection of Rifampicin resistance in Mycobacterium tuberculosis J Antimicrob Chemother 2004 54 130 133 15190018 10.1093/jac/dkh320
Sintchenko V Chew W Jelfs P Gilbert G Mutations in rpo B gene and rifabutin susceptibility of multidrug-resistant Mycobacterium tuberculosis strains isolated in Australia Pathology 1999 31 257 260 10503273 10.1080/003130299105089
Somoskovi A Song Q Mester J Tanner C Hale Y Parsons L Salfinger M Use of molecular methods to identify the Mycobacterium tuberculosis complex (MTBC) and other mycobacterial species and to detect rifampin resistance in MTBC isolates following growth detection with BACTEC MGIT 960 System J Clin Microbiol 2003 41 2822 2826 12843007 10.1128/JCM.41.7.2822-2826.2003
Gamboa F Cardona PJ Manterola JM Lonca J Matas L Padilla E Manzano JR Ausina V Correlation of mutations detected by INNO-LiPA with levels of Rifampicin resistance in Mycobacterium tuberculosis Indian J Med Res 2004 120 100 105 15347859
Tracevska T Jansone I Broka L Marga O Baumanis V Mutations in the rpo B and katG genes leading to drug resistance in Mycobacterium tuberculosis in Latvia J Clin Microbiol 2002 40 3789 3792 12354882 10.1128/JCM.40.10.3789-3792.2002
Traore H Fissette K Bastian I Devleeschouwer M Portaels F Detection of Rifampicin resistance in Mycobacterium tuberculosis isolates from diverse countries by a commercial line probe assay as an initial indicator of multidrug resistance Int J Tuberc Lung Dis 2000 4 481 484 10815743
Watterson S Wilson S Yates M Drobniewski F Comparison of three molecular assays for rapid detection of rifampin resistance in Mycobacterium tuberculosis J Clin Microbiol 1998 36 1969 1973 9650946
Whiting P Rutjes AW Reitsma JB Bossuyt PM Kleijnen J The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews BMC Med Res Methodol 2003 3 25 14606960 10.1186/1471-2288-3-25
Deeks JJ Systematic reviews in health care: Systematic reviews of evaluations of diagnostic and screening tests BMJ 2001 323 157 162 11463691 10.1136/bmj.323.7305.157
Pai M McCulloch M Enanoria W Colford JM Systematic reviews of diagnostic test evaluations: what's behind the scenes? ACP Journal Club 2004 141 11 A 15230559
Zamora J Muriel A Abraira V Meta-DiSc for Windows: A Software package for the Meta-analysis of Diagnostic Tests
Walter SD Properties of the summary receiver operating characteristic (SROC) curve for diagnostic test data Stat Med 2002 21 1237 1256 12111876 10.1002/sim.1099
Moses LE Shapiro D Littenberg B Combining independent studies of a diagnostic test into a summery ROC curve: data-analytic approaches and some additional considerations Stat Med 1993 12 1293 1316 8210827
Lijmer JG Bossuyt PM Heisterkamp SH Exploring sources of heterogeneity in systematic reviews of diagnostic tests Stat Med 2002 21 1525 1537 12111918 10.1002/sim.1185
Morcillo N Zumarraga M Alito A Dolmann A Schouls L Cataldi A Kremer K van Soolingen D A low cost, home-made, reverse-line blot hybridization assay for rapid detection of rifampicin resistance in Mycobacterium tuberculosis Int J Tuberc Lung Dis 2002 6 1 7 11931394
|
16050959
|
PMC1185540
|
CC BY
|
2021-01-04 16:28:14
|
no
|
BMC Infect Dis. 2005 Jul 28; 5:62
|
utf-8
|
BMC Infect Dis
| 2,005 |
10.1186/1471-2334-5-62
|
oa_comm
|
==== Front
BMC MedBMC Medicine1741-7015BioMed Central London 1741-7015-3-131602660710.1186/1741-7015-3-13Research ArticleChanges in standard of candidates taking the MRCP(UK) Part 1 examination, 1985 to 2002: Analysis of marker questions McManus IC [email protected] J [email protected] OL [email protected] JA [email protected] Dept of Psychology, University College London, Gower Street, London WC1E 6BT, UK2 MRCP(UK) Central Office, 11 St Andrews Place, Regents Park, London NW1 4LE, UK2005 18 7 2005 3 13 13 20 4 2005 18 7 2005 Copyright © 2005 McManus et al; licensee BioMed Central Ltd.2005McManus 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 maintenance of standards is a problem for postgraduate medical examinations, particularly if they use norm-referencing as the sole method of standard setting. In each of its diets, the MRCP(UK) Part 1 Examination includes a number of marker questions, which are unchanged from their use in a previous diet. This paper describes two complementary studies of marker questions for 52 diets of the MRCP(UK) Part 1 Examination over the years 1985 to 2001 to assess whether standards have changed.
Methods
Study 1, which used routinely collected information on the performance of 4405 marker items, used a statistical method to assess changes in performance across diets. Study 2 compared performances of individual candidates on 28 individual marker items that were shared by the 1996/2 and 2001/3 diets.
Results
Study 1 found evidence that candidate performance on the MRCP(UK) Part 1 Examination showed a gradual improvement over the period 1985 to 1997, which was followed by a sharp decline in performance until 2001. The 'dog-leg' in performance at 1997/3 was not an artefact of changed Examination Regulations, mix of UK and overseas candidates, or time from qualification until taking the Examination. Study 2 confirmed that performance in 2001/3 was significantly worse than in 1996/3, that the poorer performance was found in graduates of UK medical schools, and that candidates passing the Examination in 2001/3 performed less well than those passing in 1996/2.
Conclusion
There has been a decline in the performance of graduates from UK medical schools taking the MRCP(UK) Part 1 examination. The reasons for this are not clear, but the finding has implications for medical education, and further studies are needed of performance in other postgraduate and undergraduate examinations. The use of norm-referencing as the sole method for setting the pass mark over this period meant that candidates passing the MRCP(UK) examination also had a lower standard. The MRCP(UK) Part 1 and Part 2 examinations now have their standard set by criterion-referencing.
==== Body
Background
The role of postgraduate medical examinations is to set standards of practice and thus to assure the public and the medical profession that doctors have the knowledge and expertise required to diagnose and treat patients, and to progress in their medical careers. An important part of running "high-stakes" qualifying examinations is the setting of a pass mark, and many postgraduate examinations in the UK have relied on norm-referencing, in which a fixed proportion of candidates passes at each occasion [1].
Although administratively straightforward, norm-referencing has several problems. The absolute performance across different diets of the exam varies, firstly due to asking questions of different difficulty, and secondly as a result of candidates differing in their ability. These factors of question difficulty and candidate ability are entirely confounded when only overall examination performance is considered, and norm-referencing cannot assess whether candidate ability varies from occasion to occasion. The result, as was shown in an important and influential analysis of the American Board of Internal Medicine's examination, was a slide in standards over the time-period 1983 to 1988 [2]. On that basis the Board implemented a process of criterion-referencing, in which examiners set a pass mark by assessing the content of each individual question on the examination. Having said that, we also acknowledge that there are arguments in some situations for the use of norm-referencing [3], and as a result there are even stronger arguments for the use of compromise methods [4,5].
In this paper we primarily wish to assess whether the absolute standard of candidates taking the MRCP(UK) examination changed over the period 1985 to 2002, when the format of the examination and the method for setting its pass mark were relatively constant. The assessment of the true ability of candidates, independently of question difficulty, requires a process of equating to establish whether candidates on one occasion are of equivalent ability to those on another occasion. Statistical equating can be carried out if marker questions are available, the same questions being used on two different diets [6].
Background to Study 1
In this paper we describe two complementary analyses of marker questions used in the MRCP(UK) Part1 Examination between 1985 and the first diet of 2002 (2002/1). Three diets of the Examination were held each year, and in all diets the pass rate was norm-referenced at 35% of those candidates taking the Examination on their first four attempts at UK examination centres.
Two separate studies are described:
Study 1: Analysis of aggregate performance of marker questions, 1988/1 to 2002/1.
Study 2: Comparison of 1996/2 and 2001/3 diets
Although Study 2 follows on from Study 1, conceptually and in practice, the requirements of the journal are that the method of each is presented before the results of each, and then the discussion of each. This is somewhat confusing, and in particular it is necessary to present the background and justification to study 2 before the results of Study 1. Readers who are confused by this layout are referred to the Additional File, where the text of the paper is ordered in a more logical format.
Background to Study 2
As will become apparent, Study 1 demonstrates that performance on marker questions declined during the period 1997 to 2002. However there are several possible interpretations of that result, not least because the data are aggregated across all candidates and do not allow analysis of sub-groups of candidates, such as UK graduates on first and subsequent attempts. Study 2 therefore analysed raw data at the level of responses to individual questions from individual candidates, thereby allowing a detailed comparison of the two diets.
Methods
Methods: Study 1
The format of the Examination, which was held three times a year, was unchanged until 2002/1, consisting of 60 five-part Multiple True-False questions (a total of 300 items). Negative marking was used in scoring the examination, and results were expressed as a 'corrected percentage correct', which takes guessing into account. The only substantive change in the Examination Regulations was that candidates were allowed to make an unlimited number of attempts from 1999/2 onwards, whereas previously they had been limited to four attempts.
Marker items were defined as any items included for a second time in any of the 43 diets of the MRCP(UK) Part 1 Examination held between 1988/1 and 2002/1, and which had been used in the previous diet with the stem, the item and the correct answer unchanged. Documents for tracing markers prior to 1988 were not readily available. The dates of the two diets, and the proportion of candidates getting each item right, wrong or not answering it, were recorded for each marker item.
Statistical analysis: Study 1
The central issue in a marker question analysis is whether, on average, aggregate performance on items has increased or decreased between the first and second usage. By assessing how such a change relates to the dates when the items were used, one can estimate the changing overall performance of candidates in the Examination. The study is a variant of an incomplete paired comparison design [7,8].
If the percentages of candidates who get an item correct or wrong on the first and second occasion are c1, w1 and c2, w2, then the change in performance of the item is:
Δ = (c2 - w2) - (c1 - w1)
It should be noted that Δ is independent of the proportion of candidates choosing not to answer an item since those not answering are effectively neutral, neither gaining a mark from a correct answer nor losing one due to a wrong answer. Given the different performance of the marker questions at different diets, one can reconstruct the changing true ability of the candidates by a process analogous to triangulation, a process that can be carried out using multiple regression (for a technical explanation see below). The method takes one arbitrarily chosen date as the reference category (the first diet of 2000 was used) and estimates the ability of the candidates at each other diet relative to the reference diet. The estimates, like all regression coefficients, have a standard error and confidence intervals.
Statistical method, Study 1: details of technique
Although there are many articles and books devoted to test equating, they usually consider only the problem of a large set of overlapping items occurring in two adjacent diets of an examination. The marker questions for Part 1 are large in number but are distributed across many diets, making a different problem for estimation. It is interesting to note that the regression solution described here is very general in its applicability, and was previously used in an entirely different context [9].
The exam is taken on n occasions. Let the standard of the candidates vary, such that for diet d, the true standard, relative to some arbitrary reference diet, r, is sd (i.e. sr = 0). If the same item (question) were to be used on every diet, then on diet 1, a proportion s1 would get the item correct, on diet 2 s2 would get the item correct, etc.
A diet has a variable number of marker items, which have been used unchanged in a number of previous diets. An individual marker item, m, in diet k, will have been used previously in, say, diet j. The statistical analysis is required to derive the true standard of the candidates at the various diets, si (i = 1, n; sr = 0), from the performance on the marker items.
Let the difference in performance on marker item m between diets j and k be expressed as Δm. In the specific case of a true-false examination with negative marking, let cj % get the item correct at time j, wj % get the item wrong at time j, and naj % not answer the item at time j (and with equivalent symbols for time k). Δm is then calculated as:
Δm = (ck-wk)-(cj-wj)
Note that if the performance of the candidates on the question is the same at diets j and k then
Δm = 0.
A series of dummy variables, v1 to vn, is then created for each marker variable, which take the values :
vj = -1
vk = +1
vp = 0 [p = 1, n; p not equal to j, p not equal to k ]
sd (d = 1,n) can be estimated using multiple regression, where Δm is the dependent variable, and the predictor variables are v1 to vn (excluding the reference diet, vr). The unstandardised regression coefficients of vd (d = 1,n) are then the estimates of sd (d = 1,n), with sr = 0, and the standard errors of the regression coefficients are the standard errors of the estimates of sd (d = 1,n).
SPSS version 11.5 was used for statistical analysis of the data. For statistical analysis, diets 1, 2 and 3 in a year were set at 0, .33 and 0.67 of the year.
Methods: Study 2
Raw data were available only for the 1996/2 diet and the diets from 1997/1 onwards, and of these only the 1996/2 and 2001/3 diets shared sufficient marker items for analysis (see figure 1). Since the 1996/2 diet took place just before the apparent decline in performance seen in figure 2, whereas 2001/3 took place afterwards, a comparison is appropriate.
Figure 1 Date of previous setting (vertical axis) of marker questions used in various diets (horizontal axis). The horizontal axis shows a particular diet of the exam, and the vertical axis the previous diets from which marker items were taken. Size of points is proportional to the number of questions.
Figure 2 The estimated true ability of candidates taking the examination at each diet. The solid points are the estimates from the regression (sd). The thick black line is a fitted lowess line (locally weighted regression), and the thin black lines show lowess lines through the one standard error confidence intervals for the points. The thick dashed line is the 'dog-leg' curve (see text). The vertical dashed line indicates the date when unlimited attempts were allowed in the Part 1 examination.
The 1996/2 and 2001/3 diets had 28 items in common, which were scored on the basis of +1 for a correct answer, -1 for a wrong answer and 0 for no answer, giving a 'marker score' expressed as a corrected percentage score.
For the 1996/2 diet the pass mark, defined by the MRCP(UK) Part 1 Examining Board as the performance of the top 35% of candidates taking the Examination in the UK on their 1st to 4th attempts, was 54.84%, and for the 2001/3 diet it was 48.50%. For comparative purposes only, candidates are here divided into four groups, 'Fail, 'Bare Fail', 'Bare Pass' and 'Pass'; the threshold between Bare Fail and Bare Pass was the pass mark itself, and the thresholds between Fail and Bare Fail and between Bare Pass and Pass were set five absolute percentage points below or above the pass mark. The thresholds defining the four groups are therefore 49.84%, 54.84% and 59.84% for the 1996/2 diet, and 43.50%, 48.50% and 53.50% for the 2001/3 diet.
All analyses were restricted to 'UK graduates', defined as those with primary registrable qualifications from United Kingdom medical schools.
Results
Results: Study 1
Altogether 5332 marker questions were used in the 43 diets held between 1988/1 and 2002/1, and these had previously been used on diets extending back to 1972. Although in principle the statistical method can assess standards outside of the range of the diets actually assessed, a preliminary analysis suggested the process was unstable prior to about 1985, with large standard errors of the estimates. Analysis is therefore restricted to the 4405 marker questions that provided information on the performance of the 52 diets held between 1985/1 and 2002/1.
Figure 1 shows the timing of the two diets in which a marker item had been used. On average each diet, which consisted of 300 items, contained 124 marker items (41.3%), (SD 23.4; range 72–173). Marker items had on average been used 5.7 years previously (i.e. 17 diets; SD 2.22, range 1.33–24.33). The average number of marker items in any one diet that came from the same previous diet was 10.4 (SD 7.6, range 1 to 48).
The individual points in figure 2 show the estimated ability of candidates at each diet from 1985/1 to 2002/1, sd, calculated by regressing the dummy variables v1 to vn on the values of Δm. The ability of candidates increased slowly but consistently between 1985 and about 1996, after which ability appears to decline fairly steeply. That hypothesis was formalised by using non-linear regression to fit two straight lines to the data, each with its own independent slope, with a 'dog-leg' mid-way through the data, the date of the dog-leg itself being a free parameter. The dog-leg curve shown in figure 2 fits the data well (R2 = .468), and is a significant improvement over a simple linear regression (F(2,48) = 18.28, p < 0.001). The inflexion of the dog-leg is at 1997.3 (equivalent to the 1997/2 diet), with 95% confidence intervals of 1996.0 to 1998.6 (equivalent to diets 1996/1 to 1998/3). The slope before 1997 is +0.18%/year (95% CI 0.03% to 0.33%), and the slope after 1997 is -1.34%/year (95% CI -1.93% to -0.75%).
Results: Study 2
The 1996/2 diet was taken by 2132 candidates of whom 852 were UK graduates. The 2001/3 diet was taken by 2051 candidates of whom 557 were UK graduates. The 28 marker items were combined into a single scale for which alpha was 0.593, which is equivalent, using the Spearman-Brown formula [10], to a reliability of 0.940 for a full-length 300 item test, somewhat higher than the mean reliability across 54 diets of the examination as a whole of 0.865 [11].
Overall performance on marker items: Study 2
In the 1996/2 diet, the 852 UK graduates have a mean marker score of 69.4 (SD 14.5), compared with a mean marker score of 59.6 (SD 15.8) for the 557 UK graduates in the 2001/3 diet (t = 11.92, 1407 df, p < 0.001). The average mark in 2001/3 was therefore 14.1% lower than in 1996/2 when precisely the same items are compared. Detailed analyses of items answered correctly, incorrectly or not answered showed that the 2001/3 candidates had answered fewer of the 28 items correctly (20.54 vs 22.40; p < 0.001) and more of the items incorrectly (3.84 vs 2.98; p < 0.001), and more were not answered (3.84 vs 2.61; p < 0.001). The reduced performance in the 2001/3 diet was independent of whether candidates were taking the Examination on their first, second or subsequent attempts (see Figure 3; ANOVA: diet F(1,1362) = 96.7, p < 0.001; attempt F(4,1362) = 3.63, P = 0.006; interaction F(3,1362) = 1.22, P = 0.300).
Figure 3 Overall performance on the 28 marker items for UK graduates taking the 1996/2 and 2001/3 diets of the examination, on their first and subsequent attempts. Until diet 1999/2 candidates were only allowed to take the exam four times, whereas after 1999/2 they were allowed unlimited attempts. Fewer than 20 candidates were on their 6th or higher attempt, and they have been omitted from the analysis.
Performance in relation to pass mark: Study 2
Analysis of the marker scores of candidates in each diet according to candidates' overall performances in the Examination showed lower scores in 2001/3 than in 1996/2 for all ability levels, including those who had passed the exam (figure 4: ANOVA: diet F(1,1401) = 255.3, p < 0.001; ability group F(3,1401) = 237.30, P < 0.001; interaction F(3,1401) = 0.705, P = 0.549).
Figure 4 Performance of UK candidates on the marker items in relation to overall examination performance (see text for details of Fail, Bare Fail, Bare Pass, and Pass). Points are plotted ± 2 standard errors.
Performance on individual marker items in 1996/2 and 2001/3: Study 2
In 2001/3, UK candidates performed significantly less well (p < 0.05) on 17 of the 28 individual marker items (table 1) and better on 2 items, and performance was not significantly different for 9 items (see Figure 5). The median change in performance was -7.65% (inter-quartile range -1.42% to -15.73%).
Table 1 Performance of UK graduates on marker items in the 1996/2 and 2001/3 diets (N = 852 and 555 respectively). Significance tests are shown for each item.
1: Recognised features of infectious mononucleosis include: 1996/2 Q.13 2001/3 Q.15
A: Petechial haemorrhages on the palate [True]
(Kendall's tau = -0.038, p = 0.153) Right 81.0% 77.7%
NA 9.5% 11.9%
Wrong 9.5% 10.5%
B: a vesicular rash on the neck and trunk [False]
(Kendall's tau = -0.097, p < 0.001) Right 76.1% 67.4%
NA 12.8% 15.1%
Wrong 11.2% 17.5%
C: aseptic meningitis [True]
(Kendall's tau = -0.178, p < 0.001) Right 82.5% 67.4%
NA 10.8% 15.1%
Wrong 6.7% 17.5%
E: raised serum aspartate aminotransferase activity [True]
(Kendall's tau = -0.080, p = 0.003) Right 91.4% 86.3%
NA 6.7% 10.8%
Wrong 1.9% 2.9%
Item D was not a marker item
2: A patient with chronic osteitis of the femur associated with a discharging sinus has a haemoglobin of 10.5 g/dL, MCV 78 fl, MCH 30 pg, WBC 10.8 × 109/L, platelets 420 × 109/L, and a normal blood film. The following statements are correct: 1996/2 Q.17 2001/3 Q.21
B: Antibiotic therapy is the probable cause of the blood picture [False]
(Kendall's tau = -0.078, p = 0.004) Right 85.8% 79.6%
NA 8.1% 12.3%
Wrong 6.1% 8.1%
C: C-reactive protein levels will be normal [False]
(Kendall's tau = +0.118, p < 0.001) Right 83.3% 91.5%
NA 6.3% 4.3%
Wrong 10.3% 4.1%
D: Parenteral iron therapy would be valuable in treating the anaemia [False]
(Kendall's tau = -0.080, p = 0.004) Right 92.6% 87.9%
NA 4.5% 5.4%
Wrong 2.9% 6.7%
E: Abundant stainable iron will be found in bone marrow macrophages [True]
(Kendall's tau = -0.160, p < 0.001) Right 45.5% 27.9%
NA 34.9% 44.1%
Wrong 19.6% 27.9%
NB Item D was not a marker item
3: Recognised features of acute poisoning due to theophylline include: 1996/2 Q.27 2001/3 Q.24
A: vomiting [True]
(Kendall's tau = -0.040, p = 0.139) Right 92.7% 90.5%
NA 5.8% 7.6%
Wrong 1.5% 2.0%
B: convulsions [True]
(Kendall's tau = -0.056, p = 0.043) Right 95.9% 93.3%
NA 3.2% 5.4%
Wrong 0.9% 1.3%
C: supraventricular tachycardia [True]
(Kendall's tau = -0.086, p = 0.001) Right 88.4% 81.8%
NA 6.5% 12.8%
Wrong 5.2% 5.4%
D: hypokalaemia [True]
(Kendall's tau = -0.044, p = 0.093) Right 69.0% 65.8%
NA 18.9% 17.1%
Wrong 12.1% 17.1%
E: metabolic acidosis [True]
(Kendall's tau = +0.040, p = 0.110) Right 26.9% 30.8%
NA 39.8% 38.7%
Wrong 33.3% 30.5%
4: Recognised features of Wolff-Parkinson-White syndrome include: 1996/2 Q.33 2001/3 Q.33
A: an accessory connection between atria and the atrioventricular node [False]
(Kendall's tau = -0.177, p < 0.001) Right 85.7% 71.1%
NA 3.3% 4.9%
Wrong 11.0% 24.0%
B: prolonged PR interval [False]
(Kendall's tau = -0.096, p = 0.001) Right 94.4% 89.0%
NA 0.9% 3.6%
Wrong 4.7% 7.4%
C: prolonged QRS complex [True]
(Kendall's tau = -0.286, p < 0.001) Right 76.2% 47.5%
NA 4.8% 9.0%
Wrong 19.0% 43.5%
D: paroxysmal ventricular tachycardia [False]
(Kendall's tau = -0.223, p < 0.001) Right 65.8% 41.7%
NA 9.0% 15.2%
Wrong 25.1% 43.1%
E: dominant R waves in lead V1 of the ECG [True]
(Kendall's tau = +0.032, p = 0.213) Right 71.1% 74.0%
NA 18.8% 17.5%
Wrong 10.1% 8.5%
5: The following clinical features suggest an organic basis for psychiatric symptoms: 1996/2 Q.40 2001/3 Q.41
A: disorientation in time [True]
(Kendall's tau = -0.074, p = 0.006) Right 86.2% 80.0%
NA 3.3% 8.1%
Wrong 10.6% 11.9%
B: visual hallucinations [True]
(Kendall's tau = -0.122, p < 0.001) Right 86.9% 77.3%
NA 3.2% 5.8%
Wrong 10.0% 17.0%
C: mutism [False]
(Kendall's tau = -0.150, p < 0.001) Right 84.0% 70.6%
NA 8.9% 19.9%
Wrong 7.0% 9.6%
D: inability to retain new information [True]
(Kendall's tau = +0.071, p = 0.005) Right 67.0% 72.9%
NA 9.2% 10.8%
Wrong 23.8% 16.2%
E: perseveration [True]
(Kendall's tau = -0.004, p = 0.885) Right 45.3% 39.0%
NA 16.5% 29.2%
Wrong 38.1% 31.8%
6: Characteristic features of schizophrenia include: 1996/2 Q.43 2001/3 Q.40
A: memory impairment [False]
(Kendall's tau = +0.009, p = 0.725) Right 91.3% 91.9%
NA 5.4% 4.7%
Wrong 3.3% 3.4%
B: auditory hallucinations in clear consciousness [True]
(Kendall's tau = -0.125, p < 0.001) Right 98.5% 93.9%
NA 0.4% 1.8%
Wrong 1.2% 4.3%
C: incongruity of affect [True]
(Kendall's tau = -0.085, p = 0.001) Right 81.9% 73.8%
NA 6.8% 13.9%
Wrong 11.3% 12.3%
D: feelings of panic in buses and shops [False]
(Kendall's tau = -0.037, p = 0.178) Right 96.5% 94.9%
NA 2.0% 3.4%
Wrong 1.5% 1.6%
E: a feeling of being under the influence of an external force [True]
(Kendall's tau = -.038, p = 0.185) Right 99.1% 98.2%
NA 0.7% 1.1%
Wrong 0.2% 0.7%
Figure 5 Performance of the 28 individual marker items in the 1996/2 and 2001/3 diets. The diagonal line is the point of equality; points below the line represent items for which performance is worse in 2001/3 than in 1996/2. Solid points are statistically significant (p < .05), and open points are non-significant. Code numbers of items are those shown in table 1.
Conclusion
Conclusions: Study 1
In the analysis of marker items in Study 1, candidates showed a gradual improvement in performance on the MRCP(UK) Part 1 examination for the twelve years between 1985 and 1997. Such a change could either result from a genuine increase in candidates' knowledge or from candidates becoming more 'test-wise', perhaps due to attending crammer courses and becoming more aware of questions from previous papers. However, the latter explanation seems unlikely in view of the sharp decline in performance of candidates after 1997.
If the decline in performance after 1997 reflects decreasing candidate ability, then that has important implications for medical education and training, and possible confounders and artefacts must be excluded. Separate analyses, not reported in full here but available in the Additional File, have investigated the following possibilities:
1. Change in number of attempts allowed
From 1999/2 onwards, candidates were not restricted to four attempts at the Examination. Figure 2, however, shows that the decline in performance started well before that date.
2. Changes in strategy in relation to negative marking
The MRCP(UK) Part 1 examination was negatively marked, wrong answers incurring a greater penalty than unanswered questions. If candidates changed their strategy, that may have produced an apparent decline in performance. We have modelled the number of not-answered questions, and although this did show a decline in between 1990 and 1993, the timing was unrelated to the change shown in Figure 2.
3. Changes in the mix of candidates taking the Examination
Because the marker question statistics recorded in the Examination records are based on aggregate statistics for all candidates taking the examination, changes in the mix of candidates could cause a change in marker question performance. We have looked at the relative proportion of UK and overseas graduates, the numbers of candidates on first, second, third, fourth and later attempts, and the time between qualification and taking the MRCP(UK) Part 1 Examination, and although secular trends are visible, none shows a sudden change in 1997. (See Additional File).
This study of 4,405 marker items suggests the standard of candidates taking the Part 1 MRCP(UK) Examination may have changed over time, rising gradually until about 1997, and then declining rather more rapidly. The discontinuity around 1997 does not seem to be related to any obvious change in the structure of the Examination or the composition of the candidates taking it.
Conclusion: Study 2
Study 2 examined the performances of individual candidates on marker items and confirms the earlier finding that performance dropped between 1996/2 and 2001/3. In particular, fewer UK graduates who passed the Examination in 2001/3 gave correct answers than had equivalent candidates in 1996/2. The median change in performance on items of -7.75% was similar to the expected change of -7.1%, based on the -1.34% per year shown in figure 2. The results of figure 3 are probably not therefore distorted or biased by inadvertent differences in candidate mix between the diets.
We found that the performances on individual marker items between 1996/2 and 2001/3 had dropped for 21 of the 28 items. None of the questions is about recondite, obscure or unimportant areas of knowledge for a general physician in training, and none of the changes are likely to reflect changes in the importance of knowledge, in understanding of disease mechanisms, or in treatment strategies. They are therefore acceptable marker questions. The largest decreases were on the electrocardiography and anatomy of the Wolff-Parkinson-White syndrome, aseptic meningitis in infectious mononucleosis, and bone marrow biopsy findings in the anaemia of chronic infection, all of which are important clinical problems. The only two significant increases in knowledge are on questions concerning C-reactive protein (a relatively recently introduced clinical test), and on the diagnosis of organic brain disease.
Discussion: General
These two complementary studies have implications specifically for postgraduate medical examinations, and more generally for undergraduate medical education. The studies provide evidence that there was a sudden, relatively steep decline in the performance of candidates passing the MRCP(UK) Part 1 Examination between 1997 and 2002, which was not an artefact of changes in the mix of overseas and UK candidates, or changes in the time after qualifying of first or subsequent sittings of the examination. Study 2 confirmed that the decline had taken place in doctors graduating from UK medical schools.
i). Implications for standard setting in postgraduate medical examinations
The MRCP(UK) examination sets a standard for professional clinical practice in the UK. Our use of marker questions for assessing the standard across diets parallels a study in 1989, which described the falling standard of candidates passing the American Board of Internal Medicine (ABIM) examination [2] (although there are differences in the way marker questions were used). In both the ABIM examination and the MRCP(UK) Part 1 Examination the declining standard of candidates passing the examination probably arose from reliance on the sole use of norm-referencing for standard setting. Any other examination relying solely on norm-referencing may also be vulnerable to the same problem.
The MRCP(UK) Part1 and Part 2 written Examinations have recently changed their format, the Part 1 Examination now consisting entirely of 'best-of-five' questions. The MRCP(UK) Part 1 and Part 2 Examinations both carry out standard-setting by a process incorporating criterion referencing using the Angoff technique [12,13], with the pass mark itself set by the Hofstee compromise technique [4], which reduces the likelihood of large short-term swings in the pass rate. The pass rates in the three MRCP(UK) Part 1 Examination diets of 2003 for candidates on their first four attempts sitting the exam at UK centres were 31.5%%, 33.4% and 32.3%, somewhat less than the 35% that would have occurred using norm-referencing. Performance in the MRCP(UK) Part 1 Examination is continuing to be monitored by the use of marker questions.
ii). Implications for undergraduate and postgraduate medical education
The decline in performance of candidates from UK medical schools taking the MRCP(UK) Part 1 Examination raises questions extending beyond the Part 1 Examination itself. The examination can be taken eighteen months after graduation, and a high proportion of UK graduates take it at the earliest possible time, when they typically have five or six years of undergraduate education, a year of PRHO posts, and six months of SHO training. Several explanations need to be considered for the changes in standard that we have found.
i. Changing relevance of the examination questions
Topics once perceived as central to medical training may now no longer be important to modern medical practice. If the marker questions used were out-of-date then that may explain the apparent decline. However, not only is the content of marker questions always approved by MRCP(UK) Part 1 Examining Board before each inclusion in the Examination, but the questions shown in table 1 clearly relate to core conditions and their underlying disease mechanisms, and hence changes cannot be shrugged off as resulting from irrelevant or outmoded questions.
ii. Changing career patterns of graduates
The present results relate only to one examination, the MRCP(UK) Part 1 Examination, albeit an exam taken by over 30% of UK graduates. Corroboration of the present findings from other UK postgraduate examinations is desirable, in order to assess the generality of the findings. It is possible that around 1997 more able UK graduates candidates decided they no longer wished to take the MRCP(UK) Part 1 Examination, and instead took other career paths (and that seemed to be the explanation for the declining standard in the ABIM examination [2]). Although perhaps unlikely, the possibility can be assessed by analysing marker questions from other postgraduate examinations, which should then show an improved performance by UK graduates.
iii. Changes in clinical experience
In recent years the working hours of junior doctors have declined, in part due to changes in Government regulations, and clinical experience and hence examination performance may also have declined. In the absence of good measures of clinical experience this hypothesis is difficult to test. There is, however, evidence that the undergraduate clinical experience of UK doctors qualifying in 1996 was lower than that of doctors qualifying a decade earlier [14], and that more recent medical graduates have less knowledge of basic clinical science [15].
iv. Changes in undergraduate medical training
Undergraduate medical training in the UK has been continually changing for nearly four decades, dating back primarily to the Royal Commission of 1968 [16], and supported by subsequent recommendations from the General Medical Council [17,18]; new subjects were introduced into the curriculum, and traditional subjects such as anatomy were de-emphasised. Particularly dramatic changes followed the General Medical Council's Tomorrow's Doctors [19] of 1993, as a result of which many medical schools introduced major curricular changes, often involving problem-based learning. A number of medical schools also merged, and most medical schools became larger. Although these latter innovations might have caused changes in the knowledge-base of graduates, they are unlikely to explain changes we describe here, which began in 1997 and hence relate to students entering medical school in 1990 or 1991, before the publication of Tomorrow's Doctors. A key question concerns whether the standards of undergraduate examinations, both basic medical sciences and finals, have been maintained; however, the patchy use of marker questions, frequent changes in undergraduate examination formats, and the absence of a UK national medical licensing examination make it unlikely that the question can be answered easily. Indeed, the only reliable evidence on the absolute standard of undergraduate training may have to come from performance in postgraduate examinations.
In summary, we have provided evidence of a decline in the performance of candidates taking the MRCP(UK) Part 1 Examination between 1997 and 2001. In addition, as a result of the reliance on norm-referencing, there was also a decline in the standard of those passing the Examination. Criterion referencing is now included as a central part of the MRCP(UK) Examination standard setting process. The reasons for the declining standard of UK graduates are not clear, but on balance are more likely to reflect changes in undergraduate training than changes in postgraduate medical education. More research into the standard of other postgraduate examinations, as well as undergraduate assessments, is urgently needed.
Competing interests
ICM, JAV, OLD and JM have all been involved with the MRCP(UK) for many years. JM is a paid employee of the MRCP(UK) central office, JAV, OLD, and ICM contribute to the examination academically, and JAV and OLD receive financial compensation for the extensive work that this involves.
Authors' contributions
The original idea for the study was ICM's, arising from discussions that occurred at Board Meetings chaired by JAV. JM contributed extensive statistical and administrative support to the study, and JAV and OLD contributed to the interpretation of results and the writing of the paper. ICM wrote the first draft of the manuscript and carried out the multivariate statistical analyses. All authors have contributed to the final draft of the paper.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Changes in the composition and timing of entries to MRCP(UK) Part 1.
Click here for file
Acknowledgements
We are grateful to a number of members of the MRCP(UK) Central Office and to MRCP(UK) examiners for discussion of the issues arising from this paper.
==== Refs
Anonymous Examining the Royal Colleges' examiners [editorial] Lancet 1990 335 443 445 1968172 10.1016/0140-6736(90)90670-Z
Norcini JJ Maihoff NA Day SC Benson JA Jr Trends in medical knowledge as assessed by the certifying examination in internal medicine Journal of the American Medical Association 1989 262 2402 2404 2795825 10.1001/jama.262.17.2402
McHarg J Bradley P Chamberlain S Ricketts C Searle J McLachan JC Assessment of progress tests Medical Education 2005 39 221 227 15679690 10.1111/j.1365-2929.2004.02060.x
Hofstee WKB Anderson SB, Helmick JS The Case for Compromise in Educational Selection and Grading On Educational Testing 1983 San Francisco: Jossey-Bass 109 127
De Gruijter DNM Compromise models for establishing examination standards Journal of Educational Measurement 1985 22 263 269 10.1111/j.1745-3984.1985.tb01063.x
Kolen MJ Brennan RL Test equating: Methods and practices 1995 New York: Springer-Verlag
David HA The method of paired comparisons 1963 London: Charles Griffin
Bock RD Jones LV The measurement and prediction of judgement and choice 1968 San Francisco: Holden-Day
McManus IC Edmondson D Rodger J Balance in pictures British Journal of Psychology 1985 76 311 324
Ghiselli EE Campbell JP Zedeck S Measurement theory for the behavioral sciences 1981 San Francisco: W H Freeman
McManus IC Mooney-Somers J Dacre JE Vale JA Reliability of the MRCP(UK) Part I Examination, 1984–2001 Medical Education 2003 37 609 611 12834418 10.1046/j.1365-2923.2003.01568.x
Angoff WH Thorndike RL Scales, norms, and equivalent scores Educational Measurement 1971 Washington, DC: American Council on Education 508 600
Livingston SA Zieky MJ Passing scores: a manual for setting standards of performance on educational and occupational tests 1982 Philadelphia: Educational testing service
McManus IC Richards P Winder BC Clinical experience of UK medical students Lancet 1998 351 802 803 9519956 10.1016/S0140-6736(05)78929-7
McKeown PP Heylings DJA Stevenson M McKelvey KJ Nixon JR McCluskey DR The impact of curricular change on medical students' knowledge of anatomy Medical Education 2003 37 954 961 14629407 10.1046/j.1365-2923.2003.01670.x
Royal Commission Royal Commission on Medical Education (The Todd Report), Cmnd 3569 1968 London: HMSO
General Medical Council Recommendations on basic medical education 1980 General Medical Council: Education Committee
General Medical Council Recommendations on general clinical training 1987 General Medical Council
General Medical Council Tomorrow's Doctors: Recommendations on undergraduate medical education 1993 London: General Medical Council
|
16026607
|
PMC1185541
|
CC BY
|
2021-01-04 16:27:56
|
no
|
BMC Med. 2005 Jul 18; 3:13
|
utf-8
|
BMC Med
| 2,005 |
10.1186/1741-7015-3-13
|
oa_comm
|
==== Front
BMC Med EducBMC Medical Education1472-6920BioMed Central London 1472-6920-5-271602273810.1186/1472-6920-5-27Research ArticleWhat motivates senior clinicians to teach medical students? Dahlstrom Jane [email protected] Anna [email protected] Darryl [email protected] Cathy [email protected] Kathleen [email protected] D Ashley R [email protected] Medical School, Building 42A, ANU, Canberra, ACT 0200, Australia2005 18 7 2005 5 27 27 23 3 2005 18 7 2005 Copyright © 2005 Dahlstrom et al; licensee BioMed Central Ltd.2005Dahlstrom 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
This study was designed to assess the motivations of senior medical clinicians to teach medical students. This understanding could improve the recruitment and retention of important clinical teachers.
Methods
The study group was 101 senior medical clinicians registered on a teaching list for a medical school teaching hospital (The Canberra Hospital, ACT, Australia). Their motivations to teach medical students were assessed applying Q methodology.
Results
Of the 75 participants, 18 (24%) were female and 57 (76%) were male. The age distribution was as follows: 30–40 years = 16 participants (21.3%), 41–55 years = 46 participants (61.3%) and >55 years = 13 participants (17.3%). Most participants (n = 48, 64%) were staff specialists and 27 (36%) were visiting medical officers. Half of the participants were internists (n = 39, 52%), 12 (16%) were surgeons, and 24 (32%) were other sub-specialists. Of the 26 senior clinicians that did not participate, two were women; 15 were visiting medical officers and 11 were staff specialists; 16 were internists, 9 were surgeons and there was one other sub-specialist. The majority of these non-participating clinicians fell in the 41–55 year age group. The participating clinicians were moderately homogenous in their responses. Factor analysis produced 4 factors: one summarising positive motivations for teaching and three capturing impediments for teaching. The main factors influencing motivation to teach medical students were intrinsic issues such as altruism, intellectual satisfaction, personal skills and truth seeking. The reasons for not teaching included no strong involvement in course design, a heavy clinical load or feeling it was a waste of time.
Conclusion
This study provides some insights into factors that may be utilised in the design of teaching programs that meet teacher motivations and ultimately enhance the effectiveness of the medical teaching workforce.
==== Body
Background
Clinical teachers are central to the successful education of medical graduates. They are a precious resource with a range of competing activities like clinical care and research. In order to better recruit and retain clinical teachers, medical schools must be cognisant of the variety of factors that may motivate doctors to teach students. Medical schools have increased expectations of clinical teachers with curriculum development, small group dynamic teaching and assessment responsibilities, yet have little direct line management of clinical teachers. In Australia, universities do not pay their clinical teachers and therefore they do not "own" them. Hospitals pay doctors and implicitly expect them to teach as a service to the profession. There is no clearly stated contractual requirement. Substantial resource is put into training clinical teachers and schools are interested in minimising teacher turnover. Initial motivation and any accompanying rewards are central to remaining motivated to teach. Equally medical schools have an incentive to recruit satisfied and effective teachers in order to improve educational outcomes.
A motive can be defined as an entity that impels one to action of a particular type. In contemporary psychology, motivation encompasses three areas: drives (innate origins of behaviour that impel an individual to action), goals or purpose (a conscious plan for action that entices an individual to action) and reinforcers (entities that increase or decrease the probability of a behaviour being replicated such as pleasure/pain or reward/punishment) [1]. Motivation is a complex concept investigated in a number of ways. Maslow identified five levels of "needs" or drives that motivate behaviour: physiological needs (to satisfy hunger, thirst, shelter), safety needs (for security, order and stability), belonging and love needs (affection and identification), esteem needs (prestige, success and self-respect) and the need for self actualisation [2]. Goals may be achievement oriented (such as meeting a set of learning goals) or prosocial goals (such as peer acceptance and respectability) [3]. Herzberg, (quoted in Ellis) contrasted extrinsic rewards (pay and benefits) and the inherent intrinsic rewards (such as self respect and personal achievement) [4]. These three areas of drives, goals and rewards all overlap and influence behaviour in different ways at different times.
So what is understood of the motivations of clinical teachers? Clinical supervisors rated predetermined possible motivations to explain their volunteering to teach: personal satisfaction was highest, followed by the opportunity to attract students to one's speciality area. Less important was any sense of prestige or improved standing amongst peers. Despite focusing on an intrinsic motivator, this group still identified a need to be acknowledged by the medical school. Faculty appointments, discounted continuing education, access to computerised information and libraries along with better education as clinical teachers were all valued as suitable rewards. Supervisors in open responses did not suggest monetary compensation [5]. Work focusing on medical teachers' reluctance to teach noted ten impediments including lack of reward, perverse incentives of academic promotion by research with little value on teaching, lack of teaching skill, competitive agendas of clinical service and research, obtuse curriculum redesign and administrative blocks like high student teacher ratios [6]. Broader study on motivation for academic career progression amongst academic physicians found gender and work focus (research vs teaching) influenced motivation. Compared with male physicians, female physicians were more motivated in work by the desire to help others while clinician researchers valued self-expression as a more powerful motivator than did clinician educators [7].
A review of rewards and incentives for non-salaried clinical teachers found that most medical schools offered some incentive such as educational opportunities, academic appointments and special recognition events [8]. Medical faculty (directly paid by a university to teach) valued recognition of outstanding teaching (the Dean's teaching awards) and educational development as a teacher as suitable reward [9].
This study applied Q methodology, an established sorting method, to quantify subjective views on motivation. This method has previously successfully screened aspiring schoolteachers as to why they chose a career in teaching [10] (accessed 16/02/2002) and reviewed career satisfaction amongst nurses [11].
We aimed to investigate what motivates our senior clinicians to teach medical students in order to identify factors that may assist in the design of teaching programs that meet teacher motivations and ultimately enhance the effectiveness of the medical teaching workforce.
Methods
Study group
In our context, this work was quality improvement in education and did not require formal ethics approval. The study group was all senior clinicians registered on the medical student teaching staff list at a public, tertiary level regional teaching hospital (The Canberra Hospital, Canberra, Australia). Their appearance on the list implied some interest in teaching. Demographic data collected included gender, age group (30–40, 41–55, >56 years), employment contract (salaried staff specialist or contracting visiting medical officer), and the speciality of the clinician (internist, surgeon, or other sub-specialist – pathologists, radiologists, psychiatrists etc.).
Q methodology
This method was chosen as the assessment and statistical method for the study as it combines qualitative and quantitative research traditions [12] (accessed 07/06/2005). Q methodology can reveal the subjectivity in a situation and although initially used in personality assessment, has been applied to a range of psychological investigations [13]. The method allows a quantitative evaluation of the opinion of individuals about topics of common concern. This leads to a composite of opinions that may be aggregated into viewpoints [14].
Central to Q methodology is the ranking of single statements on a continuum detailing the degree of agreement or disagreement with the statement. Unlike single dimension questionnaires, the use of a quasinormal distribution forces participants to rank statements relative to the other statements about the question of concern. The ranking of individual opinions about the statements facilitates the formation of individual viewpoints about the overall subjective question to which the sample statements referred. The viewpoints are then compared in a correlation matrix to identify similarities between individual viewpoints.
Choosing the sample population (P-set)
Individuals are chosen to participate in a study based on their relevance to the goals of the study, as opposed to being selected for their representativeness of a larger population. This collection of individuals is referred to as a "person-set", or P-set [15]. The P-set for this study was drawn from senior clinicians on the teaching list of the Canberra Clinical School at The Canberra Hospital (n = 101). Of these, 26 were on leave or not contactable during the time agreed for data collection. A limited time frame for the data collection was required to limit staff discussion of the study between participants before providing their opinion. All 75 contactable staff agreed to participate. Apart from instructing the participant on the study objectives, the method used and how to complete the questionnaire, there was no other discussion between the participant and the researcher.
Creating the Q sample
This is the creation of statements used to examine the topic of investigation. These need to be drawn from people with expertise in the issue under study. They may be developed through focus group discussions or brainstorming [11]. The clinicians conducting the study (which represented the population of interest i.e. senior teaching clinicians) contributed statements representing reasons clinicians may teach, or not teach, medical students. All members were experienced clinical teachers undertaking further studies in medical education. From the study group deliberations and through examination of the relevant literature, a representative set of 69 statements (the Q sample) was created (Table 1). Approximately equal numbers of positive and negative statements were created. The statements were consecutively numbered and printed onto labels. The labels were then put onto note cards with one statement per card.
Table 1 Statements included in the Q sort
Statement
1. I enjoy spending time with students in small groups
2. I don't enjoy lecturing to large groups of students
3. I like the challenge of teaching students as effectively as possible
4. I am bored by teaching
5. I don't feel any sense of duty to teach
6. I teach because it sets a good example to my students to become teachers
7. I teach because I have been inspired to teach by my mentors
8. I teach because I am good at it relative to other academic skills
9. I teach because it is a requirement of my employment contract
10. I teach because I believe it is an appropriate service to my profession
11. Teaching doesn't do anything to enhance my clinical knowledge and/or skills
12. I teach because I enjoy the sense of performing in front of an audience
13. I don't get any financial reward from teaching
14. I teach because I want to help my students become good doctors
15. I don't teach because I am not the one most familiar with a given topic
16. I don't teach because my institution provides poor facilities for teaching
17. I don't teach because I have insufficient time available to teach
18. I teach because there are opportunities for 'virtual' and/or 'online' and/or remote teaching
19. I don't teach as my speciality is too 'cutting edge' to be relevant to students
20. I don't teach because there are no clearly stated learning goals in the course
21. I don't teach because there is no strong involvement of teaching staff in the design of the course
22. I don't teach because there is no recognition for what I do
23. Opportunities for academic promotion have nothing to do with my motivation to teach
24. I teach because the course allows a deep approach to learning by the students
25. I don't teach as students make me feel inadequate
26. I don't teach because opportunities are not available for me to improve my teaching skills
27. I don't teach because I receive inadequate feedback from students on my performance
28. I teach because I believe I communicate well with people
29. I don't teach because I believe the institution devalues teaching and learning
30. I don't teach because the setting in which I am expected to teach is inappropriate
31. I teach because I feel part of the continuum of learning of my students' experience
32. I teach because I feel responsible for the student learning outcomes of my efforts
33. I teach because it gives me a sense of power
34. I teach to improve my communication skills
35. I don't teach as I am not a useful role model
36. My clinical load deters me from teaching
37. My clinical load deters me from putting any time into preparation for teaching
38. I don't teach because I am not concerned about the success of the clinical and/or medical school
39. The teaching I had as a medical student has inspired me to want to teach
40. I teach as a means of reviewing a topic area unfamiliar to me
41. I teach to be challenged in my established views
42. I don't teach because I find it unenjoyable
43. It teach because of the prestige it gives me with my peers
44. I teach because my patients expect it of me
45. I don't teach because interacting with students is boring
46. I teach because of the intellectual stimulation
47. I teach because my colleagues expect me to do so
48. I teach because I was asked to do so by the Clinical and/or Medical School
49. I don't teach because it fails to keep me up to date
50. I teach students because interaction with them makes me think more critically
51. I teach students to ensure they receive a balanced clinical education.
52. I teach because I can enhance my knowledge and understanding of junior doctors
53. I teach because the interaction with students provides an opportunity for my opinions to be heard
54. I teach to ensure the students appreciate my specialty in a favourable way
55. I teach because it allows me to interact with students and show an appreciation of their position
56. I don't teach just because it is expected of me
57. I teach students to show them the correct way of clinical practice in my specialty
58. I teach to ensure any false understanding of my specialty is not perpetuated
59. I teach because I can demonstrate a healthy lifestyle to my students
60. I don't teach just because of the academic position I hold
61. I teach because I can challenge students to be more critical in their thinking.
62. I don't teach because one can't influence the behaviour of students for the better
63. I don't teach because teachers don't contribute to the formation of future doctors
64. I don't teach as I don't approve of new teaching techniques
65. I don't teach as students today lack respect
66. I don't teach as it is a waste of time
67. I teach to engage with younger people
68. I teach as it enhances my status in my profession
69. I don't teach as I feel my knowledge is out of date
Creating the Q sort and data collection
After formation of the Q sample, a quasi-normal distribution containing as many cells as there were Q sample statements was created [11]. A table (11 columns by 11 rows) was made and then reduced to a series of cells in a symmetrical forced normal distribution (Figure 1). This distribution was referred to as a Q sort and was the data collection instrument. A cover sheet also contained an outline of the purpose of the study, instructions to participants and recorded demographic data. An enlarged version of the Q sort was created to fit the note cards to facilitate the physical sorting of statements (Q sort diagram).
Figure 1 Each member of the group approached an allocated subset of the senior medical staff to request their participation. After a brief discussion of the methodology each consenting participant was then asked to undertake the study at a convenient time and place. Each participant was given the stack of 69 note cards containing the questions and instructed to sort all cards according to their level of agreement or disagreement onto the large Q sort diagram. The 69 note cards with each statement were placed by the clinician on the ranking space perceived as most appropriate for that individual statement in comparison to the other statements. This manual form of positioning of the note card with a statement on it allowed the ranking of a statement to be altered easily when comparing it to each new statement assessed by the participant. The final positioning of all statements therefore provided a ranking of the statements relative to each other in importance as determined by the individual clinician.
Demographic information was collected on each participant. The Q sample statements and data from the participants were entered into the Q methodology freeware program downloaded from the Q Method Page [16] (accessed 03/06/2002).
Data analysis
Factor analysis using the Centroid approach with varimax rotation was subsequently used to identify common themes among participants' viewpoints. The variables for analysis were the statement rankings produced by the clinicians. The purpose of the method was to identify orthogonal factors (i.e. at 90 degrees) representing different points of view among the clinicians. The clinician's loading on a factor indicated his/her shared viewpoint with other clinicians on the same factor. The clinicians loading on any one factor indicated a shared common viewpoint about their reasons for teaching by the similar rank ordering of their statement rankings. Positive scores indicated agreement, while negative scores indicated disagreement. Decreasing scores reflected less important views. Importantly, these factors were not considered as traits of the group, psychological or otherwise. They represented only common subjective opinions obtained at a single point in time in a cross-sectional survey method i.e. they reflected correlations between people not items. Q methodology yields ipsative rather then normative data in that a person reveals "individual" opinions rather than in comparison with another opinion [17]. Each identified factor was examined for distinguishing statements (those with scores that were significantly different at the p < 0.05 and p < 0.01 level from the same statement's score on other identified factors) [11].
Results
Demographic data
All 101 teachers were approached for participation. Of the 75 participants 18 (24%) were female and 57 (76%) were male with the following age distributions: 30–40 = 16 (21.3%), 41–55 = 46 (61.3%) and greater than 55 = 13 (17.3%). Most were staff specialists (n = 48, 64%) and 27 (36%) were visiting medical officers. Half of the participants were internists (n = 39, 52%), 12 (16%) were surgeons and 24 (32%) were other groups.
Of the 26 senior clinicians who did not participate 2 were women, 15 were visiting medical officers and 11 staff specialists, 16 were internists, 9 surgeons and 1 other. All but 2 of these clinicians fell in the 41–55 year age group. The participating teachers were seen to not differ significantly from the complete teaching list. Most nonparticipants were on leave during the study period.
Factor analysis
The data sorted into four factors: Factor 1 (Table 2 accounting for 68 participants' sorts), Factor 2 (Table 3, 3 sorts), Factor 3 (Table 4, 2 sorts) and Factor 4 (Table 5, 2 sorts). In summary, Factor 1 the " I teach because..." factor, represented most participants agreeing with a range of positive statements about the value of teaching and disagreeing with statements phrased negatively about teaching. Highly ranked positive items included:
Table 2 Distinguishing statements for factor 1. (n = 68)
No. Statement Rank Score
14 I teach because I want to help my students become good doctors 5 1.84*
3 I like the challenge of teaching students as effectively as possible 4 1.56*
57 I teach students to show them the correct way of clinical practice in my specialty 3 1.44*
1 I enjoy spending time with students in small groups 3 1.41*
7 I teach because I have been inspired to teach by my mentors 3 1.38*
54 I teach to ensure the students appreciate my specialty in a favourable way 2 1.04*
41 I teach to be challenged in my established views 2 1.02*
39 The teaching I had as a medical student has inspired me to want to teach 2 0.97*
32 I teach because I feel responsible for the student learning outcomes of my efforts 2 0.97*
52 I teach because I can enhance my knowledge and understanding of junior doctors 2 0.96*
58 I teach to ensure any false understanding of my specialty is not perpetuated 2 0.96*
40 I teach as a means of reviewing a topic area unfamiliar to me 1 0.90
55 I teach because it allows me to interact with students and show an appreciation of their position 1 0.77*
24 I teach because the course allows a deep approach to learning by the students 1 0.38
53 I teach because the interaction with students provides an opportunity for my opinions to be heard 1 0.31*
12 I teach because I enjoy the sense of performing in front of an audience 0 -0.22*
43 I teach because of the prestige it gives me with my peers -1 -0.44
22 I don't teach because there is no recognition for what I do -1 -0.78*
30 I don't teach because the setting in which I am expected to teach is inappropriate -1 -0.86*
29 I don't teach because I believe the institution devalues teaching and learning -2 -0.89
64 I don't teach as I don't approve of new teaching techniques -2 -0.95
69 I don't teach as I feel my knowledge is out of date -2 -1.07
62 I don't teach because one can't influence the behaviour of students for the better -2 -1.10
19 I don't teach as my speciality is too 'cutting edge' to be relevant to students -2 -0.11*
5 I don't feel any sense of duty to teach -3 -1.12*
38 I don't teach because I am not concerned about the success of the clinical and/or medical school -3 -1.15*
42 I don't teach because I find it unenjoyable -3 -1.15*
63 I don't teach because teachers don't contribute to the formation of future doctors -3 -1.15*
49 I don't teach because it fails to keep me up to date -4 -1.19*
25 I don't teach as students make me feel inadequate -4 -1.23*
45 I don't teach because interacting with students is boring -4 -1.32*
4 I am bored by teaching -5 -1.53*
66 I don't teach as it is a waste of time -5 -1.78*
NB1. All P < 0.05 (* Indicates significance at P < 0.01)
Table 3 Distinguishing statements for factor 2 (n = 3)
No. Statement Rank Score
21 I don't teach because there is no strong involvement of teaching staff in the design of the course 5 1.75*
2 I don't enjoy lecturing to large groups of students 4 1.49
66 I don't teach as it is a waste of time 1 0.50
7 I teach because I have been inspired to teach by my mentors 0 0.05
46 I teach because of the intellectual stimulation -2 -0.71
10 I teach because I believe it is an appropriate service to my profession -3 -1.33*
NB1. P < 0.05 (* Indicates significance at P < 0.01)
Table 4 Distinguishing statements for factor 3 (n = 2)
No. Statement Rank Score
42 I don't teach because I find it unenjoyable 5 1.93
66 I don't teach as it is a waste of time 3 1.48
10 I teach because I believe it is an appropriate service to my profession 0 -0.23*
46 I teach because of the intellectual stimulation -4 -1.70
NB1. P < 0.05 (* Indicates significance at P < 0.01)
Table 5 Distinguishing statements for factor 4 (n = 2)
No. Statement Rank Score
36 My clinical load deters me from teaching 5 2.61*
23 Opportunities for academic promotion have nothing to do with my motivation 5 2.30
40 I teach as a means of reviewing a topic area unfamiliar to me 4 1.66*
9 I teach because it is a requirement of my employment contract 3 1.44
22 I don't teach because there is no recognition for what I do 1 0.18
29 I don't teach because I believe the institution devalues teaching and learning 0 -0.03*
64 I don't teach as I don't approve of new teaching techniques 0 -0.21
45 I don't teach because interacting with students is boring 0 -0.34*
69 I don't teach as I feel my knowledge is out of date 0 -0.37*
66 I don't teach as it is a waste of time -1 -0.58*
56 I don't teach just because it is expected of me -4 -1.54
18 I teach because there are opportunities for 'virtual' and/or 'online' and/or remote teaching -5 -2.30
NB1. P < 0.05 (* Indicates significance at P < 0.01)
• helping students become good doctors
• enjoying the challenge of effective teaching
• valuing the presentation of one's own specialty
• enjoyment of small group teaching
• inspiration from mentors and past teachers
• liking to be challenged in one's views
• feeling responsible for students
• wanting to understand students.
The descending rank reflects less important views. These keen clinical teachers did not agree that they taught to "perform in front of an audience" and disagreed that teaching was boring, unsupported or a waste of time.
Factors 2, 3 and 4, the "I struggle to teach because ......" factor, represented 7 participant views in total, agreeing with a number of negative statements about teaching. Highly ranked negative items included:
• lack of involvement in course design
• lack of enjoyment in teaching
• clinical load deterring involvement in teaching.
It appears these 7 sorts are from those not motivated to teach. The software used prevented any further analysis of participants' demographic details and Q sorts.
Discussion
This study shows that most senior medical clinicians, of diverse discipline, are motivated to teach medical students and that the main reason appears altruistic – a desire to help students become good doctors. A small, but not insignificant group, of senior clinicians do not want to teach, citing lack of involvement in the design in the course and excessive clinical load as negative motivators. Given the participants were taken from supposed active teaching lists this imbalance is not surprising.
Our results show that the majority of senior clinicians motivated to teach (factor 1) dwelt on a number of common items: inspiration from senior mentors, the altruistic role in development of junior doctors and the opportunity to highlight a specialty area. These items draw from concepts of goal directed or purposeful drivers of behaviour. Fulkerson highlighted some of these factors in a 1997 study [5]. These authors also found that the issue of rewards was important in motivating medical graduates to teach. Questions related to the value of rewards were included (such as payment, contracts and peer pressure) but these did not appear to be a prime motivator in our study group.
Factors 2 to 4 reflected small numbers of clinicians (seven in total) apparently disinterested in teaching. They cited issues such as disengagement with course material, lack of pleasure in teaching and excessive clinical load as impediments to teaching. While small in number, their views are worth examination as the dissenting opinion. These items are reinforcers (albeit negative reinforcers) of teaching behaviours. Schormair et al, 1992 [6] also drew attention to the lack of co-ordination of courses and the heavy burden of patient care and administrative tasks underlying the lack of motivation of some medical teachers to teach. In contrast to those who valued teaching, perhaps powerful mentors did not inspire this group during their own training.
The limitations of our study include the qualitative nature of the observations, the lack of breadth of sampling (focussing on a single hospital's clinical school affiliated staff and to some degree "reaching the converted"), the lack of inter-rater reliability assessments and the forced normalisation of the questions (perhaps unrestrained, participants may have skewed all statements to one pole or the other). Each member of the research team may have introduced the study differently (although there were common instructions). Although deidentified, the fact that each researcher recorded the views of their participants may have biased away from socially undesirable responses. The statements also cannot be considered exhaustive of all the possible reasons that may motivate a clinician to teach. A computerised questionnaire may also have been more efficient and interactive. It also must be remembered that motivation is an inherently complex construct and while Q methodology capture subjective views, it may not net all aspects of motivation.
There is also the issue of social desirability bias. Participants were aware that the evaluator was a colleague and would be aware of their results, perhaps skewing answers to those apparently more favourable. This may have contributed to an artificial multi-modal distribution. The questionnaire itself may also have contributed to an information bias. For example, in the heading it was requested to either strongly agree or strongly disagree rather than to give a more open level of agreement assessment. The questions asking why " I don't teach" to a group who are all allegedly teachers may lead to confusion, and again an observation bias or a classification bias. The incorrect placement of double negatives and even triple negatives may have led to noise, obscuring real differences and opinion.
Implications
Clinical teachers are a valuable resource. They are essential to the successful teaching of medical students. Our study identified matters within the "ownership" of universities: engagement of clinical teachers in course design. This is an area ripe for action from medical schools and these data suggest that real inclusion would improve clinical teacher motivation. Rather more metaphysically, encouraging teachers to dwell on the inspirational models of their mentors may also enhance recruitment and retention. Opportunities to highlight special interests and to teach effectively would also be sensible strategies. Allowing clinical teachers to engage with small groups of students and to develop some understanding of student needs would meet the teachers' motivation to demonstrate their understanding of student experience. This sample suggests that contracts, money, a sense of duty and peer pressure play little part in motivating teachers. Previous research however, makes clear the important place of modest rewards such as Dean's teaching prizes. These data also underline the negative effect of service burden amongst clinical teachers. While this is not immediately under medical school control, universities can contribute meaningfully to discussions on balance of service and teaching commitments amongst health staff.
Future research
This method could be replicated amongst more varied clinical teachers such as community practitioners and non-medical teachers to assess consistency of motivation. Further work is required to study the impact on clinical teacher workforce recruitment and retention through meeting these expressed motivations. We plan an intervention study to understand the effectiveness of tailoring teaching experience to these identified motivators. An understanding of intrinsic motivation is only helpful if it leads to higher teacher satisfaction, better quality teaching and retention in the workforce.
Conclusion
Our inquiry suggests that promotion of teaching to senior clinicians is likely to have increased success if prospective teachers contribute to course development, sufficient time is allocated to teaching, memories of inspirational teachers are reawakened, the link between strong teaching and junior doctor outcomes is emphasised, and staff are reminded of the opportunity to 'advertise' their specialty. Medical schools face increasing difficulties staffing clinical programs and this study provides avenues to improve recruitment of senior medical staff to the teaching ranks.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
All authors contributed to design, data collection and paper preparation. DM and CO also conducted the data analysis.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We wish to acknowledge the senior clinicians who participated in this study, Angela Brew and Kim McShane (Institute for Teaching and Learning, University of Sydney), and colleagues in the Graduate Certificate of Education (Higher Studies) 2001–2 course who provided useful comments on the manuscript.
==== Refs
Bindra D Stewart J Motivation 1966 Harmondsworth Middlesex England, Penguin Books Ltd
Maslow AH Toward a Psychology of Being 1968 New York, Van Nostrand Reinhold Co
Covington MV Goal Theory, Motivation, and School Achievement: An Integrative Review Annual Review of Psychology 2000 51 171 200 10751969 10.1146/annurev.psych.51.1.171
Ellis TI Motivating teachers for excellence. Eric Digests 1984
Fulkerson PK Wang-Cheng R Community-based faculty: Motivation and rewards Family Medicine 1979 29 105 107 9048169
Schormair C Ten statements on the motivation of medical teachers to teach Medical Teacher 1992 14 283 286 1293453
Wright SM Beasley BW Motivating factors for academic physicians within departments of medicine Mayo Clinic Proceedings 2004 79 1145 1150 15357036
Kumar A Loomba D Rahangdale RY Kallen DJ Rewards and Incentives for Nonsalaried Clinical Faculty Who Teach Medical Students Journal of General Internal Medicine 1999 14 370 372 10354257 10.1046/j.1525-1497.1999.00341.x
Hoban JD Cariaga-Lo L Bennett BA Ernest JM Vanderweide SA Harrington ME Incentives for Teaching Academic Medicine 1996 71 106 107 8615913
Daniel LG Ferrel CM Clarifying reasons why people aspire to teach: An application of Q-Methodology
Chinnis AS Summers DE Doerr C Paulson DJ Davis SM Q Methodology: A New Way of Assessing Employee Satisfaction Journal of Nursing Administration 2001 31 252 259 11388161 10.1097/00005110-200105000-00005
Müller FH Kals E Die Q-methode: Ein innovatives verfahren zur erhebung subjektiver einstellungen und meinungen Forum Qualitative Sozialforschung,
Stephenson W The Study of Behaviour: Q Technique and its methodology 1953 Chicago, University of Chicago Press
Brown SR Q Methodology and Quanlitative Research Qualitative Health Research 1996 16
McKeown B Thomas D Q Methodology 1988 Newbeury Park, Sage Publications
Shomaker TS Ricks DJ Hale DC A Prospective, Randomized Controlled Study of Computer-assisted Learning in Parasitology Academic Medicine 2002 77 446 449 12010707
Anastasi A Psychological Testing 1968 3rd Edn Toronto Ontario, Macmillan Publishing Co. Inc
|
16022738
|
PMC1185542
|
CC BY
|
2021-01-04 16:30:56
|
no
|
BMC Med Educ. 2005 Jul 18; 5:27
|
utf-8
|
BMC Med Educ
| 2,005 |
10.1186/1472-6920-5-27
|
oa_comm
|
==== Front
BMC Med EthicsBMC Medical Ethics1472-6939BioMed Central London 1472-6939-6-61599240110.1186/1472-6939-6-6Research ArticleDeath, dying and informatics: misrepresenting religion on MedLine Rodríguez del Pozo Pablo [email protected] Joseph J [email protected] Division of Medical Ethics, Weill Medical College of Cornell University in Qatar (WCMC-Q). P.O. Box 24144, Education City, Doha, Qatar2 Division of Medical Ethics, Weill Medical College of Cornell University (WMCCU), New York, USA2005 1 7 2005 6 6 6 16 3 2005 1 7 2005 Copyright © 2005 del Pozo and Fins; licensee BioMed Central Ltd.2005del Pozo and Fins; 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 globalization of medical science carries for doctors worldwide a correlative duty to deepen their understanding of patients' cultural contexts and religious backgrounds, in order to satisfy each as a unique individual. To become better informed, practitioners may turn to MedLine, but it is unclear whether the information found there is an accurate representation of culture and religion. To test MedLine's representation of this field, we chose the topic of death and dying in the three major monotheistic religions.
Methods
We searched MedLine using PubMed in order to retrieve and thematically analyze full-length scholarly journal papers or case reports dealing with religious traditions and end-of-life care. Our search consisted of a string of words that included the most common denominations of the three religions, the standard heading terms used by the National Reference Center for Bioethics Literature (NRCBL), and the Medical Subject Headings (MeSH) used by the National Library of Medicine. Eligible articles were limited to English-language papers with an abstract.
Results
We found that while a bibliographic search in MedLine on this topic produced instant results and some valuable literature, the aggregate reflected a selection bias. American writers were over-represented given the global prevalence of these religious traditions. Denominationally affiliated authors predominated in representing the Christian traditions. The Islamic tradition was under-represented.
Conclusion
MedLine's capability to identify the most current, reliable and accurate information about purely scientific topics should not be assumed to be the same case when considering the interface of religion, culture and end-of-life care.
==== Body
Background
With the globalization of medical science, there is a concomitant need to better understand cross-cultural differences [1]. This ethical imperative is perhaps most critical during life's final chapter when diverse populations invoke their religious traditions and doctrines to questions of death and dying [2,3]. To meet the needs of dying patients and their families, practitioners may turn to the convenience of databases like MedLine to provide more culturally-competent care. Although MedLine has become the gold standard for scientific papers, it is less clear that it is a comprehensive source for scholarship in the broader medical humanities [4]. In this paper we sought to determine how accurate MedLine is as a source of information about how religious beliefs inform end-of-life care[5,6].
To answer these questions we have queried MedLine to assess the scope and quality of information that those conducting a search might obtain using common search phrases that working practitioners might employ. Our objectives are twofold. First, to parse out MedLine's representation of reality from wider scholarly treatment concerned with the interface of religion, medicine and death and dying. Second, to stress that MedLine's reliability is related not only to what is indexed in it, but also to how it is accessed. Our topic is more cultural than scientific: death and dying in the three major monotheistic religions. This type of issue is particularly pertinent to healthcare providers treating a heterogeneous population, since culture and religion play a central role in shaping everything from patients' notions of health and disease causation, [7-10] to their reaction to pain, to their expectations of the doctor-patient relationship [11-14].
Compiling the literature on religion and medicine raises questions about the bibliographic strengths of informatic catalogues in general [15] and MedLine in particular when we seek less scientifically informed knowledge. It also challenges us to consider how user-dependent MedLine is when the search is less than straightforward and what the implications are of incomplete or decontextualized searches. Finally, it generates doubts about how theological issues are catalogued in a database designed primarily for scientific literature.
Methods
We searched MedLine using PubMed in order to retrieve and thematically analyze full-length scholarly journal papers or case reports dealing with religious traditions and end-of-life care. Eligible papers were limited to English-language articles that included an abstract, to ascertain the articles' content prior retrieval. Editorials, book reviews, letters-to-the-editor and brief comments were excluded, because generally they are not peer-reviewed.
To assess the relative magnitude of the broader literature, a search without language or genre limitations was also run. The search was limited to articles published between January 1993 and June 30, 2004, with the actual search completed on July 2, 2004.
We defined our search as a string of words, connected by the appropriate syntax and Boolean operators [16]. The search would consist of two parts. The first contained the names of each of the three religious traditions, according to the Medical Subject Headings (MeSH) [17], the National Library of Medicine's thesaurus of terms used to index all articles in MedLine, as noted:
• christianity OR christian
• islam OR islamic OR muslims
• judaism OR jewish
The second part included the sub-strings "death and dying" and "terminal care". The former is the heading term used by the National Reference Center for Bioethics Literature (NRCBL) [18] as the higher-level descriptor for articles dealing with an ample range of issues related to death and end-of-life care. "Terminal care" is the higher-level term MeSH uses to classify those entries that deal with death and dying. PubMed automatically recognizes MeSH terms, and retrieves all articles classified under such terms. To summarize, then, the search strings were:
• (islam OR islamic OR muslims) AND ("death and dying" OR "terminal care")
• (christianity OR christian) AND ("death and dying" OR "terminal care")
• (judaism OR jewish) AND ("death and dying" OR "terminal care")
Eligible articles were assessed for publication characteristics including site of publication and country of origin as well as authorship and any institutional or denominational affiliation. Professional backgrounds of authors were also recorded from the papers' institutional websites or their home pages.
Articles that were the product of a logical – but nonsensical – Boolean hit, such as the retrieval of articles by authors with a surname of "Christian" or "Islam" or writers who worked at a "Jewish Hospital", were removed from consideration if they had nothing to do with death and dying. All other articles retrieved were studied in order to avoid the introduction of a selection bias.
The initial general searches for religion, death and dying and denomination are noted in Table 1. Narrowing the search as described above, we obtained the results condensed in Table 2.
Table 1 Articles retrieved by non-specified searches
String English, with abstracts All languages, with abstracts
Religion 6212 6894
"death and dying" OR "terminal care" 2704 3100
christianity OR christian 7492 8494
islam OR islamic OR muslims 1031 1160
judaism OR jewish 7144 7306
Table 2 Articles retrieved by the specific searches
String English, with abstracts All languages, with abstracts %
(christianity OR christian) AND ("death and dying" OR "terminal care") 46 50 61.3
(islam OR islamic OR muslims) AND ("death and dying" OR "terminal care") 9 9 12
(judaism OR jewish) AND ("death and dying" OR "terminal care") 20 23 26.3
Total 75 82 99.6
Results
Main findings
The search retrieved a total of 75 references. After eliminating duplicative results (four) and nonsensical retrievals (five), 66 articles were available for analysis. These articles were published in 36 different journals from seven countries or regions, but U.S.-edited journals accounted for the majority, with 27 publications, followed by the United Kingdom with three. Other English-speaking and European countries were represented with one or two journals at most. There was one Israeli journal and none from the Arab League or other predominantly Muslim country.
The majority of articles were in non-denominational journals, although no major medical journal such as The New England Journal of Medicine, The British Medical Journal, The Journal of the American Medical Association, The Lancet or Annals of Internal Medicine, were represented in our sample. Only two were from denomination affiliated journals: Health Progress, which is the official organ of The Catholic Health Association of the United States, and Christian Bioethics, an interdenominational, non-ecumenical publication [19]. These two denominational journals concentrated almost a third of all articles retrieved (or 21 articles). Twelve articles reviewed Christian traditions, with a high concentration in the Catholic doctrine (seven articles). Eight articles dealt with Jewish teachings, while four discussed Islamic beliefs. No articles for the Islamic tradition were written by clergy. Nearly all of the articles (61 total) had identified authors. The remaining five did not have an identified author and formed part of a series of doctrinal articles published in Health Progress. More than a third of articles were authored by physicians, with another 16% written by nurses, thus indicating that health professionals accounted for almost two-thirds of retrieved articles. The remaining articles were written by authors working in ethics, the humanities, social work, theology or the law. Nine authors were identified as clergy. Eight were noted as Christian clerics (six Catholics and two Orthodox), including one nun. One author was noted to be a rabbi. There were no Islamic clerics identified amongst the writers.
Most authors (74%) were affiliated with hospitals or universities, and more than half (56%) were based in the U.S. The rest were evenly distributed among Europe, Canada and other English-speaking countries. Only one author was based in a Muslim country (Pakistan). Most authors (25 total) belonged to non-denominational institutions, while 17 were affiliated with Christian institutions, predominantly Catholic (11 total), evenly distributed among Catholic universities and hospitals. Another 12 authors belonged to Jewish institutions, eight from Jewish hospitals or hospices and another four from universities.
With this data we then tried to assess what practices would be prohibited, obligatory and permitted according to the three monotheistic traditions as represented in the retrieved literature. Nevertheless, our efforts led to identifying papers that failed to provide more than broad categoricals within the traditions. This obscures the rich debate that occurs in theological and scholarly circles over diverse questions such as pain management and the potential hastening of death; whether suffering is redemptive; artificial nutrition and hydration; medical futility; and the role of quality-of-life considerations, to name but a few contentious topics that involve a broader scholarly community than those whose work is cited on MedLine.
Limitations
Our inquiry is limited by the search methods we employed as well as the MedLine database queried. However, our objective was, precisely, to identify such limitations in order to demonstrate the need for caution when excessively relying on database searches for topics which transcend the purely scientific.
Having noted this caveat, our study was limited by circumscribing eligible articles to those with abstracts, but this was necessary to identify articles that merited additional analysis. Our findings are only applicable to articles in English, the predominant language of MedLine, and cannot be generalized to other religious traditions or different thematic areas.
Finally, to avoid the introduction of our own biases about what would constitute an appropriate article for study, we included articles for analysis that would otherwise only marginally shed light on our designated areas of inquiry. This illustrates the limitations of relying purely on retrieved articles without an additional level of scholarly discrimination.
Discussion
While MedLine is an invaluable source of information, our study indicates the potential limits of the scholarly convenience afforded by informatics. Our data suggest that users of MedLine will gain only a partial view of the range of scholarship related to death and dying and the three major monotheistic traditions. Although a bibliographic search in MedLine on death and dying in the teachings of the three major monotheistic traditions will produce instant results and some valuable literature, the price of such convenience may be a somewhat biased and partial view of the issue.
Although religion is present as a topic among the MedLine indexed literature, articles retrieved under "ethics" are three times more numerous than "religion". Despite the relatively large number of citations concerning religion, our study indicates that there was a general imbalance between the number of articles and followers of each of the traditions [20,21]. This discrepancy is even more obvious when MedLine is compared to the widely popular Google search engine. In contrast to the predominance of Christian and Jewish hits on MedLine, the same searches on Google showed more balanced numbers with respect to religion and death and dying: Christianity (38,400); Islam (11,200); and Judaism (19,500). The preponderance of Christian and Jewish articles on MedLine is multifactorial, and is a question that warrants additional study to understand the determinants which produce and disseminate scholarship in the art and science of medicine across differing religious and cultural traditions.
Whatever the tradition, the vast majority of articles came from journals based in the U.S. A few articles came from developed European countries, with one from Israel and another from South Africa. It is notable that, despite the presence of 400 medical journals in the WHO-designated Eastern Mediterranean region [22], which includes countries from Morocco to Pakistan, none of the articles we retrieved was from a journal based in a Muslim country. Authors' residences followed almost the same pattern.
This preponderance of U.S. journals and authors can alter the scholarly landscape and distort the application of religious teachings when articles are read outside of the North American context. Because these journals have global reach and influence, most articles they carry are firmly rooted in the American context, and cannot necessarily be extrapolated to the local environment. That may limit the literature's usefulness to those practicing in other parts of the world. And when practitioners respect uncritically these texts the way they do the scientific papers found in MedLine, this may lead to a misrepresentation of local cultures and traditions.
For example, in predominantly Catholic Latin America, the question of physician-assisted suicide or euthanasia – frequently addressed in these articles – may be little more than an intellectual curiosity, the practice of which is nearly inconceivable in a Latin context. No North American Catholic author wrote about death and dying in the context of poverty, which gives rise to the most difficult ethical quandaries from the Rio Grande to Antarctica [23]. Although questions about withholding and withdrawing advanced life-sustaining technologies is a frequent theme, it is a topic that is less than relevant when so many patients in the region do not have access to basic care. Questions about the sanctity of life in this context intersect questions of social justice and the relationship of poverty to coping with death, dying and suffering [24,25].
The fact that most articles were written by doctors and nurses may also contribute to a rather narrow focus on clinical practices at the end of life, ignoring broader public health and societal context factors that may influence the care provided to dying patients and their families. Papers featuring practical guidelines for the care of the dying Catholic [26,27] or Jewish patient [28] or the importance of family to the Muslim patient [29] can enhance doctors' cultural competence [30]. Nonetheless, these contextual pieces, so firmly situated in the clinic, do not do justice to broader questions about the healthcare system or the socio-economic organization. Nor do they explain the underlying theology. In our view, the latter limitation was not remediated by religious people writing many of the doctrinal articles. These articles were more theoretical than practical and often rigidly orthodox, unable to mollify religious scripture to accommodate the quotidian needs of patients and families at the end of life [31]. It is important to note that many of the leading pastoral care journals, which take this perspective, yielded no articles in our MedLine search, even though many are indexed.
Conclusion
Although MedLine is an excellent source of the more objective reality of science, our observations indicate that MedLine is an incomplete source of information for the complex interplay of death, dying and religion.
These errors of omission reflect the orientation of the biomedical paradigm MedLine was meant to serve. Unfortunately, it has been unable to transcend its scientific origins.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Both authors contributed equally to this work.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Dr. Fins is a faculty scholar of the project on Death in America of the Open Society Institute.
==== Refs
Fadiman A The Spirit Catches You and You Fall Down 1997 New York, Farrar, Strauss and Giroux
Spiro HM (Editor) Facing Death (Where Culture, Religion, and Medicine Meet) 1996 New Haven, Yale University Press
Bigby J Judyann Bigby Beyond culture: Strategies for caring for patients from diverse racial, ethnic, and cultural groups Cross-Cultural Medicine 2003 Philadelphia, American College of Physicians 1 28
McLellan F 1966 and all that-when is a literature search done? Lancet 358 646 2001 Aug 25 11530164 10.1016/S0140-6736(01)05826-3
Singer PA Miles SH Siegler M Computer searches of the medical ethics literature J Clin Ethics 1990 1 195 8 Fall 2132010
Smith LG Schwartz JD Disproportionate use of MEDLINE searches by housestaff Acad Med 1997 72 160 1 9075416
Geertz C Local knowledge: further essays in interpretive anthropology 1983 New York, Basic Books
Nehra A Kulaksizoglu H Global perspectives and controversies in the epidemiology of male erectile dysfunction Curr Opin Urol 2002 12 493 6 12409879 10.1097/00042307-200211000-00009
Johnson JL Bottorff JL Balneaves LG Grewal S Bhagat R Hilton BA Clarke H South Asian womens' views on the causes of breast cancer: images and explanations Patient Educ Couns 1999 37 243 54 14528550 10.1016/S0738-3991(98)00118-9
Nielsen M Hoogvorst A Konradsen F Mudasser M van der Hoek W Causes of childhood diarrhea as perceived by mothers in the Punjab, Pakistan Southeast Asian J Trop Med Public Health 2003 34 343 51 12971560
Cassisi JE Umeda M Deisinger JA Sheffer C Lofland KR Jackson C Patterns of pain descriptor usage in African Americans and European Americans with chronic pain Cultur Divers Ethnic Minor Psychol 2004 10 81 9 14992632 10.1037/1099-9809.10.1.81
Núñez AE Robertson C Multicultural considerations in women's health Med Clin North Am 2003 87 939 54 14621325 10.1016/S0025-7125(03)00064-6
Silverman M Musa D Kirsch B Siminoff LA Self care for chronic illness: older African Americans and whites J Cross Cult Gerontol 1999 14 169 89 14617891 10.1023/A:1006676416452
Fukuhara S Lopes AA Bragg-Gresham JL Kurokawa K Mapes DL Akizawa T Bommer J Canaud BJ Port FK Held PJ Worldwide Dialysis Outcomes and Practice Patterns Study. Health-related quality of life among dialysis patients on three continents: the Dialysis Outcomes and Practice Patterns Study Kidney Int 2003 64 1903 10 14531826 10.1046/j.1523-1755.2003.00289.x
Baker N Annals Of Scholarship: Discards The New Yorker 64 1994; 4 Apr
PubMed Help
Medical Subjects Headings
National Reference Center for Bioethics Literature Classification Scheme
Christian Bioethics, Non Ecumenical Studies in Medical Morality
American Religious Identification Survey 2001
Barrett DB Kurian GT Johnson TM World Christian Encyclopedia: A Comparative Survey of Churches and Religions in the Modern World 2001 New York, Oxford University Press
First association of medical editors for Eastern Mediterranean Region Bull World Health Organ 2003 81 922 922
de Ferranti D Perry GE Ferreira F Walton M Inequality in Latin America: Breaking with History? 2004 Washington DC, World Bank
Simón P Couceiro A Urraca S Decisiones éticas conflictivas en torno al final de la vida: una introducción y un marco de análisis Eutanasia hoy Un debate abierto 1996 Madrid: Noesis 313 354
Rodríguez del Pozo P Etica y distribución de recursos escasos Revista Chilena de Medicina Intensiva 2001 16 226 233
Kalua PM Tan SY Bacon JG Better care for the dying. Hawaii healthcare system develops a manual for end-of-life care Health Prog 1999 80 58 61 10351504
Taylor C Ministering to persons who face death. Practical guidance for care givers of persons making end-of-life treatment decisions Health Prog 1994 75 58 62 10133753
Bonura D Fender M Roesler M Pacquiao DF Culturally congruent end-of-life care for Jewish patients and their families J Transcult Nurs 2001 12 211 20 11989036
Moazam F Families, patients, and physicians in medical decisionmaking: a Pakistani perspective Hastings Cent Rep 2000 30 28 37 11475993
Betancourt JR Cultural competence – marginal or mainstream movement? N Engl J Med 351 953 5 2004 Sep 2 15342800 10.1056/NEJMp048033
Brenner D Blanchard T Fins JJ Hirschfield B Embracing Life and Facing Death: A Jewish Guide to Palliative Care 2002 New York, CLAL- The National Jewish Center for Learning and Leadership
|
15992401
|
PMC1185543
|
CC BY
|
2021-01-04 16:31:58
|
no
|
BMC Med Ethics. 2005 Jul 1; 6:6
|
utf-8
|
BMC Med Ethics
| 2,005 |
10.1186/1472-6939-6-6
|
oa_comm
|
==== Front
BMC MicrobiolBMC Microbiology1471-2180BioMed Central London 1471-2180-5-391598242210.1186/1471-2180-5-39Research ArticleModeling Lactococcus lactis using a genome-scale flux model Oliveira Ana Paula [email protected] Jens [email protected]örster Jochen [email protected] Fluxome Sciences A/S, Søltofts Plads, Building 223, DK-2800 Kgs. Lyngby, Denmark2 Center for Microbial Biotechnology, BioCentrum-DTU, Technical University of Denmark, Building 223, DK-2800 Kgs. Lyngby, Denmark2005 27 6 2005 5 39 39 1 4 2005 27 6 2005 Copyright © 2005 Oliveira 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
Genome-scale flux models are useful tools to represent and analyze microbial metabolism. In this work we reconstructed the metabolic network of the lactic acid bacteria Lactococcus lactis and developed a genome-scale flux model able to simulate and analyze network capabilities and whole-cell function under aerobic and anaerobic continuous cultures. Flux balance analysis (FBA) and minimization of metabolic adjustment (MOMA) were used as modeling frameworks.
Results
The metabolic network was reconstructed using the annotated genome sequence from L. lactis ssp. lactis IL1403 together with physiological and biochemical information. The established network comprised a total of 621 reactions and 509 metabolites, representing the overall metabolism of L. lactis. Experimental data reported in the literature was used to fit the model to phenotypic observations. Regulatory constraints had to be included to simulate certain metabolic features, such as the shift from homo to heterolactic fermentation. A minimal medium for in silico growth was identified, indicating the requirement of four amino acids in addition to a sugar. Remarkably, de novo biosynthesis of four other amino acids was observed even when all amino acids were supplied, which is in good agreement with experimental observations. Additionally, enhanced metabolic engineering strategies for improved diacetyl producing strains were designed.
Conclusion
The L. lactis metabolic network can now be used for a better understanding of lactococcal metabolic capabilities and potential, for the design of enhanced metabolic engineering strategies and for integration with other types of 'omic' data, to assist in finding new information on cellular organization and function.
==== Body
Background
Lactic acid bacteria (LAB) are a heterogeneous group of microorganisms able to convert carbohydrates into lactic acid. They are applied worldwide in the industrial manufacture of fermented food products, mainly in the dairy industry. During fermentation LAB primarily produce lactic acid from the available carbon source, resulting in the rapid acidification of the food raw material, which is an important parameter in the preservation and extension of shelf life of food products. LAB metabolism also contributes for the development of desired product properties such as flavor and texture [1,2]. Because of their long tradition of safe use (GRAS microorganism), their capacity to grow rapidly on lactose-based media derived from milk and their potential to generate a variety of metabolic products, LAB also have the potential to be used as cell-factories in bioreactors for the in situ production of compounds that contribute to the flavor, texture or health benefits of foods [1].
Among LAB, Lactococcus lactis is, by far, the most extensively studied with respect to its physiology, metabolic pathways and regulatory mechanisms, and its genome was the first LAB genome to be completely sequenced and annotated [3]. Lactococci are nutritionally fastidious organisms with a very limited biosynthetic capacity. Anabolic precursors are primarily imported from the growth media, whereas only a minor fraction is synthesized de novo from a carbon source. The major part of the carbon from carbohydrates is converted into fermentation end-products. For example, during growth on glucose, only about 5% of the metabolized sugar is converted into biomass [4]. The very limited biosynthetic capacity of L. lactis implies that, for optimal growth, they require exogenous supply of a fermentable sugar, numerous vitamins and amino acids, phosphate, potassium and magnesium [5].
L. lactis is a facultative anaerobic bacterium. Some strains are capable of growing in the presence of oxygen and adjust their metabolism accordingly, while others are strongly inhibited under aerobic conditions. As this microorganism lacks a functional respiratory chain, the ability to grow aerobically has been linked to the presence of NADH-oxidases. Since L. lactis is not able to perform respiration, ATP is only formed through substrate level phosphorylation. Thus, in order for Lactococci to grow at a high specific growth rate, a high carbohydrate degradation rate (and, concomitantly, an efficient sugar transport system) is called for. The main function of the sugar metabolism in L. lactis is to generate the energy necessary for rapid growth and for maintenance of intracellular pH during acidification of the environment [6].
Due to its major importance as a laboratorial and industrial microorganism, and because of its relatively simple metabolism and limited biosynthetic capabilities, L. lactis has been an organism of choice for many metabolic engineering purposes [1,6,7]. Hence, the design of enhanced metabolic engineering strategies calls for models where cellular behavior can be predicted.
The reconstruction of the entire metabolic reaction network of a cell and subsequent application of genome-scale flux models has been conducted for many organisms, including bacteria, yeast, fungi and animal cells [8-12]. These models have the potential to become common modeling tools. One approach that has been used to explore the capabilities of these large metabolic networks is flux balance analysis (FBA). This is simply a linear programming posed problem in which the constraints are defined by the stoichiometry of enzymatic reactions and transport steps. A solution to the problem, i.e., a set of fluxes through all the defined reactions, can be found through specifying an objective function, which is often defined as the optimization of a certain flux of interest, e.g., the flux towards formation of biomass. Maximization of biomass production has been shown to allow description of overall metabolic behavior in a number of cases, probably because most cells have evolved, under laboratory conditions, towards the maximization of their growth performance [13]. By using appropriate constraints and a meaningful objective function, FBA has been successfully used in exploring the relationship between genotype and phenotype and in the prediction of product yields and growth rates under changing environmental and genetic conditions, at steady state [14-16]. More recently, another approach has been proposed for dealing with the effect of gene deletions in the prediction of flux distributions, based on quadratic programming [17]. This minimization of metabolic adjustment (MOMA) approach relies on the assumption that optimal growth may initially not be true for mutants generated artificially in the laboratory as usually those mutants have not yet undergone evolution towards optimal growth.
In this work the reconstruction process of the metabolic network of L. lactis is presented. Network reactions were collected using the annotated genome sequence, biochemical and metabolic pathways databases, biochemistry books and journal publications. Once the network was established, FBA and MOMA were applied to analyze the network capabilities and to model phenotypic behavior under anaerobic and aerobic conditions. Simulation results were compared with experimental data available in the literature. Furthermore, the model was used to identify possible metabolic engineering targets to design an efficient diacetyl producing strain.
Results and discussion
Characteristics of the reconstructed network
The reconstruction process resulted in a network comprising 621 reactions and 509 metabolites. The entire reaction database can be consulted in the Additional file 1 and is also available at . A total of 358 ORFs out of the detected 2310 ORFs in the sequenced genome were considered, corresponding to 476 associated reactions. The remaining 145 reactions were included based on biochemical/physiological considerations or inferred by the demands imposed on the metabolic network. Namely, reactions were included based on experimental information reported in literature to account for pathway gaps, transport steps and biomass assembly. From the entire set of reactions, 492 correspond to intracellular reactions while 129 are transport fluxes over the cytoplasmic membrane. From the 358 ORFs considered, 63 ORFs account for genes that have been isolated and identified from L. lactis, 286 ORFs have high homology with identified genes from other organisms, 2 ORFs are annotated sequences derived from low level homologues and 7 ORFs are described as probable homologues of unknown function (ORF reliability classification based on [18]). The reaction database includes a total of 509 network metabolites, from which 414 are intracellular and 95 are external metabolites. 87 out of these 95 are metabolites secreted or taken up by the cell without undergoing phosphorylation, and so it can be pointed out that the reconstructed network accounts for 422 unique metabolites. Table 1 summarizes the main characteristics of the reconstructed metabolic network of L. lactis.
Network reactions
From the 358 ORFs considered in the reconstruction process, ORFs that are assigned to energy metabolism, amino acid biosynthesis and nucleotides metabolism account for more than half of the total number. These ORFs are involved in 476 reactions. The relatively high number of associated reactions is mainly due to the existence of low substrate-specific enzyme activities catalyzing different reactions. For instances, the model includes a low-specific aminotransferase (araT) involved in the metabolism of several amino acids and defined to catalyze 18 reactions in the model. An equation on biomass formation was developed to account for the drain of precursors and building blocks into biomass. A detailed description of its assembly can be found in the Additional file 2. This equation also accounts for energy requirements associated with growth, which have been estimated through model fitting against experimental observations as being 18.15 mmol ATP gDW-1. One reaction was included to account for non-growth dependent ATP maintenance. For growth under glucose limitation this value has been previously experimentally determined and equals 1 mmol ATP gDW-1h-1 [19].
Network metabolites
The metabolic network of L. lactis contains 509 metabolites, 422 of which are unique metabolites. It is through metabolites that reactions are connected, as the product of one reaction becomes the substrate of another. At a zoom-out level, different biochemical pathways within a cell are interconnected by virtue of metabolites that participate in more than one pathway [20]. In particular, cofactors like ATP, NADH and NADPH play an important role in connecting the many different pathways. The most frequent metabolic intermediates in the reconstructed network are presented in Table 2. A total of 28 intracellular metabolites were not connected into the overall metabolic network. These non-connected metabolites take part in 25 "non-connected" reactions, catalyzed by 21 "non-connected" gene products (see Additional file 3)
Growth requirements and minimal media
L. lactis is able to metabolize a variety of different sugars and other carbon sources [2,21] to obtain energy, redox power and precursor metabolites for macromolecular biosynthesis. Fermentable carbon sources utilized by L. lactis include fructose, galactose, glucosamine, glucose, lactose, maltose, mannitol, mannose, ribose, sucrose and trehalose.
The capabilities of the in silico model to utilize different carbon sources to grow were evaluated using FBA, using comparable uptake rates and carbon-limited conditions. All the above mentioned carbon sources were supplied, one at a time, together with all amino acids and nucleotides. A sugar consumption rate of 13.6 mmol gDW-1 h-1 was considered (6.8 mmol gDW-1 h-1 in the case of disaccharides), both under aerobic and anaerobic conditions.
The specific growth rate under anaerobic conditions was always higher than at aerobic conditions when the specific oxygen consumption was set to 3.61 mmol gDW-1 h-1 [22], which is in good agreement with experimental evidence [23]. The predicted anaerobic specific growth rate on glucose was 0.79 h-1, while the aerobic specific growth rate was 0.62 h-1. In case oxygen uptake is not constrained (unlimited oxygen uptake is allowed) the specific growth rate was 0.82 h-1. Constrained aerobic growth is slightly lower than anaerobic growth due to the different flux constraints applied to the enzymes that metabolize pyruvate into acetyl-CoA (see Methods). Pyruvate-formate lyase is only active under anaerobic conditions, while the pyruvate dehydrogenase (PDH) complex is only active during aerobic growth [24]. This forces the cell to produce more NADH under aerobic conditions, since the PDH complex is a NADH producing step and is also an essential reaction for formation of the precursor metabolite acetyl-CoA. However, model results indicate that for low oxygen consumptions there is a limited capacity to recycle NADH through NADH-oxidase, therefore causing the cell to reduce the flux through the PDH complex, decreasing the amount of acetyl-CoA available for lipid metabolism and, consequently, for biomass formation.
Growth on mannose, galactose, sucrose, lactose and glucosamine was found to be the same as for growth on glucose. The capacity to grow on trehalose and maltose was slightly higher, due to a decreased ATP requirement for the synthesis of the corresponding phosphorylated sugars. Anaerobically, growth on fructose and mannitol led to a decrease of 3% and 16% in biomass formation, respectively, compared with growth on glucose. This decrease has been qualitatively observed experimentally [21,25,26]. In the case of fructose, the difference seems to be associated with a lower capacity to generate NADPH due to a lower flux through the pentose phosphate pathway. In the case of mannitol, the lower biomass formation rate seems to be related both with a decrease in the capacity of generating NADPH and with a NADH burden due to formation of an additional NADH molecule during the reaction catalyzed by mannitol-1-phosphate dehydrogenase. Probably as consequence of that burden, the simulated growth leads to the production of high amounts of ethanol and formate, which is in good agreement with experimental results reported by Neves et al. (2002) [26].
In addition to a carbohydrate source, the minimal amino acid requirements were determined by omission of each amino acid at a time. The single omission of arginine, methionine and valine was found to prevent in silico growth, even when all other amino acids are supplied and despite the presence of the biosynthetic pathways for these amino acids in the reconstructed network. In the case of valine this result is trivial, since simulations were run under the constraint that the reaction catalyzed by bcaT_1 can only occur in the catabolic direction, and there is no alternative pathway for valine synthesis. In the case of methionine, the reason appears to be the lack of available FAD (cofactor) to sustain growth. Finally, if no arginine is supplied, the linear problem becomes infeasible. Nevertheless, arginine synthesis is observed when maximizing for an arginine drain.
Growth was not observed when only these three amino acids were supplied. If glutamate (or, alternatively, glutamine) is allowed to be taken up in the presence of these three amino acids, growth occurs. Therefore, glucose, arginine, methionine, glutamate (or glutamine) and valine were found to be the minimal required medium for growth. L. lactis strains are usually auxotrophic for 7 to 9 amino acids, including these four [5,27]. Growth in the above defined minimal media is, however, 54% lower than in rich media. When all the amino acids reported as essential for L. lactis IL1403 [27] are allowed to be taken up by the model, growth rate increases to 87% of the value in rich media. This corresponds to the addition of asparagine, histidine, isoleucine and leucine to the above defined minimal media.
Single gene deletion analysis
Single gene deletion analyses were computed to predict the lethal effect of deleting each individual gene. Prediction capability was assessed by comparing simulation results with experimental data on gene lethality reported for L. lactis [28] and Bacillus subtilis [29], a related Gram-positive bacteria of low G+C content, during growth in rich media. 25 genes were found in literature as being lethal, out of the 34 predicted by the model. From these, one is a false lethal (ThyA), while no references were found for the remaining 8 (see Additional file 4). Regarding these analyses, it has to be point out that regulatory proteins and constrains were seldom included, so it is likely that we are underestimating the number of actual lethal genes.
The effect of the deletion of each reaction was also assessed. These computational studies were performed both in rich media and minimal media, and both under aerobic and anaerobic conditions. FBA and MOMA were used to compute these simulations, with MOMA predicting around 10 more lethal genes/reactions than FBA (Table 3). We focus the following analysis on MOMA results (all details can be found in the Additional file 4). The deletion of 24.6% of the genes showed to be lethal in minimal media, while in rich media this number decreases to 12.1% (percentage based on the number of ORFs included in the model, discounting the 21 "non-connected" gene products). The single deletion of reactions accounts for 23.1% lethal reactions in minimal media and 14.1% in rich media (considering 596 network reactions: all the 621 network reactions minus the 25 "non-connected" reactions). Additionally, it was observed that although the number of lethal reactions is always higher than the number of lethal genes, this is mainly due to the existence of reactions catalyzed by unknown gene products (enzymes without a corresponding annotated ORF).
Aerobic and anaerobic simulations led to similar results, with a few more genes/reactions being lethal under aerobic conditions, namely those associated to oxygen utilization, CO2 production and to the pyruvate dehydrogenase (PDH) complex. Under aerobic conditions, the PDH complex is the only pathway leading to the formation of acetyl-CoA from pyruvate, and as acetyl-CoA is an essential metabolite in many processes (such as lipid formation), deletion of the PDH complex results in a lethal phenotype.
When comparing the growth capabilities in both rich and minimal media, it can be observed that approximately 40 additional genes are lethal during growth in minimal media. These are mainly genes associated with biosynthesis of amino acids. From the 83 lethal genes in minimal media, 19 encode for gene products that catalyze more than one reaction. Only one of these 83 genes encoded for a protein that has an isoenzyme.
Modeling the shift from homolactic to heterolactic metabolism
To further evaluate the modeling capabilities of FBA we used the genome-scale flux model to simulate the shift from homolactic to heterolactic metabolism in L. lactis growing anaerobically. This process consisted in comparing simulation results with experimental observations reported by Thomas, T.D. et al. (1979) [30] for glucose limited anaerobic chemostat cultures, though refining and tuning the model through the introduction of appropriate biological meaningful constrains. Nevertheless, one should notice that although FBA is based on steady-state assumption and therefore is more suitable for simulation of metabolic behavior in continuous cultivations, model results can also be compared with batch experimental data under the assumption of pseudo steady-state during the exponential growth phase.
Experimental observations reported in literature suggest that, for L. lactis, product formation at high dilution rates during continuous cultivations is similar to product formation during batch growth at high glucose concentrations, resulting in lactic acid as the sole metabolic product [19,31]. On the other hand, growth at low dilution rates in continuous conditions or at low concentrations of glucose in batch conditions (ie, in the presence of low fluxes through glycolysis) results in a mixed-acid fermentation, where formate, ethanol and acetate are produced in a molar ratio of 2:1:1 [32]. While indicating that the shift in metabolism is due to regulation, these observations also suggest that different flux constraints would have to be introduced to model the shift from homolactic to heterolactic metabolism.
For simulations of phenotypic behavior, the first obvious objective function was thought to be "maximization of growth" while constraining for substrate uptake. However, it became clear that constraining values for amino acid uptake rates would pose a problem, as the in silico strain was auxotrophic for some amino acids, which could act as a carbon source, and no information was found available in the literature for amino acid uptake rates under chemostat conditions. To solve this problem, consumption rates were approximated to be equal to the amino acids requirement for biomass formation. In the absence of experimental information for amino acid uptake rates another approach was followed, in which phenotypic behavior was simulated by constraining the growth rate and minimizing for substrate uptake. Both approaches led to different and complementary qualitative results regarding the network capabilities, as discussed below.
Maximizing for growth
When maximizing for biomass formation, the metabolic phenotype cannot be correctly predicted from the reconstructed network by simply stating anaerobic constraints (see Methods). The process of tuning and refining the model is discussed below and detailed results are summarized in Figures 5 and 6 (see Additional file 5). Simulations in which the specific consumption rate of glucose was constrained to values equal or greater than 14.1 mmol gDW-1 h-1 revealed the occurrence of multiple solutions for products of metabolism. From these simulations it was observed that growth rate does not change with different glucose uptake rates, indicating that growth is nitrogen-limited and reaches a maximum at 0.82 h-1 (in good agreement with the fact that amino acid uptake rates were established based on the amino acid cell content at 0.8 h-1), while for glucose uptake rates lower than 14.1 mmol gDW-1 h-1 growth is glucose-limited. Since lactate production is reported under high glycolytic fluxes, and this is known to be due to regulation exerted by the NADH/NAD+ ratio on the pyruvate-formate lyase [33], additional constraints can be included in the model in order to account for regulatory information. For example, Covert, M.W. et al [34] described a Boolean on/off approach to account for regulation within the metabolic network. However, since pyruvate-formate lyase is essential to the formation of acetyl-CoA (and therefore, biomass) under anaerobic conditions, a simple on/off Boolean statement cannot be applied. Therefore, based on fitting the model against experimental data, the pyruvate-formate lyase flux was set to 2.15 and 9.8 mmol gDW-1 h-1 to simulate homolactic growth when glucose uptake rates equal 24.6 and 18 mmol gDW-1 h-1, respectively.
In all simulations it can be observed that the metabolic model predicts trace production of amino acid catabolism products (3-methyl-2-oxobutanoate, 3-methyl-2-oxopentanoate, 4-methyl-2-oxopentanoate, phenyllactate, methional), which is in good agreement with experimental results [35-37]. Interestingly, it can be observed that although the formation of these catabolic products is not directly accompanied by the formation of energy or reduced compounds, they all contribute to a gain in biomass formation.
Plotting experimental against model results for biomass, lactate, formate, ethanol and biomass formation (Figure 1) shows that simulations can reproduce the general observed tendencies for product and biomass formation, overestimating biomass formation. Also formate, ethanol and acetate are slightly overestimated for the range of the simulated conditions. At high glucose uptake rates, the model cannot predict the absence of these products. From Figure 1 it can further be observed that lactate production is poorly described with the considered constraints. Additional constraints could have been introduced to better describe lactate formation. However, this imposes too many uncertain variables, namely for amino acid consumption.
Minimizing for substrate uptake
To overcome this difficulty another approach was used: the minimization of substrates uptake while constraining the growth rate. From simulation results, the formation of end-products of the pyruvate metabolism is observed. No by-products from the amino acid catabolism were predicted using this approach, as it minimizes amino acids uptake, and therefore no excess of amino acids is available for catabolism of amino acids. Without further constraints, mixed-acid fermentation is observed, as this is energetically more favorable for the cell (more ATP is formed when acetate is synthesized). However, as mentioned above, high fluxes through glycolysis lead to regulation effects towards homolactic fermentation. Consequently, the reaction catalyzed by pyruvate-formate lyase was constrained to values between 0 and 9 mmol gDW-1 h-1 (see Table 6) by fitting simulation results against experimental data
From Figure 2 it is observed that model results for formate, ethanol and acetate formation fairly fit experimental observations, although the model underestimates ethanol and acetate production and overestimates formate production. Furthermore, glucose consumption is also underestimated by model predictions. The difference is around 14% at high substrate uptake rates and 25% for lower values. This can be due, for example, to differences in the maintenance energy between the real cell and the simulated system.
Analysis of amino acid requirements for in silico growth interestingly shows that, for most of the cases, amino acid uptake rates linearly increase with the growth rate (Figure 3). This observation only applies if metabolic families of amino acids, derived from the same precursor, are considered. Threonine was included in the serine-family (instead of the expected aspartate-family), as it is related with glycine through the reaction L-threonine <-> glycine + acetaldehyde. Even with this change, linearity is not observed for both these families. Two reasons can be hypothesized to explain that: either threonine synthesis pathway depends on the growth rate or, alternatively, it is the serine contribution for the pyruvate pool that depends on biomass formation rate.
When minimizing for substrate uptake, the predicted amino acid uptake rates corresponds to the theoretical amino acid requirements for the cell to grow at the established growth rate. However, in vivo amino acids consumption is usually higher than the theoretical needs for macromolecules biosynthesis. The excess of amino acids can then be further catabolized, resulting in the secretion of amino acid by-products.
Amino acid biosynthesis capabilities
In all simulations under the objective of minimizing the substrate uptake, it could be noted that some amino acids are not taken up from the medium even if they are present in the medium. The cell seems to prefer to synthesize some of them. Amino acids completely synthesized by the in silico strain were alanine, aspartate, glycine and phenylalanine (either if glutamate is or is not supplied). Remarkably, this observation is in very good agreement with experimental data reported by Jensen, N.B. et al. (2002) [38] for L. lactis subsp cremoris. In their work, they analyzed the capacity for the de novo biosynthesis of amino acids when all amino acids except glutamate were supplied, having observed that alanine, aspartate, phenylalanine and threonine, were synthesized de novo by the cell. Analyzing the flux distribution of the reactions involved in amino acid biosynthesis, it can be found that those preferences are associated with an increased production of 2-oxoglutarate, leading to an increased formation of L-glutamate.
Identfication of metabolic engineering targets
A relevant application of genome-scale metabolic models is the simulation of cellular behavior in response to genetic perturbations. Namely, genome-scale metabolic models can be used as tools in the design of metabolic engineering strategies, aiming at finding genetic targets leading to enhanced desired properties [20]. We exemplify here the use of the reconstructed network to predict potential ways to increase the yield of diacetyl, an important flavor compound in dairy products. Diacetyl is a by-product of L. lactis fermentative metabolism and it is produced chemically by oxidative decarboxylation of the metabolic intermediate 2-acetolactate (which is derived from the condensation of two molecules of pyruvate). Hence, knockout strategies leading to an increased yield of 2-acetolactate in glucose were investigated using FBA and MOMA.
Lactococci metabolism around pyruvate is depicted in Figure 4. From pyruvate, carbon can be redirected towards acetyl-CoA, lactate or 2-acetolactate. Common strategies to increase the flux towards 2-acetolactate have been gene knockouts around pyruvate (except of the lethal pyruvate dehydrogenase complex), over-producing the 2-acetolactate synthase and/or over-producing an heterologous NADH-oxidase [1,7,39-41]. For example, Henriksen et al (2001) [24] have succeed to convert up to 95% of glucose towards the formation of 2-acetolactate and related compounds through the deletion of lactate dehydrogenase and pyruvate-formate lyase. Hugenholtz et al (2000) [39] constructed a high-producing diacetyl strain able to redirect 80% (16%) of glucose into 2-acetolactate (diacetyl), by deleting acetolactate synthase and over-expressing an heterologous NADH-oxidase.
The first step in our modeling strategy was to find a set of appropriate constraints that lead to results comparable with experimental observation. By simply allowing unconstrained uptake of oxygen it was possible to obtain both 2-acetolactate and acetate as the main by-products of metabolism, which is in good agreement with experimental reports [39]. Oxygen uptake was about 1.4 mmol O2 / mmol glucose. The simulated growth rate and yield of 2-acetolactate on glucose are presented in Table 4. However, a plurality of solutions were observed under the chosen conditions. In order to minimize the number of solutions, we constrained the activity of lactate dehydrogenase (ldh), 2-acetolactate synthase (aldB) and alcohol dehydrogenase (adhA) to zero. This forces pyruvate to be redirected either towards acetate or 2-acetolactate (and related compounds, ie, C4 and C5 products).
Next, the impact of single gene deletions on product formation was simulated by maximizing for growth. It was found that the deletion of PTA leads to a slight yield increase. The enzyme Pta catalyzes the conversion of acetyl-CoA to acetyl-P, and its deletion eliminates the production of acetate. Furthermore, another single gene deletion was run on the "mutant" ΔldhΔaldΔadhAΔpta and three deletions leading to a higher production of 2-acetolactate were found, ΔFHS, ΔSERA and ΔZWF. The enzyme Fhs is a formyltetrahydrofolate synthetase (EC 6.3.4.3), involved in the folate metabolism. Its contribution towards a higher yield of 2-acetolactate is due to an increase in carbon availability for by-product formation, since this reaction is used by the cell to generate ATP and formate (in rich media). The enzyme SerA, a 3-phosphoglycerate dehydrogenase, is involved in serine metabolism. This part of the metabolism connects with Fhs. Hence, this deletion increases 2-acetolactate yield similarly to the deletion of SERA. Finally, ZWF deletion is only predicted by MOMA as being advantageous. ZWF encodes for glucose-6-phosphate 1-dehydrogenase, the first step of the pentose-phosphate pathway. This deletion is accompanied by a decrease in biomass formation, because the availability of NADPH decreases. Therefore, more carbon is redirected towards other products of metabolism (C4 and C5 compounds), resulting in an increase of 2-acetolactate yield. Observed yields and biomass formation rates are summarized in Table 4. One should keep in mind that these are theoretical maximum specific growth rates at which high yields of 2-acetolactate can be obtained. Other factors such as a limitation in the uptake rate of oxygen may lead to experimentally lower values.
Conclusion
The metabolic network of Lactococcus lactis was reconstructed based on genomic, physiological and biochemical information, comprising a total of 621 reactions and 509 metabolites. Lactococcal network characteristics are comparable with other bacterial genome-scale reconstructed networks.
Metabolic network analysis was carried out using FBA and MOMA. The genome-scale metabolic model for L. lactis was shown to be robust and able to predict many experimental observations, when considering additional constraints derived from available experimental data. The model proved to be a useful tool to analyze the metabolic capabilities of L. lactis and to understand how the individual components in the system interact and influence the overall cell function. For example, the model could predict that, if all the amino acids were supplied, the cell will prefer to synthesize de novo alanine, aspartate, glycine and phenylalanine. The model can now be used as a useful tool to test or develop novel metabolic engineering strategies to redirect fluxes towards the production of important products such as diacetyl, alanine and exopolysaccharides [1].
Reconstructed metabolic networks are finding many other applications than the ones described in this work. Optimization methods have also been used to assess maximum capabilities of the network and to analyze gene dispensability [42]. More recently, different methods to integrate metabolic networks with transcriptome data were described [43-45]. Patil and Nielsen (2005) [45] describe a method that represents the metabolic network as an enzyme-metabolite interaction graph and, assigning expression scores to each enzyme, it is possible to highlight which metabolites are more affected by transcriptional changes (the so-called reporter metabolites). The method also identifies the most active metabolic sub-network responding to a particular perturbation. In the near future it is expected that metabolic networks will be further used together with other types of 'omic' information and help to reveal hidden information on cellular organization and function.
Methods
Network reconstruction
The reconstruction process of the metabolic network of L. lactis involved a comprehensive search of the current knowledge of its metabolism. The process started based on the ORFs information from the annotated genome of L. lactis IL 1403 [3]. From this list, a reaction database was built using the available genomic, biochemical and physiological data accessible in databases and relevant literature. Particular focus was given to the reactions reversibility. Reactions whose reversibility could not be assessed were defined as reversible. The reaction set also includes a reaction for biomass formation defined as a drain of major building blocks into biomass. One reaction was included to account for non-growth dependent ATP maintenance. After the initial assembly of the entire metabolic network, the list was re-examined to account for metabolic and physiological details. The list of reactions was manually and carefully examined regarding reliability of gene assignment, all possible/probable catalytic activities of gene products and the in vivo reaction reversibility.
Biomass composition
An equation for biomass formation was developed to account for the drain of precursors and building blocks into biomass. Biomass synthesis was set as a linear combination of seven macromolecular components – proteins, DNA, RNA, lipids, lipoteichoid acids, peptidoglycan and polysaccharides – which were considered to account for the cell overall biomass composition. The individual composition of every component was maintained at a fixed stoichiometry, independent of the specific growth rate. Cellular energy requirements were also considered, by taking into account information on the ATP cost of polymerisation and growth and non-growth associated ATP maintenance. The detailed calculation of the biomass composition can de found in the Additional file 2.
From the developed equations, an elemental biomass composition was determined for the reconstructed in silico strain, CH1.95O0.63N0.22P0.02S0.01, corresponding to a molecular weight of 27.8 g/C-mol. This is in good agreement with values found in the literature. Novak, L. et al. (2000) have measured a value of 27.7 g/C-mol for L. lactis.
Mathematical frameworks: FBA and MOMA
Flux balance analysis (FBA) is a linear programming posed problem where equations are defined by mass balance-derived stoichiometric reactions, constrains are imposed by fluxes limitations and the objective function is established based on a biological meaningful objective.
Given a stoichiometric matrix derived from mass balance around all metabolites of a cell and assuming steady state, a system of linear equations is produced which simply states that the incoming fluxes are balanced by the outgoing fluxes. For a metabolic network comprising N metabolites and M metabolic reactions, the stoichiometric matrix can be written as:
where Sji is the stoichiometric coefficient of metabolite j in reaction i, vi represents the flux of reaction i, and bj quantifies the network's uptake (or secretion) of metabolite j. Reversible reactions are defined simply as two irreversible reactions in opposite directions, constraining all fluxes to nonnegative values, ie:
vi ≥ 0, ∀ i (2)
Some other constraints based on physiological or physicochemical aspects may be applied, such as thermodynamics considerations, regulation effects and maximal enzymatic rates:
α ≥ vi ≥ βi, ∀ i (3)
Establishing a particular objective function, Z, written as a linear combination of existing variables, the optimal solution can then be found at one corner of the set of feasible solutions. For metabolic applications, typical objective functions are maximization of biomass formation or minimization of substrate consumption.
Flux balance analysis can then be simply summarized as a linear programming problem posed as:
maximize Z = eg, biomass formation (4)
subject to
vi ≥ 0, ∀ i
α ≥ vi ≥ βi, ∀ i
Another mathematical framework that can be used to find flux distribution solutions is the so-called minimization of metabolic adjustment (MOMA) [17]. MOMA relaxes the assumption of optimal growth flux for gene deletions, displaying a suboptimal flux distribution that is intermediate between the wild-type optimum and the mutant optimum. The philosophy of MOMA can be interpreted as the projection of the FBA optimum onto the feasible space of the mutant. Therefore, MOMA can be posed as a quadratic programming problem, with the same set of linear constrains as for FBA and where the objective function is to minimize the distance between the feasible solutions space of both wild-type and mutant.
Both the FBA and the MOMA problems were solved with an in-house developed software using the following solvers: GNU Linear Programming Kit and Object-Oriented software for Quadratic Programming .
Model constraints
All simulations were run allowing free uptake of nucleotides (xanthine, uracil, cytosine, adenine, guanine and hypoxanthine), phosphate, biphosphate and sulfate. Additionally, non-growth associated ATP requirement was always set to 1 mmol ATP gDW-1 h-1, which is the value experimentally estimated for carbon-limitated chemostat cultures [19].
To simulate growth under anaerobic conditions, the oxygen uptake rate was set to zero. Experimental insight on pyruvate metabolism led to two additional constraints. Since it is known that the pyruvate dehydrogenase (PDH) complex is not active in the absence of oxygen [46], pdhA_1 and pdhB_1 fluxes were therefore set to zero.
Aerobic growth was simulated by allowing oxygen to be taken up by the model. Unless otherwise stated, the oxygen consumption was set to 3.61 mmol gDW-1 h-1 [22]. Under aerobic conditions the pyruvate dehydrogenase is active but the pyruvate-formate lyase (PFL) is strongly inhibited by oxygen [24]. pdhA_1 and pdhB_1 were therefore left unconstrained while the pfl_1 flux was set to zero.
Evaluation of growth requirements and minimal media
Network capabilities to utilize different sugar sources were evaluated using FBA. A number of different sugars were supplied, one at a time, together with all amino acids and nucleotides. A sugar consumption rate of 13.6 mmol gDW-1 h-1 was allowed, both under aerobic and anaerobic conditions (6.8 mmol gDW-1 h-1 in the case of the disaccharides sucrose, lactose, maltose and trehalose). When the specific growth rate was higher than 0.01 h-1, the sugar was considered to be used for growth. This value was established based on the observation that when supplying only amino acids but no sugar, a specific growth rate of 0.01 h-1 is predicted by the model, which is not observed experimentally [19].
Amino acid requirements for in silico cell growth were analyzed under anaerobic conditions. The dilution rate was fixed to 0.18 and 0.76 h-1 and the objective function used in this investigation was the minimization of substrates uptake rate. Beginning with all the amino acids available, single-omissions were simulated by setting the corresponding uptake rate to zero, one at a time. If no biomass is formed, the omitted amino acid was defined as essential for growth. Simulations were run with the reactions catalyzed by bcaT_1, bcaT_2 and bcaT_3 constrained to the catabolic direction.
A minimal medium was established by allowing uptake of glucose at a rate of 13.6 mmol gDW-1 h-1 and uptake of the previously determined essential amino acids at a rate of 0.5 mmol gDW-1 h-1 per amino acid. Remaining amino acids were individually supplied until growth was observed.
Single gene/reaction deletion analyses
Single gene deletion (SGD) and single reaction deletion (SRD) analyses were performed using both FBA and MOMA. Reaction deletions were simulated by setting the corresponding flux to zero. Gene deletions were simulated by setting to zero all fluxes catalyzed by the corresponding gene product. Lethality was evaluated based on the deletions that led to infeasible problems or to biomass formation lower than 0.01 h-1. For simulation of growth on rich media, glucose uptake was set at 13.6 mmol gDW-1 h-1, and all amino acids were allowed to be taken up (at a rate corresponding to the amino acid cell content assuming a specific growth rate of 0.8 h-1). Simulating growth on minimal media, only glucose at 13.6 mmol gDW-1 h-1, arginine, methionine, valine and glutamante (0.5 mmol gDW-1 h-1) were supplied.
Modeling of homolactic and heterolactic metabolism
The shift from homolactic to heterolactic metabolism was simulated under anaerobic conditions, using FBA. Appropriated flux constraints were determined by fitting the model to experimental results, which is described in the following.
Two different approaches were selected to predict cell growth and product formation. First, glucose and amino acid uptake rates were set to fixed values and biomass production was maximized. Growth rate and products formation were determined as a function of different glucose uptake rates. Values for sugar uptake rates were taken from the literature [4,30], ranging from 7 to 24.6 mmol gDW-1 h-1. Amino acids uptake rates were calculated from the amino acid cell content assuming a specific growth rate of 0.8 h-1 [4]. A second approach consisted of setting a fixed specific growth rate while minimizing for the substrate uptake rates (objective function written as a linear combination of glucose and amino acids uptake rates, all terms with coefficient one). In this case, uptake rates and product formation were calculated for different specific growth rates. Here, a constraint on the flux of the pyruvate-formate lyase had to be introduced to account for experimental evidence, as suggested by Melchiorsen et al. (2002) [31].
Design of a diacetyl overproducing mutant
Simulations were performed under aerobic conditions and on rich media (as described above), with glucose as the carbon source (13.6 mmol gDW-1 h-1). Pyruvate secretion was not allowed and oxygen uptake was unconstrained. In silico gene deletions were simulated by constraining the respective fluxes to zero and solving the FBA problem while optimizing for biomass. Single gene deletions leading to high 2-acetolactate formation rates while allowing for growth (specific growth rate higher than 0.01 h-1) were selected as targets for metabolic engineering, as described in the Results and discussion section.
List of abbreviations
LAB: Lactic Acid Bacteria
GRAS: Generally Recognized As Safe
FBA: Flux Balance Analysis
MOMA: Minimization Of Metabolic Adjustment
ORF: Open Reading Frame
PDH: Pyruvate dehydrogenase
PFL: Pyruvate-formate lyase
Authors' contributions
APO carried out all aspects of the work and drafted the manuscript. JN conceived of the study and participated in its design and coordination. JF participated in the analysis of simulation results, conceived of the study, participated in its design and coordination and helped draft the manuscript. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
List of reactions, ordered by metabolic pathway and alphabetically. List of reactions included in the metabolic network of Lactococcus lactis. Sheet "pathway_order" contains the reactions grouped by metabolic pathway, and journal references are indicated whenever reaction information is taken from journal references other than Bolotin et al. (2001). Sheet "alphabetic_order" lists the reactions alphabetically.
Click here for file
Additional File 2
Biomass composition. Derivation of the equation for biomass composition, based on literature information.
Click here for file
Additional File 3
Non-connected Metabolites. List of the non-connected metabolites and reactions.
Click here for file
Additional File 4
Single gene and reaction deletions. List of genes and reactions from the metabolic network found to be lethal, ie, no growth was observed when setting the corresponding flux to zero and applying FBA or MOMA. Results are described both under anaerobic and aerobic conditions and both during growth on rich and minimal media.
Click here for file
Additional File 5
Modeling the shift from homolactic ot heterolactic metabolism. These tables summarize the modeling procedure for inclusion of appropriate constrains when applying FBA to simulate anaerobic growth.
Click here for file
Acknowledgements
Mats Åkesson is acknowledged for fruitful discussions on modeling issues and results. Kiran R Patil is acknowledged for developing a program to perform single gene/reaction deletion using FBA and MOMA. Karin Hammer is acknowledged for contribution on insights of gene lethality in L. lactis.
Figures and Tables
Figure 1 Plotting model and experimental data for anaerobic growth, chemostat conditions, when maximizing for biomass formation. Growth rate and conversion yields for lactate (L), formate (F), ethanol (E) and acetate (A) on glucose (S) are plotted against glucose uptake rate. Model results are from simulations NX5 to NX10 (see Table 5 in Additional file 5). Experimental data is from Thomas, T.D., et al (1979) [30]. Ethanol and formate predictions are overlapped.
Figure 2 Plotting model and experimental data for anaerobic growth, chemostat conditions, when minimizing for substrates uptake. Glucose uptake rate and conversion yields for lactate (L), formate (F), ethanol (E) and acetate (A) on glucose (S) are plotted against growth rate. Model results are from simulations NS3, NS5, NS7, NS8 to NS11 (see Table 6 in Additional file 5). Experimental data is from Thomas, T.D., et al (1979) [30]. Acetate and formate predictions are overlapped.
Figure 3 Model predictions for glucose and amino acid uptake rates versus the growth rate. Model results are from simulations NS3, NS5, NS7, NS8 to NS11. Amino acids were grouped into six families: Histidine (His), Aromatic (Phe, Trp, Tyr), Serine-family (Cys, Gly, Met, Ser, Thr), Pyruvate-family (Ala, Ile, Leu, Val), Aspartate-family (Asp, Asn, Lys) and Glutamate-family (Arg, Glu, Gln, Pro) [47].
Figure 4 The pyruvate metabolism of Lactococcus lactis. LDH: lactate dehydrogenase; PDH: pyruvate dehydrogenase complex; PFL: pyruvate formate-lyase; ADHE: acetaldehyde dehydrogenase ; ADHA: alcohol dehydrogenase; PTA: phosphotransacetylase; ACKA: acetate kinase, ALS/ILV B: catabolic and anabolic 2-acetolactate synthase; ALDB-acetolactate decarboxylase; BUTA-diacetyl reductase; BUTB: acetoin reductase; NOX: NADH – oxidase. ..
Table 1 Main characteristics of the reconstructed metabolic network of Lactococcus lactis.
ORFs 358
Experimental evidence 63
Clear function (functional annotation derived from probable homologues) 286
Tentative function (functional annotation derived from tentative homologues) 2
Putative function (sequence has probable homologues of uncertain function) 7
Metabolites 509
Intracellular metabolites 414
Extracellular metabolites 95
Unique metabolites 422
Reactions 621
Internal reactions 492
Exchange fluxes 129
Reactions with ORFs assignments 476
Reactions based on biochemical evidence / physiological considerations or inferred by the demands imposed on the metabolic reaction network 145
Table 2 Connectivity of metabolites in Lactococcus lactis. Connectivity is defined as the number of reactions in which the listed metabolites participate in.
Metabolite Connectivity
ATP 117
proton 116
ADP 101
Phosphate 95
Pyruvate 60
Diphosphate 48
CO(2) 46
Glutamate 42
Phosphoenolpyruvate 39
NADP(+) 39
NAD(+) 38
NADPH 37
NADH 34
NH(3) 20
Table 3 Lethality of single gene deletions and single reaction deletions. Detailed data can be found in the Additional file 4.
Anaerobic Rich media FBA # lethal reactions 71
# lethal genes 33
MOMA # lethal reactions 81
# lethal genes 38
Minimal media FBA # lethal reactions 119
# lethal genes 72
MOMA # lethal reactions 131
# lethal genes 77
Aerobic Rich media FBA # lethal reactions 72
# lethal genes 34
MOMA # lethal reactions 84
# lethal genes 41
Minimal media FBA # lethal reactions 127
# lethal genes 77
MOMA # lethal reactions 138
# lethal genes 83
Table 4 Designed strategies to enhance 2-acetolactate production, and corresponding predictions for specific growth rates and yields of 2-acetolactate on glucose.
FBA MOMA
Growth (h-1) Y (C-mol/C-mol) Growth (h-1) Y (C-mol/C-mol)
Reference 0.82 0.31 0.82 0.31
Δldh ΔaldB ΔadhA 0.82 0.32 0.82 0.31
Δldh ΔaldB ΔadhA Δpta 0.56 0.74 0.323 0.34
Δldh ΔaldB ΔadhA Δpta Δfhs 0.56 0.76 0.53 0.75
Δldh ΔaldB ΔadhA Δpta ΔserA 0.56 0.76 0.53 0.74
Δldh ΔaldB ΔadhA Δpta Δzwf 0.38 0.71 0.38 0.75
==== Refs
Kleerebezem M Hols P Hugenholtz J Lactic acid bacteria as a cell factory: rerouting of carbon metabolism in Lactococcus lactis by metabolic engineering Enzyme Microb Technol 2000 26 840 848 10862894 10.1016/S0141-0229(00)00180-0
de Vos W Metabolic engineering of sugar catabolism in lactic acid bacteria Antonie Van Leeuwenhoek 1996 70 223 242 8879408 10.1007/BF00395934
Bolotin A Wincker P Mauger S Jaillon O Malarme K Weissenbach J Ehrlich SD Sorokin A The complete genome sequence of the lactic acid bacterium Lactococcus lactis ssp. lactis IL1403 Genome Res 2001 11 731 753 11337471 10.1101/gr.GR-1697R
Novak L Loubiere P The metabolic network of Lactococcus lactis: distribution of (14)C- labeled substrates between catabolic and anabolic pathways J Bacteriol 2000 182 1136 1143 10648541 10.1128/JB.182.4.1136-1143.2000
van Niel EWJ Hahn-Hagerdal B Nutrient requirements of lactococci in defined growth media Applied Microbiology and Biotechnology 1999 52 617 627 10.1007/s002530051569
Hugenholtz J Kleerebezem M Metabolic engineering of lactic acid bacteria: overview of the approaches and results of pathway rerouting involved in food fermentations Curr Opin Biotechnol 1999 10 492 497 10508636 10.1016/S0958-1669(99)00016-6
Kleerebezem M Boels IC Groot MN Mierau I Sybesma W Hugenholtz J Metabolic engineering of Lactococcus lactis: the impact of genomics and metabolic modelling J Biotechnol 2002 98 199 213 12141987 10.1016/S0168-1656(02)00132-3
Forster J Famili I Fu P Palsson BO Nielsen J Genome-scale reconstruction of the Saccharomyces cerevisiae metabolic network Genome Res 2003 13 244 253 12566402 10.1101/gr.234503
Edwards JS Palsson BO Metabolic flux balance analysis and the in silico analysis of Escherichia coli K-12 gene deletions BMC Bioinformatics 2000 1 1 11001586 10.1186/1471-2105-1-1
Edwards JS Palsson BO Systems properties of the Haemophilus influenzae Rd metabolic genotype J Biol Chem 1999 274 17410 17416 10364169 10.1074/jbc.274.25.17410
Schilling CH Covert MW Famili I Church GM Edwards JS Palsson BO Genome-scale metabolic model of Helicobacter pylori 26695 J Bacteriol 2002 184 4582 4593 12142428 10.1128/JB.184.16.4582-4593.2002
Sheikh K Forster J Nielsen LK Modeling Hybridoma Cell Metabolism Using a Generic Genome-Scale Metabolic Model of Mus musculus Biotechnol Prog 2005 21 112 121 15903248 10.1021/bp0498138
Edwards JS Ibarra RU Palsson BO In silico predictions of Escherichia coli metabolic capabilities are consistent with experimental data Nat Biotechnol 2001 19 125 130 11175725 10.1038/84379
Covert MW Schilling CH Famili I Edwards JS Goryanin II Selkov E Palsson BO Metabolic modeling of microbial strains in silico Trends Biochem Sci 2001 26 179 186 11246024 10.1016/S0968-0004(00)01754-0
Palsson B In silico biology through "omics" Nat Biotechnol 2002 20 649 650 12089538 10.1038/nbt0702-649
Patil KR Akesson M Nielsen J Use of genome-scale microbial models for metabolic engineering Curr Opin Biotechnol 2004 15 64 69 15102469 10.1016/j.copbio.2003.11.003
Segre D Vitkup D Church GM Analysis of optimality in natural and perturbed metabolic networks Proc Natl Acad Sci U S A 2002 99 15112 15117 12415116 10.1073/pnas.232349399
Andrade MA Brown NP Leroy C Hoersch S de Daruvar A Reich C Franchini A Tamames J Valencia A Ouzounis C Sander C Automated genome sequence analysis and annotation Bioinformatics 1999 15 391 412 10366660 10.1093/bioinformatics/15.5.391
Benthin S Growth and product formation of Lactococcus cremoris 1992 Technical University of Denmark
Stephanopoulos G Aristidou A Nielsen J Metabolic Engineering: Principles and Methodologies 1998 Academic Press, San Diego
Roissart H Luquet FM Roissart H and Luquet FM Bactéries Lactiques 1994 Lorica
Jensen NB Melchiorsen CR Jokumsen KV Villadsen J Metabolic behavior of Lactococcus lactis MG1363 in microaerobic continuous cultivation at a low dilution rate Appl Environ Microbiol 2001 67 2677 2682 11375180 10.1128/AEM.67.6.2677-2682.2001
Nordkvist M Jensen NBS Villadsen J Glucose Metabolism in Lactococcus lactis MG1363 under Different Aeration Conditions: Requirement of Acetate To Sustain Growth under Microaerobic Conditions Appl Environ Microbiol 2003 69 3462 3468 12788751 10.1128/AEM.69.6.3462-3468.2003
Henriksen CM Nilsson D Redirection of pyruvate catabolism in Lactococcus lactis by selection of mutants with additional growth requirements Appl Microbiol Biotechnol 2001 56 767 775 11601628 10.1007/s002530100694
Pedersen MB Koebmann BJ Jensen PR Nilsson D Increasing Acidification of Nonreplicating Lactococcus lactis{Delta}thyA Mutants by Incorporating ATPase Activity Appl Environ Microbiol 2002 68 5249 5257 12406711 10.1128/AEM.68.6.3010-3023.2002
Neves AR Ramos A Shearman C Gasson MJ Santos H Catabolism of mannitol in Lactococcus lactis MG1363 and a mutant defective in lactate dehydrogenase Microbiology 2002 148 3467 3476 12427938
Cocaign-Bousquet M Garrigues C Novak L Lindley ND Loubiere P Rational development of a simple synthetic medium for the sustained growth of Lactococcus lactis. Journal of Applied Bacteriology 1995 79 108 116
Lai CY Cronan JE Beta-ketoacyl-acyl carrier protein synthase III (FabH) is essential for bacterial fatty acid synthesis J Biol Chem 2003 278 51494 51503 14523010 10.1074/jbc.M308638200
Kobayashi K Ehrlich SD Albertini A Amati G Andersen KK Arnaud M Asai K Ashikaga S Aymerich S Bessieres P Boland F Brignell SC Bron S Bunai K Chapuis J Christiansen LC Danchin A Debarbouille M Dervyn E Deuerling E Devine K Devine SK Dreesen O Errington J Fillinger S Foster SJ Fujita Y Galizzi A Gardan R Eschevins C Fukushima T Haga K Harwood CR Hecker M Hosoya D Hullo MF Kakeshita H Karamata D Kasahara Y Kawamura F Koga K Koski P Kuwana R Imamura D Ishimaru M Ishikawa S Ishio I le Coq D Masson A Mauel C Meima R Mellado RP Moir A Moriya S Nagakawa E Nanamiya H Nakai S Nygaard P Ogura M Ohanan T O'Reilly M O'Rourke M Pragai Z Pooley HM Rapoport G Rawlins JP Rivas LA Rivolta C Sadaie A Sadaie Y Sarvas M Sato T Saxild HH Scanlan E Schumann W Seegers JFML Sekiguchi J Sekowska A Seror SJ Simon M Stragier P Studer R Takamatsu H Tanaka T Takeuchi M Thomaides HB Vagner V van Dijl JM Watabe K Wipat A Yamamoto H Yamamoto M Yamamoto Y Yamane K Yata K Yoshida K Yoshikawa H Zuber U Ogasawara N Essential Bacillus subtilis genes PNAS 2003 100 4678 4683 12682299 10.1073/pnas.0730515100
Thomas TD Ellwood DC Longyear VM Change from homo- to heterolactic fermentation by Streptococcus lactis resulting from glucose limitation in anaerobic chemostat cultures J Bacteriol 1979 138 109 117 108249
Melchiorsen CR Jokumsen KV Villadsen J Israelsen H Arnau J The level of pyruvate-formate lyase controls the shift from homolactic to mixed-acid product formation in Lactococcus lactis Appl Microbiol Biotechnol 2002 58 338 344 11935185 10.1007/s00253-001-0892-5
Melchiorsen CR Metabolic Engineering of Pyruvate Metabolism in Lactococcus lactis 2000 Technical University of Denmark
Garrigues C Loubiere P Lindley ND Cocaign-Bousquet M Control of the shift from homolactic acid to mixed-acid fermentation in Lactococcus lactis: predominant role of the NADH/NAD+ ratio J Bacteriol 1997 179 5282 5287 9286977
Covert MW Palsson BO Transcriptional regulation in constraints-based metabolic models of Escherichia coli J Biol Chem 2002 277 28058 28064 12006566 10.1074/jbc.M201691200
Christensen JE Dudley EG Pederson JA Steele JL Peptidases and amino acid catabolism in lactic acid bacteria Antonie Van Leeuwenhoek 1999 76 217 246 10532381 10.1023/A:1002001919720
Amarita F Fernandez-Espla D Requena T Pelaez C Conversion of methionine to methional by Lactococcus lactis FEMS Microbiol Lett 2001 204 189 195 11682200 10.1016/S0378-1097(01)00402-5
Yvon M Thirouin S Rijnen L Fromentier D Gripon JC An aminotransferase from Lactococcus lactis initiates conversion of amino acids to cheese flavor compounds Appl Environ Microbiol 1997 63 414 419 9023921
Jensen NB Christensen B Nielsen J Villadsen J The simultaneous biosynthesis and uptake of amino acids by Lactococcus lactis studied by (13)C-labeling experiments Biotechnol Bioeng 2002 78 11 16 11857275 10.1002/bit.10211
Hugenholtz J Kleerebezem M Starrenburg M Delcour J de Vos W Hols P Lactococcus lactis as a cell factory for high-level diacetyl production Appl Environ Microbiol 2000 66 4112 4114 10966436 10.1128/AEM.66.9.4112-4114.2000
Lopez F Starrenburg M Hugenholtz J The role of NADH-oxidation in acetoin and diacetyl production from glucose in Lactococcus lactis subsp. lactis MG1363 FEMS Microbiology Letters 1997 156 15 19 10.1016/S0378-1097(97)00394-7
Lopez F Kleerebezem M de Vos WM Hugenholtz J Cofactor engineering: a novel approach to metabolic engineering in Lactococcus lactis by controlled expression of NADH oxidase J Bacteriol 1998 180 3804 3808 9683475
Papp B Pal C Hurst LD Metabolic network analysis of the causes and evolution of enzyme dispensability in yeast Nature 2004 429 661 664 15190353 10.1038/nature02636
Akesson M Forster J Nielsen J Integration of gene expression data into genome-scale metabolic models Metab Eng 2004 6 285 293 15491858 10.1016/j.ymben.2003.12.002
Covert MW Knight EM Reed JL Herrgard MJ Palsson BO Integrating high-throughput and computational data elucidates bacterial networks Nature 2004 429 92 96 15129285 10.1038/nature02456
Patil KR Nielsen J Uncovering transcriptional regulation of metabolism by using metabolic network topology Proc Natl Acad Sci U S A 2005 102 2685 2689 15710883 10.1073/pnas.0406811102
Snoep JL de Graef MR Westphal AH de KA Teixeira de Mattos MJ Neijssel OM Differences in sensitivity to NADH of purified pyruvate dehydrogenase complexes of Enterococcus faecalis, Lactococcus lactis, Azotobacter vinelandii and Escherichia coli: implications for their activity in vivo FEMS Microbiol Lett 1993 114 279 283 8288104 10.1016/0378-1097(93)90284-9
Schlegel HG Allgemeine Mikrobiologie (7th Ed) 1992 7th Thieme
|
15982422
|
PMC1185544
|
CC BY
|
2021-01-04 16:03:39
|
no
|
BMC Microbiol. 2005 Jun 27; 5:39
|
utf-8
|
BMC Microbiol
| 2,005 |
10.1186/1471-2180-5-39
|
oa_comm
|
==== Front
BMC Med Inform Decis MakBMC Medical Informatics and Decision Making1472-6947BioMed Central London 1472-6947-5-191596974910.1186/1472-6947-5-19Research ArticleReal time spatial cluster detection using interpoint distances among precise patient locations Olson Karen L [email protected] Marco [email protected] Marcello [email protected] Kenneth D [email protected] Children's Hospital Informatics Program, Children's Hospital Boston, Boston, Massachusetts, USA2 Department of Pediatrics, Harvard Medical School, Boston, Massachusetts, USA3 Department of Biostatistics, Harvard School of Public Health, Boston, Massachusetts, USA4 Istituto di Metodi Quantitativi, Università Bocconi, Milano, Italy2005 21 6 2005 5 19 19 8 12 2004 21 6 2005 Copyright © 2005 Olson et al; licensee BioMed Central Ltd.2005Olson 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
Public health departments in the United States are beginning to gain timely access to health data, often as soon as one day after a visit to a health care facility. Consequently, new approaches to outbreak surveillance are being developed. When cases cluster geographically, an analysis of their spatial distribution can facilitate outbreak detection. Our method focuses on detecting perturbations in the distribution of pair-wise distances among all patients in a geographical region. Barring outbreaks, this distribution can be quite stable over time. We sought to exemplify the method by measuring its cluster detection performance, and to determine factors affecting sensitivity to spatial clustering among patients presenting to hospital emergency departments with respiratory syndromes.
Methods
The approach was to (1) define a baseline spatial distribution of home addresses for a population of patients visiting an emergency department with respiratory syndromes using historical data; (2) develop a controlled feature set simulation by inserting simulated outbreak data with varied parameters into authentic background noise, thereby creating semisynthetic data; (3) compare the observed with the expected spatial distribution; (4) establish the relative value of different alarm strategies so as to maximize sensitivity for the detection of clustering; and (5) measure factors which have an impact on sensitivity.
Results
Overall sensitivity to detect spatial clustering was 62%. This contrasts with an overall alarm rate of less than 5% for the same number of extra visits when the extra visits were not characterized by geographic clustering. Clusters that produced the least number of alarms were those that were small in size (10 extra visits in a week, where visits per week ranged from 120 to 472), diffusely distributed over an area with a 3 km radius, and located close to the hospital (5 km) in a region most densely populated with patients to this hospital. Near perfect alarm rates were found for clusters that varied on the opposite extremes of these parameters (40 extra visits, within a 250 meter radius, 50 km from the hospital).
Conclusion
Measuring perturbations in the interpoint distance distribution is a sensitive method for detecting spatial clustering. When cases are clustered geographically, there is clearly power to detect clustering when the spatial distribution is represented by the M statistic, even when clusters are small in size. By varying independent parameters of simulated outbreaks, we have demonstrated empirically the limits of detection of different types of outbreaks.
==== Body
Background
Public health departments in the United States are beginning to gain timely access to health data, often as soon as one day after a visit to a health care facility [1-3]. Consequently, new approaches to surveillance for disease outbreaks are being developed. These methods require models for baseline patterns and thresholds to detect unusual events [1]. Baseline patterns can be modeled in terms of temporal characteristics, spatial characteristics, or both. When cases are clustered geographically, such as those in the Amoy Gardens apartment complex during the 2003 SARS epidemic [4], an analysis of their spatial distribution may greatly facilitate the detection of a disease outbreak. Methods for both temporal and spatial surveillance have been recently reviewed [5,6].
One consideration regarding appropriate baseline data for spatial surveillance is whether to use individual point locations or aggregate counts by regions such as census tracts. Because aggregating points may result in a loss of precision [7], our work uses precise locations, i.e. geocoded patient addresses expressed as longitude and latitude. The novel approach of our method focuses on the detection of perturbations in the distribution of mutual distances among all the individual points in a geographical region to identify clusters [8-10]. Barring outbreaks, this distribution of interpoint distances can be quite stable over time (see Figure 1) [9]. We sought to measure the cluster detection performance of our method, and to determine factors affecting sensitivity to spatial clustering among patients presenting to hospital emergency departments (ED) with respiratory syndromes.
Figure 1 Pair-wise distances between home addresses of respiratory patients to one hospital over three years by season. The twelve curves (4 seasons × 3 years) overlap considerably, suggesting stability for the distance distribution over time. The maximum interpoint distance was 100 miles; the distribution up to 50 is shown.
Methods
This study identifies factors affecting the performance of an algorithm for measuring the degree of deviation from an expected geographic distribution of patient home addresses for a population visiting a localized site of care. The home address is only one of many possible places where a person might be exposed during an actual outbreak. However, other locations are not routinely recorded in the administrative databases typically used for syndromic surveillance. The approach was to (1) define a baseline spatial distribution of home addresses for patients visiting an emergency department with respiratory syndromes using historical data; (2) develop a controlled feature set simulation by inserting simulated outbreak data with varied parameters into authentic background noise, thereby creating semisynthetic data [11]; (3) compare the observed with the expected spatial distribution; (4) establish the relative value of different alarm strategies so as to maximize sensitivity for the detection of clustering; and (5) measure factors which have an impact on sensitivity.
Study population
Data were obtained retrospectively from hospital databases. The study was approved by the Institutional Review Board. Subjects were ED patients with respiratory syndromes treated at an urban, academic, pediatric, tertiary care hospital from December 24, 2000 to December 20, 2003. These dates were chosen to span the four seasons over three years while maintaining complete seven-day weeks. Patients with respiratory syndromes were identified by chief complaints and diagnostic codes as described in previous reports [1,12]. Of the total of 155,705 ED visits, 28% (43,156) were classified as having respiratory syndromes.
Home addresses of patients were translated to geographic coordinates using ArcGIS 8.1 (Environmental Systems Research Institute, Inc., Redlands, CA). Addresses were cleaned prior to geocoding using software (ZP4, Semaphore Corp., Aptos, CA) that matched addresses to the August 2003 United States Postal Service ZIP+4 database and made corrections. 93% (40,221) of the home addresses were successfully geocoded and patients who lived within 80 kilometers of the hospital (98%) were included in the study, for a total of 39,229 respiratory visits.
The number of visits for respiratory syndrome varied by season: 13,156 (34%) in the winter, 9,140 (23%) in the spring, 6,382 (16%) in the summer, and 10,551 (27%) in the fall. The home addresses of study patients were not evenly distributed within the study area. 14,231 (36%) lived from 0–5 km of the hospital, 16,351 (42%) from 5–15 km, 7,545 (19%) from 15–50 km, and 1,102 (3%) from 50–80 km (see Figure 2).
Figure 2 Baseline distribution of respiratory patients to the emergency department of one hospital. The study population (blue dots) lived within 80 km of the hospital (black ring). Simulated clusters were placed at 5, 15, and 50 km, along the red rings. Total population density of study patients within the four areas pictured was: 182.6 per square km within 0–5 km of the hospital, 32.6 per sq km within 5–15 km, 1.3 per sq km within 15–50 km, and 0.1 per sq km within 50–80 km.
Baseline spatial distribution of home addresses
The baseline spatial distribution was represented by a set of bins, each containing an equal proportion of pair-wise distance values. To establish this baseline, the three years of data were divided into 156 individual, one-week-long data sets. The number of respiratory visits each week ranged from 120 to 472. The average number varied by season: winter 346 (s.d. 68), spring 229 (s.d. 43), summer 164 (s.d. 28), and fall 271 (s.d. 49). For each week of data, all n(n-1)/2 pair-wise distances among patient addresses were calculated as follows:
d = 6378 × 2 × arcsin (), a = sin2 ((Y1 - Y2 )/2) + cos (Y2 ) × cos (Y1 ) × sin2 ((X1 - X2 )/2),
where Y1 and Y2 are latitude in radians for point 1 and point 2 of the pair, X1 and X2 are longitude in radians, and d is the interpoint distance in kilometers.
The sets of weekly pair-wise distances were combined into separate data sets by season, and then into a single data set with all seasons combined. Distance values in each data set were ranked in order of magnitude, and divided into ten bins, each with the same number of distance values. To maintain equal proportions, the widths varied (necessarily) across bins. Bin ranges were relatively small for the initial bins (2.6 km on average for bins 1–6), then increased somewhat (9.3 km on average for bins 7–9), with one large final bin (117 km). Next, each distance value in the individual week-long data sets was assigned the bin number into which it fell. For example, if one of the distance values between two patients during a particular winter week was 5 km, then that value fell into bin 3 because its endpoints were from 4.8 to 6.9 km. The number of records in each bin each week was then counted. This resulted in some variability in terms of how many pairs fell into each bin each week, although averaged over all of the weeks, each bin contained 10% of all distance values.
Controlled feature set simulation
Outbreaks were simulated by adding additional visits to the baseline data. 288 simulated spatial clusters were created using a cluster creation software tool [13]. Each cluster was added to each of the 156 week-long data sets, resulting in 44,928 weekly data sets containing simulated clusters. The clusters varied in size (10, 25, 40 additional visits), distance from the hospital (5, 15, 50 km), and radius of the circle within which points were randomly scattered (0.25, 0.5, 1, 3 km).
In addition, the special situation of an outbreak characterized by an increased number of visits originating over the whole geographic area of interest and, in effect, having no geographic clustering, was studied. Three data sets that varied on size (10, 25, 40 additional visits) were created to simulate this situation. Coordinates were randomly selected from the entire three years of data so that the extra visits would reflect the underlying geographic distribution of the study population. Each data set was inserted into each of the 156 week-long data sets, resulting in 468 weekly data sets containing extra visits dispersed randomly over the entire geographic area.
Comparison of observed versus expected spatial patterns
A metric, the M statistic, was used to characterize a discrepancy between an expected proportion of distance values in each bin and the actual proportions [9,10] using a nonparametric comparison based on the covariance matrix. Bins endpoints had been defined so that the expected proportions were equal for all bins. The statistic is intended to be sensitive to deviations in the geographic distribution. The M statistic was calculated as follows:
M = (obs - exp)T S- (obs - exp)
where obs is a vector of normalized observed proportions, exp is a vector of normalized expected proportions, S is a 10 × 10 variance-covariance matrix of the baseline proportions (calculated with data for 156 weeks). T refers to the transpose of the matrix, and S- refers to the Moore-Penrose generalized inverse of the S matrix. Proportions were normalized by dividing the bin frequency by the total for all bins and multiplying by 100.
Cutoff values
To evaluate the M statistic, cutoff values to indicate clustering were established for each season and for all seasons combined. A simulated baseline data set without extra visits or clusters was used to determine cutoff values at which a false positive alarm rate of .05 could be maintained. This baseline was generated from repeated random samples of patient locations from the week-long data sets described above.
Because the number of ED visits each week varied, sample sizes were generated from a list of weekly visit frequencies. For each season, 1000 frequency values were randomly selected from the weeks that comprised the season. For each of these values, that many addresses were randomly selected from the entire set of actual patient addresses for the season. The M statistic for each data set was calculated, and the 1000 values of the statistic were ranked by magnitude. In separate steps, the 1000 sample sizes were ranked, and the values of M times the sample size were also ranked. This process was repeated for each season. Finally, all seasons were combined and the entire process was repeated, using 5000 instead of 1000 samples for the all-season data. All cutoff values were based on percentile ranks.
Alarm strategies for the detection of clustering
Six alarm strategies utilizing the M statistic and the number (N) of ED respiratory visits were evaluated. These strategies are listed in Table 1. Each was designed to maintain a false positive rate of 0.05. Two strategies focused only on the number of visits and were included as a comparison to strategies that incorporated spatial information. Because N was not the focus of this study, more complex models for the time series data [14,15] were not investigated. Two strategies evaluated the geographic distribution of patient addresses, and two combined information regarding both the number of visits and the geographic distribution. Four of the six strategies ignored season, and cutoff values were established using the 5000 all-seasons samples. Two strategies required separate values for each season, and the 1000 samples for each season were used to establish these cutoffs. Each alarm strategy was applied to each semisynthetic data set. Sensitivity to the true positive alarms in the data sets with simulated clusters was expected to be high. On the other hand, false alarm rates for data with random additional visits were expected to be near 5% for strategies that evaluated the geographic distribution of visits.
Table 1 Description of alarm strategies for the detection of spatial clustering.
Alarm strategy Description
N > 95th percentile Number of ED respiratory visits is too high
N > 95th percentile, by season Number of visits is too high, separate values by season
M > 95th percentile M statistic is too high
M > 95th percentile, by season M statistic is too high, separate values for each season
MN > 95th percentile Calculate M × N, value is too high
N and MN rules N is too high (top 0.5% distribution)
Or M × N is too high (top 0.5% distribution)
Or both N is high (>80%) and M × N is high (>80%)
N = number of hospital emergency department respiratory visits, M = M statistic, used to characterize the geographic distribution.
Results
Overall performance of the alarm strategies
Overall sensitivity to detect clustering with the addition of simulated geographic clusters is listed in Table 2 by alarm strategy and time of year. Note that use of a single MN cutoff value at the 95th percentile yielded the highest overall sensitivity (62%), and the highest values by season, except for winter, where it was the second best strategy. Therefore, the presentation of results will highlight this strategy, although sensitivity for the other strategies is included in subsequent tables.
Table 2 Overall sensitivity to detect spatial clustering.
Percent of simulated outbreaks that exceeded a threshold
Alarm strategy All seasons a Winter b Spring b Summer b Fall b
N > 95th percentile 8.76 33.33 0.00 0.00 2.56
N > 95th percentile, by season 16.24 11.40 21.67 19.66 11.97
M > 95th percentile 49.17 26.68 53.34 73.71 42.26
M > 95th percentile, by season 49.13 43.61 49.42 55.35 48.01
MN > 95th percentile 62.32 55.43 63.49 70.90 59.27
N and MN rules 55.83 66.60 49.61 55.01 52.52
a The standard errors for All seasons were all less than or equal to 0.2%.
b The standard errors for Winter, Spring, Summer, and Fall were all less than or equal to 0.5%.
Evident in Table 2 is the observation that reliance only on the detection of an increased number of visits was a poor strategy when clusters were relatively small in size. Sensitivity was generally improved by instead relying on the geographic distribution of the clustered points. Sensitivity was most improved when both the increased number and the spatial distribution was incorporated into the alarm strategy.
Alarm rates when extra visits are not characterized by spatial clustering
A special situation examined in this study was an increase in the number of visits at the same three sizes as the cluster sizes. However, these additional visits were not deliberately characterized by spatial clustering. This scenario could represent either a random increase in visits or an outbreak spread over the entire region. The methods used in this study were not designed to be sensitive to situations where outbreaks do not cluster spatially. Therefore, alarm rates under these conditions could represent either a false alarm rate or a low sensitivity to widely dispersed outbreaks. Rates for each strategy by season are presented in Table 3. The strategies that consider the spatial distribution generally maintained false alarm rates near the desired rate of 5%, with a notable exception of one strategy in the winter.
Table 3 Alarm rates for extra visits that are not characterized by geographic clustering.
Percent of simulated outbreaks that exceeded a threshold
Alarm strategy # extra visits per week All seasons Winter Spring Summer Fall
N > 95th percentile 10 6.41 23.68 0.00 0.00 2.56
Overall rate = 8.76 25 8.97 34.21 0.00 0.00 2.56
40 10.90 42.11 0.00 0.00 2.56
N > 95th percentile, by season 10 9.62 7.89 12.50 10.26 7.69
Overall rate = 16.24 25 14.74 10.53 20.00 15.38 12.82
40 24.36 15.79 32.50 33.33 15.38
M > 95th percentile 10 3.21 0.00 0.00 12.82 0.00
Overall rate = 2.14 25 3.21 0.00 2.50 10.26 0.00
40 0.00 0.00 0.00 0.00 0.00
M > 95th percentile, by season 10 0.00 0.00 0.00 0.00 0.00
Overall rate = 0 25 0.00 0.00 0.00 0.00 0.00
40 0.00 0.00 0.00 0.00 0.00
MN > 95th percentile 10 4.49 2.63 2.50 12.82 0.00
Overall rate = 3.42 25 5.13 5.26 5.00 5.13 5.13
40 0.64 2.63 0.00 0.00 0.00
N and MN rules 10 7.69 26.32 2.50 0.00 2.56
Overall rate = 7.48 25 7.05 21.05 2.50 0.00 5.13
40 7.69 23.68 5.00 0.00 2.56
Overall rate is percent positive alarms, regardless of number of extra visits and season. Strategies that only considered N were expected to yield alarm rates greater than the false positive rate of 5% because extra visits were added. Strategies that considered the spatial distribution were expected to yield alarm rates near 5% because the extra visits were not spatially clustered.
Factors affecting sensitivity to detect clustering
The simulated clusters varied on several parameters. Sensitivity for the four alarm strategies that use the M statistic by cluster size, by distance from the hospital, and by density of the cluster are reported in Tables 4, 5, 6. The two strategies that use only N are not included in these tables because the geographic parameters are ignored by those strategies and results would be identical to those presented in Table 3.
Table 4 Sensitivity to detect clustering with simulated clusters of three sizes.
Percent of simulated outbreaks that exceeded a threshold
Alarm strategy # extra visits All seasons Winter Spring Summer Fall
M > 95th percentile 10 15.40 1.29 13.28 41.56 5.18
25 53.75 21.33 62.29 84.94 45.41
40 78.35 57.43 84.45 94.63 76.20
M > 95th percentile, by season 10 8.30 6.30 7.81 11.73 7.32
25 57.41 47.09 60.34 65.52 56.36
40 81.69 77.44 80.10 88.81 80.34
MN > 95th percentile 10 20.93 13.87 21.43 32.69 15.52
25 74.69 65.35 76.61 84.70 71.82
40 91.35 87.06 92.42 95.33 90.46
N and MN rules 10 14.86 35.22 6.43 9.56 8.97
25 65.12 74.67 57.68 67.17 61.38
40 87.50 89.91 84.71 88.30 87.21
Table 5 Sensitivity to detect clustering with simulated clusters at three distances from the hospital.
Percent of simulated outbreaks that exceeded a threshold
Alarm strategy km from hospital All seasons Winter Spring Summer Fall
M > 95th percentile 5 34.74 12.76 37.29 63.40 24.89
15 56.78 33.43 61.73 79.24 51.98
50 62.17 39.82 67.89 82.91 57.36
M > 95th percentile, by season 5 33.20 27.68 30.19 43.76 31.11
15 57.65 55.64 56.76 65.35 52.84
50 63.38 54.35 69.55 61.90 67.31
MN > 95th percentile 5 49.33 41.82 50.58 60.26 44.42
15 69.81 62.78 71.01 76.89 68.35
50 73.41 67.51 74.40 80.13 71.40
N and MN rules 5 42.44 56.21 34.10 40.92 39.08
15 62.87 72.02 57.47 62.91 59.43
50 67.92 76.03 63.90 67.25 64.80
Patient population density was greatest at 5 km from the hospital and declined as distance away from the hospital increased.
Table 6 Sensitivity to detect clustering with simulated clusters of four radius sizes.
Percent of simulated outbreaks that exceeded a threshold
Alarm strategy Radius All seasons Winter Spring Summer Fall
M > 95th percentile 250 m 53.94 30.15 59.06 78.45 47.33
500 m 53.22 29.86 58.23 77.53 46.55
1 km 51.30 28.07 56.04 75.57 44.80
3 km 38.22 18.64 40.03 63.28 30.38
M > 95th percentile, by season 250 m 54.75 48.10 55.56 61.47 53.67
500 m 53.82 47.48 54.24 60.58 52.81
1 km 51.73 45.72 51.81 59.05 50.18
3 km 36.24 33.15 36.08 40.31 35.36
MN > 95th percentile 250 m 66.97 59.54 68.82 75.04 64.25
500 m 66.41 59.10 67.67 74.96 63.68
1 km 64.71 57.75 65.94 73.11 61.82
3 km 51.21 45.32 51.53 60.51 47.33
N and MN rules 250 m 60.67 69.52 55.00 61.18 57.34
500 m 59.93 69.12 54.10 60.33 56.55
1 km 58.28 67.91 52.47 58.33 54.81
3 km 44.44 59.87 36.88 40.21 41.38
As shown in Table 4, clusters that are small in size produced the fewest alarms, with an overall sensitivity at size 10 of 21%. However, there was seasonal variability, from a low of 14% in the winter to a high of 33% in the summer. During the winter, when there were the most baseline visits, a cluster of size 10 was about one-seventh the size of the standard deviation for number of weekly visits. In contrast, during the summer, this cluster size was about one-third the standard deviation. With 25 points in the cluster, sensitivity improved markedly to an overall rate of 75%, and 40 clustered points yielded an overall sensitivity of 91%. Again, seasonal variability was evident with the lowest values in the winter and highest in the summer.
In Table 5, the effect of cluster location is demonstrated with clusters placed at three distances from the hospital. Those closest were in regions most densely populated by hospital patients, and were characterized by an overall sensitivity of 49% at 5 km. At greater distances, where the patient population density declined, sensitivity increased to 70% and 73% at 15 and 50 km. Seasonal variability was evident, with winter rates lowest and summer highest.
In Table 6, the effect of cluster dispersion is demonstrated with four radius sizes within which extra visits were randomly scattered. Although sensitivity declined with increasing radius size, the effect was not dramatic at 250 m, 500 m, and 1 km where overall alarm rates decreased from 67% to 65%. However, when the radius increased to 3 km, the decline in sensitivity was greater (51%). Once again, winter rates were lower than summer, and the 3 smallest radii had very similar rates by season.
Interactions among cluster parameters
To investigate the effects of interactions among the cluster parameters, a logistic regression analysis was performed. Cluster size, distance to hospital, radius size, and all higher order interactions were included in a model to predict whether or not the value of MN exceeded a threshold. All terms were significant, and the maximum-rescaled R-squared was .59. When season and all its interactions with the other variables were added to the model described above, an additional 3-way interaction was significant (cluster size × distance to hospital × season). When this interaction, season, and all two-way interactions with season were added to the first model, the maximum-rescaled R-squared was .61.
To further investigate these interactions, analyses that cross tabulated cluster size, distance from the hospital, and cluster density, and those that cross tabulated these variables with season were performed. Sensitivity values were ranked from highest to lowest to determine which type of cluster produced the least and the most alarms. Overall results are presented in Table 7, and results by season in Table 8.
Table 7 Sensitivity to detect clustering by number of extra visits, distance from the hospital, and radius of the simulated cluster.
Distance from the hospital
# extra visits Cluster radius 5 km 15 km 50 km
10 250 m 15.58 31.04 25.27
500 m 14.55 30.59 25.37
1 km 12.76 30.31 26.28
3 km 5.90 23.90 24.54
25 250 m 68.40 87.27 95.24
500 m 66.09 87.91 95.33
1 km 58.27 86.81 95.51
3 km 26.60 68.86 94.05
40 250 m 91.54 99.45 99.82
500 m 90.45 99.63 99.82
1 km 88.01 99.63 99.82
3 km 53.78 92.31 99.82
The numbers in the cells are the percentage of simulated outbreaks that exceeded the 95th percentile value of M × N (M statistic × number of visits).
Table 8 Sensitivity to detect clustering by season, number of extra visits, distance from the hospital, and radius of the simulated cluster.
Distance from the hospital
Cluster radius 5 km 15 km 50 km
WINTER
10 extra visits 250 m 11.58 18.05 16.92
500 m 11.05 18.05 16.92
1 km 10.00 18.80 18.05
3 km 5.00 12.78 16.92
25 extra visits 250 m 55.26 77.44 86.84
500 m 52.89 78.57 86.84
1 km 47.37 78.95 86.47
3 km 22.37 62.78 84.21
40 extra visits 250 m 83.95 98.50 99.25
500 m 82.63 98.87 99.25
1 km 78.16 98.87 99.25
3 km 41.58 91.73 99.25
SPRING
10 extra visits 250 m 16.50 33.93 23.93
500 m 14.00 32.86 23.93
1 km 12.50 32.50 25.00
3 km 7.25 27.14 22.86
25 extra visits 250 m 71.25 89.29 99.29
500 m 67.25 90.00 99.29
1 km 58.50 88.57 99.29
3 km 25.50 67.86 99.29
40 extra visits 250 m 95.50 99.64 100.00
500 m 94.00 99.64 100.00
1 km 92.25 99.64 100.00
3 km 52.50 91.07 100.00
SUMMER
10 extra visits 250 m 26.15 42.49 41.03
500 m 25.64 43.22 41.03
1 km 22.31 41.39 42.86
3 km 9.74 35.90 40.66
25 extra visits 250 m 84.10 95.24 98.53
500 m 83.33 95.60 98.90
1 km 75.13 94.87 99.27
3 km 38.97 77.66 99.27
40 extra visits 250 m 95.90 100.00 100.00
500 m 95.64 100.00 100.00
1 km 94.10 100.00 100.00
3 km 72.05 96.34 100.00
FALL
10 extra visits 250 m 7.95 29.30 19.05
500 m 7.44 27.84 19.41
1 km 6.15 28.21 19.05
3 km 1.54 19.41 17.58
25 extra visits 250 m 62.56 86.81 95.97
500 m 60.51 87.18 95.97
1 km 51.79 84.62 96.70
3 km 19.49 67.03 93.04
40 extra visits 250 m 90.51 98.63 100.00
500 m 89.23 100.00 100.00
1 km 87.18 100.00 100.00
3 km 48.72 90.11 100.00
The numbers in the cells are the percentage of simulated outbreaks that exceeded the 95th percentile value of M × N (M statistic × number of visits).
The simulated clusters that produced the fewest alarms were those with 10 extra visits, placed 5 km from the hospital within a circle having a radius of 3 km (sensitivity= 6%). Clusters of the same size, at the same distance, and within increasingly smaller radii also yielded few alarms. At this distance, underlying patient population density is greatest. Clusters that produced the most alarms were those with 40 extra visits, placed 50 km from the hospital, and radius size did not matter. Furthermore, sensitivity remained nearly as high (99%) when the same size clusters were placed 15 km away as long as the radius was less than 3 km. At these high rates, season had no effect on sensitivity. With a midrange cluster size (25), there was also high sensitivity (94–96%) at 50 km from the hospital with any radius size, but the effect of time of year became apparent. Winter rates (84–87%) at this distance for the four radius sizes were lower than rates for the other seasons (93–99%). Seasonal effects remained apparent as the presence of clustering became more difficult to detect. For clusters with the lowest alarm rates, those with 10 extra visits, winter sensitivity values (5–19%) were roughly half the size of summer values (10–43%). For clusters with 25 extra visits placed close to the hospital (5 km), winter rates for the different radii ranged from 22–55%, whereas summer rates ranged from 39–84%.
Discussion
This study illustrates the importance of considering spatial information for outbreak detection, and demonstrates that using an interpoint distance distribution and precise address locations is a powerful approach. Using readily available syndromic data, quite small outbreaks would generate an alarm when cases are spatially clustered. With just 10 extra visits per week, or just over one extra visit per day, spatial clustering was detected about 20% of the time. When just under six extra clustered visits were added per day, clustering was detected about 90% of the time. Although sensitivity varies with cluster parameters, the absolute values are not of primary importance when evaluating the results because some of the simulated clusters were intended to be difficult to detect. The patterns of high and low values highlight the effects of the parameters on sensitivity.
Clusters with the lowest alarm rates were those that were small in size, large in area, and located close to the hospital where underlying patient population density was greatest. Analyses in this paper were for patients at a single hospital who tend to live close to the hospital. If data from multiple hospitals were combined, the effect of distance from the hospital might be diminished as patients are spread more uniformly over the area of coverage. Also, not all hospitals are located in dense population centers. In new locations with different population density characteristics, the sensitivity of our method will likely vary.
The extreme values of the cluster parameters were chosen specifically to test the limits of detection. We found that clusters could be too small (10 extra visits) for our method to indicate clustering, and that they could be too widely dispersed (within a circle with a 3 km radius). Midrange values in terms of size and cluster radius were sensitive to clustering in our part of the country, and may appropriately characterize the parameters of clusters expected during an actual outbreak.
The effect of season is of interest because it suggests that the choice of alarm thresholds should be tailored to time of year. However, there is some arbitrariness to seasonal boundaries based on calendar dates. For example, a week that was just prior to the date that a season changed may actually be more like the season that follows that date. Hence, the strategy for the season that follows may instead be more appropriate. Furthermore, season itself is not likely the variable of interest. Instead, variables such as changing numbers of patients or changing locations from which patients come may be the factors that affect sensitivity, rather than the season itself. And while in general these variables change over seasons, the specific time at which they change varies. For example, influenza season occurs in the winter, but does not always occur on the same dates each winter.
Ideally, a detection strategy could more precisely handle changing characteristics of the baseline data, such as the onset of influenza season. In other work, we continue to develop such strategies for the M statistic. We also continue to test its performance in areas with different geographic characteristics and with different data types such as gastrointestinal syndrome or viral tests. And finally, we are working to extend the utility of our approach by developing methods to locate where the spatial clustering occurs.
This study uses the home address of the patient, readily available in hospital information systems. Should an outbreak spread through a work or school environment, or a place of common gathering, such as a baseball stadium, the distribution of patients' home addresses may not adequately reveal the appropriate clustering. However, the methods are applicable for other patient locations [16] should more complete location information be obtained from patients in clinical settings.
Conclusion
Measuring perturbation in the interpoint distance distribution is a sensitive method for detecting the presence of spatial clusters. When cases are clustered geographically, there is clearly power to detect clustering when the spatial distribution is represented by the M statistic, even when outbreaks are small in size. By varying independent parameters of simulated outbreaks, we have demonstrated empirically the limits of detection of different types of outbreaks.
List of abbreviations used
ED: Emergency Department
N: number
M: M statistic
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
KLO and KDM developed the study design. KLO was responsible for data acquisition and analysis, and writing the manuscript. MB and MP adapted the M statistic for biosurveillance. KDM participated in the analysis and wrote portions of 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
This work was funded by R01LM007970-01 from the National Library of Medicine, National Institutes of Health.
==== Refs
Mandl KD Overhage JM Wagner MM Lober WB Sebastiani P Mostashari F Pavlin JA Gesteland PH Treadwell T Koski E Implementing syndromic surveillance: a practical guide informed by the early experience Journal of the American Medical Informatics Association 2004 11 141 150 14633933 10.1197/jamia.M1356
Brillman JC Burr T Forslund D Joyce E Picard R Umland E Modeling emergency department visit patterns for infectious disease complaints: results and application to disease surveillance BMC Med Inform Decis Mak 2005 5 4 15743535 10.1186/1472-6947-5-4
Heffernan R Mostashari F Das D Karpati A Kulldorff M Weiss D Syndromic surveillance in public health practice, New York City Emerg Infect Dis 2004 10 858 864 15200820
Update 15 – Situation in Hong Kong, activities of WHO team in China http://www.who.int/csr/sarsarchive/2003_03_31/en/
Brookmeyer R Stroup DF Monitoring the health of populations: Statistical principles and methods for public health surveillance 2004 Oxford University Press
Sonesson C Bock D A review and discussion of prospective statistical surveillance in public health J Royal Statistical Soc A 2003 166 5 21 10.1111/1467-985X.00256
Bonetti M Forsberg L Ozonoff A Pagano M Banks HT, Castillo-Chaves C The distribution of interpoint distances Bioterrorism: Mathematical modeling applications in homeland security 2004 Philadelphia: SIAM
Bonetti M Pagano M The interpoint distance distribution as a descriptor of point patterns, with an application to spatial disease clustering Stat Med 2005 24 753 773 15523703 10.1002/sim.1947
Olson KL Bonetti M Pagano M Mandl KD A population-adjusted stable geospatial baseline for outbreak detection in syndromic surveillance Morbidity & Mortality Weekly Report, Syndromic Surveillance Reports 2004 53 256
Bonetti M Olson KL Mandl KD Pagano M Parametric models for interpoint distances and their use in biosurveillance Proceedings of the American Statistical Association, Biometrics Section [CDROM] 2003
Mandl KD Reis BY Cassa C Measuring outbreak-detection performance by using controlled feature set simulations Morbidity & Mortality Weekly Report, Syndromic Surveillance Reports 2004 53 130 136
Beitel AJ Olson KL Reis BY Mandl KD Use of emergency department chief complaint and diagnostic codes for identifying respiratory illness in a pediatric population Pediatric Emergency Care 2004 20 355 360 15179142 10.1097/01.pec.0000133608.96957.b9
Cassa C Olson KL Mandl KD A system to generate outbreak clusters for semisynthetic datasets to evaluate outbreak detection performance Morbidity & Mortality Weekly Report, Syndromic Surveillance Reports 2004 53 231
Reis BY Mandl KD Time series modeling for syndromic surveillance BMC Med Inform Decis Mak 2003 3 2 12542838 10.1186/1472-6947-3-2
Devine O Brookmeyer R, Stroup DF Exploring temporal and spatial patterns in public health surveillance data Monitoring the health of populations 2004 Oxford: Oxford University Press 71 98
Ozonoff A Bonetti M Forsberg L Pagano M The use of multiple addresses to enhance cluster detection Proceedings of the American Statistical Association, Biometrics Section [CDROM] 2003
|
15969749
|
PMC1185545
|
CC BY
|
2021-01-04 23:52:11
|
no
|
BMC Med Inform Decis Mak. 2005 Jun 21; 5:19
|
utf-8
|
BMC Med Inform Decis Mak
| 2,005 |
10.1186/1472-6947-5-19
|
oa_comm
|
==== Front
BMC Musculoskelet DisordBMC Musculoskeletal Disorders1471-2474BioMed Central London 1471-2474-6-371599240210.1186/1471-2474-6-37Research ArticleLumbar position sense acuity during an electrical shock stressor Hjortskov Nis [email protected] Christian [email protected] Nils [email protected] Department of Physiology, National Institute of Occupational Health, Lersø Parkallé 105, DK 2100 Copenhagen, Denmark2005 1 7 2005 6 37 37 5 11 2004 1 7 2005 Copyright © 2005 Hjortskov et al; licensee BioMed Central Ltd.2005Hjortskov 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
Optimal motor control of the spine depends on proprioceptive input as a prerequisite for co-ordination and the stability of the spine. Muscle spindles are known to play an important role in proprioception. Animal experiments suggest that an increase in sympathetic outflow can depress muscle spindle sensitivity. As the muscle spindle may be influenced by sympathetic modulation, we hypothesized that a state of high sympathetic activity as during mental stress would affect the proprioceptive output from the muscle spindles in the back muscles leading to alterations in proprioception and position sense acuity. The aim was to investigate the effect of mental stress, in this study the response to an electrical shock stressor, on position sense acuity in the rotational axis of the lumbar spine.
Methods
Passive and active position sense acuity in the rotational plane of the lumbar spine was investigated in the presence and absence of an electrical shock stressor in 14 healthy participants. An electrical shock-threat stressor lasting for approximately 12 minutes was used as imposed stressor to build up a strong anticipatory arousal: The participants were told that they were going to receive 8 painful electrical shocks however the participants never received the shocks. To quantify the level of physiological arousal and the level of sympathetic outflow continuous beat-to-beat changes in heart rate (beats*min-1) and systolic, diastolic and mean arterial blood pressure (mmHg) were measured. To quantify position sense acuity absolute error (AE) expressed in degrees was measured. Two-way analysis of variance with repeated measurements (subjects as random factor and treatments as fixed factors) was used to compare the different treatments.
Results
Significant increases were observed in systolic blood pressure, diastolic blood pressure, and heart rate during the stress sessions indicating elevated sympathetic activity (15, 14 and 10%, respectively). Despite pronounced changes in the sympathetic activity and subjective experiences of stress no changes were found in position sense acuity in the rotational plane of the lumbar spine in the presence of the electrical shock stressor compared to the control period.
Conclusion
The present findings indicate that position sense acuity in the rotational plane of the spine was unaffected by the electrical shock stressor.
==== Body
Background
Epidemiological studies have identified associations between work related psychosocial stressors and low back disorders [1,2]. The physiological mechanisms and pathways linking work stressors to low back disorders are uncertain.
In laboratory studies arousal or mental stress impaired the performance of different motor tasks. An electrical shock stressor was associated with reductions in steadiness of a pinch grip task [3,4] and the presence of mental stressors using non-supportive language and actions increased the spine compression during a standardized lifting task [5,6].
Optimal motor control of the spine depends on proprioceptive information as a prerequisite for coordination and the stability of the spine [7]. Position sense, functionally defined as the awareness of the actual position or movement of the limb, in the lumbar spine is influenced by low back pain [8-10], muscle fatigue [11] and muscle vibration [12]. The muscle spindle afferents play a major role in the sensation of position and movement [13-15] and factors altering the muscle spindle sensitivity may affect the proprioception.
Animal studies have shown the presence of sympathetic fibres [16] and adrenergic receptors inside the muscle spindle [17], and demonstrated how sympathetic stimulation during a muscle stretch reduced the muscle spindle sensitivity in rabbit jaw muscles [18,19]. In studies of human muscles, no detectable change in resting discharge of spindle firing during a period of increased muscle sympathetic activity (MSNA) was found [20]. Further, Matre & Knardahl [21] demonstrated that proprioceptive acuity was unchanged or, in one condition, improved during muscle sympathetic activation. Whether and how an increased sympathetic activation of the muscles spindles affect the proprioceptive acuity in human muscles are thus far from evident, and to the authors knowledge no studies have investigated the effect of mental stress, i.e. the response to an electrical shock stressor, on proprioceptive acuity in the low back region. The diversity of sympathetic outflow to different muscle groups during mental stress, i.e. mental stress increased MSNA in the calf but not in the forearm [22], should also be mentioned. In this regard it is unknown whether mental stress increases MSNA in the back muscles. It has however been suggested that when motor tasks requiring precision and continuous proprioceptive feedback are performed in work situations with strong excitement and stress, the enhanced muscle sympathetic outflow may affect motor performance through the muscle spindle system [18]. We examined whether a state of high sympathetic activity would affect the proprioceptive acuity of the back muscles. Therefore, the aim was to investigate the effect of mental stress, in this study the response to an electrical shock stressor, on position sense acuity in the rotational axis of the lumbar spine.
Methods
Participants
Fourteen healthy participants, 8 female and 6 male students (age 23.4 (SD 1.3) years, body mass 66.8 (SD 9.9) kg and height 173 (SD 9.5) cm) participated in the study. The participants had no history of injury or current problems with the low back. The local ethics committee of Copenhagen approved the study. All participants gave their informed consent.
Procedure
The procedure is illustrated in Figure 1. The study was performed over two days (2–4 days in between). The participants performed two position sense tasks (i.e. a passive and an active position sense task. These are described in the "position sense task section" below) during each of the following periods: a "stress period", a "novelty stress period" (day 1) and during two control periods (day 2).
Figure 1 Procedure. The experimental protocol at day one and day two. Arrows indicate the time the participants reported their subjective experiences of stress. "Stress" refers to the exposure to the electrical shock stressor.
The participants were exposed to an electrical shock stressor in the "stress period", i.e. the participants were told that they would receive 8 painful electrical shocks in this period (for further description of the stressor, see the electrical shock stressor section). The "novelty stress period" was without the exposure to the electrical shock stressor, but being anxious/nervous for participating in the experiment resulting in markedly elevations in the physiological stress markers (blood pressure and heart rate) and anxious/nervous for the stressor in the following "stress period". The "novelty stress", "stress" and the two control periods lasted 12 minutes each. Five minutes rest separated the "novelty stress" and the "stress" periods on day one and the control 1 and control 2 periods on day two. Prior to the testing periods the participants rested in 5-minutes. The procedure was the same on day two (control day) except that the "novelty stress" and the "stress" periods were changed to control periods (control 1 and 2).
It was not possible to randomise the order of the stress and control periods because pilot tests indicated that the physiological markers of stress (blood pressure and heart rate) were permanently elevated on day one due to novelty with the laboratory surroundings, and did not return to resting levels when the control and stress periods were performed in one day. This is also confirmed by the fact that there are order effects in resting blood pressure, i.e. the blood pressure was even increased in the resting period between the "novelty stress" and the "stress" periods. Lack of randomisation is in accordance with previous studies investigating the effect of mental stressors on motor performance [3-5,23].
Position sense tasks
Subjects were seated in a car-like chair of the Biodex System III Isokinetic Dynamometer (Biodex Medical Inc., Shirley, NY, USA) in a motorised rotational back attachment. The upper part of the trunk was strapped to the attachment with a belt at the level of the deltoid muscle, and the thighs were strapped to the chair. Further, the arms were strapped to the attachment in front of the participant. Horizontal rotations in the rotational plane of the lumbar spine from right to the left were performed from a starting position of 0° (relative to the sagittal plane) to target positions of 10°, 20° and 30° in a range of motion from 0° to 40°. The dynamometer was locked in the 0° position to ensure the same starting position in all the tests. Four trials were performed for each target position in each position sense task i.e. the participants performed 12 target positions in each position sense task. The order of the target positions and the passive and active tasks was randomised.
In the passive task, the trunk was passively moved at an angular velocity of 10°s-1 to a pre-determined target position. The target position was unknown to the participant to avoid the participants to predict the target position. The experimenter stopped the dynamometer at the target positions. The rotational attachment was then locked and the trunk remained at the target position for 5 s. Then the participant was passively returned at 10°s-1 to the starting position. After remaining in the starting position for 5 s the trunk was passively moved at 10°s-1 and stopped when the participant pressed a trigger. The trigger indicated recognition of the target position. In the active task the participant actively moved the trunk from the starting position until a command to stop was given. The rotational attachment was locked and the trunk remained at the target position for 5 s. Then the participant actively moved the trunk to the starting position remaining there for 5 s. The participant actively moved the trunk to match the target position and they indicated when the trunk was considered to be at the target position.
The test-retest reliability (2 days in between test and retest) of the passive and active position sense procedures was tested in a pilot study involving 10 participants. The statistical analysis showed no difference between the test and retest for the passive and active tasks. Based on the results of the intra class correlation (ICC as an estimate of reliability) (0.46 for the active task and 0.69 for the passive) and standard error of measurement (SEM as an estimate of precision) (0.64° in the active task and 0.39° in the passive task) values, the test reliability and precision for spinal position sense testing were moderate according to the criteria of Shrout & Fleiss [24]. These values are in accordance with previous studies e.g. [25-27]. The relatively low SEM values expressed in absolute error in the passive and active procedure indicate relatively good and precise test stability [28].
Sources of errors were minimized by using: the same experimenter in all trials; standardised verbal instructions, i.e. the participants received identical instructions about the proprioceptive tasks in the novelty stress, stress and control conditions; blindfolding to eliminate visual cues; randomisation of target positions; to keep each trial short (approx. 45 min), and familiarization with the principles in the testing procedure.
Measurements
The participants reported subjective experiences of stress prior to testing and after each of the periods at day one and two. The following four 11-point scales (0 = not at all, 10 = extremely) were used: 1) stressed, 2) tensed, 3) exhausted and, 4) concentrated [29,30].
To quantify the level of physiological arousal, non-invasive continuous beat-to-beat changes in heart rate (HR) (beats*min-1) and systolic (SBP), diastolic (DBP) and mean arterial blood pressure (MAP) (mmHg) were measured with an inflatable cuff placed over the proximal portion of the middle finger connected to a Finometer™ device (Finapres Medical Systems BV – TNO TPD Biomedical Instrumentation, The Netherlands) and recorded in a computer. The heart rate and blood pressure data were analysed using BeatScope software package version 1.1 (TNO TPD Biomedical Instrumentation, The Netherlands). The blood pressure was automatically corrected for hydrostatic pressure to compensate for vertical movements of the hand with respect to heart level and the concomitant pressure changes in the finger blood pressure.
The angular positions were recorded in a computer using a LabView program (National Instruments). The absolute errors (AE), i.e. the absolute values of the difference between the reproduced position and the target position, were determined using a MatLab programme. For testing differences in distance the mean AE for each target position (10°, 20° and 30°) for each subject was computed, and for testing differences in procedures (passive and active) and periods (novelty stress, stress and control 1 and control 2) the mean AE for each procedure and each period were computed.
Electrical shock-threat stressor
Prior to the stress period two circular dummy "shock electrodes" were attached to the dorsal side of the right forearm. The electrodes were attached to an electrical stimulation device. The aim was to build up a strong anticipatory arousal thereby strongly activating the sympathetic nervous system [31]. During the stress period – the participants did not receive any electrical shocks in the experiment – the participants were told that they were going to receive 8 painful electrical shocks without any presage either during the passive or active position sense task. To further heighten the arousal, the participants were told that they would receive additional electrical shocks every time the absolute error was exceeding 5°. When the first part of the stress period was terminated without any electrical shocks the participants were assured that they would receive the electrical shocks in the subsequent part of the stress period.
Statistics
Two-way analysis of variance with repeated measurements (subjects as random factor and treatments i.e. the novelty stress, stress and control 1 and 2 as fixed factors) was used to compare the different treatments. If significant differences appeared, multiple comparisons (Tukey test) were used to isolate the treatment that differed from the others. Mean AE for each subject for each target position was calculated. Mean AE, SBP, DBP, MAP and HR were dependent variables. Furthermore, median values of self-reports on stress, tenseness, exhaustion and concentration were dependent variables. Level of significance was set to P < 0.05.
Results
Subjective experience of stress
Self-reports on stress, tenseness, exhaustion, and concentration are presented in Figure 2. The participants experienced significantly more stress and tension in the stress period compared to the novelty stress, control 1 and 2 periods. Furthermore, the participants experienced significantly more tension in the novelty stress period compared to the control 1 and 2 periods. No differences were observed in the scales exhaustion and concentration between the periods.
Figure 2 Subjective experience of stress. Self-reports on stress, tenseness, exhaustion and concentration. Data are presented as means ± SD (n = 14). * P < 0.005. Stress vs. novelty stress, control 1 and 2**P < 0.05. Novelty stress vs. control 1 and 2.
Physiological arousal (blood pressure and heart rate)
SBP, DBP, MAP and HR are presented in table 1. SBP, DBP and MAP were significantly increased in the novelty stress and stress periods compared to the resting periods. SBP, DBP and MAP were significantly higher in the novelty stress and stress periods (day 1) compared to control 1 and 2 (day 2), whereas HR tended to be higher in the novelty stress and stress periods compared to control 1 and 2 (P = 0.055). No differences were observed between the novelty stress and stress periods on day 1 and between the control 1 and 2 on day 2 for SBP, DBP, MAP and HR.
Table 1 Blood pressure and heart rate. Means (SD) of systolic blood pressure, diastolic blood pressure, mean arterial pressure and heart rate during the resting period, novelty stress and stress periods on day 1 and the resting period, control 1 and 2 periods on day 2 (n = 14).
Systolic blood pressure (mmHg) Diastolic blood pressure (mmHg) Mean arterial Pressure (mmHg) Heart rate (Beats*min-1)
Day 1
Rest 132.3 (13.6) 84.0 (13.0) 102.4 (12.3) 78.7 (14.7)
Novelty Stress 142.0 (13.2)* 87.2 (13.1)* 107.8 (10.1)* 79.2 (15.5)
Stress 145.9 (12.0)* 91.3 (9.6)* 112.1 (9.0)* 80.80 (13.5)
Day 2
Rest 121 (13.1) 77.8 (8.5) 94.7 (9.0) 74.2 (13.7)
Control 1 128.9 (13.0) 80.8 (7.8) 98.8 (8.4) 77.1 (13.0)
Control 2 130.2 (11.6) 82.7 (7.0) 100.2 (7.7) 75.1 (11.4)
* P < 0.005 Control 1 and 2 vs. novelty stress and stress.
When expressed as percent change from the resting period, the SBP, DBP, MAP and HR were significantly elevated in the novelty stress and stress periods compared to the control period. The SBP, DBP, MAP and HR increased with 15.2 ± 8.9%, 13.5 ± 9.4%, 14,3 ± 8.5% and 10.2 ± 11.6%, respectively in the stress period and with 12.4 ± 7.5%, 9.3 ± 9.4%, 10.6 ± 7.9% and 9.9 ± 12.6%, respectively, in the novelty stress period compared to the resting period (Figure 3).
Figure 3 Physiological arousal. Percent change from the resting period for systolic blood pressure, diastolic blood pressure, mean arterial pressure and heart rate during the novelty stress and stress, and control 1 and 2 periods. Data are presented as means ± SE (n = 14). * P < 0.005 Stress vs. control 1 and 2. **P < 0.05 Novelty stress vs. control 1 and 2
Position sense acuity (absolute error)
Figure 4 shows the novelty stress and stress, control 1 and 2 comparisons for AE. No significant changes in AE for passive position sense acuity and active position sense acuity were found when comparing the novelty stress, stress, control 1 and 2 periods. AE was significantly lower in the active procedure compared to the passive procedure. When testing the effect of the distances no differences in AE for 10°, 20° and 30° were found in the active and passive position sense tasks.
Figure 4 Position sense acuity. Passive and active position sense acuity. Absolute errors between target position and the reproduced position in the novelty stress, stress, control 1 and 2 periods. Data are presented as means ± SE (n = 14).
Discussion
Significant changes in the cardiovascular parameters indicative of increased sympathetic activity during the electrical shock stressor and significant changes in self-reported tenseness and stress were demonstrated. However, the position sense acuity in the rotational plane of the lumbar spine was unaffected.
In view of the debate on a sympathetically mediated decrease in muscle spindle sensitivity the results provide limited information. The observation in the present study that an increase in sympathetic nervous activity, as indicated by the increase in blood pressure and heart rate, has no detrimental effect on position sense in the back has some important practical implications, but does not solve the problems on the potential role of the muscle spindle system. The maintained proprioceptive acuity may thus either reflect that a stress induced sympathetically mediated decrease in spindle sensitivity is of limited importance, or it could simply mean that a sympathetically mediated decree in spindles sensitivity does not occur in the present experimental conditions. Further, important to the hypothesis of the present study, that a state of high sympathetic activity would affect proprioception, is whether sympathetic nerves innervates muscle spindles and whether mental stress (in this study the response to the electrical shock stressor) increases muscle sympathetic activity to spinal muscles.
The hypothesis of stress induced disturbances of the sensitivity of the muscle spindle via sympathetic regulation originates from animal studies [18,19,32,33] demonstrating sympathetic fibres penetrating into the muscle spindle capsule [16] and the presence of adrenergic receptors inside the muscle spindle [17]. However, these findings may not be transferable into humans. Macefield et al. (2003)[20] failed to observe any changes in muscle spindle firing during a strong and sustained increase in MSNA in relaxed human leg muscles lending no support to the concept that the sympathetic nervous system can influence the sensitivity of human muscle spindles. This could explain the lack of effect of mental stress on position sense acuity in our study. Further, another study suggested that proprioception was unchanged or, in one condition, improved during muscle sympathetic activation lending partly support to the hypothesis of a sympathetic modulation of the muscle spindle [21]. Finally, Hjortskov et al.[34] demonstrated a facilitation of the short latency stretch reflex in the relaxed soleus muscle during manoeuvres known to increase MSNA, i.e. mental arithmetic, static handgrip exercise and post-handgrip ischemia. Although this is consistent with the idea that sympathetic nervous activity can exert a direct influence on human muscle spindles it is still unknown whether sympathetic nerves innervate human muscle spindles.
While it has consistently been shown that mental stress evokes an increase in MSNA e.g. [22,38], it is important to note the diversity of sympathetic outflow to different muscle groups i.e. during mental arithmetic an increased MSNA was seen in the lower limb but not in the upper limb [22]. Further, not all muscle spindles receive sympathetic innervations, and the proportion that does, varies between muscles [16]. Therefore, it may be that the low back muscles involved in the trunk movement in the present study, just as the arm muscles, are not under sympathetic control.
Further, if the low back muscles are under sympathetic control, the question arises as to whether MSNA was increased in the present study. In this regard, it has been shown that only high levels of mental stress increases the MSNA and that a decrease generally occurs during low levels of mental stress [35]. The effectiveness of the electrical shock stressor in activating the sympathetic system is assessed through its cardiovascular effects but it is unknown whether this stressor also elicits changes in MSNA and muscle spindle firing rate in the back muscles. However, arterial blood pressure responses have been proposed to reflect both cardiac and MSNA during mental stress. Compared to a study by Callister et al (1992) [38] reporting increases of 120–135% in MSNA, the cardiovascular changes were similar or even higher in our study. Likewise, compared to other studies using an electrical shock stressor [3,4,31], the physiological effects of the stressor were high. Therefore, if the low back muscles are under sympathetic control, it seems reasonable to assume an increased level of MSNA during the stress periods in the present study. It could however be that the level of MSNA during mental stress generally is not sufficient to influence the muscle spindle or that the low back muscles are not under sympathetic control as indicated above.
The type of proprioception tests may also explain the lack of change in position sense acuity in the present study. It has been suggested that primary muscle spindle afferents provide relatively more information on limb velocity whereas secondary muscle spindle afferents contribute mainly to limb position sense [15]. Interestingly, mental computation increased the response of the primary muscle spindle afferents while the secondary muscle spindle afferents exhibited no change in their sensitivity to stretch during mental computation [36,37]. Accordingly, it may be that proprioception testing designed to involve the primary muscle spindle afferents to a higher degree, as in replication of limb movement velocity, would be affected by mental stress.
Mental stress has been found to increase spine loadings during standardized lifting tasks [5,6,38] and to impair fine motor control [3,4]. Different mechanisms have been proposed. While Noteboom et al (2001a)[4] hypothesized elevated neuroendocrine activity during heightened arousal being responsible for the changes in motor performance, Davis et al (2002) [6] and Marras et al (2000) [5] suggested a biomechanical pathway leading to an overreaction of the musculoskeletal system i.e. in less controlled trunk movements and increases in trunk muscle coactivation. Further, it has been suggested that stress induced enhanced muscle sympathetic outflow to the muscle spindles may detrimentally affect motor performance and possible cause inefficient muscle use [18]. Contrary to that, Rossi-Durand (2002) [36] demonstrated that mental computation increased the muscle spindle sensitivity and suggested on that background that the increase in muscle spindle sensitivity could prepare the spindles to better play their role in proprioceptive information. The present study neither confirms nor disproves this suggestion.
A limitation of the study is the relatively low number of subjects participating in the study and that the order of the stress and control periods was not randomised. A larger study group may have influenced the results. Further, it could be argued that the short-term exposure to the stressor may have minimised the effects. However, despite the short-term exposure the cardiovascular stress indicators were markedly heighten during the "novelty stress" and the "stress" periods.
Conclusion
Participants presented with the shock stressor experienced significant changes in cognitive and physiological measures of stress. However, the position sense acuity in the rotational plane of the lumbar spine was unaffected during the electrical shock stressor. Further human studies on different muscle groups and with different testing procedures are needed to clarify the effect of mental stress on proprioception and motor control.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
NH, CHK and NF formulated the study idea, the choice of techniques and the experimental design and analyzed the experimental results in collaboration. NH and CHK performed the experiments. NH, CHK and NF processed the data, performed the statistics and drew figures/tables. NH wrote the draft of the manuscript, which was revised in collaboration and agreed upon in its final version by all authors.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
National Research Council and the Institute of medicine Musculoskeletal Disorders and the Workplace 2001 Washington, D.C., National Academy Press 1 492
Hoogendoorn WE Bongers PM Vet H I.L.D. H Ariens GA W. M Bouter LM Psychosocial work characteristics and psychological strain to low-back pain Scand J Work Environ Health 2001 27 258 267 11560340
J.T. N M. F Enoka RM Activation of the arousal response can impair performance on a simple motor task J Appl Physiol 2001 91 821 831 11457799
Noteboom JT Barnholt KR Enoka RM Activation of the arousal response and impairment of performance increasse with anxiety and stressor intensity J Appl Physiol 2001 91 2093 2101 11641349
Marras WS Davis KG Heaney CA Maronitis BS Allread WG The influence of psychosocial Stress, gender, and personality on mechanical loading of the lumbar spine. Spine 2000 25 3045 3054 11145816 10.1097/00007632-200012010-00012
K.G. D Marras WS C.A. H Waters TR Gupta P The impact of mental processing and pacing on spine loading. 2002 Volvo Award in Biomechanics Spine 2002 27 2645 2653 12461390 10.1097/00007632-200212010-00003
S. B Lysens R Spaepen A Lumbosacral repositioning accuracy in standing posture: a combined electrogoniometric and videographic evaluation Clin Biomech 1999 14 361 363 10.1016/S0268-0033(98)00086-2
Gill KP Callaghan MJ The measurement of lumbar proprioception in individuals with and without low back pain Spine 1998 23 371 377 9507628 10.1097/00007632-199802010-00017
Newcomer K Laskowski ER Yu B Larson DR An KN Repositioning error in low back pain Spine 2000 25 245 250 10685490 10.1097/00007632-200001150-00017
S. B Cordo P Lysens R S. V S. S The role of paraspinal muscle spindles in lumbosacral position sense in individuals with and without low back pain Spine 2000 25 989 994 10767813 10.1097/00007632-200004150-00015
Taimela S Kankaanpää M Luoto S The effect of lumbar fatigue on the ability to sense a change in lumbar position Spine 1999 24 1322 1327 10404574 10.1097/00007632-199907010-00009
S. B Lysens R S. S S. V Effect of paraspinal muscle vibration on position sense of lumbosacral spine Spine 1999 24 1328 1331 10404575 10.1097/00007632-199907010-00010
McCloskey DI Kinesthetic sensibility Physiol Rev 1978 58(4) 763 820 360251
McCloskey DI Garlick DG and Korner PI Propriocepton and control of movement in man Frontiers in physiological research 1984 1 Canberra, Australian Academy of Science 239 248
Sjölander P Johansson H Yahia L Sensory endings in ligaments: Response properties and effects on proprioception and motor control Ligaments and ligamentoplasties 1997 Berlin/Heidelberg, Springer -Verlag 39 83
Barker D Saito M Autonomic innervation of receptors and muscle fibres in cat skeletal muscle Proc Roy Soc,Ser B 1981 B212 317 332 6115396
C. G Passatore M Filippi GM Postsynaptic alpha1- og alpha2- adrenoceptors mediating the action of the sympathetic system on muscle spindles in the rabbit. Pharmacological Research Communications 1986 18 161 170 3010342 10.1016/0031-6989(86)90144-X
Roatta S Windhorst U Ljubisavljevic M Johansson H Passatore M Sympathetic modulation of muscle spindle afferent sensitivity to stretch to stretch in rabbit jaw closing muscles J Physiol 2002 540 237 248 11927683 10.1113/jphysiol.2001.014316
C. G F. D Passatore M Effect of sympathetic nervous system activation on the tonic vibration reflex in rabbit jaw closing muscles J Physiol (Lond) 1993 469 601 613 8271218
Macefield VG Y.B. S Wallin BG Resting discharge of human muscle spindles is not modulated by increases in sympathetic drive J Physiol 2003 551 1005 1011 12923218 10.1113/jphysiol.2003.040196
Matre D Knardahl S Sympathetic nerve activity does not reduce proprioceptive acuity in humans Acta Physiol Scand 2003 178 261 268 12823184 10.1046/j.1365-201X.2003.01122.x
Anderson EA Wallin BG Mark AL Dissociation of sympathetic nerve activity in arm and leg muscle during mental stress Hypertension, supplement 3 1987 9 3-114 3-119
Hjortskov N Rissén D Blangsted AK Fallentin N Lundberg U Søgaard K The effect of mental stress on heart rate variability and blood pressure during computer work Eur J Appl Physiol 2004 2004 84 89 14991326 10.1007/s00421-004-1055-z
Shrout PE Fleiss JL Intraclass correlations: uses in assessing rater reliability Psychol Bull 1979 86 420 428 10.1037//0033-2909.86.2.420
Lönn J Crenshaw AG Djupsjobacka M Johansson H Reliability of position sense testing assessed with a fully automated system Clin Physiol 2000 20 30 37 10651789 10.1046/j.1365-2281.2000.00218.x
Maffey-Ward L Jull G Wellington L Toward a clinical test of lumbar spine kinesthesia J Orthop Sports Phys Ther 1996 24 354 358 8938601
Lam SS Jull G Treleaven J Lumbar spine kinesthesia in patients with low back pain J Orthop Sports Phys Ther 1999 29 294 299 10342567
C.R. D D.W. B Assessing reliability and precision of measurement: An introduction to Intraclass correlation and standard error of mesurement Journal of Sport Rehabilitation 1993 2 35 42
Lundberg U Frankenhaeuser M Pituitary-adrenal and sympathetic-adrenal correlates of distress and effort J Psychosom Res 1980 24 125 130 7441580 10.1016/0022-3999(80)90033-1
Lundberg U Granqvist M Hansson T Magnusson M Wallin L Psychological and physiological stress responses during repetitive work at an assembly line Work & Stress 1989 3(2) 143 153
S. B H. BZ Y. B D.W. W G. L N. T S. L A. G N. L O. Z Experimental induction and termination of acute psychological stress in human volunteers: Effects on immunological, neuroendocrine, cardiovascular, and psychological parameters Brain, Behaviour, and Immunity 1998 12 34 52 10.1006/brbi.1997.0511
Passatore M F. D S. R C. G G. M Effects of cervical sympathetic nerve stimulation on the cerebral microcirculation: possible clinical implications Acta Neurobiol Exp 1996 56 117 127
Hellström F Thunberg J S. R Ljubisavljevic M Passatore M Johansson H Discharge responses of muscle spindles to stimulation of the cervical sympathetic nerve in cat neck muscles International Symposium Cairns, Australia, Sept 3-6, 2001 Movement and Sensation (IUPS) 2001
Hjortskov N Skotte J Hye-Knudsen CT Fallentin N Sympathetic outflow enhances the stretch reflex response in the relaxed soleus muscle in humans J Appl Physiol 2005 98 1366 1370 15542572 10.1152/japplphysiol.00955.2004
Callister R Suwarno NM Seals DR Sympathetic activity is influenced by task difficulty and stress perception during mental challenge in humans J Physiol (Lond) 1992 454 373 387 1474496
Rossi-Durand C The influence of increased muscle spindle sensitivity on achilles tendon jerk and Hæreflex in relaxed human subject Somatosens Mot Res 2002 19 286 295 12590830 10.1080/0899022021000037755
Ribot-Ciscar E Rossi-Durand C Roll JP Increased muscle spindle sensitivity to movement during reinforcement manoeuvres in relaxed human subjects Journal of Physiology 2000 523 271 282 10673561 10.1111/j.1469-7793.2000.t01-1-00271.x
K.G. D Heaney CA The relationship between psychosocial work characteristics and low back pain: underlying methodological issues Clin Biomech 2000 15 406
|
15992402
|
PMC1185546
|
CC BY
|
2021-01-04 16:32:05
|
no
|
BMC Musculoskelet Disord. 2005 Jul 1; 6:37
|
utf-8
|
BMC Musculoskelet Disord
| 2,005 |
10.1186/1471-2474-6-37
|
oa_comm
|
==== Front
BMC PediatrBMC Pediatrics1471-2431BioMed Central London 1471-2431-5-201598752410.1186/1471-2431-5-20Research ArticleQualitatively and quantitatively similar effects of active and passive maternal tobacco smoke exposure on in utero mutagenesis at the HPRT locus Grant Stephen G [email protected] Department of Environmental and Occupational Health, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, Pennsylvania 15260, USA2005 29 6 2005 5 20 20 11 2 2005 29 6 2005 Copyright © 2005 Grant; licensee BioMed Central Ltd.2005Grant; 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
Induced mutagenesis in utero is likely to have life-long repercussions for the exposed fetus, affecting survival, birth weight and susceptibility to both childhood and adult-onset diseases, such as cancer. In the general population, such exposures are likely to be a consequence of the lifestyle choices of the parents, with exposure to tobacco smoke one of the most pervasive and easily documented. Previous studies attempting to establish a direct link between active smoking and levels of somatic mutation have largely discounted the effects of passive or secondary exposure, and have produced contradictory results.
Methods
Data from three studies of possible smoking effects on in utero mutagenesis at the HPRT locus were compiled and reanalyzed, alone and in combination. Where possible, passive exposure to environmental tobacco smoke was considered as a separate category of exposure, rather than being included in the non-smoking controls. Molecular spectra from these studies were reanalyzed after adjustment for reported mutation frequencies from the individual studies and the entire data set.
Results
A series of related studies on mutation at the X-linked HPRT locus in human newborn cord blood samples has led to the novel conclusion that only passive maternal exposure to tobacco mutagens has a significant effect on the developing baby. We performed a pooled analysis of the complete data from these studies, at the levels of both induced mutation frequency and the resulting mutational spectrum.
Conclusion
Our analysis reveals a more commonsensical, yet no less cautionary result: both active maternal smoking and secondary maternal exposure produce quantitatively and qualitatively indistinguishable increases in fetal HPRT mutation. Further, it appears that this effect is not perceptibly ameliorated if the mother adjusts her behavior (i.e. stops smoking) when pregnancy is confirmed, although this conclusion may also be affected by continued passive exposure.
==== Body
Background
It has now been unambiguously established that cigarette smoking causes lung and other cancers, and that exposure to secondary tobacco smoke exhaled by smokers also has a causal role in carcinogenesis [1,2]. Carcinogenic smoke metabolites act primarily as genotoxicants by direct DNA adduction as well as by producing oxidative DNA damage. These DNA adducts therefore act as effective biological markers of tobacco smoke carcinogen exposure, integrating differences in metabolic capacity that modify the activation of such chemicals in the body. A subset of these adducts elude the cellular DNA repair systems, which also exhibit interindividual functional variation, persisting through DNA replication to produce mutations [reviewed in [3]]. Thus, induced mutations in surrogate reporter genes can also act as biomarkers of tobacco smoke carcinogenesis, although there is an attenuation of the genotoxic "signal". Induced mutagenesis is therefore more a measure of biological effect than a quantification of exposure [4]. The HPRT assay is the most widely applied measure of in vivo mutagenesis and has often been used as an intermediate biomarker of biological effect in exposed populations [5,6]. Although most individual studies and meta-analyses have demonstrated a significant induction of in vivo mutation in smokers [6,7], there are still exceptions [8-10]. Similarly, there are contradictory studies on the effect of maternal smoking on mutation frequencies in their unborn offspring [11,12].
Maternal tobacco smoking has been associated with premature delivery, low birth weight, deficient lung and neurological function, and increased risk of perinatal mortality [3,13-16]. During differentiation and development, specific cell types are sequentially induced to proliferate, when they become hypersensitive to cytotoxic agents, resulting in the observed dependence of teratological effects upon timing of exposure [17]. Although it has not been proven, there is great concern that the developing fetus might also be hypersensitive to genotoxic agents, producing many oncogenically "initiated" cells which might then expand during the fulfillment of the developmental program. Many studies have demonstrated that tobacco carcinogens cross the placenta [18,19], so that their mutagenic effects might be detected in the offspring. Additionally, in the particular case of the HPRT assay, cord blood mutation frequencies (Mf) measured at birth are approximately 10-fold lower than that predicted from age-dependence data in older children and adults, with a correspondingly low variance, such that cord blood HPRT mutation measurements should be uniquely sensitive to inductive effects [20].
A number of studies have been undertaken to determine whether maternal "lifestyle" factors influence in utero mutagenesis, often specifically targeting smoking as the putative source of genotoxicants. Indeed, two studies from essentially the same laboratories have come to very different conclusions on this question: first, that there was no detectable effect of maternal tobacco smoke exposure on HPRT Mf in cord blood [21], and second, that despite the fact that there was no increase in the Mf of children born to mothers who had been exposed to environmental tobacco smoke, there was a significant shift in the mutational spectrum in these children, indicating that different mechanisms of mutation were responsible for their observed Mf [22,23]. We now report a pooled reanalysis of these data, which provides evidence for a more coherent interpretation of these studies. We find that children of active smokers, women who quit smoking when they found they were pregnant and women who were exposed only to secondary smoke during their pregnancy all had similar, significant increases in T-lymphocyte HPRT Mf over offspring of women who reported neither active nor secondary exposure. These data provide a rationale for a shift in the HPRT mutational spectra in children of mothers passively exposed to environmental tobacco smoke. These data also provide evidence that secondary smoking exposure can have effects indistinguishable from active smoking. This result is discussed in the context of other attempts to document the genotoxic effects of tobacco smoke, and with regard to our own observations in women who attempted to protect their unborn child by quitting smoking during pregnancy.
Methods
Data
All HPRT assays were performed by the method of O'Neill et al [24] in the laboratories of R.J. Albertini, B.A. Finette and colleagues at the University of Vermont. Exposure histories were obtained by questionnaire at postpartum interviews. Women were considered to be passively exposed if they lived or worked in the presence of active smokers. Tobacco carcinogen biomarkers were not specifically determined in the first study [21], although biomarkers of drug abuse were concurrently monitored and corroborated interview data, and in previous studies in the same hospital population serum thiocyanate and cotinine levels corroborated histories of active smoking [25,26]. Tobacco smoke exposure in the second study [22,23] was assessed by measurement of cord blood cotinine levels [27]. The authors identified two subjects whose cotinine levels were not in agreement with their self-reported tobacco smoke exposure group (MFS72 from the passive exposure group and MFS30 from the "quitters" both had cotinine levels characteristic of active smokers) and excluded them from subsequent analysis. In addition, subject MFS99 from the actively smoking group had undetectable cotinine levels, and cotinine levels were not performed on MFS9 from the passively exposed group. These data were analyzed three ways, first retaining these samples in their self-reported categories, in keeping with the assignment of the data from Manchester et al [21], second, deleting them from the analysis as per Finette et al [22,23], and third, reassigning these samples to the category indicated by their cotinine results.
Data from the study of Manchester et al [21] was obtained by pooling their two data sets identified as "University Hospital Colorado" (tobacco smoke exposure was ascertained in only 60 of the 67 subjects analyzed) and "Private Hospital Colorado". Data from the second study is largely that of Finette et al [22] supplemented with new subjects MFS3, MFS14, MFS36, MFS83 and MFS89 in the non-smoking, non-passively exposed group, the addition of MFS12 to the passively exposed group (subsequently removed as an outlier), and the adjustment of the Mf of MFS65 in the passively exposed group, all reported in Finette et al [23].
Two outliers (defined as having HPRT Mf greater than three standard deviations higher than the population) had been previously identified, one in each of these studies. These values both remained outliers in their respective data sets after ln transformation, but only the highest outlier remained significant after pooling the transformed data from the two studies. Except where specifically mentioned in the text, inclusion or exclusion of these subjects did not affect the statistical analyses performed on these populations.
Statistical analysis
Pairwise analyses were performed on native and ln transformed data using Student's t test assuming equal variance from the statistical toolpack of Microsoft Excel. Nonparametric analyses were performed using the Mann-Whitney U test available in MiniTab. Except where noted, all three analyses yielded equivalent results regarding significance. Overall smoking effects were evaluated using single factor ANOVA from Excel.
Comparisons of distributions were performed by the chi-square test in MiniTab. Mutations were considered to be independent if they arose uniquely and/or demonstrated a unique rearrangement of the T-cell receptor β and γ genes [28]. In all three mutational spectra studies considered, mutants were derived from a subset of subjects and, in a small minority of cases, multiple clones were analyzed from the same individual. In the McGinness et al [29] and Manchester et al [21] studies molecular analyses consisted of Southern blotting with a full length human HPRT cDNA as probe. In the Finette et al [23] study, all but one mutant was defined more completely by sequencing of the base change(s) involved or of the deletion breakpoints.
Results and discussion
Study 1
The first set of data was derived from a cohort of 70 newborns born at two hospitals in Denver, Colorado [21]. HPRT Mf were determined on cord blood samples using the clonogenic assay. Smoking status was initially separated into three categories: active smokers, active smokers who quit after confirmation of pregnancy (quitters), and non-smokers. In a secondary analysis, non-smokers were then broken out into those likely to have ongoing exposure to environmental tobacco smoke and those actively avoiding passive or secondary smoking exposure. These categories were based on the paradigm that the genotoxic effects of secondary smoke should be intermediate between those of non-smokers with no secondary exposure and active smokers, and their exposure levels and therefore induced mutation frequencies were expected to be closer to those of the non-exposed population than those of the active smokers. Thus, the HPRT Mf of the total non-smoking population (with and without evidence of passive exposure, although clearly skewed towards the passively exposed population, which contributed 20 of the 28 subjects in this category) was used as the basis of comparison for the smoking and quitting groups, and neither was found to have a significant induction of HPRT mutants. In the present investigation, we have combined these primary and secondary analyses, using only the non-smokers with no evidence of secondary tobacco smoke exposure as the basis of comparison (this was the only category in the original study reported as having a significantly lower Mf). These data are summarized in Table 1, panel a.
Table 1 HPRT Mf in newborns with and without exposure to tobacco smoke metabolites in utero
HPRT Mf (× 10-6)
maternal exposure N mean ± SD median range P1,2 P2,3
a) data from Manchester et al [21]
unexposed 18 0.76 ± 0.50 0.61 0.14 – 1.9
passive only 20 1.60 ± 1.43 1.35 0.30 – 5.3 0.021
quit during pregnancy4 4 1.85 ± 1.16 1.60 0.35 – 3.2 0.004
smoked throughout 27 1.36 ± 0.99 0.98 0.28 – 3.5 0.019 0.012
b) data from Finette et al [22,23]
unexposed 26 0.72 ± 0.53 0.52 0.05 – 1.9
passive only5 22 1.18 ± 1.28 0.67 0.10 – 5.1 0.14
quit during pregnancy 8 0.79 ± 0.46 0.69 0.18 – 1.8 0.51
smoked throughout 12 0.71 ± 0.51 0.56 0.14 – 1.8 0.73 0.47
c) pooled data
unexposed 44 0.73 ± 0.51 0.60 0.05 – 1.9
passive only5 42 1.38 ± 1.36 0.87 0.10 – 5.3 0.006
quit during pregnancy4 12 1.27 ± 0.93 0.91 0.18 – 3.2 0.014
smoked throughout 39 1.16 ± 0.91 0.87 0.14 – 3.5 0.007 0.007
1specific exposed group vs. unexposed HPRT Mf from single factor ANOVA on ln transformed data
2since these four tests were performed simultaneously, to preserve an overall α of 0.05, the threshold for significance of each individual test should be set at P = 0.0125, or, if exposure is only tested for an induction of mutation, P = 0.025
3overall single factor ANOVA on ln transformed data
4excluding outlier with HPRT Mf of 14.7 × 10-6
5excluding outlier with HPRT Mf of 45.3 × 10-6
One outlier was identified in this data set, defined as an individual with an HPRT Mf greater than 3 standard deviations higher than the mean for the population. This individual was born to a women who quit smoking during her pregnancy, and had an Mf of 14.7 × 10-6, 10-fold higher than the mean of the entire population, 15-fold higher than the median value.
By breaking the "non-smoking" population into those with and without evidence of environmental tobacco smoke exposure, and using those with no evidence of such passive exposure as baseline, we now show both significant effects of tobacco smoke exposure overall in this population, as well as significant inductions in all three exposed categories. Moreover, the HPRT Mf in the three exposed populations were not significantly different from one another (pairwise P values ranged from 0.09 to 0.75), unless the outlier was included in the analysis, in which case the "quitters" were significantly higher than all three of the other groups.
Study 2
The second set of data is derived from two related publications [22,23], that were designed as follow-ups to those of McGinniss et al [11] and Manchester et al [21]. In the former study, newborns in Burlington, Vermont, demonstrated no detectable effect of maternal active smoking on cord blood HPRT Mf, although passive exposure was not considered, and therefore might have been a confounding factor. Subjects for the follow-up studies were recruited from the same university-affiliated hospital in Vermont, and had similar HPRT Mf. In this study, passive exposure was assessed by interview and ongoing tobacco smoke exposure was estimated by measurement of cotinine levels in the cord blood. In general, these cotinine measurements confirmed the smoking exposure assignments based on the interviews. These data are summarized in Table 1, panel b.
This population also contained an outlier, this one in the passively exposed group, with an Mf of 45.3 × 10-6, 30-fold higher than the average of the population and 70-fold higher than the median value.
Finette et al [22,23] reported on two different but overlapping subsets of these data, and found no evidence of any type of tobacco smoke exposure affecting HPRT Mf. Analysis of the entire data set, as summarized in Table 1, panel b, confirms these results. Indeed, only if the extreme outlier is included in the analysis is any comparison even close to significant (unexposed vs. passively exposed, P = 0.063).
Pooled data
These two studies examined similarly sized populations, and both failed initially to demonstrate an influence of tobacco smoke exposure on newborn HPRT Mf. These two sets of subjects are geographically distinct, and may differ in other ways, but this cannot be assessed from the published data. No other factor was reported to have significantly affected newborn HPRT Mf in either study, however. The Mf of the unexposed populations from the two studies are not significantly different from one another (P = 0.48), but the combined exposed population from the Colorado population is 1.5-fold higher than the equivalent population from the Vermont studies, which is significant (P < 0.001). This difference has been attributed to maternal environmental and socioeconomic factors, but nothing has been proven. The distribution of samples between the four smoking exposure categories differs significantly between the two studies (P = 0.044), with the major disparity being the proportion of active smokers (39% in the Manchester et al study [21] vs. 17% in the Finette et al studies [22,23], P = 0.006). It is tempting to invoke this difference in population distribution to explain the higher overall Mf of the Colorado population (mean 1.32 × 10-6, median 0.96 × 10-6, range 0.14–5.3 × 10-6) than the population from Vermont (mean 0.89 × 10-6, median 0.64 × 10-6, range 0.05–5.1 × 10-6) (P = 0.006) when the outliers are not included in the analysis. However, the proportion of all tobacco-exposed individuals (including active smokers, quitters and passively exposed mothers) is not significantly different between the two populations (74% vs. 62%, P = 0.66). The HPRT Mf for the pooled data set are given in Table 1, panel c.
Analysis of the pooled data from these two studies essentially reiterates the results of the reanalysis of the data from Manchester et al [21] discussed above: all three groups of tobacco exposed newborns have HPRT Mf significantly higher than the unexposed group, and there is no significant difference between the levels of induced mutation amongst the three exposed populations. These data indicate that tobacco smoke exposure in utero does induce detectable HPRT mutants in the fetus, and that passive maternal exposure has a similar teratogenic effect as active maternal smoking, a finding that is not unprecedented [30].
HPRT molecular spectra
Despite the lack of evidence for a mutagenic effect of tobacco smoke in their newborn cord bloods, Finette et al [23] nevertheless examined the molecular spectrum of HPRT mutants in two of their subpopulations, those without evidence of any maternal tobacco smoke exposure and those with passive exposure only. The mutations were classified as a) small, intragenic changes, b) gene rearrangements or deletions, or c) exon 2/3 deletions characteristic of illegitimate VDJ recombination (especially in newborn populations [23,31,32]). These data, summarized in Table 2, panel a, suggest a shift in the spectrum of the exposed population to significantly higher proportions of both small mutations and deletions attributable to VDJ recombination. Since there was no overall increase in HPRT Mf in this population, however, the exposed population also had a compensatory significantly lower proportion of non-VDJ mediated deletions and rearrangements, suggesting a protective effect of tobacco smoke exposure on these types of mutagenic events. We have found that the need to invoke such a protective effect is reduced if these data are put in perspective of the related studies mentioned above [21,23] and if mutation frequencies are used to normalize the distributions.
Table 2 HPRT mutational spectra in newborns with and without exposure to tobacco smoke metabolites in utero
a) distribution of mutant clones
maternal exposure study total independent mutants small mutations (%) deletions, rearrangements (%) VDJ recombinant deletions (%) P1
unexposed Finette et al [23] 30 10 (33) 14 (47) 6 (20)
mixed McGinniss et al [11] 41 7 (17) 14 (34) 20 (49) 0.039
mixed Manchester et al [21] 38 13 (34) 16 (42) 9 (24) 0.91
passively exposed Finette et al [23] 35 17 (49) 6 (17) 12 (34) 0.036
b) mutation frequencies for three classes of mutants based on individual studies
maternal exposure study overall mean Mf ± SD (× 10-6) small mutations Mf (× 10-6) deletions, rearrangements Mf (× 10-6) VDJ recombinant deletions Mf (× 10-6) P2
unexposed Finette et al [22,23] 0.72 ± 0.53 0.24 0.34 0.14
mixed McGinniss et al [11] 0.64 ± 0.40 0.11 0.22 0.31 0.003
mixed Manchester et al [21] 1.32 ± 1.093 0.45 0.56 0.31 < 0.001
passively exposed Finette et al [22,23] 1.18 ± 1.284 0.57 0.20 0.40 0.002
c) mutation frequencies for three classes of mutants based on pooled data5
unexposed Finette et al [22,23] 0.73 ± 0.51 0.24 0.34 0.15
mixed McGinniss et al [11] 0.99 ± 0.95 0.17 0.34 0.48 0.008
mixed Manchester et al [21] 0.99 ± 0.953 0.34 0.42 0.24 0.037
passively exposed Finette et al [22,23] 1.38 ± 1.364 0.67 0.24 0.47 < 0.001
1χ2
2t tests on ln transformed data
3excluding outlier with HPRT Mf of 14.7 × 10-6
4excluding outlier with HPRT Mf of 45.3 × 10-6
5for the purposes of this analysis the data of McGinniss et al [11] was pooled with that of Manchester et al [21] and Finette et al [22,23] to yield a single Mf.
Summaries of the HPRT mutational spectra generated from the earlier analysis of a newborn population from Vermont [29] and the Colorado UHD population [21] are also presented in Table 2a. These spectra were generated from mutants without regard for their potential tobacco smoke exposure, so are classified as "mixed". The population of McGinniss et al [11,29] contained only 20% active smokers, however, while the incidence of passive exposure of the remaining 80% of the population was not estimated. 45% of the population reported in Manchester et al [21] actively smoked throughout pregnancy, and another 33% reported ongoing exposure to secondary tobacco smoke; only 13% could be considered unexposed. These data might therefore be expected to begin to show the effects of both active cigarette smoking and passive secondary exposure on cord blood HPRT mutagenesis, although the power would not be as great as if they were derived only from defined exposed groups.
In adults, active tobacco smoke exposure has been found to increase the frequency and proportion of small base changes at the HPRT gene [33], consistent with the known mechanisms of tobacco smoke mutagens and the types of mutations found in oncogenes in smoking-associated cancers [34,35]. Illegitimate VDJ recombination is a mechanism of mutagenesis unique to T- and B-lymphocytes, and is implicated in many of the molecular events associated with leukemia and lymphoma [36,37]. The human HPRT gene contains cryptic sites for this DNA splicing event, resulting in the deletion of exons 2 and 3 [31,38], and the occurrence of this type of HPRT mutation seems to be associated with the incidence of acute lymphocytic leukemia in children [39]. Elevated levels of illegitimate VDJ recombination have been found in workers occupationally exposed to pesticides and herbicides [40], especially 2,4-dichlorophenoxyacetic acid [41,42] and in cancer patients undergoing chemotherapy [43], particularly with the DNA topoisomerase inhibiting agent etoposide [44,45].
Overall, the two newborn populations from Vermont had indistinguishable HPRT Mf (P = 0.50), and the total population data from McGinniss et al [11] was also consistent with the unexposed group reported by Finette et al [22,23] (P = 0.51), but the passively exposed group had a significantly higher level of mutation (P = 0.013). The distribution of mutants among the three mechanistic classes differed significantly in both cases, however, with the mutants from McGinniss et al [29] exhibiting less small mutation and more VDJ recombination-mediated deletion than either group from Finette et al [22,23]. The Colorado population had a significantly higher Mf than the unexposed subset of the second Vermont population (P = 0.008), but a very similar distribution of mutants (P = 0.91). On the other hand, the Colorado population had a similar mutation frequency as the passively exposed subpopulation from this study (P = 0.55), but a somewhat different mutant distribution (P = 0.068). We believe that these mutational comparisons are of little use unless both frequency and distribution are taken into account at the same time. In Table 2, panels b and c, the overall HPRT Mf from these individual studies and subpopulations (panel b), or the Mf generated from our pooled analysis (panel c), are used to calculate the frequency of each type of mutant in each population, as was done in Manchester et al [21] and Finette et al [23].
Expressing the mutational classes as frequencies makes it easier to see the general trends in these studies and their inconsistencies. The frequency of VDJ recombination-mediated deletions is now increased in all exposed populations, and the results from the mixed tobacco smoke populations are consistent with an intermediate level of exposure (remember that even though these populations should contain maternal active smoking exposures, and quitters, the meta-analysis indicated that these should have induced Mf similar to the passively exposed population). The differences in the frequencies of non-VDJ recombination-mediated deletions and rearrangements are diminished under these circumstances. The increase in frequency of small mutations observed in the passively exposed population of Finette et al [22,23] is difficult to rationalize with the low levels found in the McGinniss et al [29] study, however, the induction in the Manchester et al [21] population is again intermediate between those of the two subpopulations from Finette et al [22,23]. Significantly, none of the decreases observed in the frequencies of mutational subclasses from the unexposed population of Finette et al [23] were themselves statistically significant.
Discussion
Pooling data from studies applying similar techniques has been shown to be a useful method of investigating subtle effects in molecular epidemiology [46]. The present data was derived from a limited number of studies, however, and may contain unintended bias based on mutation detection methods, study design or uncontrolled confounders. Moreover, we stress that all of this data is based on mutation at a single locus, the X-linked HPRT gene, which may not be representative of the entire genome [5].
All cells in the embryo undergo periods of rapid differentiation and proliferation. It has long been postulated that rapidly growing cells are at increased risk of genotoxic damage; this idea is based on the hierarchy of tissues affected by ionizing radiation exposure, the response of tumors to genotoxic chemotherapy, and has been put forward as a way to rationalize hormonal carcinogenesis with the somatic mutational basis of cancer. While in utero exposures have been associated with later increases in cancer susceptibility, this research has mostly involved agents that interfere with the differentiation process [47], rather than classical mutagens [48].
Conclusion
This analysis demonstrates that, despite the conclusions of the original papers presenting the data, both active and passive tobacco smoke exposure in utero results in increased fetal mutation at the HPRT locus. The observed mutational induction by passive maternal tobacco smoke exposure clarifies the shift in the HPRT mutational spectrum previously reported [23], without requiring a complementary protective effect of tobacco smoke on certain types of mutation. The types of mutations observed are consistent with the known mechanisms of tobacco smoke mutagenesis, as well as the unique biochemistry of T-lymphocytes during in utero development. The establishment of these in utero tobacco smoke effects depended not only on the size of the pooled data set, but also on the judicious selection of a control group, and should abundantly demonstrate the long-term benefits of publishing data in a form that allows for such a re-analysis.
The observation that tobacco smoke mutational effects were not significantly ameliorated by quitting active smoking after the first trimester is troubling. It may well be consistent with the "all-or-none" quality of toxic exposures early in development, although it is doubtful that mutations arising by the mechanism of VDJ recombination-mediated deletion are possible at such an early stage of development. A more probable explanation for the persistent mutational induction observed in the quitters may involve continued passive exposure, since smoking mothers are far more likely to also be exposed to secondary smoke in the home [49]. This question should be directly addressed. We are presently analyzing data from another large set of newborns. In a preliminary report of the first third of this data maternal exposure to alcohol rather than tobacco was associated with higher HPRT Mf [50], although there was a shift in the mutational spectrum in the children of smoking mothers consistent with those described here [51].
Overall, these data suggest that further modification of residential and occupational exposures may be necessary to protect the developing fetus from tobacco smoke mutagenesis during pregnancy. If passive exposure does as much damage to the fetus as active smoking, it is imperative that workplace protection be offered to pregnant women, or better, to women who might or intend to become pregnant. This protection must also be provided in the home, where not only the mother, but any other smoking members of the household should be encouraged to quit for the duration of the pregnancy (or longer), or at least should not smoke in the presence of the pregnant woman.
Abbreviations
HPRT, hypoxanthine-guanine phosphoribosyltransferase; Mf, mutation frequency.
Competing interests
The author declares that he has no competing interests.
Authors' contributions
This study was conceived and performed by SGG.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This study was supported by grants HD33016 and AD36422 from the National Institute of Child Health and Human Development and by a grant from the University of Pittsburgh Competitive Medical Research Fund. We are particularly indebted to the authors of Finette et al [22,23] for publishing their data in sufficient detail as to allow this reanalysis.
==== Refs
Phillips DH Smoking-related DNA and protein adducts in human tissues Carcinogenesis 2002 23 1979 2004 12507921 10.1093/carcin/23.12.1979
Pfeifer GP Denissenko MF Olivier M Tretyakova N Hecht SS Hainaut P Tobacco smoke carcinogens, DNA damage and p53 mutations in smoking-associated cancers Oncogene 2002 21 7435 7451 12379884 10.1038/sj.onc.1205803
IARC Working Group on the Evaluation of Carcinogenic Risks to Humans Tobacco smoke and involuntary smoking IARC Monogr Eval Carcinogen Risks Hum 2004 83 1 1438
Grant SG Molecular epidemiology of human cancer: biomarkers of genotoxic exposure and susceptibility J Environ Pathol Toxicol Oncol 2001 20 237 253 11797833
Grant SG Jensen RH Garratty G Use of hematopoietic cells and markers for the detection and quantitation of human in vivo somatic mutation Immunobiology of Transfusion Medicine 1993 New York, Marcel Dekker 299 323
Cole J Skopek TR Somatic mutant frequency, mutation rates and mutational spectra in the human population in vivo Mutat Res 1994 304 33 105 7506357
Curry J Karnaoukhova L Guenette GC Glickman BW Influence of sex, smoking and age on human hprt mutation frequencies and spectra Genetics 1999 152 1065 1077 10388825
Tompa A Sápi E Detection of 6-thioguanine resistance in human peripheral blood lymphocytes (PBL) of industrial workers and lung cancer patients Mutat Res 1989 210 345 351 2911261
Sala-Trepat M Cole J Green MHL Rigaud O Vilcoq JR Moustacchi E Genotoxic effects of radiotherapy and chemotherapy on the circulating lymphocytes of breast cancer patients. III: Measurement of mutant frequency to 6-thioguanine resistance Mutagenesis 1990 5 593 598 2263217
Davies MJ Lovell DP Anderson D Thioguanine-resistant mutant frequency in T-lymphocytes from a healthy human population Mutat Res 1992 265 165 171 1370715
McGinniss MJ Falta MT Sullivan LM Albertini RJ In vivo hprt mutant frequencies in T-cells of normal human newborns Mutat Res 1990 240 117 126 2300072 10.1016/0165-1218(90)90015-T
Ammenheuser MM Berenson AB Stiglich NJ Whorton EB JrWard JB Jr Elevated frequencies of hprt mutant lymphocytes in cigarette-smoking mothers and their newborns Mutat Res 1994 304 285 294 7506372
Nilsen ST Sagen N Kim HC Bergsjo P Smoking, hemoglobin levels, and birth weights in normal pregnancies Am J Obstet Gynecol 1984 148 752 758 6702944
Moessinger AC Mothers who smoke and the lungs of their offspring Ann N Y Acad Sci 1989 562 101 104 2742269
Blair PS Fleming PJ Bensley D Smith I Bacon C Taylor E Berry J Golding J Tripp J Smoking and the sudden infant death syndrome: results from 1993-5 case-control study for confidential inquiry into stillbirths and deaths in infancy BMJ 1996 313 195 198 8696194
Habek D Habek JC Ivanisevic M Djelmis J Fetal tobacco syndrome and perinatal outcome Fetal Diagn Ther 2002 17 367 371 12393968 10.1159/000065387
Beckman DA Brent RL Mechanisms of teratogenesis Annu Rev Pharmacol Toxicol 1984 24 483 500 6203482 10.1146/annurev.pa.24.040184.002411
Anderson LM Hecht SS Dixon DE Dove LF Kovatch RM Amin S Hoffmann D Rice JM Evaluation of the transplacental tumorigenicity of the tobacco-specific carcinogen 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone in mice Cancer Res 1989 49 3770 3775 2736518
Coghlin J Gann PH Hammond SK Skipper PL Taghizadeh K Paul M Tannenbaum S 4-Aminobiphenyl hemoglobin adducts in fetuses exposed to the tobacco smoke carcinogen in utero J Natl Cancer Inst 1991 83 274 280 1994056
Finette BA Sullivan LM O'Neill JP Nicklas JA Vacek PM Albertini RJ Determination of hprt mutant frequencies in T-lymphocytes from a healthy pediatric population: statistical comparison between newborn, children and adult mutant frequencies, cloning efficiency and age Mutat Res 1994 308 223 231 7518049
Manchester DK Nicklas JA O'Neill JP Lippert MJ Grant SG Langlois RG Moore DH 3rdJensen RH Albertini RJ Bigbee WL Sensitivity of somatic mutations in human umbilical cord blood to maternal environments Environ Mol Mutagen 1995 26 203 212 7588645
Finette BA Poseno T Vacek PM Albertini RJ The effects of maternal cigarette smoke exposure on somatic mutant frequencies at the HPRT locus in healthy newborns Mutat Res 1997 377 115 123 9219586
Finette BA O'Neill JP Vacek PM Albertini RJ Gene mutations with characteristic deletions in cord blood T lymphocytes associated with passive maternal exposure to tobacco smoke Nature Med 1998 4 1144 1151 9771747 10.1038/2640
O'Neill JP McGinniss MJ Berman JK Sullivan LM Nicklas JA Albertini RJ Refinement of a T-lymphocyte cloning assay to quantify in vivo thioguanine-resistant mutant frequency in humans Mutagenesis 1987 2 87 94 3331707
Manchester DK Jacoby EH Sensitivity of human placental monooxygenase activity to maternal smoking Clin Pharmacol Ther 1981 30 687 692 7297026
Manchester DK Bowman ED Parker NB Caporaso NE Weston A Determinants of polycyclic aromatic hydrocarbon-DNA adducts in human placenta Cancer Res 1992 52 1499 1503 1540958
Van Vunakis H Gjika HB Langone JJ O'Neill IK, Brunnemann KD, Dodet B, Hoffmann D Radioimmunoassay for nicotine and cotinine Environmental Carcinogens: Methods of Analysis and Exposure Measurement 1987 9 Lyon: WHO/IARC 317 330
Nicklas JA O'Neill JP Albertini RJ Use of T-cell receptor gene probes to quantify the in vivo hprt mutations in human T-lymphocytes Mutat Res 1986 173 67 72 3001517 10.1016/0165-7992(86)90013-8
McGinniss MJ Nicklas JA Albertini RJ Molecular analyses of in vivo hprt mutations in human T-lymphocytes. IV. Studies in newborns Environ Mol Mutagen 1989 14 229 237 2583130
Rubin DH Krasilnikoff PA Leventhal JM Weile B Berget A Effect of passive smoking on birth-weight Lancet 1986 8504 415 417 2874412 10.1016/S0140-6736(86)92132-X
Fuscoe JC Zimmerman LJ Lippert MJ Nicklas JA O'Neill JP Albertini RJ V(D)J recombinase-like activity mediates hprt gene deletion in human fetal T-lymphocytes Cancer Res 1991 51 6001 6005 1933863
Yoshioka M O'Neill JP Vacek PM Finette BA Gestational age and gender-specific in utero V(D)J recombinase-mediated deletions Cancer Res 2001 61 3432 3438 11309304
Hackman P Hou S-M Nyberg F Pershagen G Lambert B Mutational spectra at the hypoxanthine-guanine phosphoribosyltransferase (HPRT) locus in T-lymphocytes of nonsmoking and smoking lung cancer patients Mutat Res 2000 468 45 61 10863157
DeMarini DM Shelton ML Levine JG Mutation spectrum of cigarette smoke condensate in Salmonella: comparison to mutations in smoking-associated tumors Carcinogenesis 1995 16 2535 2542 7586163
Denissenko M Pao A Tang M-S Pfeifer GP Preferential formation of benzo[a]pyrene adducts at lung cancer mutational hotspots in p53 Science 1996 274 430 432 8832894 10.1126/science.274.5286.430
Alt FW Blackwell TK DePinho RA Reth MG Yancopoulos GD Regulation of genome rearrangement events during lymphocyte differentiation Immunol Rev 1986 89 5 30 3081433
Kirsch IR Lista F Lymphocyte-specific genomic instability and risk of lymphoid malignancy Semin Immunol 1997 9 207 215 9200332 10.1006/smim.1997.0071
Fuscoe JC Vira LK Collard DD Moore MM Quantification of hprt gene deletions mediated by illegitimate V(D)J recombination in peripheral blood cells of humans Environ Mol Mutagen 1997 29 28 35 9020304 10.1002/(SICI)1098-2280(1997)29:1<28::AID-EM4>3.0.CO;2-9
Finette BA Poseno T Albertini RJ V(D)J recombinase-mediated HPRT mutations in peripheral blood lymphocytes of normal children Cancer Res 1996 56 1405 1412 8640832
Lipkowitz S Garry VF Kirsch IR Interlocus V-J recombination measures genomic instability in agriculture workers at risk for lymphoid malignancies Proc Natl Acad Sci USA 1992 89 5301 5305 1608939
Garry VF Tarone RE Kirsch IR Abdallah JM Lombardi DP Long LK Burroughs BL Barr DB Kesner JS Biomarker correlations of urinary 2,4-D levels in foresters: genomic instability and endocrine disruption Environ Health Perspect 2001 109 495 500 11401761
Knapp GW Setzer RW Fuscoe JC Quantitation of aberrant interlocus T-cell receptor rearrangements in mouse thymocytes and the effect of the herbicide 2,4-dichlorophenoxyacetic acid Environ Mol Mutagen 2003 42 37 43 12874811 10.1002/em.10168
Abdallah JM Lombardi DP Kirsch IR Genetic instability in patients with Hodgkin's disease undergoing chemotherapy J Clin Invest 1995 96 2744 2747 8675643
Chen CL Fuscoe JC Liu Q Relling MV Etoposide causes illegitimate V(D)J recombination in human lymphoid leukemic cells Blood 1996 88 2210 2218 8822941
Fuscoe JC Knapp GW Hanley NM Setzer RW Sandlund JT Pui CH Relling MV The frequency of illegitimate V(D)J recombinase-mediated mutations in children treated with etoposide-containing antileukemic therapy Mutat Res 1998 419 107 121 9804912
Le Marchand L Guo C Benhamou S Bouchardy C Cascorbi I Clapper ML Garte S Haugen A Ingelman-Sundberg M Kihara M Rannug A Ryberg D Stucker I Sugimura H Taioli E Pooled analysis of the CYP1A1 exon 7 polymorphism and lung cancer (United States) Cancer Causes Control 2003 14 339 346 12846365 10.1023/A:1023956201228
Hatch EE Palmer JR Titus-Ernstoff L Noller KL Kaufman RH Mittendorf R Robboy SJ Hyer M Cowan CM Adam E Colton T Hartge P Hoover RN Cancer risk in women exposed to diethylstilbestrol in utero JAMA 1998 280 630 634 9718055 10.1001/jama.280.7.630
Diwan BA Riggs CW Logsdon D Haines DC Olivero OA Rice JM Yuspa SH Poirier MC Anderson LM Multiorgan transplacental and neonatal carcinogenicity of 3'-azido-3'-deoxythymidine in mice Toxicol Appl Pharmacol 1999 161 82 99 10558926 10.1006/taap.1999.8782
Hruba D Kachlik P Influence of maternal active and passive smoking during pregnancy on birthweight in newborns Cent Eur J Public Health 2000 8 249 252 11125982
Bigbee WL Day RD Grant SG Keohavong P Xi L Zhang L Ness RB Impact of maternal lifestyle factors on newborn HPRT mutant frequencies and molecular spectrum – initial results from the Prenatal Exposures and Preeclampsia Prevention (PEPP) Study Mutat Res 1999 431 279 289 10635994
Keohavong P Xi L Day RD Zhang L Grant SG Day BW Ness RB Bigbee WL HPRT gene alterations in umbilical cord blood T-lymphocytes in newborns of mothers exposed to tobacco smoke during pregnancy Mutat Res 2005 572 156 166 15790499
|
15987524
|
PMC1185547
|
CC BY
|
2021-01-04 16:31:06
|
no
|
BMC Pediatr. 2005 Jun 29; 5:20
|
utf-8
|
BMC Pediatr
| 2,005 |
10.1186/1471-2431-5-20
|
oa_comm
|
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-731600017910.1186/1471-2458-5-73Research ArticlePrevalence of thyroid nodules in an occupationally radiation exposed group: a cross sectional study in an area with mild iodine deficiency Trerotoli Paolo [email protected] Anna [email protected] Giuseppe [email protected] Riccardo [email protected] Gabriella [email protected] Department of Internal Medicine and Public Medicine, Chair of Medical Statistic, University of Bari, Italy2 Department of Emergency and Transplantation, Chair of Endocrinology, University of Bari, Italy3 Radioprotection Staff, Direzione Sanitaria, Azienda Ospedaliera Policlinico, Bari, Italy2005 7 7 2005 5 73 73 15 2 2005 7 7 2005 Copyright © 2005 Trerotoli et al; licensee BioMed Central Ltd.2005Trerotoli 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
Thyroid nodules and thyroid cancer occur more frequently in people exposed to radiation for therapeutic purposes, and to nuclear fallout. Furthermore, it is known that a moderate degree of iodine deficiency may be responsible for an increased prevalence of thyroid nodules, while it is suspected that radiation exposure could induce changes in thyroid autoimmunity. The iodine intake of people resident in Bari, S. Italy, is mildly deficient, which could be presumed to cause a higher prevalence of thyroid pathology. This study was conducted to evaluate the prevalence of thyroid nodules in a population occupationally exposed to radiation, in an area of mild iodine deficiency.
Methods
A cross-sectional study was designed to evaluate the prevalence of thyroid nodules in radiation exposed workers, compared with a stratified sample of non exposed workers. After giving written consent to participate in the study, all the recruited subjects (304 exposed and 419 non exposed volunteers) were interviewed to fill in an anamnestic questionnaire, and underwent a physical examination, ultrasound thyroid scan, serum determinations of fT3, fT4 and TSH, fine needle aspiration biopsy. The sample was subdivided into one group exposed to a determined quantity of radiation (detected by counter), one group exposed to an undetectable quantity of radiation, and the non exposed control group.
Results
The prevalence of thyroid nodules <1 cm in diameter, defined as incidentalomas, in the exposed group with detected doses, was 11.28% in males and 9.68% in females, while in the exposed group with undetectable dose the prevalence was 10.39% in males and 16.67% in females. In the non exposed group the prevalence of incidentalomas was 9.34% in males and 13.20% in females. These prevalences were not statistically different when analysed by a multiple test comparison with the bootstrap method and stratification for sex.
Instead, the prevalence of thyroid nodules > 1 cm in diameter resulted statistically different in exposed and non exposed health staff: 18.68% in non exposed males vs exposed: 3.76% (determined dose) and 9.09% (undetectable dose) in males, and 20.30% in non exposed females versus 3.23% (detected dose) and 9.52% (undetectable dose) in exposed females.
There was a higher proportion of healthy staff in the exposed group than in the non exposed: (80.45% vs 68.68% in males; 80.65% vs 57.87% in females).
Conclusion
In our study, occupational exposure to radiation combined with mild iodine deficiency did not increase the risk of developing thyroid nodules. The statistically significant higher prevalence of thyroid nodules in the non exposed group could be explained by the high percentage (22%) of people with a familial history of, and hence a greater predisposition to, thyroid disease.
The endemic condition of mild iodine deficiency, demonstrated in other studies, played a major role in determining the thyroid pathology in our study groups.
==== Body
Background
Thyroid nodules and thyroid cancer occur more frequently in people exposed to radiation for therapeutic purposes, and to nuclear fallout. Furthermore, it is known that a moderate degree of iodine deficiency may be responsible for an increased prevalence of thyroid nodules [8,13,14,20] and it is also suspected that radiation exposure could induce changes in thyroid autoimmunity [5,21]. However, well-designed long term studies are needed to accurately evaluate the complex association between low-dose environmental exposure to radiation and thyroid autoimmunity [5,21].
Recent studies in the rat have shown that thyroid cells are influenced both by irradiation and iodine intake, underlining the importance of the cell proliferation caused by disorders in iodine intake, either deficient or in excess [3].
There is a general consensus in the fact that there is a high prevalence of thyroid cancer and thyroid nodule formation in workers who are occupationally exposed to radiation [1,2,4,22], although some authors have shown confidence intervals of the odds ratio that are not so consistent with the risk of radiation [6,7].
The iodine intake of people resident in Bari, S. Italy, is mildly deficient [12], which could be presumed to cause a higher prevalence of thyroid pathology, as has in fact been demonstrated [9,10]. Therefore, in conditions of prolonged occupational exposure we could expect an increased prevalence [16]. This hypothesis is also linked to the observation that prolonged exposure to radiation from a young age increases the risk of thyroid disease [15,17,18].
This study was conducted with the aim of investigating the occurrence of thyroid nodules in hospital workers exposed to radiation, living in an area of mild iodine deficiency, and comparing the results with a representative group of non exposed subjects working in the same hospital.
Methods
The study was conducted in the Azienda Ospedaliera Policlinico, Bari, and was designed as a cross-sectional study to evaluate the prevalence of thyroid nodules in staff occupationally exposed to radiation, and hence registered in the Radioprotection Service list. This group consists of 304 people, classified as belonging to the maximum risk category for radiation exposure, according to the Italian radioprotection law. They are periodically submitted to a complete physical examination by the physicians employed by the service, and the level of exposure is controlled by the physicists. We took into account the first examination of each exposed subject made in the year 2002.
A control group was set up, sending a letter inviting 986 non radiation-exposed staff, stratified for age, sex and assigned job unit, to participate in the investigation. The invitation was first sent in January 2002 and then repeated four times, until May 2003.
All the subjects who agreed to participate (n = 419; 42.5%) underwent a physical examination and a questionnaire was filled in collecting the following information: age, sex, job seniority, length of occupationally exposed period, cumulative annual radiation dose, presence of nodules confirmed by ultrasound scan, serum levels of fT3, fT4, TSH, familiarity, presence of other causes of radiation exposure (such as neck X-ray or CT scan, or radiotherapy).
To correctly define the kind of thyroid disease, a diagnostic protocol was established, including physical examination, ultrasound scan of the gland. Patients with thyroid nodules >1 cm in diameter were submitted to ultrasound examination and guided fine-needle aspiration (FNA).
Serum fT3, fT4, TSH levels were measured directly by RIA, only in the subjects with ultrasonographic thyroid alterations and CT was performed in the subjects with nodular lesions.
Statistical analysis
Data were collected on Excel for Windows spreadsheets, and analysed with SAS software version 8.2 for PC.
Data were summarized as number and percentage for the qualitative variables. To evaluate the relation between categorical variables the chi-square test, adjusted for sex (χ2MH), was performed.
Multiple comparison was performed to evaluate the differences in percentage of thyroid conditions among the different groups of exposure. The Cochrane-Armitage test was used, with a bootstrap p-value adjustment [22]. Quantitative variables, were summarised as median and interquartile range (IQR), and non parametric tests were performed, because the Gaussian distribution could not be accepted (Wilks test: p < 0.01).
Results
The percentage of the potential control group that agreed to participate in this study was 42.5% (419/986), but full information was received only from 38.8% of subjects (383/986). People were actively invited to contribute their data with repeated calls, but after one year (three repeat calls) the study was stopped.
Table 1 shows the main characteristics of the people recruited in the control group and exposed groups. Between the two groups there was a statistically significant difference for sex: the male percentage was 73.68% in exposed vs 47.52% in non exposed; the female percentage in exposed subjects was 26.32% vs 52.48% in non exposed (χ2 = 46.08; p < 0.0001). No statistically significant difference was detected for age. In the exposed no information was available about familiarity for thyroid disease, whereas a history of thyroid disease was present in 22.19% (85/383) of the control group. At the time of the study, 477 people had no thyroid nodules, 77.96% (237/304) in the exposed group and 62.66% (240/383) among the controls, while about 35 healthy people lacked a precise value for exposure.
Table 1 Distribution of subjects according to the main characteristics
Exposed Not exposed Not Included
Sex n % n % n %
Male 224 73,68 182 48,02 12 36,36
Female 80 26,32 197 51,98 21 63,64
Total 304 100,00 379 100,00 33 100,00
Missing - - 4 - 2 -
Exposed Not exposed Not Included
Age class n % n % n %
up to 40 77 25,67 84 25,53 10 29,41
41–50 157 52,33 155 47,11 13 38,24
more or equal to 51 66 22,00 90 27,36 11 32,35
Total 300 100,00 329 100,00 34 100,00
Missing 4 - 54 - 1 -
Exposed Not exposed Not Included
Work status n % n % n %
Graduate health workers 133 47,00 37 9,79 6 18,18
Not graduate health workers 137 48,41 170 44,97 23 69,70
Administrative - 0,00 65 17,20 - 0,00
Technical - 0,00 29 7,67 - 0,00
Assistant 13 4,59 77 20,37 4 12,12
Total 283 100,00 378 100,00 33 100,00
Missing 21 - 5 - 2 -
Nodular pathology, single or multiple, was detected in 18 subjects in the exposed group (5.92%; 18/304) and 76 in the non exposed group (19.85%; 76/383).
Regarding nodules with a diameter less than 1 cm (defined as incidentalomas) we found an equal prevalence in the exposed and non exposed groups (11.51%, 35/304 vs 11.49%, 44/383).
The exposed group was divided into two subgroups according to the doses detected by counter: 0 μSv and more than 0 μSv. Among the group with complete information about thyroid pathology and level of exposure we restricted the analysis to 283 exposed people, for whom we had the precise level of radiation doses for the last 8 years.
Therefore, we compared the prevalence of nodules in the three groups: one group directly exposed to radiation (G1), one group working in the radiation risk environment but without any detected radiation (G2), one group of non exposed subjects (G3).
The thyroid conditions were slightly modified, classifying people in 4 classes: healthy, affected by thyroiditis, nodular pathology (taking together single and multiple nodules), incidentalomas. (Table 2).
Table 2 Distribution of people according to sex, group of exposure and thyroid condition
MALE
Thyroid condition Exposed group with dose detected (G1) Exposed group with doses = 0 (G2) Not Exposed (G3)
N % N % N %
Incidentalomas 15 11,3 8 10,4 17 9,3
Nodules (single or multiple) 5 3,8 7 9,1 34 18,7
Thyroiditis 6 4,5 3 3,9 6 3,3
Healthy 107 80,5 59 76,6 125 68,7
Total 133 100,0 77 100,0 182 100,0
FEMALE
Thyroid condition Exposed group with dose detected (G1) Exposed group with doses = 0 (G2) Not Exposed (G3)
N % N % N %
Incidentalomas 3 9,7% 7 16,7% 26 13,2
Nodules (single or multiple) 1 3,2% 4 9,5% 40 20,3
Thyroiditis 2 6,5% 3 7,1% 17 8,6
Healthy 25 80,6% 28 66,7% 114 57,9
Total 31 100,0% 42 100,0% 197 100,0
A statistically significant association was found between levels of exposure and thyroid condition adjusted for sex (χ2MH = 24.89, p = 0.0004).
A normal thyroid was more frequent in groups G1 and G2 than G3 (G1 vs G3: p = 0.009; G1+G2 vs G3: p = 0.0167).
There was a statistically significant difference in the occurrence of nodules, in particular when we compared G1 or G2, separately or together, with G3 (G1 vs G3: p < 0.0001; G2 vs G3: p = 0.0025; G1+G2 vs G3: p < 0.0001). In fact, the G3 group showed a higher prevalence of nodules compared with the other two groups.
There was a slightly higher prevalence of incidentalomas in exposed males, at both levels of exposure (G1: 11.28%15/133; G2: 10.39 8/77; G3: 9.34% 17/182). Instead, in women there was a higher frequency of both incidentalomas and nodules in the non exposed group. These differences did not result statistically significant.
Neither the frequency of thyroiditis was significantly different among G1, G2 and G3.
Only one woman, in the non exposed group with nodular pathology, was positive for papillary carcinoma at FNA. Therefore, in our sample there was a nodule malignancy rate of 2.94%(in G3).
Data on the history of exposure were available since 1995. For G1 the median period of radiation exposure was 8 years in all the pathologic groups and the non pathologic group. The median cumulative radiation dose in people with incidentalomas resulted lower (143 μSv IQR 69–1100 μSv) compared with that of people with other conditions. No significant difference in radiation exposure was found among the groups (table 3).
Table 3 Cumulative radiation doses (μSv) in the exposed group differentiated by thyroid condition
N Min 1° quartile Median 3° quartile Max
Incidentalomas 18 21.00 69.00 143.50 1100.00 5671.00
Nodules (single or multiple) 6 136.00 144.00 751.50 1248.00 4525.00
Thyroiditis 8 5.00 230.50 444.50 2683.00 12111.00
Healthy 132 0.00 54.50 175.00 1137.50 24635.00
Discussion
A number of authors have underlined the role of radiation as a risk factor for the development of thyroid cancer, nodules or thyroiditis [2], particularly among X-ray workers and other health staff exposed to radiation in laboratories [5].
On the contrary, other authors have demonstrated that it is difficult to point to radiation as the cause of nodules or other non-malignant pathologies [13].
The Chernobyl episode gave rise to many epidemiological studies, analysing either the radiation risk for the general population or for workers. While the effect of radiation on the thyroid tissue of children is well proven [22], there is different evidence about the effects on the workers employed in the cleanup [21].
Little is known about the joint effect of iodine deficiency and radiation exposure on the risk of thyroid nodules: some authors have suggested that elimination of any iodine deficiency may be important in reducing the effects of radiation exposure on the thyroid [17].
Other authors have reported that mild iodine deficiency, or endemic iodine deficiency, can create a situation in which it is not possible to demonstrate the effect of radiation, because iodine intake is most effective as a risk factor [8,14].
Our results are not in agreement with this evidence, since we found a significantly different frequency of single nodules between the exposed groups and the control group, with a higher prevalence in the non exposed group. These results could be due to the higher percentage (22%) of people with familial thyroid disease and hence a greater predisposition to thyroid disease in the latter group.
Moreover, a selection bias could have entered the study, with a greater participation, especially in the control group, of people that were aware of the problem and decided to take advantage of the free medical tests.
We must also consider the fact that Bari and its surroundings are considered an area with mild iodine deficiency [12]. Our sample, both exposed and non exposed, is composed of people resident for more than 10 years in the Bari area, therefore environmental factors are very relevant in the development of thyroid pathologies.
Sex was found to be a significant factor for development of thyroid pathologies, both in the exposed and non exposed groups. Probably the well known characteristics of patients at risk of thyroid disease (females aged 40–60) have a stronger effect than radiation exposure in an occupational context where appropriate radioprotection measures are adopted.
Other authors have explained the lack of association of thyroid nodules and occupational radiation exposure by selection bias of the radiology service staff among healthy personnel, or by other selection filtering factors [22].
It is proven that long radiation exposure from a young age is a high risk factor for the development of thyroid cancer and nodules. In our sample, the exposure was well documented for the past 8 years, according to the Italian law, but the age of workers at the beginning of exposure, before the law came into effect, is not well documented.
In a recent meta-analysis by Tan and Gharib [19] the risk for malignancy in incidentalomas ranged between 0.45% and 13%. Contradictory attitudes have been proposed for the management of non palpable thyroid nodules. We agree with the suggestions that a systematic FNAB performed in all nodule <1 cm is not advisable, because only solid hypoechoic feature is a useful criterion to predict malignancy indipendently from the diameter of the nodule [11]. Considering that in our cohort of patients with incidentalomas, we didn't find these ecographic characteristics, we suppose the probably benign nature of most such lesions and we decided to kept them under periodical 'observation' without performing the FNAB. The real meaning of the high prevalence of incidentalomas in the exposed group requires a longer observation of these subjects and an enlargement of the casistic. The analysis of cumulative doses documented for each person did not result statistically significant. This can be considered a further element demonstrating the strict control of health conditions among exposed subjects, making it difficult to assess the risk of occupational radiation exposure because the effect of radioprotection measures can bias the conclusion.
Conclusion
Our cross-sectional study shows evidence that occupational radiation exposure does not increase the risk of developing thyroid nodules. The statistically significant prevalence of thyroid disease in the non exposed group in our study could be explained by the high percentage (22%) of subjects with a familial history of thyroid disease and hence a greater predisposition to such diseases.
A repeat cross-sectional study in the future could help to control any modifications of the prevalence of nodules and of incidentalomas in the exposed groups. In this way we could detect the onset of disease, and also take into account any job shifts to the non exposed group.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Trerotoli P., Serio G.: design of the study, data analysis and writing of the manuscript
Ciampolillo A.: patients examinations, interpretation of results of statistical analyses, writing of the manuscript
Marinelli G.: data collection and patients examination (exposed only)
Giorgino R.: revision of results and manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The Authors thank Dr Martino M., and Mrs Vulpis A. and other staff of the "Servizio di Radioprotezione" of Azienda Policlinico of Bari for their help.
They are grateful to Dr Talpo P., Servizio di Fisica, for providing supplementary information about radiation doses in the exposed group.
==== Refs
Antonelli A Silvano G Bianchi F Gambuzza C Tana L Salvioni G Baldi V Gasperini L Baschieri L Risk of thyroid nodules in subjects occupationally exposed to radiation: a cross sectional study Occup Environ Med 1995 52 500 504 7663633
Antonelli A Silvano G Gambizza C Bianchi F Tana L Banchieri L Is Occupationally Induced Exposure to Radiation a Risk Factor for Thyroid Nodule Formation? Arch Environ Health 1996 51 177 180 8687237
Boltze c Brabant G Dralle H Gerlach R Roessner A Hoang-Vu C Radiation-Induced Thyroid Carcinogenesis as a Function of Time and Dietary Iodine Supply: An in Vivo Model of Tumorigenesis in the Rat Endocrinology 2002 143 2584 2592 12072390 10.1210/en.143.7.2584
Castersen Jm Wingren G Hatschek T Fredriksson M Noorlind-Brage H Axelson O Occupational Risk of Thyroid Cancer: Data From the Swedish Cancer-Environment Register-1961–1979 Am J Ind Med 1990 18 535 540 2244626
Eheman CR Garbe P Tuttle RM Autoimmune thyroid disease associated with environmental thyroidal irradiation Thyroid 2003 13 453 64 12855012 10.1089/105072503322021115
Inskip PD Hartshorne MF Tekkel M Rahu M Veidebaum T Auvinen A Crooks LA Littlefield LG McFee AF Salomaa S Makinen S Tucker JD Sorensen KJ Bigbee WL Boice JD jr Thyroid nodularity and cancer among Chernobyl cleanup workers from Estonia Radiat Res 1997 147 225 235 9008215
Ivanov VK Tsyb AF Gorsky AI Makysutov MA Rastopchin EM Konogorov AP Bityujov AP Matyash VA Mould RF Thyroid cancer among 'liquidators' of the Chernobyl accident Br J Radiol 1997 70 937 941 9486071
Kikuchi S Perrier ND Ituarte P Siperstein AE Duh QY Clerk OH Latency period of thyroid neoplasia after radiation exposure Ann Surg 2004 239 536 43 15024315 10.1097/01.sla.0000118752.34052.b7
Knudsen N Laurberg P Perrild H Bulow I Ovesen L Jorgensen T Risk factors for goiter and thyroid nodules Thyroid 2002 12 879 88 12487770 10.1089/105072502761016502
Laurberg P Nohr SB Pedersen KM Hreidarsson AB Andersen S Bulow Pedersen I Knudsen N Perrild H Jorgensen T Ovesen L Thyroid disorders in mild iodine deficiency Thyroid 2000 10 951 63 11128722
Leenhardt L Hejeblum G Franc B Du Pasquier L Delbot T Le Guillouzic D Menegaux F Guillausseau C Hoang C Turpin G Aurengo A Indications and Limits of Ultrasoud-Guided Cytology in the Management of Non palpable Nodules J Clin Endocrinol Metab 1999 84 24 28 9920057 10.1210/jc.84.1.24
Macchia V Mariano A Nasti A Pisano G Ciampolillo A Giorgino R Pagliara S Macchia PE Luppoli G Lombardi G Fenzi G Carenza iodica e gozzo endemico nell'Italia Meridionale Il patologo clinico 1996 5 86 93
Nadolnik LI Netsetskaia ZV Vinogradov VV Effect of long term exposure to low dose gamma irradiation on the rat thyroid status Radiats Biol Radioecol 2004 44 76 80 15060946
Niedzela M Kormann E Breborwicz D Freyster E Harasymczuk J Warzywoda Rolski M Breborwicz J A prospective study of thyroid nodular disease in children and adolescents in western Poland from 1996 to 2000 and the incidence of thyroid carcinoma relative to iodine deficiency and the Chernobyl disaster Pediatr Blood Cancer 2004 42 84 92 14752799 10.1002/pbc.10421
Pacini F Agate L Molinaro E Elisei R Pinchera A Thyroid diseases around Chernobyl: from autoimmune disease to malignant tumors International Congress Series 2002 1234 175 183 10.1016/S0531-5131(01)00606-9
Schlumberger M Thyroid Cancer After Irradiation Eur J Cancer 2001 37 S133 10.1016/S0959-8049(01)80978-3
Shaktarin VV Tsyh AF Stepanenko VF Orlov MY Kopecky KJ Davis S Iodine deficiency, radiation dose and the risk of thyroid cancer among children and adolescents in the Bryansk region of Russia following the Chernobyl power station accident Int J Epidemiol 2003 32 584 91 12913034 10.1093/ije/dyg205
Shibata Y Yamashita S Masayakin VB Panasyuk GD Nagataki S 15 years after Chernobyl: new evidence of thyroid cancer Lancet 2001 358 1965 66 11747925 10.1016/S0140-6736(01)06971-9
Tan GH Gharib H Thyroid incidentalomas: management approaches to nonpalpable nodules discovered incidentally on thyroid imaging Ann Intern Med 1997 126 226 31 1 9027275
Verger P Catelinois O Tirmarche M Cherie-Chaline L Pirard P Colonna M Hubert P Thyroid cancers in France and the Chernobyl accident: risk assessment and recommendations for improving epidemiological knowledge Health Physiol 2003 85 323 9 10.1097/00004032-200309000-00008
Vermiglio F Volnova E Lo Presti VP Moleti M Violi MA Artemisio A Trimarchi F Post-Chernobyl increased prevalence of humoral thyroid autoimmunity in children and adolescents from a moderately iodine-deficient area in Russia Thyroid 1999 9 781 6 10482370
Violante FS Romano P Bonfiglioli R Lodi V Missere M Mattioli S Raffi GB Lack of association between occupational radiation exposure and thyroid nodules in healthcare personnel Int Arch Occup Environ Health 2003 76 529 532 12851827 10.1007/s00420-003-0443-8
Westfall PH Tobias Rd Rom D Wolfinger RD Hochberg Y Multiple Comparison and Multiple Testing Using the SAS System 1999 Cary, NC: SAS Institute Inc
|
16000179
|
PMC1185548
|
CC BY
|
2021-01-04 16:28:54
|
no
|
BMC Public Health. 2005 Jul 7; 5:73
|
utf-8
|
BMC Public Health
| 2,005 |
10.1186/1471-2458-5-73
|
oa_comm
|
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-751601879810.1186/1471-2458-5-75Research ArticleAre alcoholism treatments effective? The Project MATCH data Cutler Robert B [email protected] David A [email protected] Department of Psychiatry and Behavioral Sciences, University of Miami School of Medicine, Miami, Florida, USA2005 14 7 2005 5 75 75 2 9 2004 14 7 2005 Copyright © 2005 Cutler and Fishbain; licensee BioMed Central Ltd.2005Cutler and Fishbain; 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
Project MATCH was the largest and most expensive alcoholism treatment trial ever conducted. The results were disappointing. There were essentially no patient-treatment matches, and three very different treatments produced nearly identical outcomes. These results were interpreted post hoc as evidence that all three treatments were quite effective. We re-analyzed the data in order to estimate effectiveness in relation to quantity of treatment.
Methods
This was a secondary analysis of data from a multisite clinical trial of alcohol dependent volunteers (N = 1726) who received outpatient psychosocial therapy. Analyses were confined to the primary outcome variables, percent days abstinent (PDA) and drinks per drinking day (DDD). Overall tests between treatment outcome and treatment quantity were conducted. Next, three specific groups were highlighted. One group consisted of those who dropped out immediately; the second were those who dropped out after receiving only one therapy session, and the third were those who attended 12 therapy sessions.
Results
Overall, a median of only 3% of the drinking outcome at follow-up could be attributed to treatment. However this effect appeared to be present at week one before most of the treatment had been delivered. The zero treatment dropout group showed great improvement, achieving a mean of 72 percent days abstinent at follow-up. Effect size estimates showed that two-thirds to three-fourths of the improvement in the full treatment group was duplicated in the zero treatment group. Outcomes for the one session treatment group were worse than for the zero treatment group, suggesting a patient self selection effect. Nearly all the improvement in all groups had occurred by week one. The full treatment group had improved in PDA by 62% at week one, and the additional 11 therapy sessions added only another 4% improvement.
Conclusion
The results suggest that current psychosocial treatments for alcoholism are not particularly effective. Untreated alcoholics in clinical trials show significant improvement. Most of the improvement which is interpreted as treatment effect is not due to treatment. Part of the remainder appears to be due to selection effects.
==== Body
Background
A fundamental belief of addiction treatment is that therapy is effective. Addiction counselors are encouraged to use methods that have been shown to be effective in high quality clinical trials [1]. Three of the best of those methods were selected for Project MATCH, a large multicenter US trial designed to match the most effective treatment to individual patient characteristics. The criteria used to select the three treatments used in MATCH included the following: demonstrated clinical effectiveness; applicability to existing treatment programs and client populations, and distinctiveness from each other [2].
Project MATCH took great care to assure that the therapy was of the highest quality. Therapy was manualized and the three manuals [3-5] were organized into specific treatment sessions. Cognitive Behavioral Therapy (CBT) focused on handling thoughts about alcohol, dealing with urges, refusing drinks, avoiding situations that might lead to relapse, etc. Motivational Enhancement Therapy (MET) provided structured feedback about alcohol-related problems, and attempted to motivate commitment to change, to increase individual responsibility, and to enlist personal resources. Twelve Step Facilitation (TSF) was based on principles of Alcoholics Anonymous and it introduced the first three steps of AA and promoted active participation in AA. Therapists were required to have a Masters degree or Certified Addiction Counselor degree, a commitment to the particular therapeutic approach that they would provide (i.e., CBT, MET, or TSF), and at least two years experience. Therapist training was centralized at the coordinating center using seminars, required two supervised training cases, and also included some ongoing supervision of all sessions. Therapists taped their sessions with clients, and the tapes were scored at the coordinating center.
In some ways, the results of the MATCH clinical trial were disappointing. At the time it was concluded, in the late 1990s, it was the one of the most expensive clinical trials ever undertaken, costing 27 million dollars; it was conducted by the most seasoned alcoholism professional investigators, and it was designed to validate the top "cutting-edge" findings which had accumulated the strongest experimental support. Some 504 hypotheses were tested [6]. The final results did not support the hypotheses. There were essentially no matches between the therapeutic treatments and the participants above the level of random probability [6]. An analysis of the problem suggested that too many Type I errors were being made in the alcoholism literature [7]. Type I errors typically occur when an inappropriately large number of statistical tests are performed.
In announcing the disappointing MATCH results, the director of the National Institute of Alcohol Abuse and Alcoholism stated "All three treatments evaluated in Project MATCH produced excellent overall outcomes" [8]. That position was given scientific weight most notably by William Miller and colleagues in their paper "How effective is alcoholism treatment in the United States?" [9]. They suggested that treatment is extremely effective by way of presenting the MATCH results, and results of other large multi-site alcoholism trials. Table 1 shows outcome data abstracted from that paper. In addition to MATCH, the studies consisted of an analysis conducted by the Rand Corporation of treatment programs around the U.S. [10], a study at Veteran's Hospitals that compared disulfiram (Antabuse) to placebo [11], a VA study comparing lithium to placebo [12], a study on relapse [13], and a study comparing 12-step to cognitive-behavioral treatment [14]. Table 1 shows that the MATCH outcomes were very similar to the outcomes of the other studies.
Table 1 Alcoholism treatment 12 month outcome data cited by Miller to suggest that treatment is effective.
Multi-center trial Percent days abstinent Drinks per drinking day
Lithium/Placebo 83
Disulfiram/Placebo 84
Relapse prevention 85 5
MATCH 77 5
12 Step/Cognitive 8
Rand 5
Mean 82 6
The effectiveness of psychosocial therapy for alcoholism is being challenged in a number of different ways. Evidence is accumulating that extensive therapy may be no more effective that brief intervention [1,15]. Brief interventions are minimal types of therapies that can consist of simple expressions of concern about drinking delivered by a MD in a hospital trauma unit. There is a growing literature on "natural recovery" showing that many, if not most, individuals with serious alcohol consumption problems are able to recover without treatment [16]. A recently published meta-analysis [17] reported a significant improvement in untreated alcoholics enrolled in clinical trials. And there have been a few published trials that have concluded that therapy is not particularly effective [18,19].
The present study reports analyses of some overlooked data from Project MATCH. The overall relationship between treatment quantity (number of treatment sessions attended) and drinking measures was analyzed. Next, primary outcome data of participants who dropped out of treatment before receiving any therapy was compared to that of participants who attended each and every session for the full 12 weeks of therapy. Additionally we identified one anomalous group consisting of those participants who attended only one session of therapy. The data from this unusual group provides additional clues that help in interpreting the findings.
Methods
The MATCH Data set was made available to qualified researchers after the study had been completed. As an NIAAA committee member designing another large multi-center funded alcoholism study, the senior author of this study (RBC) obtained an official copy of the Project MATCH Public Data Set (V1.0) from the Coordinating Center in Farmington, Connecticut on August 24, 1998. The trial was conducted with approval from the appropriate ethics committees, with informed written consents, and the procedure was in compliance the Helsinki Declaration. This data set comprised some 256 variables on all 1726 participants in Project MATCH. Drinking data were summarized for each participant at pre-treatment, at weekly intervals for the 12 weeks of the trial, and at monthly intervals during the follow-up. Drinking data consisted of two measures, percent days abstinent (PDA) and drinks per drinking day (DDD). Participants were identified by which of the three treatments they received, Cognitive Behavioral (CBT), Motivation Enhancement (MET) or Twelve-Step (TSF). The number of treatment sessions each participant attended was also included. It is this variable that is the focus of this study.
The first step of the analysis was to compare our data to official published results. Pearson correlations were computed between number of treatment sessions attended with percent days abstinent and with drinks per drinking days at follow-up. We then computed correlations between number of sessions and the drinking variables during treatment. Both our analyses and previously published MATCH analyses substituted transformed scores for the drinking measures in order to normalize the data and thereby meet the assumptions of the statistical tests. The transformed scores were supplied in the data set. In this study we used the transformed scores for all statistical tests and the original drinking measures for display purposes.
In the next step the participants were categorized into groups based on the number of treatment sessions they attended, ranging from 0 to 12. Those alcoholics who signed up for the study but who never attended any sessions were coded zero. Those who were coded 12 attended every session and received either CBT or TSF but not MET. MET consisted of only 4 sessions. The outcome variables analyzed in this study are restricted to the two primary MATCH outcome variables.
Three groups were formed consisting of participants who attended either 0, 1 or all 12 treatment sessions. Chi square tests were used to test relationships between categorical variables. Repeated measures Manovas were used to test the differences between groups over all time points. Drinking outcome at follow-up was a mean of the entire follow-up period (from month 4 to month 15). The transformed scores were used to compute effect size. Paired t-tests were used to test for change within groups and independent groups t-tests were used to test differences between groups. Drinking data were incomplete for a number of the participants in the 0 and 1 session treatment groups. The highest level of missing data occurred at week one, where data were available for 57% of the 0 session group and 73% of the 1 session group. By follow-up, data were available for about 80% of participants in both these dropout groups.
Results
Table 2 compares our results to the Project MATCH's published results [20] on the correlations between number of treatment sessions attended (0 to 12, or 0 to 4 for MET) and the two primary outcome measures, drinks per drinking day and percent days abstinent. The table shows that our results are essentially identical to the official results. This indicates that the data used in the present study are correct. Additionally the results show rather low correlations between number of treatment sessions attended and outcome, particularly long term outcome. The percent of variance explained can be obtained by squaring the correlation coefficient. The amount of outcome that can be attributed to attending treatment ranges from 0 to 9%, with a median of approximately 3%.
Table 2 A comparison of our results to the results published by the official MATCH study group shows correlations between number of treatment sessions attended and drinking outcome.
Percent Days Abstinent
Our results Published MATCH results
Month CBT* MET TSF CBT MET TSF
6 .2556 .0269 .2678 .26 .03 .27
9 .1743 .0135 .2666 .17 .01 .27
12 .1685 .0011 .2657 .17 .00 .27
15 .1111 .0583 .2162 .11 .06 .22
Drinks per drinking day
Month CBT MET TSF CBT MET TSF
6 -.2281 -.0969 -.2832 -.22 -.10 -.28
9 -.2112 -.0856 -.2875 -.21 -.09 -.29
12 -.1372 -.0425 -.3028 -.14 -.04 -.30
15 -.1257 -.0795 -.2073 -.13 -.08 -.21
* CBT = Cognitive Behavioral Therapy
MET = Motivational Enhancement Therapy
TSF = Twelve Step Facilitation Therapy
Table 3 displays the same type of correlations as in Table 2, but this time the drinking measures were for the weeks during treatment. The table shows that there is a relationship between drinking level at these early time points and number of treatment sessions. Note that drinking level at week one predicts the total number of weeks the participant will remain in treatment.
Table 3 Correlations between number of treatment sessions attended and drinking during the course of the study.
Percent Days Abstinent Drinks Per Drinking Day
Week CBT* MET TSF CBT MET TSF
1 .207 .107 .278 -.232 -.150 -.251
2 .279 .131 .301 -.311 -.183 -.242
3 .302 .115 .373 -.340 -.152 -.309
4 .342 .108 .375 -.359 -.131 -.303
5 .356 .157 .407 -.361 -.183 -.344
6 .371 .165 .373 -.346 -.156 -.304
7 .354 .184 .391 -.358 -.209 -.346
8 .345 .181 .375 -.326 -.203 -.326
9 .352 .140 .397 -.339 -.175 -.375
10 .400 .149 .378 -.384 -.154 -.331
11 .399 .136 .418 -.432 -.156 -.395
12 .395 .121 .436 -.404 -.160 -.413
* CBT = Cognitive Behavioral Therapy
MET = Motivational Enhancement Therapy
TSF = Twelve Step Facilitation Therapy
Tables 4 and 5 display the mean follow-up data for the two primary drinking outcomes (percent days abstinent and drinks per drinking day) of Project MATCH. Outcome at follow-up is a mean of the data from 4 to 15 months. Overall the data show that the three treatments were fairly equal and that patients who attended more sessions had somewhat better outcomes. However, there was one anomalous group. Those who dropped out after one session (the 1 treatment group) had worse outcomes than those who dropped out before attending even one session. They had worse scores than this zero treatment group on 46 of 50 measures (percent days abstinent and drinks per drinking day at pre-treatment, 12 weeks of scheduled therapy, and 12 follow-up time points). All the data are shown in the attached file [see Additional file 1]. They also had worse scores than the other treatment groups (those who attended 2 to 12 sessions) on 526 of 550 measures. The one treatment group was significantly worse that the zero treatment group over all time points for both percent days abstinent (F = 4.73 (1, 130) p = .031) and drinks per drinking day (F = 4.51 (1, 130) p = .036).
Table 4 Percent days abstinent at follow-up.
Total Number of
Sessions Participant
Received Therapy Cognitive
Behavioral
(CBT) % Twelve-Step
Facilitation
(TSF) % Motivation
Enhancement
(MET) % Mean of All
0 .68 .69 .75 .72
1 .64 .61 .70 .64
2 .74 .69 .73 *
3 .68 .63 .72 *
4 .50 .76 .77 *
5 .79 .69 .72
6 .78 .82 .80
7 .73 .72 .73
8 .77 .63 .70
9 .77 .78 .78
10 .74 .86 .80
11 .83 .85 .84
12 .85 .87 .86
Note: Means were calculated for total sample available at follow-up.
* Not calculated since MET was restricted to 4 sessions
Table 5 Drinks per drinking day at follow-up.
Total Number of
Sessions Participant
Received Therapy Cognitive
Behavioral
(CBT) % Twelve-Step
Facilitation
(TSF) % Motivation
Enhancement
(MET) % Mean of All
0 6.54 7.42 7.71 7.32
1 8.22 8.66 5.98 7.84
2 6.03 6.55 6.91 *
3 9.25 7.37 5.96 *
4 7.74 4.78 4.71 *
5 3.78 6.62 5.65
6 7.30 4.19 5.51
7 5.90 6.50 6.19
8 4.70 6.00 5.35
9 5.04 5.69 5.40
10 6.14 3.72 4.95
11 4.50 2.97 3.78
12 3.23 3.18 3.21
Note: Means were calculated for total sample available at follow-up.
* Not calculated since MET was restricted to 4 sessions
Next, in order to better illustrate what is happening in these data, mean drinking levels of participants who dropped out before receiving any treatment (N = 100 at baseline) is compared to those who attended all 12 sessions of Cognitive Behavioral or Twelve-Step Therapy (N = 355 at baseline). Also included are the anomalous group identified above, i.e., the group that dropped out after attending only one treatment session (N = 121 at baseline). We can dispense with displaying the results for groups 2 through 11 without losing any information because of the linear relationship between number of treatment sessions and drinking level shown in Tables 2 and 3. Scores for these other groups, who attended between 2 and 11 treatment sessions, fell in a linear manner between the 0 and the 12 session groups (as shown in Tables 4 and 5 and the Additional File 1).
There was no relationship between membership in the groups (0, 1 or 12) and inpatient or outpatient status (Chi square = .06 (2) p = .966) or gender (Chi square = .43 (2) p = .805). However there was a strong relationship with treatment site (Chi square = 93.90 (20) p < .0001). Some sites had larger number of participants dropping out before treatment, some had larger numbers dropping out after one treatment, and some had larger numbers attending the full 12 sessions. Some of this difference may have been due to different procedures at different sites, some may have been due to characteristics of the therapists or participants.
Figures 1 and 2 show pre-treatment and follow-up outcome data for the 0, 1 and 12 session groups. The figures shows large improvements in percent days abstinent (Figure 1) and drinks per drinking day (Figure 2) for participants who attended either 0, 1 or all 12 treatment sessions. The improvement in both measures is highly significant for all groups (Table 6). Participants who received either 0, 1 or 12 treatment sessions are displayed. Shown are the raw score means and the statistical tests and effect size estimates which were calculated using the Project MATCH transformed scores. Participants in the 0 and 1 treatment session groups dropped out and received little or no treatment.
Figure 1 Percent days abstinent at follow-up. Percent days abstinent at pre-treatment and follow-up for patients who received 0, 1 or 12 treatment sessions. The 0 treatment dropout group showed great improvement from pre-treatment to follow-up. The 1 session attendance dropout group had worse scores than did the 0 treatment group. These data suggest that most of the improvement in the full 12 session attendance group cannot be due to treatment. See also Figure 2.
Figure 2 Drinks per drinking day at follow-up. Drinks per drinking day at pre-treatment and follow-up for patients who received 0, 1 or 12 treatment sessions. The 0 treatment dropout group showed great improvement from pre-treatment to follow-up. The 1 session attendance dropout group had worse scores than did the 0 treatment group. These data suggest that most of the improvement in the full 12 session attendance group cannot be due to treatment. See also Figure 1.
Table 6 Change in primary outcomes for various periods of the study for participants who dropped out early (0 or 1 week) and for those who received full treatment (12 weeks).
Total Improvement
Percent days abstinent
Group Pretreatment Follow-up t= df p= Effect
0 33 72 1.95 85 .000 1.498
1 29 64 9.30 105 .000 1.320
12 30 86 31.74 354 .000 2.042
Drinks per drinking day
0 17.9 7.3 1.75 85 .000 1.534
1 22.8 7.6 11.99 105 .000 1.552
12 15.7 3.2 34.68 354 .000 2.494
Instantaneous Improvement (week 1)
Percent days abstinent
Group Pretreatment Week 1 t= df p= Effect
0 32 81 9.17 56 .000 1.981
1 30 69 8.43 87 .000 1.447
12 30 92 35.30 339 .000 2.317
Drinks per drinking day
0 18.7 3.8 1.75 56 .000 2.097
1 22.5 8.7 9.69 87 .000 1.770
12 15.6 1.5 35.28 339 .000 2.909
Acute Treatment Phase (weeks 1 & 12)
Percent days abstinent
Group Week 1 Week 12 t= df p= Effect
0 81 81 .19 56 .852 -.025
1 69 63 1.69 87 .098 -.174
12 92 96 2.82 339 .005 .155
Drinks per drinking day
0 3.8 5.0 1.08 56 .283 -.172
1 8.7 8.8 .82 87 .415 -.090
12 1.5 1.1 2.05 339 .041 .126
The participants who dropped out after one session (the 1 session group) had worse scores than the other groups at pre-treatment as well as follow-up on both measures. Additionally, at baseline the 0 treatment group's scores were noticeably worse than the 12 treatment group's scores on drinks per drinking day (t = 3.26, 395 df, p = .001).
Effect sizes for all groups are also shown in Table 6. For percent days abstinence, those who dropped out before receiving any treatment had an effect size of 1.50, and those who attended all 12 sessions had an effect size of 2.04. For drinks per drinking day, there were effect sizes of 1.53 for the zero treatment group and 2.49 for the full treatment group. The estimated no-treatment effect sizes as proportions of the full treatment response was .73 (1.50/2.04) for percent days abstinent and .61 (1.53/2.49) for drinks per drinking day.
Figures 3 and 4 show the data for the 12 weeks of the treatment period. Improvement from pre-treatment to week 1 was statistically highly significant for all groups for both measures (Table 6). The estimated no-treatment effect sizes as proportions of the full treatment response was .85 (1.98/2.32) for percent days abstinent and .72 (2.10/2.91) for drinks per drinking day. Only the 12 session group showed significant improvement from week 1 to week 12 (Table 6).
Figure 3 Percent days abstinent during treatment. Percent days abstinent at pre-treatment and during the 12 weeks of treatment for patients who received 0, 1 or 12 treatment sessions. The effective improvement in drinking was instantaneous, evident at week 1. The improvement was maintained at the same approximate level for the 12 weeks of scheduled treatment. The effect occurred for all groups, whether they attended all 12 treatment sessions, only one treatment session, or did not attend any treatment sessions. See also Figure 4.
Figure 4 Drinks per drinking day during treatment. Drinks per drinking day at pre-treatment and during the 12 weeks of treatment for patients who received 0, 1 or 12 treatment sessions. The effective improvement in drinking was instantaneous, evident at week 1. The improvement was maintained at the same approximate level for the 12 weeks of scheduled treatment. The effect occurred for all groups, whether they attended all 12 treatment sessions, only one treatment session, or did not attend any treatment sessions. See also Figure 3.
Change from week one to follow-up showed deterioration for the zero and full treatment groups for both measures. There were no significant differences between the zero treatment dropout group and the full treatment group from week 1 to follow up in percent days abstinent (t = .25, 395 df, p = .800) or in drinks per drinking day (t = 1.22, 395 df, p = .222).
Discussion
The results suggest that treatment was not particularly effective. The following lines of evidence point to this conclusion. Correlations between treatment attendance and outcome were very small (as shown in Table 2). A median 3% of the variance in outcome might be attributed to treatment.
The correlations existed before most treatment occurred, at week 1 (Table 3). We would normally infer from the correlations in Table 2 that more treatment produces better drinking outcomes, but the Table 3 correlations suggests the reverse, that better drinking levels predict more treatment.
Nearly two thirds of the long term improvement in the full treatment group was matched by the untreated rapid dropout group (Figures 1 &2 and Table 6). Only in the remaining one third could there be a subcomponent consisting of a treatment effect.
Most of the improvement was instantaneous, occurring at week 1, before the participants had received the bulk of their treatment (Figures 3 &4 and Table 6). Although the full treatment group received 11 more therapy sessions, the additional improvement was of small magnitude. For example, at week one percent days abstinence had increased by over 60%, and the additional 11 weeks of treatment increased it by only 4%. If treatment were the causal agent we would expect that the effect would occur over the course of weeks with the administration of treatment.
There was a similar instantaneous improvement in untreated alcoholics (Figures 3 &4 and Table 6). The effect size estimates suggested that nearly three fourths of the instantaneous improvement in the full treatment group was matched by the untreated group.
Those who received zero treatment sessions had better outcomes than those who received one session (Figures 1 to 4 and Tables 4 to 6). The implication of this is discussed below.
Improvement was maintained over time even in the no treatment group (Figures 1 to 4 and Table 6). Change from week 1 to follow-up was not significantly different between the zero and full treatment groups. In both groups the week 1 to week 12 improvement was lost by follow-up. These data do not support the contention that retention of clients in treatment for as long as possible increases the chances that they will derive benefit from therapy.
A more reasonable interpretation of these data is that they illustrate the importance of selection effects, i.e., participants who reduce their alcohol consumption are more likely to enter or remain in treatment and those who continue drinking are more likely to drop out of treatment. One of the best studies of alcoholism treatment outcome was conducted by the Rand Corporation in the late 1970s [10]. Participants were patients who attended inpatient or outpatient treatment at centers across the United States. They found that "it is possible that the correlation [between attendance and outcome] arises from selection effects, such that the better motivated or more successful patients continue in treatment, whereas the more intractable cases drop out. Such a pattern could result from subject self-selection or from the operation of the treatment environment in encouraging continued participation for more responsive patients." (page 155). The likely selection effect in the current data was illustrated by the anomalous group participants who dropped out after attending only one treatment session as shown in Figures 1 and 2. Those who received zero treatment had better outcomes than those who received one session of treatment. Few would argue that this shows that treatment was harmful. A more likely explanation for this difference is some sort of self-selection. The higher drinking level of the 1 session dropout participants at baseline suggests that they may have been more dependent on alcohol than those in the 0 session dropout group. The relative higher level of dependence may have put these individuals under more pressure to do something about their drinking, explaining why they did not drop out prior to the first session. A similar logic could apply to the outcomes of the consistent attendees of the full treatment group. The likely selection effect is also shown in that participants with higher drinking levels at week 1 were more likely to drop out of treatment (Table 3).
The decreased drinking in both untreated and treated participants can be explained by a number of factors. One factor is that part of the effect is not real; many active alcoholics underreport drinking. Collateral informant interviews and other verification techniques are only partially effective in correcting the data. The Rand study [10], for example, found that 30% of the collateral informants were unable to provide information. Underreporting can make treatment appear more effective than it actually is.
Additionally, there are a number of non-treatment effects likely to result in reduced drinking [19,21-23]. In order to enter the trial participants had to first achieve a level of abstinence or reduced intake. If a participant arrives at a site in an intoxicated state immediate action is required by staff, such as admission to a detox unit, or detainment in the waiting room until the breath alcohol level returns to normal. These rules would have applied to each participant in Project MATCH at the time of enrollment and would have contributed to the rapid improvement seen in the week one data. The pre-study screening procedures used in clinical trials, both the overt criteria and the subjective criteria, are designed to select participants who are motivated to reduce their drinking. Enrolling in the trial suggests that the alcoholic has crystallized a decision to reduce or abstain from drinking. Once in the trial, the continued monitoring of drinking behavior by staff personnel may have both motivational and therapeutic benefits. For example, in one study with a 2 year follow-up [21], over half the participants indicated they liked the "caring, concern and help" follow-up telephone contact, and in another [24], the telephone interviewers reported that they usually entered in a sympathetic interaction with the study participants. Such positive empathetic contact could be of therapeutic benefit.
There are a number of limitations to these analyses. The data from two thirds of the subjects were not used in the illustration of mean drinking levels shown in the figures and in Table 6. However these data were used in Tables 2 though 5, and are presented in more detail in the attached data [see Additional File 1]. The linear relationship shown in Tables 2 and 3, and the means in Tables 4 and 5 indicated that no information was left out. Additionally, other analyses of these groups have been previously published, e.g., by the Project MATCH research group [20]. We chose to highlight several specific groups. The logic of selecting the group that received no treatment and the group that received all 12 sessions of treatment was clear – they offered an unambiguous treatment comparison. The data presented here show that the outcomes of the 12 session group were better than the outcomes of participants who received between 2 and 11 sessions, making the 12 session group a fair comparison. Additionally we identified an anomalous group, the one session rapid dropouts, and used that group in an attempt to interpret the data. The mathematical bases for the anomalous designation, and thus the selection of this group, were presented.
Analyses were primarily limited to descriptive and simple inferential statistics. This was done because the findings are likely to be extremely controversial. We have therefore presented results that are easily replicated, and easily understood.
Although the results are essentially negative, suggesting that current treatments are not effective, we do not offer suggestions for future directions. We feel we will have made a contribution if the data presented can be accepted as accurate. If they are accepted then implications for future research and treatment will naturally follow. For example, if the patient's motivations, opportunities, beliefs and hopes are the critical issues, how do we measure them? How do we influence them? How do they interact with the treatment environment?
It may be that pre-treatment patient characteristics (e.g., level of dependence, social support, etc.) have a large influence on both the number of treatment sessions attended and drinking outcome. However, even if this is true, it would not be evidence of treatment effectiveness. Only if one could show that positive prognostic factors were weighted heavily against the treatment attendees and in favor of the dropouts would these results be open to reinterpretation. The baseline and week one drinking data presented here do not support the likelihood of such a possibility. Additionally, there is no evidence in the literature to support the notion that, for example, alcoholics who lack social support are more likely to enter or remain in treatment. There are a large number of both positive and negative reasons why alcoholic participants drop out of clinical trials. Positive reasons include work commitments, pregnancy, re-location to another area and remission from drinking. Negative reasons include continued or increased drinking, abuse of other substances, attitude towards the clinical staff or environment, physical illness, hospitalization and incarceration.
Conclusions drawn from therapy delivered in clinical trials might not be applicable to therapy in other settings. We might well expect great differences in clinical effectiveness between different therapists, and between different treatment programs. However it can be argued that the large non-treatment effect seen in this study is present in other aggregated outcome studies published in the literatures. Miller and others [9] presented results from a number of such trials, in addition to Project MATCH. Table 1 summarized their findings for the two outcome variables studied here. The outcomes of the different studies are remarkably similar. The similarity in results would suggest that the non-treatment effect identified here may be present in all these studies.
The outcome variables in these analyses were the original primary MATCH outcome variables. We have been able to show that the analyses of these variables, and the treatment attendance variable, are in perfect concordance with published analyses of the Project MATCH Research group [20]. Over 60 publications have been generated by Project MATCH, but, to the best of our knowledge, all have overlooked the main finding of this study, i.e., the good outcomes of the zero treatment group when compared to the full treatment group and that the improvement in all groups occurred immediately after enrollment in the trial. Ineffective treatment would be the most parsimonious explanation for the rather surprising main findings of Project MATCH, that there was no match between patient characteristics and different types of treatment, and that all three treatments were equal.
There may be similarities between these results, for alcoholism patients, and effects seen in some other types of patients. Depressed patients sometimes report significant improvement after enrolling in clinical trials but before receiving therapy [25]. Recent time-course analyses in depression report sudden decreases in depression regardless of treatment condition [26]. These rapid responders were associated with better outcome at the end of the treatment and into follow-up [27].
It is difficult to compare the high quality follow-up data of Project MATCH to that in the alcoholism literature, much of which are collected under quite different circumstances. The zero treatment participants at the final follow-up interval (month 15) reported a mean of 25.1 drinks per week, with 45% (35/78) abstinent. These outcomes appear somewhat better than those recently summarized in the literature [17]. Of some 17 studies than included placebo or no treatment conditions, with and without prior detoxification, a mean (for studies) was 21% abstinent, and the average participant was drinking 31 drinks per week [17].
Exaggerated claims of treatment effectiveness can have undesirable consequences for patients, for therapists, and for science. Patients who fail an "effective" treatment may feel even more hopeless. This increased despair may be extremely deleterious in people with such life-threatening habits. Therapists may feel inadequate or frustrated with repeated failures. For science, exaggerated claims tend to shift focus into unproductive directions, and to obscure the pertinent facts that are necessary in order to move the science forward.
While this study shows that three of the best treatments currently available for addiction were not very effective, it remains likely that many severely dependent alcoholic individuals benefit from external help. By suggesting practical and helpful ways for dealing with the problems of addiction, therapy may help a patient regain a sense of control over his or her life. We are not suggesting that alcoholism treatment should be discontinued or even reduced. People with alcohol problems clearly need all the help our society can give them.
Conclusion
These results suggest that current psychosocial treatments for alcoholism are not particularly effective. The improvements in drinking appear to be due to selection effects. Alcoholics who decide to enter treatment are likely to reduce drinking. Those who decrease their drinking are more likely to remain in treatment. Widespread acceptance of these results would have a profound influence on alcoholism research and treatment because it would shift focus away from treatment components and toward patient characteristics and beliefs.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RBC conceived of the study, participated in drafting the manuscript and performed the statistical analysis. DAF participated in drafting the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Drinking outcomes for all groups at each time point. Shown are mean percent days abstinent and drinks per drinking day for all participants in the Project MATCH data set. Participants were categorized by number of treatment sessions attended. Participants in MET are not included in the 2, 3 and 4 treatment session means.
Click here for file
==== Refs
Miller WR Wilbourne PL Hettema JE Hester RK, Miller WR What works? A summary of alcohol treatment outcome research Handbook of Alcoholism Treatment Approaches 2003 3 Boston: Allyn and Bacon 13 63
Donovan DM Dadden RM DiClemente CC Carroll KM Longabaugh R Zweben A Rychtarik R Donovan D, Mattson ME Issues in the selection and development of therapies in alcoholism treatment matching research Alcoholism Treatment Matching Research; Methodological and Clinical Approaches Journal of Studies on Alcohol 1994 138 148
Nowinski J Baker S Carroll K Twelve Step Facilitation Therapy Manual 1999 NIH Publication No. 94-3722, Washington D.C
Miller WR Zweben A DiClemente CC Rychtarik RG Motivational Enhancement Therapy Manual 1999 NIH Publication No. 94-3723, Washington D.C
Kadden R Carroll K Donovan D Cooney N Monti P Abrams D Litt M Hester R Cognitive-Behavioral Coping Skills Therapy Manual 1995 NIH Publication No. 94-3724, Washington D.C
Longabaugh R Wirtz PW Longabaugh R, Wirtz PW Substantive Review and Critique Project MATCH Hypotheses: Results and Causal Chain Analyses 2001 NIH Publication No. 01-4238, Washington D.C 305 325
Moyer A Finney JW Elworth JT Kraemer HC Can methodological features account for patient-treatment matching findings in the alcohol field? J Stud Alcohol 2001 62 62 73 11271966
Bower B Alcoholics synonymous: Heavy drinkers of all stripes may get comparable help from a variety of therapies Sci News 1997 151 62 63
Miller WR Walters ST Bennett ME How effective is alcoholism treatment in the United States? J Stud Alcohol 2001 62 211 22 11327187
Polich JM Armor DJ Braiker HB The Course of Alcoholism Four Years after Treatment 1981 New York: John Wiley and Sons
Fuller RK Branchey L Brightwell DR Derman RM Emrick CD Iber FL James KE Lacoursiere RB Lee KK Lowenstam I Maany I Neiderheisser D Nocks JE Shaw S Disulfiram treatment of alcoholism; A Veterans Administration cooperative study JAMA 1986 256 1449 1455 3528541 10.1001/jama.256.11.1449
Dorus W Ostrow DG Anton R Cushman P Collins JF Schaefer M Charles HL Desai P Hayashida M Malkerneker U Willenbring M Fiscella R Sather MR Lithium treatment of depressed and nondepressed alcoholics JAMA 1989 262 1646 1652 2504944 10.1001/jama.262.12.1646
Lowman C Allen J Stout RL Replication and extension of Marlatt's taxonomy of relapse precipitants: Overview of procedures and results. The Relapse Research Group Addiction 1996 S51 S71 8997781 10.1080/09652149638809
Ouimette PC Finney JW Moos RH Twelve-step and cognitive-behavioral treatment for substance abuse: A comparison of treatment effectiveness J Consult Clin Psychol 1997 65 230 240 9086686 10.1037//0022-006X.65.2.230
Moyer A Finney JW Swearingen DE Vergun P Brief interventions for alcohol problems: a meta-analytic review of controlled investigations in treatment-seeking and non-treatment-seeking populations Addiction 2002 97 279 292 11964101 10.1046/j.1360-0443.2002.00018.x
Sobell LC Cunningham JA Sobel MB Recovery for alcohol problems with and without treatment: Prevalence in two population surveys Am J Public Health 1996 86 966 972 8669520
Moyer A Finney JW Outcomes for untreated individual involved in randomized trials of alcohol treatment J Subs Abuse Treat 2002 23 247 252 10.1016/S0740-5472(02)00264-7
Chick J Ritson B Connaughton J Steward A Chick J Advice versus extended treatment for alcoholism: A controlled study British J Addiction 1988 83 159 17
Edwards G Orford J Egert S Gutheir S Hawker A Hensman C Mitcheson M Oppenheimer E Taylor C Alcoholism: A controlled trial of treatment and advice J Stud Alcohol 1977 38 1004 1031 881837
Mattson ME Del Boca FK Carroll KM Cooney NL DiClemente CC Donovan D Kadden RM McRee B Rice C Rycharick RG Zweben A Compliance with treatment and follow-up protocols in project MATCH: Predictors and relationship to outcome Alcohol Clin Exp Res 1998 22 1328 1338 9756050
Sobel LC Sobel MB Frequent follow-ups and data gathering and continued care with alcoholics International Journal of the Addictions 1981 16 1077 1086 6281201
Jonson H Hermansson U Ronnberg S Gyllenhammar C Forsberg L Comments on brief interventions of alcohol problems: a review of a review Addiction 1995 90 1118 1119 7549782
Ogborne AC Annis HM The reactive effect of follow-up assessment procedures: an experimental study Addict Behav 1988 13 123 129 3369320 10.1016/0306-4603(88)90001-9
Maisto SA Sobell LC Sobell MB Sanders B Effects of outpatient treatment for problem drinkers Am J Drug Alcohol Abuse 1985 11 131 149 4061431
Fennell MJV Teasdale JD Cognitive therapy for depression: Individual differences and the process of change Cognitive Therapy and Research 1987 11 253 271 10.1007/BF01183269
Renaud J Brent DA Baugher M Birmaher B Kolko DJ Bridge J Rapid response to psychosocial treatment for adolescent depression: A two-year follow-up J Am Acad Child Adolesc Psychiatry 1998 37 1184 1190 9808930 10.1097/00004583-199811000-00018
Tang TZ DeRubeis RJ Sudden gains and critical sessions in cognitive-behavioral therapy for depression J Cons Clin Psychol 1999 67 894 904 10.1037//0022-006X.67.6.894
|
16018798
|
PMC1185549
|
CC BY
|
2021-01-04 16:28:58
|
no
|
BMC Public Health. 2005 Jul 14; 5:75
|
utf-8
|
BMC Public Health
| 2,005 |
10.1186/1471-2458-5-75
|
oa_comm
|
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-811607639610.1186/1471-2458-5-81Research ArticleHealth system outcomes and determinants amenable to public health in industrialized countries: a pooled, cross-sectional time series analysis Arah Onyebuchi A [email protected] Gert P [email protected] Diana M [email protected] Niek S [email protected] Department of Social Medicine, Academic Medical Center, University of Amsterdam, PO Box 22700, 1100 DE Amsterdam, the Netherlands2 Netherlands Institute for Health Sciences, Erasmus MC, PO Box 1738, 3000 DR Rotterdam, the Netherlands3 Center for Prevention and Health Services Research, National Institute of Public Health and the Environment (RIVM), PO Box 1, 3720 BA Bilthoven, the Netherlands4 Tranzo, Faculty of Social and Behavioural Sciences, Tilburg University, PO Box 90153, 5000 LE Tilburg, the Netherlands5 Netherlands Institute for Health Services Research (Nivel), PO Box 1568, Utrecht 3500 BN, the Netherlands2005 2 8 2005 5 81 81 9 2 2005 2 8 2005 Copyright © 2005 Arah et al; licensee BioMed Central Ltd.2005Arah 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 studies have tried to assess the combined cross-sectional and temporal contributions of a more comprehensive set of amenable factors to population health outcomes for wealthy countries during the last 30 years of the 20th century. We assessed the overall ecological associations between mortality and factors amenable to public health. These amenable factors included addictive and nutritional lifestyle, air quality, public health spending, healthcare coverage, and immunizations.
Methods
We used a pooled cross-sectional, time series analysis with corrected fixed effects regression models in an ecological design involving eighteen member countries of the Organisation for Economic Cooperation and Development during the period 1970 to 1999.
Results
Alcohol, tobacco, and fat consumption, and sometimes, air pollution were significantly associated with higher all-cause mortality and premature death. Immunizations, health care coverage, fruit/vegetable and protein consumption, and collective health expenditure had negative effects on mortality and premature death, even after controlling for the elderly, density of practicing physicians, doctor visits and per capita GDP. However, tobacco, air pollution, and fruit/vegetable intake were sometimes sensitive to adjustments.
Conclusion
Mortality and premature deaths could be improved by focusing on factors that are amenable to public health policies. Tackling these issues should be reflected in the ongoing assessments of health system performance.
==== Body
Background
Western populations currently enjoy unprecedented wealth and longevity [1]. It is held that medical care – or more broadly healthcare including public health – contributed immensely to the increased longevity. However, McKeown [2], McKinlay et al [3,4], Illich [5] and others have questioned the role of medical care in these gains. Others, like Mackenbach [6,7] and Bunker et al [8,9] contend that medical care has contributed reasonably to the mortality decline. Unfortunately, these debates have done little to further the course and relevance of health systems today. Resource constraints, poor quality, and the controversial World Health Report 2000 [10] have all led to the increased assessment of health systems in terms of effectiveness [11], safety [12], equity and responsiveness [10]. Meanwhile, the functioning of the public health aspect of health systems has received less attention. Functional public health systems are nowadays seen within the context of health promotion and integral social structures [13]. This focus on health promotion strategies that are embedded in societal structures has been described as public health coming "full circle" in the "new public health" era [14].
The new public health hopes to address major risk factors implicated in the global burden of disease. These factors, which threaten the substantial health gains made in the 20th century, include addictive behavior (such as tobacco smoking), nutritional lifestyle (e.g. fat consumption), degrading environmental quality (e.g. air pollution), and less-than-adequate public health investments, coverage and preventive interventions (e.g. immunizations) [15]. Despite the recent increase in studies that either looked primarily at such factors or controlled for them in their analyses [16-22], few studies have tried to assess the combined cross-sectional and temporal contributions of a more comprehensive set of amenable factors to population health outcomes for wealthy countries during the last 30 years of the 20th century.
Therefore, this study uses a pooled, cross-country time-series design to assess the ecological relationships of such amenable factors to population health, adjusting for demographic, national wealth, and medical care-related factors in selected industrialized countries during the period 1970–1999. Population health is a commonly used compound indicator of health system and social performance, and is captured in this study as all-cause mortality and all-cause potential years of life lost. We aim to speak to the relevance of public health-related issues, and their place in assessing health system performance.
Theoretical framework
Public Health, as an organized effort of society [23], espouses several principles, namely: (a) emphasis on collective responsibility and role of the state; (b) focus on whole populations; (c) emphasis on prevention; (d) concern for the underlying socio-economic determinants of health and disease; (e) multi-disciplinary approach (both quantitatively and qualitatively); and (f) partnerships with populations served [24]. These principles form a useful basis for evaluating the functioning of public health systems [25]. Simply put, medical care (with its emphasis on personal clinical services) and public health (with its emphasis on collective societal efforts for population health) represent the two traditional components of a health system [26].
Given public health's broad focus on population health, the theoretical framework for this study is based on a population health determinants model [27-31]. There are many determinants of population health which are commonly classified as either proximal or distal [18]. The proximal determinants have direct effects on health, and the distal determinants have indirect effects.
The proximal determinants, which act on both micro and macro levels, often include lifestyle or behavior (e.g. alcohol, fat, tobacco, fruit and vegetable consumption), and socioeconomic environment (including macro-economic measures such as wealth), demography (e.g. elderly proportion of the total population), physical environment (e.g. air pollution by oxides of sulphur, nitrogen or carbon) and host constitution. The health system, which also operates at this proximal level, shares an interface with other sectors of organized societies such as the social, political and economic systems. Health system inputs such as physicians and medical technology may be the result of intersectoral dynamics and social choices [32]. It is expected that public health systems can influence many of the proximal non-medical determinants and avert or minimize the need for expensive medical care.
Distal determinants of health include the national, institutional, political, legal, and cultural factors that indirectly influence health by acting on the more proximal factors, their interrelated mechanisms, levels, trends, and distributions. These distal factors are usually more stable than proximal determinants. Though we do not address distal factors in this study, we can roughly capture their potential impact on mortality over time by using dummy variables to account for any unmeasured time-dependent heterogeneity introduced by these distal determinants.
Methods
We use an ecological design – a pooled cross-sectional, time series analysis of secondary data [33,34] – to quantify the relationship between average population health and factors amenable to public health, taking into account demographic, medical, and macro-economic (that is, crude national wealth) determinants of health. The analysis is based on the country-year units of selected eighteen member countries of the Organisation of Economic Cooperation and Development (OECD) from 1970 to 1999.
Data and measures
The data for this study are derived from secondary sources [35,36]. The eighteen countries are a convenient sample of wealthy societies whose health experiences are different enough to allow for variation, but similar enough to support effective pooling. These countries, which were used in a recent study on primary care and mortality, are Australia, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Italy, Japan, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom, and the United States [18]. The data management and preparation follow the methods used in that study [18].
The outcome measures of health system performance used are age- and sex-standardized all-cause total mortality, and potential years of life lost (PYLL) from all-causes. Both types of measures, expressed per 100,000 population, are standardized to the 1980 OECD population. The all-cause PYLL is a measure of premature, but preventable death before age 70 years. PYLL is calculated by summing up deaths occurring at each age and multiplying this with the number of remaining years to live up to the selected 70-year limit. Both all-cause mortality and PYLL have been used in ecological studies of health outcomes [18,37,38].
The independent measures represent four convenient blocks of amenable factors: addictive behavior/lifestyle; nutritional lifestyle; environment, investment and coverage; and disease preventive measures. Alcohol (measured in liters per capita) and tobacco (in grams per capita) consumption represent the commoner addictive lifestyle factors used in studies [18]. Consumptions of fat, fruit/vegetable (both in kilograms per capita), and protein (measured in grams per capita per day) reflect nutritional lifestyle. The OECD data on nutritional factors come from the FAOSTAT database of the Food and Agriculture Organization of the United Nations [39].
Air pollution is measured as the emission of nitrogen oxide in kilograms per capita, and is expected to have negative effects on health and the environment [40]. Collective (public) health expenditure is a measure of national expenditure on prevention, public health services and health administration (excluding medical or personal care expenditure), and is hypothesized to have a negative effect on mortality. We quantified collective expenditure as a percentage of GDP to reduce collinearity. Although already at high levels in the selected OECD countries, healthcare coverage is a good structural variable for the percentage of the total population that can access public healthcare goods and services included in total public health expenditure, independent of the scope of cost sharing. Here, air pollution, collective health expenditure, and healthcare coverage represent the environment, public health investment and coverage.
Percentages of children reaching their first birthday who were immunized against measles and DTP (diphtheria, tetanus and pertussis) are used as measures of preventive functions of public health, though they could also serve as outcomes. However, we used two dummy variables to control for 1980s and 1990s fixed effects, and omitted variable bias and to eliminate any unobserved heterogeneity [34]. The 1970s served as reference.
To account for demographic, medical care, and macro-economic factors, we chose several well-documented measures. These were: percentage of the total population aged 65 year and above as a demographic measure (elderly); practicing physician density per 1,000 population and per capita doctor visits as measures of medical care input; and population standardized gross domestic product (GDP per capita) as a measure of national wealth and a proxy for medical expenditures [18]. We deflated the GDP per capita by using the constant 1995 US dollar.
Design and analysis
We employed a pooled cross-sectional time series design that entailed stacking the eighteen countries (also referred to as cross-sections) over time. This resulted in a combination of cross-sections and time series with a matrix configuration that considered variation between cross-sections before variation within cross-sections over time. The obvious advantage of using a pooled ecological design is that it increases statistical power. We used fixed effects regression models and robust statistical modeling techniques to overcome repeated measure biases, correlated errors and heterogeneity [41-44]. The final regression model is specified as follows:
Ynt = γt + β0 + βkXknt + εnt
where
n = 1....18 countries
t = 1....T time points (calendar years from 1970 through 1999)
k = 1....K number of independent variables
γt = set of time effect dummy variable(s)
β0 = constant
βk = pooled regression estimates of the effect of each independent variable
Xknt = independent variables per country for each unit year
Unfortunately, simply because the above regression equation includes both stochastic and non-stochastic variables, the expected value of the error term is not zero and the variance is not constant. By assuming all the coefficients (the βk's) to be the same for each cross-section within the regression model, we compound the problems of heteroscedasticity (that is, non-constant variances) or autoregression (that is, decaying variance due to correlated error terms over time). Autoregression can only be addressed after heteroscedasticity has been corrected for, since both anomalies cannot be visualized at the same time [33]. For pooled data with correlated errors, the ordinary least squares (OLS) method does not yield the correct standard errors on which to base the hypothesis or relationship testing under study. To correct for heteroscedasticity in this sample, we employ heteroscedasticity-consistent standard error (HCSE) estimators which use the square of the residuals from the OLS equation to approximate the variance-covariance structure of the regression estimates [41-44]. Since most of these estimators have not been fully implemented in routine statistical software packages, we use a special HCSE macro to model the heteroscedasticity-consistent covariance matrix [44]. Autoregression is then addressed via a quasi-differencing technique [34].
Alternatively, the pooled cross-sectional time series data could be treated as repeated observations on each cross-section, although strictly speaking, here, they represent average national characteristics per unit time [34]. Similar results are obtained for this study by using the Linear Mixed Models procedure of SPSS (version 12.0.2, SPSS Inc., IL, March 2004) and the SAS PROC MIXED (version 8.2, SAS Institute Inc, NC, 2001) procedure to model the means, variances and covariances of this data as repeated measures. However, these general softwares implement a heteroskedastic-consistent matrix method which uses conditional variance of the error, rather than the more robust methods which we used to adjust each squared residual by a function of how deviant the pattern of independent variables of each cross-section is [43].
We used five pre-specified nested models to address the question 'after adjusting for demographic, medical care input, and wealth, can mortality and premature death be still explained by factors amenable to public health?' The first four models used measures of addictive behavior/lifestyle; nutritional lifestyle; environment, investment and coverage; and preventive measures. Model 1 examined the effects of tobacco and alcohol on mortality. Model 2 extended model 1 to include fat, fruit/vegetable and protein consumption. Model 3 added air pollution, collective health expenditure, and health care coverage. In model 4, we included measles and DTP immunizations. Finally, model 5 further adjusted for time fixed effects using per decade time dummies with the first decade (1970s) as reference [34]. We also re-estimated models 3 to 5 after excluding the United States from the healthcare coverage variable, given the known lower levels of healthcare coverage in the American population. In addition, we re-ran these models excluding the healthcare coverage variable entirely in order to gauge its impact on the models considering that healthcare coverage also reflects access to care in general, not just public health services. We present statistics for hypothesis testing, model improvement and the proportion of total variance (as adjusted R squared) in outcomes explained in the final models, along with the regression estimates and their modeled standard errors.
Results
Between 1970 and 1999, all-cause mortality and PYLL decreased on the average by approximately 27% and 37% respectively across the selected countries (Table 1). Protein and fruit/vegetable consumption increased while alcohol, tobacco, and fat intake decreased, albeit with substantial variation across countries. Fat intake actually increased in Greece, Italy, Japan, Portugal and Spain by about 0.2 to 2% annually [35]. Immunization levels and collective health expenditure also increased. The healthcare coverage levels improved by about 2 to 4 percentage points when the United States was excluded from the pool. There were also substantial increases in the elderly population, national wealth, density of practicing physicians, and doctor visits during the 1970–1999 period.
Table 1 Descriptive statistics for variables used in the pooled cross-sectional time series analysis
Variables 1970–79 1980–89 1990–99
Mean Standard deviation Mean Standard deviation Mean Standard deviation
All-cause mortality (both sexes per 100,000 population) 938.25 113.74 789.11 79.79 687.65 83.87
Potential years of life lost (before age 70 years per 100,000 population) 7009.15 1449.67 5372.07 891.38 4424.51 798.37
Tobacco (grams per capita) 2,699.88 450.32 2411.38 455.98 2052.00 502.29
Alcohol consumption (liters per capita) 11.62 3.52 11.33 2.94 9.98 2.27
Fat consumption (kilograms of butter per capita per year) 5.42 3.76 4.90 3.30 3.56 2.42
Fruit & vegetable consumption (kilograms per capita per year) 184.69 61.97 200.14 68.71 220.75 71.03
Protein consumption (grams per capita per day) 93.48 8.03 98.62 7.81 103.06 7.52
Pollution (Nitrogen oxide emission, in kilograms per capita) 41.50 16.29 43.77 20.72 44.71 26.59
Collective health expenditure (% GDP) 0.45 0.24 0.50 0.25 0.53 0.32
Healthcare coverage (% population) [without USA] 90.49 [92.2] 13.69 [12.10] 94.16 [96.75] 13.40 [7.10] 93.84 [97.86] 17.61 [5.96]
DTP immunization (% children) 86.61 12.38 87.09 10.56 91.09 8.71
Measles immunization (% children) 59.23 9.57 72.67 12.07 86.58 11.19
Elderly (percentage of population over 65 years) 11.58 2.09 13.09 1.93 14.73 1.64
Physician density (per 1,000 population) 1.58 0.32 2.29 0.45 2.87 0.58
Doctor visits (per capita) 4.08 1.19 5.39 2.17 6.2 2.55
GDP per capita (in constant 1995 US dollars) 17,750.94 7,029.03 21,543.59 8,110.65 25,783.31 9,306.90
Table 2 details the regression results for all-cause mortality. Here, tobacco was positively associated with all-cause mortality in models 1 to 3, but was not significant in models 4 and 5 that included immunization variables and time dummies. Alcohol was significantly and positively associated with mortality in all the models, yielding mortality increases of 6.6 to 8.6 per 100,000 populations for every one-liter increase in per capita alcohol consumption. Fat consumption also showed a strong positive relationship to mortality in all cases. Fruit/vegetable consumption was only negatively related to mortality in the full model and when the model excluded environmental, collective health spending, coverage and immunizations. Protein showed a stable negative association in all models, changing only slightly in model 3. Air pollution was hardly related to mortality in all models that included healthcare coverage. Collective health expenditure, healthcare coverage, and immunizations were all negatively associated with higher mortality (P < 0.01).
Table 2 Regression estimates from the pooled cross-sectional time series analysis of all-cause mortality per 100,000 population
Variables Model 1 Estimate (S.E.) Model 2 Estimate (S.E.) Model 3 Estimate (S.E.) Model 4 Estimate (S.E.) Model 5 Estimate (S.E.)
Constant 949.32*** (45.75) 1,256.36*** (64.55) 1,487.51*** (80.04) 1,548.45*** (91.38) 1,491.87*** (93.78)
Tobacco 0.03*** (0.01) 0.03*** (0.01) 0.02** (0.01) 0.01 (0.01) 0.01 (0.01)
Alcohol 8.59*** (1.49) 8.59*** (1.50) 6.56*** (1.46) 8.03*** (1.46) 8.61*** (1.52)
Fat consumption 9.44*** (1.66) 9.83*** (1.91) 9.83*** (1.91) 8.46*** (1.96)
Fruit/vegetable consumption -0.20* (0.08) -0.03 (0.09) -0.10 (0.09) -0.19* (0.09)
Protein consumption -3.22*** (0.57) -4.25*** (0.70) -3.93*** (0.76) -3.52*** (0.72)
Air Pollution 0.13 (0.20) 0.46 (0.25) P = 0.07 0.48 (0.26) P = 0.06
Collective health expenditure -72.44*** (18.02) -68.98*** (18.08) -48.95* (19.13)
Healthcare coverage -1.89*** (0.30) -1.73*** (0.31) -1.44*** (0.32)
DTP immunization -1.39*** (0.31) -1.39*** (0.31)
Measles immunization -0.88** (0.27) -0.88** (0.27)
Elderly 4.55* (2.26) 1.47 (2.04) 0.60 (2.19) 4.83* (2.19) 4.87* (2.12)
Physician density -103.65*** (6.34) -47.81*** (7.52) -49.04*** (8.07) -20.57* (8.07) -3.92 (8.74)
Doctor visits -15.55*** (1.75) -9.57*** (1.70) -4.90** (1.80) -5.31** (1.71) -3.60* (1.68)
GDP per capita -0.003*** (0.0004) -0.006*** (0.0005) -0.006*** (0.0005) -0.005*** (0.0005) -0.005*** (0.0005)
1980s fixed effects++ -60.78*** (9.93)
1990s fixed effects++ -83.04*** (14.11)
F change - 38.44# 15.03# 26.57# 17.32#
Adjusted R2 0.57 0.66 0.79 0.72 0.74
S.E.: standard error of estimate; *P < 0.05; **P < 0.01; ***P < 0.001; ++decade time effects to eliminate omitted-variable bias due to unobserved heterogeneity with reference to 1970s baseline; #significantly better than preceding model and model 1 (P < 0.000); n = 451
Exclusion of the United States data on healthcare coverage did not substantially alter the models, nor did change its association with mortality. Although the total exclusion of the healthcare coverage variable did not improve the fit of the models (not reported in Table 2), it significantly changed the mortality associations of tobacco, fruit/vegetable consumption, and air pollution. Regression coefficients for tobacco changed from 0.02 to 0.06 (P < 0.001) and 0.01 to 0.02 (P < 0.0) in only models 3 and 4 respectively. For fruit/vegetable consumption, the estimates changed from -0.03 to -0.37 (P < 0.001) and -0.10 to -0.25 (P < 0.01) in only models 3 and 4 respectively. For air pollution, the estimates changed from 0.13 to 1.40 (P < 0.001), 0.46 to 0.79 (P < 0.01) and 0.48 to 0.74 (P < 0.01) in models 3, 4 and 5 respectively. GDP per capita remained strongly negatively associated with mortality in all models. Except for the elderly proportion, all medical input adjustment covariates were significantly associated with lower mortality. The total explained variance in all-cause mortality ranged from 57% in model 1 to 74% in the fully adjusted model 5.
Table 3 summarizes the corrected fixed effects regression results for age- and sex-standardized all-cause PYLL, a measure of premature death. Five stepwise contemporaneous adjusted models are presented, with each model compared to the preceding one and model 1 for improvements. Again, tobacco was significantly associated with higher PYLL in all models except those corrected for immunization and time effects. Alcohol and fat intake were positively related to higher PYLL. Fruit/vegetable intake was only negatively associated with PYLL in model 3 which did not correct for immunization and time effects. Air pollution tended to be significantly and positively associated with premature death in the fuller models (P < 0.001). Protein, collective health expenditure, healthcare coverage, and immunizations all strongly accounted for lower PYLL. As for total all-cause mortality, excluding the United States data on healthcare coverage did not substantially alter our estimates for factors associated with PYLL. A total exclusion of the healthcare coverage variable improved the regression coefficients of air pollution from 2.92 to 20.14 (P < 0.001) in model 3, and fruit/vegetable consumption from -0.21 to -2.21 (P < 0.05) in model 5 only. For the premature death outcome, the regression coefficients varied more in strength than was seen in Table 2 for all-cause mortality. The full, time-adjusted model accounted for 79% of the total variance in PYLL in the pooled countries.
Table 3 Regression estimates from the pooled cross-sectional time series analysis of all-cause potential years of life lost per 100,000 population
Variables Model 1 Estimate (S.E.) Model 2 Estimate (S.E.) Model 3 Estimate (S.E.) Model 4 Estimate (S.E.) Model 5 Estimate (S.E.)
Constant 7,611.05*** (489.65) 9,020.45*** (831.22) 13,168.22*** (1,221.73) 13,906*** (869.99) 13,458.57*** (890.95)
Tobacco 0.27** (0.10) 0.25* (0.10) 0.14* (0.07) 0.01 (0.07) 0.12 (0.07) P = 0.09
Alcohol 122.74*** (17.18) 120.99*** (17.65) 87.09*** (17.94) 104.86*** (13.51) 110.26*** (13.59)
Fat consumption 67.99*** (17.19) 142.39*** (15.96) 102.44*** (15.20) 91.90*** (15.05)
Fruit/vegetable consumption -0.61 (0.85) -3.08** (1.07) -1.09 (0.85) -0.21 (0.84)
Protein consumption -14.97* (6.46) -31.04*** (7.02) -22.59*** (6.62) -19.01** (6.38)
Air Pollution 2.92 (1.8) 7.31*** (2.16) 7.57** (2.30)
Collective health expenditure -1,268.53*** (261.88) -1,180.31*** (162.92) -1,002.32*** (169.98)
Healthcare coverage -36.32*** (5.32) -28.31*** (3.24) -25.37*** (3.30)
DTP immunization -21.97*** (3.79) -26.20*** (3.77)
Measles immunization -21.28*** (2.87) -14.53** (2.92)
Elderly -66.14** (23.54) -103.99*** (20.32) -70.40*** (19.75) -44.80* (18.88) -47.50** (16.73)
Physician density -768.98*** (72.85) -490.75*** (88.92) -507.20*** (87.45) -41.25 (79.42) -193.32* (89.77)
Doctor visits -129.19*** (17.28) -92.20*** (14.28) -10.50 (18.21) -30.13* (14.25) -13.71 (14.28)
GDP per capita -0.035*** (0.005) -0.052*** (0.006) -0.053*** (0.005) -0.039*** (0.005) -0.04*** (0.005)
1980s fixed effects++ -656.13*** (104.82)
1990s fixed effects++ -775.84*** (141.07)
F change - 14.02# 36.97# 71.43# 21.89#
Adjusted R2 0.57 0.60 0.69 0.77 0.79
S.E.: standard error of estimate; * P < 0.05; ** P < 0.01; *** P < 0.001; ++decade time effects to eliminate omitted-variable bias due to unobserved heterogeneity with reference to 1970s baseline; #significantly better than preceding model and model 1 (P < 0.000); n = 451
Discussion
Our contemporaneous, pooled cross-sectional series analysis suggests that a number of factors that may be amenable to public health had important associations with mortality and premature death in the selected OECD countries during the period 1970 to 1999. Most of these associations were still significant even after controlling for demographics, physician density, visits to the doctor, and GDP per capita. Tobacco, alcohol and fat intake were all positively associated with overall mortality and premature death. Protein consumption, collective health expenditure, healthcare coverage, and immunizations exhibited negative associations with both outcome measures. Though air pollution did not have significant effects on overall mortality, it was sometimes related to higher premature death when other covariates were fully taken into account. Fruit/vegetable consumption showed weak and inconsistent negative effects on both outcome measures, when partially adjusted for other covariates.
Previous studies have showed that some of the factors we studied were important in explaining mortality in OECD countries [12,18,20-22,45]. Though tobacco is known to be strongly related to higher mortality, our results support other findings that tobacco becomes insignificant in fuller models, perhaps suggesting that the tobacco consumption variable is poorly defined in the OECD dataset [18,45]. It has been suggested that a better definition would be "percent of population that smokes every day," but not the per capita use of tobacco we analyzed in this study [18]. Interestingly, in unreported lagged analyses, tobacco only became significant in fully adjusted models with at least 5 year-lags. It is, however, unclear what this might mean or how to determine what an appropriate lag period would be for all covariates in the models. We also found that tobacco was sensitive to adjustments for healthcare coverage and time effects in the total all-cause mortality models, suggesting, perhaps, that while tobacco consumption variable was limited to specific populations within countries, the healthcare coverage variable had a wider reach across populations, effectively diluting the statistical effect of tobacco. Similarly, the failure of fruit/vegetable consumption and air pollution to show the expected associations [15,16] in the fuller models in this study may be due to poor definitions and data quality, or due to their sensitivity to the effect of healthcare coverage in the models.
In its 2002 World Health Report, the World Health Organization showed that lifestyle, behavioral, and environmental risk factors, such as the ones in this study, accounted for significant proportions of the disease and mortality burden in most parts of the world including the affluent countries [15]. As much as 39–40% of the disease burden and 51–53% of mortality in developed countries were attributable to 20 selected risk factors such as tobacco, alcohol, high blood pressure, high body mass index, high cholesterol and low fruit and vegetable intake [17]. Our study is the first, however, to report a pooled time series impact of tobacco, alcohol, fat, fruit/vegetable, air pollution, collective health expenditure, healthcare coverage and immunizations on mortality and premature death, adjusting for demographics, medical care input and national wealth in the selected OECD countries during the last 30 years of the 20th century. It indirectly supports some of the aforementioned population risks, but raises questions as to what public health can actually do to curb the unhealthy associations.
Revisiting the essential functions of public health [24,46,47] and its classical paradigms of health promotion, health protection and disease prevention [14] may offer broad insights into holistic approaches for addressing the effects of health determinants. For instance, nutritional lifestyle factors are amenable to health promotion [15]. Air pollution could be addressed under health protection activities such as environmental modification and regulations. Immunizations against infections such as measles, diphtheria, tetanus and pertussis belong to the disease prevention role of public health. Although, this study is ecological in nature, uses population average variables [48,49], and recognizes the potential for the ecological fallacy [50], it is unlikely that the solutions to the problems of, say, alcohol, tobacco, and nutritional lifestyle would be entirely ecological. Solutions, such as behavioral modification, targeted at all levels of the society, from individuals to groups, would be necessary.
There is evidence that some countries such as the US [51,52] and the Netherlands [53] have ongoing initiatives aimed at tackling health determinants in their populations. It is yet to be seen how successful these programs would be if specific attention is not given to re-engineering public health systems. Increasing healthcare coverage is yet another important way of improving population health and its distribution. This is particularly important for the US where coverage is still a big problem.
Furthermore, the functions of health status monitoring, surveillance, reducing disaster impact, human resource development and public health regulation require substantial investment. Public health investment may have increased relative to GDP in many OECD countries, but the attained levels and distribution of collective health expenditure are still inadequate, given the problems of re-emerging infections, unsolved issues of poverty and inequalities, global terrorism and environmental degradation [54]. Currently, many OECD countries spend far more on the curative medical care sector [55,56] than on prevention and health promotion [35,36]. Unfortunately, many of the diseases (e.g. coronary heart disease) treated in their hospitals, for example, tend to arise from such preventable factors as excessive tobacco, fat and alcohol use [15-17]. Our study showed that even after adjusting for medical care input, there were excess mortality and premature deaths due to preventable factors.
It, therefore, seems prudent to re-focus on public health functions of health systems for at least four reasons. First, it averts health problems and minimizes subsequent morbidity and mortality. Second, public health faces a legitimacy or relevance problem when it does not deal competently with the conflict between civil liberties and health promotion [13], as well as with the new 'epidemics' such as obesity [14,57]. Third, the recent attention given to health system performance should be more comprehensive and include the optimal functioning of public health systems alongside medical care structures [10,11,30,58]. Fortunately, the US, UK, Netherlands, Australia and Canada are among the countries actively pursuing systematic evaluations of their health systems. Fourth, public policy on health and health-related social issues needs to become more integrated, and public health offers an important interface between the traditional health sector and the social sectors. There is need for integrated, intersectoral and innovative solutions beyond the prevailing narrow policy approaches [57,59]. In the light of a similar OECD study that showed that primary care had strong relationship with health outcomes [18], even after controlling for similar factors as we studied, it seems that strengthening primary care and public health may be a prudent and an effective strategy against unfavorable health outcomes. Our study further reinforces recent analyses which used the concept of 'avoidable' mortality (that is, mortality that should not occur in the presence of effective and timely healthcare) to point out the importance of appropriate public health policies as an integral part of evaluating and improving health system performance [60,61].
Limitations of this study
This study used data that may have comparability and definitional deficiencies [20,35,36]. Use of secondary data from international resources can import the attendant problems of incomparable definitions and poor data quality. The OECD Health dataset (from where we took our public health and medical care related variables) and the OECD's Annual National Accounts data (that provided the expenditure variables in our study) are no exceptions. There are likely issues of errors of observation and comparability in this database given the daunting tasks that underlie such international data collection efforts. The incomparability issues are even more likely to be more severe as the dataset tries to include more non-healthcare accounts measures such as lifestyle factors, as has been the case in recent years. Yet, one can be too apologetic about measurement errors in the OECD Health dataset given its seeming robustness for routine and political use and for guiding practical decisions so far [62]. Besides, efforts are constantly being made to increase the value and quality of the data.
The measures we used are, at best, weak proxies for more robust measures of aggregate lifestyle, environmental quality and safety, public health investment and medical care inputs [61,63]. Medical care input data tend to show mixed results, especially within the context of avoidable mortality [61,64]. Furthermore, this study does not provide clear directions as to which policies are best suited for addressing lifestyle, environment, public health investments or any of the factors we studied. The pooled nature of the statistical models limit the potential for generalizability of our findings to other countries not included in this study. Moreover, the estimated models used crude measures, ignored distributional concerns and distal determinants of health, and did not consider the possible multilevel and/or lagged nature of the explored relationships.
Conclusion
We have presented a pooled, cross-sectional time series analysis of the associations of public health interventions and investment, the environment, and lifestyle-related factors with population health in selected industrialized countries during the period 1970–1999. Given the limitations of the study, we only make broad-brush assessments of the relevance of these findings. In view of current health concerns, our findings serve to make a case for a "new public health" as a cornerstone of health systems. As such, health policies aimed at preventable factors, namely those modifiable by public health, should count in the overall assessment of health systems.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
OAA conceived of the study, collected the data, and led the analysis, interpretation and manuscript drafting. GPW assisted with the study design and strategy for analysis, interpretation of findings and critical review of the manuscript for intellectual content. DMD also assisted with the study design, interpretation of findings and critical review of the manuscript for intellectual content. NSK assisted with the study design, interpretation, and critical review of the manuscript. All authors reviewed the manuscript for intellectual content.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors thank Mary Nicolaou and Dr. Karien Stronks of the Academic Medical Center of the University of Amsterdam, Prof. Vahé Kazandjian of the Johns Hopkins Bloomberg School of Public Health and two reviewers of the Journal for their helpful comments.
==== Refs
Mathers CD Iburg KM Salomon JA Tandon A Chatterji S Ustun B Murray CJ Global patterns of healthy life expectancy in the year 2002 BMC Public Health 2004 4 66 15619327 10.1186/1471-2458-4-66
McKeown T The Role of Medicine – Dream, Mirage or Nemesis? 1976 London: Nuffield Provincial Hospitals Trust
McKinlay JB McKinlay SM Beaglehole R A review of the evidence concerning the impact of medical measures on recent mortality and morbidity in the United States Int J Health Serv 1989 19 181 208 2654039
McKinlay JB McKinlay SM The questionable contribution of medical measures to the decline of mortality in the United States in the twentieth century Milbank Mem Fund Q Health Soc 1977 55 405 428 413067
Illich I Medical Nemesis 1976 New York: Random House
Mackenbach JP Looman CW Kunst AE Habbema JD van der Maas PJ Post-1950 mortality trends and medical care: gains in life expectancy due to declines in mortality from conditions amenable to medical intervention in The Netherlands Soc Sci Med 1988 27 889 894 3227384 10.1016/0277-9536(88)90278-X
Mackenbach JP The contribution of medical care to mortality decline: McKeown revisited J Clin Epidemiol 1996 49 1207 1213 8892485 10.1016/S0895-4356(96)00200-4
Bunker JP Medicine's core values. Medical care does add to life expectancy BMJ 1994 309 1657 7819956
Bunker JP The role of medical care in contributing to health improvements within societies Int J Epidemiol 2001 30 1260 1263 11821323 10.1093/ije/30.6.1260
World Health Organization The World Health Report 2000 Health Systems: Improving Performance 2000 Geneva: WHO
Arah OA Klazinga NS Delnoij DM ten Asbroek AH Custers T Conceptual frameworks for health systems performance: a quest for effectiveness, quality, and improvement Int J Qual Health Care 2003 15 377 398 14527982 10.1093/intqhc/mzg049
Arah OA Klazinga NS How safe is the safety paradigm? Qual Saf Health Care 2004 13 226 232 15175496 10.1136/qshc.2003.007070
Hamlin C Detels R, McEwen J, Beaglehole R, Tanaka H The history and development of Public Health in developed countries Oxford Textbook of Public Health 2002 Oxford: Oxford University Press 21 37
Awofeso N What's new about the "new public health"? Am J Public Health 2004 94 705 709 15117684
World Health Organization The World Health Report 2002 Reducing Risks, Promoting Healthy Life 2002 Geneva: World Health Organization
Ezzati M Lopez AD Rodgers A Vander HS Murray CJ Selected major risk factors and global and regional burden of disease Lancet 2002 360 1347 1360 12423980 10.1016/S0140-6736(02)11403-6
Ezzati M Hoorn SV Rodgers A Lopez AD Mathers CD Murray CJ Estimates of global and regional potential health gains from reducing multiple major risk factors Lancet 2003 362 271 280 12892956 10.1016/S0140-6736(03)13968-2
Macinko J Starfield B Shi L The contribution of primary care systems to health outcomes within Organization for Economic Cooperation and Development (OECD) countries, 1970–1998 Health Serv Res 2003 38 831 865 12822915 10.1111/1475-6773.00149
Anderson G Hussey PS Comparing health system performance in OECD countries Health Aff (Millwood) 2001 20 219 232 11585171 10.1377/hlthaff.20.3.219
Retzlaff-Roberts D Chang CF Rubin RM Technical efficiency in the use of health care resources: a comparison of OECD countries Health Policy 2004 69 55 72 15484607 10.1016/j.healthpol.2003.12.002
Or Z Exploring the Effects of Health Care on Mortality across OECD Countries 2001 Paris: Organisation for Economic Cooperation and Development
Or Z Determinants of Health in Industrialized Countries: A Pooled Cross-country Time Series Analysis 2000 Paris: Organisation for Economic Cooperation and Development
Acheson D Public Health in England The Report of the Committee of Inquiry into the Future Development of the Public Health Function 1988 Cmnd 289. London: HMSO
American Public Health Association National Public Health Performance Standards Program (accessed 30 July 2004)
Mays GP McHugh MC Shim K Perry N Halverson PK Lenaway D Identifying dimensions of performance in local public health systems: results from the National Public Health Performance Standards Program J Public Health Manag Pract 2004 10 193 203 15253515
Jamrozik K Hobbs M Detels R, McEwen J, Beaglehole R, Tanaka H Medical care and public health Oxford Textbook of Public Health 2002 Oxford: Oxford University Press 213 242
Evans RG Stoddart GL Producing health, consuming health care Soc Sci Med 1990 31 1347 1363 2126895 10.1016/0277-9536(90)90074-3
Evans RG Stoddart GL Consuming research, producing policy? Am J Public Health 2003 93 371 379 12604475
Kindig D Stoddart G What is population health? Am J Public Health 2003 93 380 383 12604476
ten Asbroek AH Arah OA Geelhoed J Custers T Delnoij DM Klazinga NS Developing a national performance indicator framework for the Dutch health system Int J Qual Health Care 2004 16 i65 i71 15059989 10.1093/intqhc/mzh020
McDowell I Spasoff RA Kristjansson B On the classification of population health measurements Am J Public Health 2004 94 388 393 14998801
Hollingsworth JR Hage J Hanneman RA State Intervention in Medical Care: Consequences for Britain, France, Sweden, and the United States, 1890–1970 1990 Ithaca and London: Cornell University Press
Sayrs LW Pooled Time Series Analysis 1989 Newbury Park, CA: Sage Publications, Inc
Ashenfelter O Levine PB Zimmerman DJ Statistics and Econometrics: Methods and Applications 2003 New York: John Wiley & Sons, Inc
Organisation for Economic Cooperation and Development Health Data 2003 2003 Paris: Organization for Economic Cooperation and Development
Organisation for Economic Cooperation and Development Annual National Accounts for OECD Member Countries 2004 Paris: Organization for Economic Cooperation and Development
Shi L Starfield B The effect of primary care physician supply and income inequality on mortality among blacks and whites in US metropolitan areas Am J Public Health 2001 91 1246 1250 11499112
Shi L Macinko J Starfield B Wulu J Regan J Politzer R The relationship between primary care, income inequality, and mortality in US States, 1980–1995 J Am Board Fam Pract 2003 16 412 422 14645332
Food and Agriculture Organization of the United Nations (accessed 1 July 2004)
Anonymous Confirming need for protective national health-based air quality standards Am J Public Health 2001 91 501 502 11236440
White H A heteroskedastic-consistent covariance matrix estimator and a direct test of heteroskedasticity Econometrica 1980 48 817 838
MacKinnon JG White H Some heteroscedasticity-consistent covariance matrix estimators with improved finite sample properties J Econometrics 1985 29 53 57 10.1016/0304-4076(85)90158-7
Long JS Ervin L Using heteroscedasticity-consistent standard errors in the linear regression models The American Statistician 2000 54 217 224
Hayes AF Heteroscedasticity-Consistent Standard Error Estimates for the Linear Regression Model: SPSS and SAS Implementation 2003 Columbus, OH: The Ohio State University
Kennelly B O'Shea E Garvey E Social capital, life expectancy and mortality: a cross-national examination Soc Sci Med 2003 56 2367 2377 12742601 10.1016/S0277-9536(02)00241-1
Pan American Health Organization Essential Public Health Functions Health Sector Reform: Reassessing Implications for PAHO's Technical Cooperation 2000 PAHO Annual Managers Meeting. Washington DC: PAHO
The World Bank Public Health and Bank Operations 2002 Washington DC: World Bank
Susser M The logic in ecological: I. The logic of analysis Am J Public Health 1994 84 825 829 8179056
Susser M The logic in ecological: II. The logic of design Am J Public Health 1994 84 830 835 8179057
Schwartz S The fallacy of the ecological fallacy: the potential misuse of a concept and the consequences Am J Public Health 1994 84 819 824 8179055
Healthy People 2010 2000 2 Washington DC: US Department of Health and Human Services
Pamuk ER Wagener DK Molla MT Achieving national health objectives: the impact on life expectancy and on healthy life expectancy Am J Public Health 2004 94 378 383 14998799
Ministry of Health, Welfare and Sport Living Longer in Good Health Also A Question of A Healthy Lifestyle Netherlands Healthcare Prevention Policy 2004 The Hague: Ministry of Health, Welfare and Sport
Beaglehole R ed Global Public Health: A New Era 2003 Oxford: Oxford University Press
Anderson GF Poullier JP Health spending, access, and outcomes: trends in industrialized countries Health Aff (Millwood) 1999 18 178 192 10388215 10.1377/hlthaff.18.3.178
Anderson GF Hurst J Hussey PS Jee-Hughes M Health spending and outcomes: trends in OECD countries, 1960–1998 Health Aff (Millwood) 2000 19 150 157 10812793 10.1377/hlthaff.19.3.150
Arah OA Catastrophic failures of public health Lancet 2004 363 1551 1552 15135607 10.1016/S0140-6736(04)16157-6
van Oers JAM ed Health on Course? The 2002 Dutch Public Health Status and Forecasts Report 2002 Houten: Bohn Stafleu Van Loghum
Fischer F Reframing Public Policy: Discursive Politics and Delibrative Practices 2003 Oxford: Oxford University Press
Nolte E McKee M Measuring the health of nations: analysis of mortality amenable to health care BMJ 2003 327 1129 14615335 10.1136/bmj.327.7424.1129
Nolte E McKee M Does Healthcare Save lives? Avoidable Mortality Revisited 2004 London: The Nuffield Trust
Reinhardt UE Hussey PS Anderson GF Cross-national comparisons of health systems using OECD data, 1999 Health Aff (Millwood) 2002 21 169 181 12025981 10.1377/hlthaff.21.3.169
Mackenbach JP Health care expenditure and mortality from amenable conditions in the European community Health Policy 1991 19 245 255 10115995 10.1016/0168-8510(91)90011-L
Carr-Hill RA Hardman GF Russell IT Variations in avoidable mortality and variations in health care resources Lancet 1987 1 789 792 2882193 10.1016/S0140-6736(87)92810-8
|
16076396
|
PMC1185550
|
CC BY
|
2021-01-04 16:28:58
|
no
|
BMC Public Health. 2005 Aug 2; 5:81
|
utf-8
|
BMC Public Health
| 2,005 |
10.1186/1471-2458-5-81
|
oa_comm
|
==== Front
BMC Struct BiolBMC Structural Biology1472-6807BioMed Central London 1472-6807-5-91598515310.1186/1472-6807-5-9Research ArticleAn inactivated nuclease-like domain in RecC with novel function: implications for evolution Rigden Daniel John [email protected] School of Biological Sciences, University of Liverpool, Crown St., Liverpool L69 7ZB, UK2005 28 6 2005 5 9 9 15 4 2005 28 6 2005 Copyright © 2005 Rigden; 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 PD-(D/E)xK superfamily, containing a wide variety of other exo- and endonucleases, is a notable example of general function conservation in the face of extreme sequence and structural variation. Almost all members employ a small number of shared conserved residues to bind catalytically essential metal ions and thereby effect DNA cleavage. The crystal structure of the RecBCD prokaryotic DNA repair machinery shows that RecB contains such a nuclease domain at its C-terminus. The RecC C-terminal region was reported as having a novel fold.
Results
The RecC C-terminal region can be divided into an alpha/beta domain and a smaller alpha-helical bundle domain. Here we show that the alpha/beta domain is homologous to the RecB nuclease domain but lacks the features necessary for catalysis. Instead, the domain has a novel function within the nuclease superfamily – providing a hoop through which single-stranded DNA passes. Comparison with other structures of nuclease domains bound to DNA reveals strikingly different modes of ligand binding. The alpha-helical bundle domain contributes the pin which splits the DNA duplex.
Conclusion
The demonstrated homology of RecB and RecC shows how evolution acted to produce the present RecBCD complex through aggregation of new domains as well as functional divergence and structural redeployment of existing domains. Distantly homologous nuclease(-like) domains bind DNA in highly diverse manners.
==== Body
Background
The largest evolutionary superfamilies of proteins cover such a large range of sequence space that the relationships shared by members may not be apparent by standard means of sequence comparison, and hence are often only recognized after structural determinations. Such has frequently been the case for the PD-(D/E)xK superfamily of nucleases. Within the superfamily, structures were first obtained for four restriction enzymes, of such diverse sequences that they were initially assumed not to share homology (reviewed in [1]). Since then structures have confirmed distant and often unexpected homologies of those four with many other restriction enzymes, as well as exo- and endo-nucleases involved in such diverse cellular processes as DNA repair [2], transposition [3], Holliday junction resolution [4] and recombination [5].
The unifying catalytic site characteristic of the superfamily is the presence of one or more catalytically essential divalent cations [6,7]. The conserved acidic residues of the PD-(D/E)xK motif, which can be separated by any number of residues, bind one metal cation while the conserved lysine residue is involved in positioning water suitably to attack the DNA backbone. In some lineages of the superfamily variation on this classical motif is apparent in the substitution of the second acidic residue by a catalytically essential His residue (2), or in the migration of the second acidic residue [8] or the lysine residue [9] to other parts of the fold. Irrespective of this variation, the catalytic site is placed at one edge of the core four or five-stranded β-sheet at the heart of the α /β domain structure [1,6,7]. While an overwhelming majority of the superfamily contain one of these catalytic site variants some interesting exceptions have been noticed. Thus, while clearly containing a PD-(D/E)xK superfamily-like domain structure [10], the tRNA splicing endoribonuclease EndA, has evolved an unrelated catalytic site on the opposite side of the fold to the conventional site [11]. A catalytically inactive version of the fold has also been seen in the N-terminal domain of S. cerevisiae RPB5, an RNA polymerase subunit, where evidence suggests that it functions in protein-protein interactions [12].
Although extremely diverse in structure and sequence, modern sequence comparison methods have played their part in elucidating the full range of PD-(D/E)xK superfamily members [9,13-15]. Nevertheless, structure determinations and structure-informed bioinformatics [16] will continue to be crucial in this most diverse of superfamilies. Some five years ago it was predicted that the nuclease activity associated with the C-terminus of RecB [17] resulted from the presence of a domain homologous to that of λ-exonuclease, despite RecB not possessing a PD-(D/E)xK motif [13,14]. This prediction has been recently confirmed with the crystal structure determination of the structure of the RecBCD heterotrimer [18]. This remarkable complex (see [18] and references therein) which functions to process double-stranded breaks in DNA, contains two distinct helicase activities, contributed by RecB and RecD. Also present is a catalytically inactive subunit, RecC. Among its proposed roles is recognition of the Chi DNA sequence [18]. Remarkably, twin helicase(-like) motor domains (canonically named 1A and 2A) are present in all three subunits, although those in RecC are inactivated and only those in RecB and RecC contain α-helical insert domains in each motor domain (named 1B and 2B, respectively). As mentioned, the helicase domains of RecB are followed by a PD-(D/E)xK superfamily nuclease domain 3. In contrast, the C-terminal 'domain 3' of RecC was reported as being of novel fold [18].
Here we show that the C-terminal region ('domain 3') of RecC can actually be dissected into two domains, the first of which is clearly related to PD-(D/E)xK superfamily nuclease domains (hereafter called simply nuclease domains) and particularly to the corresponding domain of RecB. The nuclease-like domain of RecC is inactivated and therefore possesses not even the metal-ligating residues of the PD-(D/E)xK motif. Instead, it carries out a function not hitherto observed in the superfamily, providing an aperture through which one strand of newly split DNA duplex is fed. Comparisons show that nuclease(-like) domains are extraordinarily versatile in their mode of interaction with duplex DNA. Characteristics of the RecC nuclease-like domain show that RecB and RecC share a common ancestor and reveal how evolution has progressed by sequential addition of domains to the C-terminus, as well as by altering function of, and repositioning of, existing domains.
Results and discussion
An unsuspected nuclease-like domain in RecC
Domain 3 of RecC has been described as being of novel fold [18]. Structural examination suggested that it could, in fact, be divided into two domains, an α/β domain and a C-terminal all α-helical domain. Although the division was made by eye initially, analysis with Protein Domain Parser [19] produced a result that differed by just two residues. When the α/β domain (comprising residues 828–1033) was submitted to DALI [20], the most closely related structure in the database was reported as phosphoserine phosphatase but in second place was λ-exonuclease (PDB code 1avq; [5]). A root mean squared (rms) deviation between the third RecC domain and λ-exonuclease of 4.2Å for 121 Cα atoms was obtained (yielding a DALI Z score of 4.1). λ-exonuclease is the nearest structural neighbour to the nuclease domain of RecB [18]. For that pair, 131 Cα atoms can be superimposed with an rms deviation of 3.5 Å (Z score of 6.2). From these data and visual inspection (later additionally supported by PSI-BLAST results – see below), it is clear that the third RecC domain is a relative of the nuclease domain common to RecB and λ-exonuclease (Figures 1 and 2). Notably, the further division of the C-terminal RecC 'domain 3' into two domains was essential for this relationship to become apparent. In contrast, the fourth, α-helical bundle domain of RecC has no close neighbours in the present database.
The nuclease fold common to λ-exonuclease, RecB and now RecC is found in a wide variety of exo- and endonucleases, from restriction enzymes to Holliday junction resolvases, and enzymes of DNA repair [14]. Within the superfamily, conserved motifs vary with family, but all centre on acidic residues involved in binding the divalent metal cation typically required for catalysis [6,7]. These residues are the sole residues conserved across almost the whole superfamily. A calcium ion bound to RecB in the crystal structure [18] marks the binding site for the essential metal while in λ-exonuclease, soaking in manganese has revealed the corresponding site [5]. A metal-binding site, like those shown in Figure 1, is not present in RecC (Figure 3). Indeed, the overall sequence identity between the RecB and RecC sequence segments shown in Figure 3 is just 2–11 %. Thus, just as domains 1 and 2 of RecC are inactivated helicase domains [18], so its domain 3 is an inactivated nuclease.
Interestingly, comparison of the nuclease domains of RecB, RecC and λ-exonuclease shows that the Rec subunits clearly share a more recent ancestor than the common ancestor of all three structures. As Figures 1 and 2 show, a single helix present in λ-exonuclease is replaced in both RecB and RecC by a three-helix α-helical bundle. This bundle is not present in the more distant relatives of λ-exonuclease highlighted by the CE server [21] such as archaeal Holliday junction resolvase, tRNA endonuclease and the PvuII restriction enzyme. Curiously, the degree of structural superposition that can be achieved between the RecB and RecC nuclease domains and λ-exonuclease suggests no closer relationship between the former pair. For example, 71 Cα atoms of RecB nuclease domain may be superimposed on their equivalents in RecC to produce an rms deviation of 2.19 Å. In comparison, 82 Cα atoms of RecB superimpose on equivalents of λ-exonuclease with a lower rms deviation of 1.71Å. However, the superimposable three-helix α-helical bundle shared only by RecB and RecC (Figures 1 and 2) show that they are more closely evolutionarily related to each other than to other homologous structures. The closer structural superposition of RecB and λ-exonuclease seems likely to arise from their shared nuclease activity, while RecC has evolved a different function.
Novel function of the nuclease-like domain in RecC
As mentioned, nuclease domains as represented in the present PDB are extremely diverse in sequence but share conserved residues that bind essential metal ions and are almost invariably catalytically active. The recognition of the third domain of RecC as an inactivated nuclease domain highlights a wholly unexpected new function for a non-catalytic but clearly nuclease-like domain. As shown in Fig 2e, the nuclease-like domain of RecC provides a hoop through which a single strand of the newly separated DNA duplex is passed. The hoop is the entrance to the 5' channel leading to RecD in the RecBCD complex [18]. The pin responsible for separating the two DNA strands consists of a loop extending out of the α-helical bundle domain 4 of RecC.
Structural comparisons show that a series of three structural adaptations have been required in RecC in order to achieve this novel ssDNA-hoop function. These involve three regions of sequence marked on Figures 2 and 3. Region 1 comprises a long linker sequence between the extended structure that starts the domain and the three helix α-helical bundle subdomain. This linker region is very poor in regular secondary structure and adopts dramatically different conformations in the two domains. Significant sequence identity between RecB and RecC seems absent in the region. In RecB this linker lies along the surface of the remainder of the domain. In dramatic contrast, region 1 in RecC has few contacts with the rest of the domain (although it contacts other parts of RecC – see below) and forms most of the rim of the hoop through which ssDNA is passed. Region 2 is the connection between the two strands forming an antiparallel β-sheet. In E. coli RecC the connection is a minimal β-turn and connections in other RecC sequences are also very short (Figure 3). In contrast, Region 2 in RecB is usually much larger, tracing out, in the E. coli RecB structure, an 11-residue α-helix and a substantial stretch lacking regular secondary structure. Structure comparison shows the reason for the short connectors in RecC (Figure 2); larger connectors occupy the same space as the fourth domain of RecC. Thus, a larger connector would be incompatible with a RecC-style pin domain. Region 3, providing the connector between a β-strand and an α-helix, is again larger in RecB than in RecC and again contains an α-helix in RecB. Here the reason for the shorter connector in RecC is even more fundamental; were it to have the longer connector of RecB, the aperture whereby ssDNA passes through the RecC nuclease-like domain would be sterically obstructed.
DNA interactions with nuclease and nuclease-like domains
Unfortunately, no structure of λ-exonuclease in complex with DNA is yet available. However, other enzymes sharing the same fold, including many type II restriction enzymes, have been crystallized in complex with DNA. Therefore, DNA-bound structures were sought for the enzymes identified as closest structural neighbours for λ-exonuclease by the CE server [21]. This analysis pinpointed the restriction enzyme PvuII (PDB code 1pvi; [22]) and the vsr exonuclease (PDB code 1odg; [23]) involved in repair of bacterial G:T mismatches. Further analysis (not shown) showed that the mode of binding of DNA to Pvu II was, in fact, typical of many restriction enzymes, irrespective of dimeric vs tetrameric quaternary state and of differing modes of dimerization.
Remarkably, as shown in Figure 4, the axes of duplex DNA binding to PvuII and to vsr exonuclease are almost orthogonal, a difference that seems to have escaped notice. The catalytic sites of both enzymes, although differing in detail, are similarly placed at one edge of the β-sheet, defining the 'front' of catalytic nucleases. Most unexpectedly, the inactivated nuclease-like domain of RecC which also, in the context of RecBCD, binds duplex DNA, prior to strand splitting by the fourth domain, does so in a completely novel manner. First, the axis of the bound duplex DNA is approximately orthogonal to both PvuII and vsr exonuclease modes. Secondly, the binding involves the 'back' of the domain; only a single strand of the DNA arrives at the 'front' side after passing through the aperture (Figure 4). These results make clear that few assumptions can be made regarding modes of DNA binding by nuclease(-like) domains in the absence of experimental data such as structures in complex with DNA.
Homology of RecB and RecC
The observation of inactivated helicase-like domains in RecC was not considered reason enough to propose the existence of homology between RecB and RecC extending over their whole length [18]. Indeed, both sequence and structural comparisons at first suggest that RecB more closely resembles other helicases than it does RecC. For example, in the results of PSI-BLAST [24] starting with E. coli RecB, PcrA, another helicase that contains large helical-insert domains in each helicase domain [25], appears as a significant hit (e = 6 × 10-9) in the results of the first iteration. In contrast, using an e-value cut-off of 0.0001 four iterations are required before RecC sequences, including that of E. coli RecC, appear among the significant hits. While the BLAST alignments centred on the helicase(-like) domains the C-terminal nuclease(-like) domains were sometimes matched, although PSI-BLAST runs of the nuclease domain of RecB failed to hit the nuclease-like domain of RecC, and vice versa. Similarly, structural comparisons show that both helicase domains and both α-helical insert domains of RecB are more similar to their counterparts in PcrA than to the corresponding RecC domains (not shown). Nevertheless, the clear homology of the RecB and RecC nuclease(-like) domains, evident in their common three α-helical bundle (see above) strongly suggests that RecB and RecC share a more recent common ancestor than they have in common with other extant helicases. How then to explain the apparently closer relationship of RecB with PcrA than with RecC? As was proposed for the nuclease(-like) domains (see above) it seems like the dramatic functional differences between corresponding RecB and RecC domains are responsible. As discussed above, the RecC nuclease-like domain is significantly shorter than the RecB nuclease domain in two key regions, each associated with its new role as provider of an ssDNA hoop. Thus, it seems plausible that the maintenance of helicase activity by the helicase domains of PcrA and RecB is responsible for their apparently closer relationship, the structural changes accompanying evolution of the helicase-like domains in RecC for new roles having obscured their more recent shared ancestry with RecB.
The recognition of homology between RecB and RecC, and the dissection of their domains leads to an interesting comparison with PcrA. In PcrA, a duplication of an ancestral RecA-like domain, already containing an all α-helical insert domain, is evident [25]. In RecB a long linker region and following nuclease domain have been added to the PcrA template (Figure 5). A further domain addition has occurred in RecC, that of the small C-terminal α-helical bundle domain that contributes the duplex-splitting pin. This picture of aggregation of novel functionality through domain addition is complemented by alterations in function of homologous domains. Thus, as described, the nuclease-like domain of RecC continues to bind duplex DNA, but using a different surface of the domain, as well as providing the entrance to the 5' ssDNA channel leading to RecD. This modification is paralleled in the helicase-like domains by a change from catalytic helicase activity to Chi sequence recognition ([18] and references therein). The α-helical inserts into the helicase(-like) domains also have different functions in RecB and RecC [18], including, in the case of the RecC domain 1B binding to the RecC nuclease-like domain and the rim of its ssDNA aperture (Figure 5). Although homologous, the structural comparison of complete RecB and RecC subunits shows large differences in relative domain orientations and positions, most dramatically with regard to the position of the nuclease(-like) domains relative to the helicase(-like) domain cores (Figure 5).
There is an interesting parallel to be drawn between RecBCD and AddAB (also known as RexAB), a different DNA repair system found in Gram positive bacteria where RecBCD is lacking (see [26] for a review). AddA and AddB also appear homologous and each possesses helicase and nuclease motifs. Within AddAB, it is AddB that recognises the Chi sequence and therefore is the counterpart of RecC in RecBCD. Most interestingly, however, both the nuclease domains of AddA and AddB appear to be active [27]. The AddAB system may therefore resemble an evolutionarily intermediate stage, through which the RecBCD machine passed before inactivation of the RecC nuclease domain and recruitment of RecD.
In summary, the improved domain dissection of RecC presented here and its ramifications enhance our understanding of the evolutionary processes responsible for the remarkable DNA processing machinery that is the RecBCD complex [18]. It is now even more apparent that relatively straightforward addition of modular functionality has been accompanied by quite dramatic functional evolution of homologous domains.
Methods
Protein structures were retrieved from the Protein Databank (PDB; [28]). Protein structural superpositions were obtained at the CE [21] and DALI [20] servers and by using the program LSQMAN [29]. Structural relationships were also explored in the SCOP database [30]. Protein structure visualization employed O [31] and PyMOL [32], the latter also being used for production of figures. Iterative database searches were carried out using PSI-BLAST [24]. Sequences were retrieved from the COG [32] entries for RecB (COG1074) and RecC (COG1330). Maximally diverse representatives were chosen using JALVIEW [34] which was also used for general sequence manipulation. Protein sequence alignment was carried out using MUSCLE [35] and T-COFFEE [36]. Formatting of sequence alignments was done with ESPRIPT [37] using default options for colouring of sequence conservation.
Acknowledgements
I am grateful for the helpful remarks of one of the anonymous referees regarding AddAB.
Figures and Tables
Figure 1 Stereo structural superposition of the nuclease(-like) domains of RecB and RecC produced with LSQMAN [29]. RecB is coloured grey while RecC, is coloured in a spectrum from blue (N-terminus) to red (C-terminus). Structurally superposed regions are shown in cartoon representation (rms deviation of 2.19Å for 71 Cα atoms), other parts as a Cα trace. The calcium ion bound to RecB is shown as a magenta sphere.
Figure 2 Comparisons of structurally aligned nuclease(-like) domains in λ-exonuclease, RecB and RecC. The comparison in a)-c) shows how a single helix in λ-exonuclease (PDB code 1avq; [5]) (a) has been replaced by superimposable α-helical bundles in RecB (b) and RecC (c) (PDB code 1w36; [18]), indicating a more recent shared ancestor of the latter pair. The regions in question are shown as light grey. The remainders of the molecules are coloured in a spectrum from blue (N-terminus) to red (C-terminus). In a) and b), two acidic, metal-ligating residues drawn as sticks mark respective catalytic sites. RecB and RecC are compared in more detail in d) and e), respectively, again coloured from blue to red with the exception of labelled key regions 1 (black), 2 (dark grey) and 3 (grey). Bound metal is shown in b) and d) as spheres while e) additionally shows DNA (shades of pink) and the domain 4 of RecC, coloured uniformly lime green with its pin structure labelled. The DNA strand that penetrates the hoop provided by RecC is shown as a broader cartoon. The RecC "hoop" region (see text for details) is labelled in c) and e) and DNA strand termini are labelled in e).
Figure 3 Structure-based sequence alignment of the nuclease(-like) domains of RecB and RecC. Nuclease(-like) domain sequences of RecB (above, group 1) and RecC (below, group 2) were chosen from diverse representative species and extracted from complete alignments of COG database [31] entries for RecB or RecC. Purple indicates the E. coli sequences crystallized as the RecBCD complex (PDB code 1w36; [18]). Other sequences are labelled with Genbank numbers and sequence codes Bb, Borrelia burgdorferi; Cp, Chlamydophila pneumoniae; Mt, Mycobacterium tuberculosis; Xf, Xylella fastidiosa. Red colouring indicates conservation within each group while green is used for three important catalytic residues of RecB – H956, D1067 and D1080 [18]. Elements of regular secondary structure are shown above (RecB) and below (RecC) the alignment, where spirals represent α-helices and arrows β-strands. The three key regions (numbered 1–3) involved in adaptation of the RecC nuclease-like domain to its new function, as discussed in the text (see also Fig 2), are boxed and labelled. Purple underlining indicates zones that can be simultaneously structurally aligned (rms deviation of 2.19Å for 71 Cα atoms).
Figure 4 Comparison of modes of DNA binding to superimposed nuclease(-like) domains. The domain structures are those of a) PvuII (PDB code 1pvi; [22] b) vsr exonuclease (PDB code 1odg; [23]) and c) RecC (PDB code 1w36; [18]). Protein chains are coloured in a spectrum from blue (N-terminus) to red (C-terminus) while DNA is coloured uniformly pink. In order to illustrate the approximate locations of the catalytic sites, selected catalytic residues are shown for PvuII (D58 and E58) and vsr exonuclease (D51 and H69). DNA termini are labelled, as is the "hoop" in RecC.
Figure 5 Domain comparison of PcrA, RecB and RecC. The structures of a) substrate-complexed PcrA (PDB code 3pjr; [38]), b) RecB and c) RecC (PDB code 1w36; [18]), are superimposed using domain 2A. The domains 1A are coloured red, while orange is used for 1B, blue for 2A, cyan for 2B, green for 3 and yellow for 4. The same colours are used in the schematic diagram d) which illustrates how evolution progressed through addition of domains. The "pin" and "hoop" in RecC (see text for details) are labelled. In PcrA the residues defining the starts of the various (sub-)domains are 1B, 92; 1A continuation, 218; 2A, 287; 2B, 385; 2A continuation, 553. In RecB the corresponding residue numbers are 1B, 151; 1A continuation, 349; 2A, 446; 2B, 583; 2A continuation, 729; linker, 870; 3, 900. In RecC they are 1B, 79; 1A continuation, 208; 2A, 329; 2B, 443; 2A continuation, 649; linker, 784; 3, 828; 4, 1034.
==== Refs
Aggarwal AK Structure and function of restriction endonucleases Curr Opin Struct Biol 1995 5 11 19 7773740 10.1016/0959-440X(95)80004-K
Tsutakawa SE Muto T Kawate T Jingami H Kunishima N Ariyoshi M Kohda D Nakagawa M Morikawa K Crystallographic and functional studies of very short patch repair endonuclease Mol Cell 1999 3 621 628 10360178 10.1016/S1097-2765(00)80355-X
Hickman AB Li Y Mathew SV May EW Craig NL Dyda F Unexpected structural diversity in DNA recombination: the restriction endonuclease connection Mol Cell 2000 5 1025 1034 10911996 10.1016/S1097-2765(00)80267-1
Bond CS Kvaratskhelia M Richard D White MF Hunter WN Structure of Hjc, a Holliday junction resolvase, from Sulfolobus solfataricus Proc Natl Acad Sci U S A 2001 98 5509 5514 11331763 10.1073/pnas.091613398
Kovall R Matthews BW Toroidal structure of lambda-exonuclease Science 1997 277 1824 1827 9295273 10.1126/science.277.5333.1824
Kovall RA Matthews BW Type II restriction endonucleases: structural, functional and evolutionary relationships Curr Opin Chem Biol 1999 3 578 583 10508668 10.1016/S1367-5931(99)00012-5
Pingoud A Jeltsch A Structure and function of type II restriction endonucleases Nucleic Acids Res 2001 29 3705 3727 11557805 10.1093/nar/29.18.3705
Skirgaila R Grazulis S Bozic D Huber R Siksnys V Structure-based redesign of the catalytic/metal binding site of Cfr10I restriction endonuclease reveals importance of spatial rather than sequence conservation of active centre residues J Mol Biol 1998 279 473 481 9642051 10.1006/jmbi.1998.1803
Feder M Bujnicki JM Identification of a new family of putative PD-(D/E)XK nucleases with unusual phylogenomic distribution and a new type of the active site BMC Genomics 2005 6 21 15720711 10.1186/1471-2164-6-21
Bujnicki JM Rychlewski L Unusual evolutionary history of tRNA splicing the endonuclease EndA: relationship to the LAGLIDADG and PD-(D/E)XK deoxyribonucleases Protein Sci 2001 10 656 660 11344334 10.1110/ps.37101
Li H Trotta CR Abelson J Crystal structure and evolution of a transfer RNA splicing enzyme Science 1998 280 279 284 9535656 10.1126/science.280.5361.279
Todone F Weinzierl RO Brick P Onesti S Crystal structure of RPB5, a universal eukaryotic RNA polymerase subunit and transcription factor interaction target Proc Natl Acad Sci U S A 2000 97 6306 6310 10841537 10.1073/pnas.97.12.6306
Daiyasu H Komori K Sakae S Ishino Y Toh H Hjc resolvase is a distantly related member of the type II restriction endonuclease family Nucleic Acids Res 2000 28 4540 4543 11071943 10.1093/nar/28.22.4540
Aravind L Makarova KS Koonin EV Holliday junction resolvases and related nucleases: identification of new families, phyletic distribution and evolutionary trajectories Nucleic Acids Res 2000 28 3417 3432 10982859 10.1093/nar/28.18.3417
Bujnicki JM Rychlewski L Grouping together highly diverged PD-(D/E)XK nucleases and identification of novel superfamily members using structure-guided alignment of sequence profiles J Mol Microbiol Biotechnol 2001 3 69 72 11200231
Bujnicki JM Crystallographic and bioinformatic studies on restriction endonucleases: inference of evolutionary relationships in the "midnight zone" of homology Curr Protein Pept Sci 2003 4 327 337 14529527 10.2174/1389203033487072
Yu M Souaya J Julin DA The 30-kDa C-terminal domain of the RecB protein is critical for the nuclease activity, but not the helicase activity, of the RecBCD enzyme from Escherichia coli Proc Natl Acad Sci USA 1998 95 981 986 9448271 10.1073/pnas.95.3.981
Singleton MR Dillingham MS Gaudier M Kowalczykowski SC Wigley DB Crystal structure of RecBCD enzyme reveals a machine for processing DNA breaks Nature 2004 432 187 193 15538360 10.1038/nature02988
Alexandrov N Shindyalov I PDP: protein domain parser Bioinformatics 2003 19 429 430 12584135 10.1093/bioinformatics/btg006
Holm L Sander C Dali: a network tool for protein structure comparison Trends Biochem Sci 1995 20 478 480 8578593 10.1016/S0968-0004(00)89105-7
Shindyalov IN Bourne PE Protein structure alignment by incremental combinatorial extension (CE) of the optimal path Protein Eng 1998 11 739 747 9796821 10.1093/protein/11.9.739
Cheng X Balendiran K Schildkraut I Anderson JE Structure of PvuII endonuclease with cognate DNA EMBO J 1994 13 3927 3935 8076590
Bunting KA Roe SM Headley A Brown T Savva R Pearl LH Crystal structure of the Escherichia coli dcm very-short-patch DNA repair endonuclease bound to its reaction product-site in a DNA superhelix Nucleic Acids Res 2003 31 1633 1639 12626704 10.1093/nar/gkg273
Altschul SF Madden TL Schäffer AA Zhang J Zhang Z Miller W Lipman DJ Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694 10.1093/nar/25.17.3389
Subramanya HS Bird LE Brannigan JA Wigley DB Crystal structure of a DExx box DNA helicase Nature 1996 384 379 383 8934527 10.1038/384379a0
Chedin F Kowalczykowski SC A novel family of regulated helicases/nucleases from Gram-positive bacteria: insights into the initiation of DNA recombination Mol Microbiol 2002 43 823 834 11929535 10.1046/j.1365-2958.2002.02785.x
Quiberoni A Biswas I El Karoui M Rezaiki L Tailliez P Gruss A In vivo evidence for two active nuclease motifs in the double-strand break repair enzyme RexAB of Lactococcus lactis J Bacteriol 2001 183 4071 4078 11395472 10.1128/JB.183.13.4071-4078.2001
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
Kleywegt GJ Use of non-crystallographic symmetry in protein structure refinement Acta Cryst 1996 D52 842 857
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
Jones TA Zou JY Cowan SW Kjeldgaard M Improved methods for building protein models in electron density maps and the location of errors in these models Acta Cryst 1991 A47 110 119
PyMOL Home Page
Tatusov RL Natale DA Garkavtsev IV Tatusova TA Shankavaram UT Rao BS Kiryutin B Galperin MY Fedorova ND Koonin EV The COG database: new developments in phylogenetic classification of proteins from complete genomes Nucleic Acids Res 2001 29 22 28 11125040 10.1093/nar/29.1.22
Clamp M Cuff J Searle SM Barton GJ The Jalview Java alignment editor Bioinformatics 2004 20 426 427 14960472 10.1093/bioinformatics/btg430
Edgar RC MUSCLE: multiple sequence alignment with high accuracy and high throughput Nucleic Acids Res 2004 32 1792 1797 15034147 10.1093/nar/gkh340
Notredame C Higgins DG Heringa J T-Coffee: A novel method for fast and accurate multiple sequence alignment J Mol Biol 2004 302 205 217 10.1006/jmbi.2000.4042
Gouet P Courcelle E Stuart DI Metoz F ESPript: analysis of multiple sequence alignments in PostScript Bioinformatics 1999 15 305 308 10320398 10.1093/bioinformatics/15.4.305
Velankar SS Soultanas P Dillingham MS Subramanya HS Wigley DB Crystal structures of complexes of PcrA DNA helicase with a DNA substrate indicate an inchworm mechanism Cell 1999 97 75 84 10199404 10.1016/S0092-8674(00)80716-3
|
15985153
|
PMC1185551
|
CC BY
|
2021-01-04 16:03:52
|
no
|
BMC Struct Biol. 2005 Jun 28; 5:9
|
utf-8
|
BMC Struct Biol
| 2,005 |
10.1186/1472-6807-5-9
|
oa_comm
|
==== Front
Biomed Eng OnlineBioMedical Engineering OnLine1475-925XBioMed Central London 1475-925X-4-361595523510.1186/1475-925X-4-36ResearchComputational optical biopsy Li Yi [email protected] Ming [email protected] Ge [email protected] Department of Mathematics, University of Iowa, Iowa City, IA 52242, USA2 Bioluminescence Tomography Laboratory, Departments of Radiology and Biomedical Engineering, University of Iowa, Iowa City, IA 52242, USA2005 14 6 2005 4 36 36 1 4 2005 14 6 2005 Copyright © 2005 Li et al; licensee BioMed Central Ltd.2005Li 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.
Optical molecular imaging is based on fluorescence or bioluminescence, and hindered by photon scattering in the tissue, especially in patient studies. Here we propose a computational optical biopsy (COB) approach to localize and quantify a light source deep inside a subject. In contrast to existing optical biopsy techniques, our scheme is to collect optical signals directly from a region of interest along one or multiple biopsy paths in a subject, and then compute features of an underlying light source distribution. In this paper, we formulate this inverse problem in the framework of diffusion approximation, demonstrate the solution uniqueness properties in two representative configurations, and obtain analytic solutions for reconstruction of both optical properties and source parameters.
Molecular imagingoptical imagingfluorescent imagingbioluminescent imaginginverse problemdiffusion approximation
==== Body
Introduction
Gene therapy is a breakthrough in the modern medicine, which promises to cure diseases by modifying gene expression. A key for development of gene therapy is to monitor the in vivo gene transfer and its efficacy in the mouse model. Traditional biopsy methods are invasive, insensitive, inaccurate, inefficient, and limited in the extent. To map the distribution of the administered gene, reporter genes such as those producing luciferase are being used to generate light signals within a living mouse, which can be externally measured [1]. A cooled highly sensitive CCD camera has been built to take a 2D view of expression of the bioluminescent reporter luciferases. Such a 2D image of photon emission is then superimposed onto a 2D visible light picture of the mouse for localization of the reporter gene activity. In addition to gene therapy, this new imaging tool has great potentials in various other biomedical applications as well. An in vivo bioluminescence tomography system integrated with an X-ray CT/micro-CT scanner is recently reported in [2,3]. The novel concept is to collect emitted photons from multiple 3D directions with respect to a living mouse marked by bioluminescent reporter luciferases, and reconstruct an internal bioluminescent source distribution based on both the outgoing bioluminescent signals and the CT/micro-CT volume of the mouse. Then, the 3D bioluminescent source distribution and the corresponding CT/micro-CT volume are registered of anatomical and pathological structures, such as the lung and various tumors.
Optical imaging of small animals based on fluorescent/bioluminescent probes promises great opportunities for translational research and eventually clinical applications, because fluorescent/ bioluminescent signals directly reveal molecular and cellular activities, and are sensitive, specific, non-ionizing, non-invasive and cost-effective. Pure optical imaging cannot detect the molecular activities triggered by biomarkers because the light generated are generally out of the visible spectrum. Despite the progress in optical molecular imaging of small animals, little research has been done for optical molecular imaging of patients. A light source induced by either fluorescence or bioluminescence probes is usually weak, and would be often deep inside a body should it be used in patients. Optical methods for in vivo imaging are all faced with the problem of limited transmission of light through tissues [[1], p. 237]. Because the human body absorbs and scatters photons in the visible and near infrared ranges with the mean-free-path in the sub-millimeter domain, such a source cannot be effectively detected on the body surface [4].
In this paper, we propose a computational optical biopsy (COB) method [5] to supplement and enhance the capabilities of fluorescent molecular tomography and bioluminescence tomographic, especially for their potential uses in patients. In order to detect the light source in a region of interest deep inside a subject, we can use a fiber-optical probe to detect the light source directly in the subject along one or multiple biopsy paths and next to compute the parameters or features of the embedded light source.
Several optical biopsy needle systems are already in operation [6-8], which have indicated the physical feasibility of this COB project. While similar to existing optical biopsy procedures in using fiber-optic probes [6-8], the proposed COB system and methods depend on not only optical devices but also advanced modeling and computation techniques to reconstruct an underlying source intensity distribution and extract its features of interest such as source center and effective intensity. There are several distinctions that substantiate our innovations. Our COB approach relies on sophisticated signal modeling and estimation from data collected along a number of biopsy paths, while other biopsy/endoscopic techniques perform direct anatomical imaging on speci c spots only. Our COB targets source intensity distributions triggered by probes instead of tissue/vascular properties that are concerned by other biopsy/endoscopic methods. Our COB intends to sense both fluorescent and bioluminescent sources, not just fluorescent sources as some optical endoscopic/spectroscopic techniques are designed for.
Mathematically we will Consider
where u0 is the average photon flux in all directions, , μα are positive constants with μa and μs being the absorption and scattering constants respectively in and f is either a measure or a L∞ ∩ L1 function, which represents the light source.
Single point source
We will first consider the case where the source term , where δ is the Dirac operator. In this case we have
where C is a constant such that
or
i.e.,
Assume that the needle insertion follows a straight line l : , where is the direction of the insertion and one point along the insertion (Figure 1).
Figure 1 A needle insertion path, where t is the direction of the insertion, 0 one point along the insertion, v a direction in which the needle tip detects light, s1; s2 are points along the path
Let υ be a direction that needle tip detects light from. Then the measurement along the line l is given by
and = (t1, t2, t3). υ = (υ1, υ2, υ3) which has a fixed angle θ with the needle direction , i.e., cos (θ) = t1υ1 + t2υ2 + t3υ3
such that
Due to the notation and translation invariant properties of (1) and the design of the insertion needle, we may assume, without loss of generality, that
so that υ1 = 0, (υ2υ3) = (cos α, sin α) and . Then (7) becomes
Now let .
Assuming that each insertion line can detect sources from two different angles α1, α2 such that α1 - α2 ≠ kπ, k ∈ ± , then
is non-singular. Therefore appropriate combination α1 and α2 depending on γ of the measurements m(s) would single out , so that without loss of generality, we may assume that
which are the measurements such that <x - xi, υ > = , k = 2, 3 after combinations of α1 and α2 mentioned above and define
Theorem 2.1
(Single point source) If N = 1, i.e.: if , then one insertion would uniquely determine λ, x1, and μeff, provided that the insertion line does not go through the point of the source. (See remark.)
Proof
In this case since
Differentiate (13) with respect to s, we have
such that the critical point of (s) uniquely identifies x1. Next with x1 identified, we obtain
or
And by taking the derivatives of and evaluating them at x1, we obtain that
where z = μeffw.
From (18) we obtain
and plugging it into (19) we have
If f (z) is monotone in z > 0, then (21) would give us a unique solution z > 0.
We compute and obtain
i.e., we have uniquely solved by the information from the values of μeff, w and therefore λ by (17).
Next if we let x2 = w cos β, x3 = w sin β, then we get
which could uniquely determine β by the sign conditions of , k = 2.3.
That is, by now, we are able to determine, (x1, x2, x3), μeff and λ; i.e. complete information about the source.
Remark
In case that the insertion line goes through the point of the source, which is verified by having each
{some single point}
then only the two components of the point source, orthogonal to the insertion can be identified, which is the best possible. On the other hand, in practice, such events would have probability zero!
Single ball source
Next we consider ball sources, i.e., we assume that
where χΩ is the characteristic function of Ω and .
For such source we have where
where
– the total intensity of the source in the form of λχB(o, r). (24)
Again we will discuss the single ball source first, i.e. N = 1, so that we have
Theorem 3.1
(Single ball source) If N = 1 ; i.e. if u0 is given by (25), then one insertion would uniquely determine M (λ3, r1), x1 and μeff; in case the line stays outside of B (x1, r1). In case the insertion line enters the interior of B (x1, r1), then x1, r1, λ1 and μeff are uniquely determined by one insertion.
Proof
Case 1
If the insertion line (refer to Theorem 1.1 in Section 2) stays outside B (x1, r1), then the detected u0 is given by
i.e., u0 (x) behaves exactly as a single point source with an intensity of M (λ1, r1) and hence Theorem 2.1 guarantees the result.
Case 2
If the insertion line enters the support of the ball source B (x1, r1) but misses the center point x1, then again due to the rotation and translation invariant properties of (1), we may assume, without loss of generality that the insertion is given by (8) and x1 = (x1, x2, x3) and hence the detection given by (6) is now
m (s) (27)
where .
Again, assuming that each insertion line can detect sources from two different angles α1, α2 such that α1 - α2 ≠ kπ, k ∈ ± , then
is non-singular. Therefore appropriate combination of the two measurements depending only on α1 and α2 would single out and , so that without loss of generality we may assume that
mk (s) = φ (s), where
If we let x2 = w cos β, w3 = w sin β, then
which would uniquely determine β by the above question and the sign conditions if , k = 2.3.
Define
(s) = m2 (s)2 + m3 (s)2 = w2φ2 (s). (30)
Note that since in this case we assume that the insertion line enters the interior of B (x1, r1) but misses the center x1, we have 0 <w <r1 and the line would enter and leave B (x1, r1) when
which is exactly where our measurement Mk (s), or equivalently φ (s) fails to be differentiable. Therefore let sin and sout be two points where we observe the jump discontinuity of 1 (s).
Then we have
(32) together with (29) uniquely identify x1 and establish a relation between r1 and w. Then we are able to evaluate (x1) to get
i.e.,
Next by taking the derivatives of and evaluating them at x1, we obtain that
>From (35) we obtain
and plugging it into (36) we get
which can be veri ed to be an increasing function. Therefore (38) has a unique solution z and (37) then uniquely de nes the w which in turn defines λ1 uniquely by (34) and r1 uniquely by (32).
Discussions and conclusion
We have demonstrated the modeling of computational optical biopsy with the diffusion optics, the solution uniqueness properties in two typical configurations and provide explicit formulas for the reconstruction of both optical properties and source parameters. Mathematically, one single insertion will be enough to estimate the above parameters. However, physically, more measurements will guarantees the robustness of the estimates. The needle trajectory can also be dynamically restarted and optimized towards the center of the source based on the successive estimate. Further investigation on multiple point sources and multiple ball sources is undergoing. For example the double sources probelm can be handled in such a way that by using the moments defined (11) one could reduce the problem to a equivelent one with a few freedom and thus a numerical optimization technique will be used to solve the inverse problems.
Contributions
The three authors made about equal contributions in this work.
6 Acknowledgements
This work is partially supported by the National Institutes of Health (EB001685 and EB002667).
==== Refs
Contag CH Bachmann MH Advances in vivo bioluminescence imaging of gene expression Annual Review Of Biomedical Engineering 2002 4 235 260 12117758 10.1146/annurev.bioeng.4.111901.093336
Wang G Development of the first bioluminescent tomography system Radiology Suppl (Proceedings of the RSNA) 2003
Wang G Li Y Jiang M Uniqueness theorems for bioluminescent tomography Medical Physics 2004 31 2289 2299 15377096 10.1118/1.1766420
Beuthan J Mahnke C Netz U Minet O MÃller G Optical molecular imaging: Overview and technological aspects Medical Laser Application 2002 17 25 30
Wang G Li Y Jiang M Computational optical biopsy methods, techniques and apparatus. Provisional patent application filed in March 2005 (Patent disclosure filed with Univ. of Iowa Research Foundation in December 2003)
Liu Q Ramanujam N Experimental proof of the feasibility of using an angled fiber-optic probe for depth-sensitive fluorescence spectroscopy of turbid media Optics Letters 2004 29 2034 2036 15455771 10.1364/OL.29.002034
Li XD Chudoba C Imaging needle for optical coherence tomography Optics Letters 2000 25 1520 1522
Alfano RR Katz A Optical biopsy V: 27–28 January San Jose, USA Progress in biomedical optics and imaging; v 5, no 15 2004, Bellingham, Wash: SPIE ix 134
|
15955235
|
PMC1185552
|
CC BY
|
2021-01-04 16:37:35
|
no
|
Biomed Eng Online. 2005 Jun 14; 4:36
|
utf-8
|
Biomed Eng Online
| 2,005 |
10.1186/1475-925X-4-36
|
oa_comm
|
==== Front
Cancer Cell IntCancer Cell International1475-2867BioMed Central London 1475-2867-5-201600198010.1186/1475-2867-5-20Primary ResearchChanges in P-glycoprotein activity are mediated by the growth of a tumour cell line as multicellular spheroids Valeria Ponce de León [email protected]úl Barrera-Rodríguez [email protected] Depto. de Bioquímica. Instituto Nacional de Enfermedades Respiratorias-SSA México. Clza. Tlalpan, 4502, C.P. 14080, México, D.F2005 7 7 2005 5 20 20 29 3 2005 7 7 2005 Copyright © 2005 Valeria and Raúl; licensee BioMed Central Ltd.2005Valeria and Raúl; 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
Expression of P-glycoprotein (P-gp), the multidrug resistance (MDR) 1 gene product, can lead to multidrug resistance in tumours. However, the physiological role of P-gp in tumours growing as multicellular spheroids is not well understood. Recent evidence suggests that P-gp activity may be modulated by cellular components such as membrane proteins, membrane-anchoring proteins or membrane-lipid composition. Since, multicellular spheroids studies have evidenced alterations in numerous cellular components, including those related to the plasma membrane function, result plausible that some of these changes might modulate P-gp function and be responsible for the acquisition of multicellular drug resistance. In the present study, we asked if a human lung cancer cell line (INER-51) grown as multicellular spheroids can modify the P-gp activity to decrease the levels of doxorubicin (DXR) retained and increase their drug resistance.
Results
Our results showed that INER-51 spheroids retain 3-folds lower doxorubicin than the same cells as monolayers however; differences in retention were not observed when the P-gp substrate Rho-123 was used. Interestingly, neither the use of the P-gp-modulating agent cyclosporin-A (Cs-A) nor a decrease in ATP-pools were able to increase DXR retention in the multicellular spheroids. Only the lack of P-gp expression throughout the pharmacological selection of a P-gp negative (P-gpneg) mutant clone (PSC-1) derived from INER-51 cells, allow increase of DXR retention in spheroids.
Conclusion
Thus, multicellular arrangement appears to alter the P-gp activity to maintain lower levels of DXR. However, the non expression of P-gp by cells forming multicellular spheroids has only a minor impact in the resistance to chemotherapeutic agents.
P-GlycoproteinMulticellular spheroidsMulticellular drug resistanceNSCLC
==== Body
Background
Multidrug resistance to chemotherapy is one of the biggest problems in the treatment of cancer. Currently, the best understood mechanism of multidrug resistance (MDR) is associated with the overexpression of protein efflux-pump known as P-glycoprotein (P-gp), but other non-Ppg mechanisms are also involved (i.e. MRP1, topoisomerases, glutathione-S transferases, etc). The P-gp is the protein product of the MDR-1 gene and is expressed as a transmembranal protein (Mr 170 000) capable of decreasing the intracellular concentration of a broad range of cytotoxic agents in an energy-dependant mediated efflux [1,2]. Overexpression of P-gp in human cell lines confers resistance to many of the most effective chemotherapeutic agents used clinically in chemotherapy, including anthracyclines (e.g., doxorubicin (DXR)), Vinca alkaloids (e.g., vincristine), epipodophyllotoxins (e.g., etoposide), actinomycin D, paclitaxel, as well as many others non-chemotherapeutic agents like Rhodamine-123 and ethidium bromide [3]. Although the more accepted model is that the P-gp by itself extrude chemotherapeutic agents out of the cells, more recent studies suggest that P-gp activity may be modulated by cellular components such as membrane proteins, membrane-anchoring proteins or the composition of lipids themselves [4-7].
Since the first studies by Sutherland et al in 1979 [8], it was shown that tumour cells growing as multicellular spheroids resembles many of the behaviours found in solid tumours, including multicellular resistance (MCR) [9,10]. Using the model of multicellular spheroids, several authors have shown that P-gp, is more efficient to conferring resistance in cells cultivated as spheroids as compared to cells cultivated as monolayers [11-13]. From these observations, one question arises: Can tumour cells modulate its P-gp activity as a direct consequence of the environmental condition where grown? To address this question, we studied an NSCLC cell line named INER-51 that showed a P-gp-mediated resistance to DXR in the spheroid model. In early studies with INER-51 cells, we found that the formation of multicellular spheroids does not show any increase in mRNA for MDR-1 gene or in a differential P-gp expression in the specific areas of the spherule.
Results
Doxorubicin and Rhodamine-123 retention in spheroids
It is well known that multicellular spheroids are more resistant to chemotherapeutic drugs than the same cell cultures as monolayers [11,12]. With the aim to evaluate differences in P-gp activity in multicellular spheroids with respect to monolayers, two P-gp substrates were used (DXR and Rho-123). Both compounds were chosen, because they are good P-gp substrates with an autofluorescence capacity. Our results showed that in monolayers the amount of DXR retained in the cells was in direct proportion to the drug added to the medium (Figure 1). In contrast, multicellular spheroids showed a lower capacity for DXR retention (3-fold lower) than monolayers. Interestingly, this poor retention was not observed when Rho-123 was used as P-gp substrate because the Rho-123 levels retained were equal in both types of cultures (box insert in Figure 1a).
Figure 1 Dose-dependence of doxorubicin-retention in INER-51 cells. (a) Tumour cells growth as monolayers (filled circles) or multicellular spheroids (open circles) were incubated with increased concentrations of DXR or Rho-123 (Insert box) for 30 min and then intracellular fluorescence was determinated by spectrofluormetry. b) Time-course of DXR efflux in INER-51 cells. Monolayers (filled circles) and a multicellular spheroids (open circles) were previously loaded with 30 μM of DXR for 30 min and afterwards the remaining DXR was quantitated in several intervals of time. Each point represents the mean of at least 3 experiments and error bars are the standard error of the mean.
Since the drug-retention is a dynamic process that involves drug uptake (simple diffusion) as well as drug linkage (diffusion vs. expulsion mediated for an active mechanism), we tried to determine whether a more efficient P-gp-dependent drug removal was responsible for allowing lower intracellular DXR concentrations. Thus, drug-removal in monolayers previously loaded with 30 μM of DXR during 30 min showed a first order decay reaction with a half-time of drug concentration at the first 5 minutes interval that indicated the presence of an active transport mechanism (Figure 1b). However, for multicellular spheroids the presence of active transport was not evident since its was not possible to load the cells with sufficient amounts of DXR to determine the efflux values. A common observation in these experiments was that a constant amount of DXR was retained for a longer period of time (45 minutes) in cells grown as monolayers. A possible explanation could be from the presence of positively charged-DXR, which stores in acidic vesicles as chromaffin granules and lysosomes [20,21].
Circumvention of DXR retention with P-pg reversal agents
With the aim to determine how much the P-gp influences the levels of DXR retained, the modulator agent Cs-A was employed. As shown in the Figure 2, the incubation of monolayers with 5 μM of Cs-A efficiently enhances (2-fold) the intracellular DXR fluorescence in direct relation to reversal agent concentration. Surprisingly, the P-gp-modulation activity of Cs-A was not evident in multicellular spheroids because no effect in the intracellular DXR retention could be seen. Also, other reversal agents, like SDZ PSC 833 and verapamil were not able to increase DXR retention, neither in monolayers nor in multicellular spheroids (data not shown).
Figure 2 Circumvention of doxorubicin-retention by cyclosporin-A (Cs-A). INER-51 cells growth as monolayers (filled circles) or multicellular spheroids (open circles) were treated with increase concentrations of Cs-A for 1 hour previous to being loaded with 30 μM of DXR for 30 min in presence of reversal agent. Doxorubicin retained was determinate by spectrofluormetry. Each point represents the mean of at least 3 experiments and the error bars are the standard error of the mean.
Effect of ATP-depletion in DXR retention
P-gp is a protein that belongs to the ABC binding cassette protein, for which efficient drug efflux needs ATP hydrolysis. Thus, with the propose to achieve more information about the P-gp function, we decided to inhibit the P-gp activity through depletion of ATP-pools. Three metabolic poisons were used to be sure of complete ATP-depletion in multicellular spheroids. Figure 3 shows that ATP-depletion did induce a 1.5-fold increase of DXR retention in cells maintained as monolayers but none effect was again seen when in ATP-depleted multicellular spheroids.
Figure 3 Effect of ATP-depletion in doxorubicin-retention. INER-51 cells as monolayers or multicellular spheroids were pre-incubated (open symbols) or not (filled symbols) with 1 mM sodium-cyanide, 10 mM sodium-fluoride, 10 mM sodium azide before the addition of increased DXR concentrations. The intracellular DXR-fluorescence was evaluated by spectrofluormetry. Each point represents the mean of at least 3 experiments and the error bars are the standard error of the mean. The Student's t-test for paired data was performed to identify changes in DXR retention. The results were considered to be significant when p < 0.05. NS: no significant.
Achieving the PSC-1 mutant clone non-expressing P-gp
With no possibility of obtaining irrefutable evidence of P-gp modulation from multicellular spheroids, we decided to eliminate the P-gp expression from the lung cancer line. Therefore, through the co-selection of the parental cell line INER-51 with 5 μM of DXR and 5 μM of SDZ PSC 833, we were able to obtain one mutant cell clone that did not express P-gp, which was named PSC-1 line (Figure 4a). In vitro RT-PCR analysis of other MDR-related genes did not show qualitative differences between parent and mutant cells (Figure 4b). The amplification experiments showed that both cells lines express with approximately with the same intensity transcripts for topoisomerase I, topoisomerase IIα and topoisomerase IIβ but neither of them expressed the multidrug resistance-associated protein (MRP1) or glutathione-S transferase-μ.
Figure 4 RT-PCR analysis of the levels of expression for MDR genes in the parental cell line INER-51 as well as the PSC-1 cell clone. a) Amplification of MDR-1 gene visualized by ethidium bromide (left) or autoradiagraph (middle). The expression of G3PDH gene was used as a constitutive control for the integrity of the RNA molecules. b) Amplification by RT-PCR of other non-MDR-1 genes related to MDR phenotype. As control (CTL) of gene expression, the next cell lines and tissues were used: INER-37 cell line for MRP1 and GST-μ; A427 cell line for MDR-1; HeLa cell line for topoisomerase I, topoisomerase IIα and topoisomerase IIβ; finally, placental tissue was used as a control of BCRP expression.
DXR retention and drug cytotoxicity in PSC-1 cells
The similarities between INER-51 cells (P-gppos) and PSC-1 cells (P-gpneg), allowed us the possibility to evaluate whether or not the lower DXR-retention levels were mediated by a positively modulated P-gp-mechanism. The incubation of PSC-1 spheroids with increasing DXR concentrations showed a significative increase in the drug retention (1.8-fold) in comparison to INER-51 spheroids and was only 1.6-fold lower than PSC-1 growth as monolayers (Figure 5). When Rho-123 was assayed, an increase in the retention of the dye relative to the parental line INER-51 as monolayers was evident. In PSC-1 spheroids, the levels of Rho-123 retention were similar in both INER-51 monolayers and multicellular spheroids (box insert in Figure 5).
Figure 5 Comparative doxorubicin retention of PSC-1 cells and INER-51 cells. PSC-1 tumour cells growth as monolayers (filled circles) or multicellular spheroids (filled triangles) and INER-51 monolayers (open circles) or multicellular spheroids (open triangles) were incubated with increased concentrations of DXR (as in Figure 1) and then intracellular fluorescence was determined by spectrofluormetry. Insert box. Rhodamine-123-retention in PSC-1 cell growth as monolayers or multicellular spheroids.
Several techniques have been used to evaluate the MCR in multicellular spheroids. Since in previous experiments we were unable to successfully disaggregate multicellular spheroids through trypsin-based protocols, we decided to evaluate the MCR using the MTT assay (which as been previously evaluated by Furukawa et al[17]. Interestingly, this assay showed that the increase in DXR retention had only a minor impact in the MCR. Thus, PSC-1 spheroids were able to maintain their resistance to DXR or etoposide as the parental INER-51 cells (Figure 6a,b) with only minor changes in the IC50 values (Table 1). Only when PCS-1 cells were assayed against the cytotoxic effect of methotrexate (which use another detoxification via previous to P-gp), differences between INER-51 cells and PSC-1 cells could be seen (Figure 6c).
Figure 6 Cytotoxicity of doxorubicin (a), etoposide (b) and methotrexate (c) in monolayers or multicellular spheroids from PSC-1 (P-gpneg) clone cells in comparison to the parental INER-51 cell line. Cytotoxicity was measurement by the MTT colorimetric assay and expressed as percent growth inhibition in comparison with the untreated tumour cells. Each point represents the mean of at least 3 independent experiments and error bars are the standard error of the mean
Table 1 Indices of cytotoxicity (IC50) to several cancer drugs founded in multicellular spheroids and monolayer.
Drug Monolayers IC50 (μM) Multicell. spheroids IC50 (μM)
INER-51 PCS-1 RR INER-51 PCS-1 RR
Doxorubicin 6.2 9.2 1.4 15 9.8 0.6
Etoposide 2.2 4.1 1.8 >50 >50 ----
Methotrexate 0.8 2.8 3.5 >50 >50 ----
Drug exposition was continuous for 72 hours and drug cytotoxicity was performed by MTT assay. IC50: Fifty percent of the decrease in the survival rate. RR: The value of relative resistance of INER-51 cells vs PSC-1 cells as monolayers or multicellular spheroids.
Discussion
It is well known that the three-dimensional arrangement evoke deeper structural changes in the cells to maintain the integrity of the multicellular structure, some of which include: a) the expression of proteins of ECM [22]; b) membrane protein anchoring [23,24]; c) heat shock proteins [25] and d) adhesion proteins as well as changes in lipid membrane composition and hypoxia [26-29]. Another phenomenon frequently observed in tumour cells growing as multicellular spheroids is the acquisition of MCR. In spheroids cultures as well in monolayer systems one of the major mechanisms to confer resistance is shown by the expression of P-gp. However, some evidence suggests that to confer MCR, the P-gp also seems to work more efficiently [11,12]. Recent data suggests that P-gp activity can be modulated through the interaction with diverse cellular components, some of which are also altered when cells are multicellular spheroids [4-7,11,12]. Thus, it would be interesting to know if to acquire MCR, tumour cells spheroids can modulate its P-gp activity.
Our findings suggest that a more efficient P-gp-mediated efflux seems to be responsible for maintaining lower levels of DXR in INER-51 spheroids than cells in monolayer cultures. However, this P-gp mediated drug efflux did not seem to operate under the same conditions with others substrates, because when the lipophilic cation Rho-123 was used, the levels of retention were similar and independent of the culture conditions.
Under the spheroid condition, P-gp appears to obey different regulatory mechanisms since neither Cs-A treatment nor ATP-depletion were able to increase the levels of DXR into the spherules. Other non-P-gp mediated mechanisms also appear to be operating to maintain lower DXR levels in the multicellular spheroids, because the lack of P-gp expression did not reach levels of DXR comparable to the monolayer's cultures and had only a minor impact in the acquisition of MCR to chemotherapeutic agents, including DXR.
Some evidences has shown that the membrane-cell composition can modulate the transbilayer movement rate of MDR-type drugs across the membrane and consequently affects the "competition" between the active P-gp-mediated drug efflux and the passive drug uptake: i.e. retention [30]. Particularly, DXR has shown to have a lower rate of penetration through membranes due to its specific interactions with cardiolipin [31,32]. Since membrane changes affecting plasma membrane do not confer resistance themselves but could drive the P-gp function [33-35], we hypothesize a possible drawing to understand how P-gp successfully maintains lower DXR concentrations. In this picture, a delay in the rate of passive transbilayer movement of DXR through the plasma membrane results in an enhanced ability of P-gp to recognize and extrude it out of the spheroids. Thus, when P-gp is absent (as in PSC-1 cells), the saturation of lipid targets permit the increase of the DXR concentration into the multicellular spheroids. Enhanced efficiency of P-gp to extrude substrates has been fully demonstrated by different members of the DXR and rhodamine dry analogues, with each one having different lipophilic properties [36].
Several studies have shown that the use of modulator agents provoke the sensitisation of cell growth as multicellular spheroids [37-39]. The efficacy of these agents is due to the higher permeability rate in relation to P-gp substrates [40]. However, when INER-51 spheroids were pre-treated with 10 μM of the modulator agent of Cs-A, no increase in DXR retention was observed. A similar phenomenon was observed in multicellular spheroids treated with verapamil [38,39]. Today, a number of studies using various techniques have suggested a probable model of interaction between P-gp-substrates with modulators agents and how these molecules bind to different sites on P-gp [41]. Four binding sites have been identified on P-gp of which vinblastine and daunorubicin bind to the site I, whereas modulators bind to site IV. In this model, all the binding sites display allosteric interaction between each one of them, thus affecting P-gp-mediated activity. In addition, an unrecognised binding site has been proposed for Rho-123 [42-44]. Thus, the failure of Cs-A to inhibit P-gp-mediated DXR efflux may be a consequence of changes in the P-gp affecting the modulator-binding site or the allosteric inhibition with some hypothetic molecule. Recently, in cells with MDR-phenotype, it has been demonstrated that a specific interaction exists between P-gp and HSP-90 [45]. If spherules formation elicits the expression of HSP-90, which is able to bind P-gp, the failure of Cs-A inhibition might be understood. Furthermore, the allosteric inhibition of the modulator site by unknown factors could be an explanation for the common failure seen when solid tumours, including lung cancer tumours, are treated with reversal agents [46].
Another interesting observation came from the impact of ATP depletion in DXR retention. For this experiment we used a metabolic cocktail poison to be certain of the ATP pools depletion. However, this treatment was unable to increase DXR retention in multicellular spheroids but its effect was evident in monolayers. Since, there is a general consensus in the literature that P-gp is an energy-driven pump, where the energy is provided by the hydrolysis of ATP, our data are puzzle. Nevertheless, using competitions assays, Biswas EE [47], demonstrated that NBD1 of histidine permease (Hisp) and maltose transporter protein (Malk), two ABC members, can function as a general nucleotide binding domain, with a nucleotide preference CTP>GTP>ATP>>UTP. Thus, the impact of ATP depletion in DXR retention is not understood for the moment.
In recent years, it has become clear that the multiple non-P-gp mediated mechanisms may be operational in tumour spheroids to confer MCR [48]. One mechanism most recently described is the breast cancer resistance protein (BCRP), which is a drug pump efflux able to efficiently extrude DXR [49]. Since INER-51 cells express mRNA for BRCP, we cannot discard the activity of this protein as a mechanism that helps to maintain lower DXR retention levels. However, the activity of BCRP as pump efflux is higher for Rho-123 efflux than DXR, although their contribution to the drug retention in spheroids is controversial [50].
In addition to DXR resistance, INER-51 spheroids also showed resistance to other anticancer drugs such as etoposide and methotrexate. Commonly, the etoposide-resistance involves alterations in the nuclear target enzymes topoisomerases. Oloumi et al[51] indicate that alterations in subcellular localization of topoisomerases type II may have an important role in resistance to cytotoxic agents when cells are in a close contact. Whereas Luo et al[52] showed that phosphorylation of topo II alpha was reduced at least 10-fold in the outer cells of V79 spheroids relative to monolayers. However, the role of topoisomerase II alpha as a mechanism to confer MCR to etoposide in INER-51 spheroids has not been evaluated yet and will be considered for future experiments.
A different situation can be speculated for the resistance to methotrexate. Under physiological conditions, the weak acid methotrexate tends to be in the charged form and is taken up into cells largely by a folate transport mechanism. Unlike DXR, methotrexate is not sequestered in membranes nor in acidic endosomes, but it may be "trapped" inside cells by polyglutamation [53]. Thus, the differences in methotrexate resistance of INER-51 cells relative to PSC-1 cells could arise from differences in P-gp-independent metabolic pathways.
Finally, several reports have shown a lower MDR-1 expression in NSCLC [54]. However, some authors have shown an important role of P-gp expression in lung cancer tumours and particularly in lung cancer diagnosed in people whom smoke tobacco [55-57]. Due to our findings, it would be interesting to evaluate both the levels of P-gp expression as well as P-gp activity in lung cancer tumours.
Conclusion
Our results suggest that in INER-51 cells cultured as multicellular spheroids, a more efficient P-gp activity is responsible for maintaining lower retention levels of DXR in comparison to the P-gp activity of cells grown as monolayers. Interestingly, whereas the P-gp expression helps maintain lower levels of DXR in the multicellular spheroids, the mechanisms that govern the MCR seems to be different because the lack of P-gp-expression only showed a minor impact of resistance to several chemotherapeutic drugs, suggesting that other non-P-gp mechanisms are also operating.
Methods
Chemicals
Doxorubicin (DXR) was provided by Farmitalia-Carlo Erba, whereas Cyclosporin-A was provided by Sandoz Farma, whereas SDZ PSC 833 was gift by Novartis. Rhodamine-123 (Rho-123) and 3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyl tetrazolium bromide (MTT) were purchased from Sigma (St. Luis, MO, USA). All of the working solutions were initially dissolved in dimethyl sulfoxide (DMSO) and posterior dilutions were put in culture medium.
Cell lines and culture conditions
The lung cancer cell line INER-51 and its clone PSC-1 (non-expressing P-gp or P-gpneg) were grown as monolayer cultures in RPMI-1640 medium at 37°C in 5% CO2. INER-51 is a NSCLC cell line established in our laboratory from pleural effusion of patient diagnosed with primary lung cancer without previous chemotherapy treatment. The kidney cell line A498 (expressing P-gp) and the lung adenocarcinoma cell line INER-37 (expressing MRP1) were used as controls of P-gp function and MRP1 expression, respectively. The culture medium was supplemented with 10% FCS (Sigma, Co. St. Luis, MO, USA), 1% non-essential amino acids, 1 mM sodium pyruvate, 2 mM L-glutamine, 100 units/ml of penicillin and 100 μg/ml of streptomycin.
Culture of monolayers and multicellular spherules
Monolayer cells were passed a week by trypsin-EDTA solution (Invitrogen™, USA). To obtain multicellular spheroids, 3.5 × 105 exponentially tumour cells were seeded in 1% agarose-coated (24-well/plate) in RPMI-1640 complete medium [14]. Cultures were routinely grown for 72 hours to acquire multicellular spheroids of approximately of 500 μm of diameter.
P-gp non-expressing mutant clones
In order to obtain one mutant clone with the capacity to not express P-gp, INER-51 cells were treated as described by Beketic-Oreskovic, et al[15]. Briefly, 1 × 106 tumour cells were seeded in a T25 plastic tissue culture flask (Falcon, USA). When the cell culture achieved a semi-confluent grade, the culture medium was removed and fresh complete medium was added containing 5 μM DXR and 5 μM of SDZ PSC 833. The cells were maintained under this condition during two weeks until some survival clones were evident upon microscopic observation. A total of 6 survival clones were isolated, propagated for a month in absence of drugs, and tested for MDR1 expression by in vitro RT-PCR. One mutant clone that manifested stable for mdr1neg phenotype was selected and cloned again. Each clone obtained was tested for MDR-1 expression again. Finally, the clone PSC-1 was further propagated and used for studies in the present communication.
Drug resistance in monolayers or multicellular spheroids
The level of resistance to drugs was determined with the use of MTT assay as previously described by S. Cole [16] and on multicellular systems as evaluated by Furukawa, et al[17]. For monolayers, 7 × 103 cells/well were plated in 96-well/plate (Costar, USA) and drugs were added in different concentrations per well. In the case of multicellular spheroids, they were obtained as described above and were fed with fresh complete medium containing different drug concentration. After 72 hours, the culture medium was retired and MTT reagent diluted in PBS was added to obtain a final concentration of 2 mg/ml. After incubation for 4 hours, individual spheroids surrounded with formazan crystals were transferred into 1.5 ml eppendorf tubes. Cells in monolayers were washed carefully with PBS once. Both monolayers and multicellular spheroids crystals were dissolved by addition of 100% DMSO for 20 min with occasional shaking. Absorbance at 540 nm was measured using an automated microplate reader (Labsystem Multiskan MS, Finland). In each experiment, the drug determination was analysed in six individual wells. Cell survival was estimated as a percentage of the corresponding control. Drug-cytotoxicity was assayed by the IC50, corresponding to the 50% decrease in cell survival rate respective to not-drug treated cultures.
Doxorubicin and Rhodamine-123 retention assays
Exponentially growing cells (3.5 × 105) were seeded 72 hours prior to treatment with drugs in 1% agar-coated 24-well tissue culture plates as described above. Either multicellular spheroids or monolayer cells were incubated with increased concentrations of DXR or Rho-123 for 30 min and then washed with fresh drug-free medium twice and PBS once. Cell suspensions from monolayers were obtained by tripsinization whereas multicellular spheroids were recovered using micropipette tips. To exclude artefacts arising from DXR quenching by binding to DNA and accumulating in acidic organelles, DXR was extracted from spheroids with the method described by Wartenberg et al[18]. Finally, intracellular DXR or Rho-123 were determined by spectrofluormetry (Phototechnology, International, Princeton, New Jersey, USA) at λem 580 nm and λex 427 or λem 510 nm and λex 480 respectively.
Doxorubicin efflux assays
The efflux assays were based on those previously described [19]. Briefly, monolayers or multicellular spheroids were drug loaded with DXR (30 μM) throughout incubation at 37°C in 5% CO2 for 30 min. The loading cells were then washed with a complete drug-free ice-cold medium and either placed on ice or incubated at 37°C in 5% CO2 for different intervals of time. This incubation allowed drug efflux to occur, and during experimental time the remaining drug was quantitated by spectrofluormetry as described above.
Circumvention of P-gp-mediated efflux with Cs-A or ATP-depletion
For evaluation of P-gp activity, monolayers or multicellular spheroids were pre-incubated for 1 hour with increasing drug concentrations of cyclosporin-A (Cs-A). After pre-incubation, multicellular spheroids were loaded for 30 min with 30 μM of DXR in presence of the reversal agent and afterwards the DXR uptake was assessed again. For ATP-depletion, cells were pre-incubated for 20 min at 4°C in PBS/BSA (1 mg/ml) containing 1 mM sodium-cyanide, 10 mM sodium-fluoride and 10 mM sodium azide prior to the addition of DXR. After centrifugation at 1000 × g for 15 min, both monolayers and multicellular spheroids were loaded with different DXR concentrations in the presence of ATP synthesis inhibitors and the fluorescence was measured as mentioned above.
In vitro Reverse transcriptase-PCR assay
Total RNA was extracted from the cell lines with Trizol reagent (Invitrogen™, USA) according to manufacturer's instructions. Single stranded cDNA was synthesized by reverse transcription from 5 μg of total RNA using Superscript™ RNAse Reverse Transcriptase (Invitrogen™, USA) and oligo-dT16–18. The amplification was performed in a final volume of 25 μl, containing 0.5 μl cDNA, 50 pM of each oligonucleotide primer, 30 μM of each dNTPs, 2.5 units of Taq DNA polymerase, 1.5 mM MgCl2, 20 mM Tris-HCl (pH 8.4) and 50 mM KCl. Amplification was carried out in a Thermal Cycler (Programmed Thermal Controller, model PTC-100, MJ Research Inc., USA) for 35 cycles of denaturalisation at 94°C for 1 min, annealing at 55–60°C for 2 min, and polymerisation at 72°C for 3 min. The PCR primers and expected product size were as follows: For MDR-1, forward: 5'-cccatcattgcaatagcagg-3' and reverse: 5'-gttcaaacttctgctcctga-3 [150 bp]; MRP1, forward: 5'-tctctcccgacatgaccgagg-3' and reverse: 5'-ccaggaatatgatgccccgacttc-3' [140 bp]; topoisomerase IIα, forward: 5'-tttaaggcccaagtccagttaaac-3' and reverse: 5'-gtataacaatatcatcaagattgt [343 pb]; topoisomerase IIβ, forward: 5'-gaagtgttcactagtaaaatacagt-3' and reverse: 5'-cataatctttccatagcgtaaggtt-3' [336 bp]; topoisomerase I, forward: 5'-aagcagaggaagtagctacg-3' and reverse: 5'-gctcatctgtttccgagctt-3' [206 bp]; GST-μ, forward: 5'-gaactccctgaaaagctaaag-3' and reverse: 5'-gttgggctcaaatatacggtgg-3' [250 bp]; G3PDH, forward: 5'-tggggaaggtgaaggtcgga-3' and reverse: 5'-gaaggggtcattgatggcaa-3' [110 bp].
List of abbreviations
Cs-A: Cyclosporin-A
DXR: Doxorubicin
MCR: Multicellular resistance
MTT: 3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyl tetrazolium bromide
NSCLC: Non-small cell lung cancer
P-gp: P-Glycoprotein
Rho-123: Rhodamine 123
MDR: Multidrug Resistance
DMSO: Dimethyl sulfoxide
Authors' contributions
All author(s) contributed equally to this work.
Acknowledgements
We are indebted to Ms. Christine Wilson for linguistic revision and Rodolfo Ocádiz Delgado (CINVESTAV-IPN) for critical and technical review of the manuscript.
This work was submitted in partial fulfilment of the requirements for the D. Sc. Degree for Ponce de Leon
Suarez,V., at DOCTORADO EN CIENCIAS BIOMEDICAS, UNIVERSIDAD NACIONAL AUTONOMA DE MEXICO
==== Refs
Hrycyna C Molecular genetic analysis and biochemical characterization of mammalian P-glycoprotein involved in multidrug resistance Cell Dev Biol 2001 12 247 256 10.1006/scdb.2000.0250
Sauna ZE Smith MM Müller M Kerr KM Ambudkar SV The Mechanism of action of multidrug-resistance-linked P-glycoprotein J Bioenergetics Biomembranes 2001 33 481 491 10.1023/A:1012875105006
Gottesman MM Fojo T Bates SE Multidrug resistance in cancer: role of ATP-dependent transporters Nat Rev Cancer 2002 2 48 58 11902585 10.1038/nrc706
Takeshita H Kusuzaki K Ashihara T Gebhardt MC Mankin HJ Hirasawa Y Actin organization associated with the expression of multidrug-resistant phenotype in osteosarcoma cells and the effect of actin depolymerization on drug resistance Cancer Lett 1998 126 75 81 9563651 10.1016/S0304-3835(97)00539-9
Luciani F Molinari A Lozupone F Calcabrini A Lugini L Strigaro A Puddu P Arancia G Cianfriglia M Fais S P-Glycoprotein-actin association through ERM family proteins: a role in P-glycoprotein function in human cells of lymphoid origin Blood 2002 99 641 648 11781249 10.1182/blood.V99.2.641
Callaghan R van Gorkom LC Epand RM A comparison of membrane properties and composition between cell lines selected and transfected for multi-drug resistance Br J Cancer 1992 66 781 786 1358166
Le Moyec L Tatoud R Degeorges A Calabresse C Bauza G Eugene M Calvo F Proton nuclear magnetic resonance spectroscopy reveals cellular lipids involved in resistance to adriamycin and taxol by the K562 leukemia cell line Cancer Res 1996 56 3461 3467 8758912
Sutherland RM Eddy HA Bareham B Reich K Vanantwerp D Resistance to adriamycin in multicellular spheroids Int J Radiat Oncol Biol Phys 1979 5 1225 1230 528267
Mueller-Klieser W Three-dimensional cell cultures: from molecular mechanism to clinical applications Am J Physiol 1997 273 C1109 1123 9357753
Kobayashi H Man S Graham CH Kapitain SJ Teicher BA Kerbel RS Acquired multicellular-mediated resistance to alkylating agents in cancer Proc Natl Acad Sci USA 1993 90 3294 3298 8475071
Kolchinsky A Roninson IB Drug resistance conferred by MDR1 expression in spheroids formed by glioblastoma cell lines Anticancer Res 1997 17 3321 3327 9413166
Desoize B Gimonet D Jardiller JC Cell culture as spheroids: an approach to multicellular resistance Anticancer Res 1998 18 4147 4158 9891460
Kerbel RS Rak J Kobayashi H Man MS St Croix B Granham CH Multicellular resistance: a new paradigm to explain aspects of acquired drug resistance of solid tumors Cold Spring Harb Symp Quant Biol 1994 59 661 672 7587127
Yuhas JM Li AP Martinez AO Landman AJ A simplified method for production and growth of multicellular tumor spheroids Cancer Res 1977 37 3639 3643 908012
Beketic-Oreskovic L Durán GE Chen G Dumontet C Sikic BI Decrease mutation rate for cellular resistance to doxorubicin and suppression of mdr1 gene activation by the Cyclosporin PCS 833 J Natl Cancer Inst 1995 87 1593 1602 7563202
Cole SP Rapid chemosensitivity testing of human lung tumor cells using the MTT assay Cancer Chemother Pharmacol 1986 17 259 263 3742711 10.1007/BF00256695
Furukawa T Kubota T Watanabe M Takahara T Yamaguchi H Takeuchi T Kase S Kadaira S Ishibiki K Kitajima M Hoffman RM High in vitro-in vivo correlation of drug response using sponge-gel-supported three-dimensional histoculture and the MTT end point Int J Cancer 1992 51 489 498 1592540
Wartenberg M Frey C Diedershagen H Ritgen J Hescheler J Sauer H Development of an intrinsic P-glycoprotein-mediated doxorubicin resistance in quiescent cell layers of large, multicellular prostate tumor spheroids Int J Cancer 1998 75 855 863 9506530 10.1002/(SICI)1097-0215(19980316)75:6<855::AID-IJC7>3.0.CO;2-U
Hrycyna CA Ramachandra M Pastan I Gottesman MM Functional expression of human P-glycoprotein from plasmids using vaccinia virus-bacteriophage T7 RNA polymerase system Methods Enzymol 1998 292 456 473 9711574
Moriyama Y Manabe T Yoshimori T Tashiro Y Futai M ATP-dependent uptake of anti-neoplastic agents by acidic organelles J Biochem (Tokyo) 1994 115 213 218 8206870
Schindler M Grabski S Hoff E Simon SM Defective pH regulation of acidic compartments in human breast cancer cells (MCF-7) is normalized in Adriamycin resistant cells (MCF-7adr) Biochemistry 1996 35 2811 2817 8608115 10.1021/bi952234e
Nederman T Norling B Glimelius B Carlsson J Brunk U Demonstration of an extracellular matrix in multicellular tumor spheroids Cancer Res 1984 44 3090 3097 6373002
Bichat F Mouawad R Solis-Recendez G Khayat D Bastian G Cytoskeleton alteration in MCF7R cells, a multidrug resistant human breast cancer cell line Anticancer Res 1997 17 3393 3402 9413178
dit Faute MA Laurent L Ploton D Poupon MF Jardillier JC Bobichon H Distinctive alterations of invasiveness, drug resistance and cell-cell organization in 3D-cultures of MCF-7, a human breast cancer cell line, and its multidrug resistant variant Clin Exp Metastasis 2002 9 161 168 11964080 10.1023/A:1014594825502
Jakubowicz-Gil J Paduch R Gawron A Kandefer-Szerszen M The effect of heat shock, cisplatin, etoposide and quercetin on Hsp27 expression in human normal and tumour cells Folia Histochem Cytobiol 2002 40 31 35 11885806
Rosi A Grande S Luciani AM Barone P Mlynarik V Viti V Guidoni L 1H MRS studies of signals from mobile lipids and from lipid metabolites: comparison of the behavior in cultured tumor cells and in spheroids NMR Biomed 2004 17 76 91 15052555 10.1002/nbm.867
Hazlehurst LA Dalton WS Mechanisms associated with cell adhesion mediated drug resistance (CAM-DR) in hematopoitec malignances Cancer Metastasis Rev 2001 20 43 50 11831646 10.1023/A:1013156407224
Wartenberg M Ling FC Muschen M Klein F Acker H Gassmann M Petrat K Putz V Hescheler J Sauer H Regulation of the multidrug resistance transporter P-glycoprotein in multicellular tumor spheroids by hypoxia-inducible factor (HIF-1) and reactive oxygen species FASEB J 2003 17 503 505 12514119
Pedro D Beltran PJ Wang YF Bucana CD Yoon SS Deguzman ACP Fidler IJ Cell density-dependent regulation of mdr-1 gene expression in murine colon cancer cells Int J Oncol 1996 9 865 878
Tunggal JK Melo T Ballinger JR Tannock IF The influence of expression of P-glycoprotein on the penetration of anticancer drugs through multicellular layers Int J Cancer 2000 86 101 107 10728602 10.1002/(SICI)1097-0215(20000401)86:1<101::AID-IJC16>3.0.CO;2-I
Tritton TR Cell surface actions of adriamycin Pharmacol Ther 1991 49 293 309 2052627 10.1016/0163-7258(91)90060-Y
Goormaghtigh E Chatelain P Caspers J Ruysschaert JM Evidence of a specific complex between adriamycin and negatively-charged phospholipids Biochim Biophys Acta 1980 597 1 14 7370238
Martin C Walker J Rothnie A Callaghan R The expression of P-glycoprotein does influence the distribution of novel fluorescent compounds in solid tumour models Br J Cancer 2003 89 1581 1589 14562035 10.1038/sj.bjc.6601300
Alemán C Annereau JP Liang XJ Cardarelli CO Taylor B Yin JJ Aszalos A Gottesman MM P-glycoprotein, expressed in multidrug resistant cells, is not responsible for alterations in membrane fluidity or membrane potential Cancer Res 2003 63 3084 3091 12810633
Ferté J Analysis of the tangled relationships between P-glycoprotein-mediated multidrug resistance and the lipid phase of the cell membrane Eur J Biochem 2000 267 277 294 10632698 10.1046/j.1432-1327.2000.01046.x
Eytan GD Regev R Oren G Assaraf YG The role of passive transbilayer drug movement in multidrug resistance and its modulation J Biol Chem 1996 271 12897 12902 8662680 10.1074/jbc.271.6.3163
Ehrlich PH Moustafa ZA Archinal-Mattheis AE Newman MJ Bair KW Cohen D The reversal of multidrug resistance in multicellular tumor spheroids by SDZ PSC 833 Anticancer Res 1997 17 129 133 9066642
Anderson M Warr JR Expression of verapamil hypersensitivity in multidrug-resistant cells grown as multicellular spheroids Cancer Chemother Pharmacol 1990 26 151 154 2347041
Sakata K Kwok TT Gordon GR Walch NS Sutherland RM Resistance to verapamil sensitization of multidrug-resistant cells grown as multicellular spheroids Int J Cancer 1994 59 282 286 7927930
Tunggal JK Cowan DS Shaikh H Tannock IF Penetration of anticancer drugs through solid tissue: a factor that limits the effectiveness of hemotherapy for solid tumors Clin Cancer Res 1999 5 1583 1586 10389947
He L Liu GQ Interaction of multidrug resistance reversal agents with P-glycoprotein ATPase activity on blood-brain barrier Acta Pharmacol Sin 2002 23 423 429 11978192
Martin C Berridge G Higgins CF Mistry P Charlton P Callaghan R Communication between multiple drug binding sites on P-glycoprotein Mol Pharmacol 2000 58 624 632 10953057
Hicks KO Ohms SJ van Zijl PL Denny WA Hunter PJ Wilson WR An experimental and mathematical model for the extravascular transport of DNA intercalator in tumours Br J Cancer 1997 76 894 903 9328149
Cowan DSM Hicks KO Wilson WR Multicellular membranes as in vitro model for extravascular diffusion in tumours Br J Cancer 1996 74 528s 531s
Bertram J Palfner K Hiddemann W Kneba M Increase of P-glycoprotein-mediated drug resistance by hsp 90 beta Anticancer Drugs 1996 7 838 845 8991187
Millward MJ Cantwell BM Munro NC Robinson A Corris PA Harris AL Oral verapamil with chemotherapy for advanced non-small cell lung cancer: a randomised study Br J Cancer 1993 67 1031 1035 8388231
Biswas EE Nucleotide binding domain 1 of the human retinal ABC transporter functions as a general ribonucleotidase Biochemistry 2001 40 8181 8187 11444963 10.1021/bi0106686
Gottesman MM Mechanisms of cancer resistance Ann Rev Medicine 2002 53 615 657 10.1146/annurev.med.53.082901.103929
Doyle LA Ross DD Multidrug resistance mediated by the breast cancer resistance protein BCRP ABCG2 Oncogene 2003 22 7340 7358 14576842 10.1038/sj.onc.1206938
Honjo Y Hrycyna CA Yan Q-W Medina-Perez WY Robery RW van de Laar A Litman T Dean M Bates SE Acquired mutations in the MXR/BCRP/ABCP gene alter substrate specificity in MXR/BCRP/ABCP-overexpressing cells Cancer Res 2001 61 6635 6639 11559526
Oloumi A MacPhail SH Johnston PJ Banath JP Olive PL Changes in subcellular distribution of topoisomerase II alpha correlate with etoposide resistance in multicell spheroids and xenograft tumors Cancer Res 2000 60 5747 5753 11059769
Luo C Johnston PJ MacPhail SH Banath JP Oloumi A Olive PL Cell fusion studies to examine the mechanism for etoposide resistance in Chinese hamster V79 spheroids Exp Cell Res 1998 243 282 289 9743588 10.1006/excr.1998.4170
Zhao R Goldman ID Resistance to antifolates Oncogene 2003 22 7340 7358 14576842 10.1038/sj.onc.1206938
Scagliotti GV Novello S Salvaggi G Multidrug resistance in non-small-cell lung cancer Ann Oncol 1999 10 S83 86 10582146 10.1023/A:1008329010443
Yeh JJ Hsu WH Wang JJ Ho ST Kao A Predicting chemotherapy response to paclitaxel-based therapy in advanced non-small-cell lung cancer with P-glycoprotein expression Respiration 2003 70 32 35 12584388 10.1159/000068411
Volm M Mattern J Samsel B Overexpression of P-gp and glutathione S-transferase-π in resistant non-small cell lung carcinomas of smokers Br J Cancer 1991 64 700 704 1680367
Oka M Fukuda M Sakamoto A Takatani H Fukuda M Soda H Kohno S The clinical role of MDR1 gene expression in human lung cancer Anticancer Res 1997 17 721 724 9066608
|
16001980
|
PMC1185553
|
CC BY
|
2021-01-04 16:24:16
|
no
|
Cancer Cell Int. 2005 Jul 7; 5:20
|
utf-8
|
Cancer Cell Int
| 2,005 |
10.1186/1475-2867-5-20
|
oa_comm
|
==== Front
Cancer Cell IntCancer Cell International1475-2867BioMed Central London 1475-2867-5-251606096510.1186/1475-2867-5-25HypothesisOn the nature of cancer and why anticancer vaccines don't work Prehn Richmond T [email protected] Dept of Pathology, University of Washington, 5433 South Hudson Street, Seattle, WA 98118, USA2005 1 8 2005 5 25 25 7 6 2005 1 8 2005 Copyright © 2005 Prehn; licensee BioMed Central Ltd.2005Prehn; 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.
In this essay I suggest that the major difficulty in producing effective anti-cancer vaccines lies in the fact that most cancers have little immunogenicity because of a basic paucity of tumor-specific antigenicity. The lack of antigenicity, despite extensive genomic instability, could be explained if most tumor mutations occur in silenced genes. A further problem is that an immune reaction against tumor antigens, especially in moderate or low amount, may be stimulatory rather than inhibitory to tumor growth.
==== Body
It is now almost half a century since the overthrow of Ehrlich's doctrine of "horror autotoxicus" and the general acceptance of the contrary idea that animals can indeed be immunized against the growth of a transplanted syngeneic cancer. Why then is it that, despite nearly 50 years of intense investigation, attempts to use the immune reaction as a tool against cancer have, with the exception of bladder cancer, met with only moderate success? What is the realistic prospect that the next 50 years will see an improvement in this dismal state of affairs? Many investigators, myself included, have a large vested interest in the field of cancer immunology and will be reluctant to entertain any discouraging viewpoint, but the actual facts are, I believe, discouraging.
Background
As a part of the original demonstration that syngeneic anticancer immunity is possible, it was shown that, among sarcomas induced in mice by a hydrocarbon, each tumor, when transplanted, could arouse an inhibitory immune reaction against itself [1]. However, it also became very clear that each tumor was antigenically unique, even if each had been induced by identical means in one and the same animal; although cross-reactions were reported, these were the exception [2,3]. Since it is probable that the immunogenicity was caused by the mutations induced by the mutagenic carcinogen, it was obvious that these chemicals produced a different spectrum of mutations in each tumor and with very little overlap. Consequently, one had to conclude that any of a vast array of possible mutations could be found in phenotypically similar cancers, a not impossible idea. However, it was also clear that none of the carcinogen-induced mutations, at least among those identified by their resulting antigenicity, could be considered essential or causative for the induction of the cancer.
Not surprisingly, although sarcomas induced by the same concentration of oncogen displayed a wide range of immunogenic strengths, there was found to be a positive statistical correlation between the strength of the immunogenicity and the concentration of the inducer [4,5]. Unfortunately, this relationship means that tumors induced by low concentrations of inducer, or that arise without obvious cause, tend to have either a very low or a nondetectable immunogenicity [6]. It is my hypothesis, as presented in this essay, that the paucity of immunogenicity reflects, for the most part, a basic lack of tumor-specific antigenicity rather than a blocking or suppression of an immune response.
Immunosurveillance
It has been argued that the apparent general lack of tumor immunogenicity may be an artifact caused by immune selection for nonimmunogenic tumor variants. Perhaps most tumors, in accord with the immunosurveillance hypothesis, are really highly immunogenic and what we see is actually a small surviving, relatively non-immunogenic, highly selected subpopulation. This popular concept, judged by at least four types of evidence, is probably incorrect and thus cannot account for the paucity of tumor immunogenicity.
Firstly, immunological depression by any means usually has little if any facilitating effect on oncogenesis. Thus, chemical oncogenesis is not obviously altered by immunodepression [7]; the tumor immunogenicity and/or the degree of immune depression must be very carefully titrated in order to see any effect of immune depression, either positive or negative [8,9]. Both within and without the immunologically isolated environment of intraperitoneal diffusion chambers, nonimmunogenic tumors are commonly induced [10,11]. Furthermore, chemically-induced tumorigenesis may actually be lessened, not increased, in immunologically crippled, germ-free nude mice as compared with their essentially normal heterozygous nude controls [8].
Secondly, often a minute dosage of highly immunogenic transplanted cancer cells may grow when a somewhat larger inoculum fails [12]; this "sneaking through phenomenon", which can occur even in specifically immunized animals, again suggests that small incipient tumors could probably not be effectively surveyed.
Thirdly, even highly immunogenic hydrocarbon-induced sarcomas may fail to induce immunity if the tumor is left undisturbed in situ, so how could more ordinary tumors of lesser immunogenicity kindle a surveillance mechanism? In fact, to arouse an inhibitory immunity in the primary autochthonous host is difficult and, at least with hydrocarbon-induced sarcomas, requires repeated immunizations [13]. More often the challenge tumor, inoculated back into the animal in which it had originated, grew better than it did in other animals of the same strain, ie., better than it did in animals that had never before been exposed to that tumor [14,15]. This latter observation is best interpreted by the immunostimulation hypothesis [16] which I will now discuss.
Immunostimulation
As compared with normal spleen cells, low proportions of spleen cells from immunized mice stimulated rather than inhibited the proliferation of admixed tumor cells when the mixture was injected subcutaneously. In this so called "Winn test" [17], larger proportions of immune spleen cells inhibited the growth of the same admixed tumor cells. As previously discussed, most tumors, arising from low concentrations of inducing agents, are expected to have little immunogenicity. The little immune response these tumors might engender would be expected, judging by these Winn test results, to stimulate rather than inhibit tumor growth. This was emphatically confirmed in a variety of extensive studies which showed, among other things, that a newly induced in situ mouse tumor, mesenchymal or epithelial, was stimulated to grow faster (had a shorter latency) if it could engender an immune response [8,11]. Even tumors that subsequently were shown to be highly immunogenic usually grew faster than tumors of little or no immunogenicity when the tumors were left undisturbed in their primary hosts [11]. Therefore, when immunodepression does appear to favor oncogenesis, this result, in many cases, is probably not because an immunological inhibitor to tumor growth has been reduced, but rather because the immune reaction has been depressed to a more stimulatory level [18]. Also, the possibility of doing harm by attempts at immunotherapy need to be carefully cosidered. For more extensive reviews and other evidences of tumor immunostimulation see [8,11,18,19].
The fact that a moderate tumor-specific immune reaction apparently favors the growth of an undisturbed in situ tumor, seems adequate to rule out any immunosurveillance of incipient cancers. Thus, the paucity of tumor immunogenicity is probably not caused by immunoselection. Rather, most types of tumor are apparently created de novo with little tumor-specific immunogenicity unless the tumors were induced by a large concentration of carcinogen and/or by an oncogenic virus. In fact, at least among hydrocarbon-induced mouse sarcomas, whatever little immune selection there may be apparently favors, rather than inhibits, the proliferation of a nacsent tumor.
Role of Mutation
If cancer is based, in accord with the current paradigm, upon somatic mutations, why would most tumors have so little specific immunogenicity? While this question has many possible answers, it seems to me that, in the absence of immunoselection, there are two prime possibilities: either the somatic mutation idea of carcinogenesis is incorrect and/or the tumor mutations occur, for the most part, in silenced genes that are incapable of producing an antigenic product.
The hypothesis that the somatic-mutation paradigm is incorrect has been advanced over a number of years by a number of brave heretics [20-26]. They suggest that the mutations seen in neoplasia may not be causative; they are, instead, probably incidental or secondary. Primary epigenetic rather than genetic changes are postulated to result in the neoplastic phenotype. This hypothesis has the great virtue, as compared with the mutational paradigm, in that a tumor's reversion to a normal phenotype can be more easily understood. Reversion to a normal phenotype is indeed observed in a variety of laboratory experiments as well as in some clinical settings [26,27]. The real cause of most cancers is postulated by the heretics to reside in disrupted cell to cell signaling or some other epigenetic alteration which changes the spectrum of gene expression to produce a neoplastic phenotype. The many cogent arguments put forth by the heretics, entirely apart from any considerations of immunogenicity, have been well reviewed elsewhere [20-26]. However, these ideas of an epigenetic etiology fit well with the facts concerning the general lack of immunogenicity of cancers and could help to explain why immunity has not as yet lived up to its expected diva role in cancer therapy and prevention.
My personal view of the intimate details of carcinogenesis incorporates two suppositions, for each of which there is supporting experimental evidence. I will therefore, for purposes of this discussion, consider both of the following statements to be correct. The first is that DNA repair is defective or nonexistent in untranscribed or silenced genes [28]. Secondly, DNA mutation and repair both require cellular proliferation [29].
Sonnenschein and Soto have suggested that proliferative activity, rather than quiescence, is the default condition of cells [20]; thus the natural impulse of free-living cells to proliferate is, in a multicellular animal, actively regulated by aspects of the multicellular environment. It seems logical, therefore, to propose that the abnormal proliferative activity in a cancer requires that certain suppressor genes, especially those that normally, in a post-embryonic animal, suppress embryonic development and/or wound healing, be silenced. My view of carcinogenesis depends, as I have said, upon the assumption that DNA damage in untranscribed genes is not repaired [28]. Therefore, it seems reasonable to assume that, over time in a tumor, mutations might tend to accumulate in any silenced gene or in non-coding segments of a gene.
The picture that emerges from these considerations is that neoplasia is probably caused by the reexpression of proliferation-enhancing development or wound-healing genes that, outside of a neoplasm, would usually be expressed only in embryonic life or during wound healing. Such reexpression, indeed overexpression, now as oncogenes, could be caused by either interference with communication between the oncogene and the appropriate suppressor gene or by the actual silencing of the suppressor gene by either mutation or, perhaps more commonly, by epigenetic influences. Mutations that may have occurred in the development genes might often be repaired, if and when these genes were reexpressed as oncogenes within a proliferating neoplasm. Since the oncogene's function is to drive the cancer, any mutations in these genes would probably be highly selected to retain the gene's normal or near normal function; there might thus be selection to produce only normal or near normal product, a product that might then be only minimally, if at all, immunogenic. Alternatively, any mutations that did occur among the oncogenes might not produce immunogenicity if the products of such mutations were not found on the cell surface.
Recapitulation
Thus, my thesis is that most of the mutations actually found in a neoplasm would probably not be among the reexpressed proliferation-stimulating oncogenes, but would be among those suppressor genes that were newly silenced either during or after transformation and that remained silenced, unexpressed, and unrepaired throughout the life of the tumor. Because these suppressor genes had been silenced, their mutations would probably not result in new antigens.
To recapitulate, an important corollary of the previous discussion is the conclusion that the mutations that are identified by their associated antigens, are seldom cancer-inducing, but are random and incidental. The genes that actually drive the neoplasm, the so-called oncogenes, are embryonic development or wound-healing genes that are reexpressed and overexpressed in a tumor. Being highly selected in the tumor for essentially normal functions, the reexpression of these genes, even when mutated, might not produce new antigens, especially if the oncogene products did not appear on the cell surface. Mutations that occur in silenced suppressor genes go unrepaired, but because silent unexpressed genes have no product, these mutations also should fail to give rise to antigens. Thus, my view of carcinogenesis suggests that there are usually, with few exceptions, no mutations in the carcinogenic process that can produce new antigens unless a virus or an unrealistically high concentration of oncogen is involved. Mutations, induced by high concentrations of a chemical carcinogen do give rise to immunogenicity and thus they presumably occur in expressed genes, but, as already stated, such mutations are apparently not directly related to etiology and are usually unrealistic laboratory constructs.
In sum, the dismal record of the attempts to utilize anti-tumor immunity in the clinic seems entirely consistent with the idea that cancer is a disease based most often upon the silencing of suppressor genes, either by mutation or by epigenetic means. The epigenetic pathways may be the more frequent. The postulated, as well as demonstrated, paucity of antigen-producing mutations in the expressed genes of most cancers, plus the lack of antigenicity that would result from mutations that might arise in the silenced suppressor genes, explains the basic paucity of tumor-specific immunogenicity. Furthermore, the little tumor immunogenicity that may exist usually produces a reaction that is stimulatory, not inhibitory, to tumor proliferation.
Although there is certainly a role for immune suppressor cells and other blocking factors [30], in this essay I have suggested that the major difficulty lies in the fact that most cancers have little immunogenicity because of a basic lack of tumor-specific antigenicity. What weak antigens a tumor may have are usually tumor specific rather than tumor-type specific. The tissue-specific antigens involved in autoimmune reactions and the "carcino-embryonic antigens" are important realities, but have been difficult to utilize to any great extent in cancer prevention or therapy, probably in large part because of the powerful natural tolerance to such antigens. Perhaps further research will reveal ways to use the autoimmune mechanism against cancers that arise in disposible organs such as the prostate or the mammary gland where tumor specificity might be less important and organ specificity might be sufficient.
If my view of the carcinogenic process is correct, the utility of immune mechanisms in the "war on cancer" seems likely to remain, dispite some minor triumphs, rather dismal. However, the success of intravesicular BCG to treat carcinoma of the bladder offers hope [31]. There have also been encouraging reports about the probable immunogenicity of cutaneous melanoma [32,33]. Dispite my previously expressed view that immunosurveillance of cancer is unlikely, there is one observation that deserves special mention. It has long been known that cancer occurs with greatly increased frequency in patients who undergo prolonged immunodepression to facilitate kidney transplantation. The increased incidence is not general, but is largely confined to tumors of the lymphoid system or of the skin [34]. The excess of lymphoid tumors could be easily considered the ultimate consequence of compensatory hyperplasia in the damaged immunological organ, but the excess of skin tumors is more likely to have a direct immunological basis, either decreased surveillance or increased immunostimulation by the impaired immune mechanism. In either case, the fact that the excess tumor incidence is largely confined to the skin suggests that skin tumors may be unusually antigenic and/or the skin is an unusally active immunological organ.
My view of the very complex interaction of cancer and the immune reaction is certainly simplistic and perhaps overly pessimistic, but it may, I am afraid, capture the essence of the problem.
==== Refs
Prehn RT Main JM Immunity to methylcholanthrene-induced sarcomas J Natl Cancer Inst 1957 18 769 78 13502695
Globerson A Feldman M Antigenic specificity of Benzo(a)pyrene-induced sarcomas J Natl Cancer Inst 1964 32 1229 43 14191397
Basombrio MA Search for common antigenicities among twenty-five sarcomas induced by methycholanthrene Cancer Res 1970 30 2458 62 4097428
Prehn RT Relationship of tumor immunogenicity to the concentration of the oncogen J Natl Cancer Inst 1975 55 189 90 1159812
Lawler EM Prehn RT Influence of immune status of host on immunogenicity of tumors induced with two doses of methylcholanthrene Cancer Immunol Immunother 1982 13 194 7 6762248
Hewett HB Blake ER Walder AS A critique of the evidence for active host defenses against cancer, based on personal studies of 27 murine tumours of spontaneous origin Br J Cancer 1976 33 241 59 773395
Stutman O Immunodepression and malignancy Adv Cancer Res 1975 22 261 422 766581
Outzen HC Development of carcinogen-induced skin tumors in mice with varied states of immune capacity Int J Cancer 1980 26 87 92 7239715
Prehn RT Immunostimulation of chemical oncogenesis in the mouse Int J Cancer 1977 20 918 22 591130
Carbone G Parmiani G Nonimmunogenic sarcomas induced by 3-MCA treatment of murine fibroblasts in diffusion chambers J Natl Cancer Inst 1975 55 1196 7
Prehn RT Bartlett GL Surveillance, latency, and the two levels of MCA-induced tumor immunogenicity Int J Cancer 1987 39 106 10 3793267
Marchant J Sarcoma induction in mice by methylcholanthrene (Antigenicity tests of sarcomas induced in thymus grafted and control animals) Br J Cancer 1969 23 383 90 5788047
Klein G Sjogren HO Klein E Hellstrom KE Demonstration of resistance against methylcholanthrene-induced sarcomas in the primary autochthonous host Cancer Res 1960 20 1561 72 13756652
Basombrio MA Prehn RT Immune status of autochthonous and adoptlively protected mice toward spontaneous and chemically induced tumors Cancer Res 1972 32 2545 50 5082598
Stjernsward J Immune status of the primary host toward its own methylcholanthrene-induced sarcomas J Natl Cancer Inst 1968 40 13 22 5635014
Prehn RT Lappe MA The immune reaction as a stimulator of tumor growth Transpl Revs 1971 7 26 54
Prehn RT The immune reaction as a stimulator of tumor growth Science 1972 176 170 1 5014438
Prehn RT Prehn LM The autoimmune nature of cancer Cancer Res 1987 47 927 32 3542202
Prehn RT Stimulatory effects of immune reactions upon the growths of untransplanted tumors Cancer Res 1994 54 908 14 8313380
Soto A Sonnenschein C The somatic mutation theory of cancer: growing problems with the paradigm? BioEssays 2004 26 1097 1107 15382143
Farber E Rubin H Cellular adaptation in the origin and development of cancer Cancer Res 1991 51 2751 2761 2032214
Rubin H Spontaneous transformation as aberrant epigenesis Differentiation 1993 53 123 137 8359592
Pierce GB Speers WC Tumors as caricatures of the process of tissue renewal: prospects for therapy by directed differentiation Cancer Res 1988 48 1996 2004 2450643
Jaffe LF Epigenetic theories of cancer initiation Adv Cancer Res 2003 88 209 230 14710952
Prehn RT Cancers beget mutations versus mutations beget cancers Cancer Res 1994 54 5296 5300 7923156
Prehn RT The role of mutation in the new cancer paradigm Cancer Cell Internat 2005 5 8
Haas D Ablin AR Miller C Complete morphologic maturation and regression of stage IVS neuroblastoma without treatment Cancer 1988 62 818 25 3293764
Mellon I Bohr VA Smith CA Hanawalt PC Preferential DNA repair of an active gene in human cells Proc Natl Acad Sci USA 1986 83 8878 82 3466163
Bielas JH Heddle JA Proliferation is necessary for both repair and mutation in transgenic mouse cells Proc Natl Acad Sci USA 2000 97 11391 11396 11005832
Chatenoud L Salomon B Bluestone JA Suppressor T cells – they're back and critical for regulation of autoimmunity! Immunological Reviews 2001 182 149 163 11722631
Patard JJ Rodriguez A Leray E Rioux Leclercq N Guille F Lobel B Intravesical bacillus Calmette-Guerin treatment improves patient survival in T1G3 bladder tumours European Urology 2002 41 635 41 12074781
Berd D Sato T Maguire HC Kairys J Mastrangelo MJ Immunopharmacologic analysis of an autologous hapten-modified melanoma vaccine J Clinical Oncology 2004 22 1 13
Hodi FS Scott G Joseph A Withdrawal of immunosuppression contributing to the remission of malignant melanoma: a case report Cancer Immunity 2005 5 7 10 15901138
Penn i Tumors in the immunocompromised patient Ann Rev Med 1988 39 63 73 3285791
|
16060965
|
PMC1185554
|
CC BY
|
2021-01-04 16:24:16
|
no
|
Cancer Cell Int. 2005 Aug 1; 5:25
|
utf-8
|
Cancer Cell Int
| 2,005 |
10.1186/1475-2867-5-25
|
oa_comm
|
==== Front
J CarcinogJournal of Carcinogenesis1477-3163BioMed Central London 1477-3163-4-91601880010.1186/1477-3163-4-9ResearchIntercalated duct cell is starting point in development of pancreatic ductal carcinoma? Wada Ryo [email protected] Kaoru [email protected] Toshikazu [email protected] Takayuki [email protected] Michio [email protected] Department of Pathology, Juntendo University Shizuoka Hospital, Shizuoka, Japan2 Department of Pathology (I), Juntendo University, School of Medicine, Tokyo, Japan3 Department of Internal Medicine, Juntendo University Shizuoka Hospital, Shizuoka, Japan4 R & D Center, Biomedical Laboratories, Inc., Kawagoe, Saitama, Japan2005 14 7 2005 4 9 9 16 3 2005 14 7 2005 Copyright © 2005 Wada et al; licensee BioMed Central Ltd.2005Wada 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
Although it is well known that the pancreatic ductal carcinoma may develop having a relationship to the mucous gland hyperplasia (MGH) with atypia (PanIN-1B by PanIN system), the starting point of this atypical MGH is unclear. To know it, we examined the pancreas tissue using many methods described below.
Methods
1. Twenty-seven surgically resected pancreas tissue specimens, including pancreatic ductal carcinomas (PDC), chronic pancreatitis and normal pancreas, were investigated using immunohistochemical stainings for MUC1, MUC6, 45M1, Ki67 and p53. 2. DNA extraction and analysis of K-ras mutation at codon 12 using microdissection method: The paraffin blocks with 16 regions including the intercalated duct cell (IC) adjacant to the atypical MGH were prepared for DNA extraction. Mutation of K-ras codon 12 was analized and compared in enriched polymerase chain reaction-enzyme-linked minisequence assay (PCR-ELMA).
Results
1. In the normal pancreas, although no positive cell was seen in 45M1, p53, Ki67, the cytoplasm of IC were always positive for MUC1 and sometimes positive for MUC6. In the pancreas with fibrosis or inflammation, MGH was positive for MUC6 and 45M1. And atypical MGH was positive for MUC1, MUC6 and 45M1. Some IC adjacent to the atypical MGH was positive for Ki67 as well as atypical MGH. The carcinoma cells in all cases of PDC were diffusely positive for MUC1, 45M1, p53 and Ki67, and focally positive for MUC6. 2. In K-ras mutation, we examined the regions including IC adjacent to the atypical MGH, because the immunohistochemical apomucin stainings of these regions resembled those of PDC as decribed above. And K-ras mutation was confirmed in 12 of 16 regions (75%). All mutations were a single mutation, in 6 regions GTT was detected, in 4 regions GAT was detected and in 2 region AGT was detected.
Conclusion
Some intercalated duct cell may be the starting point of the pancreatic ductal carcinoma, because the exhibitions of mucin expressions, Ki67, p53 and K-ras mutation in some intercalated duct cell resembled those of mucous gland hyperplasia or pancreatic ductal carcinoma.
pancreatic cancerintercalated ductmucous gland hyperplasiaPanINK-rashistogenesismolecular analysis
==== Body
Background
The pancreatic ductal carcinoma (PDC) is fatal, even if its size is very small [1,2]. Therefore, it is very important to know the characteristics of pre-cancerous lesion of PDC for the preventive medicine and the early detection of PDC.
It has been well known that the mucous gland hyperplasia (MGH) (goblet cell metaplasia) is one of pre-cancerous lesion of PDC [3-5], and today, the histogenesis of PDC has been accepted by a model for a sequence of morphological changes, named the PanIN system [6], in which the lower grade PanIN is thought to exchange to higher grade PanIN and finally to PDC [7-9]. And MGH with no atypia is almostly equal to PanIN-1A and MGH with atypia is almostly equal to PanIN-1B in this system [6-9].
The main purpose of the present study was to investigate the starting point of MGH as pre-cancerous lesion of PDC.
Methods
Twenty-seven surgically resected pancreas tissue specimens, including 14 cases of PDC, which consisted of moderately differentiated tubular adenocarcinoma, 7 cases of the chronic pancreatitis and 6 cases of the normal pancreas, were assessed at the Department of Pathology, Juntendo University Shizuoka Hospital between 1998 and 2004. Informed consent for the medical examinations described below was obtained from all patients.
The specimens were fixed in 10% formalin solution for 1 – 5 days and prepared by cutting the lesions into 4 – 5 mm sections. Sections were embedded in paraffin, sectioned at a thickness of 4 μm and stained with hematoxylin and eosin (HE) and many immunohistochemical stainings described below. And after these stainings, some paraffin blocks were used for DNA extraction.
1. Immunohistochemical stainings were performed by the avidin-biotin-peroxidase complex method
anti-p53 oncoprotein (p53: DO7, monoclonal antibody, Novocastra Inc., UK), anti-MUC1 glycoprotein (MUC1: human CA 15-3, DF3, monoclonal antibody, DAKO, USA), anti-MUC6 glycoprotein (MUC6: CLH5, monoclonal antibody, Novocastra Inc., UK), anti-human gastric mucin-45M1 (45M1, monoclonal antibody, Novocastra Inc., UK) and anti-Ki-67 (MIB-1, monoclonal antibody, Coulter Japan Inc., Japan) (MUC1 and MUC6 at a dilution of 1:50, 45M1 at a dilution of 1:50, p53 at a dilution of 1:100 and Ki67 at a dilution of 1:100). Every staining was performed with 15-minute microwave treatment.
2. DNA extraction and analysis of K-ras mutation at codon 12
In 16 regions including the intercalated duct cells (IC) adjacent to the atypical MGH (panIN-1B), the existence or nonexistence of the K-ras codon 12 mutation was investigated, because the immunohistochemical expressions of the IC were very interesting, as described in the results, and K-ras codon 12 mutation is well known as one of the popular genetic abnormalities of PDC and MGH [10-13].
Paraffin blocks with the target foci were prepared for DNA extraction. The target foci were microdissected using a 20-gauge needle, comparing the slide with HE staining of same position. Extracted DNA was diluted with 5 ml of TaKaRa DEXPAT (for DNA Extraction from Paraffin-embedded Tissue, TaKaRa Biomedical Inc.).
Mutation of K-ras codon 12 was analized and compared in enriched polymerase chain reaction-enzyme-linked minisequence assay (PCR-ELMA) [14,15]. In PCR-ELMA, upstream for the first and second PCR were 5'-TAAACTTGTGGTAGTTGGAACT-3', downstream for the first PCR was 5'-GTTGGATCATATTCGTCCAC-3', and downstream for the second PCR was 5'-CAAATGATCTGAATTAGCTG-3'. The first PCR reaction was performed containing 1 μL of DNA lysate, 100 μM dNTP, 1.5 mM MgCl2, 1 μM each primers, 0.625 U Taq DNA polymerase and 1 × PCR buffer [containing 10 mM Tris-HCl (pH 8.3 at 25°C), 50 mM KCl and 0.001%(w/v) gelatin] in thermal cycler. And 10 μL of the denatured second PCR product was hybridized with probes for detecting the K-ras codon 12 wild-type (GGT) and six mutants (GAT, GCT, GTT, AGT, CGT and TGT) DNA were immobilized, at 55°C for 30 minutes, and 100 μL of biotinated A and 0.01 U of TdqDNA polypmerase were added and incubatuon was continued at 55°C for 30 minutes.
In development, 100 μL of avidine-horseradish peroxidase conjugate was contained and the mixture was performed at room temperature for 30 minutes, and 100 μL of tetramethylbenzidine (TMB) substrate was containd and the plates were performed in the dark at room temperature for 20 minutes. And 100 μL of stop solution was contained and the light absorbance of each sample was measured by spectrophotometry (Multiskan Multisoft, Labsystems, Tokyo) with a 450 nm filter wavelength (Figure 1).
Figure 1 Enriched polymerase chain reaction – enzyme linked mini-sequence assay (PCR-ELMA), showing K-ras codon 12 mutation in mucous gland hyperplasia of the pancreas. AGT – type mutation was seen in lane 3.
Results
1. Immunohistochemical studies of pancreas tissue (Table 1)
Table 1 Positivity of Pancreatic Epithelium for Several Stainings
MUC1 MUC6 45M1 Ki67 P53
Acinal cell (100 regions) - ± - - -
Intercalated duct (100 regions) + ± - - -
Duct (100 regions) - ± - - -
MGH with no atypia (50 regions) - + + ± -
Atypical MGH (20 regions) + + + ± -
Ductal carcinoma (15 lesions) + ± + + +
MGH: Mucous gland hyperplasia -: negative, +: positive, ±: positive or negative
In the normal pancreas (acinal cell: 100 regions, IC: 100 regions, duct: 100 regions), although no positive cell was seen in 45M1, p53, Ki67, the cytoplasm of IC were always positive for MUC1 (Figure 2) and sometimes positive for MUC6.
Figure 2 In normal pancreas, only the intercalated duct cells were positive for MUC1. (Left: HE, Right: MUC1, original magnification × 400)
In the pancreas with fibrosis or inflammation, the cytoplasm of the epitheli in MGH [16-18] with both no atypia (50 regions) and atypia (20 regions) was positive for 45M1 (Figure 3) and MUC6. And these atypical MGH (20 regions), which were PanIN-1B by PanIN system [1], were also positive for MUC1. No positive cell for p53 was seen in MGH with atypia or no atypia. Some IC adjacent to the atypical MGH (PanIN-1B) was positive for Ki67 (Figure 4).
Figure 3 The epithelium of the mucous gland hyperplasia was positive for 45M1. (Left: HE, Right: 45M1, original magnification × 200)
Figure 4 The intercalated duct cells adjacent to the atypical mucous gland hyperplasia were positive for Ki67. (Ki67, original magnification × 200)
The carcinoma cells in all cases (14 cases) of PIDC were diffusely positive for MUC1, 45M1, p53 and Ki67, and focally positive for MUC6. Namely, in all cases, the nuclei of many carcinoma cells were positive for p53 and Ki67. And in all cases, the expression of MUC1, MUC6 and 45M1 apomucins was found in the cell membranes of many carcinoma cells and the cytoplasms of some carcinoma cells.
2. K-ras mutation
In K-ras mutation, we examined the regions including IC adjacent to the atypical MGH, because the immuno-histochemical apomucins stainings of these regions were almostly equal to those of PDC described above. And K-ras mutation was comfirmed in 12 of 16 regions including IC adjacent to the atypical MGH (75%). All mutations were a single mutation, in 6 regions GTT was detected, in 4 regions GAT was detected and in 2 region AGT was detected.
Discussion
Generally, the neoplasia or pre-neoplasia will have some characteristics of its original cell and may develop from the proliferative cell in the organ. Thus, it is important to find out the normal cells which have the characteristics of the neoplasia or pre-neoplasia for detecting the starting point of the neoplasia.
Therefore, in the present study, the mucin expressions which are known as one of the characteristics of PDC [19-24] and K-ras mutation which is very famous gene abnormality in PDC [10,11] were investigated the non-neoplastic tissue or the neoplasia.
Although the results in the current study may be little novel and supportive of the known facts, these findings should reveal that the present study was correct.
That is to say, in the current study, the intercalated duct cells (IC) were positive for MUC1, although the other normal epitheli of the pancreas were negative for it. In the other reports, the positivity for MUC1 in the normal pancreas lack consistency, positive [22] or negative [21,23]. The reason of this discrepancy is unkown, however, in this study, the normal pancreas tissues were obtained from surgical resection due to non-pancreatic diseases and these specimens were quickly fixed in formalin solution. Thus, these specimens were suitable for the detailed histological examination in the normal pancreas and the results in the current study should be reliable.
And the results in the present study indicated that the mucin phenotypes of the atypical MGH (PanIN-1B by PanIN system) had a diffusely positive reaction for both anti-MUC1 staining and anti-45M1 staining, and had a spasely positive reaction for anti-Ki-67 staining, and these findings resembled those of the pancreatic ductal carcinoma (PDC), except for the expression of p53 (MGH: negative, PDC: positive). The expressions of the apomucins, Ki67 and p53 in MGH and PDC in the present study matched mostly those of other reports [20,21,23-26].
It has been well known that the K-ras mutation may confirm in MGH and PDC [10-13] and the results in the current study also showed that the K-ras codon 12 mutations were confirmed in the regions including IC adjacent to the atypical MGH of human pancreas (GGT→GTT in 6 regions, GGT→GAT in 4 regions and GGT→AGT in 2 region). And some IC adjacent to the atypical MGH was positive for Ki67.
Namely, the IC adjacent to the atypical MGH (PanIN-1B), which is sometimes proliferative cell, had the mucin phenotypic expression and high frequency of K-ras mutation as well as PDC.
Thus, we think that some IC is the starting point of MGH and may be thought to be the starting point of PDC, considering the histogenesis of PDC (from lower grade PanIN to higher grade PanIN, to PDC).
Further molecular studies concerning the intercalated duct cell in various pancreatic disease should be warranted.
Conclusion
Some intercalated duct cell may be the starting point of the pancreatic ductal carcinoma, because the exhibitions of mucin expressions, Ki67, p53 and K-ras mutation in some intercalated duct cell resembled those of mucous gland hyperplasia or pancreatic ductal carcinoma.
Abbreviations
IC, intercalated duct cell ; PDC, pancreatic ductal carcinoma ; MGH, mucous gland hyperplasia
Acknowledgements
The authors thank Mr. D. Mrozek for assistance with manuscript.
==== Refs
Tsuchiya R Tajima Y Matsuzaki S Early pancreatic cancer Pancreatology 2001 1 597 603 12120242 10.1159/000055869
Egawa S Takeda K Fukuyama S Clinicopathological aspects of small pancreatic cancer Pancreas 2004 28 235 240 15084963 10.1097/00006676-200404000-00004
Sommers SC Murphy SA Warren S Pancreatic duct hyperplasia and cancer Gastroenterology 1954 27 629 640 13210601
Kozuka S Sassa R Taki T Relation of pancreatic duct hyperplasia to carcinoma Cancer 1979 43 1418 1428 445339
Chen J Baithun SI Ramsay MA Histogenesis of pancreatic carcinomas: a study based on 248 cases J Pathol 1985 146 65 76 4009323 10.1002/path.1711460108
Hurban RH Adsay NV Albores-Saavedra J Pancreatic intraepithelial neoplasia: a new nomenclature and classification system for pancreatic duct lesions Am J Surg Pathol 2001 25 579 586 11342768 10.1097/00000478-200105000-00003
Hruban RH Wilentz RE Maitra A Identification and analysis of precursors to invasive pancreatic cancer Methods Mol Med 2004 103 1 14 15542896
Kloppel G Luttges J The pathology of ductal-type pancreatic carcinomas and pancreatic intraepithelial neoplasia: insights for clinicians Curr Gastroenterol Rep 2004 6 111 118 15191688
Hruban RH Takaori K Klimstra DS An illustrated consensus on the classification of pancreatic intraepithelial neoplasia and intraductal papillary mucinous neoplasms Am J Surg Pathol 2004 28 977 987 15252303
Almoguera C Shibata D Forrester K Most human carcinomas of exocrine pancreas contain mutant c-K-ras genes Cell 1988 53 549 554 2453289 10.1016/0092-8674(88)90571-5
Grunnewald K Lyons J Frohlich A High frequency of Ki-ras codon 12 mutations in pancreatic adenocarcinomas Int J Cancer 1989 43 889 895
Tabata T Fujimori T Maeda S The role of ras mutation in pancreatic cancer, precancerous lesions, and chronic pancreatitis Int J Pancreatol 1993 14 237 244 8113625
Yanagisawa A Ohtake K Ohashi K Frequent c-Ki-ras oncogene activation in mucous cell hyperplasias of pancreas suffering from chronic inflammation Cancer Res 1993 53 953 956 8439969
Matsubayashi H Watanabe H Yamaguchi T Difference in mucus and K-ras mutation in relation to phenotypes of tumors of the papilla of Vater Cancer 1999 86 596 607 10440687 10.1002/(SICI)1097-0142(19990815)86:4<596::AID-CNCR8>3.0.CO;2-H
Wada R Yamaguchi T K-ras codon 12 mutations of the super-minute dysplasia in Barrett's esophagus by DNA extraction using a microdissection method Dis Esophagus 2003 16 214 217 14641312 10.1046/j.1442-2050.2003.00331.x
Walters MN Goblet-cell metaplasia in ductules and acini of the exocrine pancreas J Pathol Bacteriol 1965 89 569 572 14320299 10.1002/path.1700890215
Pour PM Sayed S Sayed G Hyperplastic preneoplastic, and neoplastic lesions found in 83 human pancreas Am J Clin Pathol 1982 77 137 152 7039298
Cubilla AL Fitzgerald PJ Tumors of the exocrine pancreas Atlas of Tumor Pathology, Fascile 19, 2nd Series 1984 Armed Forces Institute of Pathology, Washington DC 71 89
Chambers JA Hollingsworth MA Trezise AE Development expression of mucin genes MUC1 and MUC2 J Cell Sci 1994 107 413 424 7515892
Terada T Ohta T Sasaki M Expression of MUC apomucins in normal pancreas and pancreatic tumours J Pathol 1996 180 160 165 8976874 10.1002/(SICI)1096-9896(199610)180:2<160::AID-PATH625>3.0.CO;2-A
Masaki Y Oka M Ogura Y Sialylated MUC1 mucin expression in normal pancreas, benign pancreatic lesions, and pancreatic ductal adenocarcinoma Hepatogastroenterology 1999 46 2240 2245 10521973
Balague C Gambus G Carrato C Altered expression of MUC2, MUC4, and MUC5 mucin genes in pancreas tissues and cancer cell lines Gastroenterology 1994 106 1054 1061 8143972
Monges GM Mathoulin-Portier MP Acres RB Differential MUC1 expression in normal and neoplastic human pancreatic tissue. An immunohistochemical study of 60 samples Am J Clin Pathol 1999 112 635 640 10549250
Yonezawa S Horinouchi M Osako M Gene expression of gastric type mucin (MUC5AC) in pancreatic tumors: its relationship with biological behavior of the tumor Pathol Int 1999 49 45 54 10227724 10.1046/j.1440-1827.1999.00823.x
Scarpa A Capelli P Mukai K Pancreatic adenocarcinomas frequently show p53 gene mutation Am J Pathol 1993 142 1534 1543 8494051
Redston MS Caldas C Seymour AB p53 mutations in pancreatic carcinoma and evidence of common involvement of homocopolymer tract in DNA microdeletions Cancer Res 1994 54 3025 3033 8187092
|
16018800
|
PMC1185555
|
CC BY
|
2021-01-04 16:39:20
|
no
|
J Carcinog. 2005 Jul 14; 4:9
|
utf-8
|
J Carcinog
| 2,005 |
10.1186/1477-3163-4-9
|
oa_comm
|
==== Front
Cerebrospinal Fluid ResCerebrospinal Fluid Research1743-8454BioMed Central London 1743-8454-2-21595338610.1186/1743-8454-2-2ResearchGenetic loci for ventricular dilatation in the LEW/Jms rat with fetal-onset hydrocephalus are influenced by gender and genetic background Jones Hazel C [email protected] Crystal F [email protected] David A [email protected] Mei [email protected] Barbara J [email protected] Department of Pharmacology and Therapeutics, University of Florida, Gainesville, FL 32610, USA2 Dr. H. C. Jones, Gagle Brook House, Chesterton, Bicester, Oxon OX26 1UF, UK2005 12 6 2005 2 2 2 16 12 2004 12 6 2005 Copyright © 2005 Jones et al; licensee BioMed Central Ltd.2005Jones 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 LEW/Jms rat strain has inherited hydrocephalus, with more males affected than females and an overall expression rate of 28%. This study aimed to determine chromosomal positions for genetic loci causing the hydrocephalus.
Methods
An F1 backcross was made to the parental LEW/Jms strain from a cross with non-hydrocephalic Fischer 344 rats. BC1 rats were generated for two specific crosses: the first with a male LEW/Jms rat as parent and grandparent, [(F × L) × L], designated B group, and the second with a female LEW/Jms rat as the parent and grandparent [L × (L × F)], designated C group. All hydrocephalic and a similar number of non-hydrocephalic rats from these two groups were genotyped with microsatellite markers and the data was analyzed separately for each sex by MAPMAKER.
Results
The frequency of hydrocephalus was not significantly different between the two groups (18.2 and 19.9 %), but there was a significant excess of males in the B group. The mean severity of hydrocephalus, measured as the ventricle-to-brain width ratio, was ranked as B group < C group < LEW/Jms. For the both rat groups, there were several chromosomes that showed possible regions with association between phenotype and genotype significant at the 5% or 1.0% level, but none of these had significant LOD scores. For the C group with a female LEW/Jms parent, there was a fully significant locus on Chr2 with a LOD score of 3.81 that was associated almost exclusively with male rats. Both groups showed possible linkage on Chr17 and the data combined produced a LOD score of 2.71, between suggestive and full significance. This locus was largely associated with male rats with a LEW/Jms male parent.
Conclusion
Phenotypic expression of hydrocephalus in Lew/Jms, although not X-linked, has a strong male bias. One, and possibly two chromosomal regions are associated with the hydrocephalus.
==== Body
Introduction
Fetal hydrocephalus occurs in humans from causes such as intraventricular hemorrhage and intrauterine infections, but in other cases the cause cannot be identified with certainty. Epidemiological studies provide evidence that hydrocephalus has a genetic component [1-3], although only one inherited form, X-linked hydrocephalus, has been characterized at the molecular level [4]. Rodent hydrocephalus mutants have been known for many years [5] and a few mouse mutants have been genetically characterized [6-8]. The publication of the first DNA assembly for the rat has enabled the rat genome to be integrated with DNA sequences from other species [9]. It is now possible to identify homologous regions between rat and human or rat and mouse, and to place disease-related genes from the human or mouse on the rat genome. Additionally, the identification of candidate genes for specific traits in rats is possible through comparative mapping. The study of disease-related genes in the rat will lead to a better understanding of inherited conditions in humans.
The LEW/Jms rat was first described in 1983 as being derived from an inbred strain of Wistar-Lewis rats and as having lethal fetal-onset hydrocephalus with a frequency that varied between litters from 12 – 25% and sometimes higher [10]. A six-generation pedigree showed that about 25% of the breeding pairs did not produce hydrocephalic pups. The authors concluded that there was a Mendelian autosomal recessive mode of inheritance.
Seven rats of the strain were received at the University of Florida in year 2000 and their DNA was tested with 87 selected microsatellite markers. All but two of these markers were both homozygous and homogeneous indicating that the strain was almost completely inbred [11]. Since 2000 the strain has been maintained by brother-sister mating and almost all successful breeding pairs have produced hydrocephalic offspring. This suggests that hydrocephalus may not be a Mendelian recessive trait [11]. Severe hydrocephalus is evident soon after birth from a domed head, with death occurring soon after weaning. Therefore the strain is maintained by breeding from apparently non-hydrocephalic rats because pups with overt disease do not survive to reproduce. It was found, however, that some adult ex-breeding rats have a milder form of hydrocephalus. These rats, however, did not produce pups with an increased frequency for hydrocephalus, which would have been expected with direct transmission of the trait. The overall frequency of hydrocephalus among pups was 27.7%, with a significant excess of affected males. Crossing to another rat strain, Fischer 344, produced a small number of pups with hydrocephalus (3%). A backcross from the F1 progeny to the LEW/Jms strain produced hydrocephalic pups, also with an excess of males and a frequency of hydrocephalus of 18.8% [11]. The presence of affected pups in the F1 generation and the high frequency of affected BC1 pups suggest that the trait may be semi dominant and controlled by one or possibly two genetic loci. This study aimed to perform a genome-wide scan and QTL analysis on backcross progeny to identify chromosomal region(s) associated with the hydrocephalus. Using gender-specific crosses, the genotyping has revealed one and possibly two loci associated with hydrocephalus.
Materials and Methods
Animals
For all experiments the 'Principles of Laboratory Animal Care' (NIH publication no. 86-33, revised 1985) was followed. All rats were pathogen free at the start of the experiment and were housed for the duration of the experiment in a single room under conventional conditions. Pathogen monitoring was performed periodically. The LEW/Jms strain was donated by Dr. K. Sudoh, University of Tokyo, to H.C.J. at King's College, London, UK, in 1987. Between 1991 and 2000, they were housed at the University of Manchester Institute of Science and Technology, Manchester, UK (C. S. Bannister). Seven animals transferred to the University of Florida in 2000 were the founder rats for the current breeding colony and for this backcross experiment. The animals used in this study were selected at random from the four breeding lines described previously [11]. Inbred Fischer 344 rats were purchased from Harlan (Harlan F344/Hsd). This strain does not develop hydrocephalus and was used in a previous genetic analysis with the H-Tx hydrocephalic strain [12].
Breeding
In the first part of the study, the LEW/Jms (L) rats were bred to Fischer 344 (F) and the F1 progeny backcrossed to LEW/Jms as described previously producing 1574 backcross (BC1) progeny [11]. A genotype analysis using the complete set of BC1 progeny did not produce meaningful results. Hence the results were examined according to the sex of the parents (see Results, Analysis of genotypes). Of 1574, 599 had LEW/Jms as the paternal parent for both generations, designated 'B' group [(F × L) × L], 114 of which had hydrocephalus. A further 365 were designated 'C' group with LEW/Jms as the maternal parent for both generations [L × (L × F)] and 68 had hydrocephalus. Additional backcross breeding was carried out to increase the number of rats within these two specific crosses. For the 'B' group progeny, 8 female F1 rats bred from LEW/Jms males and F344 females, were crossed with 8 male LEW/Jms rats and 373 BC1 progeny were generated. For the 'C' group progeny, 24 male F1 rats bred from F344 males and LEW/Jms females, were crossed with 24 LEW/Jms females and 608 BC1 progeny were generated. Records of the breeding pairs and litters born were entered into a database (Filemaker Pro, Filemaker Inc, CA, USA) for both LEW/Jms breeding colony and the backcross.
Analysis of phenotype
In a previous study it was shown that the severity of hydrocephalus as measured by the ratio of ventricle-to-brain width is independent of age between pups aged 2–23 days after birth. This was true for both the LEW/Jms colony and the BC1 rats [11]. To measure phenotype rats were euthanized with CO2 suffocation, or in the case of pups less than 10 days old, with an overdose of sodium pentobarbital (100 mg/kg). The brains were excised, fixed in 10% neutral buffered formalin and sliced coronally at 1 mm thickness using a fine blade. The slices were examined and photographed under a binocular microscope. A slice at the level of the striatum was photographed and the dilatation measured as the ratio of ventricle width-to-brain width (hydrocephalus severity, [13]). Non-hydrocephalic (control) rats were given a nominal phenotype of 0.01 because although ventricles are not visible on 1 mm slices, small ventricles are found in histological sections [14]. Phenotype measurements were entered into the MAPMAKER program for analysis. The data was compared to data from the brains of 392 rats from the LEW/Jms parental strain described previously [11].
Genotyping
With the exceptions described below, BC1 progeny were sacrificed between 2 and 23 days of age. Liver tissue was removed, frozen in liquid N2 and stored at -80°C for DNA extraction. Because it was found previously that a proportion of ex-breeding LEW/Jms rats had ventricular dilatation when examined post mortem [11], some BC1 rats were raised until 22–24 weeks of age, at which stage they were sacrificed and tissues removed; n = 54 rats from five litters for the B group and n = 60 rats from five litters for the C group. Genomic DNA was extracted from liver tissue using the standard chloroform-phenol method, amplified by PCR using primers for microsatellite markers (Rat Map Pairs, Research Genetics or Invitrogen) and separated by agarose gel electrophoresis as described previously [15].
For the initial study, DNA was extracted from all 247 hydrocephalic rats and from 168 littermates that had no ventricular dilatation (non-hydrocephalic rats). This set of BC1 rats had mixed parentage, consisting of 29% that had a female LEW/Jms parent for the first generation and a male for the backcross, 39% with both LEW/Jms parents being male rats (designated 'B' type), 23% with a female LEW/Jms parent in the first cross and also in the backcross (designated 'C' type), and 8.5% with a female parent in the first cross and a male for the backcross. Genotyping was performed in stages using a panel of 96 genome-wide microsatellite markers. The mean spacing between markers was 14.36 cM and the largest was 41.64 cM on Chr20, where informative markers were scarce. Apart from this, two other markers on Chrs7 and 18 had spacings >30 cM (30.40 and 31.75) and all other spacings were < 30 cM. The results were examined with the χ2 test for significant departure from the null hypothesis that the ratio of homozygous to heterozygous genotypes was 50:50.
For the additional rats generated in the B and C categories, DNA was extracted from the frozen liver of all overtly hydrocephalic pups an equal number of non-hydrocephalic littermates, and from pups found to have mild hydrocephalus after examination of the brains. First-stage genotyping was performed using DNA from 30 or more hydrocephalic and non-hydrocephalic rats from each group, with the same genome-wide panel of 96 microsatellite markers. The data was combined with data from the initial study, analyzed separately for each group, and examined for significance at the 5% level. The presence of significance at the 5% level, while not sufficient for likely linkage, was used as a guideline to determine the strategy for possible further genotyping. Genotyping was then continued on specific chromosomes where there was significant association, until all rats had been included. Additional markers were included to increase the density on these chromosomes. The data for each chromosome was analyzed by MAPMAKER.EXP to determine the best marker order and by MAPMAKER.QTL to calculate the LOD score [16]. Significance levels were determined using a LOD score of 1.9 (P < 0.0034) for suggestive significance and a score of 3.3 (P <0 .0001) for full significance as defined by Lander & Krugylak [17].
X Chromosome analysis
Since there was an excess of males with hydrocephalus, a possible association with ChrX was sought. Genotypes of the B type progeny with a male LEW/Jms parent at each generation, were informative for ChrX. The C progeny could not be used because the contribution from the non-hydrocephalic F344 strain came from ChrY. Seven microsatellite markers polymorphic for the LEW/Jms × F344 cross were genotyped on all 'B' rats and the data analyzed independently for each sex by the χ2 test.
Human-rat homology to search for candidate genes
Candidate genes were sought for the chromosomal regions where there was significant evidence for a hydrocephalus locus as determined by MAPMAKER. A strategy was used similar to that described previously for the H-Tx rat linkage analysis [12]. The Ensemble Rat Genome Browser, a joint project between the European Molecular Biology Laboratory-European Bioinformatics Institute and the Sanger Institute (, version 26.3 d.1, 08/02/2004) was used to identify the megabase (Mb) positions for the microsatellite markers in the vicinity of the hydrocephalus loci. The likely genetic positions for the loci were identified based on the LOD score maps generated by MAPMAKER. Possible candidate genes were selected from known rat genes and from predicted genes that were identified from homologous regions on the human and the mouse genomes. Genes were then evaluated as potential candidates using a number of different criteria [12].
Results
Breeding and expression of hydrocephalus
LEW/Jms parental strain: From breeding records kept over a period of 3.5 y, the overall frequency of overt hydrocephalus was 28% out of a total of 2401 pups (Fig. 1a, column 'all'). As reported previously, there were significantly more males with hydrocephalus than females, χ2 = 46.21, but no significant difference between the sexes of non-hydrocephalic rats. Instead there was a small, but significant excess of total males over total females, χ2 = 12.3 (Fig 1b). The frequency of hydrocephalus varied with parity, in that the percentage in second and third litters, 31.2% and 47.6% respectively, was significantly increased over that in first litters, 22.2%, P < 0.001, χ2 test (Fig 1a). Although the frequency in the fourth and fifth litters was also increased over the first litter, there was no statistical significance, possibly because the numbers were small as some females did not continue to breed after the third litter. The average litter size decreased from 9.5 pups in first litters to 5.0 in fifth litters.
Figure 1 a: Percentage of hydrocephalus in LEW/Jms rats by litter number (parity) and total percentage (All). There was a significant increase in percentage of affected rats between the 1st and 2nd and the 1st and 3rd litters, P < 0.001. This was not maintained for the 4th and 5th litters. The number of litters in the data sets is depicted above the columns.b: The ratio of males to females in the parental LEW/Jms strain and in the BC1 progeny. In the parental strain, there were significantly more males than females with hydrocephalus (black bars, P < 0.001), and also among the total pups born (striped bars). In the B group [(F × L) × L], there was a significant excess of females among the non-hydrocephalic pups, (open bars, P < 0.001 and a significant excess of males in the hydrocephalic pups, P < 0.001. In the C group [L × (L × F)], there were no significant departures from the expected 1:1 sex ratios for the hydrocephalic or the non-hydrocephalic rats (ns).c: Percentage of hydrocephalus in [(F × L) × L] BC1 rats by litter number (parity) and in total. There was a significant increase in the percentage expressing hydrocephalus between the 1st and 2nd litters, P < 0.001. This was not maintained in subsequent litters. The number of litters in the data sets is depicted above the columns. d: Percentage of hydrocephalus in [L × (L × F)] BC1 rats by litter number (parity) and in total. There was a steady and significant increase in expression between 1st litters and all subsequent litters, P < 0.001. The number of litters in the data sets is depicted above the columns.
Backcross progeny
Table 1 shows the total number of BC1 pups sacrificed at 2–23 days of age and numbers genotyped for each group. There was an excess of males over females with hydrocephalus in the B group, χ2 = 15.57 (Table 1, Fig. 1b). The same trend was also present in the C group, but not significant. Both groups also had an excess of female over male non-hydrocephalic rats, although again, it was only significant for the B group, χ2 = 11.02. The frequency of hydrocephalus in the B and C backcross progeny was not significantly different between the two groups, 18.2% and 19.9%, respectively (Figs. 1c and 1d, column 'All'). Similar to the parental LEW/Jms strain, the frequency of hydrocephalic pups depended on parity and was significantly lower in the first litters than in subsequent litters. This effect was much less evident for the B progeny than for C progeny, which had the female LEW/Jms parent (Figs. 1c and 1d). The ratio of male to female hydrocephalic pups was not significantly different between the 1st, 2nd, 3rd and 4th litters for either group, χ2 test. Among the backcross progeny sacrificed at 22–24 weeks, there were seven rats with mild hydrocephalus in the B group (n = 54) and nine in the C group (n = 60). An additional two had severe, but non-fatal disease in the B group and three in the C group. As with the pups, there was also an excess of males with hydrocephalus in these two groups. However, the sample was small and there was no statistical significance.
Table 1 Numbers of BC1 pups of each phenotype bred and genotyped (in parenthesis). In the B group there were more males than females with hydrocephalus, P < 0.001 (a) and more females than males in the non-hydrocephalic group P < 0.001 (b). The differences were not significant for the C group.
Parentage Sex Hydrocephalic Non-hydrocephalic Total
B Group [(F × L) × L] Male 109 (108)a 330 (91) 439 (198)
Female 58 (58) 421 (51)b 479 (109)
C Group [L × (L × F)] Male 102 (102) 342 (82) 444 (184)
Female 80 (79) 386 (58) 466 (137)
Analysis of Phenotypes
The mean severity of ventricular dilatation for the B group was 0.58 +/- 0.01 and for the C group was 0.61 +/- 0.01. This difference was significant, P < 0.05, Kruskal-Wallis test. There was no significant difference between the mean severity for males and females in either group (data not shown). Dilatation severity was 0.66 +/- 0.01 for the LEW/Jms parental strain, which was significantly higher than for B or C rats, P < 0.01 and 0.05, respectively. As reported previously, there was no significant difference in hydrocephalus severity between males and females in the parental strain [11].
Analysis of genotypes
The genotypes for the first backcross progeny with rats from mixed mating groups were examined using χ2 test for association between phenotype and genotype. On Chr17 at marker D17Rat17, there was significance at the level of P < 0.05. No other marker on any chromosome had a significant result. The data was re-tested after separation of the genotypes into four groups according to sex of the parental rats. However, as already stated, the number of rats was too low at this stage for the results to be meaningful. The genotypes obtained for B and C rats in the initial study were combined with data for the additional B and C rats bred subsequently. First, the genotypes were analyzed for the male and female data combined. In addition, because of the strong male bias in the expression of hydrocephalus, the data was analyzed separately for males and for females (Tables 2 and 3).
Table 2 Genotypes for B type [(F × L) × L] male and female rats showing markers for which the association between phenotype and genotype was significant at a level of 0.05 (*) or 0.01 (**).
[(F × L) × L] Males Genotypes
Hydrocephalic Non-hydro-cephalic
LL LF LL LF χ2 P value
Chr1 D01Rat56 45 63 51 40 4.09 *
D01Rat57 46 62 43 27 6.03 *
D01Rat65 47 60 52 37 4.09 *
D01Rat219 40 60 52 39 5.61 *
D01Rat67 47 59 54 38 4.06 *
D01Rat208 46 62 52 39 4.18 *
Chr 5 D05Rat49 45 33 26 38 4.10 *
Chr 11 D11Rat28 50 42 32 54 5.25 *
D11Rat73 53 40 35 56 6.33 *
Chr17 D17Rat85 65 40 40 51 6.31 *
D17Rat65 68 39 40 50 7.21 **
Chr19 D19Rat28 68 40 43 52 6.39 *
D19Rat12 66 41 42 52 5.82 *
D19Rat40 68 39 44 45 3.95 *
D19Rat95 63 43 32 41 4.22 *
L F L F χ2 P value
ChrX DXRat83 60 43 39 51 4.28 *
(F × L) × L] Females LL LF LL LF χ2 P value
Chr1 D01Rat36 24 34 31 19 4.57 *
Chr5 D05Rat36 35 23 19 32 5.79 *
D05Rat41 38 18 24 27 4.74 *
Chr13 D13Rat85 19 28 32 18 5.40 *
Chr19 D19Rat28 36 24 22 33 4.59 *
Table 3 Genotypes for C type [L × (L × F)] male and female rats showing markers for which the association between phenotype and genotype was significant at a level of 0.05 (*) or 0.01 (**).
Genotype
[L × (L × F)] Males Hydrocephalic Non-hydro-cephalic
LL LF LL LF χ2 P value
Chr2 D02Rat15 61 40 34 45 5.36 *
D02Rat34 63 41 34 48 6.71 **
D02Rat91 60 37 36 43 4.66 *
D02Rat52 64 40 37 45 4.98 *
D02Rat49 68 37 38 42 5.53 *
D02Rat46 66 38 39 41 3.99 *
D02Rat241 67 31 35 43 9.84 **
D02Rat62 65 40 34 48 7.72 **
D02Rat65 59 19 39 31 6.55 *
D02Rat250 58 38 32 48 7.28 **
Chr4 D04Rat112 37 66 47 34 8.93 **
D04Rat72 41 62 46 34 5.65 *
Chr10 D10Rat28 36 10 25 22 6.47 *
Chr16 D16Rat81 32 53 32 25 4.71 *
Chr17 D17Arb3 66 42 38 47 5.15 *
D17Arb13 66 41 38 47 5.50 *
D17Rat181 63 40 37 44 4.38 *
D17Rat17 66 40 39 45 4.75 *
[L × (L × F)] Females LL LF LL LF χ2 P value
Chr2 D02Rat21 48 25 23 35 8.87 **
D02Rat135 47 28 21 36 8.65 **
D02Rat26 48 25 25 28 4.35 *
D02Rat34 46 30 23 35 5.74 *
D02Rat91 49 25 23 35 9.25 **
D02Rat46 48 26 21 35 9.58 **
D02Rat52 52 24 26 32 7.53 **
D02Rat49 51 24 26 32 7.20 **
Chr7 D07Rat195 19 27 28 14 5.68 *
Chr17 D17Arb4 32 25 21 36 4.27 *
D17Rat181 41 34 21 36 4.13 *
D17Rat17 43 32 22 35 4.55 *
D17Arb5 43 27 25 33 4.28 *
B group QTL mapping
Analysis of the data for male and female rats combined showed two or more markers on each of four chromosomes with significance at the 5% level or above by χ2 test (Chrs 1, 5, 17, and 19). QTL analysis was performed with MAPMAKER for these chromosomes using the combined data, and also on data for each sex separately. None of the chromosomes reached the level required for full significance (3.3) for both sexes combined, but a score suggestive for significance was achieved on Chr5, LOD = 1.94, and almost achieved on 19, LOD = 1.89 (Fig 2). Separate male and female MAPMAKER analyses for chromosome 5 indicated a greater level of significance for females than males (Table 2, Figs. 2a, 2b). For chromosome 19, there was a higher significance level for the males (Table 2, Fig. 2c) and the contribution of the females to the combined LOD score was very small (Fig. 2d). On Chr17, the peak LOD for both sexes was 1.84 at D17Rat65 and again, the effect was largely on the males (Table 2). The X chromosome marker DXRat83 gave a significant result, P < 0.05 for the male rats in this B group (Table 2). In addition, the data showed that chromosomes 11 (for males) and 13 (for females) had results for one or two markers significant at the 5% level or higher (Table 2). The low significance levels and small LOD scores obtained for this group of rats did not contribute meaningful results to this genetic analysis. The exception was data for Chr17, where the results were combined for the B and C groups (see below).
Figure 2 LOD score graphs created from the MAPMAKER output for [(F × L) × L], B rats on chromosomes 5 and 19. The X-axes represent the recombination distance in centi-Morgans (cM). The microsatellite markers are positioned on the X-axis and named below. The horizontal dotted line represents the score required for suggestive significance (1.9). a: Plot for male B rats on Chr 5 and b: plot for female B rats (solid lines), dotted line represents a portion of the plot for both sexes combined, where the LOD score (1.94) reached suggestive significance. There was a strong female bias (3b). c: Plot for male B rats on Chr 19 and d: plot for female rats (solid lines). The dotted line near the centromeric end represents part of the plot for both sexes combined where the LOD score (1.89) is very close to suggestive significance for linkage. The male rats contributed to this peak almost exclusively.
C group QTL mapping
The combined analysis for male and female C rats showed quite different results to that seen for the B group. Instead, chromosomes 2 and 4 were significant for two or more markers at the 5% level or above. Of these, chromosome 2 had a peak of LOD = 3.81, indicating a locus with full significance for hydrocephalus situated near D2Rat241 (Figs 3a,b). The LOD score for chromosome 4 (1.59) did not reach the level for suggestive significance. Similar to the B group, there was a peak on Chr17 at a different location, D17Rat13, LOD = 1.73, but it was not significant. The males and females were analyzed separately (Table 3). The locus on Chr2 was gender specific in that it had a much larger effect on males with a LOD score of 3.43, whereas for the females the LOD score was only 1.41 (Table 3, Figs. 3a,b). The effect on Chr4 was almost totally on the male rats (Table 3). For Chr17, males and females were affected equally (Table 3). In addition to chromosomes 2, 4, and 17, described above, chromosomes 10 and 16 (for males) and 7 (for females) had data for one or more markers significant at the 5% level or above (Table 3). Only Chr2 was studied further, apart from Chr17 where the data was combined for both groups (see below).
Figure 3 LOD score graphs created from the MAPMAKER output for [L × (L × F)] C rats on Chr 2. The X-axes represent the recombination distance in centi-Morgans (cM). a: plot for male rats and b: for female rats (solid line). The horizontal dotted line represents the score required for suggestive significance (1.9) and the solid line the score for full significance (3.3). The two graphs are quite different for males and females with the males reaching a score indicative of full significance (3.43) in the same location as the map for both sexes combined (dotted line, LOD = 3.81).
Figure 4 LOD score graphs created from the MAPMAKER output for Chr 17, using the combined data for both B and C rats. The X-axes represent the recombination distance in centi-Morgans (cM). The horizontal dotted line
represents the score required for suggestive significance (1.9). a: plot for male rats and b: for female rats (solid lines). The combined male and female score represented a locus that was between suggestive and full significance (dotted plot, LOD = 2.71). This locus also had a male specificity that on further analysis was shown to come from the male B rats (data not shown).
B and C QTL combined
All data sets were combined for QTL analysis of Chr17. The maximum LOD score for all rats was 2.71, between suggestive and full significance, and situated near D17Rat62. For males it was 2.07, also close to D17Rat62, but for females the maximum LOD was only 1.26 and situated in a different position at D17Arb5 (Figs. 4a,b). Hence this may be a second gender-specific locus. The male bias largely came from the B group with the male LEW/Jms parent as described above.
In summary, the genotype analysis of B rats with a male LEW/Jms parent and grandparent showed no chromosomal regions indicative of a locus for hydrocephalus, with the possible exception of Chr17. On the other hand, analysis using C rats with a female LEW/Jms parent and grandparent showed a locus with significant linkage for hydrocephalus on Chr2 that chiefly affected males. The locus on rat Chr2 is situated at 217–218 Mb in band q41. The region at and around the locus is homologous to human chromosomes 1 and 4. Chromosome 17 linkage was common to both rat groups. There was a region of Chr17 that was between the suggestive and fully significant level for hydrocephalus when the data was combined. In this case, the locus acted on both sexes but more so with the male rats from the B group. The peak was situated close to D17Rat62 located at 83.5 Mb in band 17q12.3. This region is homologous to human 10p14 at 12.25 Mb.
Discussion
The LEW/Jms rat is a model for fetal-onset human hydrocephalus. In the human, ventriculomegaly, defined as dilated lateral ventricle atria, can be detected by ultrasound examination from 20 weeks of gestation and sometimes earlier [18]. In some cases the dilatation remains stable or resolves. In other cases there is progression to hydrocephalus with increased head circumference and a requirement for shunt treatment in the post-natal period. Fetal hydrocephalus is frequently associated with a poor neurodevelopmental outcome [18,19]. In many cases the primary cause is uncertain, but stenosis of the cerebral aqueduct is often a feature [20]. The LEW/Jms rat model falls into this category, having fetal-onset progressive hydrocephalus with an abnormal aqueduct [21,22]. In many respects the phenotype is similar to hydrocephalus in the H-Tx strain [23,24]. Both strains have severe fetal-onset disease associated with aqueduct stenosis and dysplasia of the subcommissural organ [14,21,22]. Hydrocephalus expression in H-Tx rats has been shown to be polygenic and influenced by at least four loci on different chromosomes [12] and by strong epigenetic effects [25]. However, neither gender nor cross-specific effects were observed in H-Tx. A surprising observation reported in this study was the increase in the frequency with parity from 22.2% in first litters to 47.6% in third litters. This appears to be a similar phenomenon to that observed in the H-Tx rat, where it was found that the frequency of hydrocephalus was lower in first litters than in subsequent litters [25]. In H-Tx hydrocephalus, the increase in hydrocephalus frequency among the pups in utero was associated with concurrent suckling by the dam of a previous litter. In the case of LEW/Jms rats, there was a progressive increase in frequency with parity but whether or not the phenomenon was related to concurrent suckling was not investigated. It does indicate, however, that there may be epigenetic effects affecting the expression of hydrocephalus in this strain as occurs in the H-Tx strain.
The results of this study suggest that there is a locus for hydrocephalus on Chr2, as shown in male rats with a female LEW/Jms parent. There is possibly a second locus on Chr17 that is associated with hydrocephalus in rats with parents of either sex, although the males with a male LEW/Jms parent made the largest contribution. This is the first time that a genetic analysis has been attempted in the LEW/Jms rat strain. It was reported previously that twice as many male as females rats are affected with hydrocephalus [11]. Although DNA samples from the BC1 progeny with a male LEW/Jms parent were tested with seven ChrX markers, only one marker, DXRat83 at position 43.2 Mb in band q21 on the X chromosome (Ensembl Rat Genome Browser ), showed a low level of significance (P < 0.05) with the male rats. It therefore seems unlikely that X chromosome linkage is involved despite the fact that X-linked hydrocephalus is well characterized in humans [4] and is due to mutations in the gene coding for L1 neural cell adhesion molecule. A more likely explanation for the specific sex effects is that gender affects phenotypic expression in this strain. Gender-specific loci have been observed in the analysis of other quantitative traits in rodents [26-28]. One explanation for this phenomenon is that genes on the autosomes, such as those coding for sex hormones, may influence the expression of disease-related genes.
In a recent genetic analysis of inherited hydrocephalus in the H-Tx rat, a region of Chr17 was identified at 25 – 55 cM that was highly significant for hydrocephalus in this strain. It was concluded that one or more loci exist in this region [12]. The equivalent genetic map positions were 71.2 – 92.6 Mb. This large region is homologous to human Chr1q43 and 10p11.21-p13. The locus identified here for LEW/Jms rats is located in the middle of this section, close to D17Rat62 located at 83.5 Mb in band 17q12.3. This region is homologous to human 10p14 at 12.25 Mb (Fig. 5). It is possible that the two strains share a common susceptibility locus for hydrocephalus on Chr17. There are three possible candidate genes in this region close to D17Rat 62. One is SPAG6 at human 10p12.2 and 22.6 Mb. This gene codes for sperm-associated antigen isoform 1 which is the murine homologue of a component of the central flagella apparatus in sperm flagellae. Spag6 knockout mice are infertile and have hydrocephalus [29]. These mice may have impaired cilia function in the brain, a potential cause for hydrocephalus [30,31]. Two other known genes located within 10 Mb of the locus are ITGA8 or integrin alpha 8 at human 10p13 and located at 15.6 Mb, and VIM coding for vimentin at human 10p13 and located at 17.3 Mb. Both of these genes are important for brain development [32,33] but have no known association with hydrocephalus. Three more candidate genes are located on this chromosome close to the largest linkage peak for the H-Tx strain but further from the LEW/Jms peak [12]. These are FDZ8, frizzled 8 precursor which codes for a Wnt receptor, MTR, methyltetrahydrofolate-homocysteine methyltransferase or vitamin B12-dependant methionine synthase, and the gene for the acetylcholine muscarinic type 3 receptor, CHRM3.
Figure 5 A schematic map for the hydrocephalus locus on rat Chr17. The scale on the left is the rat genetic length in megabases (Mb). Positions for the rat DNA markers (open squares) were identified from Rat Genome Browser, Ensemble web site . To the right of the rat chromosome are the human chromosome homologues. The positions for two possible hydrocephalus loci found in a previous study on H-Tx rats [12], and the peak for LEW/Jms rats in this study are marked (arrows). Possible candidate genes (black diamonds and arrows) are named and their human cytogenetic and Mb positions given in parenthesis.
The locus identified by MAPMAKER on Chr2 was located close to marker D2Rat241 at 217–218 Mb in band q41. This region is homologous with human Chr1 and with a section of human Chr4 (Fig 6). It is also homologous with mouse Chr3 and includes a region that is particularly rich in genes that are transcribed in the nervous system [34]. The regions at and around the locus were examined for possible candidate genes using Ensemble Genome Browser to identify rat genes and homologous human or mouse genes that are expressed in brain and might have an association with hydrocephalus. One such gene is P97582 (rat 224.2 Mb) or ANK2 (human Chr4, 114.3 Mb), which codes for brain ankyrin or ankyrinB. Ankyrins are spectrin-binding proteins on cell membranes that associate with L1 CAM, and with several ion channels. AnkyrinB deficient mice have a similar phenotype to L1 deficient mice with features that include dilated cerebral ventricles [35]. Close by at 222.9 Mb is CGT or 2 hydroxyacyl sphingosine 1-B-galactosyltransferase, a gene found in oligodendrocytes and involved in myelination [36]. Another attractive candidate is NGFB, or beta nerve growth factor precursor at position 197.2 Mb. NGF and other neurotrophins and their receptors are upregulated in brain damage including that caused by hydrocephalus [37]. Furthermore, hydrocephalic H-Tx rats have alterations in brain NGF concentrations [38] and children with hydrocephalus have elevated NGF in the CSF [39]. NOTCH2, or notch homologue protein 2 precursor, is at 192.8 Mb and close to NGFB. Notch proteins are transmembrane receptors involved in cell fate determination in the CNS and Notch2 is important for roof plate development [40]. Notch2 affects Wnt-1 expression, and in mouse the Wnt sw/sw mutant has defective SCO development and hydrocephalus [41]. Also in the same region of Chr2 is CA14 or membrane-associated carbonic anhydrase XIV precursor at 190.65 Mb. This isoform of carbonic anhydrase is expressed in choroid plexus in addition to neuronal cells but its function is not clear [42] although another isoform CAII plays an important role in CSF secretion at the choroid plexus [43].
Figure 6 A schematic map for the hydrocephalus locus on rat Chr2. The scale on the left is the rat genetic length in megabases (Mb). Positions for the rat DNA markers (open squares) were identified from Rat Genome Browser, Ensemble web site . To the right of the rat chromosome are the human chromosome homologues. The peak LOD score for LEW/Jms rats is marked (arrows). Possible candidate genes (black diamonds and arrows) are named and their human cytogenetic and Mb positions given in parenthesis.
The identification of possible candidate genes is extremely speculative because of the low resolution obtained from QTL linkage analysis and the fact that the chromosomal regions identified contain many hundreds of genes. In contrast to genetic diseases with Mendelian inheritance, QTL mapping for complex traits has not, in most cases, led to the identification of abnormal genes [44]. However, additional strategies are available such as expression profiling in disease states, DNA sequencing for polymorphisms in candidate genes and transgenic technology all of which can lead to gene identification. In conclusion, the genetic basis for hydrocephalus expression in LEW/Jms rats is associated with one or possibly two genetic loci and in addition, the phenotypic expression is strongly influenced by gender.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
HCJ conceived of the study, was responsible for its design and coordination and writing the manuscript. DAM and BJC participated in rat breeding. CFT, DAM, MY and BJC all participated in phenotype and genotype analysis. CFT was responsible for the male/female analysis, and assisted in figure, table and manuscript preparation.
Acknowledgements
This research was funded by the Maren Foundation and NIH NS40359. We are grateful for L. Morel for initial consultations. The technical assistance of E. Joy Akins, Baligh Yehia and Gin Fu Chen is gratefully acknowledged.
==== Refs
Lorber J The family history of uncomplicated congenital hydrocephalus: an epidemiological study based on 270 probands Br Med J 1984 289 281 284 6430438
Varadi V Toth Z Torok O Papp Z Heterogeneity and recurrence risk for congenital hydrocephalus (ventriculomegaly): a prospective study Am J Med Genet 1988 29 305 310 3354602 10.1002/ajmg.1320290209
Rittler M Liascovich R Lopez-Camelo J Castilla EE Parental consanguinity in specific types of congenital anomalies Am J Med Genet 2001 102 36 43 11471170 10.1002/1096-8628(20010722)102:1<36::AID-AJMG1394>3.0.CO;2-M
Weller S Gartner J Genetic and clinical aspects of X-linked hydrocephalus (L1 disease): Mutations in the L1CAM gene Hum Mutat 2001 18 1 12 11438988 10.1002/humu.1144
Gruneberg H Two new mutant genes in the house mouse J Genet 1943 45 22 28
Davy BE Robinson ML Congenital hydrocephalus in hy3 mice is caused by a frameshift mutation in Hydin, a large novel gene Hum Mol Genet 2003 12 1163 1170 12719380 10.1093/hmg/ddg122
Chae TH Kim S Marz KE Hanson PI Walsh CA The hyh mutation uncovers roles for alpha Snap in apical protein localization and control of neural cell fate Nat Genet 2004 36 264 270 14758363 10.1038/ng1302
Kume T Deng KY Winfrey V Gould DB Walter MA Hogan BLM The forkhead/winged helix gene Mf1 is disrupted in the pleiotropic mouse mutation congenital hydrocephalus Cell 1998 93 985 996 9635428 10.1016/S0092-8674(00)81204-0
Gibbs RA Weinstock GM Metzker ML Muzny DM Sodergren EJ Scherer S Scott G Steffen D Worley KC Burch PE Okwuonu G Hines S Lewis L DeRamo C Delgado O Dugan-Rocha S Miner G Morgan M Hawes A Gill R Celera Holt RA Adams MD Amanatides PG Baden-Tillson H Barnstead M Chin S Evans CA Ferriera S Fosler C Glodek A Gu Z Jennings D Kraft CL Nguyen T Pfannkoch CM Sitter C Sutton GG Venter JC Woodage T Smith D Lee HM Gustafson E Cahill P Kana A Doucette-Stamm L Weinstock K Fechtel K Weiss RB Dunn DM Green ED Blakesley RW Bouffard GG de Jong PJ Osoegawa K Zhu B Marra M Schein J Bosdet I Fjell C Jones S Krzywinski M Mathewson C Siddiqui A Wye N McPherson J Zhao S Fraser CM Shetty J Shatsman S Geer K Chen Y Abramzon S Nierman WC Havlak PH Chen R Durbin KJ Egan A Ren Y Song XZ Li B Liu Y Qin X Cawley S Worley KC Cooney AJ D'Souza LM Martin K Wu JQ Gonzalez-Garay ML Jackson AR Kalafus KJ McLeod MP Milosavljevic A Virk D Volkov A Wheeler DA Zhang Z Bailey JA Eichler EE Tuzun E Birney E Mongin E Ureta-Vidal A Woodwark C Zdobnov E Bork P Suyama M Torrents D Alexandersson M Trask BJ Young JM Huang H Wang H Xing H Daniels S Gietzen D Schmidt J Stevens K Vitt U Wingrove J Camara F Mar AM Abril JF Guigo R Smit A Dubchak I Rubin EM Couronne O Poliakov A Hubner N Ganten D Goesele C Hummel O Kreitler T Lee YA Monti J Schulz H Zimdahl H Himmelbauer H Lehrach H Jacob HJ Bromberg S Gullings-Handley J Jensen-Seaman MI Kwitek AE Lazar J Pasko D Tonellato PJ Twigger S Ponting CP Duarte JM Rice S Goodstadt L Beatson SA Emes RD Winter EE Webber C Brandt P Nyakatura G Adetobi M Chiaromonte F Elnitski L Eswara P Hardison RC Hou M Kolbe D Makova K Miller W Nekrutenko A Riemer C Schwartz S Taylor J Yang S Zhang Y Lindpaintner K Andrews TD Caccamo M Clamp M Clarke L Curwen V Durbin R Eyras E Searle SM Cooper GM Batzoglou S Brudno M Sidow A Stone EA Venter JC Payseur BA Bourque G Lopez-Otin C Puente XS Chakrabarti K Chatterji S Dewey C Pachter L Bray N Yap VB Caspi A Tesler G Pevzner PA Haussler D Roskin KM Baertsch R Clawson H Furey TS Hinrichs AS Karolchik D Kent WJ Rosenbloom KR Trumbower H Weirauch M Cooper DN Stenson PD Ma B Brent M Arumugam M Shteynberg D Copley RR Taylor MS Riethman H Mudunuri U Peterson J Guyer M Felsenfeld A Old S Mockrin S Collins F Genome sequence of the Brown Norway rat yields insights into mammalian evolution Nature 2004 428 493 521 15057822 10.1038/nature02426
Sasaki S Goto H Nagano H Furuya K Omata Y Kanazawa K Suzuki K Sudo K Collmann H Congenital hydrocephalus revealed in the inbred rat LEW/Jms Neurosurgery 1983 13 548 554 6606138
Jones HC Carter BJ Morel L Characteristics of hydrocephalus expression in the LEW/Jms rat strain with inherited disease Childs Nerv Syst 2003 19 11 18 12541080
Jones HC Yehia B Chen GF Carter BJ Genetic analysis of inherited hydrocephalus in a rat model Exp Neurol 2004 190 79 90 15473982 10.1016/j.expneurol.2004.06.019
Jones HC Carter BJ Depelteau JS Roman M Morel L Chromosomal linkage associated with disease severity in the hydrocephalic H-Tx rat Behav Genet 2001 31 101 111 11529267 10.1023/A:1010266110762
Somera KC Jones HC Reduced subcommissural organ glycoprotein immunoreactivity precedes aqueduct closure and ventricular dilatation in H-Tx rat hydrocephalus Cell Tissue Res 2004 315 361 373 14722750 10.1007/s00441-003-0843-9
Jones HC Lopman BA Jones TW Carter BJ Depelteau JS Morel L The expression of inherited hydrocephalus in H-Tx rats Childs Nerv Syst 2000 16 578 584 11048632 10.1007/s003810000330
Lander ES Green P Abramson J Barlow A Daly MJ Lincoln SE Newburg L Abrahamson J MAPMAKER: an interactive computer package for constructing primary genetic linkage maps of experimental and natural populations. Genomics 1987 1 174 181 3692487 10.1016/0888-7543(87)90010-3
Lander ES Kruglyak L Genetic dissection of complex traits: guidelines for interpreting and reporting linkage results Nat Genetics 1995 11 241 247 7581446 10.1038/ng1195-241
Bannister CM Russell SA Rimmer S Arora A Pre-natal ventriculomegaly and hydrocephalus Neurol Res 2000 22 37 42 10672579
Futagi Y Suzuki Y Toribe Y Morimoto K Neurodevelopmental outcome in children with fetal hydrocephalus Pediatr Neurol 2002 27 111 116 12213611 10.1016/S0887-8994(02)00395-8
Renier D Sainte-Rose C Pierre-Kahn A Hirsch JF Prenatal hydrocephalus: outcome and prognosis Childs Nerv Syst 1988 4 213 222 3167875 10.1007/BF00270917
Yamada H Oi S Tamaki N Masumoto S Sudo K Prenatal aqueduct stenosis as a cause of congenital hydrocephalus in the inbred rat LEW/Jms Childs Nerv Syst 1991 7 218 222 1933919 10.1007/BF00249399
Yamada H Oi S Tamaki N Matsumoto S Sudo K Histological changes in the midbrain around the aqueduct in congenital hydrocephalic rat LEW/Jms Childs Nerv Syst 1992 8 394 398 1458497 10.1007/BF00304787
Kohn DF Chinookoswong N Chou SM A new model of congenital hydrocephalus in the rat Acta Neuropathol 1981 54 211 218 7257730 10.1007/BF00687744
Jones HC Bucknall RM Inherited prenatal hydrocephalus in the H-Tx rat: a morphological study Neuropathol Appl Neurobiol 1988 14 263 274 3221976
Jones HC Depelteau JS Carter BJ Somera KC The frequency of inherited hydrocephalus is influenced by intrauterine factors in H-Tx rats Exp Neurol 2002 176 213 220 12093098 10.1006/exnr.2002.7921
Peirce JL Derr R Shendure J Kolata T Silver LM A major influence of sex-specific loci on alcohol preference in C57Bl/6 and DBA/2 inbred mice Mamm Genome 1998 9 942 948 9880657 10.1007/s003359900904
Furuya T Salstrom JL McCall-Vining S Cannon GW Joe B Remmers EF Griffiths MM Wilder RL Genetic dissection of a rat model for rheumatoid arthritis: significant gender influences on autosomal modifier loci Hum Mol Genet 2000 9 2241 2250 11001927
Clark JS Jeffs B Davidson AO Lee WK Anderson NH Bihoreau MT Brosnan MJ Devlin AM Kelman AW Lindpaintner K Dominiczak AF Quantitative trait loci in genetically hypertensive rats. Possible sex specificity Hypertension 1996 28 898 906 8901842
Sapiro R Kostetskii I Olds-Clarke P Gerton GL Radice GL Strauss III JF Male infertility, impaired sperm motility, and hydrocephalus in mice deficient in sperm-associated antigen 6 Mol Cell Biol 2002 22 6298 6305 12167721 10.1128/MCB.22.17.6298-6305.2002
Daniel GB Edwards DF Harvey RC Kabalka GW Communicating hydrocephalus in dogs with congenital ciliary dysfunction Dev Neurosci 1995 17 230 235 8575342
Ibanez-Tallon I Gorokhova S Heintz N Loss of function of axonemal dynein Mdnah5 causes primary ciliary dyskinesia and hydrocephalus Hum Mol Genet 2002 11 715 721 11912187 10.1093/hmg/11.6.715
Einheber S Pierce JP Chow D Znamensky V Schnapp LM Milner TA Dentate hilar mossy cells and somatostatin-containing neurons are immunoreactive for the alpha8 integrin subunit: characterization in normal and kainic acid-treated rats Neuroscience 2001 105 619 638 11516828 10.1016/S0306-4522(01)00205-6
Stagaard M Mollgard K The developing neuroepithelium in human embryonic and fetal brain studied with vimentin-immunocytochemistry Anat Embryol (Berl) 1989 180 17 28 2476946 10.1007/BF00321896
Karim SA Johnson KJ Griffiths IR Vouyiouklis DA A physical map of the genomic region on mouse chromosome 3 containing the hindshaker (hsh) mutation Genomics 2004 83 225 230 14706451 10.1016/j.ygeno.2003.08.014
Scotland P Zhou D Benveniste H Bennett V Nervous system defects of AnkyrinB (-/-) mice suggest functional overlap between the cell adhesion molecule L1 and 440-kD AnkyrinB in premyelinated axons J Cell Biol 1998 143 1305 1315 9832558 10.1083/jcb.143.5.1305
Kagawa T Oba A Okumura S Ikenaka K Localization of mRNA for UDP-galactose: ceramide galactosyltransferase in the brain during mouse development Dev Neurosci 1996 18 309 318 8911769
Shinoda M Hidaka M Lindqvist E Soderstrom S Matsumae M Oi S Sato O Tsugane R Ebendal T Olson L NGF, NT-3 and Trk C mRNAs, but not TrkA mRNA, are upregulated in the paraventricular structures in experimental hydrocephalus Childs Nerv Syst 2001 17 704 712 11862435 10.1007/s00381-001-0515-6
Miyajima M Sato K Arai H Choline acetyl transferase, nerve growth factor and cytokine levels are changed in congenitally hydrocephalic HTX rats Pediatr Neurosurg 1996 24 1 4 8817609
Hochhaus F Koehne P Schaper C Butenandt O Felderhoff-Mueser U Ring-Mrozik E Obladen M Buhrer C Elevated nerve growth factor and neurotrophin-3 levels in cerebrospinal fluid of children with hydrocephalus BMC Pediatr 2001 1 2 11580868 10.1186/1471-2431-1-2
Kadokawa Y Marunouchi T Chimeric analysis of Notch2 function: a role for Notch2 in the development of the roof plate of the mouse brain Dev Dyn 2002 225 126 134 12242712 10.1002/dvdy.10140
Danielian PS McMahon AP Engrailed-1 as a target of the Wnt-1 signaling pathway in vertebrate midbrain development Nature 1996 383 332 334 8848044 10.1038/383332a0
Parkkila S Parkkila AK Rajaniemi H Shah GN Grubb JH Waheed A Sly WS Expression of membrane-associated carbonic anhydrase XIV on neurons and axons in mouse and human brain Proc Natl Acad Sci U S A 2001 98 1918 1923 11172051 10.1073/pnas.98.4.1918
Speake T Whitwell C Kajita H Majid A Brown PD Mechanisms of CSF secretion by the choroid plexus Microsc Res Tech 2001 52 49 59 11135448 10.1002/1097-0029(20010101)52:1<49::AID-JEMT7>3.0.CO;2-C
Flint J Mott R Finding the molecular basis of quantitative traits: successes and pitfalls Nat Rev Genet 2001 2 437 445 11389460 10.1038/35076585
|
15953386
|
PMC1185556
|
CC BY
|
2021-01-04 16:37:38
|
no
|
Cerebrospinal Fluid Res. 2005 Jun 12; 2:2
|
utf-8
|
Cerebrospinal Fluid Res
| 2,005 |
10.1186/1743-8454-2-2
|
oa_comm
|
==== Front
Chiropr OsteopatChiropractic & Osteopathy1746-1340BioMed Central London 1746-1340-13-151607899910.1186/1746-1340-13-15Case ReportCervical stenosis in a professional rugby league football player: a case report Pollard Henry [email protected] Lotte [email protected] Wayne [email protected] Department of Health and Chiropractic, Macquarie Injury Management Group, Macquarie University, 2109, Sydney Australia2 Lotte Hansen Chiropractic, 70 Donald Street, Hamilton NSW 2303 Australia3 Department of Health and Chiropractic, Macquarie Injury Management Group, Macquarie University, 2109, Sydney Australia2005 3 8 2005 13 15 15 11 4 2005 3 8 2005 Copyright © 2005 Pollard et al; licensee BioMed Central Ltd.2005Pollard 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
This paper describes a case of C7 radiculopathy in a professional rugby league player after repeated cervical spine trauma. The report outlines the management of the patient following an acute cervical hyperflexion injury with chiropractic manipulation and soft tissue therapies. It also presents a change in approach to include distractive techniques on presentation of a neurological deficit following re-injury. The clinical outcomes, while good, were very dependent upon the athlete restricting himself from further trauma during games, which is a challenge for a professional athlete.
Case presentation
A 30-year old male front row Australian rugby league player presented complaining of neck pain after a hyperflexion and compressive injury during a game. Repeated trauma over a four month period resulted in radicular pain. Radiographs revealed decreased disc height at the C5-C6 and C6-C7 levels and mild calcification within the anterior longitudinal ligament at the C6-C7 level. MRI revealed a right postero-lateral disc protrusion at the C6-C7 level causing a C7 nerve root compression.
Conclusion
Recommendations from the available literature at the present time suggest that conservative management of cervical discogenic pain and disc protrusion, including chiropractic manipulation and ancillary therapies, can be successful in the absence of progressive neurological deficit. The current case highlights the initial successful management of a football athlete, and the later unsuccessful management. This case highlights the issues involvement in the management of a collision sport athlete with a serious neck injury.
manipulation, chiropracticnon operative treatmentcervical, radiculopathysport, injuryrugby league
==== Body
Background
Effective cervical injury management requires an understanding of the pathomechanics of injury [1]. There are numerous reports in the literature on cervical injuries in full contact collision sports such as rugby (United Kingdom), rugby league/union (Australia) and American football (United States) [2-7]. However, there is a lack of literature on treatment, particularly non-operative manual therapy. This case report proposes an approach to address discogenic pain and disc protrusion. Injury management of a professional athlete is a challenge given the pressure on the player to play each match and perform through the competitive season.
Case Presentation
Case Report
A 30 year old professional Australian rugby league player presented complaining of diffuse neck pain following a hyperflexion and compressive neck injury while being tackled with his neck in maximal flexion. The injury occurred six days prior to presentation. There had been no significant previous history of injury or concussions over a 15 year career.
On examination, pain was reproduced in the right trapezius area on forward flexion and neurological testing was normal. There was no upper limb pain present. Other testing, including vertebral artery and blood pressure testing, proved unremarkable. Treatment consisted of high velocity, low amplitude manipulative therapy directed to the C2, C5 and T1 levels and proprioceptive neuromuscular facilitation (PNF) stretching to the trapezius and cervical musculature. This resulted in full pain free cervical ROM in three treatments.
Over the next four months the patient presented on five more occasions for similar cervical injuries following hyperflexion and compressive injuries while being tackled in matches. Symptoms progressed to include frontal headaches and cervico-thoracic pain. Treatment continued to include manipulation to the cervical and thoracic spine, PNF, soft tissue massage and trigger point therapy to the trapezius, sterno-cleido mastoid (SCM) and suboccipital musculature. This resulted in symptomatic relief within two to three treatments at each presentation. Over the four month period blood pressure and upper limb neurological status was monitored which proved to be unremarkable.
Four months after the initial presentation, the patient presented again following a combined hyperflexion with left lateral rotation injury after a collision in a game. Severe pain was immediately experienced over the right lower neck with pain radiating into the right upper trapezius region. Cervical radiographic images were ordered, including lateral, anterio-posterior, antero-posterior open mouth, oblique and functional views in flexion and extension. These views revealed decreased disc height at the C5-C6 and C6-C7 levels with mild osteophytic lipping. No definite instability was demonstrated. On the oblique view the intervertebral foramina appeared normal throughout the cervical spine.
A working diagnosis of an acute C6 disc lesion was established and the patient was advised to consult the team physician for advanced imaging to determine the state of any disc disruption. He continued to receive two treatments per week to manage his symptoms. Treatment was directed toward the management of the acute symptoms with ice, soft tissue therapy and manipulation to sites around (but not on) the lesion. The patient was also advised to commence a short course of non-steroidal anti-inflammatory medication (NSAIDs). Practitioners should be mindful of their recommendations as all medications must be deemed appropriate according to the governing sport's policy on sports doping.
After each game for a two month period symptoms progressively worsened until he presented with severe constant pain in the right upper trapezius region. This time the right triceps muscle strength was significantly reduced and the pain was relieved by elevating the right arm over his head and by cervical traction, indicating possible neural impingement. Other upper limb neurological testing, including upper limb deep tendon reflexes and sensory testing, was negative.
The clinical diagnosis was altered to a C6 radiculopathy. Treatment was altered to intermittent cervical traction and pulsed ultrasound therapy to the right trapezius area twice per week. Home advice included the use of an ice pack (20 minutes on the hour) and intermittent cervical traction three times a day. NSAID use was advised.
Two weeks later the patient presented with cervico-thoracic pain after being hit on the left side of his head while being tackled. Cervical ROM was full but painful in right rotation, flexion and extension. Upper limb neurological testing was unremarkable. A non-specific soft tissue sprain/strain diagnosis was given. Treatment involving cervical and thoracic manipulation and massage was again delivered with symptomatic relief. He was advised not to play rugby league until his symptoms had settled.
The patient continued to play and one month later began experiencing numbness in the second and third digits of his right hand. The Door Bell Sign over the right lower cervical area (pressure over the IVF at the antero-lateral aspect of the neck) reproduced the radicular symptoms. Other upper limb neurological testing was normal. The patient continued to receive intermittent cervical traction, gentle mobilisation of the cervical spine and manipulation to the upper thoracic area, along with massage therapy twice per week.
Four days later team medical staff referred the patient for magnetic resonance imaging (MRI). The MRI revealed a right postero-lateral disc protrusion at the C6-C7 level. The patient was strenuously advised not to play for three weeks, or to do upper body weight training. Two weeks later after following this advice the patient had regained full cervical ROM and pain was only reproduced on passive flexion and extension. He was still experiencing numbness in the second and third digits of the right hand. He began playing games again before finishing the season three weeks later.
Follow up two months post season showed no pain or restriction to cervical ROM. He was still experiencing some numbness in the second digit of the right hand, and the right triceps muscle was slightly weak. Whilst a new symptom, the mild nature of the tricep weakness was not considered serious, but it was to be monitored closely for any signs of deterioration. The player was informed of this approach and consented to it. Other neurological and orthopaedic tests were negative (shoulder and elbow range of motion, and Phalen's test). The athlete was advised to gradually build up his weight training and to maintain cervical flexibility with strengthening and stretching exercises.
The player was still reluctant to stop football and has since been advised by his orthopaedic surgeon to have discectomy surgery and to fuse the C6-C7 level, requiring six months of rehabilitation before returning to play. The player did not have the surgery and rested until the following year where he began the process all over again. He is now retired from football.
Literature Review
This case outlines a series of cervical traumas producing neck, arm and head pain. The series of injuries involved forced flexion, compression and lateral deviation away from the painful side. This mechanism is in contrast to the mechanism of extension with lateral deviation towards the painful side as described in the majority of studies of neck injuries in American football and rugby [2,5,6], The clinical signs suggest a disc herniation following repeated trauma resulting in compression of the C7 nerve root.
There are several studies reporting chronic recurrent cervical nerve root neuropraxia (sometimes called "chronic burner syndrome"), in American football [5,6] and in rugby players [2]. This can commonly occur during blocking, tackling or engaging in a scrum. Chronic burner syndrome can be defined as:
1) a chronic recurrent neuropraxia or axonotmesis, or both, of a nerve root associated with prolonged weakness,
2) time loss from practice and games, and
3) recurrence [6].
Nerve root compression in the intervertebral foramina secondary to disc herniation or degenerative changes, or both, is the most common cause in football players seen with recurrent or chronic burners [6]. In such cases, degenerative changes frequently present with concurrent cervical canal stenosis and can predispose injury [8].
A correlation seems to exist between chronic recurrent cervical nerve root neurapraxia and cervical canal stenosis in tackled football players [1,5-7] and risk of more serious cervical spine injury increases with increasing stenosis [9]. A spinal canal-vertebral body ratio (Pavlov's ratio) on lateral radiographs of 0.80 or less (normal ratio 1:1) at one or more levels has been found in a tackle football population who have experienced an episode of cervical cord neuropraxia manifested by sensory and/or motor symptoms [10]. Despite a series of minor neurological insults, no correlation between the prodromal episodes of cord neuropraxia and occurrence of permanent quadriplegia has been found [10]. Also, the presence of uncomplicated developmental narrowing of the stable cervical spine does not predispose permanent neurological injury [1].
Absolute contraindications to continued participation in contact sports has been recommended to apply to those individuals who have had a documented episode of cervical cord neurapraxia associated with the following:
• ligamentous instability,
• intervertebral disc disease with cord compression,
• significant degenerative changes,
• MRI evidence of cord defects or swelling,
• positive neurological findings lasting more than 36 hours,
• more than one recurrence [10].
The extremely low predictive value of Pavlov's ratio (as an indicator of clinically relevant spinal stenosis) precludes its use as a screening mechanism for determining participation in contact activities [7]. To accurately assess spinal canal stenosis, cross-sectional imaging technology such as MRI, contrast positive CT, and myelography should be employed [7]. Plain radiographic identification of a narrow spinal canal in a player sustaining cervical cord neuropraxia warrants MRI investigation to rule out soft tissue based stenosis [11].
Mechanism of Injury
Most of the literature on cervical spine injuries in football, such as burner syndrome, emphasises an extension type mechanism of injury. In our case, the mechanism of injury involved both hyperflexion and a compressive force. As hyperflexion involves more compressive load to the cervical spine than extension, this combination has a greater potential for injury, particularly if a stenosis situation concurrently exists [12].
With cervical hyperflexion, the spinolaminar line of the superior vertebra and the posterior superior aspect of the vertebral body below approximate, resulting in a rapid decrease of the spinal canal with compression of the spinal cord [1]. The brief, sudden deformation of the cord is thought to produce disturbed sensory and motor function below the involved level [1,13]. In most instances of acute spinal injury, disruption of cord function is the result of local cord anoxia and increased concentration of intracellular calcium [1]. Playing with improper technique, such as spear tackling, has been associated with catastrophic injuries [14]. In the case presented in this report, the technique of running at a tackler with neck hyperflexion before impact contributed to the repetitive history of injury and should have been corrected.
Hyperflexion injuries in Whiplash-Associated-Disorders (WAD) do not involve the exact same mechanism of injury (i.e. absence of axial compression) but the soft tissue damage can be very similar [15]. For example, Grade III WAD features include: cervical herniated disc, cervicalgia with headaches and limited range of motion combined with neurologic symptoms and signs are present [15].
With compression, a force exerted through the crown of the head can be transmitted through the skull to the cervical vertebrae resulting in the crushing of the vertebrae and extrusion of the vertebral body and disc material posteriorly into the cervical vertebral canal [3]. When the cervical spine is in hyperflexion with rotation, vertebral dislocation without fracture is possible, which is more likely if the head is locked on the ground adding a compressive force [15]. The most damaging mechanisms of injury to the spine are torsional and combined motions (i.e. forward flexion and lateral rotation) with a combined axial load [17,18].
Management
Most of the literature involving treatment of patients with cervical disc herniations producing neurologic loss reports surgical outcomes. Furthermore, reports of patients with cervical spondylotic myelopathy show the symptoms progress gradually or step-wise and recovery of neurological function after conservative treatment or decompression surgery is common [19-21]. There is a lack of literature on manual therapy for an acute presentation of chronic recurrent cervical nerve root neuropraxia.
The treatment protocol in the acute phase should involve rest, ice and intermittent traction [22]. Case reports have shown successful chiropractic management of radicular and pseudoradicular pain through high-velocity, low amplitude thrust techniques and associated soft tissue therapies [23]. Manipulative therapy to the spinal segments above and below the painful levels, after the appropriate pre-manipulative provocation testing for verebrobasilar insufficiency has been performed, is believed to aid in increasing fluid infusion, and decreasing nociceptive input to the spinal cord [24]. The chiropractic literature suggests, in cases of cervical disc herniation, that manipulative therapy of the involved level is delivered only in the sub-acute phase, with the line of thrust being only in the pain-free directions [22].
The practitioner should be mindful of the potential for iatrogenic joint instability to occur. Damage to the supporting structures resulting in hypermobile joints can be aggravated by and result from repeated manipulations [25]. The recommended management protocol for Grade III WAD, which is a similar injury, is shown in Table 1 and could be viewed as a guideline for management of footballers with cervical stenosis [15].
Table 1 Management of Grade III Whiplash Associated Disorders (WAD) [15]
• Soft collars should not be used as they do not adequately immobilise the spine
• Narcotic analgesics may occasionally be needed for pain relief. Psychopharmacologic drugs and muscle relaxants should not be used
• In the few cases in which rest for the neck might be indicated, it should be limited to less than four days and followed by early activation
• The Task Force consensus is that manipulative treatments by trained persons for the relief of pain and facilitating early mobility can be used in WAD
• Long term repeated manipulation or physiotherapy without multidisciplinary evaluation is not justified
• Surgery for WAD patients is rarely indicated. Surgery is only indicated for Grade III patients with progressive neurologic deficit or persisting arm pain
The role of surgical versus non-surgical treatment of patients with cervical disc herniation has been compared in a longitudinal cohort study [26]. Twenty-six subjects with a clearly defined diagnosis of cervical herniated nucleus pulposus on MRI were evaluated for outcome with conservative treatment. This included ice, rest, hard cervical collar, NSAIDs for six to 12 weeks, manual and mechanical traction followed by home traction, and progressive strengthening exercises of the shoulder girdle and chest with training in postural control and body mechanics training. Follow up for over a year showed that 24 of the patients were successfully treated without surgery. Twenty patients achieved good or excellent results determined by symptom level, activity and function level, medication and ongoing medical care, job status and satisfaction. Only two patients required surgery suggesting that many cervical disc herniations can be successfully managed with conservative treatment.
Low level evidence from the available literature suggests that conservative management of discogenic pain and disc protrusion can be as successful as surgical treatment. However, such approaches, including corticosteroid injections [28], need to be validated by higher level evidence such as a randomised controlled trial [26,27]. Further, the current clinical trials have not included an athletic population.
It has been suggested that individuals with disc herniations will require lifelong management to ensure long-term participation in sport [26,27]. However, at present there is little evidence to support or refute such an approach. Based on a study of American football players the current recommendations for an athlete with intervertebral disc injury participating in a contact sport is as follows:
1. No contraindication to participate in contact sport in individuals with a healed disc herniation treated conservatively,
2. Relative contraindication in individuals with facet instability,
3. Absolute contraindication in individuals with:
a. Acute disc herniation (with or without neurological findings)
b. Acute or chronic disc herniation with decreased cervical range of motion,
c. Acute or chronic disc herniation with signs and symptoms of cord neuropraxia due to congenital stenosis of the cervical canal [4].
According to these recommendations the patient described in this case report would belong in group 3a) and should therefore not have played until his symptoms had subsided. With professional athletes a player can be reluctant to report injuries for fear of losing a spot in the team or losing wages. Despite not believing it is safe to play with injuries, many athletes are willing to risk doing so [29]. The authors of this study further concluded that: individuals with uncomplicated cervical cord neuropraxia can return to play without risking further damage, and various clinical manifestations are not related to the radiological findings.
The current literature agrees that the fundamental requirements for return to contact sports to include: normal strength, painless range of motion, a stable vertebral column and adequate space for the neurological elements [14]. Return to play guidelines need to be further investigated as several approaches exist and are open to interpretation. Some approaches allow a more liberal return to play criteria and would more likely be used with professional sports persons. Other return to play criteria are less liberal and would be used for amateurs and junior players
A one level surgical fusion has been suggested not to present any contraindication to participation in contact activities provided the athlete is completely asymptomatic and neurologically normal when commencing sport [4]. In a prospective study it was found that most cervical disc herniations regress with time and without the need for surgical resection [30]. Patients were finally examined and discharged from care because of sustained pain control at an average of six months. These findings are similar to four cases reporting spontaneous resolution of cervical herniation [31]. Given that surgery will require six months of rest from contact sport and a rehabilitation program, one may question whether the same outcomes with aggressive conservative therapy would have similar if not better results. Guidelines for surgical cases need to be clearly defined and randomised controlled trials comparing surgical to conservative care performed.
A review of the management of lumbar intervertebral disc injuries in athletes suggests that the high recurrence rate of low back pain may indicate that the resolution of symptoms is accompanied by restoration of function [27]. This is possibly due to functional changes that occur with injury, which can be assumed to occur with cervical disc injuries. Cervical dysfunction can be caused by failure to rehabilitate previous injuries [32,33], a scenario commonly encountered in the professional athlete who frequently returns to play sooner than the ideal. This becomes important for an athlete as performance can suffer in the absence of pain but in the presence of subtle biomechanical maladaptations [27].
Conclusion
This case report has outlined the progression of cervical injury to a disc protrusion resulting in a C7 radiculopathy in a professional rugby league player, due to numerous blows to the cervical spine after a series of hyperflexion injuries. The patient ultimately suffered a severe forced flexion combined with left lateral flexion injury to the cervical spine and experienced sensory and motor changes in the right C7 nerve root distribution. When it became apparent that there was intervertebral foramen encroachment secondary to a disc protrusion the treatment protocol changed toward a more conservative approach.
It is important that the athlete is informed of the problem with particular regard to potential risks if they continue to play. When considering surgery, the long-term consequences of the intervention should be thoroughly discussed with the athlete along with all potential management options including no treatment and retirement from the sport. Further research is needed on the chiropractic management of acute athletic injuries to the spine and the long-term outcome for surgery versus conservative management of patients wishing to continue their athletic career.
Authors' contributions
HP conceived of the study, participated in its design and helped to draft and edit the manuscript.
LH provided treatment to the subject and helped draft the manuscript
WH helped to collect literature and draft the manuscript.
All authors read and approved the manuscript.
==== Refs
Torg JS Thibault L Sennett B Pavlov H The pathomechanics and pathophysiology of cervical spinal cord injury Cl Orth Rel Res 1995 321 259 269
Wetzler MJ Akpata T Albert T Foster TE Levy AS A retrospective study of cervical spine injuries in American rugby, 1970 to 1994 Am J Sports Med 1996 24 454 458 8827303
Silver JR Stewart D The prevention of spinal injuries in rugby football Paraplegia 1994 32 442 453 7970845
Torg JS Ramsey-Emrhein JA Suggested management guidelines for participation in collision activities with congenital, developmental, or postinjury lesions involving the cervical spine Med Sci Sports Exerc 1997 29 S256 S272 9247923
Markey KL Denedetto MD Curl WW Upper trunk brachial plexopathy Am J Sports Med 1993 21 650 655 8238703
Levitz CL Reilly PJ Torg JS The pathomechanics of chronic, recurrent cervical nerve root neurapraxia Am J Sports Med 1997 25 73 76 9006696
Cantu RC Stingers, transient quadriplegia, and cervical spinal stenosis: Return to play criteria Med Sci Sports Exerc 1997 29 S233 S235 9247920
Weinberg J Rokito S Silber JS Etiology, treatment, and prevention of athletic "stingers" Clin Sports Med 2003 22 493 500 12852682 10.1016/S0278-5919(02)00057-1
Castro FP Jr Stingers, cervical cord neurapraxia, and stenosis Clin Sports Med 2003 22 483 92 12852681 10.1016/S0278-5919(02)00094-7
Torg JS Cervical spinal stenosis with cord neurapraxia and transient quadriplegia Sports Med 1995 20 429 434 8614762
Kim DH Vaccaro AR Berta SC Acute sports-related spinal cord injury: contemporary management principles Clin Sports Med 2003 22 501 12 12852683 10.1016/S0278-5919(02)00105-9
Bhatoe HS Cervical spinal cord injury without radiological abnormality in adults Neurol India 2000 48 243 8 11025628
Breslow MJ Rosen JE Cervical spine injuries in football Bull Hosp Jt Dis 2000 59 201 10 11409239
Morganti C Recommendations for return to sports following cervical spine injuries Sports Med 2003 33 563 73 12797838
Spitzer WO Skovron ML Salmi LR Cassidy JD Duranceau J Suissa S Zeiss E Scientific monograph of the Quebec Task Force on Whiplash Associated Disorders: Redefining "Whiplash" and its management Spine 1995 21S 38S
Bauze RJ Ardran GM Experimental production of forward dislocation in the human cervical spine J Bone Joint Surg Br 1978 60 239 245 659473
Quarrie KL Cantu RC Chalmers DJ Rugby union injuries to the cervical spine and spinal cord Sports Med 2002 32 633 53 12141883
Torg JS Guille JT Jaffe S Injuries to the cervical spine in American football players J Bone Joint Surg Am 2002 84 112 122 11792789
Ito T Oyanagi K Takahashi H Takahashi HE Ikuta F Cervical spondylotic myelopathy Spine 1996 21 827 833 8779013 10.1097/00007632-199604010-00010
Geck MJ Eismont FJ Surgical options for the treatment of cervical spondylotic myelopathy Orth Cl Nth Am 2002 33 329 48 10.1016/S0030-5898(02)00002-0
Emery SE Cervical spondylotic myelopathy: diagnosis and treatment J Am Acad Ortho Surg 2001 9 376 88
Hubka MJ Phelan SP Delaney PM Robertson VL Rotary manipulation for cervical radiculopathy: observations on the importance of the direction of the thrust J Manipulative Physiol Ther 1997 20 622 7 9436148
Pollard H Tuchin P Cervical radiculopathy: A case for ancillary therapies? J Manipulative Physiol Ther 1995 18 244 249 7636415
Patterson MM The spinal cord: Participant in disorder Spinal Manipulation 1993 9 2 11
Hurwitz EL Aker PD Adams AH Meeker WC Shekelle PG Manipulation and mobilisation of the cervical spine: A systematic review of the literature Spine 1996 21 1746 1758 8855459 10.1097/00007632-199608010-00007
Saal JS Saal JA Yurth EF Nonoperative management of herniated cervical intervertebral disc with radiculopathy Spine 1996 21 1877 1883 8875719 10.1097/00007632-199608150-00008
Young JL Press JM Herring SA The disc at risk in athletes: Perspectives on operative and nonoperative care Med Sci Sports Exerc 1997 29 S222 S232 9247919
Slipman CW Chow DW Therapeutic spinal corticosteroid injections for the management of radiculopathies Phys Med Rehabil Clin N Am 2002 13 697 711 12380554
Finch C Donohue S Garnham A Safety attitudes and beliefs of junior Australian football players Inj Prev 2002 8 151 4 12120836 10.1136/ip.8.2.151
Bush K Chaudhuri R Hillier S Penny J The pathomorphologic changes that accompany the resolution of cervical radiculopathy Spine 1997 22 183 187 9122798 10.1097/00007632-199701150-00009
Vinas FC Wilner H Rengachary S The spontaneous resorption of herniated cervical discs J Clin Neurosci 2001 8 542 6 11683601 10.1054/jocn.2000.0894
Herring SA Weinstein SM Nicholas JA, Hershman EB Assessment and nonsurgical management of athletic low back injury The Lower Extremity & Spine in Sports Medicine 1995 2 St. Louis: Mosby-Year Book 1171 1197
Kibler WB Clinical aspects of muscle injury Med Sci Sports Exerc 1990 22 450 2402204
|
16078999
|
PMC1185557
|
CC BY
|
2021-01-04 16:38:23
|
no
|
Chiropr Osteopat. 2005 Aug 3; 13:15
|
utf-8
|
Chiropr Osteopat
| 2,005 |
10.1186/1746-1340-13-15
|
oa_comm
|
==== Front
Chiropr OsteopatChiropractic & Osteopathy1746-1340BioMed Central London 1746-1340-13-91600017510.1186/1746-1340-13-9DebateChiropractic as spine care: a model for the profession Nelson Craig F [email protected] Dana J [email protected] John J [email protected] Gert [email protected] Stephen M [email protected] R Douglas [email protected] Kurt [email protected] Thomas [email protected] American Specialty Health 777 Front St. San Diego, CA 92101, USA2 Palmer Centre for Chiropractic Research, Palmer College of Chisopractic, 1000 Brady Street Davenport, IA 52803, USA3 Texas Back Institute 6020 W. Parker Road Plano, TX 75093, USA4 Northwestern Health Sciences University 2501 W. 84th St. Bloomington, MN 55431, USA5 University of Bridgeport 126 Park Avenue Bridgeport, CT 06604, USA2005 6 7 2005 13 9 9 20 5 2005 6 7 2005 Copyright © 2005 Nelson et al; licensee BioMed Central Ltd.2005Nelson 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
More than 100 years after its inception the chiropractic profession has failed to define itself in a way that is understandable, credible and scientifically coherent. This failure has prevented the profession from establishing its cultural authority over any specific domain of health care.
Objective
To present a model for the chiropractic profession to establish cultural authority and increase market share of the public seeking chiropractic care.
Discussion
The continued failure by the chiropractic profession to remedy this state of affairs will pose a distinct threat to the future viability of the profession. Three specific characteristics of the profession are identified as impediments to the creation of a credible definition of chiropractic: Departures from accepted standards of professional ethics; reliance upon obsolete principles of chiropractic philosophy; and the promotion of chiropractors as primary care providers. A chiropractic professional identity should be based on spinal care as the defining clinical purpose of chiropractic, chiropractic as an integrated part of the healthcare mainstream, the rigorous implementation of accepted standards of professional ethics, chiropractors as portal-of-entry providers, the acceptance and promotion of evidence-based health care, and a conservative clinical approach.
Conclusion
This paper presents the spine care model as a means of developing chiropractic cultural authority and relevancy. The model is based on principles that would help integrate chiropractic care into the mainstream delivery system while still retaining self-identity for the profession.
ChiropracticEvidence-Based Health CareHealth Care ProfessionsProfessional Ethics
==== Body
Background
It is always fashionable to speak of an issue or controversy as reaching a "crisis point," or of an organization or profession reaching a "crossroads" in its development. However such exhortations are often merely hyperbole. At the risk of committing this offense, we believe that the chiropractic profession today faces an exceptionally difficult set of challenges and, yes, a crisis. The nature of this crisis is the profession's continued inability to define itself. The chiropractic profession, more than 100 years after its founding, does not project a definition of itself that is consistent, coherent or defensible. The healthcare system is increasingly intolerant of such ambiguity and uncertainty; an intolerance which will only intensify in the future.
The primary purpose of this paper is to offer a coherent and defensible professional identity. We argue that chiropractic's identity is as a provider of spine care. We argue further that such a model is consistent with the best available scientific evidence, is consistent with the current public perception, provides benefit to both the profession and the public, and is capable of gaining for the profession the cultural authority it now lacks. In developing this model we established a set of criteria that the model must meet:
1. It must be consistent with accepted modes of scientific reasoning and knowledge.
2. It must accommodate future changes in scientific understanding.
3. It must represent a set of clinical competencies within the reach of practicing chiropractors.
4. It must be consistent, credible and communicable to external constituencies on whom the profession relies.
5. It must represent the evidence of practice experience.
6. It must find a substantial presence within the healthcare marketplace.
7. It must be compatible with the training, licensure, history and heritage of chiropractic.
Part I: The Context of the Identity Crisis
The Search for Cultural Authority
All healthcare disciplines have members who quibble over priorities and preferred belief systems. To prevent these squabbles from limiting advancement and productivity, there must be an understanding of common ground on which to build. With that in mind, it helps to ask "What are the core values/concerns held by the members of the chiropractic profession on which nearly all parties can agree?" We propose that there are a number of common factors even among the most diverse viewpoints within chiropractic.
• Patients benefit from chiropractic care.
• Over the past several decades, a substantial body of evidence has accumulated to inform decision-making on the value of chiropractic manipulation for low back, neck and headache complaints.
• A large population exists that is underserved by chiropractic.
• Extra-disciplinary competition is increasing, with greater encroachment on traditionally chiropractic domains.
• Significant barriers persist which obstruct the profession and its members from reaching their group and individual potentials.
With this common understanding we can ask, "Why is the modern evidence largely being ignored by policy makers and the access to chiropractic care being impeded by arbitrary obstacles?" To answer this question, we should step back and take a dispassionate assessment of how society invests its trust in professionals. The trivial answer identifies institutional bias as the cause; that is, policy makers rely solely on practitioners of medicine as their advisors. Although there is evidence that these attitudes are easing, stereotyping and bias toward the chiropractic profession remains pervasive. However, this is a superficial and inadequate explanation, as the sovereignty of medicine over healthcare has eroded significantly and its biases are increasingly evident to decision makers.
The more complete answer is based on the competition for cultural authority that each profession faces during its evolution. Cultural authority is granted by society based on recognition of a professional group's competency and legitimacy with respect to the domain over which it professes dominance. With cultural authority comes a certain degree of autonomy and privilege. Chiropractic has not anchored its cultural authority. Evidence of competency exists by virtue of years of practical experience and the presence of substantial evidence of effectiveness for methods of care for which the profession has held as its primary domain for the majority of the 20th Century. It is on the front of legitimacy that we have failed. This failure is fueled by a mismatch between the profession's assessment of the value the practice of chiropractic offers and society's assessments of the same. Some chiropractors lament that the profession has done a poor job of educating the public about chiropractic. They posit that if we would just do enough advertising and more effective public relations, the resistance to using chiropractic services would decrease. As enticing as the argument sounds, that experiment actually has been done and has proven not only to be false but counterproductive. Canadian chiropractors found, in two separate samples, that marketing to the public about subluxation and the adjustment resulted in a backlash against the term "subluxation" and an increase in the public's desire to consult a medical doctor if they perceived they might have a subluxation [1]. The educational materials about chiropractic ideology were created by advertising professionals and broadcast under supervision of the chiropractors. The public is clearly not interested in, or receptive to this sort of message from the chiropractic profession.
Legitimacy, as defined above, is the active battleground today. Points of contention are the credibility of clinical claims for effectiveness of chiropractic manipulation for a variety of non-spinal conditions, cost of chiropractic care versus "standard care," and the presence of real or perceived unethical practices. Certainly, there is room to argue about most of these points. The profession is further encumbered by questionable institutionalized practices. For example, some practice consultants promote the policy of withholding administration of treatment on the first visit, preferring to reschedule the patient for a report of findings on a subsequent visit. Where is the clinical rationale for such practice? Are these doctors insufficiently skilled in interpreting the history and examination findings for a routine first visit without time to confer and study? Others promote the use of x-rays on nearly every patient in order to determine biomechanical deviations from a theoretical "model" of a normal spine implying that this information is so essential to successful treatment that the benefit outweighs the very real risk of radiation exposure [2]. These and other business practices promoted across the profession are tolerated without challenge by the rank and file. These practices degrade the credibility of the profession and its members as competent clinicians and diminish the public's trust and level of cultural authority. Considering these various threats to professional legitimacy, a new model is needed. Such a model will provide the chiropractic profession with common core values that permit the development and expansion of chiropractic as future evidence arises. A significant component of this new model must take into account accepted concepts of professional ethics.
Professional Ethics and Chiropractic Identity
This discussion occurs within the context of chiropractic as a licensed healthcare profession. The status of "licensed healthcare profession" confers upon the chiropractic profession certain privileges, but it also imposes upon it a specific set of expectations and ethical obligations. Professional ethics differ from the ethics of mercantilism. For the customer, the relationship with a merchant has always been governed by the dictum caveat emptor or, let the buyer beware. Mercantilism demands that, for the merchant, pecuniary interests supersede others. Despite the fact that a chiropractic practice is typically a commercial, for-profit enterprise, the chiropractor is not governed by the dictates of mercantilism but rather by professionalism. Professions are so-called because they "profess" to have knowledge and skills beyond the comprehension of the laity. The theory of professionalism is predicated on this asymmetry of knowledge. Classically, the only professions were medicine, law, and the clergy, to which modern disciplines can be added, such as engineers, financial planners, etc. Hughes coined the expression credat emptor, let the buyer have faith, to describe the special relationship professionals have with their patient, client or parishioner [3]. Thus, chiropractors, as health professionals, are expected to make recommendations that are in the best interest of the patient, superseding the doctor's pecuniary interests.
As a result of patients' ignorance concerning the specialized knowledge of the professional, the faith a patient places in his or her doctor must extend to the information they are given by their doctor. The imbalance in knowledge means that the doctor not only must not lie to a patient (the ethical duty of veracity) but also must take pains to ensure that what they tell the patient is the truth (the ethical duty of fidelity), as best as it can be known by the doctor and understood by the patient.
At first glance, avoiding a lie and telling the truth may appear to be synonymous but they are not. If one honestly believes a piece of information told to another, then one is not lying. However, if that information is in fact not valid, one has not lied but has told an untruth. Thus, the person has erroneously transmitted incorrect information. Transmission of false information, if correct information is reasonably available to the profession, is a violation of one's duty of fidelity. The duty of fidelity is, in part, to comply with the reasonable expectations of the patient including the expectation that information given is in fact valid.
The ethics of professionalism require not only veracity, but also fidelity. Neither a chiropractor nor any other healthcare provider practicing under the protection of a licensed profession has the ethical right to promote unscientifically unreasonable beliefs. The principle of fidelity and the state of scientific knowledge regarding certain historical chiropractic beliefs should not allow the expression of these beliefs to the patient as clinical truths.
After D.D. Palmer founded chiropractic in 1895 his original body of work contained a number of postulates. Below, we will present an analysis of Palmer's Postulates. This analysis is not new and has been available to the whole profession. We do not regard this analysis as anything that should be regarded as controversial or contentious. It is merely an observation that conventional scientific methods should be applied to the principles of chiropractic. Despite the critical threats to the validity of this paradigm, a sizable proportion of the profession still holds these postulates to be valid [4]. The segment of the profession that continues to hold firmly to Palmer's Postulates do so only through a suspension of disbelief. Given that one of the philosophical pillars of science is skepticism, a suspension of disbelief or a lack of skepticism, is evidence of antiscientific thinking [5,6]. These stratagems to avoid the truth that Palmer's Postulates are unproven might be beneficial to the chiropractor, but are ethically suspect when they allow the practitioner to maintain a "faith, confidence and belief" in that paradigm to the patient's ultimate detriment.
Misplaced Optimism
Over the past two decades it has been possible to view the chiropractic profession and its prospects for advancement in an extremely optimistic light. Compared to the profession's first 85 years of existence, the period of time from, say, 1980 to 2000 saw what seemed to be an unbroken string of successes. This period saw the ongoing development of the first chiropractic scientific journal, the first evidence (through clinical trials) of effectiveness of spinal manipulation, a legal anti-trust victory over the institution of medicine in the USA (Wilk v. AMA); an explosion in the number of students enrolled in chiropractic colleges, and the publication of a United States government report supporting the use of spinal manipulation for low back pain. In addition to these concrete developments the chiropractic profession benefited from the widely documented increase in interest and utilization of what has become known as complementary and alternative medicine (CAM) [7-9]. By the end of the century, as the result of these events and trends, the profession enjoyed a level of public acceptance (including that of other healthcare professions) that was unprecedented in its history. Some analysts of the healthcare system projected that by the year 2010 there would be over 100,000 chiropractors practicing in the United States alone [10,11]. It appears that reality will fall well short of that prediction.
As propitious as these developments appeared at the time, they have not secured the future of the chiropractic profession. A recent assessment by Richard Cooper MD, identified a variety of factors that threaten the future of chiropractic [12]. Dr. Cooper's analysis has captured the attention of many in the chiropractic profession and represents a realistic set of concerns, and calls for corrective action by the leadership of this profession.
During this same period, the healthcare system as a whole has undergone profound scientific, regulatory, political and economic changes that impose new expectations and responsibilities on all healthcare providers. An unprecedented level of professional accountability, predictability, and consistency are expected from all healthcare professionals. The chiropractic profession of the 21st Century is obligated to provide a mature, ethical, and moral response as it seeks to anchor its professional jurisdiction and cultural authority.
Internal Confusion
The chiropractic profession is not currently prepared to effectively meet these challenges. More than 100 years after its origins, the chiropractic profession remains focused on the internal debate "What is chiropractic?" – a quandary shared by many other stakeholders in the healthcare system. Perhaps as testimony to some underlying strength of chiropractic, the profession has managed to survive in spite of its confused self-vision. The more important issue is the profound organizational weakness suggested by the century-old debate on fundamental identity. It is difficult to fault decision-makers within the healthcare industry for any reluctance to embrace chiropractic when they do not know what it is they are asked to embrace.
There is a lack of uniformity and consensus within the profession about the proper role of chiropractic. Depending upon whose point of view is solicited; chiropractors are subluxation-correctors, primary care physicians (PCP), neuromusculoskeletal (NMS) specialists, wellness practitioners, or holistic health specialists. Within each of these models there are many competing factions. While the many professional subgroups of medicine (pediatrics versus cosmetic surgery, for example) converge, at least in theory, on broad but common ideology and professional attributes, the same is not true among the more divergent chiropractic factions. The differing chiropractic schools of thought form competing professional models that are not mutually compatible. Moreover, the disparities are indefensible in the context of the scientific, regulatory, political and economic criteria under which healthcare delivery is expected to operate. A number of models are impractical, implausible or even indefensible from a purely scientific point of view (e.g., subluxation-based healthcare), from a professional practice perspective (e.g., the primary care model), or simply from common sense (e.g. Innate Intelligence as an operational system for influencing health).
Part II: The Failed Identities of Chiropractic
The "ACC Paradigm" document developed by the Association of Chiropractic Colleges in 1996 currently represents the closest thing to an official consensus of chiropractic identity [13]. This paradigm was formed by consensus among the 16 presidents of the member ACC institutions – a group generally believed to hold divergent beliefs and interests. We respectfully submit that this widely disseminated document does not fulfill the criteria outlined above. While perhaps a political triumph (getting all the presidents to sign on to the same document), it contributes little to the understanding of the profession's role in modern healthcare delivery by the relevant stakeholders. It is interesting that two major sources of contentious debate, the terms "subluxation" and "diagnosis," are both used in the same document. Even in that context, the reader may be left with a feeling of internal tension between them. It is otherwise a recitation of the trivial (the purpose of chiropractic is to optimize health), the obvious (doctors of chiropractic establish a doctor/patient relationship and utilize adjustive and other clinical procedures unique to the chiropractic discipline.), and of the tautological ([chiropractors]...employ the education, knowledge, diagnostic skill, and clinical judgment necessary to determine appropriate chiropractic care and management.) Experience with healthcare decision-makers at both the local and federal levels makes it appear highly unlikely that the ACC Paradigm will prove useful when these decision-makers assess the practical role of the profession.
The chiropractic profession has succeeded in a number of important ways. Foremost, it has provided an effective and much needed healthcare service; that is, the conservative management of common musculoskeletal disorders in a population of patients who would otherwise be less well treated. It has devoted its resources in creating a sizable infrastructure of schools, publications, research centers, and scientific conferences. It has succeeded in providing economically viable careers for tens of thousands of individual chiropractors. Inroads have been made in policy-making arenas and in efforts to train its members in practice protocols to facilitate a stronger interface with payers and policy makers. Interdisciplinary training has begun to establish a cadre of qualified clinical and fundamental scientists with a chiropractic background. Chiropractic has succeeded in transforming itself from a marginal discipline into one that has an opportunity (if it acts wisely) to become an integral part of the healthcare system.
The basic premise of this paper is that existing institutions within chiropractic have not expressed a model of chiropractic that empowers the granting of cultural authority, sustained economic viability, and scientific integrity. There are two particular perspectives we believe are at odds with the seven criteria outlined above: 1. The philosophical model and 2. The primary care model. In order to effectively make a case for the Spine Care model that we propose, we must first directly address these two differing points of view.
The Philosophical Model of Chiropractic
The word "philosophy" is a much used but much misunderstood term within chiropractic. Most of the time those who invoke a "philosophical" argument are using the term in its colloquial sense: "I believe in a traditional set of chiropractic beliefs (chiropractic philosophy)." This set of beliefs is probably more correctly described as the ideology of chiropractic or the hypothesis of chiropractic, rather than as a philosophy.
This model of chiropractic has continued to advance a hypothetical model of health and disease divergent from other (particularly mainstream) modes of thought among the health professions. Indeed, some aspects of the hypothesis are now known to be at odds with scientific fact. To what extent can this chiropractic hypothesis be credited with the past successes of the profession? We argue that it is incorrect to interpret the success of the chiropractic profession as evidence of the validity of this chiropractic hypothesis. The profession has recorded limited successes in spite of what is largely the failure of this hypothesis.
What is the Chiropractic Hypothesis?
Before going further it is necessary to specify exactly what is meant by the chiropractic hypothesis. While there are an abundance and variety of competing versions of this hypothesis, all of which are ferociously defended by their adherents, it is still possible to identify several principles that are both common to the majority of these, and distinct from other healing systems. These principles are:
1. There is a fundamental and important relationship (mediated through the nervous system) between the spine and health.
2. Mechanical and functional disorders of the spine (subluxation) can degrade health.
3. Correction of the spinal disorders (adjustments) may bring about a restoration of health.
For the purpose of this discussion, these three principles will be referred to as Palmer's Postulates. There are a variety of different ways in which these postulates are expressed. The structure/function metaphor is often invoked – alterations of the body's structural components will result in functional aberrations and disease. Others emphasize the neurological aspect, the spine being both the source of noxious neurological stimuli and the locus of therapy where treatment can be administered to correct such stimuli.
But in the end, all of these modes of expression converge on essentially the same end point. That is the concept that the spine is not just another conglomeration of bone and muscle like the shoulder or the knee. Rather, it occupies a unique and privileged position in the makeup of the human body, representing both a vulnerability to our health and also a means of achieving optimal health. Expressions of Palmer's Postulates are ubiquitous within the profession and are not confined to extreme or narrow elements of the profession. These principles are to be found in some form in the mission statements of every North American chiropractic college and in the curricula of those colleges. They are further embodied in the ACC Paradigm paper. With the understanding that there is a great deal of room for qualification, clarification, and interpretation, we believe that Palmer's postulates do capture the essential hypothetical premise of chiropractic, and it is an error to underestimate the degree to which this theoretical model continues to define chiropractic. Even in the context of chiropractic research, where you might not expect a great deal of sympathy for these ideologies, Palmer's Postulates continue to guide the research priorities and agenda in the chiropractic profession.
We must also consider the concept of vitalism (in chiropractic, Innate Intelligence) as a component of Palmer's Postulates. Although there is a long historical legacy of vitalism, and although it continues to be a feature within many contemporary belief systems, there really can be no compromise on its inclusion as a defining principle of chiropractic. It was precisely the rejection of vitalism in the 18th Century and the emerging understanding (through the invention of the microscope and other technological advances) of biological mechanisms that marks one of the watershed moments in the evolution of science. Chiropractic can choose to retain its vitalistic component only if it chooses to operate completely outside the scientific healthcare community. Vitalism does not require any further or more extensive analysis before rejecting it. To reject vitalism is to simply to announce that one accepts the conventional view of biology similar to the way one accepts the convention view of cosmology by rejecting a geocentric universe. In making this categorical rejection of vitalism one important distinction is necessary. While vitalism is incompatible with a valid professional model of chiropractic, it is not incompatible with an individual chiropractor's professional beliefs. An individual physician of any type may have religious convictions that inform their professional lives, and yet these convictions remain totally outside the domain of the professions' common identity. Similarly, an individual chiropractors belief (or non-belief) in vitalism can be considered to be entirely a personal matter so long as these beliefs do not distort the discharge of professional duties and obligations.
A distinction can be drawn between the "classical vitalism" described above and a "modern vitalism" that can be accommodated by conventional biomedical science. This modern vitalism is best described by the phrase vis medicatrix naturae – the healing power of nature. The truth of this proposition is indisputable. Nature, or more specifically, the body's natural healing mechanisms, is the principle mechanism by which any healing process occurs. Without these natural mechanisms (our immune system, our wound healing capacity, and countless other regulatory and corrective systems) life itself is barely possible.
This modern vitalism can also serve as a useful and valid guiding clinical principle. It implies, correctly, that these natural healing systems should be given every opportunity to operate with minimal interference by outside agencies, including by chiropractors. This sort of therapeutic minimalism is, in fact, an important part of model that we will propose.
We have asserted that Palmer's Postulates have failed. To understand our assertion, please first consider the nature of a scientific theory. A theory is an explanation. It is an effort to explain and make understandable a set of observations or facts that are otherwise confusing, paradoxical, or self-contradictory in some way, and for which our existing theoretical understanding offers an inadequate explanation. Implicitly, every theory is an answer to the question, "Why is it that...?" or, "How could it be that...?" A theory should be a solution to a puzzle. If a theory is sound it will solve the puzzle and also accurately predict as yet unobserved phenomena, thus increasing our ability to understand and manipulate our world. For example:
• William Jenner's theory of acquired immunity provides an explanation for the observation that milkmaids with cowpox scars do not contract smallpox.
• John Snow's theory of cholera transmission answers the question, "Why did almost everyone who drank from the Broad Street well contract cholera, and those who drank from other water sources did not?"
• Barry Marshall's theory of the infectious nature of ulcers answers, "Why does the occurrence of peptic ulcers, thought to be a psychogenic disease, very closely resemble that of infectious diseases?"
When looking at these and other successful theories, there are some important common elements:
• In each case, there was a riddle to be solved, a set of unexplained facts. The theories did not arise out of a vacuum. They arose out of the necessity to explain some new observations.
• The observations were accurate. The phenomena that Jenner, Snow, and Marshall were trying to explain were real. They had correctly perceived and recorded events in their world. For great scientists, observation implies a deliberate, systematic, and disciplined process, and not simply the casual perceptions of our surroundings and experiences.
• The observations could not be explained by existing theory. Each of the sets of observations described above were either at odds with our existing understanding of the world or simply not taken into account by other theories.
• All have survived repeated experimental test.
When one examines Palmer's Postulates in this light, their limitations become obvious. First, we need to ask what phenomena, exactly, are these postulates trying to explain? Particularly with respect to the first postulate that establishes the relationship between the spine and health, what observations gave rise to this hypothesis? Is there some set of facts or observations that cannot be understood without the insight provided by the postulates? D.D. Palmer might state that he was trying to explain why a deaf man with a vertebral misalignment recovered his hearing following re-alignment of that vertebra. However, there is no evidence that Palmer undertook any sort of systematic exploration of the spine/health relationship following his epiphany. What we know about D.D. Palmer suggests that patient and disciplined observation was not his forte. His method of discovery was by inspiration and revelation.
Subsequent generations of chiropractors might say that Palmer's Postulates are required to explain why there are so many healthy, happy, satisfied, apparently healed chiropractic patients. But there is nothing puzzling or mysterious about doctors having content patients – all healing systems from Ayurveda to chiropractic to medicine to therapeutic touch can make such claims. The power of natural history, regression to the mean, and non-specific treatment effects guarantee such results and unless one sets out to deliberately harm patients, it's difficult to avoid having satisfied and improved patients. Recovered patients are the inevitable consequence of having patients and no insight is gained into the validity of any of these healing systems by observing this fact.
The problem, simply, is that there is no need for Palmer's Postulates. There never has been a set of facts or phenomena concerning the relationship between the spine and health that require Palmer's postulates to understand them. The spine/health theory does not rest on any foundation of careful, comprehensive, and reliable observational data.
To illustrate this absence, the sort of observations that would require the explanations of Palmer's Postulates might look something like this:
• The observations that most persons with idiopathic scoliosis suffer from a wide range of diseases that non-scoliotics do not.
• The observation that persons with a specific spinal characteristic suffer inordinately from a particular health problem.
• The observation that back pain predictably results from certain postural defects.
The problem is that none of these observations, or any similar, are known to be true. Where evidence exists on these questions it points mostly in a direction the opposite of Palmer's Postulates. The real paradoxes and riddles are questions like, "Why is it that a scoliotic, osteophytic, degenerated spine with asymmetrical facets and collapsed discs can so often result in no clinical problems?" Or, conversely, why is it that someone with no identifiable anatomic spinal disorder can suffer from low-back disability. A disinterested party, dispassionately examining the evidence available today regarding the relationship between the spine and health, or the structure/function relationship, would arrive at the following conclusion:
The human organism is highly resilient and broadly adaptable to a wide range of structural imperfections, and it is only after a rather high threshold of deformity is surpassed, that function is degraded.
The Primary Care Model of Chiropractic
The other great divide within chiropractic concerns the question of whether or not chiropractic is a primary care profession. Unfortunately, just as the word "philosophy" is routinely misused, so is the concept of "primary care." Paradoxically, even the extremes of the profession on the philosophy question (e.g., Sherman College and National University) both endorse the notion of chiropractic as a primary care profession. This agreement does not suggest that chiropractic, as primary care is a valid and compelling concept. Rather, it suggests that the concept has been unexamined and hastily adopted. This section will examine the meaning of primary care as it applies to chiropractic.
What is Primary Care?
There are several definitions of primary care physicians (PCP), but possibly the most accepted is the definition provided by the Institute of Medicine in a 1996 report. It defines primary care as, "the provision of integrated, accessible, health care services by clinicians who are accountable for addressing a large majority of personal health care needs, developing a sustained partnership with patients, and practicing in the context of the family and the community [14]." The essence of the IOM definition, as well as others, is of a primary care physician as a generalist and not a specialist. This is most easily illustrated by the prototypical examples of PCPs as identified in the IOM report: family practitioners, pediatricians and internists. The report also identifies nurse practitioners and physician assistants who are specifically trained in providing primary care.
In each of these examples, the PCP provider sees a wide range of complaints (respiratory, cardiovascular, gastrointestinal, and musculoskeletal) within the specified patient population, treats most of these complaints directly, and refers the rest as appropriate. Even in the more limited primary care professions (nurse practitioner, physician assistant) the generalist theme is also fundamental to defining their practice. These practitioners provide more limited care than medical PCPs and act more in a triage capacity than in a therapeutic capacity depending on complexity of the case. But there is general agreement that these providers fit the primary care model when they opt for the generalist practice.
To what extent do chiropractors satisfy the generalist model? Not at all, as it turns out. The most obvious index of this is the chiropractic patient population. In the last decade there have been many studies, surveys, and analyses that have described and characterized the chiropractic patient population [15-21]. These studies all reach the same conclusion: the chiropractic patient population consists, almost in its entirety, of persons with musculoskeletal pain complaints, the overwhelming majority of which are spine related. A small subset, approximately 5%, of patients have headache as a primary complaint. Any reasonable estimate would place the percentage of chiropractic patients with somatic pain at >95%. Most of the balance of patients receive some sort of "maintenance" or "wellness" care. A very small number (<1%) have complaints that fall outside these categories.
It might be argued that the make-up of chiropractic patient population simply represents a cultural and historical artifact; that the public has not been educated as to the suitability of chiropractors as PCPs and it's simply a question of providing proper education to the public on this matter. The fundamental limitations imposed by the profession upon itself make this argument implausible.
The first limitation is therapeutic. By intent, chiropractic has limited its therapeutic armamentarium to manual and physical techniques. This limited set of therapies is well suited to the set of complaints normally seen by and successfully treated by chiropractors. This limited set of therapies also offers the advantage of a very low risk of harm. However, this limited set of effective services is poorly suited for providing primary care. Beyond musculoskeletal conditions, there are very few conditions for which manual therapies provide optimal effectiveness. The vast majority of human health problems that require an intervention do not fall within the chiropractic therapeutic spectrum. Chiropractic cannot simultaneously retain its limited set of therapies and pursue primary care status.
It might be argued that even with its therapeutic limitations chiropractic could provide the services of a diagnostic generalist and make therapeutic referrals as needed. However, the defining characteristic of any diagnostic generalist is a rigorous training and experience with the spectrum of disorders likely to be encountered. Any intellectually honest analysis of this question will not support the supposition that chiropractic training provides such rigor in this domain. The length, breadth, and depth of chiropractic clinical training do not support the claim of broad diagnostic competency required of a PCP. Studies of chiropractic intern clinical experience provides no evidence that chiropractors are trained to a level of a diagnostic generalist for non-musculoskeletal conditions [22,23]. For chiropractors to describe themselves as PCP diagnosticians is to invite comparisons to other PC diagnosticians, i.e., family practitioners, pediatricians and internists. Such comparisons will not reflect favorably on chiropractic.
Finally, it might be argued that although the chiropractic profession is not currently trained to provide PCP care, it could be and we should set ourselves to the goal of making this happen. If a chiropractor as PCP is not at this moment a reality, we can imagine a different reality in the future in which the Chiropractor/PCP model made sense. What would have to change for this reality to come true? At a minimum, the following:
1. Chiropractic would have to dramatically increase the length, breadth and depth of its clinical education at all its accredited institutions.
2. Chiropractic would have to develop an acceptable solution to its therapeutic limitations, either through changes in state licensure or by some as yet unidentified process.
3. Chiropractic would have to demonstrate its ability to deliver safe and effective care beyond its current model.
4. Having achieved goals 1-3, the chiropractic profession would have to change the view of the public and other health professions of chiropractors as back doctors.
5. And finally, the profession would have to convince the healthcare marketplace (in which there is no current or anticipated shortage of PCPs) that there is some point to expanding the number of PCPs.
These events do not appear to be likely to occur in the near future.
Part III: The Spine Care Model
In the course of discussions among the authors of this paper as well as others who were involved in the process, it became clear that there were many points of consensus. These consensus points are listed below in the approximate order of their importance to the model.
• Chiropractic as an NMS specialty, with particular emphasis on the spine.
• Chiropractic as a portal of entry (POE) physician/provider.
• Chiropractic as a willing and contributing part of the evidence based healthcare (EBHC) movement.
• Chiropractic as conservative/minimalist healthcare provider.
• Chiropractic as a fully integrated part of the healthcare system, rather than as an alternative and competing healthcare system.
Incorporating all of the above elements, chiropractic should actively market itself to the public and to the rest of the healthcare system in a sober and moderate fashion, and with a message that is completely compatible with current social, economic, political, and scientific realities. The balance of this paper will be devoted to examining these issues.
The Dental Model
As a start to defining the model it is helpful to find another profession with analogous clinical jurisdiction e.g. focused practice emphasis on a region or set of problems, limited therapeutic regimen, and broad public identification with a selected role in healthcare. We believe the dental profession is a practical and successful parallel. Consider the advantages of the dental model:
• Dentists and dental surgeons have established themselves as the absolute, undisputed authorities in tooth care, a critical and essential component of human health, and a contributor to care for orofacial disorders generally. No one suggests they should not be portal of entry providers. No other profession considers usurping the role as tooth-care expert.
• In the public's perception, dentists are among the most highly esteemed of the healthcare professions.
• Dentists are recognised with the title "doctor" and reap the social, professional and financial benefits of their reputation and training.
• Dentists, though primarily focused on the dental anatomy and disease, are also expected to understand differential diagnosis of conditions related to their area of focus.
• The services that dentists provide, focused though they are to tooth, gums, and mouth, are of immense benefit to the health and well being of the public.
As this model unfolds, this is the image we might want to keep in mind – chiropractors as dentists of the back.
The Vocational Role of Chiropractic: Treatment of Back Pain
The purpose of this essay is to define chiropractic as a profession. The term is emphasized because it is necessary to remind ourselves what this means and what are the consequences of being a profession. A profession is not defined by a set of ideas and values. Professions may have ideas and values, but these are not what distinguish or differentiate them as professions. Those organizations that are defined by ideas and values are entities like political parties, ideologies, religions, or organizations devoted to narrow issues like pro-life or pro-choice organizations. For such organizations, it is correct to state that the idea comes first, and everything else – strategy, tactics, etc. – flows from the question: what will best promote our idea?
A profession is about a specific vocational role that the profession fills. A profession is defined by the work it does and the role it fills, not by its ideas and values [24]. The ideas and values of a profession must be secondary – they exist to answer the question: "How can we best discharge our designated role in society?" Professions do not or should not exist to be champions of ideas. This is most specifically true of the licensed professions. Society grants a license, a franchise, to a profession, not so that profession can champion its ideals, but because society wants some specific work done and it feels that granting a franchise is the best way to do it. This social contract is quite explicit. In most cases the vocational role of professions is quite obvious and can be stated in a few syllables:
• Tooth and gum care.
• Design and engineering of buildings.
• Measurement of financial performance.
• Legal services.
This simple and coherent vocational role is what the chiropractic profession seems to have so much difficulty in defining, and what the ACC paradigm fails to provide. Among the reasons for this failure is that chiropractic has always been confused about the concept of a profession and has tended to view itself a champion of ideas rather than as a provider of service. This confusion is perhaps understandable in an historical context. Chiropractic didn't begin as a profession; it began as an idea or set of ideas (vitalism, subluxation). Palmer and company were champions of these ideas, competing with charlatans and learned (not scientific) professional rivals for status. Over the decades, the institutions and each individual chiropractor saw themselves as a champion of the chiropractic idea.
But, at some point over the last 100 years, and unbeknownst to the individuals and institutions of chiropractic, it became a profession with a specific vocational role. As these thousands of chiropractors over the decades were advancing the ideals of the profession through manipulation of the spine, the public, which is largely disinterested in the ideas, decided that chiropractic had a professional role to fill. Thus, creating the profession as it exists today.
The irony is that the specific professional/vocational role that chiropractic fills is obvious to the majority of patients and other non-chiropractors – it is chiropractors themselves who seem to be confused by the issue and who then provide confounding answers and contradictory testimony to policy makers. For all other mainstream healthcare professions it is easy to provide a straightforward answer to this question of role. Whether it is an optometrist, a pediatrician, a dentist, a family medical practitioner, or a psychologist, each has clinical domain that is essentially self-evident. For all other PCPs, and POE (point of entry) providers there is a relatively clearly defined patient population for whom the practitioner is an appropriate provider. This patient population may be defined by age, gender, and most importantly, by nature of healthcare problem or complaint. There may be some disagreement among various professions at the margins of this question, but only at the margins.
A somewhat different state of affairs obtains for those health professionals whose clinical purpose is not defined by a patient population, but by a specific technique or skill. For example, consider a general surgeon, pathologist or radiologist. The potential patient population of these providers is virtually everyone, as a function of their specific need for the service. To some this might seem an attractive model for chiropractic – our patient population is everyone who needs spinal correction, which is to say, everyone. In fact chiropractic has attempted this by defining itself in metaphysical terms (Innate Intelligence), as a technique (chiropractic adjustment), and as an ideology (Palmer's Postulates), rather than as a provider of specific clinical services. The failure of this approach is in fact the genesis of this paper. To define the clinical purpose of chiropractic, it is necessary only to observe what chiropractors actually do and for what purposes patients seek care from doctors of chiropractic: the provision of portal-of-entry care for the diagnosis and management of back pain, neck pain, and related disorders. In the shorthand that the public might use, chiropractors are back doctors. Restating some of the earlier points, this conclusion is based on these facts:
• The population – Over 90% of chiropractic patients seek care for back-related problems.
• The evidence – Clinical science provides a body of evidence for the effectiveness of chiropractic care for back pain, neck pain, and headache.
• The education and training – Chiropractic clinical education and training are focused almost exclusively on the conservative treatment of spine complaints.
• The public identity – The public perception of chiropractic is that of a back pain specialist and nearly a total rejection of an alternate role.
• The competition – The legitimate professional claim for chiropractic in the remainder of healthcare and public policy lies strictly within the domain of back- related pain outside the bounds of medical emergency.
• The claim of professional jurisdiction – Credibility for the claim, either diagnostically or therapeutically, for a broader role beyond the realm of this definition is lacking.
Should the chiropractic profession concern itself with what others think? It should, must and had certainly better do so as it is reliant upon its consumers for its existence. A profession is a public trust. The privileges accorded to a member of a profession are in direct exchange for professional members' service to the public. It is nonsensical to organize a profession in terms that are at odds with the public's perceptions of its interests unless a compelling and persuasive argument can be made that the public's perception is not in their best interest and is amenable to change. We maintain that there is no such argument. In fact, efforts to launch such a campaign have failed. For example, two recent public relations efforts have been attempted by chiropractic organizations. These efforts were preceded and followed by measure of the public attitudes toward the profession. In both cases, efforts to convince healthcare consumers about the role of subluxation in their lives backfired miserably. Not only were few persons encouraged to consult a chiropractor, but, the number of skeptics was increased and more respondents stated that they would seek a medical consultation first following the PR effort than before the campaign. The argument that the public can be persuaded to understand and accept the subluxation model of chiropractic has been tested and it has failed.
Finding a substantial presence within the healthcare marketplace is well satisfied by the spinal care model. A recent analysis of healthcare and productivity costs associated with specific complaints reveals the following:[25]
• Three of the top 10 conditions suffered by the US population (in terms of costs) are back pain related.
• Collectively, the annual rate of back pain-related healthcare episodes is 157 episodes per 1000 covered lives, making it the single most common complaint.
• Collectively, the annual direct healthcare cost for back pain is US $122 per person, second only to the cost of managing angina pectoris.
• Collectively the annual average cost of payment for lost work and short-term disability is US $87 per person, making back pain the most costly of all diagnostic categories in disability-related costs.
It should be noted that while some of these back pain episodes are undoubtedly not chiropractic cases (that is, they are legitimate in-patient or surgical cases) almost all are. Conservatively, at least 75% of this spine care patients potentially stand to benefit from chiropractic care, compared to the 12-17% who currently avail themselves of the services. This study, and many others, provides ample evidence that the clinical domain of back pain provides an enormous potential patient base and subsequent economic base for chiropractic.
Thus, the logic of the chiropractor as spinal care doctor proceeds as follows. First, chiropractors are de facto back pain/spine doctors seeing a limited proportion of the population, today. That is, as chiropractic is currently practiced (even given the confused message that chiropractic projects) it is entirely dependent on back pain/spine care for its economic survival. Second, the back pain market is enormous and can provide, by itself, a sufficient patient base to support the entire profession. Third, expansion of the chiropractic market share for spine-related symptoms is hindered primarily by a lack of credibility of its claims and the resistance that this lack generates among consumers and policy makers. Fourth, chiropractic has the most clinical training, expertise, and demonstrated clinical effectiveness as conservative back pain/spine doctors. Fifth, chiropractic as a spinal care specialty is the only basis on which the profession is understood and accepted by those outside the profession. Sixth, there is nothing to be lost, either in the short or long term by adopting this strategy. The state of mind regarding the profession that we would like to make is: Go to a DC for your spinal health and prevention as you would go to your dentist for your dental health and prevention. We reemphasize that there is nothing to be lost, either in the short or long term, by adopting this strategy. This model of chiropractic as the spinal care profession is in no way intended to preclude the patient population of extra-spinal musculoskeletal complaints. However there are several reasons why we feel it is reasonable to de-emphasize, relative to spinal care, this patient population.
1. It represents a very small percentage (<5%) of the current chiropractic patient population.
2. There is very little evidence of effectiveness of chiropractic care for this population and it is unlikely that a sufficient number of these patients present for care in order to conduct appropriate studies in a reasonable and timely manner.
3. It is unclear what advantage(s) chiropractic care might offer relative to other providers (physical therapists, rheumatologists, etc.) for care of these problems.
4. It is likely, with today's knowledge, that the proportion of extra-spinal MS patients for whom conservative manual therapy is the optimal approach is significantly less than is the case for spinal conditions.
5. There is far less public awareness or willingness (as reflected the utilization of services) of chiropractic as a provider of care for these conditions.
Portal of Entry Status
We suspect that among some chiropractors there is confusion about the two terms "primary care," and "portal of entry," and that this confusion is at least partially responsible for the enthusiasm for the primary care model. The American Chiropractic Association, in fact, uses both terms to describe the profession [26,27]. However, primary care, as discussed above, describes a generalist provider, while a portal of entry (POE) describes a health care provider who may practice autonomously and to whom the public has direct access. The confusion lies in the belief that in order to achieve portal of entry status one must first be a primary care provider. A primary care physician is certainly a portal of entry provider, but one need not be the former to be the latter. The examples of dentistry, optometry, and clinical psychology illustrate this point.
On this question there is virtual unanimity in the chiropractic profession and the logic of chiropractors as portal of entry providers is obvious to all but the most vociferous opponents of the profession. The POE status of chiropractic is guaranteed in all the 50 American states as well as in most countries outside the US where chiropractic is licensed. There is no credible case that can be made that in some manner the public will be better served by requiring them to go through a gatekeeper (presumably an MD) to seek care from a chiropractor. The primary impediment to full implementation of portal of entry status is not a regulatory or a statutory problem, but a problem of inter-professional trust. Within specific health care delivery and financing systems there are gatekeeper provisions that require patients to be referred for chiropractic care. These gatekeeper arrangements arise either through concern of improper diagnostic workup and clinical decision-making, or through concerns of utilization abuse. While the fairness and appropriateness of these gatekeeper requirements is certainly in doubt, the surest way for the profession to protect and expand its POE status is to establish the cultural authority, and thus, the trust, that will make these gatekeeper provisions unthinkable.
The Acceptance of Evidence-Based Healthcare
Fifteen years ago, the editor of the New England Journal of Medicine, Arnold Relman, MD, wrote an editorial in which he announced that healthcare had entered a new age, The Age of Accountability [28]. What he was describing is what we now call Evidence-Based Healthcare (EBHC). During the same period of time in which the CAM revolution was taking place, a second less visible revolution was also taking place – the establishment and application of the principles of EBHC.
Evidence-based healthcare is often ill defined, misunderstood, and a basis for concern or even fear by health-care providers. One of the best definitions we have seen appeared in an editorial in the British Medical Journal in 1996 written by some of the most prominent educators in EBHC, David Sackett and his colleagues [29]. They defined EBHC as the conscientious, explicit and judicious use of current best external scientific evidence in making decisions about the care of patients. EBHC does not mean that individual clinical experience is of limited or no value; on the contrary, EBHC offers advice on how to maximize the clinical expertise and combine that with the best available external scientific evidence that usually comes from systematic reviews and evidence-based clinical guidelines. Another important aspect of EBHC is the identification and incorporation of informed patient preferences.
The concern and fear that many health-care providers have is that EBHC will be misused by healthcare policymakers and health insurance companies to curtail the cost and limit reimbursements. Such policies would be inconsistent with the fundamental principles of EBHC. Clinicians who practice EBHC will develop the skills to identify and apply one or a combination of the most efficacious treatments, which if based on the individual patient profile will tend to maximize the benefit and minimize the risk. This may sometimes raise rather than lower the cost of their care. EBHC is not about proof or certainty. It is a method of dealing with uncertainty. It is about weighing the evidence and weighing alternatives.
There is one additional element of EBHC that requires amplification. It is important to understand EBHC does not mean care should be withheld if there is no proof of efficacy from systematic reviews or meta-analyses of randomized clinical trials. Absence of evidence of treatment efficacy does not equate with evidence of its absence. Such a standard would produce therapeutic paralysis. For example, there are virtually no clinical studies, chiropractic or otherwise, that have evaluated the effectiveness of treatment for thoracic spine pain. Obviously it is not reasonable to send a thoracic spine patient home with the apology, "Sorry, can't treat you – no evidence of efficacy." It is however essential that clinicians understand that evidence ranges from the weakest (clinical experience or expert opinion) to the strongest (high quality systematic reviews of all available relevant scientific studies). Many different systems for grading the evidence and making recommendations currently exist, and major efforts are underway internationally to standardize this process.
The Role of Clinical Experience in EBHC
The central premise of EBHC is that even the most well thought out, tightly reasoned, and scientifically plausible treatment regimen may not produce benefit to the patient. The scientific literature is overflowing with examples of commonly used treatment procedures or regimens which were based on sound pathophysiologic principles, but were ultimately found to be of no benefit or even harmful to the patient [30-33]. For the chiropractic profession the lesson is obvious. Whether its Palmer's Postulates or any of its innumerable variations (in the form of proprietary techniques) the chiropractic profession cannot predicate its clinical validity upon untested theories.
EBHC principles state that healthcare providers need to combine their clinical expertise with the best available external evidence and that neither alone is sufficient. The most difficult and counter-intuitive notion for clinicians to accept is that their everyday experience of satisfied and seemingly recovered patients is not evidence of clinical effectiveness. There are several competing explanations for this apparent success. Many of the conditions treated by chiropractors, such as back pain, neck pain and headache, have a self-limiting natural history although they may be recurrent. The nonspecific placebo effect of the doctor-patient relationship explains many of the results attributed to specific interventions. Clinicians notoriously have selective memories and tend to recall success stories and generalize from those. The lack of systematic and standardized recording of diagnoses and clinical outcomes that could be gathered in databases and summarized objectively prevents clinicians from having an unbiased knowledge of the effect of their therapeutic efforts. EBHC recognizes the limitations and inherent unreliability of uncontrolled clinical observations and impressions and the inevitability of mistaken conclusions based on those uncontrolled observations. EBHC stresses the importance of outcomes-based clinical research, of regularly consulting the scientific literature, of optimizing the clinical skills of healthcare providers, and taking patients preferences into account.
As a practical matter, many chiropractors, and medical physicians as well, fear that EBHC will result in a change and possible limitation of their individual clinical prerogatives. They are correct in this conclusion. It is in fact the precise purpose of EBHC to help define what constitutes best practices-different from what would be the case if individual providers were given free reign to continue with their habitual practice behavior based exclusively on clinical experience.
It is also important to recognize that EBHC is in its infancy. The processes of EBHC will continue to accelerate in the future. When there is enough evidence to justify it, relative to a particular condition, we see the development of "disease management" programs. Disease management represents nothing more than a highly evolved implementation of EBHC. When there is sufficient evidence available, it becomes possible to implement very specifically defined (and also, very effective) treatment protocols that take into account important differences in prognostic factors among patients. These programs already exist for congestive heart failure, asthma, urinary tract infections, diabetes and other common illnesses. It is currently not feasible but only a matter of time before disease management of back pain, for example, becomes possible and necessary. If the chiropractic profession hopes to make progress within the healthcare mainstream, it must go out of its way to be clear that it understands EBHC, that it embraces its principles, and that it is acting to advance its implementation.
Conservative/Minimalist Healthcare
One of the general truths revealed through the application of EBHC is that less is often more in healthcare. There are countless examples where clinical studies have shown that providing less healthcare, doing fewer procedures, taking a more conservative approach, even doing nothing, is superior to a more aggressive approach. This idea has always been understood at some level (it is the premise behind the "First, do no harm," doctrine), but it has been difficult for our healthcare system to act on the idea. Most incentives, economic and otherwise, propel care in the direction of more, rather than of less.
Chiropractic has a considerable advantage when it comes to implementing the doctrine of "First, do no harm." The scientific literature strongly supports the finding that chiropractic, and specifically, spinal manipulation, is generally safe. The evidence regarding spinal manipulation indicates that the incidence of serious injury is, if not trivial, extremely low. Of the more common adverse effects resulting from spinal manipulation, nearly all are transient and minor. Overall, the safety profile of spinal manipulation is excellent and more so when compared to other treatment options.
Through historical precedent, by intent and by design chiropractic has evolved using a conservative therapeutic regimen consisting of manual and physical therapies as well as exercise. The clinical effectiveness of this approach has been established, the safety profile is excellent and there are distinct cost advantages to this approach when used appropriately. We see no reason to change the therapeutic scope of chiropractic. It should be understood that this is a contingent position. It is contingent upon the continued clinical effectiveness and superior safety profile of these conservative modalities relative to other more aggressive interventions, particularly medication and surgery. None of these therapies, conservative or otherwise, will remain static and as they are improved upon in the future their relative merits may change as well. Chiropractic's allegiance to a conservative therapeutic regimen is valid only as long as it remains a clinically and economically sensible thing to do.
In order to fully exploit the advantages of its current conservative approach the chiropractic profession must take active measures to curb abuses that run counter to this approach. There is a long tradition in the profession of promoting the idea that the unadjusted spine is an invitation to disease. There are practice management procedures that attempt to maximize the number of patient visits that can be extracted from each new patient. There is nothing conservative about a treatment regimen of 3-times-a-week, forever. There is a commonly expressed notion among the public and among other health professionals that chiropractic treatment is open-ended and often never-ending. By these and other similar offenses, chiropractic has surrendered the high ground when it comes to delivering conservative healthcare. Using its current set of conservative therapies and incorporating the best published data, chiropractic can make a credible case that it offers the best combination of safety, effectiveness and cost for the management of back pain.
Integration
The spine care model will facilitate integration of the chiropractic profession into the mainstream of healthcare. Integration offers substantial advantages toward addressing professional values and resolving the concerns outlined in the beginning of this essay. It is the primary vehicle by which cultural authority can be anchored for its competencies currently supported in the scientific literature. Integration brings with it a greater responsibility, but also brings the resources and patient access necessary to answer the core issues common to all chiropractic ideological debates.
Chiropractic has operated a parallel tract of professionalization since its inception. As Abbott observed, parallel development is associated with significantly greater obstacles and opposition than a profession that evolves as a branch from common roots [24]. While there is much accomplishment to appreciate, the profession continues to be hindered by limited resources for its stability and advancement. While at least partial acceptance and licensure has been achieved in many countries, many challenges remain before the profession can establish its reputation of competence and legitimacy necessary for full cultural authority. In modern society, training and licensure is no longer sufficient to demonstrate competence. That requires continued validation, which, in turn, requires credible data and a coherent identity. Legitimacy is eroded if practice patterns are tied to reimbursement, profit margin, or professional rivalry.
Perhaps the most fundamental question that the chiropractic profession must answer to finalize its cultural authority is: "Does the chiropractic profession continue to position itself in opposition to orthodox medicine, or does it stand as an advocate of the patient's best interests, as a part of mainstream healthcare, along with medicine?" To date, the chiropractic profession has enjoyed the ability to evade that decision, occupying an ambiguous position between opposition of medicine and full participation in the mainstream. The profession and its members have often used marketing methods offering an alternative to medicine. At the same time, political activism in the USA has yielded many of the benefits of the mainstream through participation in the private third-party payment system, in Medicare and a variety of other state-sponsored programs, as well as inclusion in student loan programs and in the Veterans Administration and Defense Department programs.
The emergence of the phenomenon of CAM has also played a role over the past few decades. Analyzed in both scientific publications and the popular media, the CAM phenomenon is now a well-established and positively recognized element within our healthcare system. The difficulty is that the CAM phenomenon has reinforced the cultural authority chasm in which the profession finds itself. There is such significant evidence supporting chiropractic benefits for spine care that it is considered by policy makers to be more mainstream than CAM. Yet, professional claims over the non-musculoskeletal domain and questionable practice behaviors obstruct full consideration within the mainstream by purchasers of healthcare research and delivery. While for some, the notion of being an alternative healthcare provider has a certain cache; this notion is neither clinically nor scientifically justified. It is a cultural and political status crafted by society for the prime purpose of evaluating whether the claims made by such practitioners are of any value. In the long run, the evaluation will elevate some and will degrade others. As noted by Marcia Angel, in the special New England Journal of Medicine issue on alternative healthcare: "There is only medicine that has been adequately tested and medicine that has not, medicine that works and medicine that may or may not work."[34].
Further, the barrier to entry into CAM is too low for the profession of chiropractic. There are too many CAM-related procedures, practices and providers that lack scientific rigor. Chiropractic is by far the most mature profession among those associated with CAM. Its pre-professional requirements are the highest; the professional education the most developed; its research capacity, the most advanced; and its presence in the healthcare marketplace, the most comprehensive. Simply, the chiropractic profession undermines its legitimacy and authority by striving to remain within the CAM phenomenon.
We do not believe that this intermediate status of half-alternative/half-integrated is sustainable for much longer. The profession needs to decide in which of these two camps to plant both feet. Without the intent of its members, like the question of chiropractic's clinical purpose, this question of integration has largely been decided by default – chiropractic is an integrated part of the healthcare system, and the profession must continue to promote further integration. The benefits of integration to the profession are too great to ignore. To be a part of the system is to have access to all the resources of the system- funds for research, state supported education institutions, training opportunities in hospitals and other integrative clinical settings, access to other educational institutions and nearly universal inclusion in all reimbursement systems. We must take particular note of the recent approval and funding of a chiropractic college at Florida State University. This was a tremendously important development in this history of chiropractic and one that had the potential to profoundly deepen and accelerate the integration of chiropractic into the mainstream. What is unsettling is the fact that the college failed before it could even enroll a single student almost exclusively because of the failure of the profession to advance a coherent credible message regarding its role within the healthcare system.
For the profession, integration will insist on clinical accountability and responsibility, a demand that our members feel even now with the increased pressures of healthcare reform. The rewards of integration, however, are extensive. The experience of individuals who have broken down many of the barriers and succeeded in establishing chiropractic programs within mainstream healthcare centers is expanding. The development of chiropractic facilities for the United States Congress, within the military, and within private musculoskeletal centers has been universally positive for patients and for the participating chiropractors. Beside personal professional success, these experimental programs have bought additional trust and credibility within the system. The participants have experienced a hitherto unheard of expansion in clinical exposure. Increased patient volumes, case variation and complexity and provider satisfaction are evident. Doctors can experience a new freedom from the tyranny of personality cults and practice-builder manipulations. New opportunities for career track development are opening as healthcare policy makers, clinical and basic scientists and educators for interested individuals who are interested in cross training.
For the profession's infrastructure, integration confers enormous advantages. By functioning within the mainstream of healthcare, chiropractic will be able to gain access to a far broader population of patients and practice within a more varied set of patient care settings. The academic institutions will be able to free themselves from the stranglehold of economic dependence on tuition and the political reliance on ideological gurus who manipulate alumni and support to garner institutional control. The results will expand the jurisdiction and influence of the profession's cultural authority as warranted. The profession will be a member at the table of discussions and debate over the future of healthcare delivery. As a participant, chiropractic autonomy over its domain will be more certainly assured than in our current reactive conflict postures.
Other Issues
There are a variety of other questions that bear upon the issue of the chiropractic model.
1. The role of spinal manipulation in chiropractic. There is no foreseeable future in which spinal manipulation is not the primary therapeutic tool of chiropractic. But if or when that changes, it will be a function of the progress and evolution of clinical science, and not as a principle of chiropractic. That is, SMT should be viewed not as a defining element of chiropractic, but simply as what we happen to do. Invoking the dental analogy again, dentists do not define themselves as "implanters of dental amalgam," although that is probably what they do the most. As the discussion above on chiropractic philosophy illustrates, to do otherwise, to focus exclusively on SMT, as the chiropractic therapy will hinder our ability to pursue a more optimal treatment for back pain. We must make sure that we are prepared and equipped to identify and deliver whatever conservative therapies for back pain prove to be most effective.
2. The use of drugs. Should chiropractors seek limited prescribing rights as has been attempted in the past? Or should chiropractic promote itself as a "drugless" profession? We believe the answer to both these questions is "no." In the first instance (should chiropractors prescribe), clinical science has created a very strong case for conservative healthcare. Much of the advantage that chiropractic currently enjoys in the realm of back pain treatment (in terms of cost, safety, and satisfaction) is directly attributable to its conservative (non-drug) interventions. The US osteopathic experience is informative in this regard. Given the option of prescribing and using other more invasive interventions, it is much easier to prescribe than it is to use a manual therapy, and the role of manipulative therapy has diminished and nearly vanished from the profession.
Regarding the second question (should chiropractic promote its "drugless" nature), we should not promote the juvenile notion that drugs are bad and SMT is good. Our non-use of drugs should simply be regarded as a conscious decision to focus on a particular therapeutic approach, rather than a comprehensive rejection of drug therapy (or any other specific intervention that we do not happen to provide). Our position on drug use should be precisely the same as medicine: all drug use should be appropriate and guided by the scientific literature. And we should acknowledge that sometimes the correct treatment would involve drugs.
The decision to reject the use of drugs should always be contingent upon the scientific literature. The literature currently provides that conservative and manual therapies are legitimate treatment options for a large percentage of the patient population with spinal complaints. Until such point as it becomes clear that it is not possible to practice EBHC without drugs (and that point may never arrive) chiropractic should remain committed to conservative manual therapies.
3. Chiropractic education and licensing. The model we have proposed does not require any specific change in chiropractic education or licensure to be implemented. In fact, one criterion behind our model is that it reflects how chiropractors are educated, and how they practice. So, we already have concluded that the de facto model being taught at chiropractic colleges is that of a back pain specialist (their proclamations of primary care, notwithstanding). We do believe that a more explicit embrace of the Spine Care model would lead to a higher quality of education. We do, of course, hope that chiropractic education improves, particularly with respect to the patient care component of the education. Similarly, the Spine Care model is completely consistent with current state licensing. There will always be disputes and turf wars at the margins of the licensing process, and there are some onerous elements to some state laws, but we do not propose any wholesale revision or alteration of the statutory scope of chiropractic practice.
4. Wellness/prevention as a principle of chiropractic. Nearly all factions of the profession make the claim that chiropractic represents a "wellness" approach to health. Some factions use this term to mean, "We will prevent disease by eliminating subluxations." Others use the term to mean, "We will prevent back pain and related disorders by providing comprehensive spine care." And still others use the term to mean, "We will prevent a variety of degenerative diseases (cardiovascular, neoplastic, etc.) by advising patients on how to live a more healthy life." The first example is unproven and unlikely to be true. The second two examples are also unproven, although they are not scientifically implausible as is the first example. The question is whether the chiropractic can actually deliver on the promise to promote health and prevent disease (as opposed to treating symptomatic patients). To date chiropractic has not demonstrated that it can deliver on the promise of prevention. It is difficult to make the case that chiropractic, uniquely or distinctively among health profession, is concerned with, and capable of providing effective preventive health care. Chiropractors should certainly concern themselves with patients' behavior that may affect patients' health, and provide whatever advice, council, and encouragement they can to improve health related behavior. But, until we can demonstrate that we are effective where others are not, the proposition of chiropractic as the "wellness profession" is not defensible.
Summary
To date, the chiropractic profession has failed to develop the legitimacy necessary to defend its autonomy and cultural authority. It has not shown the will or ability to define for itself a coherent and consistent identity that takes into account the realities of the healthcare world in which we operate. If the profession fails to do so its future will be imperiled. We offer a professional model for the chiropractic profession. The essential characteristics of this model are:
• Spinal care as the defining clinical purpose of chiropractic.
• Chiropractic as a portal-of-entry provider.
• The acceptance and promotion of EBHC.
• A conservative clinical approach.
• Chiropractic as an integrated part of the healthcare mainstream
• The rigorous implementation of accepted standards of professional ethics.
Authors' contributions
The cover letter describes the general context and process by which this manuscript was created. After the first set of meetings Dr. Nelson wrote an initial draft document, which reflected the collective thoughts and analyses of the participants. All of the authors participated and contributed in the initial and subsequent meetings during which the overall themes were identified and described. All of the authors made original contributions to the content of the draft and final manuscript. And all of the authors participated in the editing and revisions of the multiple drafts that existed between the initial and final draft.
==== Refs
Ryan J Measuring subluxation; A report on public response Chiropractic Business 2000 Fall/Autumn
accessed March 23, 2005
Hughes EC Professions Daedalus 1963 92 655 68
McDonald WP Durkin K Iseman S Pfefer M Randall B Smoke L How chiropractors think and practice: the survey of North American chiropractors 2003 Ada, OH: Institute for Social Research, Ohio Northern University
Alcock JE Alternative medicine and the psychology of belief Sci Rev Alt Med 1999 3 45 52
Keating JC Toward a philosophy of the science of chiropractic: A primer for clinicians 1992 Stockton, CA: Stockton Foundation for Chiropractic Research
Eisenberg DM Kessler RC Foster C Norlock FE Calkins DR Delbanco TL Unconventional medicine in the United States. Prevalence, costs, and patterns of use N Engl J Med 1993 328 246 52 8418405 10.1056/NEJM199301283280406
Eisenberg DM Davis RB Ettner SL Appel S Wilkey S Van Rompay M Trends in alternative medicine use in the United States, 1990-1997: results of a follow-up national survey JAMA 1998 280 1569 75 9820257 10.1001/jama.280.18.1569
Tindle HA Davis RB Phillips RS Eisenberg DM Trends in use of complementary and alternative medicine by US adults: 1997-2002 Altern Ther Health Med 2005 11 42 9 15712765
Cooper RA Healthcare workforce for the twenty-first century: the impact of nonphysician clinicians Ann Rev Med 2001 52 51 61 11160767 10.1146/annurev.med.52.1.51
Cooper RA Laud P Dietrich CL Current and projected workforce of nonphysician clinicians JAMA 1998 280 788 94 9729990 10.1001/jama.280.9.788
Cooper RA McKee HJ Chiropractic in the United States: trends and issues Milbank Q 2003 81 107 38 12669653 10.1111/1468-0009.00040
accessed March 23, 2005
Donaldson MS Yordy KD Lohr KN Vanselow NA Primary care: America's health in a new era 1996 Washington, DC; National Academy Press
Cote P Cassidy JD Carroll L The treatment of neck and low back pain: who seeks care? who goes where? Med Care 2001 39 956 67 11502953 10.1097/00005650-200109000-00006
Palinkas LA Kabongo ML the San Diego Unified Practice Research in Family Medicine Network The use of complementary and alternative medicine by primary care patients. ASURF*NET study J Fam Pract 2000 49 1121 30 11132062
Sherman KJ Cherkin DC Connelly MT Erro J Savetsky JB Davis RB Eisenberg DM Complementary and alternative medical therapies for chronic low back pain: What treatments are patients willing to try? BMC Complement Altern Med 2004 19 4 9
Cherkin DC Deyo RA Sherman KJ Hart LG Street JH Hrbek A Davis RB Cramer E Milliman B Booker J Mootz R Barassi J Kahn JR Kaptchuk TJ Eisenberg DM Characteristics of visits to licensed acupuncturists, chiropractors, massage therapists, and naturopathic physicians J Am Board Fam Pract 2002 15 463 72 12463292
Sharma R Haas M Stano M Patient attitudes, insurance, and other determinants of self-referral to medical and chiropractic physicians Am J Public Health 2003 93 2111 7 14652343
Coulter ID Hurwitz EL Adams AH Genovese BJ Hays R Shekelle PG Patients using chiropractors in North America: who are they, and why are they in chiropractic care? Spine 2002 27 291 6 11805694 10.1097/00007632-200202010-00018
Hurwitz EL Coulter ID Adams AH Genovese BJ Shekelle PG Use of chiropractic services from 1985 through 1991 in the United States and Canada Am J Public Health 1998 88 771 6 9585743
Nyiendo J Phillips RB Meeker WC Konsler G Jansen R Menon M A comparison of patients and patient complaints at six chiropractic college teaching clinics J Manipulative Physiol Ther 1989 12 79 85 2523949
Nyiendo J Haldeman S A prospective study of 2,000 patients attending a chiropractic college teaching clinic Med Care 1987 25 516 27 3695660
Abbott A A System of professions: An essay on the division of expert labor 1988 Chicago, IL: University of Chicago Press 25
Goetzel RA Hawkins K Ozminkowski RJ Wang S The health and productivity cost burden of the top 10 physical and mental conditions affecting six large US employers JOEM 2003 45 5 14 12553174
accessed March 23, 2005
accessed March 23, 2005
Relman AS Assessment and accountability: the third revolution in medical care N Engl J Med 1988 319 1220 222 3173460
Sackett DL Rosenberg WM Gray JA Haynes RB Richardson WS Evidence based medicine: what it is and what it isn't Br Med J 1996 312 71 223 8555924
Daltroy LH Iversen MD Larson MG A controlled trial of an educational program to prevent low back injuries N Engl J Med 1997 337 322 8 9233870 10.1056/NEJM199707313370507
Autret-Leca E Giraudeau B Ployet MJ Jonville-Bera AP Amoxicillin/clavulanic acid is ineffective at preventing otitis media in children with presumed viral upper respiratory infection: a randomized, double-blind equivalence, placebo-controlled trial Br J Clin Pharmacol 2002 54 652 6 12492614 10.1046/j.1365-2125.2002.t01-6-01689.x
CAST II Investigators Effect of Antiarrhythimic Agent Moricizine on Survival after Myocardial Infaraction New Eng J Med 1992 327 227 33 1377359
Epstein AE Hallstrom AP Rogers WJ Mortality following ventricular arrhythmia suppression by encainide, flecainide, and moricizine after myocardial infarction. The original design concept of the Cardiac Arrhythmia Suppression Trial (CAST) JAMA 1993 270 2451 5 8230622 10.1001/jama.270.20.2451
Angell M Kassirer JP Alternative medicine – the risks of untested and unregulated remedies N Engl J Med 1998 339 839 41 9738094 10.1056/NEJM199809173391210
|
16000175
|
PMC1185558
|
CC BY
|
2021-01-04 16:38:23
|
no
|
Chiropr Osteopat. 2005 Jul 6; 13:9
|
utf-8
|
Chiropr Osteopat
| 2,005 |
10.1186/1746-1340-13-9
|
oa_comm
|
==== Front
J Inflamm (Lond)Journal of Inflammation (London, England)1476-9255BioMed Central London 1476-9255-2-61598751710.1186/1476-9255-2-6ResearchComparative effects of the herbal constituent parthenolide (Feverfew) on lipopolysaccharide-induced inflammatory gene expression in murine spleen and liver Smolinski Alexa T [email protected] James J [email protected] Department of Food Science and Human Nutrition, Michigan State University, East Lansing, Michigan, USA2 Institute for Environmental Toxicology, Michigan State University, East Lansing, Michigan, USA3 Department of Microbiology and Molecular Genetics, Michigan State University, East Lansing, Michigan, USA2005 29 6 2005 2 6 6 9 1 2005 29 6 2005 Copyright © 2005 Smolinski and Pestka; licensee BioMed Central Ltd.2005Smolinski and Pestka; 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
Parthenolide, a major sesquiterpene lactone present in extracts of the herb Feverfew, has been investigated for its inhibitory effects on mediators of inflammation, including the proinflammatory cytokines. Although parthenolide's anti-inflammatory effects have been investigated in vitro, little in vivo data are available. Moreover, the molecular mechanisms for these inhibitory effects are not fully understood. The objective of this study was to test the hypothesis that parthenolide suppresses lipopolysaccharide (LPS)-induced serum (interleukin) IL-6, tumor necrosis factor (TNF)-α, IL-1β and cyclooxygenase (COX)-2 expression in mice as indicated by reduced splenic and liver mRNA levels.
Methods
Mice were co-treated i.p. with LPS (1 mg/kg bw) and parthenolide (5 mg/kg bw) and blood, spleen and liver collected. Serum was analyzed for IL-6, TNF-α and IL-1β by ELISA. Total RNA was extracted from spleen and liver, and real-time RT-PCR was used to determine relative mRNA expression of IL-1β, IL-6, TNF-α and COX-2.
Results
LPS induced increases in serum IL-6 and TNF-α concentrations with only IL-6 being suppressed in parthenolide-treated mice. Induction of IL-6 mRNA was reduced, TNF-α and COX-2 mRNAs unchanged, and IL-1β mRNA increased in spleens of parthenolide plus LPS co-treated animals compared to LPS-only. No significant differences were observed in inflammatory gene expression between these two groups in liver samples. Overall, mRNA expression of each proinflammatory gene was much higher in spleen when compared to liver.
Conclusion
In summary, only one gene, IL-6, was modestly suppressed by parthenolide co-exposure which contrasts with many in vitro studies suggesting anti-inflammatory effects of this compound. Also, LPS evoked greater effects in spleen than liver on expression of proinflammatory genes. Further study of the effects of parthenolide and other herbal constituents on inflammatory gene expression using model animal systems as described here are critical to evaluating efficacy of such supplements as well as elucidating their mechanisms of action.
==== Body
Background
Parthenolide, the major sesquiterpene lactone derived from the feverfew extract (Tanacetum parthenium), has been studied for its inhibitory effects on inflammation in cell culture and, to a limited extent, in live animals. This constituent has been shown to attenuate a variety of inflammatory endpoints [1-12]. Recent attention has turned to the determination of the molecular mechanisms by which parthenolide imparts its effects on inflammatory responses.
Investigations of the anti-inflammatory properties of parthenolide, and feverfew have focused on suppression of primary inflammatory endpoints such as platelet aggregation [1] and carrageenan-induced mouse [2] and rat [3] paw edema. Additional studies have evaluated parthenolide's inhibitory effect on inflammatory mediators including activity and expression of cyclooxygenase (COX) [4,5], generation of prostaglandins [6,7], and leukotrienes (LT) [4] and expression of proinflammatory cytokines [5,8]. Most recently, the compound was found to inhibit activation of transcription factor nuclear factor (NF)-κB [9-12].
Previous research in our laboratory focused on the inhibitory effects of parthenolide on lipopolysaccharide (LPS)-induced proinflammatory cytokine production in the supernatant of murine cell culture and sera of animals [13]. The data showed that parthenolide impairs LPS-induced tumor necrosis factor (TNF)-α and interleukin (IL)-6 upregulation in culture and in sera of animals when parthenolide was administered via i.p. injection.
Although protein levels of LPS-induced proinflammatory cytokines are reportedly reduced by parthenolide treatment, there are limited data evaluating the effect of parthenolide on mRNA expression of these cytokines. Hwang et al. [5] showed that parthenolide suppresses LPS-induced steady state levels of TNF-α and IL-1β mRNA in cell culture. Parthenolide had no inhibitory effect on IL-6 mRNA levels in LPS-stimulated macrophages, but did attenuate IL-12 p40 and p35 mRNA expression [14] as well as the chemokine IL-8 in cultured human respiratory epithelium [15].
Parthenolide's effects on specific cytokine gene expression have been documented in vitro, but, to our knowledge, few data are available regarding effects on mRNA expression of cytokines or other inflammatory genes such as COX-2 in vivo. This is an important consideration because absorption, distribution and metabolism of this compound will likely impact how it affects inflammation in the host. The objective of this study was to test the hypothesis that parthenolide-induced suppression of serum LPS-induced IL-6 and TNF-α correlate with reduced mRNA levels for these genes, and other related proinflammatory genes, in the spleen and liver which are tissues well-known to express IL1β, IL-6, TNF-α and COX-2. Additionally, we sought to determine whether differences in expression levels of each gene existed between the spleen and liver. These organs contain macrophages and other cell types capable of responding to LPS and other inflammatory stimuli.
Methods
Chemicals
All chemicals were purchased from Sigma Chemical Co. (St. Louis, MO) unless otherwise noted. Parthenolide (Calbiochem, San Diego, CA) was dissolved in tissue culture grade dimethyl sulfoxide (DMSO). Lipopolysaccharide (LPS) from Salmonella typhimurium [1.5 EU/ng LPS; Stimulation index (SI) 3.6 @15.6 μg/ml LPS] was dissolved in endotoxin-free tissue culture grade water.
Experimental design
All animal handling was conducted in accordance with guidelines established by the National Institutes of Health. Experiments were designed to minimize the numbers of animals used. Female B6C3F1 mice (8–10 weeks) were obtained from Charles River (Portage, MI). Animals were housed 3 or 4 per cage with a 12 h light/dark cycle, provided standard rodent chow and water ad libitum, and acclimated to their environment at least one week before the start of experiments.
Chow and water were removed from cages one hour prior to the start of each experiment. Mice were co-treated with 5 mg/kg, i.p. (in 50 μl DMSO) parthenolide and 1 mg/kg, i.p. LPS (in 100 μl water). This parthenolide dose was selected based on solubility limitations and because it was the lowest dose to show consistent inhibition of IL-6 and TNF-α elevation in four preliminary studies using doses at 0.05, 0.5, 1, 5 and 10 mg/kg. Vehicle-treated mice received 50 μl DMSO, i.p. and 100 μl water, i.p. Parthenolide control animals received parthenolide 5 mg/kg, i.p. and 100 μl water, i.p. After 90 minutes, blood was collected by retro-orbital bleeding under methoxyflurane anesthesia. Animals were then immediately euthanized by cervical dislocation and spleen and liver were collected. The time interval was chosen based on preliminary studies of LPS induction in expression of the four target genes. This euthanasia method was chosen to minimize artifactual immunologic effects [16] and was approved by MSU All University Committee on Animal Research and Care.
Serum IL-6, TNF-α, and IL-1β determination by ELISA
Blood was allowed to clot overnight at 4°C. Serum was analyzed for IL-6, TNF-α and IL-1β by ELISA. IL-6 analysis was performed using purified and biotin-conjugated rat anti-mouse IL-6 antibodies from PharMingen (San Diego, CA) as described previously [13]. Streptavidin-peroxidase (Sigma) and 3,3',5,5'-tetramethylbenzidine (TMB, Fluka, Ronkonkoma, NY) were used for detection. Absorbance was read at 450 nm using a Vmax™ Kinetic Microplate Reader (Molecular Devices, Menlo Park, CA). For TNF-α analysis the OptEIA Set: Mouse TNF-α (Mono/Poly) kit was employed (PharMingen). For IL-1β analysis, a DuoSet® ELISA (R&D Systems, Minneapolis, MN) was used. The sensitivity of all three ELISAs was 20 pg/ml.
Total RNA extraction from spleen and liver
Spleens and livers were cut into small pieces and placed into TRIzol® Reagent (Invitrogen Life Technologies, Carlsburg, CA). Samples were homogenized for 30 seconds at setting 8 using a Polytron® Homogenizer (Brinkmann, Westbury, NY) and RNA extractions were completed according to manufacturer's instructions. Total RNA was quantified at 260 nm using a GeneQuant RNA/DNA Calculator (Pharmacia Biotech, Cambridge, England).
mRNA quantification from spleen and liver
Relative IL-6, TNF-α, IL-1β and COX-2 mRNA levels were determined according to manufacturer's instructions using TaqMan ® real-time reverse transcription (RT)- polymerase chain reaction (PCR), ABI Prism® 7700 Sequence Detection System (Applied Biosystems, Foster City, CA) and Applied Biosystems reagents unless indicated otherwise. The RT-PCR reaction was carried out in a total reaction volume of 25 μl containing: 1) RNase-free water (Sigma) to 25 μl; 2) 12.5 μl TaqMan® One-Step RT-PCR Master Mix Reagent; 3) 1.25 μl either IL-6, TNF-α or IL-1β Pre-Developed Assay Reagent (primer and probe sets); 4) 1.25 μl 18S rRNA Pre-Developed Assay Reagent; 5) 50 ng total RNA in RNase-free water and 6) 0.63 μl MultiScribe and RNase Inhibitor Mix. COX-2 mRNA was similarly analyzed using forward 5'-CAGAAC CGCATT GCCTCTG-3' and reverse 3'-AGCTGTACTCCTGGTCTTCAATGTT-5' primers (900 nM each) (Michigan State University Genomics Facility, East Lansing, MI) and probe 6FAM-CAACACACTCTATCACTGGCACCCCCTG-TAMRA (250 nM) designed using Primer Express™ software (Applera Corporation, Norwalk, CT). All samples were multiplexed with 18S rRNA which served as an endogenous reference for cytokine mRNA normalization. All samples were assayed in duplicate and serial dilutions of standard (total RNA from LPS-treated mouse spleen) in triplicate. No template control and no RT negative control reactions were also performed. Reaction conditions were: 48°C for 30 min; 95°C for 10 min; and 40 cycles of 95°C for 10 seconds and 60°C for 1 min.
Statistics
All statistical analyses were performed using SigmaStat Statistical Analysis Software (Jandel Scientific, San Rafael, CA). For comparison of two groups, a Student's t-test was used. For comparisons of multiple groups using parametric data, one-way analysis of variance (ANOVA) using Student-Newman-Keuls Method for all pairwise multiple comparisons was performed.
Results
Parthenolide co-treatment in vivo inhibits LPS-induced IL-6 protein production in serum
In order to determine the systemic effect of parthenolide co-treatment on LPS-induced IL-6 production, mice were treated with parthenolide (5 mg/kg, i.p.) and LPS (1 mg/kg, i.p.) for 90 minutes. Blood was collected and serum analyzed for IL-6. Animals treated with LPS alone produced 26 ± 2.6 ng/ml of IL-6 (Fig. 1). Serum concentrations of IL-6 were not detectable in vehicle and parthenolide control animals. Co-treatment with parthenolide caused a 35 percent reduction in LPS-induced IL-6 production compared to animals treated with LPS alone (P < 0.05).
Figure 1 IL-6 protein production in sera following parthenolide and LPS co-treatment. Female B6C3F1 mice were co-treated with parthenolide (5 mg/kg, i.p.) or 50 μl DMSO and LPS (1 mg/kg, i.p.) or 100 μl water. After 90 minutes, blood was collected and serum analyzed for IL-6 by ELISA. The letter (a) indicates a significant difference compared to vehicle and parthenolide controls; (b) indicates a significant difference compared to LPS. Data are mean ± SEM (n = 16, controls n = 4), and is a combination of 4 separate experiments.
Parthenolide co-treatment in vivo impairs LPS-induced IL-6 mRNA expression in spleen but not liver
Relative IL-6 mRNA expression in the spleen and liver of co-treated animals was also determined by real-time RT-PCR. IL-6 mRNA expression was significantly induced in spleen 239 ± 19-fold spleen control) and liver (117 ± 18-fold liver control) (Fig. 2). IL-6 expression in vehicle and parthenolide control animals was negligible in both spleen and liver samples. Splenic IL-6 mRNA levels of parthenolide and LPS co-treated animals (191 ± 12-fold) was 20 percent less than compared to spleens from LPS-only treated animals (p < 0.05), but not significantly different in liver (P < 0.05). Overall, IL-6 mRNA expression in spleen was 2.8-fold higher than the liver in LPS-treated animals, and was 1.4-fold higher in the spleen of animals receiving LPS plus parthenolide co-treatment compared to that of the liver.
Figure 2 IL-6 mRNA expression levels in spleen and liver following parthenolide and LPS co-treatment. Female B6C3F1 mice were co-treated with parthenolide (5 mg/kg, i.p.) or 50 μl DMSO and LPS (1 mg/kg, i.p.) or 100 μl water. Spleen and liver were collected after 90 minutes and total RNA was extracted and subjected to real-time, one-step RT-PCR using TaqMan primers and probes. IL-6 mRNA levels were normalized using 18S rRNA and related to spleen control values. (a) indicates a significant difference compared to vehicle and parthenolide controls; (b) indicates a significant difference compared to LPS. Data are mean ± SEM (n = 16, controls n = 4), and is a combination of 4 separate experiments.
Parthenolide co-treatment in vivo does not inhibit LPS-induced TNF-α protein production in serum
LPS-treated animals exhibited significantly increased TNF-α concentration (2.50 ± 0.27 ng/ml) in sera compared to both vehicle and parthenolide control animals (Fig. 3). TNF-α was not detectable in either control group. TNF-α concentrations in animals co-treated with parthenolide plus LPS (2.11 ± 0.26 ng/ml) were not significantly different from LPS-only treated animals (P < 0.05) although there was a downward trend.
Figure 3 TNF-α protein production in sera following parthenolide and LPS co-treatment. Mice were treated and sera analyzed for TNF-α as described in Fig. 1 legend. The letter (a) indicates a significant difference compared to vehicle and parthenolide controls. Data are mean ± SEM (n = 16, controls n = 4), and is a combination of 4 separate experiments.
Parthenolide co-treatment in vivo does not affect LPS-induced TNF-α mRNA in spleen and liver
TNF-α mRNA was also increased in the spleen (4.84 ± 1.4-fold) and liver (2.33 ± 0.71-fold) of LPS-treated animals over controls. In both the spleen and liver there were no differences in TNF-α mRNA expression between LPS-treated and LPS plus parthenolide-treated mice (Fig. 4) (P < 0.05). In fact, there was no statistical differences among any of the groups evaluated in this study (P < 0.05). LPS-induced splenic TNF-α mRNA levels were considerably higher (14-fold) than those in the liver.
Figure 4 TNF-α mRNA expression levels in spleen and liver following parthenolide and LPS co-treatment. Mice were treated and analyzed for TNF-α mRNA described in Fig. 2 legend. TNF-α mRNA levels were normalized using 18S rRNA and related to spleen control values. Data are mean ± SEM (n = 16, controls n = 4), and is a combination of 4 separate experiments.
Parthenolide co-treatment in vivo elevates LPS-induced IL-1β mRNA in spleen but not liver
Serum IL-1β was not detectable in any of the groups tested. However, IL-1β mRNA expression was significantly elevated in spleen (16.01 ± 1.45-fold) and liver (62.2 ± 7.43-fold) of LPS-only treated animals in comparison to vehicle and parthenolide control animals (Fig. 5). The level of IL-1β mRNA in spleen and liver of vehicle and parthenolide control animals was negligible. In the spleens of co-treated animals, there was a significant (p < 0.05) increase in IL-1β mRNA (21.48 ± 6.91-fold) compared to LPS-only treated animals. IL-1β mRNA expression was 3.2-fold higher in spleen of LPS-only treated animals, and 4.5-fold higher in LPS plus parthenolide co-treated animals, when compared to expression levels of the liver.
Figure 5 IL-1β mRNA expression levels in spleen and liver following parthenolide and LPS co-treatment. Mice were treated and analyzed for IL-1β mRNA as described in Fig. 2 legend. The letter (a) indicates a significant difference compared to vehicle and parthenolide controls; (b) indicates a significant difference compared to LPS. Data are mean ± SEM (n = 16, controls n = 4), and is a combination of 4 separate experiments.
Parthenolide co-treatment in vivo does not affect LPS-induced COX-2 mRNA in spleen and liver
Relative COX-2 mRNA expression was assessed in the spleen and liver of parthenolide plus LPS co-treated animals. COX-2 mRNA expression was markedly induced in the spleen of LPS treated animals (44.8 ± 5.0-fold over control) and increased to a lesser extent in liver (4.6 ± 0.9) (Fig. 6). In both spleen and liver samples there were no significant differences in COX-2 mRNA expression between LPS-treated and LPS plus parthenolide-treated mice (P < 0.05). Overall, COX-2 mRNA expression levels were 15.6- and 14.2-fold higher in spleen of LPS-treated mice and parthenolide plus LPS-treated mice, respectively, when compared to expression levels observed in liver.
Figure 6 COX-2 mRNA expression levels in spleen and liver following parthenolide and LPS co-treatment. Mice were treated and COX-2 mRNAs analyzed as described in Fig. 2 legend. The letter (a) indicates a significant difference compared to vehicle and parthenolide controls. Data are mean ± SEM (n = 16, controls n = 4), and is a combination of 4 separate experiments.
Discussion
Parthenolide has been demonstrated to inhibit inflammatory gene expression in vitro (4–8, 13). The results of this study are important because it is the first, to our knowledge, to evaluate parthenolide's effect on inflammatory gene expression in two primary sites of the LPS response – the spleen and the liver. The data indicate that protein concentrations in serum followed a similar trend to splenic mRNA accumulation for IL-6 in LPS and parthenolide/LPS co-treated mice. However protein and splenic mRNA levels were not consistent with liver samples. The mRNA levels of each inflammation-related gene in the liver was not changed, irrespective of parthenolide co-treatment, when compared to LPS alone.
There is a marked contrast between the robust attenuation by parthenolide of proinflammatory gene expression reported in vitro (4–12, 15) and the small responses reported in animals here and previously (13). It is possible that in the whole animal the kinetics absorption, metabolism and distribution and clearance of parthenolide preclude sufficient contact time in target immune tissue to evoke potent attenuation of LPS response. These factors must be considered when attempting to extrapolate an in vitro effect of an herbal compound to the in vivo situation.
Serum IL-6, but not TNF-α, was significantly reduced following co-treatment with parthenolide (5 mg/kg, i.p.) and LPS (1 mg/kg, i.p.) compared to animals receiving LPS alone. Similarly, IL-6 mRNA concentration in the spleens of co-treated animals was significantly reduced, whereas splenic TNF-α mRNA, and COX-2, were not changed as compared to LPS- treated animals. In contrast, the level of IL-1β mRNA in the spleen was significantly elevated in co-treated mice but no effects were observed in the liver. Serum IL-1β were also evaluated, but levels were below the limit of detection in all treatment groups (data not shown). Absence of serum IL-1β might result from delay in translation/secretion of the protein after transcription, receptor binding or degradation. Regardless of the cause, comparisons could not be made between protein production and mRNA expression for IL-1β in this study.
The observed mRNA expression levels of the proinflammatory cytokines and COX-2 might be higher in spleen than liver because of inherent phagocytic capacities of the macrophage cell populations within each organ. In the spleen, macrophages play key roles in phagocytosis, especially of nonopsonized particles, whereas the macrophage of the liver, Kupffer cells, play a major role in the removal of opsonized particles [17]. Soluble LPS injected i.p., might not become opsonized since it is not a particulate and therefore may be preferentially processed in the spleen rather than liver. This may account, at least in part, for the increased cytokine and COX-2 gene expression observed in the spleen when compared to the liver.
Specific cell populations of spleen and liver are likely to contribute to observed cytokine expression. Hepatocytes, the parenchyma cells of the liver, account for 60% of total liver cells and 80% of the liver's volume [18]. The primary functions of these cells are exocrine and metabolic in nature. Although they are capable of functioning as antigen-presenting cells in certain situations, they are not primary mediators of immune regulation in the liver [19]. Other, nonparenchymal cells of the liver include Kupffer cells, the resident macrophage, and interstitial dendritic cell types. Both Kupffer and dendritic cells are capable of producing proinflammatory cytokines. The macrophage population of liver (10%) is more than three times larger than the spleen (3%) [20]. However, the opposite is observed with respect to dendritic cell populations. In the spleen, the dendritic cell population is approximately ten times larger compared to liver [20]. It is possible that dendritic cells, which are constitutively activated, might respond more readily to antigen exposure than macrophage cell populations, and as a result, express proinflammatory cytokines to a greater extent [21]. These and other cell types including endothelial and epithelial cells might also contribute differences in spleen and liver mRNA expression. Future examination of macrophage responses at other tissue sites and responses of other cell types is clearly warranted.
Parthenolide's in vitro effects on mediators of inflammation including cytokines (TNF-α, IL-1β and IL-6) [5,13], chemokine (IL-8) [15] and lipid mediators (prostaglandins [6,7], COX [4,5] and leukotrienes [4]) have been extensively studied. Recent research has focused on the role of transcription factor NF-κB [8,10-12]. Notably, transcriptional regulation of cytokine genes including TNF-α [22], IL-6 and IL-8 [23] has been strongly linked to NF-κB activation. Interestingly, parthenolide has been shown to inhibit expression of each of these cytokines [5,13], as well as activation of NF-κB [8,10-12] in cell culture studies. Parthenolide appears to inhibit NF-κB by targeting the IκB (inhibitor of NF-κB) kinase complex [9] which might inhibit proinflammatory cytokine and chemokine gene expression. Relative to other transcription factors like CCAAT/enhancer binding protein (C/EBP)β, NF-κB plays a dominant role in regulation of IL-6 expression in other models of inflammation [24].
In contrast to the observed effects of LPS and parthenolide co-treatment on IL-6 production and gene expression, no inhibitory effect was observed for TNF-α. Although NF-κB has been implicated in the transcriptional regulation of TNF-α, the functional concert of NF-κB with other transcription factors such as activator protein (AP)-1 [25] and C/EBPβ (reviewed by [26]) may override the importance of NF-κB in LPS-induced TNF-α expression. The dual pathway of NF-κB and AP-1 has been shown to enhance production of some proinflammatory cytokines, notably TNF-α [27]. Parthenolide inhibits NF-κB, but has no effect on AP-1 [11]. Therefore, the expression of TNF-α may be compensated for by transcriptional activation by AP-1.
Similar to the effects observed for TNF-α mRNA expression, there were no significant changes in COX-2 mRNA expression of LPS versus LPS plus parthenolide co-treated animals. Hwang et al. [5] demonstrated the inhibitory effects of parthenolide on LPS-induced COX-2 protein and mRNA, however, those studies employed cultured alveolar macrophage cells rather than an in vivo model as described here. No other studies have directly evaluated the effect of parthenolide on COX-2 mRNA. The COX-2 gene is regulated by a number of transcription factors including NF-κB, C/EBPβ and AP-1 as well as cAMP response element-binding protein (CREB) and others. Site-directed mutagenesis studies of basal COX-2 expression in murine lung tumor derived cell lines highlight the role of C/EBPβ and CREB as major transcriptional regulators of COX-2 [28], whereas NF-κB appeared to have no role in COX-2 transcriptional regulation using this model. Thus the lack of inhibitory effect on COX-2 mRNA expression might be explained, in part, by the limited role of NF-κB in COX-2 transcriptional regulation.
IL-1β mRNA expression followed a different pattern than the other two cytokines. IL-6 was decreased and TNF-α was unchanged, while IL-1β levels were increased following co-treatment with LPS and parthenolide. Although IL-1β is also transcriptionally regulated by NF-κB [22,25] and C/EBPβ (reviewed by [29]), similar to IL-6 and TNF-α, it might be differentially regulated in response to LPS. In support of this hypothesis, in vivo studies by Zhou et al. [30] show that mRNA levels of IL-1β in the spleen are not affected under conditions of LPS tolerance whereas both TNF-α and IL-6 are reduced.
Conclusion
In summary, parthenolide selectively modulated proinflammatory cytokine gene expression in vivo. Only one gene, IL-6, was modestly suppressed which contrasts with many in vitro studies suggesting anti-inflammatory effects of this compound. It is possible that the differences in metabolism and/or distribution of parthenolide could explain contrasting in vivo and in vitro results. LPS also exerted greater effects in spleen than liver on expression of proinflammatory genes. Higher doses of parthenolide might have greater effects on IL-6 but there are solubility issues relative to delivery. In addition, the physiological significance of greater doses would be questionable. Further study of the effects of parthenolide and other herbal constituents on inflammatory gene expression using model animal systems as described here are critical to evaluating efficacy of such supplements as well as elucidating their mechanisms of action.
List of abbreviations
alpha, α; hour, h; LPS, lipopolysaccharide; TNF, tumor necrosis factor; IL, interleukin; COX, cyclooxygenase; PG, prostaglandin; DMSO, dimethyl sulfoxide; CREB, cAMP response element-binding protein; C/EBPβ, CCAAT/enhancer binding protein beta; NF-κB, nuclear factor kappa B
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
ATS participated in study design, carried out experiments, performed data analysis and drafted the manuscript. JJP participated in study design and coordination, as well as editing and revision of the final manuscript.
Acknowledgements
This work was supported by Public Health Service Grants E09521, ES03553 and DK058833 from the National Institutes for Health. We would like to thank H.R. Zhou and A. Thelen for technical advice and Mary Rosner for assistance with manuscript preparation. The authors thank Kate Brackney and Dan Lampen for assistance with animal experiments.
==== Refs
Groenewegen WA Heptinstall S A comparison of the effects of an extract of feverfew and parthenolide, a component of feverfew, on human platelet activity in-vitro J Pharm Pharmacol 1990 42 553 557 1981582
Schinella GR Giner RM Recio MC Mordujovich dB Rios JL Manez S Anti-inflammatory effects of South American Tanacetum vulgare J Pharm Pharmacol 1998 50 1069 1074 9811170
Jain NK Kulkarni SK Antinociceptive and anti-inflammatory effects of Tanacetum parthenium L. extract in mice and rats J Ethnopharmacol 1999 68 251 259 10624885 10.1016/S0378-8741(99)00115-4
Sumner H Salan U Knight DW Hoult JR Inhibition of 5-lipoxygenase and cyclooxygenase in leukocytes by feverfew. Involvement of sesquiterpene lactones and other components Biochem Pharmacol 1992 43 2313 2320 1319159 10.1016/0006-2952(92)90308-6
Hwang D Fischer NH Jang BC Tak H Kim JK Lee W Inhibition of the expression of inducible cyclooxygenase and proinflammatory cytokines by sesquiterpene lactones in macrophages correlates with the inhibition of MAP kinases Biochem Biophys Res Commun 1996 226 810 8 8831694 10.1006/bbrc.1996.1433
O'Neill LA Barrett ML Lewis GP Extracts of feverfew inhibit mitogen-induced human peripheral blood mononuclear cell proliferation and cytokine mediated responses: a cytotoxic effect Br J Clin Pharmacol 1987 23 81 83 3493021
Pugh WJ Sambo K Prostaglandin synthetase inhibitors in feverfew J Pharm Pharmacol 1988 40 743 745 2907548
Uchi H Arrighi JF Aubry JP Furue M Hauser C The sesquiterpene lactone parthenolide inhibits LPS- but not TNF-alpha-induced maturation of human monocyte-derived dendritic cells by inhibition of the p38 mitogen-activated protein kinase pathway J Allergy Clin Immunol 2002 110 269 276 12170268 10.1067/mai.2002.126381
Hehner SP Hofmann TG Droge W Schmitz ML The antiinflammatory sesquiterpene lactone parthenolide inhibits NF- kappa B by targeting the I kappa B kinase complex J Immunol 1999 163 5617 23 10553091
Hehner SP Heinrich M Bork PM Vogt M Ratter F Lehmann V Sesquiterpene lactones specifically inhibit activation of NF-kappa B by preventing the degradation of I kappa B-alpha and I kappa B-beta J Biol Chem 1998 273 1288 1297 9430659 10.1074/jbc.273.3.1288
Bork PM Schmitz ML Kuhnt M Escher C Heinrich M Sesquiterpene lactone containing Mexican Indian medicinal plants and pure sesquiterpene lactones as potent inhibitors of transcription factor NF-kappaB FEBS Lett 1997 402 85 90 9013864 10.1016/S0014-5793(96)01502-5
Rungeler P Castro V Mora G Goren N Vichnewski W Pahl HL Inhibition of transcription factor NF-kappaB by sesquiterpene lactones: a proposed molecular mechanism of action Bioorg Med Chem 1999 7 2343 2352 10632044 10.1016/S0968-0896(99)00195-9
Smolinski AT Pestka JJ Modulation of proinflammatory cytokine production in vitro and in vivo by the herbal constituents apigenin (chamomile), ginsenoside Rb1 (ginseng) and parthenolide (feverfew) Food Chem Toxicol 2005 41 1381 1390 12909272 10.1016/S0278-6915(03)00146-7
Kang BY Chung SW Kim TS Inhibition of interleukin-12 production in lipopolysaccharide-activated mouse macrophages by parthenolide, a predominant sesquiterpene lactone in Tanacetum parthenium: involvement of nuclear factor-kappaB Immunol Lett 2001 77 159 163 11410248 10.1016/S0165-2478(01)00211-5
Mazor RL Menendez IY Ryan MA Fiedler MA Wong HR Sesquiterpene lactones are potent inhibitors of interleukin 8 gene expression in cultured human respiratory epithelium Cytokine 2000 12 239 45 10704251 10.1006/cyto.1999.0526
Howard HL McLaughlin-Taylor E Hill RL The effect of mouse euthanasia technique on subsequent lymphocyte proliferation and cell mediated lympholysis assays Lab Anim Sci 1990 40 510 514 2170752
Schuurman H Krajnc-Franken M Kuper C van Loveren H Vos J Haschek W, Rousseaux C Immune System Fundamentals of Toxicologic Pathology 1998 San Diego: Academic Press 233 272
Popp J Cattely R Haschek W, Rousseaux C Hepatobiliary System Fundamentals of Toxicologic Pathology 1998 San Diego: Academic Press 127 151
Lau AH Thomson AW Dendritic cells and immune regulation in the liver Gut 2003 52 307 314 12524419 10.1136/gut.52.2.307
Zhang Y Shlomchik WD Joe G Louboutin JP Zhu J Rivera A APCs in the liver and spleen recruit activated allogeneic CD8+ T cells to elicit hepatic graft-versus-host disease J Immunol 2002 169 7111 7118 12471148
Banchereau J Steinman RM Dendritic cells and the control of immunity Nature 1998 392 245 252 9521319 10.1038/32588
Mercurio F Manning AM Multiple signals converging on NF-kappaB Curr Opin Cell Biol 1999 11 226 32 10209157 10.1016/S0955-0674(99)80030-1
Baldwin AS Jr The NF-kappa B and I kappa B proteins: new discoveries and insights Annu Rev Immunol 1996 14 649 83 8717528 10.1146/annurev.immunol.14.1.649
Baeuerle PA Baichwal VR NF-kappa B as a frequent target for immunosuppressive and anti-inflammatory molecules Adv Immunol 1997 65 111 137 9238509
Tak PP Firestein GS NF-kappaB: a key role in inflammatory diseases J Clin Invest 2001 107 7 11 11134171
Poli V The role of C/EBP isoforms in the control of inflammatory and native immunity functions J Biol Chem 1998 273 29279 29282 9792624 10.1074/jbc.273.45.29279
Yokoo T Kitamura M Dual regulation of IL-1 beta-mediated matrix metalloproteinase-9 expression in mesangial cells by NF-kappa B and AP-1 Am J Physiol 1996 270 F123 F130 8769830
Wardlaw SA Zhang N Belinsky SA Transcriptional regulation of basal cyclooxygenase-2 expression in murine lung tumor-derived cell lines by CCAAT/enhancer-binding protein and activating transcription factor/cAMP response element-binding protein Mol Pharmacol 2002 62 326 333 12130685 10.1124/mol.62.2.326
Wedel A Ziegler-Heitbrock HW The C/EBP family of transcription factors Immunobiology 1995 193 171 85 8530141
Zhou HR Islam Z Pestka JJ Kinetics of lipopolysaccharide-induced transcription factor activation/inactivation and relation to proinflammatory gene expression in the murine spleen Toxicol Appl Pharmacol 2003 187 147 161 12662898 10.1016/S0041-008X(02)00077-7
|
15987517
|
PMC1185559
|
CC BY
|
2021-01-04 16:36:23
|
no
|
J Inflamm (Lond). 2005 Jun 29; 2:6
|
utf-8
|
J Inflamm (Lond)
| 2,005 |
10.1186/1476-9255-2-6
|
oa_comm
|
==== Front
J Neuroengineering RehabilJournal of NeuroEngineering and Rehabilitation1743-0003BioMed Central London 1743-0003-2-191603365010.1186/1743-0003-2-19CommentaryGait variability: methods, modeling and meaning Hausdorff Jeffrey M [email protected] Laboratory for Gait & Neurodynamics, Movement Disorders Unit, Department of Neurology, Tel-Aviv Sourasky Medical Center, Tel-Aviv, Israel2 Department of Physical Therapy, Sackler School of Medicine, Tel-Aviv University, Tel-Aviv, Israel3 Division on Aging, Harvard Medical School, Boston, MA, USA2005 20 7 2005 2 19 19 7 7 2005 20 7 2005 Copyright © 2005 Hausdorff; licensee BioMed Central Ltd.2005Hausdorff; 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 gait variability, the stride-to-stride fluctuations in walking, offers a complementary way of quantifying locomotion and its changes with aging and disease as well as a means of monitoring the effects of therapeutic interventions and rehabilitation. Previous work has suggested that measures of gait variability may be more closely related to falls, a serious consequence of many gait disorders, than are measures based on the mean values of other walking parameters. The Current JNER series presents nine reports on the results of recent investigations into gait variability. One novel method for collecting unconstrained, ambulatory data is reviewed, and a primer on analysis methods is presented along with a heuristic approach to summarizing variability measures. In addition, the first studies of gait variability in animal models of neurodegenerative disease are described, as is a mathematical model of human walking that characterizes certain complex (multifractal) features of the motor control's pattern generator. Another investigation demonstrates that, whereas both healthy older controls and patients with a higher-level gait disorder walk more slowly in reduced lighting, only the latter's stride variability increases. Studies of the effects of dual tasks suggest that the regulation of the stride-to-stride fluctuations in stride width and stride time may be influenced by attention loading and may require cognitive input. Finally, a report of gait variability in over 500 subjects, probably the largest study of this kind, suggests how step width variability may relate to fall risk. Together, these studies provide new insights into the factors that regulate the stride-to-stride fluctuations in walking and pave the way for expanded research into the control of gait and the practical application of measures of gait variability in the clinical setting.
agingcognitive functiondual taskingfall riskfractalsmodelingParkinson's disease
==== Body
Introduction
Like most physiologic signals, measures of gait are not constants but rather fluctuate with time and change from one stride to the next, even when environmental and external conditions are fixed (Figure 1). In healthy adults, these stride-to-stride fluctuations are relatively small and the coefficient of variation of many gait parameters (e.g., gait speed, stride time) is on the order of just a few percent [1-3], testimony to the accuracy and reliability of the fine-tuned systems that regulate gait. Recently, the apparently "noisy" variations in stride length, stride time and gait speed have also been shown to display a hidden and unexpected fractal-like property [4-9]. These properties of gait exhibit long-range (power-law) correlations and a "memory" effect, such that fluctuations at any given moment are statistically related to those that occur over many different time scales. When the systems regulating gait are disturbed (e.g., as a result of certain diseases), movement control may be impaired leading to increased stride-to-stride fluctuations and/or alterations in their multiscale dynamics.
Figure 1 Example of the stride-to-stride fluctuations in the stride time as measured in two older adults: an older adult non-faller and an idiopathic faller. In both subjects, the stride time changes from one stride to the next. Although the mean values of the stride time are essentially identical in both subjects, the magnitude of the stride-to-stride fluctuations is much larger in the faller. SD: standard deviation; CV: coefficient of variation.
The current series of the Journal of NeuroEngineering and Rehabilitation (JNER) is dedicated to gait variability. As guest editor of a collection of nine papers on this topic, I have had the opportunity to preview the wealth of information on stride-to-stride fluctuations in gait and the manifold ways in which gait variability may be analyzed. The articles in this collection cover a wide spectrum of themes ranging from methods for evaluating gait variability, animal and mathematical models investigating the factors that influence the variability of gait, and evaluations of the clinical utility of such measures. Altogether, these reports underscore the complex and fascinating nature of gait variability.
To set the stage, it is helpful to briefly highlight previous work in this area. Earlier studies have demonstrated that:
• Gait variability is a quantifiable feature of walking that is altered (both in terms of magnitude and dynamics) in clinically relevant syndromes, such as falling, frailty, and neuro-degenerative disease (e.g., Parkinson's and Alzheimer's disease [10-19].
• The magnitude of the stride-to-stride fluctuations in stride length and step timing are unaltered in healthy older adults, whereas the dynamics of gait change with healthy aging (e.g., alterations in the fractal pattern) [1,20,21].
• Physiologic factors that affect gait dynamics include neural control, muscle function and postural control; however, more subtle alterations in underlying physiology including cardiovascular changes and mental health may also influence the variability of gait (Figure 2) [10-12,16,19,22-24].
Figure 2 Simplified block diagram of the locomotor system. Also shown are a sample of the alterations that occur in aging and disease which affect gait stability, at least as reflected in stride time variability, and fall risk. CBF: cerebral blood flow. Modified from Hausdorff et al, J Appl Physiol 2001.
• Improvements in muscle function and therapeutic interventions are associated with enhanced gait stability, but not always with more conventional measures of average gait velocity or cadence [12,16,25].
• Gait instability measures apparently predict falls in idiopathic elderly fallers and other populations who share an increased fall risk [2,16,17,19,26-30].
Thus, gait variability may serve as a sensitive and clinically relevant parameter in the evaluation of mobility, fall risk and the response to therapeutic interventions.
Gait variability: a marker of fall risk
Studies of gait variability have been motivated by a number of factors. One intriguing aspect of gait variability is its relationship to fall risk. In one of the first quantitative studies of gait variability, Guimaraes and Isaacs [31] suggested that elderly fallers walked with increased gait variability, both in terms of step length and step time, compared to non-falling older adults. Indeed, one of the "holy grails" of geriatric and rehabilitation research is the identification of markers that can be used to prospectively identify older adults at greatest risk of falling. A number of studies have demonstrated that measures of gait variability may be help achieve this end [26,27,29]. Indeed, survival analyses have also shown that subjects are significantly more likely to fall sooner if gait instability measures are relatively increased at baseline, further underscoring the potential utility of such measures.
The nature of the relationships among the average gait speed, the average stride length, and the variability of these measures are critical to the study of fall risk. Although a reduced gait speed has often been viewed as a sign of fall risk, Maki showed that, at least among certain older adults, average gait speed and related measures are related to fear of falling, but not to the risk of falling per se, while measures of variability predict future falls [27]. A number of other investigations demonstrated that the degree of variability may be more closely related to fall risk than average gait speed, average stride length, and average stride time [2,26-29]. These results suggest that measures of gait variability may sometimes be more sensitive than other measures of gait, and that these measures may provide a clinical index of gait instability and fall risk. If one views gait variability as a reflection of the inconsistency in the central neuromuscular control system's ability to regulate gait and maintain a steady walking pattern, then it makes sense that measures of gait variability would be associated with instability and fall risk. A more variable gait in which the center of pressure moves over and beyond the base-of-support in a relatively uncontrolled, unstable fashion may predispose to unsteadiness and falls.
Similarly, it is important to stress that just as the assessment of the magnitude of gait variability may provide important, independent information above and beyond average values, so, too, may the investigation of the dynamics of gait variability offer additional insights. A number of studies have demonstrated this concept. Here, we briefly describe one example in which going beyond the first (the mean) and second (the standard deviation) moments proves relevant to the understanding of a disorder.
The cause of impaired gait among many older adults defies identification, even after thorough examination. This has been termed a "higher-level gait disorder" (HLGD) or "cautious gait" [32,33]. A study of the gait dynamics of these patients found that they had significantly larger (p < .0001) gait variability (the 2nd moment) compared to controls [19] and that about 50% of them reported falling. A fractal scaling index of gait was useful in discriminating fallers from non-fallers in this patient group, while all other measures (of muscle function, balance, and gait, including gait speed and stride time variability) did not [19]. These findings illustrate how going beyond conventional statistical summaries may improve discriminatory power and provide a more complete characterization of gait changes.
In the present JNER series, Brach and colleagues study the 2nd moment to quantify the magnitude of stride-to-stride fluctuations and examine the relationship between gait variability and fall history in a population-based sample of more than 500 older adults. In what is probably the largest quantitative study on this question to date, too much or too little step width variability was associated with a fall history in a relatively healthy cohort of older adults who do not walk slowly (i.e., gait speed ≥1.0 m/sec). These findings raise a number of interesting questions about the relationship between variability and fall risk, and encourage the study of specific aspects of variability and their inter-relationships (e.g., step length vs. step width).
Gait variability and heart rate variability
The strides in knowledge gleaned from studies of other physiologic systems, particularly those on heart rate variability, have also provided valuable incentive to similarly investigate gait variability [4,34-43] (see also ). The healthy heartbeat was originally thought to be quite regular and periodic, essentially the product of a single, metronomic pacemaker. Thus, for a long time, mean heart rate was regarded as the primary outcome, and fluctuations about the mean were largely ignored. It emerged from later studies, however, that the heart rate normally fluctuates, over many time scales, in a complex manner from one beat to the next [37,44]. In fact, the cardiovascular system shows erratic beat-to-beat fluctuations resembling those found in dynamical systems that are being driven away from a single equilibrium state, even under entirely healthy, resting conditions. A large body of investigations have demonstrated that there is important information hidden in the dynamics of the heart rate that can be detected using methods that examine the variability, scaling and multi-scaling properties of the heartbeat [4,39,45]. Moreover, numerous investigations have demonstrated the clinical utility of heart rate variability measures with important diagnostic and prognostic utility including the prediction of life threatening arrhythmias and mortality [46-53].
While there are obvious fundamental differences between the regulation of heart rate and the regulation of gait, the success of research into the former has spurred dynamical investigations of the latter. In the past, the fluctuations in gait were largely ignored or erroneously viewed simply as "noise". Many of the tools for quantifying heart rate variability were applied to study the stride-to-stride fluctuations in gait [5,6,8,13,19,35,36,54-56]. Of course, while both signals do share many of the same characteristics, there are several important differences: for example, increased stride time variability (i.e., the magnitude of the fluctuations) is usually a sign of pathology, while increased heart rate variability is a healthy sign. On the other hand, many of the dynamic properties of both signals are similar: heart rate and gait timing exhibit complex fluctuations reminiscent of fractals, and this property is typically altered with aging and certain diseases [4,9,19,20,47,48,54,57-59]. Challenging reports to the contrary [60], in the current series, the findings of West and Latka suggest that gait fluctuations, like the healthy heart rate, are also multi-fractal.
The parallel between gait fluctuations and heart rate variability should be considered with some caution. It would be remiss to investigate heart rate variability and not examine the average heart rate. Similarly, it would be deficient to study gait variability and disregard mean values of stride time, stride length and gait speed. These measures offer an excellent, initial description of a person's mobility and gait [61]. The lesson from the study of heart rate is that additional information can be uncovered by examining the fluctuations around the means, both in terms of the magnitude and the dynamics. The experience with heart rate also poses a challenge: pharmacologic and intervention studies have clearly identified key components that underlie the fluctuations in heart rate (e.g., the interplay between the parasympathetic and sympathetic systems). Equally fundamental studies are needed to more completely understand the physiology and patho-physiology that underlie gait variability and its dynamics.
Methods: data acquisition and signal processing
Data acquisition and signal processing are two key areas that enable the study of gait variability. Traditional camera-based, motion analysis limits the study to a few strides and is not optimal for measuring the stride-to-stride fluctuations. A number of methods have been used to study gait under ambulatory conditions, including accelerometers, gyroscopes, foot switches, body-worn sensors and wearable computers, gait mats, and force-plate mounted treadmills or optical measurement of treadmill walking [27-29,54,62-70]. In the present series of JNER papers, Terrier and Schutz review the use of global position satellite monitoring for measuring gait. Although its time may not yet have come for routine use, this method has some important benefits, such as allowing for the determination of both the spatial and temporal measures of gait on a stride-to-stride basis.
Once the signal is acquired, questions about signal processing inevitably follow. Chau and colleagues describe challenges that arise when analyzing gait variability and present an interesting strategy for dealing with them. Their excellent review introduces the reader to different sources of variability and provides a heuristic method for summarizing various types of variability measures.
Modeling of gait variability
A number of approaches may be applied to make sense of the various measures of gait variability. In this JNER series, Amende and colleagues report on the dynamics of gait in mouse models of Parkinson's disease, Huntington's disease and amyotrophic lateral sclerosis. In this first-ever study of the stride-to-stride fluctuations of gait in animal models of neurodegenerative processes, they demonstrate that gait changes parallel those seen in clinical studies of humans (check out the gait of these animals in on-line video). This finding supports the validity of these models and sets the stage for a novel means of studying gait dynamics. While there are of course critical differences between two and four legged locomotion, these animal models enable manipulation and invasive intervention that are not feasible in human studies, thus offering a way to identify the mechanisms that underlie changes in the stride-to-stride regulation of gait.
West and Latka take a different, complementary approach toward understanding the fluctuations in gait. Using mathematical methods, they build upon earlier nonlinear dynamics models of the fluctuations in the stride time [56,71,72] and demonstrate that these fluctuations in healthy subjects can be described using a fractional Langevin equation. It remains to be seen whether this model can be applied to data collected in animal models and how disease and aging alter model parameters.
Gait variability, cognitive function, meaning and more
Another approach taken to gain insight into the factors that influence gait variability is to manipulate the locomotor system or specific components of the system by means of clinical studies. A priori, one might argue that stride-to-stride variability is regulated by automated processes and requires minimal cognitive resources. This argument is consistent with the report of Maki [27], demonstrating that variability was related to fall risk, but not to fear of falling. Indeed, studies of dual tasking found that gait speed slowed when healthy subjects, young and old, performed a secondary dual task during walking, while the variability of stride and swing timing was unchanged, even when subjects simultaneously walked and subtracted 7's serially, a challenging cognitive task [73,74]. In contrast, dual tasking not only reduced gait speed, it also increased variability among patients with impaired automaticity (e.g., Parkinson's disease patients) [17,73-75]. These findings are in line with the view that the regulation of variability is normally automated and requires minimal cognitive input. However, when automaticity is impaired (e.g., in the presence of pathology, cognitive tasks affect gait variability. One recent investigation disputed the concept of automatic regulation and suggested that stride time variability is related to specific cognitive processes, namely executive function [76]. In the present series, papers by Beauchet and colleagues and Grabiner and Troy describe the effects of a secondary, dual task on the gait variability in healthy young adults. One study suggests that there is no effect on stride length variability, while there is a small increase in stride time variability due to changes in mean gait speed. The second paper suggests that stride width variability becomes reduced during dual tasking. These interesting findings raise the question: "why?" and call for a more all-embracing understanding of the mechanisms that control gait variability and a "smooth" gait.
When dealing with this question, the complex relationships between gait speed and measures of variability of gait should be considered. When all other variables are kept constant, studies in young adults have demonstrated a U-shaped relationship between stride length (speed and/or cadence) and measures of gait variability. Minimal variability occurred near the usual walking speed and cadence [77-79], where energy costs of walking are also minimal and head stability is maximal [80,81]. Thus, when investigating gait and the factors that influence variability, it is important to take into account the possibility that any observed group differences or responses to intervention are simply a result of changes in gait speed. In many cases, however, it is possible to demonstrate that variability parameters are regulated independently of mean values (e.g., of stride length and stride time) [78]. For example, in the present series, Kessler and colleagues show that healthy controls and patients with a HLGD reduce their stride length and walk more slowly when they are asked to walk in conditions of minimal lighting. While variability measures increased among the patients, control subjects evidenced no change, even though they did walk more slowly in near darkness.
A potential way of separating values of variability from those of mean stride length and speed is described by Frenkel-Toledo and colleagues in the present series. They show that swing time variability is larger in patients with Parkinson's disease compared to healthy controls and that swing time variability is insensitive to changes in gait speed in both groups. Perhaps this measure can be used as a speed-independent measure of variability to help to unravel the mechanisms that influence the stride-to-stride fluctuations of gait and to identify measures with clinical utility that are not influenced by gait speed. Interestingly, a recent report observed a dissociation between left and right swing time variability in patients with Parkinson's disease who have a severe gait disturbance, i.e., freezing of gait [82], further demonstrating the potential utility of measures of swing time variability.
Outstanding issues
The investigations reported in this special series on gait variability advance our understanding of an intriguing aspect of gait: the ability of the healthy neural control system to fine tune the stride-to-stride fluctuations of walking to a remarkable degree. At the same time, they delineate a number of important questions that remain to be resolved by future studies. For example, several reports highlight the differences between measures of the mean, the variance and the dynamics. A theoretical framework is needed to understand these differences. One possible explanation for the difference between the results of the study by Brach and colleagues and those of previous studies relates to the question: how much is enough? In order to obtain reliable and meaningful measures of variability, how many strides need to be studied? Owings and Grabiner [64] suggest that hundreds of strides are required to accurately estimate step kinematic variability for treadmill walking. The number needed for walking on level ground is undetermined. If variability measures are to be used in the clinic, more research is required to determine the trade-off between accuracy, reliability, validity and clinical utility.
The question of how many strides to measure highlights the need for the development of standards and reference values. Standards were set to define minimum data acquisition requirements (e.g., sampling rate) to promote research and the clinical implementation of heart rate variability measures [83]. While there may be some debate about the exact values, the defining of standards greatly enhances the quality of the data and the ability to interpret and compare studies that use a given tool. Similarly, well-established reference values and norms are needed in different age groups and populations in order to promote interpretation and clinical references. Heart rate databases in different populations, complete with annotations and medical information, are widely available (e.g., see ), significantly advancing the sharing of methods and interpretation. Similar open-access database efforts would greatly help the study of gait variability and the development of clinical measures, but this must await the establishment of minimum standards for data collection and validation of the means of comparing data from different measurement systems.
The studies by West and Latka and Brach and colleagues also return us to hypotheses originally put forth by Gabell and Nayak over two decades ago [1]. They speculated that stride time variability reflects gait timing mechanisms and the pattern generator of gait, while variability of support time and step width more closely reflect balance control. Future studies are needed to unravel the various aspects of gait variability and their nonlinear interactions (in this respect, the potential of the animal models comes to fore), to identify the mechanisms that are responsible for each of the complementary measures of the stride-to-stride fluctuations in gait, and to work out the relationship between balance control and gait variability. The basal ganglia and dopamine-sensitive networks apparently participate in the regulation of gait variability while visual feedback apparently does not play a critical role in healthy adults. We lack, however, a good understanding of the neural center(s) that generates and regulates gait timing and are left to speculate why the maintenance of gait variability may be influenced by cognitive challenges, at least certain types under specific conditions. It has become clear that more sinus rhythm heart rate variability is (generally) "good", while more stride time variability is (generally) "bad" The final words on the value and interpretation of the variability of multiple other aspects of gait (e.g. step width variability), their inter-dependence and the relationship to the variability of other motor control tasks await the results of future studies [63,65,68,84-88].
Conflict of interest Statement
The author(s) declare that they have no competing interests.
Acknowledgements
This work was supported in part by NIH grants AG14100, HD39838, RR13622, and AG08812, from the National Parkinson Foundation and from the US-Israel BiNational Science Foundation. The author thanks Drs. Nir Giladi and Ary L. Goldberger for invaluable discussions.
==== Refs
Gabell A Nayak US The effect of age on variability in gait J Gerontol 1984 39 662 666 6491179
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
Terrier P Schutz Y Variability of gait patterns during unconstrained walking assessed by satellite positioning (GPS) Eur J Appl Physiol 2003 90 554 561 12905048 10.1007/s00421-003-0906-3
Goldberger AL Amaral LAN Hausdorff JM Ivanov PC Peng CK Stanley HE Fractal dynamics in physiology: Alterations with disease and aging Proceedings of the National Academy of Sciences of the United States of America 2002 99 2466 2472 11875196 10.1073/pnas.012579499
Griffin L West DJ West BJ Random stride intervals with memory Journal of Biological Physics 2000 26 185 202 10.1023/A:1010322406831
West BJ Griffin L Allometric control, inverse power laws and human gait Chaos Solitons & Fractals 1999 10 1519 1527 10.1016/S0960-0779(98)00149-0
Hausdorff JM Purdon PL Peng CK Ladin Z Wei JY Goldberger AL Fractal dynamics of human gait: stability of long-range correlations in stride interval fluctuations J Appl Physiol 1996 80 1448 1457 8727526
Hausdorff JM Peng CK Ladin Z Wei JY Goldberger AL Is walking a random walk? Evidence for long-range correlations in stride interval of human gait J Appl Physiol 1995 78 349 358 7713836
Terrier P Turner V Schutz Y GPS analysis of human locomotion: Further evidence for long-range correlations in stride-to-stride fluctuations of gait parameters Hum Mov Sci 2005 24 97 115 15896861 10.1016/j.humov.2005.03.002
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
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
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 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
Hausdorff JM Zemany L Peng C Goldberger AL Maturation of gait dynamics: stride-to-stride variability and its temporal organization in children J Appl Physiol 1999 86 1040 1047 10066721
Hausdorff JM Schaafsma JD Balash Y Bartels AL Gurevich T Giladi N Impaired regulation of stride variability in Parkinson's disease subjects with freezing of gait Exp Brain Res 2003 149 187 194 12610686
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
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 1633 1637 14687395 10.1046/j.1532-5415.2003.51516.x
Stolze H Kuhtz-Buschbeck JP Drucke H Johnk K Illert M Deuschl G Comparative analysis of the gait disorder of normal pressure hydrocephalus and Parkinson's disease J Neurol Neurosurg Psychiatry 2001 70 289 297 11181848 10.1136/jnnp.70.3.289
Herman T Giladi N Gurevich T Hausdorff JM Gait instability and fractal dynamics of older adults with a "cautious" gait: why do certain older adults walk fearfully? Gait Posture 2005 21 178 185 15639397 10.1016/j.gaitpost.2004.01.014
Hausdorff JM Mitchell SL Firtion R Peng CK Cudkowicz ME Wei JY Goldberger AL Altered fractal dynamics of gait: reduced stride-interval correlations with aging and Huntington's disease J Appl Physiol 1997 82 262 269 9029225
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
Hausdorff JM Peng CK Goldberger AL Stoll AL Gait unsteadiness and fall risk in two affective disorders: a preliminary study BMC Psychiatry 2004 4 39 15563372 10.1186/1471-244X-4-39
Hausdorff JM Herman T Baltadjieva R Gurevich T Giladi N Balance and gait in older adults with systemic hypertension Am J Cardiol 2003 91 643 645 12615286 10.1016/S0002-9149(02)03332-5
Hausdorff JM Forman DE Ladin Z Goldberger AL Rigney DR Wei JY Increased walking variability in elderly persons with congestive heart failure J Am Geriatr Soc 1994 42 1056 1061 7930329
Frenkel-Toledo S Giladi N Peretz C Herman T Gruendlinger L Hausdorff JM Treadmill walking as an external pacemaker to improve gait rhythm and stability in Parkinson's disease Mov Disord 2005
Hausdorff JM Rios D Edelberg HK Gait variability and fall risk in community-livingolder adults: a 1-year prospective study Arch Phys Med Rehabil 2001 82 1050 1056 11494184 10.1053/apmr.2001.24893
Maki BE Gait changes in older adults: predictors of falls or indicators of fear J Am Geriatr Soc 1997 45 313 320 9063277
Menz HB Lord SR Fitzpatrick RC Acceleration patterns of the head and pelvis when walking are associated with risk of falling in community-dwelling older people J Gerontol A Biol Sci Med Sci 2003 58 M446 M452 12730255
Nakamura T Meguro K Sasaki H Relationship between falls and stride length variability in senile dementia of the Alzheimer type Gerontology 1996 42 108 113 9138973
Visser H Gait and balance in senile dementia of Alzheimer's type Age Ageing 1983 12 296 301 6660138
Guimaraes RM Isaacs B Characteristics of the gait in old people who fall Int Rehabil Med 1980 2 177 180 7239777
Giladi N Herman T Reider-Groswasser II Gurevich T Hausdorff JM Clinical characteristics of elderly patients with a cautious gait of unknown origin J Neurol 2005
Nutt JG Marsden CD Thompson PD Human walking and higher-level gait disorders, particularly in the elderly Neurology 1993 43 268 279 8437689
Glass L Synchronization and rhythmic processes in physiology Nature 2001 410 277 284 11258383 10.1038/35065745
Chau T A review of analytical techniques for gait data. Part 2: neural network and wavelet methods Gait & Posture 2001 13 102 120 11240358 10.1016/S0966-6362(00)00095-3
Chau T A review of analytical techniques for gait data. Part 1: fuzzy, statistical and fractal methods Gait & Posture 2001 13 49 66 11166554 10.1016/S0966-6362(00)00094-1
Goldberger AL Is the normal heartbeat chaotic or homeostatic? News Physiol Sci 1991 6 87 91 11537649
Goldberger AL Non-linear dynamics for clinicians: chaos theory, fractals, and complexity at the bedside Lancet 1996 347 1312 1314 8622511 10.1016/S0140-6736(96)90948-4
Havlin S Buldyrev SV Goldberger AL Mantegna RN Ossadnik SM Peng CK Simons M Stanley HE Fractals in biology and medicine Chaos Solitons Fractals 1995 6 171 201 11539852 10.1016/0960-0779(95)80025-C
Stanley HE Buldyrev SV Goldberger AL Goldberger ZD Havlin S Mantegna RN Ossadnik SM Peng CK Simons M Statistical mechanics in biology: how ubiquitous are long-range correlations? Physica A 1994 205 214 253 11541307
West BJ Latka M Glaubic-Latka M Latka D Multifractality of cerebral blood flow Physica A-Statistical Mechanics and Its Applications 2003 318 453 460 10.1016/S0378-4371(02)01377-8
Latka M Glaubic-Latka M Latka D West BJ Fractal rigidity in migraine Chaos Solitons & Fractals 2004 20 165 170 10.1016/S0960-0779(03)00440-5
Peng CK Mietus JE Liu YH Lee C Hausdorff JM Stanley HE Goldberger AL Lipsitz LA Quantifying fractal dynamics of human respiration: Age and gender effects Annals of Biomedical Engineering 2002 30 683 692 12108842 10.1114/1.1481053
Glass L Complex cardiac rhythms Nature 1987 330 695 696 3696231 10.1038/330695a0
Ivanov PC Amaral LA Goldberger AL Havlin S Rosenblum MG Struzik ZR Stanley HE Multifractality in human heartbeat dynamics Nature 1999 399 461 465 10365957 10.1038/20924
Goldberger AL West BJ Applications of nonlinear dynamics to clinical cardiology Ann N Y Acad Sci 1987 504 195 213 3477116
Ho KK Moody GB Peng CK Mietus JE Larson MG Levy D Goldberger AL Predicting survival in heart failure case and control subjects by use of fully automated methods for deriving nonlinear and conventional indices of heart rate dynamics Circulation 1997 96 842 848 9264491
Huikuri HV Makikallio TH Peng CK Goldberger AL Hintze U Moller M Fractal correlation properties of R-R interval dynamics and mortality in patients with depressed left ventricular function after an acute myocardial infarction Circulation 2000 101 47 53 10618303
Makikallio TH Seppanen T Airaksinen KE Koistinen J Tulppo MP Peng CK Goldberger AL Huikuri HV Dynamic analysis of heart rate may predict subsequent ventricular tachycardia after myocardial infarction Am J Cardiol 1997 80 779 783 9315590 10.1016/S0002-9149(97)00516-X
Makikallio TH Ristimae T Airaksinen KE Peng CK Goldberger AL Huikuri HV Heart rate dynamics in patients with stable angina pectoris and utility of fractal and complexity measures Am J Cardiol 1998 81 27 31 9462601 10.1016/S0002-9149(97)00799-6
Makikallio TH Hoiber S Kober L Torp-Pedersen C Peng CK Goldberger AL Huikuri HV Fractal analysis of heart rate dynamics as a predictor of mortality in patients with depressed left ventricular function after acute myocardial infarction. TRACE Investigators. TRAndolapril Cardiac Evaluation Am J Cardiol 1999 83 836 839 10190395 10.1016/S0002-9149(98)01076-5
Baumert M Baier V Haueisen J Wessel N Meyerfeldt U Schirdewan A Voss A Forecasting of life threatening arrhythmias using the compression entropy of heart rate Methods Inf Med 2004 43 202 206 15136870
Wessel N Voss A Kurths J Schirdewan A Hnatkova K Malik M Evaluation of renormalised entropy for risk stratification using heart rate variability data Med Biol Eng Comput 2000 38 680 685 11217887
Malatesta D Simar D Dauvilliers Y Candau R Borrani F Prefaut C Caillaud C Energy cost of walking and gait instability in healthy 65- and 80-yr-olds J Appl Physiol 2003 95 2248 2256 12882986
West BJ Griffin L Allometric control of human gait Fractals-An Interdisciplinary Journal on the Complex Geometry of Nature 1998 6 101 108 10.1142/S0218348X98000122
West BJ Scafetta N Nonlinear dynamical model of human gait Physical Review e 2003 67
Kaplan DT Furman MI Pincus SM Ryan SM Lipsitz LA Goldberger AL Aging and the complexity of cardiovascular dynamics Biophys J 1991 59 945 949 2065195
Lipsitz LA Physiological complexity, aging, and the path to frailty Sci Aging Knowledge Environ 2004 2004 e16
Lipsitz LA Goldberger AL Loss of 'complexity' and aging. Potential applications of fractals and chaos theory to senescence JAMA 1992 267 1806 1809 1482430 10.1001/jama.267.13.1806
Ivanov PC Hausdorff JM Havlin S Amaral LA Arai K Shulte-Frolinde V Yoneyama M Stanley HE Levels of complexity in scale-invariant neural signals http://arxiv org/abs/cond-mat/0409545 2004
Guralnik JM Ferrucci L Pieper CF Leveille SG Markides KS Ostir GV Studenski S Berkman LF Wallace RB Lower extremity function and subsequent disability: consistency across studies, predictive models, and value of gait speed alone compared with the short physical performance battery J Gerontol A Biol Sci Med Sci 2000 55 M221 M231 10811152
Aminian K Najafi B Bula C Leyvraz PF Robert P Spatio-temporal parameters of gait measured by an ambulatory system using miniature gyroscopes Journal of Biomechanics 2002 35 689 699 11955509 10.1016/S0021-9290(02)00008-8
Brach JS Berthold R Craik R VanSwearingen JM Newman AB Gait variability in community-dwelling older adults J Am Geriatr Soc 2001 49 1646 1650 11843998
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
Grabiner PC Biswas ST Grabiner MD Age-related changes in spatial and temporal gait variables Arch Phys Med Rehabil 2001 82 31 35 11239283 10.1053/apmr.2001.18219
Hausdorff JM Ladin Z Wei JY Footswitch system for measurement of the temporal parameters of gait J Biomech 1995 28 347 351 7730393 10.1016/0021-9290(94)00074-E
Mayagoitia RE Nene AV Veltink PH Accelerometer and rate gyroscope measurement of kinematics: an inexpensive alternative to optical motion analysis systems Journal of Biomechanics 2002 35 537 542 11934425 10.1016/S0021-9290(01)00231-7
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
Zijlstra W Assessment of spatio-temporal parameters during unconstrained walking European Journal of Applied Physiology 2004 92 39 44 14985994 10.1007/s00421-004-1041-5
Bonato P Advances in wearable technology and applications in physical medicine and rehabilitation J Neuroengineering Rehabil 2005 2 2 15733322 10.1186/1743-0003-2-2
Ashkenazy Y Hausdorff JM Ivanov PC Stanley HE A stochastic model of human gait dynamics Physica A 2002 316 662 670
Hausdorff JM Ashkenazy Y Peng CK Ivanov PC Stanley HE Goldberger AL When human walking becomes random walking: fractal analysis and modeling of gait rhythm fluctuations Physica A 2001 302 138 147 12033228
Yogev G Giladi N Peretz C Springer S Simon ES Hausdorff JM Dual tasking, gait rhythmicity, and Parkinson's disease: which aspects of gait are attention demanding? Eur J Neurosci 2005
Springer S Giladi N Simon ES Hausdorff JM Deficits in executive function in idiopathic elderly fallers: Association with fall risk Neurology 2004 62 A129 A129 10.1159/000080030
Hausdorff JM Balash J Giladi N Effects of cognitive challenge on gait variability in patients with Parkinson's disease J Geriatr Psychiatry Neurol 2003 16 53 58 12641374
Hausdorff JM Yogev G Springer S Simon ES Giladi N Walking is more like catching than tapping: gait in the elderly as a complex cognitive task Exp Brain Res 2005
Danion F Varraine E Bonnard M Pailhous J Stride variability in human gait: the effect of stride frequency and stride length Gait Posture 2003 18 69 77 12855302 10.1016/S0966-6362(03)00030-4
Hausdorff JM Stride variability: beyond length and frequency Gait Posture 2004 20 304 15531178 10.1016/j.gaitpost.2003.08.002
Yamasaki M Sasaki T Torii M Sex difference in the pattern of lower limb movement during treadmill walking Eur J Appl Physiol Occup Physiol 1991 62 99 103 2022210 10.1007/BF00626763
Holt KG Hamill J Andres RO Predicting the minimal energy costs of human walking Med Sci Sports Exerc 1991 23 491 498 1905381
Holt KJ Jeng SF RR RR Hamill J Energetic Cost and Stability During Human Walking at the Preferred Stride Velocity J Mot Behav 1995 27 164 178 12736125
Plotnik M Giladi N Balash Y Peretz C Hausdorff JM Is freezing of gait in Parkinson's disease related to asymmetric motor function? Ann Neurol 2005 57 656 663 15852404 10.1002/ana.20452
Heart rate variability: standards of measurement, physiological interpretation and clinical use. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology Circulation 1996 93 1043 1065 8598068
Newell KE Corcos DM Variability and motor control 1993 Human Kinetics Publishers
Riley MA Turvey MT Variability and determinism in motor behavior Journal of Motor Behavior 2002 34 99 125 12057885
Domkin D Laczko J Djupsjobacka M Jaric S Latash ML Joint angle variability in 3D bimanual pointing: uncontrolled manifold analysis Exp Brain Res 2005 163 44 57 15668794 10.1007/s00221-004-2137-1
Latash ML Progress in motor control, volume one: Bernsteins's traditions in movement studies 1998 1st Human Kinetics Publishers
Grobstein P Ramachandran VS Variability in brain function and behavior The Encyclopedia of Human Behavior, Volume 4 1994 Academic Press Inc. 447 458
|
16033650
|
PMC1185560
|
CC BY
|
2021-01-04 16:37:40
|
no
|
J Neuroengineering Rehabil. 2005 Jul 20; 2:19
|
utf-8
|
J Neuroeng Rehabil
| 2,005 |
10.1186/1743-0003-2-19
|
oa_comm
|
==== Front
Lipids Health DisLipids in Health and Disease1476-511XBioMed Central London 1476-511X-4-141601180510.1186/1476-511X-4-14ResearchInducible nitric oxide synthase links NF-κB to PGE2 in polyunsaturated fatty acid altered fibroblast in-vitro wound healing Jia Yi [email protected] John J [email protected] Department of Basic Medical Sciences, Purdue University, West Lafayette, Indiana 47907, USA2005 12 7 2005 4 14 14 30 6 2005 12 7 2005 Copyright © 2005 Jia and Turek; licensee BioMed Central Ltd.2005Jia and Turek; 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 study investigated mechanisms of altered fibroblast collagen production induced by polyunsaturated fatty acids. 3T3-Swiss fibroblasts were grown in medium containing either eicosapentaenoic or arachidonic acid. The effects of nuclear factor-kappaB activation by lipopolysaccharide on inducible nitric oxide synthase, nitric oxide, prostaglandin E2, collagen production, and in-vitro wound healing were studied.
Results
Eicosapentaenoic acid treated cells produced less prostaglandin E2 but had increased inducible nitric oxide synthase expression, nitric oxide production, collagen formation, and recoverage area during in-vitro wound healing than cells treated with arachidonic acid. Activation of nuclear factor-kappaB with lipopolysaccharide increased inducible nitric oxide synthase expression, the production of nitric oxide, prostaglandin E2, collagen, and the in-vitro wound recoverage area. The nitric oxide synthase inhibitor, NG-nitro-L-arginine methyl ester, decreased lipopolysaccharide-induced nitric oxide, but the amount of nitric oxide was greater in eicosapentaenoic acid treated cells. NG-nitro-L-arginine methyl ester plus lipopolysaccharide treatment increased collagen production and cellular recoverage area while treatment with NG-nitro-L-arginine methyl ester alone decreased it in wounded fibroblasts.
Conclusion
The activation of the NF-κB pathway and PGE2 can be linked by the cross-talk of iNOS and NO in the PUFA altered fibroblast collagen production and wound healing. Additional studies are needed to determine how polyunsaturated fatty acids can be used as adjuvants in combination with other treatments (i.e, drugs) to design therapies to either enhance healthy collagen production or inhibit production and reduce fibrosis.
==== Body
Background
The purpose of this research was to test the hypothesis that polyunsaturated fatty acids can alter collagen formation during in-vitro healing by changing iNOS expression and NO production, which have cross-interaction with the nuclear transcription factor kappaB (NF-κB) pathway and PGE2. The control of collagen formation for optimal healing in tissues and organs is essential for numerous diseases [1-3]. The healing of skin and connective tissues such as ligaments require enhanced and effective production of healthy collagen for strength and to shorten the recovery time, but without scarring. However, collagen formation in injured vital organs needs to be minimized to prevent fibrosis and subsequent loss of organ function. Both enhanced and reduced collagen formation likely have common regulatory mechanisms that remain to be elucidated. Multiple cellular and extracellular factors can influence collagen formation, such as nitric oxide [4], PGE2 [5], as well as growth factors [6] and matrix metalloproteinases [7]. Therefore, potential therapies to control healing may require an approach targeting multiple molecules or mechanisms.
Our previous studies showed that polyunsaturated fatty acids (PUFA) alter collagen production in avian chondrocytes [8], porcine medial collateral ligament fibroblasts [9], and murine 3T3-Swiss fibroblasts [5]. Eicosapentaenoic acid (EPA, 20:5 n-3) treated porcine medial collateral ligament fibroblasts produced more collagen than those treated with arachidonic acid (AA, 20:4 n-6) [9]. In murine 3T3-Swiss fibroblasts, we have observed that collagen production could be regulated by exposure to different n-6: n-3 PUFA ratios and these effects were mediated, in part, by PGE2 and changes in the signaling via the different PGE receptor subtypes [5].
Since many collagen formation associated genes have promoter or enhancer elements for NF-κB [10], we studied the different response of NF-κB related genes to EPA and AA treatments in 3T3-Swiss fibroblasts by a gene profiling system [11]. Treatments with lipopolysaccharide (LPS), an NF-κB inducer, stimulated increased expression of several genes in the NF-κB pathway that are linked collagen production (i.e, interleukin-6 and inducible nitric oxide synthase).
The activated NF-κB dimer binds to the 5'-flanking region of the iNOS promoter and induces iNOS formation [12]. Inducible nitric oxide synthase (iNOS) and nitric oxide (NO) have an important role in collagen formation during wound healing [13,14]. Inhibition of NO synthase significantly decreased collagen synthesis in wound fibroblasts [13]. Dermal fibroblasts from iNOS-knock out murine fibroblasts proliferated more slowly and synthesized less collagen, and NO donors restored the collagen synthesis to normal level [15]. Another study implicated a role for NO in PGE2 formation and collagen deposition in rats [16]. Therefore, endogenous iNOS and NO may link the NF-κB pathway to PGE2 and regulate collagen formation during wound healing.
Results
Real-time RT-PCR for iNOS mRNA
The transcriptional level of iNOS mRNA was determined by the real-time RT-PCR (Figure 1). The expression of iNOS mRNA can be altered by stimulation of the NF-κB pathway. Incubation with the NF-κB pathway inducer, lipopolysaccharide (LPS, 10 μg/ml), significantly (P < 0.01) increased the expression of iNOS mRNA in both arachidonic acid (AA, 20:4 n-6) and eicosapentanoic acid (EPA, 20:5 n-3) treated normal or wounded 3T3-Swiss fibroblasts. However, the EPA-treated fibroblasts were more responsive to induction of the NF-κB pathway than AA-treated cells. Activation of the NF-κB pathway in EPA-treated cells with LPS resulted in significantly increased (P < 0.001) transcription of iNOS mRNA compared to AA-treated cells. The addition of the nitric oxide synthase inhibitor, NG-nitro-L-arginine methyl ester (L-NAME, 10-7 M) resulted in a significant increase (P < 0.05) in the transcription of iNOS mRNA only in EPA-treated fibroblasts with LPS stimulation.
Figure 1 Real-time RT-PCR for iNOS mRNA expression in 3T3-Swiss fibroblasts n = 3, ± SD). The expression of iNOS mRNA can be altered by stimulation of the NF-κB pathway or inhibition of nitric oxide synthase in 3T3-Swiss fibroblasts. EPA treated fibroblasts were more sensitive to the changes of NF-κB pathway than AA treated cells. Cells were incubated for 48 hr with bovine serum albumin alone as control or bovine serum albumin-soap loaded fatty acids (25 μM). Medium was then replaced with fresh fatty acid enriched medium containing lipopolysaccharide (LPS, 10 μg/ml) with or without NG-nitro-L-arginine methyl ester (L-NAME, 10-7 M) for another 24 hr. The wound was created after initial 48 hr treatments in duplicate set of plate with or without LPS. Cells were harvested at 24 hr post wounding and real-time RT-PCR was performed. The results were presented as the percentage of iNOS to β-actin mRNA expression. Bars with different letters are significantly different (P < 0.05). EPA, eicosapentaenoic acid; AA, arachidonic acid; LPS, lipopolysaccharide; L-NAME, NG-nitro-L-arginine methyl ester.
Quantification of nitrite
A stable end product of NO synthesis, nitrite, was measured to determine the NO concentration (Figure 2). The production of nitrite can be changed by altering the activity of the NF-κB pathway and nitric oxide synthase. EPA-treated normal fibroblasts without LPS stimulation produced more nitrite than control or AA-treated fibroblasts. However, when stimulated by LPS, EPA-treated normal fibroblasts produced significantly (P < 0.05) less nitrite than control or AA-treated cells. Treatment with L-NAME significantly reduced LPS-induced nitrite production, and EPA-treated cells produced more nitrite than AA-treated cells. Wounded and LPS stimulated cells decreased nitrite production for all treatments compared to LPS alone, but EPA-treated fibroblasts produced less nitrite than AA-treated or control cells.
Figure 2 NO production in 3T3-Swiss fibroblasts (n = 3, ± SD). Cells were incubated for 48 hr with bovine serum albumin alone as control or bovine serum albumin-soap loaded fatty acids (25 μM). Medium was then replaced with fresh fatty acid enriched medium containing lipopolysaccharide (LPS, 10 μg/ml) with or without NG-nitro-L-arginine methyl ester (L-NAME, 10-7 M) for another 24 hr. The wound was created after initial 48 hr treatments in a duplicate set of plate with LPS treatments. Culture supernatants (100 μl) were collected at 24 hr post wounding and quantification of nitrite was performed. The results were presented as the nitrite concentration (μM). Bars with different letters are significantly different (P < 0.05). EPA, eicosapentaenoic acid; AA, arachidonic acid; LPS, lipopolysaccharide; L-NAME, NG-nitro-L-arginine methyl ester.
Quantification of PGE2
Activation of NF-κB pathway by LPS in 3T3-Swiss fibroblasts significantly increased (p < 0.05) the production of PGE2 over control levels for all groups (Figure 3). Both basal and LPS stimulated amounts of PGE2 were lowest in EPA-treated fibroblasts compared to the control and AA-treated cells. The addition of the nitric oxide synthase inhibitor, L-NAME, to LPS stimulated fibroblasts decreased PGE2 production to basal levels for all groups. Wounding of LPS stimulated cells did not significantly change the amount of PGE2 produced by cells when compared with those stimulated by LPS alone.
Figure 3 PGE2 production in 3T3-Swiss fibroblasts n = 3, ± SD). The addition incubation of nitric oxide synthase inhibitor decreased the LPS induced PGE2 production in 3T3-Swiss fibroblasts. Cells were incubated for 48 hr with bovine serum albumin alone as control or bovine serum albumin-soap loaded fatty acids (25 μM). Medium was then replaced with fresh fatty acid enriched medium containing lipopolysaccharide (LPS, 10 μg/ml) with or without NG-nitro-L-arginine methyl ester (L-NAME, 10-7 M) for another 24 hr. The wound was created after initial 48 hr treatments in duplicate set of plate with or without LPS. Culture supernatants were collected at 24 hr post wounding and the quantification of PGE2performed. The results were presented as the PGE2 concentration (pg/ml). Bars with different letters are significantly different (P < 0.05). EPA, eicosapentaenoic acid; AA, arachidonic acid; LPS, lipopolysaccharide; L-NAME, NG-nitro-L-arginine methyl ester.
Collagen formation
Wounded fibroblasts without any PFA treatment increased both collagen production (CP) and collagen as a percentage of total proteins (C-PTP) compared to control (non-wounded, no PFA treatment) fibroblasts (Table 2). However, wounded fibroblasts treated with LPS produced less collagen and C-PTP than the control cells with LPS stimulation. Treatment with L-NAME alone decreased collagen production in control fibroblasts, while the additional of L-NAME plus LPS produced more collagen and had more collagen as a percentage of total protein in both control and wounded fibroblasts. The combination treatments of EPA, L-NAME and LPS produced higher CP and C-PTP than treatments of AA, L-NAME and LPS in normal fibroblasts.
Table 2 Collagen formation in 3T3-Swiss fibroblasts (n = 3, ± SD). Cells were incubated for 48 hr with bovine serum albumin alone as control or bovine serum albumin-soap loaded fatty acids (25 μM). The wound was then created and the medium was then replaced with fresh fatty acid enriched medium containing 50 μM ascorbic acid and 5 μCi of 3H-proline with or without lipopolysaccharide (LPS, 10 μg/ml) and NG-nitro-L-arginine methyl ester (L-NAME, 100 nM). A duplicate set of plates was made without wounding. Cells were harvested at 24 hr post wounding and assayed for collagen, total protein and DNA. The amount of collagen is expressed as DPM from 3H-proline per μg of DNA. Mean values within rows having different superscripts are significantly different (P < 0.05) by 1-way ANOVA and Tukey test. EPA, eicosapentaenoic acid; AA, arachidonic acid; LPS, lipopolysaccharide; L-NAME, NG-nitro-L-arginine methyl ester.
Treatments Collagen production (CP) (DPM/μg DNA amount) Collagen as a percentage of total protein (C-PTP) (%)
Control 2419.6 ± 318.1a 18.7 ± 1.6a
Wound 3044.8 ± 269.2b 21.8 ± 1.5b
LPS 3953.3 ± 57.3c 26.5 ± 0.8c
L-NAME 1225.6 ± 270.7d 11.4 ± 2.7d
LPS + Wound 2650.3 ± 146.9ab 19.0 ± 1.8ab
LPS + L-NAME 5393.7 ± 583.4e 37.5 ± 3.3e
LPS + L-NAME + EPA 6580.2 ± 418.3e 42.4 ± 5.6e
LPS + L-NAME + AA 3511.3 ± 561.4c 26.7 ± 4.0c
In-vitro wounding assay
The cellular recoverage area in wounded fibroblasts changed by the treatments of LPS, nitric oxide synthase inhibitor, cyclooxygenase (COX) inhibitor, and PFA (Figure 4). After 24 hr post wounding, LPS activation of the NF-κB pathway increased the percentage of cellular recoverage area compared to the control. The treatment of L-NAME alone decreased the percentage of cellular recoverage area in all groups while the treatment of L-NAME plus LPS increased the cellular recoverage area compared to LPS treatment alone. Administration of indomethacin, a cyclooxygenase inhibitor, increased the cellular recoverage area in LPS stimulated fibroblasts. The treatments of EPA and AA alone both increased the recoverage area compared to the control. EPA-treated fibroblasts with LPS had a higher the percentage of cellular recoverage area than the AA plus LPS treated cells, with or without other treatments.
Figure 4 In-vitro wounding assay in 3T3-Swiss fibroblasts (n = 3, ± SD). The addition treatment of L-NAME to LPS increased the percentages of cellular recover area while the treatment of L-NAME alone decreased it in wounded 3T3-Swiss fibroblasts. When stimulated with LPS, EPA treated fibroblasts had higher the percentage of cellular recover area than the AA treated cells. Cells were incubated for 48 hr with bovine serum albumin alone as control or bovine serum albumin-soap loaded fatty acids (25 μM). Then wound was created and the medium was replaced with fresh fatty acid enriched medium containing lipopolysaccharide (LPS, 10 μg/ml), NG-nitro-L-arginine methyl ester (L-NAME, 10-7 M) and indomethacin (10-8 M) for another 24 hr. Multiple photographs of the wound were obtained and the percentage of cellular recover areas were determined. Bars with different letters are significantly different (P < 0.05). EPA, eicosapentaenoic acid; AA, arachidonic acid; LPS, lipopolysaccharide; L-NAME, NG-nitro-L-arginine methyl ester; INDO, indomethacin.
Discussion
Polyunsaturated fatty acids affect the formation of numerous mediators of inflammation, such as eicosanoids and cytokines [9]. Thus, dietary fatty acids, those mobilized from cellular phospholipids or triacylglycerols stored in adipose tissue, and topical applications [18] can all potentially influence wound healing. A limited number of studies have examined the effects of dietary or topical applications of PUFA on wound healing. A cutaneous wound healing study found that rats fed a diet enriched in n-3 fatty acids produced wounds that were weaker in tensile strength compared to those from rats fed a diet containing n-6 fatty acids [19]. Cutaneous wounds in dogs fed an n-3 enriched diet had reduced epithelialization and contraction of open wounds and less edema in sutured wounds after 5 days than dogs fed a diet enriched in n-6 fatty acids. However, the n-3 diet did not appear to have a negative effect on wound healing [20]. A more recent study found that topical application of oleic acid (monounsaturated n-9 fatty acid) to surgically induced skin wounds in mice resulted in more rapid wound closure than applications of linolenic acid (18:3, n-3) or linoleic acid (18:2, n-6) [18]. The topical application of linolenic acid also resulted in an increased amount of connective tissue fibers compared to other fatty acid treatments. These limited studies suggest that in order to take advantage of the inflammation mediating properties of PUFA and use them as adjuvants in the healing process, it may be necessary to use the different classes of PUFA selectively during the stages of wound healing. Wound healing consists of an inflammatory phase, a proliferative (fibroblastic) phase, and a remodeling (maturation) phase [21,22]. In addition, the type of wound (e.g., acute or chouronic) and patient (e.g., diabetic, burn injured, or immunosuppressed) will require matching the appropriate PUFA to the particular stage of wound healing. For example, in immunosuppressed patients, a PUFA that enhances inflammation in the early stages may be needed to stimulate the healing process.
A component of optimal healing is regulation of collagen formation. Favorable collagen formation includes both enhanced collagen production for healing connective tissues and diminished formation in vital organs to minimize fibrosis. Our previous studies revealed that polyunsaturated fatty acids affected a number of mediators and pathways involved in collagen formation [5,9]. Many collagen formation associated genes, such as IL-6 and iNOS, have promoter or enhancer elements for NF-κB [23,24]. These genes provide ideal targets to connect various pathways involved in PUFA influenced collagen formation. The present paper studied the link between NF-κB pathway generated iNOS and nitric oxide (NO) in PUFA altered collagen formation, and the association with PGE2 production. Our hypothesis in the present study is that polyunsaturated fatty acids can alter collagen formation during in-vitro healing by changing iNOS expression and NO production, which have cross-interaction with the NF-κB pathway and PGE2 (Figure 4, 5).
Figure 5 Proposed model for regulation of collagen formation in these experiments. 1. Activation of the nuclear factor-κB (NF-κB) signaling pathway by lipopolysaccharide. 2. NO can inhibit the binding of NF-κB family members to target DNA. 3. NF-κB family binding to DNA and transcription of iNOS mRNA. 4. L-NAME inhibits the production of NO from iNOS. 5. NO regulates PGE2 production divergently through activation of COX isoforms. 6. Indomethacin inhibits COX and reduces PGE2production. 7. Collagen production and wound healing processes are changed by the activation of different PGE2 receptors subtypes and second messengers. LPS, lipopolysaccharide; Tlr4, Toll-like receptor 4; Traf6, Tumor necrosis factor receptor-associated factor 6; IκB, Inhibitor of kappa light polypeptide gene enhancer in B-cell; NF-κB, nuclear transcription factor kappaB; iNOS, inducible nitric oxide synthase; L-NAME, NG-nitro-L-arginine methyl ester; NO, nitric oxide; COX, cyclooxygenase; INDO, indomethacin; PKC, protein kinase C; cAMP, cyclic adenosine monophosphate.
NO is synthesized from the terminal guanidine nitrogen atom of L-arginine by nitric oxide synthase (NOS) and involved in many biological functions [25]. The inducible isoform of NOS (iNOS) is regulated primarily at the transcriptional level and contributes to most of the NO produced compared to the other isoforms [26]. The NF-κB signaling pathway regulates promoter regions and induces iNOS transcription [12]. In the present study using real-time RT-PCR, activation of the NF-κB pathway by LPS increased the transcription of iNOS mRNA in 3T3-Swiss fibroblasts. The increase was greatest in EPA-treated fibroblasts. Other researchers reported that NO can also regulate iNOS transcription through a negative feedback by inhibiting NF-κB binding to DNA [27]. In our experiments, the addition of a small amount of the nitric oxide synthase inhibitor, L-NAME, increased the transcription of iNOS mRNA only in EPA plus LPS treated normal fibroblasts. However, the increased iNOS expression in EPA plus LPS treated cells did not result in increased nitric oxide (nitrite) production, and EPA plus LPS treated cells produced less nitric oxide than control or AA treated cells. The decreased amount of nitric oxide with EPA plus LPS treatment may enhance iNOS expression due to decreased negative feedback on NF-κB binding to DNA (Figure 4, 5). The observation that increased iNOS expression in EPA-treated cells does not lead to increased NO production, may be explained by altered post-transcriptional regulation that leads to reduced synthesis of iNOS protein and thus decreased NO production.
Reduced NO in EPA treated cells, even though there is increased transcription of iNOS mRNA, can also be linked to PGE2 production in these cells. The amount of PGE2 produced by the 3T3 fibroblasts was inversely related to the amount of nitric oxide produced. PGE2 will bind to PGE receptors (EP1, EP2, EP3, and EP4) that activate signaling by protein kinase C (PKC) or cyclic AMP (cAMP). Cells treated with AA are mainly responsive to signaling from the EP1 receptor (PKC signaling), whereas EPA-treated cells are mainly responsive to EP2 and EP4 signaling (cAMP signaling) [5]. Arachidonic acid alone has been shown to function as a secondary messenger and activate PKC. The increased amount of PGE2 in AA-treated cells and stimulation of the PKC signaling pathway via the EP1 receptor (e.g., PKC activation) could affect post-translational synthesis of iNOS protein. Other studies identified a role for PKC in the induction of iNOS by the NF-κB pathway in murine 3T3 fibroblasts [28] and the nitric oxide action on angiogenesis [29].
Nitric oxide regulates PGE2 production in a divergent manner. Nitric oxide stimulated PGE2 release in macrophages [30,31] while endogenous NO inhibited PGE2 production in LPS-stimulated macrophages [32]. The dual effect of NO on PGE2 production may be due to the different effects on cyclooxygenase (COX) isoforms. Study of COX isoform deficient murine fibroblasts revealed that NO changed PGE2 production by activating COX-1 but inhibiting COX-2 [30]. Other studies showed that low concentrations of nitric oxide attenuate PGE2 production induced by LPS, in part, due to decreased expression of COX-2 protein [33]. However, LPS activation of the NF-κB pathway can influence COX-2 gene expression by directly altering COX-2 mRNA transcription [34]. Our results showed that inhibition of NO with L-NAME, decreased LPS-induced PGE2 production in normal 3T3-Swiss fibroblasts. Thus, PGE2production in 3T3-Swiss fibroblasts may be regulated at the COX-2 transcription level through the NF-κB pathway or the COX-2 post-translational level through the production of NO derived via nitric oxide synthase. The combination of PUFA treatment (EPA) plus blocking of nitric oxide production with L-NAME, resulted in a synergistic effect that enhanced collagen production compared to control fibroblasts.
The suppression of type I collagen gene expression by PGE2 can be mediated by both altering the amount and steady-state of collagen mRNA [35,36]. Interestingly, inhibition of iNOS activity has been shown to both slow collagen production in the healing process and also favor collagen deposition and development of fibrosis [37]. It was hypothesized that NO scavenged reactive oxygen species by formation of peroxynitrite, and iNOS inhibition blocked NO formation and contributed to the development of fibrosis [37]. Therefore, appropriate induction of iNOS from the NF-κB signaling pathway may be a critical mechanism for controlling collagen formation in both fibrosis and proper wound healing.
Changes in collagen formation have also been correlated to the concentration of NO produced. Collagen production (CP) and collagen as a percentage of total protein (C-PTP) was increased in LPS and interferon-γ (IFN-γ) incubated wound derived murine fibroblasts [13]. Additional treatments of NO synthase inhibitors decreased the CP and C-PTP in the same study. However, LPS and IFN-γ together, decreased, while inhibitors of NO restored the collagen production in human small intestinal lamina propria fibroblasts [38]. Another study in normal rat skin fibroblasts showed that a nitric oxide donor at low concentration enhanced, but at higher concentration reverted collagen synthesis back to control levels [4]. Dermal fibroblasts from iNOS-knock out mice proliferated more slowly and synthesized less collagen [15], and transfection with iNOS enhanced the collagen production in a cutaneous wounded rat model [39]. In our experiments, LPS stimulated the expression of iNOS mRNA, NO production, and collagen synthesis in 3T3-Swiss fibroblasts. The additional incubation of small amount of L-NAME to LPS decreased the NO production to a low level but increased the expression of iNOS mRNA and the collagen production.
Healing fibroblasts are phenotypically characterized by changes in collagen production, cell proliferation, and migration. Wound cells have been shown to increase the expression of iNOS and production of NO in various models [13,40]. Inhibition of iNOS delayed the reepithelialization in cutaneous wound repair [41]. The delayed wound repair in iNOS knockout mice was reversed by iNOS gene transfer [42]. Interestingly, the role of nitric oxide on cell proliferation after wounding is concentration-dependent [4]. At low concentration, NO promoted cell proliferation in murine fibroblasts [43], while at higher concentrations, NO decreased the proliferation of rat dermal skin fibroblasts [4]. The present study showed that the treatment of L-NAME alone inhibited the NO production and decreased the percentage of cellular recoverage area in 3T3-Swiss fibroblasts. However, the additional treatment of L-NAME to fibroblasts stimulated by LPS increased cell proliferation and migration compared to the LPS treatments alone. L-NAME alone may block the effect of nitric oxide at low concentration while the addition of L-NAME plus LPS decreased the LPS induced NO to a low concentration and increased the collagen formation, cell proliferation, and migration. Indomethacin, a cyclooxygenase inhibitor, increased the percentage of cellular recoverage area in fibroblasts with LPS stimulation. The decreased PGE2 induced by indomethacin also stimulated wound healing in 3T3-Swiss fibroblasts.
In summary, we demonstrated that iNOS and NO provide a cross-talk between the activation of the NF-κB pathway and PGE2 in the PUFA altered fibroblast collagen production and wound healing. The use of PUFA for regulation of collagen formation in wound healing or fibrosis will require additional experiments in order to determine how they may be used as adjuvants. Most likely it will be necessary to use a particular class of fatty acids (e.g., n-3, n-6, n-9) appropriate to the type of wound and stage of wound healing. In addition, use of fatty acids in combination with selective targeting of transcription factors and mediators (e.g., stimulation or inhibition) that regulate collagen formation will provide the level of control necessary to adapt the treatment protocol to the type of wound.
Methods
Reagents
Reagents were purchased from Sigma-Aldrich, St. Louis, MO, unless specified otherwise.
Cell culture and fatty acid enrichment
Mouse fibroblast (3T3-Swiss albino; American Type Culture Collection CCL-92, Rockville, MD) were maintained as subconfluent monolayers in six-well plates (Corning Costar, Cambridge MA) with Dulbecco's modified Eagle's medium (DMEM), 4 mM L-glutamine, 1.5 g/L sodium bicarbonate, 4.5 g/L glucose, and 10% bovine calf serum (Hyclone, Logan, UT). Subconfluent cultures grown for 24 hr in maintenance medium were washed twice and changed to fresh medium minus calf serum. In place of serum, the control medium was supplemented with 5 mg/100 ml of fatty acid free bovine serum albumin (BSA); and the fatty acid enriched test media were supplemented with BSA-loaded fatty acid soaps of arachidonic acid (AA, 20:4 n-6) or eicosapentaenoic acid (EPA, 20:5 n-3) (Nu-Chek Prep, Elysian, MN) to a final concentration of 25 μM. Cells were grown for 48 hr in the test media prior to addition of treatments described below.
In vitro wound assay
An in vitro wound healing assay was used to evaluate the migration and proliferation of 3T3-Swiss fibroblast as previous studies [9]. Briefly, a sterile pipette tip was used to make a 0.5-mm-wide wound by streaking across a monolayer of 3T3 Swiss fibroblasts. The wound was created when the cells were about 80% confluent after the initial 48 hr polyenoic fatty acid (PFA) enrichment and the migration of fibroblasts into the wound was measured at 24 hr post wounding. Multiple photographs of the wound were obtained by phase contrast microscopy, and the mean areas of cell recoverage for each sample were determined with image analysis software (Optimas 6.1, Media Cybernetics, Silver Spring, MD). The percentage of cellular recoverage area to the whole wound area was measured to evaluate the combined effects of cell proliferation and migration.
A similar set of plates was used to perform the real-time RT-PCR of iNOS, quantification of nitrite and PGE2, and collagen synthesis. For real-time RT-PCR and the quantification of nitrite and PGE2, the following protocol was used. After the initial 48 hr PFA enrichment, the media was changed with fresh fatty acid enriched medium containing Eshericia coli O55:B5 lipopolysaccharide (LPS, 10 μg/ml) with or without NG-nitro-L-arginine methyl ester (L-NAME, 10-7 M) and indomethacin (INDO, 10-8 M). Lipopolysaccharide is used to activate the NF-κB pathway and mimic some aspects of in vivo inflammation. For collagen formation, after the initial 48 hr polyenoic fatty acid (PFA) enrichment, the media was replaced with fresh fatty acid enriched medium containing 50 μM ascorbic acid and 5 μCi of 3H-proline (Amersham, Arlington Heights IL) with or without treatments. At 24 hr post wounding, a portion of the media was collected or the cells were harvested. Parallel plates were also cultured with the same treatments but without wounding. The cells for all treatments used in these experiments had greater than 99% viability based upon the standard trypan blue dye exclusion test.
RNA isolation and cDNA synthesis
After cells were washed with Hank's balanced salt solution and harvested, total RNA was isolated using a commercially available kit (RNAqueous, Ambion, Austin, TX) and digested with RNase-free DNase, as recommended by the supplier. The concentration and purity of total RNA were determined by measuring the optical density at 260 and 280 nm. The ratio of absorbance at 260 to 280 nm was 1.8–2.0. Two μg total RNA was reverse transcribed using a cDNA synthesis Kit (iScript™ cDNA Synthesis Kit, Bio-Rad Laboratories, Hercules, CA) following the manufacture's instructions. Generated cDNA were diluted with RNA-free water before usage.
Real-time RT-PCR
Synthesized cDNA encoding for murine iNOS and β-actin (as endogenous control) were amplified and analyzed by a real-time reverse transcription polymerase chain reaction (real time RT-PCR) system (Applied Biosystems 7300 Real Time PCR System, Foster City, CA). The oligonucleotide primers were designed by Primer Express software (Primer Express 2.0, Applied Biosystems) and ordered from IDT (Integrated DNA Technologies, Coralville, IA). The cDNA sequences were obtained from the Genebank database as indicated in table 1. PCR amplifications were performed according to the manufacture's protocol. Samples were prepared in a total volume of 50 μl containing 2 μl cDNA sample (diluted 1x, 10x, 100x, 200x), 25 μl SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA), 1 μl each primer, and 21 μl RNA-free water. For each reaction, the polymerase was initiated at 50°C for 2 min and 95°C for 10 min, and amplification was then performed at 35 cycles of switching between 95°C for 15 seconds, 55°C for 30 seconds, and 72°C for 15 seconds followed by melting point analysis from 60°C to 95°C. All samples were done in triplicate and the coefficient was determined by creating a standard curve from plotting CT values vs. the total RNA. The results were presented as the mRNA expression percentage of iNOS to β-actin.
Table 1 Primers for iNOS and β-actin. Primers for iNOS and β-actin were designed by Primer Express®2.0 (Applied Biosystems). iNOS, inducible nitric oxide synthase.
Gene Genebank accession # Start positions Primer sequences (5' ? 3') Amplicon length (bp)
iNOS U43428 2407 AGGGAATCTTGGAGCGAGTTGT 103
2509 AGCCTCTTGTCTTTGACCCAGT
β-actin X03672 886 TGGAATCCTGTGGCATCCATGA 91
976 AGCACTGTGTTGGCATAGAGGT
Quantification of nitrite
Nitrite (NO2-) in culture supernatants was measured to determine the synthesis of NO as other studies [17]. Culture supernatants (100 μl) were mixed with an equal volume of Griess reagent (1% sulfanilamide, 0.1% naphthylethylene diamine dihydrochloride and 2.5% phosphoric acid) and incubated at the room temperature with dimmed light for 10 min. Nitrite was measured at 550 nm by a microplate reader (EIA reader 2550, Bio-Rad, Hercules, CA) using the double-distilled water as blank and the sodium nitrite to generate a standard curve.
Quantification of PGE2
Assessment of total prostaglandin E2 (PGE2) amount from the medium of 3T3-Swiss fibroblasts incubation was preformed by an ELISA kit (Monoclonal prostaglandin E2 EIA Kit-514010; Cayman Chemical, Ann Arbor, MI), according to the manufacturer's recommended protocol.
Collagen assay
Collagen was assayed as described previously [5]. The media was collected and the cells washed twice with cold phosphate buffered saline (PBS). The cells were pelleted by centrifugation and the PBS wash combined with the media fraction. The cell pellet was suspended in 1.0 ml of ammonium hydroxide-Triton X-100 cell lysing solution (AT solution). Following 15-minute incubation at 37°C, 750 μl of the lysate is combined with the media fraction. The two fractions, cell and media, then underwent trichloroaceticic acid (TCA) precipitation (equal volume of 20% TCA added to the cell plus media fraction). The acid insoluble precipitate was then rinsed several times with 10% TCA to remove free 3H-proline. The precipitate was redissolved in 0.05 N NaOH in 0.05 M TES buffer plus 0.005 M CaCl2 and half of the solution incubated for 6 hr at 37°C with protease-free Type VII collagenase (100 U/ mL) in TES and the other half of the solution served as a control. Following the digestion, TCA was again added to precipitate the acid insoluble proteins; however, the collagen fragments generated by collagenase treatment remain in solution. The supernatant and precipitate were counted in a scintillation counter and collagen, non-collagenous, and total protein production reported as DPM/μg DNA.
DNA assay
The remaining 250 μl of the AT solution cellular-lysate from the collagen synthesis assay was used for total DNA determination. Picogreen, 50 μl (Molecular Probes, Eugene OR) was added to 50 μl of lysate or to 50 μl of known DNA standards and fluorescence of the dye binding to double stranded DNA was measured in a spectofluorometer. A DNA standard curve was generated by linear regression and sample DNA values was used to obtain the unknown values.
Statistical analysis
Data were presented as means ± standard deviation (SD) and analyzed by both one-way and two-way ANOVA procedures of SAS (SAS Institute, Cary, NC). A Tukey test was used to analyze significant main and interaction effects. A P value < 0.05 was considered statistically significant.
Acknowledgements
This study was supported in part by a grant from the State of Indiana 21st Century Research and Technology Fund. The author's thank Ingrid A. Schoenlein for technical assistance.
==== Refs
Hill C Flyvbjerg A Rasch R Bak M Logan A Transforming growth factor-beta2 antibody attenuates fibrosis in the experimental diabetic rat kidney. J Endocrinol 2001 170 647 651 11524245 10.1677/joe.0.1700647
Ghosh AK Factors involved in the regulation of type I collagen gene expression: implication in fibrosis Exp Biol Med (Maywood ) 2002 227 301 314 11976400
Kanzler S Baumann M Schirmacher P Dries V Bayer E Gerken G Dienes HP Lohse AW Prediction of progressive liver fibrosis in hepatitis C infection by serum and tissue levels of transforming growth factor-beta J Viral Hepat 2001 8 430 437 11703574 10.1046/j.1365-2893.2001.00314.x
Witte MB Thornton FJ Efron DT Barbul A Enhancement of fibroblast collagen synthesis by nitric oxide Nitric Oxide 2000 4 572 582 11139365 10.1006/niox.2000.0307
Jia Y Turek JJ Polyenoic fatty acid ratios alter fibroblast collagen production via PGE2 and PGE receptor subtype response Exp Biol Med (Maywood ) 2004 229 676 683 15229362
Molloy T Wang Y Murrell G The roles of growth factors in tendon and ligament healing Sports Med 2003 33 381 394 12696985
Daniels JT Cambrey AD Occleston NL Garrett Q Tarnuzzer RW Schultz GS Khaw PT Matrix metalloproteinase inhibition modulates fibroblast-mediated matrix contraction and collagen production in vitro Invest Ophthalmol Vis Sci 2003 44 1104 1110 12601036 10.1167/iovs.02-0412
Watkins BA Xu H Turek JJ Linoleate impairs collagen synthesis in primary cultures of avian chondrocytes Proc Soc Exp Biol Med 1996 212 153 159 8650253
Hankenson KD Watkins BA Schoenlein IA Allen KG Turek JJ Omega-3 fatty acids enhance ligament fibroblast collagen formation in association with changes in interleukin-6 production Experimental Biology & Medicine 2000 223 88 95 10.1046/j.1525-1373.2000.22312.x
Baldwin AS The NF-kappa B and Ikappa B proteins: new discoveries and insights Annual Review of Immunology 1996 14 649 681 8717528 10.1146/annurev.immunol.14.1.649
Jia Y Turek JJ Altered NF-kappaB gene expression and collagen formation induced by polyunsaturated fatty acids. Journal of Nutritional Biochemistry 2005
Xie QW Kashiwabara Y Nathan C Role of transcription factor NF-kappa B/Rel in induction of nitric oxide synthase J Biol Chem 1994 269 4705 4708 7508926
MR S PA E FJ T K K SS G A B Nitric oxide, an autocrine regulator of wound fibroblast synthetic function. J Immunol 1997 158 2375 2381 9036987
Witte MB Barbul A Role of nitric oxide in wound repair Am J Surg 2002 183 406 412 11975928 10.1016/S0002-9610(02)00815-2
Shi HP Most D Efron DT Tantry U Fischel MH Barbul A The role of iNOS in wound healing Surgery 2001 130 225 229 11490353 10.1067/msy.2001.115837
Muscara MN McKnight W Asfaha S Wallace JL Wound collagen deposition in rats: effects of an NO-NSAID and a selective COX-2 inhibitor Br J Pharmacol 2000 129 681 686 10683192 10.1038/sj.bjp.0703112
Cardoso CR Souza MA Ferro EA Jr FS Pena JD Influence of topical administration of n-3 and n-6 essential and n-9 nonessential fatty acids on the healing of cutaneous wounds Wound Repair Regen 2004 12 235 243 15086775 10.1111/j.1067-1927.2004.012216.x
Albina JE Gladden P Walsh WR Detrimental effects of an omega-3 fatty acid-enriched diet on wound healing JPEN J Parenter Enteral Nutr 1993 17 519 521 8301804
Scardino MS Swaim SF Sartin EA Hoffman CE Oglivie GK Hanson RA Coolman SL Davenport DJ The effects of omega-3 fatty acid diet enrichment on wound healing Veterinary Dermatology 1999 10 283 290 10.1046/j.1365-3164.1999.00148.x
Gilmore MA Phases of wound healing Dimens Oncol Nurs 1991 5 32 34 1823567
Bowes LE Jimenez MC Hiester ED Sacks MS Brahmatewari J Mertz P Eaglstein WH Collagen fiber orientation as quantified by small angle light scattering in wounds treated with transforming growth factor-beta2 and its neutalizing antibody Wound Repair Regen 1999 7 179 186 10417754 10.1046/j.1524-475X.1999.00179.x
Shimizu H Mitomo K Watanabe T Okamoto S Yamamoto K Involvement of a NF-kappa B-like transcription factor in the activation of the interleukin-6 gene by inflammatory lymphokines Mol Cell Biol 1990 10 561 568 2405250
Morris KR Lutz RD Choi HS Kamitani T Chmura K Chan ED Role of the NF-kappaB signaling pathway and kappaB cis-regulatory elements on the IRF-1 and iNOS promoter regions in mycobacterial lipoarabinomannan induction of nitric oxide Infect Immun 2003 71 1442 1452 12595462 10.1128/IAI.71.3.1442-1452.2003
Coleman JW Nitric oxide in immunity and inflammation Int Immunopharmacol 2001 1 1397 1406 11515807 10.1016/S1567-5769(01)00086-8
Knowles RG Moncada S Nitric oxide synthases in mammals Biochem J 1994 298 ( Pt 2) 249 258 7510950
Park SK Lin HL Murphy S Nitric oxide regulates nitric oxide synthase-2 gene expression by inhibiting NF-kappaB binding to DNA Biochem J 1997 322 ( Pt 2) 609 613 9065784
Kleinert H Euchenhofer C Ihrig-Biedert I Forstermann U In murine 3T3 fibroblasts, different second messenger pathways resulting in the induction of NO synthase II (iNOS) converge in the activation of transcription factor NF-kappaB J Biol Chem 1996 271 6039 6044 8626388 10.1074/jbc.271.11.6039
Jones MK Tsugawa K Tarnawski AS Baatar D Dual actions of nitric oxide on angiogenesis: possible roles of PKC, ERK, and AP-1 Biochem Biophys Res Commun 2004 318 520 528 15120632 10.1016/j.bbrc.2004.04.055
Clancy R Varenika B Huang W Ballou L Attur M Amin AR Abramson SB Nitric oxide synthase/COX cross-talk: nitric oxide activates COX-1 but inhibits COX-2-derived prostaglandin production J Immunol 2000 165 1582 1587 10903767
Clancy R Varenika B Huang W Ballou L Attur M Amin AR Abramson SB Nitric Oxide Synthase/COX Cross-Talk: Nitric Oxide Activates COX-1 But Inhibits COX-2-Derived Prostaglandin Production The Journal of Immunology 2000 165 1582 1587 10903767
az-Cazorla M Perez-Sala D Lamas S Dual effect of nitric oxide donors on cyclooxygenase-2 expression in human mesangial cells J Am Soc Nephrol 1999 10 943 952 10232679
Patel R Attur MG Dave M Abramson SB Amin AR Regulation of cytosolic COX-2 and prostaglandin E2 production by nitric oxide in activated murine macrophages J Immunol 1999 162 4191 4197 10201946
Rhee SH Hwang D Murine TOLL-like receptor 4 confers lipopolysaccharide responsiveness as determined by activation of NF kappa B and expression of the inducible cyclooxygenase J Biol Chem 2000 275 34035 34040 10952994 10.1074/jbc.M007386200
Varga J Diaz-Perez A Rosenbloom J Jimenez SA PGE2 causes a coordinate decrease in the steady state levels of fibronectin and types I and III procollagen mRNAs in normal human dermal fibroblasts. Biochem Biophys Res Commun 1987 147 1282 1288 3478047 10.1016/S0006-291X(87)80209-7
FB R WF L JR B LF S G K MB G Suppression of type I collagen gene expression by prostaglandins in fibroblasts is mediated at the transcriptional level Mol Med 2000 6 705 719 11055589
Ferrini MG Vernet D Magee TR Shahed A Qian A Rajfer J Gonzalez-Cadavid NF Antifibrotic role of inducible nitric oxide synthase Nitric Oxide 2002 6 283 294 12009846 10.1006/niox.2001.0421
D. C K.S.N. K Induction of cell proliferation and collagen synthesis in human small intestinal lamina propria fibroblasts by lipopolysaccharide: possible involvement of nitric oxide Biochemical and Biophysical Research Communications 1997 240 458 463 9388501 10.1006/bbrc.1997.7680
Thornton FJ Schaffer MR Witte MB Moldawer LL MacKay SL Abouhamze A Tannahill CL Barbul A Enhanced collagen accumulation following direct transfection of the inducible nitric oxide synthase gene in cutaneous wounds Biochem Biophys Res Commun 1998 246 654 659 9618268 10.1006/bbrc.1998.8681
Schaffer MR Tantry U van Wesep RA Barbul A Nitric oxide metabolism in wounds J Surg Res 1997 71 25 31 9271274 10.1006/jsre.1997.5137
Stallmeyer B Kampfer H Kolb N Pfeilschifter J Frank S The function of nitric oxide in wound repair: inhibition of inducible nitric oxide-synthase severely impairs wound reepithelialization J Invest Dermatol 1999 113 1090 1098 10594757 10.1046/j.1523-1747.1999.00784.x
Yamasaki K Edington HD McClosky C Tzeng E Lizonova A Kovesdi I Steed DL Billiar TR Reversal of impaired wound repair in iNOS-deficient mice by topical adenoviral-mediated iNOS gene transfer J Clin Invest 1998 101 967 971 9486966
Du M Islam MM Lin L Ohmura Y Moriyama Y Fujimura S Promotion of proliferation of murine BALB/C3T3 fibroblasts mediated by nitric oxide at lower concentrations Biochem Mol Biol Int 1997 41 625 631 9090471
Green LC Wagner DA Glogowski J Skipper PL Wishnok JS Tannenbaum SR Analysis of nitrate, nitrite, and [15N]nitrate in biological fluids Anal Biochem 1982 126 131 138 7181105 10.1016/0003-2697(82)90118-X
|
16011805
|
PMC1185561
|
CC BY
|
2021-01-04 16:39:18
|
no
|
Lipids Health Dis. 2005 Jul 12; 4:14
|
utf-8
|
Lipids Health Dis
| 2,005 |
10.1186/1476-511X-4-14
|
oa_comm
|
==== Front
Mol CancerMolecular Cancer1476-4598BioMed Central London 1476-4598-4-231601417610.1186/1476-4598-4-23ResearchVariation in transcriptional regulation of cyclin dependent kinase inhibitor p21waf1/cip1 among human bronchogenic carcinomas Harr Michael W [email protected] Timothy G [email protected] Erin L [email protected] Kristy A [email protected] Cheryl AM [email protected] James C [email protected] Department of Medicine, Medical University of Ohio, 219 Health Education Building, 3055 Arlington Avenue, Toledo, OH, 43614-5806, USA2 Department of Cariology, Restorative Sciences and Endodontics, University of Michigan, 2310A Dental Research Building, 1011 North University Avenue, Ann Arbor, MI, 48109-1078, USA2005 13 7 2005 4 23 23 14 2 2005 13 7 2005 Copyright © 2005 Harr et al; licensee BioMed Central Ltd.2005Harr 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
Cell proliferation control depends in part on the carefully ordered regulation of transcription factors. The p53 homolog p73, contributes to this control by directly upregulating the cyclin dependent kinase inhibitor, p21waf1/cip1. E2F1, an inducer of cell proliferation, directly upregulates p73 and in some systems upregulates p21 directly. Because of its central role in controlling cell proliferation, upregulation of p21 has been explored as a modality for treating bronchogenic carcinoma (BC). Improved understanding of p21 transcriptional regulation will facilitate identification of BC tissues that are responsive to p21-directed therapies. Toward this goal, we investigated the role that E2F1 and p73 each play in the transcriptional regulation of p21.
Results
Among BC samples (N = 21) p21 transcript abundance (TA) levels varied over two orders of magnitude with values ranging from 400 to 120,000 (in units of molecules/106 molecules β-actin). The p21 values in many BC were high compared to those observed in normal bronchial epithelial cells (BEC) (N = 18). Among all BC samples, there was no correlation between E2F1 and p21 TA but there was positive correlation between E2F1 and p73α (p < 0.001) TA. Among BC cell lines with inactivated p53 and wild type p73 (N = 7) there was positive correlation between p73α and p21 TA (p < 0.05). Additionally, in a BC cell line in which both p53 and p73 were inactivated (H1155), E2F1 TA level was high (50,000), but p21 TA level was low (470). Transiently expressed exogenous p73α in the BC cell line Calu-1, was associated with a significant (p < 0.05) 90% increase in p21 TA and a 20% reduction in E2F1 TA. siRNA mediated reduction of p73 TA in the N417 BC cell line was associated with a significant reduction in p21 TA level (p < 0.01).
Conclusion
p21 TA levels vary considerably among BC patients which may be attributable to 1) genetic alterations in Rb and p53 and 2) variation in TA levels of upstream transcription factors E2F1 and p73. Here we provide evidence that p73 upregulates p21 TA in BC tissues and upregulated p21 TA may result from E2F1 upregulation of p73 but not from E2F1 directly.
==== Body
Background
Cell cycle homeostasis in normal human bronchial epithelial cells (BEC) is highly regulated at the G1/S transition. In G1 phase of the cell cycle, formation of a heterodimeric complex between cyclin D and cyclin dependent kinases 4 or 6 (cdk 4,6) leads to the phosphorylation of the tumor suppressor retinoblastoma gene product (pRb) [1-3]. Phosphorylation causes conformational change of the pRb/E2F complex, followed by release, and activation of the E2F1, 2, and 3 transcription factors [4-6]. Free E2F proteins bind strongly to DNA and were first identified by their ability to transactivate the adenoviral E2 promoter [7]. E2F1 functions to upregulate transcription of genes required for entry into S phase, including cyclin E, c-myc and itself [8-10]. In turn, c-myc directly upregulates transcription of cyclin E and cdk4 [11,12]. Thus, phosphorylation of pRb by one or more cyclin/cdk complexes causes activation and upregulation of E2F1, upregulation of c-myc transcription by E2F1, and upregulation of cdk4 transcription by c-myc. These interactions are initiated at the restriction point of G1/S, which is associated with independence of the cell from extracellular growth factors [4,13]. The events described above contribute to a cell proliferation signal amplification cycle that would be uncontrolled in the absence of compensatory negative feedback.
Compensatory feedback signals, including the activation of p53 and transcriptional upregulation of p73 and p21waf1/cip1 (p21 hereafter) act to slow cell proliferation [14-16]. Unlike p53, p73 is not frequently mutated in human cancers [17], and thus it is not considered a classical tumor suppressor gene, as defined by Knudson's two hit hypothesis [18,19]. However, it functions to promote cell cycle arrest, DNA repair, and apoptosis much like p53 [18,20]. E2F1 (and c-myc) transactivation of p14ARF leads to stabilization of p53 [21,22] which slows cell cycling through the upregulation of p21 [23,24], and also induces apoptosis [25]. E2F1 also upregulates p73 [15,26,27] and p73 upregulates p21 [15], which in turn acts to inhibit the release of E2F1 from pRb, resulting in compensatory feedback for the loss of cell proliferation control. In some systems, E2F1 also upregulates p21 directly [28].
In previous studies, we determined that p21 transcript abundance (TA) levels vary considerably among bronchogenic carcinoma (BC) primary tissues and cell lines, and in some of these samples the p21 TA level is higher than the level observed in normal BEC [29,30]. Because p21 normally slows cell proliferation, this observation was unexpected, and counter-intuitive. If validated, these preliminary findings would have important implications for the many efforts presently underway to design gene specific cancer treatments that function to achieve cell proliferation control [31].
The purpose of this study was to better understand inter-tumor variation in the mechanisms responsible for loss of proliferation control and to better define the role of E2F1, p73 and p21 regulatory pathways as they relate to cell proliferation control in human BC. Our approach included additional descriptive studies in BC cell lines, as well as primary BC tissues and normal BEC to better quantify inter-tumor variation of p21 TA levels. With respect to the fourteen BC cell lines used in this study, because each has been extensively characterized at the genetic level [17,25,32-35], we were better able to define the regulatory relationship between E2F1, p73, and p21 after considering the known alterations in each individual cell line. Our descriptive approach was followed by experimental testing of the hypothesis posed to explain these observations.
Results
Transcript abundance (TA) levels of E2F1, p73α, and p21 in normal bronchial epithelial cells (BEC) (N = 18) and bronchogenic carcinoma (BC) tissues (N = 21) were measured by StaRT-PCR. Cultured cells and primary tissues are shown in Tables 1 and 2, respectively. For the BC cell lines, known alterations in p53, p16, p14ARF, Rb, p73, and c-myc also are presented, along with the TA values. Bivariate analysis of the TA data from these cultured human BEC samples provided data consistent with observations in other tissues that E2F1 regulates p73 transcription [15,26,27] and that p73 regulates p21 transcription [15]. However, in contrast to other cell types, they suggested that E2F1 does not directly regulate p21 transcription. To further investigate the significance of this finding, seven primary BC tissues (Table 2) were also assessed.
Table 1 Transcript abundance measurements for cultured normal BEC and BC cell lines
Sample Cell Type Medium Genetic Alterations Transcript Abundance
p53 p16 p14ARF Rb p73 c-myc E2F1 p73α p21
17378B N B 1,500 1 21,000
6F0333B N B 4,700 1 76,000
6F0395B N B 5,100 1 120,000
SW900 Sq R I 7,300 100 12,000
Calu-1 Sq R I I 8,500 28 2,000
H460 LC R I 9,000 14 18,000
H82 SC R I I I A 15,000 17 400
A549 A R I I 16,000 1 6,200
A2126 A R I 17,000 30 6,100
N417 SC R I I A 19,000 5,200 34,000
H322 BA R I I 20,000 190 31,000
H446 SC R I A 67,000 710 120,000
H1155* LC R I I I 50,000 250 470
H520 Sq R I I 56,000 120 2,300
H146 SC R I 59,000 700 19,000
H661 LC R I I 76,000 210 58,000
A427** A R I 180,000 220 100,000
N, normal; Sq, squamous carcinoma; LC, large cell carcinoma; SC, small cell carcinoma; A, adenocarcinoma; BA, bronchoalveolar carcinoma; B, BEGM medium; R, RPMI with 10% FBS; I, inactivated; A, amplification. Non-detectable TA were entered as 1 to allow plotting on a logarithmic scale. Cell lines are grouped by increasing E2F1 TA. All values are in units of molecules/106 molecules β-actin and represent the mean from three independent measurements. *H1155 has a missense mutation in the transactivation domain of p73. **A427 has a deletion in the p73 coding sequence but the protein product still retains its transactivational function [35].
Table 2 Transcript abundance measurements for primary normal BEC and BC tissues
Sample ID Diagnosis Cell Type E2F1 p73α p21
63 NC N 270 850 14,000
282 NC N 540 640 10,000
64 NC N 2,000 400 10,000
285 NC N 2,500 870 3,100
257 NC N 270 5,200 12,000
156 NC N 750 1,000 25,000
194 NC N 290 3,800 3,500
150 NC N 160 3,700 11,000
136 NC N 700 31 9,100
261 NC N 250 3,100 30,000
191 C N 150 900 13,000
158 C N 190 2,200 11,000
146 C N 230 2,600 9,100
34 C N 570 4,400 12,000
212 C N 600 5,300 35,000
274 C Sq 730 1 7,000
279 C A 1,200 1 4,000
190 C BA 6,600 3 27,000
102 C A 12,000 21 21,000
165 C SC 36,000 4,300 16,000
123 C Sq 54,000 5,500* 90,000
277* C A 120,000 100 2,000
NC, non-cancer; C, cancer; N, normal; Sq, squamous carcinoma; A, adenocarcinoma; BA, bronchoalveolar carcinoma; SC, small cell carcinoma. Non-detectable TA were entered as 1 to allow plotting on a logarithmic scale. Normal BEC are grouped by diagnosis and cell type. BC tissues are grouped by increasing E2F1 TA. All values are in units of molecules/106 molecules β-actin and represent the mean from three independent measurements. *This value represents the mean from two measurements due to insufficient sample.
The mean, median, and quartile values for each gene in normal (N = 18) compared to malignant samples (N = 21) are shown in Table 3. E2F1 TA levels were over 30-fold higher in BC relative to normal BEC (p < 0.0001). Conversely, mean p73α TA was higher in normal BEC by over 2-fold, but this difference was not statistically significant (p = 0.07), although the median p73 TA value was higher in normal BEC by nearly 9-fold. With respect to p21 TA level, there was no statistical difference in the mean value between the two groups.
Table 3 Descriptive statistics for E2F1, p73α, and p21 transcript abundance measurements
Normal BEC Bronchogenic Carcinoma
E2F1 p73α p21 E2F1 p73α p21
Mean 1,200 1,900 24,000 Mean 40,000 840 25,000
Min 150 1 3,100 Min 730 1 400
LQ 260 460 10,000 LQ 9,000 17 4,000
Median 560 950 12,000 Median 19,000 100 12,000
UQ 1,300 3,600 24,000 UQ 56,000 250 27,000
Max 5,100 5,300 120,000 Max 180,000 5,500 120,000
Min, minimum; LQ, lower quartile; UQ, upper quartile; max, maximum. Values were derived from statistical analysis of samples presented in Tables 1 and 2.
Median and quartile values were used to determine if TA levels were high or low for a given sample. For example, the median value for E2F1 in BC was 19,000. Therefore, a sample with an E2F1 TA level greater than 19,000 would be considered high and a TA level less than 19,000 would be considered low. These criteria were used to identify cell lines with low (Calu-1) or high (N417) p73α TA for use in the exogenous p73 and p73 siRNA transfection experiments.
Bivariate analysis of E2F1 and p21
There was no significant correlation between E2F1 and p21 TA levels among BC cell lines and primary tissues (N = 21) (Figure 1).
Figure 1 Lack of correlation between E2F1 and p21 TA. There was no correlation between E2F1 and p21 TA among BC cell lines (N = 14) and primary BC tissues (N = 7). Each point represents the mean value from triplicate measurements of E2F1 and p21 as shown in Tables 1 and 2 (except where indicated).
Bivariate analysis of E2F1 and p73α TA levels
E2F1 and p73α TA values were positively correlated (p < 0.001) among BC cell lines and primary tissues (N = 21) (Figure 2).
Figure 2 Bivariate analysis of E2F1 and p73α TA. E2F1 and p73α TA were positively correlated among BC cell lines (N = 14) and primary BC tissues (N = 7). Each point represents the mean value from triplicate measurements of E2F1 and p73α as shown in Tables 1 and 2 (except where indicated).
Bivariate analysis of p73α and p21 TA levels
Among BC cell lines in which p53 is known to be completely inactivated by mutation or deletion (N = 7), p73α and p21 were significantly correlated (p < 0.05), as shown in Figure 3. In contrast, there was a borderline but insignificant (p = 0.06) positive correlation between p73α and p21 TA among all BC cell lines and primary tissues (N = 21). This result may be explained by p53 regulation of p21 transcription as a confounding variable in those cell lines with wild type p53.
Figure 3 Bivariate analysis of p73α and p21 TA. p73α and p21 TA were positively correlated (p < 0.05) among BC cell lines (N = 7) where p53 was known to be inactivated by mutation or deletion and p73 was wild type. Each point represents the mean value from triplicate measurements of p73α and p21 as shown in Table 1.
E2F1 and p21 TA analysis in BC cell line with inactivated p53 and p73
In the H1155 cell line in which p53 and p73 are both inactivated, E2F1 TA level was high, yet p21 TA level was low (Table 1). Consistent with our observation that E2F1 and p73α are correlated in BC tissues (Figure 2), E2F1 and p73α TA levels were both high in H1155. However, the missense mutation in the DNA binding domain of p73 inhibits its transactivational function [35]. Thus, H1155 is a cell line with a naturally occurring p73 mutation that directly supports our hypothesis that p21 is not upregulated by E2F1 directly.
Expression of exogenous p73α is associated with increased p21 and decreased E2F1 TA in Calu-1
To test the hypothesis that p21 transcription is regulated by p73 and not by E2F1 directly, p73α was transiently expressed in the squamous carcinoma cell line Calu-1. This line expresses low levels of endogenous p73α TA (28), low levels of p21 TA (2,000), and low levels of E2F1 TA (8,500). p73α TA was induced over 1,000-fold relative to the mock, 24 hours post-transfection (Figure 4A). Exogenous p73 protein expression was confirmed by Western analysis using an anti-HA antibody specific for an amino terminal tag on p73 (Figure 4B). p21 and E2F1 TA levels were quantified 24 hours post-transfection. While p21 TA was upregulated 90% (p < 0.001) relative to mock-transfected cells, there was a 20% downregulation (p < 0.05) of E2F1 TA (Figure 4C).
Figure 4 Effect of p73α transient expression on p73α, p21, and E2F1 TA in Calu-1 cells. A) p73α TA was induced by over 3 orders of magnitude relative to mock transfected cells. Calu-1 cells were transfected with 5 μg of GFP control plasmid or HA-p73α. B) Total HA-p73 protein was analyzed in mock or p73α transfected cells. 20 μg of lysate were blotted on a PDVF membrane and incubated with an anti-hemagglutinin primary antibody conjugated to HRP. C) p21 was upregulated by 90% and E2F1 was reduced by 20% in p73α transfected cells. Results represent the mean value from triplicate measurements from three independent experiments. Error bars represent the S.E.M. RNA was extracted 24 hours post-transfection and treated with DNaseI. RNA was PCR amplified to rule out the possibility of plasmid contamination. No PCR products were detected.
Gene specific silencing of p73 associated with decreased p21 TA level
According to bivariate analysis shown in Figure 3, p73α and p21 were significantly correlated in cell lines that have inactivated p53. Therefore, we anticipated that gene specific silencing of p73 in one of these lines would directly reduce p21 TA. We used a pool of siRNAs to target all isoforms of p73 in the null-p53 small cell carcinoma N417. This cell line is an appropriate model because it expresses high p21 (34,000), E2F1 (19,000), and endogenous p73α (5,200) TA levels. p73α TA decreased by 80% relative to the non-specific siRNA control (p < 0.05), while p21 TA decreased by 70% (p < 0.01).
Discussion
Relative to normal bronchial epithelial cells (BEC), p21 is upregulated in some bronchogenic carcinoma (BC) tissues and downregulated in others (Tables 1, 2, and 3). Elevation of p21 TA in malignant cells may seem counter-intuitive because it would be expected to slow cell proliferation. However, in some tumors pRb is dysfunctional and feedback signals such as p73 upregulation by E2F1 may increase p21 transcription in an ineffective attempt to prevent phosphorylation of pRb and release of activated E2F1. It is likely that such tumors would be poor candidates for therapy intended to control cell proliferation through specific upregulation of p21 transcription. However, in other tumors, such as those with a genetic profile similar to that of A549 or Calu-1 (Table 1), pRb is intact, and TA levels of E2F1, p73, and p21 are all low. In a tumor such as this, there is reason to believe that the upregulation of p21 transcription would be an effective treatment. However, in order to develop effective p21 gene-specific therapeutics, and biomarkers that predict which tumor will respond, it is necessary to better understand p21 transcriptional regulation.
In some cell types, E2F1 directly regulates p21 [28], however, the data presented here support the hypothesis that E2F1 does not upregulate p21 directly in human BC, but rather, elevated p21 TA results from E2F1 upregulation of p73. This hypothesis, generated initially from empirical observation, is supported by experimental data. There were four supportive empirical observations. First, there was lack of correlation between E2F1 and p21 TA among BC tissues. Second, there was positive correlation among BC tissues between E2F1 and p73 TA. Third, there was positive correlation between p73 and p21 TA among BC cell lines with inactivated p53. Fourth, in the H1155 BC cell line, in which p53 and p73 are inactivated, E2F1 and p73 TA levels were high, but p21 TA level was low. If E2F1 upregulated p21 directly, it would be reasonable to expect that p21 would be upregulated in this cell line, not downregulated.
In experiments designed to directly test this hypothesis, transient exogenous expression of p73α was associated with increased p21 and decreased E2F1 TA levels and siRNA mediated silencing of p73 was associated with decreased p21 TA levels. Although the siRNA experiments support our hypothesis, they are not as supportive as the transient transfection experiments because E2F1 TA was reduced along with p73 and p21. We speculate that this decrease was due to a non-specific effect of the siRNAs. A non-specific effect of the siRNA method has been previously reported for cell cycling genes, including p53 and p21 [36].
Conclusion
In this study, we provide strong empirical and experimental evidence that in human bronchogenic carcinoma, p21 transcription is regulated by p73 but, in contrast to other cell types, not directly by E2F1. This knowledge will facilitate a) development of p21 gene-specific therapeutics necessary for individualized treatment strategies, and b) discovery of biomarkers that will predict which tumors will respond to p21 gene-specific therapeutics.
Methods
Normal cell populations
Normal BEC stock populations (lot numbers: 17378, 6F0333, 6F0395) were obtained from Clonetics (San Diego, CA) and incubated in BEGM medium.
Carcinoma cell lines
Fourteen BC cell lines (Table 1) were obtained from American Type Culture Collection (Rockville, MD) and incubated in RPMI with 10% FBS.
Culture conditions
Normal BEC and BC cells proliferate optimally under different conditions [37]. The medium that is optimal for BC cell lines, RPMI with 10% fetal bovine serum (FBS), induces terminal squamous differentiation in normal BEC [38]. In contrast, BC cell lines do not divide in serum-free media that are optimal for proliferation of normal BEC. BC cell lines were incubated in RPMI with 10% FBS and normal BEC from three individuals were incubated in bronchial epithelial growth medium (BEGM). In order to directly compare with carcinoma cell lines under the same optimal conditions, normal BEC cell populations (17378, 6F0333, 6F0395) were also incubated for 24 hours in RPMI with 10% FBS.
Primary tissue samples
Primary normal BEC and primary BC samples (Table 2) were obtained under IRB approved protocols as previously described [29,39,40]. In each case informed consent was obtained from the patient.
RNA extraction and reverse transcription
Total RNA was extracted by phenol/chloroform methods using TRI-Reagent (Molecular Research Center, Inc., Cincinnati, OH). Approximately 1 μg of total RNA was reverse transcribed using oligo dTs and MMLV-reverse transcriptase (Invitrogen, Inc., Carlsbad, CA).
Transcript abundance measurement
Transcript abundance (TA) was measured by Standardized RT (StaRT)-PCR [30,41,42]. With this method there is an internal standard, within a standardized mixture of internal standards (SMIS) for each gene amplified in the PCR reaction. This enables regular assessment of performance characteristics as recommended by recent FDA guidelines [43]. Among these performance characteristics are reproducibility, lower detection threshold, linear dynamic range, signal to analyte response, false negatives, and false positives. For each gene measured in this study, the lower detection threshold was less than 10 molecules and the linear dynamic range was less than 10 to greater than 107 molecules. False negatives are eliminated due to the presence of an internal standard and false positives are eliminated by using a water control to ensure that there is no contamination within the PCR reaction.
The reagents for analysis of E2F1 and p21 were commercially prepared (Gene Express, Inc., Toledo, OH). To analyze p73, a SMIS containing internal standards for only p73 and β-actin were prepared in this laboratory, because p73 is not in a commercially available SMIS. p73 forward and reverse primers amplify at least four distinct isoforms including α-δ, but do not distinguish between the full-length and ΔN transcripts. A separate pair of primers published by Kartasheva, et al. [44] were used to determine the presence of ΔNp73. The internal standard for p73 was prepared using the forward primer and a competitive template (CT) primer. The CT primer hybridizes upstream of the reverse primer but retains its sequence at the 5' end [45]. This enables the simultaneous amplification of the internal standard and endogenous native template (NT) using only the forward and reverse primers. p73 forward and reverse primer sequences are as follows: p73 F, 5' ACT TTG AGA TCC TGA TGA AG 3' R, 5' CAG ATG GTC ATG CGG TAG TG 3'. Primer sequences for p21 and E2F1 were previously reported [30].
Six SMISs (A-F) were used for all TA measurements, with p73 CT at concentrations ranging from 10-12 (SMIS A) to 10-17 M (SMIS F) and β-actin CT constant at 10-13 M in all six SMISs. The dilution of each cDNA sample that contained 60,000 molecules of β-actin cDNA was determined through calibration to 1 μL of SMIS F. Such calibrated samples were then used in all StaRT-PCR experiments. In some experiments, if the amount of cDNA sample available was low, the cDNA and the SMIS were both diluted 10-fold. Equal volumes of cDNA and SMIS were combined in a master mix along with 30 mM MgCl2 (Idaho Technology, Inc., Idaho Falls, ID), 2 mM dNTPs, Taq Polymerase (Promega, Madison, WI), and RNase-free water. For each TA measurement, a 10 μL reaction volume was PCR amplified in a Rapidcycler (Idaho Technology, Inc., Idaho Falls, ID) for 35 cycles. PCR reactions were denatured for 5 seconds at 94°C, annealed for 10 seconds at 58°C, and elongated for 15 seconds at 72°C.
Plasmid DNA
A pcDNA 3.0 (Invitrogen Inc., Carlsbad, CA) expression vector was kindly provided from the laboratory of Vincenzo DeLaurenzi (University of Rome, Italy). The p73 gene contains a hemagglutinin tag sequence and is regulated by a CMV promoter sequence. An expression vector encoding a CMV regulated green fluorescent protein was obtained from (Gene Therapy Systems, San Diego, CA), and used as a negative control (mock) and determinant for transfection efficiency.
Transient transfection assays
Calu-1 cells were incubated in RPMI supplemented with 10% FBS and grown to confluence. Twenty-four hours prior to transfection, cells were trypsinized and transferred to 60 mm dishes. For transfections, 5 μg of plasmid DNA was diluted in 0.5 mL of serum-free medium with the appropriate concentration of Lipofectamine 2000 transfection reagent (Invitrogen, Inc., Carlsbad, CA). Cells were incubated with DNA-lipid complexes in serum-containing medium for 4–8 hours and subsequently treated with fresh medium. RNA was isolated 24 hours post-transfection and analyzed by StaRT-PCR. 60 molecules of p73 internal standard were sufficient to quantify endogenous p73α TA and 6,000 molecules were required to quantify the combined endogenous and exogenous transcript. To exclude the possibility that the high level measured was partly due to amplification of contaminating plasmid DNA, RNA from p73α transfected Calu-1 cells was PCR-amplified with p73 specific primers. No PCR product was observed.
Western blot analysis
Calu-1 cells were lysed 24 hours post-transfection by three consecutive freeze-thaws in a 0.25 M Tris lysis buffer (Invitrogen, Inc., Carlsbad, CA). Total protein concentration was determined colorimetrically using the bicinchoninic acid (BCA) assay (Pierce, Inc., Rockford, IL). 20 μg of total protein from Calu-1 cells were loaded on a 7% SDS Tris Acetate NuPage gel (Invitrogen, Inc., Carlsbad, CA). Proteins were transferred to a PVDF membrane and incubated with an anti-HA primary antibody conjugated to horseradish peroxidase (Santa Cruz Biotechnology, Santa Cruz, CA). The PVDF was then incubated with chemiluminescent substrates (Santa Cruz Biotechnology, Santa Cruz, CA) and visualized by autoradiography.
siRNA (RNAi)
Approximately 1 million N417 cells were incubated in a six-well plate with RPMI supplemented with 10% FBS. Five siRNA oligonucleotides specific for the p73 gene or a non-specific pooled duplex control (Dharmacon, Inc., Lafayette, CO) were diluted in serum-free media and added to the appropriate concentration of TKO transfection reagent (Mirus Corp., Madison, WI). Untransfected cells were treated with transfection reagent but not siRNA. Transfected cells were incubated continuously with siRNA complexes at a final concentration of 100 nM. RNA was isolated 24 to 48 hours post-transfection and analyzed by StaRT-PCR.
Bivariate and statistical analysis
Pearson correlation and paired-sample and independent T-tests were performed using SPSS 11.5.1 for Windows (SPSS, Chicago, IL) Due to inter-sample variation it was necessary to normalize the data by logarithmic transformation. For each test, a p-value of less than 0.05 was considered statistically significant. Bivariate graphs were created using Microsoft Excel 2000 (Microsoft Corp, Redmond, WA).
Competing interests
ELC, KAW, and JCW each have significant equity interest in Gene Express, Inc., which produces and markets StaRT-PCR reagents used in these studies.
Authors' contributions
MWH was responsible for TA measurement of p73 in all primary samples, TA measurement of E2F1 and p21 in primary samples, siRNA and transient transfection experiments, statistical analysis, and was the primary author of this manuscript. TGG was responsible for the preparation of the p73 SMIS, TA measurement of p73, E2F1, and p21 in all cultured cell lines, and contributed to the research design of this study. ELC, KAW, and CAMR were responsible for TA measurement of E2F1 and p21 and were involved in the acquisition and preparation of primary samples. JCW coordinated and obtained funding for this study and drafted and revised this manuscript. MWH, TGG, and JCW jointly conceived the experiments.
Acknowledgements
We would like to thank Dr. Vincenzo DeLaurenzi for providing the p73 plasmid DNA. We would also like to thank Drs. Dawn-Alita Hernandez, Yongsook Yoon, Jeffrey Hammersley, Ragheb A. Assaly, and Stacie L. Roshong-Denk for helping us acquire primary samples, Dr. Sakik Khuder for helping with statistical analysis, and Charles Knight, Bradley Austermiller, and D'Anna Mullins for their technical assistance.
==== Refs
Zetterberg A Larsson O Wiman KG What is the restriction point? Curr Opin Cell Biol 1995 7 835 842 8608014 10.1016/0955-0674(95)80067-0
Kato J Matsushime H Hiebert SW Ewen ME Sherr CJ Direct binding of cyclin D to the retinoblastoma gene product (pRb) and pRb phosphorylation by the cyclin D-dependent kinase CDK4 Genes Dev 1993 7 331 342 8449399
Ewen ME Sluss HK Sherr CJ Matsushime H Kato J Livingston DM Functional interactions of the retinoblastoma protein with mammalian D-type cyclins Cell 1993 73 487 497 8343202 10.1016/0092-8674(93)90136-E
Weinberg RA The retinoblastoma protein and cell cycle control Cell 1995 81 323 330 7736585 10.1016/0092-8674(95)90385-2
Chellappan SP Hiebert S Mudryj M Horowitz JM Nevins JR The E2F transcription factor is a cellular target for the RB protein Cell 1991 65 1053 1061 1828392 10.1016/0092-8674(91)90557-F
Lees JA Saito M Vidal M Valentine M Look T Harlow E Dyson N Helin K The retinoblastoma protein binds to a family of E2F transcription factors Mol Cell Biol 1993 13 7813 7825 8246996
Kovesdi I Reichel R Nevins JR Identification of a cellular transcription factor involved in E1A trans-activation Cell 1986 45 219 228 2938741 10.1016/0092-8674(86)90386-7
Stevaux O Dyson NJ A revised picture of the E2F transcriptional network and RB function Curr Opin Cell Biol 2002 14 684 691 12473340 10.1016/S0955-0674(02)00388-5
DeGregori J Kowalik T Nevins JR Cellular targets for activation by the E2F1 transcription factor include DNA synthesis- and G1/S-regulatory genes Mol Cell Biol 1995 15 4215 4224 7623816
Stewart MJ Litz-Jackson S Burgess GS Williamson EA Leibowitz DS Boswell HS Role for E2F1 in p210 BCR-ABL downstream regulation of c-myc transcription initiation. Studies in murine myeloid cells Leukemia 1995 9 1499 1507 7658719
Hermeking H Rago C Schuhmacher M Li Q Barrett JF Obaya AJ O'Connell BC Mateyak MK Tam W Kohlhuber F Identification of CDK4 as a target of c-MYC Proc Natl Acad Sci USA 2000 97 2229 2234 10688915 10.1073/pnas.050586197
Jansen-Durr P Meichle A Steiner P Pagano M Finke K Botz J Wessbecher J Draetta G Eilers M Differential modulation of cyclin gene expression by MYC Proc Natl Acad Sci USA 1993 90 3685 3689 8386381
Pardee AB G1 events and regulation of cell proliferation Science 1989 246 603 608 2683075
Harper JW Adami GR Wei N Keyomarsi K Elledge SJ The p21 Cdk-interacting protein Cip1 is a potent inhibitor of G1 cyclin-dependent kinases Cell 1993 75 805 816 8242751 10.1016/0092-8674(93)90499-G
Zhu J Jiang J Zhou W Chen X The potential tumor suppressor p73 differentially regulates cellular p53 target genes Cancer Res 1998 58 5061 5065 9823311
Xiong Y Hannon GJ Zhang H Casso D Kobayashi R Beach D p21 is a universal inhibitor of cyclin kinases Nature 1993 366 701 704 8259214 10.1038/366701a0
Yoshikawa H Nagashima M Khan MA McMenamin MG Hagiwara K Harris CC Mutational analysis of p73 and p53 in human cancer cell lines Oncogene 1999 18 3415 3421 10362363 10.1038/sj.onc.1202677
Yang A Kaghad M Caput D McKeon F On the shoulders of giants: p63, p73 and the rise of p53 Trends Genet 2002 18 90 95 11818141 10.1016/S0168-9525(02)02595-7
Yang A Walker N Bronson R Kaghad M Oosterwegel M Bonnin J Vagner C Bonnet H Dikkes P Sharpe A p73-deficient mice have neurological, pheromonal and inflammatory defects but lack spontaneous tumours Nature 2000 404 99 103 10716451 10.1038/35003607
Jost CA Marin MC Kaelin WG Jr p73 is a simian [correction of human] p53-related protein that can induce apoptosis Nature 1997 389 191 194 9296498 10.1038/38298
Zindy F Eischen CM Randle DH Kamijo T Cleveland JL Sherr CJ Roussel MF Myc signaling via the ARF tumor suppressor regulates p53-dependent apoptosis and immortalization Genes Dev 1998 12 2424 2433 9694806
Bates S Phillips AC Clark PA Stott F Peters G Ludwig RL Vousden KH p14ARF links the tumour suppressors RB and p53 Nature 1998 395 124 125 9744267 10.1038/25867
el-Deiry WS Tokino T Velculescu VE Levy DB Parsons R Trent JM Lin D Mercer WE Kinzler KW Vogelstein B WAF1, a potential mediator of p53 tumor suppression Cell 1993 75 817 825 8242752 10.1016/0092-8674(93)90500-P
Waldman T Kinzler KW Vogelstein B p21 is necessary for the p53-mediated G1 arrest in human cancer cells Cancer Res 1995 55 5187 5190 7585571
Nicholson SA Okby NT Khan MA Welsh JA McMenamin MG Travis WD Jett JR Tazelaar HD Trastek V Pairolero PC Alterations of p14ARF, p53, and p73 genes involved in the E2F-1-mediated apoptotic pathways in non-small cell lung carcinoma Cancer Res 2001 61 5636 5643 11454718
Stiewe T Putzer BM Role of the p53-homologue p73 in E2F1-induced apoptosis Nat Genet 2000 26 464 469 11101847 10.1038/82617
Rodicker F Stiewe T Zimmermann S Putzer BM Therapeutic efficacy of E2F1 in pancreatic cancer correlates with TP73 induction Cancer Res 2001 61 7052 7055 11585734
Hiyama H Iavarone A Reeves SA Regulation of the cdk inhibitor p21 gene during cell cycle progression is under the control of the transcription factor E2F Oncogene 1998 16 1513 1523 9569018 10.1038/sj.onc.1201667
DeMuth JP Jackson CM Weaver DA Crawford EL Durzinsky DS Durham SJ Zaher A Phillips ER Khuder SA Willey JC The gene expression index c-myc × E2F-l/p21 is highly predictive of malignant phenotype in human bronchial epithelial cells Am J Respir Cell Mol Biol 1998 19 18 24 9651176
Willey JC Crawford EL Jackson CM Weaver DA Hoban JC Khuder SA DeMuth JP Expression measurement of many genes simultaneously by quantitative RT-PCR using standardized mixtures of competitive templates Am J Respir Cell Mol Biol 1998 19 6 17 9651175
Vassilev LT Vu BT Graves B Carvajal D Podlaski F Filipovic Z Kong N Kammlott U Lukacs C Klein C In vivo activation of the p53 pathway by small-molecule antagonists of MDM2 Science 2004 303 844 848 14704432 10.1126/science.1092472
Minna JD Battey JF Birrer M Brooks BJ Cuttitta F DeGreve J Gazdar AF Johnson BE Nau MM Sausville EA Chromosomal deletion, gene amplification, alternative processing, and autocrine growth factor production in the pathogenesis of human lung cancer Princess Takamatsu Symp 1986 17 109 122 3332004
Wistuba II Gazdar AF Minna JD Molecular genetics of small cell lung carcinoma Semin Oncol 2001 28 3 13 11479891 10.1053/sonc.2001.25738
Olivier M Eeles R Hollstein M Khan MA Harris CC Hainaut P The IARC TP53 database: new online mutation analysis and recommendations to users Hum Mutat 2002 19 607 614 12007217 10.1002/humu.10081
Huqun Endo Y Xin H Takahashi M Nukiwa T Hagiwara K A naturally occurring p73 mutation in a p73-p53 double-mutant lung cancer cell line encodes p73 alpha protein with a dominant-negative function Cancer Sci 2003 94 718 724 12901798
Scacheri PC Rozenblatt-Rosen O Caplen NJ Wolfsberg TG Umayam L Lee JC Hughes CM Shanmugam KS Bhattacharjee A Meyerson M Short interfering RNAs can induce unexpected and divergent changes in the levels of untargeted proteins in mammalian cells Proc Natl Acad Sci USA 2004 101 1892 1897 14769924 10.1073/pnas.0308698100
Willey JC Moser CE JrLechner JF Harris CC Differential effects of 12-O-tetradecanoylphorbol-13-acetate on cultured normal and neoplastic human bronchial epithelial cells Cancer Res 1984 44 5124 5126 6488172
Lechner JF Haugen A McClendon IA Shamsuddin AM Induction of squamous differentiation of normal human bronchial epithelial cells by small amounts of serum Differentiation 1984 25 229 237 6698333
Warner KA Crawford EL Zaher A Coombs RJ Elsamaloty H Roshong-Denk SL Sharief I Amurao GV Yoon Y Al-Astal AY The c-myc × E2F-l/p21 interactive gene expression index augments cytomorphologic diagnosis of lung cancer in fine-needle aspirate specimens J Mol Diagn 2003 5 176 183 12876208
Crawford EL Khuder SA Durham SJ Frampton M Utell M Thilly WG Weaver DA Ferencak WJ Jennings CA Hammersley JR Normal bronchial epithelial cell expression of glutathione transferase P1, glutathione transferase M3, and glutathione peroxidase is low in subjects with bronchogenic carcinoma Cancer Res 2000 60 1609 1618 10749130
Crawford EL Peters GJ Noordhuis P Rots MG Vondracek M Grafstrom RC Lieuallen K Lennon G Zahorchak RJ Georgeson MJ Reproducible gene expression measurement among multiple laboratories obtained in a blinded study using standardized RT (StaRT)-PCR Mol Diagn 2001 6 217 225 11774186 10.1054/modi.2001.29789
Crawford EL Warner KA Weaver DW Willey JC Quantitative endpoint RT-PCR expression measurement using the Agilent 2100 Bioanalyzer and standardized RT-PCR Agilent Application 2001 1 8
Guidance for Industry 2005 Pharmacogenomic Data Submissions. U.S. Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research, Center for Biologies Evaluation and Research, Center for Devices and Radiological Health
Kartasheva NN Contente A Lenz-Stoppler C Roth J Dobbelstein M p53 induces the expression of its antagonist p73 Delta N, establishing an autoregulatory feedback loop Oncogene 2002 21 4715 4727 12101410 10.1038/sj.onc.1205584
Celi FS Zenilman ME Shuldiner AR A rapid and versatile method to synthesize internal standards for competitive PCR Nucleic Acids Res 1993 21 1047 8451177
|
16014176
|
PMC1185562
|
CC BY
|
2021-01-04 16:36:35
|
no
|
Mol Cancer. 2005 Jul 13; 4:23
|
utf-8
|
Mol Cancer
| 2,005 |
10.1186/1476-4598-4-23
|
oa_comm
|
==== Front
Mol PainMolecular Pain1744-8069BioMed Central London 1744-8069-1-201598750310.1186/1744-8069-1-20ResearchSubstance P-driven feed-forward inhibitory activity in the mammalian spinal cord Nakatsuka Terumasa [email protected] Meng [email protected] Daisuke [email protected] Christopher [email protected] Jennifer [email protected] Hong [email protected] Toyofumi [email protected] Charles [email protected] Robert [email protected] Jianguo G [email protected] McKnight Brain Institute, University of Florida, Gainesville, Florida 32610, USA2 Department of Oral & Maxillofacial Surgery and Diagnostic Sciences, College of Dentistry, University of Florida, Gainesville, Florida 32610, USA3 Department of Orthodontics, College of Dentistry, University of Florida, Gainesville, Florida 32610, USA4 Department of Neuroscience, College of Medicine, University of Florida, Gainesville, Florida 32610, USA5 Comprehensive Center for Pain Research, University of Florida, Gainesville, Florida 32610, USA2005 29 6 2005 1 20 20 6 5 2005 29 6 2005 Copyright © 2005 Nakatsuka et al; licensee BioMed Central Ltd.2005Nakatsuka 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.
In mammals, somatosensory input activates feedback and feed-forward inhibitory circuits within the spinal cord dorsal horn to modulate sensory processing and thereby affecting sensory perception by the brain. Conventionally, feedback and feed-forward inhibitory activity evoked by somatosensory input to the dorsal horn is believed to be driven by glutamate, the principle excitatory neurotransmitter in primary afferent fibers. Substance P (SP), the prototypic neuropeptide released from primary afferent fibers to the dorsal horn, is regarded as a pain substance in the mammalian somatosensory system due to its action on nociceptive projection neurons. Here we report that endogenous SP drives a novel form of feed-forward inhibitory activity in the dorsal horn. The SP-driven feed-forward inhibitory activity is long-lasting and has a temporal phase distinct from glutamate-driven feed-forward inhibitory activity. Compromising SP-driven feed-forward inhibitory activity results in behavioral sensitization. Our findings reveal a fundamental role of SP in recruiting inhibitory activity for sensory processing, which may have important therapeutic implications in treating pathological pain conditions using SP receptors as targets.
==== Body
Feedback/feed-forward inhibitory modulation driven by glutamate has been well studied in the dorsal horn of the spinal cord [1-3]. Little is know whether feedback/feed-forward inhibitory active may be driven in a glutamate-independent manner. A number of neuropeptides including substance P (SP) are also released from nociceptive primary afferent fibers [4]. SP has been regarded as a pain substance for decades [5-7], as supported by studies, including chemical ablation of lamina I neurons expressing the SP receptors [8], genetic disruption of the genes encoding substance P [9] and its receptors [10]. The nociceptive function of SP is mainly attributed to the activation of NK1 receptors (NK1R) that are expressed on nociceptive projection neurons located in lamina I of the dorsal horn [8,11,12]. It is unknown whether endogenously released SP can directly drive, in a glutamate-independent manner, inhibitory activity within the spinal cord to control nociceptive responses.
We performed patch-clamp recordings from dorsal horn neurons in lamina V (Figure 1a), a region important for nociceptive transmission and modulation [1,2]. When primary afferent fibers (dorsal roots) were briefly stimulated electrically (500 μA, 5 stimuli in 2.5 sec), EPSCs (excitatory postsynaptic currents) were recorded from lamina V neurons (Figure 1b). All EPSCs were blocked by ionotropic glutamate receptor antagonists 20 μM CNQX plus 50 μM APV (Figure 1b) or 3 mM kynurenic acid [3]. Brief stimulation of primary afferent fibers also evoked IPSCs (inhibitory postsynaptic currents). These immediate IPSCs (Figure 1c top) were driven by glutamatergic synaptic input, or glutamate-driven feed-forward inhibitory activity, because they were completely abolished in the presence of CNQX plus APV (Figure 1c bottom). However, when prolonged stimulation was applied (500 μA, 20 Hz, 1 min), a robust and long-lasting increase of IPSC frequency and amplitude was recorded in the presence of CNQX plus APV (n = 5, Figure 1d–f) or kyurenic acid (KA, n = 7, Figure 2c). These results revealed a feed-forward inhibitory (FFI) pathway not driven by glutamate.
Figure 1 Feed-forward inhibitory activity in the absence of glutamatergic driving force. a, Rat spinal cord slice with attached dorsal root. A portion of the root is sucked into a stimulation electrode. Recordings were made in lamina V. b, Five consecutive traces show EPSCs evoked by electrical stimulation (top). The EPSCs were abolished in the presence of 20 μM CNQX plus 50 μM APV (bottom). Vh = -60 mV. c, In the same cell, stimulation evoked IPSCs (top), which were abolished in the presence of 20 μM CNQX and 50 μM APV (bottom). Vh = -10 mV. d, In the same cell, trains of stimulation (20 Hz for 1 min) increased IPSCs in the presence of 20 μM CNQX plus 50 μM APV. Top trace was IPSCs recorded before and after electrical stimulation. The bottom 3 traces are at an expanded scale. Vh = -10 mV. e&f, Time course of IPSC frequency (e) and amplitude (f). Horizontal bars indicate stimulation. Overall, at peak responses, IPSC frequency increased to 376 ± 47% of control (n = 5, P < 0.05); IPSC amplitude increased to 228 ± 74% of control (n = 5, P < 0.05). Similar results were also obtained in the presence of 3 mM kynurenic acid (see Figure 2c). g–j, Capsaicin-induced increases in inhibitory activity in the absence of glutamatergic driving force. g, The top trace is a continuous recording of IPSCs from a rat lamina V neuron before and following the application of 2 μM capsaicin in the presence of 3 mM kynurenic acid. The bottom 2 traces are at an expanded scale. h, The time course of IPSC frequency in (g). bin width: 10s. i&j, Capsaicin-induced increases in IPSC frequency (i) and amplitude (j) recorded from 6 rat lamina V neurons in the presence of 3 mM kynurenic acid.
Figure 2 Feed-forward inhibitory activity driven by SP through NK1 receptor activation. a, Effects of exogenously applied neuropeptides on IPSCs recorded in rat lamina V neurons. Neuropeptides tested include galanin (0.3 μM, n = 4), somatostatin (2 μM, n = 4), NPY (1 μM, n = 4), CGRP (0.5 μM, n = 5), and substance P (1 μM, n = 6). b, Antagonism of capsaicin-induced increases in inhibitory activity in rat lamina V neurons. Capsaicin was applied in the presence of neurokinin receptor antagonists APTL (10 μM, n = 5), L-733, 060 (2 μM, n = 4), L-732,138 (100 μM, n = 5), SB222200 (2 μM, n = 8), and the Gi/o protein blockers pertussis toxin (2 μg/ml, n = 4) and NEM (100 μM, n = 5). c, Antagonism of electrical stimulation-induced increases in inhibitory activity by the SP antagonist APTL (10 μM, n = 7). Recordings were from rat lamina V neurons. d, Effects of capsaicin and SP on IPSCs in NK1R+/+ mice (n = 8 for capsaicin, n = 10 for SP) and NK1R-/- mice (n = 21 for cap, n = 11 for SP). Experiments were performed in the presence of 3 mM kynurenic acid (a–d) or 20 μM CNQX plus 50 μM APV (some experiments in a&b). e, Images show a cultured GIN mice EGFP neuron (arrow indicated) before (left) and after loading the Ca2+ indicator Fluor-3 (middle), and following application of 100 nM SP (Right). The experiment was performed in the presence of 500 nM TTX and 3 mM kynurenic acid. Similar results were obtained from 22 EGFP neurons. f, The florescence image shows a spinal cord slice obtained from a GIN mouse. An EGFP neuron in lamina V is indicated by a small box and enlarged in a bigger box. g, Non-adaptive action potential firing induced by depolarizing current (50 pA) in the EGFP neuron. h, Application of 1 μM SP produced a prolonged depolarization and action potential firings (top) in the same cell. The dotted line (bottom) shows, at expanded scale, the membrane depolarization (action potentials are omitted for clarity). Kynurenic acid (3 mM) was present throughout the experiments (n = 7).
We used capsaicin, the active ingredient of hot chili peppers, to stimulate primary afferent fibers. Capsaicin is widely used as a natural stimulant for studying nociception. It excites nociceptive primary afferent fibers to release glutamate and neuropeptides including substance P through activation of TRPV1 receptors [13-15]. Capsaicin (2 μM) produced a robust and long-lasting increase in IPSC frequency and amplitude in the presence of 3 mM kynurenic acid (Figure 1g–j). The capsaicin effects were similar in the presence of kynurenic acid or other glutamate receptor antagonists (Additional file: 1, Figure 1a–c), indicating that the effects were unlikely due to an incomplete block of glutamate-driven FFI. Inhibitory neurons in lamina V use both GABA and glycine as co-transmitters [16], and increases of IPSCs by capsaicin were completely abolished in the presence of 20 μM bicuculline and 2 μM strychnine (n = 8).
It is unknown whether, transmitters, other than glutamate released from primary afferent fibers can directly drive inhibitory circuitry in the spinal cord. If a transmitter can drive FFI, exogenous application should increase inhibitory activity. We examined neuropeptides thought to be released from primary afferent fibers. Galanin (300 nM), NPY (neuropeptide Y, 1 μM), somatostatin (2 μM), and CGRP (calcitonin gene-related peptide, 0.5 μM) were tested, but none increased IPSCs (Figure 2a). However, SP significantly increased inhibitory activity under conditions when ionotropic glutamate receptors were blocked; SP increased IPSC frequency to ~350% of control (Figure 2a, n = 6) and amplitude to ~200% of control (n = 6).
If endogenously SP drives FFI following capsaicin stimulation, SP receptor antagonists should attenuate FFI. APTL (D-Arg1, D-Pro2, D-Trp7,9, Leu11]-Substance P, 10 μM), a neurokinin receptor antagonist, substantially blocked capsaicin-induced increases in IPSCs (Figure 2b). L-733,060 (2 μM) and L-732,138 (100 μM), two NK1 receptor (NK1R) antagonists, also inhibited capsaicin-induced increases in IPSCs. NK3 receptors are expressed in the dorsal horn, but SB22200 (2 μM), a NK3 receptor antagonist, did not significantly attenuate capsaicin-induced increases of IPSCs. Similar to capsaicin stimulation, we found that FFI elicited by electrical stimulation was largely abolished by the NK1R antagonist ATPL (Figure 2c). These results suggest that endogenous SP drives inhibitory activity.
NK1Rs couple with either the pertussis toxin (PTX)-insensitive Gq/G11 family [17] or PTX-sensitive Gi/Go family depending on cell types [18,19]. To elucidate which type of G-proteins was involved in SP-driven FFI, PTX was tested. We found that capsaicin-induced increases in inhibitory synaptic activity were completely abolished when spinal cord slices were pretreated with PTX (Figure 2b). Capsaicin-induced increases of inhibitory synaptic activity were also completely blocked in the presence of NEM (N-Ethylmaleimide), a Gi/Go protein inhibitor (Figure 2b). Thus, PTX-sensitive G-protein is involved in SP-driven FFI.
To confirm the involvement of NK1Rs, we used spinal cord slice preparations obtained from both wild type (NK1R+/+) and NK1R knockout mice (NK1R-/-). While capsaicin increased IPSCs in NK1R+/+ mice, it had no effect in NK1R-/- mice (Figure 2d). Consistent with this result, SP (1 μM) did not increase IPSCs in NK1R-/- mice, but did substantially increase IPSCs in NK1R+/+ mice (Figure 2d, Additional file: 1, Figure 2a,b). Thus, endogenous SP released from primary afferent fibers drives inhibitory activity (SP-driven FFI).
Possible cellular mechanisms of SP-driven FFI include i) direct excitation of inhibitory neurons; ii) via intermediate steps; and/or iii) through synaptic modulation. If NK1Rs are expressed on dorsal horn inhibitory interneurons [20], SP may directly excite inhibitory neurons. To test this possibility, we used dorsal horn neuron cultures made from GIN mice, a strain of transgenic mice that express EGFP (enhanced green fluorescent protein) under control of a promoter for GAD67 [21]. In GIN mice, almost all EGFP neurons examined in the dorsal horn are inhibitory neurons [22]. As shown in Figure 2e, SP (100 nM) increased intracellular Ca2+ in ~30% (23/77) of EGFP neurons tested in the presence of 500 TTX and 3 mM kynurenic acid. We determined whether EGFP neurons in lamina V responded to SP using spinal cord slices prepared from GIN mice (Figure 2f). Most EGFP neurons recorded (64%) showed non-adaptive action potential firing in response to membrane depolarization (Figure 2g). Of 22 EGFP neurons examined, 7 (~30%) responded to 1 μM SP with prolonged membrane depolarization (5 ± 1 mV, n = 7) and action potential firing (Figure 2h). These results suggest that a cellular mechanism of SP-driven FFI is direct excitation of inhibitory interneurons by SP.
We found that SP (Additional file: 1, Figure 3a–c) and capsaicin (n = 12) had no effect on mIPSCs. SP also did not affect paired-pulse eIPSC ratio and corresponding eIPSC ratio (Additional file: 1, Figure 3d–f). These results suggest that SP/NK1R-mediated increases of IPSCs represent feed-forward neuronal activity rather than pre- or post-synaptic modulation at inhibitory synaptic junction sites.
We evaluated the extent SP-driving inhibitory activity contributes to the total inhibitory activity under normal conditions, i.e. without blocking glutamate-driven FFI. We also compared temporal phases between SP-driven FFI and glutamate-driven FFI. In NK1R+/+ mice, IPSC frequency and amplitude were increased after trains of electrical stimulation (Figure 3a,c,d), similar to the results when glutamatergic driving force was blocked (Figure 1a–f & Figure 2c). In contrast, in NK1R-/- mice, IPSCs were not significantly changed after the same trains of stimulation (Figure 3b,c,d). We examined IPSCs during electrical stimulation and found that, in both NK1R+/+ and NK1R-/- mice, IPSCs were elicited pulse-by-pulse immediately following each stimulus. These immediate IPSCs represented glutamate-driven FFI because they could be blocked by ionotropic glutamate receptor antagonists (see Figure 1c). Since the pulse-by-pulse inhibitory activity was seen in both NK1R+/+ and NK1R-/- mice, but the long-lasting increases in IPSCs after trains of stimulation were only observed in NK1R+/+ mice, it suggests that the latter is driven by substance P through NK1R activation. Similar to electrical stimulation, a large and long-lasting increase in inhibitory synaptic activity was observed in NK1R+/+ mice but not in NK1R-/- mice after capsaicin stimulation in the absence of ionotropic glutamate receptor antagonists (Figure 3e). Thus, SP-driven FFI and glutamate-driven FFI have distinct temporal phases.
Figure 3 SP-driven inhibitory activity under conditions when glutamatergic driving force is intact. All experiments were performed in bath solution without glutamate receptor antagonists. a, A continuous recording of IPSCs from a lamina V neuron of a NK1R+/+ mouse. Four traces (bottom) show, at an expanded scale, the IPSCs before, during, 1 sec after, and 9 min after trains of stimulation. The trace during stimulation is at a more expanded scale to show pulse-by-pulse eIPSCs. b, Same as a except the experiment was performed on a NK1R-/- mouse. The pulse-by-pulse eIPSCs (second trace of lower panel) were similar to those of NK1R+/+ mice, but ISPCs returned to the basal level immediately after termination of the train stimulation (third trace of lower panel). c, Time course of IPSC frequency (top) and amplitude (bottom). IPSCs during stimulation are not included. d, Peak IPSC frequency and amplitude after trains of stimulation in NK1R+/+ (n = 6) and NK1R-/- mice (n = 8). In a–d, stimulation was applied at intensity of 500 μA and a frequency of 20 Hz. e, Capsaicin-induced changes of IPSCs in lamina V neurons of NK1R+/+ mice (n = 6) and NK1R-/- mice (n = 16).
One physiological role of SP-driven FFI may be to balance neuronal activity and counteract SP-mediated nociceptive responses in the dorsal horn. To examine this potential physiological function, a behavioral model was used to see if blockade of SP-driven FFI, using an NK1R antagonist, causes behavioral sensitization to nociceptive stimuli. However, an NK1R antagonist will also block SP-mediated nociceptive response, thus interfering with the observation of a functional change following blockade of SP-driven FFI. To solve this complication, we chemically ablated NK1R-expressing neurons in the superficial lamina (Figure 4a&b); most of these neurons are nociceptive projection neurons responsible for SP-mediated nociception [8] and SP-evoked descending modulation [23]. Ablating NK1R-expressing neurons in the superficial lamina was achieved by intrathecally applying substance P-conjugated saporin (SP-SAP) [8], a targeted toxin for NK1R-expressing neurons (Figure 4a&b). In these animals, NK1R-expressing neurons in deep laminae remain intact or less affected [8,12,24]. To verify that SP-driven FFI remains intact, we used spinal cord slices prepared from SP-SAP treated animals and made recordings from lamina V neurons. Capsaicin was found to increase IPSCs to a similar degree in animals with (Figure 4c) or without SP-SAP treatment (Figure 1g–j, Supplementary Figure 1), indicating that the ablation did not affect SP-driven FFI in lamina V.
Figure 4 Assessment of the role of SP-driven inhibitory activity in behavioral responses to noxious stimuli. a&b, Micrographs show NK1 receptor immunoreactivity in the lamina I region of a normal rat (a) and 14 days following intrathecal application of SP-SAP (b). c, Capsaicin-induced increases of IPSCs recorded from lamina V neurons of SP-SAP treated rats (n = 5). d, The first set of bars show baseline of reflexive lick/guard response to heat stimuli at 44.5°C in normal (open bar, n = 8) and SP-SAP rats (solid bar, n = 8). The second set of bars show sensitization of behavioral responses by capsaicin in normal rats (n = 8) and attenuation of the behavioral response in SP-SAP rats (n = 8). The third set of bars show that the NK1 receptor antagonist CP97 attenuated behavioral responses in normal rats (n = 8) but sensitize behavioral response in SP-SAP rats (n = 8) (see Additional file: 1).
SP-SAP treated animals were used to access if the SP-driven FFI plays a role in controlling nociceptive behavioral responses. Reflexive lick/guard (L/G) responses to nociceptive heat stimuli at 44.5°C [24] were determined. Both the control and SP-SAP groups showed similar baseline responses to noxious stimuli (Figure 4d) [8]. Control rats showed behavioral sensitization following application of capsaicin cream to the planter surface, but a substantial attenuation of behavioral sensitization was observed in parallel experiments carried out in SP-SAP animals [8]. To examine whether the NK1-expressing neurons in deeper laminae of SP-SAP animals may intrinsically control behavioral responses to nociceptive heat stimuli, the behavioral responses were determined following blockade of NK1Rs by its antagonist CP-96,345 (36 nmol). Nociceptive reflexes showed sensitization when CP-96,345 was applied in SP-SAP animals, but behavioral hypersensitivity was attenuated by CP-96,345 in control animals (Figure 4d). The opposite effects of NK1R antagonists between normal and SP-SAP animals indicate a dual function of NK1Rs in nociceptive processing in vivo. The behavioral sensitization by the NK1R antagonist in SP-SAP animals revealed a role of SP-driven FFI in controlling nociceptive responses.
SP-driven FFI is a novel sensory processing mechanism. The unique feature is its temporal phase that extends long time after stimulation. This is distinct from glutamate-driven feedback/feed-forward inhibitory activity. Compromising SP-driven FFI can result in sensory hypersensitivity, providing implications in sensory pathology and therapeutics that targets neurokinin system [8,12].
Methods
Electrophysiology recordings were performed on lamina V neurons in transverse spinal cord slices prepared from rats, NK1R+/+ and NK1R-/- mice, and GIN mice. Sprague Dawley rats were used at the age of 35 ± 7 days. Balb/c NK1R knockout mice (NK1R-/-) and GIN mice [21] (Jackson Laboratory) were used at the age of 21–35 days. Transverse slices were sectioned (600 μm in thickness) from spinal cord L5 segments of these animals [25]. In each experiment, a spinal cord slice was transferred to a recording chamber. The slice was superfused with a bath solution containing (in mM) 117NaCl, 3.6KCl, 2.5CaCl2, 1.2MgCl2, 1.2NaH2PO4, 25NaHCO3, and 11glucose, equilibrated with 95% O2 and 5% CO2, pH 7.35, 24°C. For voltage-clamp recordings, electrodes (~5 MΩ) were filled with a solution containing (in mM): Cs2SO4 110, CaCl2 0.5, MgCl2 2, Tea-Cl 5, EGTA 5, HEPES 5, pH 7.2. For current-clamp recordings, electrodes were filled with a solution containing (in mM): potassium gluconate 120, KCl 20, MgCl2 2, Na2ATP 2, NaGTP 0.5, HEPES 20, EGTA 0.5, pH 7.2. In experiments to determine EPSCs, cells were held at -60 mV. When IPSCs were recorded, cells were held at -10 mV. Unless otherwise indicated, IPSCs were recorded in the presence of 3 mM kynurenic acid. Miniature IPSCs (mIPSCs) were recorded in the presence of 500 nM TTX.
To stimulate primary afferent fibers, capsaicin (2 μM) was bath applied for 1 min. Capsaicin-induced increases in inhibitory activity were characterized pharmacologically with APTL (D-Arg1, D-Pro2, D-Trp7,9, Leu11]-Substance P, 10 μM), L-733, 060 (2 μM), L-732,138 (100 μM), SB222200 (2 μM), pertussis toxin (PTX, 2 μg/ml), NEM (N-Ethylmaleimide; 100 mM). Except for PTX, all compounds were applied through bath solution; all antagonists and blockers were pre-applied for 10 min. In experiments using PTX, spinal cord splices were pretreated with 2 μg/ml PTX for 2–4 hours.
To elicit feed-forward inhibitory activity by electrical stimulation, dorsal roots were stimulated electrically through a suction electrode. Stimulation was applied at an intensity of ~500 μA and pulse duration of 100 μsec. Unless otherwise indicated, stimulation was applied in a train of pulses that had a frequency of 20 Hz and duration of 1 min. Recordings were performed in the bath solution containing (in mM) 117 NaCl, 3.6 KCl, 4 CaCl2, 0.5 MgCl2, 1.2 NaH2PO4, 25 NaHCO3, 11 glucose, equilibrated with 95% O2 and 5% CO2.
To examine whether SP had effects on evoked IPSCs, paired-pulse evoked IPSCs were examined before and following application of 1 μM SP. Paired-pulse evoked IPSCs were elicited by focal stimulation in lamina V near the recorded neurons. Stimuli were applied at the intensity of 50–150 μA, pulse duration of 100 μs, and paired-pulse interval of 100 ms. The interval between two sets of paired-pulses was 10 s.
Calcium Imaging was performed on dorsal horn neuron cultures (5–7 days) made from neonatal GIN mice [26]. Cells were perfused with bath solution containing (in mM): 150 NaCl, 5 KCl, 2 MgCl2, 2 CaCl2, 10 glucose, 10 HEPES, pH 7.4; 500 nM TTX and 3 mM kynurenic acid. EGFP neurons were first identified and an image was taken. Cells were then loaded with the Ca2+ indicator Fluo-3 on the stage of microscope. Subsequently, calcium imaging was performed [27], the effect of SP (100 nM) on EGFP neurons was tested.
To chemically ablating NK1R-expressing lamina I neurons with SP-SAP [8,24], a 32 g catheter was inserted into the lumbosacral subarachnoid space (L6-S1) of adult rats (250–300 g) [28] and SP-SAP (300 ng, substance P-conjugated saporin, Advanced Targeting System) was injected through the catheter to the lumbar enlargement. Fourteen days after this procedure, animals were used for in vitro electrophysiological recordings or in vivo behavioral tests. Controls were animals after sham operation.
Behavioral tests were performed on 8 SP-SAP treated animals and 8 control animals. Reflexive lick/guard responses were assessed in two consecutive ten-minute trials involving 36.0°C (pre-test) trial and then a 44.5°C (test) [24,29]. Lick responses were defined as a stereotyped lifting of the hindlimb followed by holding and licking the hindpaw. Guard responses were defined as an exaggerated raising of the hindlimb. Peripheral sensitization of behavioral responses was induced by application of capsaicin cream (1%) to the planter surface of one hindpaw. Reflexive responses were assessed three hours after application. To test the effects on behavioral responses following blockade of NK1 receptors, CP-96,345 (36 nmol), an NK1 antagonist was applied through the catheter 10 min before behavioral tests.
NK1 receptor immunostaining was performed after behavioral tests to confirm the effective removal of NK1R-expressing lamina I neurons in SP-SAP treated animals. NK1R immunostaining was performed using a polyclonal anti-NK1R serum (1:3000) on a series of sections (100 μm in thickness) cut from L5 of the spinal cord.
Analysis of synaptic events, including threshold setting and peak identification criteria, were performed according to a method previously described [26]. For calcium imaging experiments, responsive neurons are defined as ΔF/Fo > 20%. The duration of behavioral responses were collected by custom software (EVENTLOG) across testing sessions for all rats [24,29]. Unless otherwise indicated, data represent Mean ± SEM, * p < 0.05, student-t test. Statistical analysis of behavioral responses was performed by ANOVA, followed by Newman-Keuls post-tests.
Supplementary Material
Additional file 1
Substance P-driven inhibitory activity in the mammalian spinal cord
Click here for file
Acknowledgements
We thank Drs. A. MacDermott, MW. Salter, L. Wang, and M. Zhuo for comments on an early version of the manuscript, J. Palmer for his assistance in behavioral testing. This work was supported by National Science Foundation Grant 0237317 (J.G.G) and National Institute of Health Grant NS38254 (J.G.G).
==== Refs
Melzack R Wall PD Pain mechanisms: a new theory Science 1965 150 971 9 5320816
Willis WD JrCoggeshall RE Sensory mechanisms of the spinal cord 2004 Kluwer Academic/Plenum Publishers, New York 271 388
Yoshimura M Jessell T Amino acid-mediated EPSPs at primary afferent synapses with substantia gelatinosa neurones in the rat spinal cord J Physiol 1990 430 315 35 1982314
Hokfelt T Ljungdahl A Terenius L Elde R Nilsson G Immunohistochemical analysis of peptide pathways possibly related to pain and analgesia: enkephalin and substance P Proc Natl Acad Sci U S A 1977 74 3081 5 331326
Potter GD Guzman F Lim RK Visceral pain evoked by intra-arterial injection of substance P Nature 1962 193 983 4 14488287
Jessell TM Neurotransmitters and CNS disease. Pain Lancet 1982 2 1084 8 6182432 10.1016/S0140-6736(82)90014-9
Iversen L Substance P equals pain substance? Nature 1998 392 334 5 9537317 10.1038/32776
Mantyh PW Rogers SD Honore P Allen BJ Ghilardi JR Li J Daughters RS Lappi DA Wiley RG Simone DA Inhibition of hyperalgesia by ablation of lamina I spinal neurons expressing the substance P receptor Science 1997 278 275 9 9323204 10.1126/science.278.5336.275
Cao YQ Mantyh PW Carlson EJ Gillespie AM Epstein CJ Basbaum AI Primary afferent tachykinins are required to experience moderate to intense pain Nature 1998 392 390 4 9537322 10.1038/32897
De Felipe C Herrero JF O'Brien JA Palmer JA Doyle CA Smith AJ Laird JM Belmonte C Cervero F Hunt SP Altered nociception, analgesia and aggression in mice lacking the receptor for substance P Nature 1998 392 394 7 9537323 10.1038/32904
Ikeda H Heinke B Ruscheweyh R Sandkuhler J Synaptic plasticity in spinal lamina I projection neurons that mediate hyperalgesia Science 2003 299 1237 40 12595694 10.1126/science.1080659
Nichols ML Allen BJ Rogers SD Ghilardi JR Honore P Luger NM Finke MP Li J Lappi DA Simone DA Mantyh PW Transmission of chronic nociception by spinal neurons expressing the substance P receptor Science 1999 286 1558 61 10567262 10.1126/science.286.5444.1558
Caterina MJ Schumacher MA Tominaga M Rosen TA Levine JD Julius D The capsaicin receptor: a heat-activated ion channel in the pain pathway Nature 1997 389 816 24 9349813 10.1038/39807
Theriault E Otsuka M Jessell T Capsaicin-evoked release of substance P from primary sensory neurons Brain Res 1979 170 209 13 466404 10.1016/0006-8993(79)90957-0
Yaksh TL Farb DH Leeman SE Jessell TM Intrathecal capsaicin depletes substance P in the rat spinal cord and produces prolonged thermal analgesia Science 1979 206 481 3 228392
Maxwell DJ Todd AJ Kerr R Colocalization of glycine and GABA in synapses on spinomedullary neurons Brain Res 1995 690 127 32 7496799 10.1016/0006-8993(95)00613-U
Khawaja AM Rogers DF Tachykinins: receptor to effector Int J Biochem Cell Biol 1996 28 721 38 8925404 10.1016/1357-2725(96)00017-9
Laniyonu A Sliwinski-Lis E Fleming N Different tachykinin receptor subtypes are coupled to the phosphoinositide or cyclic AMP signal transduction pathways in rat submandibular cells FEBS Lett 1988 240 186 90 2461321 10.1016/0014-5793(88)80365-X
Quartara L Maggi CA The tachykinin NK1 receptor. Part I: ligands and mechanisms of cellular activation Neuropeptides 1997 31 537 63 9574822 10.1016/S0143-4179(97)90001-9
Littlewood NK Todd AJ Spike RC Watt C Shehab SA The types of neuron in spinal dorsal horn which possess neurokinin-1 receptors Neuroscience 1995 66 597 608 7543982 10.1016/0306-4522(95)00039-L
Oliva AA JrJiang M Lam T Smith KL Swann JW Novel hippocampal interneuronal subtypes identified using transgenic mice that express green fluorescent protein in GABAergic interneurons J Neurosci 2000 20 3354 68 10777798
Heinke B Ruscheweyh R Forsthuber L Wunderbaldinger G Sandkuhler J Physiological, neurochemical and morphological properties of a subgroup of GABAergic spinal lamina II neurones identified by expression of green fluorescent protein in mice J Physiol 2004 560 249 66 15284347 10.1113/jphysiol.2004.070540
Suzuki R Morcuende S Webber M Hunt SP Dickenson AH Superficial NK1-expressing neurons control spinal excitability through activation of descending pathways Nat Neurosci 2002 5 1319 26 12402039 10.1038/nn966
Vierck CJ JrKline RH Wiley RG Intrathecal substance p-saporin attenuates operant escape from nociceptive thermal stimuli Neuroscience 2003 19 223 32 12763083 10.1016/S0306-4522(03)00125-8
Takeda D Nakatsuka T Papke R Gu JG Modulation of inhibitory synaptic activity by a non-alpha4beta2, non-alpha7 subtype of nicotinic receptors in the substantia gelatinosa of adult rat spinal cord Pain 2003 101 13 23 12507696 10.1016/S0304-3959(02)00074-X
Gu JG Albuquerque C Lee J MacDermott AB Synaptic strengthening through activation of Ca2+-permeable AMPA receptors Nature 1996 381 793 796 8657283 10.1038/381793a0
Tsuzuki K Xing H Ling J Gu JG Menthol-induced Ca2+ release from presynaptic Ca2+ stores potentiates sensory synaptic transmission J Neurosci 2004 24 762 761 14736862 10.1523/JNEUROSCI.4658-03.2004
Jasmin L Ohara PT Long-term intrathecal catheterization in the rat J Neurosci Methods 2001 110 81 89 11564527 10.1016/S0165-0270(01)00420-4
King CD Devine DP Vierck CJ Rodgers J Yezierski RP Differential effects of stress on escape and reflex responses to nociceptive thermal stimuli in the rat Brain Res 2003 987 214 22 14499966 10.1016/S0006-8993(03)03339-0
|
15987503
|
PMC1185563
|
CC BY
|
2021-01-04 16:40:08
|
no
|
Mol Pain. 2005 Jun 29; 1:20
|
utf-8
|
Mol Pain
| 2,005 |
10.1186/1744-8069-1-20
|
oa_comm
|
==== Front
RetrovirologyRetrovirology1742-4690BioMed Central London 1742-4690-2-441600460910.1186/1742-4690-2-44Short ReportDetermination of the relative amounts of Gag and Pol proteins in foamy virus particles Cartellieri Marc [email protected] Wolfram [email protected]öder Ottmar [email protected] Dirk [email protected] Axel [email protected] Institut für Virologie, Medizinische Fakultät, Technische, Universität Dresden, Germany2 Institut für Virologie und Immunbiologie, Universität Würzburg, Germany2005 8 7 2005 2 44 44 18 4 2005 8 7 2005 Copyright © 2005 Cartellieri et al; licensee BioMed Central Ltd.2005Cartellieri 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.
We determined the relative ratios of Gag and Pol molecules in highly purified virions of spumaretroviruses or foamy viruses (FVs) using monoclonal antibodies and bacterially expressed reference proteins. We found that the cleaved p68Gag moiety dominates in infectious FVs. Furthermore, approximate mean ratios in FV are 16:1 (pr71Gag plus p68Gag:p85RT),12:1 (p68Gag:p85RT), and 10:1 (pr71Gag plus p68Gag:p40IN). Thus, the results indicate that FVs have found a way to incorporate approximately as much Pol protein into their capsids as orthoretroviruses, despite a completely different Pol expression strategy.
==== Body
One of the central features of Spumaretrovirinae, which distinguishes them from Orthoretrovirinae, is the expression of a Pol precursor protein independently of the Gag protein from a spliced mRNA [1-3]. This mechanism of Pol generation raises several interesting questions: (i) How is Pol expression regulated? (ii) How is the Pol protein incorporated into the virion? (iii) And how much Pol protein is actually present in infectious viruses? While question one has, to our knowledge, not been investigated yet, answers to question two are emerging [4,5]. Here we tried to address question three.
Theoretical lines of argument favor the view that only a few molecules of Pol may be incorporated into a FV particle. The reverse transcriptase (RT) is the main enzymatic subunit of the Pol precursor [6]. This enzyme has been shown to be of much higher processivity than orthoretroviral RTs [7,8]. Therefore, it was argued that FVs probably encapsidate less of their highly active Pol protein compared to orthoretroviruses [7,8]. Following this line of argument, it is noteworthy that the FV protease (PR) is contained within the 85 kD Pol subunit, which also bears the RT/RNaseH [6]. However, in contrast to orthoretroviruses, the FV PR cleaves the cognate Gag protein only once prior to or during budding [6]. Therefore, FV may need less amounts of PR enzyme than orthoretroviruses.
Furthermore, experiments aimed to elucidate the mechanism of Pol protein particle incorporation (the above raised question two) indicated that Pol interacts with specific sequences on the (pre-) genomic RNA and that RNA serves as a bridging molecule between Gag (capsid) and Pol [4,5]. Two distinct elements on the RNA have been identified, which probably facilitate this interaction [4]. This can be regarded as another argument in support of only trace amounts of encapsidated Pol protein.
Here we wanted to investigate the approximate relative ratio of Pol to Gag molecules in infectious virions on a biochemical level to get an estimate of the FV particle composition using the prototypic FV (PFV) as a model. We did not attempt to determine absolute numbers of Gag and Pol molecules per particle.
Prokaryotic expression and purification of viral proteins
The cloning strategy [9,10] and the purified recombinant proteins are depicted in Fig. 1. pETgag2 was made by digestion of pETgagl [11] with AdeI, T4 DNA polymerase treatment, and recutting with NdeI. A 1.9 kb gag gene (aa 1–625 of 648 aa) was inserted into pET22b (Novagen) in-frame to the C-terminal histidine tag after SacI, T4 DNA polymerase, and NdeI treatment. The PFV pol domain encoding the 85 kD PR, RT, and RNaseH subunits was amplified with primers #1217 (5'tc cacatatgaatcctcttcagctgttacagccgc) and #1414 (5'tattacactcgagcacataacttccttg), which bear NdeI and XhoI restriction sites (underlined). pETpol2 was made from pET22b and the amplimer using these enzymes. The integrase (IN; aa 751–1143) construct pETpol3 was made alike with #1219 (5'gttatgtgcatatgtgtaataccaaaaaacc) and #1413(5'tgcgctctcgagatttttttccaaatg). All plasmids were sequenced in their FV parts to verify correct insertions and to exclude PCR artifacts.
Figure 1 Bacterial expression of PFV gag and pol genes. (A) Strategy to insert the gag and pol open reading frames into the bacterial expression vector pET22b. The FV gene fragments are placed in frame to a C-terminal histidine (HIS) tag. (RBS), prokaryotic ribosomal binding site. (B) Coomassie stain of recombinant proteins which were purified via the C-terminal HIS-tag over Ni2+-chelate matrices. Two examples per protein are shown.
BL21(DE3)pLys (Novagen) served as a host strain for recombinant proteins. Expression was induced with 1 mM isopropyl-β-D-thiogalactopyranoside. The proteins were purified on Ni2+-chelate columns under denaturing conditions with 6 M urea. After renaturation in dialysis buffer (150 mM NaCl, 1 mM EDTA pH 5,0, 20 mM Tric-HCL pH 7,5) the amounts of purified proteins in the eluted fractions were determined by a BCA assay (Pierce). Proteins were subjected to sodium-dodecyl-sulfate-containing 7.5% polyacrylamide gel electrophoresis (SDS-PAGE) and Coomassie-blue stain. The purity was analyzed by digital imaging (Phoretix 1D Advanced Version 4.01).
Pol protein is abundant in cells lytically infected with FV
We first estimated the amount of Pol proteins present in FV infected cells. In addition, we determined the sensitivity of the MABs in detecting Gag and Pol protein species. A cellular lysate was prepared from BHK-21 cells lytically infected with PFV, which was obtained by transfection of 293T cells with the pcHSRV2 infectious molecular clone by calcium phosphate coprecipitation [12]. Proteins in the lysates were analysed with the Gag and Pol hybridomas SGG1 (recognizing Gag), 15E10 (PR/RT/RnaseH), and 3E11 (IN) [11,13] in an immunoblot along with defined amounts of recombinant Gag and Pol proteins purified from bacteria. As shown in Fig. 2, the MAB 3E11 has a detection limit of approx. 10 ng of IN protein expressed in bacteria, while the RT (15E10) and Gag (SGG1) MABs were able to detect 20 ng and 40 ng of the respective proteins from bacteria. This experiment further revealed that the method to detect FV Gag and Pol by the ECLplus reagent (Amersham-Pharmacia) was in a linear range from 10 to more than 100 ng of recombinant protein (Fig. 2 and data not shown). The IgG concentrations of the hybridomas used in this particular experiment were determined, following a published protocol (Mouse-IgG-ELISA, Roche), to be 3.2μg/ml (3E11), 10.5 μg/ml (15E10), and 10.1 μg/ml (SGG1). In conclusion, the IN MAB was at least 12 times more sensitive than the Gag MAB and approx. 6.5 times more than the RT antibody.
Figure 2 Immunoblot of a dilution series of recombinant Gag and Pol proteins, a cellular lysate (C), and extra-cellular virus (V) detected with the MABs SGG1 (Gag), 15E10 (RT), and 3E11 (IN). (C) was obtained by harvesting lytically infected BHK-21 cells, and (V) prepared by concentrating the supernatant of lytically infected cells through a sucrose cushion. On the right side the indicated amounts of recombinant proteins, specifying FV Gag and Pol proteins as shown in Fig. 1, were mixed and loaded onto an SDS-PAGE.
Due to the presence of five Gag and Pol molecule species of different molecular weights (pr71Gag, p68Gag, pr127Po1, p85RT, and p40IN) it was not possible to calculate exactly the respective molecule numbers present in infected cells. However, the comparison of the intensity of the lanes corresponding to Gag (pr71/p68) and Pol (pr127/p85/p40) proteins, which were detected by the MABs in the lysates, indicated that high amounts of Pol are expressed upon lytic infection in BHK-21 cells. This correlates well with the published amount of pol-specific mRNA, reported to equal the full-length or gag-specific mRNA in the bovine FV system [14]. The ease, with which Pol proteins can be detected in FV infected cells is indicative of their relatively high expression level compared to Gag. This finding questions the theoretical assumption of only trace amounts of Pol in FV particles. Obviously, FV utilizes distinct ways to avoid overloading infected cells with Pol protein. High cellular loads of retroviral Pol proteins can be associated with cell toxicity [15]. Although not necessary to incorporate high amounts of RT in FV particles, this abundance of FV Pol proteins in infected cells may have other yet undiscovered reasons in FV biology.
Determination of the Pol protein amounts relative to Gag in FV particles
We generated highly purified virus by consecutive centrifugation through a sucrose cushion and a linear gradient made of iodixanol. BHK-21 cells were infected with the supernatant from transfected 293T cells and cell-free virus was harvested when productive infection was ongoing, usually after 3–5 days. The supernatant was clarified from cellular debris by low-speed centrifugation and filtered through a 0.45μm pore-size filter (Sartorius). Virus was concentrated by centrifugation through a 20% sucrose cushion in TNE buffer (20 mM TRIS-HC1, pH 7.5, 150 mM NaC1, 1 mM EDTA) in a SW28 rotor (Beckman) at 25,000 rpm, 4°C for 1 hr. The sediment was resolved in Dulbecco's minimal essential medium (DMEM) and placed on a 2 ml 10–40% continuous iodixanol (OptiPrep from Axis-Shield) gradient for further virus purification. The gradient was cast in a gradient mixer (SG30 from Hoefer) the day before use. Following centrifugation in a TLS-55 rotor (Beckman) at 48,000 rpm and 4°C for 4 hrs, 200 μl fractions were taken from the top. From each fraction 30 μl were used for the determination of the refraction index, 20 μ1 for infectivity assay on BHK/LTR(PFV)lacZ cells [16], and l00 μl for immunoblotting.
As exemplified in Fig. 3A, fractions 5 and 6 were the main gradient fractions in which viral Gag and Pol proteins were detected by immunoblotting. Fraction 6 was also the main fraction of viral infectivity as shown in Fig. 3B. A mean density of 1.119 g/ml (± 0.011) was found for infectious PFV particles. This value is slightly lower than previous results with sucrose gradients [3,17,18]. Defined amounts of bacterially-expressed Gag and Pol proteins were also applied to the gel. The intensities of the bands were determined with a LAS-3000 (Fujifilm) and the relative amounts of Gag and Pol proteins were calculated using the software Image Gauge 3.01 (Fujifilm). A regression curve was formed, in which the total amounts of recombinant protein loaded in each lane were related to the optical densities of the individual protein bands which were produced after blotting, reaction with MABs, and ECLplus staining. In Fig. 4 an example is depicted, which was derived from the same samples shown in Fig. 3. The ability to build a regression curve from the sample detection also illustrates that the assay was linear over the protein range analyzed.
Figure 3 Representative example of the determination of the relative amounts of Gag and Pol proteins in purified PFV. (A) Extracellular virus was centrifuged through a sucrose cushion and the sediment was loaded onto a linear iodixanol gradient. Fractions were taken from the top and analyzed by immunoblotting with the Gag- and Pol-specific MABs. Defined amounts of recombinant PFV Gag and Pol proteins were also loaded onto the gel and simultaneously incubated with the MAB solutions. The blot was developed with the ECLplus reagent from Amersham-Pharmacia. (P), Pellet of the gradient. (B) Density and infectivity of the gradient fractions shown in (A). The infectivity was determined by a blue cell assay [16].
Figure 4 Relation of the intensities of the bands in the lanes with recombinant PFV proteins shown in Fig. 3 and amounts of protein loaded onto the gel. The latter was expressed as the number of molecules. Band intensities were determined with a LAS-3000 and calculated using the Image Gauge 3.01 software (Fujifilm). Over the protein range analyzed the band intensities were found to be in a linear relation to the protein amounts.
A total of 36 gradient fractions were analyzed with three independent quantifications for the individual gradients. The results are summarized in Table 1. We found that purified FV virions had a mean pr71Gag to p68Gag ratio of 1 to 4.2, which indicated that the cleaved p68Gag protein is the dominant capsid protein species in infectious PFV particles. The SGG1 MAB binding site is located N-terminal of the Gag cleavage site that generates p68 Gag and the 3 kD C-terminal peptide from the pr71 Gag precursor (our unpublished results). Therefore, the antibody detects both, the uncleaved and the cleaved protein equally well. The 127 kD Pol precursor protein was barely detected in the virus preparations, which indicated almost complete cleavage into the 85 kD RT and 40 kD IN subunits. Importantly, the relation of Gag proteins (pr71 plus p68) to p85RT was determined to be 15.8 to 1. This illustrates that PFV has found an independent way to incorporate as much Pol protein relative to Gag into progeny virus as typically found in orthoretroviruses [19]. With respect to the amount of IN protein, a ratio of 9.8 Gag molecules (pr71Gag plus p68Gag) to 1 IN molecule was revealed. Considering only the cleaved moiety, the p68Gag/p40IN ratio was determined to be 7.8 to 1 (Table 1). Thus, we constantly detected approximately 1.6 to two times more IN than RT protein in infectious virions. FV initially encapsidate the 127 kD Pol precursor protein which is cleaved into its subunits after packaging [4]. It may, therefore, be surprising not to find equal amounts of the two subunits in virions. The reason for this is presently unclear. It may be that different blotting efficiencies of the two proteins account for differences in detectability. Alternatively, different amounts of RT and IN enzymes in viral particles may be a consequence of the particular FV replication pathway. FVs reverse transcription takes place to a significant extent in the cytoplasm before progeny virus release [12,20,21]. The conditions of this reverse transcription late in the replication cycle are not understood. Gag gene expression appears to be required [22,23], but complete assembly of viral capsids may be not. While IN enzyme will be needed by the virus for the next round of replication, the RT subunit may be dispensable to the extent reverse transcription has already been completed and there is no need for RT to be actively encapsidated.
Table 1 Relative amounts of Gag and Pol proteins in foamy viruses
pr71/p68Gag:p85RT p68Gag:p85RT Pr71/p68Gag:p40IN p68Gag:p40IN p68Gag:pr71Gag
Mean 15.8 : 1 12.3 : 1 9.8 : 1 7.8 : 1 4.2 : 1
SD1 5.6 4.8 7.8 6.9 2.0
Maximum 26.3 : 1 22.7 : 1 41.3 : 1 35.8 : 1 8.0 : 1
Minimum 6.8 : 1 5.2 : 1 3.0 : 1 2.3 : 1 1.3 : 1
1SD, standard deviation
As detailed above, the reasons to assume that only trace amounts of Pol protein are encased in spumaretrovirus virions were hitherto largely theoretical. We provide here experimental evidence that many more Pol molecules per capsid can be found in purified FVs than was previously thought, even when taking into account that we did not determine the absolute numbers of molecules per virion, but only the relative Gag to Pol ratios. How can this finding be explained in the light of recent results in which two distinct RNA structures were identified to be essential for Pol protein incorporation into FV particles [4]? Firstly, with respect to this study only the minimal RNA sequence requirements for Pol protein encapsidation using subgenomic constructs have been determined, and not the relative ratios between Gag and Pol using a full-length viral genome. Secondly, it may be that the presence of the RNA domains, found to be responsible for Pol packaging, leads to the encapsidation of not only two Pol molecules per viral RNA, but of a larger complex which consists of many more protein molecules. This complex may be stabilized by protein-protein interactions between Pol and Gag, the individual Pol molecules, or a combination of both.
Authors' contributions
MC performed all experiments described in this manuscript. WR assisted in bacterial expression and purification of recombinant proteins. The experiments were designed and supervised by OH, DL and AR. AR wrote the manuscript together with MC.
Acknowledgements
We are indebted to Jürgen Helbig for the determination of the IgG concentration in MAB preparations.
This study was supported by grants from the DFG to A.R. (SFB479 and RE627/6-4) and to D.L. (LI621/3-1).
==== Refs
Enssle J Jordan I Mauer B Rethwilm A Foamy virus reverse transcriptase is expressed independently from the Gag protein Proc Natl Acad Sci USA 1996 93 4137 4141 8633029 10.1073/pnas.93.9.4137
Bodem J Löchelt M Winkler I Flower RP Delius H Flügel RM Characterization of the spliced pol transcript of feline foamy virus: the splice acceptor site of the pol transcript is located in gag of foamy viruses J Virol 1996 70 9024 9027 8971036
Yu SF Baldwin DN Gwynn SR Yendapalli S Linial ML Human foamy virus replication: a pathway distinct from that of retroviruses and hepadnaviruses Science 1996 271 1579 1582 8599113
Peters K Wiktorowicz T Heinkelein M Rethwilm A RNA and Protein Requirements for the Incorporation of Pol Protein into Foamy Virus Particles J Virol 2005 79 7005 7013 15890940 10.1128/JVI.79.11.7005-7013.2005
Heinkelein M Leurs C Rammling M Peters K Hanenberg H Rethwilm A Pregenomic RNA is required for efficient incorporation of pol polyprotein into foamy virus capsids J Virol 2002 76 10069 10073 12208988 10.1128/JVI.76.19.10069-10073.2002
Flügel RM Pfrepper KI Proteolytic processing of foamy virus Gag and Pol proteins Curr Top Microbiol Immunol 2003 277 63 88 12908768
Rinke CS Boyer PL Sullivan MD Hughes SH Linial ML Mutation of the catalytic domain of the foamy virus reverse transcriptase leads to loss of processivity and infectivity J Virol 2002 76 7560 7570 12097569 10.1128/JVI.76.15.7560-7570.2002
Boyer PL Stenbak CR Clark PK Linial ML Hughes SH Characterization of the polymerase and RNase H activities of human foamy virus reverse transcriptase J Virol 2004 78 6112 6121 15163704 10.1128/JVI.78.12.6112-6121.2004
Ausubel FM Brent R Kingston RE Moore D Seidman JG Smith JA Struhl K Current protocols in molecular biology 1987 New York, NY: John Wiley
Sambrook J Russell DW Molecular cloning: a laboratory manual 2001 3 Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press
Heinkelein M Dressler M Jarmy G Rammling M Imrich H Thurow J Lindemann D Rethwilm A Improved primate foamy virus vectors and packaging constructs J Virol 2002 76 3774 3783 11907217 10.1128/JVI.76.8.3774-3783.2002
Moebes A Enssle J Bieniasz PD Heinkelein M Lindemann D Bock M McClure MO Rethwilm A Human foamy virus reverse transcription that occurs late in the viral replication cycle J Virol 1997 71 7305 7311 9311807
Imrich H Heinkelein M Herchenröder O Rethwilm A Primate foamy virus Pol proteins are imported into the nucleus J Gen Virol 2000 81 2941 2947 11086125
Holzschu DL Delaney MA Renshaw RW Casey JW The nucleotide sequence and spliced pol mRNA levels of the nonprimate spumavirus bovine foamy virus J Virol 1998 72 2177 2182 9499074
Orlova M Yueh A Leung J Goff SP Reverse transcriptase of Moloney murine leukemia virus binds to eukaryotic release factor 1 to modulate suppression of translational termination Cell 2003 115 319 331 14636559 10.1016/S0092-8674(03)00805-5
Schmidt M Rethwilm A Replicating foamy virus-based vectors directing high level expression of foreign genes Virology 1995 210 167 178 7793069 10.1006/viro.1995.1328
Hooks JJ Gibbs CJ Jr The foamy viruses Bacteriol Rev 1975 39 169 185 51620
Gelderblom H Frank H Nermut MV, Steven AC Spumavirinae Animal Virus Structure 1987 3 Amsterdam, New York, Oxford: Elsevier 305 311
Vogt VM Coffin JM, Hughes SH, Varmus HE Retroviral virions and genomes Retroviruses 1997 Cold Spring Harbor: Cold Spring harbor Laboratory Press 27 69
Roy J Rudolph W Juretzek T Gärtner K Bock M Herchenröder O Lindemann D Heinkelein M Rethwilm A Feline foamy virus genome and replication strategy J Virol 2003 77 11324 11331 14557618 10.1128/JVI.77.21.11324-11331.2003
Yu SF Sullivan MD Linial ML Evidence that the human foamy virus genome is DNA J Virol 1999 73 1565 1572 9882362
Enssle J Fischer N Moebes A Mauer B Smola U Rethwilm A Carboxy-terminal cleavage of the human foamy virus Gag precursor molecule is an essential step in the viral life cycle J Virol 1997 71 7312 7317 9311808
Heinkelein M Pietschmann T Jarmy G Dressler M Imrich H Thurow J Lindemann D Bock M Moebes A Roy J Efficient intracellular retrotransposition of an exogenous primate retrovirus genome 2000 19 3436 3445 10880456
|
16004609
|
PMC1185564
|
CC BY
|
2021-01-04 16:36:39
|
no
|
Retrovirology. 2005 Jul 8; 2:44
|
utf-8
|
Retrovirology
| 2,005 |
10.1186/1742-4690-2-44
|
oa_comm
|
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-711601880710.1186/1465-9921-6-71ResearchExhaled volatile organic compounds in patients with non-small cell lung cancer: cross sectional and nested short-term follow-up study Poli Diana [email protected] Paolo [email protected] Massimo [email protected] Matteo [email protected] Olga [email protected] Bruno [email protected] Luca [email protected] Michele [email protected] Antonio [email protected] National Institute of Occupational Safety and Prevention Research Center at the University of Parma, Via Gramsci 14, 43100 Parma, Italy2 Laboratory of Industrial Toxicology, Dept. of Clinical Medicine, Nephrology and Health Sciences, University of Parma, Via Gramsci 14, 43100 Parma, Italy3 Unit of Thoracic Surgery, University of Parma, Via Gramsci 14, 43100 Parma, Italy4 Respiratory Dept. and Lung Function Unit of Maugeri Foundation, Via Pinidolo 23, 25064 Gussago (Bs), Italy2005 14 7 2005 6 1 71 71 22 3 2005 14 7 2005 Copyright © 2005 Poli et al; licensee BioMed Central Ltd.2005Poli 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
Non-invasive diagnostic strategies aimed at identifying biomarkers of lung cancer are of great interest for early cancer detection. The aim of this study was to set up a new method for identifying and quantifying volatile organic compounds (VOCs) in exhaled air of patients with non-small cells lung cancer (NSCLC), by comparing the levels with those obtained from healthy smokers and non-smokers, and patients with chronic obstructive pulmonary disease. The VOC collection and analyses were repeated three weeks after the NSCLC patients underwent lung surgery.
Methods
The subjects' breath was collected in a Teflon® bulb that traps the last portion of single slow vital capacity. The 13 VOCs selected for this study were concentrated using a solid phase microextraction technique and subsequently analysed by means of gas cromatography/mass spectrometry.
Results
The levels of the selected VOCs ranged from 10-12 M for styrene to 10-9 M for isoprene. None of VOCs alone discriminated the study groups, and so it was not possible to identify one single chemical compound as a specific lung cancer biomarker. However, multinomial logistic regression analysis showed that VOC profile can correctly classify about 80 % of cases. Only isoprene and decane levels significantly decreased after surgery.
Conclusion
As the combination of the 13 VOCs allowed the correct classification of the cases into groups, together with conventional diagnostic approaches, VOC analysis could be used as a complementary test for the early diagnosis of lung cancer. Its possible use in the follow-up of operated patients cannot be recommended on the basis of the results of our short-term nested study.
==== Body
Background
Breath analysis seems to be a promising approach to identify new biomarkers of inflammatory and oxidative lung processes, and different volatile organic compounds (VOCs) of endogenous or exogenous origin have been analyzed to study lung diseases [1] and characterize environmental and occupational exposure to chemical pollutants [2].
During the 1970s, Pauling et al.[3] determined more than 200 components in human breath, some of which have subsequently been associated with different pathological conditions on the basis of their effect and/or their metabolic origin.
In 1985, Gordon et al. identified several alkanes and monomethylated alkanes in the exhaled air of lung cancer patients [4], an observation that aroused interest because of the possible use of exhaled biomarkers for early detection of the disease. Classical screening procedures, such as chest radiography and sputum cytology, have not decreased the number of deaths due to lung cancer [5], but promising results have recently been obtained using novel imaging techniques such as low-dose helicoidal computed tomography [6], although cost effectiveness and possible over-diagnosis seem to be serious issues. There is therefore a considerable need for non-invasive diagnostic procedures aimed at identifying lung cancer at an early stage and adding specificity to imaging techniques.
In 1999, Phillips et al. [7] selected 22 VOCs – mainly alkanes and benzene derivatives – to distinguish subjects with and without lung cancer, and have recently modified the VOC pattern subject to statistical analysis by reducing them to nine [8]. Selected alkanes and methylated alkanes have proved to be highly discriminating in distinguishing lung cancer patients from healthy controls, but breath analyses can be affected by both clinical and analytical confounding variables [9]. The published studies have included mixed groups of patients with primary small or non-small cell lung cancer (NSCLC) and lung metastases, and did not compare VOC levels in lung cancer patients with those in asymptomatic smokers or subjects suffering from chronic obstructive pulmonary disease (COPD), both of which may precede or be associated with the development of lung cancer and which may characterise the people undergoing screening procedures [10,11]. Furthermore, there are no data supporting the usefulness of VOC analysis in the follow-up of patients after tumour resection. Finally, only a qualitative approach has been used to identify selected VOCs, without any attempt to quantify the individual components. Actual breath concentrations could increase the statistical power of comparisons aimed at identifying differences between groups and between repeated measurements in the same individuals.
The aim of this study was to set up a new method for identifying and quantifying selected VOCs in exhaled air, and apply it to a cross-sectional study of NSCLC and COPD patients, and healthy control smokers and non-smokers, and a short-term follow-up study of patients undergoing surgery for NSCLC.
Methods
Study design
The design of the present study included a cross-sectional investigation during which 13 selected VOCs were measured in air exhaled by NSCLC and COPD patients, and asymptomatic control smokers and non-smokers. A subsequent nested short-term follow-up study of the NSCLC patients was carried out with repeat VOC sampling and analysis about three weeks (range 2 – 4) after they had undergone tumor resection (T1).
Subjects
We enrolled 36 patients who underwent tumor resection because of histological evidence of NSCLC at the University of Parma's Department of Thoracic Surgery. The assessments of tumour size and nodes were based on the International Union Against Cancer TNM staging system [12], and all of the patients were classified as having stage Ia, Ib and IIa lung cancer. None of the patients received radiation or chemotherapy before surgery.
The study also included 25 subjects with clinically stable, mild to moderate COPD, all of whom were diagnosed on the basis of the GOLD guidelines [13]. In brief, the entry criteria, consisted of a post-bronchodilator FEV1 of <80% the predicted value, an FEV1/FVC ratio of <70%, β2-agonist-reversibility at baseline FEV1 of <200 ml and/or 15%, and the absence of clinical asthma or other significant respiratory diseases. None of them had experienced any worsening in symptoms over the previous eight weeks.
The asymptomatic controls were 35 smokers and 50 non-smokers. The smokers had to have normal spirometry values (FEV1 and FEV1/FVC) and not be suffering from chronic bronchitis; the non-smokers had to have no pulmonary symptoms or a history of pulmonary disease, and normal lung spirometry results. The smokers did not smoke for at least one hour before breath collection.
Twenty-six of the NSCLC patients agreed to repeat the breath collection during a follow-up visit 15–30 days after surgery; the other 10 were excluded from the nested follow-up study because their clinical condition had significantly worsened.
Table 1 shows the characteristics of the study subjects, all of whom gave their informed consent.
Table 1 Demographic characteristics of studied groups.
NSCLC COPD Controls Smokers
Subjects (n°) 36 25 50 35
Age (median, years) 67.2 70.2 55.7 54.1
Sex (male/female) 28/8 18/7 27/23 30/5
Current smokers 2 1 0 35
Ex smokers 28 21 0 0
Ever smokers 6 3 50 0
*Pack-years 20 20 n.a. 25 ± 2.6
FEV1 (% predicted) 69.8 ± 15.2 61.7 ± 13.4 105.6 ± 9.1 101.8 ± 10.2
The ex-smokers subjects had stopped smoking for at least one year. * Pack-years (mean ± SD) among current smokers. NSCLC = non-small cell lung cancer; COPD = chronic obstructive pulmonary disease; n.a. = not applicable.
Breath collection
After carrying out a series of experiments in order to establish a reliable sampling procedure, we modified the breath sampling procedure recommended by the manufacturer of a commercially available device (Bio-VOC® sampler, Markes International Ltd, Rhondda Cynon Taff, UK) (Figure 1). Briefly, after 60 minutes' rest, the subjects were asked to perform a single slow vital capacity breath into a one-way valve connected to a Teflon®-bulb, which traps the last portion of exhaled air (150 ml).
Figure 1 Breath collection and VOC extraction. The subjects performed a single slow vital capacity into a Teflon® bulb (Bio-VOC® breath sampler) (a) which traps the last portion of exhaled air (150 mL); the VOCs were extracted by directly inserting a 75 mm Carboxen/PDMS SPME fiber (30 min) into the bulb (b).
Twenty environmental samples were taken from the rooms in which the subjects performed the test in order to compare breath and ambient air VOC levels.
VOC extraction and analysis
After breath collection, 1 μL of n-heptane-d16 and styrene-d8 methanolic solution (1.5 × 10-5 M) was added to each sample as internal standard (IS) for respectively aliphatic and aromatic compounds. The exhaled VOCs and IS were extracted by means of SPME using a 75 μm Carboxen/PDMS fibre (Supelco, Bellefonte, PA, USA), which was put into the Bio-VOC® breath sampler for 30 min at room temperature and then thermally desorbed in GC injection port at 280°C. The GC/MS analysis was carried out using a Hewlett-Packard HP 6890 gas chromatograph coupled with an HP 5973 mass selective detector (Palo Alto, CA, USA). The VOCs were separated on an Equity™-1 column (30 m, 0.25 mm i.d., 1.0 μm film, Supelco) and acquired in full-scan mode in 40–350 m/z range.
Thirteen VOCs (seven aliphatic and six aromatic compounds) were selected, each of which was identified by means of its mass spectrum and confirmed by comparing its retention time with that of pure standard and characteristic fragment ions; only the substances that did not interfere with co-eluting compounds were chosen.
The preliminary experiments addressed methodological issues, defined standard operating procedures, and validated analytical methods of VOC collection and analysis. The factors affect the SPME process, such as adsorption and desorption times and sampling temperature, were optimized. The extraction time profile at room temperature (22°C) was 30 min and not markedly different among the compounds. The SPME fibre was immediately transferred to the GC-injector port in order to avoid the loss of the extracted substances and avoid analyte evaporation [14]. No carry-over effects were observed when desorption was performed at 280°C for 5 min.
The method was validated by studying the linear range, and the limits of detection and precision. Linearity was established over four orders of magnitude (1012-10-8 M, r2>0.98) and the limits of detection, calculated as a signal/noise ratio of about 3, was about 10-12 M for all the compounds. Analytical precision, calculated as % RSD, was within 3.1–13.7% for all of the intra- and inter-day determinations on standards. The gaseous standards were directly prepared in the Bio-VOC® bulb filled with helium, 1 μL of VOC methanolic standard solution, 1 μL of IS (1.5 × 10-5 M), and 6 μL of deionised water. The standards were stabilised at room temperature for almost one hour and remained stable up to 60 hours.
Statistical analysis
As the benzene and toluene levels had a log-normal distribution (the Kolmogorov-Smirnov normality test) parametric tests were used for the cross-sectional study (one-way ANOVA followed by the Games Howell post-hoc test). Non-parametric statistics (Kruskal-Wallis test followed by Dunn's Post Hoc test) were used for the other VOCs, whose distribution was not normal even after log-transformation. The cases were classified by means of multinomial logistic regression using group codes as the dependent variable and all of the VOC concentrations (except total xylenes because of their high correlation with ethylbenzene: r>0.95) as predictors. Interpretable factors based on VOC levels were obtained by means of principal component analysis (Varimax rotation with Kaiser's normalization) [15]. The Keiser Meyer Olkin (KMO) test was used to test sample adequacy (considered acceptable if the KMO constant was >0.60), and the number of factors was chosen on the basis of the flex point of the graph of decreasing eigenvalues; the percentage of variance explained was also recorded.
In the case of the follow-up study, Student's t test for repeated measures was applied to the benzene and toluene levels; Wilcoxon's test was used for all of the other VOCs.
A p value of <0.05 was considered significant for all of the statistical analyses. SPSS 13.0 (SPSS inc. Chicago, IL, USA) and PRISM 3.0 (Graphpad, San Diego, CA) were used for the statistical analyses.
Results
Tables 2 and 3 respectively summarise the VOC levels and the statistical significances of the between-group differences. As all of the VOCs showed significant differences between at least two group pairs, the overall p values of the Kruskal-Wallis and ANOVA tests for individual VOCs fell between 7.5 × 10-13 (for Ethylbenzene) to 1.6 × 10-3 (isoprene). For these highly significant differences, adjustments for multiple testing calculated using Holm's test (less conservative than Bonferroni's test [16]) did not affect the results. The levels of 10 of the 13 substances were significantly higher in the NSCLC patients than in control non-smokers; the levels of 9 were higher in the COPD patients and control smokers than in control non-smokers.
Table 2 Exhaled VOC levels in studied groups
Controls (10-12 M) NSCLC (10-12 M) COPD (10-12 M) Smokers (10-12 M)
Isoprene 3789 (1399 – 6589) 6041 (3130 – 8863) 1758 (453 – 4981) 7243 (1361 – 16968)
2-Methylpentane 27.7 (3.4 – 50.3) 139.5 (65.7 – 298.8) 44.7 (21.7 – 63.8) 109.8 (62.8 – 173.5)
Pentane 268.0 (107.7 – 462.7) 647.5 (361.3 – 1112.5) 477.7 (261.5 – 1547.4) 511.4 (241.3 – 1128.3)
Ethylbenzene 13.6 (10.8 – 15.1) 24.0 (13.6 – 32.6) 51.1 (26.9 – 132.7) 39.7 (21.7 – 74.1)
Xylenes total 31.1 (21.1 – 56.4) 68.9 (43.6 – 108.4) 94.8 (49.7 – 131.9) 85.8 (60.1 – 185.2)
Trimethylbenzene 6.2 (4.7 – 11.0) 14.9 (9.3 – 22.1) 18.5 (10.4 – 25.4) 18.9 (11.9 – 44.9)
Toluene 80.8 (58.9 – 140.0) 158.8 (118.7 – 237.5) 158.5 (103.5 – 269.7) 453.5 (169.6 – 745.7)
Benzene 44.7 (27.7 – 68.6) 94.5 (62.2 – 132.2) 73.3 (51.8 – 95.4) 269.2 (84.6 – 745.1)
Heptane 8.4 (5.0 – 15.3) 13.5 (1.5 – 34.0) 47.3 (13.9 – 98.0) 98.0 (40.3 – 161.7)
Decane 208.7 (14.3 – 405.5) 568.0 (277.9 – 1321.6) 737.3 (524.6 – 1177.6) 239.2 (60.0 – 884.0)
Styrene 12.3 (5.3 – 21.8) 17.9 (8.5 – 37.2) 87.6 (56.0 – 148.8) 7.2 (2.8 – 41.6)
Octane 20.2 (4.0 – 50.8) 61.0 (22.4 – 112.9) 52.5 (31.9 – 147.2) 33.5 (19.7 – 57.8)
Pentamethylheptane 0.9 (0.1 – 2.6) 2.5 (1.2 – 9.7) 2.0 (1.2 – 7.6) 5.8 (1.2 – 16.5)
Concentrations expressed as median values(25th -75th percentile).
Table 3 Statistical differences between groups.
NSCLC vs. Controls COPD vs. Controls Smokers vs. Controls NSCLC vs. COPD NSCLC vs. Smokers COPD vs. Smokers
Isoprene n.s. n.s. n.s. p < 0.05 n.s. P < 0.01
2-Methylpentane p < 0.001 p < 0.05 p < 0.001 p < 0.001 n.s. P < 0.05
Pentane p < 0.001 p < 0.05 p < 0.05 n.s. n.s. n.s.
Ethylbenzene p < 0.01 p < 0.001 p < 0.001 p < 0.05 n.s. n.s.
Xylenes total p < 0.001 p < 0.001 p < 0.001 n.s. n.s. n.s.
Trimethylbenzene p < 0.01 p < 0.001 p < 0.001 n.s. n.s. n.s.
Toluene p < 0.001 n.s. p < 0.001 n.s. p < 0.001 P < 0.01
Benzene p < 0.001 n.s. p < 0.001 n.s. p < 0.001 P < 0.05
Heptane n.s. p < 0.01 p < 0.001 n.s. p < 0.001 n.s.
Decane p < 0.001 p < 0.01 n.s. n.s. n.s. n.s.
Styrene n.s. p < 0.001 n.s. p < 0.001 n.s. P < 0.001
Octane p < 0.001 p < 0.01 n.s. n.s. n.s. n.s.
Pentamethylheptane p < 0.001 n.s. p < 0.001 n.s. n.s. n.s.
The significance of the multiple comparisons inside the individual univariate tests. ANOVA followed by Games Howell Post Hoc test for benzene and toluene, Kruskal-Wallis test followed by Dunn's Post Hoc test for all the other VOCs were performed.
The NSCLC patients had significantly higher 2-methylpentane and isoprene levels and significantly lower ethylbenzene and styrene levels than the COPD patients, and significantly lower benzene, heptane and toluene levels than the control smokers. In comparison with the control smokers, the COPD patients had lower 2-methylpentane, benzene and toluene levels, and higher styrene levels.
Exhaled breath of non-smoking controls had higher levels of isoprene and heptane than the environmental air, whereas NSCLC and COPD patients and control smokers showed higher levels of almost all substances (data not shown).
Principal component analysis (table 4), with a KMO constant of 0.83, distinguished three factors with eigenvalues >1, of which the third was the flex point of the graph of decreasing eigenvalues. The first grouped benzene, heptane, toluene, ethylbenzene, trimethylbenzene with an explained variance of 27.5% (total xylenes were excluded because of their high correlation with ethylbenzene: r>0.95); the second grouped octane, styrene, pentamethylheptane and decane with an explained variance of 20%, and the third grouped pentane, isoprene and methylpentane with an explained variance of 19%. The total explained variance of the model was therefore 66.5%.
Table 4 Principal Components analysis of variables.
Factors
Group 1 2 3
Isoprene 1 0.797
2-Methylpentane 1 0.562
Pentane 1 0.531
Ethylbenzene 2 0.851
Trimethylbenzene 2 0.794
Toluene 2 0.773
Benzene 2 0.728
Heptane 2 0.629
Decane 3 0.878
Styrene 3 0.704
Octane 3 0.643
Pentamethylheptane 3 0.592
In order to test the discriminant power of the exhaled VOC pattern, a multinomial logistic regression was made using the coding group as the output variable and the concentration of all of the VOCs except total xylenes as predictors: concentrations were used because they are direct measures with an intrinsic experimental error and therefore more appropriate than the ratio between exhaled breath and air VOC concentration, a function derived from two different experimental measures by means of mathematical manipulations. Figure 2 shows the correct classification of cases into four groups as the Cox and Snell pseudo R-square of the model was 0.83 (goodness-of-fit test). In general, 82.5% of subjects were correctly classified: a maximum of 87.8% for control non-smokers and a minimum of 72.2% for the NSCLC patients. Analysis of residuals did not reveal any particular cases with an undue influence on the model or the overall classification. On the basis of these results, the overall sensitivity (calculated as NSCLC true positive/ true positive + false negative) was 72.2% and overall specificity (calculated as NSCLC true negative/ true negative + false positive) was 93.6%.
Figure 2 Classification of cases with multinomial logistic regression analysis. ** Correctly classified cases. 82.5% of the subjects were correctly classified.
In the follow-up study of the NSCLC patients, only isoprene and decane significantly decreased after surgery (p < 0.05, table 5).
Table 5 VOCs levels at T0 (before surgery) and T1 (after surgery).
T0 T1
Isoprene 6121 (4069–9031) *4125 (2415–7407)
2-Methylpentane 139.5 (68.8–291.6) 123.5 (81.1–227.6)
Pentane 647.5 (388.5–1013) 529.5 (329.6–960.0)
Ethylbenzene 24.0 (14.8–28.0) 19.7 (15.7–34.5)
Xylenes total 69.0 (45.8–105.6) 67.8 (51.2–129.4)
Trimethylbenzene 15.2 (10.1–22.3) 13.2 (10.2–22.5)
Toluene 161.9 (118.7–232.5) 160.3 (119.0–232.7)
Benzene 95.7 (62.9–132.2) 99.6 (60.0–119.2)
Heptane 15.1 (0.9–34.6) 18.7 (9.5–39.5)
Decane 625.0 (322.6–1392) *443.0 (197.0–920.7)
Styrene 22.1 (11.5–38.1) 18.0 (12.1–43.1)
Octane 65.7 (45.8–131.4) 49.7 (28.5–102.5)
Pentamethylheptane 2.6 (1.7–10.0) 2.5 (1.1–8.8)
* means a statistically significant difference (p < 0.05). The data are expressed as median (25th -75th percentile).
Discussion
Non-invasive diagnostic strategies aimed at identifying biomarkers of early lung cancer probably require the use of a panel rather than single substances [17]. The main finding of our study was that none of selected VOCs alone distinguished the NSCLC patients from the other study groups (i.e. non of them was a specific biomarker of NSCLC), but overall VOC concentrations were highly discriminant (>70%). Owing to the limited sensitivity and specificity of VOC analysis, a NSCLC diagnosis only based only VOC concentrations in exhaled breath cannot be recommended at this stage. We did not calculate positive and negative predictive values, as they are highly dependent on the prevalence of the condition being examined in the population at hands. Owing to the low prevalence of NSCLC even in selected groups at high risk, the positive predicted value of exhaled VOCs is expected to be low, and should probably be used to rule out, rather than to confirm NSCLC in subjects with suspect nodules.
Moreover, exhaled breath analysis is a particularly interesting strategy but is still hampered by the lack of a standardised breath collection system and putative exhaled biomarkers.
Our simple method of breath collection has a number of advantages: i) it samples a fixed volume of air and discards anatomic dead space air; ii) its fixed resistance allows a reasonably constant expiratory flow; iii) it has no carry-over effects and permits the addition of internal standards to the breath samples, which improves data reproducibility; and iv) it is a well-tolerated, suitable for screening purpose, and also applicable to difficult clinical and psychological conditions such as those observed in NSCLC patients.
Further studies are needed to evaluate the VOC levels obtained from repeated expirations or tidal breathing, but the collection procedures require respiratory devices equipped with instruments that control ventilatory pattern [18], and this may limit their widespread application.
We selected 13 VOCs from the chromatographic profile of exhaled breath on the basis of the detectability of the peak and their biological significance, ten of which have been previously used for discriminant lung cancer analysis by Phillips et al. [7]; the other three were markers of oxidative stress such as pentane with its methylated form (2-methylpentane), and toluene, which is closely related to cigarette smoke.
The fact that we identified fewer VOCs than Phillips et al. [7] may have been partially due to differences in our breath sampling procedures: rather than concentrating the breath sample in a sorbent trap [19], we collected breath VOCs from a single expiration and extracted them using SPME fibre. The SPME technique may be less sensitive, but has the advantages of not requiring sample preparation or any specific equipment for GC analysis [20]; furthermore, it allowed us to measure most of the substances of interest proposed in the literature. Another reason for the difference in VOC identification may be the different clinical characteristics of lung cancer patients: we enrolled early-stage NSCLC patients because they may benefit more from early detection strategies.
There were no significant differences between the level of most of the VOCs in the exhaled air of the control non-smokers and those in the ambient air, which suggests that ambient levels may influence the VOCs exhaled by healthy non-smokers (data not shown). However, the VOC levels in diseased patients were not explainable solely by ambient VOC concentrations during breath collection, because the samples of all of the study subjects were collected in the same place. The NSCLC and COPD patients and the control smokers had generally higher levels of all of the exhaled VOCs than the control non-smokers (except for isoprene in the COPD group), which reflects differences in exhaled air composition due to pathological conditions or smoking rather than environmental contamination.
Various approaches have been adopted in an attempt to distinguish endogenous substances from exogenous contaminants, such as correcting exhaled VOC concentrations by subtracting inspiratory VOC levels or by calculating alveolar gradients [7]. However, although these methods are easy to perform, they do not take into account the complexity of pulmonary adsorption and exhalation of volatile substances [2].
Although the exact origin of exhaled VOCs remains to be demonstrated, principal components analysis (PCA) factorised the compounds into three groups (table 4) and suggests some fascinating hypoteses. It may be particularly relevant in distinguishing substances of endogenous origin from those influenced by confounding factors mainly related to tobacco smoke.
Isoprene, pentane and 2-methylpentane are grouped together (group 1, factor 3). These substances can be considered mainly endogenous compounds even though pentane and its methylated forms are also present in vehicle engine exhausts [21] and isoprene is also a constituent of tobacco smoke [22]. In humans, isoprene is formed from acetilCoA and is the basic molecule in cholesterol biosynthesis [23], and pentane comes from human lipid peroxidation [24]. The grouping of these with 2-methylpentane is in line with the results of a previous study that considered methylated alkanes as a secondary product of human oxidative stress [25], although the exact source of methylated alkanes is still debated [26].
Of the group 1 substances, 2-methylpentane levels were higher in NSCLC patients than in the control non-smokers and COPD patients, which suggests its potential usefulness in screening procedures (probably in combination with other relevant biomarkers). In line with previous observations [27], pentane levels were higher in the exhaled air of the patients with NCSLC and COPD and asymptomatic smokers than in the control non-smokers, but did not differentiate the first three groups from each other.
Also in line with previously published studies [27,28], isoprene levels were significantly higher in the breath than in the environmental samples (data not shown), and higher in the NSCLC patients and control smokers than in the COPD patients. The between-group differences are difficult to interpret, but are probably related to the moderate effect of cigarette smoke on isoprene levels, and partially to the lung destruction (emphysema) often affecting COPD patients. In this regard, although no studies have compared breath isoprene levels in NSCLC and COPD patients, lower levels have been observed in the exhaled breath of patients with acute respiratory distress syndrome (ARDS) in comparison with those without ARDS [29].
The substances belonging to group 2 (factor 1) could be classified mainly as smoking-related exogenous compounds because their levels were higher in the control smokers than control non-smokers. Ethylbenzene may be of particular interest because of its ability to distinguish NSCLC and COPD patients, and control non-smokers.
The substances belonging to group 3 (factor 2) are heterogeneous and it is therefore more difficult to interpret the between-group differences in the levels of the individual substances.
The results of the VOC analysis of our nested short-term follow-up study of surgically treated NSCLC patients showed that only isoprene and decane levels significantly decreased after surgery (Table 5), thus indicating that breath VOC analysis cannot be recommended as a short-term follow-up procedure in such patients.
Conclusion
Although none of the individual exhaled VOC alone was specific for lung cancer, a combination of 13 VOCs does allow the classification of cases into groups. Exhaled VOC analysis may therefore be useful in improving the specificity and sensitivity of conventional diagnostic approaches to lung cancer. However, these findings will require validation in larger clinical studies.
List of abbreviation used
COPD = Chronic Obstructive Pulmonary Disease; GC/MS = Gas Chromatography/Mass Spectrometry; IS = internal standard; NSCLC = Non-Small Cells Lung Cancer; PCA = Principal Components Analysis; SPME = Solid Phase Microextraction; VOC = Volatile Organic Compound; trimethylbenzene = 1,2,4- trimethylbenzene; pentamethylheptane = 2,2,4,6,6-pentamethylheptane.
Competing interests
All authors excluded any competing interest.
Authors' contributions
DP: substantial contribution to conception and design, acquisition of data, analysis and interpretation of data, involved in drafting the articles.
PC: substantial contribution to conception and design, collection of samples, revision of the draft critically for important intellectual content.
MC: substantial contribution to conception and design, analysis and interpretation of data, involved in drafting the articles.
MG: substantial contribution to conception and design, statistical analysis and interpretation of data, involved in drafting the articles.
OA: collection of samples, revision of the draft critically for important intellectual content.
BB: substantial contribution to conception and design, collection of samples, revision of the draft critically for important intellectual content.
MR: substantial contribution to conception and design, collection of samples, revision of the draft critically for important intellectual content.
AM: substantial contribution to conception and design, statistical analysis and interpretation of data, involved in drafting the articles, final approval of the version to be published.
Acknowledgements
This study was supported in part by Ricerca Finalizzata 2003 from Italian Ministry of Health and in part by grant R01 HL72323 from the National Heart, Blood and Lung Institute (NHLBI; Bethesda, USA). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NHLBI or National Institute of Health.
We thank E. Zaffignani for her cooperation during the study.
==== Refs
Miekisch W Schubert JK Noeldge-Schomburg GF Diagnostic potential of breath analysis – focus on volatile organic compounds Clin Chim Acta 2004 347 25 39 15313139 10.1016/j.cccn.2004.04.023
Imbriani M Ghittori S Gases and organic solvents in urine as biomarkers of occupational exposure: a review Int Arch Occup Environ Health 2005 78 1 19 15592680 10.1007/s00420-004-0544-z
Pauling L Robinson AB Teranishi R Cary P Quantitative analysis of urine vapor and breath by gas-liquid partition chromatography Proc Natl Acad Sci U S A 1971 68 2374 2376 5289873
Gordon SM Szidon JP Krotoszynski BK Gibbons RD O'Neill HJ Volatile organic compounds in exhaled air from patients with lung cancer Clin Chem 1985 31 1278 1282 4017231
Humphrey LL Teutsch S Johnson M U.S. Preventive Services Task Force Lung cancer screening with sputum cytologic examination, chest radiography, and computed tomography: an update for the U.S. Preventive Services Task Force Ann Intern Med 2004 140 740 753 15126259
Gohagan JK Marcus PM Fagerstrom RM Pinsky PF Kramer BS Prorok PC Ascher S Bailey W Brewer B Church T Engelhard D Ford M Fouad M Freedman M Gelmann E Gierada D Hocking W Inampudi S Irons B Johnson CC Jones A Kucera G Kvale P Lappe K Manor W Moore A Nath H Neff S Oken M Plunkett M Price H Reding D Riley T Schwartz M Spizarny D Yoffie R Zylak C THE LUNG SCREENING STUDY RESEARCH GROUP. Final results of the Lung Screening Study, a randomized feasibility study of spiral CT versus chest X-ray screening for lung cancer Lung Cancer 2005 47 9 15 15603850 10.1016/j.lungcan.2004.06.007
Phillips M Gleeson K Hughes JM Greenberg J Cataneo RN Baker L McVay WP Volatile organic compounds in breath as markers of lung cancer: a cross-sectional study Lancet 1999 353 1930 1933 10371572 10.1016/S0140-6736(98)07552-7
Phillips M Cataneo RN Cummin AR Gagliardi AJ Gleeson K Greenberg J Maxfield RA Rom WN Detection of lung cancer with volatile markers in the breath Chest 2003 123 2115 2123 12796197 10.1378/chest.123.6.2115
Ost D Shah RD Fein D Fein AM To screen or not to screen: a volatile issue in lung cancer Chest 2003 123 1788 1792 12796150 10.1378/chest.123.6.1788
Bach PB Elkin EB Pastorino U Kattan MW Mushlin AI Begg CB Parkin DM Benchmarking lung cancer mortality rates in current and former smokers Chest 2004 126 1742 1749 15596668 10.1378/chest.126.6.1742
Papi A Casoni G Caramori G Guzzinati I Boschetto P Ravenna F Calia N Petruzzelli S Corbetta L Cavallesco G Forini E Saetta M Ciaccia A Fabbri LM COPD increases the risk of squamous histological subtype in smokers who develop non-small cell lung carcinoma Thorax 2004 59 679 681 15282388 10.1136/thx.2003.018291
Sobin LH Witterkind C TNM classification of malignant tumours International Union against Cancer 2002 6 New York: Wiley-Liss
Pauwels RA Buist AS Calverley PM Global strategy for the diagnosis, management and prevention of chronic obstructive pulmonary disease: NHLBI/WHO Global Initiative for Chronic Obstructive Lung Disease (GOLD) Workshop summary Am J Respir Crit Care Med 2001 163 1256 1276 11316667
Arthur CL Pawliszyn J Solid phase microextraction with thermal desorption using fused silica optical fibers Anal Chem 1990 62 2145 2148 10.1021/ac00218a019
Edwards RD Jurvelin J Koistinen K Saarela K Jantunen M VOC source identification from personal and residential indoor, outdoor and workplace microenvironment samples in EXPOLIS-Helsinki, Finland Atmos Environ 2001 35 4829 4841 10.1016/S1352-2310(01)00271-0
Roback PJ Askins RA Judicious Use of Multiple Hypothesis Tests Conserv Biol 2005 19 261 267 10.1111/j.1523-1739.2005.00269.x
Andriani F Conte D Mastrangelo T Leon M Ratcliffe C Roz L Pelosi G Goldstraw P Sozzi G Pastorino U Detecting lung cancer in plasma with the use of multiple genetic markers Int J Cancer 2004 108 91 96 14618621 10.1002/ijc.11510
Cope KA Watson MT Foster WM Sehnert SS Risby TH Effects of ventilation on the collection of exhaled breath in humans J Appl Physiol 2004 96 1371 1379 14672964 10.1152/japplphysiol.01034.2003
Phillips M Method for the collection and assay of volatile organic compounds in breath Anal Biochem 1997 247 272 278 9177688 10.1006/abio.1997.2069
Lord H Pawliszyn J Evolution of solid-phase microextraction technology J Chromatogr A 2000 885 153 193 10941672 10.1016/S0021-9673(00)00535-5
Tsai JH Chiang HL Hsu YC Weng HC Yang CY The speciation of volatile organic compounds (VOCS) from motorcycle engine exhaust at different driving modes Atmos Environ 2003 37 2485 2496 10.1016/S1352-2310(03)00177-8
Darrall KG Figgins JA Brown RD Phillips GF Determination of benzene and associated volatile compounds mainstream cigarette smoke Analyst 1998 123 1095 1101 9709493 10.1039/a708664d
Stone BG Besse TJ Duane WC Evans CD DeMaster EG Effect of regulating cholesterol biosynthesis on breath isoprene excretion in men Lipids 1993 28 705 708 8377584
Pitkänen OM Hallman M Andersson SM Determination of ethane and pentane in free oxygen radical-induced lipid peroxidation Lipids 1989 24 157 159 2569148
Phillips M Cataneo RN Greenberg J Grodman R Gunawardena R Naidu A Effect of oxygen on breath markers of oxidative stress Eur Respir J 2003 21 48 51 12570108 10.1183/09031936.02.00053402
Mitsui T Kondo T Inadequacy of theoretical basis of breath methylated alkane contour for assessing oxidative stress Clin Chim Acta 2003 333 93 94 10.1016/S0009-8981(03)00173-6
Mitsui T Naitoh K Tsuda T Hirabayashi T Kondo T Is endogenous isoprene the only coeluting compound in the measurement of breath pentane? Clin Chim Acta 2000 299 193 198 10900304 10.1016/S0009-8981(00)00275-8
Jones AW Lagesson V Tagesson C Determination of isoprene in human breath by thermal desorption gas chromatography with ultraviolet detection J Chromatogr B 1995 672 1 6
Miekisch W Schubert JK Vagts DA Geiger K Analysis of volatile disease markers in blood Clin Chem 2001 47 1053 1060 11375291
|
16018807
|
PMC1185565
|
CC BY
|
2021-01-04 16:23:26
|
no
|
Respir Res. 2005 Jul 14; 6(1):71
|
utf-8
|
Respir Res
| 2,005 |
10.1186/1465-9921-6-71
|
oa_comm
|
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-721602272910.1186/1465-9921-6-72ResearchAirway cellularity, lipid laden macrophages and microbiology of gastric juice and airways in children with reflux oesophagitis Chang AB [email protected] NC [email protected] J [email protected] JM [email protected] PJ [email protected] GJ [email protected] LC [email protected] GD [email protected] MK [email protected] J [email protected] Department of Paediatrics, University of Queensland, Brisbane, Australia2 Department of Respiratory Medicine, Royal Children's Hospital, Brisbane, Australia3 Department of Gastroenterology, Royal Children's Hospital, Brisbane, Australia4 Department of Anatomical Pathology and Cytopathology, Queensland Health Pathology Service, Royal Brisbane Hospital, Brisbane, Australia5 Department of Microbiology, Queensland Health Pathology Service, Royal Brisbane Hospital, Brisbane, Australia2005 15 7 2005 6 1 72 72 29 3 2005 15 7 2005 Copyright © 2005 Chang et al; licensee BioMed Central Ltd.2005Chang 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
Gastroesophageal reflux disease (GORD) can cause respiratory disease in children from recurrent aspiration of gastric contents. GORD can be defined in several ways and one of the most common method is presence of reflux oesophagitis. In children with GORD and respiratory disease, airway neutrophilia has been described. However, there are no prospective studies that have examined airway cellularity in children with GORD but without respiratory disease. The aims of the study were to compare (1) BAL cellularity and lipid laden macrophage index (LLMI) and, (2) microbiology of BAL and gastric juices of children with GORD (G+) to those without (G-).
Methods
In 150 children aged <14-years, gastric aspirates and bronchoscopic airway lavage (BAL) were obtained during elective flexible upper endoscopy. GORD was defined as presence of reflux oesophagitis on distal oesophageal biopsies.
Results
BAL neutrophil% in G- group (n = 63) was marginally but significantly higher than that in the G+ group (n = 77), (median of 7.5 and 5 respectively, p = 0.002). Lipid laden macrophage index (LLMI), BAL percentages of lymphocyte, eosinophil and macrophage were similar between groups. Viral studies were negative in all, bacterial cultures positive in 20.7% of BALs and in 5.3% of gastric aspirates. BAL cultures did not reflect gastric aspirate cultures in all but one child.
Conclusion
In children without respiratory disease, GORD defined by presence of reflux oesophagitis, is not associated with BAL cellular profile or LLMI abnormality. Abnormal microbiology of the airways, when present, is not related to reflux oesophagitis and does not reflect that of gastric juices.
==== Body
Introduction
Gastroesophageal reflux (GOR) is very common and defined as the passage of gastric contents into the oesophagus. GOR disease (GORD) is defined as symptoms or complications of GOR [1]. GORD includes the presence of oesophagitis, histologically defined on oesophageal biopsy, and increased reflux index on pHmetry in association with appropriate symptoms [2,1]. GOR and GORD is associated with pulmonary disease and postulated mechanisms include aspiration of gastric components, tracheo-gastric reflex, and sensory nerve stimulation [3]. Secondary aspiration relates to aspiration of gastric contents, which contain oral and ingested micro-organisms as well as gastric juices. With no gold standard of defining recurrent aspiration [4], current tests include nuclear medicine tests).)[5] and, quantification of lipid laden macrophages in bronchoalveolar lavage (BAL) fluid [6]. It is controversial whether an increased lipid laden macrophages index (LLMI) is a useful indicator for recurrent pulmonary aspiration [7,8]. Although LLMI is increased in aspiration lung disease, it is also found in other lung diseases [6,9,10]. Increased LLMI has been well documented in highly selected groups of children with lung disease [11] but there is little data regarding BAL of children without lung disease and GORD.
Assessment of airway profile is increasingly used in research as well as in clinical medicine for supportive (but not definitive) diagnosis of respiratory diseases in children [12,13]. The airway cellular profile of children with GORD without chronic lung disease is unknown and this knowledge will be useful for comparative clinical and research purposes.
The aims of the study were to compare, (1) BAL cellularity and lipid laden macrophage index (LLMI) and, (2) microbiology of BAL and gastric juices of children with GORD (G+) to those without (G-). We hypothesised that airways of children with reflux oesophagitis were more likely to have increased LLMI and neutrophilia from recurrent small volume aspiration, and that the bacterial flora in the lungs would be similar to that in the gastric aspirate fluid.
Methods
Children aged 0.75–14 years undergoing elective flexible upper endoscopy were invited to participate in the study (August 2002 till June 2004). All children undergoing flexible upper endoscopy had seen a consultant paediatric gastroenterologist and the procedure performed under general anaesthesia including endotracheal intubation. Children were enrolled for the study on the morning of their procedure. Medical history was obtained from a parent on a standardised proforma for all children. Exclusion criteria were; children with neuro-developmental abnormalities, known underlying cardiorespiratory disease other then asthma and those with a clinical history of primary aspiration (coughs and chokes with feeds at least twice a week). GORD was considered present if histology of distal oesophageal biopsy showed reflux oesophagitis determined by pathologists blinded to the child's respiratory history [14]. Written consent was obtained and the study approved by our institution's human ethics committee.
At commencement of the flexible upper endoscopy, gastric juice was obtained directly under vision and suctioned into a mucus trap. When <0.5 mls was obtained, 5 or 10 mls of saline flush was used and colony count of an organism (if cultured) was corrected by the same factor. To obtain BAL fluid, a non-bronchoscopic standardised and repeatable [15]. technique was utilised. Briefly, with the child's head turned to the left, an 8F catheter was passed as far as possible through the endotracheal tube, ensuring that it went beyond the estimated carina site. Sterile normal saline (1 ml/kg to maximum of 20 mls) was instilled and suctioned into a mucus trap and this specimen was used for microbiology examination. A further 1 ml/kg (maximum of 20 mls) was instilled and the 2nd collection utilised for cytology and lipid laden macrophage count. Cell count was performed on the cell suspension, cytocentrifuge slides were prepared and stained (modified Wright's stain, Diff Quik, Lab Aids, Narrabeen, NSW, Australia) for cell differential profile (400 cells counted when possible). Additional slides were prepared for LLMI using Oil Red O stain (Sigma Chemicals) where 100 macrophages counted and scored 0–4 [16,4]. LLMI (range 0–400) was obtained by the addition of these scores. All cellular and LLMI examinations were performed by cytologists blinded to the children's medical history.
Quantitative aerobic cultures of bacteria were undertaken on BAL and gastric juices using standard sterile loops (10 & 100 ul) on blood and chocolate agar plates for detection of aerobic bacteria. Plates were incubated at 35°C for 48-hours and isolates counted and identified to the genus level. Positive bacterial culture was defined as growth of ≥104 colony forming unit/ml [17]. Viral studies were also performed on BAL; direct immunofluorescence antigen (DFA) was used to detect RSV, adenovirus, parainfluenza viruses 1,2,3 and influenza A and B. If viral direct immunofluorescent antigen testing was negative, nucleic acid amplification (NAA) tests were undertaken for all the above viruses using multiplex PCR [18]
Statistical analysis
Children were categorised having GORD (G+) and not having GORD (G-). Chi square was used to compare categorical variables between groups and odds ratio described. Data were not normally distributed and thus non parametric analyses were used; Mann-Whitney for comparisons between 2 groups and Kruskal Wallis when >2 groups were compared. Medians and inter-quartile range (IQR) were used for all descriptive data. Two tailed p value of <0.05 was considered significant. SPSS ver 11 was utilised for all statistical calculation.
Results
Median age of the 150 children (91 boys and 56 girls) recruited was 8.2 years (IQR 7). The primary indications for oesophago-gastroscopy were abdominal pain (n = 77), recurrent vomiting (n = 35), poor weight gain (n = 20), review of previous lesion (n = 19) and choking (n = 17); some children had more than one primary indication for oesophago-gastroscopy. Most (n = 136, 90.7%) children were clinically suspected of having GORD and oesophagitis was present in 77 (51.3%) children. There were 77 children in G+ category, 73 in G-. Only 17 children had tobacco smoke exposure and as numbers were small, comparisons were not made.
G- group (median 7, IQR 13) had a significant but small increase in BAL neutrophil % when compared to the G+ group (5, 7.5), p = 0.002. There was no significant difference in percentages of macrophages, lymphocytes, eosinophils and LLMI in BAL between G+ and G- groups (table 1, p range 0.23 to 0.78). When children whose BAL showed positive bacterial culture were excluded (n = 31), BAL neutrophil % was still significantly higher in G- (n = 52) than G+ (n = 67) children, p = 0.009; median of 7 vs 4% respectively. Percentages of macrophages, lymphocytes, eosinophils and LLMI in BAL between G+ and G- groups remained not significantly different when those with BAL positive culture were excluded.
Table 1 Cellular profile of children grouped by presence and absence of GORD
G+ N = 77 G- N = 73
Lymphocyte %
Median, IQR 4.0, 3 4.4, 4
Neutrophil %
Median, IQR 5, 7.5 7.5, 14
Macrophage %
Median, IQR 89, 16 88.3, 14
Eosinophil %
Median, IQR 0, 0 0, 0
Total cell count
Median, IQR 96, 175 116, 83.5
LLM index
Median, IQR 42, 47 38, 30
G+ = group with reflux oesophagitis
G- = group without reflux oesophagitis
Viral studies (DFA and NAA) were negative in all the BAL samples. Positive bacterial cultures was found in the BALs of 31 children (20.7%) and the gastric aspirates from 8 (5.3%) children, table 2. Eighteen children had a significant growth of S. pneumoniae in their BAL but only one of these children had significant growth of S pneumoniae in their gastric aspirate (105). S. aureus was found in the BAL of 2 children and one of these also had a low count(102) of S. aureus in their gastric aspirate. The cellular profile of these 31 children (median %neutrophils was 20, IQR 34; %lymphocytes 6, 8.5; %macrophages 65.5, 34) was significantly different to those without positive BAL culture (%neutrophils 5, 4; %lymphocytes 3, 4; %macrophages 90, 9); p of 0.00001 for all cell types. Children who were G+ were no more likely to have positive BAL culture than the G- group (table 3). Thirty seven children had recent use (within a week) of proton pump inhibitors (PPI). Use of PPI did not influence G+/G- status (p = 0.452) and also had no significant effect on BAL positive culture state (p = 0.762) or gastric aspirate culture (p = 0.092).
Table 2 Positive bacteria culture in BAL and gastric aspirate
Growth of organism of ≥104 colony forming unit BAL n Gastric aspirate n
S. pneumoniae 19 6
H. influenzae 10 0
M. catarrhalis 7 0
S. aureus 3 0
Candida 0 2
Table 3 Comparisons of groups with positive bacterial culture in BAL
BAL culture
Group category Negative (<104 cfu/ml) Positive (≥104 cfu/ml) p value OR, 95% CI
G- 52 17
G+ 67 14 0.27 0.64, 0.29–1.42
G+ = group with reflux oesophagitis
G- = group without reflux oesophagitis
Discussion
In 150 children, we have shown that children with GORD and without an underlying lung problem have no abnormality in their airway cellular profile or LLMI. Indeed the percentage of neutrophils was significantly higher in the G- group than in the G+ group (the difference between the groups was however small and not clinically significant). We have also shown that positive bacterial culture with recognised respiratory pathogens was relatively high at 20.6% and was not influenced by G+ state. Lastly we showed that the microbiology of airways does not reflect that of gastric aspirates in children with and without GORD.
This is the first study that has examined airway cellularity and microbiology of airways of children without an underlying respiratory illness in relation with GORD. Although the percentage of neutrophils in BAL of G- children was higher than that in G+ children, the difference between the groups was small (median difference of 2%) and in the clinical context this is not significant. These BAL values are very close to the range described in normal children [12]. There is a paucity of data on airway findings in patients with GORD without an underlying lung disease. Our findings are similar to a small (n = 11) study in adults with GORD (without lung disease) which also described that GORD was not associated with airway neutrophilia. Our findings of absence of airway neutrophilia in children with GORD are in contrast to studies that have examined BAL in children with chronic respiratory disease [11]. It is likely that airway neutrophilia in the presence of respiratory illness is from the respiratory disease itself rather than from stimulation of the tracheo-gastric reflex. Indeed in our study, the percentage of neutrophils in BAL of children with GORD was lower than that in children without GORD. However, the absence of airway neutrophilia does not mean absence of neutrophilic inflammation as we did not examine for neutrophilic markers such as IL-8.
We did not find any difference in LLMI in children with or without GORD. In our study, LLMI is not a useful marker of presence of reflux oesophagitis in children without an underlying respiratory illness. The plausible explanations include; none of these children had secondary aspiration or/and LLMI is not a sensitive test for secondary aspiration. Indeed the utility of LLMI as a sensitive and specific marker of aspiration in children has been questioned [6,9,10]., and Colombo reported that LLMI was highest in patients underdoing chemotherapy and graft vs host disease [4]. Krishnan and colleagues recently described the poor specificity and sensitivity of LLMI for aspiration [19]. However they used tracheal aspirates[4], which is not representative of BAL [20,21]. BAL fluid is the common standardised method for examining airway cellularity in the respiratory field [12].
We found a high incidence of positive bacterial culture (of common respiratory bacteria in the BAL but not in gastric aspirate) in our cohort of children, defined on a chosen threshold of ≥104 cfu/ml based on previous studies [17]. However, the diagnostic threshold for quantitative culture on BAL for bronchitis in children (as opposed to pneumonia) is unknown and caution is needed in the interpretation of BAL microbiology [12]. As airway neutrophilia was also present in children with positive bacterial culture it is likely that culture results were significant. We did not find any relationship between airway and gastric aspirate microbiology, suggesting that aspiration of gastric pathogens was not significant in these children. Our findings are in support of a study showing that the stomach is not a source for colonization of the upper respiratory tract and pneumonia [22]. Swallowing of respiratory secretions of common respiratory bacteria is unlikely to result in positive gastric aspirate culture as gastric juice has a bactericidal effect; S. aureus is killed within 30–45 minutes of inoculation whereas P. aeruginosa is killed in 60–90 mins [23].
Our data may not be extrapolated to other definitions of GORD. We examined GORD defined by oesophageal biopsy, a common method of diagnosing GORD in children in Australia. In our institution, oesophago-gastroscopy is more commonly performed than pHmetry (approximately 800 and 300 respectively per year). A variety of techniques are utilised to confirm or support the diagnosis of GORD. However there is no single perfect method for the objective definition of all GORD types. Each modality has its advantages and disadvantages/limitations. Nevertheless, arguably oesophagitis is the gold standard definition but is also likely the least sensitive diagnostic method especially when GORD is mild and/or occasional or intermittent. The American Gastroenterology Association (AGA) guidelines on pHmetry stated "In the absence of oesophagitis, there is no gold standard for the definition of GERD..." [2]. pHmetry has been reported to be more sensitive but there is considerable disagreement what constitutes an abnormal pHmetry [1]. Also, some gastroenterologists argue that pHmetry alone cannot be used to diagnose GORD, given the known problems outlined in the AGA guidelines. Thus, similar findings using GORD based on pHmetry and perhaps multi-channel intraluminal electrical impedance monitoring, (arguably a more sensitive method for evaluating GORD variants ie acid and non-acid reflux [24]) would be necessary to conclude that all types of GORD are associated with normal airway cellular profile in otherwise well children. However, given that there was no significant abnormality in both groups of children examined in this study, it is unlikely that any type of acid related GORD will be associated with abnormal airway cellularity in children without lung disease.
Conclusion
We conclude that, in children without respiratory disease, presence of GORD defined by reflux oesophagitis, is not associated with BAL cellular profile or LLMI abnormality. Abnormal microbiology of the airways, when present, does not reflect that of gastric juices and, is not associated with reflux oesophagitis.
Acknowledgements
We are grateful to J Gaffney, Mary DaSilva and the nurses of the Department of Gastroenterology, Royal Children's Hospital as well as the anaesthetists in particular Drs Chris Beem, A Newton, I Webb, D Hill, and M Pabari, without whom this project would not be possible. This study was partially funded by the Royal Children's Hospital Foundation and by Sylvia and Charles Viertel Charitable Foundation. ABC is supported by the Royal Children's Hospital Foundation and by a Practitioner Fellowship from the Australian National Health and Medical Research Council.
==== Refs
Rudolph CD Mazur LJ Liptak GS Baker RD Boyle JT Colletti RB Gerson WT Werlin SL Guidelines for evaluation and treatment of gastroesophageal reflux in infants and children: recommendations of the North American Society for Pediatric Gastroenterology and Nutrition J Pediatr Gastroenterol Nutr 2001 32 Suppl 2 S1 31 11525610 10.1097/00005176-200100002-00001
American Gastroenterological Association medical position statement: guidelines on the use of esophageal pH recording Gastroenterology 1996 110 1981 8964427 10.1053/gast.1996.1101981
Daoui S Agostino B Gallelli L Emonds X Rossi F Advenier C Tachykinins and airway microvascular leakage induced by HCl intra-oesophageal instillation Eur Respir J 2002 20 268 273 12212954 10.1183/09031936.02.00250902
Colombo JL Pulmonary aspiration and lipid-laden macrophages: In search of Gold (standards) Pediatr Pulmonol 1999 28 79 82 10423305 10.1002/(SICI)1099-0496(199908)28:2<79::AID-PPUL1>3.0.CO;2-A
Bar-Sever Z Connolly LP Treves ST The radionuclide salivagram in children with pulmonary disease and a high risk of aspiration Pediatr Radiol 1995 25 Suppl 1 S180 S183 8577521
Bauer ML Lyrene RK Chronic aspiration in children: evaluation of the lipid-laden macrophage index Pediatr Pulmonol 1999 28 94 100 10423308 10.1002/(SICI)1099-0496(199908)28:2<94::AID-PPUL4>3.0.CO;2-0
Colombo JL Hallberg TK Recurrent aspiration in children: lipid-laden alveolar macrophage quantitation Pediatr Pulmonol 1987 3 86 89 3588061
Lopez-Andreu JA Roques-Serradilla JM Cortell-Aznar I Fat-laden macrophages as a marker of reflux aspiration: still of some help J Pediatr Gastroenterol Nutr 2003 36 652 653 12717094 10.1097/00005176-200305000-00017
Knauer-Fischer S Ratjen F Lipid-laden macrophages in bronchoalveolar lavage fluid as a marker for pulmonary aspiration Pediatr Pulmonol 1999 27 419 422 10380094 10.1002/(SICI)1099-0496(199906)27:6<419::AID-PPUL9>3.0.CO;2-U
Kazachkov MY Muhlebach MS Livasy CA Noah TL Lipid-laden macrophage index and inflammation in bronchoalveolar lavage fluids in children Eur Respir J 2001 18 790 795 11757629 10.1183/09031936.01.00047301
Sacco O Fregonese B Silvestri M Sabatini F Mattioli G Rossi GA Bronchoalveolar lavage and esophageal pH monitoring data in children with "difficult to treat" respiratory symptoms Pediatr Pulmonol 2000 30 313 319 11015132 10.1002/1099-0496(200010)30:4<313::AID-PPUL7>3.0.CO;2-H
de Blic J Midulla F Barbato A Clement A Dab I Eber E Green C Grigg J Kotecha S Kurland G Pohunek P Ratjen F Rossi G Bronchoalveolar lavage in children. ERS Task Force on bronchoalveolar lavage in children. European Respiratory Society Eur Respir J 2000 15 217 231 10678650
Riedler J Grigg J Robertson CF Role of bronchoalveolar lavage in children with lung disease Eur Respir J 1995 8 1725 1730 8586129 10.1183/09031936.95.08101725
Yellon RF Coticchia J Dixit S Esophageal biopsy for the diagnosis of gastroesophageal reflux-associated otolaryngologic problems in children The American Journal of Medicine 108 131 138 6-3-2000 10.1016/S0002-9343(99)00352-6
Warke TJ Kamath S Fitch PS Brown V Shields MD Ennis M The repeatability of nonbronchoscopic bronchoalveolar lavage differential cell counts Eur Respir J 2001 18 1009 1012 11829083 10.1183/09031936.01.00203101
Corwin RW Irwin RS The lipid-laden alveolar macrophage as a marker of aspiration in parenchymal lung disease Am Rev Respir Dis 1985 132 576 581 4037530
Labenne M Poyart C Rambaud C Goldfarb B Pron B Jouvet P Delamare C Sebag G Hubert P Blind protected specimen brush and bronchoalveolar lavage in ventilated children Crit Care Med 1999 27 2537 2543 10579277 10.1097/00003246-199911000-00035
Syrmis MW Whiley DM Thomas M Mackay IM Williamson J Siebert DJ Nissen MD Sloots TP A sensitive, specific, and cost-effective multiplex reverse transcriptase-PCR assay for the detection of seven common respiratory viruses in respiratory samples J Mol Diagn 2004 6 125 131 15096568
Krishnan U Mitchell JD Tobias V Day AS Bohane TD Fat laden macrophages in tracheal aspirates as a marker of reflux aspiration: a negative report J Pediatr Gastroenterol Nutr 2002 35 309 313 12352518 10.1097/00005176-200209000-00013
Malikides N Hughes KJ Hodgson DR Hodgson JL Comparison of tracheal aspirates and bronchoalveolar lavage in racehorses. 2. Evaluation of the diagnostic significance of neutrophil percentage Aust Vet J 2003 81 685 687 15086110
Mentec H May-Michelangeli L Rabbat A Varon E Le Turdu F Bleichner G Blind and bronchoscopic sampling methods in suspected ventilator-associated pneumonia. A multicentre prospective study Intensive Care Med 2004 30 1319 1326 15098088 10.1007/s00134-004-2284-7
Bonten MJ Gaillard CA van Tiel FH Smeets HG van der GS Stobberingh EE The stomach is not a source for colonization of the upper respiratory tract and pneumonia in ICU patients Chest 1994 105 878 884 8131556
Williams C Occurrence and significance of gastric colonization during acid-inhibitory therapy Best Pract Res Clin Gastroenterol 2001 15 511 521 11403543 10.1053/bega.2001.0191
Shay S Tutuian R Sifrim D Vela M Wise J Balaji N Zhang X Adhami T Murray J Peters J Castell D Twenty-four hour ambulatory simultaneous impedance and pH monitoring: a multicenter report of normal values from 60 healthy volunteers Am J Gastroenterol 2004 99 1037 1043 15180722 10.1111/j.1572-0241.2004.04172.x
|
16022729
|
PMC1185566
|
CC BY
|
2021-01-04 16:23:27
|
no
|
Respir Res. 2005 Jul 15; 6(1):72
|
utf-8
|
Respir Res
| 2,005 |
10.1186/1465-9921-6-72
|
oa_comm
|
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-751603364010.1186/1465-9921-6-75ResearchCCR2 and CXCR3 agonistic chemokines are differently expressed and regulated in human alveolar epithelial cells type II Pechkovsky Dmitri V [email protected] Torsten [email protected] Corinna [email protected] Antje [email protected] Ekkehard [email protected]üller-Quernheim Joachim [email protected] Gernot [email protected] Department of Pneumology, Medical Center, Albert-Ludwigs University, Freiburg, Germany2 Research Institute for Pulmonary Diseases and Tuberculosis, Minsk, Belarus3 Division of Infectious Diseases, University of British Columbia, Vancouver, British Columbia V5Z 3J5, Canada4 Division of Clinical and Experimental Pathology, Research Center Borstel, Borstel, Germany5 Department of Thoracic Surgery, Albert-Ludwigs University, Freiburg, Germany6 Lungenklinik, Krankenhaus Merheim, Kliniken der Stadt Köln, Köln, Germany2005 20 7 2005 6 1 75 75 16 2 2005 20 7 2005 Copyright © 2005 Pechkovsky et al; licensee BioMed Central Ltd.2005Pechkovsky 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 attraction of leukocytes from circulation to inflamed lungs depends on the activation of both the leukocytes and the resident cells within the lung. In this study we determined gene expression and secretion patterns for monocyte chemoattractant protein-1 (MCP-1/CCL2) and T-cell specific CXCR3 agonistic chemokines (Mig/CXCL9, IP-10/CXCL10, and I-TAC/CXCL11) in TNF-α-, IFN-γ-, and IL-1β-stimulated human alveolar epithelial cells type II (AEC-II). AEC-II constitutively expressed high level of CCL2 mRNA in vitro and in situ , and released CCL2 protein in vitro . Treatment of AEC-II with proinflammatory cytokines up-regulated both CCL2 mRNA expression and release of immunoreactive CCL2, whereas IFN-γ had no effect on CCL2 release. In contrast, CXCR3 agonistic chemokines were not detected in freshly isolated AEC-II or in non-stimulated epithelial like cell line A549. IFN-γ, alone or in combination with IL-1β and TNF-α resulted in an increase in CXCL10, CXCL11, and CXCL9 mRNA expression and generation of CXCL10 protein by AEC-II or A549 cells. CXCL10 gene expression and secretion were induced in dose-dependent manner after cytokine-stimulation of AEC-II with an order of potency IFN-γ>>IL-1β ≥ TNF-α. Additionally, we localized the CCL2 and CXCL10 mRNAs in human lung tissue explants by in situ hybridization, and demonstrated the selective effects of cytokines and dexamethasone on CCL2 and CXCL10 expression. These data suggest that the regulation of the CCL2 and CXCL10 expression exhibit significant differences in their mechanisms, and also demonstrate that the alveolar epithelium contributes to the cytokine milieu of the lung, with the ability to respond to locally generated cytokines and to produce potent mediators of the local inflammatory response.
==== Body
Background
Many pulmonary disorders are characterized by accumulation and activation of inflammatory cells within the lung, followed by the release of regulatory mediators, resulting in macrophage/lymphocyte alveolitis. Sarcoidosis, tuberculosis, hypersensitivity pneumonitis, eosinophilic pneumonia, and usual interstitial pneumonia represent such lung diseases that have in common the selective recruitment and activation of different types of leukocytes, and therefore, exhibit distinct forms of alveolitis [1-5]. The inflammatory phase of alveolitis is initiated by epithelial and/or endothelial injury involving the structures of the alveolar wall. The alveolar surface area of the lung is covered with a layer of alveolar epithelial cells type I and type II. Type I cells function as a physical barrier, whereas type II cells produce surfactant and act as progenitors to replace injured alveolar epithelial cells type I [6]. Thus, located at the boundary between the alveolar airspace and the interstitium, alveolar epithelial cells type II (AEC-II) are ideally situated to regulate the recruitment and activation of different types of leukocytes through the production of chemokines/cytokines in response to inflammatory stimulation from the alveolar space. Recent studies have suggested that AEC-II secrete a variety of mediators, including proinflammatory cytokines and chemokines important for the recruitment of monocytes / macrophages and T cells into the lung interstitium and alveolar space [7-10].
Although leukocyte recruitment is a complex and multistep process with involvement of different types of cells, cell-surface adhesion molecules, and soluble inflammatory mediators, the prominent role of the attractant molecules such as chemokines has widely been appreciated [11,12]. Chemokines are a superfamily of small, secreted proteins that direct the recruitment of leukocytes to the sites of inflammation. They are classified into four subfamilies on the basis of the primary sequence of the first two of four invariant cysteine residues, and named according to the recommendation for new systematic nomenclature for human chemokines [11]. CC chemokines/CCL attract monocytes, eosinophils, basophils, dendritic and T cells and signal through chemokine receptors CCR1 to CCR10. In contrast to CC chemokines, the CXC chemokines (CXCL) are divided into two classes depending on the presence of the glutamate-leucine-arginine motif (ELR) in the NH2-terminal domain. The CXC chemokines signal through the chemokine receptors CXCR1 to CXCR5 (reviewed in [11]). The CC chemokine, monocyte chemoattractant protein-1/CCL2 (CCL2), has been shown in vitro and in vivo to target preferentially monocytes and memory T cells through the CCR2 [13-16]. Monokine induced by IFN-γ (Mig/CXCL9), IFN-induced protein of 10 kDa (IP-10/CXCL10), and IFN-inducible T-cell α-chemoattractant (I-TAC/CXCL11) are all members of the non-ELR CXCL class and target preferentially memory T cells and natural killer cells through the single and shared receptor CXCR3 [17,18]. Recently, it has been reported that some chemokine receptors are associated with human Th1 or Th2 cells, and therefore the respective agonists can selectively attract the respective Th cell subset into inflammatory sites (reviewed in [12]).
In this context, we hypothesized that AEC-II are an important source of CCL2 and the CXCR3 agonistic chemokines in the lung, and through expression of these mediators involved in the homing of immune effector cells during lung inflammatory processes. As a model we investigated the gene expression and production of chemokines, important for the recruitment of CCR2 and CXCR3 bearing mononuclear leukocytes, by human primary AEC-II and airway epithelial like cell line A549 after exposure of the cells to the proinflammatory cytokines TNF-α, IFN-γ, and IL-1β. A striking result was the difference between spontaneous and cytokine-induced CCL2, CXCL9, CXCL10, and CXCL11 mRNA expression and/or protein production in both human AEC-II and A549 cell cultures. Finally, we provide evidence of selective CCL2 and CXCL10 mRNA expression of human AEC-II in vivo .
Materials and Methods
Reagents
The following materials were purchased from GIBCO BRL (Paisley, Scotland): PBS, RPMI 1640 medium with 2 mM L-glutamine, FCS, HEPES, TRIZOL Reagent, SuperScript™ RNase H- reverse transcriptase (RT), oligo (dT)12–18 primer and agarose; penicillin/streptomycin solution and sodium pyruvate from Biochrom (Berlin, FRG); trypsin/EDTA solution from Boehringer-Mannheim (Mannheim, FRG); collagen R from Serva (Heidelberg, FRG); chloroform and isopropanol from Merck (Darmstadt, FRG); recombinant human IFN-γ (specific activity 3 × 107 U/mg) and recombinant human IL-1β (specific activity 2 × 108 U/mg) from Biotrend (Cologne, FRG); recombinant human TNF-α was a courtesy of Dr. E. Schlick (Knoll AG, Ludwigshafen, FRG); dexamethasone from Sigma (St. Louis, MO); 100 mm plastic dishes, 75 cm2 tissue culture flask and 24-well cell culture plates from NUNC (Wiesbaden, FRG). All reagents used were of the highest available grade and were dissolved in pyrogen-free water.
Human Lung Tissue
Lung tissue samples were obtained from subjects with lung cancer undergoing lobectomy or pneumectomy. Twelve patients with bronchogenic carcinoma, without any other systemic or pulmonary diseases, were enrolled in this study. All subjects were smokers and have had no respiratory tract infection within the last month. None of them was taking immunosuppressants within one month before surgery. In addition, lung tissue samples were obtained from 3 patients with pulmonary sarcoidosis who had undergone diagnostic wedge biopsies and from 3 patients with pulmonary tuberculosis who had undergone upper lobectomy due to destructive tuberculoma. Informed consents were obtained from all subjects. The study was approved by the medical ethics committees of the involved institutions.
Primary Human Alveolar Epithelial Cells Type II
Samples from macroscopically tumor-free lung tissue were cut from the surgical specimens and used for cell isolation procedure as described previously [19]. In brief, the lung tissue was first sliced and slices were washed three times at 4°C in PBS. The washed slices were incubated in sterile dispase solution at 37°C for 45 min. After dispase digestion the lung tissue slices were cut into small, pipetable pieces, and thoroughly pipetted for several min. Crude tissue and cell suspensions were filtered through nylon gauze with meshes of 100 μm, 50 μm, and 20 μm. The resulting single cell suspension was placed on Ficoll separating solution and centrifuged at 800 × g for 20 min. The AEC-II-enriched cells from the interphase were incubated in 100 mm plastic dishes at 37°C in humidified air containing 5% CO2 for 15, 20 and 30 min with seeding of non-adherent cells on fresh dishes for each time interval to remove adherent cells (alveolar macrophages, monocytes, fibroblasts, and endothelial cells). To remove remaining monocytes/macrophages and lymphocytes, antibodies against CD3 (OKT3, ECACC 86022706) and CD14 (HB-246 ATCC) were added and the antibody-binding cells were removed by anti-mouse IgG coated magnetic beads and Magnetic Activated Cell Sorting (MACS) system (Miltenyi Biotec, Bergisch Gladbach, FRG) as suggested by the supplier. Identity of type II alveolar epithelial cells was confirmed by a modified Papanicolaou staining, their alkaline phosphatase activity, and SP-A mRNA expression in RT-PCR (see below). Cell purity was assessed by immunoperoxidase staining with monoclonal antibodies directed against CD3 and CD14 (Immunotech, Marseille, France) as previously described [20]. Viability of the AEC-II after isolation was > 97% as determined by trypan blue exclusion. After the final step of MACS purification, the AEC-II preparations included in this report were free of CD14+ and CD3+ cells as determined by immunocytochemistry. 98 ± 1.3% of cells were identified as AEC-II by the presence of dark blue inclusions as revealed by modified Papanicolaou staining and 93 ± 2.1% of cells were positive for alkaline phosphatase (data not shown). All RNA samples isolated from these AEC-II preparations contained SP-A mRNA, and CD3 and CD14 mRNA were found in four of twelve samples by RT-PCR (data not shown). In order to avoid false positive results from contaminated cells, these four AEC-II preparations were excluded from further experimental data analysis.
A549 Cell Line
A549 cells were used as the positive control for CCL2, CXCL9, CXCL10, and CXCL11 mRNA expression and protein production upon stimulation with proinflammatory cytokines. Experiments were performed with cells after 7, 8 and 9 passages after thawing and inoculation in culture. Cells were grown on 75 cm2 tissue culture flask in culture medium (CM) (RPMI1640 medium, 10% heat inactivated FCS, 1% penicillin/streptomycin solution, 1% sodium pyruvate solution and 20 mM HEPES) in a humidified atmosphere containing 5% CO2 at 37°C for 5 days. After this culture period, cells were removed from plastic surfaces by treatment with trypsin/EDTA solution (0.05/0.02% in PBS) for 10 min at 37°C, washed twice in PBS and suspended in CM.
Cell Cultures
Immediately after purification, AEC-II were suspended in CM (1 × 106 cells/ml) and treated with TNF-α (1 – 10 ng/ml), IFN-γ (10 – 100 U/ml) or IL-1β (10 – 100 U/ml) in collagen R-coated 24-well plates at 37°C, 5% CO2 atmosphere. A549 cells were plated at 1 × 106 /ml in 24-well plates in the same culture condition as for AEC-II and stimulated with TNF-α (1 – 10 ng/ml), IFN-γ (50 – 500 U/ml) or IL-1β (50 – 500 U/ml) in different combinations as indicated in the Results section. At the indicated time, cell-free supernatants were harvested and stored at -70°C, and cell pellets were extracted for total RNA. The cell viability after culture always exceeded 95% in both AEC-II and A549 cells as determined by trypan blue exclusion. For samples of RNA from freshly isolated AEC-II or harvested A549 cells, they were subjected to RNA isolation procedures before cultures, henceforth referred to as non-cultured controls.
Reverse Transcriptase Polymerase Chain Reaction (RT-PCR)
Total RNA was extracted from cells using TRIzol according to manufacturer's instructions (GIBCO BRL). Equal amounts of total RNA from each sample were primed with oligo dT and reverse-transcribed with SuperScript™ RT for 1 h at 37°C to produce complementary DNA (cDNA). The resulting cDNAs (volume of 2.5 μL) were used for the amplification by PCR of specific targets: CCL2, CXCL10, CXCL11, CXCL9, SPA, and the housekeeping gene β-actin. To demonstrate that RNA samples from AEC-II were not contaminated by RNAs from other types of cells (lymphocytes or alveolar macrophages (AM)) CD3- and CD14-specific primers were also used. All primers were intron-spanning to avoid false positive results by contamination with genomic DNA (Table 1). Target cDNA was amplified using a three-step PCR and an automated thermocycler (Biometra, Göttingen, FRG) according to Murray et al. [21] with primer pairs for CD3 and CD14, and as previously described [19] with primer pairs for β-actin. PCR conditions for CCL2 amplification included: 95°C for 5 min, 95°C for 30 s, 60°C for 30 s, 72°C for 1 min, and 72°C (terminal extension) for 5 min; for CXCL10, CXCL11, and CXCL9: 94°C for 1 min, 53°C for 1 min, 72°C for 2 min; and for SP-A: 94°C for 1 min, 54°C for 1 min, 72°C for 1 min 30 s, and 72°C (terminal extension) for 15 min. The numbers of cycles were the same for all targets (35 cycles), with the exception for SP-A (30 cycles). PCR products (for predicted sizes see Table 1) were electrophoresed on 1.5% agarose gels and stained with GelStar® stain (FMC BioProducts, Rockland, ME). Gel analysis was done densitometrically with "Gel Doc 2000" gel documentation system and "Quantity One 4.0.3" software (Bio-Rad Laboratories, Hercules, CA). To ensure that RNA was effectively reverse transcribed to cDNA for each condition and that stimulation with cytokines by itself did not have any effect on the housekeeping gene β-actin expression, the β-actin PCR was routinely performed in each experiment. To assure the identity of the PCR-amplified fragments, the size of each amplified mRNA fragment was compared with DNA standards (100 bp DNA Ladder; GIBCO BRL, Paisley, Scotland) electrophoresed on the same gel. Additionally, the PCR products were sequenced by the dideoxynucleotide chain-termination method with an autosequencer (ABI PRISM-377, Perkin-Elmer), and their specificity was further confirmed by comparing with the sequence data from the GenBank database (accession numbers M68519 for SP-A, X14768 for CCL2, AF030514 for CXCL11, NM002416 for CXCL9, and NM001565 for CXCL10) (data not shown). Results are expressed as percent of signal intensities assigned to the target mRNA of the corresponding signal produced by the amplimers for the β-actin gene using the same cDNA specimen.
Table 1 Primers used in RT-PCR analysis
cDNA Primer Sequence* Product Size (bp)
CCL2 F† : 5'-CAA ACT GAA GCT CGC ACT CTC GCC-3'
R† : 5'-ATT CTT GGG TTG TGG AGT GAG TGT TCA-3' 356
CXCL9 F: 5'-CGT GGT AAA ACA CTT GCG GAT ATT-3'
R: 5'-CAA TCA TGC TTC CAC TAA CCG ACT-3' 376
CXCL10 F: 5'-CCA TGA ATC AAA CTG CGA TTC TG-3'
R: 5'-CTT GGA AGC ACT GCA TCG ATT T-3' 338
CXCL11 F: 5'-AAA GGC TGG TTA CCA TCG GAG T-3'
R: 5'-RTGT TGC CAG TAT CCC ATA GCG T-3' 444
CD3 F: 5'-GGC TGT CCT CAT CCT GGC TAT CAT-3'
R: 5'-ACT GGT TTC CTT GAA GGT GGC TGT-3' 517
CD14 F: 5'-ACT CCC TCA ATC TGT CGT TCG CTG-3'
R: 5'-CTG AAG CCA AGG CAG TTT GAG TCC-3' 341
SP-A F: 5'-TCT TTG GAT GCC AAC TCA GC-3'
R: 5'-CTT TAT TCA GCT CAG GGG TG-3' 666
β-actin F: 5'-AGC GGG AAA TCG TGC GTG-3'
R: 5'-CAG GGT ACA TGG TGG TGCC-3' 309
*All primers were synthesized by MWG-Biotech (MWG-Biotech AG, Ebersberg, FRG); † F and R denote forward and reverse primer respectively
Measurement of CCL2 and CXCL10 Concentrations
Chemokines concentrations in A549 cell and primary cultured AEC-II supernatants were measured in duplicate by commercial available ELISA kits. Human CCL2 and CXCL10 ELISA kits were from HyCult biotechnology (Uden, the Netherlands). The assays were performed as suggested by the suppliers. Optical density readings were obtained with a MRX Microplate Reader and analyzed with Revelation 2.0 software (both from Dynex Technologies, FRG). The lower detection limit of the assays was 10 pg/ml for CCL2 and 20 pg/ml for CXCL10. For duplicate samples an intra assay coefficient of variation (CV) of < 10% and interassay CV of < 20% was accepted.
In Situ Hybridization (ISH)
Paraffin embedded lung tissue samples were prepared from the same surgical specimens as described above and used for ISH. These tissue samples showed normal architecture with few intra-alveolar macrophages and edema. Some lung tissue explants were placed in CM alone or with IFN-γ (500 U/ml) and IL-1β (500 U/ml), and/or 10-4 M dexamethasone and incubated at 37°C in humidified air containing 5% CO2 for 24 h. After incubation, these lung tissue explants were further used for ISH. The cDNA probes corresponding to CCL2 and CXCL10 mRNAs were produced by PCR as described before, filtered through Centri-Sep spin columns (Applied Biosystems, Foster City, CA), and labeled with digoxigenin (DIG) following the manufacturer's instructions (Dig-High-Prime, Roche, FRG). After deparaffinization, in situ hybridization was carried out overnight and, after washing at high stringency, detection was performed by application of Anti-Dig/alkaline-phosphatase-conjugate and new-fuchsin as substrate for alkaline phosphatase [22]. Slides were counterstained with Mayers hemalum and mounted with Kayser's glyceringelatine. For negative control, sections were hybridized with hybridization buffer in the absence of labeled cDNA probes. Hybridization of a probe targeting the mRNA of SP-A, a specific product of AEC-II, served as an additional positive control.
Statistical Analysis
Data are expressed as means ± SEM. Statistical comparisons were made by ANOVA with post hoc Fisher's protected least significant difference (PLSD) for each agent separately. Probability values were considered significant if they were less than 0.05. All testing was done using StatView 5.0 program (SAS Institute Inc., Cary, NC) for Macintosh computers.
Results
Chemokine mRNA expression by A549 cells
In preliminary experiments, RT-PCR was performed on the AEC-II-like cell line A549 to assess the spectrum of chemokine mRNA expression at baseline and in response to 24-h stimulation by TNF-α, IFN-γ, and IL-1β at different concentrations. In the same experiments, we also investigated the effects of the combinations of the above mentioned cytokines and different culture periods on chemokine mRNA expression by A549 cells. A549 cells spontaneously expressed mRNA for CCL2 (Figure 1A), and there was a moderate enhancement within 24 h of culture (Figure 1B). Stimulation with TNF-α, IFN-γ or IL-1β resulted in modulation of the steady-state level of CCL2 mRNA within 24 h, and at the end-time point of cultures proinflammatory cytokines slightly increased CCL2 mRNA expression level in a concentration-dependent manner (Figure 1B). Although the differences of CCL2 mRNA accumulation in non-stimulated and TNF-α-, IFN-γ-, or IL-1β-stimulated A549 cells were not obvious, probably due to the high baseline level of CCL2 expression, stimulation with the combination of TNF-α, IFN-γ, and IL-1β led to higher levels of CCL2 mRNA accumulation in a time-dependent fashion (Figure 1B). In contrast, CXCL10, CXCL11, and CXCL9 transcripts were not detected in non-stimulated A549 cells. As shown in Figure 1, resting A549 cells, as well as TNF-α- or IL-1β-treated cells, do not express detectable amounts of CXCL10 or CXCL9 mRNA. Although no detectable amount of CXCL11 transcripts was found in non-stimulated A549 cells, the stimulation with TNF-α, IL-1β or IFN-γ strongly induced CXCL11 mRNA expression (Figure 1A and 1C). IFN-γ alone induced mRNA expression of CXCL10, but not CXCL9, in a dose- and time-dependent manner (Figure 1A and 1B). A considerable accumulation of CXCL10 and CXCL9 mRNA was observed in A549 cells stimulated with IFN-γ plus, either IL-1β or TNF-α, with maximal expression levels being reached by 16 h for CXCL9 and by 24 h for CXCL10, respectively (Figure 1A and 1B). CXCL10, CXCL11, and CXCL9 transcripts were also highly increased by stimulation with combinations of IFN-γ, IL-1β, and TNF-α at different concentrations (Figure 1A and 1B). CXCL11 gene appears to be more sensitive on cytokine mediated induction than CXCL10 and CXCL9. The level of CXCL11 mRNA was increased within 8 h, and declined to baseline at 24 h in the presence of TNF-α or IL-1β in a time- and dose-dependent manner. IFN-γ clearly up-regulated the accumulation of CXCL11 mRNA at all concentrations tested (Figure 1C). Although kinetics of CXCL10, CXCL11, and CXCL9 mRNA expression in IFN-γ-stimulated A549 cells differed greatly from those of IFN-γ plus IL-1β plus TNF-α cells (as in the former conditions, CXCL11 and CXCL10 transcripts reached a maximum at 16 or 24 h, whereas in the latter relatively high levels of chemokine mRNA were detected at 4 or 8 h), it is evident that IFN-γ represents the most potent stimulus to induce mRNA expression of all three CXCR3 agonistic chemokines and that IL-1β and TNF-α exaggerate the up-regulatory effect of IFN-γ in A549 cell line (Figure 1B).
Figure 1 mRNA expression of CCL2 and CXCR3 agonistic chemokines by A549 cells. A : Dose response of proinflammatory cytokine-induced CCL2, CXCL10, CXCL11, and CXCL9 mRNA accumulation. The representative gel images from one out of three independent experiments are shown. Expression of β-actin in the same samples demonstrates equal loading of lanes. B : Densitometric analysis of the CCL2, CXCL10, CXCL11, and CXCL9 mRNA expression. RT-PCR was performed with total RNA obtained from A549 cells stimulated for the indicated times with 50 U/ml IFN-γ (IFN50), 50 U/ml IFN-γ and 5 ng/ml TNF-α (IFN50+TNF5) or 50 U/ml IL-1β (IFN50+IL50), and a combination of cytokines (CTMX) 50 U/ml of IFN-γ and IL-1β, and 5 ng/ml TNF-α. The mRNA-amplificates from each culture was quantitated individually. The distinct dots on the lines represent the mean percentages of β-actin density of duplicate determinations at each individual time-point for different concentrations/combinations of cytokines. Data are from one representative experiment out of three. C : Dose- and time-dependent effects of TNF-α, IFN-γ, and IL-1β at indicated concentrations on CXCL11 mRNA expression by A549 cells are shown.
Chemokine mRNA Expression by AEC-II in Primary Culture
Next we examined the effects of proinflammatory cytokines on the expression of chemokine genes expression by human AEC-II to determine whether a similar pattern of mRNA expression and induction as in A549 cells is also detectable in primary AEC-II. As experiments employing A549 cells showed that chemokine mRNA expression levels peaked 24 h after stimulation with proinflammatory cytokines, we used this time point to study the effect of different doses of TNF-α, IFN-γ, and IL-1β on CCL2, CXCL10, CXCL11, and CXCL9 mRNA accumulation in primary cultured AEC-II. We found that non-cultured AEC-II expressed detectable amounts of CCL2 mRNA, which were significantly increased by culture with or without cytokine stimulation (Figure 2, P < 0.01, n = 8). TNF-α and IL-1β slightly increased CCL2 mRNA accumulation in a dose-dependent fashion, but this was not statistically significant compared with non-stimulated cells (Figure 2, P > 0.05, n = 8). The maximum level of CCL2 mRNA expression was seen in cells stimulated with 10 U/ml of IFN-γ gradually decreasing to baseline values with increasing of IFN-γ concentration (Figure 2).
Figure 2 CCL2 mRNA expression by primary cultured AEC-II in response to proinflammatory cytokine stimulation for 24 h. Upper part of figure shows representative images of CCL2 mRNA amplificates in AEC-II derived from one of eight identical experiments. Expression of β-actin in the same samples demonstrates equal loading of lanes. Line 0 – 10 represent cells non-cultured, non-stimulated and stimulated with TNF-α, IFN-γ, or IL-1β, respectively. Line M indicates the molecular weight marker. The lower part of figure shows the results of densitometric analysis of the CCL2 mRNA expression. The mRNA-amplificates from each culture were quantitated individually. Values presented are the mean percentages of β-actin density ± SEM calculated from eight independent experiments. *P < 0.05 compared with non-cultured cells.
The CXCL mRNA expression pattern of primary AEC-II was similar to that of A549 cells, with some peculiarities in cytokine-stimulated cells. As shown in Figure 3, neither non-cultured nor non-stimulated AEC-II expressed detectable amount of CXCL9 mRNA in all experiments performed. In contrast to A549, CXCL9 mRNA was detected in AEC-II after TNF-α, IL-1β and, more strongly, after IFN-γ treatment. CXCL11 and CXCL10 mRNA were expressed in non-stimulated AEC-II after 4 h of culture, peaked at 16 h and slightly decreased thereafter (Figure 3 and not shown). However, both CXCL10 and CXCL11 were expressed at relatively higher levels after IL-1β, and especially, after IFN-γ treatment in a dose-dependent manner (Figure 3). We did not study the effects of different cytokine combinations on the chemokine mRNA expression patterns due to the strait in amounts of pure human AEC-II isolated from lung tissue samples. No changes in SP-A mRNA expression of non-stimulated or cytokine-stimulated AEC-II were detected after 24 h cultures (data not shown).
Figure 3 Effect of TNF-α, IFN-γ, and IL-1β stimulation at indicated concentrations on CXCL10, CXCL11, and CXCL9 mRNA expression by primary cultured AEC-II. One representative image of eight independent experiments for each chemokine is shown in the upper part of figure. Expression of β-actin in the same samples demonstrates equal loading of lanes. Line 0 – 10 represent cells non-cultured, non-stimulated and stimulated with TNF-α, IFN-γ or IL-1β respectively. Line M indicates the molecular marker. The lower part of figure shows the results of densitometric analysis of CXCL10, CXCL11, and CXCL9 mRNA expression in AEC-II isolated from one individual. Each panel shows the mean values of duplicate assays for each condition from one experiment representative of eight.
Production of CCL2 and CXCL10 by AEC-II in Primary Culture
Because CXCL10 mRNA was strongly up-regulated in AEC-II after cytokine stimulation and CCL2 mRNA was detected even in non-stimulated cells, we measured the concentrations of these chemokines in supernatants of AEC-II cultures in the presence or absence of proinflammatory cytokines. In accordance with mRNA expression patterns of CCL2 and CXCL10, AEC-II spontaneously release CCL2 at concentration of 12.7 ± 2.0 ng/ml/106 cells (Figure 4A), and no detectable amounts of CXCL10 were released by non-stimulated AEC-II after 24 h of cultures (Figure 4B). As shown in Figure 4A, treatment of the AEC-II with IL-1β caused a significant increase in the production of CCL2 (10 U/ml of IL-1β: 25.5 ± 4.2; 50 U/ml: 24.4 ± 3.6; 100 U/ml: 23.8 ± 4.4 ng/mL/106 cells respectively, P < 0.05, n = 12). TNF-α slightly increased the CCL2 release at concentrations of 1 ng/ml (18.4 ± 4.2 ng/ml/106 cells), 5 ng/ml (18.2 ± 3.8 ng/ml/106 cells) and 10 ng/ml (19.2 ± 4.4 ng/ml/106 cells), however, this effect was statistically significant only for a TNF-α concentration of 10 ng/ml (P < 0.05; Figure 4A). However, IFN-γ did not significantly change CCL2 protein levels in AEC-II cultures compared with non-stimulated controls (Figure 4A). In marked contrast, AEC-II generated ng/ml quantities of CXCL10 upon stimulation with IFN-γ after 24 h, but TNF-α and IL-1β exerted only marginal effects. As seen in Figure 4B, 10 U/ml of IFN-γ was sufficient to induce a significant increase in the CXCL10 generation by AEC-II being maximal at 100 U/ml of cytokine. A TNF-α concentration of 10 ng/ml slightly, but statistically significantly increased CXCL10 generation by AEC-II compared with non-stimulated controls (Figure 4B).
Figure 4 (A ) CCL2 and (B ) CXCL10 immunoreactivity in AEC-II supernatants after TNF-α, IFN-γ, and IL-1β stimulation at indicated concentrations for 24 h. Values presented are means ± SEM (n = 12). Statistically significant differences from non-stimulated cells assessed by ANOVA with PLSD separately for each cytokine are indicated by an asterisk (*P < 0.05).
Since IL-1β and IFN-γ disclosed the highest differences in the stimulatory capacity for CCL2 and CXCL10 releases in 24-h AEC-II cultures, the effects of both cytokines were analyzed in more detail. As shown in Figure 5A, IL-1β increased CCL2 release in a time- and dose-dependent manner. The IL-1β-induced increase in CCL2 production could be detected as early as 4 h after stimulation and significantly increased with time (Figure 5A). Just within the first 4 h of AEC-II cultures IFN-γ induced a modest CCL2 release, which did not differ statistically significantly from controls. Conversely, with the increase of culture time IFN-γ concentration-dependently decreased CCL2 release of AEC-II in a non-significant magnitude (Figure 5A). The dose- and time-dependent increases of IFN-γ and IL-1β on CXCL10 generation are shown in Figure 5B. A clear-cut dose- and time dependency as seen for stimulation with IFN-γ could also be observed for IL-1β. The low CXCL10 background release increased significantly in the presence of 50 or 100 U/ml IL-1β at time points 16 and 24 h. However, this increase is about 10-fold lower compared with the CXCL10 levels induced by IFN-γ at the same concentrations and time points (Figure 5A and 5B). Experiments with the cell line A549 demonstrated that non-stimulated cells generate significantly lower levels of immunoreactive CCL2 (P < 0.01; 2.3 ± 0.9 ng/ml/106 cells after 24 h, n = 6) compared with primary cultured AEC-II (data not shown). Additionally, TNF-α, but not IFN-γ or IL-1β, up-regulated CCL2 release, and this effect was only seen after 4 h of culture (data not shown). A549 cells also released CXCL10, and consistent with the mRNA data, a combination of IFN-γ with IL-1β and/or TNF-α significantly up-regulated CXCL10 release by these cells. IFN-γ/IL-1β/TNF-α-stimulated A549 cells generated 5.3 ± 1.9 ng/ml/106 cells (n = 3) of CXCL10 protein for 24 h, which was a 50-fold increase over IFN-γ-stimulated cells (data not shown).
Figure 5 Time course of spontaneous and cytokine-induced (A ) CCL2 and (B ) CXCL10 production by primary cultured AEC-II. IL-1β, but not IFN-γ increases CCL2 release, whereas both IL-1β and IFN- increase CXCL10 release in a time- and dose-dependent manner. The increase induced by IL-1β is about 10-fold lower compared with the CXCL10 levels induced by IFN-γ at the same concentrations and time points. The distinct dots on the lines represent means of triplicate determinations at each time-point for different concentrations of cytokine. *P < 0.05 compared to spontaneously produced chemokine levels by AEC-II. Data are representative results from three independent experiments.
CCL2 and CXCL10 mRNA expression by AEC-II in vivo
To determine if AEC-II expression of those chemokines can also be regulated in vivo , we took advantage of an in situ hybridization (ISH) method. ISH using DIG-labeled cDNA probes detected specific signals for CCL2 mRNA mainly in intra-alveolar macrophages in all lung tissue preparations included in the present study. Positive signals for CCL2 mRNA were also detected in AEC-II, which were typically localized at alveolar corners and exhibited cuboidal morphology (Figure 6A, arrowheads). After treatment with IL-1β almost all AEC-II displayed strong positive signal for CCL2 mRNA (Figure 6C, arrowheads). The same pattern of CCL2 mRNA expression was observed in both macrophages located in the alveolar lumen and those adjacent to alveolar epithelium (Figure 6C, inset, arrows). Interestingly, a weak positive signal was also detected in AEC type I (Figure 6C, sharp arrowheads). Dexamethasone treatment markedly inhibited IL-1β-induced CCL2 expression, but did not change basal levels compared to non-stimulated (data not shown) or non-cultured samples (Figure 6E and 6A). In contrast to CCL2, no positive signals for CXCL10 mRNA were detected in tissue explants from normal lungs (Figure 6B). However, after stimulation of whole lung tissue explants for 24 h in vitro with IL-1β and IFN-γ, in AEC-II (Figure 6D, inset, arrowheads) as well as in AM clear positive signals for CXCL10 mRNA could be detected (Figure 6D, arrows). Treatment with dexamethasone almost completely suppressed cytokine-induced CXCL10 mRNA in AEC-II and AM (Figure 6F, arrows). In situ hybridization was also performed on lung tissue preparations obtained from patients with pulmonary sarcoidosis and tuberculosis. The strong positive signals of CXCL10 mRNA were observed in AM and AEC-II on the perifocal zones of sarcoid granulomas (Figure 6G) and in the alveolar epithelium on tuberculous lung tissue preparations (Figure 6H). The specific signals were not detected in control preparations, in which specific DNA probes were substituted by hybridization buffer (not shown). For control purposes SP-A mRNA predominantly localized in AEC-II was detected in all lung tissue preparations (data not shown).
Figure 6 Localization of CCL2 and CXCL10 mRNA in lung tissue explants non-cultured and cultured with IL-1β and IFN-γ, in the presence or absence of 10-4 M dexamethasone (DEX) for 24 h. CCL2 mRNA is constitutively produced in normal lung tissue by AEC-II (arrowheads) and AM (arrows) (A ), and stimulation with 500 U/ml of human recombinant IL-1β and IFN-γ (C , inset) significantly up-regulates CCL2 mRNA expression by these cells compared to non-cultured (A ) or 24-h cultured controls (data not shown). AEC-I also show weak positive signal for CCL2 mRNA (C , sharp arrowheads). DEX treatment leads to inhibition of the IL-1β-induced CCL2 mRNA production in AEC-II (E and C ), but does not change the basal expression (E and A ). Panels B , D , and F show the same lung tissue explants treated as indicated above and hybridized with a CXCL10 specific probe. Panel D (inset) illustrates a prominent inducible effect of cytokine-stimulation on CXCL10 mRNA expression in situ by AEC-II (arrowheads) and AM (arrows), and panel F depicts the inhibitory effect of DEX on IFN-γ/IL-1β-induced CXCL10 mRNA expression. Data of one representative experiment from five are shown. Panels G and H illustrate CXCL10 mRNA localization in the lung from patients with pulmonary sarcoidosis and tuberculosis respectively. Control slides, in which hybridization buffer alone was applied, display no reactivity in all experiments (data not shown).
Discussion
To increase our knowledge in mechanisms controlling the recruitment and activation of inflammatory cells in the alveolar space and the role of alveolar epithelial cells type II in the cytokine network of the lung, we investigated the effects of proinflammatory cytokines on chemokine gene expression and production by human primary AEC-II. We examined CCL2, a CC chemokine that attracts predominantly monocytes/macrophages and activated T cells by binding to CCR2, and CXCL9, CXCL10, and CXCL11, T cell-specific chemokines binding to CXCR3. In this work, we demonstrate that CCL2 mRNA is present in freshly isolated AEC-II and its level is significantly up-regulated during culture. The proinflammatory cytokines IL-1β, TNF-α, and IFN-γ increased the accumulation of CCL2 mRNA in 24 h cultured AEC-II in a dose-dependent manner, however these effects were not statistically significant compared with non-stimulated cells. CCL2 mRNA patterns of resting and cytokine-stimulated A549 cells, which were used as control, disclosed the same expression profiles as observed in primary AEC-II. The highest CCL2 mRNA level was detected in A549 cells stimulated with a combination of IL-1β, TNF-α, and IFN-γ, and this effect was time-dependent.
In agreement with mRNA data, we also found that significant amounts of CCL2 protein were spontaneously secreted from primary cultured AEC-II, and IL-1β or TNF-α, but not IFN-γ, up-regulated CCL2 production, confirming a previous study [8]. In contrast to studies of IFN-γ effects on CCL2 release by human bronchial and endothelial cells and fibroblasts we could not demonstrate that IFN-γ up-regulates the CCL2 protein production by AEC-II and A549 cells [23-25]. Although, IFN-γ modulates CCL2 mRNA expression in AEC-II, time course experiments showed that IFN-γ does not significantly influence CCL2 release. These results are in line with other observations of IFN-γ being rather an inhibitor than promoter of spontaneous and LPS-induced tissue-specific CCL2 releases [26,27]. Interestingly, IFN-γ also selectively inhibits the CCR2 expression on human monocytes [28]. It seems that differences in IFN-γ regulation of CCL2 may result from differential responses of target cells to pro- and anti-inflammatory stimuli and to cell type-specific patterns of stimulus sensitivity. In contrast to IFN-γ, IL-1β was the most potent stimulus on CCL2 release by AEC-II. More than 25 ng/ml/106 cells of CCL2 were detected in supernatants collected from AEC-II, activated with IL-1β for 24 h. This is the highest level reported to date for cytokine-stimulated CCL2 secretion from airway epithelial cells. In comparison, in a 24-h period, human bronchial epithelial cells treated with IL-1β released 25-fold less CCL2 [23], and in our experiments, A549 cells maximally stimulated with TNF-α secreted 10-fold less protein. Furthermore, Sadek and associates demonstrated that human AM retrieved from BAL generate the same levels of CCL2 after 72-h stimulation with LPS in culture [29]. Thus, spontaneous CCL2 level produced by AEC-II is of biological importance since on a per cell basis it is three-fold higher than baseline CCL2 level generated by human AM for 72 h in vitro [29]. Although we did not determine the capacity of IL-1β-stimulated AEC-II to attract mononuclear cells, previous studies have shown that AEC-II-derived CCL2 is strongly chemotactic for CD14+ and CD3+ cells in vitro and in vivo [8,10,15].
Under normal condition the alveolar space contains a low number of leukocytes with AM forming about 95% of total cell population. As AEC-II are uniquely positioned in the borders between the microvascular compartment and the alveolar space, and constitutively generate considerable amounts of CCL2, one may hypothesize that this AEC-II-derived chemokine is responsible, at least partially, for basal recruitment of monocytes and their differentiation into AM, in healthy humans. In this respect Gunn and colleagues have demonstrated that CCL2 overexpression in the lung of transgenic mice leads to a marked increase in the number of BAL monocytes, but does not cause inflammatory activation of cells [15]. According with our in vitro and in vivo results, it is necessary to have additional inflammatory agonists such as macrophage-derived IL-1β and/or TNF-α for increasing and amplifying CCL2 expression by AEC-II, which in turn might be important factors for further development and manifestation of lung inflammation.
In contrast to CCL2, we could demonstrate that IFN-γ induces the expression of CXCL9, CXCL10, and CXCL11 by AEC-II. In fact, IFN-γ significantly up-regulated CXCL9, CXCL10, and CXCL11 mRNA accumulation and CXCL10 production of AEC-II and A549 cells. Albeit at low levels, CXCL10, CXCL11, and CXCL9 were also induced directly by IL-1β and TNF-α in primary cultured AEC-II. Interestingly, the kinetics of CXCL11 mRNA expression in IFN-γ- or IFN-γ plus TNF-α and IL-1β-treated A549 cells were faster than those of CXCL9 or CXCL10 mRNA expression. Furthermore, the effect of TNF-α and IL-1β was more pronounced on CXCL11 mRNA accumulation in A549 cells compared with CXCL10 or CXCL9, suggesting that different pathways might be involved in CXCL9 and CXCL10 expression at one side, and CXCL11 on the other side, even though they are all induced by IFN-γ. The significance of a sequential regulated expression of CXCL10, CXCL11, and CXCL9 by AEC-II is only a matter of speculation, but a similar pattern has been found in other cell types, for instance, in endothelial cell, bronchial epithelial cells, and neutrophils [30-32]. CXCL10 mRNA accumulation in primary cultured AEC-II and in the control A549 cells closely reflected levels of secreted CXCL10 protein, as previously shown in human bronchial epithelial cells [31], suggesting that the CXCL10 production is strongly regulated and dependent on the transcriptional mechanisms.
To date, a number of cellular sources of CCL2 and CXCL10 in the inflamed lung have been identified, including macrophages, fibroblasts, airway epithelium, and endothelial cells. In the present study, we have evaluated the ability of human AEC-II to express and produce CCL2 and CXCL10 in vitro , and express CCL2 and CXCL10 mRNA in situ . ISH studies revealed intense positive signals for CXCL10 mRNA in AEC-II, as well as in interstitial and alveolar macrophages in lung tissue explants after in vitro treatment with IL-1β and/or IFN-γ. In contrast, no positive signal was detected in non-cultured or cultured lung tissue explants with medium alone or with IFN-γ/IL-1β and 10-4 M dexamethasone. On the other hand, CCL2 mRNA expression was detected in AEC-II and AM in all lung tissue explants, and IL-1β, alone or in combination with IFN-γ, markedly up-regulated CCL2 mRNA expression of AEC-II and AM in situ . Treatment with dexamethasone attenuated the signal intensities in cytokine-stimulated, but not in non-stimulated preparations. These results from in vitro stimulation and dexamethasone inhibition experiments demonstrated that AEC-II were capable of expressing significant quantities of CXCL10 mRNA in situ only under local IFN-γ or IL-1β activation, in contrast to CCL2, which is expressed constitutively, and proinflammatory cytokines only up-modulated the steady state mRNA levels of CCL2. These data corroborate well observations from van der Velden et al. and Witowski et al., which have shown that dexamethasone and actinomycin D strongly inhibit cytokine-driven but not constitutive CCL2 release in human bronchial epithelium and peritoneal fibroblasts [23,33].
Our in situ hybridization data are consistent with observations from Jaffe and associates, and Martin et al., which have shown that only after local, but not after systemic administration of recombinant human IFN-γ in healthy volunteers, CXCL10 mRNA was induced in AM [34,35]. Moreover, production of CXCL10 and CXCL9 in the injured lung was not completely suppressed in mice deficient for IFN-γ or the IFN-γ receptor [36], and CXCL10, but not CCL2, failed in the human T cell transendothelial migration model [37]. All these data suggest that in contrast to CCL2, CXCL10 is directly induced by IFN-γ released from cells within the lung rather than by IFN-γ derived from distant sites, and that alternate agonists are present in the alveolar compartment, which may together with IFN-γ or separately amplify CXCL10 expression, and subsequently promote local accumulation of CXCR3+ cells in the alveolar space. In addition, ISH data strongly suggest that effects of the proinflammatory cytokines on CCL2 and CXCL10 mRNA accumulation and protein generation in primary cultured AEC-II were not due to artifacts elicited by cell isolation procedure or culture conditions.
The quantitative differences, time and cytokine-inducing profile dependencies in CCL2, CXCL9, CXCL10, and CXCL11 expression of AEC-II suggest that there are several cytokine/chemokine cascades within the injured lung, which in turn, determine the flexible programs of recruitment and activation of inflammatory cells into the alveolar space. For instance, alveolar macrophage-derived pro-inflammatory cytokines such as IL-1 and TNF-α, but not IFN-γ, directly up-regulate expression by AEC-II of certain chemokines including CCL2 and CXCL8 [8,10,19,38]. In our previous report we demonstrated that in close similarity with CCL2, TNF-α and IL-1β, but not IFN-γ, up-modulated constitutive interleukin-8/IL-8 (CXCL8) release in primary cultured AEC-II, and cytokine-derived increase of CXCL8 basal level was only two-fold lower than LPS-induced CXCL8 release in AM [19]. Our results demonstrating high levels of CCL2, CXCL8 and CXCL10 production by AEC-II in vitro indicate that AEC-II have the potential to participate in physiologic and pathologic macrophage, neutrophil and T cell responses within the alveolar space. By chemoattracting IL-1-producing monocytes/macrophages, TNF-α-producing neutrophils and IFN-γ-producing CD4+ or CD8+ cells in close proximity to the epithelium, AEC-II-derived CCL2, CXCL8, CXCL9, CXCL10, and CXCL11, which themselves are up-regulated by IL-1β and/or TNF-α, and IFN-γ, may activate several positive feedback loops. In addition, it has been reported that CXCL10 selectively activated and enhanced antigen-driven IFN-γ gene expression in T cells [39]. Thus, it is tempting to speculate that locally produced CCL2 and CXCL10 by AEC-II attract activated memory T cells into alveoli and further amplify antigen-driven IFN-γ response of Th1 cells. The strong confirmation of this hypothesis results from animal models of the lung inflammation and some clinical observations in pulmonary diseases. It has been demonstrated that bronchoalveolar lavage (BAL) cells disclosed the Th1-dominated pattern in transgenic rats that express CXCL10 in the lung, and BAL lymphocytes of HIV-infected patients with T-cell alveolitis were CD8+ T cells expressing high levels of CXCR3 and IFN-γ, which exhibited a high migratory capability in response to CXCL10 and CXCL9 [40,41]. Similarly, over-expression of human CCL2 in transgenic AEC-II or intra-tracheal administration of murine MCP-1 caused a substantial accumulation of activated monocytes within the bronchoalveolar space of mice [15,16]. Additionally, Miotto et al. reported that CXCL10 expression was strongly up-regulated in Th1-mediated lung diseases, whereas increased CCL2 expression was not specifically associated with Th1 or Th2 patterns[3]. Our in situ hybridization data confirm that in contrast with normal lung tissue preparations a strong expression of CXCL10 can be detected in sarcoidosis and tuberculosis – diseases in which the Th1-type cytokine IFN-γ is up-regulated.
Beside the pro-inflammatory effects of CXCL9, CXCL10, and CXCL11 they also exhibit down-regulation e.g. on the migration of eosinophils [42] or on angiogenesis [43]. This is of interest, because it could be shown in a mouse model that CXCL11 attenuates fibrosis by inhibition of vascular remodeling [44]. In addition, it could be demonstrated that CXCL9 down-regulates IL-4 expression but up-regulates IFNγ expression by T cells. This illustrates that AEC-II not only induce the migration of Th1 cells by the release of CXCR3-ligands but they also participate in T cell activation [20] and Th1 polarization of lung T cells.
Since it is known that AEC-II also express CXCR3A and CXCR3B [45] it is tempting to speculate about a possible activity of CXCL9, CXCL10, and CXCL11 on AEC-II in an autocrine fashion, e.g. in re-epithelialization after lung injury. Therefore, induction of CXCR3 agonists may in some respects of benefit and may represent a possible therapeutic mechanism of IFNγ therapy of IPF patients [46].
Conclusion
Taken together our data indicate that although AEC-II constitute about 15% of all lung cells, they may play a prominent role in the pathogenesis of inflammatory lung diseases. This study suggests that in human AEC-II, CXCR3 agonistic chemokines are expressed and regulated differently from CCL2. This pattern of differential expression and regulation was also seen when we explored the effect proinflammatory cytokines on the IL-8/CXCL8 expression by human AEC-II [19]. We proposed that CXCL8 and CCL2 constitutively produced by AEC-II may mediate the „steady state" recruitment of blood cells, i.e. monocytes/macrophages and neutrophils, which provide a first line of the host defense in the lung (Figure 7A). Bacteria, endotoxin, viruses, and other pathogens may enhance the production of CXCL8 and CCL2 by AEC-II directly or with monocyte/macrophage derived TNF-α and IL-1β as an intermediary. Cytokine-mediated increases of CCL2 or CXCL8 levels may therefore be considered as an alert signal for leukocytes in the pulmonary vascular bed and alveolar space (Figure 7B). The increased numbers of leukocytes, including activated T cells, and the presence of specific antigen can further elicit the development of specific immune response within alveoli. Subsequently, activated T cells (Th1) released high amounts of IFN-γ resulting in additional CXCL9, CXCL10, or CXCL11 induction by resident cells (AEC-II) of the alveoli. This positive feedback loop establishes a new chemotactic gradient inducing an increased migration of CXCR3+ T cells from the blood stream through the endothelium into both the interstitium and alveolar space (Figure 7C). Under certain pathologic circumstances, several auto-regulatory loops involving alveolar epithelium might be exaggerated, and may contribute to acute and chronic inflammatory injuries of the lung.
Figure 7 Alveolar epithelial cells type II drive a leukocyte trafficking in host defense and specific immune response. The schematic drawing shows spatial distribution of chemokine expression and leukocyte recruitment: in healthy condition (A ) two chemokines, MCP-1/CCL2 and IL-8/CXCL8, are expressed constitutively by AEC-II and provided the physiological influx of monocytes (Mn), neutrophils (PMN), and lymphocytes (Ly) into alveolar space from pulmonary vasculature (VSL); in the initial phase of alveolar inflammation caused by inflammatory agents (B ) increasing alveolar macrophage (AM) and particularly epithelial expression of CCL2 and CXCL8 as well as the expression of CXCL10 and CXCL11 is paralleled by pronounced recruitment of Mn, PMN, and Ly to the alveolar compartment; further development of the immune response (C ) leads to the expression of high levels of CXCL9, CXCL10, and CXCL11 in AEC-II induced by IFN-γ and/or TNF-α, and IL-1β and associated with recruitment of activated T cells (Th1) to the interstitium and alveolar space. Mn and PMN reside in the interstitium and vascular compartment, which may be because of an IFN-γ inhibition of CCL2 and CXCL8 expression by AEC-II.
List of abbreviations
AEC-II – alveolar epithelial cells type II, BAL – bronchoalveolar lavage, CCL2 – CC-ligand 2 (= MCP-1 (monocyte chemoattractant protein-1)), CXCL9 – CXC-ligand 9 (= MIG (monokine induced by interferon-γ)), CXCL10 – CXC-ligand 10 (= IP-10 (Interferon-inducible protein-10)), CXCL11 CXC-ligand 11 (= I-TAC (Interferon-inducible T-cell alpha chemoattractant)), IFN-γ – interferon-γ, IL-1β-interleukin-1β, ISH – in situ hybridization, SPA – surfactant protein A, TNF-α – tumor necrosis factor α,
Authors' contributions
DVP isolated the cells, performed the cell culture and PCR, developed specific vectors for ISH and drafted the manuscript. TG performed the ISH, EV conducted the pathological part of the study, CL, AP JMQ performed the clinical part and were involved in the coordination of the study. GZ conceived the study, carried out the ELISAs and was involved in preparing the manuscript.
Acknowledgements
The authors would like to thank Dr. E. Richter, National Reference Center for Mycobacteria, for sequencing of the PCR-products, S. Adam, D. Bubritzki, H. Kühl, and N. Husmann for technical assistance.
This work was supported by grant from the Deutsche Forschungsgemeinschaft (No. Mu 692/5-5). D.V.P. is the recipient of a Research Fellowship from the Alexander von Humboldt Foundation.
==== Refs
Bergeron A Bonay M Kambouchner M Lecossier D Riquet M Soler P Hance A Tazi A Cytokine patterns in tuberculous and sarcoid granulomas: correlations with histopathologic features of the granulomatous response J Immunol 1997 159 3034 3043 9300729
Robinson DS Ying S Taylor IK Wangoo A Mitchell DM Kay AB Hamid Q Shaw RJ Evidence for a Th1-like bronchoalveolar T-cell subset and predominance of interferon-gamma gene activation in pulmonary tuberculosis Am J Respir Crit Care Med 1994 149 989 993 8143065
Miotto D Christodoulopoulos P Olivenstein R Taha R Cameron L Tsicopoulos A Tonnel AB Fahy O Lafitte JJ Luster AD Wallaert B Mapp CE Hamid Q Expression of IFN-gamma-inducible protein; monocyte chemotactic proteins 1, 3, and 4; and eotaxin in TH1- and TH2-mediated lung diseases J Allergy Clin Immunol 2001 107 664 670 11295656 10.1067/mai.2001.113524
Walker C Bauer W Braun RK Menz G Braun P Schwarz F Hansel TT Villiger B Activated T cells and cytokines in bronchoalveolar lavages from patients with various lung diseases associated with eosinophilia Am J Respir Crit Care Med 1994 150 1038 1048 7921434
Yamasaki H Ando M Brazer W Center DM Cruikshank WW Polarized type 1 cytokine profile in bronchoalveolar lavage T cells of patients with hypersensitivity pneumonitis J Immunol 1999 163 3516 3523 10477626
Mason RJ Williams MC Crystal RG and West JB Alveolar Type II Cells The Lung, scientific foundations 1991 New York., Raven Press 235 246
Barrett EG Johnston C Oberdorster G Finkelstein JN Silica-induced chemokine expression in alveolar type II cells is mediated by TNF-alpha-induced oxidant stress Am J Physiol 1999 276 L979 88 10362723
Eghtesad M Jackson HE Cunningham AC Primary human alveolar epithelial cells can elicit the transendothelial migration of CD14+ monocytes and CD3+ lymphocytes Immunology 2001 102 157 164 11260320 10.1046/j.1365-2567.2001.01172.x
Koyama S Sato E Nomura H Kubo K Nagai S Izumi T Type II pneumocytes release chemoattractant activity for monocytes constitutively Am J Physiol 1997 272 L830 7 9176245
Paine R Rolfe MW Standiford TJ Burdick MD Rollins BJ Strieter RM MCP-1 expression by rat type II alveolar epithelial cells in primary culture J Immunol 1993 150 4561 4570 8482848
Rossi D Zlotnik A The biology of chemokines and their receptors Annu Rev Immunol 2000 18 217 242 10837058 10.1146/annurev.immunol.18.1.217
Sallusto F Mackay CR Lanzavecchia A The role of chemokine receptors in primary, effector, and memory immune responses Annu Rev Immunol 2000 18 593 620 10837070 10.1146/annurev.immunol.18.1.593
Boring L Gosling J Chensue SW Kunkel SL Farese RVJ Broxmeyer HE Charo IF Impaired monocyte migration and reduced type 1 (Th1) cytokine responses in C-C chemokine receptor 2 knockout mice J Clin Invest 1997 100 2552 2561 9366570
Carr MW Roth SJ Luther E Rose SS Springer TA Monocyte chemoattractant protein 1 acts as a T-lymphocyte chemoattractant Proc Natl Acad Sci U S A 1994 91 3652 3656 8170963
Gunn MD Nelken NA Liao X Williams LT Monocyte chemoattractant protein-1 is sufficient for the chemotaxis of monocytes and lymphocytes in transgenic mice but requires an additional stimulus for inflammatory activation J Immunol 1997 158 376 383 8977213
Maus U Herold S Muth H Maus R Ermert L Ermert M Weissmann N Rosseau S Seeger W Grimminger F Lohmeyer J Monocytes recruited into the alveolar air space of mice show a monocytic phenotype but upregulate CD14 Am J Physiol Lung Cell Mol Physiol 2001 280 L58 68 11133495
Cole KE Strick CA Paradis TJ Ogborne KT Loetscher M Gladue RP Lin W Boyd JG Moser B Wood DE Sahagan BG Neote K Interferon-inducible T cell alpha chemoattractant (I-TAC): a novel non- ELR CXC chemokine with potent activity on activated T cells through selective high affinity binding to CXCR3 J Exp Med 1998 187 2009 2021 9625760 10.1084/jem.187.12.2009
Farber JM Mig and IP-10: CXC chemokines that target lymphocytes J Leukoc Biol 1997 61 246 257 9060447
Pechkovsky DV Zissel G Ziegenhagen MW Einhaus M Taube C Rabe KF Magnussen H Papadopoulos T Schlaak M Muller-Quernheim J Effect of proinflammatory cytokines on interleukin-8 mRNA expression and protein production by isolated human alveolar epithelial cells type II in primary culture Eur Cytokine Netw 2000 11 618 625 11125305
Zissel G Ernst M Rabe K Papadopoulos T Magnussen H Schlaak M Müller-Quernheim J Human alveolar epithelial cells type II are capable of regulating T cell activity J Investig Med 2000 48 66 75 10695271
Murray PI Clay CD Mappin C Salmon M Molecular analysis of resolving immune responses in uveitis Clin Exp Immunol 1999 117 455 461 10469047 10.1046/j.1365-2249.1999.00993.x
Goldmann T Wiedorn KH Kuhl H Olert J Branscheid D Pechkovsky D Zissel G Galle J Muller-Quernheim J Vollmer E Assessment of transcriptional gene activity in situ by application of HOPE-fixed, paraffin-embedded tissues Pathol Res Pract 2002 198 91 95 11928870
van der Velden VH Verheggen MM Bernasconi S Sozzani S Naber BA van der Linden-van Beurden CA Hoogsteden HC Mantovani A Versnel M Interleukin-1beta and interferon-gamma differentially regulate release of monocyte chemotactic protein-1 and interleukin-8 by human bronchial epithelial cells Eur Cytokine Netw 1998 9 269 277 9831176
Brown Z Gerritsen ME Carley WW Strieter RM Kunkel SL Westwick J Chemokine gene expression and secretion by cytokine-activated human microvascular endothelial cells. Differential regulation of monocyte chemoattractant protein-1 and interleukin-8 in response to interferon-gamma Am J Pathol 1994 145 913 921 7943180
Struyf S Van Collie E Paemen L Put W Lenaerts JP Proost P Opdenakker G Van Damme J Synergistic induction of MCP-1 and -2 by IL-1beta and interferons in fibroblasts and epithelial cells J Leukoc Biol 1998 63 364 372 9500525
Oh JW Schwiebert LM Benveniste EN Cytokine regulation of CC and CXC chemokine expression by human astrocytes J Neurovirol 1999 5 82 94 10190694
Kopydlowski KM Salkowski CA Cody MJ van Rooijen N Major J Hamilton TA Vogel SN Regulation of macrophage chemokine expression by lipopolysaccharide in vitro and in vivo J Immunol 1999 163 1537 1544 10415057
Penton-Rol G Polentarutti N Luini W Borsatti A Mancinelli R Sica A Sozzani S Mantovani A Selective inhibition of expression of the chemokine receptor CCR2 in human monocytes by IFN-gamma J Immunol 1998 160 3869 3873 9558092
Sadek MI Sada E Toossi Z Schwander SK Rich EA Chemokines induced by infection of mononuclear phagocytes with mycobacteria and present in lung alveoli during active pulmonary tuberculosis Am J Respir Cell Mol Biol 1998 19 513 521 9730880
Ebnet K Simon MM Shaw S Regulation of chemokine gene expression in human endothelial cells by proinflammatory cytokines and Borrelia burgdorferi Ann N Y Acad Sci 1996 797 107 117 8993355
Sauty A Dziejman M Taha RA Iarossi AS Neote K Garcia-Zepeda EA Hamid Q Luster AD The T cell-specific CXC chemokines IP-10, Mig, and I-TAC are expressed by activated human bronchial epithelial cells J Immunol 1999 162 3549 3558 10092813
Gasperini S Marchi M Calzetti F Laudanna C Vicentini L Olsen H Murphy M Liao F Farber J Cassatella MA Gene expression and production of the monokine induced by IFN-gamma (MIG), IFN-inducible T cell alpha chemoattractant (I-TAC), and IFN-gamma-inducible protein-10 (IP-10) chemokines by human neutrophils J Immunol 1999 162 4928 4937 10202039
Witowski J Thiel A Dechend R Dunkel K Fouquet N Bender TO Langrehr JM Gahl GM Frei U Jorres A Synthesis of C-X-C and C-C chemokines by human peritoneal fibroblasts: induction by macrophage-derived cytokines Am J Pathol 2001 158 1441 1450 11290562
Jaffe HA Buhl R Mastrangeli A Holroyd KJ Saltini C Czerski D Jaffe HS Kramer S Sherwin S Crystal RG Organ specific cytokine therapy. Local activation of mononuclear phagocytes by delivery of an aerosol of recombinant interferon-gamma to the human lung J Clin Invest 1991 88 297 302 1905329
Martin RJ Boguniewicz M Henson JE Celniker AC Williams M Giorno RC Leung DY The effects of inhaled interferon gamma in normal human airways Am Rev Respir Dis 1993 148 1677 1682 8256919
Neumann B Emmanuilidis K Stadler M Holzmann B Distinct functions of interferon-gamma for chemokine expression in models of acute lung inflammation Immunology 1998 95 512 521 9893039 10.1046/j.1365-2567.1998.00643.x
Roth SJ Carr MW Springer TA C-C chemokines, but not the C-X-C chemokines interleukin-8 and interferon-gamma inducible protein-10, stimulate transendothelial chemotaxis of T lymphocytes Eur J Immunol 1995 25 3482 3488 8566041
O'Brien AD Standiford TJ Christensen PJ Wilcoxen SE Paine R Chemotaxis of alveolar macrophages in response to signals derived from alveolar epithelial cells [see comments] J Lab Clin Med 1998 131 417 424 9605106 10.1016/S0022-2143(98)90142-1
Gangur V Simons FE Hayglass KT Human IP-10 selectively promotes dominance of polyclonally activated and environmental antigen-driven IFN-gamma over IL-4 responses Faseb J 1998 12 705 713 9619449
Palmer K Emtage PC Strieter RM Gauldie J Transient gene transfer of non-ELR chemokines to rodent lung induces mononuclear cell accumulation and activation J Interferon Cytokine Res 1999 19 1381 1390 10638707 10.1089/107999099312849
Agostini C Facco M Siviero M Carollo D Galvan S Cattelan AM Zambello R Trentin L Semenzato G CXC chemokines IP-10 and mig expression and direct migration of pulmonary CD8+/CXCR3+ T cells in the lungs of patients with HIV infection and T-cell alveolitis Am J Respir Crit Care Med 2000 162 1466 1473 11029363
Fulkerson PC Zimmermann N Brandt EB Muntel EE Doepker MP Kavanaugh JL Mishra A Witte DP Zhang H Farber JM Yang M Foster PS Rothenberg ME Negative regulation of eosinophil recruitment to the lung by the chemokine monokine induced by IFN-gamma (Mig, CXCL9) Proc Natl Acad Sci U S A 2004 101 1987 92. Epub 2004 Feb 09. 14769916 10.1073/pnas.0308544100
Belperio JA Keane MP Arenberg DA Addison CL Ehlert JE Burdick MD Strieter RM CXC chemokines in angiogenesis J Leukoc Biol 2000 68 1 8 10914483
Burdick MD Murray LA Keane MP Xue YY Zisman DA Belperio JA Strieter RM CXCL11 Attenuates Bleomycin-induced Pulmonary Fibrosis via Inhibition of Vascular Remodeling Am J Respir Crit Care Med 2005 171 261 8. Epub 2004 Oct 22. 15502109 10.1164/rccm.200409-1164OC
Kelsen SG Aksoy MO Yang Y Shahabuddin S Litvin J Safadi F Rogers TJ The chemokine receptor CXCR3 and its splice variant are expressed in human airway epithelial cells Am J Physiol Lung Cell Mol Physiol 2004 287 L584 91. Epub 2004 May 21. 15155273 10.1152/ajplung.00453.2003
Strieter RM Starko KM Enelow RI Noth I Valentine VG Effects of interferon-gamma 1b on biomarker expression in patients with idiopathic pulmonary fibrosis Am J Respir Crit Care Med 2004 170 133 40. Epub 2004 Mar 24. 15044205 10.1164/rccm.200312-1670OC
|
16033640
|
PMC1185567
|
CC BY
|
2021-01-04 16:23:25
|
no
|
Respir Res. 2005 Jul 20; 6(1):75
|
utf-8
|
Respir Res
| 2,005 |
10.1186/1465-9921-6-75
|
oa_comm
|
==== Front
Theor Biol Med ModelTheoretical Biology & Medical Modelling1742-4682BioMed Central London 1742-4682-2-251602661610.1186/1742-4682-2-25ResearchIdentification and isolation of embryonic stem cells in reproductive endocrinology: theoretical protocols for conservation of human embryos derived from in vitro fertilization Sills Eric Scott [email protected] Takumi [email protected] Noriko [email protected] Queenie V [email protected] Gianpiero D [email protected] Georgia Reproductive Specialists LLC, Division of Reproductive Endocrinology and Infertility, Department of Obstetrics and Gynecology, Atlanta Medical Center; Atlanta, Georgia 30342 USA2 Cornell Center for Reproductive Medicine and Infertility, Weill Medical College of Cornell University, New York, New York 10021 USA3 HT-336, 505 East 70th Street, New York, New York 10021 USA2005 18 7 2005 2 25 25 1 4 2005 18 7 2005 Copyright © 2005 Sills et al; licensee BioMed Central Ltd.2005Sills 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
Embryonic stem cells (ESC) are pluripotent cells obtained from the inner cell mass (ICM) of blastocysts derived from in vitro culture associated with reproductive endocrinology therapy. Human ESCs are regarded as highly significant since they retain the capacity to differentiate into any of approximately 200 unique cell types. Human ESC research is controversial because to acquire such cells, the ICM of human blastocysts must be manipulated in a way that renders embryos nonviable and unsuitable for transfer in utero. Techniques to yield competent ESCs with conservation of source blastocysts would satisfy many objections against ESC research, but at present such approaches remain largely untested.
Results and discussion
We contrast experimental culture of single blastomeres obtained by 1) non-destructive biopsy of embryos destined for transfer, and 2) isolation of karyotypically normal blastomeres from disaggregated ("dead") embryos considered unsuitable for transfer, and evaluate these approaches with regard to production of ESCs. Pluripotency was confirmed by morphological criteria and by quantification of divergent homeodomain proteins specific to undifferentiated cell development. Following ESC isolation and identification, assessment was conducted according to a novel ESC grading system, also proposed here.
Conclusion
The role of reproductive endocrinology in ESC research remains paramount. In this report, we hypothesize new and expand on existing strategies having the potential to enhance human ESC isolation, identification and in vitro maintenance.
==== Body
Background
While the definitive characterization of murine embryonic stem cells was first reported in 1981, embryonic stem cells (ESC) were not isolated and fully described in humans until much later [1]. Without question, the scarce supply of human ESCs combined with the technical challenges associated with interspecies translation of stem cell derivation contributed to the long interval between these reports. To obtain human ESCs, embryos produced during in vitro fertilization (IVF) are maintained in extended culture to the blastocyst stage (4–5d post fertilization) when the polarized inner cell mass (ICM) develops. The outer trophoectoderm is removed via immunosurgery, thus exposing the ICM for disaggregation and plating on a feeder cell layer for further culture. Importantly, this disruptive process renders the embryo non-viable and unsuitable for in utero transfer [2]. Homogenous human ESC colonies may be derived from subsequent isolation and re-plating of the ICM cells, which are then screened for stemness by a variety of recognized markers.
Once in stable culture, ESCs are capable of either symmetric (clonogenic) or asymmetric fission. Symmetric ESC division yields a self-renewing supply of pluripotent ESCs, while asymmetric division produces one cell identical to the parent ESC plus one differentiated cell. The mechanism(s) responsible for modulating these specific ESC fission patterns remain poorly understood. In any case, since it is not yet possible to de-differentiate committed somatic cells to reacquire pluripotency, embryos associated with IVF have thus far been the only source for human ESCs. The fact that live human embryos must be destroyed to produce human ESCs presents a substantial ethical obstacle for the advancement of human ESC research. With the vast therapeutic promise of human ESC seen against the destruction of human embryos required to realize such aspirations, compelling arguments have been articulated both in support of and in opposition to human ESC research [3,4].
It must be admitted that thus far human ESCs have provided no reproducible, safe, unique and previously unattainable treatment for any human disease. Nevertheless, interest in exploration of the full therapeutic possibility of human ESCs continues to grow and is no longer confined to medical scientists and reproductive endocrinologists – indeed, it now includes public opinion leaders and medical consumers as well [5]. Yet the concerns of human ESC research opponents are not without ethical justification; these objections could be substantially assuaged if safe and effective laboratory protocols could be developed that offered human ESCs whilst preserving (or at least not destroying) the blastocysts from which they originated. In this paper we present results from pilot studies based on some theoretical approaches, in a manner to facilitate human ESC research and to promote respect for human embryos obtained from clinical reproductive endocrinology practice.
Human embryonic stem cells: theoretical approaches
Blastomere biopsy and culture
Prior to the blastocyst stage, a human embryo at 2–3d post fertilization consists of just 4–8 cells, all of which are totipotent. In contrast to the pluripotent cells obtained from the blastocyst ICM, one or two of these blastomeres may be biopsied without compromising the integrity of the sampled embryo [6]. Blastomeres obtained for PGD are generally fixed and processed with fluorochromes to detect aneuploidy by partial karyotype analysis, although the process has more recently advanced to testing for single gene disorders via polymerase chain reaction [7] and single cell whole genome amplification by multiple displacement amplification [8]. Such processing irrevocably alters the blastomere destined for PGD – the viability of this cell is sacrificed in return for the vital genomic information provided by PGD. However, assuming two distinct blastomeres were extracted at a well-timed embryo biopsy for PGD, and since in the absence of mosaicism each blastomere retains the potential to develop into a complete organism [9], the possibility exists that at least one sampled blastomere obtained for PGD could be maintained in culture specifically for ESC production. As with traditional PGD protocols, genetic data needed from PGD could still be obtained and inform embryo transfer decisions, while the second blastomere could provide a potential source of human ESCs with no measurable adverse affect on the developmental integrity of the biopsied embryo.
Utilizing a murine embryo model, we evaluated this concept where two blastomeres were isolated from a single 8-cell embryo via standard microsurgical biopsy techniques [10]. Zona-free murine blastomeres were then washed, individually plated and cultured as previously described [11]; embryos from which the biopsies were taken were maintained in standard culture (control group). All cells were monitored × 12 h to assess cleavage, differentiation, and attachment to the feeder cell monolayer, as applicable (Figure 1 and 2). With proper culture conditions we observed advancement to morphologically normal blastocyst stage in both groups. Next, cells resembling an ICM that originated from the intact/source embryo group and the single blastomere culture group were disaggregated from their respective blastocysts and re-plated on to fresh feeder cells for confirmation and further analysis; no cells were frozen. This work carries forward a theoretical approach suggested more than a decade ago [12], and demonstrates that a blastomere biopsy and culture approach can supply a single totipotent cell for subsequent ESC culture without harming the source embryo.
Blastomere donation from non-viable ("dead") embryos
Human embryo assessment plays a central role in IVF to identify embryos with the best prognosis for transfer, but what is less clear is the fate of embryos judged not suitable for transfer or cryopreservation due to arrested growth or gross developmental abnormality. Despite the absence of formal guidelines governing human embryology practice, many IVF centers carefully monitor embryos over several days before making the determination that they should be neither transferred nor cryopreserved based on non-viability. Indeed, even with cryopreservation as late as post-fertilization day 7, human livebirths have been achieved [13]. However, as previous investigators have noted [2,14], a consensus definition of embryo non-viability or death remains elusive and it is reasonable to expect that the concept of embryo death will formalize gradually in a process similar to that which led to the 1981 Uniform Determination of Death Act [15]. In the meantime, most major IVF clinics already obtain written informed consent from patients to discard any human embryos deemed non-viable or dead.
Interestingly, IVF laboratories have confronted this challenge and produced an informal if not exactly uniform process to declare a human embryo "dead". Since the life of any developing organism is more than the sum of its cellular parts, it has been suggested that the defining vital characteristics of a 4- or 8-cell human embryo must include continued and integrated cellular division, growth, and differentiation [16]. And by extension, embryos that have irreversibly lost this basic capacity (even if individual constituent cells may remain alive) should be properly regarded as organismically dead. Therefore our investigations were based on assessment of fresh (non-cryopreserved) 4–8 cell embryos demonstrating developmental arrest observed over an 8-day in vitro culture interval. Among such non-viable embryos destined for discard, a high rate of chromosomal error has been found in some, but not all, blastomeres [17]. It is the salvage of any normal blastomeres within a "dead" embryo that holds particular promise for human ESC research. Specifically, if embryos classified as non-viable and unsuitable for transfer or cryopreservation were disaggregated (rather than discarded) and plated as single totipotent blastomeres as described above, then the possibility exists that at least some karyotypically normal cell colonies could develop and serve as a reliable human ESC source. While the attempt to produce blastocysts from isolated blastomeres in vitro is not new [18], we feel this approach has received limited attention and merits further exploration to advance human ESC research, particularly since this source of ESCs would not derive from human embryos otherwise destined for transfer or cryopreservation.
We investigated the efficacy of a novel methodology with murine embryos that failed to meet viability standards, and were therefore unsuitable for transfer or cryopreservation. Embryos used in this pilot study displayed arrested growth and were classified as nonviable no later than the 8-cell stage. Embryos were disaggregated into single blastomeres by brief exposure to trypsin under micromanipulation control. Next, blastomeres were individually plated on a feeder cell layer and cultured in an experimental medium supplemented with β-mercaptoethanol, amino acids, nucleosides, antibiotics, L-glutamine with 2000 IU/ml mouse recombinant leukemia inhibiting factor in 6% CO2 at 37°C. Fully-expanded or hatching mouse blastocysts were plated as controls. The salvaged blastomeres and normal blastocysts were monitored daily to evaluate differentiation, cleavage and attachment to the feeder cell layer. Although some blastomeres obtained from the dead embryos failed to progress, a few ICM-like clusters developed from single blastomeres. These were isolated (as were ICMs derived from the intact blastocysts) and disaggregated into single cells by trypsinization and replated on to fresh feeder cell layers. These ESC lines were assessed for pluripotency by morphological criteria as well as alkaline phosphatase activity, Oct-4, and TROMA-1 [19], which validated stemness in this experiment.
Impact of mosaicism on ESC derivation
Soon after the first clinical experience with preimplantation genetic diagnosis was reported [20], it was suggested that blastomere mosaicism might contribute to the clinical error rate observed in PGD [21]. The precept that not all blastomeres are necessarily equivalent has subsequently emerged as a recognized tenet in human embryology; it figures prominently in the informed consent for patients contemplating PGD [22]. Currently, a technique to determine the extent of embryo mosaicism without disassembling the embryo (and thus rendering it nonviable) does not exist. Accordingly, mosaicism presents potentially serious weaknesses for the two proposed ESC techniques described here, since the effectiveness of each approach is affected by the extent of blastomere mosaicism, which cannot be known a priori.
Nevertheless, for the two theoretical ESC protocols we present, the impact of embryo mosaicism is not the same and each instance deserves separate consideration. For example, if the PGD+blastomere biopsy and culture method were applied to embryos with extensive blastomere heterogeneity, this approach would be unlikely to produce chromosomally normal cells for subsequent in vitro ESC culture. If, however, embryos with very limited or no mosaicism are used for the proposed PGD+blastomere culture process, human ESC production could proceed with much greater likelihood given the uniformity of all sampled cells. Given the unknown extent of embryo mosaicism, limitations of single blastomere biopsy have been recognized [22] and some researchers have recommended confirmatory PGD by sampling a second blastomere [23]. In contrast, among blastomeres obtained from the disaggregation of nonviable embryos, it would be reasonable to expect a higher frequency of mosaicism. In such a setting, even limited mosaicism would yield the desirable result based on the presence of at least one genetically normal constituent blastomere within an organismically dead embryo.
Objective assessment of ESC colonies
Although considerable resources are required to harvest and propagate ESCs, effective methods to verify stemness and monitor quality in such cells are also needed to bring the full range of therapeutic possibility into focus. In an effort to develop an assessment system for ESCs, our center cultured murine blastocysts on mouse fibroblasts in experimental media supplemented with 2000 IU/ml mouse recombinant leukemia inhibitory factor. At 4–5d, the ICMs were mechanically isolated and disaggregated by trypsin. Cell passages were repeated × 2–3d as needed, according to colony confluency.
Our ESC colonies were then graded on the basis of three factors: 1) colony number, 2) colony density, and 3) colony quality. We determined colony character by morphological assessment using an inverted microscope with phase-contrast optics. Typically, ESCs are large and demonstrate a high nuclear:cytoplasm ratio (Figure 3). Each colony was classified according to the proportion of stem cells present within the colony, where >70% (good), 40–70% (average), or <40% (poor) were the three divisions. For all ESC colonies, alkaline phosphatase activity and Oct-4 were used as markers of totipotency. TROMA-1 antibody (monoclonal) directed against cytokeratin-like filaments of trophectoderm and endodermal cells served as a negative marker. As an additional control these markers were tested on expanded blastocysts. Specimens were fixed with 4% paraformaldehyde and permeabilized with 0.2% Triton X-100. Alkaline phosphatase activity in fixed cells was detected via azo-dye with Texas-red filter under fluorescence microscopy. ESCs were exposed to Oct-4 polyclonal antibody (1:100 dilution) and monoclonal TROMA-1 antibody (1:6 dilution), followed by rinse with PBS/BSA to remove unbound antibody. From these experiments, we obtained 16 murine ESC lines from 164 source blastocysts. Assessments via alkaline phosphatase and Oct-4 to verify pluripotency of the ESC lines were in agreement (χ2 = 0.105), while TROMA-1 identified endoderm and trophoblast. Pluripotency was successfully confirmed in at least some cells from each colony studied, and we were able to establish concordance between morphological criteria and marker activity. Further studies will be helpful to show if additional parameters can refine this ESC scoring system.
Stem cell research: social and political factors
Public sharing of information about the basic science of ESCs has proven to be important, since those who are aware of the stem cell debate tend to be more supportive of research in this area compared to those less familiar with the topic [24]. A national public opinion study (n = 1,512) conducted in 2004 found that a narrow majority (52%) of Americans regard the advancement of ESC research as more important than embryo destruction; when this question was posed just two years earlier, support for ESC research was 43% [24].
Discussion on human ESC research occurs against a stormy sociopolitical background – a circumstance not unlike earlier breakthroughs in sperm banking, oral contraception, in vitro fertilization, and intracytoplasmic sperm injection. For ESC, the debate was vocal and escalated quickly to the highest levels of government. In response to the 2005 Presidential State of the Union address, some leaders in the U.S. Congress have proposed standards to authorize broader use of human embryos in research. Specifically, legislation is planned to permit research involving human embryos but only if such embryos have been "ethically derived" – i.e. embryos developed for the purpose of IVF that would otherwise be discarded [25]. This legislative initiative is endorsed by the Coalition for the Advancement of Medical Research, Association of American Universities, Juvenile Diabetes Research Foundation, and Parkinson's Action Network, as well as the Cancer Research and Prevention Foundation.
This type of federal research funding strategy involving existing embryonic stem cell lines is consistent with the President's belief in the fundamental value and sanctity of human life. The President's decision reflects a commitment to preserve the value and sanctity of human life balanced with a desire to promote vital medical research. The Executive Order permits federal funding of research involving the more than 60 existing stem cell lines, but will not sanction or encourage destruction of additional human embryos. Indeed, the embryos from which the existing ESCs originated have already been destroyed. Federal funding of medical research on these existing stem cell lines will promote the sanctity of life "without undermining it" and will allow scientists to explore the potential of this research to benefit the lives of millions of people who suffer from life destroying diseases.
Only certain cell lines will be considered eligible for federally-sponsored human ESC research. These cell lines must be derived (1) with the informed consent of the donors, (2) from non-transferred embryos created solely for reproductive purposes, and (3) without any financial inducements to the donors. In order to ensure that federal funds are used to support only ESC research that is scientifically sound, legal, and ethical, the NIH will examine the derivation of all existing stem cell lines and create a registry of those lines that satisfy these criteria. Thus far, 23 ESC lines have proven viable and have met the NIH criteria, although up to 60 existing ESC lines from genetically diverse populations are eligible for federally-funded research [26].
Conclusion
While other investigators have noted that it is not currently possible to transform a single blastomere into stem cells without recourse to formation and intentional destruction of whole blastocysts [16], in our studies de novo embryos were not generated. Indeed, only ICM-like cell clusters were obtained for further analysis. Refinement and increased efficiency of these two ESC protocols brings the potential to offer a reliable supply of embryonic stem cells without production of whole embryos or compromising extant source blastocysts otherwise selected for embryo transfer or cryopreservation.
Our work confirms the importance of identifying factors that facilitate growth and inhibit differentiation of ESCs. Liberation from a feeder cell requirement may be essential for certain types of experiments, as well as for production of cells for transplantation [27]. Despite extensive experience with mouse embryonic fibroblast feeder cell layers for "culture support", exactly what these feeder cells provide for ESCs has not yet been characterized at the molecular level. Microarray technology applied to early embryo biology promises to answer many of these questions. However, as continued effort is applied toward elucidating these mediators, an adequate supply of source ESCs will prove pivotal to the entire process. From this perspective, our work in a murine model offers some novel approaches to allow progress in ESC research that would not damage or destroy extant human embryos. It is hoped that further study of embryo biopsy+culture as well as blastomere donation from non-viable embryos will permit ethical ESC research to reach its full potential.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
ESS, TT, NT, QVN and GDP contributed equally to this work. GDP conceived of the project, coordinated the research, and edited the manuscript.
Figure 1 Blastomere isolation sequence. An intact 8-cell mouse embryo (a) was subjected to pronase digestion to remove the zona pellucida (b). Single blastomeres were disaggregated by microdissection (c) and after stabilization in culture were monitored for further treatment (d). Scale bar = 100 microns.
Figure 2 Evolution of experimental blastomere growth observed on feeder cell layer on culture days 2, 3, and 4. Top row shows a single blastomere undergoing cleavage (a) and forming a "unilaminar vesicle" on day 3 (b). Cellular arrest and degeneration were evident by day 4 (c). Bottom row shows another cleaving blastomere (d), which formed a cellular aggregate on day 3 (e) and later developed an inner cell mass-like structure (f).
Figure 3 Experimental embryonic stem cell colonies derived from a single blastomere.
==== Refs
Thomson JA Itskovitz-Eldor J Shapiro SS Waknitz MA Swiergiel JJ Marshall VS Jones JM Embryonic stem cell lines derived from human blastocysts Science 1998 282 1145 1147 9804556 10.1126/science.282.5391.1145
Fischbach GD Fischbach RL Stem cells: science, policy, and ethics J Clin Invest 2004 114 1364 1370 15545983 10.1172/JCI200423549
Donovan PJ Gearhart J The end of the beginning for pluripotent stem cells Nature 2001 414 92 97 11689953 10.1038/35102154
Doerflinger RM The ethics of funding embryonic stem cell research: a Catholic viewpoint Kennedy Inst Ethics J 1999 9 137 150 11657324
Coalition urges passage of stem cell legislation Boston Globe:B3 March 17, 2005.
Hardy K Martin KL Leese HJ Winston RM Handyside AH Human preimplantation development in vitro is not adversely affected by biopsy at the 8-cell stage Hum Reprod 1990 5 708 714 2254404
Shahine LK Caughey AB Preimplantation genetic diagnosis Gynecol Obstet Invest 2005 60 39 46 15722632 10.1159/000083483
Hellani A Coskun S Benkhalifa M Tbakhi A Sakati N Al-Odaib A Ozand P Multiple displacement amplification on single cell and possible PGD applications Mol Hum Reprod 2004 10 847 852 15465849 10.1093/molehr/gah114
Chesne P Heyman Y Peynot N Renard JP Nuclear transfer in cattle: birth of cloned calves and estimation of blastomere totipotency in morulae used as a source of nuclei C R Acad Sci III 1993 316 487 491 8053996
Spielmann H Jacob-Muller U Beckord W Immunosurgical studies on inner cell mass development in rat and mouse blastocysts before and during implantation in vitro J Embryol Exp Morphol 1980 60 255 269 6895522
Shamonki M Tanaka N Takeuchi T Neri QV Baek KJ Rosenwaks Z Palermo GD Embryonic stem cell harvesting is enhanced by a chemically-defined medium (abstract O-019) Hum Reprod 2005 20 i7
Jones HW Reflections on the usefulness of embryo cloning Kennedy Inst Ethics J 1994 4 205 207 11645277
Sills ES Sweitzer CL Morton PC Perloe M Kaplan CR Tucker MJ Dizygotic twin delivery following in vitro fertilization and transfer of thawed blastocysts cryopreserved at day 6 and 7 Fertil Steril 2003 79 424 427 12568858 10.1016/S0015-0282(02)04687-3
Grinnell F Defining embryo death would permit important research Chron High Educ 2003 49 B13 15287127
President's Commission for the Study of Ethical Problems in Medicine and Biomedical and Behavioral Research Defining death: a report on the medical, legal and ethical issues in the determination of death 1981 USGPO Washington DC
Landry DW Zucker HA Embryonic death and the creation of human embryonic stem cells J Clin Invest 2004 114 1184 1186 15520846 10.1172/JCI200423065
Voullaire L Slater H Williamson R Wilton L Chromosome analysis of blastomeres from human embryos by using comparative genomic hybridization Hum Genet 2000 106 210 217 10746563 10.1007/s004390051030
Eckert J Tao T Niemann H Ratio of inner cell mass and trophoblastic cells in blastocysts derived from porcine 4- and 8-cell embryos and isolated blastomeres cultured in vitro in the presence or absence of protein and human leukemia inhibitory factor Biol Reprod 1997 57 552 560 9282990
Simonsson S Gurdon JB Changing cell fate by nuclear reprogramming Cell Cycle 2005 4 4
Handyside AH Lesko JG Tarin JJ Winston RM Hughes MR Birth of a normal girl after in vitro fertilization and preimplantation diagnostic testing for cystic fibrosis N Engl J Med 1992 327 905 909 1381054
Munne S Lee A Rosenwaks Z Grifo J Cohen J Diagnosis of major chromosome aneuploidies in human preimplantation embryos Hum Reprod 1993 8 2185 2191 8150922
Baart EB Van Opstal D Los FJ Fauser BC Martini E Fluorescence in situ hybridization analysis of two blastomeres from day 3 frozen-thawed embryos followed by analysis of the remaining embryo on day 5 Hum Reprod 2004 19 685 693 14998971 10.1093/humrep/deh094
Emiliani S Gonzalez-Merino E Englert Y Abramowicz M Comparison of the validity of preimplantation genetic diagnosis for embryo chromosomal anomalies by fluorescence in situ hybridization on one or two blastomeres Genet Test 2004 8 69 72 15140376 10.1089/109065704323016058
The Pew Forum on Religion & Public Life "August 2004 New Interest Index Survey" 2004 Washington DC
Castle MN DeGette D Cosponsor the "Stem Cell Research Enhancement Act" [letter to Members of the Congress of the United States] January 19, 2005.
Fact Sheet-Embryonic Stem Cell Research U.S. Department of Health & Human Services. Washington, DC (Originally issued July 14, 2004; revised August 18, 2004).
Hipp J Atala A Tissue engineering, stem cells, cloning, and parthogenesis: new paradigms for therapy J Exp Clin Assist Reprod 2004 1 3 15588286 10.1186/1743-1050-1-3
|
16026616
|
PMC1185568
|
CC BY
|
2021-01-04 16:39:26
|
no
|
Theor Biol Med Model. 2005 Jul 18; 2:25
|
utf-8
|
Theor Biol Med Model
| 2,005 |
10.1186/1742-4682-2-25
|
oa_comm
|
==== Front
Virol JVirology Journal1743-422XBioMed Central London 1743-422X-2-471592706810.1186/1743-422X-2-47ResearchPossible active origin of replication in the double stranded extended form of the left terminus of LuIII and its implication on the replication model of the parvovirus Diffoot-Carlo Nanette [email protected]élez-Pérez Lisandra [email protected] Jesús-Maldonado Idaris [email protected] Department of Biology, University of Puerto Rico, P.O. Box 9012, Mayagüez, Puerto Rico 006802005 31 5 2005 2 47 47 14 4 2005 31 5 2005 Copyright © 2005 Diffoot-Carlo et al; licensee BioMed Central Ltd.2005Diffoot-Carlo 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 palindromic termini of parvoviruses have proven to play an essential role as origins of replication at different stages during the replication of their viral genome. Sequences from the left-end telomere of MVM form a functional origin on one side of the dimer replicative form intermediate. In contrast, the right-end origin can operate in its closed replicative form hairpin configuration or as a fully duplex linear sequence derived from either arm of a palindromic tetramer intermediate. To study the possibility that the LuIII left hairpin has a function in replication, comparable to that described for MVM, the replication of a minigenome containing two copies of the LuIII left terminus (LuIII Lt-Lt) was studied.
Results
The data presented demonstrates that LuIII Lt-Lt was capable of replicating when NS1 helper functions were provided in trans. This extended hairpin, capable of acting as an origin of replication, lacks the arrangement of the specific domains present in the dimer duplex intermediate of MVM, the only active form of the left hairpin described for this parvovirus.
Conclusions
These findings suggest that the left hairpin of LuIII has an active NS1 driven origin of replication at this terminus in the double stranded extended form. This difference between LuIII and MVM has great implications on the replication of these viruses. The presence of origins of replication at both the left and right termini in their natural hairpin form can explain the unique encapsidation pattern observed for LuIII hinting on the mechanism used by this virus for the replication of its viral genome.
==== Body
Background
Parvoviral DNA replication is a complex process that proceeds by a rolling hairpin mechanism [1-3]. Autonomous parvovirus replication and assembly occurs in the nucleus and is dependent upon host enzymes and cellular functions occurring during the S phase of the cell cycle [4-6]. MVM has been studied as a model for the replication of autonomous parvoviruses [7]. Replication initially proceeds rightward from the terminal 3' hydroxyl of the hairpin stem. The 3' hairpin serves as a primer, which allows a host polymerase to synthesize a complementary copy of the internal sequence of the viral genome until the growing strand reaches the folded back 5' terminus at the right end, resulting in a covalently closed DNA replicative form (cRF) [8]. Further processing involves the opening of the cRF at its right end by the non structural protein 1 (NS1). NS1 attaches covalently to the 5' end at the nick site via a phosphotyrosine bond [9], followed by displacement and copying of the right end hairpin, giving rise to an extended molecule designated 5' eRF [1,9,10]. Rearrangement of the copied right hand palindrome into hairpin structures creates the so-called rabbit-ear replicative form (5' reRF) [11]. This provides a primer for strand-displacement synthesis, leading to the formation of a dimer duplex intermediate (dRF) in which two unit length copies of the genome are joined by a single duplex copy of the original 3' palindrome [8,10,12,13]. In the bridge arrangement of the dRF, the mismatched doublet GA and triplet GAA are now based paired to their complementary sequences. The sequence surrounding the doublet is a potent origin, but the analogous region containing the triplet is considered completely inactive [5]. The actual sequence of the GA doublet is unimportant, but insertion of any third nucleotide here inactivates the origin, suggesting that it represents a critical spacer element [5]. The junction region thus formed contains an active NS1 driven origin [14,15].
Genetic mapping studies revealed that the minimal active MVM-3' [Genbank NC 001510] replication origin is a multi-domain structure of approximately 50 base pair (bp) sequence derived from the outboard arm of the palindromic dimer bridge structure [5,12,14]. It contains three distinct recognition elements: an NS1 binding site (ACCA)1–3; an NS1 nick site (CTWW↓TCA-); and a region containing a consensus activated transcription factor (ATF/CREB) binding site, essential for origin activity. NS1 binds the minimal origin in an ATP-dependent manner but is unable to initiate replication [16]. A cellular factor termed PIF, for parvovirus initiation factor, acts as an essential cofactor for NS1 in the replication initiation process allowing efficient and specific nicking of the 3' minimal origin and leaving NS1 covalently attached to the 5' end of the DNA at the nick site [16,17]. The region containing the PIF binding site is highly conserved in the 3' hairpin of other parvoviruses related to MVM, such as LuIII, H1 and MPV [16]. Once the dimer junction is formed, it is resolved asymmetrically by NS1 which introduces a single-stranded nick into the active origins generating two types of replicative form DNA: an extended palindromic form, and a turnaround form that recreates the left-hand termini [3,14,18]. The turnaround molecule generated in this way re-enters the cycle, while the extended molecule is thought to lead to the displacement of single-stranded genomic DNA, which is then packaged into pre-formed empty capsids [19].
Although the two viral telomeres are very different from each other in size, primary sequence and secondary structure, they both contain elements that become rearranged to create an NS1 dependant origin of replication, activated by different cellular cofactors. Sequences from the left-end telomere form a functional origin only on one side of the dRF intermediate [5,14]. In contrast, the right-end origin can operate in its cRF hairpin configuration and as a fully duplex linear sequence derived from either arm of a palindromic tetramer intermediate [20,21]. Unlike PIF heterocomplex, the essential cofactor for the right end origin is a non sequence- specific DNA-binding protein from the high-mobility group 1/2 (HMG 1/2) family of chromatin-associated polypeptides [20].
To study the possibility that the LuIII [Genbank M81888] left hairpin has a function in replication, comparable to that described for MVM, a minigenome containing two copies of the LuIII-3' terminus (LuIII Lt-Lt) was constructed. The sequences were cloned into the Bam HI site of the pUC19 vector in the head to tail-tail to head orientation, [LuIII nucleotides (nt.) 1-278/278-1]. The data presented demonstrates that LuIII Lt-Lt was capable of replicating when helper functions were provided in trans by pGLu883ΔXba, the genomic clone of LuIII, or with pCMVNS1, an NS1 expressing vector, suggesting that this LuIII sequences contain all the cis-acting sequences required for excision and DNA replication. The replication of this minigenome demonstrates that the left hairpin of LuIII has an active NS1 driven origin of replication that does not have the arrangement of the dimer duplex intermediate described for MVM.
Results and Discussion
A plasmid (LuIII Lt-Lt) containing two copies of the LuIII 3' termini flanking an E. coli stuffer sequence, was constructed (figure 1). In anticipation of the difficulties expected in manipulating the left end hairpin and to increase the chances of obtaining the desired construct two copies of the left end termini were successfully ligated in vitro, in a tail (nt 278) to head (nt 1) -head to tail orientation, this prior to cloning into pUC19. Sequencing of all recombinants obtained, with an exception, revealed a single copy of the left hairpin of LuIII ligated to E. coli sequences of ~250 bp. These recombinants all contained the LuIII hairpin sequence in the same orientation in pUC19 with respect to the Reverse and Forward Primers, conserving the LuIII sequence at the 5' end and the E. coli sequence at the 3' end. Cotmore and Tattersall [22] reported that the palindromic inserts had a greater tendency for deletions, even in recombination-deficient strains of E. coli, this probably due to the complex structures assumed by the inserts. Liu et al. [3] also reported inherent difficulties in cloning hairpins, resulting in many incorrect and presumably rearranged clones. The LuIII sequences may have formed a complex hairpin structure in vivo, due to its palindromic nature that was removed by slipped mispairing during replication [23]. Difficulty in the sequencing of these clones, particularly with the Reverse primer (M13R), supports this observation.
Figure 1 Strategy Used to Construct LuIII Lt-Lt. White, grey and dotted regions represent LuIII, pUC19 vector and E. coli sequences, respectively. Restriction enzyme sites used are indicated. PGLU883 corresponds to the LuIII infectious genomic clone.
LuIII Lt-Lt was cotransfected with pGLu883ΔXba, the genomic clone of LuIII, by electroporation into HeLa cells. pGLu883ΔXba provides the trans acting factors necessary for replication of the minigenome. Southern blot analysis of the transfection assays are shown in figure 2. The blot was hybridized with the LuIII Lt-Lt Bam HI fragment labeled by random primed Digoxigenin-11-dUTP. Cotransfection of pGLu883ΔXba/LuIII Lt-Lt (lane 2), resulted in three sets of doublet bands. These doublets were of ~1.8, ~1.2 and ~.8 kb. These bands do not appear for the replication of the LuIII genomic clone, pGLu883ΔXba (lane 1) nor for the transfection of LuIII Lt-Lt (lane 3) for which only input plasmid was observed since the plasmid was not capable of replicating in the absence of helper functions. When DNA samples were digested with Mlu I (lanes 4–6) pGLu883ΔXba resulted in a strong band of ~278 bp (lane 4) corresponding to the left terminus of LuIII. Given the probe used (exclusively the LuIII Lt-Lt insert) the large fragment corresponding to nts 279-5135 of the LuIII genome was not observed on this gel. The presence of this fragment was confirmed by southern blot analysis using the full length genome of LuIII (Data not shown). Cotransfection of pGLu883ΔXba/LuIII Lt-Lt digested with Mlu I (lane 5) resulted in two bands, one migrating with the ~278 bp band of pGLu883ΔXba/ Mlu I (lane 4) and a band of greater intensity migrating slightly faster. Digestion of the cotransfection sample with Mlu I (lane 5) also eliminated the three sets of doublets observed in the uncut sample (lane 2) of the cotransfection suggesting that these molecules likely represent concatemers of a single molecule. Digestion of a monomer molecule resulting from the replication of LuIII Lt-Lt with Mlu I is expected to generate two fragments, one of ~278 bp corresponding to the left hairpin of LuIII and a band corresponding to the E. coli stuffer sequence which has a size of ~250 bp; two molecules of the hairpin should be generated for every copy of the stuffer sequence, therewith the intensity of the band corresponding to the hairpin is expected to be greater than the band corresponding to the stuffer sequence. Two bands were observed for this digestion (lane 5); the larger band migrates along side the band observed for pGLu883ΔXba likely representing the left end hairpin of LuIII in double stranded form. The smaller of the two bands, of greater intensity, may represent the left hairpin with an alternate conformation. A faint band of similar migration is observed for pGLu883ΔXba when digested with Mlu I (lane 4). The band corresponding to the stuffer sequence is not obvious, this likely due to its similar migration to the LuIII left end with a different conformation. Lane 6, containing the transfection sample of only LuIII Lt-Lt shows a band of ~250 bp resulting from the digestion of input plasmid that was not capable of replicating, this confirms our assumption that the stuffer sequence observed in lane 6 migrates similar to the left hairpin with an altered conformation hence its greater intensity when compared to the migration of the double stranded left hairpin.
Figure 2 DNA Samples Recovered From Transfection Assays of LuIII Lt-Lt Digested With Mlu I. Samples correspond to DNA isolated from transfection assays. Lanes 2 and 5 represent cotransfections with pGLu883ΔXba and LuIII Lt-Lt. White lines indicate DNA fragments recovered from the replication of LuIII Lt-Lt. Sizes of the 2 log ladder (Roche) are shown. The probe used consisted of the insert of LuIII Lt-Lt labeled by the DNA random primed labeling method.
LuIII Lt-Lt was also cotranfected with pCMVNS1, an expression vector for the MVM nonstructural protein NS1 (figure 3). LuIII Lt-Lt was capable of replicating when only NS1 was present in trans (lane 7) resulting in the same banding pattern as observed in figure 2 (lane2). It has been suggested that the non-structural protein NS1 makes the excision [4] by introducing a single-stranded nick, possibly at the 5' end of the viral genome. If the minigenome could replicate under these conditions, it contains all the cis-acting sequences required for excision and DNA replication. These results suggest that LuIII Lt-Lt was capable of excision and replication when pGLu883ΔXba or pCMVNS1 was provided in trans and that only NS1 viral functions appear to be required for the excision and replication of LuIII Lt-Lt.
Figure 3 DNA Recovered from Transfection Assays of LuIII Lt-Lt with pCMVNS1. DNA samples shown correspond to: 1. the full length insert isolated from LuIII Lt-Lt, 2. negative control of transfection, 3–7. DNA isolated from transfection assays of the indicated samples. Arrow heads point to DNA fragments recovered from the replication of LuIII Lt-Lt. Sizes of the 2 log ladder (Roche) are indicated. The probe used consisted of the insert of LuIII Lt-Lt labeled by the DNA random primed labeling method.
A possible mechanism for the replication of LuIII Lt-Lt is shown in figure 4. The model proposes a nick at the NS1 nick site present at the left hairpin (step 1); this generates two origins of replication running in opposite directions (step 2) that lead to strand displacement. The new hairpins assume secondary structures and continue DNA synthesis (step 4), generating a close-end molecule. This step generates two copies of a molecule estimated to be ~664 nts in length. Both molecules can now generate a monomer length molecule of ~806 bp (step 5). As a result of replication, the arrangement of the arms in the hairpin change resulting in hairpins with the GAG triplet present at the 5'end of the molecules. This forces the molecule to go through a dimer intermediate (steps 7 and 8) generating a molecule with a turn around end of ~1192 nts in length. This dimer is then resolved to generate monomer length double stranded molecules (step 9). The sizes of the DNA molecules obtained from this model on the replication of LuIII Lt-Lt correspond very closely with the sizes of the products predicted by the model (figure 2 and 3) for the replication of LuIII Lt-Lt.
Figure 4 Proposed Model for the Rescue and Replication of LuIII Lt-Lt. Restriction sites and their positions with respect to the LuIII sequence are indicated. Grey, white and zigzag regions represent pUC19, LuIII left terminus and E. coli sequences respectively. In steps 4 and 5 the molecules generated (a/b and aa/bb) are identical, for this reason only one molecule at each step is continued throughout the scheme. The estimated sizes of some of the molecules (boxed) are indicated.
Parvovirus DNA replication starts when the 3' hydroxyl group at the left end of the viral genome primes the synthesis of a complementary strand, leading the formation of a double stranded replicative form known as the cRF. In vitro studies have shown that the cRF of autonomous parvovirus like MVM terminates in closed hairpins at both ends, making cRF a major, or even obligatory intermediate of parvovirus replication [8], but only the right-end hairpin is resolved in the presence of NS1 [8,24]. The cRF is re-opened and copied, producing a right end extended form (5' eRF) followed by unfolding of the hairpin and copying of the terminal sequence. This leads to the formation of dimeric RF (dRF) and higher-order concatamers that would be resolved into monomeric (mRF) RF DNA. If the wild type LuIII virus replicates using the mechanism described for MVM and forms the cRF, the replication of two copies of the left end such as in LuIII Lt-Lt should result in a dead molecule that could not be resolved by NS1. Although the terminal palindromic sequences are essential for the replication of the APVs genome, the right and left terminal sequences are not equivalent in function [25,26]. According to the modified rolling hairpin model of MVM replication, the right end origin is active in the covalently closed hairpin configuration and also in the extended right end telomere [14,24]. In contrast, the MVM left end inverted repeat does not constitute a replication origin in the hairpin configuration and needs to be copied in the form of a left-to-left end bridge to be subsequently resolved at the multimeric RF DNA stage [1,3,8,14,27].
When the dimer bridge origin of MVM is compared to the left end arrangement in LuIII Lt-Lt (figure 5), it becomes apparent that the left terminus is an incomplete origin of replication based on the origin proposed for MVM replication. A competent replication origin contains, among other things an NS1 nick site. If like MVM, the left end terminus of LuIII is only processed when present as a bridge in the dimer RF but not as a hairpin in monomeric replicative form, neither of the left end termini in LuIII Lt-Lt would be recognized by NS1. As a result, the LuIII insert would not be excised from the plasmid pUC19, and hence no replication would be expected to occur. Comparison of the sequences present in LuIII Lt-Lt with the junction bridge in the dimer replicative form of MVM [28] (figure 4) illustrates that the A and B arms of the LuIII left end are organized differently from that proposed for the active origin of replication for MVM. Unlike the dimer arrangement described for MVM, in LuIII Lt-Lt the CT doublet is positioned at the 5'end and the CTC triplet is positioned inboard at the 3'end in both hairpins. In the hairpin arrangement an NS1 nick site is not present at the 5' end of the CT bubble as described in the MVM dimer bridge. Nevertheless, LuIII Lt-Lt was capable of replication suggesting that the left hairpin of LuIII does constitute a replication origin in the extended double stranded hairpin configuration.
Figure 5 Comparison of the MVM Dimer Bridge (A) with the Hairpin Arrangement in LuIII Lt-Lt (B). Hairpins and NS1 recognition nick sites are indicated by dark bold lines and arrows respectively. The grey patterned boxes correspond to pUC19 sequences.
Given the functionality of the left hairpin of LuIII as an origin of replication in the extended double stranded form a replication model of LuIII can be predicted resulting in equivalent amounts of plus and minus DNA viral strands (figure 6). In this model the plus and minus DNA strands, independently initiate replication from the right and left hairpins respectively (step 1). The NS1 nick sites present at the left and right termini in LuIII differ from each other; there is an insertion of an Adenine residue in the NS1 nick site present at the 5' terminus of LuIII. This additional adenine is also not present in the NS1 nick site described for MVM [29].
Figure 6 Proposed Model for the Replication of Parvovirus LuIII. A model for the replication of the (+) and the (-) strand of LuIII is shown. The NS1 nick site and its complementary sequence (*) are indicated. The unpaired sequences present at the left hairpin are shown. The arrows point to NS1 nick sites. A corresponds to the insertion in the NS1 nick site present at the right terminus of LuIII.
This replication model for LuIII predicts flip/flop conformations at both termini. Earlier studies [30] in which the left and right termini of the minus and plus strands, respectively, were labeled at the 3' hydroxyl group and subsequently digested with Hha I suggested that the left terminus of the LuIII minus strand exists only in the flip conformation, and the right terminus of the plus strand exists in both the flip and flop conformations. Numerous bands were observed when the left terminus of the minus strand was digested with Hha I yet these were justified as alternate secondary structures of the hairpin in the flip conformation. The expected fragments for the digestion of the flip and flop conformations of the left hairpin are very similar in size, any slight variation in migration due to the secondary structures assumed by these fragments could have impaired the interpretation of the results. The conformation present at the left end of the plus strand still remains unknown.
Conclusion
The data presented demonstrates that LuIII Lt-Lt contains all the cis-acting sequences required for excision and DNA replication when NS1 viral functions are provided in trans. These findings suggest that the left hairpin of LuIII has an active NS1 driven origin of replication at this terminus in the double stranded extended form. This extended hairpin, capable of acting as an origin of replication, lacks the arrangement of the specific domains present in the dimer duplex intermediate of MVM, the only active form of the left hairpin described for MVM. This difference between LuIII and MVM has great implications on the replication of these viruses. The presence of origins of replication at both the left and right termini can explain the unique encapsidation pattern observed for LuIII hinting on the mechanism used by LuIII for the replication of its viral genome.
Methods
Construction of LuIII Lt-Lt
The LuIII Lt-Lt minigenome (figure 1) has two copies of the left end palindrome of the autonomous parvovirus LuIII (nt. 1-278) cloned into the Bam HI site of pUC19 [29] [Genbank L09137]. The 3' hairpin of LuIII was obtained from pGLu883 [30], the full-length genomic clone of LuIII cloned into the pUC19 vector. pGLu883 was digested with both Bam HI (pUC19 nt. 417) and Mlu I I (LuIII nt. 278) for two hours at 37°C and then electrophoresed on a 1.2% agarose gel in 1X TBE buffer at 75 V. The Bam HI / Mlu I I digestion generated three fragments of approximately 278, 2686, and 4861 bp. The 278 bp fragment corresponding to the left end hairpin was isolated and purified using the Geneclean Spin Kit® (QBio-gene, Carlsbad, CA), and then were ligated through the Mlu I site in an overnight reaction at 4°C using 1 U of T4 DNA ligase. The ligation was digested with Bam HI generating a fragment of 568 bp corresponding to the two copies of the 3' hairpin in a "head to tail-tail to head" conformation (nts 1-278, 278-1). The fragment generated was purified as described and ligated into the Bam HI site of pUC19 that was previously treated with calf intestinal alkaline phosphatase (CIAP) (Roche Applied Science, Indianapolis, IN) for one hour at 37°C.
Preparation of Competent Cells
Two different strains of Escherichia coli were used as competent cells: DH5α [(lacZ.M15. (lacZYA-argF) recA1 endA1 hsdR17 (rkmk+) phoA supE44 thi gyrA96 relA1)] (ATCC, Rockville, MD) and SURE®2 super competent cells [(e14- (McrA-). (mcrCB-hsd SMR-mrr) 171 endA1 supE44 thi-1 gyrA96 relA1 lac recB recJ sbcC umuC::Tn5 (Kanr) uvrC (F' proAB lacIqZ.M15 Tn10 (Tetr) Amy Camr)] (Stratagene, La Jolla, CA). Competent cells were prepared by the calcium chloride method [31].
Transformation of Competent Cells
The recombinant molecules were transformed in both DH5α and SURE®2 competent cells. Competent cells were thawed on ice for 15 minutes (min.). The DNA was added to the cells and incubated on ice for 30 min. Cells were heat-shocked in a 42°C water bath and subsequently incubated on ice for 2 min. DH5α and SURE®2 competent cells were heat-shocked for 2 min. and 30 seconds respectively. 100 μL of preheated (42°C) LB broth was added to both cell samples and incubated at 37°C for 1 hour (h) with shaking at 225 rpm. DH5α transformed cells were spread on LB agar plates containing 50 mg/mL ampicillin and 80 μL of 2% X-gal. SURE®2 transformed cells were spread on LB plates containing 50 mg/mL ampicillin, 100 μL of 2% X-gal and 100 μL of 10 mM IPTG.
Isolation of DNA Recombinants
The resultant plasmids from DH5α and SURE®2 transformed cells were purified by the alkaline lysis miniprep method, described by Ausubel et al. [31] and analyzed with restriction enzymes. Sequencing was performed at the New Jersey Medical School, Molecular Resource Facility.
Tissue Culture
HeLa (ATCC, Rockville, MD) cells were grown in Minimal Essential Medium (MEM Eagle) (MP Biomedicals, Aurora, OH) supplemented with 10% fetal bovine serum (FBS) (HyClone, Logan, UT) and PSG (8 mM Penicillin G, 3 mM Streptomycin Sulfate, 200 mM L-Glutamine). They were incubated at 37°C in 25 and/or 75 cm2 plastic tissue culture flasks. For sub-culturing, the cells were rinsed twice with Phosphate-Buffered Saline (1X PBS) and incubated in 1X Trypsin (Difco, Detroit, MI) for 5 min. at 37°C. Cells were harvested by centrifugation at 3800 rpm for 5 min. at 4°C. The resultant pellet was resuspended in the medium described above and seeded into culture flasks at a proportion of 1:3.
Transfection Assay
HeLa cells were grown to 100 % confluency in a 75 cm2 flask. They were washed three times with 1X PBS and then tripsinized at 37°C for 5 minutes. Cells were harvested by centrifugation at 3,800 rpm for 5 min. at 4°C and washed in 10 ml of PBS. Cells were resuspended and split at a proportion of 1:9. Approximately, 5 μg of pGLu883ΔXba, LuIII Lt-Lt minigenome and pCMVNS1 were added to the corresponding tubes and incubated at 37°C for 10 min. Cells were transferred to sterile cuvettes with a 4-mm gap width, and electroporated individually at 230 V and 950 μF using a capacitance discharge machine (Gene Pulser, Bio-Rad Laboratories, Hercules CA). After each pulse, 700 μL of MEM-10% FBS were added to the cuvette and the cells were resuspended carefully. The electroporated cells were incubated for 45 min. at 37°C and then transferred to 25 cm2 flasks containing 3 mL MEM-10% FBS. After an overnight incubation at 37°C, the medium was changed, and the cells were incubated at 37 °C until the low molecular weight DNAs were isolated at five days post-transfection, as described by Tam and Astell [25]. DNA samples were resuspended in 30 μL TE (10 mM Tris-HCl, 1 mM EDTA, pH 8.0).
Southern Blot Analysis
Samples were electrophoresed on a 1.2% agarose gel in 1X TAE buffer at 80 V, and passively transferred onto a Zeta Probe nylon membrane (Bio-Rad Laboratories, Hercules, California) as described by Ausubel et al [31]. Probes were labelled by the random primed DNA labeling method with Digoxigenin-11-dUTP (Roche Applied Science, Indianapolis, IN). The blot was hybridized at 50°C and washed at 55°C. Detection was performed according to manufacturer's instructions (Roche Applied Science, Indianapolis, IN).
Competing interests
The author(s) declare that they have no competing interests
Authors' contributions
NDC drafted and revised critically the manuscript, had the intellectual idea of the study and its design, contributed significantly in the analysis and interpretation of the data, proposed the replication models presented and gave the final approval of the version to be published.
LVP constructed LuIII Lt-Lt, collected the data resulting from the transfection of LuIII Lt-Lt/pGluΔXba, contributed in the analysis and interpretation of the data, participated in the idea and design of the models proposed and in the drafting and revision of the manuscript.
IDM collected the data resulting from the transfections of LuIII Lt-Lt/pGluΔXba and, LuIII Lt-Lt/pCMVNS1, contributed in the analysis and interpretation of the data, participated in the design of the models proposed and in the drafting and revision of the manuscript.
All authors read and approved the final manuscript.
Acknowledgements
We thank Dr. David Pintel and Dr. Ian Maxwell for the pCMVMNS1 and pGLuΔXba clones respectively and Omayra Rivera-Denizard for her helpful suggestions in the design of the models.
This work was supported by the Minority Biomedical Research Support, National Institute of Health Grant SO6GM08103 and the College of Arts and Sciences, University of Puerto Rico at Mayaguez.
==== Refs
Astell CR Chow MB Ward D Sequence analysis of the termini of virion and replicative forms of Minute Virus of Mice DNA suggests a modified rolling hairpin model for autonomous parvovirus DNA replication J Virol 1985 54 171 177 3973977
Cotmore SF Tattersall P Depamphilis ML Parvovirus DNA Replication DNA Replication in Eukaryotic Cells 1996 New York: Cold Spring Harbor Lab 799 813
Liu Q Yong CB Astell CR In vitro resolution of the dimmer bridge of the Minute Virus of Mice (MVM) genome supports the modified rolling hairpin model for MVM replication Virol 1994 201 251 262 10.1006/viro.1994.1290
Berns KI Fields BN, Knipe DM, Howley PM Parvoviridae: the viruses and their replication Fundamental Virology 1996 3 Pennsylvania: Lippincott-Raven 1017 1036
Cotmore SF Tattersall P An asymmetric nucleotide in the parvoviral 3' hairpin directs segregation of a single active origin of DNA replication Embo J 1994 13 4145 4152 8076610
Faust EA Rankin CD In vitro conversion of MVM virus single-stranded DNA to the replicative form by DNA polymerase alpha from Ehrlich ascites tumor cells Nucl Acids Res 1982 10 4181 4201 6812024
Faisst S Rommelaere J (eds) Schmidt A Parvoviruses From Molecular Biology to Pathology and Therapeutic Uses Contrib Microbiol 2000 4 New York: Karger Press
Baldauf AQ Willwand K Mumtsidu E Nüesch JP Rommelaere J Specific initiation of replication at the right-end telomere of the closed species of Minute Virus of Mice replicative-form DNA J Virol 1997 71 971 980 8995615
Cotmore SF Tattersall P The NS-1 polypeptide of Minute Virus of Mice is covalently attached to the 5' termini of duplex replicative-form DNA and progeny single strands J Virol 1988 62 851 860 3339715
Willwand K Mumtsidu E Kuntz-Simon G Rommelaere J Initiation of DNA replication at palindromic telomeres is mediated by a duplex-to-hairpin transition induced by the Minute Virus of Mice nonstructural protein NS1 J Biol Chem 1998 273 1165 1174 9422783 10.1074/jbc.273.2.1165
Kuntz-Simon G Bashir T Rommelaere J Willwand K Neoplastic transformation-associated stimulation of the in vitro resolution of concatemer junction fragments from Minute Virus of Mice DNA J Virol 1999 73 2552 2558 9971842
Cotmore SF Tattersal P DNA replication in the autonomous parvoviruses Semin Virol 1995 6 271 281 10.1006/smvy.1995.0033
Wilson GM Hindal HK Yeung DE Chen W Astell CR Expression of Minute Virus of Mice major nonstructural protein in insect cells: Purification and identification of ATPase and helicase activities Virol 1991 185 90 98 10.1016/0042-6822(91)90757-3
Cotmore SF Nüesch JPF Tattersall P Asymmetric resolution of a parvovirus palindrome in vitro J Virol 1993 67 1579 1589 8437230
Majaniemi I Siegl G Early events in the replication of parvovirus LuIII Arch Virol 1984 81 285 302 6089705 10.1007/BF01309999
Christensen J Cotmore SF Tattersall P A novel cellular site-specific DNA-binding protein cooperates with the viral NS1 polypeptide to initiate parvovirus DNA replication J Virol 1997 71 1405 1416 8995666
Christensen J Cotmore SF Tattersall P Parvovirus initiation factor PIF: a novel human DNA-binding factor which coordinately recognizes two ACGT motifs J Virol 1997 71 5733 5741 9223459
Cotmore SF Nüesch JP Tattersall P In vitro excision and replication of 5' telomeres of Minute Virus of Mice DNA from cloned palindromic concatemer junctions Virol 1992 190 365 377 10.1016/0042-6822(92)91223-H
Muller DE Siegl G Maturation of Parvovirus LuIII in a subcellular system. I. Optimal conditions for in vitro synthesis and encapsidation of viral DNA J Gen Virol 1983 64 1043 1054 6842185
Cotmore SF Tattersall P High-mobility group 1/2 proteins are essential for initiating rolling-circle-type DNA replication at a parvovirus hairpin origin J Virol 1998 72 8477 8484 9765384
Cotmore SF Christensen J Tattersall P Two widely spaced initiator binding sites create an HMG1-dependent parvovirus rolling-hairpin replication origin J Virol 2000 74 1332 1341 10627544 10.1128/JVI.74.3.1332-1341.2000
Cotmore SF Tattersall P In vivo resolution of circular plasmids containing concatemer junction fragments from Minute Virus of Mice DNA and their subsequent replication as linear molecules J Virol 1992 66 420 431 1530771
Tam P Astell CR Multiple cellular factors bind to cis-regulatory elements found inboard of the 5' palindrome of Minute Virus of Mice J Virol 1994 68 2840 2848 8151755
Willwand K Baldauf AQ Deleu L Mumtsidu E Costello E Beard P Rommelaere J The Minute Virus of Mice (MVM) nonstructural protein NS1 induces nicking of MVM DNA at a unique site of the right-end telomere in both hairpin and duplex conformations in vitro J Gen Virol 1997 78 2647 2655 9349487
Tam P Astell CR Replication of Minute Virus of Mice minigenomes: novel replication elements required for MVM DNA replication Virol 1993 193 812 824 10.1006/viro.1993.1190
Cotmore SF Tattersall P Genome packing sense is controlled by the efficiency of the nick site in the right-end replication origin of parvoviruses Minute Virus of Mice and LuIII J Virol 2005 79 2287 2300 15681430 10.1128/JVI.79.4.2287-2300.2005
Cotmore SF Tattersall P The autonomously replicating parvoviruses of vertebrates Adv Virus Res 1987 33 91 173 3296697
Cotmore SF Tattersall P Resolution of parvovirus dimer junction proceeds through a novel heterocruciform intermediate J Virol 2003 77 6245 6254 12743281 10.1128/JVI.77.11.6245-6254.2003
Diffoot N Chen KC Bates RC Lederman M The complete nucleotide sequence of parvovirus LuIII and localization of a unique sequence possibly responsible for its encapsidation pattern Virol 1993 192 339 345 10.1006/viro.1993.1040
Diffoot N Shull BC Chen KC Stout ER Lederman M Bates R Identical ends are not required for the equal encapsidation of plus- and minus- strand parvovirus LuIII DNA J Virol 1989 63 3180 3184 2542625
Ausubel FM Roger B Kingston RE Moore DD Seidman JG Smith JA Struhl K Short protocols in Molecular Biology 1999 New York: John Wiley & Sons
|
15927068
|
PMC1185569
|
CC BY
|
2021-01-04 16:39:00
|
no
|
Virol J. 2005 May 31; 2:47
|
utf-8
|
Virol J
| 2,005 |
10.1186/1743-422X-2-47
|
oa_comm
|
==== Front
Virol JVirology Journal1743-422XBioMed Central London 1743-422X-2-551602273010.1186/1743-422X-2-55ResearchHuman cytomegalovirus uracil DNA glycosylase associates with ppUL44 and accelerates the accumulation of viral DNA Prichard Mark N [email protected] Heather [email protected] Gregory M [email protected] Chengjun [email protected] Zhaoti [email protected] Melissa [email protected] George [email protected] Earl R [email protected] Department of Pediatrics, University of Alabama at Birmingham, Birmingham AL, USA2 Department of Research, MedImmune Vaccines Inc., Mountain View, CA, USA2005 15 7 2005 2 55 55 18 5 2005 15 7 2005 Copyright © 2005 Prichard et al; licensee BioMed Central Ltd.2005Prichard 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 cytomegalovirus UL114 encodes a uracil-DNA glycosylase homolog that is highly conserved in all characterized herpesviruses that infect mammals. Previous studies demonstrated that the deletion of this nonessential gene delays significantly the onset of viral DNA synthesis and results in a prolonged replication cycle. The gene product, pUL114, also appears to be important in late phase DNA synthesis presumably by introducing single stranded breaks.
Results
A series of experiments was performed to formally assign the observed phenotype to pUL114 and to characterize the function of the protein in viral replication. A cell line expressing pUL114 complemented the observed phenotype of a UL114 deletion virus in trans, confirming that the observed defects were the result of a deficiency in this gene product. Stocks of recombinant viruses without elevated levels of uracil were produced in the complementing cells; however they retained the phenotype of poor growth in normal fibroblasts suggesting that poor replication was unrelated to uracil content of input genomes. Recombinant viruses expressing epitope tagged versions of this gene demonstrated that pUL114 was expressed at early times and that it localized to viral replication compartments. This protein also coprecipitated with the DNA polymerase processivity factor, ppUL44 suggesting that these proteins associate in infected cells. This apparent interaction did not appear to require other viral proteins since ppUL44 could recruit pUL114 to the nucleus in uninfected cells. An analysis of DNA replication kinetics revealed that the initial rate of DNA synthesis and the accumulation of progeny viral genomes were significantly reduced compared to the parent virus.
Conclusion
These data suggest that pUL114 associates with ppUL44 and that it functions as part of the viral DNA replication complex to increase the efficiency of both early and late phase viral DNA synthesis.
==== Body
Background
The enzymatic removal of uracil from DNA occurs in all free-living organisms. Both the misincorporation of dUTP by DNA polymerase and the spontaneous deamination of cytosine are relatively frequent events and give rise to uracil residues covalently linked to the genome, with the latter resolving into A:T transition mutations in one of the nascent strands [4,42]. Human herpesviruses, poxviruses and retroviruses either encode or recruit uracil DNA glycosylase (UNG) homologs, presumably to remove uracil bases from genomic DNA [5]. A number of studies used site directed mutagenesis to characterize the function of this gene in the life cycle of these viruses and most have described unexpected facets of the phenotype that involve DNA (or RNA) replication [5]. Studies described here with human cytomegalovirus (CMV) suggest that the UNG is part of the replication complex and that it functions in the replication of the viral genome.
Highly conserved mechanisms have evolved to minimize the presence of uracil in genomic DNA, presumably to prevent damage to the genome [30,44,46]. In humans, at least five base excision repair enzymes are capable of removing uracil bases incorporated in DNA. The human UNG gene expresses distinct nuclear and mitochondrial forms of this enzyme, designated UNG2 and UNG1, respectively [18]. In addition, a thymine(uracil) DNA glycosylase, a cyclin-like UNG, and a new gene SMUG1 have all been shown to possess this activity [24,26,27]. The relative function of each of these molecules remains to be characterized, but it appears that these molecules have developed specialized roles in mammals. Recent studies describing the phenotype of UNG knockout mice did not identify a greatly increased spontaneous mutation rate, in contrast to studies in both prokaryotes and sacharomyces [18]. SMUG1 appears to be responsible for recognizing and repairing uracil residues resulting from the spontaneous deamination of cytosine [26], whereas UNG2 colocalizes with replication foci in dividing cells and is thought to remove uracil during the replication process [18]. An ancillary role for this enzyme in mammalian DNA replication is also supported by the fact that UNG2 interacts physically with both replication protein A [25], as well as proliferating cell nuclear antigen (PCNA) which is a central regulator of DNA synthesis [28]. Further, these interactions suggest that UNG2 participates in the PCNA-requiring 2–8 bp patch base excision repair pathway [39].
A number of virus families appear to recruit UNG2, or to encode UNG2 homologs for use in the replication process. In human immunodeficiency virus (HIV) type 1, the vpr gene product interacts specifically with UNG2 [3]. The Vpr from simian immunodeficiency virus also binds UNG2 in a similar manner, however, it doesn't appear to impact the phenotype of cell cycle arrest associated with Vpr [38]. UNG2 is packaged inside retrovirus virions by an integrase dependent mechanism [45], and physically associates with integrase as well as reverse transcriptase in the pre-integration complex [33]. Lysates from purified virions demonstrated that UNG2 remained functional and was capable of directing the repair of uracil from a synthetic oligonucleotide template in conjunction with reverse transcriptase in a manner that is independent of apurinic/apyrimidinic endonuclease [33]. The function that UNG2 serves in HIV replication is unclear. However, the misincorporation of dUTP in a RT/RNAse H assay does not appear to affect first strand DNA synthesis by RT, but rather, it affects the specificity of cleavage by RNAse H resulting in reduced second strand synthesis from the RNA primers [17]. Poxviruses also encode a UNG2 homologs that perform an essential function in the replication of this virus [22,41,43] and are thought to act at the level of DNA synthesis [8]. More recent studies confirmed that D4R is essential for vaccinia DNA synthesis, and that its essential function is unrelated to its ability to excise uracil from DNA [7].
Herpesviruses all encode UNG homologs that do not appear to be required for replication in cell culture [23,31,36], although the deletion of the homolog in herpes simplex virus appears to reduce neuroinvasiveness in animal models [35]. CMV is unique among these viruses in that the deletion of this ORF results in a distinct phenotype characterized by a marked delay in the onset of DNA synthesis despite the normal temporal expression of early genes involved in this process [29,31]. The phenotype is less apparent in rapidly dividing cells, suggesting that a cellular gene might compensate at least to some degree [6]. Another interesting aspect of the UNG- phenotype occurs late in infection where the mutant virus fails to initiate robust DNA synthesis and concurrently fails to incorporate uracil in the genome, suggesting that the removal of these moieties may be related to the switch to late phase DNA synthesis [6]. It is unclear why this phenotype is observed in CMV and not in other herpesviruses, but it may be related to the distinct mechanisms that this virus has evolved to replicate its genome that is independent of origin binding proteins encoded by most other herpesviruses.
To help understand how the UL114 gene product functions in viral DNA synthesis, a complementing cell line was constructed and recombinant viruses in which this gene product was epitope tagged were used to characterize its expression and localization in the context of a viral infection. Herein, we demonstrate that pUL114 localizes to the viral replication compartments and associates with the accessory factor of the DNA polymerase (ppUL44, ICP36), and that the absence of this molecule results in delayed onset of viral DNA synthesis as well as inefficient replication of the viral genome.
Results
Restoration of UL114
Recombinant viruses with deletions in UL114 express early gene products with normal kinetics, yet exhibit a marked delay in the onset of DNA synthesis [6,31]. This phenotype was assigned to UL114, since two independent isolates of the recombinant virus exhibited the same phenotype. To formally ascribe the observed phenotype to this locus, the lesion was repaired with an Eag I DNA fragment (AD169 coordinates 162693–164080) that spans the deletion in the mutant virus (Fig. 1). Plaques resistant to high concentrations of xanthine were isolated and were shown to have restored the deleted sequences as determined by Southern analysis (data not shown). Kinetics of viral DNA synthesis were examined in HEL cells infected with the parent virus, the mutant (RC2620) and the rescued virus (RQ2620) to determine if the restoration of the UL114 locus reverted the phenotype of delayed DNA synthesis. As observed previously, the mutant exhibited very little DNA synthesis in the first three days of infection (Fig 2A). In contrast, the rescued virus appeared to synthesize DNA with the same kinetics as the parent virus suggesting that the defect was due to the engineered mutation rather than to mutations elsewhere in the genome. These data were confirmed in HEL cells in an experiment in which single-step replication kinetics were examined. Delayed viral replication was observed in the mutant virus, whereas, no difference was observed between the wt virus and the recombinant virus in which the UL114 lesion was repaired (Fig. 2B). Thus, two facets of the described phenotype (DNA synthesis and replication kinetics) were reverted upon restoration of this gene and we formally assigned this phenotype to the engineered mutation. This phenotype was also reproduced in Towne strain of CMV when the UL114 open reading frame was disrupted.
Figure 1 Recombinant viruses. The top line represents the CMV genome with the region surrounding UL114 expanded below. The second line represents the structure of the region in the parent virus (AD169). The third line labeled "RC2620" depicts the 1.2 kb insertion containing the E. coli gpt gene (white arrow) that replaces most of the UL114 ORF. The final three lines represent the same region in Towne and depict the placement of the 35 aa ICP4 epitope tags in the ORF. The entire ORF was also deleted in Towne as a control and resulted in the same slow replication phenotype as was observed in the AD169 strain.
Figure 2 Repair of RC2620. (A) HEL cells were infected at an MOI of 5 PFU/cell and total DNA was harvested at the indicated times. The quantity of viral DNA for AD169 (black squares), RC2620 (black circles), and RQ2620 (open circles) were determined by dot blot hybridization as described in materials and methods. (B) Titers of AD169 (black squares), RC2620 (black circles), and RQ2620 (open circles) are shown. The time point at 0 hpi represents the titer of the input virus.
Complementation of the UNG deficient mutant in trans and the effect of uracil content on the phenotype
Previous work demonstrated that virion DNA from the mutant virus contained modestly elevated levels of uracil compared to the wt virus, which is a predicted phenotype [31]. Thus, it is possible that the delay in DNA synthesis simply reflects the time required to repair misincorporated uracil residues in the input viral genomes, and once this is accomplished, DNA synthesis proceeds normally. To test this hypothesis, a cell line that could complement the mutant virus in trans was constructed by methods described previously [32]. Virus stocks produced in the complementing cell line (HL114) were determined to possess normal levels of uracil, suggesting that the cell line was able to compensate for the deficiencies in the deletion mutant (data not shown). Thus, subsequent infection of HEL cells with these complemented virus stocks should reveal effects that are related to the genetic differences of the viruses, rather than the physical characteristics of the input genomes.
Complemented virus stocks were used to infect both HEL cells and HL114 cells at an MOI of 5 PFU/cell and kinetics of viral DNA synthesis were determined. In HEL cells, the mutant virus failed to induce detectable DNA synthesis at 72 hpi, whereas cells infected with repaired virus synthesized large quantities of viral DNA (Fig. 3). A similar result was obtained when uncomplemented virus stocks were used to infect these cells (data not shown). This suggested that the defect in DNA synthesis was likely related to a deficiency in pUL114 rather than the uracil content of the input viral genomes. As a control, both viruses were used to infect the complementing cells and both viruses produced similar quantities of DNA by 72, hpi, indicating that pUL114 supplied in trans could complement the observed defect in DNA synthesis. The complementation did not appear to be complete however, and there does appear to be a slight lag in DNA synthesis by the mutant virus. These results were confirmed by titering progeny virus at 96 hpi, when the mutant virus exhibits titers that are more than ten-fold lower than the parent virus in primary fibroblasts. Infection of complementing cells produced indistinguishable titers of both the mutant and restored viruses, while titers of the deletion virus were reduced more than ten-fold in primary fibroblasts (data not shown). Thus, the physical characteristic of the deletion mutant's genome appear to be unrelated to the observed phenotype and it appears more likely that the observed defects are due to a deficiency in pUL114 during the lytic replication cycle.
Figure 3 Kinetics of DNA synthesis and viral replication in complementing cells. Virus stocks of the parent virus and the mutant virus were produced in the complementing cell line (HL114) and used to infect either HEL cells or IHL114 cells at an MOI of 5 PFU/cell. Circular and square symbols represent quantities of DNA from RC2620 and the repaired virus respectively while solid and open symbols represent DNA isolated from HEL cells and HL114 cells respectively. The average of triplicate values are shown.
Construction of epitope tagged viruses
To investigate a potential role for pUL114 in viral DNA replication, it was necessary to characterize the expression and intracellular localization of this gene product during the replication cycle. Site directed mutagenesis in very large constructs is difficult to accomplish using standard techniques, so a rapid method for epitope tagging viral genes was developed. Homologous recombination in Saccharomyces cerevisae was conducted by methods similar to those described earlier in yeast artificial chromosomes [19]. A previous report described a method for recycling the KanMX selectable marker in yeast, through the induction of CRE recombinase that resulted in the loxP dependent excision of this marker. This construct was modified such that a precise deletion of the marker would yield an in frame 35 aa insertion including the ICP4 epitope tag. Amplification of pkanMX-ICP4 allowed the insertion of this epitope tag anywhere in the viral cosmid with primers containing 40 bp 5' extensions to target the desired locus in the DNA (Fig. 4). This technique was used to construct three cosmids in which UL114 was tagged at the amino (UL114NTAG) and the carboxyl (UL114CTAG) termini, as well as the precise replacement of UL114 with the 35 aa ORF containing the epitope tag (UL114KO) (Fig. 1). Resulting cosmids were used in a standard cotransfection to generate three tagged recombinant viruses by methods described previously [15].
Figure 4 Rapid epitope tagging strategy in yeast. The top line represents the target ORF in the context of a large yeast plasmid or YAC. Line 2 shows a PCR product containing the epitope tagging cassette with 40 bp targeting sequences homologous to the regions designated by the dashed lines. Line 3 shows the site-specific integration of the cassette resulting from homologous recombination in yeast. The final line represents UL114 in the YAC with an in frame 35 aa amino terminal insertion containing the ICP4 epitope and a single loxP site. This strategy can be used to place the epitope tag anywhere in the ORFs on the YAC by changing the targeting sequences on the PCR primers.
Localization to replication compartments and association with ppUL44
Previous work used immunofluorescence microscopy to examine the nature and distribution of CMV replication components at various times in the virus life cycle [29]. This work suggested that various members of the viral replication complex, including ppUL44, the DNA polymerase processivity factor, localize into specific replication compartments in patterns that are characteristic of a given point in the replication cycle. In light of the putative role of the UL114 gene product in viral DNA replication, similar studies were undertaken, using the epitope-tagged viruses described above to determine the location of pUL114 in infected fibroblasts. HEL cells were infected with the recombinant viruses and were examined by fluorescence microscopy using anti-ICP4 and anti-UL44 monoclonal antibodies. At 48 hpi, ppUL44 localized to the nucleus in small foci in a pattern that was very similar to that for pUL114 (Fig 5A–C). By 72 hpi, epitope tagged pUL114 expressed from the CTAG virus partitioned to the replication compartments within the nucleus as defined by ppUL44 staining (Fig. 5D–F) and light punctate cytoplasmic staining was also observed in some cells. The recombinant UL114 NTAG virus did not exhibit the strong nuclear localization observed with UL114 CTAG and it is possible that fusing the ICP4 epitope to this part of the molecule may have interfered with its normal localization (data not shown).
Figure 5 Localization of pUL114 in infected HEL cells. Cells were infected with a recombinant virus with an epitope tag in the carboxyl terminus of UL114. Monolayers were fixed and stained with an anti-ppUL44 monoclonal antibody (FITC) and an anti-ICP4 mouse monoclonal antibody (Texas Red). Cells were fixed at 48 hpi and images of FITC, Texas Red, and a merged image with DAPI are shown(A-C). Cells were fixed at 72 hpi and images of FITC, Texas Red, and a merged are shown (D-F). All images were captured digitally and prepared in Adobe Photoshop.
The localization pattern exhibited by the tagged versions of pUL114 suggested that it might be physically interacting with the viral DNA replication machinery. We hypothesized that pUL114 might interact with ppUL44 analogous to the UNG2 interaction with PCNA that occurs the human DNA replication complex [28]. Extracts of cells infected with the epitope tagged viruses and a wt virus were immunoprecipitated with a monoclonal antibody to ppUL44. Precipitated proteins were separated on denaturing polyacrylamide gels, transferred to nitrocellulose and a monoclonal antibody specific for the ICP4 epitope was used to detect the tagged pUL114 molecules. A protein with a predicted molecular weight of 32 kDa was specifically detected from the recombinant virus in which pUL114 was tagged at the carboxyl terminus (Fig. 6A). A very light band with the same migration rate was detected from UL114 NTAG-infected cells upon long exposure, consistent with its reduced localization to the nucleus. No specific species were detected in extracts prepared from the wt virus. The reverse experiment was performed with pUL114-EGFP fusion proteins that were precipitated with a monoclonal antibody specific for GFP and the monoclonal antibody to ppUL44 was used to detect the coprecipitated protein. This experiment confirmed the earlier result and demonstrated that it was also possible to specifically coprecipitate ppUL44 with pUL114 fusion proteins (Fig. 6B). Consistent with the previous result, the coprecipitation appeared to be less efficient for pUL114 labeled at the amino terminus.
Figure 6 Coprecipitation of pUL114 and ppUL44. (A) Primary foreskin fibroblast cells were infected either with ICP4-tagged recombinant viruses or Towne at an MOI of approximately 1 PFU/cell. Cells were lysed at 48 hpi, and extracts were immunoprecipitated with a monoclonal antibody to ppUL44 and separated on an SDS-PAGE gel. Proteins were transferred to a membrane and a monoclonal antibody to the ICP4 epitope was used to detect coprecipitated poteins in the immunoblot. (B) EGFP.373 and C1-114.373 cells were infected with AD169 at an MOI of 2 PFU/cell and harvested at 24 hpi. Fusion proteins were precipitated with a monoclonal antibody to EGFP, separated on non-denaturing SDS PAGE gels, transferred to nitrocellulose, and immmunoblotting was performed with monoclonal antibody to ppUL44. Arrows designate the specific bands.
To confirm these results, plasmids expressing ppUL44 (pMP62) and pUL114 with a carboxyl terminal EGFP tag were transfected into monolayers of primary foreskin fibroblast cells. In cells transfected with pMP62 alone, ppUL44 localized exclusively to the nucleus and is shown merged with DAPI image (Fig 7A), which was similar to the localization observed in infected cells early in infection. Cells expressing either the full length pUL114-EGFP fusion protein (pMP39), or the fusion protein in which aa 3–24 were deleted from pUL114 (pMP41) exhibited punctate cytoplasmic fluorescence (Fig 7B, C). This localization pattern was distinct from the nuclear staining observed with the UL114 CTAG recombinant virus. However, when ppUL44 and full length pUL114 fusion proteins were coexpressed in the same cell, pUL114 was recruited to the nucleus with ppUL44 (Fig 7D–F), consistent with its nuclear localization in the context of infected cells. A small quantity of ppUL44 also appeared to localize to a subset of the cytoplasmic punctae containing pUL114. Deletion of aa 3–24 from the pUL114 fusion protein eliminated its recruitment to the nucleus by ppUL44, suggesting that this domain is required for the interaction ppUL44 (Fig 7G–I). This interpretation of the data is consistent with the impaired nuclear localization observed with UL114 NTAG-infected cells, in which the amino terminal domain of pUL114 was altered through the addition of the ICP4 epitope tag (data not shown). Also consistent with this result, is the inefficient coprecipitation of ppUL44 with pUL114 fusion proteins when the tags were fused to the amino terminus (Fig. 6). These data suggest that these proteins associate in a manner that is dependent on aa 3–24 of pUL114, and independent of other viral proteins or viral DNA. These experiments do not, however, eliminate the possibility that they might associate in an indirect manner through cellular proteins.
Figure 7 Recruitment of pUL114 to the nucleus by ppUL44. Plasmids expressing ppUL44 or pUL114-EGFP fusion proteins were transfected into primary fibroblast cells and visualized by immunofluorescent staining. In the first row of images, ppUL44 stained with Texas Red exhibited strong nuclear localization as evidenced by the colocalization with DAPI in the merged image (violet). The pUL114-EGFP fusion protein and a similar protein containing a 25 aa amino terminal deletion (green) both localized to the cytoplasm and are shown merged with DAPI staining (blue). In the second row of images, the coexpression of ppUL44 (Texas red), pUL114-EGFP (green) and a merged image show that ppUL44 can recruit pUL114 to the nucleus. In third row of images ppUL44 (Texas red) and pUL114-EGFP containing a 25 aa amino terminal deletion (green) did not colocalize to the nucleus when co-expressed.
Characterizing the defect in DNA synthesis
The localization of pUL114 to replication compartments, and its apparent association with ppUL44, which is known to interact with the DNA polymerase [9] imply that this molecule is part of the viral DNA replication complex. This interpretation of the data is consistent with the observed phenotype of delayed DNA synthesis in the UL114 deletion virus [6,31], and is also consistent with results reported for the human UNG2 that has been shown to localize to replication complexes [28]. If this assumption is correct and the viral UNG is an important part of the replication complex, then the defect in viral DNA synthesis should be apparent throughout the viral DNA replication process. To characterize the affect of pUL114 on DNA synthesis, triplicate monolayers of replicating primary foreskin fibroblasts were infected with either Towne, or an isogenic recombinant virus without UL114 and the accumulation of viral DNA was quantified with a TaqMan-based assay. Input copy number following infection was determined at 2 hpi and yielded average values of 4.2 × 104 and 2.3 × 104, for the wt and mutant viruses respectively with standard deviations of <15% for both values. During the course of infection, genome copy number was determined in total DNA and the data were normalized relative to the input copy number (Fig. 8). During the first 18 h of infection, copy number of the wt and deletion virus genomes decreased at the same rate with a half-life of approximately 8 h (Fig 8B). This is consistent with data presented earlier, which suggested that increased uracil levels did not substantially affect genomic integrity and were unlikely to be responsible for the observed defects in DNA synthesis. This analysis also revealed two features of the defect in DNA synthesis. First, the accumulation rate of viral DNA synthesis was significantly reduced in the recombinant virus with a deletion in UL114 (Fig. 8A). A 7-fold increase in copy number was attained in the parent virus at 25 hpi, but this same level was not achieved in the mutant until 48 hpi. By this time, the wt virus had attained a 300-fold amplification of the input genome, which was not attained by the mutant even after an additional 48 h of incubation. Exponential growth rates were calculated from curves fitted to the experimental data for both viruses. The wt rate (r) was determined to be approximately 0.2 h-1, whereas the copy number of the mutant expanded at a rate of about 0.1 h-1. This decreased rate of DNA accumulation is consistent with the observed decrease in viral DNA described previously [31] and also with the data showing a defect in the transition to late phase DNA synthesis reported recently by Courcelle et al. [6]. A second defect in DNA synthesis was also observed. The initial doubling of the wt genome was detected at 21 hpi and the copy number increased exponentially to a 7-fold increase by 25 hpi (Fig. 8B). During this period of time, no increase in the copy number of mutant virus genomes was observed. Thus, the initial phase of DNA synthesis also appears to be compromised in the absence of pUL114, despite the fact that early genes are expressed at normal levels at this point in time [29,31]. If viral DNA synthesis in the mutant had initiated at the same time as the parent virus, the increased copy number should have been easily detectable by 25 hpi, even at the reduced rate of accumulation we report here. Thus, either the initiation or the early theta-type DNA replication postulated for this family of viruses appears to be compromised in absence of pUL114. These data suggest that pUL114 acts during both the onset and the subsequent expansion phase of viral DNA synthesis and suggests that this gene product functions as part of the viral DNA replication machinery.
Figure 8 Defects in DNA synthesis associated with pUL114. Triplicate wells of HEL cells were infected at an MOI of 0.01 with Towne (black circles) or the isogenic deletion virus, UL114 KO tag, (shaded squares). Total DNA was harvested at the indicated times, and the genome copy number was determined with a TaqMan assay using a standard curve of virion DNA. Copy number was normalized to the quantity of input genomes determined at 2 hpi with error bars representing the standard deviation of the triplicate samples. (A) The log of the accumulated viral DNA copy number is shown versus time post infection. The wt exponential rate of accumulation (r) was determined to be approximately 0.2 h-1, whereas the copy number of the mutant expanded only at a rate of about 0.1 h-1. (B) Data for the first 24 h replotted on a linear scale show the delayed onset of DNA synthesis during the first duplication of the viral genome.
Discussion
Perhaps the simplest explanation of the observed phenotype associated with UL114 deletion viruses is that the recombinant virus fails to remove uracil residues from its genome and that these lesions decrease genome stability and impede DNA synthesis. Two lines of evidence argue against this interpretation of the data. First, input genomes of the recombinant virus in infected cells appeared to be as stable as the wt genomes in infected cells and had similar initial half lives (Fig 8B). Second, the complementing cell line reduced the uracil content of the mutant genomes to levels indistinguishable from the parent virus, yet the observed phenotype of these complemented virus stocks in non-complementing cells was unaffected. Thus, it appears that the viral UNG plays a more direct role in the synthesis of viral DNA. However, these data do not exclude the possibility that the removal of uracil may be important late in infection. We suggest that the HCMV UNG2 homolog functions as part of the DNA replication machinery and that it significantly accelerates the synthesis of genomic DNA.
The parallels between this system and the recent results for human UNG2 are striking. PCNA and ppUL44 are thought to perform a similar function and associate with human DNA polymerase δ and the HCMV DNA polymerase, respectively. Despite the fact that these processivity factors do not share significant aa sequence homology and exhibit different 3-D structures [1], they retain interactions with their respective DNA polymerases [21], as well as an association with their respective UNG homologs. The fact that the amino terminal domains of both pUL114 and UNG2 are required to mediate these interactions suggests that this might be a common feature among all UNG2 homologs. This relationship is also conserved in vaccinia virus where the viral UNG2 homolog (D4R) was shown to physically associate with the A20R DNA polymerase processivity factor [14]. In this system, the viral UNG was shown to be essential for viral DNA synthesis, and this requirement was unrelated to the ability of the molecule to excise uracil [7]. A potential role for UNG in DNA replication was also noted in Epstein Barr Virus where the UNG2 homolog (BKRF3) increased the efficiency of replication of a transfected plasmid containing the origin of replication [10] and was absolutely required when the core essential genes were supplied on a set of cosmid clones [11]. Less analogous but equally compelling, is the recruitment of UNG2 to the preintegration complex in HIV and its specific interaction with both the integrase as well as the reverse transcriptase [33]. The conserved relationship between UNG2 homologs and DNA replication complexes in these diverse systems suggests that it performs a conserved function in mammals. It is unclear if this function is related to UNG enzymatic activity, and it is likely that these molecules perform an additional function replication that remains uncharacterized. This view is supported by the fact that the UNG enzymatic activity can be eliminated without severely affecting the replication of vaccinia virus, whereas larger mutations are lethal [7]. A specialized role for UNG2 has also been proposed in mammalian systems since UNG-/UNG- mice are viable and do not exhibit the phenotype of highly elevated mutation frequency that would be predicted by earlier studies in prokaryotes and Sacharomyces. Information garnered in future studies with HCMV will be particularly helpful in shaping our understanding of the function of UNG2 in the DNA replication foci of mammalian cells. The unique phenotype associated with pUL114 in HCMV infection and the fact that this simple system closely resembles that in humans make it an attractive system to probe the unique function of mammalian UNG2 homologs in DNA synthesis.
In HSV, the deletion of the UNG homolog (UL2) affects the ability of the virus to replicate in mice, particularly the CNS. The deletion of UL2 resulted in a 100,000-fold reduction in the neuroinvasiveness and may represent a potential attenuating mutation in candidate vaccines [34] Previous studies with UNG deletion mutants in HSV were not shown to affect replication in tissue culture, they replicated to lower titers in vivo and were orders of magnitude less neuroinvasive than control viruses [34]. To investigate the possibility that the phenotype might be more pronounced in vivo, we infected human fetal retinal tissue implanted in a SCID-hu mouse [2,16]. In this model, a deficiency in pUL114 resulted in a decreased infection rate (P = 0.015) as well as significantly reduced titers in infected animals (P = 0.0063). However, the observed defects in vivo were not more pronounced that the replication defects in cell culture and were not similar to results observed with HSV.
Conclusion
The work presented here suggests that pUL114 is part of the DNA replication machinery and that it significantly accelerates the synthesis of genomic DNA. This interpretation of the data is consistent with the early expression kinetics and the nuclear localization exhibited by this molecule in infected cells, which are both predicted characteristics of an enzyme presumed to act in DNA repair. Equally consistent is the observed intranuclear localization to viral replication compartments at a time when viral DNA synthesis is known to occur [29]. The fact that pUL114 appears to associate with ppUL44 is intriguing, because of the central role that ppUL44 plays in the synthesis of viral DNA [9,20,21]. These data taken together with the observed defects in the onset and expansion of viral DNA synthesis suggest that it functions as part of the DNA replication machinery.
We propose a model in which pUL114 functions as part of the viral DNA polymerase complex and is required for the efficient establishment and expansion of viral DNA synthesis. Results presented here suggest that the performance of the DNA replication machinery is significantly impaired without pUL114. The precise mechanism that this molecule uses to affect DNA synthesis is unclear but it may or may not be related to its ability to excise uracil from DNA. The interaction with ppUL44 suggest that this molecule might be close to the replication forks where it might help destabilize double stranded DNA through a scanning and pinching base flipping mechanism similar to that described for the human homolog [12]. Additional experiments in this system will be required to determine the correlation between uracil excision activity and the efficiency of viral DNA replication.
The evolving view of UNG function in the life cycle of viruses increases its appeal as a target for antiviral chemotherapy, particularly in poxviruses where it is essential for virus replication. This approach may also be valuable in herpesviruses given its proximity to the replication complex as well as its important role in vivo. It is certainly possible to obtain specific inhibitors of viral UNG molecules based on their ability to block the enzyme's ability to excise uracil, however at present, it is unclear that this enzymatic activity is responsible for the interesting affects observed both in vitro and in vivo. Rational drug strategies should be possible, but their development is dependent upon a better understanding of the biological functions of this molecule in virus replication.
Methods
Plasmids
Construction of pON2619 and pON2620 were described previously [31]. To construct a retroviral vector, a 1782 bp EcoRI fragment (coordinates 163071 to 164853 AD169 genome) containing the UNG open reading frame was inserted into the MfeI site in pLXIN (Clontech, Palo Alto, CA) to yield pON2159. EGFP fusion constructs were constructed by amplifying an 800 bp DNA fragment containing the UL114 open reading frame using the forward primer 5'-GGA CTC AGA TCT ATG GCC CTC AAG CAG TGG ATG-3' and the reverse primer 5'-GTC GAC TGC AGA GAA TCT CCC ACA GAG TCG CCA GTC C-3'. The resulting fragment was purified from an agarose gel and cloned into the Bgl II and Pst I sites of pEGFPC1 and pEGFPN3 (Clontech, Palo Alto, CA) to generate plasmids pEGFPC1/UL114 and pEGFPN3/UL114. Plasmids pMP39 and pMP41 were constructed by amplifying with forward primers 5'-ATG GCC CTC AAG CAG TGG-3' and 5'-ATG GCC GCT CGC GTG TTT TGT CTG AGC-3' respectively, with reverse primer 5'-TCA TCT GAG TCC GGA CTT GTA CA-3' using pEGFPN3/UL114 as a template and cloning into pcDNA3.1. The resulting plasmids were sequenced and express proteins of the predicted molecular weight. The UL114 open reading frame in pMP41 contains a deletion of aa 3 to 24. Primers 5'-CAC CAT GGA TCG CAA GAC GCG C-3' and 5'-CTA GCC GCA CTT TTG CTT CT-3' were used to amplify UL44 and the PCR product was cloned into a eukaryotic expression vector to yield pMP62. The plasmid pUG6 contains a recyclable genetic marker for site directed mutagenesis in yeast [13]. This cassette was amplified with the forward primer 5'-CAG GTC GAC AAC CCT TAA TAT AAC TTC GTA TAA TGT ATG CTA TAC GAA GTT ATT AGG TCT AGA GAT CTG TTT AGC TTG C-3' and the reverse primer 5'-TCC TGG AGC TCG ATC TCC TGC TGC ATC TGC TGC ATC ATC ATA TTC ATC ACC TAA TAA CTT CGT ATA GCA TAC ATT ATA CGA AGT TAT ATT AAGGGT TCT CG-3'. The resulting product was TOPO-cloned into pcDNA3.1 (Invitrogen, Carlsbad, CA) to yield pKan-ICP4 and used as a template for subsequent amplifications. The Sma I – Sca I fragment of pRS413, which contains ARS4, CEN6 and the HIS3 selectable marker, was ligated into the XmnI site of pACYC184. A single EcoRI site in the resulting intermediate construct was converted to a unique PacI site by ligation to EcoRI PacI adapters to produce pACYC ars cen. PacI fragments from cosmids described previously [15] were cloned in the PacI site for subsequent experiments.
Mutagenesis in Yeast
PCR products for site directed mutagenesis were generated using the forward primer 5'-AGG TCG ACA ACC CTT AAT ATA ACT-3' and reverse primer 5'-TCC TGG AGC TCG ATC TCC TGC TGC AT-3' and were targeted by adding 40 bp of homologous sequence to the 5' end of each primer. PCR products were cotransformed by a standard lithium acetate protocol with target viral cosmids in Saccharomyces cerevisiae strain CGY2570 carrying plasmid pSH47 [13] which expresses CRE recombinase under control of the GAL promoter. Recombinants were selected on yeast complete medium plates containing 400 μg/ml G418 and the selectable marker was excised through the galactose-dependent expression CRE recombinase to yield 35 aa in frame insertions containing the HSV ICP4 epitope tag.
Cells and virus
Primary human foreskin fibroblast (HFF) cells and human embryonic lung (HEL) cells were grown in monolayer cultures in Dulbecco's modified Eagle medium (Gibco BRL, Gaithersberg, MD) supplemented with 100 units/ml penicillin G, 100 μg/ml streptomycin sulfate and 10% fetal bovine serum (FBS). Parental virus (AD169) was obtained from the ATCC and virus stocks were obtained and titered as described previously [40]. Cell lines expressing EGFP fusion proteins were constructed by transfecting 10 μg of linearized pEGFPC1/UL114, pEGFPN3/UL114 or pN3EGFP inU373 cells with Lipofectin (Gibco BRL, Gaithersberg, MD) according to the manufacturers recommendations. Stably transfected C1-114.373, N3-114.373 and EGFP.373 cells were selected with 1 mg/ml G418, and resulting colonies were isolated and frozen at passage 5. The construction and propagation of RC2620 as well as the production of high MOI growth curves were described previously [31]. The construction of RQ2620 was performed as described previously [31] and the resulting repaired virus was plaque purified 3 times after it was shown to be free of the contaminating parent virus by Southern analysis. Epitope tagged viruses were constructed by cotransfecting a set of 8 cosmids derived from the Towne strain of HCMV [15], including one cosmid that was subjected to site directed mutagenesis in yeast as described above. The epitope tagged viruses replicated to high titers in HFF cells and do not appear to be replication impaired.
Construction of HL114 cells
pON2159 was tranfected into PA317 cells to produce defective retrovirus stocks that were subsequently used to transduce the UL114 gene into low passage primary HEL by methods described previously [32]. Transduced cells were selected with 400 μg/ml G418 starting at 24 hpi. Surviving cells were passaged in G418 and were used as a mixed population.
DNA synthesis kinetics
Confluent monolayers of HFF cells in 6-well cluster dishes were infected at an MOI of 5 PFU/cell with either the AD169, RC2620, or RQ2620. Total DNA was extracted as described previously [31], diluted and transferred to a Hybond N+ membrane in a dot-blot manifold and probed with a plasmid containing viral sequences (89797-94860 in the AD169 genome). The resulting film was captured digitally and quantified with Scan Analysis (Biosoft, Cambridge, UK).
Determination of genome copy number
Towne and an isogenic recombinant virus containing a deletion in UL114 were used to infect dividing HFF cells at an MOI of 0.02 PFU/cell in 12-well plates. Monolayers were rinsed three times at 2 hpi and supplemented with fresh media. At harvest, monolayers were rinsed twice with fresh media, and frozen at -80°C in a final volume of 0.2 ml. Total DNA was extracted using a QIAamp DNA blood minikit according to the manufacturers recommendations (Qiagen, Valencia, CA). Copy number was determined using the ABI PRISM 7700 sequence detection system and TaqMan Universal PCR Master Mix. Forward primer (50 nM), 5'-CCG AGG TGG GTT ACT ACA ACG-3', reverse primer (300 nM), 5'-GGA AGG GTA GAG GCT GGC A-3', and fluorogenic probe (75 nM), 5' FAM-CCC CGT GGC CGT GTT CGA CT-3' TAMRA were used in a 50 μl reaction volume with conditions as follows: 2 min at 50 ?C, 10 min at 95 ?C, and 40 cycles of (15 sec at 95 ?C, 1 min at 60 ?C,). The DNA templates from triplicate wells were analyzed in a volume of 5 μl per reaction. Copy number was compared to a standard curve generated from HCMV genomic DNA.
Immunofluorescence microscopy
Immunofluorescence staining was performed as previously described [37]. 8-well chamber slides of confluent HFF cells (LF1043) were infected with recombinant HCMV strains UL114 NTAG, UL114 CTAG, and Towne (control) at an MOI of 0.5 PFU/cell. Slides were washed once with PBS, fixed with 1% formalin for 15 min, and washed again three times with PBS + 0.2% BSA. Permeablization was performed in PBS with 0.2% Triton X-100 for 15 min, followed by one wash step with PBS + 0.2% BSA and a blocking step in 5% normal horse serum (Vector Labs, Burlingame, CA) also for 15 min. Monolayers of HFF cells on coverslips were transfected with Lipofectamine 2000 according to the manufacturer's protocol (Invitrogen). Transfected cells were fixed and permeabilized by the same methods described above. Monoclonal antibodies to ICP4 (Rumbaugh-Goodwin Institute, Plantation, FL), or ppUL44 (gift from Bill Britt, University of Alabama at Birmingham) were incubated with cells for 1 h at 37°C. Monolayers were washed and incubated with a goat anti-mouse secondary antibody conjugated to FITC (Southern Biotechnology Associates, Birmingham, AL). For dual labeling experiments, monoclonal antibodies were directly labeled with a Zenon Texas Red labeling kit as per manufacturer's directions (Molecular Probes, Eugene, OR). After three washes, Vectashield mounting medium containing DAPI (Vector Labs, Burlingame, CA) was added to each slide along with a glass coverslip. Cells were examined with a Nikon TE2000 Microscope using a 40 × objective. Fluorescence images of stained cells were captured with Hamamatsu ORCFA-100 digital camera and recorded using Simple PCI software. All photographs were prepared in Adobe Photoshop CS.
Immunoprecipitations
T-25 flasks or 6-well dishes of confluent HFF cells, C1-114.373, N3-114.373, or EGFP.373 cells were infected with AD169, Towne, UL114 CTAG, or UL114 NTAG at an MOI of 5 PFU/cell. At 24 and 48 hpi, the monolayers were washed with PBS and lysed on ice in 500 μl of lysis buffer containing 50 mM HEPES, pH 7.5, 150 mM NaCl, 2.5 mM EGTA, 1 mM EDTA, 1% Triton X-100, 1 mM PMSF, and 1 proteinase inhibitor tablet (Boehringer Mannheim). Cell lysates were preadsorbed for 1 h at 4°C with 30 μl of a 50% suspension of protein A-Sepharose in lysis buffer. Proteins were precipitated with 30 μl of the protein A-Sepharose suspension, and 1 μl of a monoclonal antibody to the EGFP domain (Clontech, Palo Alto, CA), the ICP4 epitope tag, or ppUL44 (Rumbaugh-Goodwin Institute, Plantation, FL). The tubes were rocked overnight at 4°C. The protein A-Sepharose beads were washed twice in lysis buffer and once in lysis buffer with the addition of 0.1% SDS, and 1% sodium deoxycholate (RIPA). Samples were boiled 5 min in 80 mM Tris, pH 6.8, 2% SDS, 10% sucrose, and 0.004% bromophenol blue and the proteins were separated by SDS polyacrylamide electrophoresis (48) and transfered to an Immobilon-P membrane (Millipore) at 200 mA for 1 h. Blots were blocked at room temperature for 30 min with 5% (w/v) skim milk in 50 mM Tris, pH7.5, 0.2 M NaCl, 0.01% Tween 20, and incubated at room temperature for 1 h with a monoclonal antibody specific for ppUL44, ICP4, or EGFP diluted 1:1000 in washing buffer (1% (w/v) skim milk, 50 mM Tris, pH7.5, 0.2 M NaCl, 0.01% Tween 20). Blots were washed three times with washing buffer, and incubated at room temperature for 1 h with anti-Mouse IgG(H+L) HRP conjugate (New England Biolabs) diluted 1:2500 in washing buffer. The unbound secondary antibody was removed in three washes with 50 mM Tris, pH7.5, 0.2 M NaCl. The membrane was subsequently developed with LumiGLO or ECL according to the manufacturer's recommendations and used to expose Kodak biomax film.
List of Abbreviations
HCMV human cytomegalovirus
HIV human immunodeficiency virus
dUTP deoxyuridine triphosphate
RT reverse transcriptase
HEL human embryonic lung
UNG uracil DNA glycosylase
PCNA proliferating cell nuclear antigen
EGFP green fluorescent protein
SCID severe combined immunodeficiency
FITC fluoroscein isothiocyanate
MEM minimal essential medium
TR texas red
DNA deoxyribonucleic acid
Competing interests
Some of the authors of this publication are supported financially by salary and shares of MedImmune Inc. The authors declare that they have no other competing interest.
Authors' contributions
MNP generated the recombinant viruses and cell lines described here, performed the complementation experiments and made the yeast mutagenesis constructs. HL performed the immunoprecipitations. GMD and MD worked on the yeast mutagenesis system and assisted in the production of recombinant viruses. CM worked on the immunofluorescence studies. ZW performed the TaqMan analyses. GK and ERK provided critical intellectual input.
Acknowledgements
We thank Giesela Mosig, and Charmain Tan Courcelle for helpful discussions and William J. Britt for his gift of monoclonal antibodies and critical reading of the manuscript. This work was supported in part by the contract NO1-AI-30049 from the National Institute for Allergy and Infectious Diseases.
==== Refs
Brynolf K Eliasson R Reichard P Formation of Okazaki fragments in polyoma DNA synthesis caused by misincorporation of uracil Cell 1978 13 573 580 207436 10.1016/0092-8674(78)90330-6
Tye BK Lehman IR Excision repair of uracil incorporated in DNA as a result of a defect in dUTPase J Mol Biol 1977 117 293 306 342701 10.1016/0022-2836(77)90128-0
Chen R Wang H Mansky LM Roles of uracil-DNA glycosylase and dUTPase in virus replication J Gen Virol 2002 83 2339 2345 12237414
Percival KJ Klein MB Burgers PM Molecular cloning and primary structure of the uracil-DNA-glycosylase gene from Saccharomyces cerevisiae J Biol Chem 1989 264 2593 2598 2644266
Varshney U Hutcheon T van de Sande JH Sequence analysis, expression, and conservation of Escherichia coli uracil DNA glycosylase and its gene (ung) J Biol Chem 1988 263 7776 7784 2836397
Yang H Chiang JH Fitz-Gibbon S Lebel M Sartori AA Jiricny J Slupska MM Miller JH Direct interaction between uracil-DNA glycosylase and a proliferating cell nuclear antigen homolog in the crenarchaeon Pyrobaculum aerophilum J Biol Chem 2002 277 22271 22278 11927597 10.1074/jbc.M201820200
Krokan HE Otterlei M Nilsen H Kavli B Skorpen F Andersen S Skjelbred C Akbari M Aas PA Slupphaug G Properties and functions of human uracil-DNA glycosylase from the UNG gene Prog Nucleic Acid Res Mol Biol 2001 68 365 386 11554311
Muller SJ Caradonna S Isolation and characterization of a human cDNA encoding uracil-DNA glycosylase Biochim Biophys Acta 1991 1088 197 207 2001396
Nilsen H Haushalter KA Robins P Barnes DE Verdine GL Lindahl T Excision of deaminated cytosine from the vertebrate genome: role of the SMUG1 uracil-DNA glycosylase Embo J 2001 20 4278 4286 11483530 10.1093/emboj/20.15.4278
Nilsen H Otterlei M Haug T Solum K Nagelhus TA Skorpen F Krokan HE Nuclear and mitochondrial uracil-DNA glycosylases are generated by alternative splicing and transcription from different positions in the UNG gene Nucleic Acids Res 1997 25 750 755 9016624 10.1093/nar/25.4.750
Nagelhus TA Haug T Singh KK Keshav KF Skorpen F Otterlei M Bharati S Lindmo T Benichou S Benarous R Krokan HE A sequence in the N-terminal region of human uracil-DNA glycosylase with homology to XPA interacts with the C-terminal part of the 34-kDa subunit of replication protein A J Biol Chem 1997 272 6561 6566 9045683 10.1074/jbc.272.10.6561
Otterlei M Warbrick E Nagelhus TA Haug T Slupphaug G Akbari M Aas PA Steinsbekk K Bakke O Krokan HE Post-replicative base excision repair in replication foci Embo J 1999 18 3834 3844 10393198 10.1093/emboj/18.13.3834
Simbulan-Rosenthal CM Rosenthal DS Hilz H Hickey R Malkas L Applegren N Wu Y Bers G Smulson ME The expression of poly(ADP-ribose) polymerase during differentiation-linked DNA replication reveals that it is a component of the multiprotein DNA replication complex Biochemistry 1996 35 11622 11633 8794742 10.1021/bi953010z
Bouhamdan M Benichou S Rey F Navarro JM Agostini I Spire B Camonis J Slupphaug G Vigne R Benarous R Sire J Human immunodeficiency virus type 1 Vpr protein binds to the uracil DNA glycosylase DNA repair enzyme J Virol 1996 70 697 704 8551605
Selig L Benichou S Rogel ME Wu LI Vodicka MA Sire J Benarous R Emerman M Uracil DNA glycosylase specifically interacts with Vpr of both human immunodeficiency virus type 1 and simian immunodeficiency virus of sooty mangabeys, but binding does not correlate with cell cycle arrest J Virol 1997 71 4842 4846 9151883
Willetts KE Rey F Agostini I Navarro JM Baudat Y Vigne R Sire J DNA repair enzyme uracil DNA glycosylase is specifically incorporated into human immunodeficiency virus type 1 viral particles through a Vpr-independent mechanism J Virol 1999 73 1682 1688 9882380
Priet S Navarro JM Gros N Querat G Sire J Functional role of HIV-1 virion-associated uracil DNA glycosylase 2 in the correction of G:U mispairs to G:C pairs J Biol Chem 2003 278 4566 4571 12458223 10.1074/jbc.M209311200
Klarmann GJ Chen X North TW Preston BD Incorporation of uracil into minus strand DNA affects the specificity of plus strand synthesis initiation during lentiviral reverse transcription J Biol Chem 2003 278 7902 7909 12458216 10.1074/jbc.M207223200
Millns AK Carpenter MS DeLange AM The vaccinia virus-encoded uracil DNA glycosylase has an essential role in viral DNA replication Virology 1994 198 504 513 8291232 10.1006/viro.1994.1061
Stuart DT Upton C Higman MA Niles EG McFadden G A poxvirus-encoded uracil DNA glycosylase is essential for virus viability J Virol 1993 67 2503 2512 8474156
Upton C Stuart DT McFadden G Identification of a poxvirus gene encoding a uracil DNA glycosylase Proc Natl Acad Sci U S A 1993 90 4518 4522 8389453
Ellison KS Peng W McFadden G Mutations in active-site residues of the uracil-DNA glycosylase encoded by vaccinia virus are incompatible with virus viability J Virol 1996 70 7965 7973 8892920
De Silva FS Moss B Vaccinia virus uracil DNA glycosylase has an essential role in DNA synthesis that is independent of its glycosylase activity: catalytic site mutations reduce virulence but not virus replication in cultured cells J Virol 2003 77 159 166 12477821 10.1128/JVI.77.1.159-166.2003
Mullaney J Moss HW McGeoch DJ Gene UL2 of herpes simplex virus type 1 encodes a uracil-DNA glycosylase J Gen Virol 1989 70 ( Pt 2) 449 454 2567340
Prichard MN Duke GM Mocarski ES Human cytomegalovirus uracil DNA glycosylase is required for the normal temporal regulation of both DNA synthesis and viral replication J Virol 1996 70 3018 3025 8627778
Reddy SM Williams M Cohen JI Expression of a uracil DNA glycosylase (UNG) inhibitor in mammalian cells: varicella-zoster virus can replicate in vitro in the absence of detectable UNG activity Virology 1998 251 393 401 9837803 10.1006/viro.1998.9428
Pyles RB Thompson RL Evidence that the herpes simplex virus type 1 uracil DNA glycosylase is required for efficient viral replication and latency in the murine nervous system J Virol 1994 68 4963 4972 8035495
Penfold ME Mocarski ES Formation of cytomegalovirus DNA replication compartments defined by localization of viral proteins and DNA synthesis Virology 1997 239 46 61 9426445 10.1006/viro.1997.8848
Courcelle CT Courcelle J Prichard MN Mocarski ES Requirement for uracil-DNA glycosylase during the transition to late-phase cytomegalovirus DNA replication J Virol 2001 75 7592 7601 11462031 10.1128/JVI.75.16.7592-7601.2001
Prichard MN Gao N Jairath S Mulamba G Krosky P Coen DM Parker BO Pari GS A recombinant human cytomegalovirus with a large deletion in UL97 has a severe replication deficiency J Virol 1999 73 5663 5670 10364316
Larionov V Kouprina N Solomon G Barrett JC Resnick MA Direct isolation of human BRCA2 gene by transformation-associated recombination in yeast Proc Natl Acad Sci U S A 1997 94 7384 7387 9207100 10.1073/pnas.94.14.7384
Kemble G Duke G Winter R Spaete R Defined large-scale alterations of the human cytomegalovirus genome constructed by cotransfection of overlapping cosmids J Virol 1996 70 2044 2048 8627734
Ertl PF Powell KL Physical and functional interaction of human cytomegalovirus DNA polymerase and its accessory protein (ICP36) expressed in insect cells J Virol 1992 66 4126 4133 1318399
Appleton BA Loregian A Filman DJ Coen DM Hogle JM The cytomegalovirus DNA polymerase subunit UL44 forms a C clamp-shaped dimer Mol Cell 2004 15 233 244 15260974 10.1016/j.molcel.2004.06.018
Loregian A Appleton BA Hogle JM Coen DM Specific residues in the connector loop of the human cytomegalovirus DNA polymerase accessory protein UL44 are crucial for interaction with the UL54 catalytic subunit J Virol 2004 78 9084 9092 15308704 10.1128/JVI.78.17.9084-9092.2004
Ishii K Moss B Role of vaccinia virus A20R protein in DNA replication: construction and characterization of temperature-sensitive mutants J Virol 2001 75 1656 1663 11160663 10.1128/JVI.75.4.1656-1663.2001
Fixman ED Hayward GS Hayward SD Replication of Epstein-Barr virus oriLyt: lack of a dedicated virally encoded origin-binding protein and dependence on Zta in cotransfection assays J Virol 1995 69 2998 3006 7707526
Fixman ED Hayward GS Hayward SD trans-acting requirements for replication of Epstein-Barr virus ori-Lyt J Virol 1992 66 5030 5039 1321285
Pyles RB Sawtell NM Thompson RL Herpes simplex virus type 1 dUTPase mutants are attenuated for neurovirulence, neuroinvasiveness, and reactivation from latency J Virol 1992 66 6706 6713 1328686
Bidanset DJ Rybak RJ Hartline CB Kern ER Replication of human cytomegalovirus in severe combined immunodeficient mice implanted with human retinal tissue J Infect Dis 2001 184 192 195 11424017 10.1086/322015
Kern ER Rybak RJ Hartline CB Bidanset DJ Predictive efficacy of SCID-hu mouse models for treatment of human cytomegalovirus infections Antivir Chem Chemother 2001 12 Suppl 1 149 156 11594682
Loregian A Appleton BA Hogle JM Coen DM Residues of human cytomegalovirus DNA polymerase catalytic subunit UL54 that are necessary and sufficient for interaction with the accessory protein UL44 J Virol 2004 78 158 167 14671097 10.1128/JVI.78.1.158-167.2004
Fuxreiter M Luo N Jedlovszky P Simon I Osman R Role of base flipping in specific recognition of damaged DNA by repair enzymes J Mol Biol 2002 323 823 834 12417196 10.1016/S0022-2836(02)00999-3
Guldener U Heck S Fielder T Beinhauer J Hegemann JH A new efficient gene disruption cassette for repeated use in budding yeast Nucleic Acids Res 1996 24 2519 2524 8692690 10.1093/nar/24.13.2519
Spaete RR Mocarski ES Regulation of cytomegalovirus gene expression: alpha and beta promoters are trans activated by viral functions in permissive human fibroblasts J Virol 1985 56 135 143 2993644
Sakakibara A Furuse M Saitou M Ando-Akatsuka Y Tsukita S Possible involvement of phosphorylation of occludin in tight junction formation J Cell Biol 1997 137 1393 1401 9182670 10.1083/jcb.137.6.1393
|
16022730
|
PMC1185570
|
CC BY
|
2021-01-04 16:39:00
|
no
|
Virol J. 2005 Jul 15; 2:55
|
utf-8
|
Virol J
| 2,005 |
10.1186/1743-422X-2-55
|
oa_comm
|
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-481602951710.1186/1477-7819-3-48ResearchManagement of renal cell carcinoma with solitary metastasis Thyavihally Yuvaraja B [email protected] Umesh [email protected] Ravichand S [email protected] Srinivas G [email protected] Hemant B [email protected] Department of Genito-urinary oncology, Tata Memorial Hospital, Mumbai, India2 Radiation Oncology, Tata Memorial Hospital, Mumbai, India3 Preventive oncology, Tata Memorial Hospital, Mumbai, India2005 20 7 2005 3 48 48 20 12 2004 20 7 2005 Copyright © 2005 Thyavihally et al; licensee BioMed Central Ltd.2005Thyavihally 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
Distant metastasis are common in Renal cell carcinoma (RCC) nearly one forth of the patients have metastasis at presentation while another 50% develop metastasis during the follow-up. A small percentage of these are solitary metastasis. We describe survival after surgical excision or radiotherapy of solitary metastatic lesion from renal cell carcinoma
Patients and methods
Between 1988–2001, 43 patients with solitary metastasis to different sites from renal cell carcinoma underwent either surgical excision or radiotherapy were analyzed. The solitary nature of the lesions was confirmed by investigations. All patients have had radical nephrectomy for the primary lesion. Survival analysis was carried out by Kaplan Meier Method.
Results
All solitary metastatic lesions were treated with intent of cure either by excision or radiotherapy. Of these, 13 patients had solitary metastasis at the time of presentation in whom 3-year overall median survival was 26 months. The survival of those who developed solitary metastases during follow-up after nephrectomy for primary was 45 months. The patients with long interval between diagnosis and development of metastasis, early stage and low grade of the primary tumor had better prognosis.
Conclusion
Complete resection of either synchronous or metachronous solitary metastases from renal cell carcinoma is justified and can contribute to a long-term survival in this select group of patients.
==== Body
Background
Nearly 20–25% of patients with renal cell carcinoma (RCC) have distant metastasis at presentation. Another 50% develop metastasis or local recurrence during follow-up after the treatment of the primary [1]. RCC can recur at any time after nephrectomy and usually metastasizes via venous and lymphatic routes. Frequent sites of metastasis include the lung parenchyma (50% to 60%), [2] bone (30% to 40%), [3] liver (30% to 40%), and brain (5%) [4]. Other rare sites of metastasis include pancreas [5,6], adrenal gland [7,8], parotid gland [9], maxilla, pharynx etc. Even with early-stage disease, late metastases can occur after complete resection. However, after the development of metastasis the prognosis is often poor, as non-operative modalities for advanced renal carcinoma have failed to improve survival significantly. The average survival of metastatic RCC is about 4 months and only 10% of these survive for one year. Chemotherapy, hormonal therapy, and radiotherapy have generally proved ineffective for primary lesion or metastatic deposits, but an improved survival is seen with the use of immunotherapy with either interleukins or interferon [10,11]. There is a small subset of patients where solitary metastasis is present either at the time of presentation or develops during follow-up after nephrectomy these patients have a better survival. An autopsy series showed that 8 to 11% of the patients with metachronous metastases have a solitary lesion [12]. With the introduction of immunotherapy or immunochemotherapy for advanced renal cell carcinoma, survival has improved for a subset of patients. Surgical resection of the renal tumor and solitary metastases if present is still the treatment of choice for these lesions [13]. By modern perioperative management, even extended resections of metastatic lesions can be performed with limited morbidity and mortality [14]. The present study reports the result of retrospective analysis of the records of 43 patients treated for solitary synchronous or metachronous metastases from renal cell carcinoma.
Patients and methods
Between 1988 and 2001, 1863 patients of RCC were treated of these 43 cases had solitary metastasis to different sites. Thirteen patients had synchronous solitary metastsis, whereas 30 patients developed solitary metastasis during follow-up after the treatment of the primary in a metachronous fashion. The location and extent of the metastatic disease was evaluated by various diagnostic methods, which include chest radiograph, isotope bone scan, liver function tests, and serum alkaline phophatase. All patients underwent either ultrasound (USG) or computerized tomography (CT) of abdomen to rule out local recurrence. The patients with lymph node metastasis were excluded from the analysis. An attempt was made to obtain histological proof of the metastatic lesion either by fine needle aspiration cytology (FNAC) or USG/CT guided biopsy. The analysis was performed separately for patients with synchronous and metachronous metastatic lesions. Survival analysis was calculated by using Kaplan-Meier test and survival was compared using Log-rank test.
Results
Synchronous group
Thirteen patients with synchronous solitary metastatic disease included 10 males and 3 females. The age of the patients ranged from 38 to 69 years with mean of 57 years. The site of solitary metastases were bone in 6, lung in 3, liver in 2, and 1 each in brain and opposite adrenal gland (Table 1). All patients underwent radical nephrectomy for the primary in the same sitting except in one patient with brain metastasis, which was irradiated. One patient with bone metastasis had positive surgical margin and received radiotherapy. The median disease free survival was 25 months and the median overall survival was 26 months with 3-year survival being 27% and none of the patients surviving 5 years (Figure 1). Five patients who could afford received interferon alpha treatment for relapse to multiple sites. Due to small number of patients a subgroup analysis was not carried out for these patients.
Table 1 Patient and tumor characteristics- synchronous solitary metastasis
Total No. of patients 13
Age 38–69 years (mean = 57)
Sex Male-10, Female-3
Site of metastasis Bone-6, Lung-3, Liver-2, Brain-1, Opposite adrenal-1
Pathological stage of primary pT1–3, pT2–6, pT3–4
Fuhrman's Grading Grade I-1, Grade II-4, Grade III-5, Grade IV-3
Figure 1 Overall survival in patients with synchronous solitary metastasis.
Metachronous group
Thirty patients developed metachronous solitary lesion during follow-up after definitive treatment of the primary renal cell carcinoma of the kidney. The age of the patients ranged from 34 to 75 with the mean of 54 years. There were 21 males and 9 female patients. The metastatic interval varied between 3 to 33 months with mean of 16 months. The distribution of site of solitary metastasis was bone in 13, lung in 6, liver in 3, brain in 2, and one each in parotid, maxilla, pharyngeal wall, soft tissue shoulder, opposite adrenal and gall bladder (Figure 2). Eleven patients with bone metastases underwent excision of the lesion and other 2 patients received definitive radiotherapy. All patients with lung and liver metastases underwent surgical excision. One patient with brain metastasis received radiotherapy and other one had surgical excision of the lesion. The patient with parotid lesion underwent parotidectomy and the one with metastasis in soft tissue shoulder had forequarter amputation. Adrenalectomy and cholecystectomy was done for adrenal and gall bladder metastasis respectively. Primary tumor stage was pT1in 9, pT2 in 13, and pT3 in 8 patients. The distribution of histological types and the nuclear grading is shown in the table 2.
Figure 2 Ultrasonography showing solitary metachronous metastasis to gall bladder.
Table 2 Patient and tumor characteristics- metachronous solitary metastasis
Total No. of patients 30
Age 34–75 Years (Mean 54)
Sex Male-21, Female-9
Metastatic interval 3–33 months (Mean 16)
Site of metastasis Bone-13, Lung-6, Liver-3, Brain-2, Parotid-1, Maxilla-1, Pharyngeal wall-1, Gall bladder-1, Opposite adrenal-1, Soft tissue shoulder-1.
Histology of primary Conventional Clear cell carcinoma -25
Papillary-4
Chromophobe -1
Pathological stage of primary pT1–9, pT2–13, pT3–8
Fuhrman's Grading Grade I-5, Grade II-12, Grade III-7, Grade IV-6
Post metastatectomy overall median survival was 45 months. The 3-year survival rate was 60 % and 5-year survival was 38 % (Figure 3). The overall survival for those patients in whom solitary metastasis occurred within one year (n = 15) was 31 months compared to 63 months in those who developed metastasis after one year (n = 15) the difference was statistically significant (log rank test p = 0.03) (Figure 4). The three patients with liver metastasis and 2 with brain metastasis behaved poorly with survival of 6 to 20 months. Two patients with bone metastasis and one patient with brain metastasis received radiotherapy after excision due to positive surgical margins.
Figure 3 Overall survival of the patients with metachronous solitary metastasis.
Figure 4 Overall survival comparing patients with metachronous solitary metastasis occurring within 12 months versus those occurring after 12 months.
The overall post metastatectomy median survival was 63 months for pT1 tumor, 45 months for pT2 and 24 months for pT3 tumors and the difference in the survival was significant (log rank test p = 0.004) (Figure 5). There was also significant difference in the survival (log rank test p = 0.001) among patients with Fuhrman's grade I to grade IV disease (Figure 6). The patients with lung and bone metastasis had overall mean survival of 62 months and those with brain and liver metastasis was 22 months. Thirteen patients received immunotherapy after relapse at multiple sites following treatment of the solitary metastasis. However, the number of patients in each group was small, hence subgroup analysis could not be carried out for these patients.
Figure 5 Overall survival according to pathologic stage of the primary tumor.
Figure 6 Overall survival according to Fuhrman's nuclear grading.
Discussion
RCC is known for its varied presentation and its propensity to metastasize by way of both venous and lymphatic routes. Ritchie and deKernion [15] found that 23% of patients with RCC present initially with metastatic disease and 25% develop metastatic disease within 5 years of nephrectomy. RCC can, however, metastasize to virtually any organ, including the thyroid, pancreas, skeletal muscle, and skin or underlying soft tissue.
Despite the early promising results with immunotherapy, a complete response occurs in less than 15% and is rarely durable, emphasizing that metastatic disease best treated with complete resection of the primary and metastatic lesions where possible. Solitary metastases are identified at the time of diagnosis of renal cell carcinoma in 2% to 4% of patients. Depending on the location of the solitary metastasis, radical nephrectomy [16] with excision of the solitary metastatic lesion has been advocated, with 20% to 30% 5-year survival rates being reported.
Barney in 1961 [17] did nephrectomy and lobectomy for lung metastasis and patient survived for 23 years. This influenced many workers to perform excision of solitary metastasis. Middleton [18] reviewed the literature and concluded that the survival after surgical excision of solitary metastatic lesions was 45% at 3-years and 34% at 5-years. Tolia and Whitmore [19], Morrow et al [20], and Jett et al [21] also reported similar 5-year survival after the treatment of solitary metastatic tumors. Skinner et al [22] reported a 29% 5-year survival in a series of 41 patients in whom one or two metastases were excised surgically in addition to nephrectomy. In our series, the 3 and 5-year post metastatectomy survival was 58% and 35% respectively in metachronous group of patients whereas 3-year survival was 20% in the synchronous group, the higher survival could be due to number of patients being loss to follow-up and hence censored from the survival analysis.
Middleton [18] found significant difference in survival between synchronous and metachronous solitary metastases, synchronous group being the worst. O'Dea et al [23] and Rafla [24] also agreed that the patients with synchronous solitary metastasis have poor survival when compared to patients who develop metastasis after removal the primary tumor. Although the site of metastasis was not significant in many reports we found patients with lung and bone metastases fared much better (median 62 months) than with liver and brain metastases (median 22 months). This could be due to the fact that not all patients with brain metastasis had undergone surgical resection, and more use of adjuvant treatment in patients with lung or bone metastasis.
Most metachronous metastases are identified in the first or second year after nephrectomy. Tally et al [25] Skinner et al [22] and Tongaonkar et al [26] found that survival in metachronous metastases which appeared before one year after the nephrectomy was poor (median overall survival 33 months) when compared to one in whom metastases appeared after one year (median overall survival 55 months). Our results also suggests that over all survival in patients with disease free interval less than one year was 31 months when compared to 63 months in those disease free intervals was more than one year.
We also noted significant difference in the survival among patients of different primary tumor stage and Fuhrman's nuclear grading with early stage and low grade having better survival. Thus, the groups of patients who do better with solitary metastasis excision are those with disease free interval of more than one year, low primary tumor stage, low grade and those with bone and pulmonary parenchymal metastasis. However, the effect of treatment modality and adjuvant immunotherapy can not be calculated due to smaller number of cases in the present series.
Conclusion
A small subset of patients with a solitary metastasis from renal cell carcinoma and good general condition may benefit from nephrectomy and resection/radiotherapy of the metastatic lesion. Surgical excision can locally control the tumor, and relieve pain. Patients with disease free interval of more than one year before development of solitary metastasis, early primary tumor stage, low-grade and patients with bone and pulmonary parenchymal metastasis appear to have better survival. We believe that the relatively prolonged survival of these small cohort of patients with solitary metastatic lesions from RCC justify aggressive surgical excision involving multispecialty specialists with long-term survival, however, prospective studies are needed before this can be recommended as standard practice.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
TY designed the study, prepared the manuscript, and drafted the manuscript.
MU gave radiation details and participated in the preparation of manuscript.
RS participated in the design of the study and performed the statistical analysis.
CR participated in its design and coordination and helped to draft the manuscript.
TH overall monitoring of the analysis and editing of the final version for publication.
All authors read and approved the manuscript
==== Refs
Matveev VB Gurarii LL Began-Bogatskii KM Surgical treatment of late metastases of kidney cancer Urol Nefrol (Mosk) 1999 2 51 52 12434447
Cozzoli A Milano S Cancarini G Zanotelli T Cosciani Cunio S Surgery of lung metastases in renal cell carcinoma Br J Urol 1995 75 445 447 7788253
Kollender Y Bickels J Price WM Kellar KL Chen J Merimsky O Meller I Malawer MM Metastatic renal cell carcinoma of bone: indications and technique of surgical intervention J Urol 2000 164 1505 1508 11025692 10.1097/00005392-200011000-00011
Ritchie AWS Chisholm GD The natural history of renal carcinoma Semin Oncol 1983 10 390 400 6665566
Andoh H Kurokawa T Yasui O Shibata S Sato T Resection of a solitary pancreatic metastasis from renal cell carcinoma with a gallbladder carcinoma: report of a case Surg Today 2004 34 272 275 14999544 10.1007/s00595-003-2680-6
Kassabian A Stein J Jabbour N Renal cell carcinoma metastatic to the pancreas: a single-institution series and review of the literature Urology 2000 56 211 215 10925080 10.1016/S0090-4295(00)00639-7
Chris WE Frank SD Solitary metachronous contralateral adrenal metastasis from renal cell carcinoma Urology 1999 54 162 164 10.1016/S0090-4295(99)00023-0
Huisman TK Sands JP Renal cell carcinoma with solitary metachronous contralateral adrenal metastasis Urology 1991 38 364 368 1755149 10.1016/0090-4295(91)80155-Z
Park YW Hlivko TJ Parotid gland metastasis from renal cell carcinoma Laryngoscope 2002 112 453 456 12148853 10.1097/00005537-200203000-00009
Motzer RJ Bander NH Nanus DM Renal cell carcinoma N Engl J Med 1996 335 865 866 8778606 10.1056/NEJM199609193351207
Rosenberg SA Yang JC White DE Steinberg SM Durability of complete responses in patients with metastatic cancer treated with high dose interleukin-2: identification of the antigen mediating response Ann Surg 1998 228 307 319 9742914 10.1097/00000658-199809000-00004
Hajdu SI Thomas AG Renal cell carcinoma at autopsy J Urol 1967 97 978 982 6028330
Swanson DA Surgery for metastases of renal cell carcinoma Scand J Surg 2004 93 150 155 15285568
Kozlowski JM Management of distant solitary recurrence in the patient with renal cancer Urol Clin North Am 1994 21 601 624 7974893
Ritchie AWS deKernion JB The natural history and clinical features of renal carcinoma Semin Nephrol 1987 7 131 139 3306862
Montie JE Stewart BH Straffon RA Banowsky LH Hewitt CB Montague DK The role of adjunctive nephrectomy in patients with metastatic renal cell carcinoma J Urol 1977 117 272 275 65479
Barney JD Churchill EJ Adenocarcinoma of the kidney with metastasis to the lung cured by nephrectomy and lobectomy J Urol 1961 42 298 300
Middleton RG Surgery for metastatic renal cell carcinoma J Urol 1967 97 973 977 6030034
Tolia BM Whitmore WF Jr Solitary metastasis from renal cell carcinoma J Urol 1975 114 836 838 1195458
Morrow CE Vassilopoulos P Grage TB Surgical resection for metastatic neoplasms of the lung Cancer 1980 45 2981 2985 6155986
Jett JR Hollinger CG Zinsmeister AR Pairolero PC Pulmonary resection of metastatic renal cell carcinoma Chest 1983 84 443 445
Skinner DG Colvin RB Vermllion CD Pfister RC Leadbetter WF Diagnosis and management of renal cell carcinoma: a clinical and pathologic study of 309 cases Cancer 1971 28 1155 1177
O' Dea MJ Zinke H Utz DC Bernatz PE The treatment of renal cell carcinoma with solitary metastasis J Urol 1978 120 540 542 712892
Rafla S Renal cell carcinoma: natural history and results of treatment Cancer 1970 25 26 40 5410313
Talley RW Moorhead EL Tucker WG San Diego EL Brennan MJ Treatment of metastatic hypernephroma JAMA 1969 207 322 328 5818154 10.1001/jama.207.2.322
Tongaonkar HB Kulkarni JN Kamat MR Solitary metastases from renal cell carcinoma: A review J Surg Oncol 1992 49 45 48 1548881
|
16029517
|
PMC1185571
|
CC BY
|
2021-01-04 16:39:04
|
no
|
World J Surg Oncol. 2005 Jul 20; 3:48
|
utf-8
|
World J Surg Oncol
| 2,005 |
10.1186/1477-7819-3-48
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 10.1371/journal.pcbi.001001505-PLCB-RA-0031R1journal-pcbi-0010015Research ArticleBioinformatics - Computational BiologyGenetics/GenomicsGenetics/Gene ExpressionHomo (Human)Mus (Mouse)Recognition of Unknown Conserved Alternatively Spliced Exons Discovering Unknown Alternative ExonsOhler Uwe *¤Shomron Noam Burge Christopher B *Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States of AmericaHenikoff Steven EditorFred Hutchinson Cancer Research Center, United States of America*To whom correspondence should be addressed. E-mail: [email protected] (UO); [email protected] (CBB)¤ Current address: Institute for Genome Sciences and Policy, Duke University, Durham, North Carolina, United States of America
7 2005 8 7 2005 1 2 e1516 2 2005 6 6 2005 Copyright: © 2005 Ohler 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 split structure of most mammalian protein-coding genes allows for the potential to produce multiple different mRNA and protein isoforms from a single gene locus through the process of alternative splicing (AS). We propose a computational approach called UNCOVER based on a pair hidden Markov model to discover conserved coding exonic sequences subject to AS that have so far gone undetected. Applying UNCOVER to orthologous introns of known human and mouse genes predicts skipped exons or retained introns present in both species, while discriminating them from conserved noncoding sequences. The accuracy of the model is evaluated on a curated set of genes with known conserved AS events. The prediction of skipped exons in the ~1% of the human genome represented by the ENCODE regions leads to more than 50 new exon candidates. Five novel predicted AS exons were validated by RT-PCR and sequencing analysis of 15 introns with strong UNCOVER predictions and lacking EST evidence. These results imply that a considerable number of conserved exonic sequences and associated isoforms are still completely missing from the current annotation of known genes. UNCOVER also identifies a small number of candidates for conserved intron retention.
Synopsis
Alternative splicing is a process in which more than one protein variant can be produced from one gene: Specific parts of the mRNA precursor are included or excluded during the processing into the mature transcript. It is very prevalent in mammalian genomes, and variants are often specific for particular cell types, developmental states, or environmental changes. The identification of such variants has until recently relied solely on the sequencing and comparison of expressed sequence tags (ESTs), but the number of available ESTs is not large enough to cover all variants under all conditions.
Ohler et al. have now devised a comparative genomics algorithm based on a pair hidden Markov model, which identifies parts of genes that are alternatively spliced and have not been observed in ESTs. Starting from known annotated genes conserved in human and mouse, they scan corresponding intron pairs of these genes to identify conserved sequences that match the model. Experimental validation of a number of new predictions show that the approach can successfully uncover splice variants that are as yet unknown and not part of the large libraries of ESTs. Together with recently proposed complementary computational methods, this approach helps us to complete our knowledge about the transcript diversity created by alternative splicing.
Citation:Ohler U, Shomron N, Burge CB (2005) Recognition of unknown conserved alternatively spliced exons. PLoS Comp Biol 1(2): e15.
==== Body
Introduction
Almost all protein-coding genes of humans and other mammals have a split structure with several exons and introns. Intronic sequences are removed from the primary transcript by the process of pre-mRNA splicing [1], an essential step in eukaryotic gene expression. The number of functional variants generated from one transcript can be greatly increased by alternative splicing (AS), in which one or more exons or parts thereof are skipped, or an intron is retained, when compared to a different transcript from the same gene [2–4]. By this mechanism, an organism can generate several protein isoforms from a single gene, potentially leading to huge numbers of protein variants, and AS is an important means of gene regulation, being frequently used during development or in differentiation. It is also a very common event: even conservative estimates put the fraction of human genes with more than one isoform at 40% [5], with similar rates estimated in all animals [6]. The basic types of AS are exon skipping, intron retention, and alternative 5′ and 3′ splice site usage, with exon skipping being the most prevalent in mammals [7–10]. To date, AS events have been identified on a large scale primarily from comparisons and alignments of expressed sequence tag (EST) and cDNA sequences, and databases based on these alignments have been described [8,10,11]. Ab initio prediction of AS events from one genomic sequence alone has been attempted only rarely: Computational screening of introns for sequences similar to neighboring exons revealed candidate duplicated exons, which may be involved in mutually exclusive splicing [12]. A hidden Markov model (HMM) sampling approach can detect likely variants of complete gene structures [13]. These studies, however, did not experimentally verify predicted AS events.
Despite the large number of ESTs that have been sequenced from a variety of organisms and tissues, the coverage of the transcriptome still remains limited, especially for genes expressed at lower levels or under limited conditions. It becomes increasingly hard to distinguish functional but rare EST-detected variants from nonfunctional isoforms and artifactual sequences contained in the libraries [7]. At least in mammals, the less common isoforms are often not conserved, in contrast to the high degree of conservation seen for the more common variants [14,15]. Exons subject to nonconserved skipping events are significantly different from alternative conserved exons (ACEs, pairs of orthologous human/mouse exons both subject to exon skipping), being less likely to preserve reading frame and more likely to contain in-frame stop codons, suggesting that a significant fraction does not lead to functional proteins [3]. ACEs also tend to be flanked by long, highly conserved intronic sequences, possibly because of the presence of sequence elements required to regulate inclusion of the exons in specific cell types or conditions. Regions containing ACEs are thus often among the most highly conserved segments in the human genome [16,17]. Two non-EST-based computational approaches have made use of these specific features to successfully classify conserved exons as to whether they are subject to skipping or not [17,18], which confirmed that current coverage of splicing isoforms by ESTs alone is still limited, but that most EST-derived skipping events may in fact not be conserved. Regarding intron retention events (IREs), a recent study estimated that they occur in about 15% of human genes [19]; however, stricter requirements lower this estimate to 5% [7,19], and half of these cases are within the untranslated region and thus do not alter the encoded protein. Only ten of the reliably determined IREs, coding or noncoding, were found to be clearly conserved between human and mouse, based on currently available EST and cDNA evidence, suggesting that this mode of regulation is not common in mammalian genes.
Results
Design of a Pair HMM to Discover Conserved AS
Given that conserved coding AS events have the potential to alter protein isoforms under tightly regulated circumstances, these sequences should be among the functionally most important segments of the genome. Computational approaches to predict ACEs [17,18] demonstrated that inferring AS events from ESTs alone will miss a considerable fraction of conserved skipping events: those for which current EST libraries contain only those isoforms that include the exon. Presumably, this is caused by the fact that the majority of isoforms include the ACE under consideration. Here, we set out to develop a complementary approach to predict those conserved AS events in which the majority of isoforms do not include the ACE or retained intron, and for which the exonic sequence subject to AS is thus completely absent in available EST sequences and has not been described before.
To systematically identify such previously unknown ACEs and IREs missed by comparative gene-finding and cDNA and EST alignments, we developed a system for comparative prediction of mammalian conserved coding AS events termed UNCOVER (for “unknown conserved variable exon recognition”). UNCOVER is based on a pair HMM (pHMM) [20–22], a probabilistic model that can be used to obtain an optimal alignment and simultaneous annotation of two sequences. A pHMM consists of states that can be either pair states, which contain a probability distribution on the occurrence of pairs of aligned nucleotides, or single states, which model nucleotides in one sequence but not the other, thus describing insertions and deletions. Different states are used to model different patterns of conservation, e.g., the distribution in a state for the third codon position will typically contain higher probabilities for mismatches than the one for the first or second position, and those mismatches not changing the encoded amino acid will be more frequent than others. While computing the optimal alignment, a pHMM labels the alignment with the states that were used in the process, and the aligned sequence can be parsed into functional categories based on the labels.
The UNCOVER pHMM was designed to specifically align one orthologous human/mouse intron pair at a time and to predict whether it potentially harbors undiscovered AS events (Figure 1). The model states describe the probability of aligned nucleotide pairs in the 3′ and 5′ splice sites, coding regions, and noncoding alignable regions, as well as single nucleotides in nonalignable regions (i.e., insertions or deletions in the human sequence when compared to mouse). The transition probabilities of the model connect the states in different ways corresponding to submodels for none or any one of four basic AS events: skipping, retention, and alternative 5′ and 3′ exons. A labeled UNCOVER alignment can thus predict whether a conservation pattern seen in the intron pair fits better to conserved noncoding sequence, to coding sequence throughout—suggesting intron retention—or to the conserved sequence of {3′ splice site, coding exon, 5′ splice site} somewhere within the intron pair, suggesting the presence of an ACE (see Figure 1 for a detailed description of the model, and Figure 2 for an example alignment). The UNCOVER submodels for alternative 5′ and 3′ exons are at this point used only to achieve better discrimination between the different AS types, and are not analyzed in detail in this study.
Figure 1 Structure of the UNCOVER pHMM
The model is used to globally align a pair of orthologous human/mouse introns and detect conserved coding AS events.
(A) A schematic overview of the model architecture, with circles indicating groups of functionally related states. For accurate splicing, the two ends of an intron must be precisely determined by the splicing machinery. The prominent sites for this process are the 5′ splice site (5′ss) at the junction between the upstream exon and the intron and the 3′ splice site (3′ss) at the junction of the intron and the downstream exon. As reference, pictograms of the mammalian 5′ splice site and 3′ splice site are depicted, in which the letters at individual positions are scaled according to their frequency. We restrict ourselves to U2-type splice sites with perfectly conserved GT–AG dinucleotides. The alignment always starts with the conserved 5′ splice site after the initial GT dinucleotide. The transitions of the model then allow it to pursue several paths, corresponding to different types of AS, indicated by small icons. (1) The “default” is to observe conserved or nonconserved noncoding sequence, possibly alternating between these two. (2) Transitions to an ACE sequence of conserved {3′ splice site, skipped exon, 5′ splice site} are possible at any time, and can also occur more than once. (3) An IRE is modeled by going from a 5′ splice site to a 3′ splice site by only passing through a coding submodel. (4) and (5) An early exit from this codon model through another 5′ splice site leads to an alternative 5′ exon at the beginning of the sequence, or correspondingly to an alternative 3′ exon at the end. The alignment is fixed on the right side by the 3′ splice site at the end of the intron. All splice site states are first-order pair states not allowing for insertions or deletions. The 5′ splice site part of the model covers 9 nt (3 nt in the exon, plus the conserved GT and the following 4 nt in the intron); the 3′ splice site is 23 nt long (18 nt and the conserved AG in the intron plus 3 nt in the exon).
(B and C) A detailed view of the noncoding intronic submodel (B) and a close-up of the coding submodel (C), with closed circles representing pair states and dashed circles representing single states. Thick straight arrows indicate the allowed start and end states of the submodels. The noncoding conservation (B) is modeled by a first-order pair state, allowing insertions and deletions of individual nucleotides. The null model contains single first-order states representing nonconserved human and mouse intronic sequences. The coding states (C) comprise three second-order pair states for nucleotides in the three codon positions, as well as three second-order single states each for human and mouse to capture species-specific codon insertion/deletion events. The transition matrix ensures that only those insertion/deletion events covering complete codons are admissible.
Figure 2 Experimental Validation of UNCOVER Predictions
(A) RT-PCR validation of newly identified alternative exons with no prior EST evidence. Lane numbers are given in Arabic numerals below the gel; sample numbers of new verifications and negative controls are in Roman numerals above. Lanes 2–5 were verified using flanking primers and therefore show two bands each, the larger one corresponding to the event including the newly identified ACE. Lanes 6–9 used a primer internal to the newly identified exon and therefore only show one band each. Lanes 10–13 are typical examples of ten randomly selected introns in the ENCODE target regions that were not predicted to harbor AS events. Lane 14 shows a blank reaction control without adding template. Lanes 1 and 15 contain size markers spaced at 100 nt intervals, with the strong bands corresponding to 1,000 and 500 nt. Ensembl ID pairs for the known exon upstream of the validated new one and the corresponding gene are as follows: internal exons, lanes 2–6: ENSE00000881911.1:ENSG00000004866.5, ENSE00000862512.1:ENSG00000126217.3, ENSE00001201432.1:ENSG00000168781.5, ENSE00001146476.1:ENSG00000168781.5, and ENSE00001084095.4:ENSG00000164402.2; terminal exons, lanes 7–9: ENSE00001379673.1:ENSG00000159140.5, ENSE00001046164.1:ENSG00000067369.1, and ENSE00000952769.2:ENSG00000142183.3 (a known case as positive control); random negative controls, lanes 10–13: ENSE00001321652.4:ENSG00000161980.2, ENSE00000868377.2:ENSG00000102125.4, ENSE00001239587.1:ENSG00000100220.2, and ENSE00001307891.1:ENSG00000185721.1.
(B) Example UNCOVER alignment of a newly detected ACE. Aligned nucleotides are connected with a vertical dash in case of identity, a colon in case of a transition, and a dot in case of a transversion. The alignment is labeled with the types of the states that lead to the most likely alignment: C, conserved noncoding sequence; F, 5′ splice site; I, nonconserved intronic sequence; T, 3′ splice site; 1, 2, and 3, coding sequence, with the number giving the position in a codon. The detected ACE is flanked by highly conserved noncoding sequence, a characteristic of true ACEs. The sequence shown corresponds to the event in sample i in (A).
Candidate structures scored by current ab initio gene-finding algorithms are limited in that they have to fulfill the restrictions of the whole gene model—including presence of an open reading frame throughout and distributions on the expected length of exons. The UNCOVER model as shown does not impose these restrictions and thus has the potential to detect conserved events missed by computational gene finders. It can be used to predict new AS events in two species simultaneously, or to provide additional evidence for a conserved AS event in case of limited EST coverage or ESTs from only one of two species. An advantage of our pHMM is that, in addition to identifying AS events, it also identifies conserved noncoding sequences, potentially containing cis-regulatory elements for splicing or transcription. UNCOVER per se identifies any kind of coding sequence fitting the pHMM model, which means that predicted skipped exons may in fact simply be exons missed by the existing annotation that are conserved but not alternatively spliced. In practice, however, the pipeline to determine the input intronic regions uses annotations of conserved gene structures, which are generally inferred from EST and cDNA evidence, and by definition most true positive predicted exons are thus skipped exons.
Application of UNCOVER on a Curated Dataset of Known Skipped Exons
To establish a baseline for how well ACEs can be detected with our approach, we collected 241 orthologous introns containing known ACEs derived from human and mouse EST and cDNA alignments, ranging in length from about 250 nt to about 93,000 nt. UNCOVER made a total of 309 predictions with 210 true positives (Table 1), successfully pinpointing the exact location of the ACEs: 89% of true positive UNCOVER predictions identified at least one splice site exactly. The inexactness of the remaining 11% reflected the strong sequence conservation around ACEs, which makes it difficult to infer the exact location of the correct splice sites in some cases. For comparison, we performed a simple BLASTN [23] analysis, keeping all hits longer than 30 nt with E values smaller than 10−10. This resulted in 667 predictions, out of which 253 overlapped with 233 known exons. However, not a single hit corresponded to the exact exon boundaries. BLASTN can thus detect the rough locations of the great majority of ACEs in this set, but in an extremely unspecific manner; using TBLASTX instead of BLASTN gave highly similar results. Retaining only the best hit with at least 70% sequence identity but independent of E value resulted in 212 hits (88%) overlapping ACEs. The UNCOVER detection rate is thus virtually identical to the best BLAST hit analysis, but without making any unrealistic assumptions as to whether or how many ACEs may be present in an intron (and, importantly, UNCOVER predictions usually have one or both splice sites correct).
Table 1 Prediction Results on a Known Set of 241 Conserved Skipped Exons
We compare results obtained by BLASTN analysis with those of the UNCOVER approach. Predicted regions overlapping with a known ACE in the human sequence are counted as true positives, and the fractions are given for which the locations of the 5′ splice site, the 3′ splice site, or both are correct. Sensitivity is calculated as the number of true positives divided by the total number of known exons, and specificity as the number of true positives divided by the total number of predictions. We also show the total number of nucleotides spanned by all predictions, and the number of nucleotides overlapping the known ACEs.
As an alternative to probabilistic sequence models, the Ka/Ks test has recently been applied to the problem of comparative gene finding. This is an established method to detect adaptive molecular evolution, based on the observation that coding sequences are generally under selection to conserve amino acid sequence. In an application of the Ka/Ks test to gene finding [24], 92% of internal exons passed the test at a p-value of 0.05. However, only 47% of the tested conserved skipped exons in our set of 241 exons passed the test at the same p-value, even under the assumption of knowing the exact exon boundaries. This is apparently due to the smaller size of the skipped exons (median 84 nt compared to 123 nt in the set of constitutive exons used in [24]) and the higher rate of synonymous sequence conservation of ACEs compared to constitutive exons (see Protocol S1 for details). The Ka/Ks test therefore has inherent limitations when applied to detect alternatively spliced exons.
Analysis of the ENCODE Target Regions
As an application of UNCOVER on a genomic scale, we focused on the 1% subset of the human genome known as the ENCODE (Encyclopedia of DNA Elements) target regions, currently the subject of comprehensive experimental and computational analyses [25]. UNCOVER made 135 predictions in 73 out of a total of 1,776 orthologous introns (4.1%), located in 40 out of 323 genes (12.4%). In comparison, there were 982 BLAST hits to 321 introns with the thresholds set as above, more than seven times as many hits at a similar level of sensitivity. A total of 42 UNCOVER predictions corresponded to either annotated human skipped exons or sequences covered by human ESTs in dbEST (as of August 23, 2004): 15 matched annotated ACEs in known Ensembl genes; seven matched annotated Ensembl EST genes or VEGA (the manually curated Vertebrate Genome Annotation database [26]; http://vega.sanger.ac.uk) genes; and three matched spliced ESTs not corresponding to any annotation, indicating the presence of yet unannotated ACEs in the genes LUC7L, C16orf35, and CDH2. The remaining predictions matched unspliced ESTs corresponding to 11 intronic regions. Many of these ESTs were polyadenylated, and one of the matches was annotated as an alternative terminal exon of an EST gene. Indeed, we observed that with only one exception, these UNCOVER predictions were located in the 3′ terminal region of the genes. The location of these putative terminal exons cannot be expected to be exactly predicted by UNCOVER, as they do not end with a 5′ splice site and contain 3′ untranslated sequence.
For experimental validation, we selected those 20 introns containing predicted ACEs without any EST evidence that were flanked on both sides by strong splice sites. We followed an RT-PCR sequencing protocol in a set of eight adult human tissues and HeLa cells, and confirmed expression of the flanking exons for 15 out of the 20 tested introns (i.e., in five cases, we could not observe any expression in the selected tissues using multiple sets of primers). For five out of these 15, we repeatedly observed two PCR bands, with the sequence of the smaller product matching the exons flanking the predictions, and in one additional case, we saw expression of a product using primers placed inside the predicted ACE and a neighboring exon. In three of these six cases (including the gene ST7), the sequence of the alternative product included the UNCOVER predicted exons, showing that our approach led to the discovery of new ACEs expressed at low levels that had not yet been covered despite the availability of more than 5 million human ESTs (Figure 2; see also Dataset S1 for details). One case (CRAT) corresponded to a skipped exon in which only a small part in the middle was conserved between human and mouse, and which could therefore not be predicted by UNCOVER. In the remaining two cases (including MCF2L), the included alternative sequence did not match any sequence in the nonredundant GenBank database, suggesting gaps or misassemblies in these introns. Furthermore, we confirmed two of the ten potential new alternative terminal exons, using primers placed inside the predicted exon and the immediately upstream exon. Not counting the cases with nonmatching sequence, we therefore report here eight conserved AS events—five verified by RT-PCR based on de novo predictions plus three with spliced EST evidence—in addition to 15 known events present in the Ensembl annotation of the ENCODE regions (as of August 2004), and provide additional support for eight more ACEs that have only been annotated as part of Ensembl EST genes or cross-species homology.
A Genome-Wide Search for Conserved Retained Introns
Turning to conserved intron retention, we extended the UNCOVER analysis to the whole genome. Our analysis spanned a total of 84,233 orthologous intron pairs, 46 times the number within the ENCODE region, which covers 1% of the nucleotides in the genome but is somewhat gene rich. Despite this large number of introns, and without assumptions on the reading frame of the upstream exon, only 23 were predicted to be more likely to be conserved IREs with coding potential than to harbor conserved noncoding sequence (see Table 2 and Dataset S2). Out of these, 12 were covered by human ESTs (as of October 25, 2004), with a total of ten annotated as known or EST genes. The length of 12 candidates was a multiple of three, and 13 out of 19 for which we could determine the open reading frame from full-length cDNAs were predicted to continue in the frame of the upstream exon. Given evidence of length, reading frame, EST coverage, and the presence of protein domains spanning the candidate IRE, another four (among them PAX6 and PCDH17) in addition to the ten already annotated can be considered highly likely IREs, and an additional two involve splicing of an alternative 5′ splice site in a mutually exclusive fashion to one of the neighboring exons.
Table 2 Predicted Conserved Coding IREs and Their Evidence
For each candidate, the table shows Ensembl IDs of gene, transcript, and upstream exon; the HUGO gene ID (if available); whether the predicted retained intron is covered by spliced ESTs; whether it is predicted to continue in-frame with upstream exon; whether it has a size that is a multiple of three; and whether a protein domain detected by InterPro [40] spans across it. For four genes, the frame of the upstream exon could not be uniquely determined from the human and mouse annotations.
Discussion
We propose a comparative sequence analysis approach to detect hitherto unknown and alternatively spliced conserved exons, followed by experimental validation. Considering the 53 introns with UNCOVER predictions in the ENCODE region that do not contain annotated skipped exons, and adjusting the number by the sensitivity (87%) and specificity (68%) of UNCOVER on the curated ACE dataset, leads to an estimated total of 53(0.68)/(0.87) = 41 ENCODE introns containing ACEs not currently annotated. This shows that even for known and well-studied genes, current EST coverage is far from providing a complete picture of AS. Scaled up to the whole genome, which contains 46 times the number of introns in the ENCODE region, about 1,900 introns may harbor as yet unknown ACEs. Since specificity may be overestimated somewhat using the curated ACE dataset, as a lower estimate, a straight extrapolation of the so far experimentally verified ACEs suggests that at least several hundred ACEs are currently still awaiting discovery. We expect that UNCOVER will therefore be especially useful when turning to regions of the genome less covered by ESTs [27] than the ENCODE targets. On the other hand, retention of translatable introns does not appear to play a major role in generating conserved protein-coding isoforms in mammals. We do not rule out a common role for nonconserved regulated intron retention or conserved IREs in UTRs or in species other than mammals.
Considering the RT-PCR results, the isoforms that include the newly verified exons are expressed at lower levels than the isoforms in which the exons are skipped. This is in accordance with the lack of support in EST data: were the isoform with the exon included the major one, we would expect it to have been observed in the EST data. Detailed measurements of the frequency of individual AS events, such as for the well-studied cell-surface receptor CD44, showed that the inclusion of functional alternatively spliced exons can indeed be much less common than skipping [28]. A number of points argue for the functional relevance of our newly detected minor isoforms: we are usually able to amplify them by placing primers in the flanking exons (see Figure 2); they are expressed in a tissue-specific manner in human, and we observe expression in mouse as well (Figure S1); and their sequence is conserved not only in mouse but in a number of other vertebrate species (see Dataset S1).
In its current state, UNCOVER is designed to predict only fully coding exons. In addition to an easy adaptation to pairs of non-mammalian species such as nematodes or insects, further development of UNCOVER could lead to removal of this restriction to include exons with in-frame stop codons and noncoding 3′ ends. This should enable us to better predict terminal exons that are only partly coding: we showed that these can be predicted by the current version of UNCOVER, but an explicit model of noncoding conservation and polyadenylation sites would undoubtedly lead to improvements. Furthermore, including in-frame stop codons may allow predictions of additional ACEs subject to nonsense-mediated decay (NMD), a mechanism that degrades transcripts containing premature termination codons [29]. NMD has been proposed as an important mechanism for gene regulation in conjunction with AS [9]. A PCR verification of predictions subject to NMD could be done following knockdown of the important enzymes in the NMD pathway, to be able to accumulate and amplify the transcripts. To gain additional confidence in such predictions, UNCOVER ought to be extended to more than two species, which should additionally allow reliable prediction of alternative 5′ and 3′ splicing that may lead to isoforms differing by only a few nucleotides. This can be done in a way similar to an approach based on probabilistic phylogenetic models [30].
Recent independent methods based on comparative genome analysis [17,18,31], which can be regarded as complementary to the work described here, have been successful in classifying known conserved exons as skipped or constitutive. These approaches are based on methodology from statistical learning theory, and a true integration with a probabilistic approach such as UNCOVER is not straightforward. However, they could be easily used to filter our predictions. A genome-wide verification of such predictions is planned, which should contribute to completing our picture of the extent and prevalence of conserved AS.
Materials and Methods
Training and test datasets.
A comprehensive set of reliably annotated exon–intron structures of mammalian genes, including information about alternative structures as well as conservation in multiple species, was a crucial starting point for our research. The gene annotation system GENOA is a suite of programs for the spliced alignment of sets of mRNA sequences and ESTs against a whole genome and was used to align human and mouse ESTs and cDNA sequences (described in more detail elsewhere [32]). GENOA detects matches between a repeat-masked cDNA sequence and genomic DNA using BLASTN and maps the original cDNA to the assembled human genome using the spliced alignment algorithm mRNAvsGen. Subsequently, it detects BLASTN matches between a repeat-masked cDNA sequence and EST sequences and maps ESTs to regions with cDNA-aligned genomic DNA using SIM4 [33] to ensure a high quality of annotation. SIM4 aligns ESTs with genomic sequences containing the cognate genes, allowing for introns in the genomic DNA sequence and a relatively small number of sequencing errors.
We obtained chromosome assemblies (hg13) of the human genome from the University of California at Santa Cruz Web server (http://genome.ucsc.edu), transcript data in the form of about 94,000 human cDNA sequences from the combined GenBank files of gpri and gbhtc (release 134), and human ESTs from the database dbEST in repository 032703. Overall, GENOA aligned about 86,000 cDNAs and 890,000 ESTs, which resulted in about 20,800 gene regions within the human genome that exhibited multi-exon structures. The relatively low number of alignments was due to enforcement of stringent alignment criteria. Only ESTs that had at least partial overlap with a cDNA were aligned to the genome, and only those alignments that spanned at least one intron and that met stringent coverage (>90%) and identity levels (>90%) were considered. In the same manner, GENOA was applied to the mouse genome, taking version 3 of the assembly and the same releases of GenBank and dbEST as above. With the same criteria as used for the human data, we successfully aligned about 19,000 cDNAs and 480,000 ESTs, leading to 14,800 gene regions.
For candidate gene regions with alternative exon–intron structures, the spliced alignments were compared for each exon. Annotated 5′ terminal and 3′ terminal exons were separated from internal exons and excluded from further analysis. Internal exons were classified as constitutive, alternative 3′ splice site, alternative 5′ splice site, skipped, overlapping, and containing retained introns. With these alignments and the annotation of orthologs from Ensembl [34], we determined orthologous gene pairs containing conserved AS events. Applying stringent quality filters, we identified a set of 241 skipped exons with corresponding U2-type splice sites in both species that had no other detected AS events involving the skipped exon. This set constituted our test set of known ACEs. Out of the 241 exons, five were masked when applying RepeatMasker (A. Smit and P. Green, unpublished data), showing that some classes of conserved mammalian repeats can lead to conserved alternative exons. Among these five, two were SINEs of the mammalian interspersed repeat (MIR) type, one was an L3/CR1 LINE, one was an ERV class I LTR, and one was a small RNA. A larger number of human skipped exons are derived from primate-specific repetitive elements and therefore not conserved between human and mouse [3].
In the same manner, 5,066 conserved constitutive exons in genes exhibiting AS events elsewhere were identified. From these, we took the 5′ and 3′ splice sites to train the pair splice site output distributions in the model. For a training set for the coding states, orthologous human/mouse coding sequences were extracted from Ensembl, and those coding sequences annotated with start and stop codons in both human and mouse were retained. This set consisted of 5,377 orthologous sequences with known reading frame, totaling 7,140,008 nt in human and 7,005,234 nt in mouse. For the pair states, these sequences were aligned with BLASTN [23]. To prevent predicted exons from including stop codons, stop codons were removed from all coding training sequences, which effectively led to an emission probability of zero for stop codons. Finally, a study on the classification of conserved functional versus nonfunctional sequences provided alignments of 63 conserved functional noncoding regions with a total length of 28,959 nt in human and 28,167 nt in mouse [35].
The analysis of the ENCODE target regions (http://www.ensembl.org/Homo_sapiens/encode.html) was based on the 323 genes located in those regions and annotated by Ensembl as reciprocal best hit orthologs in human and mouse (Ensembl v. 22; June 2004). Our analyses used the Ensembl gene structure annotations of these genes. Orthologous introns were determined by concatenating the flanking 30 nt of both the upstream and downstream exons and blasting these exon junction sequences (EJSs) against all other EJSs from the orthologous gene. The EJS pairs were kept if the alignment extended across the junction and included sequences from both upstream and downstream exons. Identical EJS pairs coming from different transcripts of the same gene were consolidated. Introns were not considered if the intron length was smaller than 40 nt, or if at least one of the flanking exons was shorter than 30 nt. This analysis resulted in 1,823 intron pairs, out of which 1,776 were smaller than 30 kb in both species and subject to our analysis by UNCOVER.
For the analysis of intron retention, we focused on intron pairs in which each sequence was shorter than 1,000 nt, and the difference in length did not exceed 20% of the length of the longer sequence. The retained part together with the flanking exons constitutes one large exon, which is subject to the length restrictions observed for mammalian exons. This is a likely reason why the few known conserved cases of intron retention in mammals all involve relatively short introns of less than 500 nt [19]. In addition to the ENCODE target regions, we determined orthologous introns in the complete human and mouse genomes as annotated by Ensembl. Of the 84,233 orthologous introns, 25,074 satisfied these length restrictions and were analyzed by UNCOVER.
pHMMs: Structure, implementation, and training.
HMMs provide a probabilistic approach to a large number of problems in computational biology, and have been applied successfully to diverse topics ranging from gene finding to protein domain modeling [20]. A discrete HMM contains a set of states that emit symbols from an alphabet (here, the four nucleotides) according to a probability distribution. The states are connected by transitions, to which probabilities are assigned. A state in such a HMM has an associated probability of observing each residue, and the transitions determine the possible order of the states. A number of dynamic programming algorithms for HMM training and application are well known. The forward algorithm calculates the total probability that a sequence can be generated by a model, and can be applied to classification problems, with several HMMs representing different classes. The Viterbi algorithm yields the parse of a sequence with the highest likelihood, thus assigning the symbols to model states that may represent different functional categories such as exons and introns. pHMMs are extensions of HMMs, originally described to perform local or global alignments of two sequences [20]. In general, the states of the model now contain probability distributions for an alignment of two residues, and by using several different states, a pHMM can be used to model different patterns of conservation. For example, pHMM systems to identify protein-coding genes [36,37] include different states corresponding to pairs of aligned coding and noncoding nucleotides as well as splice sites. The standard HMM algorithms have been generalized and described in more detail for pHMMs [22,37] or, more generally, phylogenetic HMMs [30,38]. When applying the pHMM Viterbi algorithm, we obtain the optimal parse of the alignment into different functional classes along with the alignment, based on the sequence of states used to generate the optimal alignment.
The pHMM data structures and algorithms were implemented in C++ under Linux, with classes for individual model states and the model itself. A command-line interface allows for convenient training of model states, assembling states into a model, and applying the model to align two sequences. States can be either standard single HMM states or pair states and have an associated output distribution that may have k-order Markov dependence for a small value of k. All single and pair output distributions in the skipped exon model were independently estimated by maximum likelihood using datasets described above. Pseudocounts were added to prevent likelihoods of zero for unseen events, with the exception of the fully conserved U2 splice site dinucleotides and the codon positions (to exclude alignments with stop codons or substitutions of codons of amino acids with very different properties). The Markov order of the output distributions was usually set to one (i.e., the emission probabilities were conditional on the previous nucleotide), with the exception of codon states, where the conditioning was on the previous two nucleotides. As the model topology includes many linearly connected states with a probability of one, only few transition probabilities had to be determined. We derived the gap parameters for functional coding and noncoding sequences from the respective datasets, and manually set the remaining parameters.
With N being the number of states in the model, and L the length of one input sequence, the run time complexity of the pair Viterbi algorithm to compute the globally best alignment is of order N
2
L
2. Thus, many applications of pHMMs, such as comparative gene finding in mammalian genomes, where genes may span across hundreds of kilobases or more, often have to rely on precomputed approximate alignments as input and use the pHMM only to classify and possibly refine the alignment. For the size of most introns, it was practically possible to use the pHMM to compute the optimal global alignment. An optimal pairwise alignment of sequences is usually determined by traceback through the whole dynamic programming matrix. This requires considerable memory resources: the space complexity is O(NL
2), growing quadratically with the size of the input sequences, and for sequences longer than 1,000–2,000 bp each, the forward matrix cannot be fully stored in currently standard main memory any longer. For such sequence pairs, we therefore switched to a divide-and-conquer version of dynamic programming known as the Hirschberg algorithm [39], which reduces the memory requirement to O(NL) at the cost of doubling the run time: the Viterbi algorithm is started twice in both directions from the beginning and end of the sequences, filling the alignment matrix from both ends up to the center column. During this step, only the currently computed and the previous columns need to be kept, discarding columns computed earlier and thus effectively reducing the memory complexity by one dimension. The sum of the two center columns then contains the score of the optimal alignment, and determines one state transition and pair of symbols within the best alignment. The algorithm is then applied recursively to two subproblems, the alignment from the beginning in the upper left corner to the center split point, and from the center split point to the lower right end of the matrix, reducing the size of the problem by half at each step, which leads to a total doubling in run time.
To increase speed, we used the logarithm of the output and transition probabilities, scaled by −100 and rounded to the nearest integer to limit all operations on probabilities to sums of positive integers. This also ensured that no over- or underflow of numbers occurred. Furthermore, summations in the Viterbi matrix were not taken over all states but only over a list of potential predecessors (those with positive transition probabilities). This list was generated upon loading the model, and provided considerable speedup for sparse transition matrices. We aligned all 241 orthologous intron pairs from the ACE set with the pHMM, ranging in size up to about 90,000 nt each. For practical reasons, we restricted the analysis of the ENCODE region to pairs in which both sequences were smaller than 30,000 nt, setting aside 47 intron pairs longer than that.
The polypyrimidine tract upstream of the 3′ splice site sometimes appears as low-complexity sequence, as do parts of protein-coding regions. We therefore masked only repetitive elements and not low-complexity DNA sequences. In addition, masked sequence was unmasked at both ends by 30 nt, to prevent functional elements from being masked because of neighboring repeats. Repetitive sequences are masked with strings of the letter N, which is treated as a special unalignable character that can only be emitted from single (but not paired) pHMM states. This effectively excludes the possibility that any conserved sequence segments cross masked sequence.
Experimental RT-PCR validation.
Primer pairs were first designed to the exonic regions flanking the predicted skipped exon (up to 150 nt on each side). We used the Primer3 software (http://fokker.wi.mit.edu/primer3) with the following typical parameter settings: primer length minimum, 18 nt, desired, 21 nt, and maximum, 24 nt; melting temperature minimum, 55 °C, desired, 58 °C, and maximum, 61 °C; product length, 150–250 nt; and prefiltering of potentially mispriming sequences with the provided library of human repeats. A second round of primers included one primer placed within the predicted ACE and one primer in either the up- or downstream exon. Primer sequences were ordered from Invitrogen (Carlsbad, California, United States).
PCR was carried out with the Invitrogen Taq DNA polymerase kit on an ABI GeneAmp 9700 (Applied Biosystems, Foster City, California, United States), with 40 cycles of separation (30 s at 94 °C), annealing (30 s at 55 °C), and extension (45 s at 72 °C). We used BD Biosciences (San Jose, California, United States) Human MTC Panel I normalized cDNA libraries for eight human tissues and HeLa cell line cDNA. For the latter, first strand cDNA synthesis was carried out by incubating total RNA, isolated using TRIzol reagent (Invitrogen), with an oligo(dT) primer at 65 °C for 5 min for denaturing and then placed on ice for annealing. SuperScript III reverse transcriptase (Invitrogen) was used for reverse transcription. We first tested for presence of the predicted alternative spliced exon in brain and liver cDNA, since these tissues were reported to have the highest levels of AS [32]. If not detected or weak, we tested for expression in the six remaining tissues of MTC Panel I (heart, placenta, lung, skeletal muscle, kidney, and pancreas) and in HeLa cells.
PCR products were separated in 2% agarose gels supplemented with ethidium bromide, DNA was visualized under a UV light, and bands were excised and extracted for sequencing on an ABI 3730 DNA Analyzer (Applied Biosystems) using the QIAquick Gel Extraction Kit (Qiagen, Valencia, California, United States) according to the manufacturer's protocol. For weak bands, we performed a second PCR amplification on the extracted bands as described above, to increase the amount of DNA to levels needed for successful sequencing.
Supporting Information
Dataset S1 Click here for additional data file.
Detailed Information on UNCOVER Predictions Verified by RT-PCR or EST Evidence
(5 KB TXT)
Dataset S2 Detailed Information on Predicted Conserved Coding IREs
(20 KB DOC)
Click here for additional data file.
Figure S1 Expression of Newly Detected Skipped Exons in Different Human and Mouse Tissues
(A) Expression of the newly validated exons in human liver tissue and a HeLa cell line. The sample numbering (i–vii) corresponds to the numbers in Figure 2A; the samples, except for sample v, show expression in brain tissue cDNA. Compared with Figure 2A, it can be seen that the inclusion of the skipped exon is tissue specific rather than ubiquitous. Since the PCR product of sample v in Figure 2A was carried out on HeLa cell line cDNA, the reaction shown here (denoted with an asterisk) was carried out on brain tissue cDNA.
(B) PCR results of newly detected isoforms in other human and in mouse tissues (see Figure 2) and validation of the orthologous mouse exons. Roughly half of these were additionally verified by sequencing of the mouse PCR products.
(20 KB DOC)
Click here for additional data file.
Protocol S1 Detailed Information on Application and Results of the Ka/Ks Test
(29 KB DOC)
Click here for additional data file.
Accession Numbers
The RT-PCR verified sequences described in this paper were deposited in Genbank, under accession numbers DQ102766–DQ102772.
We thank Dirk Holste and Gene Yeo for helpful discussions and for providing the set of 241 skipped exons, and Rong Kong, Vivian Tung, Zefeng Wang, Aniket Schneider, and Grace Zheng for assistance with the experimental validation. This work is supported by a grant from the National Science Foundation to CBB and National Institutes of Health grant R03-LM08536–01.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. UO and CBB conceived and designed the experiments. UO and NS performed the experiments. UO and NS analyzed the data. UO and CBB wrote the paper.
Abbreviations
ACEalternative conserved exon
ASalternative splicing
EJSexon junction sequence
ESTexpressed sequence tag
HMMhidden Markov model
IREintron retention event
NMDnonsense-mediated decay
pHMMpair hidden Markov model
==== Refs
References
Burge C Tuschl T Sharp P 1999 Splicing of precursors to mRNAs by the spliceosomes Gesteland R Cech T Atkins J The RNA world, 2nd ed Cold Spring Harbor (New York) Cold Spring Harbor Laboratory Press 525 560
Black DL 2003 Mechanisms of alternative pre-messenger RNA splicing Annu Rev Biochem 72 291 336 12626338
Ast G 2004 How did alternative splicing evolve? Nat Rev Genet 5 773 782 15510168
Maniatis T Tasic B 2002 Alternative pre-mRNA splicing and proteome expansion in metazoans Nature 418 236 243 12110900
Mironov AA Fickett JW Gelfand MS 1999 Frequent alternative splicing of human genes Genome Res 9 1288 1293 10613851
Brett D Pospisil H Valcarcel J Reich J Bork P 2002 Alternative splicing and genome complexity Nat Genet 30 29 30 11743582
Kan Z States D Gish W 2002 Selecting for functional alternative splices in ESTs Genome Res 12 1837 1845 12466287
Lee C Atanelov L Modrek B Xing Y 2003 ASAP: The Alternative Splicing Annotation Project Nucleic Acids Res 31 101 105 12519958
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
Thanaraj TA Stamm S Clark F Riethoven JJ Le Texier V 2004 ASD: The Alternative Splicing Database Nucleic Acids Res 32 D64 D69 14681360
Kan Z Rouchka EC Gish WR States DJ 2001 Gene structure prediction and alternative splicing analysis using genomically aligned ESTs Genome Res 11 889 900 11337482
Letunic I Copley RR Bork P 2002 Common exon duplication in animals and its role in alternative splicing Hum Mol Genet 11 1561 1567 12045209
Cawley SL Pachter L 2003 HMM sampling and applications to gene finding and alternative splicing Bioinformatics 19 II36 II41 14534169
Modrek B Lee CJ 2003 Alternative splicing in the human, mouse and rat genomes is associated with an increased frequency of exon creation and/or loss Nat Genet 34 177 180 12730695
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
Bejerano G Pheasant M Makunin I Stephen S Kent WJ 2004 Ultraconserved elements in the human genome Science 304 1321 1325 15131266
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
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
Galante PA Sakabe NJ Kirschbaum-Slager N de Souza SJ 2004 Detection and evaluation of intron retention events in the human transcriptome RNA 10 757 765 15100430
Durbin R Eddy SR Krogh A Mitchison G 1998 Biological sequence analysis: Probabilistic models of proteins and nucleic acids Cambridge Cambridge University Press 356 p.
Jareborg N Birney E Durbin R 1999 Comparative analysis of noncoding regions of 77 orthologous mouse and human gene pairs Genome Res 9 815 824 10508839
Alexandersson M Cawley S Pachter L 2003 SLAM: Cross-species gene finding and alignment with a generalized pair hidden Markov model Genome Res 13 496 502 12618381
Altschul SF Madden TL Schaffer 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 Makova KD Li WH 2002 The K(A)/K(S) ratio test for assessing the protein-coding potential of genomic regions: An empirical and simulation study Genome Res 12 198 202 11779845
ENCODE Project Consortium 2004 The ENCODE (ENCyclopedia Of DNA Elements) Project Science 306 636 640 15499007
Ashurst JL Chen CK Gilbert JG Jekosch K Keenan S 2005 The Vertebrate Genome Annotation (Vega) database Nucleic Acids Res 33 D459 D465 15608237
Schmutz J Martin J Terry A Couronne O Grimwood J 2004 The DNA sequence and comparative analysis of human chromosome 5 Nature 431 268 274 15372022
Zhu J Shendure J Mitra RD Church GM 2003 Single molecule profiling of alternative pre-mRNA splicing Science 301 836 838 12907803
Maquat LE Carmichael GG 2001 Quality control of mRNA function Cell 104 173 176 11207359
Siepel A Haussler D 2004 Computational identification of evolutionarily conserved exons Proceedings of the Eighth Annual International Conference on Research in Computational Biology (RECOMB) New York ACM Press 177 186
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 G Holste D Kreiman G Burge CB 2004 Variation in alternative splicing across human tissues Genome Biol 5 R74 15461793
Florea L Hartzell G Zhang Z Rubin GM Miller W 1998 A computer program for aligning a cDNA sequence with a genomic DNA sequence Genome Res 8 967 974 9750195
Birney E Andrews D Bevan P Caccamo M Cameron G 2004 Ensembl 2004 Nucleic Acids Res 32 D468 D470 14681459
Elnitski L Hardison RC Li J Yang S Kolbe D 2003 Distinguishing regulatory DNA from neutral sites Genome Res 13 64 72 12529307
Dewey C Wu JQ Cawley S Alexandersson M Gibbs R 2004 Accurate identification of novel human genes through simultaneous gene prediction in human, mouse, and rat Genome Res 14 661 664 15060007
Meyer IM Durbin R 2002 Comparative ab initio prediction of gene structures using pair HMMs Bioinformatics 18 1309 1318 12376375
McAuliffe JD Pachter L Jordan MI 2004 Multiple-sequence functional annotation and the generalized hidden Markov phylogeny Bioinformatics 20 1850 1860 14988105
Hirschberg D 1975 A linear space algorithm for computing maximal common subsequences Commun ACM 18 341 343
Mulder NJ Apweiler R Attwood TK Bairoch A Bateman A 2005 InterPro, progress and status in 2005 Nucleic Acids Res 33 D201 D205 15608177
|
16110330
|
PMC1185642
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Jul 8; 1(2):e15
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010015
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 10.1371/journal.pcbi.001001605-PLCB-RA-0064R2journal-pcbi-0010016Research ArticleBioinformatics - Computational BiologyNeurosciencePhysiologyEukaryotesAnimalsVertebratesTeleost FishesTheoretical Analysis of Pre-Receptor Image Conditioning in Weakly Electric Fish Image Conditioning in Weakly Electric FishMigliaro Adriana 1Caputi Angel A 2*Budelli Ruben 11 Sección Biomatemática, Instituto de Biología, Facultad de Ciencias, Montevideo, Uruguay
2 Departamento de Neurociencias Integrativas y Computacionales, Instituto de Investigaciones Biológicas Clemente Estable, Montevideo, Uruguay
Friston Karl J. EditorUniversity College London, United Kingdom* To whom correspondence should be addressed. E-mail: [email protected] 2005 15 7 2005 1 2 e164 4 2005 13 6 2005 Copyright: © 2005 Migliaro 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.Electroreceptive fish detect nearby objects by processing the information contained in the pattern of electric currents through the skin. The distribution of local transepidermal voltage or current density on the sensory surface of the fish's skin is the electric image of the surrounding environment. This article reports a model study of the quantitative effect of the conductance of the internal tissues and the skin on electric image generation in Gnathonemus petersii (Günther 1862). Using realistic modelling, we calculated the electric image of a metal object on a simulated fish having different combinations of internal tissues and skin conductances. An object perturbs an electric field as if it were a distribution of electric sources. The equivalent distribution of electric sources is referred to as an object's imprimence. The high conductivity of the fish body lowers the load resistance of a given object's imprimence, increasing the electric image. It also funnels the current generated by the electric organ in such a way that the field and the imprimence of objects in the vicinity of the rostral electric fovea are enhanced. Regarding skin conductance, our results show that the actual value is in the optimal range for transcutaneous voltage modulation by nearby objects. This result suggests that “voltage” is the answer to the long-standing question as to whether current or voltage is the effective stimulus for electroreceptors. Our analysis shows that the fish body should be conceived as an object that interacts with nearby objects, conditioning the electric image. The concept of imprimence can be extended to other sensory systems, facilitating the identification of features common to different perceptual systems.
Synopsis
This paper analyzes the contribution of the body and skin conductance of weakly electric fish in shaping the electric image, using a realistic computational model.
Object recognition, a relevant issue in sensory systems, is not yet fully understood. How pre-receptor mechanisms and interactions between objects shape images is of interest in all sensory systems, leading to general concepts and a specialized jargon. The authors rescue the generality of two concepts for understanding sensory systems. These concepts were introduced early in electroreception research: object perturbed field (change in the basal field generated by the presence of an object), and imprimence (equivalent sources produced at the location of the object). The fish body is an object: generating its imprimence, modifying the basal field, and interacting with other objects. Analogously, the human body interferes, reflecting light or sound in the generation of visual and acoustic images.
The contribution of internal and skin conductivities in image generation has been controversial since the seminal work of Lissmann and Machin in 1958. We show that the high internal conductance of the fish increases and redirects the currents that illuminate objects, increasing and shaping the electric image. Skin resistance mainly influences image magnitude.
Citation:Migliaro A, Caputi AA, Budelli R (2005) Theoretical analysis of pre-receptor image conditioning in weakly electric fish. PLoS Comp Biol 1(2): e16.
==== Body
Introduction
Electroreceptive fish detect nearby objects by processing the information contained in the pattern of electric currents through the skin. In weakly electric fish, these currents result from a self-generated field, produced by the electric organ discharge (EOD). Local transepidermal voltage or current density is the effective stimulus for electroreceptors. The distribution of voltage or current on the sensory surface of the fish's skin is the electric image of the surrounding environment [1–3]. From this image, the brain constructs a representation of the external world. Therefore, to understand electrolocation it is necessary to know the image-generation strategy used by electrolocating animals.
Theoretical analysis of image generation has yielded realistic models that predict with acceptable accuracy the electrosensory stimulus [4–12]. One general conclusion of previous reports is that the skin conductance and the conductivity difference between the internal tissues of the fish and the water are the main factors shaping the electric image: the seminal paper by Lissmann and Machin [13] started a long-lasting controversy about the roles of these factors. Lissmann and Machin argued that if “… the fish has approximately the same conductivity as the water and that it does not appreciably distort the perturbing field (i.e., does not produce an image of the image), the potential distribution around the fish due to the perturbing field can be calculated.” However, several reports [3,7,14] have indicated that the internal conductivity of freshwater fish is high with respect to the surrounding water, and that the high conductance of internal tissues is critical for enhancing the local EOD field as well as for generating the centre-surround opposition pattern that characterizes electric images and that is coded by primary afferents [15].
Experimental studies in pulse gymnotids have confirmed theoretical predictions, showing that the high conductivity of the fish body funnels the self-generated current to the perioral region, where an electrosensory fovea has been described on the basis of electroreceptor density, variety, and central representation [16]. This funnelling effect enhances the stimulus at the foveal region. In addition, two different types of skin have been described in some electric fish of the family Mormyridae: the low-conductance mormyromast epithelium where electroreceptors are present, and the high-conductance non-mormyromast epithelium where electroreceptors are absent [7,17]. The mormyromast epithelium is found on the head in front of the gills, as well as along the dorsum of the back and along the ventral surface of the trunk. The non-mormyromast epithelium is found along the sides of the trunk. This heterogeneity of skin conductance introduces another factor shaping physical electric images.
This article describes a realistic modelling study of the effect of the internal and skin conductance on electric image generation in G. petersii. We have calculated the electric image of a metal object on a simulated fish having different magnitudes of conductances for internal tissues and skin. While the high conductivity of the fish body enhances the electric image by a combination of mechanisms, the skin conductance appears to optimize the transcutaneous voltage modulation by nearby objects. In contrast, transcutaneous current increases monotonically with skin conductivity. These results suggest that transcutaneous voltage is the critical proximal stimulus for electroreceptors.
We generalize two concepts: “object perturbing field” and “imprimence,” introduced early in electroreception research [13], to other sensory systems. An object perturbs an electric field as if it were adding a new field to the basal one. This perturbing field can be considered as equivalent to a certain distribution of electric sources. This distribution is referred to as an object's imprimence.
Results
Electric fields and images generated by metal objects were described in previous reports (reviewed by [3]). In Figure 1, we present results obtained with a realistic fish model and a metal cube, as a reference for the following simulations. Figure 1A shows the basal field, i.e., the field generated by the EOD in the absence of objects. Since all the components of the scene are purely resistive elements, the electric field generated by the EOD has a constant spatial distribution and thus can be described with a static analysis. Therefore, the EOD has been represented by a DC current flowing caudal to rostral along the electric organ (EO). The isopotential lines run closely parallel to the skin, and the distance between them diminishes close to the tip of the “barbillon,” a finger-like extension of the lower lip found in some mormyrid fish. This indicates that field strength and, consequently, current magnitudes are larger at the tip of the barbillon, due to edge effects. The barbillon may be thought of as an “electrosensory fovea” [16,17]. Figure 1B shows the distortion of the basal field (i.e., the object perturbing field) produced by the presence of the cube. This distortion depends on the characteristics of the object and the basal electric field in its neighbourhood. The object perturbing field shown in Figure 1 could also be produced by a set of dipoles oriented almost perpendicular to the fish skin at the point closest to the cube (object imprimence) [13]. The electric image is the difference between the current densities through the skin in the presence and the absence of the object (Figure 1C). Note that the currents increase (positive values) in the skin facing the cube and decrease in a larger surrounding region, producing a “Mexican hat”-type image that can be seen in the graph of Figure 1D, which shows a profile of the image along a line on the sagittal plane. Thus the object image is distributed over a large part of the sensory surface and is not restricted to just the area of skin facing it.
Figure 1 Image Generation in a Fish with Realistic Internal Conductivity and Homogeneous Highly Conductive Skin
(A) The coloured background represents the difference in voltage between each point surrounding the fish and an infinitely distant point, using a non-linear arctangent colour scale (used to highlight values close to zero) shown in the colour bar below for the basal field (in the absence of objects). The black line shows the zero equipotential surface, which is perpendicular to the axis of the EO equivalent dipole distribution.
(B) A similar coloured representation shows the perturbing field (i.e., the field in the presence of the object minus the basal field) produced by a metal cube (1 cm3) close to the skin (0.5 mm). The black line shows the zero equipotential surface, which is perpendicular to the axis of the object equivalent dipole distribution.
(C) Electric image of the metal object depicted in a colour map on the modelled realistic fish from a scorci view.
(D) Electric image along the intersection of the skin with the sagittal plane, illustrating its “Mexican hat” profile.
To study the effect of the skin and internal conductances on the generation of the electric image, we departed from the situation proposed by Lissmann and Machin (1958), in which all fish tissues have the same conductivity as the water. Secondly, we studied the effect of changing internal conductivity, while maintaining a skin conductance that was very high and therefore of negligible effect. Thirdly, for an internal conductivity similar to that experimentally determined, we studied the effect of changing skin conductance as if it were uniform along the fish surface. Finally, we compared results obtained with homogeneous skin conductances and those obtained with the heterogeneous distribution of the skin conductances that is present in G. petersii.
Images as a Function of Fish Internal Conductivity
We have proposed that the low resistivity of the fish body is a very important factor for the shaping of the electric image. To assess its contribution, we simulated electric images for fish having a high skin conductance but with different internal conductivities.
We first modelled a fish with an internal conductivity equal to that of the water, as assumed by Lissmann and Machin [13]. This is described as a “transparent fish.” In this case, the conductivity of the surrounding medium is homogeneous except for the object. Thus, the images calculated as the distribution of the current density across a virtual sensory epithelium are the perturbing fields at the surface of the skin. The basal field generated by the EO is similar to the field of a dipole in a homogeneous medium (Figure 2A). Consequently, and in contrast to the real situation, the isopotentials lines do not run closely parallel to the skin, and the field at the tip of the barbillon does not show an edge effect. Figure 2B shows the perturbing field (the field in the presence of the object minus the basal field) produced by a metal cube close to the skin. The imprimence of the object is equivalent to a certain distribution of dipoles located at the object site, no longer oriented perpendicular to the skin, in contrast to the naturally realistic case (see Figure 1B).
Figure 2 Image Generation in a Fish with Internal Conductivity like That of Water and with a Homogeneous Highly Conductive Skin
The black bars show the zero equipotential surfaces as in Figure 1.
(A) Basal field (in the absence of objects). (B) Perturbing field produced by the same scene as in Figure 1B.
(C) Electric image of the metal object depicted in a colour map on the modelled transparent fish from a scorci view.
(D) Electric image along the intersection of the skin with the sagittal plane.
Comparison of Figures 1 and 2 shows that the direction of the field is nearly parallel to the transparent fish body, and nearly perpendicular to the real fish body. This indicates that the conductivity of the fish body distorts the field produced by the EO. It is worth noting that the internal conductivity of the fish not only funnels the current rostrally but also exerts an effect on the field direction generated at the object location: as a consequence, the electric image is more symmetric. Comparison of image profiles along a sagittal plane (Figures 1D and 2D) shows an enhancement of the image amplitude produced by the presence of the fish body. The body exerts this effect in two ways: a) by increasing the local field in the vicinity of the object, therefore increasing the perturbing field and its imprimence, and b) by introducing an impedance gradient at the site of the sensory surface. Previous research has shown that the amplitude of the image generated by a dipole increases up to two times when the fish/water conductivity ratio increases [18]. To test this mechanism, we calculated the image of a dipole perpendicular to the skin of a transparent fish (Figure 3A, with the negative pole facing the skin), comparing it with the image of the same dipole on a fish with normal internal impedance (Figure 3B). While the waveform remains similar as shown in the current profiles along the skin intersecting with the sagittal and coronal planes, the amplitude of the profile for the realistic fish is twice that for the transparent fish (Figure 3C).
Figure 3 The Effect of Internal Conductivity on the Image Generation of a Dipole
(A) Electric image of a dipole placed at 0.5 mm from a “transparent” fish seen from a scorci view; the modelled dipole axis is perpendicular to the longitudinal axis of the fish.
(B) Same scene as (A) for fish with realistic internal conductivity.
(C) Electric image (transcutaneous current density) along the intersection of the skin with the sagittal plane (left), and the coronal plane (right), for the same dipole as in (A) and (B). Red traces show the images on a transparent fish, while blue traces correspond to a fish with realistic internal conductivity. Note that the ordinate for the realistic fish (left) is twice that for the transparent fish (right).
Figure 4A shows the normalized electric image of a metal cube calculated for fish with different body conductivities. In order to maintain a constant electric source, the tail region was modelled as an independent compartment with realistic internal conductivity. As shown in the normalized images (Figure 4A), both edges shift rostrally with a predominant shift of the rostral border, so that the image becomes wider as body conductivity increases. In addition, the shape of the profile, which initially consists of two main deflections (one caudal positive and one rostral negative), becomes more symmetric, resembling a Mexican hat. The amplitude of the image is an increasing function of body conductance (Figure 4B). These changes are correlated with an increase in the magnitude and a change in the direction of the basal field around the object.
Figure 4 The Effect of Internal Conductivity on Electric Image Generation
(A) Normalized electric images of the same metal cube (identical position) on fish with different internal conductivities. Red: 16.5 μScm−1 (the same as water conductivity), cyan: 165 μScm−1, blue: 1,650 μScm−1, black: 16,500 μScm−1 (normal conductivity), magenta: 165,000 μScm−1. The skin is modelled for all cases, with a homogeneous conductivity of 500,000 μScm−1. The dashed line shows the case of a fish with realistic internal conductivity and skin conductivity distribution. rl, realistic internal conductivity; rlh, realistic internal conductivity, heterogeneous skin distribution.
(B) Peak amplitude of the electric image of a metal cube (1 cm3) placed at 0.5 mm from the fish, as a function of body internal conductivity. The difference in the peak amplitude of the electric image corresponding to the realistic internal conductivity fish shown in this figure and that shown in Figure 1 is due to the use of two compartment bodies (see Materials and Methods).
The Effect of Skin Conductance
To assess the contribution of the skin to image formation, we studied the effect of different uniform skin conductances for a fish with normal internal conductivity. For very low skin conductivity, the transepithelial currents produced by the EO are negligible (Figure 5): the current short-circuits inside the fish because it cannot flow through the skin. The transepithelial current increases with the skin conductivity, approaching an asymptotic value (red trace in Figure 5A and 5C). Since transcutaneous voltage is the quotient of the current density divided by skin conductance, voltage increases differently with skin conductance, rising to a maximum and then decreasing (blue trace in Figure 5B and 5C). The value of skin conductance at which voltage reaches a maximum, 100 μScm−2, is close to the actual measured value for the mormyromast epithelium. This suggests that electroreceptors operate in a voltage detection mode rather than in a current detection mode. The normalized curves in Figure 5D show that the image is smoother and wider as the skin conductance decreases. Continuous traces correspond to uniform skin conductances, where the cyan one is the closest to the mormyromast epithelium value. Realistic electric images were calculated as a reference, using the distribution of conductivities determined experimentally (dotted traces) [7].
Figure 5 The Effect of Skin Conductivity on Electric Image Generation
(A) Transcutaneous current density (electric image) of a metal cube (1cm3) placed at 0.5 mm from the skin, modelled on skin with different conductivities. Red: 10 μScm−2, cyan: 100 μScm−2 (similar to mormyromast epithelium), blue: 1,000 μScm−2, black: 10,000 μScm−2, magenta: 100,000 μScm−2. All these fish have an internal conductivity of 3,300 μScm−1. Dashed line shows the case of a fish with realistic internal conductivity and skin conductivity distribution.
(B) Transcutaneous voltage calculated from the transcutaneous current densities shown in (A), using the same colour code.
(C) Current peak (right axis, red trace) and voltage peak (left axis, blue trace) as a function of skin conductivity for fish with homogeneous skin.
(D) Normalized plot for (B), using same colour code. mel, mormyromast epithelium-like conductivity; rlh, realistic internal conductivity, heterogeneous skin distribution.
Discussion
Animals extract information from the environment and from their own bodies by analyzing changes in the patterns of energy impinging on their sensory surfaces. In that sense, it can be affirmed that to see is to reconstruct visual scenes from a light pattern on the retina or to hear is to extract auditory scenes from sound patterns at the cochlea [19]. Similarly, electric sensing is to reconstruct electric scenes from the pattern of electric currents through the skin.
In electrosensory perception, each object generates a signal that results from the deformation that its presence causes in an electric field. This deformation is a virtual field, called “object perturbing field” by Lissmann and Machin [13]. The object perturbing field is not directly measurable, but computable as the electric field in the presence of the object minus the electric field in its absence, also called “basal field.” As any electric field, the object perturbing field can be considered as caused by an electric source, which is equivalent to the presence of the object.
The “imprimence” of an object, an expression also coined by Lissmann and Machin [13] referring to the electric sources equivalent to the object, not only generates an image but also a change in the field that interacts with other objects. Thus, the effect of a given object not only generates its own image but also modifies the images of other objects [10]. There are theoretical and experimental reasons indicating that the fish body is also an object, and that this is of particular importance since it is an object that is always present as a major determinant of sensory imaging [5,7,18,20]. This leads to the proposition that the fish body, by its presence and movements, constitutes a critical pre-receptor mechanism that conditions sensory signals [16,21]. We, therefore, discuss here the effect of relevant components of the fish's body on image generation.
The Effect of the Fish's Internal Conductivity
The imaging process consists of two steps: imprimence generation (yellow boxes in Figure 6) and image generation (purple boxes in Figure 6). The simplest example occurs in a “transparent” fish, isoconductive with water. The electric image (green arrow in Figure 6A) is the difference between the electrosensory stimulus generated in the presence of an object (light-blue arrow in Figure 6A) and the electrosensory stimulus generated in the absence of that object (dark-blue arrow in Figure 6A). The latter is referred to as the basal stimulus because it is caused by the basal field. Since, in the case of the transparent fish, the basal field is not distorted by the fish body, the electric image results from the projection on the skin of a field perturbation induced only as a consequence of interaction of the object with the basal field (green arrows in Figure 6A). Then, in a transparent fish, image formation can be described as a simple process consisting of two steps: a) the generation of a field by the EO and b) the deformation of this field by the presence of the object.
Figure 6 Schematic Representation of Electric Image Generation
First row, generation of stimulation in the presence of the object; second row, basal stimulation in the absence of objects; third row, sensory image.
(A) Fish with water-like internal conductivity. Imprimence generation (yellow boxes) precedes image generation (purple boxes). A field perturbation (green arrows) is induced as a consequence of the object interaction with the basal field (dark-blue arrows). The electric image is the difference between the perturbing (light-blue arrow) and the basal fields at the skin.
(B) Fish with realistic internal conductivity. The interaction of the body with the field perturbed by the object (red arrows) introduces another component (orange arrow) to the electrosensory stimulus (magenta arrow). The electric image (yellow arrow) is the electrosensory stimulus minus the basal field (blue arrow, representing the sum of the effects of the fish body and the object in the presence of each other). (See Discussion for explanation.)
However, in nature, the basal field is different from that produced by the EO in a homogeneous medium, because it is affected by the inextricable presence of the fish body. Similarly, the object perturbing field is also affected by the fish's body. This interaction (red arrows) produces two extra components that add to the basal electrosensory stimulus (dark-blue arrow): the perturbing field of the object (green arrow) and the perturbing field of the fish body (orange arrow). This resulting field acting on the skin is the electrosensory stimulus (magenta arrow). To calculate the electric image, we subtracted the effect of the basal stimulus (fish body alone, blue arrow). Thus, the electric image (yellow arrow) results from the addition of the perturbing field of the fish's body in the presence of the object (orange arrow) plus the perturbing field of the object in the presence of the fish's body (green arrow). When the object is large enough and surrounds the fish, its effect becomes very important, having a strong influence on the overall pattern of current flow. This is the case when the fish chooses to stay in confined spaces that are frequently its preference in the natural habitat, or in the tube-shaped shelters commonly used in captivity. The fish's positioning of its body in this manner strongly affects the electric images of objects and electrosensory responses [22].
When the object is relatively small or far from the fish body, the loop between the object and the fish body opens, because the influence of the field of the object on the fish body becomes negligible compared to the basal field. Consequently, the scheme of image generation is the same as in the case of the transparent fish. However, the basal field illuminating the object is different than that in the case of a transparent fish and so is the image.
The Effect of Skin Conductivity
The skin conductance is the other important factor shaping the electric image. A homogeneous decrease of the skin conductance causes: a) a decrease of the transepithelial current density, b) an increase of the transepithelial voltage up to a maximum at the range of natural skin conductivity, c) a decrease of the relative slope of the flanks of the image, and d) an increase of the centre region of the “Mexican hat” profile.
For measuring electrosensory stimulus, either local field (equivalent to current flow) or transcutaneous voltage has often been used indiscriminately [1,20,23]. However, our results indicate that current density and voltage are not equivalent stimuli. The transepithelial change in voltage caused by an object is the maximum within the range of skin conductances that are actually measured in the mormyromast epithelium of G. petersii (70–500 μScm−2 [7]), suggesting that the low conductance observed in the mormyromast epithelium might be an adaptation for optimizing voltage sensing. This low conductance of the mormyromast epithelium is caused by a thin layer of tightly packed epithelial cells [24], which makes the mormyromast epithelium up to ten times more resistant than the non-mormyromast epithelium [7]. If receptors electrically shunt their low conductance non-sensitive surroundings, transcutaneous voltage could be considered to be the meaningful parameter of the stimulus. Experimental measurements testing this hypothesis should be done.
Our study also shows a consistent decrease in the relative slope of the flanks of the image and an increase in the centre region of the “Mexican hat” profile with increasing skin resistance (see Figure 5D). In this study, restricted to single conductive objects close to the skin, these changes are rather small, indicating that the main factor for determining object image shape is the internal conductance of the fish body.
The Generality of the Concepts of Object Perturbing Field and Imprimence
Object recognition is an important issue in all sensory systems (including electrical perception), but it is not well understood. The comparative study of different sensory systems leads to general concepts and a language that could potentially be shared by researchers in different systems. In this paper, we focus on peripheral imaging mechanisms, a subject common to sensory systems. We focus in particular on the way in which pre-receptor mechanisms and interactions between different objects in a given scene shape the image.
We emphasize two concepts that were introduced early on in electroreception research [13]: object perturbing field and imprimence. An object perturbs an electric field as if it were a distribution of electric sources. The equivalent distribution of electric sources is referred to as an object's imprimence. Object perturbing field is a concept that relates to reflections and refractances in vision, echoes in audition, etc. In the same way that objects of different impedance than the water modify an electric field, objects in a visual scene modify the illumination. Similarly, echoes and resonances produced by objects modify the distribution of sound in an auditory scene. For example, the sound of a pulsed string on a guitar is greatly modified by the resonance of the box, giving the sound a characteristic timbre. The concept of imprimence can be extended in the same way. Objects producing reflexions, refractances, echoes, and resonances can be considered new sources of energy.
The imprimence produced by the animal's own body acts as a pre-receptor mechanism. The fish body can be considered as an object that interacts with other objects in the scene, generating an imprimence that through the perturbing field modifies the basal field of the scene and, consequently, the imprimence of the other objects. Many species of fish (including Mormyrids) hear underwater due to the imprimence produced by air-filled sacs such as the swimbladder [25]. In a less fundamental way, our body changes the visual image by interfering with and reflecting light, modifying the images of nearby objects. In addition, the interactions between objects and the perturbations of the fields by the imprimences of other objects are used to extract information from a scene. For example, the imprimence of the external ear modifies the incoming sound, allowing for the computation of the altitude of a source [26].
Animal senses explore nature using a limited number of types of energy and receptors with limited dynamic ranges. This constrains and conditions the representation of external reality according to the capabilities of each animal. Humans circumvent these limitations by creating artificial systems, such as radar or sonar, which expand the repertoire of representable qualities of objects. The concepts of imprimence and perturbing field may be applied to the design of artificial sensory systems. It is a common practice to deal with interactions between objects and the perturbations of the fields by the imprimences of other objects as undesirable interference. Nevertheless, evolution has developed neural operations that use images resulting from object interferences as a source of information, in some cases using this to infer object attributes. In these cases, interference between objects may increase the amount of available information contained in the image. Development of the theory of peripheral imaging is a necessary step for the design of computational procedures, allowing the extraction of a larger amount of information from the same signals.
Conclusions
The electric image of an object results from the projection on the skin of a virtual field caused by the presence of an object, in a given electrosensory scene.
The fish's large internal conductance (compared with water) causes a rostral funnelling of the current. This leads to an increase in the imprimence of objects close to the rostral regions of the fish and, consequently, to an increase in the amplitude of their images.
The large difference in conductivity between the inside and outside of the fish forces the field to be almost perpendicular to the sensory surface and, consequently, makes the shape of the image more symmetrical.
An object modifies the field of other objects immersed in the same global field. The fish body itself is a major object, inherent to the process of image generation. Thus, a global field results from the reciprocal interaction between the fish body and nearby objects.
The conductance of the skin changes the shape of the image only slightly, but drives the amplitude (considered as the distribution of transepithelial voltages) close to its maximum, for a given set of other electrical parameters. This result suggests that the high resistance of the mormyromast surface, a property conferred by a thin layer of tightly packed epithelial cells, may serve to optimize object images.
The use of a realistic computational model has allowed us to settle the controversy about the relative importance of the internal and skin conductivities in the determination of the magnitude and shape of the electric image, an issue that has been debated since the seminal paper by Lissmann and Machin [13].
We propose that the concepts of perturbing field and imprimence [13] may be usefully applied to the analysis of other sensory systems and the design of artificial ones.
Materials and Methods
The model.
Simulations were run using a program written to simulate the electric image in weakly electric fish (i.e., the currents through the fish skin), which uses the Boundary Element Method (BEM [27], as proposed by Assad [4], and has been described previously [10,28]). This program allows the determination of the electric field and the electrosensory image in a given environment (scene), calculating the currents through the skin. A scene may include objects (other than the fish) of different conductivity, shape, and size, and is defined by setting the geometry and location of one or more electric fish and objects. Water, internal, and skin conductivities can be specified as required. When the skin conductivity is not homogeneous, different regions can be defined using a graphic interface. Complex shapes, including the fish body, are approximated by a surface composed by triangles. Although the fish shape is kept constant throughout this article, the model allows its modification if required. Once the scene is determined, the potentials and current densities through the skin of the fish and through the objects are calculated. The graphic presentations were made by Matlab standard subroutines.
Changes in internal conductivity and skin conductance.
We studied the effect of the skin and internal conductivity on the electric image in the presence of a metallic (high conductance) cube placed symmetrically to the sagittal plane and facing the dorsal skin 0.5 mm away. Water conductivity was kept at 16.5 μScm−1.
To assess the influence of the internal conductivity of the fish body, different values ranging from that equal to surrounding water conductivity (in which case the fish may be considered transparent) to 16,500 μScm−1 were examined, including the value experimentally determined (3,300 μScm−1). In order to maintain a constant electric source, tail and body regions were modelled as independent compartments, maintaining the tail with a realistic internal conductivity while applying different values for the body. In these cases, the conductance across the model skin was set low enough to be considered irrelevant.
To study the influence of the skin conductance, we explored the effect of different skins with homogeneously distributed conductances ranging from 10–100,000 μScm−2 and a natural-like skin with heterogeneous conductance distribution. The internal conductivity in this case was close to that experimentally determined (3,300 μScm−1).
Two singular conditions were used for comparison purposes: a) when the fish model has experimentally determined conductances (where the fish body exerts its normal effect on the electric image); and b) when it has water-like conductances (i.e., where the fish body exerts no effect on the electric image).
The authors would like to thank Dr. Kirsty Grant and the anonymous reviewers for their helpful comments and suggestions of improvement. This work was partially financed by the Comision Sectorial de Investigación Científica (CSIC), Universidad de la República, Montevideo, Uruguay (fellowship for AM and equipment), and a grant for international cooperation from the French Ministère des Affaires Etrangères, (ECOS-Sud U03B01).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. AM is a postgraduate student (PEDECIBA, Uruguay) whose thesis is being advised by AAC and RB. AAC and RB conceived and designed the experiments, and contributed reagents/materials/analysis tools. AM performed the experiments. AM, AAC, and RB analyzed the data and wrote the paper.
Abbreviations
EOelectric organ
EODelectric organ discharge
==== Refs
References
Bastian J, editor 1986 Electrolocation Bullock TH Heiligenberg W Electroreception New York Wiley & Sons 722 p.
Bell CC 1989 Sensory coding and corollary discharge effects in mormyrid electric fish J Exp Biol 146 229 253 2689564
Budelli R Caputi A Gomez L Rother D Grant K 2002 The electric image in Gnathonemus petersii
J Physiol Paris 96 421 429 14692490
Assad C 1997 Electric field maps and boundary element simulations of electrolocation in weakly electric fish Pasadena (California) California Institute of Technology
Budelli R Caputi AA 2000 The electric image in weakly electric fish: Perception of objects of complex impedance J Exp Biol 203 481 492 10637177
Caputi A Budelli R 1995 The electric image in weakly electric fish: I. A data-based model of waveform generation in Gymnotus carapo
J Comput Neurosci 2 131 147 8521283
Caputi A Budelli R Grant K Bell C 1998 The electric image in weakly electric fish: II. Physical images of resistive objects in Gnathonemus petersii
J Exp Biol 201 2115 2128 9639586
Heiligenberg W 1973 Electrolocation of objects in the electric fish Eigenmannia rhamphichthyidae Gymnotoidei
J Comp Physiol 87 137 164
Lissmann HW 1958 On the function and evolution of electric organs in fish J Exp Biol 35 156 191
Rother D Migliaro A Canetti R Gomez L Caputi A 2003 Electric images of two low resistance objects in weakly electric fish Biosystems 71 169 177 14568217
Rasnow B 1996 The effects of simple objects on the electric field of Apteronotus leptorhynchus
J Comp Physiol A 178 397 411
Assad C Rasnow B Stoddard PK 1999 Electric organ discharges and electric images during electrolocation J Exp Biol 202 1185 1193 10210660
Lissmann HW Machin KE 1958 The mechanisms of object location in Gymnarchus niloticus and similar fish J Exp Biol 35 457 486
Caputi AA Aguilera PA Castello ME 2003 Probability and amplitude of novelty responses as a function of the change in contrast of the reafferent image in G. carapo
J Exp Biol 206 999 1010 12582142
Gomez L Budelli R Grant K Caputi AA 2004 Pre-receptor profile of sensory images and primary afferent neuronal representation in the mormyrid electrosensory system J Exp Biol 207 2443 2453 15184516
Castello ME Aguilera PA Trujillo-Cenoz O Caputi AA 2000 Electroreception in Gymnotus carapo : Pre-receptor processing and the distribution of electroreceptor types J Exp Biol 203 3279 3287 11023848
von der Emde G Schwarz S 2002 Imaging of objects through active electrolocation in Gnathonemus petersii
J Physiol Paris 96 431 444 14692491
Sicardi EA Caputi A Budelli R 2000 Physical basis of distance discrimination in weakly electric fish Physica A 283 86 93
Bregman AS 2001 Auditory scene analysis: The perceptual organization of sound Cambridge (Massachusetts) MIT Press 773 p.
Aguilera PA Castello ME Caputi AA 2001 Electroreception in Gymnotus carapo : Differences between self-generated and conspecific-generated signal carriers J Exp Biol 204 185 198 11136605
Caputi AA 2004 Contributions of electric fish to the understanding sensory processing by reafferent systems J Physiol Paris 98 81 97 15477024
Pereira AC Centurion V Caputi AA 2005 Contextual effects of small environments on the electric images of objects and their brain evoked responses in weakly electric fish J Exp Biol 208 961 972 15755894
von der Emde G 1990 Discrimination of objects through electrolocation in the weakly electric fish, Gnathonemus petersii
J Comp Physiol A 167 413 421
Quinet P 1971 Etude systématique des organes sensoriels de la peau des Mormyriformes (Pisces, Mormyriformes) Annls Musée Royal Afrique Centrale, Tervuren (Belgium) 190 1 97
Yan HY Curtsinger WS 2000 The otic gasbladder as an ancillary auditory structure in a mormyrid fish J Comp Physiol [A] 186 595 602
Hudspeth AJ Konishi M 2000 Introduction—Auditory neuroscience: Development, transduction, and integration Proc Natl Acad Sci U S A 97 11690 11691 11050196
Hunter P Pullan A 2002 FEM/BEM notes Available: http://lola.unimo.it/~fonda/DISPENSA_Tb_html/Programma/Prog_TB_4/Dipolo/fembemnotes.pdf . Accessed 15 June 2005.
Rother D 2003 Simulación de imágenes eléctricas en peces eléctricos de descarga débil Montevideo PEDECIBA-Universidad de la República 45 p.
|
16110331
|
PMC1185643
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Jul 15; 1(2):e16
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010016
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 10.1371/journal.pcbi.001001805-PLCB-PV-0115R1plcb-01-02-06PerspectivesBioinformatics - Computational BiologyEcologyNoneComputational Ecology: From the Complex to the Simple and Back PerspectivesPascual Mercedes Mercedes Pascual is in the Department of Ecology and Evolutionary Biology, University of Michigan, Ann Arbor, Michigan, United States of America. E-mail: [email protected] 2005 29 7 2005 1 2 e18Copyright: © 2005 Mercedes Pascual.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Citation:Pascual M (2005) Computational ecology: From the complex to the simple and back. PLoS Comp Biol 1(2): e18.
==== Body
In 1958, when ecology was a young science and mathematical models for ecological systems were in their infancy, Elton [1] wrote of the “neolithic days of animal ecology, that is to say about twenty-five years ago.” Acknowledging the influence of Lotka and Volterra, he noted, “Being mathematicians, they did not attempt to contemplate a whole food-chain with all the complications of five stages. They took two: a predator and its prey.”
Today, in the era of computational ecological modeling, deterministic systems for two variables—and even a whole food chain—appear like simple idealizations well removed from the complexity of nature. We now consider predator–prey interactions as “consumer–resource” interactions embedded within the large ecological networks that underlie biodiversity (Figure 1) [2]. Consequently, the scale of the problems we model has grown to reflect the world as we now need to observe it. For example, the interplay between ecosystem dynamics and the physical environment that influences global change occurs over a tremendous range of spatial and organizational scales (e.g., [3]). Similarly, the population dynamics of the transmission of infectious diseases often involve spatial or social networks with large numbers of individuals, but the interactions of each individual involve only a subset of the network and can span from local to global distances (e.g., [4–6]).
Figure 1 The Network of Trophic Interactions for Little Rock Lake, Wisconsin
Figure shows 997 feedings links (lines) among 92 taxa (nodes) [2]. The node color indicates the trophic level of the taxon, including (from bottom to top) algae, zooplankton, insects, and fishes; the link color corresponds to the type of feeding link, including herbivory and primary and secondary carnivory. This image was produced using FoodWeb3D software written by R. J. Williams and provided by the Pacific Ecoinformatics and Computational Ecology Lab (www.foodwebs.org).
These examples illustrate the current view of ecological systems as complex adaptive systems [7,8]. Complex adaptive systems are distinguished not only by the multiplicity of components within them, but also by interactions that can be local or distributed among these components and whose rates vary as nonlinear functions of the state of the system itself. One obvious role of computation in the science of complex systems is simply one of synthesis: to reconstruct the whole from the parts as we learn more and more about the components and their interactions. There are obvious limitations to this approach, evident in the famous image of those imperial cartographers who produced a map of the empire of the same size as the empire itself [9]. I argue here that an alternative and more useful role of computation is to address questions on the relationship between dynamics at different temporal, spatial, and organizational scales, that is, to address the importance of variability at small, local scales to the dynamics of aggregated quantities measured at large, global scales. If small-scale “details” matter, we need to ask how much complexity we need to incorporate into large-scale models if we seek to both understand and predict the dynamics of global quantities.
Is it possible to incorporate the effect of small-scale variability without resorting to the “brute force” approach of using higher and higher resolution? I start with examples from theoretical ecology that illustrate problems and approaches related to these scaling questions; I then present more specific examples related to global change and ecosystem dynamics, and end with a series of related problems on the dynamics of large food webs, the ultimate networks of ecological interactions.
From Individuals to Populations: How Local Effects Translate into Global Results
Lotka–Volterra equations and their many descendants assume that individuals are well mixed and interact at mean population abundances. They are mean-field equations that use the mass-action law to describe the dynamics of interacting populations, and ignore both the scale of individual interactions and their spatial distribution. A key question therefore is: can the spatial variability generated at small, individual scales influence the dynamics at larger, population scales? If so, can the effect of smaller scales be represented by simply modifying mean-field models?
Stochastic models such as interacting particle systems [10] can help us examine approaches for scaling up individual-based dynamics [4,11–14]. One of the most useful lessons learned from scaling up detailed models is also pertinent to the related but opposite problem: the formulation of simple models for global variables that still account for the effects of local interactions but do not represent them explicitly (Figure 2). In particular, can we formulate these systems without having to first simplify detailed models? The problem with simplification is that it assumes we know about the components and interactions at “microscopic” levels, but, unfortunately, this information is not often available. For example, we may know that the social network underlying the propagation of an infectious disease is important, but we may not know all the interactions and the specific contacts that led to transmission of the disease from one individual to another. If we were to start with an aggregated model at the population level (one in which the population is aggregated into a few variables describing the total number of infected, susceptible, and immune individuals), how would we formulate it to incorporate variability at spatial or organizational levels that are not explicitly represented?
Figure 2 Bridging Dynamics across Organizational Scales
On the left is a detailed model in which individual interactions in a network are described explicitly. On the right, typical “mean field” models aggregate the population into compartments (here for the three subpopulations of susceptible, infected, and recovered individuals in the dynamics of an infectious disease with permanent immunity). Computational approaches can help us understand the relationship between dynamics at these two different scales, from the individual to the population level. We can start with a stochastic individual-based model and develop approximations that simplify it (A). From this process, we can learn about the opposite direction of formulating simple models directly without sufficient knowledge to first specify the detailed interactions and components (B). These simple models represent implicitly the effect of smaller scale variability.
Recent findings on individual-based models for predator–prey dynamics in a spatial lattice indicate that simpler, low-dimensional models can still be applied at the population level [13,14]. Specifically, the temporal dynamics of global population abundances, aggregated over the whole lattice, can be approximated by mean-field-type equations in which the functional forms specifying the rates of growth and interaction have been modified as power functions. Similar results hold for disease dynamics on spatial and social networks ([15]; M. M. Maule and J. A. N. Filipe, unpublished data; M. Roy and M. Pascual, unpublished data). The rates of transmission of the original mean-field equation are modified to account for deviations from mass action by incorporating nonlinear mixing terms between susceptible and infected populations in which global abundances are raised to a power. Thus, the effect of interactions at local, individual scales can be represented implicitly by changing the shape of the functions describing interactions at global, population levels; that is, the modified framework is structured as if mass action applied when in fact it does not, yet the subtleties of nonrandom mixing are captured at the higher scale. The generality of these findings and the reasons why power-law functional forms yield successful approximations remain to be determined. Another approach based on moment closure techniques has been applied to simplify detailed models by incorporating the effects of variances and covariances on the dynamics of mean (global) quantities [5,11,16–18]; here again, the utility of this approach when the details at small scales are not known remains to be examined, as does the development of statistical methods to fit the models when data are only available for aggregated “mean” quantities.
From Physiology to Ecosystem Dynamics: Global Change Ecology
The problem of incorporating sub-grid-scale processes into large-scale models is found in many other scientific fields in which nonlinearity allows variability to interact across spatial or organizational scales. It also applies to other ecological contexts, in particular to global change ecology and to the spatiotemporal ecosystem models used to represent feedbacks between the biota and the physical environment. At large spatial and temporal scales, the question of essential biological detail quickly becomes computationally intractable. In a recent review on ecosystem–atmosphere interactions, Moorcroft [3] emphasized the problem of scaling from the level of plant physiology to ecosystem-level dynamics to address global climate change questions. One limitation of existing approaches is that within the typical grid cell of existing models, different types of plants compete for resources that are highly homogeneous or averaged over space, generating a tendency for monocultures in the simulated biology [3]. This means that the fine-scale heterogeneity present in real plant communities, which is important for buffering the systems against perturbations, is lost [3,19,20]. Such heterogeneity is generated by small-scale ecological interactions among individuals and their interplay with stochastic physical disturbances, such as fire and gap formation from fallen adult trees. While models that simulate patterns of species community composition from individual plant interactions have been created, their application to ecosystem–atmosphere interactions is computationally prohibitive. To circumvent this problem, approximations to these individual-based models have been formulated in the form of size- and age-structured partial differential equations [21], which are close in spirit to the ideas discussed in the previous section (see also [22]).
Similar connections are being addressed in aquatic environments for phytoplankton, the unicellular primary producers of lakes and oceans estimated to contribute 45% of global net primary production (i.e., the amount of carbon fixed by plants per unit area over time via photosynthesis; E. A. Litchman, C. A. Klausmeier, J. R. Miller, O. Schofield, P. G. Falkowski, unpublished data). The amount of net primary production depends on biological heterogeneity in the form of a taxonomically diverse group of species [23]. A recently developed ecosystem model incorporates different phytoplankton functional groups and their competition for light and multiple nutrients (E. A. Litchman, C. A. Klausmeier, J. R. Miller, O. Schofield, P. G. Falkowski, unpublished data). Simulations of the model at specific sites to explore future scenarios suggest that global environmental change, including global-warming-induced changes, will alter phytoplankton community structure and hence alter global biogeochemical cycles (A. Litchman, C. A. Klausmeier, J. R. Miller, O. Schofield, P. G. Falkowski, unpublished data). The coupling of this type of ecosystem model to global climate models raises again a series of open questions on model complexity and relevant spatial scales of resolution. In fact, similar questions arise not just in the context of climate change but for the general coupling of ecosystem models to large-scale physical models of ocean circulation (e.g., [24]; Figure 3).
Figure 3 Phytoplankton Biomass Generated with a Coupled Biological–Physical Model Developed to Examine the Impact of Nitrogen Fixation in the Atlantic Ocean
In this large simulation [24], the ecosystem model consists of six variables and includes two different functional groups within the phytoplankton, for nitrogen and non-nitrogen fixers. The physical model includes 19 vertical layers but only a coarse horizontal resolution (2° × 2°). In particular, it does not resolve the mesoscale variability of the flows, at characteristic scales of 1 to 100 km, known as the “weather” of the ocean. The lower left panel illustrates the variability of phytoplankton at these smaller turbulent scales, with a simulation of a coupled ecosystem–eddy model (K. Boushaba, G. Flierl, and M. Pascual, unpublished data). We can ask how the effects of these smaller scales can be incorporated in models with a coarser resolution for larger oceanic regions. Even more fundamentally, what are the relevant spatial scales of coupling?
In short, the computational and conceptual challenge is to bridge not only highly disparate temporal and spatial scales, but also organizational ones, from individual physiology to ecosystem biogeochemistry, via community structure and functional diversity (Figure 3). An understanding of how the structure of ecological communities, composed of a diverse array of species, responds to perturbations is a critical intermediate step, which brings me to the next section.
From Structure to Dynamics in Large Ecological Networks
The nonlinear dynamics of large networks is a major challenge in computational biology and complex systems in general [25]. In ecology, the study of networks of species interactions, particularly food webs composed of trophic links, has a long history at the interface of theory and empirical patterns [26–30]. Food webs are the subject of renewed attention today, with improved datasets and the explosion of research on network structure across science (see [31] for a review in ecology). However, the relationship between structure and dynamics in systems that are not just nonlinear and high-dimensional, but also adaptive, remains poorly understood. Food web structure includes the diversity of species, the patterns of connections among them, and the distribution of interaction strengths on these patterns; dynamics encompasses different measures of stability that describe the response of the system to perturbations such as robustness and resilience, which impact the persistence of species [32,33]. This is not a new area but many open questions remain [34]. For example, do rare species and those that interact weakly with others matter to overall species persistence in ecosystems? Simple models with only a few players suggest that rare or weakly interacting species can be important [35]; however, whether these findings still hold true in the bigger, more complex networks we observe in nature is not yet clear. Recent findings suggest a more complex picture in which not just the intensity of interactions, but also their location in the network, matters to (linear) stability and to the invasibility of the community by other species [36,37]. Another recent development is the consideration of other kinds of interactions such as mutualism and parasitism, which can play an important role in ecosystem persistence and bioenergetics [38,39]. The adaptive character of interactions via phenotypic plasticity and evolutionary change challenges traditional dynamical models and our view of structure itself. Adaptive change in the interactions between species influences dynamics and species' persistence, but again, this has been shown primarily in small networks with only a few players [40,41] (but see [42,43]). The spatial dimension has been largely ignored in the dynamics of large ecological networks, although it is clearly a key component of habitat loss and habitat fragmentation (but see [44] for a static treatment).
Stochastic assembly models are perhaps the best candidates to develop a general dynamical theory not only to address open questions on the relationship between structure and dynamics, but also to generate the macroscopic community patterns that ecologists observe in nature and characterize diversity (such as species–area curves and species-rank abundance curves). In these models, macroscopic patterns in diversity arise from the dynamic tension between extinction (as the result of ecological interactions and environmental perturbations) and innovation (as the result of evolution and the immigration of new species from outside the system) (e.g., [45]). Instability at one level of organization can provide the basis for robustness to change at higher levels. For example, in the species assembly model of Solé et al. [45], macroscopic quantities, such as the number of species, reach a stationary state, while at the microscopic level instability is rampant, with recurrent species extinctions and unpredictable population fluctuations. Computational developments are needed to tackle the large parameter space of this type of model and to study the model's behavior using methods that interface mathematical analysis and numerical simulation. Similar issues arise in complementary approaches to link structure and dynamics in food webs, including those that map nonlinear dynamical equations upon a static structure of links between species [46,47]. One promising direction to help us constrain parameter space and build more realistic models involves another fundamental area of ecology, the study of allometric scalings (e.g., [48]). Allometric scalings describe how biological rates vary as a function of size and can be used in the formulation of dynamical models for ecological interactions [49,50]. Ultimately, a better understanding of the critical properties of ecological networks that sustain diverse ecosystems and their functions is of fundamental importance, particularly at this time of rapid environmental change, when perturbations of structure and loss of biological diversity are unavoidable.
The topics described here only begin to illustrate some of the many rich areas for research in computational ecology. We have moved beyond learning more details about the components of complex systems in order to reconstruct their dynamics. Instead, a more fundamental role for computation is found in exploring the relationship between dynamics across scales, in the constant dialogue between simplicity and complexity. Perhaps this is best expressed by what a famous mathematician had to say about a famous macroscopic law of physics: “If our means of investigation became more and more incisive, we would discover the simple under the complex, then the complex under the simple, then again the simple under the complex, and so on, without being able to predict which state would ultimately prevail” [51].
I thank Simon Levin and Andy Dobson for comments on an earlier version of this manuscript, the Pacific Ecoinformatics and Computational Ecology Lab for the food web image, and Victoria Coles, Khalid Boushaba, and Manojit Roy for their help with the other figures. I also thank the James S. McDonnell Foundation for a Centennial Fellowship on Global and Complex Systems.
==== Refs
References
Elton CS 1958 The ecology of invasions by animal and plants Chicago University of Chicago Press 181 p.
Martinez ND 1991 Artifacts or attributes? Effects of resolution on the Little Rock Lake food web Ecol Monogr 61 367 392
Moorcroft PR 2003 Recent advances in ecosystem–atmosphere interactions: An ecological perspective Proc R Soc Lond B Biol Sci 270 1215 1227
Eames KT Keeling MJ 2003 Contact tracing and disease control Proc R Soc Lond B Biol Sci 270 2565 2571
Eames KT Keeling MJ 2002 Modeling dynamic and network heterogeneities in the spread of sexually transmitted diseases Proc Natl Acad Sci U S A 99 13330 13335 12271127
Meyers LA Pourbohloul B Newman MEJ Skowronski DM Brunham RC 2005 Network theory and SARS: Predicting outbreak diversity J Theor Biol 232 71 81 15498594
Levin SA 1998 Ecosystems and the biosphere as complex adaptive systems Ecosystems 1 431 436
Levin SA 1999 Fragile dominion: Complexity and the commons Reading (Massachusetts) Perseus Books Group 250 p.
Borges JL 1964 Dreamtigers. Boyer M, Morland H, translators Austin University of Texas Press 95 p.
Durrett R Levin SA 1994 On the importance of being discrete and (spatial) Theor Popul Biol 46 363 394
Keeling MJ 1999 Correlation equations for endemic diseases Proc R Soc Lond B Biol Sci 266 953 961
Iwasa Y 2000 Lattice models in ecology and pair-approximation analysis Dieckmann U Law R Metz JAJ The geometry of ecological interactions: Simplifying ecological complexity Cambridge Cambridge University Press 227 251
Pascual M Mazzega P Levin S 2001 Oscillatory dynamics and spatial scale in ecological systems: The role of noise and unresolved pattern Ecology 82 2357 2369
Pascual M Roy M Franc A 2002 Simple models for ecological systems with complex spatial patterns Ecol Lett 5 412 419
Gubbins S Gilligan CA 1997 A test of heterogeneous mixing as a mechanism for ecological persistence in a disturbed environment Proc R Soc Lond B Biol Sci 264 227 232
Bolker B Pacala SW 1997 Using moment equations to understand stochastically driven spatial pattern formation in ecological systems Theor Popul Biol 52 179 197 9466960
Pacala SW Levin SA 1997 Biologically generated spatial pattern and the coexistence of competing species Tilman D Kareiva P Spatial ecology: The role of space in population dynamics and interspecific interactions Princeton (New Jersey) Princeton University Press 185 203
Law R Dieckmann U 2000 Moment approximations of individual-based models Dieckmann U Law R Metz JAJ The geometry of ecological interactions: Simplifying ecological complexity Cambridge Cambridge University Press 252 269
Tilman D Knops J Wedin D Reich P Ritchie M 1997 The influence of functional diversity and composition on ecosystem processes Science 277 1300 1302
Kinzig AP Pacala SW Tilman D 2001 The functional consequences of biodiversity: Empirical progress and theoretical extensions Princeton (New Jersey) Princeton University Press 365 p.
Moorcroft PR Hurtt GC Pacala SW 2001 A method for scaling vegetation dynamics: The ecosystem demography model (ED) Ecol Monogr 71 557 586
Pascual M Levin SA 1999 Spatial scaling in a benthic population model with density-dependent disturbance Theor Popul Biol 56 106 122 10438672
Falkowski PG Laws EA Barber RT Murray JW 2003 Phytoplankton and their role in primary, new, and export production In: Fasham MJR, editor Ocean biogeochemistry The role of the ocean carbon cycle in global change 99 121
Coles VJ Hood RR Pascual M Capone DG 2004 Modeling the impact of Trichodesmium and nitrogen fixation in the Atlantic Ocean J Geophys Res 109 C06007
Strogatz SH 2001 Exploring complex networks Nature 410 268 275 11258382
May RM 2001 Stability and complexity in model ecosystems, 1st Princeton landmarks in biology ed Princeton (New Jersey) Princeton University Press 265 p.
Cohen JE 1978 Food webs and niche space Princeton (New Jersey) Princeton University Press 189 p.
Yodzis P 1981 The stability of real ecosystems Nature 289 674 676
Pimm SL 2002 Food webs Chicago University of Chicago Press 219 p.
Polis GA Winemiller KO 1996 Food webs: Integration of patterns and dynamics New York Chapman and Hall 472 p.
Dunne JA 2005 The network structure of food webs Pascual M Dunne JA Ecological networks: Linking structure to dynamics in food webs Oxford Oxford University Press In press.
Pimm SL Lawton JH Cohen JE 1991 Food web patterns and their consequences Nature 350 669 674
Holling CS Gunderson LH 2002 Resilience and adaptive cycles Gunderson LH Holling CS Panarchy: Understanding transformations in human and natural systems Washington (DC) Island Press 25 62
Pascual M Dunne J Levin SA 2005 Challenges for the future: Integrating ecological structure and dynamics Pascual M Dunne JA Ecological networks: Linking structure to dynamics in food webs Oxford Oxford University Press In press.
McCann K Hastings A Huxel G 1998 Weak trophic interactions and the balance of nature Nature 395 794 798
Kokkoris GD Jansen VAA Loreau M Troumbis AY 2002 Variability in interaction strength and implications for biodiversity J Anim Ecol 71 362 371
Neutel AM Heesterbeek JAP de Ruiter PC 2002 Stability in real food webs: Weak links in long loops Science 296 1120 1123 12004131
Lafferty KD Hechinger RF Shaw JC Whitney KL Kuris AM 2005 Food webs and parasites in a salt marsh ecosystem Collinge S Ray C Disease ecology: Community structure and pathogen dynamics Oxford Oxford University Press In press.
Bascompte J Jordano P Melián CJ Olesen JM 2003 The nested assembly of plant–animal mutualistic networks Proc Natl Acad Sci U S A 100 9383 9387 12881488
Hartvigsen G Levin SA 1997 Evolution and spatial structure interact to influence plant–herbivore population and community dynamics Proc R Soc Lond B Biol Sci 264 1677 1685
Peacor S Riolo RL Pascual M 2005 Phenotypic plasticity and species coexistence: Modeling food webs as complex adaptive systems Pascual M Dunne JA Ecological networks: Linking structure to dynamics in food webs Oxford Oxford University Press In press.
McKane AJ Drossel B 2005 Models of food web evolution Pascual M Dunne JA Ecological networks: Linking structure to dynamics in food webs Oxford Oxford University Press In press.
Kondoh M 2003 Foraging adaptation and the relationship between food-web complexity and stability Science 299 1388 1391 12610303
Brose U Ostling A Harrison K Martinez ND 2004 Unified spatial scaling of species and their trophic interactions Nature 428 167 171 15014497
Solé RV Alonso D McKane A 2002 Self-organized instability in complex ecosystems Philos Trans R Soc Lond B Biol Sci 357 667 671 12079528
Chen X Cohen JE 2001 Global stability, local stability and permanence in model food webs J Theor Biol 212 223 235 11531387
Williams RJ Martinez ND 2004 Stabilization of chaotic and non-permanent food-web dynamics Eur Phys J B 38 297 303
Enquist BJ Brown JH West GB 1998 Allometric scaling of plant energetics and population density Nature 395 163 166
De Leo GA Dobson AP 1996 Allometry and simple epidemic models for microparasites Nature 379 720 722 8602216
Emmerson MC Rafaelli D 2004 Predator–prey body size, interaction strength and the stability of a real food web J Anim Ecol 73 399 409
Poincaré H 1902 Science et hypothèse Paris Flammarion 252 p.
|
16110332
|
PMC1185644
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Jul 29; 1(2):e18
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010018
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 10.1371/journal.pcbi.001001905-PLCB-RA-0040R3plcb-01-02-03Research ArticleObstetrics - GynecologyHomo (Human)Inflammatory Aetiology of Human Myometrial Activation Tested Using Directed Graphs Directed Graphs and Uterine ActivationBisits Andrew M 1Smith Roger 1*Mesiano Sam 2Yeo George 3Kwek Kenneth 3MacIntyre David 1Chan Eng C 11 Mothers and Babies Research Centre, Hunter Medical Research Institute, John Hunter Hospital, Newcastle, Australia
2 Departments of Reproductive Biology and Ob/Gyn, Case School of Medicine, University Hospitals of Cleveland, Ohio, United States of America
3 KK Women's and Children's Hospital, Singapore, Singapore
Bourne Philip Academic EditorUniversity of California at San Diego, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 22 7 2005 1 2 e192 3 2005 22 6 2005 Copyright: © 2005 Bisits 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.There are three main hypotheses for the activation of the human uterus at labour: functional progesterone withdrawal, inflammatory stimulation, and oxytocin receptor activation. To test these alternatives we have taken information and data from the literature to develop causal pathway models for the activation of human myometrium. The data provided quantitative RT-PCR results on key genes from samples taken before and during labour. Principal component analysis showed that pre-labour samples form a homogenous group compared to those during labour. We therefore modelled the alternative causal pathways in non-labouring samples using directed graphs and statistically compared the likelihood of the different models using structural equations and D-separation approaches. Using the computer program LISREL, inflammatory activation as a primary event was highly consistent with the data (p = 0.925), progesterone withdrawal, as a primary event, is plausible (p = 0.499), yet comparatively unlikely, oxytocin receptor mediated initiation is less compatible with the data (p = 0.091). DGraph, a software program that creates directed graphs, produced similar results (p
= 0.684, p
= 0.280, and p = 0.04, respectively). This outcome supports an inflammatory aetiology for human labour. Our results demonstrate the value of directed graphs in determining the likelihood of causal relationships in biology in situations where experiments are not possible.
Synopsis
This paper describes how novel computational approaches have been used to test hypotheses for important physiological events when the traditional approaches of animal studies and experiment are not possible. The processes that regulate the onset of human labour are presently unknown, principally because there are no good animal models for human pregnancy and because it is unethical to conduct experiments on pregnant women undergoing labour. However, several hypotheses have been advanced to explain the trigger for labour, including: a functional withdrawal of the hormone progesterone, increased inflammation in the uterus, and increased signalling through the hormone oxytocin. To test these hypotheses the researchers used data on the messenger RNA concentrations of critical variables in samples of uterine muscle taken from 12 women undergoing caesarean section prior to labour and 12 women during labour. Directed graphs for each of the proposed hypotheses were then generated, where the graphs represent the direction of causal influence between different variables. Statistical testing determined how well the graphs of each hypothesis matched the experimental data. The results strongly support an inflammatory origin for the onset of human labour. This approach could be applied to other problems in human biology where the traditional approaches of experiments and animal models are not possible.
Citation:Bisits AM, Smith R, Mesiano S, Yeo G, Kwek K, et al. (2005) Inflammatory aetiology of human myometrial activation tested using directed graphs. PLoS Comp Biol 1(2): e19.
==== Body
Introduction
In most mammals, pregnancy is maintained by high maternal plasma concentrations of progesterone, and labour occurs when progesterone concentrations fall. However, parturition in humans is unusual, as maternal progesterone levels remain high until delivery. The mechanisms regulating human parturition remain obscure. Experimental studies to resolve this uncertainty are restricted because of ethical considerations. In this setting it is not surprising that predictors of premature birth and treatments for preterm labour are generally of limited efficacy [1]. The ignorance regarding human parturition is costly, as rates of premature birth are increasing and premature birth is a major cause of neonatal death and cerebral palsy [2]. Recent advances in causal pathway modelling using directed graphs raise the possibility of advancing knowledge in this area without the need for interventional experiments [3].
We began by gathering knowledge regarding potential variables that might play a part in a causal pathway leading to delivery in humans. Three major hypotheses have been advanced for the onset of normal parturition in humans: an endocrine pathway commencing with a functional progesterone withdrawal, an inflammatory stimulated process, and an oxytocin-mediated mechanism.
In the functional progesterone withdrawal pathway, an unknown factor stimulates expression of the A type of the progesterone receptor (PRA), which acts as a dominant negative repressor of the progesterone-signalling B type receptor (PRB) [4–6]. Functional progesterone withdrawal leads to increased expression of estrogen receptor α (ERα) and, hence, activation of contraction-associated proteins such as oxytocin receptor (OTR) [7] and the prostaglandin synthetic enzyme cyclooxygenase-2 (COX-2) [8]. Support from the literature for this pathway includes the effectiveness of the drug RU486 in initiating labour [9] and recent studies with progesterone administration resulting in a decreased risk of preterm delivery [10,11]. An alternative pathway begins with immune activation and the production of cytokines such as interleukin-8 (IL-8), prostaglandins, and other inflammatory factors such as manganese superoxide dismutase (MnSOD) [12,13]. These inflammatory factors lead to a functional progesterone withdrawal, possibly mediated by nuclear factor-kappa B (NF-κB) [8]. Support for this hypothesis is derived from (1) studies into preterm labour where infection of the genital tract has been implicated as a trigger [14], (2) the use of prostaglandins and mechanical methods (both stimulating inflammation) for the induction of labour [15], (3) the association between systemic infections in the mother in later stages of pregnancy and the onset of labour [16], and (4) evidence for physiological inflammation of the myometrium initiated by foetal alveolar proteins in late pregnancy [17]. For the third alternative, oxytocin has also been suggested as a mediator of myometrial activation, especially since the discovery of local production of oxytocin in the endometrium [18], the marked up-regulation of oxytocin receptors at labour [13], and the introduction of oxytocin receptor antagonists for the treatment of preterm labour [19].
Results
Using the LISREL structural equations modelling approach (with Monte Carlo analysis; Figure 1), the causal pathway modelled in the directed graph in which inflammation as represented by COX-2, IL-8, and MnSOD (p = 0.925; see Figure 1B) as an initiating event was almost twice as likely as the model incorporating progesterone functional withdrawal as an initiating step (p = 0.499; see Figure 1A), and far more likely than the model with an oxytocin receptor-mediated pathway (p = 0.091; see Figure 1C). The directed graph approach to assess the postulated pathways produced similar results: An inflammatory initiation pathway generated an exact p value of 0.684, while the progesterone withdrawal value was p = 0.280, and the oxytocin receptor pathway model p value, 0.040.
Figure 1 Directed Graphs of Messenger RNA Abundances in Human Myometrium
Three models were generated and are represented in the following graphics. (A) Does progesterone withdrawal initiate labour? (B) Does inflammation initiate labour? (C) Does oxytocin receptor mediate onset of labour?
Discussion
Our results lend support to the hypothesis that immune stimulation plays a role in the final weeks of pregnancy, eventually leading to the onset of the sustained coordinated contractions required for normal human labour. Further in vitro support for this pathway has recently been reported; in a myometrial cell line, prostaglandin PGF2α stimulated expression of PRA as predicted by the model in Figure 1B [23]. Additional strengthening of the hypothesis would come from confirmation that protein concentrations for individual inflammatory factors parallel the changes that have been observed in mRNA species. The data do not determine the aetiology of the immune activation, but such a pathway can be relatively easily extended. For example, the role of NF-κB, amniotic fluid surfactant protein A, and stretch can be tested by adding these variables to the causal pathway when data become available. Such a pathway indicates potential sites for therapeutic intervention to alter the process of labour. Labour can also be seen as a withdrawal of the factors maintaining uterine quiescence. From this perspective, inflammation can be seen as a likely factor that extinguishes uterine quiescence. It is also important to note that the cervix and other parts of the uterus may behave differently from the lower segment of the uterus from which our samples are derived. In the future, samples from these sites may provide data on additional variables to extend our knowledge of the pathways of human birth. More generally, the data illustrate the value of causal pathway modelling and directed graphs in biological situations for which experimental studies are problematic for ethical or practical reasons.
Materials and Methods
To explore the alternative hypotheses, we used data obtained from previous quantitative RT-PCR studies of relevant mRNA expression in samples of human myometrium obtained at caesarean section performed either prior to the onset of labour or during active labour (for grouped data and variables see Figure 2) [4,13]. The non-labouring samples were all taken at term, but each woman was almost certainly at different stages of a continuum leading to labour. Tissue slivers (0.5 cm × 1 cm) were obtained from the upper margins of the lower uterine segments (n = 12 N, n = 12 L). QRT-RTPCR was performed as previously described to measure the relative mRNA abundances of 11 genes that have been linked to parturition by previous studies [4,13]. QRT-RT-PCR assays used either SYBR Green (Applied Biosystems, Foster City, California, United States) as a nonspecific intercalating fluorescent dye or specific Taqman probes 5′-fluorescent labelled with either 6-FAM or VIC in a thermal cycler (ABI Prism 7700 Sequence Detector system, Applied Biosystems) linked to a Macintosh G4 (Apple Computer, Sunnyvale, California, United States).
Figure 2 Messenger RNA Abundances in Human Myometrium
Box and whisker plots of mRNA abundances of PRA and PRB, ERα, ERβ, IL-8, COX-2, MnSOD, β2-microglobulin (β2μ), connexin-43 (CX-43), OTR, and the homeobox gene HoxA10 in labouring (L) and non-labouring (N) women.
Data were initially subjected to principal component analysis, which was performed using STATA (Stata Corporation, Collegeville, Texas, United States). Using the raw data, two factors were extracted that explained 61% of the total variance in the data. Factor 1, comprising cDNAs for PRA, ERα, CX-43, IL-8, and COX-2, accounted for 46% of the total variance in the data. Factor 2, comprising cDNAs for HoxA10, OTR, MnSOD, and β2μ, contributed another 15%. Each subject was scored on the basis of these factors, resulting in a graphical plot (Figure 3). Results indicated that 61% of the variance was attributable to nine variables contained in two factors, and this analysis led to a tight grouping of non-labouring samples, while labouring samples exhibited a much larger variability. The wide variability of data in the labouring tissues suggested that this condition was heterogeneous in nature. We therefore focused our pathway analysis on the more homogenous data from the non-labouring samples.
Figure 3 Principal Component Analysis of Messenger RNA Abundances in Human Myometrium
Labouring subjects are shown by diamonds and non-labouring by squares.
The data were transformed using normal equivalent deviates to meet the assumptions of normality while still retaining the variation of the original data (MLwiN Version 1.10.0006, Multilevel Models Project, Institute of Education, University of London, United Kingdom). Using data from the non-labouring samples, we created a directed graph for each of the literature-derived hypotheses according to the methods described by Shipley and Pearl [3,20] and based on earlier work by Wright [21] (see Figure 1).
To assess the relative likelihoods of the alternative models, we used an established program for structural equations modelling, LISREL [22], and an alternative approach, DGraph [3]. The steps for structural equations modelling are to (1) specify a causal pathway, (2) generate a series of equations that are implied by the causal network or pathway, (3) calculate parameter estimates for the equations using maximum likelihood where the objective is to choose parameter estimates that minimize the difference between observed and predicted covariance matrices, (4) calculate a variance-covariance matrix predicted by these equations and calculate a variance-covariance matrix directly from the observed data, (5) calculate the difference between these two variance-covariance matrices, and (6) calculate a probability value for the causal network based on the aforementioned difference, which follows a Chi2 distribution. Where there is a significant discrepancy between the observed and expected covariance matrices, the proposed causal network is unlikely. In order to deal with the relatively small sample size available for this study, using difficult-to-obtain human myometrial samples, a confidence limit for the p-value was calculated using Monte Carlo methods [3].
The second method involves the use of directed graphs and more straightforward calculations. It is also more appropriate for small sample sizes [3]. The idea of directed graphs evolved from work in artificial intelligence. The first step in this process of inference is to formally specify a causal network, known as a directed graph, and shown in Figure 1. The rationale for the term is clear, since an explicit direction of influence is proposed. The second step is to formally acknowledge the causal implications of this graph with a series of independence statements termed “D-separation statements” (Figure 1). Central to the understanding of such causal networks is the concept of conditional independence (Figure 4), i.e., that two variables connected by a third variable, through which the path of influence is mediated A→B→C (Figure 4B), will be independent if the value of the variable B is held constant. The independence statement directly implies the causal path. In an alternative situation A→B←C (Figure 4A), A and C will be independent but will become related if B is held constant. As there are eight variables in each of our proposed networks, there are 8! possible arrangements or independence statements. Because of redundancies, a smaller number of independence statements can specify the entire causal structure. This finite set of independence statements is termed the basis set [3,20].
Figure 4 Interpreting Directed Graphs
(A) Observations on rainfall, hosepipe water, and depth of mud are made. If this causal pathway is correct, the rainfall will correlate with mud depth, and hosepipe water will correlate with mud depth, but rain will not correlate with hosepipe water. However, if mud depth is fixed (also known as “conditioned”), then rainfall will correlate with hosepipe water; rainfall and hosepipe water are said to be conditionally dependent.
(B) Observations are made on the level of water in a tank, water flow in a hosepipe, and depth of mud. Each of these three variables will be correlated. If, however, water flow is fixed, tank water and depth of mud will no longer be correlated.
Each step of a directed graph can be statistically tested in these ways and either accepted or rejected.
The third step is to statistically test the conditional independence statements listed by regressing A on B and C on B, and the residuals generated from these two equations are checked for independence using Pearson correlation or nonparametric tests, depending on the nature and distribution of the data. Since we transformed the raw data into normal equivalent deviates, Pearson correlation was appropriate, producing an exact p value (see Table 1).
Table 1 D-Separation Claims and Associated Partial Correlation Coefficients and Corresponding Probability Values
Variables in each claim are: 1, PRA; 2, PRB; 3, ERα; 4, COX-2; 7, CX-43; 8, OTR; 9, IL-8; 10, MnSOD.
IS, independence statement; p, probability; r, partial correlation coefficient
Finally, the overall plausibility of the model is assessed using a Fisher's C statistic [3]
This follows a Chi2 distribution with 2k degrees of freedom, k being the number of independence statements.
Supporting Information
Accession Numbers
The Swiss-Prot (http://www.ebi.ac.uk/swissprot) accession numbers for the proteins discussed in this paper are β2μ (P61769), COX-2 (P35354), CX-43 (P17302), ERα (P03372), ERβ (Q92731), HoxA10 (P31260), IL-8 (P10145), MnSOD (Q6LEN1), OTR (P30559), PRA (P06401), and PRB (P06401).
This work was supported by the Andrew Thyne Reid Trust and the Macquarie Bank Foundation.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. AMB and ECC conceived, designed, and performed the experiments. AMB, DM, and ECC analyzed the data. AMB, SM, GY, KK, and ECC contributed reagents/materials/analysis tools. RS, AMB, and ECC wrote the paper.
Abbreviations
β2μβ2-microglobulin
COX-2cyclooxygenase-2
CX-43connexin-43
ERestrogen receptor
IL-8interleukin-8
MnSODmanganese superoxide dismutase
OTRoxytocin receptor
PRAA type progesterone receptor
PRBB type progesterone receptor
==== Refs
References
Cole S Smith R Bisits A 2001 Pharmacotherapy for preterm labour Front Horm Res 27 279 307 11450433
Stanley FJ 1994 The aetiology of cerebral palsy Early Hum Dev 36 81 88 8200323
Shipley B 2002 Cause and correlation in biology A user's guide to path analysis, structural equations and causal inference Cambridge Cambridge University Press 29 33 71 79
Mesiano S Chan EC Fitter JT Kwek K Yeo G Smith R 2002 Progesterone withdrawal and estrogen activation in human parturition are coordinated by progesterone receptor A expression in the myometrium J Clin Endocrinol Metab 87 2924 2930 12050275
Haluska GJ Wells TR Hirst JJ Brenner RM Sadowsky DW 2002 Progesterone receptor localization and isoforms in myometrium, decidua, and fetal membranes from rhesus macaques: evidence for functional progesterone withdrawal at parturition J Soc Gynecol Investig 9 125 136
Pieber D Allport VC Hills F Johnson M Bennett PR 2001 Interactions between progesterone receptor isoforms in myometrial cells in human labour Mol Hum Reprod 7 875 879 11517295
Challis JRG Matthews SG Gibb W Lye SJ 2000 Endocrine and paracrine regulation of birth at term and preterm Endocr Rev 21 514 550 11041447
Allport VC Pieber D Slater DM Newton R White JO 2001 Human labour is associated with nuclear factor-kappaB activity which mediates cyclo-oxygenase-2 expression and is involved with the “functional progesterone withdrawal.” Mol Hum Reprod 7 581 586 11385114
Frydman R Taylor S Paoli C Pourade A 1992 [RU 486 (mifepristone): A new tool for labor induction of women at term with live fetus] Contracept Fertil Sex (Paris) 20 1133 1136 12344709
da Fonseca EB Bittar RE Carvalho MH Zugaib M 2003 Prophylactic administration of progesterone by vaginal suppository to reduce the incidence of spontaneous preterm birth in women at increased risk: A randomised placebo-controlled double-blind study Am J Obstet Gynecol 188 419 424 12592250
Meis PJ, Klebanoff M, Thom E, Dombrowski MP, Sibai B et al. 2003 Prevention of recurrent preterm delivery by 17 alpha-hydroxyprogesterone caproate N Engl J Med 348 2379 2385 12802023
Kelly RW 1996 Inflammatory mediators and parturition Rev Reprod 1 89 96 9414445
Chan EC Fraser S Yin S Yeo G Kwek K 2002 Human myometrial genes are differentially expressed in labor: A suppression subtractive hybridization study J Clin Endocrinol Metab 87 2435 2441 12050195
Hay PE Lamont RF Taylor-Robinson D Morgan DJ Ison C 1994 Abnormal bacterial colonisation of the genital tract and subsequent preterm delivery and late miscarriage BMJ 308 295 298 8124116
Chung JH Huang WH Rumney PJ Garite TJ Nageotte MP 2003 A prospective randomized controlled trial that compared misoprostol, Foley catheter, and combination misoprostol-Foley catheter for labor induction Am J Obstet Gynecol 189 1031 1035 14586350
Jeffcoat MK Geurs NC Reddy MS Cliver SP Goldenberg RL 2001 Periodontal infection and preterm birth: results of a prospective study J Am Dent Assoc 132 875 880 11480640
Condon JC Jeyasuria P Faust JM Mendelson CR 2004 Surfactant protein secreted by the maturing mouse fetal lung acts as a hormone that signals the initiation of parturition Proc Natl Acad Sci U S A 101 4978 4983 15044702
Chibbar R Miller FD Mitchell BF 1993 Synthesis of oxytocin in amnion, chorion and decidua may influence the timing of human parturition J Clin Invest 91 185 192 8423217
Thornton S Vatish M Slater D 2001 Oxytocin antagonists: Clinical and scientific considerations Exp Physiol 86 297 302 11429647
Pearl J 2000 Causality: Models, reasoning, and inference Cambridge Cambridge University Press 384 p.
Wright S 1921 Correlation and causation J Agric Res 10 557 585
Jöreskog KG Sörbom D 2003 LISREL, version 8.54 [computer program]. Scientific Software International Available: http://www.ssicentral.com/lisrel/index.html . Accessed 30 June 2005.
Madsen G Zakar T Ku CY Sanborn BM Smith R 2004 Prostaglandins differentially modulate progesterone receptor-A and -B expression in human myometrial cells: Evidence for prostaglandin-induced functional progesterone withdrawal J Clin Endocrinol Metab 89 1010 1013 14764828
|
16110333
|
PMC1185645
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Jul 22; 1(2):e19
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010019
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 10.1371/journal.pcbi.001002005-PLCB-RA-0074R1plcb-01-02-04Research ArticleBiochemistryBioinformatics - Computational BiologyCell BiologyNeuroscienceSystems BiologyEukaryotesAnimalsVertebratesMammalsRattus (Rat)Molecular Switches at the Synapse Emerge from Receptor and Kinase Traffic Interacting Synaptic SwitchesHayer Arnold 12¤Bhalla Upinder S 1*1 National Centre for Biological Sciences, Bangalore, India
2 École Supérieure de Biotechnologie de Strasbourg, Strasbourg, France
Friston Karl J EditorUniversity College London, United Kingdom*To whom correspondence should be addressed. E-mail: [email protected]¤ Current address: Institute of Biochemistry, ETH Zürich, Zürich, Switzerland
7 2005 29 7 2005 1 2 e2015 4 2005 24 6 2005 Copyright: © 2005 Hayer and Bhalla.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.Changes in the synaptic connection strengths between neurons are believed to play a role in memory formation. An important mechanism for changing synaptic strength is through movement of neurotransmitter receptors and regulatory proteins to and from the synapse. Several activity-triggered biochemical events control these movements. Here we use computer models to explore how these putative memory-related changes can be stabilised long after the initial trigger, and beyond the lifetime of synaptic molecules. We base our models on published biochemical data and experiments on the activity-dependent movement of a glutamate receptor, AMPAR, and a calcium-dependent kinase, CaMKII. We find that both of these molecules participate in distinct bistable switches. These simulated switches are effective for long periods despite molecular turnover and biochemical fluctuations arising from the small numbers of molecules in the synapse. The AMPAR switch arises from a novel self-recruitment process where the presence of sufficient receptors biases the receptor movement cycle to insert still more receptors into the synapse. The CaMKII switch arises from autophosphorylation of the kinase. The switches may function in a tightly coupled manner, or relatively independently. The latter case leads to multiple stable states of the synapse. We propose that similar self-recruitment cycles may be important for maintaining levels of many molecules that undergo regulated movement, and that these may lead to combinatorial possible stable states of systems like the synapse.
Synopsis
One of the key cellular changes that accompanies memory formation is a change in the efficacy of synaptic connections between nerve cells. Such changes may arise from long-lasting changes in the number of receptor ion channels at the synapse, and also from changes in their conductance. The authors ask how the cell maintains these changes despite molecular turnover, traffic, and biochemical noise. They use computer simulations as an “in silico” microscope to extrapolate biochemical and light microscopy measurements down to sub-synaptic volumes.
Based on these computer models, the authors propose that there is a self-sustaining switch involving the movement of receptors (AMPA receptors) to and from the synaptic membrane. The switch works because the presence of sufficient receptors at the membrane biases the trafficking machinery to recruit still more receptors. This switch has suggestive parallels with experimental observations of the conversion of synapses from silent to active, which involves AMPA receptor insertion. The authors show that yet more conductance states may arise through interactions with a biochemical switch involving a synaptic kinase (CaMKII).
This receptor switch illustrates how the cell may harness molecular turnover and traffic to maintain, rather than wash out, cellular structures and states.
Citation:Hayer A, Bhalla US (2005) Molecular switches at the synapse emerge from receptor and kinase traffic. PLoS Comp Biol 1(2): e20.
==== Body
Introduction
Long-term storage of neuronal information is believed to occur through alterations in synaptic efficacy. Many mechanisms have been identified for changes in synaptic strength, including modulation of neurotransmitter release, conductivity changes in receptors, changes in numbers of receptors or active synapses, and structural alterations of the synapse. Among these, the insertion of glutamate receptors of the alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionate (AMPA) subtype into the postsynaptic membrane and the modulation of receptor conductance by phosphorylation are key events in modulating synaptic efficacy. A fundamental issue challenges all of these mechanisms: how can they last a lifetime?
Synaptic memories can decay in at least three ways: turnover, diffusive exchange, and stochasticity. Turnover of major postsynaptic molecules ranges from periods of a few minutes to a few days and may be further enhanced by synaptic activity [1]. One solution to loss of memory due to molecular turnover is the concept of self-sustaining molecular switches [2]. These typically involve some form of molecular feedback giving rise to chemical systems, which can stably settle into one of two states. Such two-state, or bistable, systems can store information in a binary manner. Provided there is a steady supply of replacement molecules, molecular turnover can be tolerated, since newly synthesised, naïve molecules become entrained to the current state of the system. Some current proposals for such bistable synaptic switches include the calcium calmodulin type II kinase (CaMKII) hypothesis [3], a mitogen-activated protein kinase (MAPK) feedback loop [4,5], and, recently, the mammalian target of rapamycin (mTOR) protein synthesis loop [6]. Of these, the CaMKII model has been posed in the most detail with the most complete structural correlates. According to this model, CaMKII at the synapse can undergo autophosphorylation, which leads to activation of the kinase. The activated kinase molecules catalyze the phosphorylation of yet more CaMKII molecules, resulting in a self-sustaining cycle. The MAPK feedback loop model also involves a self-sustaining cycle, but in this case several intermediate molecules participate in the loop. The protein synthesis loop model is based on the observation of local protein translation machinery associated with synapses. Messenger RNA for several proteins, including the ribosomes themselves, is also present. Thus, high local protein synthesis creates the machinery for maintaining high levels of synthesis. This protein synthetic loop is regulated by mTOR.
The second mechanism for decay of synaptic memory is diffusive exchange of synaptic proteins, leading to washout of specific states in the synapse. Extrapolations from free diffusion constants suggest that diffusive exchange between the synaptic spine and dendrite is likely to be rapid, under 10 s even for proteins [7]. The postsynaptic density (PSD) is an elaborate cytoskeletal and signalling complex that provides anchors for synaptic proteins close to the region of presynaptic neurotransmitter release. This anchoring solves the problem of free diffusion and washout of active molecules, but introduces the problem of regulating the insertion of molecules into the correct locations. There is considerable evidence for targeted trafficking of molecules to and from the PSD. One such trafficking cycle is the insertion and removal of glutamate receptors of the AMPA subtype into the synaptic membrane [8]. A striking and physiologically important example of receptor insertion is the conversion of silent synapses, lacking AMPA receptors (AMPARs), into active synapses with a full complement of receptors (reviewed in [8]). The delivery of AMPARs to the synaptic membrane involves two streams: a constitutive pathway involving glutamate receptor heteromers 2 and 3 (GluR23), and an activity-dependent pathway involving GluR12 [9]. Based on current evidence, the activity-dependent insertion of GluR12 into the synaptic membrane is stimulated by phosphorylation on Ser845 [10,11]. There is also evidence for such phosphorylation being implicated in synaptic plasticity [12,13]. CaMKII also translocates to the PSD upon calmodulin (CaM) binding and stimulation [14]. Thus, in addition to their known involvement in synaptic plasticity, AMPARs and CaMKII have mechanisms for activity-dependent recruitment to the PSD in a manner that acts counter to washout processes [15]. This combination of attributes makes these molecules interesting candidates for analysing molecular memory mechanisms. Nevertheless, over the long term, even anchoring events are reversible and additional processes must be considered for stability.
A third important obstacle to stable memory formation is biochemical stochasticity. This causes uncertainty (noise) in the outcome of biochemical reactions involving small numbers of molecules. Such fluctuations are severe at the synapse, where many important signalling molecules are present in low numbers, that is, less than 100 molecules. In a typical synaptic volume of 0.1 fl [16] there are an estimated five free Ca2+ ions. Under stochastic conditions, there is a finite probability of spontaneous state flips in bistable molecular switches [7,17]. The lifetime of the stable states depends both on reaction rates and on the number of molecules. For example, the proposed MAPK switch does not fare well in synaptic volumes, and spontaneously flips state on the time scale of minutes [7]. Nevertheless, these time estimates are highly dependent on assumptions about diffusion, anchoring, and the levels of noise in other synaptic pathways.
Putting these themes together, a plausible synaptic memory mechanism might look like a bistable molecular switch that is resistant to turnover, incorporates traffic of molecules to and from the PSD, and is unlikely to spontaneously flip state even when small synaptic molecule numbers are taken into account. In this study we report a novel glutamate (AMPA) receptor–based switch that emerges from a consideration of its traffic and satisfies these criteria. We also examine a possible CaMKII switch in the context of these criteria. Finally, we integrate these switches to explore how multiple synaptic states may arise [18].
Results
Our study proceeded in three stages: model construction and exploration, then examination of regulation and bistability, and finally consideration of interactions between the two forms of bistability.
Model Construction
The key molecules in the simulation were CaMKII, AMPAR, and their main regulators: protein kinase A (PKA), protein phosphatase 2B (PP2B, also known as calcineurin), and protein phosphatase 1 (PP1). As discussed in the methods, parameters for these pathways were derived from previously published models available in the DOQCS database [19] and were refined using known synaptic parameters. Resting Ca2+ concentrations were 80 nM in all models, and at these levels there is very little PP2B or CaM activation. We expanded the models to include translocation steps using simple binding steps with synaptic anchor proteins. These binding steps are assumed to include diffusive movement of molecules between the cytosol and PSD. As shown in Figure 1, activation of CaMKII or its Thr286-phosphorylated state through CaM binding causes it to move toward the PSD. Likewise, phosphorylation of AMPAR on Ser845 causes it to move from internal pools to the synaptic membrane.
Figure 1 Model Structure
(A) Overview of model, indicating key trafficking steps for AMPAR and CaMKII.
(B–H) Chemical reaction schemes for pathways in model. Curved lines with arrows are enzymatic reactions catalyzed by molecules at the curves. Straight lines represent binding or unimolecular reactions.
(B) Details of AMPAR model. The modelled AMPAR is a tetramer with two subunits each of GluR1 (circles) and GluR2 (triangles). There are 16 phosphorylation states each in the cytosol and synaptic membrane. These are represented in expanded form in the lower portion of (B), which shows the internalised pools of receptors. Black filling of the left half of the GluR1 circle indicates phosphorylation of Ser845, and of the right half indicates phosphorylation of Ser831. Endocytosis occurs for the receptors with no GluR1-Ser845 phosphorylation, and exocytosis and degradation occur for the receptors with both GluR1 subunits phosphorylated on the Ser845 site. Exchange of receptors with the bulk AMPAR pool occurs for the unphosphorylated state only, outlined in black.
(C) CaMKII model. The dashed line for phosphorylation of CaMKII–PSD is applicable only for the bistable CaMKII models described in Figure 7.
(D) CaM activation.
(E) PP1 activation.
(F) PP2B (calcineurin) activation.
(G) cAMP formation. The unstimulated phosphodiesterase molecules (PDEs) also degrade cAMP, but at a lower rate than the activated forms illustrated. In the cAMP model we include diffusive exchange of cAMP with a dendritic compartment.
(H) PKA activation.
AMP, adenosine monophosphate; ATP, adenosine triphosphate; I1, inhibitor of PP1; Ng, neurogranin; PKA_inhib, inhibitor of PKA; PKC, protein kinase C; PP2A, protein phosphatase 2A.
Our simulations use six models.
Model 0.
This model includes all the reactions in Figure 1, with three exceptions: (1) the CaMKII autophosphorylation step shown by the dashed line in Figure 1C, (2) the exchange of synaptic AMPAR with the bulk AMPAR in the dendrite in Figure 1B, and (3) the receptor degradation in Figure 1B. This model was used to fit the receptor trafficking curves, and to explore basic regulatory mechanisms.
Model 1: AMPAR bistability.
This model includes all reactions in Figure 1, with the exception of the CaMKII autophosphorylation step in the dashed line in Figure 1C. All parameters are identical to those in model 0. This is the reference bistable model.
Model 2: Skeletal version of AMPAR bistability.
This model is used to understand bistability mechanisms.
Model 3: CaMKII bistability.
This model includes reactions from Figure 1C–1F. It explicitly includes the CaMKII autophosphorylation step in Figure 1C. Rates for a few of the CaMKII reactions are somewhat modified compared to model 1.
Model 4: Combined AMPAR and CaMKII bistability.
This model combines models 1 and 3, using the model 3 parameters where there are differences. The concentration of bulk AMPAR is reduced slightly. This version uses the same PP1 enzyme for AMPAR and CaMKII in the PSD.
Model 5: Combined AMPAR and CaMKII bistability.
This model combines models 1 and 3, using model 3 parameters where there are differences. The concentration of bulk AMPAR is reduced slightly. This version uses distinct PP1 enzymes for AMPAR and CaMKII in the PSD.
Following the initial model construction based on existing models, we wished to parameterize receptor trafficking rates. We represented movement of AMPAR to the PSD as a single binding reaction with an anchor protein. The reverse movement was modelled as a similar reaction releasing AMPAR from the anchor. These simple reactions are approximations to several cellular events, presumably including binding to anchor and diffusive or active movements of the receptor–anchor complex. As we did not have direct biochemical rates for these steps, we fit the reaction rates to published observations of AMPAR movement from labelling and microscopy studies in live-cell preparations. We used model 0 for these calculations since we wished to model AMPAR movement only between the internal and surface-membrane-anchored states. Model 0 does not consider exchange of AMPAR with the bulk, or receptor degradation, and is therefore easier to match to the AMPAR recycling experiments.
To represent the experimental pulse-labelling of receptor, we needed to monitor the movement of a small amount of receptor (the pulse) without perturbing the activity levels of the regulatory molecules in the AMPAR trafficking cycle. Experimentally this is possible because the labelled receptors seamlessly participate in the same reactions as the unlabelled ones. In the simulations, however, a pulse of receptors would result in a displacement from steady state. We therefore simulated these experiments by first computing the steady-state levels of all molecules interacting with the receptor: CaMKII, PKA, PP1, and anchor proteins. We then numerically fixed the levels of each of these interacting molecules to their steady-state levels. Finally, we introduced the test pulse of receptors into the PSD or internal compartment and simulated its movement time course. Since the interacting molecules were held fixed at their steady-state levels, this procedure had the same effect as a pulse of labelled receptors that did not perturb the steady state.
We first matched rates for surface targeting of GluR12 subunits from experimental studies, shown in Figure 2A. The original experiments used fluorescent antibodies [20] and biotinylation [10] to monitor levels of surface and internalised receptors and hence to compute recycling rates. Despite these divergent techniques, the experimental curves for receptor exocytosis matched each other well, and the simple AMPAR movement model was a good fit to both curves.
Figure 2 Matching Models to Trafficking Time Course
(A) AMPAR exocytosis time course; experiments from [10,20].
(B) AMPAR endocytosis time course; experiments from [10,21].
(C) CaMKII internalisation time course; experiments from [14].
(D) CaMKII traffic to PSD; experiments from [14].
In a similar manner, we matched rates for receptor internalisation from Ehlers [10] and Lin [21], shown in Figure 2B. Again, we were able to use a simple model of translocation to obtain a good fit to both sets of experiments.
A similar process was used for CaMKII to match the time course of activity-dependent PSD translocation and removal as monitored by light microscopy [14]. In the case of CaMKII, the role of the anchor protein was served by the N-methyl-D-aspartate receptor (NMDAR) (see Figure 1C). Again, a simple binding reaction was sufficient to fit the experimental data points (Figure 2C and 2D).
Model Exploration
At this point we had a model (model 0) that fit biochemical and cell-biological observations on AMPAR and CaMKII trafficking. This model was descriptive in that it was effective in replicating the data that had been used to set it up in the first place. We then asked whether the model was sufficiently complete to match more complex system effects that had not been used to set parameters for the model. This is often a valuable way to assess whether a model is likely to be a reasonable representation of a complex system.
We first considered AMPAR behaviour at different concentrations of steady Ca2+ (Figure 3A). The resting level of Ca2+ was 80 nM. From the simulations, we found that there was a small decline in the number of AMPA receptors at the synapse when Ca2+ exceeded 300 nM. This was due to the activation of calcineurin, which dephosphorylates the receptor on the PKA site, and hastens its return to the cytosol.
Figure 3 AMPAR and CaMKII Trafficking and Dependence on Steady Ca2+ Concentrations
(A) Number of AMPARs in internal and synaptic membrane pools; AMPARs complexed to enzymes are not counted.
(B) Number of CaMKII molecules in the cytosol and PSD. The activity in the cytosol and PSD starts to rise at about 0.5 μM Ca2+, but translocation occurs around 1 μM.
(C) Conductance of membrane-inserted AMPARs. Receptor conductance is calculated by assuming that CaMKII phosphorylation of a single GluR1-Ser831 of the tetramer gives 1.5-fold basal conductance, and of two Ser831 gives 2-fold basal conductance. The conductance dips at around 300 nM Ca2+, when PP2B is active but CaMKII has yet to become fully active.
We also computed a similar calcium-dependent curve for CaMKII (Figure 3B). Here the cytoplasmic activity of the kinase rose, followed by translocation to the PSD. The initial dip due to PP2B activity was not present in our model of CaMKII. This is because we had very little autophosphorylated CaMKII in the PSD in the basal state, so there was little substrate present for PP2B-activated PP1 to act upon.
AMPAR conductance is a function both of the number of receptors at the surface membrane, and of their phosphorylation state (see Materials and Methods). In Figure 3C we computed conductance. At low Ca2+, it closely tracked the number of membrane-inserted AMPARs. At near 1 μM Ca2+ the conductance of the receptor rose, because CaMKII was activated and phosphorylated the receptor on Ser831. The net effect of these competing events was that AMPAR conductance first declined below baseline, and then rose above baseline. This is consistent with the theoretical Bienenstock-Cooper-Munro (BCM) curve [22], experiments using electrical stimuli [23], and Ca2+-induced plasticity [24]. Thus, model 0 was consistent with a number of experimental observations for which it had not been tuned. However, model 0 did not have the capacity to retain these output changes when the inputs returned to resting levels.
We then used model 0 to analyse AMPAR responses to sustained changes in four parameters that might act as sites of regulation of AMPAR trafficking. The parameters were the activities of CaMKII, PKA, and PP1, and recycling rates of receptor to and from surface membrane. Each of these parameters is implicated in synaptic change, and is a possible upstream control signal for AMPAR conductance. It is known that appropriate stimuli can lead to changes in synaptic conductance over a range of approximately 50% to 200% of basal synaptic transmission levels [25]. We asked which of these parameters could control AMPAR conductance over this range.
In our simulated experiment, we scaled each of the four parameters from 0.1 to ten times its basal value, one at a time. In the case of CaMKII, PKA, and PP1, this scaling was done by numerically buffering the level of the active form of the enzyme to the desired value. In the case of the recycling rates, we scaled endocytosis and exocytosis rates as described below. In each case, the concentrations of the remaining molecular parameters (CaMKII, PKA, and PP1) were allowed to settle to new steady-state concentrations. In biological terms, this would correspond to applying inhibitors or activators of the selected parameter. We recorded the resulting steady-state number of synaptic AMPARs (Figure 4). As in Figure 3, we also computed the AMPAR conductance as a percent of the maximal conductance. The maximal conductance is the conductance if all the receptors were in the membrane in the doubly Ser831-phosphorylated state. In each of these calculations we maintained the total number of internal plus synaptic membrane receptors at 80 molecules, and Ca2+ concentration was at its resting level of 80 nM.
Figure 4 AMPAR Synaptic Membrane Localisation and Conductance in Response to Sustained Inputs
Each panel is computed from a series of steady-state calculations where the activity of the selected input pathway was scaled with respect to its basal activity. The x-axis is this scaling ratio. The conductance is calculated as in Figure 3 and is expressed as the secondary y-axis, as a percentage of maximal conductance (see Materials and Methods).
(A) Changing activity of CaMKII leads to small changes in synaptic membrane localisation of AMPARs, but phosphorylation of GluR1 on Ser831 gives a doubling of synaptic conductance when CaMKII activity is scaled above basal levels.
(B) Low concentrations of PKA result in reduced exocytosis of AMPAR. Basal concentrations of PKA (ratio ~ 1) are required to localise AMPAR to the synaptic membrane, and higher concentrations cause a conductance increase. This occurs because of phosphatase saturation leading indirectly to a rise in Ser831 phosphorylation due to CaMKII. The net effect is that changes in PKA activity can lead to a large change in AMPAR conductance in either direction.
(C) Changes in PP1 concentrations have little effect on AMPAR localisation. However, low PP1 leads to high phosphorylation of GluR1-Ser831 by CaMKII, and hence high conductance.
(D) Lower rates of receptor recycling to the internal pool lead to a small increase in synaptic membrane localisation. High rates bring most of the receptor to the internal pool.
The plots in Figure 4 show the results over the entire range of scaled active inputs, from a ratio of 0.1 to ten times the basal concentration of the input. CaMKII and PP1 (Figure 4A and 4C) had rather little effect on synaptic membrane localisation of receptor. Instead they acted in a complementary manner in changing synaptic conductance through phosphorylation and dephosphorylation of GluR1 on Ser831.
PKA (Figure 4B) had the largest total effect on synaptic conductance, spanning a range from nearly zero to a conductance of nearly 70% of maximal. At low PKA activity there was little insertion of AMPARs into the synaptic membrane, so the conductance was small. At high PKA activity, most of the receptors were inserted. Additionally, the PKA active input indirectly activated CaMKII, leading to receptor phosphorylation on Ser831, resulting in a further increase in conductance. This indirect activation occurs through two successive inhibitory steps. First PKA inhibits PP1 because phosphorylated inhibitor 1 of PP1 binds to and blocks PP1. Second, PP1 itself inhibits CaMKII by dephosphorylation of the kinase (see Figure 1E).
Receptor recycling has been suggested as a mechanism for altering synaptic conductance [9,10]. In the simulations we scaled the AMPAR endocytosis rate by the specified recycling ratio, and simultaneously the exocytosis rate by its inverse. Thus, a ratio of 0.1 would have an endocytosis rate of 0.1 times basal, and an exocytosis rate of ten times basal. Interestingly, it was not easy to drive more receptors into the synaptic membrane (Figure 4D). This was partly because the model already had most of its receptors in the membrane. At high values of receptor recycling rate, the endocytosis rate was greater than the exocytosis rate, so the amount of synaptic-membrane-bound receptor was strongly depleted.
Thus, as an initial prediction, our simulations pointed to either PKA or some combination of CaMKII, PP1, and recycling as being a sufficient long-term control signal to account for bidirectional AMPAR changes, even in a regime where receptor counts did not change. These control effects were not surprising, as these interactions with AMPAR recycling were specifically included into our model. Nevertheless, the model did illustrate the amount of AMPAR insertion or removal to be expected from different manipulations. A similar combination of regulatory inputs has been implicated in learning (e.g., [26]). However, the simulations at this stage did not address the question of how such long-term control signals might be maintained.
AMPAR Bistability
The above explorations of model responses had suggested that the model was reasonably consistent with a range of experimental findings, including several that it had not been tuned for. However, these initial tests used model 0, which did not consider molecular turnover. How might the inclusion of receptor synthesis and degradation alter AMPAR trafficking?
In preliminary simulations (not shown) we added or removed AMPAR molecules from model 0 at the time of starting the model, and asked how the receptors redistributed when the model was run out to steady state. The AMPAR molecules were added to the doubly Ser845-phosphorylated internal pool of receptors, but the site of addition did not affect the final steady-state distribution. Unexpectedly, we found that the addition of receptors to model 0 actually decreased the number of native receptors. Native receptors are defined as unphosphorylated receptors in the endocytosed pool. This suggested that receptors were being redistributed in the model synapse in a manner that might lead to two stable states. We proceeded to test this using a series of simulations on the complete model including AMPAR synthesis and degradation, that is, model 1. Details of model 1 parameters are in Protocol S1.
In model 1 we assumed that newly synthesised AMPARs are present in the dendrite (referred to as bulk AMPAR), and that they exchange with the native receptors in the spine (Figures 1A, 1B, and 5A). We also assumed that there is a slow degradation of the doubly Ser845-phosphorylated receptor pool. Model 1 has 164 anchor proteins located in the PSD. We performed several tests on model 1 to examine whether it indeed exhibited two stable states with different numbers of receptors inserted into the synaptic membrane. We examined influx of receptors into the spine under different conditions. We then performed steady-state analyses, including a parameter sensitivity analysis, to show bistability. We simulated state changes of the model in response to stimuli and stochasticity. These results are described below and in Figure 5, and cumulatively characterise the bistable properties of the model.
Figure 5 AMPAR Translocation and Bistability for Model 1
(A) Simplified schematic of receptor recycling.
(B) Bistability analysis. The flux of AMPARs from the bulk AMPAR pool to the native AMPAR pool is plotted against the total number of synaptic receptors. Receptor influx into the spine occurs both at very low and at high numbers of synaptic AMPARs.
(C) States of the system as a function of Ca2+ concentration. Upper curve is obtained by starting system in state with high numbers of AMPARs in the synapse; lower curve with low numbers of AMPARs. The intermediate threshold curve is calculated using successive bisection as described in the Materials and Methods. Bistability is present when Ca2+ concentration is less than 0.12 μM. Between 0.12 and 0.6 μM the system settles to a state of low AMPAR numbers. Above 0.8 μM the system is in the high state.
(D) States of the system as a function of cAMP concentration. Steady-state number of AMPARs is calculated as in (C). There is hysteresis as the high and low states coexist for the bistable region of the curve. Threshold points (open circles) complete the characteristic S-shaped curve for a bistable system.
(E) Parameter sensitivity analysis. The bars represent the range of parameter values over which the system remains bistable. Other than key regulators, the system tolerates a 2-fold or greater range of most parameters without losing bistability.
(F) Time course of AMPAR showing two stable states. A pulse of 0.2 μM cAMP is applied for 1,000 s to trigger translocation of AMPARs to the synaptic membrane. Following this, cAMP is restored to resting levels, and the system settles to the state of high membrane AMPAR. The “off” stimulus is provided by reducing cAMP to 0 μM for 6,000 s. Following this, the system settles back to the basal state of AMPAR.
(G) Stochastic runs in low and high states. The high state is triggered by an initial cAMP pulse from t = 0 to 4,000 s. The state spontaneously turns off at around 20 h in this run, but the low state does not flip.
(H) Average stability time of low and high states for different numbers of bulk receptors (mean ± standard error of the mean). Twenty-four simulations for each state were run, as in (G). Stability time is calculated as total simulation time in selected state, divided by number of transitions out of that state. Large symbols represent cases where no transitions occurred over the entire set of simulations. As expected, a higher level of bulk receptor increases the likelihood that the spine will spontaneously turn “on”, and vice versa.
(I) Stability time in low and high states for different numbers of anchor proteins. As the number of anchor proteins increases the stability time for both states also rises. At very large numbers of anchor proteins the synapse occasionally turns “on” spontaneously. Symbols and calculations as in (H).
We first asked whether there were two states in which there was an influx of receptors into the spine. This influx would be a necessary condition to offset degradation, and the presence of two such states would be an indication of bistability. We computed receptor flux between the dendrite and the native receptors in the spine as a function of the total number of AMPARs in the synapse (Figure 5B). We defined the total number of synaptic AMPARs as the sum of AMPARs in the internal synaptic pool and in the synaptic membrane (Figure 5A). In order to compute the flux we performed the following manipulation: AMPAR molecules were added to the doubly Ser845-phosphorylated internal pool of receptors. Receptor exchange with bulk AMPAR was disabled to allow the system to settle to steady state for 5,000 s without loss of receptors. Then receptor exchange was re-enabled and the simulation was run for a further 1,000 s to settle. The flux of receptors between bulk AMPAR and the native receptors was calculated at this time point to obtain Figure 5B. This calculation confirmed our preliminary observations. We found that there were two distinct and widely separated regions of receptor influx, one where there were fewer than 20 total synaptic AMPARs, and one where there were 180 or more.
This suggested that there were indeed two stable regimes where receptor influx from the dendrite might balance out receptor degradation. In the regime of low numbers of synaptic AMPARs, there was a simple equilibration between the bulk AMPA receptor pool and the native receptor pool. In the high-number regime, the receptors in the spine were redistributed to the membrane so that the native receptor pool was depleted, again leading to receptor influx. The presence of two regimes of receptor influx, depending on the number of synaptic AMPARs, may be an experimentally testable prediction.
To confirm that this formed a bistable system, we computed stable states of the system under different regulatory conditions (Figure 5C and 5D). We used Ca2+ and cyclic adenosine monophosphate (cAMP) as regulatory inputs. We obtained the stable states by running model 1 to steady state (120,000 s), starting from either a low state (low numbers of synaptic AMPARs) or a high state (high numbers of synaptic AMPARs). In cases where there was only one stable state, the two runs converged to the same steady value. In cases where there were distinct stable states, the two runs settled to their respective high or low stable states. We also found the unstable fixed point (the threshold for state switching) by using an iterative bisection method described in Materials and Methods.
These simulations give us dose-dependence curves that illustrate the bistable nature of the system. The Ca2+ dose dependence of the switch is interesting and unusual (Figure 5C). In the low Ca2+ regime, the system is bistable. The system settles into either a high or low state depending on initial conditions. In the 0.12 to 0.6 μM range, the system goes into a single low state of activity because of the action of calcineurin (PP2B). Calcineurin is activated at these concentrations of Ca2+, and is able to rapidly dephosphorylate AMPAR. The unphosphorylated receptor moves into the native receptor pool, and then out to the dendrite. At Ca2+ concentrations over 0.8 μM, the system goes into a single state of high activity because of CaM activation. The activity of CaM leads to increased AMPAR phosphorylation both through PKA and CaMKII. CaM-activated adenylyl cyclase (see Figure 1) produces cAMP, increasing PKA activity. CaM also directly activates CaMKII. These events are similar to those described for Figure 3.
In this manner, applied Ca2+ can flip the state of the switch in either direction, depending on Ca2+ amplitudes. The switch is unusual because both states are reached by an increase in the regulatory concentrations of Ca2+. As considered below for cAMP, it is much more common for one state to be triggered by low regulator concentrations, and the other state by high regulator concentrations. This bidirectional regulation by increases in Ca2+ is also observed in the findings shown in Figure 3A, but is not as striking. It is possible that there may be a very narrow bistable region at around 0.7 μM Ca2+ in the model, as suggested by the small separation between the low and high curves, but this was not within the numerical resolution of our calculations. We do not expect that such a fine separation would be biologically relevant. We also estimated the thresholds for the bistable switch (open circles in Figure 5C). These were not very dependent on Ca2+ concentrations. In biological terms, the synapse could be switched to the low state by raising Ca2+ to the low regime (0.12 to 0.6 μM), allowing the flux of receptors to be initiated, and rapidly lowering Ca2+ back to the bistable regime. To attain the high state, the Ca2+ input should be over 0.8 μM for long enough for the switch to settle, and then Ca2+ should rapidly fall below 0.1 μM into the bistable region. As we discuss later, the details of biological Ca2+ dynamics are beyond the scope of our current steady-state models.
The cAMP dose dependence of the switch was more conventional. It took the shape of a simple hysteresis curve, where the low state resulted from a decrease in cAMP and the high state from an increase (Figure 5D). The bistable region of the switch is in an intermediate range of cAMP. In this case the synapse switches to the low state if cAMP is reduced below 20 nM, and to the high state if cAMP is raised above 35 nM. To complete the analysis, we found the unstable fixed points of the bistable switch (open circles in Figure 5D). These points can be interpreted as thresholds for switching from one state to the other. As expected for a bistable system, these unstable fixed points curve back toward the limits of the hysteresis cycle in an S-shaped curve (Figure 5D).
How “robust” is the bistability of model 1? One measure of this is to ask whether the bistable effects persist when important model parameters are varied. We systematically varied important model parameters and looked for bistability. To test for bistability, we started the model off in either the upper or lower state, then ran it out to steady state with the altered parameters. If the model switched state it was no longer bistable. We found that model 1 retained its bistable behaviour over a wide range of most parameters, illustrated in Figure 5E. The key “sensitive” parameters were exactly those identified in Figure 4 as key regulators of steady-state synaptic conductance: CaMKII, PKA, recycling, and Ca2+. Most parameters were able to scale a factor of at least two up or down without losing bistability.
As a simpler signature of bistability we ran a time-course simulation in which the model explicitly switched between two steady states (Figure 5F). Here the model first settled to the low state where there were few AMPARs in the synaptic membrane. Following a cAMP pulse (0.2 μM, 1,000 s), the model switched to the high state, with many AMPARs inserted into the synaptic membrane. The system was switched back to the low state by numerically reducing cAMP to zero (6,000 s). While this switching of the number of membrane-inserted AMPARs is possibly a testable prediction, it only indicates the presence of bistability and does not shed much light on the mechanism, which is analysed below.
The time course of switching was slow, of the order of an hour. This was consistent with the receptor trafficking rates in the model, which were derived from steady-state measurements. More rapid transient rates may be applicable during synaptic stimulation, as we consider in the Discussion.
Another manifestation of robustness is the ability of the model to retain state information despite stochasticity. Stochasticity in the synapse originates from the probabilistic occurrence of reactions among small numbers of molecules and gives rise to apparent biochemical noise. This noisiness imposes severe constraints on the reliability of any proposed synaptic signalling mechanism [7]. In particular, bistable biochemical systems are subject to spontaneous state flips because of biochemical noise [7,17]. We tested stochastic responses by simulating model 1 using the Gillespie exact stochastic method [27]. The entire model was simulated stochastically, including all molecules in the dendrite, spine head, and PSD (details in Materials and methods). We started the model in the low state, where few AMPARs were inserted into the synaptic membrane (Figure 5G). In half the runs we applied a cAMP stimulus to switch the model to the high state at around 1 h (black line in Figure 5G). We then simulated the model for a period of 120,000 s (>33 h) to test its resistance to spontaneous switching from either state. In the example in Figure 5G, the low state (gray line) did not change, but the high state (black line) spontaneously turned off at around 16 h.
Based on our analysis shown in Figure 5B, the stability of each state should be a function of the concentration of bulk AMPAR. If the concentration of bulk AMPAR is high, then receptor influx increases. Under these conditions, a relatively small fluctuation should push the system past the efflux regime into the upper influx regime (Figure 5B). Conversely, at low bulk AMPAR it should be easy to flip from the high to the low state. We repeated our stochastic simulations for a range of bulk AMPAR concentrations while otherwise retaining the parameters of model 1. At each concentration of bulk AMPAR we repeated the simulations at least 24 times to build up a profile of the switching times (Figure 5H). At our reference range of 11.11 nM of bulk AMPAR, the off state was stable for more than 360 h on average, and the on state was stable for about 42 h. As we discuss below, other synaptic processes may take over the job of maintaining state information, within this time frame.
At low bulk AMPAR the high state was very short-lived, but the low state did not flip at all during the entire duration of our simulations (indicated by the large symbols in Figure 5H). Conversely, at high bulk AMPAR, the low state was unstable but the high state lasted for very long times.
We then considered how the lifetime of model states might scale with the number of anchor proteins in the PSD. As the number of anchor proteins sets the maximum number of receptors that can be inserted into the membrane, this parameter is important for the robustness and stability of the model. We simulated the stability time of the switch for a range of anchor protein numbers (Figure 5I). The default number of anchor proteins in the model was 164. At lower anchor protein numbers the switch lifetime was rather short, of the order of a few hours. At higher anchor numbers the switch lifetime increased rapidly for both states. The lifetime of the high state continued to rise and exceeded 2 mo when there were more than 320 anchor proteins (large symbols in Figure 5I indicate that the model did not change state during the entire duration of our simulations). The lifetime of the low state was greatest (approximately 2 mo) at 240 anchor proteins, and then declined to around 200 h when more anchor proteins were present. This may have occurred because the presence of additional anchor proteins biased the movement of the receptor toward the spine. Biologically, an increase in anchor protein number may correlate with the size of the spine head. Thus, our model predicts that larger spines should be more stable, an observation that has some experimental support [28,29].
At this stage of the study we had extensively analysed the properties of the bistability of model 1 with respect to the number of AMPARs at the synapse. We had considered state dependence on flux and regulators. We had shown that the bistability was robust, and in particular had shown how long the model could retain state information when stochasticity was taken into account. Based on these calculations, we suggest that the model might be a candidate for retaining synaptic state information for hours to months.
How does the bistability arise? Model 1 is quite complex and involves 16 phosphorylation states of AMPAR each in the internal pool and in the synaptic membrane. For simplicity, we made a skeletal model with the same general topology. The skeletal model retained only two phosphorylation/dephosphorylation steps involving AMPAR (Figure 6A). This is called model 2. We used the same parameters as for the full model, with the exception of lower concentrations of PP1 (0.333 μM). The concentration of PP1 was reduced as its other substrate, CaMKII, was not present in model 2. The K
m for PKA and PP1 was halved as compared to model 1, as each receptor phosphorylation site in model 2 corresponds to two receptor phosphorylation sites in model 1. For model simplicity we had the degradation steps feeding into the bulk AMPAR pool, which was numerically buffered to a steady value. The full parameters for this model are presented in Protocol S1. We used this simpler model to analyse the mechanistic basis of this form of bistability.
Figure 6 Simplified AMPAR Bistable Model
IR represents internal receptor, MR represents synaptic-membrane-localised receptor and asterisks indicate phosphorylation at Ser845. PKA is protein kinase A.
(A) Complete reaction diagram.
(B) Bistability analysis for simplified model. AMPAR flux between the bulk AMPAR and the IR state from (A) is plotted against the total number of synaptic receptors (IR + IR* + IR** + MR + MR* + MR**). As seen in the complete AMPAR model in Figure 5B, there are two regions of receptor influx into the spine, at low and high numbers of synaptic AMPARs. The zero crossings are stable states where there is no net flux of receptor.
(C–F) Time courses of key molecules in simple model. After an initial settling period, PKA is raised to 40 molecules for 2,400 s to trigger receptor influx. After the system settles into the high state, PKA is set to zero molecules for 3,600 s, to return the system to basal levels. These stimuli are indicated by horizontal bars along the time axis.
(C) Internal receptor numbers. The number of receptors in the unphosphorylated form (IR) remains very close to the bulk receptor level except during transitions, when receptors enter or leave the system.
(D) Synaptic-membrane-bound receptor levels.
(E) Numbers of free PP1 decline sharply during the high state, because of enzyme saturation.
(F) Numbers of PP1 complexes with substrates. The high amount of PP1–MR** complex is complementary to the decline in free PP1, showing the saturation of the phosphatase.
We repeated the analysis of receptor flux as a function of the number of total synaptic receptors. Model 2 also had two regions of AMPAR influx separated by a region of efflux from the synapse (Figure 6B). This indicated that it shared the same mechanism for bistability as model 1. We then performed a simulation of the time course of stimulus-triggered transitions between the stable states of the model. We took advantage of the smaller number of molecular species in model 2 to monitor all the molecular concentrations during PKA-triggered transitions between the low and high states. In this simulation the steady-state amount of PKA was one molecule. The switch was turned on using a stimulus of 40 active PKA molecules for 2,400 s, and turned off using zero active PKA molecules for 5,000 s. The responses of several molecules are illustrated in Figure 6C–6F. The low stable state (0 to 3 h) is characterised by low numbers of all forms of the receptor, and consequently little saturation of PP1. During the switch to the high state at 3 h, the concentration of unphosphorylated internal receptor drops, leading to a large influx of receptors. These are rapidly phosphorylated by the high numbers of PKA, and the number of IR** (see Figure 6 legend for explanation of abbreviations) rises sharply.
As the IR pool exchanges rapidly with bulk AMPAR, its concentration rapidly returned to basal levels after the PKA stimulus ended. Once in the IR** state, the receptor translocated to the synaptic membrane, into the MR** pool. Due to the large numbers of MR**, the PP1 became saturated (Figure 6E and 6F). Thus, the combination of PKA phosphorylation of MR, and translocation from the IR** pool, formed MR** at a greater rate than the PP1 could dephosphorylate it.
At 7 h we switched the system back to the low state by removing all PKA. This allowed PP1 to complete dephosphorylation of the phosphorylated receptor pools. There was a large transient rise in MR due to the dephosphorylation of MR** and MR*, and the slow traffic back to the IR pool. Finally, when we restored active PKA to its resting level the system settled back into the lower stable state.
The translocation step appears to be important—we were unable to obtain bistability without it—but it was not possible to completely explore parameter space so it was not clear whether this is an absolute requirement. We were also able to obtain bistability with a single phosphorylation step, provided that the translocation step was second order in the receptor (data not shown). In all these processes, it was assumed that the bulk AMPAR was constant, that is, that the balance of synthesis and degradation was sufficient to rapidly add and remove receptors from the spine. In the biological system the situation is more complex and synthesis itself may be activity dependent [30–32].
In summary, the simple model retains the fundamental features of the AMPAR translocation-based bistability and facilitates an analysis of its mechanism. The low state of this form of bistability occurs when few synaptic AMPAR molecules are present, so that PKA can act on only a few substrate molecules, and PP1 is not saturated. Therefore, few internal AMPARs are in the phosphorylated state and only a few AMPARs are translocated to the surface. The upper state of activity is characterised by high numbers of AMPARs in the phosphorylated states both internally and in the synaptic membrane. This state persists because of the higher basal activity of PKA as compared to PP1. This leads to PP1 saturation. The final step in maintaining the upper state is the translocation of phosphorylated receptor to the membrane. This removes receptors from the internal pools and keeps the native receptor (internal receptor) at sufficiently low levels that receptor influx is favoured. A similar PP1 saturation effect is seen in the complete model (model 1) when it is in the high state (data not shown). This is a possible testable prediction of the model.
CaMKII Bistability
Having characterised AMPAR translocation bistability, we wished to examine the well-studied CaMKII autophosphorylation system in the context of translocation. Although model 1 included CaMKII translocation, the CaMKII portion of the model was not bistable. A key assumption for CaMKII bistability is that upon binding to NMDAR, the CaMKII becomes susceptible to autophosphorylation even in the absence of bound CaM [33]. In our model, as in others (e.g., [34]), this assumption is also linked to PP1 saturation as we now describe. In order to analyse the CaMKII responses in a simpler context, we derived a reduced model from the basic model discussed above. This reduced model contained only CaMKII in the cytosol and PSD, and its immediate regulators, PP1, CaM, and PP2B. PKA was included only as the active enzyme, without regulatory steps. We modified this reduced model of CaMKII by including the autophosphorylation in the absence of CaM. These changes are illustrated by the dashed line in Figure 1C and the bold lines in Figure 7A. This model is called model 3.
Despite the simplicity of our CaMKII model as compared to previous work [34,35], we were able to obtain bistability. To do so we used somewhat different phosphorylation and dephosphorylation rates in the PSD as compared to the model 1 (see Protocol S1). Since CaMKII phosphorylates itself, we had to adapt our previous analysis, which relies on separate input and output molecules [4]. We did this by numerically bifurcating the autonomously active CaMKII–PSD (Aut-CaMKII; Figure 7A) into two simulated molecular pools: Aut-CaMKII enzyme and Aut-CaMKII readout. We set the number of Aut-CaMKII enzyme pools to specified values, and monitored the number of molecules in the Aut-CaMKII readout pool (Figure 7B). This manipulation was facilitated because the level of Aut-CaMKII is computed as the sum of the autonomously active states of Thr286-phosphorylated and Thr286/Thr305-phosphorylated CaMKII, indicated in gray in Figure 7A. So our enzyme assignment bypassed this summation, and directly set the number of Aut-CaMKII molecules. Our readout number was the sum of Thr286-phosphorylated and Thr286/Thr305-phosphorylated CaMKII.
The results of these calculations for a range of Aut-CaMKII enzyme values are shown in Figure 7B. The intersection points of this curve with the 45° line are stable points of the system, because at these points the enzyme and readout activities of Aut-CaMKII are identical. In other words, at these points the autonomous CaMKII would exactly sustain its own activity. The upper and lower intersection points define the stable numbers of Aut-CaMKII, and the intermediate point is a transition point. This behaviour can be seen by considering a small increase in the number of Aut-CaMKII enzyme above one of the stable points. The resulting Aut-CaMKII readout (read off from the y-axis) would be smaller than the new input number. This would tend to restore the CaMKII activity toward the stable point. A similar argument applies to small negative deflections from the stable points. Around the transition point the situation is reversed: any small deflection will be amplified until the system converges to either the upper or lower stable point.
We confirmed the presence of bistability by simulating a time series in which the system was turned on with a calcium pulse of 2.7 μM for 500 s and later turned off by raising the k
cat of the PSD-localised PP1 by 5-fold for 500 s (Figure 7C). Two distinct stable states were observed, which corresponded to the upper and lower intersection points in Figure 7B. There is a small offset between the two calculations because Figure 7C reports all forms of CaMKII in the PSD, whereas Figure 7B refers only to Aut-CaMKII.
We evaluated the robustness of the CaMKII model (model 3) using the same approach as for the AMPAR model (model 1). We found that the CaMKII bistability was highly robust with respect to variation of parameters (Figure 7D). Many parameters could be varied from 0.1 to ten times the reference value without losing bistability. Most of the remaining parameters could be varied from 0.5 to two times the reference value, and only PP1 and PKA were more sensitive. This sensitivity reflects the key role of PP1 in dephosphorylating CaMKII, and the role of PKA in controlling the activity of PP1.
We checked the robustness of model 3 in synaptic volumes by simulating it using stochastic numerical methods. Model 3 was resistant to spontaneous switching and we did not observe any switches to either state in over 300 cumulative hours of simulation time. A 33-h sample of the high and low states is shown in Figure 7E. This state stability turned out to be an artefact of our reduced model for CaMKII, which only used the final active concentration of PKA as one of the key regulatory inputs. The PKA pathway model output was quite noisy in small volumes [36]. When we incorporated the full PKA pathway into model 3, we found it introduced a considerable amount of additional stochasticity into the system and did lead to bidirectional state flips (Figure 7F). We repeated these simulations 50 times and found that the off state endured for 17.3 ± 2.5 h, and the on state for 37.2 ± 6.9 h (Figure 7G). Thus, there may be a marked reduction in bistable state lifetimes when noisy inputs are taken into account. This issue is considered in the Discussion.
Figure 7 Bistability in the CaMKII Model
CaMKII-thr286* indicates Thr286-phosphorylated CaMKII, and CaMKII-thr286*-thr305* indicates Thr286/Thr305-phosphorylated CaMKII.
(A) Schematic of CaMKII autophosphorylation in the PSD. Reactions (in bold) and PP1 concentrations are altered in the PSD from the basal model to achieve bistability. Block arrows indicate the CaMKII states that translocate. There are two shaded sets of molecules, indicating states that are summed to give rise to an enzyme activity. The Tot-CaM-CaMKII activity is the sum of concentrations of CaM-CaMKII and CaM–Thr286-phosphorylated CaMKII. The Aut-CaMKII activity is the sum of the calcium-autonomous states Thr286-phosphorylated and Thr286/Thr305-phosphorylated CaMKII.
(B) Bistability analysis. The Aut-CaMKII molecular species has been bifurcated into Aut-CaMKII enzyme, and Aut-CaMKII readout. The enzyme activity is buffered numerically (x-axis). The readout (y-axis) remains the sum of Thr286-phosphorylated and Thr286/Thr305-phosphorylated CaMKII. Fixed points are given by the intersection points of the amount of Aut-CaMKII with the 45° line. These fixed points indicate the number of Aut-CaMKII molecules that would exactly sustain their own activity through autophosphorylation. The upper and lower points are stable, and the middle point is the transition point.
(C) Time course of bistable response. The first arrow is a Ca2+ stimulus of 2.7 μM for 500 s that switches on the CaMKII bistable loop. The second arrow is a 5-fold increase in k
cat of PSD-localised PP1 for a period of 500 s, which switches CaMKII off.
(D) Parameter sensitivity analysis. Key parameters are scaled up and down and the model is tested for bistability. Most parameters can be varied 2-fold or more in either direction without the model losing bistability.
(E) Stochastic run showing stability in both high and low states, when PKA is buffered.
(F) Stochastic run showing spontaneous state flips in either direction, with the complete PKA model.
(G) Statistics of spontaneous state flips with the complete PKA model. Average turn on and turn off times are both over 15 h.
Overall, our CaMKII translocation model (model 3) also exhibited bistability coupled with translocation, such that the active state led to accumulation of CaMKII in the PSD. Under stochastic conditions, the lifetime of stable states in the model was sensitive to noise from the PKA input pathway.
Bistability Interactions: Tight Coupling
At this point we had a reasonably constrained model of AMPAR and CaMKII trafficking, and had shown that under certain conditions both molecules could be bistable. In the final part of the study we considered interactions between these two forms of bistability. We first made model 4 by merging model 1 and model 3, while sharing the same PP1 molecule in the PSD (Figure 8A). This scenario assumes that the PP1 is free to move between its CaMKII and AMPAR substrates while remaining in the PSD. Thus, there is a tight coupling between the two forms of bistability, mediated both by PP1 and by CaMKII phosphorylation of Ser831 of AMPAR, as in model 1. We asked how the system would respond to stimuli designed to activate the AMPAR and CaMKII switches independently.
Figure 8 Bistability for Tightly Coupled Switches
(A) Schematic of PSD-localised PP1 acting on both CaMKII and AMPAR substrates in the PSD. The asterisks on CaMKII and AMPAR represent phosphate groups.
(B) Time course of response to Ca2+ (2.7 μM, 500-s duration), then cAMP (0.108 μM, 2,000-s duration) stimuli. The initial Ca2+ stimulus turns on CaMKII transiently, but it eventually returns to baseline. The subsequent cAMP stimulus turns on both switches.
(C) Time course of response to cAMP (0.108 μM, 2,000-s duration), then Ca2+ (2.7 μM, 500-s duration) stimuli. The initial AMPAR stimulus (cAMP elevation) is sufficient to turn both the AMPAR and the CaMKII switches on.
(D) Stochastic run in the low state. The figure illustrates a transient event that did not result in complete turn on.
(E) Stochastic run in the high state. There is a spontaneous turn off, but the average on time is over 100 h.
In our first test we stimulated the model 4 with Ca2+ (2.7 μM, 500 s) then allowed the system to settle for 3 h, then stimulated it with cAMP (108 nM, 2,000 s) (Figure 8B). Following the Ca2+ stimulus, the CaMKII switch turned on transiently, but soon returned to baseline. The AMPAR switch did not turn on until the cAMP stimulus was applied, and at this time CaMKII also turned on. When the cAMP stimulus was applied first, it rapidly turned on both the AMPAR and CaMKII switches (Figure 8C). Together, these simulations show that in model 4 the two forms of bistability function in lockstep. That is, sustained activation of CaMKII is contingent upon the activity of AMPAR. If AMPAR is activated, it saturates PP1, and this leads to activation of the CaMKII bistable switch. As shown in Figure 8C, CaMKII is also activitated by the cAMP stimulus independently of the PP1-mediated cross-activation, leading to rapid turn on. This is because cAMP activates PKA, which relieves the PP1 inhibition of CaMKII. We ran separate simulations (not shown) that showed that even in the absence of this cAMP activation of CaMKII, the activation of AMPAR caused CaMKII to turn on as well.
We ran model 4 using stochastic methods to test its propensity to spontaneously change state. The off state was very stable though it did exhibit occasional transient spikes of activity (Figure 8D). The model spontaneously turned on only once in a cumulative total of over 1,000 h of simulation time. As before, we tested high-state durations by applying an initial cAMP stimulus at about 1 h to turn the system on, and then ran the simulation for about 33 h. We repeated these runs 24 times to obtain the distribution. The high state was not as stable, and was subject to occasional spontaneous flips to the off state with an average time of 101 ± 79 h (mean ± standard error of the mean). As expected from the lockstep mechanism, both CaMKII and AMPAR underwent a state flip at nearly the same time (Figure 8E).
Bistability Interactions: Weak Coupling
In our final model (model 5), we considered the situation where CaMKII and AMPAR interacted only weakly. This is in contrast to model 4, where CaMKII and AMPAR were tightly coupled through a shared pool of PP1. In model 5, we combined the CaMKII and AMPAR bistable models (models 1 and 3) while keeping an independent pool of PP1 for each (Figure 9A). Such a scenario might arise if PP1 were bound to distinct scaffold proteins for each of its targets, and were restricted in its mobility across targets. The CaMKII-coupled pool of PP1 was treated as independent of PKA activity. There was one indirect form of coupling still present from CaMKII to the AMPAR bistability, since CaMKII phosphorylates AMPAR on Ser831. While this does not alter traffic rates directly, the Ser831 is a substrate for the AMPAR-associated pool of PP1 in model 5. Thus, CaMKII activity did contribute to the saturation of the AMPAR-associated PP1.
Figure 9 Nested Bistability for Weakly Coupled Switches
(A) Schematic of independent PSD-localised PP1 enzyme activities for CaMKII and AMPAR. The two PP1 activities are labelled PP1-PSD-CaMKII and PP1-PSD-AMPAR, respectively. The asterisks represent phosphorylation.
(B) Time course of system response to Ca2+ (2.7 μM, 500-s duration), then cAMP (0.108 μM, 2,000-s duration) stimulus. The initial activation of CaMKII leads to a slow turn on of the AMPAR system.
(C) Time course of system response to cAMP (0.108 μM, 2,000-s duration), then Ca2+ (2.7 μM, 500-s duration) stimulus. First the AMPAR system turns on, then, following the Ca2+ stimulus, the CaMKII turns on. The conductance of the synapse has different levels in each of these states.
(D) Stochastic run for 60 h, showing resting, AMPAR only, and AMPAR + CaMKII activity states.
As before, we examined the interactions between the CaMKII and AMPAR switches using stimuli designed to turn each switch on independently of the other. When we first turned on the CaMKII switch alone using a Ca2+ stimulus, we observed a slow activation of the AMPAR switch (Figure 9B). The Ca2+ stimulus indirectly increased AMPAR insertion through the following steps: Ca2+ → CaM → CaMKII → phosphorylation of Ser831 → saturation of AMPAR-specific PP1 → reduced endocytosis of AMPAR. We did not model any changes in recycling or internalisation rates due to receptor phosphorylation on Ser831.
A particularly interesting effect was seen when the AMPAR switch was activated first using cAMP (Figure 9C). This stimulus turned on the AMPAR switch without affecting the state of CaMKII. At about 4 h we applied a Ca2+ stimulus that turned on the CaMKII switch as well. Thus, in model 5, the two switches were able to coexist in three combinations of states: both off, only AMPAR on, or both on. The fourth possible combination, of CaMKII on and AMPAR off, was not stable because of weak coupling between CaMKII and AMPAR, which slowly turned the latter on as shown in Figure 9B. The weak coupling is due to the phosphorylation of GluR1 on Ser831 by CaMKII. While this does not directly affect translocation, it does engage the AMPAR-specific pool of PP1, leading to eventual phosphatase saturation and turn on of the AMPAR switch.
Overall, this model exhibited nested bistability. The fundamental switch took place when AMPAR turned on or off. Nested within this was the capacity for CaMKII to turn on or off. A possible physiological outcome of the nesting of CaMKII activation is that the phosphorylation state and hence the conductance of the synapse can settle to three levels: (1) no AMPAR, (2) AMPAR with low levels of Ser831 phosphorylation, and (3) AMPAR with high Ser831 phosphorylation and consequently a higher conductance (Figure 9C). As shown in Figures 3 and 4, the equivalent conductance is expressed in terms of the number of unphosphorylated AMPAR channels that would have the same conductance.
The three states of the model were quite stable under stochastic conditions (Figure 9D). The only state that showed any transitions over the entire cumulative duration of simulations tested was that in which AMPAR was on and CaMKII was off (time to switch was 24.8 ± 3.5 h). We were particularly interested in the long-term stability of the states with both switches off, and both switches on. To examine these we performed several hundred independent stochastic simulation runs on a cluster, each representing 120,000 s (33 h) of simulated time. No state transitions were observed in either direction, over a cumulative duration of over a year of simulated time for each state.
Transient Responses of Models
How does the model respond to stimuli that induce changes in synaptic efficacy? We tested two synaptic plasticity protocols on each of the models 1, 3, 4, and 5 (Figures S1 and S2). These tests were only qualitative, as the models in the current study were parameterized using steady-state rather than transient experiments. Nevertheless, they are useful in showing model behaviour under transient conditions. The first protocol had been used to elicit long-term potentiation (LTP) of synaptic efficacy and consisted of three bursts of 100 impulses at 100 Hz, each separated by 600 s. The second protocol was used to induce long-term depression (LTD) at the synapse, and consisted of 900 impulses at 1 Hz. We represented each stimulus as a computed calcium transient with an exponential build-up and decay of Ca2+ using the formulation of Zhabotinsky [34] (Figure S1). The LTP stimulus gave calcium peaks of 12 μM, and the LTD stimulus had peaks of 0.5 μM. We found that the LTP stimulus was able to cause a switch to the on state only in the CaMKII model (model 3) that incorporated the PKA activation pathway (Figure S2). This is interesting, as it suggests that even in its current form the CaMKII model is reasonably sensitive to physiological stimuli that may play a role in synaptic plasticity. The LTD stimulus did not turn off any of the models, indicating that the current models are missing some key interactions. These tests highlight some of the unknowns in our models, in particular, the specific transient interactions that are needed to trigger the steady-state effects we have analysed. As discussed below, more experimental detail will be needed to extend the models to include transient response characteristics.
Summary of Bistable Behaviour
In summary, we analysed four bistable synaptic trafficking models (models 1, 3, 4, and 5). The remaining two models in this study (models 0 and 2) were used to characterise model 1. Model 3 included CaMKII alone, but models 1, 4, and 5 included both AMPAR and CaMKII. As described above, we performed a number of steady-state and transient tests to characterise the stable states of each model. Properties of these models are summarised in Table 1.
Table 1 Summary of Model Properties
Discussion
In this study we have developed a model of the movement of a glutamate receptor (AMPAR) and a calcium-activated kinase (CaMKII) to and from the synaptic membrane, using steady-state trafficking rates as a major experimental constraint. We find that the AMPAR trafficking cycle may lead to a form of switching or bistability where the presence of sufficient receptors at the synapse leads to recruitment of more receptors. This process may be a candidate for the transition from silent to active synapses, and early phases of their subsequent maintenance. When combined with a previously proposed mechanism for a CaMKII molecular switch, we observe interesting interactions between these two forms of synaptic bistability. Depending on the degree of coupling between these switches, we predict that AMPAR recruitment and CaMKII activation may either occur in a tightly coordinated manner, or nearly independent of each other. The latter may give rise to multiple stable synaptic states. Stochastic calculations suggest that these stable states persist for many hours, and in some cases over a year, despite biochemical fluctuations due to small numbers of molecules in the synapse.
Structural Bistability
Long-term storage of information at the synapse is intimately connected to structural changes [8,15]. Such changes arise from the insertion, removal, and reorganisation of synaptic molecules. For example, many types of glutamatergic synapses initially lack postsynaptic AMPARs and are unresponsive to moderate synaptic input. These “silent synapses” become active when AMPARs are inserted. The change from silent to active synapses is a major mechanism for increases in synaptic efficacy [8]. Similarly, formation of dendritic spines is facilitated by the presence of the GluR2 subunit of the AMPAR [29]. Several signalling events are known that may affect these structural processes, but it is not always clear how persistent changes may be maintained. A single brief pulse of insertion of molecules, or formation of a synaptic spine, will have a transient effect unless some self-sustaining mechanism is also available to keep the changes in place. Formally, bistable systems are a possible mechanism for such structural memory, as such systems can withstand molecular turnover [2]. The AMPAR model analysed in the current paper shows that synaptic membrane insertion (a form of structural change) can be self-sustaining even when molecular turnover and noise are considered.
There are several proposed forms of synaptic bistability, and each has some structural correlates. The classical form of synaptic bistability is the CaMKII autophosphorylation system (reviewed in [3]). This is a biochemical bistability that relies on autophosphorylation leading to self-activation of CaMKII. This self-activation leads to bistability if PP1 is present at sufficiently low levels that it can be saturated. The involvement of PP1 saturation is also an important aspect of our model for AMPAR bistability. CaMKII activation has several structural correlates, including translocation to the PSD and formation of complexes with NMDA and other PSD proteins [33]. There is also recent evidence that CaMKII activation leads to an increase in AMPAR numbers [37]. Another synaptic biochemical bistability involving a MAPK feedback loop has been proposed [4] and tested in a fibroblast model system [38]. MAPK is known to be important in synaptic plasticity and may also play a role in structural changes at the synapse [39,40]. However, the MAPK bistable feedback mechanism is vulnerable to biochemical noise and may not be plausible in synaptic volumes [7]. A more recent proposal for a self-sustaining synaptic plasticity mechanism involves the mTOR system and local protein synthesis. mTOR phosphorylation increases protein synthesis at the dendrites, and the synthesis machinery itself is one of the products. Several other proteins have also been identified that may participate in such a feedback loop [6]. Such a protein-synthesis-dependent bistability would be very interesting for structural change at the synapse, as it could account for increased availability of many synaptic and PSD proteins.
In the current study we propose a novel form of synaptic bistability involving self-recruitment of AMPARs to the synapse. The mechanism is particularly interesting for the synapse in three ways: (1) it has parallels with the conversion of silent to active synapses, (2) it works at basal levels of activity of synaptic enzymes, and (3) it intimately involves a translocation and synaptic membrane insertion process. As analysed in Figures 5 and 6, this form of bistability is a function of the number of molecules of receptor in the synapse, rather than biochemical activation. The bistability involves phosphatase saturation of PP1 due to its action on AMPAR in the PSD. This is a specific prediction of this study. However, the prediction of AMPAR-specific PP1 activity is yet to be tested in detail. We discuss the issue of PP1 access to other substrates below.
It should be stressed that this mechanism for synaptic state maintenance is by no means exclusive. We discuss several other possible mechanisms above. There are also important details about AMPAR cycling that invite further analysis. For example, the AMPARs in Ser831/Ser845 double phosphomutant mice would not sustain this form of bistability. Nevertheless, these animals form synapses and retain some degree of synaptic plasticity [12]. Furthermore, our model considers activity-dependent changes in GluR12, and does not account for a presumed hand-over of synaptic state to some long-term process involving GluR23 insertion. Mechanisms for maintenance of GluR23 levels are still poorly understood, but we speculate that this too involves a self-recruiting bistable process.
Phosphatase saturation is a key part of our self-recruitment model. This has parallels with a distinct form of bistability analysed for the MAPK system by Markevich et al. [41]. In this MAPK model, there are two stable states of MAPK activity. The high state is sustained in part because of the saturation of phosphatases that reverse its activity. The distinction, again, is that the AMPAR bistability involves translocation without sustained biochemical activation whereas the Markevich model involves biochemical activation without intrinsic structural effects.
In a broader context, the AMPAR structural bistability could be generalised as a state-dependent translocation of molecules coupled to a saturable interconversion between these states. Many cellular trafficking events have a similar general form, including the Rab-mediated system of small GTPases and nuclear transport control. By our analysis in Figures 5 and 6, translocation bistability should exhibit two clearly separated regimes in which traffic occurs into one of the compartments, and a regime in which addition of the translocated molecule actually decreases its number in one of the cellular compartments. This might be an experimentally accessible signature for such behaviour in the cell.
Stochasticity and Robustness
The typical synapse has a volume of 0.1 fl [16], and contains rather small numbers of key signalling molecules. This introduces fluctuations in reactions taking place in the synapse. We have previously analysed small-volume signalling for several pathways using simple assumptions about scaling and diffusion, and find that stochastic effects are so severe that some conventional signalling mechanisms simply do not work in these volumes [7,36]. Stable retention of synaptic state is a potential victim of stochasticity, as biochemical noise can lead to spontaneous state transitions. Bialek [17] and Miller et al. [42] have previously analysed synaptic bistability and suggest that in principle an autophosphorylation mechanism can give molecular stability of the order of hundreds of years even with a small number of CaMKII holoenzymes. Our current study is both coarser and more detailed than these analyses. Our representation of CaMKII does not consider individual holoenzymes, but on the other hand we explicitly represent translocation and several important signalling interactions at the synapse. In some models, and for some states, we did not observe any transitions over a year of simulation time. For other states the predicted lifetime is of the order of tens of hours (Table 1). Thus, in terms of state stability, our models are quite robust.
While it is heartening that our models are stable for many hours, it is clear that this may change in either direction if additional interactions are considered. For example, we ignore the diffusive exchange of most spine molecules with the PSD and dendrite. Over the time scales of our simulations, all the spine molecules would in fact exchange or degrade. This introduces further challenges to synaptic stability mechanisms that are beyond the scope of this study. A further illustration of the importance of context on stability comes from the CaMKII analysis in Figure 7. Here the inclusion of a more detailed PKA signalling model reduces the predicted lifetime of CaMKII states from hundreds of hours to tens. Additional input pathways may contribute to increased noise, but scaffold anchoring of enzymes such as PKA may create local signalling environments that could potentially reduce stochastic effects. Direct experimental measurements of synaptic stochasticity are difficult to perform [43], but such measurements would be invaluable for comparison with our simulated estimates of stochasticity.
Another commonly used measure of robustness is to ask whether the model retains some critical attribute over a wide range of parameters. In our analysis, we use bistability as the attribute of interest. We perform parameter sensitivity analyses by varying key reaction rates and enzyme parameters over a 100-fold range. Our parameter sensitivity analyses for the AMPAR and CaMKII systems suggest that our models are fairly robust by this measure. Most parameters can be varied at least 0.5- to 2-fold their original values without the model losing bistability (see Figures 5 and 7).
A particularly stringent measure of the robustness of a mechanism is to see how well essential features are preserved as reaction details are changed. We have performed a rather severe pruning of the AMPAR bistable model to a bare skeleton (see Figure 6) and with very little parameter tuning the simplified model is also bistable. Thus, by several measures, the bistable processes we have analysed are resistant to stochasticity, parameter uncertainty, and even changes in reaction mechanisms.
Coupled Bistable Switches
Individual synapses are likely to exist in many states [18]. Given the short life of synaptic molecules discussed above, it seems possible that one mechanism for stabilising such states might be to associate bistable switches with them. Multiple states may be achieved if the individual switches are coupled loosely, so that combinations of states become possible. Here we have shown (model 5; Figure 9) that distinct forms of bistability may coexist to give rise to three possible synaptic states. The AMPAR switch is the major one, as it brings the receptors to the synaptic membrane in the first place. The CaMKII switch is nested within this as it can fine-tune the conductance of the receptors. This situation of nested bistable states is possible if the mobility of PP1 at the synapse is limited so that each PP1 molecule has exclusive access either to CaMKII or to AMPAR. The alternate assumption (model 4) is that PP1 is mobile within the PSD, and can access both CaMKII and AMPAR. This assumption causes the two forms of bistability to function in lockstep, where the activation of the AMPAR switch causes the CaMKII switch to turn on. This occurs because the two switches share the same pool of PP1 enzymes, and phosphatase saturation resulting from the AMPAR switch activates CaMKII. It is likely that the biological situation involves different degrees of PP1 mobility between multiple possible synaptic targets, and may even differ for the same synapse in different contexts. In our study, we obtain distinct outcomes for two cases of PP1 mobility and targeting. This sensitivity of synaptic state to PP1 mobility is a testable prediction and highlights the possible importance of subtle details of PSD anchoring on synaptic function.
An alternative proposal for multiple levels of synaptic activation is that the CaMKII–NMDAR complex may act in a highly modular manner (reviewed by Lisman and McIntyre [3]). In this scenario, each CaMKII–NMDAR complex can independently persist in a high or low state of activity. This situation would make it possible for an individual synapse to present graded levels of conductance depending on the number of active CaMKII complexes. Our model of CaMKII is at the bulk rather than holoenzyme level, and is too coarse-grained to address this possibility. The CaMKII phosphorylation of AMPAR plays two roles in our study. First, it directly increases the conductance of individual receptor tetramers. Second, it indirectly leads to an increase in the number of receptors at the synapse, by producing additional phosphorylation states of AMPAR for PP1 to act on. In the model, this leads to further saturation of the phosphatase and ultimately to an increase in surface AMPAR. There is recent evidence that CaMKII activity may also affect AMPAR numbers [37]. This would provide another mechanism for coupling between our proposed bistable mechanisms.
In addition to forming multiple synaptic states, our simulations show that coexisting bistable mechanisms may function to “hand-over” information about synaptic state from one switch to another. For example, in model 5 (weakly coupled synaptic switches), a rather brief Ca2+ input is sufficient to activate CaMKII, which can then turn on the AMPAR switch over a timescale of hours (Figure 9B). This is loosely analogous to different forms of computer information storage, where information is initially stored in fast but volatile form (e.g., RAM) and is later transferred to slow but stable forms of memory (e.g., hard disk).
Our study illustrates how two mechanisms for synaptic bistability may coexist to give rise to multiple possible synaptic states. We propose that the synapse may exhibit a combinatorial set of states through the interactions of several molecular switches. These may include local protein synthesis feedback loops involving mTOR, self-assembly processes at the synapse, or presynaptic switches. From the cell-biological perspective, we have considered synaptic recruitment mechanisms for only two of the hundreds of postsynaptic molecules. All these molecules undergo turnover, and many experience regulated movement similar to that in the switches we have analysed. We suggest that there are many forms of self-recruitment, coordinated self-assembly, and other potentially switch-like processes that contribute to the maintenance of different constituents and states of the synapse.
Materials and Methods
Model development.
Our model was developed to closely tie with experimental observations and to build on existing, well-documented, and experimentally constrained models. Two molecular trafficking cycles form the core of the model: (1) the trafficking of AMPARs, and in particular GluR12, between internal vesicular pools and the synaptic membrane associated with the PSD and (2) the movement of CaMKII to and from the spine cytosol and the PSD (see Figure 1). As elaborated below, we developed the models using published experimental observations on these trafficking processes, and considerable specific data on the biochemistry of the phosphorylation of these molecules. A few regulatory pathways were also modelled to provide signalling input and context. Reactions in the model take place in two compartments: the PSD and the bulk cytosolic volume of the spine. The receptors are membrane-associated, so the PSD-associated synaptic membrane is included in the PSD compartment. Likewise, the internalised, vesicular pool of receptors is included in the cytosolic compartment. The PSD volume is taken as 0.01 fl, and the spine head volume as 0.09 fl. There is a third, dendritic compartment of 5 fl that is occupied only by diffusible cAMP and by a bulk AMPAR pool. The bulk AMPAR pool is assumed to be at a steady level and is meant to represent synthesis and degradation of the receptor. No reactions occur in the dendritic compartment as it is meant only to couple diffusively with the spine.
AMPAR model.
Due to a combinatorial proliferation of states, the reaction diagram of the AMPAR steps appears complex. As described below, we modelled 16 phosphorylation states of the receptor each in the internal and synaptic-membrane-associated pools. However, most of the reactions involving these 16 states were symmetric as they involved independent phosphorylation sites. This simplifies the model definition. We assumed that symmetric reactions had the same rates, so our model relies on only a few trafficking and phosphorylation parameters. AMPARs occur as tetrameric structures [44] with most AMPARs composed of two subunits each of GluR1 and GluR2 (GluR12) or GluR2 and GluR3 (GluR23). GluR23 receptors show constitutive trafficking and are responsible for basal synaptic transmission whereas GluR12 receptor insertion can be altered by stimuli [10,20]. We considered only GluR12 receptors in the model to focus on activity-stimulated events. The dynamics of GluR12 AMPAR trafficking were determined by kinase/phosphatase activities at the Ser845 sites of the two GluR1 subunits in the tetrameric GluR12 complex. Dephosphorylation of these sites by phosphatases triggers endocytosis whereas phosphorylation by PKA is required for synaptic membrane targeting [10,11]. We modelled the phosphorylation/dephosphorylation as a two-step reaction, where phosphorylation or dephosphorylation of both GluR1 subunits is necessary for synaptic membrane targeting or internalisation, respectively. Through simulations we found that basal activities of PP1 and PKA can account for the constitutive cycling of receptors in our model, consistent with experimental studies [8,45]. We assumed that PKA and PP1 were the relevant enzymes, but the model does not exclude the possibility that the same cycling effects might be mediated by other phosphorylation enzymes.
There is evidence that the membrane-associated PP1 dephosphorylates AMPARs only in the PSD, as loading neurons with active PP1 does not alter basal synaptic strength transmission [46]. Hence, we assumed that PP1 acts on GluR1 only in the PSD whereas PKA phosphorylates GluR1 in both compartments. In both compartments, Ser845 of GluR1 was also dephosphorylated by PP2B [10,47,48], which itself is inactive at basal Ca2+ concentrations.
Phosphorylation of Ser831 of the GluR1 subunit by CaMKII alters channel properties of the receptor in that the phosphorylation increases channel conductance approximately 2-fold [49]. As for the Ser845 sites, we modelled Ser831 phosphorylation of GluR1 so that both sites of a tetrameric complex could be phosphorylated individually by CaMKII. Dephosphorylation of Ser831 was modelled to occur only in the PSD, as internalisation was reported not to alter the phosphorylation state of AMPARs at Ser831 [10].
In the bistable models we explicitly modelled protein turnover through activity-dependent degradation [1]. The activity dependence was introduced by restricting the turnover to the doubly Ser845-phosphorylated states in the internal pool of AMPARs (see Figure 5A). There are around 150 receptors in an active dendritic spine [50]. We represented this constraint in the model as an anchor protein (possibly GRIP [8]) required for AMPAR insertion into the synaptic membrane.
The synaptic AMPAR conductance is a function both of the number of synaptic membrane receptors, and of their phosphorylation state. We assumed that if a single GluR1 subunit was phosphorylated on the CaMKII site (Ser831), the channel conductance was 1.5 times the basal level, and if two GluR1 subunits were phosphorylated the channel conductance doubled. In the figures, conductances are expressed as percent maximal conductance. We obtained the maximal conductance by considering that all the anchor protein was occupied, and that all the AMPARs were doubly phosphorylated and hence had double the basal conductance.
CaMKII model.
The CaMKII model was derived closely from a previously developed single-compartment model of CaMKII activity [4,51]. This model is duplicated for the cytosol and the PSD and trafficking steps included. There is evidence that PP2A dephosphorylates CaMKII in the cytosol and PP1 in the PSD [52–54]. Because of limited data about the PP2A activity, we represented the cytosolic dephosphorylation step as involving a distinct phosphatase from the PP1 in the PSD, but using the same kinetics as PP1.
Binding of Ca2+/CaM is necessary and sufficient for the kinase to translocate to the PSD [55], where it binds to the NMDAR [56,57]. As we lacked direct association constants between CaMKII and NMDAR, we used time course information to constrain translocation of CaMKII to the PSD [14]. NMDAR was modelled as a putative binding site within the PSD [58]. Robust translocation away from the PSD occurs upon removal of the Ca2+ stimulus, and phosphorylation of Thr305 is required in this process [14]. However, only simultaneous dephosphorylation at Thr286 is sufficient for effective dissociation of CaMKII from the PSD [14,59].
Other pathways.
There are numerous regulatory inputs, which are taken from a pre-existing library of signalling pathway models ([19]; the DOQCS database [http://doqcs.ncbs.res.in]). The parameters of these models are substantially the same, with the exception of PP2B (calcineurin), the cAMP pathway, and some scaling of phosphatase activities.
In the case of PP2B we did not vary any rates, but we eliminated the catalytic activity of two substates (Ca2.CaM.Ca4.CaN and Ca3.CaM.Ca4.CaN) as their contribution to the total was small (data not shown), and since the inclusion of these additional phosphorylation steps for all states of AMPAR would have substantially increased the number of reactions in the model.
In the case of cAMP we increased cyclase concentrations by a factor of approximately 4-fold, to get integral numbers of molecules in the model and to compensate for the reduction in assumed ATP concentrations from earlier model values of 5 mM to 2 mM. Phosphodiesterase concentrations have also been scaled up to maintain effective cAMP concentrations. A diffusion step is modelled for cAMP exchange with the dendrite, using a diffusion constant from frog olfactory neurons [60].
The phosphatase rates were scaled to obtain correct steady-state phosphorylation levels of inhibitor 1 of PP1 and CaMKII. PP1 rates and concentrations were also scaled, as described in the Results section, for the model of CaMKII bistability.
Most molecules in the simulation were modelled as independent pools for the PSD and cytosol. Only PKA was assumed to have access to both the cytosolic and PSD volumes. The adenylyl cyclase pathway was modelled only in the cytosolic volume. Because of its rapid diffusion, cAMP was modelled as exchanging between the spine head cytosol and the dendrite. We make an implicit assumption that the concentration of spine head constituents is maintained over the long periods of our simulations, through unspecified trafficking or other processes.
Computations.
Simulations were performed on Linux workstations and on a Linux cluster (Atipa Technologies, Lawrence, Kansas, United States) for stochastic calculations. Models were implemented using Kinetikit/GENESIS [61], and solved using the Exponential Euler method [61]. Enzyme reactions were modelled with an explicit enzyme–substrate complex, with the exception of the adenylyl cyclase activity (see Figure 1G), which used the Michaelis-Menten form to improve numerical stability.
Stochastic calculations were done using an adaptive stochastic method [62] and using the Gillespie exact stochastic method as implemented in GENESIS 3/MOOSE [27]. The exact stochastic calculations used the Mersenne Twister random number generator [63]. When using the exact stochastic method, the entire model was simulated with the Gillespie method. Thus, all reactions led to integral changes in the numbers of the variable molecules. A few molecules in the model are buffered. The numbers of these buffered molecules were folded into the corresponding rate terms for efficiency. For example, if we have the reaction A <==> B and A is buffered, then the propensity of formation of B is dn
B/dt = kf.n
A − kb.n
B, where n
A and n
B are the numbers of molecules of A and B respectively. Since A is buffered, the value of n
A is fixed and we replaced this equation with dn
B/dt = kf′ − kb.n
B, where kf′ = kf.n
A. This substitution also meant that it was possible to use nonintegral numbers for buffered molecules. This was meant to represent situations where the chemical buffering system on average gave rise to a nonintegral number of molecules.
Stochastic transition time calculations.
Several lengthy stochastic runs were performed to estimate transition times between states of the models. In order to obtain longer samples, we set off many independent simulations in parallel on a cluster using distinct random number seeds, typically for a simulation time of 120,000 s (approximately 33 h) each. Transition times were estimated for a set of independent simulations as follows. Let T be the time to first transition in a given run, or total time of the run if there were no transitions. Let N be the number of runs where there were transitions. Then transition time for the entire sample is ΣT/N. In some cases there were no transitions at all, even for a large sample of runs. In these cases N was zero, so we could only set a lower bound to the transition time to be of the order of ΣT.
In some cases (e.g., Figure 7G) we estimated individual transition times by summing T for successive runs until the first run that had a transition. This sum gave the estimated transition time. Then the sum was reset to zero and the process repeated for the next transition.
Estimation of thresholds (unstable fixed points).
Thresholds for transition between lower and upper states of AMPAR in the spine were estimated using an iterative bisection method. The range of possible values was known from the upper and lower steady states of the bistable model (model 1). These were set as the upper and lower limits U and L, respectively. The first estimate E of the threshold was halfway between U and L: E = (U + L)/2. The model was equilibrated at E receptors by blocking the AMPAR exchange with the bulk and AMPAR degradation. Then the exchange and degradation reactions were unblocked, and the model was run out for 10,000 s. Depending on whether E was above or below the actual threshold point, the model settled toward U or L, respectively. If E was high, then U was reassigned to E. If E was low, then L was reassigned to E. This process was repeated seven times to obtain an approximately 1% accurate estimate of the threshold.
Model and simulator availability.
Complete model parameters and reaction schemes are presented in Protocol S1. All models, demonstration simulations, and the GENESIS/Kinetikit simulator are freely available at http://www.ncbs.res.in/~bhalla/AMPAR_switch/index.html. Models 0 to 5 (including the buffered PKA version of model 3) are deposited in the DOQCS database (http://doqcs.ncbs.res.in) as accession numbers 59 to 65.
Supporting Information
Figure S1 Transient Stimuli and Responses of Model 0 with the Total Number of AMPARs Set to 80
(51 KB PDF)
Click here for additional data file.
Figure S2 Responses of Different Bistable Models to Transient Inputs
(60 KB PDF)
Click here for additional data file.
Protocol S1 Model Equations and Parameters
(240 KB PDF)
Click here for additional data file.
We acknowledge many helpful discussions with Mike Ehlers and Dennis Bray. AH received support from the Conseil Regional Alsace, USB from the Wellcome Trust and NCBS/TIFR. The development of GENESIS 3/MOOSE was supported in part by Biophase Systems.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. AH and USB conceived and designed the experiments, performed the experiments, and analysed the data. USB contributed reagents/materials/analysis tools. AH and USB wrote the paper.
Abbreviations
AMPAalpha-amino-3-hydroxy-5-methyl-4-isoxazole propionate
AMPARalpha-amino-3-hydroxy-5-methyl-4-isoxazole propionate receptor
Aut-CaMKIIautonomously active CaMKII–PSD
CaMcalmodulin
CaMKIIcalcium calmodulin type II kinase
cAMPcyclic adenosine monophosphate
GluR12glutamate receptor heteromer of subtypes 1 and 2
GluR23glutamate receptor heteromer of subtypes 2 and 3
LTDlong-term depression
LTPlong-term potentiation
MAPKmitogen-activated protein kinase
mTORmammalian target of rapamycin
NMDARN-methyl-D-aspartate receptor
PKAprotein kinase A
PP1protein phosphatase 1
PP2Bprotein phosphatase 2B
PSDpostsynaptic density
==== Refs
References
Ehlers MD 2003 Activity level controls postsynaptic composition and signaling via the ubiquitin-proteasome system Nat Neurosci 6 231 242 12577062
Roberson ED Sweatt JD 1999 A biochemical blueprint for long-term memory Learn Mem 6 381 388 10509708
Lisman JE McIntyre CC 2001 Synaptic plasticity: A molecular memory switch Curr Biol 11 R788 R791 11591339
Bhalla US Iyengar R 1999 Emergent properties of networks of biological signaling pathways Science 283 381 387 9888852
Kuroda S Schweighofer N Kawato M 2001 Exploration of signal transduction pathways in cerebellar long-term depression by kinetic simulation J Neurosci 21 5693 5702 11466441
Blitzer RD Iyengar R Landau EM 2005 Postsynaptic signaling networks: Cellular cogwheels underlying long-term plasticity Biol Psychiatry 57 113 119 15652868
Bhalla US 2004 Signaling in small subcellular volumes. II. Stochastic and diffusion effects on synaptic network properties Biophys J 87 745 753 15298883
Malinow R Malenka RC 2002 AMPA receptor trafficking and synaptic plasticity Annu Rev Neurosci 25 103 126 12052905
Shi S Hayashi Y Esteban JA Malinow R 2001 Subunit-specific rules governing AMPA receptor trafficking to synapses in hippocampal pyramidal neurons Cell 105 331 343 11348590
Ehlers MD 2000 Reinsertion or degradation of AMPA receptors determined by activity-dependent endocytic sorting Neuron 28 511 525 11144360
Esteban JA Shi SH Wilson C Nuriya M Huganir RL 2003 PKA phosphorylation of AMPA receptor subunits controls synaptic trafficking underlying plasticity Nat Neurosci 6 136 143 12536214
Lee HK Takamiya K Han JS Man H Kim CH 2003 Phosphorylation of the AMPA receptor GluR1 subunit is required for synaptic plasticity and retention of spatial memory Cell 112 631 643 12628184
Wikstrom MA Matthews P Roberts D Collingridge GL Bortolotto ZA 2003 Parallel kinase cascades are involved in the induction of LTP at hippocampal CA1 synapses Neuropharmacology 45 828 836 14529720
Shen K Teruel MN Connor JH Shenolikar S Meyer T 2000 Molecular memory by reversible translocation of calcium/calmodulin-dependent protein kinase II Nat Neurosci 3 881 886 10966618
Meyer T Shen K 2000 In and out of the postsynaptic region: Signalling proteins on the move Trends Cell Biol 10 238 244 10802539
Harris KM Kater SB 1994 Dendritic spines: Cellular specializations imparting both stability and flexibility to synaptic function Annu Rev Neurosci 17 341 371 8210179
Bialek W 2001 Stability and noise in biochemical switches Adv Neural Inf Process Sys 13 103 109
Montgomery JM Madison DV 2004 Discrete synaptic states define a major mechanism of synapse plasticity Trends Neurosci 27 744 750 15541515
Sivakumaran S Hariharaputran S Mishra J Bhalla US 2003 The database of quantitative cellular signaling: Repository and analysis tools for chemical kinetic models of signaling networks Bioinformatics 19 408 415 12584128
Passafaro M Piech V Sheng M 2001 Subunit-specific temporal and spatial patterns of AMPA receptor exocytosis in hippocampal neurons Nat Neurosci 4 917 926 11528423
Lin JW Ju W Foster K Lee SH Ahmadian G 2000 Distinct molecular mechanisms and divergent endocytotic pathways of AMPA receptor internalization Nat Neurosci 3 1282 1290 11100149
Bienenstock EL Cooper LN Munro PW 1982 Theory for the development of neuron selectivity: Orientation specificity and binocular interaction in visual cortex J Neurosci 2 32 48 7054394
Dudek SM Bear MF 1993 Bidirectional long-term modification of synaptic effectiveness in the adult and immature hippocampus J Neurosci 13 2910 2918 8331379
Yang SN Tang YG Zucker RS 1999 Selective induction of LTP and LTD by postsynaptic [Ca2+]i elevation J Neurophysiol 81 781 787 10036277
Bliss TVP Collingridge GL 1993 A synaptic model of memory: Long-term potentiation in the hippocampus Nature 361 31 39 8421494
Genoux D Haditsch U Knobloch M Michalon A Storm D 2002 Protein phosphatase 1 is a molecular constraint on learning and memory Nature 418 970 975 12198546
Gillespie DT 1977 Exact stochastic simulation of coupled chemical reactions J Phys Chem 81 2340 2361
Grutzendler J Kasthuri N Gan WB 2002 Long-term dendritic spine stability in the adult cortex Nature 420 812 816 12490949
Passafaro M Nakagawa T Sala C Sheng M 2003 Induction of dendritic spines by an extracellular domain of AMPA receptor subunit GluR2 Nature 424 677 681 12904794
Nayak A Zastrow DJ Lickteig R Zahniser NR Browning MD 1998 Maintenance of late-phase LTP is accompanied by PKA-dependent increase in AMPA receptor synthesis Nature 394 680 683 9716131
Ju W Morishita W Tsui J Gaietta G Deerinck TJ 2004 Activity-dependent regulation of dendritic synthesis and trafficking of AMPA receptors Nat Neurosci 7 244 253 14770185
Smith WB Starck SR Roberts RW Schuman EM 2005 Dopaminergic stimulation of local protein synthesis enhances surface expression of GluR1 and synaptic transmission in hippocampal neurons Neuron 45 765 779 15748851
Bayer KU De Koninck P Leonard AS Hell JW Schulman H 2001 Interaction with the NMDA receptor locks CaMKII in an active conformation Nature 411 801 805 11459059
Zhabotinsky AM 2000 Bistability in the Ca(2+)/calmodulin-dependent protein kinase-phosphatase system Biophys J 79 2211 2221 11053103
Holmes WR 2000 Models of calmodulin trapping and CaM kinase II activation in a dendritic spine J Comput Neurosci 8 65 85 10798500
Bhalla US 2004 Signaling in small subcellular volumes. I. Stochastic and diffusion effects on individual pathways Biophys J 87 733 744 15298882
Tomita S Stein V Stocker TJ Nicoll RA Bredt DS 2005 Bidirectional synaptic plasticity regulated by phosphorylation of stargazin-like TARPs Neuron 45 269 277 15664178
Bhalla US Ram PT Iyengar R 2002 MAP Kinase phosphatase as a locus of flexibility in a mitogen-activated protein kinase signaling network Science 297 1018 1023 12169734
Wu GY Deisseroth K Tsien RW 2001 Spaced stimuli stabilize MAPK pathway activation and its effects on dendritic morphology Nat Neurosci 4 151 158 11175875
Ajay SM Bhalla US 2004 A role for ERKII in synaptic pattern selectivity on the time-scale of minutes Eur J Neurosci 20 2671 2680 15548210
Markevich NI Hoek JB Kholodenko BN 2004 Signaling switches and bistability arising from multisite phosphorylation in protein kinase cascades J Cell Biol 164 353 359 14744999
Miller P Zhabotinsky AM Lisman JE Wang XJ 2005 The stability of a stochastic CaMKII switch: Dependence on the number of enzyme molecules and protein turnover PLoS Biol 3 e107. DOI: 10.1371/journal.pbio.0030107 15819604
Groc L Heine M Cognet L Brickley K Stephenson FA 2004 Differential activity-dependent regulation of the lateral mobilities of AMPA and NMDA receptors Nat Neurosci 7 695 696 15208630
Rosenmund C Stern-Bach Y Stevens CF 1998 The tetrameric structure of a glutamate receptor channel Science 280 1596 1599 9616121
Rosenmund C Carr DW Bergeson SE Nilaver G Scott JD 1994 Anchoring of protein kinase A is required for modulation of AMPA/kainate receptors on hippocampal neurons Nature 368 853 856 8159245
Morishita W Connor JH Xia H Quinlan EM Shenolikar S 2001 Regulation of synaptic strength by protein phosphatase 1 Neuron 32 1133 1148 11754843
Zeng H Chattarji S Barbarosie M Rondi-Reig L Philpot BD 2001 Forebrain-specific calcineurin knockout selectively impairs bidirectional synaptic plasticity and working/episodic-like memory Cell 107 617 629 11733061
Beattie EC Carroll RC Yu X Morishita W Yasuda H 2000 Regulation of AMPA receptor endocytosis by a signaling mechanism shared with LTD Nat Neurosci 3 1291 1300 11100150
Derkach V Barria A Soderling TR 1999 Ca2+/calmodulin-kinase II enhances channel conductance of alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionate type glutamate receptors Proc Natl Acad Sci U S A 96 3269 3274 10077673
Matsuzaki M Ellis-Davies GC Nemoto T Miyashita Y Iino M 2001 Dendritic spine geometry is critical for AMPA receptor expression in hippocampal CA1 pyramidal neurons Nat Neurosci 4 1086 1092 11687814
Bhalla US 2002 Mechanisms for temporal tuning and filtering by postsynaptic signaling pathways Biophys J 83 740 752 12124261
Colbran RJ Bass MA McNeill RB Bollen M Zhao S 1997 Association of brain protein phosphatase 1 with cytoskeletal targeting/regulatory subunits J Neurochem 69 920 929 9282913
Sim AT Ratcliffe E Mumby MC Villa-Moruzzi E Rostas JA 1994 Differential activities of protein phosphatase types 1 and 2A in cytosolic and particulate fractions from rat forebrain J Neurochem 62 1552 1559 8133283
Strack S Barban MA Wadzinski BE Colbran RJ 1997 Differential inactivation of postsynaptic density-associated and soluble Ca2+/Calmodulin-dependent protein kinase II by protein phosphatases 1 and 2A J Neurochem 68 2119 2128 9109540
Shen K Meyer T 1999 Dynamic control of CaMKII translocation and localization in hippocampal neurons by NMDA receptor stimulation Science 284 162 166 10102820
Gardoni F Caputi A Cimino M Pastorino L Cattabeni F 1998 Calcium/calmodulin-dependent protein kinase II is associated with NR2A/B subunits of NMDA receptor in postsynaptic densities J Neurochem 71 1733 1741 9751209
Gardoni F Schrama LH van Dalen JJ Gispen WH Cattabeni F 1999 AlphaCaMKII binding to the C-terminal tail of NMDA receptor subunit NR2A and its modulation by autophosphorylation FEBS Lett 456 394 398 10462051
Suzuki T Okumura-Noji K Tanaka R Tada T 1994 Rapid translocation of cytosolic Ca2+/calmodulin-dependent protein kinase II into postsynaptic density after decapitation J Neurochem 63 1529 1537 7931307
Elgersma Y Fedorov NB Ikonen S Choi ES Elgersma M 2002 Inhibitory autophosphorylation of CaMKII controls PSD association, plasticity, and learning Neuron 36 493 505 12408851
Chen C Nakamura T Koutalos Y 1999 Cyclic AMP diffusion coefficient in frog olfactory cilia Biophys J 76 2861 2867 10233102
Bhalla US 2002 Use of Kinetikit and GENESIS for modeling signaling pathways Hildebrandt JD Iyengar R Methods in enzymology, Volume 345 San Diego Academic Press 3 23
Vasudeva K Bhalla US 2004 Adaptive stochastic-deterministic chemical kinetic simulations Bioinformatics 20 78 84 14693812
Matsumoto M Nishimura T 1998 Mersenne Twister: A 623-dimensionally equidistributed uniform pseudorandom number generator ACM Trans Model Comput Simul 8 3 30
|
16110334
|
PMC1185646
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Jul 29; 1(2):e20
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010020
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 10.1371/journal.pcbi.001002105-PLCB-RA-0057R2plcb-01-02-05Research ArticleBioinformatics - Computational BiologyCell BiologySystems BiologyEukaryotesThe RNA Silencing Pathway: The Bits and Pieces That Matter The RNA Silencing PathwayGroenenboom Marian A. C *Marée Athanasius F. M Hogeweg Paulien Theoretical Biology and Bioinformatics, University of Utrecht, Utrecht, The NetherlandsLevin Simon EditorPrinceton University, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 29 7 2005 1 2 e2118 3 2005 29 6 2005 Copyright: © 2005 Groenenboom 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.Cellular pathways are generally proposed on the basis of available experimental knowledge. The proposed pathways, however, may be inadequate to describe the phenomena they are supposed to explain. For instance, by means of concise mathematical models we are able to reveal shortcomings in the current description of the pathway of RNA silencing. The silencing pathway operates by cleaving siRNAs from dsRNA. siRNAs can associate with RISC, leading to the degradation of the target mRNA. We propose and analyze a few small extensions to the pathway: a siRNA degrading RNase, primed amplification of aberrant RNA pieces, and cooperation between aberrant RNA to trigger amplification. These extensions allow for a consistent explanation for various types of silencing phenomena, such as virus induced silencing, transgene and transposon induced silencing, and avoidance of self-reactivity, as well as for differences found between species groups.
Synopsis
Current descriptions of cellular and molecular pathways, proposed on the basis of available experimental knowledge, are often inadequate in describing the phenomena they are supposed to explain. The authors use mathematical models to reveal shortcomings in the current description of the pathway of RNA silencing. Understanding the mechanism of RNA silencing is of utmost importance, since it is rapidly evolving into a powerful tool in biology and medicine. The authors prove that the generally accepted pathway cannot explain sustained silencing, the prevention of the destruction of a cell's own RNA, and the process of transgene-induced silencing. They propose and analyze specific extensions to the pathway, which do allow for a consistent explanation for these various types of silencing phenomena. These extensions differ in the dynamics they predict, which now can be used to experimentally discriminate between them. The authors' results demonstrate that it is indispensable to check with mathematical models the feasibility of plausible models of cellular and molecular pathways, a step that is usually left out.
Citation:Groenenboom MAC, Marée AFM, Hogeweg P (2005) The RNA silencing pathway: The bits and pieces that matter. PLoS Comp Biol 1(2): e21
==== Body
Introduction
RNA silencing protects the eukaryotic cell against viruses and transposons. Viruses produce double-stranded RNA (dsRNA) during reproduction, which can trigger the silencing of viral RNA [1,2]. RNA silencing can also be triggered by a sufficiently high expression of transgenes, a mechanism known as co-suppression or transgene induced silencing [3–6]. The activation of transgene induced RNA silencing is directly linked to the activity of RNA directed RNA polymerase (RDR): overexpression of RDR significantly reduces the number of transgenes needed to induce RNA silencing [7]. RNA silencing deficient mutants show enhanced expression of transposons [8,9]. Transposons could trigger RNA silencing for two possible reasons: they often have multiple inverted repeats (IRs) that form dsRNA transcripts [10], and their high copy number could trigger silencing.
The currently proposed pathway of RNA silencing is shown in Figure 1. Generally, the process is initiated by the cleavage of dsRNA by Dicer. Dicer, an RNase III-class enzyme, processes dsRNA into small interfering RNAs (siRNAs) 21–25 nucleotides long. siRNAs can then be incorporated into the RNA induced silencing complex (RISC) and “guide” the complex via antisense base-pairing. This results in cleavage of the target mRNA near the center of the siRNA. We refer to the aberrant pieces of RNA after cleavage as “garbage RNA”. Sijen et al. [11] found that a substantial fraction of the siRNAs in Caenorhabditis elegans is not derived directly from the introduced dsRNA. To explain this, two amplification routes have been proposed: primed and unprimed amplification [12–14]. In both cases, RDR synthesizes dsRNA: in the case of primed amplification siRNA binds to mRNA to initiate dsRNA synthesis, whereas in the case of unprimed amplification the mere presence of aberrant garbage RNA is sufficient to trigger RDR. In short, the generally accepted pathway of RNA silencing consists of the degradation of mRNA via RISC and an amplification pathway through RDR.
Figure 1 The Standard Pathway of RNA Silencing
The figure is based upon Figure 1 in Hutvágner et al. [48].
Although it sounds reasonable that such a pathway would suffice to mount responses against both viruses and transposons, we show that the proposed pathway has severe limitations. We will show that it cannot correctly describe observations on transient and sustained silencing and dose dependency. Moreover, such a pathway would be extremely vulnerable for mounting responses against self. Finally, we will show that it cannot describe transgene induced silencing at all. We will then propose three different additions to the mechanism: (i) a siRNA degrading RNase; (ii) primed amplification of garbage RNA; and (iii) activation of RDR dependent on the number of garbage RNAs. The proposed models each give a consistent explanation for various types of silencing phenomena, that is, virus induced silencing, transgene and transposon induced silencing, protection against self-reactivity, as well as for differences found between species groups. The extensions, however, do differ in the dynamics they predict, which could be used to experimentally discriminate between them.
Results
Biological Background and Description of Core Model
We study the RNA silencing pathway using concise differential equation models with mass action kinetics. There is strong evidence that there is a common core pathway of RNA silencing present in all organisms capable of RNA silencing. We focus on the experimentally derived common core of RNA silencing (Figure 1 and Introduction), which is the basis for our model. We directly translate this pathway into a system of four coupled ordinary differential equations,
in which M, D, S, and G describe the number of mRNA, dsRNA, siRNA, and aberrant garbage pieces, respectively. mRNA is transcribed with a rate i and degraded with a rate dm. dsRNA is synthesized from mRNA by RDR with a small rate p, and is cleaved into n siRNAs with rate a. siRNA can associate with mRNA via RISC with rate b. For simplicity, we do not implement the formation of RISC explicitly in our model; instead, the siRNA–mRNA complex is directly degraded into aberrant garbage RNA. ds and dg describe the degradation of siRNAs and aberrant garbage pieces, respectively.
dsRNA can also enter the pathway in ways other than through RDR: a virus can produce dsRNA, dsRNA can be introduced or injected, or a transcript with IRs can form dsRNA. We simulate the introduction of dsRNA by a stepwise intracellular increase of the amount of dsRNA.
To allow for the formation of secondary siRNAs, we extend the model with the two amplification pathways:
The underlined term g1G describes the unprimed amplification—the synthesis of dsRNA from aberrant garbage RNA by RDR; and the bold term g2SM describes primed amplification—the synthesis of dsRNA primed by the presence of a siRNA on mRNA. We consider the pathway with and without the amplification terms.
Dynamics of Core Model
The behavior of the pathway as modeled above is shown in Figure 2. The upper panels show the effect of introducing dsRNA, homologous to an endogenous gene. We first study the model without amplification, which should be representative for mammals, in which RDR has not been found [15]. In mammals, dsRNA or siRNAs have to be continuously supplied to keep a gene silenced. In accordance, the model without amplification allows only for transient responses: siRNAs derived from the dsRNA cause a strong decrease in the amount of mRNA, after which the default equilibrium is re-established (Figure 2A). Since the system has only one attractor, the cell will always return to this attractor, which is the state with normal levels of mRNA. Only when dsRNA is continuously supplied, the gene stays silenced.
Figure 2 Dynamics of the Standard Models
(A) and (B) show that after dsRNA introduction, only transient responses are possible for the standard model without or with low amplification, whereas (C) shows that with high amplification, an arbitrary amount of dsRNA causes sustained silencing. Grey lines indicate dsRNA levels, black lines mRNA levels. (D) and (E) show that an increase in copy number leads to a proportional increase in mRNA levels for the model without or with low amplification, whereas (F) shows that mRNA levels have become independent of copy number in the model with high amplification. RNA levels are expressed in number of molecules per cell. Parameter values can be found in Table 1.
Table 1 Parameter Values Used in the Models (mol−1 Means per Molecule)
The default values used for the basic model are also used for the extensions, except when indicated otherwise.
amRNA half-lives vary greatly between different species. Yeast mRNA half-lives vary from minutes to 1.5 h [49]. In humans, the median half-life is 10 h [50], and there are mRNAs that are stable for more than 24 h. Plant mRNA half-lives vary from less than 1 h to several days, with an average of several hours [51].
bWe here take siRNA half-life to be 21 min, as is measured in human cells [25].
Amplification of the response via RDR is observed in nematodes, plants, slime molds, and fungi. The dynamics of the core model with primed or unprimed amplification are very similar; we therefore show only the results obtained for primed amplification. At a low amplification rate, the dynamics do not differ from the model without amplification (Figure 2B), but at a high amplification rate the default equilibrium becomes unstable, resulting in perpetual silencing (Figure 2C). Although the cell will remain in the default state as long as siRNAs and dsRNA are completely absent, a single dsRNA strand or siRNA suffices to trigger silencing.
A model of RNA silencing should also be able to explain transgene induced silencing. We therefore analyzed the effect of increasing the number of gene copies. We here assume that each gene copy has the same transcription rate, given by parameter i. In the model without or with a low rate of amplification, an increasing copy number leads to a proportional increase in mRNA levels (Figure 2D and E). In contrast, when the amplification rate is high, the amount of mRNA does not depend on the number of gene copies (Figure 2F). In this regime, the cell is always in the silenced state, and therefore the amount of mRNA per cell cannot increase. Thus, transgene induced silencing is not possible in the core model, whether or not amplification is taken into account.
Deficiencies of Core Model
The core model without amplification is capable of explaining only transient responses. In contrast, in plants RNA silencing can be sustained even after removal of the trigger [16,17], and in C. elegans silencing can persist for even more than one generation [18]. Intuitively it seemed plausible that amplification of the response could solve this problem. The core pathway with amplification, however, results in all-or-none type of behavior: either sustained silencing is impossible, or a single dsRNA strand or siRNA is sufficient to trigger perpetual silencing. This actually means that the dynamics of the core pathway with amplification imply inevitable destruction of self.
This problem of self-destruction has also been observed by Bergstrom et al. [19]. In their model study, they added unidirectional amplification, to obtain a transient silencing response. Amplification in plants, however, can be bidirectional [20], so unidirectionality cannot be the sole mechanism that prevents responses to self. Moreover, although unidirectional amplification can prevent sustained responses, it will not prevent transient responses directed against self, implying the unrealistic scenario of an infinite series of auto-destructive responses.
Another major deficiency of the core pathway is that it cannot describe or explain transgene induced silencing. Mathematical analysis of the equations shows that the incapability of transgene induced silencing and the all-or-none type of behavior are inherent properties of the core pathway (see Materials and Methods): the qualitative dynamics do not change when some or all of the mass action terms in the models are replaced by Michaelis-Menten kinetics.
We conclude, that to alleviate the limitations discussed above, the core model should be qualitatively altered. A qualitative difference could be either a missing step in the pathway, or some cooperative effect between RNAs. On the other hand, taking, for example, more details of the RISC complex formation into account, would not make the model qualitatively different, and, therefore, the model would still suffer from the same limitations. That is, this model study shows that the core pathway, which is generally presented as being the basic mechanism, with extensions of the pathway simply being (subtle) modifications of it, is essentially incomplete, and can therefore not be considered to be the core of the pathway.
Biological Background and Description of Extended Models
We aim to find extensions to the core pathway that are able to provide insight in the type of interactions needed to explain the complexity of RNA silencing. These extended pathways should be able to describe dose dependent responses; the possibility of both transient and sustained responses; transgene or transposon induced silencing; and avoidance of self-reactivity. All extended models need to include at least one of the amplification pathways in order to account for secondary siRNAs and to allow for sustained silencing.
In the first extension, we propose that in addition to the non-specific siRNA degradation a specific siRNA degrading RNase with saturating kinetics is involved (“RNase model”). Such a protein has recently been found in C. elegans [21]. We assume that the RNase has Michaelis-Menten kinetics:
The maximum rate of siRNA degradation by the RNase is given by . The non-specific degradation of siRNAs has to be included in the RNase model: since the RNase has a saturated response, the siRNA levels would go to infinity without this non-specific degradation.
In our second extension, we generalize the primed amplification process. Whereas in the standard model the process was limited to the amplification of mRNA, we assume here that siRNAs can also bind to garbage mRNA to trigger dsRNA synthesis (“garbage model”):
The rate of dsRNA synthesis by primed amplification of garbage RNA is given by g3.
As a third extension, we consider a revised, unprimed amplification. We explore the possibility that either RDR is activated by the presence of garbage RNA, or that there is another form of cooperation between garbage RNA pieces and RDR. This has been implemented by replacing the mass action unprimed amplification by a sigmoid (unprimed) amplification (“sigmoid model”):
The maximum rate of unprimed dsRNA synthesis by RDR is given by .
Dynamics of Extended Models
The problem with the primed and unprimed amplification in the core pathway is that the number of secondary siRNAs per primary siRNA is basically independent of the initial dose. Consequently, amplification either results in explosion of the reaction, in the case that the number of secondary siRNAs per primary siRNA is larger then one, or the reaction will die out, in the case that the number of secondary siRNAs per primary siRNA is smaller than one. In contrast, in the extended pathways the number of secondary siRNAs becomes dose dependent by introducing a positive feedback into the system. In the RNase model, dose dependency is caused by the saturation of the siRNA degrading RNase: small numbers of siRNAs are rapidly degraded by the enzyme, while at larger numbers the enzyme becomes saturated, which leads to larger amounts of secondary siRNAs. In the garbage and sigmoid model, the cooperation between garbage and siRNAs, and between garbage pieces themselves, respectively, lead to dose dependency.
The behavior of the extended models is more complex than the core model. We can distinguish three main regions of qualitatively different behavior. One way to switch the system to another qualitatively different behavior is by changing the number of gene copies present in the cell. The bifurcation diagrams with the three regions for all three extended models are shown in Figure 3A, B, and C. Plotted is the equilibrium amount of mRNA against the number of copies of a gene; a stable equilibrium is indicated with a solid line, an unstable with a dashed line. In region I, there is only one attractor: after a perturbation, the system will always return to this attractor. In region II, there are two attractors, and the system can end up in either one of them. In region III, there is again only one attractor. We will first discuss each region separately, with the corresponding types of dsRNA induced silencing, and then we will continue discussing the bifurcation diagram as a whole, to understand the process of transgene induced silencing.
Figure 3 Bifurcation Diagrams of the Extended Models Showing Transgene Induced Silencing
Solid lines indicate stable equilibria; dashed lines unstable equilibria; open circles Hopf bifurcations; and closed circles fold bifurcations. The dynamic behaviors in regions I, II, and III are shown in Figure 4.
Figure 4 Dynamics of the Proposed Models
Grey lines indicate dsRNA levels, black lines mRNA levels. (A), (B), and (C) show transient silencing after dsRNA introduction in the RNase, garbage, and sigmoid model, respectively. (D), (E), and (F) show timeplots of the behavior in the bistable region after introduction of dsRNA: a low dose has only a small effect (dashed lines), but a high dose of dsRNA causes sustained silencing or, in the RNase model, large oscillations (solid lines). (G), (H), and (I) show bar graphs of transgene induced silencing, in the RNase, garbage, and sigmoid model, respectively. Parameter values can be found in Table 1.
In the first region, when there are few copies present, there is only one stable equilibrium. In this default equilibrium, there are low numbers of siRNAs and dsRNA. In this region, mRNA can be silenced transiently by the introduction of homologous dsRNA (Figure 4A, B, and C): when dsRNA is introduced, siRNAs derived from dsRNA cause a strong, rapid decrease of the amount of mRNA, after which the default equilibrium is slowly re-established. Transient silencing after dsRNA injection has been observed in nematodes, flies, and zebrafish [22–24]. Unlike the core pathway, the extended pathways are stable in face of responses against self: a low dose of dsRNA will cause a smaller response than a high dose, and a single dsRNA strand has a negligible effect. This is due to the fact that the amplification in all extended pathways is flux dependent. It means that as long as the copy number is not too high, a low dose of dsRNA will always result in only a small response, and sustained silencing cannot be triggered.
The second region, with an intermediate copy number, is bistable; that is, there are two attractors: the default state and the silenced state. (There is a third equilibrium, which is of the saddle type. The stable manifold of the saddle separates the basins of attraction of the two stable equilibria.) When starting in the default state (dsRNA and siRNAs are almost completely absent), the introduction of a small dose of dsRNA will cause a transient silencing response, after which the default equilibrium is re-established (see Figure 4D, E, and F, dashed lines). A high dose of dsRNA, however, can bring the system from the default equilibrium into the basin of attraction of the silenced equilibrium, which means that sustained silencing is triggered (Figure 4D, E, and F, solid lines). Sustained silencing has been demonstrated in C. elegans, where silencing can persist and even be transmitted to the next generation [18]. Also in plants infected with a virus carrying a gene homologous to a plant gene, silencing of the endogenous gene persists even after removal of the virus [16]. The silencing response in plants can also be transmitted via grafting with very high efficiency from silenced stocks to non-silenced stocks [17].
The existence of two attractors prevents undesired sustained responses: only when the amount of dsRNA exceeds a threshold value is the sustained response mounted. Unfortunately, until now few experiments have focused on the correlation between the dsRNA dose and the duration of the silencing response. Lipardi et al. [12] showed that in Drosophila embryo extract, doses below a threshold concentration failed to induce RNA silencing, while ten times higher doses were able to trigger silencing. This study indicates the existence of a threshold; the duration of the response, however, has not been investigated. The results of Li et al. [24] also indicate the existence of a threshold: they showed that in zebrafish (Danio rerio) embryos small doses of dsRNA lead to partial phenotypic changes only, while high doses of dsRNA lead to more than 50% partial and 35% full phenotypic changes. The partial phenotypic effects could indicate that there was only a transient response in the embryos, while the full phenotypic changes triggered by a high dose of dsRNA could indicate a sustained response.
In the garbage model, the amounts of mRNA, dsRNA, and siRNAs are stable in both attractors, but in the RNase and the sigmoid model, oscillations can occur in this region. The oscillations around the default equilibrium are always of small amplitude, but around the silenced equilibrium they can become large. The region with oscillations is much smaller in the sigmoid model than in the RNase model. We therefore show dynamics with oscillations for the RNase model and without oscillations for the sigmoid model (Figure 4D and F, respectively).
Finally, in the third region, with a high copy number, only the silenced state, with low levels of mRNA and high levels of siRNAs, is stable. The introduction of additional dsRNA will have only a small effect on the already largely reduced amount of mRNA. When a gene is present at very high copy numbers, its mRNA will be silenced continuously.
Technically, the transitions between the regions can be characterized by different bifurcations. In the garbage model, the default state and the silenced state disappear due to fold bifurcations (Figure 3, closed circles) when the number of transgenes is respectively increased or decreased. In the other two models, shortly before the fold bifurcations the equilibria become unstable due to Hopf bifurcations (Figure 3, open circles), which leads to oscillatory behavior around the equilibria. These oscillations then disappear due to homoclinic connection bifurcations. When, by changing the number of transgenes, a fold or homoclinic connection bifurcation is passed, the dynamics immediately jump to the other equilibrium (or around it, in the case that there are oscillations).
The bifurcation diagram depicts the process of transgene induced silencing. In Figure 4G, H, and I we plot bar graphs, where each bar indicates the equilibrium amount of mRNA for a certain number of gene copies, to compare our results to the graphs obtained experimentally in Drosophila by Pal-Bhadra et al. [5]. They inserted one to ten copies of a full Adh transgene using different insertion sites. Stocks with the same copy number showed similar mRNA levels, independent of where the genes were inserted into the genome. One copy resulted in a normal amount of mRNA, while up to five copies, an extra copy resulted in a proportional increase in mRNA levels. However, when a sixth copy was inserted, RNA silencing was triggered, indicated by the presence of siRNAs, and dramatically decreased mRNA levels. At even higher copy numbers, the amount of mRNA was lying around the amount expected for one or two copies only. The bar graphs we obtained with our models are in close correspondence: the amount of mRNA in the cell initially increases with increasing numbers of transgenes; however, when the number of transgenes is increased beyond a threshold level, RNA silencing is triggered.
Parameters Dependence and Predictions
There is only a limited amount of experimentally measured parameters available, which are generally obtained for different model organisms. Moreover, the range of measured values can often be very large. We, therefore, do not focus on specific parameter values, but instead use mean values to show the qualitative dynamics. We then vary the parameter values and infer what kind of qualitative and quantitative changes are to be expected to accompany such parameter changes. When data are available, we compare these model predictions with experiments in which specific parameters have been varied.
The default parameters are given in Table 1. We assumed stable mRNA (half-life 5 h), a 20× faster decay of garbage pieces (half-life 15 min), and slightly more stable siRNAs (half-life 21 min, as measured in human cells [25]). The other parameters are chosen such as to depict the full capabilities of the models.
We distinguish five types of qualitatively different effects that can be caused by changing parameter values (Figure 5). The changes in behavior can be described in terms of the threshold, which is the number of transgenes needed to trigger silencing; and in terms of the bistable point, which is the lower bound of the bistable region. Note that in the sigmoid model, the amount of mRNA in the cell always increases again at high copy numbers.
Figure 5 Changes in the Bifurcation Diagrams of Figure 3 Due to Changing Parameter Values
The black lines indicate the standard parameter values, the blue lines a lower value, and the red lines a higher value for the corresponding parameter.
Type I behavior occurs in the garbage model when changing parameters n, b, and g, and in all three models when changing p. Changing these parameters does not influence the bistable point (that is, the value above which sustained silencing can be triggered), but moves the threshold to different copy number and mRNA levels. This means that the size of the bistable region changes: when the threshold is lower than the bistable point, there is no bistable region, and transgene silencing is triggered at low copy numbers. When the threshold becomes very high, the bistable region is very large: only a very high copy number triggers silencing, while sustained silencing triggered by the introduction of dsRNA becomes possible in a large region.
Type II behavior is typical for the RNase and sigmoid model. In both models, changes in the parameters n, b, g1,2, and k result in type II behavior, as well as dr in the RNase model, and dg in the sigmoid model. Parameter changes move the bistable point and the threshold in the following manner: a low threshold coincides with an even lower bistable point; a high threshold with an even higher bistable point. This means that, in contrast to type I behavior, the bistable region disappears when the threshold is high, and sustained silencing becomes impossible. In the sigmoid model, also the possibility of transgene induced silencing disappears, since there is no noticeable decrease in mRNA levels beyond the threshold. In the RNase model, this is not the case: mRNA levels will always decrease with sufficiently high copy number, although the threshold number of transgenes can sometimes lie outside of the graphs in Figure 5.
Type III behavior occurs for changes in dm in all three models. In this case, the threshold moves to different copy numbers, but the amount of mRNA at the threshold remains the same. Instead, the initial slope of the mRNA level changes. The bistable point is not affected, and since the threshold moves, the bistable region can increase or decrease in size (and even disappear).
Type IV behavior is typical for changes in ds and dg in the garbage model. These parameters scale the complete bifurcation diagram.
Type V behavior occurs in the RNase and the sigmoid model for changes in ds. In this case, the only thing that changes is the amount of mRNA just after transgene silencing is triggered.
Changes in the threshold value of gene copies have been experimentally observed. Several studies suggest that the amount of transcribed mRNA plays an important role in the ability of transcripts to trigger transgene induced silencing. When a transgene is under control of a 35S promoter with a double enhancer, the gene is transcribed at such a high rate that a single transgene can be sufficient to trigger silencing [26]. Also, in petunia the strength of the promoter is correlated with the frequency and degree of silencing [27], and plants homozygous for a transgene are much more often silenced than hemizygous plants [27–32].
These observations are consistent with our models: when a gene is more highly expressed (in our models described by a higher value of i, the transcription rate), fewer copies are needed to trigger silencing. This is because changing i effectively rescales the x-axis of the bifurcation diagrams.
Our models suggest that the amount of dsRNA per mRNA is a major factor determining the threshold. When more dsRNA per mRNA is produced, the threshold to trigger transgene silencing will be lower. For example, an increase in the parameter p, the rate of dsRNA synthesis by RDR, results in all models in a decrease of the threshold. An increase in amplification (g) will have a similar effect. Our results are in line with experimental observations: Forrest et al. [7] showed that in strains that overexpressed RDR, the number of transgenes required to induce silencing is decreased. Likewise, transcripts with tandem IRs, which produce much more dsRNA per mRNA, have shown to be very efficient inducers of RNA silencing [10,33]. Finally, transposons often contain long IRs and are present in high copy numbers, both of which we have shown to induce silencing.
These experimental results are consistent with our models, but they do not make it possible to distinguish between them. Instead, we need experiments in which certain specific parameters are varied. For example, the models predict a completely different effect of the overexpression of RISC (all its components have to be overexpressed). In the garbage and RNase model, RISC overexpression leads to an increase and ultimately disappearance of the threshold. In the sigmoid model, we find the complete opposite: the disappearance of the threshold is not caused by RISC overexpression, but by RISC underexpression (Figure 5).
Discussion
The extended models each provide a unified framework for different RNA silencing phenomena. They provide consistent explanations for (i) dose dependent dsRNA induced silencing; (ii) stability against self-directed responses; (iii) the dependence of transgene induced silencing on RDR; (iv) the effect of IRs; (v) multiple copies; (vi) efficient promoters; and (vii) the ability of transposons to trigger silencing.
Previously it has been proposed that transgene induced silencing is triggered only if the number of transgenes exceeds a threshold level [5,6,29–31]. It has also been observed that the overexpression of RDR reduces the threshold [7]. Our models support the threshold hypothesis and give a mechanistic explanation for it: we propose that the amount of dsRNA per transcript matters.
The extensions also explain differences in RNA silencing phenomena in different species groups. According to our extended models, in organisms that have RDR homolog(s), such as plants, fungi, nematodes, and cellular slime molds, silencing can be induced by transgenes, IRs, transposons, and dsRNA. In contrast, we have shown here that organisms without RDR are unable to trigger transgene induced silencing. Accordingly, experiments have shown that plants with a mutation in RDR are no longer able to bring about transgene induced silencing, while virus (dsRNA) induced silencing is still possible in these strains [34]. The presence of an RDR in Drosophila is currently disputed. Some experiments strongly suggest the presence in Drosophila of an RDR or a protein that functions as an RDR [5,12,35]. Other experiments, however, argue against the presence of such an enzyme (a BLAST search, for example, does not yield an RDR homolog) [36–38]. Since high transgene numbers (without IRs) are capable of inducing RNA silencing in Drosophila, our model suggests that a protein with the same function as RDR must be present. Mammals lack RDR, and in agreement with our models, they are capable of only transient silencing induced by siRNAs [39] (in mammals, dsRNA triggers several non-specific responses [40,41]; sustained silencing in mammals can be accomplished only by continuous expression of siRNAs).
We did not include the effect of siRNAs on DNA chromatin, which is referred to as transcriptional silencing or heterochromatinization. Transcriptional silencing plays a role in transposon silencing [42–45]. Nolan et al. [46], however, recently showed that in the fungus Neurospora crassa the LINE1-like transposon, Tad, is post-transcriptionally silenced and not significantly methylated, indicating that transposon induced silencing in N. crassa can be independent of DNA methylation. Moreover, also in Drosophila transgene induced silencing has been shown to be solely post-transcriptional [5]. Although heterochromatinization can play an important role in transposon silencing, our model study indicates that the addition of heterochromatinization alone—that is, the stop in transcription—to the core pathway will not make transgene silencing possible. Heterochromatinization will only decrease mRNA transcription, and does not provide the necessary positive feedback.
Although it has been shown recently that RISC can perform multiple rounds of cleavage [47], we assume only one cleavage per RISC complex. Adding multiple turnover of RISC, however, does not affect the qualitative behavior of the model (unpublished data).
We proposed three different additions to the pathway. We here suggest some ways of testing or rejecting experimentally the predictions made by the different extensions. In the parameter section, we have already discussed the different behavior of the sigmoid model when RISC is overexpressed. Another difference is that only in the sigmoid model, after silencing is triggered, even higher copy numbers will cause an increase in mRNA levels again. Such an increase, however, can also indicate other sigmoid responses in the pathway, for example in Dicer or RISC. Sigmoid kinetics alone are not able to allow for low mRNA levels when copy numbers become very large. It can be argued, however, that a combination of heterochromatinization with a sigmoid response will be able to keep mRNA levels silenced.
Recently, a siRNA degrading RNase has been found in C. elegans [21]. The pathway with the siRNA degrading RNase, however, is able to cause transgene silencing only when it degrades siRNAs very rapidly and when it saturates quickly. In fact, even when this is the case, this pathway only limitedly allows for sustained RNA silencing, because the parameter range for sustained silencing is very small. We would like to see experiments that focus on the correlation between the dose of dsRNA and the duration of the silencing response. Such experiments will give insights into the existence of a positive feedback and will show if there is a bistable region. The next step would be to investigate the dependence of both the threshold and the size of the bistable region on gene copy numbers, and how this depends on changes in parameters. These observations can then be compared with the dependencies predicted in the parameter section.
The garbage model could be tested by investigating the possibility of siRNAs to serve as primers for RDR on aberrant garbage pieces. When that is possible, we expect that this primed amplification of garbage is a missing step in the RNA silencing pathway. It represents only a small addition to the currently known pathway, but it has a large impact on the dynamics, making transgene and transposon silencing, as well as dose dependent sustained silencing, possible for a wide range of parameters. Therefore, we conclude that in RNA silencing it is “the bits and pieces” that matter.
Materials and Methods
Mathematical proof of limitations of core pathway.
In this section, we prove that the core pathway, with or without amplification, is incapable of transgene induced silencing and sustained silencing.
In transgene induced silencing, the amount of mRNA in the cell initially increases when the number of transgenes is increased, but a further increase in the number of transgenes leads to a sudden drop in the equilibrium amount of mRNA, due to RNA silencing [5]. This implies that two different copy numbers can lead to exactly the same equilibrium amount of mRNA (see Figure 1 or Figure 3 in [5]). We refer to the equilibrium for a low copy number as the default state, and to the equilibrium for a high copy number as the silenced state. Despite the high transcription of mRNA, in the silenced state the amount of mRNA remains low, due to the high levels of dsRNA and siRNAs that are present. Consequently, when it would be possible to keep the amount of mRNA in the cell constant (cf, patch clamp techniques in neuroscience), there should exist mRNA levels for which there are at least two stable equilibria in the system, each with a different amount of dsRNA. Mathematically, this requirement can be studied by considering the variable M as a fixed parameter m. Then for a certain interval of values of m, there should exist at least two stable equilibria, which implies the existence of at least three equilibria, since one equilibrium will be unstable. We will show here that the previously proposed pathway cannot fulfill this requirement and, consequently, is not able to describe and explain transgene induced silencing.
The requirement is analyzed by studying the equilibrium dsRNA (D) level. Without amplification, the dsRNA dynamics consist of two parts only: a positive influx (pm) and a decay by Dicer (aD). In Figure 6, we depict the influx and breakdown of D as a function of D. To obtain an equilibrium, the influx should balance the breakdown—that is, the depicted lines should cross. Hence, without amplification there will be one equilibrium only (Figure 6A). Thus, each gene copy number leads to a unique mRNA level, so it can therefore only be the case that the equilibrium mRNA level monotonically rises with increasing transgene copy number (see Figure 2D). The addition of either primed or unprimed amplification does not allow for an increase in the number of stable equilibria. This can be derived by solving for equilibria by means of putting both dS/dt and dG/dt to zero:
Figure 6 Equilibrium Analysis of the Models Based on the Previously Proposed Pathway
Transgene induced silencing requires the existence of two stable equilibria with different dsRNA (D) levels for one fixed value of mRNA (m). Shown here are the influx (dashed line) and breakdown (solid line) of D as a function of D. Equilibria are found when influx balances breakdown—that is, where these lines cross. Both without and with amplification, there exists only one stable equilibrium; when mass action terms are replaced by saturated responses, there can be zero, one, or two equilibria, of which at most one is stable. The pathway is therefore unable to describe transgene induced silencing.
This means that the amplification terms can be written as a linear function of D, which can never result in more than one stable equilibrium (Figure 6B):
When mass-action terms are replaced by Michaelis-Menten kinetics, the amplification can be rewritten as a sum of saturated responses as well: . When the breakdown is not saturated, this still trivially leads to one equilibrium only. However, when breakdown by Dicer shows a saturated response, more than one equilibrium can be found (but also no equilibria, in the case that the saturation is very rapid). To allow for at least two stable equilibria, the line should cross the line at least three times, because a second equilibrium will always be unstable. (Note that since in this analysis m can be treated as a constant, the term pm could as well have a more complicated—for example, saturated—dependency on m.) This, however, is never possible. The model can be rescaled to . Consider the ratio of both functions:
Equilibria are found when R(D) = 1; to obtain three equilibria, the derivative of R(D) should have at least two roots:
However, by multiplying both sides with the monotonically increasing function (1+D)2, it can be shown that there is at most one root, since the left-hand side is a monotonically increasing function of D:
Consequently, dR/dD has only one root, and there will be at most two equilibria, from which at most one is stable (Figure 6C). Thus, in the previously proposed pathway transgene induced silencing is impossible.
Likewise, model dynamics describing sustained responses require multiple steady states for a unique set of parameters. This requirement is implicitly equivalent to the previous one, since the existence of a second stable equilibrium (with a lower mRNA level) automatically implies that the same equilibrium mRNA level can be found for a lower transgene copy number. That is, the previously proposed pathway can neither describe nor explain transgene induced silencing, nor sustained silencing triggered by injecting dsRNA.
Programs used.
The timeplots in Figures 2 and 3 are produced with GRIND, a computer program for the study of differential equation models by means of numerical integration, steady state analysis, and phase space analysis (http://theory.bio.uu.nl/rdb/software.html). The bifurcation diagrams are produced with CONTENT, an integrated environment for bifurcation analysis of dynamical systems (http://www.math.uu.nl/people/kuznet/CONTENT/).
We thank M. van Hoek and T. Sijen for helpful comments on the manuscript. This research was funded by the Netherlands Organization for Scientific Research (NWO) through Grant 050.50.202 of the BioMolecular Informatics program.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. MACG, AFMM, and PH conceived and designed the models. MACG performed the numerical computations. MACG, AFMM, and PH wrote the paper.
Abbreviations
dsRNAdouble-stranded RNA
IRinverted repeat
RDRRNA directed RNA polymerase
RISCRNA induced silencing complex
siRNAsmall interfering RNA
==== Refs
References
Ratcliff F Harrison BD Baulcombe DC 1997 A similarity between viral defense and gene silencing in plants Science 276 1558 1560 18610513
Covey SN Al-Kaff NS Lángara A Turner DS 1997 Plants combat infection by gene silencing Nature 385 781 782
Napoli C Lemieux C Jorgensen R 1990 Introduction of a chimeric chalcone synthase gene into petunia results in reversible co-suppression of homologous genes in trans Plant Cell 2 279 289 12354959
Van der Krol AR Mur LA Beld M Mol JN Stuitje AR 1990 Flavonoid genes in petunia: Addition of a limited number of gene copies may lead to a suppression of gene expression Plant Cell 2 291 299 2152117
Pal-Bhadra M Bhadra U Birchler JA 2002 RNAi related mechanisms affect both transcriptional and posttranscriptional transgene silencing in Drosophila
Mol Cell 9 315 327 11864605
Lechtenberg B Schubert D Forsbach A Gils M Schmidt R 2003 Neither inverted repeat T-DNA configurations nor arrangements of tandemly repeated transgenes are sufficient to trigger transgene silencing Plant J 34 507 517 12753589
Forrest EC Cogoni C Macino G 2004 The RNA-dependent RNA polymerase, QDE-1, is a rate-limiting factor in post-transcriptional gene silencing in Neurospora crassa
Nucleic Acids Res 32 2123 2128 15090622
Wu-Scharf D Jeong B Zhang C Cerutti H 2000 Transgene and transposon silencing in Chlamydomonas reinhardtii by a DEAH-box RNA helicase Science 290 1159 1162 11073454
Tabara H Sarkissian M Kelly WG Fleenor J Grishok A 1999 The rde-1 gene, RNA interference, and transposon silencing in C. elegans
Cell 99 123 132 10535731
Sijen T Plasterk RH 2003 Transposon silencing in the Caenorhabditis elegans germ line by natural RNAi Nature 426 310 314 14628056
Sijen T Fleenor J Simmer F Thijssen KL Parrish S 2001 On the role of RNA amplification in dsRNA-triggered gene silencing Cell 107 465 476 11719187
Lipardi C Wei Q Paterson BM 2001 RNAi as random degradative PCR: siRNA primers convert mRNA into dsRNAs that are degraded to generate new siRNAs Cell 107 297 307 11701121
Makeyev EV Bamford DH 2002 Cellular RNA-dependent RNA polymerase involved in posttranscriptional gene silencing has two distinct activity modes Mol Cell 10 1417 1427 12504016
Schiebel W Haas B Marinkovic S Klanner A Sanger HL 1993 RNA-directed RNA polymerase from tomato leaves. II. Catalytic in vitro properties J Biol Chem 268 11858 11867 7685023
Stein P Svoboda P Anger M Schultz RM 2003 RNAi: Mammalian oocytes do it without RNA-dependent RNA polymerase RNA 9 187 192 12554861
Ruiz MT Voinnet O Baulcombe DC 1998 Initiation and maintenance of virus-induced gene silencing Plant Cell 10 937 946 9634582
Palauqui JC Elmayan T Pollien JM Vaucheret H 1997 Systemic acquired silencing: Transgene-specific post-transcriptional silencing is transmitted by grafting from silenced stocks to non-silenced scions EMBO J 16 4738 4745 9303318
Fire A Xu S Montgomery MK Kostas SA Driver SE Mello CC 1998 Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans
Nature 391 806 811 9486653
Bergstrom CT McKittrick E Antia R 2003 Mathematical models of RNA silencing: Unidirectional amplification limits accidental self-directed reactions Proc Natl Acad Sci U S A 100 11511 11516 12972639
Klahre U Crete P Leuenberger SA Iglesias VA Meins F Jr. 2002 High molecular weight RNAs and small interfering RNAs induce systemic posttranscriptional gene silencing in plants Proc Natl Acad Sci U S A 99 11981 11986 12181491
Kennedy S Wang D Ruvkun G 2004 A conserved siRNA-degrading RNase negatively regulates RNA interference in C. elegans
Nature 427 645 649 14961122
Montgomery MK Xu S Fire A 1998 RNA as a target of double-stranded RNA-mediated genetic interference in Caenorhabditis elegans
Proc Natl Acad Sci USA 95 15502 15507 9860998
Kennerdell JR Carthew RW 1998 Use of dsRNA-mediated genetic interference to demonstrate that frizzled and frizzled 2 act in the wingless pathway Cell 95 1017 1026 9875855
Li YX Farrell MJ Liu R Mohanty N Kirby ML 2000 Double-stranded RNA injection produces null phenotypes in zebrafish Dev Biol 217 394 405 10625563
Chiu YL Rana TM 2003 sirna function in RNAi: A chemical modification analysis RNA 9 1034 1048 12923253
Elmayan T Vaucheret H 1996 Expression of single copies of a strongly expressed 35s transgene can be silenced post-transcriptionally Plant J 9 787 797
Que Q Wang HY English JJ Jorgensen RA 1997 The frequency and degree of cosuppression by sense chalcone synthase transgenes are dependent on transgene promoter strength and are reduced by premature nonsense codons in the transgene coding sequence Plant Cell 9 1357 1368 12237385
Dehio C Schell J 1994 Identification of plant genetic loci involved in a posttranscriptional mechanism for meiotically reversible transgene silencing Proc Natl Acad Sci U S A 91 5538 5542 8202523
De Carvalho F Gheysen G Kushnir S Van Montagu M Inze D 1992 Suppression of beta-1,3-glucanase transgene expression in homozygous plants EMBO J 11 2595 2602 1378394
Jorgensen RA Cluster PD English J Que Q Napoli CA 1996 Chalcone synthase cosuppression phenotypes in petunia flowers: Comparison of sense vs. antisense constructs and single-copy vs. complex T-DNA sequences Plant Mol Biol 31 957 973 8843939
Palauqui JC Vaucheret H 1995 Field trial analysis of nitrate reductase co-suppression: A comparative study of 38 combinations of transgene loci Plant Mol Biol 29 149 159 7579160
Dorlhac de Borne F Vincentz M Chupeau Y Vaucheret H 1994 Co-suppression of nitrate reductase host genes and transgenes in transgenic tobacco plants Mol Gen Genet 243 613 621 8028577
Rohr J Sarkar N Balenger S Jeong BR Cerutti H 2004 Tandem inverted repeat system for selection of effective transgenic RNAi strains in Chlamydomonas
Plant J 40 611 621 15500475
Dalmay T Hamilton A Rudd S Angell S Baulcombe DC 2000 An RNA-dependent RNA polymerase gene in Arabidopsis is required for posttranscriptional gene silencing mediated by a transgene but not by a virus Cell 101 543 553 10850496
Wei Q Lipardi C Paterson BM 2003 Analysis of the 3′-hydroxyl group in Drosophila siRNA function Methods 30 337 347 12828948
Zamore PD Tuschl T Sharp PA Bartel DP 2000 RNAi: Double-stranded RNA directs the ATP-dependent cleavage of mRNA at 21 to 23 nucleotide intervals Cell 101 25 33 10778853
Celotto AM Graveley BR 2002 Exon-specific RNAi: A tool for dissecting the functional relevance of alternative splicing RNA 8 718 724 12088145
Roignant JY Carre C Mugat B Szymczak D Lepesant JA 2003 Absence of transitive and systemic pathways allows cell-specific and isoform-specific RNAi in Drosophila
RNA 9 299 308 12592004
Caplen NJ Parrish S Imani F Fire A Morgan RA 2001 Specific inhibition of gene expression by small double-stranded RNAs in invertebrate and vertebrate systems Proc Natl Acad Sci U S A 98 9742 9747 11481446
Clemens MJ Elia A 1997 The double-stranded RNA-dependent protein kinase PKR: Structure and function J Interferon Cytokine Res 17 503 524 9335428
Player MR Torrence PF 1998 The 2–5A system: Modulation of viral and cellular processes through acceleration of RNA degradation Pharmacol Ther 78 55 113 9623881
Mette MF Aufsatz W Van der Winden J Matzke MA Matzke AJ 2000 Transcriptional silencing and promoter methylation triggered by double-stranded RNA EMBO J 19 5194 5201 11013221
Morris KV Chan SW Jacobsen SE Looney DJ 2004 Small interfering RNA-induced transcriptional gene silencing in human cells Science 305 1289 1292 15297624
Kawasaki H Taira K 2005 Transcriptional gene silencing by short interfering RNAs Curr Opin Mol Ther 7 125 131 15844619
Hamilton A Voinnet O Chappell L Baulcombe D 2002 Two classes of short interfering RNA in RNA silencing EMBO J 21 4671 4679 12198169
Nolan T Braccini L Azzalin G De Toni A Macino G 2005 The post-transcriptional gene silencing machinery functions independently of DNA methylation to repress a LINE1-like retrotransposon in Neurospora crassa
Nucleic Acids Res 33 1564 1573 15767281
Haley B Zamore PD 2004 Kinetic analysis of the RNAi enzyme complex Nat Struct Mol Biol 11 599 606 15170178
Hutvágner G Zamore PD 2002 RNAi: Nature abhors a double-strand Curr Opin Genet Dev 12 225 232 11893497
Wang Y Liu CL Storey JD Tibshirani RJ Herschlag D 2002 Precision and functional specificity in mRNA decay Proc Natl Acad Sci U S A 99 5860 5865 11972065
Yang E Van Nimwegen E Zavolan M Rajewsky N Schroeder M 2003 Decay rates of human mRNAs: Correlation with functional characteristics and sequence attributes Genome Res 13 1863 1872 12902380
Taylor CB Green PJ 1995 Identification and characterization of genes with unstable transcripts (GUTs) in tobacco Plant Mol Biol 28 27 38 7787185
|
16110335
|
PMC1185647
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Jul 29; 1(2):e21
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010021
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 10.1371/journal.pcbi.001002205-PLCB-RA-0104R2plcb-01-02-07Research ArticleBioinformatics - Computational BiologyGenetics/GenomicsGenetics/Genome ProjectsDrosophilaEukaryotesNoneCombined Evidence Annotation of Transposable Elements in Genome Sequences Combined Evidence TE AnnotationQuesneville Hadi 1*Bergman Casey M 2¤Andrieu Olivier 1Autard Delphine 1Nouaud Danielle 1Ashburner Michael 2Anxolabehere Dominique 11 Laboratoire Dynamique du Génome et Evolution, Institut Jacques Monod, Paris, France
2 Department of Genetics, University of Cambridge, Cambridge, United Kingdom
Stormo Gary EditorWashington University in St. Louis, United States of America*To whom correspondence should be addressed. E-mail: [email protected]¤Current address: Faculty of Life Sciences, University of Manchester, Manchester, United Kingdom
7 2005 29 7 2005 1 2 e2211 5 2005 30 6 2005 Copyright: © 2005 Quesneville 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.Transposable elements (TEs) are mobile, repetitive sequences that make up significant fractions of metazoan genomes. Despite their near ubiquity and importance in genome and chromosome biology, most efforts to annotate TEs in genome sequences rely on the results of a single computational program, RepeatMasker. In contrast, recent advances in gene annotation indicate that high-quality gene models can be produced from combining multiple independent sources of computational evidence. To elevate the quality of TE annotations to a level comparable to that of gene models, we have developed a combined evidence-model TE annotation pipeline, analogous to systems used for gene annotation, by integrating results from multiple homology-based and de novo TE identification methods. As proof of principle, we have annotated “TE models” in Drosophila melanogaster Release 4 genomic sequences using the combined computational evidence derived from RepeatMasker, BLASTER, TBLASTX, all-by-all BLASTN, RECON, TE-HMM and the previous Release 3.1 annotation. Our system is designed for use with the Apollo genome annotation tool, allowing automatic results to be curated manually to produce reliable annotations. The euchromatic TE fraction of D. melanogaster is now estimated at 5.3% (cf. 3.86% in Release 3.1), and we found a substantially higher number of TEs (n = 6,013) than previously identified (n = 1,572). Most of the new TEs derive from small fragments of a few hundred nucleotides long and highly abundant families not previously annotated (e.g., INE-1). We also estimated that 518 TE copies (8.6%) are inserted into at least one other TE, forming a nest of elements. The pipeline allows rapid and thorough annotation of even the most complex TE models, including highly deleted and/or nested elements such as those often found in heterochromatic sequences. Our pipeline can be easily adapted to other genome sequences, such as those of the D. melanogaster heterochromatin or other species in the genus Drosophila.
Synopsis
A first step in adding value to the large-scale DNA sequences generated by genome projects is the process of annotation—marking biological features on the raw string of adenines, cytosines, guanines, and thymines. The predominant goal in genome annotation thus far has been to identify gene sequences that encode proteins; however, many functional sequences exist in non-protein-coding regions and their annotation remains incomplete. Mobile, repetitive DNA segments known as transposable elements (TEs) are one class of functional sequence in non-protein-coding regions, which can make up large fractions of genome sequences (e.g., about 45% in the human) and can play important roles in gene and chromosome structure and regulation. As a consequence, there has been increasing interest in the computational identification of TEs in genome sequences. Borrowing current ideas from the field of gene annotation, the authors have developed a pipeline to predict TEs in genome sequences that combines multiple sources of evidence from different computational methods. The authors' combined-evidence pipeline represents an important step towards raising the standards of TE annotation to the same quality as that of genes, and should help catalyze their understanding of the biological role of these fascinating sequences.
Citation:Quesneville H, Bergman CM, Andrieu O, Autard D, Nouaud D, et al. (2005) Combined evidence annotation of transposable elements in genome sequences. PLoS Comp Biol 1(2): e22.
==== Body
Introduction
Transposable elements (TEs) are mobile, repetitive DNA sequences that constitute a structurally dynamic component of genomes. The taxonomic distribution of TEs is virtually ubiquitous: they have been found in nearly all eukaryotic organisms studied, with few exceptions. TEs represent quantitatively important components of genome sequences (e.g., 44.4% of the human genome; [1]), and there is no doubt that modern genomic DNA has evolved in close association with TEs. TEs show high species specificity, and the number and types of TE can differ quite dramatically between even closely related organisms. There is abundant circumstantial evidence that TEs may transfer horizontally between species by mechanisms that remain obscure. The forces controlling the dynamics of TE spread within a species are also poorly understood, as are the systemic effects of the elements on their host genomes. Insertions of individual TEs may lead to genome restructuring (e.g., the occurrence of inversions), mutations in genes, or changes in gene regulation. Some TE insertions may even have become domesticated to play roles in the normal functions of the host (see [2] for review). Despite their manifold effects, abundance, and ubiquity, we understand very little about most aspects of TE biology.
One way of furthering our knowledge of TE biology is through the computational analysis of TEs in the growing number of complete genomic sequences. By detailed comparison of the abundance and distribution of TEs in entire genomes, we can infer the fundamental biological properties of TEs that are shared or that differ among species. However, meaningful inferences about TE biology based on computationally derived TE annotations can only be done if we are confident about the results of these analyses. The hallmark of a strong result in computational biology should be its robustness to the particular method used. The annotation of TEs, however, typically relies on the results of a single computational program, RepeatMasker (http://www.repeatmasker.org/), which recent studies indicate may be “neither the most efficient nor the most sensitive approach” for TE annotation [3]. By contrast, recent advances in the field of gene annotation indicate that high-quality gene models can be produced by combining multiple independent sources of computational evidence [4–9]. With the recent development of several new methods for TE and repeat detection [10–16], it is now possible to apply a similar “combined evidence” approach to elevate the quality of TE annotations to a level comparable to that of gene models.
To achieve this aim, we have developed a TE annotation pipeline that integrates results from multiple homology-based and de novo TE identification methods. Currently, our pipeline uses the combined computational evidence derived from RepeatMasker (http://www.repeatmasker.org/), BLASTER [13], TBLASTX [17], all-by-all BLASTN [17], RECON [10], TE-HMM [14], and previously published TE annotations [18]. We have designed our system to use an “evidence-model” framework and the Apollo genome annotation tool [19], allowing computational evidence to be manually curated in an efficient manner to produce reliable “TE models”. The pipeline allows rapid and thorough annotation of complex TE models, providing key structural details that allow insights into the origin of highly deleted and/or nested elements. In contrast to simply masking repeats, our method provides the means to a complete and accurate annotation of TEs, supported by multiple sources of computational evidence, a goal that has important implications for experimental studies of genome and chromosome biology.
As a test case we have chosen to annotate the euchromatic genomic sequence of the fruit fly, Drosophila melanogaster. The 116.8-Mb Release 3 genome sequence of D. melanogaster is among the highest quality genome sequences and is a particularly well suited sequence for genome-wide studies of TEs, since repetitive DNA sequences have been finished to high quality and systematically verified by restriction fingerprint analysis [20]. Moreover, the Release 3.1 annotation of D. melanogaster includes a manually curated set of TE annotations [18] that can be used as a benchmark for developing and refining TE annotation methodologies. Controlled tests performed here on the Release 3 sequence show that a combined-evidence approach has superior performance over individual TE detection methods, and that a substantially larger fraction of the genome is composed of TEs than previously estimated. We have applied our pipeline to the new 118.4-Mb Release 4 sequence (http://www.fruitfly.org/annot/release4.html), which has closed several of the gaps in Release 3 and has extended the sequence of the pericentomeric regions, to produce a systematic re-annotation of TEs in the D. melanogaster genome. The euchromatic TE fraction is now estimated at 5.3% (cf. 3.86% in Release 3.1), and we found a substantially higher number of TEs (n = 6,013) than previously identified (n = 1,572). We also estimated that 518 TE copies (8.6%) are inserted into at least one other TE, forming a nest of elements. Our pipeline can be easily adapted to other genome sequences, and could markedly increase the efficiency of annotating genomic regions with complex or abundant TE insertions such as heterochromatic sequences.
Results
Evaluation of Methods
The first step in the development of our pipeline was to evaluate the abilities of different computational tools that are available to annotate TEs in order to assess the strengths and weaknesses of each method. To do this we re-annotated the D. melanogaster Release 3 sequence using different TE detection methods and compared these results to the FlyBase Release 3.1 annotation (http://www.flybase.org/annot/release3.html), which includes the results of a manually curated set of TE annotations published previously by Kaminker et al. [18].
Methods for TE annotation fall into two general classes: (i) methods designed for the annotation of known TE families, which utilize a specific reference sequence (also called a canonical sequence) and (ii) de novo methods designed for the annotation of anonymous TE families, for which no reference sequence has yet been identified. This distinction is necessary since it determines the relevant measures to evaluate different methods for TE detection.
Methods for the Annotation of Known TE Families
To allow direct comparison with previous results [18], we used the Release 3 genomic sequence as a query to be scanned for similarity to reference sequences in version 7.1 of the Berkeley Drosophila Genome Project (BDGP) TE dataset (http://www.fruitfly.org/p_disrupt/TE.html), the same version that was used for the Release 3.1 FlyBase annotation. We initially tested three methods for TE prediction (see Materials and Methods for details): (i) BLASTER using BLASTN followed by chaining with MATCHER (BLRn), (ii) RepeatMasker using default parameters (RM), and (iii) RM using default parameters followed by chaining with MATCHER (RMm). The last method was used to test the benefit of the “chaining algorithm” implemented in MATCHER.
We compared predictions to annotations by calculating sensitivity and specificity values for the number of nucleotides of TE sequence predicted by a method that overlapped (or did not overlap) TEs in the Release 3.1 FlyBase annotation (see Materials and Methods). Note that the computation of specificity is biased here, since it assumes that all TEs in the Release 3.1 FlyBase annotation are known, which is certainly not true. We also compared different categories of overlap between prediction and annotation boundaries to gain deeper insight into the details of TE detection methods (see Materials and Methods). These results are summarized in Table 1.
Table 1 Results of Comparisons between TE Prediction Methods That Use Reference Sequences and the Release 3 FlyBase TE Annotations
Relationships of predictions to annotations can be categorized as 1-to-1, 1-to-n,
n-to-1, 1-to-0, or n-to-n (where n > 1; see Materials and Methods for details).
We found that both the sensitivity and the specificity to predict Release 3.1 TEs were higher for BLRn (96.9% and 99.7%, respectively) than for RM (94.3% and 99.1%, respectively). In addition, 28% more Release 3.1 TEs were predicted exactly by BLRn (n = 854) than by RM (n = 664). BLRn also made well over an order of magnitude fewer “method not joined” errors (n = 3) than RM (n = 110), indicating that the BLRn strategy makes high-quality automatic decisions about joining fragments of TEs. RMm had intermediate performance with respect to RM and BLRn for exactly predicting Release 3.1 annotations (n = 711), but, like BLRn, had few “method not joined” errors (n = 6). These results may be explained partly by the fact the Release 3.1 annotation was produced using BLAST-based methods [18], and that the local alignment stop criterion significantly differs between the BLAST algorithm and the Smith and Waterman algorithm used by RM (in the final search phase). Thus, the good performance of BLRn for predicting Release 3.1 TE boundaries could result from the fact that the same local alignment stop criterion was used. However, differences in local alignment matching cannot explain these results entirely, since RMm outperformed RM to recover exact matches, indicating that the chaining algorithm implemented in MATCHER is a significant improvement over raw RM results for predicting Release 3.1 TE annotations.
RM identified approximately 3-fold more new TEs than BLRn, and thus appears to be a more sensitive method for the detection of previously unannotated TEs. But here also RMm had a better performance for detecting new TEs than RM, so the effects of chaining can also improve RM in this regard. The putative TEs predicted by RM in general were short, as can be seen by the relatively limited effect that an additional 3,000+ predictions had on the genome-wide specificity of RM and RMm.
Given the different performance of these approaches, we developed and tested a fourth strategy that attempts to capitalize on the strengths of both RM and BLRn. This method, called RepeatMasker-BLASTER (RMBLR), combined hits from both BLRn and RM and gave them to MATCHER for chaining. To do this, we normalized alignment scores from BLRn and RM to be the hit length for chaining. As shown in Table 1, an optimized RMBLR had higher sensitivity than RM, RMm, or BLRn alone, produced the highest number of putative new TE annotations, and otherwise retained performance features similar to RMm and BLRn. These results show that a combined approach to TE annotation is more efficient at both recovering known TE annotations and predicting new ones than each method alone.
The results shown in Table 1 also suggest that there were errors in the Release 3.1 FlyBase annotation. Among them, the tools predicted cases where two annotations could be joined automatically (category “annotation not joined” in Table 1) and others where an annotation might be split (category “annotation over-joined” in Table 1). Using the Apollo annotation editor [19] to inspect these errors visually, we found that the fragmented and the nested structures of TEs often could be recovered better with these tools than in the Release 3.1 FlyBase annotation. In addition, using Apollo we found that the many new copies appear to be bona fide remnants of TEs missing from the previous annotation; however, a detailed analysis of Release 4 revealed that many of these new TEs may result from spurious hits to simple repeats in the reference sequence (see below).
Methods for the Annotation of Anonymous TE Families
We also tested de novo methods to predict TEs that do not use a specific reference sequence, and evaluated the ability of these methods to find TEs in the Release 3.1 D. melanogaster annotation. These results serve to determine the ability of each method to identify anonymous TEs, and are important for the annotation of genome sequences where a manually curated reference set of TEs is not available. Individually, we found that these methods have lower performance than those that use specific reference sequences, but together they provide additional evidence that can be used to evaluate TE models in the final manual curation step.
TEs have been predicted anonymously using four different methods: (i) an all-by-all genome comparison with BLASTER using BLASTN followed by chaining with MATCHER and grouping with GROUPER (BLRa), (ii) RECON, using default parameters, (iii) BLASTER using TBLASTX with the entire Repbase Update as the database, followed by chaining with MATCHER (BLRtx), and (iv) a hidden Markov model that detects TE sequences based on nucleotide composition (TE-HMM). Note that for BLRa, we compared coordinates of the group of sequences obtained by GROUPER with a coverage of zero (i.e., all overlapping matches were merged; see Materials and Methods for details).
As above, sensitivity, specificity, and the comparison of boundaries between predictions and annotations were used to evaluate the performance of each method. Note again that, as previously, specificity is here biased because it assumes for its computation that all TEs in the genome are known. Here, specificity may be less meaningful than above, since the ability of these methods to detect new TEs is enhanced, and methods detecting many new TEs would have a correspondingly low specificity. Therefore, we must be careful to interpret specificity here as the ability to detect only already known TEs.
Table 2 shows that all de novo methods had relatively high overall specificity (>88%) to detect Release 3.1 TE annotations, but that RECON gave the best performance to recover Release 3.1 TEs exactly. BLRtx had the highest overall sensitivity to detect Release 3.1 TEs (97.2%), which may be explained by the fact that this method uses Repbase Update, which includes most of the Drosophila TEs. This can be shown by a similar analysis with Drosophila TEs removed from the Repbase Update (see BLRtxNoDros in Table 2), which gave lower sensitivity (44.2%), fewer new TEs (n = 8,110), and no “exact”, “near exact”, or “equivalent” cases. BLRtx and TE-HMM detected thousands more new putative TEs than RECON, BLRa, and the other methods detailed in Table 1, indicating that many new TE families may remain to be described in the D. melanogaster genome [13]. These new families are probably low in copy number and represented by nonoverlapping fragments, as suggested by the smaller number of new TEs found by BLRa and RECON. In fact RECON could only detect TEs that are repeated and have copies that are more or less well conserved to their extremities. BLRtx and TE-HMM would be able to detect TEs in few copies (even unique elements) that could be highly diverged and/or degenerate. It is perhaps surprising that BLRtx predicts the highest number of new TEs, since TE-HMM would be able to detect copies for which no distant TE reference sequence is known. However, the high number of BLRtx and BLRtxNoDros predictions may result from an under-joining of fragments of the same TE, as suggested by the large number of “method not joined” cases: n = 1,172 (BLRtx) and n = 3,587 (BLRtxNoDros). In contrast, the high number of predictions resulting from TE-HMM do not appear to result from under-joining (“method not joined”; n = 42), but rather (with their relatively low sensitivity and specificity) suggest a tendency to overpredict using the current parameters. Together these results demonstrate that de novo methods provide specific evidence that can be used to support TE models, but additional development is necessary to fine-tune these approaches to generate accurate TE annotations directly.
Table 2 Results of Comparisons between TE Prediction Methods That Do Not Use Reference Sequences and the Release 3 FlyBase TE Annotations
Relationships of predictions to annotations can be categorized as 1-to-1, 1-to-n,
n-to-1, 1-to-0, or n-to-n (where n > 1; see Materials and Methods for details).
The Annotation Pipeline
Based on these results, we designed an integrated pipeline to compute and store evidence and TE annotations for genome sequences (Figure 1). Our annotation pipeline is composed of (i) TE detection software such as BLASTER, RepeatMasker, TE-HMM, and RECON; (ii) satellite detection software such as RepeatMasker, Tandem Repeat Finder (TRF) [21], and Mreps [22], (iii) a MySQL database (http://www.mysql.com/) to manage the results of these methods and the annotations generated from them; and (iv) Open Portable Batch System (http://www.openpbs.org/) for distributing jobs on a computer cluster. The flexible architecture of this system easily allows other methods for TE detection to be added to this pipeline in the future.
Figure 1 Schematic of Our TE Annotation Pipeline
The pipeline is composed of (i) known TE family detection methods such as BLRn, RM, and RMBLR; (ii) satellite detection software such as RM, TRF, and Mreps; (iii) anonymous TE detection methods such BLRa, TE-HMM, RECON, and BLRtx; and (iv) a MySQL database called REPET to manage the results and the annotations. GAME-XML files are then generated from the results stored in the database and loaded into the Apollo genome annotation tool, allowing automatic results to be manually curated to produce a reliable annotation. To facilitate manual curation, we automatically promoted RMBLR results to be the candidate annotation.
To save computer time and reduce software memory requirements, we segmented the Release 4 genomic sequences into chunks of 200 kb overlapping by 10 kb. Each chunk was then independently analyzed by the different analysis programs, and the results were stored in the MySQL database. GAME-XML (http://www.fruitfly.org/annot/apollo/game.rng.txt) files were then generated from the results stored in the database and loaded into the Apollo genome annotation tool, allowing automatic results to be manually curated to produce a reliable annotation. For this curation we used as evidence tiers (i) the Release 3.1 FlyBase annotations with coordinates mapped to the Release 4 sequences, (ii) BLRn, RM, and RMBLR results using version 9.0 of the BDGP TE reference set, (iii) BLRtx using Repbase Update 8.12, and (iv) RECON, BLRa, and TE-HMM (see Materials and Methods for details). We required all annotations to be supported by at least one of the methods for detecting known TEs—BLRn, RM, or RMBLR. We did not include annotations based solely on anonymous prediction methods since these methods potentially suffer from high false positive rates (Table 2), even though our analyses on Release 3 suggested that there may be additional families of TEs yet to be discovered in the D. melanogaster genome. We note that our pipeline is currently designed with the goal of achieving the best possible annotation set of known TEs in a genome sequence, not the discovery of new TE families, an important endeavor in its own right but outside the scope of the current work.
To facilitate manual curation, we automatically promoted the results of RMBLR to be the candidate annotation (defined as a set of one or more joined fragments), which could then be validated or modified by the curator in Apollo according to the evidence available in the GAME-XML file (see Figure 2 for an example). In addition, we generated a candidate list of mis-joined matches that were contiguous but not joined by MATCHER because of the size of the deletion or the insertion in the genomic sequence. This list identified potential problem cases to be considered carefully for manual joins in Apollo. Moreover, we used RMBLR with conservative settings (gap penalty of 0.05), intentionally under-joining contiguous matches compared to the optimal setting (gap penalty of 0.04; see Table 1). Hence, the join decision of the most difficult cases is left to the curator. Another consequence of this conservative approach is that only a few annotations were manually split. This happened when two small and distant fragments (generally neighboring copies of INE-1 [23]) were automatically joined, and the insert between the two fragments did not correspond to another TE (as would be the case for a nested TE). We considered these joins excessive because of the lack of knowledge about the biology of the INE-1 TE family, for which it is difficult to find a reliable reference sequence. We initially split the five major chromosome arms among five curators for a first-pass manual curation, which was completed in less than 2 wk. Subsequent to this, a single curator performed a second-pass manual curation in order to improve the consistency of manual edit decisions. We examined 10,348 annotations, and only 523 (5%) of them needed to be edited. Finally, we obtained 9,053 unique TE annotations after merging annotations in the overlaps between chunks.
Figure 2 Screenshot of an Apollo View for a Peri-Centromeric Region with Extreme TE Density
Curated annotations on both forward strand (top) and reverse strand (bottom) are displayed in the light blue panels. Evidence tiers are shown in the black panels: TE-HMM (yellow), RECON (light purple), BLRa (violet), BLRtx (red), BLRn (teal), RM (blue), RMBLR (light green), and Release 3.1 FlyBase annotations (peach).
During the manual curation step, we encountered an unexpectedly large number of apparently spurious hits to particular TE families resulting from similarity to simple repeats present in the reference sequence. For example, 236 of 373 predicted TEs for the roo family [24] were generated only by matches to the [CA(A/G)]n repeat in the roo reference sequence. Since the number of spurious hits resulting from simple repeats is potentially quite large, we considered several alternative strategies for their automatic removal. We rejected the possibility of masking the reference sequences and/or the genome for simple repeats, because that could have decreased dramatically the sensitivity of the detection of TEs that have many simple repeats in their reference sequence. Moreover, this strategy does not guarantee the removal of simple repeats that are too degenerate from a regular pattern to be detected, but that could still produce spurious hits because of differences in simple repeat detection versus TE detection.
Instead, we settled on a two-step post-processing of our curated predictions that first identified all annotations that were less than a length threshold after removing regions that overlapped simple repeat regions. These putative spurious hits were then used as queries in a filtered BLAST against the BDGP TE reference set to “rescue” false spurious hits (i.e., real TEs) from true spurious hits. To develop this method, we used the roo family as a training set, for which we could easily partition spurious from real TE annotations. We tested the ability of three methods for simple repeat detection—RepeatMasker, Mreps, and TRF—to discriminate real from spurious roo annotations as a function of length remaining after simple repeat removal. We found that using RepeatMasker with a length threshold of 170 bp allowed all 236 spurious roo annotations to be identified with no real annotations identified as spurious (data not shown).
Using this threshold we detected 3,058 putative spurious hits, which were then searched with BLASTN (E-value > 1 × 10−15) using the “dust” filtering option against our reference TE sequence set. We found that only 18 of the 3,058 putative spurious hits were rescued as real annotations, indicating that our simple repeat filtering thresholds have high specificity. These 3,040 putative spurious hits were removed from the final set of Release 4 TE annotations submitted to FlyBase. Finally, to understand the source of these spurious hits in the auto-promoted TE models, we analyzed the overlap of the 3,040 spurious hits with Release 4 predictions generated individually by BLRn and RM. We find that 2,898 (95%) of the spurious hits overlapped a RM prediction, whereas only 1,255 (41%) of the spurious hits overlapped a BLRn prediction, indicating that RM generated a greater proportion of the spurious hits than BLRn.
Discussion
We have developed and implemented a combined-evidence pipeline to annotate TEs in genome sequences and applied this novel system to detecting TEs in the Release 4 sequence of D. melanogaster. Our work fulfills the demand for a unified approach to TE annotation that capitalizes on the strength of multiple TE detection methods [3] and places TE annotation on common conceptual framework with gene annotation [5–9]. Compared with annotations generated for the Release 3 sequence [18], we confirmed precisely 743 out of 1,572 TE annotations. We adjusted the boundaries of 488, joined 80, changed the strand of 66, changed the name of 14, split 16, and described 4,573 new TE annotations. (Note that the number of modifications does not total 1,572 since multiple Release 3 elements were incorporated in a single join). These 4,573 new TE annotations are all supported by significant nucleotide homology to previously recognized families of TEs in Drosophila. According to our annotation the euchromatic TE fraction is now estimated to be 5.3% (cf. 3.86% in Release 3.1), and we found a substantially higher number of TEs (n = 6,013) than previously identified (n = 1,572). Most of the new TEs derive from small fragments of about a few hundred nucleotides long, and from highly abundant families not previously annotated (e.g., INE-1). Taking into account the heterochromatic TE fraction estimated by Hoskins et al. [25] and the fraction of this compartment (1/3 of the genome), we can estimate that in D. melanogaster TEs represent about 20% of the whole genome (about 5% of the euchromatin and about 50% of the heterochromatin). The pipeline allows rapid and thorough annotation of even the most complex TE models, including highly deleted and/or nested elements. We now estimate that 518 TE copies (8.6%) are inserted into at least one other TE, forming a nest. A detailed description of abundance and distribution of TEs in Release 4 based on the result of this annotation is in preparation. The full annotation is available through FlyBase (http://www.flybase.org) and the REPET database (http://dynagen.ijm.jussieu.fr/repet/).
Performance
Our studies on the Release 3 sequence provide a first detailed genome-wide analysis of different methods for TE detection relative to a manually curated reference set of TE annotations. These results (see Tables 1 and 2) provide insight into the strengths and weaknesses of each method and therefore a deeper understanding of the consequences of algorithmic differences for TE detection. In general, our results suggest that BLRn can outperform RM with respect to the precise determination of TE boundaries, and that much of this improvement derives from the joining algorithm implemented in MATCHER. On the other hand, RM appears to be more sensitive for the detection of small and divergent TE copies. RM can detect small copies with less than 80% of identity with the reference sequence, while BLRn misses these small copies. This increase in sensitivity comes with a cost, as RM predicts many spurious hits for TE families with simple repeats in their reference sequence. Overall, we found that the differences between BLRn and RM make them very complementary for TE annotation when hits from both methods are chained with MATCHER, and that a simple-repeat-filtered version can be used to promote reliable TE models automatically.
There are many reasons why BLRn and RM perform differently. One obvious reason is that the initial word length used to seed the alignments is shorter for RM than for BLRn (nine for Cross_match versus 11 for BLASTN). Another reason is that RM chooses its scoring scheme (a match–mismatch matrix) according to the background percent guanine/cytosine composition. A third explanation could come from the final Smith–Waterman alignment performed by RM, allowing it to produce longer alignments in low identity regions. Likewise, in some particularly difficult cases where a genomic TE copy has a duplicated segment, BLRn gives a better annotation because it relies only on BLASTN hits that allow a small level of overlap between adjacent hits. The final Smith–Waterman alignment performed by RM is disturbed in these cases, at best placing a gap to face the duplicated segment. The first two reasons are a matter of parameter values, and the differences may simply be due to our use of default parameters. The more sensitive parameter set of RM has a cost in term of speed, and the trade-off between speed and sensitivity between BLRn and RM is not the same (BLRn is at least 3-fold faster). Using different parameter values could improve either BLRn sensitivity and/or RM speed. It remains to be determined to what degree the sensitivity of BLRn can be improved to a level equivalent to RM just by changing the BLASTN parameters, since the use of different match–mismatch matrices (each optimal for a background percent guanine/cytosine level) is an important difference between the two methods, and may limit BLRn sensitivity gains.
Pitfalls
From our manual edits, we were able to identify some pitfalls that could be avoided in future attempts at a fully automated TE annotation process. One of the most important problems arises from the annotation of symmetrical structures, such as terminal inverted repeats (TIRs) and long terminal repeats (LTRs). There may be palindromic structures, such as in the FB element [26]. Often the two TIRs of a genomic FB element are detected on different strands, i.e., the 5′ TIR on the positive and the 3′ TIR on the negative strand. This happens because the two TIRs are not identical in the reference sequence. Thus, if the two TIRs of the genomic copy are more similar to each other than to the appropriate TIR in the reference sequence, only one TIR of the reference (the most similar one) is used to detect the two genomic TIRs, but on different strands. To avoid this type of manual edit, we suggest using a reference sequence with identical TIRs. A similar pitfall occurs with LTR retrotransposons. If the two LTRs are not identical on the reference, a genomic copy can be detected with two 5′ LTRs (or 3′ LTRs) if its LTRs are more similar to each other than to the appropriate LTRs of the reference sequence. If a join is necessary because of an indel in the genomic copy, our algorithm fails since the coordinates on the reference sequence are not collinear. To avoid this, we suggest using reference sequences with identical LTRs.
Some non-LTR retrotransposon genomic copies have to be extended in 3′ direction to encompass the entire polyadenine (poly[A]) tail. This occurs because the reference sequence has a shorter poly(A) tail than a particular genomic copy. In general, these cases are easily identified by observing an overlapping poly(A) simple repeat at the 3′ end of the element. One solution to this problem is to extend the poly(A) tail of non-LTR retrotransposons in the reference set to the length of the longest observed genomic copy.
The biggest pitfall we have encountered is the problem posed by simple repeats that exist in TE reference sequences. Without a specific treatment of this problem we would have included 3,040 spurious hits—approximately one-third of our original set of annotations. Filtering simple repeats on the genomic or reference sequences without affecting the sensitivity of TE detection is not easy. We have developed an effective (but ad hoc) two-step filtering strategy, but the magnitude of this problem leaves room for future improvements. Currently we employ RM to detect simple repeats, although refined parameter optimization may reveal that other more specialized simple repeat detection software, such as TRF [21], Mreps [22], or other methods (e.g., [27]) might be more appropriate. A careful evaluation of methods and parameters for simple repeat detection may allow us to decrease our 170-bp threshold and avoid the rescue step.
Regardless of the best method or criteria to detect simple repeats, the existence of simple repeats in TE reference sequences raises an important problem, since it is difficult to unambiguously determine whether a simple repeat with homology to a TE is a spurious hit or reflects a true remnant of that TE in the genome. Our methods guarantee that if we leave a spurious hit in the annotation because of homology with a simple repeat, it is more than 170 bp long. Moreover, any potentially real TE labeled as spurious that did not survive our rescue strategy bears no unique hallmarks of being generated by a TE. Nevertheless, the possibility of the involvement of TEs in the genesis of microsatellites [28] highlights the fundamental biological difficulty in resolving real from spurious simple repeats in a whole-genome TE annotation.
Conclusions and Future Directions
We have shown in this work that a combined-evidence framework can improve the quality and confidence of TE annotations in D. melanogaster. Our automated pipeline allows us to annotate TEs on a genomic scale quickly and accurately, and the integration of our pipeline with the Apollo annotation tool also allows rapid evaluation and manual editing of TE annotations for even complex TE models. Based on the lessons learned in this study, we are continuing to develop and improve our pipeline. We are automating several classes of the manual edits that we have identified and expect that progressively fewer manual edits will be necessary in the future, allowing application of our pipeline to larger genome sequences such as the human sequence. One possible solution to the simple repeat problem is to develop a “combined sensor” model that would seek to resolve competing signals between simple repeats and TE models. It may also be possible to predict nested elements that require manual edits by using a stochastic context-free grammar [29] approach to model the different components of TE nests more generally; stochastic context-free grammars may also be useful in resolving problems encountered in annotating TEs with terminal repeats. The annotations presented here could be used as a training set to estimate the utility of these types of models.
We have observed several cases in the genome annotation where one or more de novo methods (RECON, BLRa, BLRtx, and TE-HMM) simultaneously support a potential sequence belonging to a new TE family. In addition, results of our analyses with tools that detect anonymous TEs (see Table 2) suggest that there may be many additional families of TEs yet to be discovered in the D. melanogaster genome. Since the methods that support these predictions potentially suffer from a high false positive rate, we have chosen not to include them in our current annotation, since more work needs to be done to validate these potential new TE families. Nevertheless the combined evidence for some of these elements is compelling and such cases are available for mining in our current results.
In general, the problem of TE discovery remains a major challenge for TE annotation. A good TE annotation relies critically on an expertly assembled reference sequence set, data that currently cannot be obtained in an automatic fashion. This crucial step is now the bottleneck in any method or pipeline to annotate TEs in genome sequences (see also [3]). The task to assemble such reference sets will be most difficult in genomes where only a few TE families are known. In these situations, we will need good de novo TE detection procedures [10–16] that can only be trained and evaluated properly using high-quality TE annotations in well-studied systems such as Drosophila. We hope that the TE annotations presented here will serve to further the development and refinement of TE discovery and annotation methods in general, as the Release 3.1 annotations have served for the development of our current methods.
Finally, we are also developing our pipeline to include methods for the detailed annotation of the structural features (open reading frames, LTRs, etc.) in TE sequences. Development of such detailed annotation methodologies will allow a detailed evaluation of the coding and expression potential of individual TE annotations in genomic sequences. Moreover, the ability to automatically annotate structural features of TEs will facilitate the manual curation and validation of candidate TE sequences resulting from one or several different de novo TE discovery methods [10–16]. Continued development of this pipeline, together with other advances in the field of TE genome informatics, will lead to a robust computational framework that can shed light on the origin and impact of TEs in modern genomes.
Materials and Methods
Data.
The D. melanogaster genomic sequences and TE reference sets are available from BDGP (http://www.fruitfly.org/). The Release 3.1 D. melanogaster genomic sequences and their TE annotations have been extracted from the GAME-XML files. The Release 4 D. melanogaster genomic sequences have been downloaded as fasta files. TE reference sequence sets v.7.1 (used by Kaminker et al. [18]) and v.9.0 have been downloaded from BDGP.
Sequences of the TEs were also obtained from the Repbase Update database release 8.12 [30], which contains all known repeated sequences including TEs (downloaded from http://www.girinst.org). We used them to detect unknown families by similarity with TEs from other species.
Sequence analysis software.
We have improved three C++ programs: BLASTER, MATCHER, and GROUPER, previously presented in Quesneville et al. [13]. BLASTER can compare two sets of sequences: a query databank against a subject databank. For each sequence in the query databank, BLASTER launches one of the BLAST programs (BLASTN, TBLASTN, BLASTX, TBLASTX, BLASTP, or MegaBLAST) [17,31–33] to search the subject databank. Each BLAST search is launched in parallel on a computer cluster. BLASTER is not limited by the length of sequences. It cuts long sequences before launching BLAST and reassembles the results afterwards. Hence, it can work on whole genomes, in particular, to compare a genome with itself to detect repeats. The results of BLASTER can then be treated by the MATCHER and GROUPER programs described below. For the experiments conducted here, NCBI-BLAST2 (ftp://ftp.ncbi.nlm.nih.gov/blast/) programs were used with default parameters, using as a query genomic fragments of 50 kb, overlapping by 100 bp.
MATCHER has been developed to map match results onto query sequences by first filtering overlapping hits. When two matches overlap on the genomic (query) sequence, the one with the best alignment score is kept; the other is truncated so that only nonoverlapping regions remain on the match. As a result of this procedure a match is totally removed only if it is included in a longer one with a best score. All matches that have E-value greater than 1 × 10−10 or length of 20 or less are eliminated.
Long insertions (or deletions) in the query or subject could result in two matches, instead of one with a long gap. Thus, the remaining matches are chained by dynamic programming. A score is calculated by summing match scores and subtracting a gap penalty (0.05 times the gap length) and a mismatch penalty (0.2 times the mismatch length region) as in [34].
The chaining algorithm ([35], pp. 325–329) is modified to produce local alignments. A match is chained with a chain of other matches only if the resulting score is greater than the score of the match alone. Thus, the chaining is stopped if the score of the resulting chain of matches is less than if the match is not chained. The best-scoring chain is kept. Then to identify other match chains, the chain previously found is removed, and we search again for the next best match chain. This is done iteratively until no chain is found. This algorithm is repeated independently for match on strand +/+, +/−, and −/+. A maximum of 20% of overlap between the matches is allowed. The chaining algorithm allows the recovery of TEs containing long insertions, and therefore can identify nested elements accurately: they appear as a long insertion inside another TE.
GROUPER uses matches (or chained matches) to gather similar sequences into groups by simple link clustering. A match belongs to a group if one of the two matching sequence coordinates overlaps a sequence coordinate of this group by more than a given length coverage percentage threshold (a program parameter). If the two matches overlap with this constraint, their coordinates are merged, taking the extremum of the both. Groups that share sequence locations—not previously grouped because of a too low length coverage percentage—are regrouped into what we call a cluster. As a result of these procedures, each group contains sequences that are homogeneous in length. A given region may belong to several groups, but all of these groups belong to the same cluster.
RepeatMasker (http://www.repeatmasker.org) screens for TEs and low-complexity DNA sequences. It detects TEs in nucleic acid sequences by nucleic sequence alignment with previously characterized elements using the program Cross_match (http://www.phrap.org/phredphrapconsed.html) or WU-BLAST (http://blast.wustl.edu) with the script MaskerAid [36]. Both alignment programs perform their Smith–Waterman alignments by first identifying exact word matches and restricting the alignment to a band or matrix surrounding this exact match or matches. According to the background percent guanine/cytosine composition, different similarity matrices (each optimal for a background percent guanine/cytosine level) are used. RepeatMasker annotates the parts of sequences that are very similar to an element from a reference set of “known elements”. Low-complexity DNA regions are detected when stretches of nucleotides are GC- or AT-rich. Simple repeats are detected by searching all di- to pentameric and some hexameric repeats, allowing for possible variation within repeats.
RECON [10] is an automated process for de novo identification of new repeat sequence families in sequenced genomes. It searches genomic sequences for long repeats and clusters them in groups of similar sequences. TE copies from a given family are expected to cluster together. Its algorithm clusters repeats obtained by an all-by-all sequence comparison (here using BLASTER with BLASTN) and redefines the clusters by the aggregation of endpoints in a multiple alignment of the identified regions. In this way it tends to distinguish true TE copies from copies in a segmental duplication.
We have shown previously how base compositional differences can be used as a tool for detection and analysis of novel TE sequences [14]. Hidden Markov models are used to take into account the base composition of the sequences and the heterogeneity between coding and noncoding parts of sequences. We use three sets of sequences from D. melanogaster containing class I TEs, class II TEs, and cellular genes. Each of these sets has a distinct, homogeneous composition, enabling us to distinguish between the two classes of TEs and the genes. This approach can be used to detect and annotate TEs in genomic sequences and complements the current homology-based TE detection methods. Furthermore, the hidden Markov model method is able to identify the parts of a sequence in which the nucleotide composition resembles that of a coding region of a TE. This is useful for the detailed annotation of TE sequences, which may contain an ancient, highly diverged coding region that is no longer fully functional.
Comparison of predictions and annotations.
We automatically compared predictions obtained with different computational methods to the Release 3.1 TE reference annotations in two ways, each implemented in a custom Python script.
The first calculated the nucleotide overlaps between the predictions and reference annotations, and computed the genome-wide sensitivity and the specificity. These values were obtained from equations (1) and (2) and the counts of true positive (TP—correctly annotated as belonging to a TE), false positive (FP—falsely predicted as belonging to a TE), true negative (TN—correctly annotated as not belonging to a TE), and false negative (FN—falsely predicted as not belonging to a TE) nucleotides.
A high sensitivity indicates that a method misses few TE nucleotides (few false negatives). A high specificity indicates that a method finds few false positive nucleotides.
The second Python script compared the boundaries of predictions to the boundaries of the reference annotations. For each prediction under test, we searched the reference annotations that overlapped on the same genomic region. Different cases could be distinguished according to one-to-one, one-to-many, many-to-one, or many-to-many relationships (see Figure 3 for details).
Figure 3 Categories of Possible Boundary Comparisons between Predictions and Reference Annotations
The different cases taken into account can be grouped according to one-to-one (1-to-1), one-to-many (1-to-n), many-to-one (n-to-1), or many-to-many (n-to-n) relationships.
For those that had a one-to-one correspondence with the same TE family, we calculated the difference in distances between predictions and annotations for their respective 5′ and 3′ coordinates. We categorized the differences in distance into three classes: ≤1 bp, ≤10 bp, or >10 bp. We called “exact” annotations those that had distances at both extremities ≤ 1 bp, “near exact” those for which the distance at one extremity was ≤ 1 bp and that of the other was >1 bp and ≤10 bp, and “one side exact” those for which one extremity was ≤1 bp and the other was >10 bp. Cases where both distances were > 1 bp and ≤ 10 bp were called “equivalent”; if one distance was > 1 bp and ≤ 10 bp and the other was >10 bp, the case was “near equivalent”; and if both distances were > 10 bp, the case was “similar”.
We also considered many-to-one relationships. Some were method errors in which a genomic copy (given by the reference annotation) had a large insertion or deletion. In this case, the two fragments (flanking the indel) were predicted as two separate copies, and the fragments were not joined. We called this error class “method not joined”. We also found cases in which two predictions were falsely considered as one in the reference annotation. Here, a long region of mismatch separated two fragments and the most parsimonious explanation was the independent insertion of two copies. These were “annotation over-joined” cases. We also found cases considered as one copy by the reference annotation, but that were in fact copies with a self-duplicated region. If the duplication was nested we call it “same TE nested”, or if not nested, “TE duplication”.
One-to-many relationships were cases in which two annotations in the reference were found joined by the method. We called this “annotation not joined”.
One-to-zero relationships corresponded to cases in which a prediction did not correspond to a reference annotation. “New TE” cases were copies identified by the method under test but not present in the reference annotation, and “different TE” cases were those overlapping a reference annotation but with a different TE family name. A TE prediction included in a prediction of a different family already involved in a given relationship with reference annotations, was called “new nest” if no corresponding reference annotation could be found. Annotation correspondence of the same TE family but on different strand was called “other strand” if the relationship was one-to-one; otherwise they were “new TE”.
Finally we had a “complex structure” case when the relation was many-to-many.
The script could be also used in an anonymous mode to test boundaries of de novo predictions that do not use a specific reference sequence. The information used for such comparisons is of poorer quality since we do not have alignment coordinates on the reference sequence (i.e., RECON and TE-HMM), which renders several categories meaningless (e.g., “different TE”, but also “new nest”, “other strand”, and “TE duplication”).
We thank Clémentine Vitte for help testing of BLASTER, Chris Mungall for early comparative analyses on Release 3, Emmanuel Mongin for the initial Release 4 mapping of Release 3.1 TE annotations, and Dave Emmert for assistance loading our Release 4 TE annotations into FlyBase. We thank Sima Misra, Chelsea Scholl, and three anonymous reviewers for helpful suggestions on the manuscript. This work was supported by the Centre National de Recherche Scientifique (CNRS), the Universities P. and M. Curie and D. Diderot (Institut Jacques Monod, UMR 7592, Dynamique du Génome et Evolution) and by the Programme Bio-Informatique (CNRS). CMB was supported by a USA Research Fellowship from the Royal Society. Work in MA's laboratory was supported by an MRC Programme Grant to MA and S. Russell.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. H. Quesneville, C. M. Bergman, M. Ashburner, and D. Anxolabehere conceived and designed the experiments. H. Quesneville, C. M. Bergman, and O. Andrieu performed the experiments. H. Quesneville, C. M. Bergman, O. Andrieu, D. Autard, and D. Nouaud analyzed the data. H. Quesneville, C. M. Bergman, O. Andrieu, D. Autard, and M. Ashburner contributed reagents/materials/analysis tools. H. Quesneville and C. M. Bergman wrote the paper.
Abbreviations
BDGPBerkeley Drosophila Genome Project
BLRaall-by-all genome comparison with BLASTER using BLASTN followed by chaining with MATCHER and grouping with GROUPER
BLRnBLASTER using BLASTN followed by chaining with MATCHER
BLRtxBLASTER using TBLASTX with the entire Repbase Update as the database, followed by chaining with MATCHER
BLRtxNoDrosBLASTER using TBLASTX with the Repbase Update with Drosophila TEs removed as the database, followed by chaining with MATCHER
LTRlong terminal repeat
poly(A)polyadenine
RMRepeatMasker using default parameters
RMBLRRepeatMasker-BLASTER
RMmRepeatMasker using default parameters followed by chaining with MATCHER
TEtransposable element
TE-HMMhidden Markov Model that detects TE coding sequences based on nucleotide composition
TIRterminal inverted repeat
TRFTandem Repeat Finder
==== Refs
References
Lander ES Linton LM Birren B Nusbaum C Zody MC 2001 Initial sequencing and analysis of the human genome Nature 409 860 921 11237011
Kidwell MG Lisch DR 2001 Perspective: Transposable elements, parasitic DNA, and genome evolution Evolution Int J Org Evolution 55 1 24
Juretic N Bureau TE Bruskiewich RM 2004 Transposable element annotation of the rice genome Bioinformatics 20 155 160 14734305
Misra S Crosby MA Mungall CJ Matthews BB Campbell KS 2002 Annotation of the Drosophila melanogaster euchromatic genome: A systematic review Genome Biol 3 RESEARCH0083 12537572
Mungall CJ Misra S Berman BP Carlson J Frise E 2002 An integrated computational pipeline and database to support whole-genome sequence annotation Genome Biol 3 RESEARCH0081 12537570
Allen JE Pertea M Salzberg SL 2004 Computational gene prediction using multiple sources of evidence Genome Res 14 142 148 14707176
Ding L Sabo A Berkowicz N Meyer RR Shotland Y 2004 EAnnot: A genome annotation tool using experimental evidence Genome Res 14 2503 2509 15574829
Ashurst JL Chen CK Gilbert JG Jekosch K Keenan S 2005 The Vertebrate Genome Annotation (Vega) database Nucleic Acids Res 33 D459 D465 15608237
Haas BJ Wortman JR Ronning CM Hannick LI Smith RK Jr 2005 Complete reannotation of the Arabidopsis genome: Methods, tools, protocols and the final release BMC Biol 3 7 15784138
Bao Z Eddy SR 2002 Automated de novo identification of repeat sequence families in sequenced genomes Genome Res 12 1269 1276 12176934
Biedler J Tu Z 2003 Non-LTR retrotransposons in the African malaria mosquito, Anopheles gambiae: Unprecedented diversity and evidence of recent activity Mol Biol Evol 20 1811 1825 12832632
McCarthy EM McDonald JF 2003 LTR_STRUC: A novel search and identification program for LTR retrotransposons Bioinformatics 19 362 367 12584121
Quesneville H Nouaud D Anxolabehere D 2003 Detection of new transposable element families in Drosophila melanogaster and Anopheles gambiae genomes J Mol Evol 57 S50 S59 15008403
Andrieu O Fiston AS Anxolabehere D Quesneville H 2004 Detection of transposable elements by their compositional bias BMC Bioinformatics 5 94 15251040
Edgar RC Myers EW 2005 PILER: Identification and classification of genomic repeats Bioinformatics 21 I152 I158 15961452
Price AL Jones NC Pevzner PA 2005 De novo identification of repeat families in large genomes Bioinformatics 21 I351 I358 15961478
Altschul SF Gish W Miller W Myers EW Lipman DJ 1990 Basic local alignment search tool J Mol Biol 215 403 410 2231712
Kaminker JS Bergman CM Kronmiller B Carlson J Svirskas R 2002 The transposable elements of the Drosophila melanogaster euchromatin: A genomics perspective Genome Biol 3 RESEARCH0084 12537573
Lewis SE Searle SM Harris N Gibson M Lyer V 2002 Apollo: A sequence annotation editor Genome Biol 3 RESEARCH0082 12537571
Celniker SE Wheeler DA Kronmiller B Carlson JW Halpern A 2002 Finishing a whole genome shotgun sequence assembly: Release 3 of the Drosophila euchromatic genome sequence Genome Biol 3 RESEARCH0079 12537568
Benson G 1999 Tandem repeats finder: A program to analyze DNA sequences Nucleic Acids Res 27 573 580 9862982
Kolpakov R Bana G Kucherov G 2003 mreps: Efficient and flexible detection of tandem repeats in DNA Nucleic Acids Res 31 3672 3678 12824391
Locke J Howard LT Aippersbach N Podemski L Hodgetts RB 1999 The characterization of DINE-1, a short, interspersed repetitive element present on chromosome and in the centric heterochromatin of Drosophila melanogaster
Chromosoma 108 356 366 10591995
Meyerowitz EM Hogness DS 1982 Molecular organization of a Drosophila puff site that responds to ecdysone Cell 28 165 176 6279311
Hoskins RA Smith CD Carlson JW Carvalho AB Halpern A 2002 Heterochromatic sequences in a Drosophila whole-genome shotgun assembly Genome Biol 3 RESEARCH0085 12537574
Potter S Truett M Phillips M Maher A 1980 Eucaryotic transposable genetic elements with inverted terminal repeats Cell 20 639 647 6251970
Sagot MF Myers EW 1998 Identifying satellites and periodic repetitions in biological sequences J Comput Biol 5 539 553 9773349
Wilder J Hollocher H 2001 Mobile elements and the genesis of microsatellites in dipterans Mol Biol Evol 18 384 392 11230539
Durbin R Eddy SR Krogh A Mitchison G 1999 Biological sequence analysis: Probabilistic models of proteins and nucleic acids Cambridge Cambridge University Press 368 p.
Jurka J 2000 Repbase update: A database and an electronic journal of repetitive elements Trends Genet 16 418 420 10973072
Gish W States DJ 1993 Identification of protein coding regions by database similarity search Nat Genet 3 266 272 8485583
Altschul SF Madden TL Schaffer 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
Zhang Z Schwartz S Wagner L Miller W 2000 A greedy algorithm for aligning DNA sequences J Comput Biol 7 203 214 10890397
Chao KM Zhang J Ostell J Miller W 1995 A local alignment tool for very long DNA sequences Comput Appl Biosci 11 147 153 7620986
Gusfield D 1997 Algorithms on strings, trees, and sequences: Computer science and computational biology New York Cambridge University Press 534 p.
Bedell JA Korf I Gish W 2000 MaskerAid: A performance enhancement to RepeatMasker Bioinformatics 16 1040 1041 11159316
|
16110336
|
PMC1185648
|
CC BY
|
2021-01-05 09:18:23
|
no
|
PLoS Comput Biol. 2005 Jul 29; 1(2):e22
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010022
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 10.1371/journal.pcbi.001002405-PLCB-RV-0069R2plcb-01-02-08ReviewBioinformatics - Computational BiologyEubacteriaBioinformatics for Whole-Genome Shotgun Sequencing of Microbial Communities ReviewChen Kevin *Pachter Lior *Kevin Chen is in the Department of Electrical Engineering and Computer Science and Lior Pachter is in the Department of Mathematics at the University of California, Berkeley, California, United States of America.
*To whom correspondence should be addressed. E-mail: [email protected] (KC), [email protected] (LP)7 2005 12 7 2005 1 2 e24Copyright: © 2005 Chen and Pachter.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 application of whole-genome shotgun sequencing to microbial communities represents a major development in metagenomics, the study of uncultured microbes via the tools of modern genomic analysis. In the past year, whole-genome shotgun sequencing projects of prokaryotic communities from an acid mine biofilm, the Sargasso Sea, Minnesota farm soil, three deep-sea whale falls, and deep-sea sediments have been reported, adding to previously published work on viral communities from marine and fecal samples. The interpretation of this new kind of data poses a wide variety of exciting and difficult bioinformatics problems. The aim of this review is to introduce the bioinformatics community to this emerging field by surveying existing techniques and promising new approaches for several of the most interesting of these computational problems.
Citation:Chen K, Pachter L (2005) Bioinformatics for whole-genome shotgun sequencing of microbial communities. PLoS Comp Biol 1(2): e24.
==== Body
Introduction
Metagenomics is the application of modern genomics techniques to the study of communities of microbial organisms directly in their natural environments, bypassing the need for isolation and lab cultivation of individual species [1–6]. The field has its roots in the culture-independent retrieval of 16S rRNA genes, pioneered by Pace and colleagues two decades ago [7]. Since then, metagenomics has revolutionized microbiology by shifting focus away from clonal isolates towards the estimated 99% of microbial species that cannot currently be cultivated [8,9].
A typical metagenomics project begins with the construction of a clone library from DNA sequence retrieved from an environmental sample. Clones are then selected for sequencing using either functional or sequence-based screens. In the functional approach, genes retrieved from the environment are heterologously expressed in a host, such as Escherichia coli, and sophisticated functional screens employed to detect clones expressing functions of interest [10–12]. This approach has produced many exciting discoveries and spawned several companies aiming to retrieve marketable natural products from the environment (e.g., Diversa [http://www.diversa.com] and Cubist Pharmaceuticals [http://www.cubist.com]). In the sequence-based approach, clones are selected for sequencing based on the presence of either phylogenetically informative genes, such as 16S, or other genes of biological interest [13–17]. The most prominent discovery from this approach thus far is the discovery of the proteorhodopsin gene from a marine community [14].
Recently, facilitated by the increasing capacity of sequencing centers, whole-genome shotgun (WGS) sequencing of the entire clone library has emerged as a third approach to metagenomics. Unlike previous approaches, which typically study a single gene or individual genomes, this approach offers a more global view of the community, allowing us to better assess levels of phylogenetic diversity and intraspecies polymorphism, study the full gene complement and metabolic pathways in the community, and in some cases, reconstruct near-complete genome sequences. WGS also has the potential to discover new genes that are too diverged from currently known genes to be amplified with PCR, or heterologously expressed in common hosts, and is especially important in the case of viral communities because of the lack of a universal gene analogous to 16S.
Nine shotgun sequencing projects of various communities have been completed to date (Table 1). The biological insights from these studies have been well-reviewed elsewhere [3,6]. Here, we highlight just two studies that exemplify the exciting possibilities of the approach. The acid mine biofilm community [18] is an extremely simple model system, consisting of only four dominant species, so a relatively miniscule amount of shotgun sequencing (75 Mbp) was enough to produce two near-complete genome sequences and detailed information about metabolic pathways and strain-level polymorphism. At the other end of the spectrum, the Sargasso Sea community is extremely complex, containing more than 1,800 species [19,20]. Nonetheless, with an enormous amount of sequencing (1.6 Gbp), vast amounts of previously unknown diversity were discovered, including over 1.2 million new genes, 148 new species, and numerous new rhodopsin genes. These results were especially surprising given how well the community had been studied previously, and suggest that equally large amounts of biological diversity await future discovery.
Table 1 Published Microbial Community Shotgun Sequencing Projects
aThe deep-sea sediment project used an additional 20 Mbp of fosmid sequence and also a filter to reduce the complexity of the community prior to sequencing.
bThe viral projects used linker-amplified shotgun libraries.
In this review, we survey several of the most interesting computational problems that arise from WGS sequencing of communities. Traditional approaches to classic bioinformatics problems such as assembly, gene finding, and phylogeny need to be reconsidered in light of this new kind of data, while new problems need to be addressed, including how to compare communities, how to separate sequence from different organisms in silico, and how to model population structures using WGS assembly statistics. We discuss all these problems and their connections to other areas of bioinformatics, such as the assembly of highly polymorphic genomes, gene expression analysis, and supertree methods for phylogenetic reconstruction.
Although we have chosen to focus on the shotgun sequencing approach, we stress that this is only one piece of the exciting field of metagenomics, and that the integration of other techniques such as large-insert clone sequencing, microarray analysis, and proteomics will be vital to achieve a comprehensive view of microbial communities.
Assembling Communities
The retrieval of nearly complete genomes from the environment without prior lab cultivation is one of the most spectacular results of metagenomics to date. A fundamental limit on the WGS approach is that we can only expect to assemble genomes that constitute a significant fraction of the community [21]. Filtration and normalization techniques that enrich the library for certain low-abundance species, a common technique in the sequencing of symbionts, are thus of vital importance when genome assembly is a primary goal [22,23].
When a closely related, fully sequenced genome is available, comparative assembly can easily be performed by extracting the homologous sequence and assembling it with either a comparative assembler [24] or an alignment program that can handle draft sequence [25,26]. This approach is standard and has been used many times for mixed sequence from multiple species ([19,27]; E. Allen, unpublished data).
In the absence of an appropriate template genome, traditional overlap–layout–consensus assembly [28] can be done, augmented by an additional binning step, in which scaffolds (contiguous sequence with gaps of approximately known size) are separated into species-specific “bins.” The first issue that needs to be overcome is the increased amount of polymorphism, since each read will typically be sampled from a different individual in the population. Second, highly conserved sequence shared between different species can seed contigs and cause false overlaps. In some communities, even phylogenetically distant genomes can share a large number of genes [29]. Careful study of the optimal overlap parameters for separating out sequences at different phylogenetic distances is important, and has been carried out for viral communities [30], but not yet for prokaryotes.
The assembly of communities has strong similarities to the assembly of highly polymorphic diploid eukaryotes, such as Ciona savigny [26] and Candida albicans [31], if we view prokaryotic strains as analogous to eukaryotic haplotypes. The main difference is that in a microbial community, the number of strains is unknown and potentially large, and their relative abundance is also unknown and potentially skewed, while in most eukaryotes we know a priori the number of haplotypes and their relative abundance. This disadvantage is mitigated somewhat by the small size and relative lack of repetitive sequence in prokaryotic and viral genomes, so that the issue of distinguishing alleles from paralogs and polymorphism from repetitive sequence is less acute.
Thus far, both community assembly and polymorphic eukaryotic assembly have been performed by running a single-genome assembler, such as the Celera assembler [32] or Jazz [33], and then manually post-processing the resulting scaffolds to correct assembly errors. Contigs erroneously split apart because of polymorphism are reconnected, and contigs based on false overlaps are broken apart. Not surprisingly, ad hoc heuristics must be employed to adapt programs optimized for single-genome assembly: the Celera assembler, for instance, treats high-depth contigs associated with abundant species as repetitive sequence.
A promising direction for both these problems is co-assembly, in which two very closely related genomes (or even two assemblies of the same genome) are assembled concurrently, using alignment information to complement mate-pair information in ordering scaffolds and correcting assembly errors in a structured, automated way. Thus far, the only published work on this problem is that of Sundararajan et al. [26], and even then, only for two genomes. For three or more genomes, even the multiple alignment problem for draft sequence is not solved. Large-insert clone sequence will also be very useful since the entire clone comes from a single strain or haplotype [22,34].
After scaffolds have been constructed, the next step is to bin the scaffolds according to species or phylogenetic clade. The gold standard for binning is the presence of a phylogenetically informative gene. 16S rRNA, though universal, is decidedly not single copy, so it is important to also consider other genes, such as RecA, EFG, EFTu, and HSP70 [19]. In the absence of one of these genes, genome signatures such as dinucleotide frequencies, codon bias, and GC-content, developed by Karlin and others in a long series of papers [35–38], can be used. These signatures appear to work for scaffolds on the order of 50 kbp in length, and, importantly, they seem to correlate only with phylogenetic relatedness and not with the environment [36]. There is a web server, Tetra, that computes tetranucleotide frequencies for metagenomics projects [39,40].
An additional source of evidence unique to WGS data is scaffold read depth, which is expected to be proportional to species abundance and thus can be used to separate high-abundance from low-abundance species. Subtleties can arise, however, since a variable polymorphism rate across a genome can cause conserved regions to be covered at high depth and variable regions to be covered at low depth.
For some applications, completely accurate binning may not be required. For example, gene finders based on hidden Markov models (HMMs) require training data from closely related species. The accuracy of the gene finder might be improved by additional training data, even if it is not from exactly the same species. One could even imagine running the following iterative algorithm: find a set of putative genes, construct gene trees with them, use the trees to crudely bin the scaffolds, retrain the gene finder, and repeat.
To conclude our discussion of assembly, we consider the important question of determining how much to sequence in order to assemble genomes. When sequencing a single genome, the Lander–Waterman model based on the assumptions of independent and random reads implies that the coverage of each base is distributed according to a Poisson distribution with parameter c (the coverage). Defining nk to be the number of bases covered exactly k times and G to be the genome size, we have
First consider the problem of assembling the most abundant genome at, say, 8× coverage. In the worst case, all species are present in equal abundance. The Lander–Waterman equation holds with G replaced by the sum of the sizes of all genomes of species in the community (sometimes called the metagenome). For the soil community, we have n
2 = 300,000 and G = 108/c, so the equation implies a coverage of 0.006 and a total of 133 Gbp of sequence needed to assemble the most abundant genome at 8× coverage, disregarding the problem of binning. The total metagenome size predicted is G = 16.7 Gbp, corresponding to 2,800 E. coli–sized genomes, which is consistent with previous estimates of soil microbial diversity and the 16S survey.
For the lower bound, we make the additional assumptions that all genomes have length 6 Mbp and that a single dominant species contributes all the overlaps in the assembly. The same equation implies that 2 Gbp of additional sequence is required for assembly at 8× coverage. This number is about twice that calculated from the 16S survey, but this might be explained by preferential amplification bias in PCR.
We performed similar calculations for the three whale fall communities. In addition, we considered the problem of assembling all genomes in these communities. Since the 16S survey indicated that three dominant species constitute approximately half the total abundance and all other species have roughly equal abundance, the Lander–Waterman model implies that the expected coverage should be distributed as the mixture of two Poissons with equal weight. The results of these calculations are summarized in Table 2. Similar results were obtained by Venter et al. [19] and Breitbart et al. [30], and there is also software for performing such calculations (http://phage.sdsu.edu/phaccs) [41].
Table 2 Bounds on Amount of Sequence Needed to Assemble Genomes (in Mbp)
Comparative Metagenomics
Gene finding is a fundamental goal in virtually all metagenomics projects, regardless of whether complete genome sequences can be assembled or not. If large scaffolds can be retrieved and binned, excellent HMM-based microbial gene finders such as FGENESB (http://www.softberry.com) and GLIMMER [42,43] can be used, in combination with expectation-maximization (EM) techniques for unsupervised training of the HMM parameters [44,45]. At the other extreme, we have unassembled reads of roughly 700 bp. These make up 50% of the total reads in the Sargasso Sea dataset and 100% in soil. Since prokaryotic genes are typically short, lack introns, and occur at high density (roughly one in 1,000 bp), each read is likely to contain a significant portion of a gene. For these reads, HMM techniques are unlikely to be successful, leaving BLAST search against a protein database or the community itself as the only realistic alternative.
There have been two simulation studies verifying the accuracy of BLAST for gene finding with single reads [21,46], though it is difficult to make this kind of experiment convincing, since the accuracy of the method is almost entirely dependent on the availability of closely related sequences in the database. We are not aware of any studies on the accuracy of HMM-based techniques on sequences significantly shorter than a whole genome, so we undertook a simple experiment ourselves. We sampled simulated “contigs” of length 10 kb from the complete genome sequence of Thermoplasma volcanium [47]. For each, we predicted genes using GLIMMER trained only on long open reading frames in the contig, and compared these to the GLIMMER predictions when trained on long open reading frames from the entire genome. We found that the results were surprisingly good. Of 92 genes completely contained in the ten simulated contigs, 86 were predicted exactly correctly. There were 16 genes that crossed the boundaries of the contigs, and GLIMMER was able to find truncated genes for seven of these. On the other hand, five of the completely spurious predictions all came from the same contig, which suggests that HMM accuracy may not be uniform over the length of the genome. More detailed studies on this problem are needed to relate the length of assembled contigs to the accuracy of the gene finder. An interesting direction is to attempt to recover more partial genes that overlap contig boundaries, firstly, by making the gene finder aware that genes on the boundary may be truncated and, secondly, by taking advantage of base quality scores for lower quality sequence at the ends of contigs. Another interesting research problem is to fine-tune gene finders for viral genomes.
The gene complement of a microbial community can be used as a fingerprint of a community, allowing us to compare different communities in a gene-centric, as opposed to genome-centric, fashion [21]. In this method, predicted genes are blasted against the COGs [48] or KEGG [49,50] databases and each community is assigned a fingerprint vector with entries corresponding to the number of hits to each COGs or KEGG category. It is also possible to cluster the COGs hits by function in order to compare the communities at a higher level.
Fingerprint vectors are analogous to gene-expression-level vectors in microarray analysis and any of the standard gene expression clustering methods can be used [51]. We first replicated the result of [21] by directly applying popular the off-the-shelf gene expression tools, CLUSTER and TreeView [52], to perform single-linkage hierarchical clustering on the KEGG vectors from several communities (Figure 1).
Figure 1 Blue-Yellow Microarray Figure Applied to KEGG Vectors for Four Metagenomics Projects
The whale-fall and Sargasso sea data are partitioned into three different samples each. The rows correspond to the different datasets and the columns to the 137 KEGG categories. Blue corresponds to underrepresentation and yellow to overrepresentation. Note that some branch lengths have been adjusted for visualization purposes and do not correspond to an actual meaningful distance.
Although the neat tree structure of the blue-yellow microarray figure (Figure 1) looks appealing, it can also be misleading at times because of the properties of UPGMA (unweighted pair group method with arithmetic mean) clustering. To check this, we applied principle components analysis to the fingerprint vectors (Figure 2). While the high-level result is similar, the principle components analysis shows that the clustering of the communities is somewhat more ambiguous than Figure 1 might suggest. For instance, note the surprising proximity of whale-fall sample 1 to the soil sample.
Figure 2 Projection of the KEGG Vectors on the First Two Principle Components
In addition to clustering, principle components analysis has the additional advantage that dimensions of the principle components with high magnitude may correspond to COGs or KEGG sequences of interest, and the principle components themselves may correspond to interesting pathways or functions. This has not yet been fully explored and could potentially be a source of new functional pathways in communities.
Finally, since fingerprinting has been advocated as an alternative to genome assembly when the amount of sequence required for assembly is very high [21], an important issue that needs to be discussed is how much sequence is required to fingerprint. In the same spirit as our Lander–Waterman calculations (equation 1), we estimate this quantity using the observation that the number of genes per shotgun read is very close to one [21,46]. Assuming a uniform species abundance distribution, we get the classic coupon collector's problem [53], in which the number of reads needed to collect a fraction f of the N genes in the community is exactly
Applying equation 2 to the soil community, if we assume 4,000 genes per genome and 3,000 genomes, then sampling half the genes would require 6 Gbp of sequencing, comparable to the lower bound on the amount of sequence needed to assemble the dominant genome (Table 2).
Based on these observations, it seems that it may be too early to conclude that fingerprinting is a powerful way of comparing communities. We also note that fingerprinting is difficult for viruses, since 65% of predicted genes from the viral community sequencing projects have no homolog in the databases [6]. However, similar techniques have been used to compare the species, as opposed to their gene complements, across different viral communities [54].
Phylogeny and Community Diversity
If complete gene sequences can be recovered from the community, classic multiple sequence alignment (MSA) [55] and phylogeny algorithms [56] can be applied. If only partial genes are available, phylogenetic reconstruction is still reasonably straightforward if there is already a database of nearly complete sequences, as with 16S [57] or RecA (http://www.tigr.org/_jeisen/RecA/RecA.html). The partial sequences can then be aligned against the complete ones, and the phylogenetic assignment performed by finding the closest sequences in the database [58]. Even for such genes, however, it is plausible to imagine a future in which the majority of genes in the database are in fact partial environmental sequences—at one point, for instance, the Sargasso Sea dataset made up 5% of the total genes in GenBank and a large number of these were unassembled reads. Alternatively, metagenomics projects may discover a highly diverged group of species that may not align well to existing sequences. In these scenarios, it will be necessary to have good MSA and phylogeny tools for partial sequences, even for these “universal” genes.
The case of viral phylogeny is more complex, firstly, because it is not clear that all viruses are related by a tree, and, secondly, because viral taxonomy has traditionally not been based on molecular sequence data, though the Phage Proteomic Tree [59] represents a step in the direction of sequence-based taxonomy. Viral taxonomy is at a very early stage of development, and there is no doubt that culture-independent methods will play an important role in the growth of the field.
Partial sequences are the crux of the phylogeny problem in the context of metagenomics. We are particularly interested in methods for such sequences because they will also be applicable for low-coverage sequencing projects of vertebrates and other species [46,60]. We are not aware of any MSA tools and phylogeny programs that are able to cope with short partial gene fragments, any two of which may fail to have significant overlap. At the alignment stage, we require a semi-global multiple alignment (i.e., terminal gaps are not penalized). The most widely used alignment tools are based on global or local alignments and do not correctly handle partial sequences (an exception is MAP [61]). Since most MSA tools are based on progressive alignment according to a guide tree, it is also important to construct this tree based on pairwise semi-global alignments and conserved terminal k-mers, as opposed to the pairwise global or local alignments currently used.
We studied 40 phosphoglycerate kinase genes from the soil study and aligned them with MUSCLE [62]. Though not optimized for partial sequences, MUSCLE did a reasonable job, as ascertained by several criteria: the number of internal gaps was small, sequences shorter than the read length had either no beginning gaps or no ending gaps (since the gene length is greater than the read length), and the total length was comparable to related proteins.
Of the 780 pairs of sequences, 95 pairs had overlap of less than 50 amino acids, and of these, 48 pairs had no overlap at all. Thus, we have an extreme instance of the missing data problem, which has been extensively discussed in the phylogenetics literature [63,64]. However, this literature has mostly studied consensus tree methods, and the effect of adding incomplete taxa and/or characters on the accuracy of traditional methods, like maximum likelihood. Relatively little effort has gone into actually finding better methods for tree reconstruction with this kind of data. Supertree methods [65], which attempt to construct trees from multiple subtrees, present one such alternative. One reason these methods have not been widely used in the past in the context of molecular data is the relative lack of maturity of the field as compared with parsimony or likelihood methods. However, encouraging new algorithmic results and software in this area [66–68] should spur renewed work on these types of methods. Supertree methods have also been criticized because incomplete data matrices (e.g., from fossil data) usually do not fit a random and independent missing data model. On the other hand, shotgun sequencing does fit this model and thus would seem an ideal setting for supertree methods. While the data might be too limited to provide completely resolved phylogenies, as previous discussed in the context of binning, even crude trees may be sufficient for certain applications, such as training HMMs.
Finally, with regards to community diversity, one of the advantages of the WGS approach is that it is less biased then PCR, which is known to suffer from a host of problems [69]. Community modeling based on analysis of assembly data within the Lander–Waterman model is beginning to show that species abundance curves are not lognormal as previously thought [41,70], so new methods that take into account these naturally occurring distributions are needed.
Conclusion
The number of new community shotgun sequencing projects continues to grow, promising to provide vast quantities of sequence data for analysis. Samples are being drawn from macroscopic environments such as the sea and air, as well as from more contained communities such as the human mouth (Table 3). Exciting advances in our understanding of ecosystems, environments, and communities will require creative solutions to numerous new bioinformatics problems. We have briefly mentioned some of these: assembly (can co-assembly techniques be used to assemble polymorphic genomes and complex communities?), binning (what is the best way to combine diverse sources of information to bin scaffolds?), gene finding (how should gene finding programs, which were designed for complete genes and genomes, be adapted for low-coverage sequence?), fingerprinting (which clustering techniques are best suited for discovering novel pathways and functional groups that allow communities to adapt to their environments?), and MSA and phylogeny (how can we best construct trees and alignments from fragmented data?).
Table 3 Examples of Ongoing Community WGS Sequencing Projects
Countless more challenges will likely emerge as WGS sequencing approaches are used to tackle increasingly complex communities. The reward for computational biologists who work on these problems will be the satisfaction of contributing to the grand enterprise of understanding the total diversity of life on our planet.
We thank Eric Allen, Jill Banfield, Susannah Tringe, and Gene Tyson for introducing us to the field of metagenomics and for helpful discussions while preparing the manuscript. We also thank Richard Karp and Satish Rao for useful discussions on bioinformatics issues, and the anonymous reviewers for their comments on an earlier version of this paper. Some of the data we have used were provided by JGI and EMBL. KC was supported by National Science Foundation (NSF) grant EF 03–31494. LP was supported by a Sloan Research Fellowship, NSF grant CCF 03–47992, and National Institutes of Health grant R01-HG02362–03.
Abbreviations
HMMhidden Markov model
MSAmultiple sequence alignment
WGSwhole-genome shotgun
==== Refs
References
DeLong EF 2002 Microbial population genomics and ecology Curr Opin Microbiol 5 520 524 12354561
Handelsman J 2004 Metagenomics: Application of genomics to uncultured microorganisms Microbiol Mol Biol Rev 68 669 684 15590779
Riesenfeld CS Schloss P Handelsman J 2004 Metagenomics: Genomic analysis of microbial communities Annu Rev Genet 38 525 552 15568985
Rodriguez-Valera F 2004 Environmental genomics, the big picture? FEMS Microbiol Lett 231 153 158 15027428
Streit WR Schmitz RA 2004 Metagenomics—The key to the uncultured microbes Curr Opin Microbiol 7 492 498 15451504
Edwards RA Rohwer F 2005 Viral metagenomics Nat Rev Microbiol 3 504 510 15886693
Olsen GJ Lane DJ Giovannoni SJ Pace NR Stahl DA 1986 Microbial ecology and evolution: A ribosomal RNA approach Annu Rev Microbiol 40 337 365 2430518
Hugenholtz P 2002 Exploring prokaryotic diversity in the genomic era Genome Biol 3 REVIEWS0003 11864374
Rappe M Giovannoni S 2003 The uncultured microbial majority Annu Rev Microbiol 57 369 394 14527284
Courtois S Cappellano CM Ball M Francou F Normand P 2003 Recombinant environmental libraries provide access to microbial diversity for drug discovery from natural products Appl Environ Microbiol 69 49 55 12513976
Riesenfeld CS Goodman RM Handelsman J 2004 Uncultured soil bacteria are a reservoir of new antibiotic resistance genes Environ Microbiol 6 981 989 15305923
Uchiyama T Abe T Ikemura T Watanabe K 2005 Substrate-induced gene-expression screening of environmental metagenomic libraries for isolation of catabolic genes Nat Biotechnol 23 88 93 15608629
Stein JL March TL Wu KY Shizuya H DeLong EF 1996 Characterization of uncultivated prokaryotes: Isolation and analysis of a 40-kilobase-pair genome fragment from a planktonic marine archaeon J Bacteriol 178 591 599 8550487
Beja O Aravind L Koonin EV Suzuki MT Hadd A 2000 Bacterial rhodopsin: Evidence for a new type of phototrophy in the sea Science 289 1902 1906 10988064
Liles MR Manske BF Bintrim SB Handelsman J Goodman RM 2003 A census of rRNA genes and linked genomic sequences within a soil metagenomic library Appl Environ Microbiol 69 2684 2691 12732537
Beja O 2004 To BAC or not to BAC: Marine ecogenomics Curr Opin Biotechnol 15 187 190 15193325
Sabehi G Beja O Suzuki MT Preston CM DeLong EF 2004 Different SAR86 subgroups harbour divergent proteorhodopsins Environ Microbiol 6 903 910 15305915
Tyson GW Chapman J Hugenholtz P Allen EE Ram RJ 2004 Community structure and metabolism through reconstruction of microbial genomes from the environment Nature 428 37 43 14961025
Venter JC Remington K Heidelberg JF Halpern AL Rusch D 2004 Environmental genome shotgun sequencing of the Sargasso Sea Science 304 66 74 15001713
Acinas SG Klepac-Ceraj V Hunt DE Pharino C Ceraj I 2004 Finescale phylogenetic architecture of a complex bacterial community Nature 430 551 554 15282603
Tringe S von Mering C Kobayashi A Salamov A Chen K 2005 Comparative metagenomics of microbial communities Science 308 554 557 15845853
Hallam SJ Putnam N Preston CM Detter JC Rokhsar D 2004 Reverse methanogenesis: Testing the hypothesis with environmental genomics Science 305 1457 1462 15353801
Dale C Dunbar H Moran NA Ochman H 2005 Extracting single genomes from heterogenous DNA samples: A test case with Carsonella ruddii , the bacterial symbiont of psyllids (Insecta) J Insect Sci 5 3 16299593
Pop M Philippy A Delcher AL Salzberg SL 2004 Comparative genome assembly Brief Bioinform 5 237 248 15383210
Bray N Pachter L 2004 MAVID: Constrained ancestral alignment of multiple sequences Genome Res 14 693 699 15060012
Sundararajan M Brudno M Small K Sidow A Batzoglou S 2004 Chaining algorithms for alignment of draft sequence. Fourth Workshop on Algorithms in Bioinformatics; 2004 25–27 May; Bergen, Norway Available: http://ai.stanford.edu/~serafim/wabi_finalSerafim.pdf . Accessed 7 July 2005.
Salzberg S Hotopp J Delcher A Pop M Smith D 2005 Serendipitous discovery of Wolbachia genomes in multiple Drosophila species Genome Biol 6 R23 15774024
Batzoglou S 2005 Algorithmic challenges in mammalian genome sequence assembly Dunn M Jorde L Little P Subramaniam S Encyclopedia of genomics, proteomics and bioinformatics Hoboken (New Jersey) John Wiley and Sons In press.
Ruepp A Graml W Santos-Martinez M Koretke KK Volker C 2000 The genome sequence of the thermoacidophilic scavenger Thermoplasma acidophilum
Nature 407 508 513 11029001
Breitbart M Salamon P Andresen B Mahaffy J Segal A 2002 Genomic analysis of an uncultured marine viral community Proc Natl Acad Sci U S A 99 14250 14255 12384570
Jones T Federspiel NA Chibana H Dungan J Kalman S 2004 The diploid genome sequence of Candida albicans
Proc Natl Acad Sci U S A 101 7329 7334 15123810
Myers EW Sutton GG Delcher AL Dew IM Fasulo DP 2000 A whole-genome assembly of Drosophila
Science 287 2196 2204 10731133
Aparicio S Chapman J Stupka E Putnam N Chia J 2002 Whole genome shotgun assembly and analysis of the genome of Fugu rubripes
Science 297 1301 1310 12142439
DeLong EF 2005 Microbial community genomics in the ocean Nat Rev Microbiol 3 459 469 15886695
Abe T Kanaya S Kinouchi M Ichiba Y Kozuki T 2003 Informatics for unveiling hidden genome signatures Genome Res 13 693 702 12671005
Campbell A Mrazek J Karlin S 1999 Genome signature comparisons among prokaryote, plasmid and mitochondrial DNA Proc Natl Acad Sci U S A 96 9184 9189 10430917
Deschavanne PJ Giron A Vilain K Fagot G Fertil B 1999 Genomic signature: Characterization and classification of species assessed by chaos game representation of sequences Mol Biol Evol 16 1391 1399 10563018
Karlin S Campbell AM Mrazek J 1998 Comparative DNA analysis across diverse genomes Annu Rev Genet 32 185 225 9928479
Teeling H Meyerdierks A Bauer M Amann R Glöckner FO 2004 Application of tetranucleotide frequencies for the assignment of genomic fragments Environ Microbiol 6 938 947 15305919
Teeling H Waldmann J Lombardot T Bauer M Glöckner FO 2004 TETRA: A web-service and a stand-alone program for the analysis and comparison of tetranucleotide usage patterns in DNA sequences BMC Bioinformatics 5 163 15507136
Angly F Rodriguez-Brito B Bangor D McNairnie P Breitbart M 2005 PHACCS, an online tool for estimating the structure and diversity of uncultured viral communities using metagenomic information BMC Bioinformatics 6 41 15743531
Salzberg S Delcher A Kasif S White O 1998 Microbial gene identification using interpolated Markov models Nucleic Acids Res 26 544 548 9421513
Delcher AL Harmon D Kasif S White O Salzberg SL 1999 Improved microbial gene identification with GLIMMER Nucleic Acids Res 27 4636 4641 10556321
Audic S Claverie J 1998 Self-identification of protein-coding regions in microbial genomes Proc Natl Acad Sci U S A 95 10026 10031 9707594
Hayes W Borodovsky M 1998 How to interpret an anonymous bacterial genome: Machine learning approach to gene identification Genome Res 8 1154 1171 9847079
Goo Y Roach J Glusman G Baliga N Deutsch K 2004 Low-pass sequencing for microbial comparative genomics BMC Genomics 5 3 14718067
Kawashima T Amano N Koike H Makino S Higuchi S 2000 Archaeal adaptation to higher temperatures revealed by genomic sequence of Thermoplasma volcanium
Proc Natl Acad Sci U S A 97 14257 14262 11121031
Tatusov R Koonin E Lipman D 1997 A genomic perspective on protein families Science 278 631 637 9381173
Kanehisa M 1997 A database for post-genome analysis Trends Genet 13 375 376 9287494
Kanehisa M Goto S 2000 KEGG: Kyoto Encyclopedia of Genes and Genomes Nucleic Acids Res 28 27 30 10592173
Quackenbush J 2001 Computational analysis of microarrary data Nat Rev Genet 2 418 427 11389458
Eisen MB Spellman PT Brown PO Botstein D 1998 Cluster analysis and display of genome-wide expression patterns Proc Nat Acad Sci U S A 95 14863 14868
Feller W 1968 An introduction to probability theory and its applications, Volume 1 Hoboken (New Jersey) John Wiley and Sons 528 p.
Breitbart M Hewson I Felts B Mahaffy J Nulton J 2003 Metagenomic analysis of an uncultured viral community from human feces J Bacteriol 185 6220 6223 14526037
Durbin R Eddy SR Krogh A Mitchison G 2004 Biological sequence analysis: Probabilistic models of proteins and nucleic acids Cambridge Cambridge University Press 368 p.
Felsenstein J 2004 Inferring phylogenies Sunderland (Massachusetts) Sinauer Associates 664 p.
Cole JR Chai B Farris RJ Wang Q Kulam SA 2005 The Ribosomal Database Project (RDP-II): Sequences and tools for high-throughput rRNA analysis Nucleic Acids Res 33 D294 D296 15608200
Ludwig W Strunk O Westram R Richter L Meier H 2004 ARB: A software environment for sequence data Nucleic Acids Res 32 1363 1371 14985472
Rohwer F Edwards R 2002 The phage proteomic tree: A genome-based taxonomy for phage J Bacteriol 184 4529 4535 12142423
Margulies EH Vinson JP Miller W Jaffe DB Lindblad-Toh K 2005 An initial strategy for the systematic identification of functional elements in the human genome by low-redundancy comparative sequencing Proc Natl Acad Sci U S A 102 4795 4800 15778292
Huang X 1994 On global sequence alignment Comput Appl Biosci 10 227 235 7922677
Edgar RC 2004 MUSCLE: Multiple sequence alignment with high accuracy and high throughput Nucleic Acids Res 32 1792 1797 15034147
Wiens JJ 2003 Incomplete taxa, incomplete characters and phylogenetic accuracy: Is there a missing data problem? J Vertebr Paleontol 23 297 310
Kearney M 2002 Fragmentary taxa, missing data, and ambiguity: Mistaken assumptions and conclusions Syst Biol 51 369 381 12028738
Bininda-Emonds ORP 2004 Phylogenetic supertrees: Combining information to reveal the tree of life New York Springer 550 p.
Chen D Eulenstein O Fernandez-Baca D 2004 Rainbow: A toolbox for phylogenetic supertree construction and analysis Bioinformatics 20 2872 2873 15145807
Pachter L Speyer D 2004 Reconstructing trees from subtree weights Appl Math Lett 7 615 621
Pachter L, Sturmfels B, editors 2005 Algebraic statistics for computational biology Cambridge: Cambridge University Press In press.
Forney L Zhou X Brown C 2004 Molecular microbial ecology: Land of the one-eyed king Curr Opin Microbiol 7 210 220 15196487
Curtis TP Sloan WT Scannell JW 2002 Modelling prokaryotic diversity and its limits Proc Natl Acad Sci U S A 99 10494 10499 12097644
Breitbart M Felts B Kelley S Mahaffy JM Nulton J 2004 Diversity and population structure of a near-shore marine-sediment viral community Proc Biol Sci 271 565 574 15156913
Cann AJ Fandrich SE Heaphy S 2005 Analysis of the virus population present in equine faeces indicates the presence of hundreds of uncharacterized virus genomes Virus Genes 30 151 156 15744573
|
16110337
|
PMC1185649
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Jul 12; 1(2):e24
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010024
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1781602274010.1186/1471-2105-6-178SoftwareApproaching the taxonomic affiliation of unidentified sequences in public databases – an example from the mycorrhizal fungi Nilsson R Henrik [email protected] Erik [email protected] Martin [email protected] Karl-Henrik [email protected] Göteborg University, Botanical Institute, Box 461, 405 30 Göteborg, Sweden2 Chalmers University of Technology, Mathematical Sciences, 412 96 Göteborg, Sweden2005 18 7 2005 6 178 178 24 3 2005 18 7 2005 Copyright © 2005 Nilsson et al; licensee BioMed Central Ltd.2005Nilsson 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
During the last few years, DNA sequence analysis has become one of the primary means of taxonomic identification of species, particularly so for species that are minute or otherwise lack distinct, readily obtainable morphological characters. Although the number of sequences available for comparison in public databases such as GenBank increases exponentially, only a minuscule fraction of all organisms have been sequenced, leaving taxon sampling a momentous problem for sequence-based taxonomic identification. When querying GenBank with a set of unidentified sequences, a considerable proportion typically lack fully identified matches, forming an ever-mounting pile of sequences that the researcher will have to monitor manually in the hope that new, clarifying sequences have been submitted by other researchers. To alleviate these concerns, a project to automatically monitor select unidentified sequences in GenBank for taxonomic progress through repeated local BLAST searches was initiated. Mycorrhizal fungi – a field where species identification often is prohibitively complex – and the much used ITS locus were chosen as test bed.
Results
A Perl script package called emerencia is presented. On a regular basis, it downloads select sequences from GenBank, separates the identified sequences from those insufficiently identified, and performs BLAST searches between these two datasets, storing all results in an SQL database. On the accompanying web-service , users can monitor the taxonomic progress of insufficiently identified sequences over time, either through active searches or by signing up for e-mail notification upon disclosure of better matches. Other search categories, such as listing all insufficiently identified sequences (and their present best fully identified matches) publication-wise, are also available.
Discussion
The ever-increasing use of DNA sequences for identification purposes largely falls back on the assumption that public sequence databases contain a thorough sampling of taxonomically well-annotated sequences. Taxonomy, held by some to be an old-fashioned trade, has accordingly never been more important. emerencia does not automate the taxonomic process, but it does allow researchers to focus their efforts elsewhere than countless manual BLAST runs and arduous sieving of BLAST hit lists. The emerencia system is available on an open source basis for local installation with any organism and gene group as targets.
==== Body
Background
Mycorrhiza is the term used to denote the root-associated symbiosis between fungus and plant where resources otherwise unattainable or very costly (chiefly carbohydrates and mineral nutrients, respectively) are exchanged. Whereas the identification of the plant partner of the symbiosis often is comparatively straightforward, the identity of the fungal component is typically much more elusive, largely so due to the dearth of information obtainable from the root samples of the plant (Figure 1). The traditional approaches to identification of mycorrhizae include studies based on light microscopy, isozyme assays, mating behaviour experiments, and somatic compatibility tests. However, all of these are associated with drawbacks such as low to moderate precision, high time consumption, or the requirement that the fungus be isolated and grown in culture, which is impossible for many mycorrhizal fungi [1-4]. Alternatively, it is sometimes possible to establish a hyphal connection between the fungal mycelium of the root tips and nearby fungal fruiting-bodies [5]; traditional fungal taxonomy rests to a large extent on the morphology of fruiting-bodies or other spore producing structures, and there is abundant literature information available for many groups of fungi. The factors triggering the formation of fruiting-bodies of mycorrhizal fungi are, however, poorly understood. This is reflected in the large number of root-associated fungi for which fruiting-bodies have never been found, suggesting that any attempt to characterize the below-ground mycoflora through collection and identification of above-ground fruiting-bodies is likely to give a skewed and incomplete picture [6-8].
Figure 1 The fly agaric: a common mycorrhizal fungus. a) Fruiting-bodies of the ectomycorrhizal fly agaric (Amanita muscaria). b) Root-tip mycelia of the Amanita type.
The advent of PCR-based methodologies has in many ways revolutionized the field of mycorrhizal research by providing a means by which, at least in theory, every fungus could be identified to species level. It has also served to highlight further the patchiness in our understanding of mycorrhizal fungi; DNA-based studies of mycorrhizal fungi typically contain a considerable proportion of sequences that lack identified matches in public sequence repositories such as GenBank and UNITE [9], and that hence cannot be identified to species or even genus level [10-13]. The prerequisite of most journals that all sequences used in a publication be made public in GenBank naturally leads authors to submit such unidentified sequences under names like "Fungus: Environmental sample", "Unidentified mycorrhizal basidiomycete", "Uncultured root-tip fungus", and "Cortinarius sp.". Thus, even if an unidentified mycorrhizal sequence turns out to have one or more identical (or nearly so) matches in GenBank, the unidentified nature of the latter still precludes identification to species level.
As more and more sequences from well-identified fruiting-bodies are submitted to and accumulate in GenBank, one can expect some of the new sequences to hint and even resolve the identity of previously unmatched sequences. Any such relation is likely to impart important information to the studies that employed the unidentified sequences in question. Yet there is no generic mechanism on GenBank to alert the sequences' authors – or anyone else who might be interested – of these new matches. In addition, many sequence authors are inefficient in keeping the annotations of their submitted sequences up-to-date, even when new taxonomic progress and knowledge have been attained. As a result, important information goes by undetected, and obsolete annotations remain and are propagated through the databases and literature via subsequent BLAST runs [14]. The present study introduces emerencia, a Perl script package to facilitate the keeping track of the identity of insufficiently identified GenBank sequences over time. A web-service demonstrating emerencia for mycorrhizal fungi and the nuclear ITS region is presented; emerencia is also available on an open source-basis and can be downloaded from the web page for local installation with any gene and organism group as targets.
Implementation
The web-service provided at is set to monitor the identity of fungal ITS sequences whose taxonomic annotations are poorly resolved. The ITS region offers three sub-loci of very different conservation levels (ITS1- very variable; 5.8S- very conserved; ITS2- semi-conserved [15]) and is as such the prevalent region of choice when fungi are to be identified [16,17,3].
The main script of the emerencia package is written in Perl [18] and runs in a UNIX environment. On a regular basis, the script is evoked and connects to GenBank through BioPerl routines [19] to retrieve all fungal ITS sequences deposited since the last update using the query phrase
(("Fungi" [Organism] AND (200 [SLEN] : 3000 [SLEN])) AND (((ITS1 [titl] OR ITS2 [titl]) OR 5.8S [titl]) OR "internal transcribed spacer" [titl] OR "internal transcribed spacers" [titl]))
For each downloaded sequence, the GenBank format annotation is parsed to determine whether the sequence should be regarded as fully identified (i.e., identified to species level) or insufficiently identified (i.e., identified only to phylum Fungi (e.g., "Unidentified fungus"), identified to ordinal level (e.g., "Thelephorales sp."), or identified to generic level (e.g., "Amanita sp.")). The criteria for this decision can be found in full detail at ; for example, a sequence is regarded as insufficiently identified if its species annotation (GenBank format reference: SOURCE) contains words like "unidentified", "uncultured", "sp.", "mycorrhizal isolate", or "unknown" (this section of the script was repeatedly fine-tuned to minimize the number of false positives and negatives). All new sequences are appended to one of two tables of a local MySQL database [20] such that one table holds the identified sequences and the other the insufficiently identified ones. The structure of the database is provided at . Local BLAST search indices [21] are computed to allow for similarity searches in both tables.
Two sets of local BLAST searches are then run using default settings:
1. all insufficiently identified sequences are BLAST:ed against all identified sequences
2. all insufficiently identified sequences are BLAST:ed against all insufficiently identified sequences.
For the first BLAST run (1), details of the best BLAST match are inspected. If, for any insufficiently identified sequence, the best BLAST match to the table of identified sequences is found to have changed (i.e., a change in the accession number of the best BLAST match in combination with an improved E-value or identical E-value and an improved score) – or if the unidentified sequence lacks any previous significant match – details of the new best BLAST match (accession number, date, and BLAST score and E-value) are appended to the entry of the insufficiently identified sequence. Similarly, for the second BLAST run (2), the best non-self match of each insufficiently identified sequence to the table of insufficiently identified sequences is noted and saved. To retain a BLAST history, the former best BLAST match of each insufficiently identified sequence is also kept. The end product of the main script of emerencia is, thus, two updated, mutually exclusive MySQL tables – one with identified fungal ITS sequences and one with insufficiently identified ITS sequences, and both with cross-linked indices on best and former best BLAST matches in each table.
The web-service enables visitors to interact with the database in a number of ways. Four major search categories are provided (Table 1):
Table 1 Functions of the emerencia web-service. Functions of the emerencia web-service at ; some examples and an informal walkthrough are also given at this address. The output of the functions features relayed hyperlinks to GenBank, Google, and Tree of Life for quick retrieval of additional information; where applicable, insufficiently identified sequences are also hyperlinked to the SEARCH FOR INSUFFICIENTLY IDENTIFIED SEQUENCE BY ACCESSION NUMBER function for a more detailed description of the sequence and its matches.
SEARCH FOR INSUFFICIENTLY IDENTIFIED SEQUENCE BY ACCESSION NUMBER For any given accession number of an insufficiently identified sequence, this function shows the present and previous best BLAST matches from the table of identified sequences together with match scores and relevant annotation. A Clustal W multiple alignment [34] of the sequences is generated and shown as an important aid in interpreting the BLAST match values. In addition, all the above is shown for the present and previous best BLAST matches in the table of insufficiently identified sequences. This function requires that the accession number provided by the user be present in the table of insufficiently identified sequences.
CHECK SPECIFIC PUBLICATION FOR INSUFFICIENTLY IDENTIFIED SEQUENCES AND THEIR IDENTITY This function retrieves all insufficiently identified sequences stemming from the user-specified publication and shows the present best identified BLAST match (and some additional information) for those sequences. The function expects 5–10 distinct key words from the title / author /journal fields of the publication and requires that at least one insufficiently identified sequence was released together with the publication in question.
SEARCH FOR INSUFFICIENTLY IDENTIFIED SEQUENCES MATCHING ACCESSION NUMBER OF IDENTIFIED TAXA For any given accession number in the table of identified sequences, this function retrieves and details all entries in the table of insufficiently identified sequences for which this accession number represents the best BLAST match. It requires that the specified accession number be present in the table of identified sequences and will proceed only if that accession number indeed constitutes the best BLAST match of at least one insufficiently identified sequence.
SEARCH FOR INSUFFICIENTLY IDENTIFIED SEQUENCES BY KEY WORD This function lets the user query the species annotation field of the table of insufficiently identified sequences using 2–5 key words, and displays all insufficiently identified sequences matching the key words. For those sequences, the best BLAST match to the table of identified sequences will be shown together with some additional information.
• SEARCH FOR INSUFFICIENTLY IDENTIFIED SEQUENCE BY ACCESSION NUMBER
• CHECK SPECIFIC PUBLICATION FOR INSUFFICIENTLY IDENTIFIED SEQUENCES AND THEIR IDENTITY
• SEARCH FOR INSUFFICIENTLY IDENTIFIED SEQUENCES MATCHING ACCESSION NUMBER OF IDENTIFIED TAXA
• SEARCH FOR INSUFFICIENTLY IDENTIFIED SEQUENCES BY KEY WORD
The different search functions produce output pages that are extensively hyperlinked to facilitate further queries against the web-service itself as well as external information resources such as GenBank and the Tree of Life project [22]. Apart from querying the database through a web browser and bookmarking or saving the results, users can subscribe to accession numbers of insufficiently identified sequences. This enables immediate notification by e-mail when the best BLAST matches of those sequences change.
The web-service is hosted on a MacOS X server running the Apache web server [23]. The databases are queried using dedicated CGI scripts written in Perl; parts of the source code of the galaxieEST and mor packages [24,17] were used for this purpose. The extensively annotated source code of the emerencia core is freely available at the web-service. Local installation and additional technical aspects are described in the documentation.
Results and Discussion
The last decade has seen a dramatic improvement of our understanding of mycorrhizal diversity, largely due to the advent of fast and comparatively cheap PCR-based methods. The discovery of new, previously unsequenced mycorrhizal fungi poses something of a taxonomic problem, particularly when fruiting-bodies and other distinguishing characteristics are absent. Many of these sequences are submitted ad tempus as "environmental samples". Unfortunately, the absence of generic mechanisms – and the apparent lack of motivation of the authors of the sequences – to refine the identity of these sequences as more information is amassed force other researchers to put in a great deal of effort (typically countless manual BLAST runs) in order to make sense of the sequences and the relation of their own sequences to those. With more than a handful of sequences to monitor, the task quickly becomes unreasonably time-consuming. The authors present a prototype software package to minimize the amount of work needd to stay updated on the identity of such insufficiently identified sequences. The web-service provided allows users to subscribe to accession numbers (sequences) with automatic email notification upon identity changes; alternatively, the same – and additional – information can be obtained through the search functions of the web-service. Furthermore, to install emerencia locally and modify it to run with other organism and gene groups should not pose any problem to anyone with a reasonable experience of UNIX-type environments. Such a local installation can stretch from a private, shell window-only tool to a public, user-oriented web-service such as the one presented here. For a local installation, parameters such as how often the script should be started, the BLAST settings, and what information to store locally, can be set as seen fit.
As of May 2005, emerencia has fetched about 29000 identified and 7500 insufficiently identified ITS sequences (Table 2). The identified sequences belong to some 8000 distinct species, which corresponds to approximately 0.5% of the estimated 1.5 million extant species of fungi [25]. While the number of fungal sequences in GenBank is expected to increase drastically over the next few years, it will take a long time before all gaps are filled [26], leaving taxon sampling a tangible problem for emerencia as well as for other tools used for sequence identification. In addition, the poor state of many taxonomic annotations in GenBank and other databases [27,28] complicates the above percentage estimates and poses a challenge to users of emerencia. As with other identification tools, it is crucial that the results obtained be viewed and used as guidance for further studies rather than accepted as true and final; emerencia is a tool to refine iteratively the identity of insufficiently identified sequences in public databases and to promote the flow of information pertaining to those sequences. It is not intended – and should never be used – as a shortcut to unequivocally correct species names and annotations.
Table 2 A brief summary of the sequence data underlying the emerencia web service at as of May 2005. The threshold BLAST E-values for "good" and "poor" matches were arbitrarily set to 0.0 and 1e-100, respectively. Graphical illustrations showing the population of the database over time and additional aspects of emerencia are generated automatically on a monthly basis and are available at the above address.
NUMBER OF INSUFFICIENTLY IDENTIFIED SEQUENCES 7528 (21 % of total)
NUMBER OF IDENTIFIED SEQUENCES 28959 (79% of total)
NUMBER OF INSUFFICIENTLY IDENTIFIED SEQUENCES WITH GOOD MATCHES (E-VALUE = 0.0) 4791 (64 % of the insufficiently identified sequences)
NUMBER OF INSUFFICIENTLY IDENTIFIED SEQUENCES WITH POOR MATCHES (E-VALUE >1E-100) 1135 (15 % of the insufficiently identified sequences)
TOTAL NUMBER OF SEQUENCES LAST UPDATED BEFORE 1995-01-01 180 (0.5%)
TOTAL NUMBER OF SEQUENCES LAST UPDATED BEFORE 2000-01-01 3651 (10 %)
TOTAL NUMBER OF SEQUENCES LAST UPDATED BEFORE 2005-01-01 31858 (87%)
NUMBER OF INSUFFICIENTLY IDENTIFIED SEQUENCES LAST UPDATED BEFORE 2000-01-01 264 (3.5 % of the insufficiently identified sequences)
NUMBER OF INSUFFICIENTLY IDENTIFIED SEQUENCES LAST UPDATED BEFORE 2000-01-01 AND WITH POOR MATCHES (E-VALUE > 1E-100) 17 (0.2 % of the insufficiently identified sequences)
NUMBER OF IDENTIFIED SEQUENCES HAVING AT LEAST ONE INSUFFICIENTLY IDENTIFIED COUNTERPART AS IDENTIFIED BY BLAST 2981 (10 % of the identified sequences)
NUMBER OF IDENTIFIED SEQUENCES WITHOUT INSUFFICIENTLY IDENTIFIED COUNTERPARTS 25978 (90 % of the identified sequences)
As with BLAST searches in general, several factors impede the interpretation of the result. The aforementioned problem with taxonomic annotations in GenBank calls, in itself, for subsequent hands-on verification of the results. Furthermore, BLAST explores – and tries to expand – local regions of sequence similarity, and it takes manual inspection of the BLAST results to find out whether the entire, or only a portion of, the query sequence was successfully matched to anything in the database. A match to only a part of the target sequence (such as the very conserved 5.8S sub-locus of the ITS region) is, for identification purposes, tantamount to no match at all [16]. It is also important to keep in mind that BLAST provides a measure of similarity, but similarity does not in turn provide a sound measure of relatedness [29,30]. Finally, it is notoriously difficult to tell an identified sequence apart from an insufficiently identified one on an automated basis; indeed, the present set-up is likely to yield a small proportion of false positives as well as false negatives (presently less than 0.1%). Such problems would largely have been avoided had there been an accepted standard for annotation of unidentified – and identified – sequences.
emerencia bears some resemblance to tools like Sequence Alerting System [31], Swiss-Shop [32], and ReHab [33], but a number of features set emerencia apart from these. emerencia is primarily a taxonomic utility designed to add an integrative aspect to GenBank data; its automated separation of identified and insufficiently identified sequences paves the way for researchers seeking reliable identification of species rather than merely the best possible match scores. emerencia can be installed locally or accessed over the Internet; in the latter case, the user will need nothing but a web browser. The data structure of emerencia allows many types of interesting questions to be asked; for instance, insufficiently identified sequences and their present best identified matches can be listed publication-wise, or all insufficiently identified sequences that constitute the best BLAST match of some given identified sequence can be listed (essentially amounting to a BLAST run in reverse) (Table 1). emerencia is tailored for variable nucleotide sequences, whereas proteins represent the target for Swiss-Shop and ReHab. Finally, the e-mail subscription utility provides a convenient way for users to stay taxonomically updated on select insufficiently identified sequences with a minimum of effort.
Conclusion
Insufficiently identified sequences generally add little to the studies in which they are included, and it is important to estimate their identity as correctly as possible. The magnitude of this manual task increases with the number of sequences, but this process can fortunately be automated. However, in spite of computational advances, the taxonomic process itself lies beyond automation, alluding to the importance of both good species knowledge and the inherent need to always approach hypothesized identifications in a critical way.
Availability and requirements
Project name: emerencia
Project home page:
Operating system(s): Primarily UNIX type platforms
Programming language: PERL, SQL
Other requirements: BLAST, Apache httpd, BioPerl, Clustal W (optional)
License: GNU GPL version 2
Any restrictions to use by non-academics: None other than those imposed by GNU GPL version 2
List of abbreviations
BLAST – Basic Local Alignment Search Tool
ITS- Internal Transcribed Spacers
SQL – Structured Query Language
Authors' contributions
All authors contributed to the structure and functions of emerencia. RHN and EK wrote most parts of the emerencia core, the CGI scripts, and the database handlers. MR was responsible for the mycological part, including literature comparison, integrity testing, and data verification. KHL contributed with advice on fungal taxonomy and systematics. All authors drafted the manuscript and approved the final version.
Acknowledgements
Tom Bruns, University of California at Berkeley, is acknowledged for initiating the project. The photos of Figure 1 were kindly provided by Leif & Anita Stridvall and Ellen Larsson, respectively. Anders Sjögren is acknowledged for valuable advice on technical matters. The manuscript benefited from helpful comments by Brandon Matheny and Manfred Binder. Financial support to RHN from the Lars Hierta Foundation and from the Helge Ax: son Johnson Foundation are gratefully acknowledged. emerencia was created using only freely available software .
==== Refs
Gardes M White TJ Fortin JA Bruns TD Taylor JW Identification of indigenous and introduced symbiotic fungi in ectomycorrhizae by amplification of nuclear and mitochondrial ribosomal DNA Can J Botany 1991 69 180 190
Marmaisse R Debaud JC Casselton LA DNA probes for species and strain identification in the ectomycorrhizal fungus Hebeloma MycolRes 1992 96 161 165
Horton TR Bruns TD The molecular revolution in ectomycorrhizal ecology: peeking into the black-box Mol Ecol 2001 10 1855 1871 11555231 10.1046/j.0962-1083.2001.01333.x
Kõljalg U Tammi H Timonen S Agerer R Sen R ITS rDNA sequence-based phylogenetic analysis of Tomentellopsis species from boreal and temperate forests, and the identification of pink-type ectomycorrhizas Mycol Prog 2002 1 81 92
Agerer R Characterization of ectomycorrhiza Method Microbiol 1991 23 25 73
Gardes M Bruns TD Community structure of ectomycorrhizal fungi in a Pinus muricata forest: above-and below-ground views Can J Botany 1996 74 1572 1583
Dahlberg A Community ecology of ectomycorrhizal fungi: an advancing interdisciplinary field New Phytol 2001 150 555 562 10.1046/j.1469-8137.2001.00142.x
Valentine LL Fiedler TL Hart AN Petersen CA Berninghausen HK Southworth D Diversity of ectomycorrhizas associated with Quercus garryana in southern Oregon Can J Botany 2004 82 123 135 10.1139/b03-117
Kõljalg U Larsson K-H Abarenkov K Nilsson RH Alexander IJ Eberhardt U Erland S Høiland K Kjøller R Larsson E Pennanen T Sen R Taylor AFS Vrålstad T Tedersoo L Ursing BM UNITE – a database providing web based methods for the molecular identification of ectomycorrhizal fungi New Phytol 2005 166 1063 1068 15869663 10.1111/j.1469-8137.2005.01376.x
Rosling A Landeweert R Lindahl BD Larsson K-H Kuyper TW Taylor AFS Finlay RD Vertical distribution of ectomycorrhizal fungal taxa in a podzol soil profile New Phytol 2003 159 775 783 10.1046/j.1469-8137.2003.00829.x
Tedersoo L Kõljalg U Hallenberg N Larsson K-H Fine scale distribution of ectomycorrhizal fungi and roots across substrate layers including coarse woody debris in a mixed forest New Phytol 2003 159 153 165 10.1046/j.1469-8137.2003.00792.x
Nielsen KB Kjøller R Olsson PA Schweiger PF Andersen F∅ Rosendahl S Colonisation and molecular diversity of arbuscular mycorrhizal fungi in the aquatic plants Littorella uniflora and Lobelia dortmanna in southern Sweden Mycol Res 2004 108 616 625 15323243 10.1017/S0953756204000073
Kaldorf M Renker C Fladung M Buscot F Characterization and spatial distribution of ectomycorrhizas colonizing aspen clones released in an experimental field Mycorrhiza 2004 14 295 306 14534850 10.1007/s00572-003-0266-1
Schüßler A Schwarzott D Walker C Glomeromycota rRNA genes – the diversity of myths? Mycorrhiza 2003 13 233 236 12845513 10.1007/s00572-003-0250-9
Álvarez I Wendel JF Ribosomal ITS sequences and plant phylogenetic inference Mol Phylogenet Evol 2003 29 417 434 14615184 10.1016/S1055-7903(03)00208-2
Bruns TD Shefferson RP Evolutionary studies of ectomycorrhizal fungi: recent advances and future directions Can J Botany 2004 82 1122 1132 10.1139/b04-021
Hibbett DS Nilsson RH Snyder M Fonseca M Costanzo J Shonfeld M Automated phylogenetic taxonomy: An example in the Homobasidiomycetes (mushroom-forming fungi) Syst Biol
The Perl Documentation Project
BioPerl
MySQL AB
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
The Tree of Life Project
The Apache httpd server project
Nilsson RH Rajashekar B Larsson K-H Ursing BM galaxieEST: addressing EST identity through automated phylogenetic analysis BMC Bioinformatics 2004 5 87 15236648 10.1186/1471-2105-5-87
Hawksworth DL The magnitude of fungal diversity: the 1.5 million species estimate revisited Mycol Res 2001 105 1422 1432
Berney C Fahrni J Pawlowski J How many novel eukaryotic 'kingdoms'? Pitfalls and limitations of environmental DNA surveys BMC Biol 2004 2 13 15176975 10.1186/1741-7007-2-13
Bridge PD Roberts PJ Spooner BM Panchal G On the unreliability of published DNA sequences New Phytol 2003 160 43 48 10.1046/j.1469-8137.2003.00861.x
Vilgalys R Taxonomic misidentification in public DNA databases New Phytol 2003 160 4 5 10.1046/j.1469-8137.2003.00894.x
de Queiroz K Phylogenetic definitions and taxonomic philosophy Biol Philos 1992 7 295 313 10.1007/BF00129972
Nilsson RH Larsson K-H Ursing BM galaxie – CGI scripts for sequence identification through automated phylogenetic analysis Bioinformatics 2004 20 1447 1452 14976034 10.1093/bioinformatics/bth119
Sequence Altering System
Swiss-shop
Whitney J Esteban DJ Upton C Recent hits acquired by BLAST (ReHAB): A tool to identify new hits in sequence similarity searches BMC Bioinformatics 2005 6 23 15701178 10.1186/1471-2105-6-23
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
|
16022740
|
PMC1186019
|
CC BY
|
2021-01-04 16:27:25
|
no
|
BMC Bioinformatics. 2005 Jul 18; 6:178
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-178
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-771600461410.1186/1471-2407-5-77Research ArticleMicroarray comparative genomic hybridization detection of chromosomal imbalances in uterine cervix carcinoma Hidalgo Alfredo [email protected] Michael [email protected] Iver [email protected] Hugo [email protected]ña Patricia [email protected]ázquez-Ortiz Guelaguetza [email protected]ández Dulce [email protected]ález José [email protected] Minerva [email protected]ópez Ricardo [email protected]érez Carlos [email protected]ía José [email protected]ázquez Karla [email protected] Brenda [email protected] Mauricio [email protected] Laboratorio de Oncología Genómica, Unidad de Investigación Médica en Enfermedades Oncológicas, Centro Médico Nacional Siglo XXI-IMSS, México2 Division of Pediatric Haematology/Oncology, University of Florida, Gainesville, USA3 Institute of Pathology, University Hospital Charité, Berlin, Germany4 Servicio de Epidemiología, Hospital de Oncologia, Centro Médico Nacional Siglo XXI-IMSS, México5 Clínica de Displasias, Hospital de Gineco-Obstetrica No. 4, Luis Castelazo Ayala-IMSS, México6 Departamento de Patología, Facultad de Medicina UNAM-Hospital General de México, SS, México7 Laboratorio de Biología Teórica, Departamento de Investigación, Universidad La Salle, México8 Instituto Nacional de Medicina Genomica, Secretaria de Salud, Mexico2005 9 7 2005 5 77 77 22 2 2005 9 7 2005 Copyright © 2005 Hidalgo 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
Chromosomal Comparative Genomic Hybridization (CGH) has been applied to all stages of cervical carcinoma progression, defining a specific pattern of chromosomal imbalances in this tumor. However, given its limited spatial resolution, chromosomal CGH has offered only general information regarding the possible genetic targets of DNA copy number changes.
Methods
In order to further define specific DNA copy number changes in cervical cancer, we analyzed 20 cervical samples (3 pre-malignant lesions, 10 invasive tumors, and 7 cell lines), using the GenoSensor microarray CGH system to define particular genetic targets that suffer copy number changes.
Results
The most common DNA gains detected by array CGH in the invasive samples were located at the RBP1-RBP2 (3q21-q22) genes, the sub-telomeric clone C84C11/T3 (5ptel), D5S23 (5p15.2) and the DAB2 gene (5p13) in 58.8% of the samples. The most common losses were found at the FHIT gene (3p14.2) in 47% of the samples, followed by deletions at D8S504 (8p23.3), CTDP1-SHGC- 145820 (18qtel), KIT (4q11-q12), D1S427-FAF1 (1p32.3), D9S325 (9qtel), EIF4E (eukaryotic translation initiation factor 4E, 4q24), RB1 (13q14), and DXS7132 (Xq12) present in 5/17 (29.4%) of the samples.
Conclusion
Our results confirm the presence of a specific pattern of chromosomal imbalances in cervical carcinoma and define specific targets that are suffering DNA copy number changes in this neoplasm.
==== Body
Background
Uterine cervix carcinoma (UCC) represents the second cause of death among the female population worldwide. The fact that more than 99% of all the cervical invasive tumors are positive for infection with high risk human papillomavirus (HPV) suggests that this is one of the most important factors for the development of this neoplasm [1,2]. These viruses can induce cellular transformation by several mechanisms; the viral oncoproteins E6 and E7 can interact with cellular proteins involved in important cellular functions, such as tumor suppression, apoptosis, cell cycle control, genomic instability, transcriptional regulation and immune evasion [3].
The induction of genomic instability by HPV seems to be particularly important for the establishment and development of an invasive tumor [4,5] since this increased genomic plasticity would generate cellular clones with enhanced transforming and invasive potential [6].
Metaphase comparative genomic hybridization (mCGH) has been applied to study different stages of this tumor [4,7-19], detecting specific patterns of chromosomal imbalances that arises very early during the development of cervical carcinoma, suggesting that the gain of chromosome 3q is one of the most important genetic alteration that defines the transition from a pre-malignant lesion to an invasive carcinoma [4]. Some of these imbalances have been related to specific clinical behaviors, such as the presence of lymph node metastases [9]. However, given the spatial resolution of mCGH [20], little is known about the identity of specific genes that might be the targets of regional chromosomal imbalances. Matrix-based CGH or array CGH overcomes this problem increasing the sensitivity for the detection of DNA copy number changes at specific loci, through the use of well defined genomic DNA fragments whose mapping location is known, arrayed onto a solid surface [21-23], thereby achieving a resolution of copy number imbalances up to the single gene level.
In order to refine the patterns of chromosomal imbalances present in cervical carcinoma, and trying to identify specific genes that might be targets of copy number changes in this tumor, we applied microarray CGH on 20 uterine cervix-derived samples (three pre-malignant lesions, 10 invasive tumors and seven UCC derived cell lines) to detect DNA copy number changes at the single gene level.
Methods
Cervical tissues
All described procedures have been evaluated and approved by the local committee of ethics of the Mexican Institute of Social Security (IMSS), and all samples were taken after informed consent from the patients. The pre-malignant lesions and the invasive tumors were collected by colposcopy-directed biopsies at the Gynecology Department of the Hospital General de México, Mexico City. The biopsies were divided in three sections. The central part was used for genomic DNA extraction using the Wizard Genomic kit (Promega, Madison, WI, USA), and the extremes were fixed with 70% ethanol overnight and paraffin embedded. Hematoxilin-eosin stained sections from these biopsies were analyzed in order to confirm the presence of at least 70% tumoral cells in the samples.
Cell lines
The cell lines included in this study were: CasKi, SiHa, both positive for HPV16, and HeLa (HPV18) The CaLo and ViBo cell lines were established from stage IIB invasive tumors, while INBL and RoVa from a stage IVA tumor. These cells are HPV18 positive and were established from tumor explants at the laboratory of Cell differentiation and Cancer of the National University of Mexico [24]. The chromosomal CGH profiles of CaLo, ViBo, INBL and RoVa have been published recently [19].
HPV detection and typing
HPV detection was carried out by PCR using the consensus primers MY09 and MY11 for the L1 region of the viral genome. After a 5 min. denaturation at 94°C, 100 ng of DNA were subjected to 35 amplification cycles with the following parameters: 94°C for 1 min., 55°C for 2 min. and 73°C for 3 min., with a final extension step of 7 min. at 72°C. The amplicon was labeled using the Big Dye sequencing kit and sequenced on an ABI371 sequencer (Applied Biosystems, Foster City, CA, USA). BLAST sequence comparison was used in order to define the viral type.
Microarray CGH
Microarray CGH was performed using the GenoSensor Array 300 system, following the manufacturer's instructions (ABBOT-Vysis, Downers Grove, IL, USA). Each array contains 861 spots, representing 287 chromosomal regions that are commonly altered in human cancer, such as telomeres, regions involved in microdeletions, oncogenes, and tumor suppressor genes. Briefly, 100 ng of genomic DNA were labeled by a random primer reaction during two hours. Tumor DNA was labeled with Cy3 and the normal female reference DNA with Cy5. After the labeling reaction, the probes were digested with DNAse at 15°C for one hr., followed by two ethanol-purifications; finally the probe size was checked by gel electrophoresis. The hybridization mixture consisted of 2.5 μl of each of the differentially labeled DNAs plus 25 μl of hybridization buffer provided in the kit. This mixture was denatured at 80°C for 10 min. at 80°C, followed by incubation at 37°C for one hr. Five μl of this probe were applied onto the spotted area of the array under a coverslip and hybridized in a humid chamber containing 50% formamide (FA)/2XSSC at 37°C for 72 hrs. After hybridization, the arrays were washed 3X in 50%FA/2XSSC at 40°C for 10 min/wash, followed by four 5 min. washes in 1XSSC at room temperature. Finally, the arrays were briefly rinsed in distilled water, mounted and counterstained in the dark for 45 min. with DAPI (4,6-diamino-2-phenylindole).
Image capture and analysis
Array analysis was performed immediately after counterstaining using the GenoSensor scanner and software. This system generates a "genomic analysis report", indicating which chromosome regions in the array are involved in copy number changes, as well as a spreadsheet containing the data generated by a single experiment. In order to compare all the experiments, a database was created using the normalized, bias corrected, tumor/normal ratio value of each experiment [see additional file 1]. Since each spot in the array is present in triplicates, the median of the three spots of each probe in the array was calculated and its log2 transformed value was used for further analysis. A fluorescence ratio >1.25 (log2 = 0.32) was considered as a DNA gain, while DNA losses were scored when the ratio was <0.75 (log2 = -0.41). A ratio >2 (log2 = 1) was considered as a high copy number amplification.
Results
HPV detection and typing
One of the premalignant lesions was positive for HPV16 infection; one for HPV31 and the other for HPV58. In the invasive tumors, seven were positive for HPV16 and in three cases, we were not able to detect HPV sequences with the oligonucleotides we used for PCR amplification. As expected, CasKi and SiHa were positive for HPV16, while HeLa, INBL, CaLo, ViVo and RoVa were positive for HPV18.
Microarray comparative genomic hybridization
All our samples, except one pre-malignant lesion, presented alterations, ranking from 1/287 (alterations/total targets in the array) in a pre-malignant lesion to 175/287alterations in the cell line RoVa. We found almost twice the number of DNA gains than DNA losses (571 vs. 298) and the average number of copy number alterations (ANCA=total number of alterations in the sample collective/total number of cases) was 43.45 per case.
One of the pre-malignant lesions did not show any alteration, while amplifications at MSH2-KCNK12 (2p22.3-2p22.1), TCL1A (14q32.1) and TOP1 (20q12) were found in a second pre-malignant lesion and DMBT1 (10q25.3), ERBB2 (17q12), and 4qTEL11 (4qtel) amplification was found in the third sample from this group.
In the invasive tumors and the cell lines, the most common amplifications (58.8% of the samples) were found at the clones RBP1-RBP2 (retinol binding protein 1 and 2, 3q21-q22), present as a high copy number amplification (HCNA) in two samples; DAB2 (disabled homolog 2, mitogen-responsive phosphoprotein (Drosophila), 5p13; C84C11/T3 (5ptel) and D5S23 (5p15.2), followed by gains of Tp63 (3q27-q29, 2 HCNA); EGFR (Epidermal growth factor receptor, 7p12.3-p12.1, 4 HCNA) and D5S2064 (5p15.2), in 52.9% of the invasive samples and amplification of INS (Insulin, 11ptel) in 47% of the samples.
The most common deletion was found at the clone corresponding to the FHIT (Fragile histidine triad) gene (3p14.2), present in 47% of the invasive samples, followed by deletions at D8S504 (8p23.3), CTDP1-SHGC- 145820 (18qtel), KIT (4q11-q12), D1S427-FAF1 (1p32.3), D9S325 (9qtel), EIF4E (eukaryotic translation initiation factor 4E, 4q24), RB1 (13q14), and DXS7132 (Xq12) present in 29.4% of the samples. A histogram of the DNA copy number alterations detected in the tumor samples analyzed by array CGH is presented in figure 1. The results of these experiments can be accessed through the Progenetix CGH database [25].
Discussion
Previous studies using chromosomal CGH have delimited a specific pattern of chromosomal imbalances in cervical carcinoma. However, there is little knowledge regarding the identity of particular genes that might be the targets for these copy number changes, making microarray CGH an attractive method in order to define these particular gene targets.
There was concordance between the alterations detected by microarray CGH and the pattern of chromosomal alterations already described by chromosomal CGH. Although the array that we used did not cover the entire genome, we were able to detect alterations at particular genes and genetic markers that might be related to the transformation process in the cervical epithelium.
It is important to notice that the limited number of pre-malignant lesions analyzed did not allowed us to detect any particular region that might be related with this stage of the disease. However, an interesting candidate gene amplified in one pre-malignant sample and in 5 invasive tumors was MSH2-KCNK12 (2p22.3-2p22.1). This gene is the human homolog of the E. coli mismatch repair gene mutS, and has been found mutated in hereditary nonpolyposis colon cancer. Higher MSH2 expression has been described in cervical intraepithelial neoplasias and invasive cervical carcinomas than in non-neoplastic cervical lesions. An altered expression of this gene has also been proposed as an important event during cervical carcinogenesis [26,27]. Interestingly, the invasive samples showing MSH2 amplification presented with a high number of alterations (>40), suggesting a possible connection between increased copy number of this gene and chromosomal instability in invasive cervical carcinomas.
One of the most important genetic events during cervical carcinoma progression is the gain of 3q. This alteration has been detected in early stages of cervical transformation and in cooperation with other imbalances, seems to play an important role in tumor development. Microarray analysis identified among the most prevalent alterations in cervical tumors and cell lines the amplification of the RBP1- RBP2 (Retinol binding protein 1 and 2, 58.8%) and Tp63 (52.9% of the samples) genes, located at 3q21-q22 and 3q27-q29, respectively.
Tp63 is a homolog of the p53 tumor suppressor gene. Its protein is inactivated by the E6 HPV oncoprotein and plays a primordial role in the development of squamo-stratified epithelia. Tp63 is highly expressed in the basal stratum of these epithelia with diminished expression in the differentiated strata, suggesting that the presence of this protein preserves the self-renewal capacity of the epithelial stem cells after an asymmetric division, in which one of the daughter cell must conserve its epithelial stem-cell properties and the other daughter cell is committed to the differentiation process [28]. This protein has been detected in human cervical tissues in the basal and parabasal layers of the ectocervical squamous epithelium, and it is not present in the differentiated layers. In premalignant lesions and invasive squamous tumors, a strong p63 expression has been described [29,30]. The presence of this protein has also been associated with poor survival and locoregional failure after radiation and chemotherapy [31]. Expression of the epidermal growth factor receptor (EGFR, 7p12.3-p12.1), which was found amplified in 52.9% of the invasive tumors that we analyzed, was found to be a prognostic predictor of extrapelvic failure after treatment, and the expression of both molecules was found to be a very good risk factor measurement in patients with stage IIB squamous cell carcinoma of the uterine cervix, who had received radiotherapy and concurrent chemotherapy [31].
DAB2 on 5p13 was amplified in 58.8% of the invasive cases. The DAB2 gene has been identified as a potent tumor suppressor gene in prostate and ovarian carcinoma [32], and loss of expression of this gene has been associated with the transition of ovarian epithelial cells to premalignant states [33]. DAB2 has been implicated in cell positioning control and seems to mediate the requirement for basement membrane attachment of epithelial cells [34]. To our knowledge, there are no available reports analyzing the expression of this gene in the uterine cervix or in cervical carcinoma. The amplification of this gene seems contrary to its putative role as a tumor suppressor gene. A possible explanation for this observation might be the loss of one allele followed by the amplification of the remaining chromosome. However, since array CGH does not offer any type of information regarding the parental origin of the amplified chromosome, this situation can not be confirmed.
Detection of the TERC gene amplification has been recently proposed as a potential marker for the evaluation of cervical carcinoma progression [35]; however, we detected amplification of the clone representing this gene in less than 10% of the samples analyzed by CGH arrays. FISH analysis, as described by Heselmayer et al., detected a higher prevalence of nuclei with a diploid pattern than those with a tetraploid pattern, even in the high grade lesions. Furthermore, the percentage of nuclei with more than 2 copies of 3q, including the tetraploid cells ranged between 3.3 to 50% of the CIN3 (cervical intraepithelial neoplasia grade 3). Dellas et al., [9] used in situ hybridization to analyze the prevalence of 3q amplifications in cervical cancer tissue arrays, detecting low level amplifications in most of the tumors studied. These results suggest that these low copy number gains might not be adequately detected by chromosome or even array CGH, due to the contamination with normal cells and/or the presence of a high number of diploid or tetraploid cells in the sample.
Regarding DNA losses, the FHIT (fragile histidine triad, 3p14.2) gene suffered losses in 47% of the cases. Aberrant expression of this gene has been well documented in cervical carcinoma and has been related to lymph node metastasis, parametrial invasion, and vaginal involvement in invasive tumors [36]. An association between FHIT gene abnormalities and infection with particular HPV types has been suggested, since 87% of the cases with absent FHIT expression were positive for HPV16 infection [37]. Furthermore, abnormal expression of this gene has been found in significantly younger patients than those with normal expression, suggesting that abnormalities in the regulation of this gene might be accelerating carcinogenesis in cooperation with HPV [37]. These observations might be related to the preferential integration of HPV into fragile sites, particularly FRA3B, where FHIT is located [38].
Conclusion
In conclusion, microarray CGH allowed the detection of particular genes located in regions with common DNA copy number changes in cervical carcinoma. Further studies using CGH arrays with a higher resolution and the possibility to combine LOH with copy number changes, might be useful for the detection of gene specific targets that are relevant for the genesis and progression of cervical carcinoma.
Abbreviations
CGH: Comparative Genomic Hybridization, UCC: Uterine cervix carcinoma, HPV: Human papilloma virus, DAPI: 4,6-diamino-2-phenylindole.
Competing interests
The author(s) declare that they do not have any competing interests.
Authors' contributions
AH: Performed the microarray CGH experiments, data analysis and paper writing; MB: Help with data submission to the Progenetix database, data analysis; IP: provided training for the experiments, CGH data analysis; PP: tissue processing; GV: HPV typing; DH: sample collection; JG: provided access to the samples; ML: access to samples, histopathological analysis; RL: sample collection; CP: DNA extraction; JG: Help with data analysis; KV: DNA extraction; BA: HPV typing; MS: project coordinator.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Raw data from the CGH microarray experiments. This is text file containing the raw data from the microarrays, the first column denotes the identification of the clone, column 2 represents the cytogenetic position of the clone, sample data begins in column 3.
Click here for file
Acknowledgements
This work was partially funded through the 7114 and 34686 grants from the Mexican Council of Science and technology (CONACyT) and the Mexican Institute for Social Security (IMSS-FOFOI FP- 2001–2003). AH, GVO, CP, RL were recipients of scholarships from the CONACyT, IMSS and DGEP-UNAM. We would like to thank Abbott-Vysis for providing the CGH array system for this analysis. This work was submitted in partial fulfillment of the requirements for the Ph.D. degree of HA at the Ph.D. in Biomedical Sciences, National University of Mexico.
Figures and Tables
Figure 1 Histogram showing the incidence of alterations in the invasive tumors and cell lines in each of the targets printed on the CGH array. The incidence value is shown at the bottom of the figure; negative values indicate DNA losses, positive values DNA gains.
Table 1 Clinical stage and HPV status of the analyzed samples.
Sample Stage HPV
Pre-malignant
G60 CIN I 58
G67 CIN III 16
G42 CIN III 31
Invasive tumors
T56 Ib 16
T55 Ib 16
T46 Ib 16
T24 IIb Undetected
SXA IIb 16
LRL IIb 16
ASJ IIb 16
A69 IIIb Undetected
VGR IIb Undetected
T49 IVa 16
Cell lines
CaLo IIb 18
ViBo IIb 18
INBL IVa 18
RoVa IVa 18
HeLa 18
SiHa 16
CasKi 16
==== Refs
Walboomers J Jacobs M Manos M Bosch X Kummer A Shah K Snijders P Peto J Meijer C Muñoz N Human Papilloma Virus is a necessary cause of invasive cervical cancer worldwide J Pathol 1999 189 12 19 10451482 10.1002/(SICI)1096-9896(199909)189:1<12::AID-PATH431>3.0.CO;2-F
Bosch X Muñoz N The viral etiology of cervical cancer Virus Res 2002 89 183 190 12445658 10.1016/S0168-1702(02)00187-9
Münger K Howley P Human papillomavirus immortalization and transforming functions Virus Res 2002 89 213 228 12445661 10.1016/S0168-1702(02)00190-9
Ried T Hesselmeyer K Blegen H Schröck E Auer G Genomic changes defining the genesis, progression and malignancy in solid human tumors: a phenotype/genotype correlation Genes Chrom Cancer 1999 25 195 204 10379865 10.1002/(SICI)1098-2264(199907)25:3<195::AID-GCC1>3.0.CO;2-8
Pihan G Wallace J Zhou Y Doxsey S Centrosome abnormalities and chromosome instability occur together in pre-invasive carcinomas Cancer Res 2003 63 1398 1404 12649205
Cahill DP Kinzler KW Vogelstein B Lengauer C Genetic instability and darwinian selection in tumours Trends Cell Biol 1999 9 M57 M60 10611684 10.1016/S0962-8924(99)01661-X
Heselmeyer K Schröck E Du Manoir S Blegen H Shah K Steinbeck R Auer G Ried T Gain of chromosome 3q defines the transition from severe dysplasia to invasive carcinoma of the uterine cervix Proc Natl Acad Sci USA 1996 93 479 484 8552665 10.1073/pnas.93.1.479
Heselmeyer K Macville M Schröck E Blegen H Hellström A Shah K Auer G Ried T Advanced stage cervical carcinomas are defined by a recurrent pattern of chromosomal aberrations revealing high genetic instability and a consistent gain of chromosome arm 3q Genes Chrom Cancer 1997 19 233 240 9258658 10.1002/(SICI)1098-2264(199708)19:4<233::AID-GCC5>3.0.CO;2-Y
Dellas A Torhorst J Jiang F Proffit J Schultheiss E Holzgreve W Sauter G Mihatsch M Moch H Prognostic value of genomic alterations in invasive cervical squamous cell carcinoma stage IB detected by comparative genomic hybridization Cancer Res 1999 59 3475 3479 10416613
Kirchhoff M Rose H Petersen B Maahr J Gerdes T Lundsteen C Bryndorf T Kryger-Baggesen N Christensen L Engelholm S Philip J Comparative genomic hybridization reveals a recurrent pattern of chromosomal aberrations in severe dysplasia/carcinoma in situ of the cervix and in advanced-stage cervical carcinoma Genes Chrom Cancer 1999 24 144 150 9885981 10.1002/(SICI)1098-2264(199902)24:2<144::AID-GCC7>3.0.CO;2-9
Allen D White D Hutchins A Scurry J Tabrizi S Garland S Armes J Progressive genetic aberrations detected by comparative genomic hybridization in squamous cell carvical cancer Br J Cancer 2000 83 1659 1663 11104563 10.1054/bjoc.2000.1509
Kirchhoff M Rose H Petersen B Maahr J Gerdes T Philip J Lundsteen C Comparative genomic hybridization reveals non-random chromosomal aberrations in early preinvasive cervical lessions Cancer Genet Cytogenet 2001 129 47 51 11520566 10.1016/S0165-4608(01)00424-1
Umayahara K Numa F Suehiro Y Sakata A Nawata S Ogata H Suminami Y Sakamoto M Sasaki K Kato H Comparative Genomic Hybridization detects genetic alterations during early stages of cervical cancer progression Genes Chrom Cancer 2002 33 98 102 11746992 10.1002/gcc.1215
Harris CP Lu XY Narayan G Singh B Murty VV Rao PH Comprehensive molecular cytogenetic characterization of cervical cancer cell lines Genes Chrom Cancer 2003 36 233 241 12557223 10.1002/gcc.10158
Narayan G Pulido HA Koul S Lu XY Harris CP Yeh YA Vargas H Posso H Terry MB Gissmann L Schneider A Mansukhani M Genetic analysis identifies putative tumor supressor sites at 2q35-q36.1 and 2q36.3-q37.1 involved in cervical cancer progression Oncogene 2003 22 3489 3499 12776201 10.1038/sj.onc.1206432
Rao PH Arias-Pulido H Lu XY Harris CP Vargas H Zhang FF Narayan G Schneider A Terry MB Murty VV Chromosomal amplifications, 3q gain and deletions of 2q33-q37 are the frequent genetic changes in cervical carcinoma BMC Cancer 2004 4 5 15018632 10.1186/1471-2407-4-5
Solinas-Toldo S Dürst M Lichter P Specific chromosomal imbalances in human papillomavirus transfected cells during progression toward immortality Proc Natl Acad Sci USA 1997 94 3854 3859 9108068 10.1073/pnas.94.8.3854
Hidalgo A Schewe C Petersen S Salcedo M Gariglio P Schlüns K Dietel M Petersen I Human papilloma virus status and chromosomal imbalances in primary cervical carcinomas and tumor cell lines Eur J Cancer 2000 36 542 548 10717534 10.1016/S0959-8049(99)00323-8
Hidalgo A Monroy A Arana R Taja L Vázquez G Salcedo M Chromosomal imbalances in four new uterine cervix carcinoma derived cell lines BMC Cancer 2003 3 8 12659655 10.1186/1471-2407-3-8
Lichter P Joos S Bentz M Lampel S Comparative genomic hybridization: uses and limitations Semin Hematol 2000 37 348 357 11071357 10.1053/shem.2000.16594
Solinas-Toldo S Lampel S Stilgenbauer S Nickolenko J Benner A Dohner H Cremer T Lichter P Matrix-based comparative genomic hybridization: biochips to screen for genomic imbalances Genes Chrom Cancer 1997 20 399 407 9408757 10.1002/(SICI)1098-2264(199712)20:4<399::AID-GCC12>3.0.CO;2-I
Pinkel D Segraves R Sudar D Clark S Poole I Kowbel D Collins C Kuo WL Chen C Zhai Y Dairkee SH Ljung BM High resolution analysis of DNA copy number variation using comparative genomic hybridization to microarrays Nature Genet 1998 20 207 211 9771718 10.1038/2524
Pollack JR Perou CM Alizadeh AA Eisen MB Pergamenschikov A Williams CF Jeffrey SS Botstein D Brown PO Genome-wide analysis of DNA copy-number changes using cDNA microarrays Nature Genet 1999 23 41 46 10471496 10.1038/14385
Caceres-Cortes JR Alvarado-Moreno JA Waga K Rangel-Corona R Monroy-Garcia A Rocha-Zavaleta L Urdiales-Ramos J Weiss-Steider B Haman A Hugo P Brousseau R Hoang T Implication of tyrosine kinase receptor and steel factor in cell density-dependent growth in cervical cancer and leukemias Cancer Res 2001 61 6281 6289 11507083
Baudis M Cleary ML Progenetix.net: an online repository for molecular cytogenetic aberration data Bioinformatics 2001 17 1228 1229 11751233 10.1093/bioinformatics/17.12.1228
Giarnieri E Mancini R Pisani T Alderisio M Vecchione A Msh2, Mlh1, Fhit, p53, Bcl-2, and Bax expression in invasive and in situ squamous cell carcinoma of the uterine cervix Clin Cancer Res 2000 6 3600 3606 10999751
Kwasniewska A Gozdzicka-Jozefiak A Postawski K Miturski R Evaluation of DNA mismatch repair system in cervical dysplasias and invasive carcinomas related to HPV infection Eur J Gynaecol Oncol 2002 23 231 235 12094961
Yang A Schweitzer R Sun D Kaghad M Walker N Bronson RT Tabin C Sharpe A Caput D Crum C McKeon F p63 is essential for regenerative proliferation in limb, craniofacial and epithelial development Nature 1999 398 714 718 10227294 10.1038/19539
Quade BJ Yang A Wang Y Sun D Park J Sheets EE Cviko A Federschneider JM Peters R McKeon FD Crum CP Expression of the p53 homologue p63 in early cervical neoplasia Gynecol Oncol 2001 80 24 29 11136565 10.1006/gyno.2000.5953
Wang T Chen B Yang Y Chen H Wang Y Cviko A Quade B Sun D Yang A Mckeon F Crum C Histologic and immunophenotypic classification of cervical carcinomas by expression of the p53 homologue p63: A study of 250 cases Hum Pathol 2001 32 479 486 11381365 10.1053/hupa.2001.24324
Cho NH Kim YB Park TK Kim GE Park K Song KJ P63 and EGFR as prognostic predictors in stage IIB radiation-treated cervical squamous cell carcinoma Gynecol Oncol 2003 91 346 353 14599865 10.1016/S0090-8258(03)00504-3
Chen H Toyooka S Gazdar AF Hsieh JT Epigenetic regulation of a novel tumor suppressor gene (hDAB2IP) in prostate cancer cell lines J Biol Chem 2003 278 3121 3130 12446720 10.1074/jbc.M208230200
Yang DH Smith ER Cohen C Wu H Patriotis C Godwin AK Hamilton TC Xu XX Molecular events associated with dysplastic morphologic transformation and initiation of ovarian tumorigenicity Cancer 2002 94 2380 2392 12015763 10.1002/cncr.10497
Sheng Z Sun W Smith E Cohen C Sheng Z Xu XX Restoration of positioning control following Disabled-2 expression in ovarian and breast tumor cells Oncogene 2000 19 4847 4854 11039902 10.1038/sj.onc.1203853
Huang LW Chao SL Chen TJ Reduced Fhit expression in cervical carcinoma: correlation with tumor progression and poor prognosis Gynecol Oncol 2003 90 331 337 12893195 10.1016/S0090-8258(03)00318-4
Butler D Collins C Mabruk M Leader MB Kay EW Loss of Fhit expression as a potential marker of malignant progression in preinvasive squamous cervical cancer Gynecol Oncol 2002 86 144 149 12144820 10.1006/gyno.2002.6712
Takizawa S Nakagawa S Nakagawa K Yasugi T Fujii T Kugu K Yano T Yoshikawa H Taketani Y Abnormal Fhit expression is an independent poor prognostic factor for cervical cancer Br J Cancer 2003 88 1213 1216 12698186 10.1038/sj.bjc.6600892
Thorland EC Myers SL Gostout BS Smith DI Common fragile sites are preferential targets for HPV16 integrations in cervical tumors Oncogene 2003 22 1225 1237 12606949 10.1038/sj.onc.1206170
|
16004614
|
PMC1186020
|
CC BY
|
2021-01-04 16:03:05
|
no
|
BMC Cancer. 2005 Jul 9; 5:77
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-77
|
oa_comm
|
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-911595523810.1186/1471-2164-6-91Methodology ArticleEvaluation of the similarity of gene expression data estimated with SAGE and Affymetrix GeneChips van Ruissen Fred [email protected] Jan M [email protected] Gerben J [email protected] Lida [email protected] Danny A [email protected] Marcel [email protected] Frank [email protected] Department of Neurogenetics, Academic Medical Center, University of Amsterdam, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands2 Department of Anatomy and Embryology, Academic Medical Center, University of Amsterdam, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands3 Department of Human Genetics, Academic Medical Center, University of Amsterdam, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands2005 14 6 2005 6 91 91 28 10 2004 14 6 2005 Copyright © 2005 Ruissen et al; licensee BioMed Central Ltd.2005Ruissen 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
Serial Analysis of Gene Expression (SAGE) and microarrays have found awidespread application, but much ambiguity exists regarding the evaluation of these technologies. Cross-platform utilization of gene expression data from the SAGE and microarray technology could reduce the need for duplicate experiments and facilitate a more extensive exchange of data within the research community. This requires a measure for the correspondence of the different gene expression platforms. To date, a number of cross-platform evaluations (including a few studies using SAGE and Affymetrix GeneChips) have been conducted showing a variable, but overall low, concordance. This study evaluates these overall measures and introduces the between-ratio difference as a concordance measure pergene.
Results
In this study, gene expression measurements of Unigene clusters represented by both Affymetrix GeneChips HG-U133A and SAGE were compared using two independent RNA samples. After matching of the data sets the final comparison contains a small data set of 1094 unique Unigene clusters, which is unbiased with respect to expression level. Different overall correlation approaches, like Up/Down classification, contingency tables and correlation coefficients were used to compare both platforms. In addition, we introduce a novel approach to compare two platforms based on the calculation of differences between expression ratios observed in each platform for each individual transcript. This approach results in a concordance measure per gene (with statistical probability value), as opposed to the commonly used overall concordance measures between platforms.
Conclusion
We can conclude that intra-platform correlations are generally good, but that overall agreement between the two platforms is modest. This might be due to the binomially distributed sampling variation in SAGE tag counts, SAGE annotation errors and the intensity variation between probe sets of a single gene in Affymetrix GeneChips. We cannot identify or advice which platform performs better since both have their (dis)-advantages. Therefore it is strongly recommended to perform follow-up studies of interesting genes using additional techniques. The newly introduced between-ratio difference is a filtering-independent measure for between-platform concordance. Moreover, the between-ratio difference per gene can be used to detect transcripts with similar regulation on both platforms.
==== Body
Background
Methods for the analysis of gene expression profiles have gone through progressive development over recent years. Traditionally, the level of transcribed mRNA has been analyzed using methods such as Northern blots, quantitative RT-PCR, differential display [1,2], representational difference analysis [3], total gene expression analysis [4] and suppressive subtractive hybridization [5,6]. All these methods, although fruitful and still in use, have a limited scope with regard to the number of genes that can be analyzed simultaneously. Because of this limitation, new methods have been developed, including serial analysis of gene expression (SAGE) [7], massive parallel signature sequencing (MPSS) [8], cDNA and oligo microarray chip technologies [9-13] and Affymetrix GeneChips [11].
SAGE is based on the high-throughput sequencing of concatemers of short (13–14 bp; recently 21–25 bp) sequence tags that originate from a known position within a transcript and therefore theoretically contain sufficient information to identify a transcript [7]. In contrast to microarrays, SAGE estimates the abundances (expression levels) of thousands of transcripts without prior knowledge of the transcripts being expressed. The proportion of the tag within the total number of tags in the library gives a direct estimate of the abundance of the transcript within a biological sample. The advantage of the SAGE technique is that it performs a random sampling from the pool of all expressed transcripts (also called a transcriptome) allowing the discovery of new transcripts. The proportional nature of the data enables easy exchange among researchers thus allowing the creation of large public SAGE data sets for numerous human tissues, both normal and diseased [14,15]. Disadvantages of SAGE are that the technique is expensive, labor-intensive and prone to sequencing errors. Moreover, the annotation of the short 10 bp sequence tags may identify more than one transcript. This can be overcome by using LongSAGE libraries that contain 17 bp tags which can be more reliably mapped to Unigene clusters or the complete genome sequence [16]. However, SAGE is not suitable for high-throughput analyses of multiple samples.
In contrast to SAGE, DNA microarrays are used to measure relative expression levels between samples of thousands of known transcripts. Currently, three array variants are being used (for reviews see [17,18]) i.e. spotted cDNA microarrays, spotted oligonucleotide microarrays and synthesized oligonucleotide microarrays (Affymetrix GeneChips). The advantages of Affymetrix GeneChips are that they are highly sensitive enabling the detection of mRNAs present at levels as low as 1 transcript in 100000 [11] when the probe labeling step is not considered [19]. They are suitable for high-throughput analyses of multiple samples, and data can easily be shared and used for comparisons with other researchers using the same chips. Disadvantages of Affymetrix GeneChips are that they are only commercially available, are costly and require expensive specialized equipment and are inflexible in design (although custom design is possible at high cost). Furthermore, GeneChips only measure the expression of genes represented on the chip in contrast to SAGE, in which the expression profile of the complete transcriptome can be mapped.
At present, SAGE, oligo microarrays, cDNA microarrays and Affymetrix GeneChips are the most widely used techniques for determining gene expression levels and gene expression ratios in different disease states and in cells under different physiological conditions or environmental stimuli. Often these different gene-expression profiling platforms are being used in parallel and data generated with the different techniques need to be compared, and possibly interchanged, within and between laboratories. Due to the overall difference in platform design, transcript level estimation, and gene annotation, direct comparisons are difficult and only a few attempts have been made to compare these different platforms (Figure 6). To determine the overall correspondence between expression levels or expression ratios of two different platforms several methods have been used in literature (Figure 1A,B and 1C). These include the parametric (Pearson) or non-parametric (Spearman) correlation coefficients between platforms, and contingency tables with varying numbers of classes for each platform. For the latter a correspondence measure can be calculated as the percentage of transcripts falling in the cells on the diagonal (Figure 1B). An extreme form of the contingency table has only 2 classes per platform (ratios above and ratios below 1) and therefore only 4 cells. This form of concordance estimation is dubbed "Up/Down classification" (Figure 1A). None of these correspondence measures was deemed satisfactory because they either treat very different ratios as similar (points A and B in figure 1A). This, in our view, makes the Up/Down classification very unreliable as an agreement measure. The use of contingency tables with more classes is already a better approach, but still some genes will be considered to be "in disagreement" while they have nearly corresponding expression ratios (points A and B in figure 1B). The Pearson correlation coefficient is a measure for the fraction of variation in Y that is explained by the variation in X, and as such, only gives a measure for the tendency of the plotted points to increase simultaneously (solid line, Figure 1C). Because of the large number of points, a slight linear regression of Y on X will give a highly significant correlation coefficient. However, when studying the correspondence between gene expression platforms, the expected linear relation has a slope of 1, when the results of both platforms are in complete correspondence (dashed line, Figure 1C), and the deviation of the observed scatter plot from this expected relation should be tested. Neither the linear Pearson, nor the Spearman rank correlation coefficient is suited for such a test. Although the fit of the point cloud to the Y = X relation can be easily calculated, the resulting statistic would still only provide a goodness of fit measure for the whole data set without giving any information on the correspondence per gene. To remedy these pitfalls we will introduce a correspondence measure based on the difference between the log(ratio) values in the two platforms for each individual transcript. Apart from serving as the basis for a measure for overall platform concordance, this method also provides the user with an agreement measure for each individual transcript which is of more interest than the overall correlation.
Figure 1 Illustration of the methods used for the comparison of expression profiles from different platforms. A: Up/Down classification: The points A and B with very different ratios are both considered to reflect a common tendency; B: contingency table diagonal: The points A and B, with very similar ratios, end up in different classes; C: correlation coefficients: The solid line fits to the point cloud which has a significant correlation coefficient between X and Y. However, the dashed line (Y = X) is the expected line when both platforms show identical expression patterns.
In the current study we have determined the similarity between SAGE- and Affymetrix GeneChips-generated gene expression profiles of two independent RNA samples. One RNA sample is isolated from a Wilms' tumor; the other is the Stratagene Universal reference RNA. These expression data were then used to evaluate the annotation problems when comparing different gene profiling platforms and the methods that can be used to compare two different platforms with respect to individual gene expression measurements and with respect to between-sample gene expression ratios. Finally, it is demonstrated that the between-ratio difference can be applied to select those transcripts that display similar expression changes in both platforms.
Results
SAGE data analysis
In order to compare SAGE with other gene expression profiling techniques we created a SAGE library with 69792 tags from a Wilms' tumor sample. SAGE data (51954 tags) for the Stratagene Universal reference RNA (GSM1734;[20]) were obtained from the NCBI website. All tag counts are after removal of duplicate dimers and linker sequences. Within the SAGE libraries we could identify 25052 and 17497 unique SAGE 10 bp tags, for the Wilms tumor sample and the Stratagene sample, respectively. Tags can be divided into tags with low abundance (1–5 tags per 100000), intermediate abundance (6–50 tags per 100000), and high abundance (more than 50 tags per 100000). In each of the libraries, these categories contained on average 84%, 15% and 1% of the total number of unique tags (Data not shown). In addition, we created a LongSAGE library of the Wilms tumor sample for annotation purposes (as described below) and not for the comparison with Affymetrix GeneChips. This library could be used as a technical replicate of the 'short' SAGE library. Comparison of the SAGE and LongSAGE libraries showed a Pearson Correlation coefficient of 0.651 (P < 0.01) and using Z-test statistics [21] the two libraries only differed significantly from each other in 3% (α = 0.05) or 0.6% (α = 0.001) of the tags (Figure 2A). The observed differences in the LongSAGE library versus the normal SAGE library might be due to treatment with different linkers, tagging enzyme (MmeI instead of BsmF1) and elimination of a blunt-end ligation. The pattern of variation in figure 2A closely resembles the variation predicted by the binomial distribution [22] of SAGE tag counts with only the 3% significantly different tag counts (blue dots; α = 0.05), falling outside the range of critical values. Overall SAGE and LongSAGE give identical results.
Figure 2 Evaluation of gene expression in Wilms' tumor tissue. The comparison of SAGE and Affymetrix duplicate samples demonstrates the reproducibility in both platforms (A, B). In addition, gene expression was compared between platforms (C) and showed a wide range of variation. The frequency distributions of gene expression values the final data sets do not differ from the total distributions (D). A: Comparison of a SAGE versus a LongSAGE library of the same sample Blue dots represent tag counts that are significantly different between the two libraries (according to the Z-test, Kal et al. 1999), green and red dots represent tag counts that do not differ between libraries. The red spots are tag counts that do not significantly differ from tag count 0 within the specified library (See also Table 4). B: Comparison of a duplicate analysis of one Wilms tumor sample using Affymetrix HG-U133A GeneChips. Gray spots represent probe sets that have an absent call. C: Comparison between SAGE and Affymetrix GeneChips for the Wilms' tumor sample. Red spots represent the total matching data set (n = 6408) and black spot represent the final selection (n = 1094). D: Frequency distribution of the Affymetrix intensity and SAGE tag counts from the final matched data set (1094 Unigene clusters) and the total matching data set. The smoothed line represents the distributions of the total data set in each platform. For both Affymetrix (classes with an intensity width of 10) and SAGE (classes based on tag counts) the distributions of the final data set and the total data set do not differ from each other (Chi-square values of 327 (df = 323; P = 0.412) and 104 (df = 105; P = 0.506), respectively).
Microarray analyses
Microarray experiments were performed using Wilms' tumor RNA and the Stratagene Reference RNA. Results of biological replicas of each sample, with independent cRNA synthesis and hybridizations, showed a good reproducibility (Pearson correlation coefficients of 0.982 (n = 11938) and 0.979 (n = 10489);both P < 0.01) using intensity values for all probe sets with a "present" signal (on average 54%; absent = 44% and marginal = 2%) (Figure 2B; black spots). This indicates that two identical RNA samples perform very similar within the pre-processing and final hybridization reactions. Although, in contrast to SAGE, the intensity signals on the array do not represent the actual abundance of mRNA molecules, we classified the Affymetrix data to get an impression of the signal distribution. These distributions are similar to those of the SAGE data. The majority (~90%) of the probe sets showed low signal intensity.
Annotation problems
In the comparison of data obtained by SAGE and Affymetrix GeneChips only reliably annotated tags can be included (as described in the 'Matching of platforms' paragraph of the Material and Methods section; see also Shippy et al.[23]). Annotation of SAGE tags to genes and their corresponding Unigene cluster numbers revealed that on average 30% of all tags (including low abundant tags) could be reliably annotated based on the SAGE Genie principles [24]. Annotation improves to an average of 70% for tags that have an intermediate to abundant expression level. The remainder of the tags could not reliably be associated with a gene or Unigene cluster because they were not available through the SAGE Genie site, annotated to unclustered ESTs, or their reliability was below 67% (according to the SAGE Genie principles). Additionally, we performed LongSAGE for the Wilms' tumor sample, which allows the identification of 17 bp tags instead of 10 bp tags. Theoretically, over 99.8% of the 17 bp tags are expected to occur only once in the human genome. However, analyses based on actual sequences have demonstrated that only 75% of the 17 bp tags occur only once in the human genome, with the remaining tags matching duplicated genes or repeated sequences [16]. Complete annotation of LongSAGE tags using SAGE Genie data and principles revealed that 28% of all tags could be assigned a reliable Unigene cluster. Similar to SAGE, the annotation improves to approximately 70% for tags that have an intermediate to abundant expression level.
The Affymetrix HG-U133A GeneChips contained probe sets for 13727 Unigene clusters that could be identified, whereas eight percent of the probe sets (i.e. 1795 probe sets) could not be linked to a Unigene cluster because these sequences are withdrawn or because these sequences are currently under revision. Figure 3 gives a schematic representation of the matching of SAGE and Affymetrix HG-U133A GeneChips data with additional information about the number of Unigene clusters within each platform, number of unambiguous Unigene clusters in each comparison and the Unigene clusters included in the final comparison. This final comparison contains 13% of the SAGE Unigene clusters and 8% of the Affymetrix Unigene clusters. These data represent 32% of the unambiguous Unigene clusters. Because of the above-mentioned problems and restrictions, only 1094 tags and probe sets were uniquely matched to the same Unigene clusters and were 'present' in both tissue samples and platforms. This relatively low number underscores the major problem in "how to merge different expression platforms". However, in view of the following quantitative comparison of gene expression platforms it is important to note that a comparison of frequency distributions of all clusters and of the selected clusters showed that the final selection of 1094 Unigene clusters does not represent a biased sample neither for the SAGE tag counts, nor for the Affymetrix array intensities. This is illustrated in figure 2D in which the frequency distributions are given for Affymetrix intensities and SAGE tag counts from the final data set of 1094 Unigene clusters. The smoothed line, which represents the frequency distribution of all SAGE tag counts and all Affymetrix intensity data (only present calls), does not differ from the distribution of the subset included in the comparison of the two platforms.
Figure 3 Flow chart for matching data from two gene expression platforms. SAGE tags were converted into Unigene clusters using data from the CGAP website. Accession numbers from Affymetrix GeneChips were also converted to their corresponding Unigene cluster. Platforms are matched according to their Unigene cluster and only unambiguous Unigene clusters are selected. Finally, data are filtered for tag counts >0 and present calls on microarray platforms. 1. In the complete process of annotation a large number of tags or probe sets lost due to the following reasons: SAGE: 11733 tags with no annotation, 13113 tags with no reliable annotation, 913 tags with multiple Unigene Clusters, 80 tags belonging to linker sequences, 20 tags belonging to repetitive sequences, 22 tags belonging to mitochondrial DNA; Affymetrix: 1795 Probe sets no longer belong to a Unigene Cluster (Build 160). The remaining 20488 probe sets represent 13727 unique Unigene clusters. 2. Unambiguous Unigene clusters refer to those clusters that occur only once within each platform.
Comparison of gene expression levels
In the comparison of platforms, we first analyzed the similarity of gene expression levels between SAGE and Affymetrix data in one tissue sample. Both datasets were matched according to their Unigene cluster numbers. Figure 2C shows a scatter plot of SAGE and Affymetrix gene expression values of the 6408 Unigene clusters before exclusion of ambiguous matches (red spots). For multiple matches, the highest tag count or intensity value per cluster was plotted. In this scatter plot the black spots represent the final selection of 1094 unambiguous and filtered Unigene clusters. Note that high Affymetrix expression levels are observed for low SAGE tag counts (spots in top-left quadrant of figure 2C), but that no high tag counts are found for low Affymetrix data (few spots in bottom-right quadrant). Overall, the correlation between SAGE tag counts and Affymetrix intensity levels of the 6408 matching Unigene clusters seemed to be modest. This was confirmed by mapping the distribution of the top 100 highly expressed genes in SAGE in the distribution of the Affymetrix dataset, and vice versa (Data not shown, but this can be inferred from figure 2C). In both comparisons, only halve of the genes from the top 100 of one platform have a rank in the top 100 of the other platform, whereas approximately 10% are matched to genes with ranks of over 1000 in the other platform. This already shows that the correlation of expression levels between platforms is modest.
Comparison of between-sample expression ratios
In most gene expression studies, alterations of expression levels are expressed in relation to the simultaneously determined expression level of a reference sample and conclusions are drawn based on these ratios. To this end, expression ratios were calculated between the reference RNA and the Wilms' tumor data for the SAGE tag counts as well as for Affymetrix HG-U133A GeneChips spot intensities. In this comparison the final data set containing only the between-sample ratios for unambiguous transcripts was used (Figure 3), allowing effective comparison of the two platforms.
To enable direct comparisons of ratio measurements using different gene expression platforms, the ratios of the Affymetrix platform were scaled to those of the SAGE platform as described in Figure 4 ("scaling of two platforms"). In addition, different approaches were used to describe the correlation of the resulting scaled gene expression ratios between platforms (Figure 5). For the comparison of gene expression ratios based on contingency tables we used two approaches, i.e. Up/Down classification (Figure 5A) and a contingency table diagonal based on intensity classes (Figure 5B). These comparisons lead to an agreement of 63% and 76% between platforms, respectively. Furthermore, the Pearson correlation coefficient, calculated as a measure for the agreement between platforms, was 0.453 (P < 0.01). Regression analysis shows a linear trend with a slope of 0.477 for Affymetrix versus SAGE, which according to the correlation coefficient differs significantly from a slope of 0. However, this slope also deviates significantly from the slope value of 1 which is expected when the platforms are identical (t-test for slopes; P <0.001; Figure 5C). Finally, we compared SAGE and Affymetrix data using our proposed classification based on the difference between the two ratios per Unigene cluster. When we accept a 0 to 3-fold difference as indicative for agreement between the two platforms (red points in figure 5D), this approach showed that the two platforms have an agreement of 78%.
Figure 4 Scaling of two gene expression profiling platforms. Illustration of the steps involved in the scaling of values in each of the platforms to a common scale. The procedure takes the ratio distribution in one of the platforms and scales the other to match the same range of ratio values using a quadratic equation based on ratio 1 and the 10th and 90th percentile values of each platform. The (scaled) ratio values are then used to calculate between-platform ratio differences per transcript. In addition, it is demonstrated how the ratio differences can be used to calculate the standardized between-platform log(ratio) difference and a probability value. For further details: see the Materials and Methods section.
Figure 5 Comparison of SAGE and Affymetrix HG-U133A GeneChips results using the scaled ratios between Wilms' tumor and Stratagene Universal Reference RNA expression levels. A: Up/Down classification. The red points in the upper-right and lower-left were considered to be in agreement between the platforms. B: contingency table diagonal based on the classification of gene expression ratios into log (10-fold) classes. The genes falling in the classes on the diagonal were considered to be in agreement between the platforms. C: Pearson correlation coefficient. The correlation coefficient was 0.472 and corresponds to a linear regression line with a slope of 0.492 (solid line) The Y = X line with a slope of 1 (dashed line) is the expected line when both platforms have identical expression patterns. D: absolute between-platform ratio differences (see Figure 4) were calculated and classified: 0–0.5 (red), 0.5–1.0 (green), 1.0–1.5 (blue), 1.5–2.0 (magenta), 2.0–2.5 (light blue). These classes represent an approximate less then 3, 10, 30, 100, and 300-fold difference, respectively, between the two platforms. The points in the 0.5 zone were considered to be in agreement between the platforms.
Figure 6 Literature overview of platform comparisons
Like others have demonstrated (Figure 6) the overall agreement between platforms improves when only highly expressed transcripts (based on their tag counts) are included (Table 1). When only lowly expressed genes were included the concordance based on the contingency table diagonal and correlation coefficient steeply decreased whereas the other measures were both hardly affected. Inclusion of only those tags that were significantly differentially expressed between the two samples markedly improved the Up/Down classification and correlation coefficient based measures. Note that the concordance measure based on the between-ratio difference was least affected by these selections. This indicates that this new measure is robust and less dependent on filtering than the other overall measures.
Table 1 Summary of similarities between SAGE and Affymetrix HG-U133GeneChips for the final dataset (= 1094)
UP/DOWN classification Contingency table diagonal Pearson Correlation coefficient3 0–3 fold between-ratio difference N
All transcripts 63% 76% 0.453 78% 1094
Low expresssion1 57% 81% 0.222 78% 572
High expresssion1 69% 81% 0.578 90% 226
Significant difference2 86% 47% 0.636 70% 167
1. Based on the binomial sampling error of SAGE tags, tags counts below 5.7 and 7.7 (per 100,000) for the WT and Stratagene sample, respectively, are not significantly different from tag count 0. When a tag falls below these thresholds in both libraries it is included in the "Low expression" group (line 2); when a tag counts is above these thresholds in both libraries it is included in the "High expression" group (line 3). The thresholds were calculated as the 95% confidence interval of the tag proportion: CI95%) = n ± 1.96* with n = tag count; N = Library size and p = n/N (proportion)
2.Significant difference between the two SAGE libraries is defined as a significant P-value (α<0.05) according to the Z-test between two libraries [21].
3.All observed correlation coefficients are significant at P < 0.01
Sources of differences in gene expression ratios
In an attempt to explain the difference in gene expression between SAGE and Affymetrix GeneChips we summarize different sources. Variation due to "noisy fold ratios" generated from low-intensity transcripts is a widespread cause of error when computing statistics on ratios without accounting for the intensities from which the ratios were derived [25]. Within our data set we have shown that the final data set is an unbiased selection of the total data set (Figure 2D). Additionally, the mean intensity signals for both SAGE and Affymetrix GeneChips appear to be randomly distributed over the ratio distribution (data not shown). This indicates that the difference in expression ratios between platforms is not caused by low intensity values.
In addition, it has been suggested that the GC-content of the transcripts could influence the correspondence between platforms [26]. To test this hypothesis for the final data set (n = 1094) we retrieved all transcript sequences (mostly Refseq sequences [27]) and probe set sequences and calculated the GC-content for each transcript and the average GC-content of the corresponding probe sets. The GC-contents were divided into classes (30–35%; 35–40%; 45–50% etc.) and the correlation between GC-content and the differences in expression ratios between platforms was tested. Statistical analysis showed that ratio differences did not depend on the GC-content of the transcript (Chi-square value of 25.69; df = 35; P = 0.875). However, Unigene clusters showing good agreement between platforms tend to depend on the high GC-content of the corresponding probe sets (Chi-square value of 61.114; df = 30; P = 0.001). This GC-analysis indicates that expression data from probe sets with a higher GCcontent show a better agreement with their corresponding SAGE data and are more reliable. Note in this respect that for a Unigene cluster the GC content of a probe set is not necessarily the same as that of a transcript.
Discussion
To answer the question whether gene expression data generated by SAGE and by Affymetrix HG-U133A GeneChips can be used interchangeably, data from these two techniques were compared using two independent RNA samples. Analysis of intra-platform variation shows good correlation for both SAGE and Affymetrix; this is also observed by others (see Figure 6). The inter-platform comparison depends on reliable annotation of the SAGE tags for which we used the tag annotation from SAGE Genie [24]. A reliable association could be made only for 30% of all tags, which increases to 70% for intermediate and high abundant tags. This indicated that SAGE tag annotation requires improvement, especially for low abundant tags. Based on literature findings, the use of LongSAGE should refine annotation of SAGE tags [16]. However, the current study showed an annotation profile similar to the above-mentioned percentages, indicating that LongSAGE is still not sufficient for unique gene identification. Similar disappointing improvements in annotation efficiency have been found in other studies [19]. Further comparison of SAGE and LongSAGE requires a study that falls beyond the scope of this paper; such a study has recently been published [19]. For the annotation in Affymetrix GeneChips, accession numbers had to be converted to Unigene clusters, which was hampered by the fact that 8% of the transcripts present on Affymetrix HG-U133A GeneChips were no longer present in a Unigene cluster. Moreover, some probe sets might represent a different transcript than initially reported (see for an example [28]).
A first impression about the agreement between SAGE and Affymetrix HG-U133A GeneChips was obtained from the evaluation of the top100 of highly abundant transcripts in one RNA sample in each platform. This comparison showed that approximately 50% of the top100 of highly expressed transcripts showed a corresponding expression within the top100 of highly expressed transcripts of the other platform. This is in line with the findings of Ishii et al. [29] who compared SAGE with Affymetrix GeneChips containing approximately 6000 transcripts, and Iacobuzio-Donahue et al. [30] who showed that only genes that display robust changes in gene expression were identified by both platforms. In our current study, approximately 80% of transcripts detected in the top100 of one platform were mapped within the top1000 of the competing platform. A similar figure was presented by Evans et al. [31] who used the RG-U34A Affymetrix GeneChips. Recently, Kim [32] suggested that absolute expression analyses of SAGE and oligonucleotide microarray technology reliably detected medium-to-high abundant transcripts.
For a more extensive comparison between the individual gene expression profiling platforms we used gene expression ratios between Wilms' tumor and Stratagene Universal reference RNA as determined by SAGE and Affymetrix GeneChips. The use of ratios might have the disadvantage of losing information about individual expression values. However, it corrects for platform specific variations (i.e. probe design, hybridization efficiencies etc.). By matching SAGE and Affymetrix data, an unambiguous data set was generated. On average about 30% of the unambiguous genes were observed to be expressed by both SAGE and Affymetrix GeneChips and could be included in the final comparison. Although this comparison comprised only 13% of all SAGE Unigene clusters and only 8% of the Affymetrix Unigene clusters, it was demonstrated that this selection was unbiased with respect to gene expression levels in each of the platforms. This allows the extrapolation of the conclusions to the whole platform.
We looked for the correspondence in gene expression results between the two techniques using Up/Down classification (Figure 1A), the contingency table diagonal (Figure 1B) and correlation coefficients (Figure 1C). In addition, an approach was introduced in which differences between scaled ratios were calculated. The latter measure was introduced to circumvent pitfalls of Up/Down classification, contingency tables and correlation coefficient that were discussed in the background section. To this end, we introduced an approach in which the scaling of the ratio data enables the calculation of individual ratio differences between platforms. These ratio differences can then be used to determine to which extend and in which range (e.g. 0–3 fold difference) two platforms differ in their expression ratio estimation. In this study we show that, as opposed to the other overall concordance measures, the between-ratio difference is hardly sensitive to filtering of noisy data. From the current analysis, we conclude that contingency tables and, preferably, calculation of ratio differences between two platforms should be used to compare gene expression profiles from different platforms. Moreover, the between-ratio difference provides the user with a correspondence measure per individual gene that can be used to select those genes for which a predetermined correspondence level is reached. The approximately normal distribution of the between-ratio differences (Figure 4) allows the calculation of a standardized difference value for each gene from which a P-value can be obtained. Note that this P-value cannot be used to test whether the ratio difference equals zero. Such a test requires a gene specific variance estimate in the denominator of the standardized difference and such a variance estimate cannot be obtained from the four non-replicated expression values that are used to calculate the ratio difference. However, the standardized difference and its P-value can be used as a measure for the position of a specific gene within the distribution of between-platform ratio differences and as such they can serve as a statistical threshold to determine which genes can be confidently interchanged between platforms. For instance, in the current study, the transcripts with a less than 0.5 fold between-ratio difference (red dots in figure 5D) have a chance of at least 0.8 that they show similar gene expression on both platforms. Some of the choices in the scaling procedure can be considered to be ad-hoc. However, given the current state of understanding of the causes for within and between platform variability it was deemed best to opt for a simple quadratic scaling equation to convert the distribution of ratios, which is asymmetric around 1 to a common scale. When the knowledge on the physics, chemistry, and sampling statistics increases, better conversion functions will present themselves.
The overall similarity between SAGE and Affymetrix GeneChips is modest when expression ratios are compared. The correspondence improves to 90% when only highly expressed transcripts are included which means that noise is filtered out for both platforms. The differences between SAGE and Affymetrix GeneChips were not caused by a biased selection of the final data set, differences in GC-content of the included transcripts or extreme ratios resulting from low gene expression values. The observed cross-platform differences, arise from intrinsic properties of the platforms themselves, differences in the principle of determining the expression levels, such as absolute (SAGE) versus quantitative (microarray) mRNA levels, and/or processing and analytical evaluation [33]. These disparities of the two technical approaches are summarized in table 2 and may all contribute to the modest overall correlation of SAGE and microarray data. We cannot conclude which of the platforms performs best. These results show, as also argued by Tan and co-workers [33], that it is important to validate the results obtained with SAGE or Affymetrix GeneChips with subsequent northern blots or quantitative PCR analysis [34-36]. It was beyond the scope of our analysis to perform such a verification of expression data. Anyway, such a validation is impractical for large numbers of genes. However, it seems that the divergence of the SAGE and Affymetrix platforms in this study is for a large part due to the wide range of Affymetrix gene expression values observed for transcripts with a low gene expression level in SAGE (Figure 2C). A similar over-representation of high Affymetrix expressions for low SAGE tag counts has been published by Lu et al. [19]. We currently showed that a SAGE and LongSAGE library from the same RNA sample showed nearly identical expression profiles (Figure 2A). These findings confirm the results found within direct comparisons of SAGE libraries [37-39]. In addition, the differences between SAGE and LongSAGE can be fully explained by the binomial distribution of the sampling error in individual SAGE tag counts [22]. Therefore, it can be ruled out that many low SAGE tag counts originate from high abundant transcripts. This is also confirmed by Sun et al. who demonstrate that 70% of the low-copy SAGE tags represent real low level transcripts [40]. The Affymetrix platform showed highly reproducible intensity values when applied twice to the same tissue sample. However, because of the variation between probe sets per Unigene cluster [25] it cannot beruled out that some Affymetrix probe sets provide systematically biased intensity levels and expression ratios. It is a known problem that different probe sets belonging to the same transcript show variation in expression detection. Several explanations have been given for this variation: (1) probe sets may represent splice variants or may cross-hybridize to different members that belong to a highly similar gene family or transcripts with different poly-A sites; (2) one probe set is more 5' located than the other and (3) one probe set is better designed than the other [41]. Such a bias might explain the weak correspondence between the SAGE and Affymetrix platform observed in this and other studies [19,23,25]
Future studies should be aimed on improving the efficiency of SAGE tag annotation and avoidance of systematic bias in microarray techniques. Only then, measurements of various technologies can be directly compared and transformed to a universal gene expression catalogue. SAGE has the advantage that a whole transcriptome is analyzed, but is limited to the analysis of a small number of samples. For screening of large sets of samples SAGE cannot be the favored choice and Affymetrix GeneChips might be a good alternative. Therefore, we think that the future lies in combining the data from SAGE with Affymetrix GeneChips, custom cDNA or oligo arrays. This gives the advantage of complete expression profiling using SAGE and high-throughput array screening of a larger panel of samples allowing rapid identification and for instance validation of clinical relevant genes involved in disease onset [42,43]. Finally, the proposed ratio difference between platforms using an universal reference sample (as also indicated in [25]) can serve as a measure for interplatform correspondence per individual gene.
Table 2 Disparities of the technical approaches
SAGE
• Sequence errors (although it has been shown that most of the single-copy SAGE tags are not generated from experimental sequence errors, but that they are novel tags derived from novel transcripts [53])
• Tag annotation difficulties
• Missing transcripts due to absence of a recognition site for the anchoring enzyme (approximately 0.7%) or GC-content bias [24,54]
• Incorrect tags arise from incomplete digestion or alternative poly-adenylation [55]
• Sequence polymorphisms resulting in multiple tags for a single transcript
Affymetrix HG-U133 GeneChips
• Probe design issues (such as distance of the target sequence from the poly-A tail; secondary structures within the target sequence; cross-reactivity of the probe with other transcripts, nucleic acid structure)
• Differences in hybridization efficiencies between probe sets
• Incorrect annotation of transcripts (no sequence verification)
• Efficiencies in dye incorporation
Conclusion
This paper evaluates several approaches for the comparison of different gene expression platforms, outlined using SAGE and Affymetrix GeneChips. We demonstrate that for both SAGE and Affymetrix GeneChips the intra-platform correlations are extremely good, but that the inter-platform agreement based on an unbiased selection of transcripts is modest. The agreement between platforms increases if only transcripts are included with high tag counts and high hybridisation intensities. It appears that the expression distributions are similar for each of the platforms, but that the correlation between platforms is modest due to intrinsic differences, like sensitivity, levels of noise, and gene annotation. Finally, we introduce a novel, filtering-independent approach for data analysis based on the calculation of differences between expression ratios observed in SAGE and Affymetrix GeneChips for each individual transcript. The statistical probability value that can be assigned to each individual betweenratio difference, allows the selection of individual transcripts that display similar regulation on both platforms.
Methods
Tissue and RNA extraction
Wilms' tumor tissue was obtained from a single individual after resection of the tumor. Tissue was immediately frozen in liquid nitrogen. Informed consent to use this material for scientific research was obtained. After homogenization, total RNA was extracted using Trizol (Invitrogen, Breda, The Netherlands), dissolved in RNase free water and stored at -80°C. The Stratagene Universal reference RNA was obtained from Stratagene (Stratagene, Amsterdam, The Netherlands, catalog #740000-41). Purity and integrity of the RNA samples was confirmed on the Agilent 2100 Bioanalyzer (Agilent Technologies Netherlands B.V., Amstelveen, The Netherlands), using the LabChip® approach.
Construction of SAGE libraries
The SAGE library of the Wilms' tumor RNA was generated using the I-SAGE kit according to the manufacturer's instructions (Invitrogen, Breda, The Netherlands; cat. #T5000-03). A detailed protocol may be obtained as a free download [44]. For LongSAGE minor modifications were implemented in the protocol of the I-SAGE kit; i.e. the restriction enzyme BsmFI was replaced by MmeI, linkers were adapted for LongSAGE and ditags were created using sticky-end ligation. All sequence files were processed using the SAGE2000 software provided by Dr. K.W. Kinzler (see also [45]). The SAGE library from the Stratagene Universal reference RNA was obtained from the NCBI website. This library can be retrieved in the Gene Expression Omnibus under code GSM1734 [14,20]).
Annotation of tags
Extracted SAGE tags were annotated based on the SAGE Genie principles [24] through several stringent filters using data from the CGAP website [15]. Several databases (i.e. HsMap.txt, HsRepetitive.txt and HsDatasets.txt) were combined to a final dataset containing all information necessary for tag annotation. Tags matching to unclustered EST's were considered to be no-matches. Tags matching to Unigene clusters retrieved from low ranked databases (<67%; according to the rules set by CGAP) were not included in our comparisons. During this process tags are matched to no, one unique, or more than one Unigene cluster (Unigene Build 160, March 2003). To further identify tags matching more than one Unigene cluster, we extracted the 11th base from our original sequence files using the SAGE2000 software. This 11th base can be used to match against the deposited sequences (Genbank, EMBL etc.) and in this way one may be able to exclude Unigene clusters that contain a different 11th base in their sequence and thereby minimize the number of multiple matches. In the final comparison tags matching to multiple Unigene clusters were excluded. For annotation of LongSAGE tags we used the data available at the CGAP site for Unigene Build 170 (July 2004). These annotations were not available for Unigene Build 160.
Affymetrix
Affymetrix HG-U133A GeneChips were used and the hybridizations were performed according to the manufacturer's protocols and carried out at the Micro-array Department (MAD; Institute for Life Sciences, Faculty of Science, University of Amsterdam). For analysis, the MAS 5.0 software suite was used and comparisons between duplicate Wilms' tumor hybridizations and duplicate Stratagene Universal reference RNA hybridizations were made (data were deposited into the GEO under accession GSE1158). This gives four comparisons (2 Log ratios), from which the geometric mean gene expression ratio between the two samples was calculated. Probe sets on the Affymetrix chips were matched with Unigene clusters (Unigene Build 160, March 2003).
Matching of platforms
The matching of data from two different gene expression profiling platforms (as illustrated in figure 3) poses a couple of problems. On the one hand, a SAGE tag may link to more than one Unigene cluster which results in matches with multiple different Affymetrix probe sets. On the other hand multiple tags originating from one Unigene cluster might match with one Affymetrix probe set. Examining all multiple matches for each individual transcript is extremely laborious and beyond the scope of this study. To circumvent these and other problems we included in our comparison only those clusters for which a one-to-one relation between the two platforms was found. These clusters are called unambiguous Unigene clusters. This matching step already results in a considerable reduction of data available for the comparison. In addition, data were filtered for the presence of gene expression (tag count>0 in both SAGE libraries and present signal on the arrays for both RNA samples).
Comparison of expression ratios between samples
For each platform and each transcript that full-filled the matching criteria an expression ratio between Wilms' tumor and Stratagene Reference RNA was calculated. With these ratios the correspondence between platforms was estimated using the Pearson correlation coefficient, Up/Down classification and a contingency table (Figure 1A, 1B, 1C). Because none of these measures was deemed satisfactory as overall correspondence measure (see background section) we developed a new measure based on the difference between the log(ratio) values in the two platforms for each individual transcript (Figure 4). The chemistry, physics and statistics of the detection technique make that in each platform the observed gene expression is a non-linear transformation of the real gene expression level. For instance, saturation of the array hybridization makes that the high expression levels are truncated. However, because such artifacts affect genes in both tissues in the same way, an observed expression ratio of 1 can still be expected to be observed for genes that are not differentially expressed in the studied tissues. On the other hand, these saturation effects, as well as the relatively larger Poisson error in the detection of low intensity values will affect the ratios on both sides of the ratio distribution in an unpredictable way. Similarly, the sampling error in SAGE will affect ratios for lowly expressed genes, despite the fact that SAGE tag counts are linearly related to transcript abundance. The substitution of zero tag counts that is required for the calculation of ratios will also skew the ratios [46]. Finally, the discrete nature of tag counts, combined with the necessary normalization of tag counts to tags per 50000, will have non-linear effects on the observed ratio distribution in the SAGE platform. Therefore, the relation between the gene expression ratios observed in the SAGE and Affymetrix platform cannot be assumed to be a simple linear Y = X relation. This is already clear from the difference ranges of ratio values in each platform. To directly compare the ratios observed in both platforms at least the range of observed ratios should be similar. The nature of the relation is unknown and fully obscured by the variability in both platforms. However, because in each platform the observed ratio of 1 can be assumed to be true, the simplest function to scale the range of ratio of one platform to that of the other platform is a quadratic equation. Such a scaling function can be based on three values from each ratio distribution. These are the ratio of 1 and, to avoid undue influence of the extreme ratios, the 10th and 90th percentile values. The quadratic scaling takes into account that the ratio distribution is not symmetrical around ratio 1. The full scaling procedure is illustrated and detailed in Figure 4. Note that the scaling uses log(ratio) values. After scaling, the absolute difference between the log(ratios) per individual gene was calculated. The resulting differences of log(expression ratios) were classified into classes of width 0.5, which corresponds to an approximate 3-fold difference in expression ratio between platforms. These classes were used to label the genes in scatter plots of two different platforms (Figure 2D). As illustrated in Figure 4, the distribution of between-ratio differences is approximately normal. Therefore, the mean and standard deviation of this distribution can be used to calculate a standardized difference value (Diffst) per gene and a P-value for this standardized difference can be obtained from the normal distribution. This P-value can then serve as a measure for the position of each gene in the distribution of between-platform ratio differences. Note that this P-value should not be interpreted as a significance value for the ratio difference between platforms. Such a test requires a gene specific variance estimate in the denominator of the standardized difference, which cannot easily be derived from the available data.
Authors' contributions
MK, FB and FVR planned and designed the study. JMR and FVR analyzed the data, generated the figures and drafted the manuscript. MK and FB helped by editing the manuscript, providing overall technical guidance and coordination. LA, DAZ and FVR created the LongSAGE and SAGE libraries, and performed cloning and sequencing of the concatemers. JMR developed the new approach for the comparison of multiple platforms, performed calculations with FVR and provided guidance with the statistical analyses. GJS and FVR performed the annotation of SAGE tags. All authors read and approved the final manuscript.
Grants
This work was supported by the Stichting Kindergeneeskundig Kankeronderzoek (SKK) and the Dutch Cancer Society (KWF; grant UVA 2001–2558)
Acknowledgements
We would like to thank Dr. A.H.C. van Kampen for reading the manuscript and helpful discussions and Raymond J. Waaijer for his bioinformatics support (Bioinformatics Laboratory, Academic Medical Center, the Netherlands).
==== Refs
Liang P Pardee AB Differential display of eukaryotic messenger RNA by means of the polymerase chain reaction Science 1992 257 967 971 1354393
Martin KJ Pardee AB Principles of differential display Methods Enzymol 1999 303 234 258 10349648
Lisitsyn N Wigler M Cloning the differences between two complex genomes Science 1993 259 946 951 8438152
Sutcliffe JG Foye PE Erlander MG Hilbush BS Bodzin LJ Durham JT Hasel KW TOGA: an automated parsing technology for analyzing expression of nearly all genes Proc Natl Acad Sci U S A 2000 97 1976 1981 10681428 10.1073/pnas.040537997
Diatchenko L Lau YF Campbell AP Chenchik A Moqadam F Huang B Lukyanov S Lukyanov K Gurskaya N Sverdlov ED Siebert PD Suppression subtractive hybridization: a method for generating differentially regulated or tissue-specific cDNA probes and libraries Proc Natl Acad Sci U S A 1996 93 6025 6030 8650213 10.1073/pnas.93.12.6025
Wang X Feuerstein GZ Suppression subtractive hybridisation: application in the discovery of novel pharmacological targets Pharmacogenomics 2000 1 101 108 11258592 10.1517/14622416.1.1.101
Velculescu VE Zhang L Vogelstein B Kinzler KW Serial analysis of gene expression Science 1995 270 484 487 7570003
Brenner S Johnson M Bridgham J Golda G Lloyd DH Johnson D Luo S McCurdy S Foy M Ewan M Roth R George D Eletr S Albrecht G Vermaas E Williams SR Moon K Burcham T Pallas M DuBridge RB Kirchner J Fearon K Mao J Corcoran K Gene expression analysis by massively parallel signature sequencing (MPSS) on microbead arrays Nat Biotechnol 2000 18 630 634 10835600 10.1038/76469
DeRisi JL Iyer VR Brown PO Exploring the metabolic and genetic control of gene expression on a genomic scale Science 1997 278 680 686 9381177 10.1126/science.278.5338.680
Lashkari DA DeRisi JL McCusker JH Namath AF Gentile C Hwang SY Brown PO Davis RW Yeast microarrays for genome wide parallel genetic and gene expression analysis Proc Natl Acad Sci U S A 1997 94 13057 13062 9371799 10.1073/pnas.94.24.13057
Lipshutz RJ Fodor SP Gingeras TR Lockhart DJ High density synthetic oligonucleotide arrays Nat Genet 1999 21 20 24 9915496 10.1038/4447
Lockhart DJ Winzeler EA Genomics, gene expression and DNA arrays Nature 2000 405 827 836 10866209 10.1038/35015701
Schena M Shalon D Davis RW Brown PO Quantitative monitoring of gene expression patterns with a complementary DNA microarray Science 1995 270 467 470 7569999
Gene Expression Omnibus (GEO)
SAGEGenie
Saha S Sparks AB Rago C Akmaev V Wang CJ Vogelstein B Kinzler KW Velculescu VE Using the transcriptome to annotate the genome Nat Biotechnol 2002 20 508 512 11981567 10.1038/nbt0502-508
Heller MJ DNA microarray technology: devices, systems, and applications Annu Rev Biomed Eng 2002 4 129 153 12117754 10.1146/annurev.bioeng.4.020702.153438
Triche TJ Schofield D Buckley J DNA microarrays in pediatric cancer Cancer J 2001 7 2 15 11269644
Lu J Lal A Merriman B Nelson S Riggins G A comparison of gene expression profiles produced by SAGE, long SAGE, and oligonucleotide chips Genomics 2004 84 631 636 15475240 10.1016/j.ygeno.2004.06.014
Edgar R Domrachev M Lash AE Gene Expression Omnibus: NCBI gene expression and hybridization array data repository Nucleic Acids Res 2002 30 207 210 11752295 10.1093/nar/30.1.207
Kal AJ van Zonneveld AJ Benes V van den Berg M Koerkamp MG Albermann K Strack N Ruijter JM Richter A Dujon B Ansorge W Tabak HF Dynamics of gene expression revealed by comparison of serial analysis of gene expression transcript profiles from yeast grown on two different carbon sources Mol Biol Cell 1999 10 1859 1872 10359602
Ruijter JM Van Kampen AH Baas F Statistical evaluation of SAGE libraries: consequences for experimental design Physiol Genomics 2002 11 37 44 12407185
Shippy R Sendera TJ Lockner R Palaniappan C Kaysser-Kranich T Watts G Alsobrook J Performance evaluation of commercial short-oligonucleotide microarrays and the impact of noise in making cross-platform correlations BMC Genomics 2004 5 61 15345031 10.1186/1471-2164-5-61
Boon K Osorio EC Greenhut SF Schaefer CF Shoemaker J Polyak K Morin PJ Buetow KH Strausberg RL De Souza SJ Riggins GJ An anatomy of normal and malignant gene expression Proc Natl Acad Sci U S A 2002
Park PJ Cao YA Lee SY Kim JW Chang MS Hart R Choi S Current issues for DNA microarrays: platform comparison, double linear amplification, and universal RNA reference J Biotechnol 2004 112 225 245 15313001 10.1016/j.jbiotec.2004.05.006
Kuo WP Jenssen TK Butte AJ Ohno-Machado L Kohane IS Analysis of matched mRNA measurements from two different microarray technologies Bioinformatics 2002 18 405 412 11934739 10.1093/bioinformatics/18.3.405
Pruitt KD Tatusova T Maglott DR NCBI Reference Sequence project: update and current status Nucleic Acids Res 2003 31 34 37 12519942 10.1093/nar/gkg111
Gilbertson RJ Clifford SC PDGFRB is overexpressed in metastatic medulloblastoma Nat Genet 2003 35 197 198 14593398 10.1038/ng1103-197
Ishii M Hashimoto S Tsutsumi S Wada Y Matsushima K Kodama T Aburatani H Direct comparison of GeneChip and SAGE on the quantitative accuracy in transcript profiling analysis Genomics 2000 68 136 143 10964511 10.1006/geno.2000.6284
Iacobuzio-Donahue CA Ashfaq R Maitra A Adsay NV Shen-Ong GL Berg K Hollingsworth MA Cameron JL Yeo CJ Kern SE Goggins M Hruban RH Highly expressed genes in pancreatic ductal adenocarcinomas: a comprehensive characterization and comparison of the transcription profiles obtained from three major technologies Cancer Res 2003 63 8614 8622 14695172
Evans SJ Datson NA Kabbaj M Thompson RC Vreugdenhil E De Kloet ER Watson SJ Akil H Evaluation of Affymetrix Gene Chip sensitivity in rat hippocampal tissue using SAGE analysis. Serial Analysis of Gene Expression Eur J Neurosci 2002 16 409 413 12193183 10.1046/j.1460-9568.2002.02097.x
Kim HL Comparison of oligonucleotide-microarray and serial analysis of gene expression (SAGE) in transcript profiling analysis of megakaryocytes derived from CD34+ cells Exp Mol Med 2003 35 460 466 14646601
Tan PK Downey TJ Spitznagel EL JrXu P Fu D Dimitrov DS Lempicki RA Raaka BM Cam MC Evaluation of gene expression measurements from commercial microarray platforms Nucleic Acids Res 2003 31 5676 5684 14500831 10.1093/nar/gkg763
Taniguchi M Miura K Iwao H Yamanaka S Quantitative assessment of DNA microarrays – comparison with Northern blot analyses Genomics 2001 71 34 39 11161795 10.1006/geno.2000.6427
Al Moustafa AE Alaoui-Jamali MA Batist G Hernandez-Perez M Serruya C Alpert L Black MJ Sladek R Foulkes WD Identification of genes associated with head and neck carcinogenesis by cDNA microarray comparison between matched primary normal epithelial and squamous carcinoma cells Oncogene 2002 21 2634 2640 11965536 10.1038/sj.onc.1205351
Barczak A Rodriguez MW Hanspers K Koth LL Tai YC Bolstad BM Speed TP Erle DJ Spotted long oligonucleotide arrays for human gene expression analysis Genome Res 2003 13 1775 1785 12805270 10.1101/gr.1048803
Dinel S Bolduc C Belleau P Boivin A Yoshioka M Calvo E Piedboeuf B Snyder EE Labrie F St-Amand J Reproducibility, bioinformatic analysis and power of the SAGE method to evaluate changes in transcriptome Nucleic Acids Res 2005 33 e26 15716308 10.1093/nar/gni025
Trendelenburg G Prass K Priller J Kapinya K Polley A Muselmann C Ruscher K Kannbley U Schmitt AO Castell S Wiegand F Meisel A Rosenthal A Dirnagl U Serial analysis of gene expression identifies metallothionein-II as major neuroprotective gene in mouse focal cerebral ischemia J Neurosci 2002 22 5879 5888 12122050
Yamamoto M Wakatsuki T Hada A Ryo A Use of serial analysis of gene expression (SAGE) technology J Immunol Methods 2001 250 45 66 11251221 10.1016/S0022-1759(01)00305-2
Sun M Zhou G Lee S Chen J Shi RZ Wang SM SAGE is far more sensitive than EST for detecting low-abundance transcripts BMC Genomics 2004 5 1 14704093 10.1186/1471-2164-5-1
Affymetrix
Nacht M Ferguson AT Zhang W Petroziello JM Cook BP Gao YH Maguire S Riley D Coppola G Landes GM Madden SL Sukumar S Combining serial analysis of gene expression and array technologies to identify genes differentially expressed in breast cancer Cancer Res 1999 59 5464 5470 10554019
Gnatenko DV Dunn JJ McCorkle SR Weissmann D Perrotta PL Bahou WF Transcript profiling of human platelets using microarray and serial analysis of gene expression Blood 2003 101 2285 2293 12433680 10.1182/blood-2002-09-2797
Invitrogen
Serial Analysis of Gene Expression
Schaaf GJ Ruijter JM van Ruissen F Zwijnenburg DA Waaijer R Valentijn LJ Benit-Deekman J van Kampen AH Baas F Kool M Full transcriptome analysis of rhabdomyosarcoma, normal and fetal skeletal muscle: statistical comparison of multiple SAGE libraries Faseb J 2005
Haverty PM Hsiao LL Gullans SR Hansen U Weng Z Limited agreement among three global gene expression methods highlights the requirement for non-global validation Bioinformatics 2004
Jurata LW Bukhman YV Charles V Capriglione F Bullard J Lemire AL Mohammed A Pham Q Laeng P Brockman JA Altar CA Comparison of microarray-based mRNA profiling technologies for identification of psychiatric disease and drug signatures J Neurosci Methods 2004 138 173 188 15325126 10.1016/j.jneumeth.2004.04.002
Lee JK Bussey KJ Gwadry FG Reinhold W Riddick G Pelletier SL Nishizuka S Szakacs G Annereau JP Shankavaram U Lababidi S Smith LH Gottesman MM Weinstein JN Comparing cDNA and oligonucleotide array data: concordance of gene expression across platforms for the NCI-60 cancer cells Genome Biol 2003 4 R82 14659019 10.1186/gb-2003-4-12-r82
Yuen T Wurmbach E Pfeffer RL Ebersole BJ Sealfon SC Accuracy and calibration of commercial oligonucleotide and custom cDNA microarrays Nucleic Acids Res 2002 30 e48 12000853 10.1093/nar/30.10.e48
Iacobuzio-Donahue CA Maitra A Shen-Ong GL van Heek T Ashfaq R Meyer R Walter K Berg K Hollingsworth MA Cameron JL Yeo CJ Kern SE Goggins M Hruban RH Discovery of novel tumor markers of pancreatic cancer using global gene expression technology Am J Pathol 2002 160 1239 1249 11943709
Feldker DE Datson NA Veenema AH Meulmeester E De Kloet ER Vreugdenhil E Serial analysis of gene expression predicts structural differences in hippocampus of long attack latency and short attack latency mice Eur J Neurosci 2003 17 379 387 12542675 10.1046/j.1460-9568.2003.02440.x
Chen J Sun M Lee S Zhou G Rowley JD Wang SM Identifying novel transcripts and novel genes in the human genome by using novel SAGE tags Proc Natl Acad Sci U S A 2002 99 12257 12262 12213963 10.1073/pnas.192436499
Margulies EH Kardia SL Innis JW Identification and prevention of a GC content bias in SAGE libraries Nucleic Acids Res 2001 29 E60 60 11410683 10.1093/nar/29.12.e60
Pauws E van Kampen AH van de Graaf SA de Vijlder JJ Ris-Stalpers C Heterogeneity in polyadenylation cleavage sites in mammalian mRNA sequences: implications for SAGE analysis Nucleic Acids Res 2001 29 1690 1694 11292841 10.1093/nar/29.8.1690
|
15955238
|
PMC1186021
|
CC BY
|
2021-01-04 16:32:48
|
no
|
BMC Genomics. 2005 Jun 14; 6:91
|
utf-8
|
BMC Genomics
| 2,005 |
10.1186/1471-2164-6-91
|
oa_comm
|
==== Front
BMC MicrobiolBMC Microbiology1471-2180BioMed Central London 1471-2180-5-411600017610.1186/1471-2180-5-41Research ArticleA novel Geobacteraceae-specific outer membrane protein J (OmpJ) is essential for electron transport to Fe (III) and Mn (IV) oxides in Geobacter sulfurreducens Afkar Eman [email protected] Gemma [email protected] Marianne [email protected] Derek R [email protected] Department of Microbiology, University of Massachusetts, Amherst, Massachusetts 01003, USA2 Biosciences Division, Argonne National Laboratory, Argonne, Illinois 60439, USA2005 6 7 2005 5 41 41 28 1 2005 6 7 2005 Copyright © 2005 Afkar 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
Metal reduction is thought to take place at or near the bacterial outer membrane and, thus, outer membrane proteins in the model dissimilatory metal-reducing organism Geobacter sulfurreducens are of interest to understand the mechanisms of Fe(III) reduction in the Geobacter species that are the predominant Fe(III) reducers in many environments. Previous studies have implicated periplasmic and outer membrane cytochromes in electron transfer to metals. Here we show that the most abundant outer membrane protein of G. sulfurreducens, OmpJ, is not a cytochrome yet it is required for metal respiration.
Results
When outer membrane proteins of G. sulfurreducens were separated via SDS-PAGE, one protein, designated OmpJ (outer membrane protein J), was particularly abundant. The encoding gene, which was identified from mass spectrometry analysis of peptide fragments, is present in other Geobacteraceae, but not in organisms outside this family. The predicted localization and structure of the OmpJ protein suggested that it was a porin. Deletion of the ompJ gene in G. sulfurreducens produced a strain that grew as well as the wild-type strain with fumarate as the electron acceptor but could not grow with metals, such as soluble or insoluble Fe (III) and insoluble Mn (IV) oxide, as the electron acceptor. The heme c content in the mutant strain was ca. 50% of the wild-type and there was a widespread loss of multiple cytochromes from soluble and membrane fractions. Transmission electron microscopy analyses of mutant cells revealed an unusually enlarged periplasm, which is likely to trigger extracytoplasmic stress response mechanisms leading to the degradation of periplasmic and/or outer membrane proteins, such as cytochromes, required for metal reduction. Thus, the loss of the capacity for extracellular electron transport in the mutant could be due to the missing c-type cytochromes, or some more direct, but as yet unknown, role of OmpJ in metal reduction.
Conclusion
OmpJ is a putative porin found in the outer membrane of the model metal reducer G. sulfurreducens that is required for respiration of extracellular electron acceptors such as soluble and insoluble metals. The effect of OmpJ in extracellular electron transfer is indirect, as OmpJ is required to keep the integrity of the periplasmic space necessary for proper folding and functioning of periplasmic and outer membrane electron transport components. The exclusive presence of ompJ in members of the Geobacteraceae family as well as its role in metal reduction suggest that the ompJ sequence may be useful in tracking the growth or activity of Geobacteraceae in sedimentary environments.
==== Body
Background
Fe (III) oxide is the most abundant metal electron acceptor in most soils and sediments and its microbial reduction greatly contributes to the degradation of organic matter in many sedimentary environments as well as to the degradation of organic contaminants in polluted groundwater [1,2]. Fe (III) also is the primary electron acceptor supporting the growth of dissimilatory metal-reducing microorganisms when metal reduction is stimulated by adding a suitable electron donor to promote the in situ bioremediation of soluble metal contaminants [3] such as uranium [4] and vanadium [5]. Thus, understanding the mechanisms of electron transfer to insoluble Fe (III) oxides could greatly aid in the study of dissimilatory metal reduction in various sedimentary environments.
Despite the wide phylogenetic diversity of microorganisms capable of dissimilatory metal-reduction [1], molecular analyses of moderate temperature sedimentary environments in which Fe (III) reduction is important have routinely found that microorganisms in the Geobacteraceae are prevalent whereas other well-studied Fe (III)-reducing microorganisms, such as Shewanella species, are not detected [1]. This has been attributed, at least in part, to different mechanisms for Fe (III) reduction in these organisms. Geobacter species need to directly contact Fe (III) oxides in order to reduce them [6], have a highly specialized strategy for searching for Fe (III) oxides [7], and use pili as conductive nanowires to transfer electrons to the insoluble electron acceptor [8]. In contrast, Shewanella [9,10] and Geothrix species [11] produce soluble electron shuttles and Fe(III) chelators which alleviate the need for direct contact with Fe (III) oxides.
Because of the insoluble nature of Fe(III) and Mn(VI) oxides, metal reduction in dissimilatory metal-reducing organisms is thought to occur at or near the outer membrane. Most studies on the mechanisms for Fe (III) reduction in Geobacter species have focused on the role of c-type cytochromes [12-14]. Over 110 putative c-type cytochrome genes have been identified in the G. sulfurreducens genome [15]. Many of these cytochrome genes are more highly expressed during growth on Fe (III) than with fumarate as the electron acceptor and deletion of some of these cytochrome genes greatly reduces the capacity for Fe (III) reduction [13,14]. However, the importance of c-type cytochromes in the final electron transfer to Fe (III) has been questioned because Pelobacter species, which are phylogenetically intertwined with Geobacter and Desulfuromonas species in the Geobacteraceae [16], can reduce Fe (III), yet do not appear to contain c-type cytochromes [17]. If, as expected, reduction of Fe(III) takes place at, or near, the outer membrane surface, then there may be outer membrane proteins other than c-type cytochromes, which have a role in electron transfer to Fe(III). In support of this, Geobacter's pili have recently been found to play a key role in electron transfer to insoluble metals by acting as microbial nanowires that extend the electron transfer capabilities of the bacterium beyond the cell surface [8].
Geobacter sulfurreducens has been routinely used as a model organism for investigations into Fe (III) reduction in Geobacteraceae because its complete genome sequence [15] and a genetic system [18] are available. Here we report that the most abundant protein in the outer membrane of G. sulfurreducens is not a cytochrome, yet this protein is required for Fe (III) reduction.
Results
Identification and characterization of OmpJ
One protein, designated outer membrane protein J (OmpJ), was much more abundant than any other protein in the outer membrane fraction of cells grown with fumarate as the sole electron acceptor (Fig. 1A). This protein also was present in cultures grown with other electron acceptors such as Fe(III) citrate, Fe(III) oxides and and Mn(IV) oxides (data not shown). MALDI-TOF mass spectrometry analyses identified eight peptides (MGDATVALGFAR, VDFGGWAANATAK, LITHFEIDSTWGK, FDPVTIDGFLLYQR, NVYLDENIPSTPLNVK, AFAIANVGFVAADKDNTTYCNAR, ALVYNVQNVIGGFVGYNANITSK, VFDNLTASVQGAYVILGDYFKDTAGTAANPEDPR) that uniquely corresponded to the protein encoded by ORF GSU3304 (gi-39998393), annotated as a putative LamB porin family protein in the genome of G. sulfurreducens. OmpJ had a predicted average molecular mass of 48.9 kDa, in accordance with its mobility in denaturing gels (Fig. 1A), and a theoretical pI of 6.25 compared with a pI of 6.7 determined by isoelectric focusing gel electrophoresis. The PSORT algorithm predicted OmpJ to be localized in the outer membrane, consistent with its presence in outer-membrane preparations (Fig. 1A).
Several secondary structure prediction methods predicted that OmpJ consists mainly of extended beta-chain fragments. Six of the first 20 predictions by 3D-PSSM [19], which uses a threading algorithm to predict the fold of a protein most homologous to structures deposited in the Protein Data Bank, suggest that OmpJ is a porin. The other predicted folds were for soluble proteins, which have large numbers of beta segments and thus appear to be inappropriate for a membrane-bound protein.
Several pieces of biochemical evidence also indicate that OmpJ may function as a porin. First, its localization in the outer membrane. OmpJ was associated to the outer membrane fraction of G. sulfurreducens and was isolated in a highly purified form after removal of the cytoplasmic membrane (Fig. 1A). Also, the apparent molecular weight of OmpJ in the absence of heating (118 kDa) was roughly twice that of the heat-treated protein (Fig. 1B), indicative of heat modifiability [20,21]. In addition, in situ proteinase K treatment of intact cells of G. sulfurreducens did not digest OmpJ (Fig. 1C) while OmpJ integrity was affected after proteinase K treatment of outer membrane fractions (data not shown). Incubation of intact cells with proteinase K leads to the degradation of exposed outer membrane proteins, while subsurface proteins such as porins remain protected against proteolysis. Taken together, these data greatly strengthen our assertion that (i) OmpJ is located in the outer membrane, (ii) it is tightly embedded within the membrane and (iii) it is not significantly exposed on the cell surface, as expected of a porin.
Presence of ompJ in other Geobacteraceae
Homologs of ompJ were found in the genomes of the two other members of the Geobacteraceae for which complete sequences are available,Geobacter metallireducens and Pelobacter carbinolicus (Fig. 2). A hypothetical protein in the genome of G. metallireducens (gi-48845525) had the highest identity at the amino acid level, 70.2% (out of 476 amino acids), and 80% similarity to OmpJ of G. sulfurreducens. A protein also was identified in P. carbinolicus (GenBank, accession number DQ004247) that shared 34% identity (out of 513 amino acids) and 51% similarity with G. sulfurreducens OmpJ. No close OmpJ homologs were found in the NCBI database outside the Geobacteraceae family. OmpJ had 21% identity and 33.3% similarity at the amino acid level with Omp35 of S. oneidensis MR-1 (Fig. 2), a porin-like protein found to have an indirect effect in metal reduction in this organism [22]. However, while OmpJ had no similarity to other porins in the database, Omp35 was most closely similar to known bacterial porins such as PorA of Neisseria meningitidis [22]. These results suggest that OmpJ is a novel type of porin that is unique to the Geobactereaceae.
Characterization of an OmpJ-deficient mutant
The presence of the ompJ gene in the Geobacteraceae family but not in any other group of bacteria, suggested that OmpJ might play a key role in the physiology of these metal reducers. In order to elucidate the physiological role of G. sulfurreducens OmpJ, a deletion mutant was constructed by gene replacement with a kanamycin cassette (Fig. 3A). SDS-PAGE analyses confirmed the absence of the OmpJ protein band in outer membrane preparations of the OmpJ-deficient mutant but also showed significant differences in the protein composition of the mutant outer membrane fraction when compared to the wild-type, with some absent proteins and some new proteins present in the outer membrane fractions of the mutant (Fig. 3B).
The OmpJ-deficient mutant strain reduced fumarate nearly as well as the wild-type strain (Fig. 4A). However, it was markedly deficient in the reduction of metals such as soluble Fe (III) citrate (Fig. 4B), insoluble poorly crystalline Fe (III) oxide (Fig. 4C), as well as Mn (IV) oxides (Fig. 4D) when compared to the wild-type strain. Attempts to complement the mutation in trans yielded a strain that produced OmpJ only at levels ca. 10% of the wild-type, as visualized on denaturing gels (data not shown). Such suboptimal levels of complementation have previously been observed in previous studies on the function of other genes in G. sulfurreducens [13,14]. However, even with suboptimal levels of OmpJ, 16% of the Fe (III) citrate reduction capacity was restored (data not shown).
The loss of the capacity for Fe (III) and Mn (IV) reduction in the ompJ mutant was associated with a significant loss of c-type cytochromes from the cell. The c-type cytochrome content in cell extracts of the wild-type was 20 ± 0.7 μmol per 100 mg protein whereas the mutant's c-type cytochrome content was 10 ± 0.2 μmol per 100 mg protein. Heme c staining of proteins separated by SDS-PAGE confirmed there was a loss of a number of both soluble and membrane-associated cytochromes in the OmpJ-deficient mutant strain (Fig. 5). There was also an apparent increase in the abundance of a 70 kDa c-type cytochrome associated with the inner membrane.
Because porins also have been shown to play a structural role in the integrity of the bacterial cell surface [23,24], cells were examined with transmission electron microscopy (TEM). TEM analyses of negatively-stained cells and of thin sections of cells from fumarate-grown cultures showed that the appearance of the mutant cells was remarkably different from the wild-type cells and was consistent with an enlarged periplasm (Fig. 6).
Discussion
The results presented in this work suggest that the most abundant outer-membrane protein in G. sulfurreducens is a probable porin and, despite its apparent lack of any moieties capable of participating in electron transfer, its presence is required in order for G. sulfurreducens to reduce extracellular electron acceptors, such as soluble Fe(III) citrate and insoluble Fe (III) and Mn (IV) oxides. These results, coupled with the recent discovery of the importance of a porin-like protein, designated Omp35, in Fe (III) reduction in Shewanella oneidensis MR-1 [22], emphasize that porins play an important role in the maintenance of the physical integrity and function of the cell surface in dissimilatory metal-reducing microorganisms.
Physiological role of OmpJ
OmpJ's annotation as a porin is consistent with the predicted beta-barrel structure of the protein [25] and with several pieces of biochemical evidence such as its location and abundance in the outer membrane (Fig. 1A) and the results of our proteinase K study, which suggest that it is embedded in the membrane (Fig. 1C). Also, porins often assemble in the outer membrane as multimeric structures composed of several porin monomers [24] and, as a result, many porins exhibit heat modifiability [20,21]. As was shown in Fig. 1B, OmpJ also showed heat modifiability.
Bacterial outer membrane porins have been studied extensively [24,25] and porin-like proteins also are found in chloroplasts and mitochondria [26]. Porins represent an unusual class of membrane proteins in that they exhibit no hydrophobic stretches in their amino acid sequences and form hollow beta-barrel structures with a hydrophobic outer surface. The barrel structure encompasses the trans-membranous pore that allows the passive diffusion of hydrophilic solutes across the (outer) membrane [25]. The enlarged periplasm observed in the OmpJ-deficient mutant, as compared to that of the wild type cells, further suggests some role in solute transport. Although OmpJ is annotated in the genome of G. sulfurreducens as a putative protein of the LamB porin superfamily, this classification does not necessarily imply a physiological role analogous to that of LamB channels. The LamB superfamily comprises channels involved in the spontaneous diffusion of specific nutrients. The LamB protein of E. coli, which also serves as the phage lambda receptor, is a prototype of this class of channels and catalyzes the influx of maltose and maltodextrins and also facilitates the influx of various carbohydrates when in low concentrations in the extracellular environment [24]. However, Geobacter species are not known to use sugars as substrates. The finding that the ompJ deletion mutant reduced fumarate at rates comparable to the wild-type suggests that the transport of the electron donor, acetate, or essential nutrients, such as ammonia, phosphate, sulfur, amino acids and trace metals, which are necessary in central metabolic reactions, was not inhibited. Morevover, the enlarged periplasmic space observed upon deletion of the ompJ gene is consistent with a channel involved in the efflux, rather than influx, of some nutrient.
It is unlikely that OmpJ is involved in the final electron transfer to Fe (III) and Mn (IV) because it lacks any apparent electron transfer moieties and because it appears to be embedded within the membrane rather than exposed externally. In addition, other non-cytochrome proteins, type IV pili, have recently been shown to be involved in the final electron transfer to insoluble metals [8]. Thus, the effect of the ompJ mutation in the inhibition of Fe (III) and Mn (IV) reduction appears to be indirect, as a result of the general reduction in the production of c-type cytochromes. Previous studies have suggested that a number of c-type cytochromes are required in order for G. sulfurreducens to effectively reduce metals and the loss of just one of these may inhibit Fe (III) reduction [13,14,27]. In some instances, outer-membrane proteins may be required for the proper localization of outer-membrane c-type cytochromes [28] and, if not properly localized and folded, cytochromes may be proteolytically degraded. However, a direct role for OmpJ in the proper localization of c-cytochromes is unlikely because OmpJ homologs are also present in Pelobacter species, which reduce Fe (III) but do not contain detectable c-type cytochromes [17]. Alternatively, OmpJ may be required for transport of some constituent such as iron that is required for c-cytochrome production. The swelling of the periplasm in fumarate-grown cells suggests, however, that transport of one or more solutes out of the cell might be inhibited in the absence of OmpJ. This abnormal periplasm is likely to interfere with protein folding and localization, thus inducing the extracytoplasmic stress responses, mediated by the RpoE sigma factor and Cpx systems in E. coli [29], and triggering a proteolytic cascade that relieves the accumulation of misfolded proteins in the periplasm [30].
Porins and metal reduction
OmpJ of G. sulfurreducens is the first porin ever to be described in detail in any member of the delta-Proteobacteria [24]. Similar to Omp35, a porin-like protein from Shewanella oneidensis MR-1 [22], it appears to have an indirect role in metal reduction. While an Omp35-deficient mutant in S. oneidensis had slower rates of reduction of fumarate and soluble or insoluble Fe(III) [22], deletion of OmpJ of G. sulfurreducens did not affect fumarate reduction but dramatically inhibited metal reduction such as soluble or insoluble Fe(III) and Mn(VI) oxides. This is not surprising inasmuch as fumarate reduction occurs intracellularly in G. sulfurreducens [31] while the S. oneidensis terminal reductase is periplasmic [32]. Interestingly, the levels and distribution of key components of electron transport in S. oneidensis such as quinones and cytochromes were normal in the Omp35-deficient mutant [22], whereas the OmpJ-deficient mutant of G. sulfurreducens had a substantial reduction in heme c and c-cytochrome abundance, which may have indirectly contributed to the marked decrease of its metal reduction potential. The lack of homology and differences in function between Omp35 and OmpJ may reflect the profound differences in the mechanisms for Fe(III) reduction in Shewanella and Geobacter species [1].
Most importantly, these studies emphasize the role of non-electron transport proteins in electron transfer to a variety of electron acceptors. OmpJ of G. sulfurreducens is of special significance because, unlike Omp35, it is widespread in members of the Geobacteraceae, but no homologs are found in any other bacterial groups. This suggests that screening for ompJ-like sequences may be a good strategy for identifying Geobacteraceae sequences in libraries of environmental genomic DNA. Furthermore, ompJ provides another gene in the short, but growing, list of Geobacteraceae-specific sequences that might be used to quantify the number of Geobacteraceae in Fe (III)-reducing environments or to infer rates of activity of these organisms from quantitative mRNA analysis [33,34].
Conclusion
In summary, OmpJ is an abundant outer-membrane protein in G. sulfurreducens and it is required for metal reduction in this organism. OmpJ also is required to keep the structural integrity of the periplasmic space, which is necessary for proper folding and functioning of electron transport components. Thus, the role of OmpJ in metal reduction may, in fact, be indirect. While the actual role of this apparent porin is still uncertain, further studies on interactions of OmpJ with other proteins in G. sulfurreducens may help to better elucidate its function.
Methods
Bacterial strains and culture conditions
All strains used in this work were isogenic with the wild-type G. sufurreducens strain PCA (ATCC 51573) [35], obtained from our laboratory culture collection. A deletion mutant in the ompJ gene (GSU3304) was constructed by cross-over PCR replacing the +311 to +1245 coding region with a kanamycin (KM) cassette, as previously described [27,36]. Briefly, the upstream region of ompJ was amplified with primers upF (5'-GCGTTGACAGACAAACTC-3') and upR (5'-GCCATCGTTCGATCTTCCG-3') and the downstream region with primers dwnF (5'-CAAGGTGTTTGACAACCTG-3') and dwnR (5'- CAAGGTGTTTGACAACCTG-3'). The KM-resistance cassette in plasmid pBBR1MCS-2 [37] was PCR-amplified with primers kanF (5'- CGGAAGATCGAACGATGGCACCTGGGATGAATGTCAGC-3') and kanR (5'-CAGGTTGTCAAACACCTTGATGGCAGGTTGGGCGTCGC-3'). The three PCR-amplified fragments were used as templates in a recombinant PCR reaction [38] using primers upF and dwnR to amplify a 1.951 kb DNA fragment. Electroporation and mutant isolation were performed as described previously [18]. The mutation was confirmed by PCR and Southern blotting [38]. The ompJ mutation was complemented in trans using plasmid pCM66-OmpJ, a pCM66 [39] derivative carrying a wild-type copy of the ompJ coding region previously amplified using primers OmpJ01 (5'-GGAAGCTTCCATGCTGTTTTATCATACCC-3') and OmpJ02 (5'-GGGAATTCGGTGATGCAATTAGAATG-3').
Cells were routinely cultured under strict anaerobic conditions in freshwater (FW) medium, as previously described [40], with 20 mM sodium acetate as the electron donor and either 55 mM Fe (III)-citrate, 40 mM fumarate, poorly crystalline Fe (III) oxide (100 mmol/l), or MnO2 (3 mol/l) as the electron acceptor.
Preparation of outer membrane proteins and PAGE analyses
Cells for protein analyses were grown in 1-liter bottles or, for mass cultivation, in 10-liter carboys to late-exponential phase, harvested by centrifugation (12,000 xg for 10 min at 4°C), and washed with 10 mM Tris-HCl (pH 8.0) containing 1 mM EDTA and 10 μM phenol methyl-sulfonyl fluoride (PMSF) to inhibit proteolytic activities. Cell samples were routinely stored at -20°C. Before use, cell samples were suspended in 10 mM Tris-HCl buffer (pH 7.5) at 4°C, and subjected to three passes through a French pressure cell at 10,000 psi to lyze the cells (cell-free extract). Cell debris and intact cells were separated by centrifugation (12,000 g for 20 min) and the cell-free extract was further centrifuged (100,000 × g for 1 h) to separate the soluble fraction (SF) from the crude membrane pellet. The pellet was resuspended in 10 mM Tris-HCl buffer (pH 7.5) and lauroyl sarcosine-sodium salt was added to a final concentration of 1% (w/v) in order to selectively solubilize cytoplasmic membrane proteins [41]. The mixture was stirred at room temperature for 1 to 2 h and then centrifuged at 100,000 × g for 1 h. The supernatant, containing the solubilized cytoplasmic membrane protein fraction (CM), was collected and stored at -20°C. The pellet, which contains the outer membrane fraction (OM), was washed three times in 10 mM Tris-HCl buffer (pH 7.5) containing 10 mM MgCl2, 2% NaCl, and 10 μM PMSF.
Protein profiles in the various fractions were analyzed by SDS-Polacrylamide Gel Electrophoresis (SDS-PAGE, 15 %) in a Mini-Protean® 3 Cell (Bio-Rad). Prior to electrophoresis, OM protein samples were mixed (1:2) in Laemmli sample buffer (Bio-Rad), containing β-mecaptoethanol and boiled for 10 min at 100°C. Protein separation prior to staining for heme c-containing proteins with N,N,N',N'-tetramethylbenzidine [42,43] was performed in the same manner except that β-mecaptomethanol was omitted from the sample buffer and boiling time was reduced to 5 min. The most abundant protein band from the OM fraction was excised from the gel, digested with trypsin in the presence of 0.01% n-octylglucopyranoside, and analyzed with matrix-assisted laser desorption ionization- time of flight (MADI-TOF) mass spectrophotometry (Kratos Axima CFR; Kratos Analytical, Manchester, England) [44].
In situ proteinase K treatment
Surface exposure of the OmpJ protein was assayed with proteinase K as a modification of a previously described protocol [45]. Cells were grown to mid-exponential phase in FW medium containing 20 mM acetate and 40 mM fumarate and harvested by centrifugation (20 min at 13,000 rpm, 4°C). After two washes in a 10 mM Tris-HCl buffer (pH 8.0) containing 10 mM MgCl and 2% NaCl, cells were suspended in the same buffer to a final concentration of 63 mg of wet cells per ml. Examination of the bacterial suspension by phase-contrast microscopy did not indicate detectable bacterial lysis. Proteinase K was gradually added to the cell suspension at concentrations ranging from 0 to 40 U ml-1. A negative control with added buffer, instead of proteinase K, and another using outer membrane fractions, instead of intact cells, also were included. Samples were stirred at room temperature for 15 min with various concentrations (0 to 40 U ml-1) of proteinase K before a protease cocktail inhibitor (Roche) was added to stop the proteolytic reaction. The cells were pelleted by centrifugation at 14,000 rpm for 5 min, washed twice in 10X volumes of the same buffer containing the protease inhibitor cocktail, and resuspended in the same buffer with the protease inhibitor cocktail. Aliquots of the cell suspensions (10 mg protein ml-1) were diluted 1:2 in SDS-sample buffer and boiled for 10 min prior to separation by SDS-PAGE, as described above.
C-type cytochromes analyses
Cells of the wild-type and mutant strains were grown to mid-exponential phase with fumarate as the electron acceptor and harvested. Cell-free extract, soluble, CM, and OM protein fractions were prepared as described above. The total heme c content of each fraction (100 μg protein) was determined by the pyridine ferrohemochrome method [46]. In addition, every fraction (10 μg protein) was separated by SDS- PAGE (15 %) at 125 V for 2–3 hrs and the gel was stained for heme c-containing proteins as previously described [42,43].
Transmission Electron Microscopy (TEM)
Cells for TEM analyses were grown to mid-exponential phase in FW medium with fumarate, fixed on to the surface of carbon-coated copper grids with glutaraldehyde and negatively stained with 2% uranyl acetate. For thin sections, cells were fixed in glutaraldehyde and stained with osmium tetraoxide following standard microscopy procedures. Samples were analyzed in a JEOL 100S transmission electron microscope operated at 60–80 V.
Analytical methods
Fe(II) production was monitored by the ferrozine assay at 12 hr intervals [40]. Protein concentration was determined by the bicinchoninic acid method with Bovine Serum Albumin (BSA) as a standard [47].
Sequence analyses
The PSORT algorithm was used to predict the cell localization of the OmpJ protein. The genome sequences of Geobacter metallireducens and Pelobacter carbinolicus can be found at . The genome sequence of G. sulfurreducens can be found at .
Authors' contributions
EA conducted most of the experiments and analyzed some of the data. GR conducted the microscopy experiments and some phenotypic characterization, analyzed and organized the data, and drafted most of the manuscript. MS conducted the analyses of the OmpJ porin amino acid sequence. DRL conceived the overall project, provided experimental guidance, and drafted portions of the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors would like to thank Maddalena Coppi and Ching Leang for their advice during the development of different parts of this work and John Leszyk (Proteomic Mass Spectrometry Lab, UMass Medical School) for assistance with MALDI-TOF mass spectrometry analyses. Also thanks to Laurie Didonato, Richard Glaven, Richard Ding, and Evgenya Shelobolina for technical assistance and helpful discussions. MS acknowledges support from the Office of Science (BER), U. S. Department of Energy NABIR program under Contract W-31-109-Eng-38. This research was supported by the Office of Science (BER), U. S. Department of Energy, and Cooperative Agreement No. DE-FC02-02ER634, and by the Office of Science (BER), U. S. Department of Energy, Grant No. DE-FG02-02ER63423 to DRL.
Figures and Tables
Figure 1 Localization and characterization of OmpJ. A: OmpJ is the most abundant protein in the outer membrane of G. sulfurreducens. Subcellular fractions (cell-free extracts (lane 1), soluble (lane 2), cytoplasmic membrane (lane 3) and outer membrane (lane 4) fractions; 5 μg protein per lane), of G. sulfurreducens grown with fumarate as electron acceptor were analyzed by SDS-PAGE. The most abundant protein in the outer membrane (indicated by an arrow) was designated OmpJ. B: Heat modifiability of OmpJ. OmpJ migrated as a dimer (white arrow) in an SDS-PAGE gel in the absence of heat treatment (lane 1) but migrated as a monomer (solid arrow) after heat treatment at 100°C for 5 minutes (lane 2). Protein, 5 μg per lane. C: Effect of proteinase K treatment on OmpJ integrity. Outer membrane fractions were treated with different concentrations (0 to 40 U ml-1) of proteinase K (+) to analyze the surface exposure of OmpJ and their protein composition was analyzed by denaturing PAGE. A negative control without proteinase K (-) also is shown. First lane in all panels corresponds to molecular weight standards. Numbers at left are molecular masses in kDa.
Figure 2 Amino acid sequence alignment of OmpJ, OmpJ homologs in other members of the Geobacteraceae and Omp35, a porin-like protein from S. oneidensis MR-1. Identical residues are highlighted in dark grey and similar residues in light grey. Abbreviations: Gsul; Geobacter sulfurreducens OmpJ; Gmet, G. metallireducens OmpJ homolog; Pcar, P. carbinolicus OmpJ homolog; Sone, S. oneidensis MR-1 Omp35.
Figure 3 Generation of an OmpJ-deficient mutant of G. sulfurreducens. A: An OmpJ-deficient mutant of G. sulfurreducens was constructed by gene replacement with a kanamycin cassette. B: Protein composition of outer membrane fractions of the wild-type (lane 1) and an OmpJ-deficient mutant (lane 2) showing the absence of the OmpJ band in the mutant strain (indicated by an arrow). First lane corresponds to molecular weight standards. Numbers at left are molecular masses in kDa.
Figure 4 Physiological characterization of an OmpJ-deficient mutant. The growth of the wild-type (WT) and OmpJ-deficient mutant (OmpJ-) strains with fumarate (A), Fe(III) citrate (B), and insoluble Fe(III) (C) or Mn(VI) (D) oxides was studied. Growth on fumarate was quantified by following the optical density of the culture at 600 nm (OD600), while growth with soluble or insoluble Fe(III) was monitored by measuring the levels of soluble Fe(II) present in the medium as a result of Fe(III) reduction. Error bars are the standard deviation from the average of triplicate determinations. The reduction of Mn(VI) oxides was visually tested as the oxidized Mn(VI) oxides turn from black to a whitish precipitate upon reduction. Uninoculated controls also were included in B–D.
Figure 5 Heme-stained protein composition of a WT (A) and an OmpJ-deficient mutant (B). A: Distribution of heme-stained proteins in subcellular fractions of G. sulfurreduces wild-type. Lane 1, total cell free extract; lane 2, soluble fractions; lane 3, inner membrane; and lane 4, outer membrane; 10 μg protein per lane. B: Distribution of heme-stained proteins in subcellular fractions of an OmpJ-deficient mutant. Lane 1, total cell free extract; lane 2, soluble fractions; lane 3, inner membrane; and lane 4, outer membrane; 10 μg protein per lane. Numbers at left are molecular masses in kDa.
Figure 6 Microscopic analyses of an OmpJ-deficient mutant. Transmission Electron Microscopy analyses of cells of the wild-type (A and C) and an OmpJ-deficient mutant (B and D) showing the enlargement of the periplasmic space in the OmpJ-mutant cells when compared with the wild type. A and B show negatively stained cells while C and D show micrographs of thin sections of, respectively, cells of the wild-type and OmpJ-deficient mutant strains. Bars, 1 μm.
==== Refs
Lovley DR Holmes DE Nevin K Dissimilatory Fe(III) and Mn(IV) reduction Adv Microb Physiol 2004 49 219 286 15518832
Lovley DR Cleaning up with genomics: applying molecular biology to bioremediation Nat Rev Microbiol 2003 1 35 44 15040178 10.1038/nrmicro731
Finneran KT Housewright ME Lovley DR Multiple influences of nitrate on uranium solubility during bioremediation of uranium-contaminated subsurface sediments Environ Microbiol 2002 4 510 516 12220407 10.1046/j.1462-2920.2002.00317.x
Anderson RT Vrionis HA Ortiz-Bernad I Resch CT Long PE Dayvault R Karp K Marutzky S Metzler DR Peacock A White DC Lowe M Lovley DR Stimulating the in situ activity of Geobacter species to remove uranium from the groundwater of a uranium-contaminated aquifer Appl Environ Microbiol 2003 69 5884 5891 14532040 10.1128/AEM.69.10.5884-5891.2003
Ortiz-Bernad I Anderson RT Vrionis HA Lovley DR Vanadium respiration by Geobacter metallireducens: novel strategy for In situ removal of vanadium from groundwater Appl Environ Microbiol 2004 70 3091 3095 15128571 10.1128/AEM.70.5.3091-3095.2004
Nevin KP Lovley DR Lack of production of electron-shuttling compounds or solubilization of Fe(III) during reduction of insoluble Fe(III) oxide by Geobacter metallireducens Appl Environ Microbiol 2000 66 2248 2251 10788411 10.1128/AEM.66.5.2248-2251.2000
Childers SE Ciufo S Lovley DR Geobacter metallireducens accesses insoluble Fe(III) oxide by chemotaxis Nature 2002 416 767 769 11961561 10.1038/416767a
Reguera G McCarthy KD Mehta T Nicoll JS Tuominen MT Lovley DR Extracellular electron transfer via microbial nanowires Nature 2005 435 1098 1101 15973408 10.1038/nature03661
Nevin KP Lovley DR Mechanisms for Fe(III) oxide reduction in sedimentary environments Geomicrobiol J 2002 19 141 159 10.1080/01490450252864253
Newman DK Kolter R A role for excreted quinones in extracellular electron transfer Nature 2000 405 94 97 10811225 10.1038/35011098
Nevin KP Lovley DR Mechanisms for accessing insoluble Fe(III) oxide during dissimilatory Fe(III) reduction by Geothrix fermentans Appl Environ Microbiol 2002 68 2294 2299 11976100 10.1128/AEM.68.5.2294-2299.2002
Magnuson TS Hodges-Myerson AL Lovley DR Characterization of a membrane-bound NADH-dependent Fe(3+) reductase from the dissimilatory Fe(3+)-reducing bacterium Geobacter sulfurreducens FEMS Microbiol Lett 2000 185 205 211 10754249 10.1016/S0378-1097(00)00081-1
Leang C Coppi MV Lovley DR OmcB, a c-type polyheme cytochrome, involved in Fe(III) reduction in Geobacter sulfurreducens J Bacteriol 2003 185 2096 2103 12644478 10.1128/JB.185.7.2096-2103.2003
Butler JE Kaufmann F Coppi MV Nunez C Lovley DR MacA, a diheme c-type cytochrome involved in Fe(III) reduction by Geobacter sulfurreducens J Bacteriol 2004 186 4042 4045 15175321 10.1128/JB.186.12.4042-4045.2004
Methé BA Nelson KE Eisen JA Paulsen IT Nelson W Heidelberg JF Wu D Wu M Ward N Beanan MJ Dodson RJ Madupu R Brinkac LM Daugherty SC DeBoy RT Durkin AS Gwinn M Kolonay JF Sullivan SA Haft DH Selengut J Davidsen TM Zafar N White O Tran B Romero C Forberger HA Weidman J Khouri H Feldblyum TV Utterback TR Van Aken SE Lovley DR Fraser CM Genome of Geobacter sulfurreducens: metal reduction in subsurface environments Science 2003 302 1967 1969 14671304 10.1126/science.1088727
Holmes DE Nevin KP Lovley DR Comparison of 16S rRNA, nifD, recA, gyrB, rpoB and fusA genes within the family Geobacteraceae fam. nov Int J Syst Evol Microbiol 2004 54 1591 1599 15388715 10.1099/ijs.0.02958-0
Lovley DR Phillips EJP Lonergan DJ Widman PK Fe(III) and S0 reduction by Pelobacter carbinolicus Appl Environ Microbiol 1995 61 2132 2138 7793935
Coppi MV Leang C Sandler SJ Lovley DR Development of a genetic system for Geobacter sulfurreducens Appl Environ Microbiol 2001 67 3180 3187 11425739 10.1128/AEM.67.7.3180-3187.2001
Kelley LA MacCallum RM Sternberg MJ Enhanced genome annotation using structural profiles in the program 3D-PSSM J Mol Biol 2000 299 499 520 10860755 10.1006/jmbi.2000.3741
Nikaido H Nakae T The outer membrane of Gram-negative bacteria Adv Microb Physiol 1979 20 163 250 394591
Wexler HM Pore-forming molecules in gram-negative anaerobic bacteria Clin Infect Dis 1997 25 Suppl 2 S284 6 9310708
Maier TM Myers CR The outer membrane protein Omp35 affects the reduction of Fe(III), nitrate, and fumarate by Shewanella oneidensis MR-1 BMC Microbiol 2004 4 23 15212692 10.1186/1471-2180-4-23
Koebnik R Locher KP Van Gelder P Structure and function of bacterial outer membrane proteins: barrels in a nutshell Mol Microbiol 2000 37 239 253 10931321 10.1046/j.1365-2958.2000.01983.x
Nikaido H Molecular basis of bacterial outer membrane permeability revisited Microbiol Mol Biol Rev 2003 67 593 656 14665678 10.1128/MMBR.67.4.593-656.2003
Schulz GE The structure of bacterial outer membrane proteins Biochim Biophys Acta 2002 1565 308 317 12409203
Benz R Permeation of hydrophilic solutes through mitochondrial outer membranes: review on mitochondrial porins Biochim Biophys Acta 1994 1197 167 196 8031826
Lloyd JR Leang C Hodges-Myerson AL Coppi MV Ciufo S Methe B Sandler SJ Lovley DR Biochemical and genetic characterization of PpcA, a periplasmic c-type cytochrome in Geobacter sulfurreducens Biochem J 2003 369 153 161 12356333 10.1042/BJ20020597
Myers CR Myers JM MtrB is required for proper incorporation of the cytochromes OmcA and OmcB into the outer membrane of Shewanella putrefaciens MR-1 Appl Environ Microbiol 2002 68 5585 5594 12406753 10.1128/AEM.68.6.2781-2793.2002
Duguay AR Silhavy TJ Quality control in the bacterial periplasm Biochim Biophys Acta 2004 1694 121 134 15546662 10.1016/j.bbamcr.2004.04.012
Wilken C Kitzing K Kurzbauer R Ehrmann M Clausen T Crystal structure of the DegS stress sensor: How a PDZ domain recognizes misfolded protein and activates a protease Cell 2004 117 483 494 15137941 10.1016/S0092-8674(04)00454-4
Butler JE Glaven RH Esteve-Nunez A Nunez C Shelobolina ES Bond DR Lovley DR A single bifunctional enzyme for fumarate reduction and succinate oxidation in Geobacter sulfurreducens and Geobacter metallireducens
Myers CR Myers JM Fumarate reductase is a soluble enzyme in anaerobically grown Shewanella putrefaciens MR-1 FEMS Microbiol Lett 1992 98 13 20 10.1016/0378-1097(92)90125-8
Chin KJ Nunez A Leang C Lovley DR Direct correlation between rates of anaerobic respiration and levels of mRNA for key respiratory genes in Geobacter sulfurreducens Appl Environ Microbiol 2004 70 5183 5189 15345398 10.1128/AEM.70.9.5183-5189.2004
Holmes DE Nevin KP Lovley DR In situ expression of nifD in Geobacteraceae in subsurface sediments Appl Environ Microbiol 2004 70 7251 7259 15574924 10.1128/AEM.70.12.7251-7259.2004
Caccavo FJ Lonergan DJ Lovley DR Davis M Stolz JF McInerney MJ Geobacter sulfurreducens sp. nov., a hydrogen- and acetate-oxidizing dissimilatory metal-reducing microorganism Appl Environ Microbiol 1994 60 3752 3759 7527204
Murphy KC Campellone KG Poteete AR PCR-mediated gene replacement in Escherichia coli Gene 2000 246 321 330 10767554 10.1016/S0378-1119(00)00071-8
Kovach ME Elzer PH Hill DS Robertson GT Farris MA Roop RM Peterson KM Four new derivatives of the broad-host-range cloning vector pBBR1MCS, carrying different antibiotic-resistance cassettes Gene 1995 166 175 176 8529885 10.1016/0378-1119(95)00584-1
Sambrook JE Fritsch F T. M Molecular cloning: a laboratory manual 1989 2nd New York, Cold Spring Harbor Laboratory Press
Marx CJ Lidstrom ME Development of improved versatile broad-host-range vectors for use in methylotrophs and other Gram-negative bacteria Microbiology 2001 147 2065 2075 11495985
Lovley DR Philips EJ Organic matter mineralization with the reduction of ferric iron in anaerobic sediments Appl Environ Microbiol 1986 51 683 689 16347032
Nikaido H Isolation of outer membranes Methods Enzymol 1994 235 225 234 8057896
Thomas PE Ryan D Levin W An improved staining procedure for the detection of the peroxidase activity of cytochrome P-450 on sodium dodecyl sulfate polyacrylamide gels Anal Biochem 1976 75 168 176 822747 10.1016/0003-2697(76)90067-1
Francis RTJ Becker RR Specific indication of hemoproteins in polyacrylamide gels using a double-staining process Anal Biochem 1984 136 509 514 6202169 10.1016/0003-2697(84)90253-7
Clauser KR Baker P Burlingame AL Role of accurate mass measurement (+/- 10 ppm) in protein identification strategies employing MS or MS/MS and database searching Anal Chem 1999 71 2871 2882 10424174 10.1021/ac9810516
El-Hage N Babb K Carroll JA Lindstrom N Fisher ER Miller JC Glimore RD Lamine JM Stevenson B Surface exposure and protease insensitivity of Borrelia burgdorferi Erp (OspEf-related) lipoproteins Microbiology 2001 147. 821 830 11283278
Berry EA Trumpower BL Simultaneous determination of hemes a, b, and c from pyridine hemochrome spectra Anal Biochem 1987 161 1 15 3578775 10.1016/0003-2697(87)90643-9
Smith PK Krohn RI Hermanson GT Mallia AK Gartner FH Provenzano MD Fujimoto EK Goeke NM Olson BJ C. KD Measurement of protein using bicinchoninic acid Anal Biochem 1985 150 76 85 3843705 10.1016/0003-2697(85)90442-7
|
16000176
|
PMC1186022
|
CC BY
|
2021-01-04 16:03:40
|
no
|
BMC Microbiol. 2005 Jul 6; 5:41
|
utf-8
|
BMC Microbiol
| 2,005 |
10.1186/1471-2180-5-41
|
oa_comm
|
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-6-461601881810.1186/1471-2202-6-46Research ArticleThe recombination activation gene 1 (Rag1) is expressed in a subset of zebrafish olfactory neurons but is not essential for axon targeting or amino acid detection Feng Bo [email protected] Sarada [email protected] Emre [email protected] Rainer W [email protected] Suresh [email protected] Developmental Neurobiology Group, Temasek LifeSciences Laboratory, 1 Research Link, The National University of Singapore, 117604, Singapore2 Max Planck Institute for Medical Research, Dept. of Biomedical Optics, Jahnstr. 29, D-69120 Heidelberg, Germany2005 15 7 2005 6 46 46 28 2 2005 15 7 2005 Copyright © 2005 Feng et al; licensee BioMed Central Ltd.2005Feng 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
Rag1 (Recombination activation gene-1) mediates genomic rearrangement and is essential for adaptive immunity in vertebrates. This gene is also expressed in the olfactory epithelium, but its function there is unknown.
Results
Using a transgenic zebrafish line and immunofluorescence, we show that Rag1 is expressed and translated in a subset of olfactory sensory neurons (OSNs). Neurons expressing GFP under the Rag1 promoter project their axons to the lateral region of the olfactory bulb only, and axons with the highest levels of GFP terminate in a single glomerular structure. A subset of GFP-expressing neurons contain Gαo, a marker for microvillous neurons. None of the GFP-positive neurons express Gαolf, Gαq or the olfactory marker protein OMP. Depletion of RAG1, by morpholino-mediated knockdown or mutation, did not affect axon targeting. Calcium imaging indicates that amino acids evoke chemotopically organized glomerular activity patterns in a Rag1 mutant.
Conclusion
Rag1 expression is restricted to a subpopulation of zebrafish olfactory neurons projecting to the lateral olfactory bulb. RAG1 catalytic activity is not essential for axon targeting, nor is it likely to be required for regulation of odorant receptor expression or the response of OSNs to amino acids.
==== Body
Background
Animals possess a number of chemosensory systems that enable them to perceive diverse stimuli in the environment. One such system is the olfactory system, which detects chemicals by a large number of olfactory sensory neurons (OSNs) in the nose. In mammals, each OSN expresses a single allele of one odorant receptor [1] on a dendrite that is exposed to the external world, and on an axon terminal that extends into the brain [2]. The projection of OSNs is highly ordered: all neurons expressing a given receptor converge to the same region in the ipsilateral olfactory bulb [3-5], terminating in a single glomerulus, i. e., a spherical area of dense synaptic neuropil. Guidance of axons is determined by a combination of factors, including the odorant receptors [6,7]. As a result of this well-ordered projection, chemical information is presented to the brain as spatial activity patterns across the array of glomeruli in the olfactory bulb [8,9].
Olfactory sensory neurons are morphologically diverse, consisting of ciliated, microvillous and crypt neurons. In fish, there appears to be some correlation between neuronal morphology, receptor class, G-protein type, and ligand spectrum [10,11]. Ciliated OSNs express receptors similar to those found in the main olfactory system of mammals express the Gαolf subunit and respond to amino acids or nucleotides. Microvillous neurons, on the other hand, express receptors from the V2R family found in the vomeronasal system of mammals, Gαo, Gαq or Gαi-3, and respond to amino acids or bile acids. In mammals, microvillous neurons detect pheromones, but can also respond to other odors [12]. Crypt neurons, which have a distinct rounded morphology, also contain Gαq in the apical region. Different classes of OSNs project to different regions of the olfactory bulb [10,13].
The recombination activation gene, Rag1, is expressed in the olfactory epithelium of mice and zebrafish [14,15]. The Rag1 gene, together with Rag2, is thought to have entered the genome of an ancestral organism 450 million years ago, soon after the divergence of jawed and jawless vertebrates; both genes have remained adjacent to one another throughout evolution. Acting together, RAG1 and RAG2 proteins function similarly to bacterial transposases such as Tn10 and are able to cleave DNA in a sequence-specific manner [16]. They mediate V(D)J recombination, and are thus responsible for the generation of antibodies and T-cell receptors [17]. As a result, each Rag1 expressing immune cell has a different identity, characterized by a permanent change to its genome, as well as the proteins expressed on its surface. Mutations in Rag1 [18] or Rag2 [19] lead to immunodeficiency.
The function of Rag1 in the olfactory system is unclear, although there has been some speculation that DNA rearrangement could be involved in odorant receptor expression [20]. As a step towards understanding the role of Rag1, we have initiated a study using the zebrafish, a model system with a relatively well-characterized olfactory system [21-23]. The transparency of the zebrafish larva enables gene expression analysis at single cell resolution, while genetics and morpholinos provide tools for assessing gene function using anatomical or physiological methods. Our results suggest that RAG1 protein is present in a subset of OSNs, including some microvillous olfactory neurons, but is not required for axon targeting.
Results
Rag1 expression in the zebrafish olfactory system
Using a transgenic zebrafish line in which the coding sequence of Rag1 within a PAC was replaced with GFP, Shuo Lin and colleagues have reported the expression of Rag1 in olfactory sensory neurons [15]. To confirm the fidelity of GFP expression, double-labelling with antibodies to zebrafish RAG1 and GFP was carried out. The specificity of the RAG1 antibody was first tested by preabsorption and immunofluorescence on thymocytes (Fig. 1A, B). Olfactory neurons from 4 day-old Rag1:GFP fish were then isolated, fixed and doubled-labelled. In these cells, two classes of GFP expression was seen – bright and dim. An overlap of RAG1 (red) and GFP (green) immunofluorescence (Fig. 1C–E) was observed. From 3 double-labelling experiments, 100 out of 105 cells with high levels of GFP were found to be RAG1-positive, with label predominantly in the nucleus. 91 out of 102 cells that expressed a lower level of GFP were also RAG1-positive. These observations establish that RAG1 protein is present in olfactory neurons, and that GFP expression in the transgenic line can serve as an indicator of endogenous RAG1.
Figure 1 Immunofluorescent labelling of RAG1. (A) Isolated zebrafish thymocytes, labelled with the antibody to RAG1 (in red) and DAPI (blue). (B) After pre-absorption with the peptide used for immunization, no labelling was detected. (C-E) Double-label of olfactory epithelial cells isolated from a Rag1:GFP transgenic fish. RAG1 protein (C, D; red) is present in the GFP-positive (C, E; green) neurons (arrow and arrowhead). Neurons with high (arrow) or low (arrowhead) GFP levels contain RAG1. DAPI (C; blue) is used to stain nuclei. Bar = 5 μm.
We imaged the Rag1:GFP fish at high resolution to determine in more detail the expression of Rag1 in the zebrafish olfactory system. GFP was first detectable at 22 hours post-fertilization, and both bright and dim cells could be seen in the olfactory epithelium (Fig. 2A). Initially, only short axons were visible. By 72 hours, the number of fluorescing cells increased and axons reached the olfactory bulb (OB; Fig. 2B). Fluorescing neurons projected axons to the lateral region of the OB (Fig. 2C–E). Intense labelling was detected in one target (Fig. 2E), which is a glomerular neuropil structure, as identified by labelling with the synaptic vesicle marker SV2 [24] (Fig. 2F). Synaptic vesicles co-localize with the termini of GFP-positive axons in this glomerular structure. (Fig. 2G).
Figure 2 Expression of Rag1 in zebrafish olfactory system. (A) At 22 hours post-fertilization, a few olfactory sensory neurons express GFP under the Rag1 promoter. (B) At 3 dpf, fluorescing axons have reached the bulb. A single target (arrowhead) is innervated by brightly labelled axons. (C, D) Frontal view of an 8-day old larva, with DiI labeling of olfactory sensory neurons (red) and Bodipy labeling of inter-cellular spaces (dim green). (C) A shallow section of the labelled forebrain, with DiI-labelled olfactory axons visible (arrows). In a deeper section (D) the GFP-containing axons (arrowhead) can be seen, along with other DiI-labelled axons (arrow). (E) Dorsal view of an isolated forebrain from a 4 day-old fish, showing the left olfactory bulb. Anterior is to the top and the midline is indicated by the dotted line. Strong GFP fluorescence is seen in axon terminals in a single region of the lateral bulb (arrowhead), while axons with lower levels of GFP innervate other regions of the lateral bulb (arrows). The inset shows one optical section, colour-coded according to fluorescence intensity. Termini with high (arrowhead) and low (arrow) intensity are indicated. (F) Frontal view of glomerular structures in the olfactory bulb of a 4-day fish, labelled with an antibody to synaptic vesicles. Only one lateral structure is innervated by OSNs with strong GFP expression (arrow). Lateral is to the left, while dorsal is to the top. (G) A single optical section through the glomerular target containing GFP-expressing neurons. The marker for synaptic vesicles (red) and GFP appear to co-localize, as indicate by the linescan. (H) An olfactory pit labeled with DiI. The GFP-expressing cells (green) have not taken up DiI (red). (I, J) A Di8ANEPPQ-labeled olfactory system of a Rag1:GFP transgenic fish. In the olfactory bulb (I), some axons with strong GFP expression (green) are also labeled with Di8ANEPPQ (red; arrow), whereas others are not (arrowhead). (J) In this section through the olfactory epithelium, a GFP-expressing neuron was labeled with Di8ANEPPQ (arrow), whereas two others were not (arrowheads). (K) The olfactory pit of a fish transgenic for Rag1:GFP and omp:tauDsRed. GFP expressing cells (green) are distinct from those labeled with DsRed. Panels A, B, E, F and K are projections reconstructed from Z stacks. ob: olfactory bulb; op: olfactory pit. Bar = 20 μm (A, B, E, F, H, I, K); 50 μm (C, D,); 5 μm (G, J).
When the fluorescent lipophilic dye DiI was placed in the fish water, many OSNs took up the dye (Fig. 2C, D). However, few (13.8%; n = 58) strongly GFP-expressing neurons were labelled (Fig. 2H). This suggests a low accessibility to the external environment within this subpopulation. With Di8ANEPPQ, a lipophilic dye with higher solubility, a higher proportion (53.8%; n = 26) of strong GFP-positive cells were labelled (Fig. 2I, J), confirming that they do have some access to the water. We then examined Rag1 expression in fish expressing tauDsRed under the control of a promoter fragment of the olfactory marker protein (OMP) gene [25]. In this line, tauDsRed is expressed in most OSNs except for a subpopulation projecting to a lateral area in the developing OB [25]. In Rag1:GFP, omp:tauDsRed double transgenic fish, GFP was detected in OSNs that were tauDsRed-negative (Fig. 2K). The GFP-positive neurons appeared to have a shorter cell body than those expressing tauDsRed.
When labelled with antibodies to different G-alpha subunits, cells with strong GFP expression were Gαolf, Gαo and Gαq-negative (Fig. 3). Crypt cells or other neurons that were labelled by the Gαq antibody were GFP-negative (Fig. 3D, E). 36 out of 85 neurons containing Gαo expressed GFP at low levels (Fig. 3A–C). The remaining cells had no GFP. An additional 13 cells had low GFP expression, but did not express Gαo. These observations indicate that the GFP-expressing cells are heterogenous in nature.
Figure 3 Expression of G alpha subunits in larval olfactory neurons.(A-C) Gαo (red fluorescence, C) is present in a subset of Rag1-expressing cells (green fluorescence, B; arrowheads) isolated from the olfactory epithelium of a Rag1:GFP zebrafish. A cell with bright GFP (arrow), however, does not contain Gαo. (D) Gαq-positive neurons (arrows) do not express GFP, and vice-versa. Axons are strongly labelled (arrowhead). (E) A crypt cell (arrowhead), with Gαq label in the dendrites, lacks GFP. Two cells with bright GFP (arrows), in contrast, lack Gαq label. (F) Gαolf label, seen here in axons and some cell bodies (arrowheads), was not detected in cells strongly expressing GFP (arrows). Bar = 5 μm (A-E); 10 μm (F).
In the adult, GFP expression could still be detected in the olfactory system. Most GFP-positive neurons had their cells bodies close to the apical surface of the olfactory epithelium, with dendrites of intermediate length; a few cells, however, had their soma deep in the epithelium (Fig 4A, B). GFP-positive axons project to the lateral olfactory bulb only, as can be seen in dissected forebrains, or using immunofluorescence on sections (Fig. 4C). A subset of axons projecting to one target in the lateral bulb contained particularly high levels of GFP (Fig. 4D).
Figure 4 Expression of Rag1 in the adult zebrafish olfactory system.(A) An olfactory rosette isolated from a 1.5 year-old adult. GFP-expressing cells are located in the sensory region of the lamella and midline raphe. Most cell bodies are located apically. (B) High magnification of lamella from another olfactory rosette, also from a 1.5 year-old adult. GFP-expressing cells have differing morphologies. Many have the cell body close to the apical surface (arrowheads). The arrow indicates one neuron with a deep cell body. (C) Anti-GFP immunofluorescence on a horizontal section through the olfactory bulb of an adult. Label is detectable only in olfactory sensory neurons innervating the lateral bulb. The incoming olfactory nerve is visible at the upper left (arrow). Anterior is to the top-left; lateral to the bottom-left. Nuclei are labelled with propidium iodide (red). (D) Lateral view of an olfactory bulb dissected from a 1.5-year-old Rag1:GFP male. Axons innervate a large portion of the lateral bulb, but axons with the highest levels of GFP (black arrowheads) form two bundles that innervate a single region (arrow). Other glomeruli are innervated by dimmer axons. Bar = 50 μm (A, C, D); 20 μm (B).
The effect of RAG1 knockdown on axon targeting
The expression of GFP in a subpopulation of neurons that project to one region of the olfactory bulb raises the possibility that Rag1 could be involved in specifying the identity, and hence the axonal targeting, of these neurons. To test this idea, three morpholinos were used to knock down Rag1 in embryos. The first morpholino (mo1) targeted the start codon of RAG1 and was tested using a fusion of the 5' end of RAG1 to GFP. Injection of mRNA for the fusion protein alone led to a strong fluorescence (Fig. 5A), whereas co-injection with the morpholino suppressed fluorescence (Fig. 5B), indicating that mo1 knocked down Rag1 translation. The second morpholino (mo2) was targeted to the splice donor site of the first exon (Fig. 5E). Injection of this morpholino led to aberrant splicing (Fig. 5F), resulting in an mRNA with a premature stop codon (data not shown). The third morpholino (mo3) was designed to the donor site of the second exon (Fig. 5E), and this resulted in loss of the normal transcript (not shown). Splice donor morpholinos were tested because the use of internal ATGs, which has been reported for some forms of Omenn's syndrome [26], might obscure the effects of mo1.
Figure 5 The effect of RAG1 depletion on the olfactory projection. (A) An embryo injected with mRNA encoding the 5' end of Rag1 fused to EGFP. (B) An embryo co-injected with the Rag1-EGFP fusion mRNA and mo1. (C) Olfactory neurons labelled with GFP under the Rag1 promoter, with brightly labeled axons projecting to a single target (arrowhead), at 3 dpf. (D) In a transgenic embryo injected with mo1, axons still project to the same target (arrowhead), but the intensity of GFP fluorescence is reduced. (E) A schematic diagram of the Rag1 gene, showing the location of morpholinos and primers that were used to analyse morpholino-injected fish. (F) RT-PCR on control or mo2 injected embryos. Abnormal splicing occurs in the morpholino-injected fish, leading to a premature stop codon, as indicated by sequencing of the upper band. (G) RT-PCR after injection of a mixture of Rag1 mo2 and mo3, showing loss of the normal transcript. (H, I) A subset of Di8ANEPPQ-labeled olfactory sensory neurons in 7-day old wild type and Rag1 mutant fish. Axons innervate all target structures detectable in this optical plane in the mutant. (J-L) SV2-labelled 4 day-old Rag1:GFP transgenic (J, K) and Rag1 mutant (L) forebrains, shown in dorsal view. The images are colour-coded according to depth. The glomerulus innervated by the strong GFP-positive neurons (J) is indicated by the arrowhead. Bar = 100 μm (A, B); 20 μm (C, D, H – K). The colour bars in J and K indicate depth. Embryos in panels H-L are shown in dorsal view, with anterior to the left.
Injection of mo1 caused a reduction in the level of GFP driven by the Rag1 promoter. This occurred despite there being an 8-nucleotide mismatch in the transgene (since GFP replaces the Rag1 coding sequence), presumably because the mismatch is contiguous and exclusively at one end of the morpholino. The convergence of brighter axons to one lateral target persisted (Fig. 4C, D; n = 15). Aside from the decrease in GFP expression, no other defect could be detected. With mo2, injections of 0.8 μM caused slight non-specific morphological abnormalities, such as small eyes. Injection with 0.8 μM mo3 caused stronger abnormalities. Less abnormality was seen when a mixture of 0.6 μM mo2 and 0.6 μM mo3 was injected. This led to a dramatic loss of normally spliced Rag1 mRNA (Fig. 5G). Injection of either mo2 or mo3 or the mixture did not cause any defect in the targeting of the GFP-expressing neurons (n = 50 embryos each). Hence strong Rag1 expression is not required for innervation of the lateral target.
As an additional step in analysing the role of Rag1 in establishing the olfactory projection, a line carrying a point mutation, leading to a STOP codon within the RAG1 catalytic domain [27], was examined. The entire projection was labeled with the lipophilic tracer Di8ANEPPQ. No obvious difference could be seen between mutants and wild types (Fig. 4H, I). To further analyse mutants, the forebrains of 4 day-old larvae were labelled with the SV2 antibody. The lateral neuropil structure that is innervated by neurons with strong GFP expression could be visualized in mutants, as in wild types (Fig. 5 J-L). Together with the morpholino data, these observations indicate that Rag1 is not required for pathfinding of OSN axons to the olfactory bulb, or for establishment of glomeruli.
Odor-evoked activity maps in the Rag1 mutant
The spatial organization of OSN axons in the olfactory bulb and their functionality was further examined by calcium imaging of odor-evoked activity maps in 3 month-old fish. OSN axons were loaded with the calcium indicator, Calcium Green-1-dextran. Odor-evoked activity in OSN axons leads to calcium influx at axon terminals, causing a localized change in indicator fluorescence. This method allows for the optical detection of odor-evoked activity selectively in OSN axons within glomeruli [9]. In wild type fish, amino acids evoke activity predominantly in the ventro-lateral region of the OB that contains densely packed, small glomeruli [9]. Within this region, clusters of glomeruli can be identified by their response to amino acids sharing particular chemical features (e. g., a short neutral side chain, a long neutral side chain, or a basic side chain). Response patterns vary considerably between individuals, but functionally defined clusters are found at similar relative positions along the anterior-to-posterior axis in different individuals, thus establishing a conserved chemotopic map [9].
The distribution and magnitude of response patterns evoked by a diagnostic set of L-amino acids with different chemical features (Met, Ala, Val, Lys, Phe, His, Trp; each 10-5 M) were generally similar in wild type and Rag1 mutant fish (Fig. 6A, B). However, in both wild type and Rag1 mutants, the fine structure of these activity patterns varied between individuals. To examine the chemotopic organization in more detail, we therefore defined three glomerular regions by characteristic response properties, as done previously to identify chemotopically organized glomerular clusters [9]. The first region was defined by responses to the basic amino acid, Lys, and partially to Trp (Fig. 6A, B; red arrowheads), the second region was defined by responses to all short-chain neutral stimuli (Phe, Trp and Ala; Fig. 6A, B; yellow arrowheads), and the third region was defined by the response to the long-chain neutral amino acids, Val and Met (Fig. 6A, B; light blue arrowhead). Regions with these functional properties were identifiable in each wild type and Rag1 mutant fish. Their spatial arrangement was analyzed by overlaying the outlines of the regions from different individuals. In both wt (n = 7) and Rag1 mutant (n = 10) fish, the three clusters occurred in a characteristic sequence along the anterior-to-posterior axis (Fig. 6C, D), as described previously [9]. Hence, the chemotopic organization of glomerular response maps was comparable in wild type and Rag1 mutant fish. Other response properties, such as wide-spread responses to food extracts [28], were also similar between wild types and Rag1 mutants (not shown). We also quantitatively analyzed the similarity relationships between activity patterns evoked by different stimuli using correlation and factor analysis [9] and observed no obvious differences between wild type and Rag1 mutant fish. Hence, physiological responses of OSNs to amino acids and their chemotopically organized projections to the OB were not detectably altered in Rag1 mutants.
Figure 6 Chemotopic organization of glomerular activity patterns in wild type and Rag1 mutant fish. (A) Glomerular activity patterns evoked by different amino acid stimuli (10-5 M) in the ventro-lateral OB of a wild type fish. Changes in fluorescence intensity report activity of OSN axon terminals and are colour coded. Dashed line depicts lateral edge of the OB. Arrowheads indicate conserved clusters of glomeruli with defined response properties. (B) Glomerular activity patterns evoked by the same stimuli in a Rag1 mutant fish. The same general areas of the bulb show a response. The difference in the intensity seen here between wild type and mutants is not statistically significant. (C) Overlay of positions of identifiable glomerular regions in wild type of the line from which Rag1 mutants were derived (n = 7). Regions were outlined manually in activity maps in each fish. Outlines from different individuals were centered on the central cluster (yellow). (D) Overlay of cluster positions determined in 10 Rag1 mutant fish.
Discussion
To delineate possible functions of Rag1 in the zebrafish olfactory system, we have examined its expression pattern in detail and examined the effect of RAG1 loss. Immunofluorescence using an antibody to zebrafish RAG1 indicates that the protein is present in OSNs. Imaging of a transgenic line, which is a sensitive method to monitor gene expression in neurons [29], suggests that Rag1 is expressed only in a subset of OSNs. Some of the neurons expressing Rag1 are likely to be microvillous neurons, as indicated by their morphology and Gαo expression. Neurons expressing GFP project to the lateral bulb, which is a region of the bulb known to contain termini of amino-acid sensitive OSNs [9]. Their projection pattern appears to be complementary to neurons expressing a reporter under the OMP promoter, as these project to the medial bulb [30]. Rag1 expression thus appears to mark a subset of distinct OSNs. These neurons are likely to be heterogenous, given that not all of them express Gαo.
A feature of the Rag1:GFP transgenic line analyzed here is that one glomerular target in each bulb is more brightly labelled than other targets. One interpretation of this observation, given that some cell bodies, axons and termini are brighter than others, is that neurons expressing GFP at high levels all have a particular identity and project to a single target. These neurons do not express any of the markers tested, possibly because of their distinct identity. Alternatively, they may not be fully mature when GFP is expressed strongly. This would be consistent with the finding that not all bright neurons take up DiI or Di8ANEPPQ from the external environment, which would be the case if their dendrites were immature. These neurons may eventually have reduced levels of GFP, and express Gαo. It should be noted that high levels of GFP expression cannot be a feature of all immature OSNs. If this were the case, it would be expected that there would be bright neurons projecting to all glomeruli; this was never seen. Although we cannot yet make any firm conclusions as to why some axons are brighter than others, their reproducible targeting can serve as a useful marker in genetic screens or in physiological analysis.
In the olfactory system, sensory neurons are distinguished from one another on the basis of odorant receptor expression, and targeting to a particular glomerulus is dependent on the odorant receptor [6], amongst other axon guidance cues. It has been suggested that odorant receptor expression could be regulated by a mechanism similar to V(D)J recombination, given the genomic organization of the receptors. We found that axon targeting and chemotopic activity maps in the OB were not affected by knockdown of Rag1 or by a mutation that abolishes the catalytic activity of Rag1. Hence, it appears that RAG1 does not have a role in defining the identity of OSNs. Recent cloning experiments show that the genome of mammalian OSNs is not rearranged [31,32]. Those experiments were carried out with neurons expressing receptors of the OR class, i.e. those expressing Gαolf, which is probably not expressed in Rag1-positive OSNs in zebrafish. The observations described here suggest that this conclusion also holds for other olfactory neurons.
RAG1 contains the active site for V(D)J recombination [33,34] and can cleave DNA in vitro on its own, although it does so with low efficiency [35]. Once DNA has been cleaved, RAG1 can bind to the broken ends and serve a protective role [36]. Given the relatively high rate of DNA breaks in neurons [37], one possibility is that RAG1 protects the genome in some neurons. However, this view does not explain why only a subset of neurons expresses Rag1. Another function of RAG1, provided by the RING-finger, is as an E3 ligase [38], raising the possibility that this activity may be required in a subset of receptor neurons. Additionally, RAG1 can independently act as an endonuclease [39]. It is apparent that RAG1 can carry out several functions, but it is unknown at present whether the expression of Rag1 in the olfactory epithelium reflects any of these abilities. The expression described here suggests that any function of Rag1 must be restricted to a subset of OSNs only.
Conclusion
This study demonstrates that Rag1 is expressed and translated in a subset of zebrafish olfactory neurons projecting to the lateral olfactory bulb. Some of these neurons express a marker of microvillous neurons. Rag1 does not appear to be essential for axon targeting or receptor expression.
Methods
RAG1 antibody
A rabbit polyclonal antibody against the C-terminus of zebrafish Rag1 (amino acids 1042~1057) was generated by ZYMED Laboratory Inc. The peptide antigen (CEETPEEADNSLDVPDF) was synthesized by Tufts University Core Facility. To test for specificity, the antibody was incubated overnight at 4°C with the peptide at a concentration of 66.6 μg/ml. After a 30 minute spin at 16 000 g, the supernatant was used for labelling thymocytes, which had been dissected from freshly killed 2 week-old zebrafish using tungsten needles.
Immunofluorescent labelling of olfactory neurons
4 day-old Rag1:GFP fish were fixed in 4% PFA-PBS for 10 minutes. The olfactory epithelium was dissected out in Ringer's solution and placed on Superfrost/Plus slides using a mouth pipette (Sigma A5177). The olfactory epithelium was allowed to semi-dry and adhere to the slide. A drop of Ringer's solution was added and a cover slip was used to squash the epithelium gently, thus dispersing the cells and allowing them to adhere to the slide. The cover slip was gently removed and the cells were re-fixed in 4% PFA-PBS for another 5 minutes. They were then rinsed with PBS and permeabilised with 1% Triton X-100. The following antibodies were used: anti-zf RAG1 (rabbit, 1:200), anti-GFP (mouse, 1:50; Molecular Probes), anti Gαo (guinea pig, 1:200), Gαq and Gαolf (both rabbit, 1:200; Santa Cruz Biotech). The Gαo antibody [40] was made to a region of the protein that is 100% conserved between Drosophila (residues 345–354) and zebrafish. For detection, 1:500 Cy3 anti-guinea pig, 1:300 AlexaFluor 568 anti rabbit and 1:300 AlexaFluor 488 anti mouse (Molecular Probes) were used. Nuclei were stained with 100 ng/ml DAPI. Cells were imaged with confocal microscopy. At least 50 neurons with strong GFP expression were analysed for each G-alpha subunit labelling.
Immunofluorescent labeling of glomeruli
The brain of 4-day old Rag1:GFP fish larvae was dissected out in Ringers and fixed for 1 hour in 4% PFA-PBS. Wholemount antibody labelling was carried out using standard procedures, using the SV2 antibody (Developmental Studies Hybridoma Bank) at 1:500 dilution, and the AlexaFluor 546 goat anti-mouse antibody (Molecular Probes) at 1:500 dilution.
Immunostaining of cryo-sectioned tissue
The brains of 3 month-old Rag1:GFP transgenic fish were dissected out in PBS and fixed in 4% PFA, then embedded in tissue freezing medium (Jung) and sectioned. The sections were incubated in rabbit anti-GFP (1:200) followed by AlexaFluor 488 anti-rabbit (1:300). Nuclei were stained with propidium iodide.
Constructs
5'RAG1-EGFP: A 350 bp 5' Rag1 DNA fragment was amplified by RT-PCR with primer Rag1a: CTCTCAATTCATAAAAAATAAATCTTAC and Rag1b: GGTCCACTCTCCCTCGAG, digested with Hae III and inserted into Smal I site of pEGFP-N1 (Clontech). For in vitro transcription of the fusion mRNA, the 5'RAG1-EGFP fragment was cut out and cloned into pCS2 [41].
Imaging
Live zebrafish embryos and larvae were embedded in 1.5 % low-melting temperature agarose (BioRad) and imaged with Zeiss LSM 510 laser scanning confocal microscopy, using 40 × (0.8 NA) or 63 × (1.2 NA) water immersion objectives. Isolated olfactory bulbs from larvae were imaged with a 20 × (0.5 NA) water immersion objective.
Dissected olfactory bulbs and rosettes from adults were imaged using widefield fluorescence, with 10x or 40xW objectives. Images were deconvolved using AutoDeblur (AutoQuant Inc.)
Lipophilic tracing of olfactory neurons
A saturated stock solution of DiI or Di8ANEPPQ (Molecular Probes) was prepared in ethanol. DiI was diluted 1:1000 in E3, while Di8ANEPPQ was diluted 1:5000, just before use. Larvae were placed in this solution in mesh baskets for 3 minutes, and then rinsed several times in fresh E3. Bodipy labelling was carried out as described [21].
Morpholino injection
Morpholinos (Gene-Tools) were diluted in H2O and injected into one-cell stage embryos. 50~100 embryos were injected with each oligo, with single morpholino concentrations ranging from 0.2 to 1.0 μM. Injection volume was approximately 1 nl. The following morpholinos were used:
mo1: 5'-TTCTCCATGGCGTCAGCTTATTCTC-3' (targets the start codon). mo2: 5'-TATTATACTCACTTGAGAAGATTCA-3' (targets the donor site of the first intron). mo3: 5'-TCTTGGCAGTACCTTGCATCATTGC-3'(targets the donor site of the second intron).
Genotyping of the Rag1 mutant
Allele specific PCR [42] was carried out using the following primers:
Rag1f1: CACTggCCCATgCTCCgATAgACC;
Rag1r1: TCCGGGGCACAGGCTATGATGAGAA;
Rag1wtr: GCTTAGCAGAAACACCTTTGACTCg;
Rag1mutr: GCTTAGCAGAAACACCTTTGACTCa.
Imaging of activity in OSN axon terminals
OSNs of adult zebrafish were loaded with Calcium Green-1 dextran (10 kD; Molecular Probes) and imaged as described previously [9]. Briefly, 6–8% of the dye in 3mM NaCl and 0.1%Triton X-100 was applied to each naris while the fish was anesthetised with 0.01% MS-222. Following a 5-minute incubation, the dye was washed away and the fish were allowed to recover. 3–5 days post labelling, the brain and nose were dissected, mounted upside-down in a custom-made perfusion chamber, and superfused with teleost artificial cerebrospinal fluid (ACSF). Amino acid solutions and extracts of commercially available fish foods were applied through a constant stream of ACSF to one naris using a computer controlled, pneumatically actuated HPLC injection valve (Rheodyne) as described [9,28].
Ca2+ imaging
Fluorescence in the olfactory bulb was recorded using a CCD camera (CoolSnapHQ; Photometrics) mounted on a custom-built upright epifluorescence microscope equipped with a 20x lens (NA 0.95; Olympus) [43]. Fluorescence was excited with a stabilized 150 W Xe arc lamp attenuated to 1.5% of full intensity by neutral density filters to minimize bleaching and phototoxicity. Images were binned to 87 × 65 pixels, acquired at 2 – 10 Hz, and digitized with 12 bits. Each pixel value was converted to a value representing the relative change in fluorescence (ΔF/F) after stimulus application. The baseline fluorescence (F) was calculated by averaging the frames before stimulus onset. Response maps were obtained by averaging ΔF/F frames over 2–4 s after response onset and mild spatial filtering.
Authors' contributions
FB documented the specific targeting of Rag1-expressing neurons in embryos and carried out morpholino experiments. SB conducted the immunofluorescence. SB and EY performed the calcium imaging, with critical help from RWF. RWF designed and analysed the physiological experiments, and helped draft the manuscript. SJ conceived the study, carried out the glomeruli labelling and imaging of adults, and drafted the manuscript.
Acknowledgements
We would like to express our thanks to Masayoshi Mishina for providing the OMP:tauDsRed construct, Stefan Schulte-Merke for the Rag1 mutant line, Shuo Lin for the Rag1:GFP line and the Qian Hu Fish Co. for help with fish transport. The SV2 antibody developed by Kathleen Buckley was obtained from the Developmental Studies Hybridoma Bank. This work was supported by the Temasek Life Sciences Laboratory, Singapore Millenium Foundation, Max-Planck Society, DFG (SFB 488) and the Boehringer Ingelheim Fonds.
==== Refs
Chess A Simon I Cedar H Axel R Allelic inactivation regulates olfactory receptor gene expression Cell 1994 78 823 834 8087849 10.1016/S0092-8674(94)90562-2
Barnea G O'Donnell S Mancia F Sun X Nemes A Mendelsohn M Axel R Odorant receptors on axon termini in the brain Science 2004 304 1468 15178793 10.1126/science.1096146
Mombaerts P Wang F Dulac C Chao SK Nemes A Mendelsohn M Edmondson J Axel R Visualizing an olfactory sensory map Cell 1996 87 675 686 8929536 10.1016/S0092-8674(00)81387-2
Ressler KJ Sullivan SL Buck LB Information coding in the olfactory system: evidence for a stereotyped and highly organized epitope map in the olfactory bulb Cell 1994 79 1245 1255 7528109 10.1016/0092-8674(94)90015-9
Vassar R Chao SK Sitcheran R Nunez JM Vosshall LB Axel R Topographic organization of sensory projections to the olfactory bulb Cell 1994 79 981 991 8001145 10.1016/0092-8674(94)90029-9
Wang F Nemes A Mendelsohn M Axel R Odorant receptors govern the formation of a precise topographic map Cell 1998 93 47 60 9546391 10.1016/S0092-8674(00)81145-9
Cutforth T Moring L Mendelsohn M Nemes A Shah NM Kim MM Frisen J Axel R Axonal ephrin-As and odorant receptors: coordinate determination of the olfactory sensory map Cell 2003 114 311 322 12914696 10.1016/S0092-8674(03)00568-3
Stewart WB Kauer JS Shepherd GM Functional organization of rat olfactory bulb analysed by the 2-deoxyglucose method J Comp Neurol 1979 185 715 734 447878 10.1002/cne.901850407
Friedrich RW Korsching SI Combinatorial and chemotopic odorant coding in the zebrafish olfactory bulb visualized by optical imaging Neuron 1997 18 737 752 9182799 10.1016/S0896-6273(00)80314-1
Hansen A Rolen SH Anderson K Morita Y Caprio J Finger TE Correlation between olfactory receptor cell type and function in the channel catfish J Neurosci 2003 23 9328 9339 14561860
Hansen A Anderson KT Finger TE Differential distribution of olfactory receptor neurons in goldfish: structural and molecular correlates J Comp Neurol 2004 477 347 359 15329885 10.1002/cne.20202
Sam M Vora S Malnic B Ma W Novotny MV Buck LB Neuropharmacology: Odorants may arouse instinctive behaviours Nature 2001 412 142 11449261 10.1038/35084137
Lipschitz DL Michel WC Amino acid odorants stimulate microvillar sensory neurons Chem Senses 2002 27 277 286 11923189 10.1093/chemse/27.3.277
Chun JJ Schatz DG Oettinger MA Jaenisch R Baltimore D The recombination activating gene-1 (RAG-1) transcript is present in the murine central nervous system Cell 1991 64 189 200 1986864 10.1016/0092-8674(91)90220-S
Jessen JR Willett CE Lin S Artificial chromosome transgenesis reveals long-distance negative regulation of rag1 in zebrafish Nat Genet 1999 23 15 16 10471489
Agrawal A Eastman QM Schatz DG Transposition mediated by RAG1 and RAG2 and its implications for the evolution of the immune system Nature 1998 394 744 751 9723614 10.1038/29457
Tonegawa S Somatic generation of antibody diversity Nature 1983 302 575 581 6300689 10.1038/302575a0
Mombaerts P Iacomini J Johnson RS Herrup K Tonegawa S Papaioannou VE RAG-1-deficient mice have no mature B and T lymphocytes Cell 1992 68 869 877 1547488 10.1016/0092-8674(92)90030-G
Shinkai Y Rathbun G Lam KP Oltz EM Stewart V Mendelsohn M Charron J Datta M Young F Stall AM Alt FW RAG-2-deficient mice lack mature lymphocytes owing to inability to initiate V(D)J rearrangement Cell 1992 68 855 867 1547487 10.1016/0092-8674(92)90029-C
Chess A Olfactory receptor gene regulation Adv Immunol 1998 69 437 447 9646850
Dynes JL Ngai J Pathfinding of olfactory neuron axons to stereotyped glomerular targets revealed by dynamic imaging in living zebrafish embryos Neuron 1998 20 1081 1091 9655497 10.1016/S0896-6273(00)80490-0
Hansen A Zeiske E Development of the olfactory organ in the zebrafish, Brachydanio rerio J Comp Neurol 1993 333 289 300 8345108 10.1002/cne.903330213
Korsching SI Argo S Campenhausen H Friedrich RW Rummrich A Weth F Olfaction in zebrafish: what does a tiny teleost tell us? Semin Cell Dev Biol 1997 8 181 187 15001094 10.1006/scdb.1996.0136
Buckley K Kelly RB Identification of a transmembrane glycoprotein specific for secretory vesicles of neural and endocrine cells J Cell Biol 1985 100 1284 1294 2579958 10.1083/jcb.100.4.1284
Yoshida T Ito A Matsuda N Mishina M Regulation by protein kinase A switching of axonal pathfinding of zebrafish olfactory sensory neurons through the olfactory placode-olfactory bulb boundary J Neurosci 2002 22 4964 4972 12077193
Santagata S Gomez CA Sobacchi C Bozzi F Abinun M Pasic S Cortes P Vezzoni P Villa A N-terminal RAG1 frameshift mutations in Omenn's syndrome: internal methionine usage leads to partial V(D)J recombination activity and reveals a fundamental role in vivo for the N-terminal domains Proc Natl Acad Sci U S A 2000 97 14572 14577 11121059 10.1073/pnas.97.26.14572
Wienholds E Schulte-Merker S Walderich B Plasterk RH Target-selected inactivation of the zebrafish rag1 gene Science 2002 297 99 102 12098699 10.1126/science.1071762
Tabor R Yaksi E Weislogel JM Friedrich RW Processing of odor mixtures in the zebrafish olfactory bulb J Neurosci 2004 24 6611 6620 15269273 10.1523/JNEUROSCI.1834-04.2004
Gong S Zheng C Doughty ML Losos K Didkovsky N Schambra UB Nowak NJ Joyner A Leblanc G Hatten ME Heintz N A gene expression atlas of the central nervous system based on bacterial artificial chromosomes Nature 2003 425 917 925 14586460 10.1038/nature02033
Miyasaka N Sato Y Yeo SY Hutson LD Chien CB Okamoto H Yoshihara Y Robo2 is required for establishment of a precise glomerular map in the zebrafish olfactory system Development 2005 132 1283 1293 15716341 10.1242/dev.01698
Li J Ishii T Feinstein P Mombaerts P Odorant receptor gene choice is reset by nuclear transfer from mouse olfactory sensory neurons Nature 2004 428 393 399 15042081 10.1038/nature02433
Eggan K Baldwin K Tackett M Osborne J Gogos J Chess A Axel R Jaenisch R Mice cloned from olfactory sensory neurons Nature 2004 428 44 49 14990966 10.1038/nature02375
Kim DR Dai Y Mundy CL Yang W Oettinger MA Mutations of acidic residues in RAG1 define the active site of the V(D)J recombinase Genes Dev 1999 13 3070 3080 10601033 10.1101/gad.13.23.3070
Landree MA Wibbenmeyer JA Roth DB Mutational analysis of RAG1 and RAG2 identifies three catalytic amino acids in RAG1 critical for both cleavage steps of V(D)J recombination Genes Dev 1999 13 3059 3069 10601032 10.1101/gad.13.23.3059
Oettinger MA Schatz DG Gorka C Baltimore D RAG-1 and RAG-2, adjacent genes that synergistically activate V(D)J recombination Science 1990 248 1517 1523 2360047
Tsai CL Drejer AH Schatz DG Evidence of a critical architectural function for the RAG proteins in end processing, protection, and joining in V(D)J recombination Genes Dev 2002 16 1934 1949 12154124 10.1101/gad.984502
Karanjawala ZE Murphy N Hinton DR Hsieh CL Lieber MR Oxygen metabolism causes chromosome breaks and is associated with the neuronal apoptosis observed in DNA double-strand break repair mutants Curr Biol 2002 12 397 402 11882291 10.1016/S0960-9822(02)00684-X
Yurchenko V Xue Z Sadofsky M The RAG1 N-terminal domain is an E3 ubiquitin ligase Genes Dev 2003 17 581 585 12629039 10.1101/gad.1058103
Kim DR Recombination activating gene 1 product alone possesses endonucleolytic activity J Biochem Mol Biol 2003 36 201 206 12689520
Wolfgang WJ Quan F Thambi N Forte M Restricted spatial and temporal expression of G-protein alpha subunits during Drosophila embryogenesis Development 1991 113 527 538 1782864
Rupp RA Snider L Weintraub H Xenopus embryos regulate the nuclear localization of XMyoD Genes Dev 1994 8 1311 1323 7926732
Glisic S Alavantic D A simple PCR method for detection of defined point mutations Trends Genet 1996 12 391 392 8909130 10.1016/S0168-9525(96)90094-3
Wachowiak M Denk W Friedrich RW Functional organization of sensory input to the olfactory bulb glomerulus analyzed by two-photon calcium imaging Proc Natl Acad Sci U S A 2004 101 9097 9102 15184670 10.1073/pnas.0400438101
|
16018818
|
PMC1186023
|
CC BY
|
2021-01-04 16:39:09
|
no
|
BMC Neurosci. 2005 Jul 15; 6:46
|
utf-8
|
BMC Neurosci
| 2,005 |
10.1186/1471-2202-6-46
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1611033810.1371/journal.pgen.001001105-PLGE-RA-0031R1plge-01-02-03Research ArticleDevelopmentNeuroscienceOphthalmologyPathologyGenetics/Disease ModelsMus (Mouse)Homo (Human)A Hybrid Photoreceptor Expressing Both Rod and Cone Genes in a Mouse Model of Enhanced S-Cone Syndrome Cone Gene Derepression in the
rd7 Mouse
Corbo Joseph C 123¤Cepko Constance L 13*1 Department of Genetics, Harvard Medical School, Boston, Massachusetts, United States of America
2 Department of Pathology, Brigham and Women's Hospital, Boston, Massachusetts, United States of America
3 Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts, United States of America
Barsh Gregory EditorStanford University School of Medicine, United States of America*To whom correspondence should be addressed. E-mail: [email protected]¤Current address: Department of Pathology and Immunology, Washington University School of Medicine, St. Louis, Missouri, United States of America
8 2005 5 8 2005 1 2 e1125 2 2005 2 5 2005 Copyright: © 2005 Corbo and Cepko.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.Rod and cone photoreceptors subserve vision under dim and bright light conditions, respectively. The differences in their function are thought to stem from their different gene expression patterns, morphologies, and synaptic connectivities. In this study, we have examined the photoreceptor cells of the retinal degeneration 7
(rd7) mutant mouse, a model for the human enhanced S-cone syndrome (ESCS). This mutant carries a spontaneous deletion in the mouse ortholog of NR2E3, an orphan nuclear receptor transcription factor mutated in ESCS. Employing microarray and in situ hybridization analysis we have found that the rd7 retina contains a modestly increased number of S-opsin–expressing cells that ultrastructurally appear to be normal cones. Strikingly, the majority of the photoreceptors in the rd7 retina represent a morphologically hybrid cell type that expresses both rod- and cone-specific genes. In addition, in situ hybridization screening of genes shown to be up-regulated in the rd7 mutant retina by microarray identified ten new cone-specific or cone-enriched genes with a wide range of biochemical functions, including two genes specifically involved in glucose/glycogen metabolism. We suggest that the abnormal electroretinograms, slow retinal degeneration, and retinal dysmorphology seen in humans with ESCS may, in part, be attributable to the aberrant function of a hybrid photoreceptor cell type similar to that identified in this study. The functional diversity of the novel cone-specific genes identified here indicates molecular differences between rods and cones extending far beyond those previously discovered.
Synopsis
Vision begins with light entering the eye. This light is projected onto the retina, a thin neural structure lining the inside of the eye. Photoreceptors, among the most important cell types in the retina, are the first to receive the incoming rays of light. In mammals, there are two types of photoreceptors: rods and cones. Rods are specialized for nighttime vision, and cones for daytime and color vision. In this study, the authors examined the photoreceptors of a mouse with a gene mutation that causes photoreceptors to develop abnormally. Humans with a similar mutation have a form of blindness called enhanced S-cone syndrome (ESCS). Surprisingly, the majority of photoreceptors in this mutant mouse were found to have features of both normal rods and cones. It is possible that the abnormal features of these photoreceptors predispose them to undergo premature death. If this model accurately reflects the situation in human patients with ESCS, it may provide an explanation for the loss of vision seen in this disease. This study also elucidated previously unknown molecular differences between normal rods and cones. This new knowledge may contribute to a better overall understanding of the mechanisms underlying night, day, and color vision.
Citation:Corbo JC, Cepko CL (2005) A hybrid photoreceptor expressing both rod and cone genes in a mouse model of enhanced s-cone syndrome. PLoS Genet 1(2): e11.
==== Body
Introduction
Enhanced S-cone syndrome (ESCS) is an unusual disease of photoreceptors that includes night blindness (suggestive of rod dysfunction), an abnormal electroretinogram (ERG) with a waveform that is nearly identical under both light and dark adaptation, and an increased sensitivity of the ERG to short-wavelength light [1,2]. The disease is caused by mutations in the orphan nuclear receptor transcription factor NR2E3 (also known as photoreceptor nuclear receptor), which is expressed exclusively in rods [3,4]. Recent human genetic studies have also demonstrated mutations in this gene in Goldmann-Favre syndrome and many cases of clumped pigmentary retinal degeneration [5].
The initial reports of patients with ESCS attributed the unusual ERG to an abnormally functioning rod photoreceptor system with persistent activity under light adaptation [6–8]. Subsequent studies, however, concluded that the ERG was due to supernumerary short-wavelength (“blue”) cone photoreceptors (S-cones) in these patients [1,2,9–11]. Histopathologic analysis of a retina from a human patient with ESCS and extensive retinal degeneration demonstrated an absence of rhodopsin-positive cells and an increase in the number of S-cone opsin-expressing cells. Nevertheless, the overall density of cones was only modestly increased in this patient (approximately 2-fold), suggesting that there might be additional factors that contribute to the very large, light-adapted ERG seen in this disease. In addition to the ERG findings, patients with ESCS have dysmorphic retinas with rosette formation in the outer nuclear layer (ONL) where photoreceptor cell bodies reside, and a slow retinal degeneration that can ultimately lead to complete blindness [12–14].
Mutations in the mouse ortholog of NR2E3 have been identified in the spontaneous mutant retinal degeneration 7
(rd7) [15]. This mutant demonstrates slow retinal degeneration and abnormal lamination of the ONL with rosette formation [15,16]. Curiously, the ERG of the mouse under both light and dark adaptation has been reported to be normal, showing progressive attenuation with time, presumably due to degenerative cell loss [15]. A prior study showed a 2- to 3-fold increase in the number S-opsin–positive cells in the rd7 retina compared to wild type [17]. In addition, two groups recently reported derepression of additional cone genes in the rd7 mutant [18,19].
In order to better understand the mechanistic basis of ESCS, we undertook a molecular and ultrastructural analysis of the photoreceptors of the rd7 mutant mouse. Microarray and in situ hybridization analyses revealed a modest increase in the number of S-opsin–positive cells and widespread derepression of many cone-specific genes within rod photoreceptor cells. Ultrastructural studies demonstrated that the cells that coexpress rod and cone genes in the rd7 retina represent a morphologically hybrid cell type, intermediate between normal rods and cones.
Results
Widespread Up-Regulation of Cone Genes in the rd7 Mutant Retina
In an initial analysis of the rd7 mutant, homozygous mutant retinas were compared with wild-type controls at multiple postnatal time points using both cDNA and Affymetrix microarrays. The cDNA microarray used in this study contains approximately 12,000 different cDNAs largely derived from the retina and nervous system, and the Affymetrix microarray contains over 34,000 genes. Experiments at all timepoints were carried out in triplicate, and stringent criteria were applied in deciding whether a given gene was up- or down-regulated in the mutant (see Materials and Methods for details).
These experiments demonstrated widespread up-regulation of cone-specific and cone-enriched genes in the rd7 retina, especially by postnatal day 14 (P14) and P21 (Figure 1). Most known cone-specific or cone-enriched genes were found to be up-regulated in the mutant (Figure 1, genes G1–G15). The majority of these genes represent components of the phototransduction cascade (e.g., opsins, transducins, and phosphodiesterase subunits). In addition to these genes, several novel cone-specific genes of unknown function recently identified in our lab were also up-regulated (Figure 1, genes G16, G17, G21, and G24; unpublished data). Finally, a wide range of other genes, most with no previously recognized role in the retina, were found to be up-regulated in the rd7 mutant (Figure 1, G26–G53; Tables S1 and S2; Figures S1–S7).
Figure 1 Cone-Specific and Cone-Enriched Genes Evaluated in the rd7 Mutant by Microarray and In Situ Hybridization
The color coding of text in the column “Gene Name” is as follows: light blue (G1–G15), genes previously reported in the literature to have cone-specific or cone-enriched patterns of expression; yellow (G16–G25), novel cone genes identified in an unrelated study (unpublished data); dark green (G26–G36), novel cone genes identified in the present study that were up-regulated in rd7; light green (G37–G46), additional genes found to be up-regulated in rd7 by microarray in the present study but that had either weak or inapparent cone-specific signal on in situ hybridization; white (G47–G53), additional genes up-regulated by microarray at two different timepoints but with either unusual expression patterns or nonconfirmatory in situ hybridizations. The column “ID” contains identifiers used in the present paper to refer to specific genes. “GenBank ID” contains the GenBank accession number of the clone used to make the probe for in situ hybridization. Within this column, “lab clone” indicates that the probe used for in situ hybridization derived from a clone in our laboratory. The region of the gene to which it corresponds is indicated in Table S1. Columns “P0” through “P21” contain the results of microarray experiments at the given postnatal dates. P0, P6, and P14 time points represent analyses on cDNA microarrays; the P21 time point represents data from an Affymetrix microarray (mouse genome 432 2.0). A red cell with a single up arrow indicates that the gene in question was up-regulated in three out of three microarrays at that time point (as described in Materials and Methods). Those cells labeled orange with a single up arrow and asterisk indicate that the gene in question was up-regulated in two out of three microarrays at that time point. The column “In Situ” lists the type of derepression seen for the gene in question in the rd7 mutant retina (type I and type II are described in the main text). Genes designated “unclassified” represent patterns of derepression that were difficult to classify as either type I or type II (see main text for more details). “Wild type” in this column indicates that the in situ hybridization pattern in the rd7 mutant retina was not different from the wild-type pattern; and “special” indicates a special pattern of expression discussed more fully in the main text. The column “Expression Pattern” contains a concise description of the wild-type expression pattern of the gene in question. In the case of genes for which no signal was obtained on in situ hybridization in the present study, the specified expression pattern derives from reports in the literature. Within this column, “cone > rod” indicates that the gene is expressed in all photoreceptors, but at higher levels in cones than rods; “cone?” indicates very weak staining in a cone-like distribution.
BP, bipolar cells; EP, early photoreceptor expression pattern; IS, inner core segment localization; MG, Müller glia; N/A, not available on the microarray; NS, no signal detected on in situ hybridization; RPE, retinal pigment epithelium.
Nr2e3 expression is first detectable by in situ hybridization around embryonic day 18 (E18); it then peaks around P6 and subsequently decreases to adult levels by P21 (unpublished data). In accordance with this time course of expression, almost no gene expression changes were found at P0, with progressively more changes at later timepoints (Figure 1). One exception to this statement is the gene RIKEN cDNA 4933409K07 (Figure 1, gene G47), which was the only gene shown to be up-regulated at all timepoints examined. Additional discussion of this gene and its unusual expression pattern will be presented below.
Two Distinct Patterns of Cone Gene Derepression in rd7
In order to confirm these microarray results, an in situ hybridization analysis of the putative up-regulated cone genes was carried out in which the rd7 mutant retina was compared with age-matched, wild-type controls. We found that the majority of the cone-specific genes that were up-regulated in microarray experiments were derepressed when assessed by in situ hybridization (Figure 2). There were two major patterns of cone gene derepression. The more common pattern (type I) manifested itself as ectopic gene expression throughout the ONL, consistent with gene expression in all photoreceptors (Figure 2; upper left photomicrographs). Typical examples of this pattern of derepression are shown in Figure 2, and many more are available in Table S1. This pattern of expression contrasts sharply with the usual pattern of cone gene expression, which consists of scattered cells localized to the scleral edge of the ONL (Figure 2).
Figure 2 Cone and Rod Gene Expression in the rd7 Mutant at P14
The upper sets of photomicrographs demonstrate examples of type I and type II cone gene derepression in the rd7 mutant retina as explained in the main text. The bottom left images show several rod-specific genes that are essentially unchanged in the rd7 background at P14. The bottom right images show the expression pattern of three photoreceptor transcription factors in the rd7 mutant. Abbreviations in the lower left hand corner of each pair of panels represent the gene symbols summarized in Figure 1.
The second category of cone gene derepression (type II) consisted of a patchy, salt-and-pepper pattern of ectopic expression in which individual positive cells were scattered throughout the ONL (Figure 2, upper right photomicrographs; Table S1). Although numerous positive cells were present in the rd7 retina (particularly in the ventral portion), there were clearly many interspersed cells that showed a complete absence of expression. In order to rule out the possibility that these scattered positive cells were simply the normal complement of cones that had failed to localize their cell bodies to the scleral edge of the ONL, the number of positive cells in the rd7 retina was quantitated by dissociated cell in situ hybridization.
Dissociated cell in situ hybridization was performed using a probe for the S-cone opsin gene (Opn1sw), which shows type II derepression (Figures 2 and 3A–3C). S-opsin was expressed in 3.2% of retinal cells in the rd7 mutant (66 S-opsin–positive cells out of 2,056 6-diamidino-2-phenylindole [DAPI]-positive cells). This value is approximately 2-fold greater than the percentage of S-opsin–positive cells identified in wild-type control retinas, 1.65% (54 S-opsin-positive cells out of 3,271 DAPI-positive cells), and accords well with the previously reported value of 2- to 3-fold more S-opsin–positive cells in rd7 compared to wild type arrived at by antibody staining of tissue sections [17].
Figure 3 S-Opsin Dissociated Cell In Situ Hybridization and S-Opsin/Rhodopsin Antibody Staining on rd7 Mutant Retina
(A–C) A dissociated cell in situ hybridization with an S-opsin probe (red) on dissociated rd7 mutant retinal cells stained with DAPI (blue). (C) shows the merged images.
(D–F) The outer nuclear layer of an rd7 mutant retina stained by antibody for S-opsin (red) and rhodopsin (green). The scleral edge of the outer nuclear layer is up. DAPI staining is in blue. (F) shows the merged images. Insets are higher-power images of the outer segments showing non-overlap of S-opsin and rhodopsin staining in the mutant.
Previous studies have estimated that the total number of cones in the mouse retina is 2% of all retinal cells [20], and that S-opsin is largely repressed in cones in the dorsal third of the retina [21]. The estimate of 1.65% S-opsin–positive cells in the wild-type retina is in agreement with these data. The fact that only 3.2% of all retinal cells are S-opsin–positive in the rd7 mutant also confirms that the majority of the photoreceptors (which make up just over 70% of the cells in the adult mouse retina) do not express this gene. In order to assess whether these supernumerary S-opsin–expressing cells coexpressed rod-specific markers, a double antibody staining for S-opsin and rhodopsin was performed. This study showed mutually exclusive domains of expression of S-opsin and rhodopsin in the photoreceptor outer segments (Figure 3D–3F). This finding suggests that the supernumerary S-opsin–expressing cells in the rd7 retina may represent normal “blue” cones.
Novel Cone-Specific Genes Are Derepressed in rd7
Given that the majority of known cone-specific genes showed marked derepression in the rd7 mutant, additional candidate genes up-regulated on microarray analysis were evaluated for cone-specific expression. In situ hybridization was performed on an additional 45 up-regulated genes, confirming that 21 of them were derepressed. Of these, at least ten showed a definite cone-specific or cone-enriched pattern of expression in the wild-type retina (Figure 1, genes G26–G35). Several examples are given in Figure 4. Note that in the wild-type retina, there is a relatively weak pattern of scattered positive cells at the scleral edge of the ONL, consistent with a cone-specific pattern of expression. All of these genes show marked derepression in the rd7 retina. A number of these novel cone-specific genes showed a striking localization of their transcripts to the photoreceptor inner segment (e.g., Bub1b and Tcta). This localization manifests in a section in situ hybridization as a dark band of staining just beyond the outer edge of the ONL immediately underlying the outer segment layer. Although such a pattern of transcript localization is commonly seen in many rod-specific genes (e.g., Rho in Figure 2; Pcdh21,
Rbp3, and Cnga1 in Table S2), it is not easily appreciated in cone-specific genes, possibly due to the relative scarcity of cones in the mouse. In the rd7 mutant retina in which such genes are widely derepressed, such a pattern of transcript localization often becomes apparent.
Figure 4 Expression Patterns of Several Novel Cone Genes Up-Regulated in rd7
In the wild-type images (wt), note the scattered, weakly positive cells at the scleral edge of the outer nuclear layer in a cone distribution. All of the genes show marked up-regulation in the rd7 mutant. Bub1b and Tcta show transcript localization predominantly to the inner segment of the photoreceptors. Retinas are oriented such that the scleral edge is up.
In addition to the ten genes that showed cone-specific expression in the wild-type retina, another 11 novel genes were derepressed in the rd7 retina by in situ hybridization (Figure 1, genes G36–G46). Some of these genes showed faint expression in a cone-like distribution (see Table S1, genes G36, G40, and G44), and one appeared to be expressed throughout the ONL but at greater levels in cones than in rods (Table S1, gene G37). The remainder of the up-regulated genes did not have detectable cone staining in the wild-type retina. Despite this apparent absence of cone staining, the pattern of derepression in rd7 suggests that these genes may also be novel cone-specific genes, albeit expressed at levels below the sensitivity threshold of our in situ hybridization assay.
In most cases, the novel cone genes identified in this study appear to have a type I pattern of derepression. However, due to the weakness of the signal in some cases, or transcript localization to the inner segment in others, it was not always possible to determine with confidence which of the two patterns of derepression (if either) each of these genes displayed. In terms of functional categorization, the novel cone genes cover a broad range including glucose metabolism (Pygm and Glo1), fatty acid metabolism (Elovl2), DNA repair (Smug1), cell cycle/chromosome segregation (Bub1b), carcinogenesis (Tcta), endothelial biology (Ece1), cytoskeletal function (Ebp4.1l1), and even otolith formation (Otop3).
A relatively frequent finding among both previously identified cone-specific genes, as well as in some of those identified in the present study, is the occurrence of gene expression in an early photoreceptor precursor pattern (Figure 5). This pattern of expression consists of positive staining by in situ hybridization specifically at the scleral border of the retina during prenatal timepoints (in the range of E13–E18). Gnb3 and Thrb2 are two examples of known cone genes with this early pattern of expression (Figure 5). Two of the 11 novel cone genes identified in this study also have this early photoreceptor pattern of expression (Ece1 and Otop3). Intriguingly, three genes shown to be up-regulated in rd7 on microarray, but that had either no detectable signal by in situ hybridization at adult stages or no apparent change in expression by in situ hybridization between wild type and rd7, also showed this early photoreceptor pattern (Figure 1, genes G48–G50). The embryonic expression pattern of two of these genes is shown in Figure 5 (the embryonic in situ hybridization for the third, G50, can be found in Table S1). Although the significance of such early photoreceptor expression is not known, it is possible that these genes may also be cone-specific but are expressed at undetectably low levels in the adult.
Figure 5 Some of the Genes Up-Regulated in rd7 Show an Early Photoreceptor Pattern of Expression
Gnb3 and Thrb2 are both previously characterized cone genes that show staining at the scleral edge of the embryonic mouse retina in cells that will differentiate into photoreceptors. Ece1 and Otop3 are two novel cone genes identified in this study that were up-regulated in rd7 and also showed an early photoreceptor pattern of expression. Prdm1 and RIKEN cDNA 1300018I05 (Figure 1, genes G48 and G49, respectively) are two other genes that had either undetectable signal (Prdm1) or no apparent change in expression pattern in adult rd7 mutants (RIKEN cDNA 1300018I05), but which also showed staining in the embryonic retina in a presumptive photoreceptor pattern. All images are from E17.5 retina except Gnb3, which was from E16.
M-Opsin and Thyroid Hormone Receptor β2 Are Unchanged in the rd7 Mutant
Only two cone-specific genes failed to show any change in expression by either microarray or in situ hybridization in the rd7 mutant: M-opsin (Opn1mw) and thyroid hormone receptor β2 (Thrb2) (Figure S8). This result is particularly notable because Thrb2 is absolutely required for the expression of M-opsin [22]. Furthermore, the repression of S-opsin expression in the dorsal third of the mouse retina is thought to depend, at least in part, on Thrb2 since S-opsin shows dorsal derepression in the Thrb2 mutant [22]. Despite the derepression of S-opsin seen in the ventral portion of the rd7 retina, the normal dorsal repression of this gene is still present in this mutant (unpublished data). This finding is consistent with the normal expression pattern and function of Thrb2 in the rd7 mutant.
One further finding to note is that the cell bodies of the M-opsin–positive cells appear to be scattered throughout the ONL in the rd7 mutant at P14 (Figure S8). Despite this fact, their overall number does not appear to be increased relative to wild-type. In addition, by P28, the M-opsin-positive cell bodies in rd7 appear to have relocated to their normal position at the scleral edge of the ONL (Figure S8). It is known that until P11, the cell bodies of cone photoreceptors in the mouse are normally dispersed throughout the ONL, only to relocate subsequently to the scleral edge of the ONL around P12 [23]. It is possible that in the rd7 mutant retina, there is a short delay in the relocation of the M-opsin–expressing cone cell bodies to the scleral edge of the ONL.
Rod Genes Are Only Modestly and Temporarily Affected in rd7
In sharp contrast to changes in cone gene expression, rod-specific genes were much less severely affected in the rd7 mutant. Microarray and in situ hybridization analysis of numerous rod genes failed to reveal marked changes in expression levels at P14 and P21 (see Figure 2, lower left photomicrographs; Table S2). In addition to the three rod genes depicted in Figure 2, in situ hybridization analysis on an additional 19 rod-specific and pan-photoreceptor genes demonstrated only a very mild diminution of expression in two of these genes, gucy2e and Rgs9, at P14, and an increase in expression in two, Nr2e3, and Cnga1 (Table S2).
Despite the minimal changes in rod gene expression at later postnatal timepoints, there was evidence of a significant delay in the onset of rhodopsin (Rho) expression in rd7 mutants relative to wild-type. Microarray analysis at P6 demonstrated five cDNA spots that were down-regulated in three out of three experiments. Of these spots, three corresponded to rhodopsin (Table S3). In situ hybridization analysis of several rod-specific genes at P6 revealed that rhodopsin alone showed a markedly lower level of expression compared to wild type (Figure 6; unpublished data). Despite this modest delay in the onset of rhodopsin expression, by P14 the gene had attained nearly normal levels in the rd7 mutant (see Figure 2, lower left photomicrographs). This latter finding suggests that all the rod- and many cone-specific genes are coexpressed in the majority of photoreceptors in the rd7 mutant.
Figure 6 The Onset of rhodopsin Expression Is Delayed in the rd7 Mutant Retina
Note the nearly undetectable staining for rhodopsin in this P6 mutant retina (top right). The majority of rod-specific genes did not show this delay in expression onset, as indicated by the normal amount of staining for Pde6a in the mutant at P6 (bottom images).
Changes in Retinal Transcription Factor and Müller Glial Gene Expression in rd7
Analysis of several photoreceptor transcription factors in the rd7 mutant indicated that the levels of Crx and Nrl are unaffected in the mutant at P14 (see Figure 2, lower right photomicrographs). Nrl is a rod-specific, basic leucine zipper transcription factor required for the activation of many rod-specific genes and the repression of most cone-specific genes in rods [24]. Nrl is known to be genetically upstream of Nr2e3 and is required for its expression [24]. Crx is a homeobox transcription factor expressed in both rods and cones and is required for the expression of a variety of rod- and cone-specific genes [25]. In contrast to what is seen in the Nrl mutant, Nr2e3 expression is unchanged in Crx mutant homozygotes (unpublished data).
Strikingly, Nr2e3 itself was markedly up-regulated in the rd7 mutant both by microarray and in situ hybridization (see Figure 2, lower right photomicrographs; Table S2). The rd7 mutant carries a deletion in Nr2e3 that removes portions of both the DNA-binding and ligand-binding domains [15]. Although this deletion most likely creates a null allele, a residual transcript is clearly present and up-regulated in the rd7 mutant. This finding suggests that Nr2e3 is required for repression of its own transcription.
One gene, RIKEN cDNA 4933409K07 (Figure 1, gene G47), was found to be up-regulated on microarray at all four timepoints examined. This gene showed a unique pattern of expression in the adult rd7 mutant retina. Whereas there was only a barely detectable hint of expression in the inner nuclear layer (INL) in the wild-type retina, this gene showed strong expression in the middle and vitreal thirds of the INL as well as patchy expression in the ganglion cell layer (GCL) and at the scleral edge of the ONL in rd7 (see Table S1). This in situ hybridization pattern is consistent with staining in Müller glia, the principal glial cell type of the mouse retina. One possible interpretation of this unusual expression pattern is that it represents an early reaction of Müller glia to injury in this mutant.
The Majority of the Photoreceptors in the rd7 Retina Represent a Morphologically Hybrid Cell Type
In order to assess the morphologic effects of the above gene expression changes, the ultrastructure of the photoreceptor cell bodies in the rd7 mutant was examined. The cell bodies were chosen for evaluation rather than the outer segments, since in the mouse, the ultrastructural differences between rod and cone somata are much greater than are the differences between the outer segments [26]. In the wild-type mouse, cone cell bodies are aligned along the scleral border of the ONL, and they are larger than those of rods. They have a smaller, more irregular mass of nuclear heterochromatin that is often broken up into multiple discrete clumps connected by thin threads. They also have more abundant electron-lucent euchromatin than rods. Lastly, they frequently have a patch of organelle-rich cytoplasm next to their nuclei, usually containing large mitochondria [26].
Analysis of toluidine blue-stained semi-thin sections revealed that such cone-like cells were present in greater abundance in the rd7 retina than in wild-type, and that their somata were scattered throughout the ONL (Figure 7). A comparison between the distribution of these cells and those expressing S-cone opsin strongly suggests that they represent the same cell population (compare Figure 7D and 7F). Analysis of the nuclear morphology of dissociated retinal cells stained for S-cone opsin by dissociated cell in situ hybridization confirmed that this is the case (unpublished data). These findings, along with the absence of rhodopsin staining in these cells (see Figure 3D–3F), suggest that these “cone-like” cells in the rd7 mutant retina may represent supernumerary normal cones with an abnormal localization of their cell bodies.
Figure 7 The rd7 Mutant Retina Contains a Morphologically Hybrid Photoreceptor Cell Type in Addition to Supernumerary S-Opsin–Positive Cones
(A and B) Toluidine blue-stained semi-thin sections of the outer nuclear layer (scleral edge oriented up).
(C and D) Hand-drawn diagrams of the cells in (A) and (B), respectively. Cells with the nuclear features of cones are highlighted in blue. Note that the number of such cells is greater in the mutant, and their cell bodies are scattered throughout the outer nuclear layer. In addition, the overall columnar architecture of the outer nuclear layer seen in the wild type is disrupted in this portion of the mutant retina. Other portions of the mutant retina with fewer supernumerary cone cells, however, retain the normal columnar appearance (unpublished data).
(E and F) Images of the outer nuclear layer (scleral edge up) stained by in situ hybridization for S-opsin. Note the typical pattern of staining at the scleral edge of the outer nuclear layer in the wild type. The rd7 mutant retina shows supernumerary S-opsin–positive cells scattered throughout the outer nuclear layer in a distribution very similar to the supernumerary cone cells seen in (B). Since images (E) and (F) derive from different retinas than those depicted in (A) and (B), the location of the individual cells do not correspond.
(G and H) Electron micrographs of the outer nuclear layer (10,000× magnification). Note the uniform distribution of rod cell bodies in the wild type (G). The cell bodies are nearly round and consist almost exclusively of a nucleus with a single, dense mass of heterochromatin. In the rd7 mutant (H), two types of cell are shown. The ovoid one with a lesser quantity of heterochromatin, paler euchromatin, and two juxtanuclear mitochondria (yellow arrow) represents a typical cone cell body. The adjacent cell with a more “rod-like” mass of heterochromatin and a single juxtanuclear mitochondrion (red arrow) represents one of the hybrid photoreceptors discussed more fully in the main text.
In contrast to the cone cell body, the wild-type rod soma is small and nearly round. It has a single, large clump of dense heterochromatin, a thin rim of moderately electron-dense euchromatin, and very scant juxtanuclear cytoplasm without organelles [26,27]. The second cell population in the ONL of the rd7 retina has some of the nuclear features of normal rods, such as a single, dense mass of heterochromatin and moderately electron-dense euchromatin (Figure 7H); yet these cells also show features of cones. First, the euchromatin is, on average, more abundant in these cells than in wild-type rods (compare Figure 7G and 7H). In addition, comparison of the diagrammatic representation of the wild-type and rd7 ONLs suggests that the average area of the S-opsin–negative cells in rd7 is greater than in the wild-type (Figure 7C and 7D). In order to confirm this impression, we quantitated the area of 50 wild-type and 50 mutant rod-like cell bodies (see Materials and Methods for details). This experiment confirmed that the average area of the rod-like somata in rd7 is approximately 30% larger than that of wild-type rod somata (mean area in rd7 was 9.75 ± 1.36 (standard deviation) μm2, compared to wild-type rods, with 7.53 ± 0.72 μm2 ; n = 50; p = 7.6 × 10–16, Student's t-test). It is also notable that the standard deviation of the somal area is nearly twice as great in rd7 than in wild-type, confirming the subjective impression of greater variability in somal size and shape in the mutant compared to the wild-type (compare Figure 7C and 7D).
Lastly, 38% (19/50) of the rd7 photoreceptors selected for somal area quantitation had prominent juxtanuclear mitochondria (red arrow in Figure 7H; unpublished data). Such juxtanuclear organelles are only very rarely seen in normal rods (1.5%; six out of 399 cells counted), but are common in cones (yellow arrow in Figure 7H). In conclusion, it is clear that this second cell population in the rd7 retina has morphological features of both normal rods and cones consistent with the coexpression of many rod- and cone-specific genes in these cells.
Discussion
In this paper we have determined that the primary role of the rod-specific transcription factor, Nr2e3, is to maintain cone genes transcriptionally silent within rods. We have identified two patterns of cone gene derepression in the rd7 mutant retina, in agreement with a previous report by Chen et al. [18]. The first pattern of derepression identified (type I) consists of ectopic expression of cone genes in the vast majority of cells in the ONL. These cells were also shown to coexpress all rod genes tested. Consistent with the hybrid pattern of gene expression in these cells, electron microscopic analysis revealed them to be morphologically intermediate between normal rods and cones.
Although genes showing type I derepression demonstrated staining in the majority of cells in the ONL, two lines of evidence suggest that these genes are not completely derepressed in these cells when compared to their expression in S-opsin–expressing cones. First, close evaluation of the staining pattern of a number of type I genes in the rd7 mutant retina (e.g., see Table S1, genes G9, G19, and G24), reveals that, in addition to the background staining throughout the ONL, there is a more darkly staining subpopulation of cells scattered throughout this layer in a distribution corresponding to that of the supernumerary S-cone opsin-expressing cells. This pattern of staining suggests that these genes are more highly expressed in S-opsin expressing cells than in the hybrid cells of the rd7 retina.
The second line of evidence derives from a comparison of the expression pattern of many type I genes in rd7 and Nrl
−/− mutant backgrounds. As mentioned above, Nrl is a retinal transcription factor that, when mutated, results in en masse conversion of rods into S-opsin–expressing cones [24]. It can be inferred from this fact that Nrl is absolutely required for the normal silencing of cone-specific genes in rods. In the Nrl homozygous mutant, there is a stronger and more uniform derepression of many cone-specific genes throughout the ONL than is seen in the rd7 retina (unpublished data). This finding further suggests that, in addition to its repression of cone gene expression via induction of Nr2e3 expression, Nrl exerts an additional level of negative control over cone genes either directly or via a second, as yet uncharacterized, repressor.
The second pattern of derepression seen in the rd7 retina (type II), is represented by a scattered population of cells throughout the ONL that shows derepression of several cone-specific genes, including S-cone opsin. By ultrastructural criteria, these cells appear to be normal cones, albeit with displaced cell bodies. Quantitation of these supernumerary S-cone opsin-positive cells indicates that they are approximately 2-fold more abundant than in normal retina, consistent with previous antibody studies [17].
Two recent studies have presented data that are consistent with many of the findings in our study [18,19]. Both studies showed that cone genes in addition to S-cone opsin are derepressed in the mouse rd7 mutant. In addition, Peng et al. [19] found by RT-PCR that the levels of several rod-specific genes, including rhodopsin, were modestly reduced in rd7 at P28. Our in situ hybridization data suggest that rhodopsin expression is markedly reduced at P6, but that it attains levels indistinguishable from wild-type by P14. Since the change in rhodopsin levels identified by Peng et al. were relatively small (an approximately 15% reduction), it is not surprising that such a difference was not detected by in situ hybridization. The overall finding of modest reductions in rod-specific gene expression is entirely in keeping with the results of the present study.
In addition to demonstrating derepression of a range of known cone-specific genes in rd7 mutants, Chen et al. [18] showed up-regulation by Northern blot of two additional genes in the rd7 mutant, Elovl2 and Fabp7. These two genes were also found to be up-regulated in rd7 in the present study (see Figure 1; Table S1). Although we found Elovl2 to have a cone-enriched pattern of expression (see Figure 1), in situ hybridization analysis of Fabp7 failed to show a signal in wild-type or mutant retina (unpublished data). Nevertheless, previous studies have suggested that Fabp7 is expressed in radial glia and immature astrocytes in the brain [28–30]. Given the expression pattern elsewhere in the nervous system, it is possible that Fabp7 is up-regulated in Müller glia in the rd7 retina in response to injury in a manner akin to the novel Müller glial gene identified in this study, RIKEN cDNA 4933409K07 (Figure 1, gene G47). Indirect support for this idea is provided by the observation that Fabp7 is up-regulated by microarray analysis in Nrl and Crx mutant retinas as well (unpublished data), suggesting that this change may represent a generalized response to injury in the retina rather than derepression of a cone-enriched gene.
The study by Chen et al. [18] made two further observations worthy of note. First, they identified a zebrafish homolog of Nr2e3 and showed it to be expressed in photoreceptors. Interestingly, they showed that this homolog appears to have a pan-photoreceptor pattern of expression early in development that later resolves into a rod-specific pattern of expression. This early expression in cones may represent a mechanism whereby the expression of cone-specific gene products is temporarily repressed. It will be important to determine the extent to which the function of Nr2e3 has been conserved in the retina of such a distantly related organism. Secondly, Chen et al. [18] used an in vitro oligonucleotide selection protocol to determine the preferred binding site for Nr2e3. This information will be very useful for future bioinformatic analyses of Nr2e3 target genes.
The gene expression changes identified in the rd7 mutant retina in the present study suggest the scheme of gene regulation in mouse rods depicted in Figure 8. As this diagram implies, there appear to be at least two different repressors of cone genes within rods, Nr2e3 and either Nrl itself or an additional unknown transcription factor downstream of Nrl, here termed “transcription factor X.” In fact, it appears that the differences between type I and type II cone genes may simply depend on which repressor—Nr2e3 or transcription factor X—is primarily responsible for the regulation of the gene in question. In the case of type I genes, Nr2e3 is the primary repressor and transcription factor X is of secondary importance. In the case of type II genes, transcription factor X exerts the major repressive effect on transcription, and Nr2e3 plays a lesser, but still important role.
Figure 8 A Partial View of the Rod Photoreceptor Transcriptional Regulatory Network
Note that green lines indicate activation, and yellow and red lines indicate weak and strong repression, respectively. The dotted lines associated with a question mark indicate that it is not known whether Nrl directly represses the target genes in question or whether its repression is mediated by a downstream transcription factor (“X”). Note that Nr2e3 appears to negatively regulate its own transcription. The regulatory linkages depicted in this diagram are not necessarily direct. The weak activation of some rod-specific genes by Nr2e3 is omitted from this diagram for clarity. Also not shown is the role of other photoreceptor transcription factors, such as Crx.
In contrast to the marked derepression of the vast majority of cone-specific genes in the rd7 mutant, two genes stand out as being unaffected by the mutation: the gene encoding M-opsin and Thrb2. As Thrb2 is known to be required for the expression of M-opsin [22], the absence of supernumerary M-opsin–positive cells may simply be a consequence of the fact that Thrb2 expression is unchanged in the rd7 mutant. Further support for this idea has been provided by recent work in our lab showing widespread derepression of cone genes in the Notch1
−/− retina (unpublished data). In contrast to the rd7 mutant, Notch1
−/− retinas show marked derepression of Thrb2 and a corresponding derepression of the gene that encodes M-opsin. An additional observation suggesting that M-opsin and S-opsin are controlled by different mechanisms comes from in vitro experiments [31,32]. While explanted P3 retinas express S-opsin and M-opsin with normal kinetics, explanted P0 retinas express only S-opsin [32]. The factor(s) controlling these differences are unknown, but may be intrinsic, as cocultures of older and younger retinas, conditioned media from older retinas, and addition of a variety of small molecules were unable to promote the expression of M-opsin in the P0-initiated cultures [32].
In contrast to our findings, Peng et al. [19] reported that M-opsin expression is mildly increased in the rd7 mutant retina. It is possible that a subtle increase in M-opsin transcript levels does occur in the rd7 retina, and that this difference could not be detected by in situ hybridization. Since virtually all M-opsin–expressing cells are localized at the outer edge of the ONL by P28 in the rd7 mutant (Figure S8), any increase in M-opsin transcript in the mutant must have occurred in cells in that location.
A variety of novel cone-specific or cone-enriched genes were characterized in this study. One of these genes, Pygm, is involved in glycogen/glucose metabolism, and a second, Glo1, is required for detoxification of methylglyoxal, a byproduct of glycolysis [33]. A third gene involved in glucose metabolism, hexokinase 2
(Hk2), is also derepressed in the rd7 mutant and shows a pattern of expression in the wild-type retina, suggesting greater expression in cones than in rods (see Figure 1; Table S1). A fourth gene involved in glucose metabolism, glucokinase regulatory protein (Gckr), was found to be increased in three out of three microarrays at P21 but was not tested by in situ hybridization (Table S4). The increased expression of Gckr in rd7 mutant retina suggests that it too may be a cone-enriched gene. A previous study found that two of these genes, Pygm and Hk2, have markedly elevated tag levels in an ONL-specific serial analysis of gene expression library consistent with their being highly enriched in wild-type photoreceptors [34]. Furthermore, prior studies have suggested differences in glycogen and glucose metabolism between primate rods and cones [35]. Our findings lend further support to this concept. Interestingly, Pygm has been implicated in human disease. Mutations in this gene underlie McArdle's disease (glycogen storage disease type V), the symptoms of which include exercise intolerance, muscle cramps, and myoglobinuria [36]. To our knowledge, no abnormalities of retinal function have been reported.
One of the most curious findings in the rd7 mutant retina was the occurrence of two different types of changes: an increase in the number of S-opsin–expressing cones and a transformation of rods into hybrid photoreceptors. It is known that Nr2e3 is expressed only in rods, and the transcript is first detectable in postmitotic cells (J. Trimarchi and CLC, unpublished data). Assuming that Nr2e3 acts cell autonomously, we can conclude that the supernumerary S-cone–positive cells and the hybrid photoreceptors identified in the rd7 retina were redirected to these fates from postmitotic cells that were destined to become rods. This conclusion raises this question: Why does loss of a single transcription factor within rod precursors lead to two alternative fates—a hybrid cell type on the one hand and apparently normal S-cones on the other? There are at least two possible explanations for these differences.
First, it is possible that there are two distinct types of rod precursor; loss of Nr2e3 in one leads to S-cone fate and in the other results in a hybrid cell type. In fact, there is experimental evidence from the rat to support the conclusion that early-born and late-born rods are intrinsically different [37]. One test of the hypothesis that there are two temporally distinct rod precursor populations would be to carry out birthdating experiments to determine whether the supernumerary S-opsin–positive cells in the rd7 retina derived exclusively from an early- or late-born population. Of course, if this were not the case, this experiment could not rule out the possibility that molecularly distinct populations of rod precursors are present simultaneously in the developing retina.
An alternative explanation would be that there is only a single, homogeneous population of postmitotic rod precursors in the mouse, and a stochastic event triggers assumption of the S-cone fate in a small subpopulation of these cells in the rd7 mutant. Recent studies in a variety of experiment systems suggest that such a stochastic, all-or-none mechanism of gene activation is commonplace [38–44]. In this scenario, the absence of Nr2e3 would alter the probability that an unknown master control gene is expressed in rod precursors. Once this event takes place, it would initiate an irreversible program of differentiation toward S-cone fate, albeit at a relatively low frequency. In this way, a subset of cells from an initially homogeneous population would select the S-cone fate in an entirely probabilistic manner.
Human patients with ESCS display three types of abnormality attributable to the retina: (1) an atypical ERG waveform that is preferentially sensitive to short-wavelength light, (2) slowly progressive retinal degeneration, and (3) abnormal retinal lamination with rosette formation [1,12,13]. The rd7 mutant mice also demonstrate the latter two defects, but have a normal ERG [15,45]. These similarities and differences between the two species help to explain the possible mechanistic basis of the ESCS.
The fact that the rd7 mouse has a normal ERG strongly suggests that the aberrant ERG in ESCS is not attributable to the activity of a hybrid photoreceptor identical to that found in this study. Namely, the signal is unlikely to derive from a population of cells coexpressing both rod and cone genes but whose photopigment is rhodopsin and not S-cone opsin. This conclusion is consistent with the evidence from human ESCS patients indicating a markedly reduced rod system and a lack of measurable rhodopsin by reflection densitometry [1,2,10,11]. It is also unlikely that we would fail to detect an ESCS-like ERG signal in mice if it were present, as such a signal has been demonstrated in the Nrl mutant mouse, which has a near total transformation of all its rods into blue cones [24].
These findings, however, do not rule out the possibility that the abnormal human ERG derives from a hybrid photoreceptor cell type that also expresses S-opsin. It is possible that there are gene regulatory differences between mice and humans such that in human NR2E3 mutants, S-opsin shows a type I pattern of derepression rather than a type II as in seen in the rd7 mouse, and is therefore expressed in all of the hybrid photoreceptor cells. Alternatively, the ratio of supernumerary S-cones to hybrid photoreceptors produced in the retina of ESCS patients might be such that a higher percentage of the presumptive rods in ESCS patients become S-cones rather than hybrid photoreceptors. As discussed above, this ratio could depend either on the relative percentages of two distinct rod precursor populations or on stochastic effects on regulatory gene expression.
In contrast to the ERG differences between mouse rd7 and human NR2E3 mutants, both species demonstrate slow retinal degeneration. It is possible that this degeneration is attributable to the abnormal function of the hybrid photoreceptor cell type characterized in the present study. The coexpression of both rod and cone genes in the same cell could predispose the cell to apoptosis.
The final common feature between mouse rd7 and human NR2E3 mutants is the presence of an abnormally laminated retina with waviness and rosette formation in the ONL [12–15]. The cause of this abnormality is not known, but it is possibly related to defects in photoreceptor cell polarity in the rd7 mutant. Rosette formation and abnormally wavy epithelia are common sequelae of defects in pathways controlling cell polarity [46,47]. In particular, loss-of-function mutations in the polarity gene crumbs
(CRB1) have been shown to cause morphological abnormalities of the ONL in both humans and mice, including rosette formation in mice very similar to that seen in the rd7 mutant [48,49]. Interestingly, Sharon et al. [5] have recently pointed out additional features shared by patients with CRB1 mutations and mutations in NR2E3, including hyperopic refractive errors and a distinctive pattern of clumped pigmentation in the retina.
In the present study we found the mouse crumbs ortholog to be up-regulated in the rd7 mutant retina by microarray, consistent with its higher expression level in cones than in rods [50]. Although we were unable to confirm this finding by in situ hybridization due to the weakness of the signal, it is possible that the up-regulation of crumbs in the retina is the cause of the lamination defects seen in the rd7 mutant. Overexpression of wild-type crumbs in Drosophila has been shown to cause polarity defects leading to waviness of epithelia and even to misalignment of nuclei within photoreceptors analogous to what is seen in the rd7 retina [47,51]. Future experiments will address this question by overexpressing full-length Crb1 in a wild-type background.
One further point worthy of note is the striking similarity between the hybrid photoreceptor identified in this study and a naturally occurring photoreceptor found in ground squirrels. The “rods” of this species have electrophysiologic, molecular, and ultrastructural features of both rods and cones [52–58]. Although these unusual findings have been difficult to interpret under the usual assumptions of “duplicity theory” [56], we would like to suggest that ground squirrels may have experienced a naturally occurring down-regulation or loss of Nr2e3 expression in their “rods” that transformed them into a hybrid photoreceptor cell type. The adaptive significance of such a change, if any, is unknown, and it may simply be due to relaxation of selective pressure for night vision in this strictly diurnal species.
Materials and Methods
Mutant mice.
Nr2e3rd7 mutant mice were obtained from Jackson Laboratories (Bar Harbor, Maine, United States; stock #004643) and maintained on a C57BL/6 background. All control mice were C57BL/6.
Microarray analysis.
Total retinal RNA samples were isolated from P0, P6, P14, and P21 Nr2e3 mutant mice using the Trizol reagent (Gibco, San Diego, California. United States). Total retinal RNA samples from age-matched wild-type C57BL/6 mice were used as controls. Individual total RNA samples were derived from four retinas (pooled from two animals). All microarray experiments were performed in triplicate, in each case with separate RNA preparations. Microarray experiments with cDNAs were performed with the P0, P6, and P14 derived samples. Probes were labeled with either Cy3 or Cy5 using the Array 900 kit from Genisphere (Hatfield, Pennsylvania, United States) starting with 5 μg of total RNA according to the manufacturer's instructions. Wild-type control probes were compared to mutant on the same microarray. In two of the three replicates, the mutant probe was labeled with Cy3 and the wild type with Cy5, and in the third replicate the dyes were swapped. Labeled probe was hybridized to microarray slides spotted with approximately 11,500 cDNA clones from the brain molecular anatomy project library (kind gift of B. Soares, University of Iowa; see http://trans.nih.gov/bmap/index.htm for details) and 500 cDNA clones from our lab collection. Slides were printed and hybridized as described [59,60]. After hybridization and washing of slides according to the manufacturer's instructions (Genisphere), the slides were scanned on an Axon Instruments (Union City, California, United States) GenePix 4000 scanner and images were analyzed using the accompanying GenePix Pro software package. The complete raw cDNA microarray data set are available in Tables S6–S14.
Two types of normalization were performed on the raw intensity scores derived from the GenePix Pro analysis. First, for a given experiment, the average intensity of all the spots in the weaker of the two channels (Cy3 or Cy5) was normalized to those in the stronger channel. Second, in a given set of experiments done in triplicate at a particular time point, the two experiments with the weaker average signal intensity over all spots were normalized to those in the third microarray with the strongest average signal intensity. All spots with signal levels equal to or below background were removed from the analysis. The resulting files contained on average about 6,000 spots. These files were then sorted according to Cy3/Cy5 signal intensity, and those spots with the 10% highest and 10% lowest intensity ratios (approximately 600 spots/experiment) were compared across the three experiments at a given time point using custom Perl scripts (available upon request from JCC). All spots which were present in the top 10% most up- or down-regulated genes in two out of three or three out of three experiments were recorded (the latter are listed in Table S3).
Microarray analysis of the P21 retinas was performed on Affymetrix mouse genome 430 2.0 GeneChip array (Affymetrix, Santa Clara, California, United States). A total of six microarray hybridizations were performed: three with probes derived from mutant RNA and three from wild-type. Probes were synthesized starting with 10 μg of total RNA for each sample according to manufacturer's instructions (Affymetrix). Hybridization, washing, and scanning of the microarrays were all performed at the Bauer Center for Genomics Research at Harvard University according to manufacturer's instructions (Affymetrix). Initial data analysis was carried out using the GeneChip Operating System (GCOS) software from Affymetrix. Pairwise comparisons were made between individual mutant microarray results and controls. All genes were removed from the analysis for which “absent” calls were made by the software for both the wild-type and mutant samples being compared. The remaining gene lists contained approximately 26,000 transcripts. These lists were then sorted according to the mutant-to-wild-type “signal log ratio” in order to identify the most markedly up- and down-regulated genes. The top 500 most up- and down-regulated transcripts (approximately 2% of all genes in each case) from each of the three pairwise comparisons between mutant and wild-type were compared using custom Perl scripts (available upon request from JCC) to identify those genes that were present in two or three out of three lists. Those genes that were up- or down-regulated in three out of three experiments were recorded (Tables S4 and S5). The complete pairwise Affymetrix microarray datasets are available in Tables S15–S17.
In Situ hybridization.
Section in situ hybridization was performed as previously described [61] using 20-μm cryosections from OCT-embedded tissue or 4-μm paraffin sections. All in situ hybridizations were performed with the mutant and wild-type control sections on the same glass slide. Riboprobes labeled with digoxygenin-tagged UTP (Roche, Basel, Switzerland) were detected with NBT/BCIP (Sigma, St. Louis, Missouri, United States). The sources of the individual riboprobes used in this study are described in Tables S1 and S2. Dissociated cell in situ hybridization was performed as described previously [62] using the same S-opsin digoxygenin-labeled probe used for section in situ hybridization. All images were captured on a compound microscope (Eclipse e1000; Nikon, Tokyo, Japan) using a CCD camera (DXM1200F, Nikon). S-opsin positive cells were quantitated on dissociated cell in situ hybridization as previously described [62]. Twenty fields were quantitated in this manner at 200× magnification for both rd7 and wild-type retinas.
Immunohistochemistry.
For antibody staining, cryosections were prepared and stained as described previously [63]. Primary antibodies used were a polyclonal anti-blue opsin raised in rabbit (1:300; Chemicon International, Temecula, California, United States; AB5407) and a mouse monoclonal anti-rhodopsin (1:200; rho4D2). Secondary antibodies used were Cy2- or Cy3-conjugated goat anti-rabbit and anti-mouse (1:500; Jackson Immunologicals, West Grove, Pennsylvania, United States). Following antibody staining, 4′-DAPI was applied to stain nuclei (Sigma), and the sections were coverslipped and mounted in Gel/Mount (Biomeda, Foster City, California, United States).
Electron microscopy.
This protocol was adapted from one used by Raviola [64] with some modifications derived from Carter-Dawson and Lavail [26]. Four adult wild-type and four mutant animals were deeply anesthetized by intraperitoneal injection of Avertin and the eyes were then removed. The cornea was gently punctured with sharp forceps and excised with iridectomy scissors. The eye was then transferred to a solution of 2% paraformaldehyde/2% glutaraldehyde in cacodylate buffer (0.1 M cacodylic acid; 0.1% calcium chloride). The lens was gently removed and the eyecup allowed to fix for 2 h at room temperature. The fixed eye was then placed on dental wax and sectioned in the midline with a fresh razor blade (half of the retinas were sectioned along the D-V axis and half along the nasal-temporal axis). The retinas (and attached retinal pigment epithelium) were carefully dissected away from the sclera, which was discarded.
The retinas were then rinsed four times for 15 min each with Sorenson's buffer (pH 7.4). They were then stained for 2 h at 4 °C with 1% osmium tetroxide in 1.5% potassium ferrocyanide. Next, the retinas were rinsed four times for 15 min in maleate buffer (pH 5.1) and then stained for 2 h at room temperature with 1% uranyl acetate in maleate buffer (pH 6.2). They were then washed four times for 15 min with maleate buffer (pH 5.1); once for 10 min with 70% ethanol; once for 10 min with 95% ethanol; and four times for 30 min with 100% ethanol. Next, the retinas were washed three times over one hour with propylene oxide and then embedded in TAAB 812 Resin (Marivac, Quebec, Canada) for 1–2 d in a 60 °C oven. Semi-thin sections were cut at a thickness of 0.5 μm and stained with 1% toluidine blue in 1% sodium borate buffer. Images of semi-thin sections from the mutant retina were taken within the ventral two-thirds of the retina where the majority of the supernumerary S-opsin–expressing cells reside. Wild-type images were taken in comparable regions. Sections for electron microscopy were cut at a thickness of 95 nm, placed on grids, and poststained with 2% uranyl acetate followed by 0.2% lead citrate. All sections were cut in a plane perpendicular to the plane of the photoreceptor layer. They were then visualized on a Jeol 1200EX electron microscope (Jeol, Tokyo, Japan). The electron microscopic images in Figure 7 derive from the ventral two-thirds of the wild-type and mutant retinas.
The area of the cell bodies of the rods and “rod-like” cells in the wild-type and mutant ONLs, respectively, were quantitated in the following manner. First, ten fields within the ONL were chosen at random at 1,000× magnification and then photographed at 4,000× magnification for each of the two genotypes. Such images typically contained 35–45 cell bodies. In order to quantitate only the area of those cell bodies that were cut as near to the midline of the cell as possible (i.e., in order to obtain the maximal cross-sectional area) the five cells with the largest apparent area in each photograph were chosen by eye and the outline of the cell membranes were traced onto white paper. These tracings were scanned along with the size bar from the electron micrographs, and the areas of the resulting digital images were quantitated using Scion Image software (NIH Image, http://rsb.info.nih.gov/nih-image). A total of 50 cells of each genotype were quantitated in this manner. In addition, in order to evaluate the percentage of rods in the wild-type retina that had any juxtanuclear mitochondria, all rods within all ten images were counted as were the number of cells showing juxtanuclear organelles. A total of six out of 399 wild-type rods (1.5%) possessed a juxtanuclear organelle. This analysis permitted us to evaluate the wild-type rods for juxtanuclear mitochondria at multiple planes of section; however, serial sections of individual cell bodies were not performed.
Supporting Information
Figures S1–S7 show the in situ hybridization images of all genes discussed in the paper (see Tables S1 and S2). All paired images (which show the wild-type control on the left and the rd7 mutant retina on the right) are labeled in the lower left-hand corner with the gene symbol followed by the age of the retinas in question (P6, P14, P28, or adult). Unpaired images represent prenatal time points and are labeled with the gene symbol of the gene in question (“wt” indicates that the retina is from a wild-type animal) and a designation of the embryonic day from which the retina derives (e.g., e17.5 = embryonic day 17.5).
Figure S1 In Situ Hybridization Images for G1–G13 in Table S1
(3.1 MB JPG)
Click here for additional data file.
Figure S2 In Situ Hybridization Images for G14–G25 in Table S1
(2.4 MB JPG)
Click here for additional data file.
Figure S3 In Situ Hybridization Images for G26–G35 in Table S1
(2.1 MB JPG)
Click here for additional data file.
Figure S4 In Situ Hybridization Images for G36–G46 in Table S1
(2.0 MB JPG)
Click here for additional data file.
Figure S5 In Situ Hybridization Images for G47–G53 in Table S1
(1.7 MB JPG)
Click here for additional data file.
Figure S6 In Situ Hybridization Images for Genes 1–11 in Table S2
(2.9 MB JPG)
Click here for additional data file.
Figure S7 In Situ Hybridization Images for Genes 12–22 in Table S2
(3.2 MB JPG)
Click here for additional data file.
Figure S8 In Situ Hybridization Results for M-Opsin and Thrb2
Note that the M-opsin–positive cells are scattered throughout the ONL at P14, but appear to have migrated to the scleral edge of the ONL by P28. There is no change in the number of M-opsin– or Thrb2-positive cells in the rd7 mutant.
(2.1 MB PDF)
Click here for additional data file.
Table S1 Cone-Specific and Cone-Enriched Genes Evaluated in the rd7 Mutant by Microarray and In Situ Hybridization
This table is a supplemental version of Figure 1. “Figure Number” indicates which figure (Figure S1–S5) contains the in situ hybridization images corresponding to the gene in question. “Lab Clone Information” indicates the region of the gene in question from which the probe used for in situ hybridization was derived. All abbreviations are as indicated in Figure 1.
(26 KB XLS)
Click here for additional data file.
Table S2 Rod Genes Evaluated in the rd7 Mutant by In Situ Hybridization
This table contains details about the in situ hybridization patterns of 22 genes (many of which are rod-specific) evaluated in the rd7 mutant retina. “Figure Number” indicates which figure (Figure S6 or S7) contains the in situ hybridization images corresponding to the gene in question. “Lab Clone Information” indicates the region of the gene in question from which the probe used for in situ hybridization was derived. The color coding of the in situ hybridization results under “In Situ Pattern” is as follows: dark green, markedly down-regulated; light green, mildly down-regulated; red, markedly up-regulated; orange, mildly up-regulated.
(20 KB XLS)
Click here for additional data file.
Table S3 Summary of cDNA Microarray Results from P0, P6, and P14
The spots listed in this table represent those that were either up- or down-regulated in three out of three microarray experiments as described in Materials and Methods. The Cy3/Cy5 signal ratios are indicated for all three microarray experiments at each time point. Note that the Cy3/Cy5 ratios for “Microarray #3” are reversed relative to the other two, since the fluorescent tag used to label wild-type and mutant RNA was swapped as described in Materials and Methods.
(22 KB XLS)
Click here for additional data file.
Table S4 Summary of Genes Up-Regulated in rd7 Mutant Retina at P21 by Affymetrix Microarray
Only genes that were found to be up-regulated in three out of three microarray experiments (as described in Materials and Methods) are listed. “Nr2e3 signal” and “C57BL/6 signal” represent the average signal for that transcript in all three microarray experiments.
(92 KB XLS)
Click here for additional data file.
Table S5 Summary of Genes Down-Regulated in rd7 Mutant Retina at P21 by Affymetrix Microarray
Only genes that were found to be down-regulated in three out of three microarray experiments (as described in Materials and Methods) are listed. “Nr2e3 signal” and “C57BL/6 signal” represent the average signal for that transcript in all three microarray experiments.
(58 KB XLS)
Click here for additional data file.
Table S6 Raw cDNA Microarray Data for rd7 versus Wild-Type Comparison at P0 (I)
(5.6 MB XLS)
Click here for additional data file.
Table S7 Raw cDNA Microarray Data for rd7 versus Wild-Type Comparison at P0 (II)
(5.6 MB XLS)
Click here for additional data file.
Table S8 Raw cDNA Microarray Data for rd7 versus Wild-Type Comparison at P0 (III)
(5.6 MB XLS)
Click here for additional data file.
Table S9 Raw cDNA Microarray Data for rd7 versus Wild-Type Comparison at P6 (I)
(5.6 MB XLS)
Click here for additional data file.
Table S10 Raw cDNA Microarray Data for rd7 versus Wild-Type Comparison at P6 (II)
(5.6 MB XLS)
Click here for additional data file.
Table S11 Raw cDNA Microarray Data for rd7 versus Wild-Type Comparison at P6 (III)
(5.6 MB XLS)
Click here for additional data file.
Table S12 Raw cDNA Microarray Data for rd7 versus Wild-Type Comparison at P14 (I)
(5.5 MB XLS)
Click here for additional data file.
Table S13 Raw cDNA Microarray Data for rd7 versus Wild-Type Comparison at P14 (II)
(5.6 MB XLS)
Click here for additional data file.
Table S14 Raw cDNA Microarray Data for rd7 versus Wild-Type Comparison at P14 (III)
(5.6 MB XLS)
Click here for additional data file.
Table S15 Raw Affymetrix Microarray Data for rd7 versus Wild-Type Comparison at P21 (I)
(18 MB XLS)
Click here for additional data file.
Table S16 Raw Affymetrix Microarray Data for rd7 versus Wild-Type Comparison at P21 (II)
(18 MB XLS)
Click here for additional data file.
Table S17 Raw Affymetrix Microarray Data for rd7 versus Wild-Type Comparison at P21 (III)
(18 MB XLS)
Click here for additional data file.
We are grateful to E. Raviola and T. Reese for help with electron microscopy and to A. Jadhav and J. Trimarchi for access to unpublished data and reagents. Thanks to A. Jadhav, J. Trimarchi, D. Kim, and T. Cherry for helpful comments on the manuscript. This work was supported by the Howard Hughes Medical Institute and grants from the National Institutes of Health (EY014822 to JCC and EY009676 to CLC). Thanks to A. Swaroop for providing us with Nrl mutant mice.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. JCC and CLC conceived and designed the experiments. JCC performed the experiments. JCC and CLC analyzed the data. JCC contributed reagents/materials/analysis tools. JCC and CLC wrote the paper.
Abbreviations
DAPI6-diamidino-2-phenylindole
E[number]embryonic day [number]
ERGelectroretinogram
ESCSenhanced S-cone syndrome
GCLganglion cell layer
INLinner nuclear layer
ONLouter nuclear layer
P[number]postnatal day [number]
==== Refs
References
Jacobson SG Marmor MF Kemp CM Knighton RW 1990 SWS (blue) cone hypersensitivity in a newly identified retinal degeneration Invest Ophthalmol Vis Sci 31 827 838 2335450
Hood DC Cideciyan AV Roman AJ Jacobson SG 1995 Enhanced S cone syndrome: Evidence for an abnormally large number of S cones Vision Res 35 1473 1481 7645276
Haider NB Jacobson SG Cideciyan AV Swiderski R Streb LM 2000 Mutation of a nuclear receptor gene, NR2E3, causes enhanced S cone syndrome, a disorder of retinal cell fate Nat Genet 24 127 131 10655056
Bumsted O'Brien KM Cheng H Jiang Y Schulte D Swaroop A 2004 Expression of photoreceptor-specific nuclear receptor NR2E3 in rod photoreceptors of fetal human retina Invest Ophthalmol Vis Sci 45 2807 2812 15277507
Sharon D Sandberg MA Caruso RC Berson EL Dryja TP 2003 Shared mutations in NR2E3 in enhanced S-cone syndrome, Goldmann-Favre syndrome, and many cases of clumped pigmentary retinal degeneration Arch Ophthalmol 121 1316 1323 12963616
Fishman GA Peachey NS 1989 Rod-cone dystrophy associated with a rod system electroretinogram obtained under photopic conditions Ophthalmology 96 913 918 2787013
Marmor MF 1989 Large rod-like photopic signals in a possible new form of congenital night blindness Doc Ophthalmol 71 265 269 2789128
Perlman I Leibu R Barth J 1993 Night blindness: A new type with abnormal properties of the electroretinogram Clinical Vision Sciences 8 159 169
Jacobson SG Roman AJ Roman MI Gass JD Parker JA 1991 Relatively enhanced S cone function in the Goldmann-Favre syndrome Am J Ophthalmol 111 446 453 2012146
Marmor MF Jacobson SG Foerster MH Kellner U Weleber RG 1990 Diagnostic clinical findings of a new syndrome with night blindness, maculopathy, and enhanced S cone sensitivity Am J Ophthalmol 110 124 134 2378376
Roman AJ Jacobson SG 1991 S cone-driven but not S cone-type electroretinograms in the enhanced S cone syndrome Exp Eye Res 53 685 690 1743268
Jacobson SG Sumaroka A Aleman TS Cideciyan AV Schwartz SB 2004 Nuclear receptor NR2E3 gene mutations distort human retinal laminar architecture and cause an unusual degeneration Hum Mol Genet 13 1893 1902 15229190
Milam AH Rose L Cideciyan AV Barakat MR Tang WX 2002 The nuclear receptor NR2E3 plays a role in human retinal photoreceptor differentiation and degeneration Proc Natl Acad Sci U S A 99 473 478 11773633
Peyman GA Fishman GA Sanders DR Vlchek J 1977 Histopathology of Goldmann-Favre syndrome obtained by full-thickness eye-wall biopsy Ann Ophthalmol 9 479 484 301373
Akhmedov NB Piriev NI Chang B Rapoport AL Hawes NL 2000 A deletion in a photoreceptor-specific nuclear receptor mRNA causes retinal degeneration in the rd7 mouse Proc Natl Acad Sci U S A 97 5551 5556 10805811
Yanagi Y Takezawa S Kato S 2002 Distinct functions of photoreceptor cell-specific nuclear receptor, thyroid hormone receptor beta2 and CRX in one photoreceptor development Invest Ophthalmol Vis Sci 43 3489 3494 12407160
Haider NB Naggert JK Nishina PM 2001 Excess cone cell proliferation due to lack of a functional NR2E3 causes retinal dysplasia and degeneration in rd7/rd7 mice Hum Mol Genet 10 1619 1626 11487564
Chen J Rattner A Nathans J 2005 The rod photoreceptor-specific nuclear receptor Nr2e3 represses transcription of multiple cone-specific genes J Neurosci 25 118 129 15634773
Peng GH Ahmad O Ahmad F Liu J Chen S 2005 The photoreceptor-specific nuclear receptor Nr2e3 interacts with Crx and exerts opposing effects on the transcription of rod versus cone genes Hum Mol Genet 14 747 764 15689355
Jeon CJ Strettoi E Masland RH 1998 The major cell populations of the mouse retina J Neurosci 18 8936 8946 9786999
Applebury ML Antoch MP Baxter LC Chun LL Falk JD 2000 The murine cone photoreceptor: A single cone type expresses both S and M opsins with retinal spatial patterning Neuron 27 513 523 11055434
Ng L Hurley JB Dierks B Srinivas M Salto C 2001 A thyroid hormone receptor that is required for the development of green cone photoreceptors Nat Genet 27 94 98 11138006
Rich KA Zhan Y Blanks JC 1997 Migration and synaptogenesis of cone photoreceptors in the developing mouse retina J Comp Neurol 388 47 63 9364238
Mears AJ Kondo M Swain PK Takada Y Bush RA 2001 Nrl is required for rod photoreceptor development Nat Genet 29 447 452 11694879
Furukawa T Morrow EM Li T Davis FC Cepko CL 1999 Retinopathy and attenuated circadian entrainment in Crx-deficient mice Nat Genet 23 466 470 10581037
Carter-Dawson LD LaVail MM 1979 Rods and cones in the mouse retina. I. Structural analysis using light and electron microscopy J Comp Neurol 188 245 262 500858
Cohen AI 1960 The ultrastructure of the rods of the mouse retina Am J Anat 107 23 48 13694328
Kurtz A Zimmer A Schnutgen F Bruning G Spener F 1994 The expression pattern of a novel gene encoding brain-fatty acid binding protein correlates with neuronal and glial cell development Development 120 2637 2649 7956838
Godbout R, Bisgrove DA, Shkolny D, Day RS, 3rd 1998 Correlation of B-FABP and GFAP expression in malignant glioma Oncogene 16 1955 1962 9591779
Feng L Hatten ME Heintz N 1994 Brain lipid-binding protein (BLBP): A novel signaling system in the developing mammalian CNS Neuron 12 895 908 8161459
Soderpalm A Szel A Caffe AR van Veen T 1994 Selective development of one cone photoreceptor type in retinal organ culture Invest Ophthalmol Vis Sci 35 3910 3921 7928189
Wikler KC Szel A Jacobsen AL 1996 Positional information and opsin identity in retinal cones J Comp Neurol 374 96 107 8891949
Thornalley PJ 2003 Glyoxalase I—Structure, function and a critical role in the enzymatic defence against glycation Biochem Soc Trans 31 1343 1348 14641060
Blackshaw S Fraioli RE Furukawa T Cepko CL 2001 Comprehensive analysis of photoreceptor gene expression and the identification of candidate retinal disease genes Cell 107 579 589 11733058
Nihira M Anderson K Gorin FA Burns MS 1995 Primate rod and cone photoreceptors may differ in glucose accessibility Invest Ophthalmol Vis Sci 36 1259 1270 7775103
Tsujino S Shanske S DiMauro S 1993 Molecular genetic heterogeneity of myophosphorylase deficiency (McArdle's disease) N Engl J Med 329 241 245 8316268
Morrow EM Belliveau MJ Cepko CL 1998 Two phases of rod photoreceptor differentiation during rat retinal development J Neurosci 18 3738 3748 9570804
van Roon MA Aten JA van Oven CH Charles R Lamers WH 1989 The initiation of hepatocyte-specific gene expression within embryonic hepatocytes is a stochastic event Dev Biol 136 508 516 2479586
Newlands S Levitt LK Robinson CS Karpf AB Hodgson VR 1998 Transcription occurs in pulses in muscle fibers Genes Dev 12 2748 2758 9732272
Rossi FM Kringstein AM Spicher A Guicherit OM Blau HM 2000 Transcriptional control: Rheostat converted to on/off switch Mol Cell 6 723 728 11030351
Walters MC Fiering S Eidemiller J Magis W Groudine M 1995 Enhancers increase the probability but not the level of gene expression Proc Natl Acad Sci U S A 92 7125 7129 7624382
Ko MS Nakauchi H Takahashi N 1990 The dose dependence of glucocorticoid-inducible gene expression results from changes in the number of transcriptionally active templates EMBO J 9 2835 2842 2167833
Fiering S Whitelaw E Martin DI 2000 To be or not to be active: The stochastic nature of enhancer action Bioessays 22 381 387 10723035
Bhat PJ Venkatesh KV 2005 Stochastic variation in the concentration of a repressor activates GAL genetic switch: Implications in evolution of regulatory network FEBS Lett 579 597 603 15670814
Chang B Hawes NL Hurd RE Davisson MT Nusinowitz S 2002 Retinal degeneration mutants in the mouse Vision Res 42 517 525 11853768
Klezovitch O Fernandez TE Tapscott SJ Vasioukhin V 2004 Loss of cell polarity causes severe brain dysplasia in Lgl1 knockout mice Genes Dev 18 559 571 15037549
Tepass U Theres C Knust E 1990
crumbs encodes an EGF-like protein expressed on apical membranes of Drosophila epithelial cells and required for organization of epithelia Cell 61 787 799 2344615
Mehalow AK Kameya S Smith RS Hawes NL Denegre JM 2003 CRB1 is essential for external limiting membrane integrity and photoreceptor morphogenesis in the mammalian retina Hum Mol Genet 12 2179 2189 12915475
Jacobson SG Cideciyan AV Aleman TS Pianta MJ Sumaroka A 2003 Crumbs homolog 1 (CRB1) mutations result in a thick human retina with abnormal lamination Hum Mol Genet 12 1073 1078 12700176
Pellikka M Tanentzapf G Pinto M Smith C McGlade CJ 2002 Crumbs, the Drosophila homologue of human CRB1/RP12, is essential for photoreceptor morphogenesis Nature 416 143 149 11850625
Fan SS Chen MS Lin JF Chao WT Yang VC 2003 Use of gain-of-function study to delineate the roles of crumbs in Drosophila eye development J Biomed Sci 10 766 773 14631116
West RW Dowling JE 1975 Anatomical evidence for cone and rod-like receptors in the gray squirrel, ground squirrel, and prairie dog retinas J Comp Neurol 159 439 460 1127139
Fisher SK Jacobs GH Anderson DH Silverman MS 1976 Rods in the antelope ground squirrel Vision Res 16 875 877 960615
Jacobs GH Fisher SK Anderson DH Silverman MS 1976 Scotopic and photopic vision in the California ground squirrel: Physiological and anatomical evidence J Comp Neurol 165 209 227 1245613
Jacobs GH Tootell RB Fisher SK Anderson DH 1980 Rod photoreceptors and scotopic vision in ground aquirrels J Comp Neurol 189 113 125 7351444
Jacobs GH 1990 Duplicity theory and ground squirrels: linkages between photoreceptors and visual function Vis Neurosci 5 311 318 2134853
Szel A Rohlich P 1988 Four photoreceptor types in the ground squirrel retina as evidenced by immunocytochemistry Vision Res 28 1297 1302 3256146
von Schantz M Szel A van Veen T Farber DB 1994 Expression of phototransduction cascade genes in the ground squirrel retina Invest Ophthalmol Vis Sci 35 2558 2566 7512947
Dyer MA Livesey FJ Cepko CL Oliver G 2003 Prox1 function controls progenitor cell proliferation and horizontal cell genesis in the mammalian retina Nat Genet 34 53 58 12692551
Livesey FJ Young TL Cepko CL 2004 An analysis of the gene expression program of mammalian neural progenitor cells Proc Natl Acad Sci U S A 101 1374 1379 14734810
Murtaugh LC Chyung JH Lassar AB 1999 Sonic hedgehog promotes somitic chondrogenesis by altering the cellular response to BMP signaling Genes Dev 13 225 237 9925646
Blackshaw S Harpavat S Trimarchi J Cai L Huang H 2004 Genomic analysis of mouse retinal development PLoS Biol 2 E247 15226823
Chen CM Cepko CL 2002 The chicken RaxL gene plays a role in the initiation of photoreceptor differentiation Development 129 5363 5375 12403708
Morrow EM Furukawa T Raviola E Cepko CL 2005 Synaptogenesis and outer segment formation are perturbed in the neural retina of Crx mutant mice BMC Neurosci 6 5 15676071
|
16110338
|
PMC1186732
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Aug 5; 1(2):e11
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010011
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1611033910.1371/journal.pgen.001001305-PLGE-RA-0057R2plge-01-02-01Research ArticleEvolutionGenetics/GenomicsArabidopsis (Thale Cress)ArabidopsisGametophytic Selection in Arabidopsis thaliana Supports the Selective Model of Intron Length Reduction Gametophytic Selection in
ArabidopsisSeoighe Cathal 1*Gehring Chris 2Hurst Laurence D 31 Institute of Infectious Disease and Molecular Medicine, University of Cape Town, Rondebosch, South Africa
2 Department of Biotechnology, University of the Western Cape, Bellville, South Africa
3 Department of Biology and Biochemistry, University of Bath, Somerset, United Kingdom
Gibson Greg EditorNorth Carolina State University, United States of America* To whom correspondence should be addressed. E-mail: [email protected] 2005 5 8 2005 1 2 e1325 3 2005 31 5 2005 Copyright: © 2005 Seoighe 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.Why do highly expressed genes have small introns? This is an important issue, not least because it provides a testing ground to compare selectionist and neutralist models of genome evolution. Some argue that small introns are selectively favoured to reduce the costs of transcription. Alternatively, large introns might permit complex regulation, not needed for highly expressed genes. This “genome design” hypothesis evokes a regionalized model of control of expression and hence can explain why intron size covaries with intergene distance, a feature also consistent with the hypothesis that highly expressed genes cluster in genomic regions with high deletion rates. As some genes are expressed in the haploid stage and hence subject to especially strong purifying selection, the evolution of genes in Arabidopsis provides a novel testing ground to discriminate between these possibilities. Importantly, controlling for expression level, genes that are expressed in pollen have shorter introns than genes that are expressed in the sporophyte. That genes flanking pollen-expressed genes have average-sized introns and intergene distances argues against regional mutational biases and genomic design. These observations thus support the view that selection for efficiency contributes to the reduction in intron length and provide the first report of a molecular signature of strong gametophytic selection.
Synopsis
Genes are odd things. Small proteins are often encoded by big genes. In the process, much of the excess material has to be cut out and thrown away. The size of the parts that are discarded (introns) differs greatly between genes. Why should this be so? The authors test three different ideas, making use of the unusual fact that in plants genes are expressed in pollen. As pollen has only one copy of every gene, natural selection is expected to work somewhat better. The authors find that the non-coding parts of genes that are especially active in pollen are particularly small. They also find that being active in pollen tends to make introns small. This provides strong support for the idea that small introns are the result of selection to reduce costs of making too much material that is only going to be thrown away.
Citation:Seoighe C, Gehring C, Hurst LD (2005) Gametophytic selection in Arabidopsis thaliana supports the selective model of intron length reduction. PLoS Genet 1(2): e13.
==== Body
Introduction
Selection for efficiency has been proposed to explain the reduced intron lengths of broadly or highly expressed genes in several animal systems [1–5]. Because of the energetic cost of transcription [4–6], which is proportional to the length of the transcript and the amount of the transcript that is produced, highly expressed genes are likely to experience greater selective pressure for a reduction in transcript length. This model sees long introns in weakly expressed genes as the result of weakened negative selection. This interpretation of the negative correlation between intron size and gene expression level [1–5] has recently been challenged. The genomic design hypothesis suggests that the shorter introns of highly expressed genes may not be the result of purifying selection, but instead reflect a reduced level of epigenetic regulation in housekeeping genes, which are often expressed at high levels [7]. Under this hypothesis, selection actively favours the accumulation of longer introns in less highly expressed genes because many of these genes are tissue specific and require greater levels of epigenetic regulation. This is supported by the fact that intergenic distances also tend to be reduced in the vicinity of highly expressed genes [2,7,8], an observation that is not explained by the transcriptional efficiency model. Moreover, if one controls for intergene distance, it is as yet unclear whether, in humans, highly expressed genes have small introns as reports are contradictory [2,7]. Hence, the relevance of the transcriptional efficiency model is currently uncertain. The correlation between intergene distance and intronic size has also been interpreted as evidence for a regional mutational bias, coupled with neutral evolution [2]. Indeed, regions of high compaction tend to be GC rich [2,8] and hence regions of high recombination rates [9]. If recombination induces deletions, then a simple mutational bias/neutral model can be considered.
Owing to the fact that it has abundant genes that are haploid expressed, Arabidopsis thaliana provides a novel testing ground to examine these conflicting viewpoints. Strong selection at the gametophytic stage, owing to haploid exposure of recessive mutations and/or to strong pollen competition [10–12], has been proposed as a key aspect of plant evolutionary biology resulting in the purging of deleterious mutations in genes that are transcribed in the growing pollen tube [13,14]. A transcriptional cost view of intron length variation predicts that this strong purifying selection should cause a reduction in intron lengths in genes that are expressed in pollen compared to genes that are expressed elsewhere and that this reduction should be most pronounced in the most highly expressed genes.
Results/Discussion
Introns, particularly those toward the 5′ ends of genes, may often have regulatory functions [15]. The lengths of introns in animals have been shown to decrease as a function of the position of the intron, counting from the 5′ end, and to depend to some extent on the breadth of expression (i.e., the number of tissues in which the gene is expressed [5]). We find a similar reduction in intron length as a function of intron position in Arabidopsis (Figure 1). In order to reduce the impact of positional effects and regulatory elements associated with proximal introns, we restricted our analysis of intron lengths of genes that are expressed in pollen to distal introns (from intron 5 to intron 10) since they are less likely to have a role in regulation (our analyses were not sensitive to the cut-off used to classify distal introns).
Figure 1 Mean Intron Length as a Function of Intron Position, Counting from the 5′ End of the Gene
Intron length was nearly constant for introns 5 to 10. Proximity to the 3′ end of the gene was not correlated with intron length. Error bars show ± twice the standard error. The data shown are for genes with exactly ten introns so that positional effects from the 3′ ends can also be assessed.
Using publicly available serial analysis of gene expression (SAGE) data, we compared intron lengths between genes that are expressed in pollen and the sporophyte. A summary of the dataset is provided in Table 1. The average intron length for the pollen genes was 107.7 base pairs (bp), compared to 123.4 bp for introns from genes expressed in at least one of four sporophyte conditions (p = 0.0002). In spite of significant differences in means, the mode of the distribution of intron lengths remained approximately the same in both groups and for all intron positions. Comparison of the distributions of intron lengths shows that there were fewer longer introns among the genes that were expressed in the gametophyte compared to the sporophyte, as indicated by curvature away from the diagonal in a quantile–quantile plot (Figure 2). We also compared intron lengths between genes expressed in the sporophyte and gametophyte with expression level as a covariate, using expression levels from the pollen SAGE dataset and the largest [16] of the four sporophyte SAGE datasets in the study. We found significant evidence for both a negative correlation between intron length and gene expression level (p = 0.01) and a reduced intron length in genes expressed at a given level in the gametophyte compared to the sporophyte (p = 0.001). This latter result suggests that introns from genes expressed in pollen remain shorter than introns from genes in the sporophyte when we control for expression level.
Table 1 Summary of the Dataset
Each cell represents the mean value of the quantity in the column for the subset of genes indicated in the row. The complete dataset used is available as Dataset S1 with this article.
aSAGE data
bMicroarray data
Figure 2 Histograms and Quantile–Quantile Plots of Mean Distal Intron Length
(A) Histogram for genes expressed in the sporophyte microarray datasets.
(B) Histogram for genes expressed in the gametophyte but not the sporophyte microarray datasets.
(C) Quantile–quantile plot of introns from all pollen-expressed and all sporophyte-expressed genes derived from the SAGE dataset. Quantiles of the intron length distributions for genes expressed in the gametophyte and sporophyte are on the x- and y-axes, respectively.
(D) Quantile–quantile plot of introns from pollen-specific and sporophyte-specific genes derived from the microarray dataset.
Might the reduced intron lengths of genes expressed in pollen be sensitive to the method of measurement of gene expression? We compared intron lengths between genes that are expressed in pollen but not in the sporophyte and vice versa using microarray data from the Expression Atlas of Arabidopsis Development [17]. The mean intron lengths for pollen-specific genes was 109.4 bp compared to 134.7 bp for the genes expressed in the sporophyte but not in pollen (p = 3 × 10−9). The expression level of the pollen-specific genes was higher, on average, than for genes that were expressed in pollen and the sporophyte. If expression level in pollen is included as a covariate, the length of introns remained significantly lower in genes that are specific to pollen compared to genes that are specific to sporophyte (p = 5 × 10−5). Introns from genes that were highly expressed in pollen were also significantly shorter than introns from genes that were highly expressed in at least one sporophyte sample, regardless of whether the gene was specific to pollen or expressed in both pollen and sporophyte (p = 0.0007).
The reduction in intron lengths in genes expressed in the pollen SAGE dataset did not appear to be affected by whether the genes were also expressed in the sporophyte, illustrating the potential impact of strong gametophytic selection on sporophyte evolution. In the SAGE dataset, genes that were specific to pollen and genes that were expressed in pollen as well as one of the sporophyte datasets had similar average intron lengths (99.1 bp and 109.7 bp; n = 13 and n = 58, respectively; p = 0.93), while in both cases the introns were significantly or marginally significantly shorter than introns of genes expressed in the sporophyte but not expressed in pollen (p = 0.06 and p = 0.0009). This additionally provides evidence that the observed difference in intron lengths between genes expressed in pollen and the sporophyte is not the result of a lack of intronic regulatory elements in genes that are expressed exclusively in pollen. Contrary to the results from SAGE, the reduction in intron lengths was confined to genes that were specific to pollen in the microarray datasets, possibly due to hybridisation cross-reactivity between homologous genes. Pollen has a high proportion of genes that appear to be expressed in pollen only [18]. The reduced intron lengths observed in such genes is not in keeping with the genomic design argument that suggests that regulation of narrowly expressed genes is responsible for their longer introns compared to broadly and highly expressed genes.
To test whether altered rates of insertion or deletion or a higher gene density in the genomic regions containing the genes that are expressed in pollen could be responsible for the reduced intron lengths, we calculated the average intron lengths of the closest genomic neighbours of the pollen genes from the SAGE dataset. The mean intron length of the neighbouring genes was not significantly different to the mean for all genes (132.5 bp compared to 134.8 bp, p = 0.14). The mean intron length for genes expressed in pollen remained significantly below the mean for genes expressed in the sporophyte, considering only the closest sporophyte-expressed neighbour for each gene expressed in pollen (p = 0.01). Thus, regional genomic effects evoked by the genomic design hypothesis [7] and the mutational bias hypothesis are not likely to be the cause of the reduced intron lengths of genes expressed in pollen. Furthermore, although the mean length of flanking regions was slightly greater for genes that were expressed in the sporophyte compared to pollen, the difference was not statistically significant (1,946 bp and 1,762 bp, for sporophyte and pollen, respectively; p = 0.57). Restricting to genes with at least five introns (the genes that contributed to this study), this difference is reduced, and the pollen genes, in fact, have slightly longer intergenic regions, although, again, the difference is not statistically significant (1,826 bp and 1,913 bp for sporophyte and pollen, respectively; p = 0.15).
The introns of genes that were highly expressed in at least one of the sporophyte expression sites in the study were significantly reduced in length compared to all genes expressed in the sporophyte (111.1 bp compared to 123.4 bp, p = 0.004). Under the genomic design hypothesis, this might be explained by the fact that highly expressed genes are often ubiquitous and do not require much regulatory information in introns or flanking regions [7]. A regional mutational bias model could explain the reduced introns if highly expressed genes are associated with high rates of deletions. Both of these hypotheses are supported by a positive correlation between the lengths of introns and flanking intergenic regions in human [2,7,8]. In contrast, for most genes there is very little correlation between intron length and the mean length of 5′ and 3′ flanking regions in Arabidopsis (Spearman ρ = 0.02, p = 0.09). Furthermore, the length of intergenic regions was not significantly correlated with mean expression in the sporophyte, and the intergenic regions flanking genes that were highly expressed in the sporophyte were not reduced in length (2,037.8 bp, compared to the mean of 1,986.6, p = 0.10). Thus, we find no evidence of a contribution from gene regulation or regional mutational effects to intron length variation in Arabidopsis, whereas the reduced intron lengths of genes expressed in pollen strongly support the transcriptional efficiency model.
Genes that were expressed in pollen had significantly lower intron densities (number of introns per kilobase of exon) than genes that were expressed in at least one of the sporophyte conditions (2.4 introns per kb compared to 3.0 introns per kb; p = 0.001). The genes that were the most highly expressed in pollen had an average intron density that was lower still (1.8 introns per kb), significantly lower than for the genes that were highly expressed in at least one of the sporophyte conditions (2.6 introns per kb; p = 0.01). It is possible that the reduced intron densities result from a disproportionate number of partially processed retroposed genes in the pollen gene dataset rather than from selection for efficiency. Although we cannot rule this out, if retroposition is indeed responsible we might expect an increased density of introns toward the 3′ end [19]. However, we find that the relative density of introns in the 3′ and 5′ halves of genes expressed in pollen is no different to genes expressed in the sporophyte (data not shown).
The debate over whether selection to reduce the cost of transcription is indeed responsible for the shorter intron lengths observed in highly and broadly expressed animal genes has remained unresolved [2,7]. Very high levels of competition at the gametophytic stage of plants provide a useful system in which selection hypotheses can be explored. Although natural selection acting on genes that are expressed at the haploid stage is thought to be an important aspect of plant evolutionary biology [13,14], the reduction in intron lengths that we observe in the genes that are expressed in pollen represents the first well-demonstrated example of the impact of gametophytic selection on the genome of a plant. At least in the case of the genes that are expressed in pollen, there is strong evidence that selection for efficiency, rather than genomic design or regional mutational bias, plays a major role in shaping intron content.
Patterns of genome evolution can differ significantly between outcrossing organisms and self-fertilizing organisms, such as A. thaliana [20]. Because Arabidopsis has most probably become highly self-fertilizing in the relatively recent past [21] and because insertions and deletions occur on a much longer time-scale than base substitutions [22,23], we expect the evolution of intron lengths in Arabidopsis to be dominated by outcrossing reproduction. However, even though heterozygosity is greatly reduced in self-fertilizing organisms so that most gametophytes carry identical alleles, gametophytic competition between transient heterozygotes resulting from de novo mutations still occurs and may well be sufficient to cause the observed reduction of intron lengths in genes expressed in pollen.
Is it conceivable that a slightly different model might apply, one in which speed rather than cost of transcription was important, owing to the fact that only one copy of the genome is present in pollen? If the increased time required to transcribe long introns rather than the energetic cost of transcription [3,6,24] is the primary selective force acting to reduce intron lengths, then it could be argued that the reduced availability of template in the gametophyte, rather than gametophytic selection, could explain intron length reduction in pollen genes. However, because several polymerases can be attached to the same template simultaneously [25,26], gene length need not have much, if any, impact on the steady-state capacity of the template to produce messenger RNA. The additional time required to transcribe longer genes may increase the activation time of a gene but is not expected to have a disproportionately large impact on highly expressed genes. In contrast, the energetic cost of transcription is a linear function of the amount of the transcript that is produced, irrespective of whether transcription is from one or two templates. Because the energetic cost of transcribing longer genes is the same in the gametophyte and sporophyte, we consider that the increased sensitivity to slight differences in fitness caused by strong gametophytic selection is responsible for the reduced length of the introns from genes that are expressed in pollen.
Materials and Methods
Models of A. thaliana genes were extracted from version 5 of the annotated Arabidopsis genome downloaded from TIGR (ftp://ftp.tigr.org/pub/data/a_thaliana/ath1/PSEUDOCHROMOSOMES/). Genes for which more than one gene model was available (corresponding to alternative transcript isoforms) were omitted. SAGE gene expression data derived from pollen [27], seedlings [28], seedling roots [29], root [30], and seedling aerial tissue [16] were downloaded from the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo) or from the source data of the original manuscripts. Only tags that were mapped to a single gene and genes to which only a single tag had been mapped were retained for each dataset. In each of the SAGE datasets, the 20% of genes with the highest tag counts were defined as highly expressed. We used only expression data from tags with counts of at least five for each dataset in order to ensure robust results and that the data were comparable between all of the datasets. All statistical tests were carried out in the R statistical computing environment (http://www.R-project.org).Two-tailed Wilcoxon Rank Sum tests were performed for all of the comparisons between sample means. We used robust regression to fit a linear model to intron lengths as a function of expression level and site of expression (sporophyte or gametophyte) considering all genes expressed in pollen and genes from the SAGE dataset representing the largest number of genes (constructed from the aerial part of the plant [16]). For the linear model, only genes from the sporophyte dataset that were not also present in the gametophyte dataset were considered. Gene expression levels in pollen and a range of sporophyte conditions (root, leaf, stem, hypocotyl, and seedling), estimated using the Affymetrix (Santa Clara, California, United States) ATH1 Arabidopsis Genome Array Gene Chip as part of the Expression Atlas of Arabidopsis Development [17], were obtained prior to publication with the kind permission of the authors. The data are available from the NASCArrays database (http://affymetrix.arabidopsis.info/narrays/experimentbrowse.pl; slide Ids ATGE_73_A/B/C, ATGE_3_A/B/C, ATGE_91_A/B/C, ATGE_28_A2/B2/C2, ATGE_2_A/B/C, ATGE_96_A/B/C). We used the mean value of the signal for each gene that was called present in the original analysis. For each condition, the top 20% of most highly expressed genes were defined as highly expressed.
Supporting Information
Dataset S1 Gene Expression and Intron Length Data
(2.2 MB TXT)
Click here for additional data file.
CS and CG acknowledge the support of the South African National Research Foundation and National Bioinformatics Network.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. CS conceived and designed the experiments and analyzed the data. CS, CG, and LDH suggested analyses and wrote the paper.
Abbreviations
bpbase pair
SAGEserial analysis of gene expression
==== Refs
References
Hurst LD McVean G Moore T 1996 Imprinted genes have few and small introns Nat Genet 12 234 237 8589711
Urrutia AO Hurst LD 2003 The signature of selection mediated by expression on human genes Genome Res 13 2260 2264 12975314
Castillo-Davis CI Mekhedov SL Hartl DL Koonin EV Kondrashov FA 2002 Selection for short introns in highly expressed genes Nat Genet 31 415 418 12134150
Eisenberg E Levanon EY 2003 Human housekeeping genes are compact Trends Genet 19 362 365 12850439
Comeron JM 2004 Selective and mutational patterns associated with gene expression in humans: Influences on synonymous composition and intron presence Genet 167 1293 1304
Ucker D Yamamoto K 1984 Early events in the stimulation of mammary tumor virus RNA synthesis by glucocorticoids. Novel assays of transcription rates J Biol Chem 259 7416 7420 6330056
Vinogradov AE 2004 Compactness of human housekeeping genes: Selection for economy or genomic design? Trends Genet 20 248 253 15109779
Versteeg R van Schaik BD van Batenburg MF Roos M Monajemi R 2003 The human transcriptome map reveals extremes in gene density, intron length, GC content, and repeat pattern for domains of highly and weakly expressed genes Genome Res 13 1998 2004 12915492
Fullerton SM Bernardo Carvalho A Clark AG 2001 Local rates of recombination are positively correlated with GC content in the human genome Mol Biol Evol 18 1139 1142 11371603
Mascarenhas JP 1993 Molecular mechanisms of pollen tube growth and differentiation Plant Cell 5 1303 1314 12271030
Lord EM Russell SD 2002 The mechanisms of pollination and fertilization in plants Annu Rev Cell Dev Biol 18 81 105 12142268
McCormick S 2004 Control of male gametophyte development Plant Cell 16 142 153
Walbot V Evans MM 2003 Unique features of the plant life cycle and their consequences Nat Rev Genet 4 369 379 12728279
Bernasconi G Ashman TL Birkhead TR Bishop JD Grossniklaus U 2004 Evolutionary ecology of the prezygotic stage Science 303 971 975 14963320
Majewski J Ott J 2002 Distribution and characterization of regulatory elements in the human genome Genome Res 12 1827 1836 12466286
Robinson SJ Cram DJ Lewis CT Parkin IA 2004 Maximizing the efficiency of SAGE analysis identifies novel transcripts in Arabidopsis
Plant Physiol 136 3223 3233 15489285
Schmid M Davison TS Henz SR Pape UJ Demar M 2005 A gene expression map of Arabidopsis thaliana development Nat Genet 37 501 506 15806101
Becker JD Boavida LC Carneiro J Haury M Feijo J 2003 A. Transcriptional profiling of Arabidopsis tissues reveals the unique characteristics of the pollen transcriptome Plant Physiol 133 713 725 14500793
Mourier T Jeffares DC 2003 Eukaryotic intron loss Science 300 1393 12775832
Marais G Charlesworth B Wright SI 2004 Recombination and base composition: The case of the highly self-fertilizing plant Arabidopsis thaliana
Genome Biol 5 R45 15239830
Charlesworth D Vekemans X 2005 How and when did Arabidopsis thaliana become highly self-fertilising Bioessays 27 472 476 15832382
Saitou N Ueda S 1994 Evolutionary rates of insertion and deletion in noncoding nucleotide sequences of primates Mol Biol Evol 11 504 512 8015443
Ophir R Graur D 1997 Patterns and rates of indel evolution in processed pseudogenes from humans and murids Gene 205 191 202 9461394
Izban MG Luse DS 1991 Transcription on nucleosomal templates by RNA polymerase-Ii in vitro—Inhibition of elongation with enhancement of sequence-specific pausing Genes Dev 5 683 696 2010092
Hawley DK Roeder RG 1987 Functional steps in transcription initiation and reinitiation from the major late promoter in a HeLa nuclear extract J Biol Chem 262 3452 3461 2434502
Femino AM Fay FS Fogarty K Singer RH 1998 Visualization of single RNA transcripts in situ Science 280 585 590 9554849
Lee JY Lee D 2003 H. Use of serial analysis of gene expression technology to reveal changes in gene expression in Arabidopsis pollen undergoing cold stress Plant Physiol 132 517 529 12805584
Du Z Scott AD May GD 2003 Amplification of high-quantity serial analysis of gene expression ditags and improvement of concatamer cloning efficiency Biotech 35 70 72
Ekman DR Lorenz WW Przybyla AE Wolfe NL Dean JF 2003 SAGE analysis of transcriptome responses in Arabidopsis roots exposed to 2,4,6-trinitrotoluene Plant Physiol 133 1397 1406 14551330
Munos S Cazettes C Fizames C Gaymard F Tillard P 2004 Transcript profiling in the chl1–5 mutant of Arabidopsis reveals a role of the nitrate transporter NRT1 in the regulation of another nitrate transporter, NRT2 Plant Cell 16 2433 2447 15319483
|
16110339
|
PMC1186733
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Aug 5; 1(2):e13
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010013
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1611034010.1371/journal.pgen.001001605-PLGE-RA-0030R2plge-01-02-02Research ArticleBioinformatics - Computational BiologyDiabetes - Endocrinology - MetabolismMus (Mouse)Glucocorticoid Receptor-Dependent Gene Regulatory Networks Glucocorticoid Receptor NetworksPhuc Le Phillip 1Friedman Joshua R 12Schug Jonathan 3Brestelli John E 1Parker J. Brandon 1Bochkis Irina M 1Kaestner Klaus H 1*1 Department of Genetics, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania, United States of America
2 Department of Pediatrics, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania, United States of America
3 Center for Bioinformatics, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania, United States of America
Gibson Greg EditorNorth Carolina State University, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 5 8 2005 1 2 e1623 2 2005 16 6 2005 Copyright: © 2005 Le 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.While the molecular mechanisms of glucocorticoid regulation of transcription have been studied in detail, the global networks regulated by the glucocorticoid receptor (GR) remain unknown. To address this question, we performed an orthogonal analysis to identify direct targets of the GR. First, we analyzed the expression profile of mouse livers in the presence or absence of exogenous glucocorticoid, resulting in over 1,300 differentially expressed genes. We then executed genome-wide location analysis on chromatin from the same livers, identifying more than 300 promoters that are bound by the GR. Intersecting the two lists yielded 53 genes whose expression is functionally dependent upon the ligand-bound GR. Further network and sequence analysis of the functional targets enabled us to suggest interactions between the GR and other transcription factors at specific target genes. Together, our results further our understanding of the GR and its targets, and provide the basis for more targeted glucocorticoid therapies.
Synopsis
Glucocorticoids are essential steroid hormones, and synthetic glucocorticoids are widely prescribed for a variety of medical conditions. Understanding the mechanism by which glucocorticoids act requires knowing the direct target genes whose expression levels are modulated by the glucocorticoid signaling pathway. In this publication, Le and colleagues have utilized two high-throughput techniques to determine genes directly regulated in vivo by the glucocorticoid receptor (GR). RNA and chromatin were extracted from the livers of mice injected with the synthetic glucocorticoid dexamethasone and compared to control littermates. The analysis of RNA expression levels generated a list of genes differentially expressed after addition of dexamethasone. The analysis of the chromatin produced a list of gene promoter sequences where the GR was bound to DNA. By intersecting the two lists, the researchers obtained a list of genes that are directly controlled by the GR, including several previously known targets. This list of direct targets was then used as the basis for complex pathways and sequence analyses, which suggested several interactions between the GR and other transcription factors. This study provides an evaluation of a medically important signaling pathway and serves as a model for future analyses of transcriptional regulation.
Citation:Le PP, Friedman JR, Schug J, Brestelli JE, Parker JB, et al. (2005) Glucocorticoid receptor-dependent gene regulatory networks. PLoS Genet 1(2): e16.
==== Body
Introduction
Glucocorticoids are essential steroid hormones that are secreted by the adrenal cortex and affect multiple organ systems. Among these effects are the ability to depress the immune system, repress inflammation, and help mobilize glucose in the fasting state. Glucocorticoids and their synthetic analogs are widely prescribed for adrenocortical insufficiency and as an immune suppressant/anti-inflammatory agent, but their systemic effects can often be debilitating. An understanding of the genes regulated by the glucocorticoid signaling pathway may lead to more targeted therapies, thereby preventing unwanted side effects.
Glucocorticoids act via a signaling pathway that involves the glucocorticoid receptor (GR), a member of the nuclear receptor superfamily of ligand-activated transcription factors [1,2]. In the absence of glucocorticoids, the GR is sequestered in the cytoplasm by a protein complex that includes heat shock protein 70 (HSP70) and HSP90. When glucocorticoids are present, they traverse the plasma membrane and bind to the GR, allowing the GR to dissociate from its chaperone proteins and translocate to the nucleus. Within the nucleus, the ligand-bound GR can bind to DNA as a monomer or as a dimer to palindromic glucocorticoid response elements (GREs) and modulate transcription [3–7].
The mechanisms of action of the ligand-bound GR are fairly complex, including the ability to both activate and repress transcription, and to interact with other transcriptional regulators such as activating protein-1 (AP-1) and nuclear factor kappa B (NF-κB) (reviewed in McKay and Cidlowski [5]). The net effect of glucocorticoid administration on a particular target gene is likely dependent upon the other transcription factors present on the target gene's promoter or enhancer(s). Specifically, the integration of multiple signaling pathways can occur at glucocorticoid response units (GRUs), which consist of a combination of a GRE and other transcription factor binding sites. Examples of these include GRUs in the promoters of the phosphoenolpyruvate carboxykinase
(Pck) and carbamoylphosphate synthetase
(Cps) genes [8–10]. Thus, understanding the complete nature of glucocorticoid action requires knowing not only the set of genes bound and regulated by the GR, but also the transcription factors that may interact with the GR, and the loci where these interactions occur.
To better understand glucocorticoid signaling, RNA expression profiling after glucocorticoid administration has been performed by several groups [11–16]. However, it is impossible to establish which differentially expressed genes are direct targets of the GR and which are controlled by downstream effectors. To address this limitation, Wang and colleagues have developed a technique termed “ChIP scanning,” which involves screening the promoter region of each putative target gene individually using chromatin immunoprecipitation (ChIP) and quantitative real-time PCR (QPCR) [17]. Unfortunately, this technique is not high-throughput, but instead involves designing multiple primer sets for each potential target gene individually. Thus, this technique is not suitable for global network analysis.
A modification of microarray technology utilizes spotted promoter regions instead of cDNA sequences. After ChIP with an antiserum raised against a particular transcription factor, the immunoprecipitated DNA is amplified, labeled, and hybridized against these promoter microarrays. Spots that are brighter in the immunoprecipitated channel than the control represent promoter sequences to which the transcription factor is bound. This technique, termed “genome-wide location analysis,” or “ChIP-on-Chip” has been utilized to determine binding sites for several transcription factors in yeast [18,19], and higher organisms [20–25]. However, binding data alone do not prove that the transcription factor of interest is important in the regulation of a particular target gene. This is especially true in higher eukaryotes, where many genes are regulated by dozens of partly redundant transcription factors [26]. Therefore, in order to identify direct targets of a given transcription factor that are dependent on the regulatory protein in question, a combination of location analysis and expression profiling is required.
Given the importance of the glucocorticoid signaling pathway in biology and medicine, we have undertaken a study to determine functional GR targets. We performed parallel mRNA expression profiling and location analysis on livers of fasted mice injected with glucocorticoids and compared the results with those of fed controls. Individually, the expression analysis and the location analysis each produced lists of genes that included many known or suspected GR targets. By combining the expression and binding data, we were able to identify 53 direct functional GR targets, many of which are novel. In addition, network and sequence analysis of the GR targets independently suggested functional interactions between the GR and several other transcription factors. Through these experiments we have extended the understanding of the complexity of the genetic networks modulated by glucocorticoids.
Results
Identification of GR Targets by Expression Profiling
The experimental paradigm we chose to use for parallel expression profiling and location analysis is outlined in Figure 1. Treated mice were fasted overnight and injected with dexamethasone, a synthetic glucocorticoid. Control mice were fed ad libitum and not injected with vehicle, as this would generate a stress response. Treated and control mice were sacrificed, and the left lobe of their livers was removed for parallel expression and genome-wide location analysis.
Figure 1 Experimental Paradigm for Orthogonal Analysis of Glucocorticoid Receptor Targets
Treatment mice were fasted overnight, then given intraperitoneal injections with dexamethasone. Treatment and control liver lobes were split in half and processed for both RNA and chromatin. RNA was subjected to microarray analysis using the PancChip 5.0 cDNA microarray, which contains over 13,000 transcripts. Location analysis was performed by immunoprecipitating with antiserum raised against the GR. ChIP material was amplified, fluorescently labeled, and hybridized against sheared genomic DNA using the Mouse PromoterChip BCBC-3.0 promoter microarray, which contains approximately 7,000 genomic promoter elements.
Several alternative experimental designs were initially considered and ultimately discarded. First, we opted to evaluate the networks controlled by glucocorticoids in hepatocytes in mice and not in hepatoma cell lines grown in culture. While this approach was technically more challenging, it was necessary because available cell lines do not reflect the metabolic regulation of hepatic gene expression accurately [27,28]. Second, it was important to compare mice in different feeding states due to the significant differences in the levels of insulin, glucagon, and glucocorticoids that are present in each state. One alternative design would have been to compare dexamethasone-injected fed mice to fed controls. When we performed a preliminary expression analysis using this design, the levels of many well-known GR targets, such as Pck and insulin-like growth factor binding protein 1
(Igfbp1), were unchanged (unpublished data). These results are consistent with the well-described ability of insulin signaling to inhibit glucocorticoid activation of many targets, such as Pck, Igfbp1,
glucose-6-phosphatase (G6pc), and 6-phospho-2-fructokinase (Pfk2) (reviewed in [29]). In addition to the presence of the inhibitory effects of insulin, an experimental design in which both groups are fed risks a significant loss of sensitivity due to a lack of fasting-induced signals that act synergistically with glucocorticoids, such as glucagon. Another alternative design would have been to compare mice fasted and injected with dexamethasone to fasted control mice. However, since endogenous glucocorticoids are released in the fasted state, the chromatin from the control samples could not serve as negative controls for the location analysis. Therefore, we elected to compare dexamethasone-injected fasted mice to fed controls.
Treated and control liver lobes were split in half. RNA was extracted from one half, while chromatin was prepared from the remaining half. Reverse-transcription QPCR (RT-QPCR) was performed to measure relative expression levels of several targets known to be induced by glucocorticoids in fasted mice. As expected, the mRNA levels of these targets, including Pck, tyrosine aminotransferase (Tat), and Igfbp1, were induced between 6- and 60-fold relative to the levels in the fed control mice (unpublished data).
The RNA was then used to perform a microarray hybridization using the PancChip 5.0 expression microarray [30]. Of the 13,000 transcripts on the array, approximately 1,300 unique genes were differentially expressed between the two conditions at a false discovery rate of 10% (complete dataset available in Datasets S1 and S2). Of those, about 30% were up-regulated in the dexamethasone-injected livers, while the majority was repressed. Again, the list of differentially expressed genes included many known GR targets, including Pck, Igfbp1, and metallothionein 2 (Mt2) (unpublished data). However, it is unclear which of these differentially expressed genes are direct targets of the GR.
Global Analysis of GR Occupancy In Vivo
Chromatin immunoprecipitation, performed using an antiserum raised against the GR, was compared to preimmune rabbit IgG to assess the specificity of the antibody at two known GR targets, Mt2 and Tat (Figure 2A). This antibody was then utilized in immunoprecipitations with chromatin from both the dexamethasone-treated and the fed control samples. GR occupancy on those same two targets, as measured by QPCR, was increased by approximately 4-fold and 13-fold, respectively, confirming the efficiency of the ChIP (Figure 2B). Next, we amplified the immunoprecipitated DNA by two rounds of ligation-mediated PCR, and confirmed that the enrichment of the known GR targets was maintained after amplification (unpublished data). Amplified samples were then fluorescently labeled and hybridized against sheared genomic DNA on the Mouse PromoterChip BCBC-3.0 promoter microarray, which contains PCR amplicons of two promoter regions for over 3,300 genes important in liver function. The first PCR amplicon, called tile 1, is approximately 1 kilobase (kb) in length and is located immediately upstream of the putative transcriptional start site. The second amplicon, tile 2, is approximately 2 kb in length and is located immediately upstream of tile 1. In total, we spotted onto the microarray approximately 3 kb of genomic promoter sequence for each of the 3,300 genes.
Figure 2 ChIP Identifies Known GR Targets in Liver
(A) Agarose gel electrophoresis of PCR products for the known GREs in Mt2 and Tat confirm the specificity of the anti-GR antibody (sc-1002, Santa Cruz) compared to the control preimmune IgG. The PCR product for the genomic locus encoding the 28S ribosomal RNA shows that equal amounts of DNA were loaded into each reaction.
(B) QPCR was used to measure the enrichment of Mt2 and Tat in chromatin immunoprecipitated with the anti-GR antiserum from five dexamethasone-treated samples compared to five fed controls, as described in Materials and Methods. The 28S PCR product was used to normalize the samples.
The location analysis identified 318 promoter regions, representing 302 distinct genes, enriched in the immunoprecipitations from dexamethasone-treated samples. This list of GR targets (Table S1) contains many genes previously shown to be regulated by GR, including Pck, Igfbp1, tumor necrosis factor (Tnf), and hormone-sensitive lipase. The Gene Ontology (GO) Biological Function categories of the GR-bound genes are shown in Figure 3. The GR targets were also assessed for statistical enrichment of GO Biological Function categories. The ten most enriched GO Biological Function categories are shown in Table 1. These categories are dominated by genes important in metabolism, consistent with the function of glucocorticoids in the liver. It is also important to note that within the GO function hierarchy, some genes belong to multiple categories.
Figure 3 Functional Categories of the Genes Generated by Location Analysis
Location analysis was performed using antiserum against the GR. Three treated and five control samples were amplified, fluorescently labeled, and hybridized to the Mouse PromoterChip BCBC-3.0 promoter microarray in a common reference design. Standard statistical methods identified 302 promoter regions significantly enriched in the treated samples compared to the fed controls (see Materials and Methods). The GO level 4 functional category for each gene was retrieved and the top 20 categories are shown. Note that some genes belong to multiple GO categories.
Table 1 Enriched GO Biological Functional Categories
The EASE analysis tool was used to determine enriched GO Biological Functional Categories within the set of 302 GR targets. The top ten categories are shown, along with the EASE score, which represents a p-value corrected for variations in category sizes.
We confirmed several of the targets identified in our location analysis by measuring their enrichment in the original immunoprecipitated DNA using QPCR. Two computational programs, NUBIScan [31] and TESS (http://www.cbil.upenn.edu/tess) were used to identify likely GR binding sites within the spotted promoter regions. QPCR was used to calculate the fold-enrichment of these genomic loci in the unamplified, immunoprecipitated DNA of treated samples compared to controls. The results are shown in Figure 4. Of the 14 samples, 12 (85%) randomly chosen promoters were enriched to the level of significance, confirming the validity of our location analysis. Furthermore, it is possible that the remaining promoters have true GR binding sites, but at loci that do not match the consensus sequence well and thus were not assessed. We also noted that the fold-enrichment measured by QPCR was generally greater than that measured by the promoter microarray. This “compression effect” has been previously described in cDNA and oligonucleotide microarrays [32].
Figure 4 Quantitative PCR Confirms GR Targets Identified by Location Analysis
The enrichment of GR targets identified by location analysis was measured in the original immunoprecipitated material by QPCR. The graph shows the fold-enrichment of five dexamethasone-treated samples compared to five fed controls. Of fourteen randomly selected GR targets, 12 showed statistically significant enrichment. Fold-enrichment and p-values were calculated as described in Materials and Methods. *p < 0.05, **p < 0.01.
To further evaluate our list of GR-bound promoters, we obtained a list of more than 50 previously published GR targets [5], 12 of which contain one or more GRE consensus sites within the sequences that are present on our promoter array. If the location analysis had produced a random set of 302 genes (302/3300 [approximately 9%]), then of these 12 known targets, approximately one gene could be expected to be on our list. However, of these 12, eight (67%) were identified by our location analysis as occupied by the GR in the liver in vivo (p < 2 × 10−6), confirming the usefulness of our approach.
Integrating Expression and Binding Data Results in Functional GR Targets
In order to determine the subset of genes that are direct and meaningful targets (i.e., are changed in their expression level) of the activated GR in the liver, we identified the overlap of the GR-bound and GR-regulated genes (Figure 5). There are approximately 2,500 genes common to both array platforms used. Of these, 498 were differentially expressed and 235 were bound by the GR. Intersecting the two sets resulted in 53 genes that were both differentially expressed and bound by the GR. These represent direct, functional targets of the activated GR, which we hereafter refer to as the differentially expressed, GR bound (DEB) set. All of these genes are listed in Table 2, and include several published GR targets, including alcohol dehydrogenase 1 (Adh1),
Pck, and Igfbp1, as well as likely GR target genes such as catalase (Cat) [33]. A more thorough search of the literature and use of expression data from a previously published experiment in a different tissue [34], indicates that 22 of the 53 genes have been shown to be differentially expressed after the addition of glucocorticoid. Thus, 31 of our target genes appear to be novel GR targets.
Figure 5 Intersection of Expression Data and Location Analysis
Of the genes common to both the cDNA microarray and the promoter microarray, 498 were differentially expressed and 235 were bound by the GR. Intersecting the two lists produced 53 genes in common. This list represents direct, functional targets of the GR in hepatocytes.
Table 2 DEB Set
The set of genes differentially expressed between treatment and control mice was intersected with the set of genes generated by the location analysis, resulting in this list of 53 genes. These genes represent direct, functional targets of the GR in mouse hepatocytes.
Network and Sequence Analysis Suggest Several Potential GR-Protein Interactions and Sites
Pathway analysis can be a useful tool to help uncover relationships among genes [35], such as the finding that multiple members of the list may be regulated by the same transcription factor. When seeded with the genes in the DEB set plus the GR itself, pathway analysis produced two networks of genes with functional relationships. These were merged together and are shown in Figure 6. Our data suggest a direct interaction between the GR (NR3C1, large red box; Figure 6) with the DEB gene set (genes in boldface).
Figure 6 A Regulatory Network for the GR
Pathway analysis was seeded with the 53 differentially expressed and GR-bound genes, plus the GR itself, as described in Materials and Methods. Genes in colored, bold text were in the seed set, while all others were brought into the network by the pathway analysis program based on their known relationships to the genes in the seed set. Color indicates induction (red) or repression (green) of expression.
Closer inspection revealed that many of the genes added to the network encode transcription factors that are known to physically interact with the GR on the promoters of particular genes. Two well-studied examples include RelA and Jun, which are members of the NF-κB and AP-1 transcriptional complexes, respectively. It has been suggested that the activated GR modulates inflammation and the immune response by physically interacting with NF-κB and AP-1, thereby repressing the activation of many of their targets [5,7]. In fact, recent work has shown that the LIM protein TRIP6 is required for this interaction [36]. Other transcription factors identified by the pathways analysis that are known to interact with the GR to modulate particular genes include Stat5 [37,38], FoxA family members [39,40], HNF6 [41], PPARγ [42], and Ets family members [43,44]. This suggests that the other transcription factors in the network, such as Myc and HNF4α, may also interact with the GR to co-regulate target genes. Thus, by combining a database of genetic relationships with a set of direct GR targets, we are able to suggest interactions between the GR and other transcription factors.
Next, we analyzed the sequences of the promoters in the DEB set, searching for enriched transcription factor binding sites. First, we scored the set of vertebrate transcription factor position-specific weight matrices (PWMs) in the TRANSFAC database [45] and in JASPAR [46] against the entire set of tiles on the promoter array. We then determined the ability of individual PWMs (representing single transcription factor binding sites) to distinguish between the DEB set and the background set. The GR PWM was enriched in the DEB set compared to the entire set of promoters on the microarray (p < 0.05), as well as compared to the promoters of the 445 genes differentially expressed but not bound by the GR (p < 0.01). For instance, at the particular score threshold, 66% of the DEB set contained a sequence surpassing the threshold, compared to only 48% of the 445 genes differentially expressed but not bound by GR.
Interestingly, the GR weight matrices (full-site and half-site) were not the most significantly enriched matrices. Among the remaining matrices representing vertebrate transcription factors, the scores of the matrices for HNF4α and the GATA family of transcription factors were more significant than the GR matrices. It is likely that the significance of the half-site GR consensus sequence, AGAACA [47], is hampered by the high frequency with which it occurs in the background set of DNA, while the full-site matrix scores probably suffer from the fact that true GREs can vary significantly from the published GRE consensus. For example, the functional GREs in the promoter region of Pck scored poorly with the consensus full-site matrix. As for the enrichment of other matrices, it may be that in our DEB set the presence of HNF4α binding sites is providing tissue specificity to the glucocorticoid response.
Because it is known that the GR can interact with other transcription factors to modulate gene expression, we searched for significant combinations of binding sites for either GR monomers or dimers and other nearby transcription factors. This analysis resulted in several combinations that were significantly enriched in the DEB set compared to the set of all mouse promoter sequences on the array (Table 3). Once again, the interaction between the GR and the AP-1 transcriptional complex presented itself. Specifically, the combination of a GR monomer (not the dimer) and an AP-1 site occurring within 34 base pairs (bp) was highly enriched in the DEB set (p < 10−7). This is consistent with published reports that the GR monomer, and not the dimer, physically interacts with AP-1 to repress certain AP-1 target genes [48]. Several other combinations discovered using this analysis have previously been shown to occur, such as the interaction between GR and CCAAT/enhancer binding protein beta (C/EBPβ) [39], GR and YY1 [38] , and GR and Oct-1 [49].
Table 3 Significant Combinations of GR Monomer/Dimer and Other Transcription Factors
The promoter sequences from the DEB set (Table 2) were searched for enriched combinations of the GR monomer or dimer binding site and another transcription factor binding site within a short distance, as described in Materials and Methods. Shown is the companion transcription factor, its consensus weight matrix (taken from TRANSFAC), the distance from the GR site, and the calculated p-value (uncorrected for multiple testing).
We were particularly interested in the potential interaction between the GR and the C/EBP family of transcription factors, since we had previously performed location analysis for C/EBPβ in mouse liver [21]. Of the promoters in the DEB set that contained a high-scoring combination of GR and nearby C/EBP binding site, 11 had been analyzed previously for C/EBPβ binding. Strikingly, five (45%) of these genes, namely beta-2 microglobulin (β2m), protein-tyrosine sulfotransferase 2 (Tpst2), HSP 1 beta (Hspcβ), ATP-binding cassette sub-family A member 1 (Abca1), and hydroxysteroid (17-beta) dehydrogenase 12 (Hsd17β12), had shown enrichment in that analysis of C/EBPβ ChIP compared to null controls. In other words, of the 11 promoters that were bound and regulated by the GR, had a computationally predicted C/EBP site near a GR site, and were evaluated in an independent experiment, almost half are true C/EBPβ targets. The remaining promoters might be bound by one of the other C/EBP family members expressed in hepatocytes. In any case, this example validates our approach to computationally identifying transcription factors binding to complex GRUs in vivo.
Discussion
Glucocorticoids are widely used in medical therapy for immunosuppression and as potent anti-inflammatory agents. However, their broad effects on different organ systems often result in debilitating side effects such as bone loss and glucose dysregulation. Knowing the direct, functional targets of the GR increases our understanding of the different mechanisms by which glucocorticoids act and aids the development of directed therapies toward different “arms” of the glucocorticoid response. To approach that aim, we have utilized two high-throughput techniques to obtain orthogonal data sets, allowing us to determine the set of genes regulated by the GR in the liver in vivo. We have found hundreds of genes whose promoters are occupied by the ligand-bound GR and, of those, dozens that are differentially expressed in hepatocytes in the presence of exogenous glucocorticoid. These functional GR targets span the known range of glucocorticoid action, containing both induced and repressed genes, as well as genes involved in metabolism, signal transduction, and the immune response/inflammation. By applying pathway and sequence analysis, we have also generated a list of transcription factors that may interact with the GR to modulate transcription of target genes.
Our expression analysis resulted in approximately 1,300 differentially expressed genes. Of these genes, many are differentially expressed solely due to the change in the feeding state of the animal. However, the alternative experimental designs were unacceptable. In our preliminary comparison of RNA from livers of mice that were fed versus fed and dexamethasone-injected, the lack of differential expression of known GR targets such as Pck and Igfbp1 suggested that an orthogonal analysis with this data set would miss many potential targets. Since the goal of the study was not to produce a list of genes differentially expressed after glucocorticoid administration, which has been previously published by others, but rather to perform an orthogonal analysis utilizing both expression and location data from the same tissue, we deemed it acceptable to trade-off specificity for sensitivity. Moreover, by intersecting the expression and location analysis data, it is likely that false positives within the expression data were removed.
The list of 302 GR-bound promoters contains many genes that are known or suspected GR targets. One important category of target genes are transcriptional regulators, including the homeobox genes Hoxa13, Hoxb4, Hoxc5, Hoxc9, and Hesx1;
transcription factor 15 (Tcf15), transcription factor 21 (Tcf21), and transcription factor AP-4 (Tfap4); Krüppel-like factor 3 (Klf3) and Klf15; as well as Trp53, Creb1, and Fos. In addition, many metabolic enzymes are present, consistent with the known role of glucocorticoids in regulating glucose metabolism.
One well-known effect of glucocorticoid administration is decreased bone density [50]. Bone density is controlled by a delicate balance between osteoclasts, which resorb bone, and osteoblasts, which mineralize bone. Our location analysis identified many GR targets that modulate bone remodeling, including TNFα and TGFβ signaling pathway members, adrenomedullin (Adm), calumenin (Calu), and annexin 2 (Anxa2) [51–56]. Of particular interest is the TNF receptor superfamily member osteoprotegerin (OPG), encoded by Tnfrsf11b. OPG is a secreted glycoprotein that acts as a decoy receptor for receptor activator of NF-κB ligand, impairing its interaction with receptor activator of NF-κB and thereby inhibiting osteoclast differentiation [57]. Thus, increased levels of OPG promote increased bone density via decreased osteoclast differentiation, while mice that are null for OPG have decreased total bone density, severe bone porosity, and high incidence of fractures [58]. In humans, it has been shown that a deletion of the OPG gene can cause juvenile Paget's disease, an autosomal recessive osteopathy characterized by debilitating fractures and deformities resulting from increased bone turnover [59]. While it has previously been shown that OPG is expressed in liver [60] and that glucocorticoid administration lowers OPG mRNA levels [61], our location analysis provides the first evidence that glucocorticoids repress OPG expression by directly binding to the OPG promoter. Interestingly, the protein encoded by core binding factor beta (cbfβ), another GR target identified by our analysis, also plays an important role in bone development, possibly via its interaction with core binding factor alpha, a known regulator of OPG [62,63]. Overall, we have demonstrated that our location analysis identified many known and suspected GR targets that directly affect bone remodeling, one of the major side effects associated with prolonged glucocorticoid administration.
While our microarray-based location analysis was highly accurate, with 85% of the targets confirmed by RT-QPCR, it appears somewhat surprising that only about a quarter of the GR-bound genes (53 of 235 for which we have expression data) were differentially expressed. A small fraction of these targets may be enriched due to nonspecific binding of the ligand-bound GR to DNA, or false positives introduced by the ligation-mediated PCR method used to amplify the material prior to hybridization. Another possibility is that the expression level of the gene may not have been high enough to detect in our expression analysis, and that the use of a more sensitive assay, such as RT-QPCR, would show more GR-bound genes to be differentially expressed. We believe, however, that the majority of these GR targets are likely to be functional, but in a different context, such as another tissue. The specific combination of other transcriptional regulators present, as well as higher-order chromatin structure, likely play a role in determining which glucocorticoid target genes are expressed and regulated. The complex regulatory mechanism present in the GRU of the Pck gene is a well-studied example [8]. Insulin signaling represses Pck transcription, and this effect is dominant over the transcriptional activation that is promoted by both glucocorticoid and glucagon signaling pathways. Furthermore, Pck expression is activated by glucocorticoids in the liver, but repressed by glucocorticoids in adipose tissue [64]. Similar mechanisms may be at work on many of the GR-occupied targets that did not show expression level changes in our particular experiment. It may be possible to expand our list of functional GR targets by incorporating other expression data.
The use of weight matrices to predict transcription factor binding sites without any prior knowledge is highly nonspecific, due to the degenerate nature of the binding sites and the complexity of the eukaryotic genome [65]. Some attempts to refine these predictions have utilized evolutionary conservation [66], while others have used functional information such as coexpression [67]. Our analysis results demonstrate that performing complex sequence analysis on a functional set of promoters can result in more specific predictions regarding transcription factor binding sites and can even allow the prediction of binding sites for other transcription factors that cooperatively regulate target genes.
An alternative to location analysis using a promoter microarray is a technique developed by Impey and colleagues called serial analysis of chromatin occupancy (SACO) [68]. It involves ChIP followed by long serial analysis of gene expression. The difference between the two techniques is significant. Promoter microarrays can only detect ChIP enrichment in the region immediately within and around the sequences spotted on the array. In contrast, SACO is unbiased in the binding sites that it can detect, thus allowing investigators to interrogate the whole genome, including introns and intergenic sequences. However, while the development of the promoter microarray has a large initial expense, all future experiments are relatively inexpensive. This contrasts with SACO, in which the sequencing cost must repeatedly be borne for each additional experiment. Thus, to perform SACO on groups of treated and untreated animals, as we have done here, is prohibitively expensive and time-consuming.
Although this work has expanded the set of genes known to be regulated by the GR, the list remains incomplete. Since glucocorticoids exert different effects on the various organ systems of the body, it is likely that the functional targets of the GR are different in each tissue. This would suggest that repeating this experiment on other glucocorticoid-responsive organs would yield new targets. Alternatively, the target list might remain the same, but the intersection between differentially expressed genes and GR-bound promoters might be different. In the future, it would be extremely useful to determine the sets of genes regulated by individual transcription factors versus those regulated by combinations of factors. It may be that those genes controlled by a more complex regulatory network are key genes in the processes in which they participate. This determination could be achieved by performing location analysis with antisera against different transcription factors and then intersecting the lists generated by each analysis, or by performing a single location analysis after sequential immunoprecipitations with different antisera [69].
In summary, we have performed parallel expression analysis and location analysis on the livers of mice treated with dexamethasone in order to determine direct, functional targets of the GR. Our analysis identified many genes previously shown to be GR targets, many genes that were suspected GR targets, and some novel GR target genes. Some of these genes may be critical for particular aspects of the glucocorticoid response, and would therefore make attractive candidates for targeted therapies. Other target genes may be control points where the integration of multiple signaling pathways occurs via GRUs, such as on the Pck promoter. We believe that these targets provide many opportunities for future research, and that this work has moved us one step closer to understanding the complete genetic network modulated by the GR and the set of transcriptional regulators with which it interacts. In addition, this work establishes a paradigm for similar orthogonal analyses, and demonstrates that pathway and sequence analyses can be used to suggest functional interactions between transcription factors on the promoters of particular genes.
Materials and Methods
The complete expression and location analysis data sets are available as Datasets S1 and S2, respectively. The microarray feature location annotation for the two platforms (Mouse PancChip5.0 and Mouse PromoterChip BCBC-3.0) are downloadable from the EpconDB website (http://www.cbil.upenn.edu/EPConDB/Downloads.shtml).
Mouse treatment.
Ten adult male CD1 mice were randomly split into two groups. The control mice were fed ad libitum, while the treatment group was fasted overnight and then given intraperitoneal injections of 0.1 mg/g body weight dexamethasone (Sigma-Aldrich, St. Louis, Missouri, United States) diluted in PBS 3 h prior to sacrifice. Control mice were not injected with vehicle (PBS), as this would have caused the release of endogenous glucocorticoid. Control and treatment animals were housed similarly with standard day/night cycles.
Expression analysis.
RNA was extracted from one half of each liver using TRIZOL (Invitrogen, Carlsbad, California, United States) and lithium chloride precipitation and then analyzed on an Agilent Bioanalyzer 2100 for quality and quantity. All samples showed intact ribosomal bands with a minimum 28S to 18S ratio of 2.0.
For RT-QPCR, 3 μg of total RNA was reverse transcribed using Superscript II (Invitrogen) primed with an oligo-dT primer and then diluted to 300 μL. Each reaction contained 1 μL of diluted cDNA. RT-QPCR was performed in triplicate using a Stratagene MX4000 QPCR machine and Stratagene Brilliant SYBR mix, as per the manufacturer's instructions (Strategene, La Jolla, California, United States). Cycling conditions were 95 °C for 10 min, then 40–45 cycles of 30 s at 95 °C, 1 min at 60 °C, and 30 s at 72 °C, followed by a dissociation curve analysis. Fold enrichments and p-values were calculated as using the Relative Expression Software Tool (REST) [70] with 2,000 randomizations and using the expression of TATA box binding protein (Tbp) as the normalizing gene. Primer sequences are provided in Table S2.
Microarray hybridizations were performed as previously described [30]. A common reference design was utilized, where the common reference was a mixture of all ten samples. This resulted in ten microarray hybridizations. For each sample, 20 μg of total RNA was reverse transcribed using Superscript III (Invitrogen), oligo-dT primers, and amino-allyl dUTP. The RNA was then degraded and the cDNA coupled to fluorescent Cy3 or Cy5. The samples were hybridized to the PancChip 5.0 cDNA microarray, which contains over 13,000 transcripts (http://www.epcondb.org). Slides were scanned on an Agilent (Palo Alto, California, United States) scanner and analyzed with GenePix 5.0 software. Data were normalized using the SMA add-on [71] in the “R” software package [72] and differentially expressed genes were identified using the SAM software package at 10% false discovery rate [73].
Chromatin immunoprecipitation.
Chromatin immunoprecipitations were performed as previously described [21]. Half of each mouse liver sample was minced in cold PBS, pushed through a 21-gauge needle, and then crosslinked using 1% formaldehyde for 10 min. The crosslinking was quenched by adding glycine to a final concentration of 0.125 M. The samples were then washed in PBS and Dounce homogenized in ChIP cell lysis buffer (5 mM Pipes [pH 8.0], 85 mM KCl, 0.5% Nonidet P-40, 10 μM leupeptin, and 1 mM PMSF). Nuclei were sedimented and separated from the cellular debris, then placed into nuclear lysis buffer (50 mM Tris HCl [pH 8.1], 10 mM EDTA, 1% SDS, 10 μM aprotinin, 10 μM leupeptin, and 1 mM PMSF). After 10 min on ice, the lysate was sonicated on ice (Sonic Dismembrator Model 100; Fisher, Pittsburgh, Pennsylvania, United States) using three pulses for 20 s at 4–6 W. Insoluble debris was removed by centrifugation, and the supernatant was collected and flash frozen in liquid nitrogen. An input fraction was generated by taking a portion of nonimmunoprecipitated chromatin material and reversing the crosslinking. When visualized on an agarose gel, the DNA produced a smear of approximately 200–600 bp in length.
For each immunoprecipitation, 2 μg of chromatin was used. The chromatin was precleared by incubating with protein-G agarose at 4 °C for 1 h. Antiserum raised against the GR (sc-1002 Santa Cruz, utilized in ChIP assay by [74]) or pre-immune IgG were added (2 μg each) and the samples rotated at 4 °C overnight. Immunoprecipitates were isolated by incubating with blocked protein-G agarose and washed extensively. Chromatin was eluted from the antibody by incubating for 10 min at room temperature with elution buffer (0.1 M NaHCO3 and 1% SDS). The crosslinking was reversed by adding NaCl to 0.2 M and incubating at 65 °C for at least 4 h. Samples were then digested with 40 ng of proteinase K, and the DNA was isolated via phenol/chloroform extraction followed by ethanol precipitation. DNA concentrations were calculated by measuring A260 on a NanoDrop ND-1000 spectrophotometer (NanoDrop, Wilmington, Delaware, United States). Sheared genomic DNA was generated by sonicating genomic DNA purified from CD1 mouse liver.
QPCR to determine enrichment of genomic loci was performed as for expression analysis, except that the reaction was performed on immunoprecipitated DNA. All occupancy calculations are the average of five samples in both the treated and fed control groups, with all QPCR reactions performed in triplicate. The fold-enrichment and p-value calculations were performed with the REST software package, with 2,000 randomizations. As our normalizing sequence, we used the loci encoding the 28S ribosomal RNA, a sequence of DNA present in multiple copies in the mouse genome. If no product was detected in a control sample in two of three triplicate wells after 45 cycles of PCR, 45 was used as the threshold cycle crossing point. Primer sequences are provided in Table S2.
Location analysis.
For location analysis, a common reference design was utilized, where the common reference was sheared genomic DNA. The three immunoprecipitated treatment samples showing the largest enrichment for the known GR targets in the promoters/enhancers of the Tat and Mt2 genes were used for location analysis and were compared to immunoprecipitations from five fed control samples. This resulted in a total of eight microarray hybridizations. Approximately one half of the material from an immunoprecipitation was amplified using an LM-PCR protocol modified from that developed by Oberley and colleagues [75]. The DNA was blunt-ended in a 50 μl reaction using 5 U of T4 DNA Polymerase (Promega, Madison, Wisconsin, United States), 1× T4 DNA polymerase buffer, and 400 μM dNTPs. The samples were incubated at 11 °C for 15 min, then purified using MinElute columns (Qiagen, Valencia, California, United States) per the manufacturer's protocol. The oligonucleotides OJW102 and OJW103 were annealed to produce a directional linker, which was blunt-ligated to the blunt-ended ChIP DNA in a 50 μl reaction containing 1 μl of high-concentration T4 DNA ligase (New England Biolabs, Beverly, Massachusetts, United States), 1× T4 Ligase buffer, and 0.9 μM annealed OJW102/OJW103. The reaction was incubated at 16 °C overnight and then purified using QiaQuick columns (Qiagen). The samples were PCR amplified in a 50 μl PCR reaction containing 5 U Taq DNA Polymerase (Promega), 1× Taq Polymerase buffer, 2 mM MgCl2, 400 μM dNTPs, and 1 μM OJW102 at the following cycling parameters: 1 cycle at 55 °C for 2 min and 72 °C for 5 min; then 15 cycles of 95 °C for 30 s, 55 °C for 30 s, and 72 °C for 1 min; then a final extension at 72 °C for 5 min. This first round of LM-PCR was purified using QiaQuick columns and then quantified using spectrophotometry. The second round of LM-PCR was performed with 100 ng of the first round product and identical PCR settings except that only ten amplification cycles were performed. This produced 3–4 μg of DNA.
Amplified material (1 μg) was labeled and hybridized against 1 μg of sheared genomic DNA. Samples were labeled using Ready-To-Go DNA labeling beads (Amersham, Little Chalfont, United Kingdom) per manufacturer's instructions. Briefly, water was added to the sample to a volume of 45 μl. The DNA was denatured at 95 °C for 3 min, then cooled on ice. Next, 5 μl of dCTP coupled to Cy3 or Cy5 dye (Amersham) was added, along with the Ready-To-Go labeling bead. This was gently mixed and incubated at 37 °C for 30 min. The Cy3 and Cy5 samples were combined and purified using MinElute columns (Qiagen). After purification, 500 ng of Cot1 Mouse DNA was added to each sample and denatured at 95 °C for 5 min. The samples were then cooled to 42 °C and an equal volume of 2× hybridization buffer (50% formamide, 10× SSC, and 0.2% SDS) was added, mixed, and applied to the Mouse PromoterChip BCBC-3.0 microarray slide.
Microarray slides were hybridized overnight, then washed and scanned with Agilent G2565BA Microarray Scanner. Images were analyzed with GenePix 5.0 software (Axon Instruments). Median background subtracted intensities were obtained for each spot and imported into the mathematical software package “R.” M and A values were calculated as log2(sample/control) and (log2 [sample] + log2 [control])/2, respectively. M values were Lowess-normalized with the SMA package developed by Speed and colleagues [71], and M values for the duplicate spots present on the array were averaged whenever both spots were present.
We used two criteria to determine whether the binding data indicated that a particular promoter region was enriched in the GR-immunoprecipitated material from the treatment samples. We first used SAM to determine the promoter regions differentially enriched between the treatment and control samples. SAM produced a list of differentially bound promoters, the majority of which were enriched in the treatment samples. Then we required that the dexamethasone-injected samples had to show enrichment compared to the sheared, unenriched genomic DNA by utilizing an average M value cutoff of M > 0.
Statistical significance.
The probability of obtaining eight or more GR targets from the list of 12 known targets present on the promoter array was calculated using the hypergeometric distribution. The family of cyclin-dependent kinases and cytochrome P450 enzymes were not considered, because the source did not list individual family members.
EASE functional category analysis.
The Expression Analysis Systematic Explorer (EASE) annotation tool provided by the National Center for Biotechnology Information (NCBI; http://apps1.niaid.nih.gov/david/) was utilized to determine GO Biological Function categories that are enriched in the set of 302 GR target genes compared to the entire set of genes on the promoter microarray. Also shown is a metric known as the “EASE score.” This is calculated as the upper bound of the distribution of jackknife Fisher exact probabilities. This score is a conservative adjustment of p-values generated by the Fisher exact test that penalizes the significance of categories supported by few genes and negligibly penalizes categories supported by many genes.
Pathways analysis.
The network was generated using Ingenuity Pathways Analysis (Ingenuity Systems, http://www.ingenuity.com). The NCBI reference sequence (RefSeq) identifier for the 53 DEB genes, plus the GR itself, was utilized to seed the network generation algorithm. The algorithm produced only two networks with more than four genes, which were then merged into a single network.
Sequence analysis.
The set of vertebrate transcription factor PWMs in TRANSFAC v7.3 and JASPAR was scored against the promoters in the DEB set and a background set of 1,120 other promoters randomly selected from those present on the Mouse PromoterChip BCBC-3.0 promoter microarray. The background set of promoters contained the same proportion of 1-kb tiles (tile 1) and 2-kb tiles (tile 2) as the DEB set. The results are the average of three runs with different background sets. The ability of each PWM, or combination of PWMs, to differentiate between the DEB and the background sets was calculated by measuring the area under the curve in a plot of sensitivity versus 1 − specificity (false positive rate) at different score thresholds. In the case of a combination of PWMs, the threshold scores of both PWMs, as well as the distance between the two, were varied to generate the curve. The optimal spacing and scoring were determined by choosing the parameter values that yielded the largest positive difference between the sensitivity and the false positive rate. The spacing reported is the distance between the outer ends of the binding sites. Values of p were calculated using the one-sided two-sample proportion test (“prop.text” function in the software package “R”) to determine the significance of the difference between the true positive and false positive rates. For the analysis of all combinations of GR and other matrices, we corrected for multiple testing by using a Bonferroni correction, setting the threshold for significance to p < 4 × 10−5 (0.05/1,145 combinations).
Supporting Information
Table S1 Location Analysis Results
Immunoprecipitations with antiserum against GR were performed on chromatin from livers of dexamethasone-treated mice and fed controls. The samples were amplified using LM-PCR and then hybridized to the Mu7K promoter microarray, using sheared genomic DNA as common control. Location analysis resulted in 318 promoters representing 302 genes that were differentially bound by the activated GR. We have listed the RefSeq identifier, a short description of the gene, and the increase in occupancy as measured by the microarray. Some genes are listed more than once, due to both tile 1 and tile 2 showing enrichment. This indicates either multiple GR binding sites or a single GR binding site near the overlap of the two tiles.
(119 KB DOC)
Click here for additional data file.
Table S2 Primer Sequences
Primers were designed to amplify putative GREs within the promoter regions enriched in the location analysis. These were utilized in QPCR reactions as described in Materials and Methods.
(40 KB DOC)
Click here for additional data file.
Dataset S1 Complete Expression Data
This table contains the annotated expression data for the five treatment (D6–D10) and control (F6–F10) samples. Hybridizations were carried out with a common reference design, where the common reference was a mixture of all samples. Values are expressed as Log (red/green), where the common reference sample was labeled in red and the test samples were labeled in green. Also included are fold-change calculations and a column indicating whether SAM software indicated that the spot was differentially expressed.
(10.5 MB XLS)
Click here for additional data file.
Dataset S2 Complete Location Analysis Data
This table contains the annotated location analysis data for the three treatment (D8–D10) and control (F6–F10) samples used in our calculations. Hybridizations were carried out with a common reference design, where the common reference was sheared genomic DNA. Values are expressed as Log (red/green), where the common reference sample was labeled in red and the test samples were labeled in green.
(2.3 MB XLS)
Click here for additional data file.
The authors wish to thank Rana Gupta for critical reading of the manuscript and James Fulmer for care of the animals. This work was supported by grants DK49210 and DK56947 to KHK from the National Institute of Diabetes and Digestive and Kidney Diseases.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. PPL, JRF, and KHK conceived and designed the experiments. PPL, JRF, JEB, JBP, and IMB performed the experiments. PPL, JS, JEB, and IMB analyzed the data. JS contributed reagents/materials/analysis tools. PPL, JRF, and KHK wrote the paper.
Abbreviations
AP-1activating protein-1
ChIPchromatin immunoprecipitation
DEBdifferentially expressed, GR bound
EASEExpression Analysis Systematic Explorer
GOGene Ontology Biological Function
GRglucocorticoid receptor
GREglucocorticoid response element
GRUglucocorticoid response unit
HSPheat shock protein
kbkilobase(s)
NCBINational Center for Biotechnology Information
NF-κBnuclear factor kappa B
OPGosteoprotegerin
PWMposition-specific weight matrix
QPCRquantitative real-time PCR
RefSeqNCBI reference sequence
RT-QPCRreverse-transcription quantitative real-time PCR
SACOserial analysis of chromatin occupancy
==== Refs
References
Yamamoto KR 1985 Steroid receptor regulated transcription of specific genes and gene networks Annu Rev Genet 19 209 252 3909942
Beato M Herrlich P Schutz G 1995 Steroid hormone receptors: Many actors in search of a plot Cell 83 851 857 8521509
Yudt MR Cidlowski JA 2002 The glucocorticoid receptor: Coding a diversity of proteins and responses through a single gene Mol Endocrinol 16 1719 1726 12145329
Karin M 1998 New twists in gene regulation by glucocorticoid receptor: Is DNA binding dispensable? Cell 93 487 490 9604923
McKay LI Cidlowski JA 1999 Molecular control of immune/inflammatory responses: Iinteractions between nuclear factor-kappa B and steroid receptor-signaling pathways Endocr Rev 20 435 459 10453354
Saklatvala J 2002 Glucocorticoids: Do we know how they work? Arthritis Res 4 146 150 12010562
De Bosscher K Vanden Berghe W Haegeman G 2003 The interplay between the glucocorticoid receptor and nuclear factor-kappaB or activator protein-1: Molecular mechanisms for gene repression Endocr Rev 24 488 522 12920152
Hanson RW Reshef L 1997 Regulation of phosphoenolpyruvate carboxykinase (GTP) gene expression Annu Rev Biochem 66 581 611 9242918
Stafford JM Wilkinson JC Beechem JM Granner DK 2001 Accessory factors facilitate the binding of glucocorticoid receptor to the phosphoenolpyruvate carboxykinase gene promoter J Biol Chem 276 39885 39891 11518712
Schoneveld OJ Gaemers IC Lamers WH 2004 Mechanisms of glucocorticoid signalling Biochim Biophys Acta 1680 114 128 15488991
Rogatsky I Wang JC Derynck MK Nonaka DF Khodabakhsh DB 2003 Target-specific utilization of transcriptional regulatory surfaces by the glucocorticoid receptor Proc Natl Acad Sci U S A 100 13845 13850 14617768
Bladh LG Liden J Dahlman-Wright K Reimers M Nilsson S 2004 Identification of endogenous glucocorticoid repressed genes differentially regulated by a glucocorticoid receptor mutant able to separate between NF-kappa B and AP-1 repression Mol Pharmacol
Galon J Franchimont D Hiroi N Frey G Boettner A 2002 Gene profiling reveals unknown enhancing and suppressive actions of glucocorticoids on immune cells FASEB J 16 61 71 11772937
Webster JC Huber RM Hanson RL Collier PM Haws TF 2002 Dexamethasone and tumor necrosis factor-alpha act together to induce the cellular inhibitor of apoptosis-2 gene and prevent apoptosis in a variety of cell types Endocrinology 143 3866 3874 12239098
Ishibashi T Takagi Y Mori K Naruse S Nishino H 2002 cDNA microarray analysis of gene expression changes induced by dexamethasone in cultured human trabecular meshwork cells Invest Ophthalmol Vis Sci 43 3691 3697 12454038
Wu W Chaudhuri S Brickley DR Pang D Karrison T 2004 Microarray analysis reveals glucocorticoid-regulated survival genes that are associated with inhibition of apoptosis in breast epithelial cells Cancer Res 64 1757 1764 14996737
Wang JC Derynck MK Nonaka DF Khodabakhsh DB Haqq C 2004 Chromatin immunoprecipitation (ChIP) scanning identifies primary glucocorticoid receptor target genes Proc Natl Acad Sci U S A 101 15603 15608 15501915
Ren B Robert F Wyrick JJ Aparicio O Jennings EG 2000 Genome-wide location and function of DNA binding proteins Science 290 2306 2309 11125145
Lee TI Rinaldi NJ Robert F Odom DT Bar-Joseph Z 2002 Transcriptional regulatory networks in Saccharomyces cerevisiae
Science 298 799 804 12399584
Odom DT Zizlsperger N Gordon DB Bell GW Rinaldi NJ 2004 Control of pancreas and liver gene expression by HNF transcription factors Science 303 1378 1381 14988562
Friedman JR Larris B Le PP Peiris TH Arsenlis A 2004 Orthogonal analysis of C/EBP beta targets in vivo during liver proliferation Proc Natl Acad Sci U S A 101 12986 12991 15317935
Martone R Euskirchen G Bertone P Hartman S Royce TE 2003 Distribution of NF-kappaB-binding sites across human chromosome 22 Proc Natl Acad Sci U S A 100 12247 12252 14527995
Kirmizis A Farnham PJ 2004 Genomic approaches that aid in the identification of transcription factor target genes Exp Biol Med (Maywood) 229 705 721 15337825
Wells J Yan PS Cechvala M Huang T Farnham PJ 2003 Identification of novel pRb binding sites using CpG microarrays suggests that E2F recruits pRb to specific genomic sites during S phase Oncogene 22 1445 1460 12629508
Mao DY Watson JD Yan PS Barsyte-Lovejoy D Khosravi F 2003 Analysis of Myc bound loci identified by CpG island arrays shows that Max is essential for Myc-dependent repression Curr Biol 13 882 886 12747840
Kel OV Romaschenko AG Kel AE Wingender E Kolchanov NA 1995 A compilation of composite regulatory elements affecting gene transcription in vertebrates Nucleic Acids Res 23 4097 4103 7479071
Nyberg SL Remmel RP Mann HJ Peshwa MV Hu WS 1994 Primary hepatocytes outperform Hep G2 cells as the source of biotransformation functions in a bioartificial liver Ann Surg 220 59 67 8024360
Jiang H Lucy MC 2001 Involvement of hepatocyte nuclear factor-4 in the expression of the growth hormone receptor 1A messenger ribonucleic acid in bovine liver Mol Endocrinol 15 1023 1034 11376119
Pierreux CE Rousseau GG Lemaigre FP 1999 Insulin inhibition of glucocorticoid-stimulated gene transcription: Requirement for an insulin response element? Mol Cell Endocrinol 147 1 5 10195686
Kaestner KH Lee CS Scearce LM Brestelli JE Arsenlis A 2003 Transcriptional program of the endocrine pancreas in mice and humans Diabetes 52 1604 1610 12829622
Podvinec M Kaufmann MR Handschin C Meyer UA 2002 NUBIScan, an in silico approach for prediction of nuclear receptor response elements Mol Endocrinol 16 1269 1279 12040014
Yuen T Wurmbach E Pfeffer RL Ebersole BJ Sealfon SC 2002 Accuracy and calibration of commercial oligonucleotide and custom cDNA microarrays Nucleic Acids Res 30 e48 12000853
Grier DG Halliday HL 2004 Effects of glucocorticoids on fetal and neonatal lung development Treat Respir Med 3 295 306 15606220
Clerch LB Baras AS Massaro GD Hoffman EP Massaro D 2004 DNA microarray analysis of neonatal mouse lung connects regulation of KDR with dexamethasone-induced inhibition of alveolar formation Am J Physiol Lung Cell Mol Physiol 286 L411 419 14607780
White P Brestelli JE Kaestner KH Greenbaum LE 2005 Identification of transcriptional networks during liver regeneration J Biol Chem 280 3715 3722 15546871
Kassel O Schneider S Heilbock C Litfin M Gottlicher M 2004 A nuclear isoform of the focal adhesion LIM-domain protein Trip6 integrates activating and repressing signals at AP-1- and NF-kappa B-regulated promoters Genes Dev 18 2518 2528 15489293
Tronche F Opherk C Moriggl R Kellendonk C Reimann A 2004 Glucocorticoid receptor function in hepatocytes is essential to promote postnatal body growth Genes Dev 18 492 497 15037546
Bergad PL Towle HC Berry SA 2000 Yin-yang 1 and glucocorticoid receptor participate in the Stat5-mediated growth hormone response of the serine protease inhibitor 2.1 gene J Biol Chem 275 8114 8120 10713133
Schoneveld OJ Gaemers IC Das AT Hoogenkamp M Renes J 2004 Structural requirements of the glucocorticoid-response unit of the carbamoyl-phosphate synthase gene Biochem J 382 463 470 15196051
O'Brien RM, Noisin EL, Suwanichkul A, Yamasaki T, Lucas PC, et al. 1995 Hepatic nuclear factor 3- and hormone-regulated expression of the phosphoenolpyruvate carboxykinase and insulin-like growth factor-binding protein 1 genes Mol Cell Biol 15 1747 1758 7532283
Pierreux CE Stafford J Demonte D Scott DK Vandenhaute J 1999 Antiglucocorticoid activity of hepatocyte nuclear factor-6 Proc Natl Acad Sci U S A 96 8961 8966 10430878
Nie M Corbett L Knox AJ Pang L 2005 Differential regulation of chemokine expression by peroxisome proliferator-activated receptor gamma agonists: Interactions with glucocorticoids and beta2-agonists J Biol Chem 280 2550 2561 15531761
Geng CD Vedeckis WV 2004 Steroid-responsive sequences in the human glucocorticoid receptor gene 1A promoter Mol Endocrinol 18 912 924 15044598
Aurrekoetxea-Hernandez K Buetti E 2000 Synergistic action of GA-binding protein and glucocorticoid receptor in transcription from the mouse mammary tumor virus promoter J Virol 74 4988 4998 10799572
Matys V Fricke E Geffers R Gossling E Haubrock M 2003 TRANSFAC: Transcriptional regulation, from patterns to profiles Nucleic Acids Res 31 374 378 12520026
Sandelin A Alkema W Engstrom P Wasserman WW Lenhard B 2004 JASPAR: An open-access database for eukaryotic transcription factor binding profiles Nucleic Acids Res 32 D91 94 14681366
Lefstin JA Yamamoto KR 1998 Allosteric effects of DNA on transcriptional regulators Nature 392 885 888 9582068
Heck S Kullmann M Gast A Ponta H Rahmsdorf HJ 1994 A distinct modulating domain in glucocorticoid receptor monomers in the repression of activity of the transcription factor AP-1 Embo J 13 4087 4095 8076604
Prefontaine GG Lemieux ME Giffin W Schild-Poulter C Pope L 1998 Recruitment of octamer transcription factors to DNA by glucocorticoid receptor Mol Cell Biol 18 3416 3430 9584182
Boling EP 2004 Secondary osteoporosis: Underlying disease and the risk for glucocorticoid-induced osteoporosis Clin Ther 26 1 14 14996513
Kirsch T 2005 Annexins—Their role in cartilage mineralization Front Biosci 10 576 581 15574394
Nakazawa T Nakajima A Seki N Okawa A Kato M 2004 Gene expression of periostin in the early stage of fracture healing detected by cDNA microarray analysis J Orthop Res 22 520 525 15099630
Wan M Cao X 2005 BMP signaling in skeletal development Biochem Biophys Res Commun 328 651 657 15694398
Reddy SV 2004 Regulatory mechanisms operative in osteoclasts Crit Rev Eukaryot Gene Expr 14 255 270 15663356
Nanes MS 2003 Tumor necrosis factor-alpha: Molecular and cellular mechanisms in skeletal pathology Gene 321 1 15 14636987
Cornish J Callon KE Coy DH Jiang NY Xiao L 1997 Adrenomedullin is a potent stimulator of osteoblastic activity in vitro and in vivo Am J Physiol 273 E1113 1120 9435526
Horowitz MC Xi Y Wilson K Kacena MA 2001 Control of osteoclastogenesis and bone resorption by members of the TNF family of receptors and ligands Cytokine Growth Factor Rev 12 9 18 11312114
Bucay N Sarosi I Dunstan CR Morony S Tarpley J 1998 Osteoprotegerin-deficient mice develop early onset osteoporosis and arterial calcification Genes Dev 12 1260 1268 9573043
Whyte MP Obrecht SE Finnegan PM Jones JL Podgornik MN 2002 Osteoprotegerin deficiency and juvenile Paget's disease N Engl J Med 347 175 184 12124406
Yun TJ Chaudhary PM Shu GL Frazer JK Ewings MK 1998 OPG/FDCR-1, a TNF receptor family member, is expressed in lymphoid cells and is up-regulated by ligating CD40 J Immunol 161 6113 6121 9834095
Vidal NO Brandstrom H Jonsson KB Ohlsson C 1998 Osteoprotegerin mRNA is expressed in primary human osteoblast-like cells: Down-regulation by glucocorticoids J Endocrinol 159 191 195 9795357
Thirunavukkarasu K Halladay DL Miles RR Yang X Galvin RJ 2000 The osteoblast-specific transcription factor Cbfa1 contributes to the expression of osteoprotegerin, a potent inhibitor of osteoclast differentiation and function J Biol Chem 275 25163 25172 10833509
Kundu M Javed A Jeon JP Horner A Shum L 2002 Cbfbeta interacts with Runx2 and has a critical role in bone development Nat Genet 32 639 644 12434156
Olswang Y Blum B Cassuto H Cohen H Biberman Y 2003 Glucocorticoids repress transcription of phosphoenolpyruvate carboxykinase (GTP) gene in adipocytes by inhibiting its C/EBP-mediated activation J Biol Chem 278 12929 12936 12560325
Bulyk ML 2003 Computational prediction of transcription-factor binding site locations Genome Biol 5 201 14709165
Tan K Moreno-Hagelsieb G Collado-Vides J Stormo GD 2001 A comparative genomics approach to prediction of new members of regulons Genome Res 11 566 584 11282972
Bussemaker HJ Li H Siggia ED 2001 Regulatory element detection using correlation with expression Nat Genet 27 167 171 11175784
Impey S McCorkle SR Cha-Molstad H Dwyer JM Yochum GS 2004 Defining the CREB regulon: A genome-wide analysis of transcription factor regulatory regions Cell 119 1041 1054 15620361
Geisberg JV Struhl K 2004 Quantitative sequential chromatin immunoprecipitation, a method for analyzing co-occupancy of proteins at genomic regions in vivo Nucleic Acids Res 32 e151 15520460
Pfaffl MW Horgan GW Dempfle L 2002 Relative expression software tool (REST) for group-wise comparison and statistical analysis of relative expression results in real-time PCR Nucleic Acids Res 30 e36 11972351
Yang YH Dudoit S Luu P Lin DM Peng V 2002 Normalization for cDNA microarray data: A robust composite method addressing single and multiple slide systematic variation Nucleic Acids Res 30 e15 11842121
R Development Core Team 2005 R: A language and environment for statistical computing, version 1.9.1 [computer program] Available: http://www.R-project.org . Accessed 30 June 2004.
Tusher VG Tibshirani R Chu G 2001 Significance analysis of microarrays applied to the ionizing radiation response Proc Natl Acad Sci U S A 98 5116 5121 11309499
Pascussi JM Busson-Le Coniat M Maurel P Vilarem MJ 2003 Transcriptional analysis of the orphan nuclear receptor constitutive androstane receptor (NR1I3) gene promoter: Identification of a distal glucocorticoid response element Mol Endocrinol 17 42 55 12511605
Oberley MJ Tsao J Yau P Farnham PJ 2004 High-throughput screening of chromatin immunoprecipitates using CpG-island microarrays Methods Enzymol 376 315 334 14975315
|
16110340
|
PMC1186734
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Aug 5; 1(2):e16
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010016
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1611034110.1371/journal.pgen.001001805-PLGE-RA-0054R3plge-01-02-04Research ArticleBioinformatics - Computational BiologyEvolutionGenetics/GenomicsHumanChimpanzeeGorillaOrangutanOscillating Evolution of a Mammalian Locus with Overlapping Reading Frames: An XLαs/ALEX Relay Oscillating EvolutionNekrutenko Anton 123*Wadhawan Samir 123Goetting-Minesky Paula 24Makova Kateryna D 41 Department of Biochemistry and Molecular Biology, Pennsylvania State University, University Park, Pennsylvania, United States of America
2 Center for Comparative Genomics and Bioinformatics, Pennsylvania State University, University Park, Pennsylvania, United States of America
3 The Huck Institutes for Life Sciences, Pennsylvania State University, University Park, Pennsylvania, United States of America
4 Department of Biology, Pennsylvania State University, University Park, Pennsylvania, United States of America
Gojobori Takashi EditorNational Institute of Genetics, Japan*To whom correspondence should be addressed. E-mail: [email protected] 2005 12 8 2005 1 2 e1825 3 2005 23 6 2005 Copyright: © 2005 Nekrutenko 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.XLαs and ALEX are structurally unrelated mammalian proteins translated from alternative overlapping reading frames of a single transcript. Not only are they encoded by the same locus, but a specific XLαs/ALEX interaction is essential for G-protein signaling in neuroendocrine cells. A disruption of this interaction leads to abnormal human phenotypes, including mental retardation and growth deficiency. The region of overlap between the two reading frames evolves at a remarkable speed: the divergence between human and mouse ALEX polypeptides makes them virtually unalignable. To trace the evolution of this puzzling locus, we sequenced it in apes, Old World monkeys, and a New World monkey. We show that the overlap between the two reading frames and the physical interaction between the two proteins force the locus to evolve in an unprecedented way. Namely, to maintain two overlapping protein-coding regions the locus is forced to have high GC content, which significantly elevates its intrinsic evolutionary rate. However, the two encoded proteins cannot afford to change too quickly relative to each other as this may impair their interaction and lead to severe physiological consequences. As a result XLαs and ALEX evolve in an oscillating fashion constantly balancing the rates of amino acid replacements. This is the first example of a rapidly evolving locus encoding interacting proteins via overlapping reading frames, with a possible link to the origin of species-specific neurological differences.
Synopsis
One of the possible ways to achieve tight co-expression of two proteins is to encode them within a single mRNA. The GNAS1 gene in mammals does just that: it encodes two interacting signaling polypeptides within a single transcript using nested reading frames shifted one nucleotide relative to each other. The exceptionally high GC content of the region where the two reading frames overlap diminishes the probability of encountering stop codons but makes the locus highly mutable. To preserve their ability to interact functionally with each other despite the high mutation rate, the two polypeptides appear to evolve in an oscillating fashion, trying to maintain approximately equal rates of amino acid substitutions. This unexpected observation provides new insights into the evolution of mostly overlooked overlapping coding regions in eukaryotic genomes.
Citation:Nekrutenko A, Wadhawan S, Goetting-Minesky P, Makova KD (2005) Oscillating evolution of a mammalian locus with overlapping reading frames: An XLαs/ALEX relay. PLoS Genet 1(2): e18.
==== Body
Introduction
The GNAS1 locus encodes the stimulatory G-protein subunit α, a key element of the classical signal transduction pathway linking receptor–ligand interactions with the activation of adenylyl cyclase and a variety of cellular responses [1–3]. The gene is subject to complex imprinting, producing a spectrum of maternally, paternally, and biallelically derived transcripts [4]. The major paternally imprinted transcript of the gene is expressed primarily in neuroendocrine tissues and includes an unusually large upstream exon (the XL-exon) comprising over 50% of the protein-coding region. The XL-exon contains two completely overlapping reading frames in the same orientation but shifted one nucleotide relative to each other so that codon positions 1, 2, and 3 of the first frame overlap with positions 3, 1, and 2 of the second frame. In humans the first frame of the exon encodes 388 N-terminal amino acids of a 736-residue extra large form of Gα (XLαs) [5–10]. The second frame encodes all 322 amino acids of alternative gene product encoded by the XL-exon (ALEX) and terminates exactly at the end of the exon. The internal section of the XL exon contains imperfect repeated units of variable length translated into amino acid repeats averaging 13 residues in both XLαs and ALEX [4]. The repeat number varies in a studied human population (n = 276), with the majority carrying a 13-unit allele, while an insertion of an additional unit (the 14-unit allele) is found in 2.2% of surveyed individuals [11]. Heterozygous individuals with a maternally inherited 14-unit allele and 13-unit homozygotes are normal. Conversely, carriers of a paternally inherited 14-unit allele exhibit hyperactivity of the G-protein pathway and suffer from a variety of pathological conditions such as mental retardation, brachymetacarpia, hypertrichosis, hypotonia, growth deficiency, or prolonged trauma-induced bleeding [12]. Binding assays showed a decreased affinity between XLαs and ALEX in individuals carrying the 14-unit allele that leads to an elevated concentration of free XLαs (unbound to ALEX) capable of activating adenylyl cyclase [12]. As a result, the intracellular cAMP concentration rises to over 600% of the normal level. Thus, ALEX regulates the intracellular cAMP level by specifically binding XLαs and preventing it from interacting with the receptors and adenylyl cyclase [12,13]. Loss-of-function mutations involving XLαs also lead to severe adverse effects. Mice lacking XLαs expression show poor postnatal development with the majority dying within 48 h of birth [5].
The functional importance of XLαs and ALEX suggested by these examples implies that this locus should be under considerable selective constraint. Yet the XL-exon evolves at a remarkable pace: the nucleotide identity between human and mouse XL-exon is only 71%, and the amino acid identity between human and mouse ALEX is 53% [13]. For comparison, the average nucleotide and amino acid identities between human and mouse protein-coding genes and their protein products are 86% and 89%, respectively [14]. Why would a locus encoding two essential signaling proteins evolve so rapidly?
Results/Discussion
To take a closer look at the evolutionary dynamics of XLαs and ALEX, we sequenced the XL-exon from eight primates and immediately found striking differences within the repeat-containing region (we used XL-exon boundaries as described in Hayward et al. [4]; also see Methods). All studied species that included human, apes (chimpanzee, gorilla, orangutan, and gibbon), Old World monkeys (colobus and macaque), and a New World monkey (squirrel monkey) varied in the number and/or sequence of repeated units (Figure 1). Human had the smallest number of repeats, while the remaining taxa contained at least one additional repeat unit between positions I and N (Figure 1), a region where an insertion in humans is linked to disease. Taxa closest to human (chimpanzee, gorilla, and orangutan) carried the largest number of repeat units and an additional insertion at position B. Gibbon, colobus, macaque, and squirrel monkey contained an additional insertion at position H. Assuming that the sequenced alleles are fixed in the respective primate populations, XL-exon experienced an episode of repeat expansion in the greater ape lineage followed by a dramatic repeat loss on the branch leading from the human/chimpanzee ancestor to modern humans (Figure 2). Note that in all sampled species both reading frames remain intact regardless of the insertion/deletion events. The observed pattern may have implications for the evolution of species-specific neurological and metabolic differences (discussed below) since the variation in the number of repeats has profound developmental and physiological effects [5,11,12].
Figure 1 Alignment of Internal Repeat Region in XLαs and ALEX Polypeptides
Black boxes highlight the position of the disease-linked repeat in the 14-unit human allele (Hs*). Sequences upstream and downstream of the shown region can be aligned unambiguously. Species abbreviations as follows: Hs, Homo sapiens (human); Pt, Pan troglodytes (chimpanzee); Gg, Gorilla gorilla (gorilla); Pp, Pongo pygmaeus (orangutan); Hl, Hylobates lar (gibbon); Ca, Colobus angolensis (colobus monkey); Mm, Macaca mulatta (macaque); Sb, Saimiri boliviensis (squirrel monkey).
Figure 2 Evolutionary Oscillation between XLαs and ALEX Frames Revealed from Nucleotide Substitution Analysis
The ratio of maximum likelihood estimates of nonsynonymous rates between the two frames (XL
K
A/ALEX
K
A) is shown on each branch. A series of colored bars above each branch shows the number of nucleotide substitutions at each codon position reconstructed using parsimony. Each bar represents a single substitution. The codon positions are numbered as follows: black letters on white background (XLαs frame); white letters on black background (ALEX frame). Boxes at the ends of external branches show repeat structure of the XL-exon in each species (white = deletion). Species abbreviations are as in Figure 1.
Next, we analyzed the pattern of nucleotide substitutions within the XL-exon (excluding the repeat-containing region) and observed a striking oscillation of amino acid replacement rates between the XLαs and ALEX. The interaction between the two proteins imposes a unique constraint: if one protein changes the other needs to rapidly “evolve” a compensatory substitution to preserve the mutual affinity. Although this cannot be observed directly in our data because such changes are likely to occur within each lineage in rapid succession, the overall effect of this process should result in similar rates of amino acid replacements in the two proteins. To test this hypothesis we compared nucleotide substitutions between XLαs and ALEX frames in sequenced species. Classical measures of nucleotide substitution rates such as K
S and K
A [15] are not directly applicable here because of the interdependence of the two overlapping frames [16–18]. However, these measures can be used in a relative context. Specifically, the ratio of nonsynonymous rates between the two frames (XL
K
A/ALEX
K
A) can be used to test the equality of amino acid replacement rates between the two proteins. To carry out this analysis we reconstructed a phylogenetic tree using unambiguously aligning portions of the XL-exon. For every branch of the tree we computed the XL
K
A/ALEX
K
A ratio using maximum likelihood estimates of nonsynonymous rates for each frame (Figure 2). Ratios vary considerably among branches. For example, branches originating from node 3 (3→Pp and 3→2) show opposing XL
K
A/ALEX
K
A ratios. However, none of the ratios is significantly different from 1 (p-values from Fisher's exact test are between 0.14 and 0.77), supporting our hypothesis that the two proteins constantly co-evolve and maintain XL
K
A/ALEX
K
A of approximately 1.
A possible caveat of this analysis is the use of internal nodes because the likelihood method we employed to estimate branch-specific rates was not intended to handle coding sequences with multiple reading frames. To address this, we estimated pairwise K
A between XLαs and ALEX reading frames. For this purpose we developed a neighbor-dependent modification of the Nei–Gojobori (NG) method [19]. Unlike the classical NG, our method estimates the number of synonymous and nonsynonymous changes in a given frame (i.e., XLαs) without considering any pathways that would create nonsense codons in the other frame (i.e., ALEX). The resulting estimates were only slightly different from the NG, Yang-Nielsen [20], and likelihood [21] methods, as the high GC content of XL-exon (68% in human) decreases the chance of encountering pathways that contain nonsense codons (Table 1). We used the new K
A estimates to calculate the XL
K
A/ALEX
K
A ratio for each pair of species. Again, although the ratios varied substantially, none was significantly different from 1 (at 1% level; Table 1). The observed oscillation of the XL
K
A/ALEX
K
A ratio around 1 likely implies constant adjustment between the two proteins aimed at maintaining mutual affinity.
Table 1 Pairwise Synonymous and Nonsynonymous Rates in XLαs and ALEX Protein-Coding Regions
XL
K
A/ALEX
K
A, the ratio of nonsynonymous rates estimated using mNG; YN, Yang-Nielsen method [20]; ML, maximum likelihood method [21]; mNG, modified neighbor-dependent NG method.
The phenomenon of oscillation is also confirmed by the pattern of nucleotide substitutions at different codon positions. Third codon positions of the XLαs frame, where most changes are synonymous, correspond to second codon positions of the ALEX frame where all substitutions lead to amino acid replacements. Similarly, third codon positions of the ALEX frame overlap with first codon positions of the XLαs where most substitutions are nonsynonymous. To visualize the substitution process at the level of codon positions, we used maximum parsimony to reconstruct ancestral sequences at the internal nodes of the tree in Figure 2. We modified the original parsimony algorithm by omitting ancestral states that may create stop codons in either of the two frames. Although ancestral sequences reconstructed using parsimony cannot be used as observed data [22], this analysis once again shows evolutionary fluctuation between the two frames (Figure 2). For example, the majority of substitutions on branches leading to Ca, Mm, and Sb are in the third codon position of the XLαs frame (corresponding to the 0-fold degenerate second codon position of the ALEX frame). This is also the case for the branch leading to Pp, while other branches within the human/ape clade show the opposite pattern—most substitutions are now in mostly 0-fold degenerate first and second codon positions of the XLαs frame. In addition, there are examples of recurrent substitutions leading to the same amino acids in different lineages (Table 2), thus, suggesting that multiple optimal variants of the two proteins are allowed.
Table 2 An Example of Recurrent Substitutions in Human and Apes
Dots indicate residues identical to human. Translations of XLαs and ALEX start with the first and the second nucleotides of shown mRNA segment.
The high GC content of the XL-exon (ranging from 68% in human to 71% in squirrel monkey) is “the blessing and the curse” of the locus: it appears to be required for the maintenance of the two reading frames, but inevitably leads to a high substitution rate. A consequence of the high GC content is the abundance of GC-rich codons in the XLαs and ALEX frames. For instance, the most abundant codons in XLαs and ALEX frames are GCC (10.6%) and CCG (8.9%), respectively (Figure 3). For comparison, average frequencies of these codons in humans (estimated from RefSeq genes) are 2.8% and 0.7%, respectively. The GC content may be driven up by a selection acting against mutations to A and T, as these can lead to the formation of stop codons (TAA, TAG, TGA) in either of the two frames. To test this hypothesis, we simulated the eight sequences in our dataset using three different codon frequency tables compiled from (1) all human RefSeq genes, (2) XLαs reading frame, and (3) ALEX reading frame. All other parameters (phylogenetic tree, branch lengths, transition/transversion ratio, codon number, and the K
A/K
S ratio as estimated from the original dataset) were fixed, and each simulation was performed 1,000,000 times. Each set of simulated sequences was examined for the presence of alternative reading frames. For example, for every set of sequences simulated using XLαs codon frequencies, we looked for the presence of an alternative reading frame in +1 phase. None of the sets from the first simulation (RefSeq codon frequencies) contained such frames, whereas approximately 1% of sets in each of the second (XLαs codon frequencies) and the third (ALEX codon frequencies) simulations contained alternative frames in +1 and −1 positions, respectively. Thus the high GC content allows for overlapping reading frames.
Figure 3 Codon Usage Estimated from Human RefSeq Genes, XLαs Coding Region, and ALEX Coding Region
Green indicates human RefSeq genes, yellow indicates XLαs coding region, and red indicates ALEX coding region.
The high GC content also leads to an excess of CpG dinucleotides, which occupy approximately 20% of XL-exon (108–119 CpG sites or 18%–21% of the sequence length, depending on the species). This is significantly higher than in the majority of primate sequences (empirical p = 0.0013): the proportion of CpG sites in human protein-coding regions from the RefSeq database have narrow distribution with a mean of 7% (99% confidence interval: [7.17%; 7.43%]). In mammals, mutation rate at CpG dinucleotides is 10–20 times higher than at other sites [23–25]. As a result, although CpG sites occupy only approximately 20% of the sequence in our dataset, approximately 50% of the observed nucleotide substitutions (responsible for approximately 30% of amino acid replacements) occur at these sites (Table 3). In this analysis we do not correct for multiple substitutions because existing models cannot be used in the context of XLαs/ALEX locus. Thus, the actual rate of evolution of XL-exon is even higher than observed. Remarkably, the majority of potential deamination events at CpG sites (CpG → CpA and CpG → TpG transitions) do not create stop codons in either of the two reading frames. Indeed, the in silico deamination of all CpG sites (109–118 replacements, depending on the species) to either TpG or CpA created only four stop codons in the XLαs and none in the ALEX frame in each species. In contrast, the simulated deamination caused on average 140 and 129 amino acid changes in XLαs and in ALEX, respectively. Therefore, high GC content leads to the high intrinsic mutability of the XL-exon but allows avoidance of stop codons.
Table 3 Nucleotide and Amino Acid Replacements at CpG Sites
These results suggest the following model of XLαs/ALEX evolution that favors purifying selection acting on the two proteins. The benefit in encoding the two signal transduction proteins within the same mRNA molecule might be the tight expression coupling: it guarantees that the two proteins are made at the same place and at the same time. To maintain two long, overlapping reading frames the XL-exon must contain an excess of GC-rich codons, but this also leads to the elevated frequency of mutation-prone CpG dinucleotides. Because the two proteins physically interact, they must accumulate amino acid substitutions in concert: neither can change too much relative to the other as their mutual affinity may become adversely affected. Therefore, a nonsynonymous mutation causing a deleterious change in affinity must be quickly corrected by either reversal or compensatory change [26]. The high mutation rate of the XL-exon, which is due to the high frequency of CpG sites, may allow such “corrective” changes to occur quickly. The reversals and/or compensatory changes likely occur in rapid succession, keeping the overall ratio of nonsynonymous changes (XL
K
A/ALEX
K
A) close to 1 for a given lineage, a phenomenon observed in our data (see Figure 2). The shortcoming of this stochastic process is that by constantly adjusting to each other, XLαs and ALEX may drift beyond the acceptable level of mutual affinity. One way to overcome this situation might be by changing the number of internal repeat units that may serve as sandbags on an air balloon—allowing rapid changes in affinity in a single step (e.g., an addition/deletion of a single repeat unit in humans causes a significant change in affinity [12]). This may explain remarkable variation in the number of internal repeat units in human and apes. This simple model implies that the two proteins evolve under a purifying selection scenario and that the observed high substitution rate is a consequence of the high GC content imposed by the need to maintain two reading frames.
We cannot rule out an alternative adaptive evolution explanation of the variation in the number of repeats and the pattern of amino acid changes in XLαs and ALEX. XLαs and ALEX are predominantly expressed in neuroendocrine tissues where they likely play a role in the development and maintenance of neurological functions [5,12,27]. In particular, XLαs expression is evident in distinct regions of the brain controlling processing of sensory information (locus coeruleus) and innervation of orofacial muscles (i.e., facial nucleus) [5]. Individuals with disrupted XLαs/ALEX interactions have multiple neurological complications, including feeding motility problems, psychomotor retardation, and disturbed behavior [12]. It is therefore plausible that amino acid replacements and the variation in the internal repeat number may have been associated with the adaptation of G-protein signaling to specific neurological functions, perhaps specific to humans. However, to reliably distinguish between the possibilities of purifying and positive selection, it is necessary to experimentally measure XLαs/ALEX affinities in primates—a direction currently pursued by our laboratories.
Is the XLαs/ALEX locus the only example of extensively overlapping reading frames in mammals? Only three additional cases are known where protein products of both reading frames were biochemically characterized. These include genes for the cyclin D-dependent kinase inhibitor INK4a [28], X-box protein 1 [29], and a region of overlap between 4E-BP3 and MASK [30]. Discovery of genes with alternative reading frames is hampered by our disbelief in their existence. For example, ALEX was discovered long after the XLαs gene had been identified [9,13]. Early results from our laboratories indicate that there are many more genes (possibly hundreds) potentially encoding multiple proteins via alternative reading frames. In each case the alternative reading frame is conserved in all known mammalian orthologs of a gene. Similarly to XLαs, most of these genes have been known for some time but the presence of the alternative reading frame has never been discovered. Biochemical characterization of these alternative products is underway and may assist us in discerning yet another facet of mammalian gene organization and evolution.
Materials and Methods
Amplification and sequencing of XL-exon.
The entire XL-exon was amplified from genomic DNA in all eight species, using primers 990F and 2954R or 2428R (Table 4). These primers were designed using published human sequence [4]. Specifically positions 318 and 511 within XL-exon were considered to be starts of XLαs and ALEX coding regions, respectively (as defined in [31] and [13]). PCR conditions were as follows: 1.75 U Taq (Expand High Fidelity PCR System; Roche Diagnostics, Mannheim, Germany), 0.2 mM dNTPs, 300 nM of each primer, 1 ng/μl template DNA, PCR buffer with MgCl2 (Expand High Fidelity PCR System), and 7% DMSO. Hot start reaction was carried out using an ABI Thermocycler 9700 (Applied Biosciences, Foster City, California, United States) under the following conditions: 94 °C for 5 min (initial denaturation), followed by 30 cycles of denaturation at 94 °C for 30 s, annealing at 61 °C for 30 s, elongation at 72 °C for 2 min, and final extension at 72 °C for 5 min. The amplified products were purified using the QIAquick PCR purification kit (Qiagen, Valencia, California, United States). In each taxon amplification products were sequenced in both directions using species-specific primers (Table 4). Sequencing reactions were carried out using 1 μM of primers, 7% DMSO, 35–50 fmol of template DNA, and CEQ DTCS Quick Start Kit (Beckman Coulter, Allendale, New Jersey, United States) in an ABI Thermocycler 9700 under the following conditions: 40 cycles of 96 °C for 20 s, 50 °C for 20 s, and 60 °C for 4 min. Traces were obtained using Beckman Coulter CEQ 8000 sequencer. Sequence traces were manually analyzed using the DNAStar software package (http://www.dnastar.com/web/index.php).
Table 4 Amplification and Sequencing Primers
Species abbreviations as in Figure 1.
Data analysis.
Reliable alignment was generated by first translating nucleotide sequences from each taxa, aligning the translations using ClustalW [32], refining these alignments manually, and then reconstructing nucleotide alignments, using the protein alignment as a guide. Phylogenetic tree and most statistics were calculated using the PAML software package [33]. All analyses were performed on the region of overlap between the two reading frames, excluding the repetitive region. Synonymous and nonsynonymous rates were apportioned among the branches of the tree using the codeml program of the PAML package under the free ratio model [34].
The neighbor-dependent modification of the NG method was written in PERL programming language and is available from the authors upon request. The only difference from the classical NG algorithm [19] is that pathways creating stop codons in the alternative reading frame are ignored by our method. For example, let us consider the alignment in Table 5.
Table 5 Sample Alignment Parameters for Neighbor-Dependent Modification of the NG Method
The alignment contains two reading frames: frame 0 starting at position 0 and frame 1 starting at position 1. The second codon of frame 0 contains two substitutions, and so there are two possible parsimonious pathways:
Pathway 2 would convert the second codon of frame 1 into a stop (TAG), and so it is not considered by our method.
To test whether the GC content of the XL-exon is required for the coexistence of the two reading frames, we first estimated codon frequencies in (1) human RefSeq genes, (2) XLαs reading frame, and (3) ALEX reading frame. This procedure was performed using a custom-designed PERL script. Coding regions of human RefSeq genes were downloaded from the National Center for Biotechnology Information ftp site (ftp://ftp.ncbi.nlm.nih.gov). We then used the evolver program of the PAML package to simulate 1,000,000 sequence sets, using the three codon frequency tables. Each set contained eight sequences corresponding to primate species used in this study. All other parameters accepted by evolver (phylogenetic tree, branch lengths, transition/transversion ratio, codon number, and the K
A/K
S ratio) were taken from codeml output generated during nucleotide substitution analysis of our data and were fixed in all three simulations. Each set of simulated sequences was then inspected for the presence of +1 and −1 overlapping reading frames. A set of simulated sequences was considered to have an overlapping reading frames if such frame was greater than or equal to 1,000 bp and was conserved in all eight sequences within the set.
Analysis of substitutions at CpG sites was carried out using a collection of PERL script, which can be obtained upon request.
Supporting Information
Accession Numbers
Sequences reported in this paper have been deposited in GenBank (http://www.ncbi.nlm.nih.gov/Genbank) under the following accession numbers: Homo sapiens (AJ224868), Colobus angolensis (AY771990), Gorilla gorilla,
Macaca mulatta,
Saimiri boliviensis,
Homo sapiens, and Pan troglodytes (AY898801–AY898805), Hylobates lar (AY4787144), and Pongo pygmaeus (AY787145).
We thank Ross Hardison, Webb Miller, Davis Ng, and the members of the Center for Comparative Genomics and Bioinformatics for helpful insights and discussions. Genomic DNA for chimpanzee and macaque was obtained from the Coriell Institute for Medical Research. The study was supported by funds from the Pennsylvania State University, the Huck Institutes for Life Sciences, and the National Institutes of Health.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. AN, SW, and KM conceived and designed the experiments. SW and PGM performed the experiments. AN analyzed the data. KM contributed reagents/materials/analysis tools. AN wrote the paper.
Abbreviations
ALEXalternative gene product encoded by the XL-exon
NGNei–Gojobori
XLαsextra large form of Gα
==== Refs
References
Harris BA 1988 Complete cDNA sequence of a human stimulatory GTP-binding protein alpha subunit Nucleic Acids Res 16 3585 3131741
Levine MA Modi WS O'Brien SJ 1991 Mapping of the gene encoding the alpha subunit of the stimulatory G protein of adenylyl cyclase (GNAS1) to 20q13.2—q13.3 in human by in situ hybridization Genomics 11 478 479 1769666
Kozasa T Itoh H Tsukamoto T Kaziro Y 1988 Isolation and characterization of the human Gs alpha gene Proc Natl Acad Sci U S A 85 2081 2085 3127824
Hayward BE Kamiya M Strain L Moran V Campbell R 1998 The human GNAS1 gene is imprinted and encodes distinct paternally and biallelically expressed G proteins Proc Natl Acad Sci U S A 95 10038 10043 9707596
Plagge A Gordon E Dean W Boiani R Cinti S 2004 The imprinted signaling protein XL alpha s is required for postnatal adaptation to feeding Nat Genet 36 818 826 15273686
Klemke M Pasolli HA Kehlenbach RH Offermanns S Schultz G 2000 Characterization of the extra-large G protein alpha-subunit XLalphas. II. Signal transduction properties J Biol Chem 275 33633 33640 10931851
Pasolli HA Klemke M Kehlenbach RH Wang Y Huttner WB 2000 Characterization of the extra-large G protein alpha-subunit XLalphas. I. Tissue distribution and subcellular localization J Biol Chem 275 33622 33632 10931823
Zakut H Ehrlich G Ayalon A Prody CA Malinger G 1990 Acetylcholinesterase and butyrylcholinesterase genes coamplify in primary ovarian carcinomas J Clin Invest 86 900 908 2394839
Kehlenbach RH Matthey J Huttner WB 1994 XL alpha s is a new type of G protein Nature 372 804 809 7997272
Kehlenbach RH Matthey J Huttner WB 1995 XL-alpha-s is a new type of G protein. CORRECTION Nature 375 253
Freson K Hoylaerts MF Jaeken J Eyssen M Arnout J 2001 Genetic variation of the extra-large stimulatory G protein alpha-subunit leads to Gs hyperfunction in platelets and is a risk factor for bleeding Thromb Haemost 86 733 738 11583302
Freson K Jaeken J Van Helvoirt M de Zegher F Wittevrongel C 2003 Functional polymorphisms in the paternally expressed XLalphas and its cofactor ALEX decrease their mutual interaction and enhance receptor-mediated cAMP formation Hum Mol Genet 12 1121 1130 12719376
Klemke M Kehlenbach RH Huttner WB 2001 Two overlapping reading frames in a single exon encode interacting proteins—A novel way of gene usage EMBO J 20 3849 3860 11447126
Waterston RH Lindblad-Toh K Birney E Rogers J Abril JF 2002 Initial sequencing and comparative analysis of the mouse genome Nature 420 520 562 12466850
Li WH 1997 Molecular evolution Sunderland (Massachusetts) Sinauer 481 p.
Rogozin IB Spiridonov AN Sorokin AV Wolf YI Jordan IK 2002 Purifying and directional selection in overlapping prokaryotic genes Trends Genet 18 228 232 12047938
Pedersen AM Jensen JL 2001 A dependent-rates model and an MCMC-based methodology for the maximum-likelihood analysis of sequences with overlapping reading frames Mol Biol Evol 18 763 776 11319261
Krakauer DC 2000 Stability and evolution of overlapping genes Evolution Int J Org Evolution 54 731 739
Nei M Gojobori T 1986 Simple methods for estimating the numbers of synonymous and nonsynonymous nucleotide substitutions Mol Biol Evol 3 418 426 3444411
Yang Z Nielsen R 2000 Estimating synonymous and nonsynonymous substitution rates under realistic evolutionary models Mol Biol Evol 17 32 43 10666704
Goldman N Yang Z 1994 A codon-based model of nucleotide substitution for protein-coding DNA sequences Mol Biol Evol 11 725 736 7968486
Yang Z 1998 Likelihood ratio tests for detecting positive selection and application to primate lysozyme evolution Mol Biol Evol 15 568 573 9580986
Giannelli F Anagnostopoulos T Green PM 1999 Mutation rates in humans. II. Sporadic mutation-specific rates and rate of detrimental human mutations inferred from hemophilia B Am J Hum Genet 65 1580 1587 10577911
Ebersberger I Metzler D Schwarz C Paabo S 2002 Genomewide comparison of DNA sequences between humans and chimpanzees Am J Hum Genet 70 1490 1497 11992255
Krawczak M Ball EV Cooper DN 1998 Neighboring-nucleotide effects on the rates of germ-line single-base-pair substitution in human genes Am J Hum Genet 63 474 488 9683596
Kondrashov AS Sunyaev S Kondrashov FA 2002 Dobzhansky-Muller incompatibilities in protein evolution Proc Natl Acad Sci U S A 99 14878 14883 12403824
Abramowitz J Grenet D Birnbaumer M Torres HN Birnbaumer L 2004 XLalphas, the extra-long form of the alpha-subunit of the Gs G protein, is significantly longer than suspected, and so is its companion Alex Proc Natl Acad Sci U S A 101 8366 8371 15148396
Quelle DE Zindy F Ashmun RA Sherr CJ 1995 Alternative reading frames of the INK4a tumor suppressor gene encode two unrelated proteins capable of inducing cell cycle arrest Cell 83 993 1000 8521522
Calfon M Zeng H Urano F Till JH Hubbard SR 2002 IRE1 couples endoplasmic reticulum load to secretory capacity by processing the XBP-1 mRNA Nature 415 92 96 11780124
Poulin F Brueschke A Sonenberg N 2003 Gene fusion and overlapping reading frames in the mammalian genes for 4E-BP3 and MASK J Biol Chem 278 52290 52297 14557257
Hayward BE Moran V Strain L Bonthron DT 1998 Bidirectional imprinting of a single gene: GNAS1 encodes maternally, paternally, and biallelically derived proteins Proc Natl Acad Sci U S A 95 15475 15480 9860993
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 Bielawski JP 2000 Statistical methods for detecting molecular adaptation Trends Ecol Evol 15 496 503 11114436
|
16110341
|
PMC1186735
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Aug 12; 1(2):e18
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010018
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1608950410.1371/journal.pbio.0030269Correspondence and Other CommunicationsEcologyEvolutionGenetics/Genomics/Gene TherapyHomo (Human)Comment on “Recent Origin and Cultural Reversion of a Hunter–Gatherer Group” CorrespondenceWaters Tony
1
1California State University at ChicoChico, CaliforniaUnited States of America8 2005 16 8 2005 16 8 2005 3 8 e269Copyright: © 2005 Tony Waters.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 Origin and Cultural Reversion of a Hunter-Gatherer Group
Authors' Reply
==== Body
I read the article “Recent Origin and Cultural Reversion of a Hunter–Gatherer Group” [1] with interest. The article raises questions about the nature of contemporary hunter–gatherer groups like the Mlabri of Thailand that are important. But I am concerned that the authors, in demonstrating the elegance of their genetic technique, have reduced the anthropological question about socioecology to an “either–or” one of descent from an ancient isolated group versus a relatively recent “flight to the forest” by a small founder group from a horticultural society. The authors claim that genetic, linguistic, and folkloric data come down solidly on the side of the latter conclusion. I think that as likely an explanation is that the Mlabri are a product of the socioecological world of highland Southeast Asia, where most groups have varying elements of both modes of subsistence.
No Southeast Asian highlanders are strictly horticulturalists or hunter–gatherers. Most Southeast Asian highlanders are horticulturalists who supplement their diet through foraging. A few of them also trade with groups like the Mlabri, who are at one extreme of the horticulturalist–forager continuum. Sometimes, trade occurs between linguistic groups, using shared knowledge of each other's languages. Other times, trade is within the same ethnic group. Indeed, the Khmu of Laos, who are linguistically most closely related to the Mlabri, have traditionally practiced this mixed strategy.
When observed in both the 1930s by Bernatzik [2], and in the late 20th century by missionaries and anthropologists, the Mlabri were in contact with other ethnic groups, primarily the highland Hmong, Northern Thai, and Lao. Indeed, Mlabri men spoke these languages well enough to trade forest products for scraps of cloth and rice. It is also probable that, as with many other such groups, women were captured or married, and Mlabri children were occasionally taken for adoption. Checking for evidence of Mlabri mtDNA in these populations could verify whether this is the case. However, this raises a second problem with the approach the authors took. The DNA of the hill tribes presented in the article did not include those groups that the Mlabri have had contact with, such as the Hmong, northern Thai, Htin, Lao, and Khmu of the remoter areas of Nan (Thailand), Phrae (Thailand), and Sayaboury (Laos) provinces, where they have lived during at least the last 70–80 years. Instead, the authors used blood samples from different hill tribes speaking Sino-Tibetan languages and currently living in the Chiang Rai and Mae Hong Son provinces of Thailand, hundreds of kilometers to the west. These tribes have had no known contact with the Mlabri during the last 80 years, or before. In such a context, perhaps it is not surprising that the authors concluded that the Mlabri were isolated from these groups.
This opens up another explanation for how the Mlabri might have persisted in Southeast Asia during the last 600 years. They could have been skilled hunter–gatherers who 600 years ago began living in symbiotic trading relationships with more settled groups. There is no reason that such relationships could not have been persistent, even though it does not fit neatly into the old hunter–gatherer versus horticulturalist dichotomy, favored by the authors. Nevertheless, I think that this is an interesting relationship to explore. While, as the authors point out, the Mlabri may have little to teach us about how humans subsisted before the dawn of agriculture, they may well have much to say about the socioecology of how horticulturalists and hunter–gatherers coexisted since the emergence of agriculture 10,000 years ago.
Citation: Waters T (2005) Comment on “Recent origin and cultural reversion of a hunter–gatherer group.” PLoS Biol 3(8): e269.
==== Refs
References
Oota H Pakendorf B Weiss G von Haeseler A Pookajorn S Recent origin and cultural reversion of a hunter–gatherer group PLoS Biol 2005 3 e71 10.1371/journal.pbio.0030071 15736978
Bernatzik H Dickson EW The spirits of the yellow leaves 1958 London R. Hale 222
|
16089504
|
PMC1187857
|
CC BY
|
2021-01-05 08:21:25
|
no
|
PLoS Biol. 2005 Aug 16; 3(8):e269
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030269
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1608950610.1371/journal.pbio.0030288FeatureBioinformatics/Computational BiologyNonePrices for Ingenuity FeatureO'Neill Bill 8 2005 16 8 2005 16 8 2005 3 8 e288Copyright: © 2005 Bill O'Neill.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.Competitions and community exchanges are spurring scientific progress.
==== Body
Martin Reese proved that he was more than a fly on the wall in the Drosophila Genome Center at the University of California at Berkeley when, as a young PhD student, he encouraged colleagues to approach their work in an entirely new way. The goal was to develop efficient computational tools to predict individual genes from the masses of genetic code being sequenced. “The problem was people published papers on their algorithms…on different data sets,” recalls Reese, “so it was very hard to get a really good assessment on actually whose is the best and what are really the right methods and the right underlying signs to be used to do this job right.” Borrowing an idea from protein biologists, Reese brought together 12 teams under the umbrella of the Genome Annotation Assessment Project (GASP), provided them with the same Drosophila sequence to hone their prediction programs, and invited them to defend their results at a workshop in Heidelberg, Germany, a few months later, in July 1999. What the participants lacked in funding, they made up for, in abundance, with bonhomie and commitment, says Reese. And the timing could not have been better. The newly available Drosophila genome allowed Reese to test the workshop's top two programs on the front line: “We learned some tricks through this experiment [and] the Drosophila was one of the best annotated genomes as a result” (Figure 1).
Figure 1
Drosophila melanogaster Genome Sequence from the 1999 GASP Project (http://www.fruitfly.org/GASP)
The expert annotations of the Berkeley Drosophila Genome Project groups are on top and on the bottom of the display (yellow). The submitted predictions from 12 different groups are in different colours, and close to the centre, general DNA sequence features are shown.
(Image: Reese et al. [1])
Exploiting competitive spirit to advance research is not new (Box 1), though it more often comes with the prospect of lucrative prizes from entrepreneurs eager to pit technical brains against each other to secure fast-track solutions. Such competitions are now more common and quirkily diverse, from battlefield strategists looking for a new generation of robotic navigators, to transport planners desperate for cooler travel underground, to NASA chiefs eager to extract oxygen from moon dust. Since 2003, there's been the Methuselah Mouse Prize (M Prize) for extending life span, and the X Prize Foundation, established in 1995, has just introduced an annual cup competition for space-travel innovations. The 55-year-old challenge to find a computer that can respond like a human still attracts competitors for an annual award, and even the world's climate-change negotiators proposed a prize for innovation ahead of the G8 Economic Summit in July, before seeing geopolitics squash the idea (along with a boost in research and development). The granddaddy of competitions—the prize launched in 1714 to determine longitude at sea—is these days heralded as the coming of the modern business world's entrepreneurial spirit to science.
Box 1. Prize Competitions
X Prize Cup
A prospective annual award for space-travel innovations from the X Prize Foundation, to replace its Ansari X Prize, whose $10 million purse for flying a private vehicle at least twice into space and back again within a fortnight went to US aviator Burt Rutan in October 2004. The award follows a long line of aviation prizes that go back to and beyond Charles Lindbergh's historic solo flight across the Atlantic in Spirit of St. Louis in 1927, which won him $25,000 from New York hotelier Raymond Orteig.
M Prize
An ongoing challenge that consists of two awards: a Longevity Prize for the oldest Mus musculus (currently standing at 1,819 days), and a Rejuvenation Prize for the best late-onset intervention (based on the rejuvenated mouse's age at death, currently standing at 1,356 days). Awards come from a fund, currently valued at around $1.3 million, to which anyone may contribute. A leading sponsor is “The 300”—modelled on the 300 Spartans who crucially delayed the invasion of Greece by hordes of Persians in 480 B.C.—whose members pledge regular contributions to the fund and whose names will be etched in history (as those of the Spartans were engraved on a stone tablet in Sparta).
Loebner Prize for Turing Test
British mathematician Alan Turing postulated, in 1950, that a “thinking” computer could produce responses to fool an interrogator that it was human; the prize, created by New Jersey industrialist Hugh Loebner in 1990, keeps the Turing Test a live challenge. Loebner has pledged $100,000 (plus a gold medal) for the first computer whose responses are indistinguishable from a human's. In the meantime, an annual prize of $3,000 (plus a bronze medal) goes to the most human computer that year. In 2005, according to Loebner, the award of $25,000 (plus a silver medal) looks likely to be won for the first time.
Longitude Prize
This prize was established in 1714 by the British government to determine longitude at sea. Instead of relying on astronomical sightings, watchmaker John Harrison built a precision clock to keep the time of a home port (of known longitude). Denied the £20,000 prize by assessors (wary that astronomy had been bypassed), Harrison petitioned King George III to circumvent them and to persuade Parliament to award him directly. Harrison was finally rewarded in 1773, 12 years late and 45 years after he began work on his “marine chronometer”. He died on his birthday in 1776, at the age of 83.
Moon Regolith Oxygen Challenge
Just this year, NASA announced a deadline stretching into 2008 for its third Centennial Challenge competition, the Moon Regolith Oxygen Challenge, to extract breathable oxygen from simulated lunar soil, and is dangling a purse of $250,000 in front of likely duellists.
Such prizes can bask in ambitious publicity campaigns and cash incentives, running into millions of dollars, in an effort to tap the broadest seam of ingenuity, yet money and marketing, even military muscle, are no guarantee of success. The Defense Advanced Research Projects Agency is having to rerun the Grand Challenge it staged late last year to find the fastest “autonomous ground vehicle” that can cover around 175 miles of treacherous desert track in under ten hours. Unperturbed, the Defense Advanced Research Projects Agency has doubled the prize money to $2 million. London Underground, which runs the capital's subterranean train service, has just abandoned its Cooling the Tube contest, after spending nearly two years appraising 3,500 entries from 60 countries and failing to find any feasible innovation to cope with summer heat waves that leave Tube travellers sweltering. And London Mayor Ken Livingstone, who promised £100,000 to the winner in 2003, has kept his hands in his pockets.
Against this background, the astonishing success of GASP and of a flourishing brand of similar low-key and often low-budget exchanges among scientific researchers, each vying to devise the perfect algorithm for some knotty problem, in fields that range from protein biochemistry to information retrieval to statistical genetics, seems all the more remarkable. And yet more remarkable still about these research exchanges is the avowed disdain among participants for the term “competition” to describe what they prefer to call a “community experiment”, and their claimed satisfaction (on coming first) with fame and glory among peers rather than with any cash award.
Critical Assessments—“It's Not a Competition”
As someone who could be in the running for ageing research's M Prize, Jim Carey, a population biologist and Professor of Entomology at the University of California at Davis, explains the problem with many such contests: “It is not so much that legitimate ageing researchers do not want to be seen as actively seeking a prize, as it is that a research strategy built on the goal of winning the prize would be way too high a risk.” Carey has considered his chances though. “When I heard about [the M Prize] we were in the midst of our ovary transplant mouse studies and it crossed my mind momentarily that maybe we'd be in the running. But it had no influence on our thinking about the type of research we do; this is not what drives us. If we were to ever win this prize (hypothetically), it would be by default rather than by inspiration; that is, we would claim the prize (since why not?) but that would not have been the driving force.”
At the community-experiment level, however, goals can be much more clearly tied to practical research agendas (Box 2). Among the most successful participants is David Baker, Professor of Biochemistry at the University of Washington in Seattle. Baker is involved in exchanges to develop computational methods to predict three-dimensional protein structure, the Critical Assessment of Techniques for Protein Structure Prediction (CASP), and protein–protein interactions, the Critical Assessment of Predicted Interactions (CAPRI), in which the correct results have been obtained experimentally but are known only to assessors. The most accurate prediction is declared winner.
Box 2. Community Experiments
CASP—Critical Assessment of Techniques for Protein Structure Prediction
Research teams make blind predictions about the structures of the same set of proteins from given sequences of amino acids. Started in 1994 and staged every two years, the experiments are coordinated by Lawrence Livermore National Laboratory in Livermore, California. For CASP6 last year, more than 200 teams from 24 countries provided over 30,000 predictions on 90 protein domains. John Moult founded CASP in response to the “clear inadequacy of the peer reviewed publication system in this area of biology [and to] new ways of doing things made possible by cheap universal electronic communication”. An associated Web-based community discussion arena, FORCASP (Forum for CASP), provides an online meeting place and an intense discussion venue for the CASP community.
CAPRI—Critical Assessment of Predicted Interactions
Research teams make blind predictions about the structures of protein–protein complexes from given structures of the individual proteins. CAPRI aims to do for macromolecular interaction, a central theme in functional genomics, what CASP has done for protein structure. Started in 2001 and staged whenever an experimentalist offers an adequate target, according to cofounder Joël Janin, CAPRI is coordinated by the European Bioinformatics Institute at Hinxton, United Kingdom. Round seven began in May. Over four years, X-ray crystallographers have provided 21 targets, including two that were cancelled.
CAFASP—Critical Assessment of Fully Automated Structure Prediction
Evaluates the performances of automatic prediction servers to determine how accurately they predict protein structures without the intervention of experts (as allowed in CASP), such that nonexperts could use them with confidence.
CAMDA—Critical Assessment of Microarray Data Analysis
Research teams analyse the same standard datasets and compare notes on the different techniques to mine microarray data. Modelled on CASP, CAMDA was founded in 2000 at Duke University Bioinformatics Shared Resource in Durham, North Carolina, and has staged conferences every year since then.
TREC—Text Retrieval Conference
Founded in 1992 by the National Institute of Standards and Technology and the US Department of Defense, TREC is a series of workshops (TREC 2005 reports in November) to encourage research in information retrieval from large text collections, notably for the benefit of the intelligence community. A more recent initiative, supported by the US National Science Foundation, focuses on the study of the retrieval of genomic data, which is broadly interpreted to mean not just gene sequences but also supporting documentation such as research papers and laboratory reports.
BioCreAtIvE—Critical Assessment of Information Extraction Systems in Biology
Established in 2003 at the National Center of Biotechnology (Centro Nacional de Biotecnología) in Madrid, Spain, BioCreAtIvE claims to be the “first very biologically motivated evaluation of text mining systems” [2].
GAWs—Genetic Analysis Workshops
Started in 1982 and now under the auspices of the International Genetic Epidemiology Society, GAWs bring genetic epidemiologists together to evaluate statistical methods on real or computer-simulated data that organizers distribute to investigators about six or seven months before the next meeting. GAW15 is scheduled for November 2006.
GASP—Genome Annotation Assessment Project
Some 12 groups participated in a one-off experiment in 1999, coordinated by the Drosophila Genome Center at the University California at Berkeley to assess gene and functional site predictions in genomic DNA using a Drosophila sample. Earlier this year, the Municipal Institute of Medical Research (Institut Municipal d'Investigavió Mèdica) in Barcelona, Spain, launched E-GASP (in association with the Encyclopedia of DNA Elements project), which challenged 18 teams to do the same for the human genome.
Success with computational modelling promises much. Better algorithms to predict the locations of genes would make finding them much less time-consuming and would lead to the discovery of more of them, insists Reese. The problem for geneticists is the lack of a robust comparison against which to gauge the accuracy of their predictions. “The protein people have the three-dimensional structure, which is clear,” he says. For Baker, determining protein structures experimentally is expensive and time-consuming, and “cannot keep up with the explosion of DNA sequences,” he says. “If we could accurately and consistently predict protein structures and interactions, it would have a huge impact on biology.
“The most exciting results so far in any of these things for me personally were our results in the last CAPRI test,” says Baker. “The predictions were so stunningly accurate that if we'd made [them] inside my research group…I'd have been convinced that we must have cheated somehow,” he recalls. “Several predictions were much more accurate than any predictions of anything in structural biology have ever been” (Figure 2).
Figure 2 Structure Prediction with RosettaDock in CAPRI
Prediction of the structure of the dockerin–cohesin complex (Target 12). Superposition of predicted (blue) and X-ray (red and orange) protein complex structures. A side chain whose conformation was correctly predicted to change upon complex formation is shown in green. The upper panel shows the whole complex; the lower panel shows details of the interface. In addition to the rigid-body orientation, the conformation of most of the side chains is predicted correctly, leading to the correct identification of 87% of the contacts in the crystal structure.
(Illustration: Created by Ora Furman, University of Washington, using the PyMOL Molecular Graphics System [http://www.pymol.org])
Revolutionising Research (in Some Areas at Least)
Like Reese with respect to genetics, Baker sees the evolution of community experiments, in the form of CASP and CAPRI, as rescuing structural biologists from primary research papers that cannot be trusted. Too many published models rely on known results: “It's like trying to predict yesterday's weather from conditions that you knew the day before,” he says. “It's not conscious cheating, it's just that if you're trying to reproduce some set of [known] results with a computational model, if you try hard enough and you're smart enough then you'll figure out a way to do it.” The issue, he adds, is “whether you actually have captured some essential truth about how things work or whether you've just managed to twiddle all the numbers so that you reproduce a certain set of results.”
He's now convinced that the days of depending on experimentation alone to determine protein biochemistry are numbered, as the benefits of CASP and CAPRI kick in: “And a good measure of when we are there is these types of experiments.” In a manuscript in review, Baker reports that “for about a third of the small proteins we looked at—very small proteins, less than 85 amino acids—we could predict their structures quite accurately. You'd like it to be 100%…but it's a lot better than it was a few years ago when it'd have been zero.” For the moment, he says, lack of sufficient computational power is the problem.
John Moult, who founded CASP, agrees that reliable prediction is not far off. “I always say five years. Been saying that for about 20 years now,” he notes. “Seriously—if we can get effective refinement methods, then homology models based on high sequence identity (say, more than 30%) could quickly become competitive with experiment. However, we are only just beginning to progress on that problem, so it is hard to call.”
CASP is succeeding where similar collaborations to resolve other biological questions could easily struggle, notes Moult, Professor of Computational Biology at the Center for Advanced Research in Biotechnology at the University of Maryland Biotechnology Institute in Rockville. “The CASP model requires new experimental data to become available on an appropriate time scale [and] that's fairly uncommon,” he says. “On the other hand, it is my strong conviction that new communication methods will allow a whole range of new ways of collaborating on a community scale.”
Granted, There Can Be Issues
But the collaboration is not always an easygoing affair. “Overall it works well, but there can of course be tensions in various forms,” says Moult. Participants who feel that the evaluation criteria are unfair to their predictions present the most common complaint: “Though this has happened, it is the exception,” he stresses. And growing sensitivity at funding agencies about the value of community experiments increases the tension. Involvement in CASP improves a researcher's chances of securing a grant, “or rather, not being involved in CASP may damage prospects,” says Moult. “This is not a good thing. It puts pressure on people to participate whether they think it's a good idea or not. I also suspect the significance of the results is sometimes overestimated by review committees.”
Such friction may be more of an issue in the US than in Europe, according to Joël Janin, Professor of Biophysics at the Centre National de la Recherche Scientifique Laboratory of Structural Enzymology and Biochemistry near Paris, France, and CAPRI's cofounder. “CAPRI wasn't planned to be a competition, and I do my best to keep the 'community-wide experiment' spirit in it,” he notes. “This seems to work in Europe and Asia for the moment, but American participants tell me they feel pressure from grant agencies.” The National Institutes of Health funds CASP, while CAPRI runs on a shoestring. “When Shoshana Wodak [from the Free University of Brussels] and I launched CAPRI, there was skepticism from our colleagues in the US that it could be run from Europe,” recalls Janin. “That skepticism was partly justified—we failed to get EU funding, and in the end Shoshana is moving from Brussels to Toronto.”
On the edge of these life sciences communities, looking in, is Ellen Voorhees, a computer scientist, who runs the Text Retrieval Conference (TREC) at the National Institute of Standards and Technology in Gaithersburg, Maryland, US. Since 2003, a decade after it was established, TREC has expanded its annual research workshops on improving the effectiveness of information retrieval systems, notably for the benefit of the intelligence community, to include assessments of methods for recovering genomics data.
Voorhees appreciates the tensions with community experiments. While TREC assessors do evaluate different retrieval systems and publish scores, she says, “TREC offers no award, and names no winners.” But some people still call TREC a competition, and there is an undeniable competitive element to it, she admits. “I used to try to correct people who called TREC a competition, but have given that up as a hopeless task.”
Benchmarking Research
Voorhees appreciates the rewards of community experiments. “TREC has created a series of retrieval test collections that define benchmark tasks that drive the research. These collections simply could not have been built without a collaborative effort because the collections depend on the pooled results of many different retrieval systems. A single organisation trying to build a collection of similar size could not obtain a collection of equivalent quality because of the bias introduced by a single system.
“Retrieval effectiveness doubled on the basic ‘ad hoc’ task over the first six years of TREC,” notes Voorhees. “TREC introduced the first large-scale evaluations of cross-language retrieval, and retrieval of recordings of speech. More than 250 groups from more than 20 countries have participated in at least one TREC. Many groups have participated multiple times. These groups must see some value in participating.”
With community experiments proving to be such a valuable tool across computational research, could they also help to solve questions other than algorithmic ones? Baker is far from convinced: “Prediction experiments are special for prediction problems, which will generally be computational.” Janin agrees: “I cannot imagine how to organize a wet bench experiment in the same way, but who knows?” Moult is more accommodating: “Things like CASP have so far focussed on testing how well computational methods succeed at reproducing experimental reality. In that mode, [it is] hard to see how experiment might fit in,” he says, adding wryly, “experiment reproducing experiment?” But the next generation of community experiments could see more of an overlap with experiment, he concedes: “For example, asking the small molecule docking community to contribute suggestions as to what might be the ligand binding specificity of proteins of unknown function. These suggestions would then help guide experimental binding studies. This sort of community-wide computational experiment does have more similarity with some community-based experimental projects, for example, Tom Terwilliger's TB structure consortium [http://www.doe-mbi.ucla.edu/TB/], where he manages target and results lists, and in principle, anyone can do the needed experimental work.” But for Reese, who now runs Omicia, a prognostic genetics start-up he founded in 2002 in Emeryville, California, community experiments have a finite life. “Once we have all the genes found, then [GASP] will become redundant.”
Citation: O'Neill B (2005) Prices for ingenuity. PLoS Biol 3(8): e288.
Bill O'Neill is a freelance writer based in London, United Kingdom. E-mail: [email protected]
Abbreviations
CAPRICritical Assessment of Predicted Interactions
CASPCritical Assessment of Techniques for Protein Structure Prediction
GASPGenome Annotation Assessment Project
M PrizeMethuselah Mouse Prize
TRECText Retrieval Conference
==== Refs
References
Reese MG Hartzell G Harris NL Ohler U Abril JF Genome annotation assessment in Drosophila melanogaster
Genome Res 2000 10 483 10779488
Centro Nacional de Biotecnología BioCreAtIvE: Critical Assessment of Information Extraction Systems in Biology 2003 Available: http://www.pdg.cnb.uam.es/BioLINK/BioCreative.eval.html . Accessed 24 June 2005
|
16089506
|
PMC1187858
|
CC BY
|
2021-01-05 08:21:26
|
no
|
PLoS Biol. 2005 Aug 16; 3(8):e288
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030288
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1608950710.1371/journal.pbio.0030292PrimerDevelopmentAnimalsThe Left–Right Polarity Puzzle: Determining Embryonic Handedness PrimerWood William B 8 2005 16 8 2005 16 8 2005 3 8 e292Copyright: © 2005 William B. Wood.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.
De Novo Formation of Left-Right Asymmetry by Posterior Tilt of Nodal Cilia
How Do Embryos Know Left from Right?
Whenever symmetry is broken in nature to yield only one of two equally probable outcomes, there is an intriguing problem to be solved.
==== Body
Whenever symmetry is broken in nature to yield only one of two equally probable outcomes, whether in physics, chemistry, or biology, there is an intriguing problem to be solved. Physicists from M. and P. Curie to T.-S. Lee and C.-D. Yang puzzled over such phenomena at the atomic level. Organic chemists puzzled over the handedness of molecules for many years after Pasteur showed that grape juice contained only one of the possible right- and left-handed mirror-image forms (enantiomers) of tartaric acid (Figure 1; see Glossary [Box 1]). And biologists continue to puzzle over the handedness of organisms.
Box 1. Glossary
Bilaterian: having bilateral symmetry.
Chiral: having chirality.
Chirality: the screw sense (handedness) of a helix. Its mirror image will have the opposite chirality.
Dextral and sinistral: right- and left-handed, respectively (referring to laterality or chirality). For chiral structures these terms are absolute (a right- or left-handed screw axis); for laterality, the predominant handedness is often called dextral arbitrarily.
Enantiomeric: handed.
Enantiomers: the two possible mirror image forms of an asymmetric object (usually applied to molecules) with no bilateral symmetry.
Handed: having handedness.
Handedness: the difference between two objects that are mirror images of each other, such as the right and left hands.
Laterality: the handedness of situs, the arrangement of internal organs (viscera) in the body.
Situs inversus viscerum: a condition in which the normal laterality is mirror-image reversed, that is, it develops with opposite to normal handedness. Individuals with this condition may be functionally completely normal.
Figure 1 Mirror-Image Symmetry of the Enantiomeric Molecules D- and L-Tartaric Acid
Pasteur discovered that a solution of tartaric acid from grape juice (now known to contain only the D form) rotated plane-polarized light, whereas chemically synthesized tartaric acid did not. Pasteur solved this puzzle by showing that the chemically synthesized compound was a mixture of the two forms, which when separated could rotate light in opposite directions.
Why is there a puzzle? The embryos of most, probably all, bilaterians exhibit obvious polarities from head to toe (along the anterior–posterior axis) and back to front (along the dorsal–ventral axis), but they also exhibit less obvious left–right (L-R) differences. That is, although bilaterally symmetrical on the outside, they are L-R asymmetric on the inside. The polarity of the L-R axis determines the laterality of the body plan, for example, whether the human heart will be on the left side or on the right. There are two possible mirror-image forms of the animal body plan, just as there are for tartaric acid, differing only in L-R polarity. Almost without exception, however, the body plans of individuals in any given species develop as only one of the two possible “enantiomers.” This means that at some point during embryonic development, just as in the grape's synthesis of tartaric acid, L-R symmetry must be broken in a unique manner, so that all individuals develop with the same handedness—for example, with the heart on the left. Establishment of asymmetry in embryos is no longer a mystery; we know of several mechanisms by which a cell or a tissue can become asymmetrically polarized. The puzzle here lies in the mechanism of initial choice between two opposite polarities that should be equally probable. We now know that stereospecific synthesis of biomolecules like tartrate comes about because the enzymes that catalyze their synthesis are also stereospecific, handed molecules. As the organic chemist F. R. Japp stated in 1898: “only asymmetry can beget asymmetry” (quoted in [1]). Where does the stereospecific cue in embryonic handedness choice come from?
L-R Asymmetry Can Be Established Early in Development
Over 100 years ago, H. E. Crampton [2] observed that the handedness, or chirality, of snail shell coiling, dextral (right-handed) for some species and sinistral (left-handed) for others, could be predicted from the handed orientation of the two mitotic spindles prior to second cleavage of the embryo. A. Sturtevant [3], working with a mutation that caused sinistrality in a normally dextral snail species, showed that handedness is controlled by a maternal-effect gene, suggesting that some maternal gene product, incorporated into the oocyte, could influence spindle orientation and the chirality of subsequent shell coiling (although recent work has shown that the early embryonic stages of sinistral mutants are not strict mirror images of the corresponding normal stages [4]). This suggestion was borne out by the finding that sinistral zygotes could be “rescued” to become dextral embryos by injection of cytoplasm from a dextral oocyte [5]; unfortunately, the active substance has not been identified. So snails establish handedness very early; what bilateral symmetry they exhibit must be superimposed later (see Box 2).
Box 2. Intriguing Questions about L-R Asymmetry and Handedness
(1) What is the L-R ground state of the embryo? Is it bilateral symmetry, which must be broken to create the L-R asymmetry of the viscera, or is it very early L-R asymmetry, on which later external bilateral symmetry must be imposed?
(2) Why has one handedness been selected by evolution over the other for each species? We know of many examples, from nematodes to humans, showing that when rare mirror-image individuals of the opposite handedness arise, they can be functionally entirely normal [25]. So internal handedness doesn't appear to matter, and yet only one of the two possible body plans is observed in almost all individuals of a species.
(3) Why have animals evolved to be L-R asymmetric internally, rather than bilaterally symmetric as on the outside?
Apparent answers to the first question for at least some animals will come up in the text. About the second and third questions we can still only speculate.
The nematode Caenorhabditis elegans, which exhibits clear laterality of internal organs, has also established handed L-R asymmetry by the six-cell stage [6] and probably earlier (W. B. Wood, unpublished data). The external bilateral symmetry of the animal is imposed during embryonic development by cell signaling [7], which changes the relationship of cell lineage to cell fate on the two sides of the animal to compensate for the physically asymmetric placement of lineally homologous cells in the ectoderm [6].
What about vertebrates? Much has been learned from the study of molecular markers that exhibit L-R asymmetry in their expression (reviewed in [8,9]). In embryos of the frog Xenopus, there is a clear L-R asymmetry of the maternally expressed TGF-β family member Vg1 in vegetal blastomeres as early as the eight-cell stage. Vg1 is seen predominantly on the left side, and injection of Vg1 on the right side leads to random choice of laterality in the resulting embryos (that is, about 50% dextral and 50% sinistral) [10]. More recently, a second, even earlier asymmetry was found at the four-cell stage [11] in localization of the mRNA for a maternally expressed H+/K+-ATPase. Asymmetric localization of this proton pump is important: pharmacological blocking of the ATPase results in randomization of laterality. Recent results with zebrafish [12] and chick [11] embryos have shown ionic potential differences across the midline prior to gastrulation, resulting from asymmetric proton pump activity, and these differences also appear to be required for normal handedness choice. These results suggest that in lower vertebrates, as well as invertebrates, handed L-R asymmetry is established early in embryogenesis, even though morphological L-R asymmetry is not apparent until gastrulation.
Analysis of other signaling molecules in mouse embryos also revealed L-R molecular asymmetries, but not until around the time of early gastrulation, when thousands of cells are present. Subsequent studies showed that these embryos, as well as all the vertebrates mentioned above, have elaborate, presumably homologous asymmetric signaling pathways that function from this point onward to maintain L-R differences on either side of the midline and thereby control laterality of heart looping and asymmetric development of the viscera (reviewed in [9]). (Incidentally, recent work has shown that in vertebrates, too, the symmetry of somite development along the dorsal midline must be superimposed on the underlying pattern of L-R asymmetry by additional signaling [12,13].)
A Mechanical Polarity Generator
The elucidation of later L-R signaling in mammals does not address the question of when or how L-R asymmetry with the correct handedness is initially established. The first clues to a surprising possible answer to this question came from human, and then from mouse, genetics. Among individuals with Kartagener syndrome, caused by one of several human dynein defects that result in ciliary dysfunction (leading to bronchial problems and male infertility), laterality was found to be randomized; that is, half of these patients exhibited “situs inversus viscerum” (reversed body plan) while the rest had the normal body plan. The iv gene in mice, mutation of which also causes randomized laterality, was found to encode a new member of the dynein family, which was named left-right dynein, or Lrd.
The significance of dynein involvement in handedness choice became clear through a remarkable series of discoveries, beginning with the demonstration in mouse embryos that monocilia, present on the node (corresponding to the amphibian Spemann organizer) in early gastrulation and previously thought to be immotile, did in fact beat. Moreover, their beating could move fluorescent beads consistently to the embryo's left, suggesting that they could be providing an asymmetrical cue for handedness determination [14]. Consistent with this view, iv mutant mouse nodal cilia appeared to be immotile, and mouse knockout mutations of the Kif3 kinesin genes, resulting in lack of nodal cilia, also randomized laterality. Artificially created rightward flow resulted in embryos with reversed laterality, and artificial leftward flow with iv mutant embryos rescued the mutant defect, strong evidence that the directional flow itself was causative for correct handedness determination [15]. Presumably, the normal direction of the flow was somehow dependant on the intrinsic chirality of the cilia themselves, thus providing a possible physical basis for choice of the correct handedness.
But how the cilia might actually provide such a cue remained an unanswered question until recently. What was being moved? Nonaka et al. [14] originally proposed that the cilia might move an unidentified morphogen, which could trigger asymmetric establishment of the previously defined left and right signaling cascades. Later evidence suggested that the more immotile cilia around the edges of the node could be mechanosensors, containing the polycystic-kidney-disease (PCKD) ion channel protein. It was proposed that these sensory cilia could be activated by fluid flow on only the left side to initiate an observed asymmetric release of Ca++, which in turn could activate subsequent asymmetric signaling [16,17].
Another unanswered question was how the nodal cilia could cause leftward flow. Monocilia lack the central-pair microtubules that define the beating direction of “conventional” cilia, and consequently, monocilia move with a uniform rotating motion. Conventional cilia, by contrast, exhibit a back-and-forth beat with defined power and return strokes that can push surrounding fluid in one direction. Rotating cilia should cause local vortices, not a directional flow. Attempts to explain this directionality by the geometry of the nodal depression were unsatisfying [14].
Three recent papers have provided some answers to these questions. What's being moved? Tanaka et al. [18] present evidence that fibroblast growth factor (FGF) in the region of the node stimulates the release of 0.3- to 5-µm vesicles that contain the signaling molecules and possible morphogens Sonic Hedgehog (SHH) and retinoic acid (RA). These “nodal vesicular particles” are swept by nodal cilia to the left edge of the node, where they fragment to release their cargo, which might be the trigger for the previously observed asymmetric rise in local Ca++ concentration.
And leftward flow? A group of fluid dynamicists proposed a simple solution [19]: just tilt the cilia toward the posterior! In this configuration, when the clockwise-rotating cilia stroke to the embryo's right, they will be close to the nodal cell surface, which locally impedes fluid flow, and when they stroke to the left at the top of their arc, they will be away from the surface, where fluid flow is unimpeded. The result will be leftward fluid flow. Reporting in this issue of PLoS Biology, Nonaka et al. [20] have used high-speed video microscopy to experimentally validate the predicted posterior tilt. This work accords with a recent independent study [21] and moreover shows that the rightward stroke of each cilium actually brushes along the nodal surface, so that the trajectory of the ciliary tip is a D-shaped rather than a circular arc (Figure 2). As a further test of the fluid dynamic theory, Nonaka et al. [20] built a working model with tilted wire cilia rotating through a viscous medium to approximate the fluid dynamics of the nodal environment. They show that it indeed moved suspended particles in the predicted direction only.
Figure 2 Ventral View of Monocilia on the Mouse Node in Early Gastrulation
The diagram shows how clockwise-rotating cilia on the nodal cells can move a fluid suspension of small vesicles containing signaling molecules (nodal vesicular particles [NVPs]; red spheres) toward the left, creating a right-to-left asymmetric gradient across the midline. Key to the cilia's function is the posterior tilt of their rotational axes, as explained in the text. Connection arrows show the trajectory of the tip of one cilium as it rotates.
These recent papers provide answers to two major questions about how nodal cilia can cause directional flow and how this, in turn, can initiate L-R asymmetric signaling. While they do not rule out the mechanosensory model, they do show that asymmetric transport of putative morphogens occurs as well. Perhaps both mechanosensors and morphogens are involved in activating subsequent laterality pathways.
A General Mechanism?
Where does this leave our understanding of handedness choice? There is still a major caveat regarding the mammalian mechanism and its relationship to the presumably homologous mechanisms in other vertebrates. Rotating cilia, transiently present on the node, or equivalent structures in early gastrulation have now been demonstrated or implicated in embryos of mouse, rabbit, chick, zebrafish, medaka fish, and frog [21–23]. In the two mammals and the two fish, this rotation has been shown to move nodal fluid to the left, suggesting that all these embryos, despite very different embryonic and nodal geometries, may use a conserved mechanism for regulating subsequent laterality pathways that is dependent on the inherent chirality of cilia. Still unclear, however, is whether this ciliary rotation is the initial event that breaks L-R symmetry to establish handedness, or whether it serves as an amplifying mechanism for an initial choice that was made earlier in embryogenesis. Most of the vertebrate researchers cited above assume the former possibility, based on experiments showing that directional flow of nodal fluid is both necessary and sufficient for handedness determination. However, if we consider elaboration of L-R asymmetry as a stepwise process or pathway, necessity and sufficiency are to be expected of a downstream component, and they do not preclude the possibility that there are required upstream components as well.
In all but the mammals, L-R asymmetries are known to be present before the node develops. Levin [24] has convincingly reviewed arguments for early laterality cues that could be amplified by the action of nodal cilia. Among these early asymmetries, the potential difference across the midline, in particular, is common to zebrafish, frog, and chick and is necessary for normal development of laterality. There is, to my knowledge, no similar evidence for necessary asymmetries preceding nodal flow in the mouse, but few attempts have been made to find them [24]. Conceivably, the need for earlier cues was lost during the evolution of mammals. But at least the existence of such cues should be rigorously tested in the mouse embryo before assuming they are not present or play no role.
And so, with the possible exception of the mammalian mechanism, the nature of the initial symmetry-breaking cue that dictates correct handedness choice in invertebrates and most vertebrates still eludes us. Parts of the L-R asymmetry picture have become clearer, but there are still several pieces of the puzzle to be put in place.
Citation: Wood WB (2005) The left–right polarity puzzle: Determining embryonic handedness. PLoS Biol 3(8): e292.
William B. Wood is Distinguished Professor of Molecular, Cellular, and Developmental Biology at the University of Colorado, Boulder, Colorado, United States of America. He is currently at the Max-Planck Institute for Cell Biology and Genetics in Dresden, Germany, doing research on handedness choice in Caenorhabditis elegans embryos with support from the Alexander von Humboldt Foundation. E-mail: [email protected]
Abbreviation
L-Rleft–right
==== Refs
References
Weyl H Symmetry 1952 Princeton (New Jersey) Princeton University Press 168
Crampton H Reversal of cleavage in a sinistral gasteropod Ann N Y Acad Sci 1894 8 167 170
Sturtevant AH Inheritance of direction of coiling in Limnea
Science 1923 58 269 270 17837785
Shibazaki Y Shimizu M Kuroda R Body handedness is directed by genetically determined cytoskeletal dynamics in the early embryo Curr Biol 2004 14 1462 1467 15324662
Freeman G Lundelius JW The developmental genetics of dextrality and sinistrality in the gastropod Lymnaea peregra
Rouxs Arch Dev Biol 1982 191 69 83
Sulston J Schierenberg E White J Thomson J The embryonic cell lineage of the nematode Caenorhabditis elegans
Dev Biol 1983 100 64 119 6684600
Wood WB Evidence from reversal of handedness in C. elegans embryos for early cell interactions determining cell fates Nature 1991 349 536 538 1992354
Vogan KJ Tabin CJ A new spin on handed asymmetry Nature 1999 397 295 298 9950421
Mercola M Levin M Left-right asymmetry determination in vertebrates Annu Rev Cell Dev Biol 2001 17 779 805 11687504
Hyatt BA Yost HJ The left-right coordinator: The role of Vg1 in organizing left-right axis formation Cell 1998 93 37 46 9546390
Levin M Thorlin T Robinson K Nogi T Mercola M Asymmetries in H(+)/K(+)-ATPase and cell membrane potentials comprise a very early step in left-right patterning Cell 2002 111 77 89 12372302
Kawakami Y Raya A Raya RM Rodriguez-Esteban C Belmonte JC Retinoic acid signalling links left-right asymmetric patterning and bilaterally symmetric somitogenesis in the zebrafish embryo Nature 2005 435 165 171 15889082
Vermot J Pourquie O Retinoic acid coordinates somitogenesis and left-right patterning in vertebrate embryos Nature 2005 435 215 220 15889094
Nonaka S Tanaka Y Okada Y Takeda S Harada A Randomization of left-right asymmetry due to loss of nodal cilia generating leftward flow of extraembryonic fluid in mice lacking KIF3B motor protein Cell 1998 95 829 837 9865700
Nonaka S Shiratori H Saijoh Y Hamada H Determination of left-right patterning of the mouse embryo by artificial nodal flow Nature 2002 418 96 99 12097914
Tabin CJ Vogan KJ A two-cilia model for vertebrate left-right axis specification Genes Dev 2003 17 1 6 12514094
McGrath J Somlo S Makova S Tian X Brueckner M Two populations of node monocilia initiate left-right asymmetry in the mouse Cell 2003 114 61 73 12859898
Tanaka Y Okada Y Hirokawa N FGF-induced vesicular release of Sonic hedgehog and retinoic acid in leftward nodal flow is critical for left-right determination Nature 2005 435 172 177 15889083
Cartwright JH Piro O Tuval I Fluid-dynamical basis of the embryonic development of left-right asymmetry in vertebrates Proc Natl Acad Sci U S A 2004 101 7234 7239 15118088
Nonaka S Yoshiba S Watanabe D Ikeuchi S Goto T De novo formation of left–right asymmetry by posterior tilt of nodal cilia PLoS Biol 2005 3 e268 10.1371/journal.pbio.0030268 16035921
Okada Y Takeda S Tanaka Y Belmonte JC Hirokawa N Mechanism of nodal flow: A conserved symmetry breaking event in left-right axis determination Cell 2005 121 633 644 15907475
Essner JJ Vogan KJ Wagner MK Tabin CJ Yost HJ Conserved function for embryonic nodal cilia Nature 2002 418 37 38 12097899
Essner JJ Amack JD Nyholm MK Harris EB Yost HJ Kupffer's vesicle is a ciliated organ of asymmetry in the zebrafish embryo that initiates left-right development of the brain, heart and gut Development 2005 132 1247 1260 15716348
Levin M Motor protein control of ion flux is an early step in embryonic left-right asymmetry Bioessays 2003 25 1002 1010 14505367
Wood WB Left-right asymmetry in animal development Annu Rev Cell Dev Biol 1997 13 53 82 9442868
|
16089507
|
PMC1187859
|
CC BY
|
2021-01-05 08:21:26
|
no
|
PLoS Biol. 2005 Aug 16; 3(8):e292
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030292
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1608950810.1371/journal.pbio.0030297Community PageBioinformatics/Computational BiologyEcologyGenetics/Genomics/Gene TherapyNoneBioinformatics and Data Management Support for Environmental Genomics Community PageField Dawn [email protected] Bela Snape Jason 8 2005 16 8 2005 16 8 2005 3 8 e297Copyright: © 2005 Field 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 UK Natural Environment Research Council has funded the creation of a dedicated bioinformatics centre as part of a £26m Environmental Genomics initiative.
==== Body
As concerns over pollution and climate change increase, understanding the impact of environmental change on living organisms is coming to the fore as never before. Research in the area of environmental genomics, through the application of genomic technologies, is shedding light on fundamental processes by which organisms evolve and adapt to both the biotic and abiotic aspects of their environments. Understanding how organisms perceive and react to changes in their environments also has many direct applications, for example, in the field of bioremediation.
The United Kingdom Natural Environment Research Council (NERC) has recently invested over £26 million in this area under the Environmental Genomics and Post-Genomics and Proteomics Science Programmes. As part of the Environmental Genomics initiative, NERC funded the creation of a dedicated bioinformatics centre with a remit to provide bioinformatics consultation, training, and data management.
The Environmental Genomics Thematic Programme Data Centre (EGTDC) was established in 2002 and works to develop and implement bioinformatics and data management solutions for environmental genomics researchers. The EGTDC team of bioinformaticians and data managers work together to develop a variety of open-source projects including Bio-Linux, a Linux computing platform customized for bioinformatics research, Partigene, an EST analysis pipeline, and Maxd, a transcriptomics software suite that specializes in aiding users in annotating experiments to MIAME standards. The EGTDC also presents courses on a range of topics, has developed an extensive set of Web-based documents, and aims to make all of the resources it develops available to the wider public.
The EGTDC is particularly interested in the use and development of data standards in this area and has developed an “ENV” extension to MIAME to capture information about microarray experiments relevant to environmental samples. This activity is now formally recognized by the Microarray Gene Expression Data Society.
Through the combined provisioning of computers, software, appropriate data standards, and bioinformatics consultation, the EGTDC aims to help researchers more easily collect, store, and interpret their genomic data (Figure 1). To maximize the collective value of this data, the EGTDC has recently developed a public data catalogue (Figure 2). Discovery-level metadata are stored along with accession numbers for data held in public databases (dbEST, ArrayExpress, EMBL, etc). Therefore, the data catalogue is a way for users to “spider out” to the digital datasets that may be held in primary public databases and in specialized databases generated by the environmental genomics research community, the EGTDC, or individual researchers. The catalogue currently holds information for 28 environmental genomic grants and a total of 314 data holdings of a variety of types, including “omic” data (genomics, arrays, and proteomics data), single nucleotide polymorphisms, microsatellites, genetic maps, libraries, publications, and other documents, such as protocols and phenotypic data. This catalogue can be searched by keyword, data type, and a variety of other fields and provides accession numbers that can be cited in publications or linked to (accessed) directly by URL. Anyone interested in submitting information on existing or future datasets can contact the EGTDC helpdesk at E-mail: [email protected].
Figure 1 The Bio-Linux Computing Platform
(A) Bio-Linux is a Linux distribution customized to be user-friendly that contains approximately 60 popular bioinformatics packages. The Bio-Linux system is freely available, though researchers not supported by the EGTDC must provide their own hardware. Software developed at the EGTDC is included on the Bio-Linux system, making it easy for researchers not well versed in computing to try out these packages. The graphical menus and wide range of bioinformatics software installed on Bio-Linux makes it an ideal system for all levels of users, including beginners, power users, and developers.
(B) MaxdLoad2, one of the tools installed on Bio-Linux, provides an intuitive interface to annotate experiments to MIAME standards. When completed, the experiment annotation can be exported in MAGE-ML for submission to ArrayExpress. MaxdLoad2 has been recently engineered to capture data from the “ENV” extension to MIAME.
Figure 2 The EGTDC Data Catalogue
The EGTDC data catalogue currently contains descriptions of 28 environmental genomic grants, each of which can be viewed by its accession number (http://envgen.nox.ac.uk/). Projects include the following studies.
(A) How viral infections affect natural phenomena like marine algal blooms (egcat:000010). A virus-infected bloom of the microalga Emiliania huxleyi in the English Channel. Up to 50 million viruses per millilitre were observed in this bloom.
(B) The use of earthworms as sentinels of heavy metal pollution in soils (egcat:000024).
(C) The genes responsible for circadian and tidal rhythmicity in marine worms (egcat:000029). The “Worm Team” from left: Cas Kramer, Thierry Bailhache, Peter Olive, and Kim Last, ready for the collection of king ragworms in the Blyth Estuary.
([A] Image: Remote Sensing Data Analysis Service/Plymouth Marine Laboratory; [B] Image: Dr. A. John Morgan; [C] Image: Kim Last)
The EGTDC is currently expanding its remit to include future support for proteomics and metabolomics data management and integration of solutions in a systems biology context, with funding from the NERC Post-Genomics and Proteomics Science Programme. The EGTDC is currently funded until early 2009 and hopes to continue, in collaboration with others, to improve the range of bioinformatics tools and databases available for researchers working in environmental genomics. Its vision for the future includes the ability to integrate its genomic holdings with biological and environmental datasets held across the NERC and beyond. The EGTDC expects to change its name to the NERC Environmental Bioinformatics Centre in the near future, but its main remit will remain the collection, curation, and management of genomic data of environmental relevance.
More information can be found on the EGTDC Web site at http://envgen.nox.ac.uk.
Citation: Field D, Tiwari B, Snape J (2005) Bioinformatics and data management support for environmental genomics. PLoS Biol 3(8): e297.
Dawn Field is Head of the Molecular Evolution and Bioinformatics Section and Director of the Environmental Genomics Thematic Programme Data Centre (EGTDC) at the Oxford Centre for Ecology and Hydrology, Oxford, United Kingdom. Bela Tiwari is the EGTDC's Lead Bioinformatician. Jason Snape is Science Co-Ordinator for the Environmental Genomics Thematic Programme at the Brixham Environmental Laboratory, AstraZeneca, Devon, United Kingdom.
Abbreviations
EGTDCEnvironmental Genomics Thematic Programme Data Centre
NERCUnited Kingdom Natural Environment Research Council
|
16089508
|
PMC1187860
|
CC BY
|
2021-01-05 08:21:26
|
no
|
PLoS Biol. 2005 Aug 16; 3(8):e297
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030297
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030270Correspondence and Other CommunicationsEcologyEvolutionGenetics/Genomics/Gene TherapyHomo (Human)Authors' Reply CorrespondenceStoneking Mark Pakendorf Brigitte
1
Oota Hiroki
2
1Max Planck Institute for Evolutionary AnthropologyLeipzigGermany2University of TokyoKashiwaJapan8 2005 16 8 2005 16 8 2005 3 8 e270Copyright: © 2005 Stoneking 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.
Recent Origin and Cultural Reversion of a Hunter-Gatherer Group
Comment on "Recent Origin and Cultural Reversion of a Hunter-Gatherer Group"
==== Body
Waters [1] makes a number of points concerning our article [2], which, in our view, require clarification. First, Waters states that classifying Southeast Asian highland groups as either strictly horticultural or strictly foraging is overly simplistic, as most groups practice horticulture supplemented with some degree of foraging. While we are sympathetic with the view that subsistence strategy is more complicated than a simple dichotomy (indeed, one of the main messages of our paper is that a strictly foraging group such as the Mlabri may have practiced horticulture in the past), we wish to emphasize that the Mlabri are, indeed, quite different from the other Southeast Asian highland groups in that they have never, in either their recorded or oral history, practiced horticulture. It is this distinction, coupled with their extreme paucity of genetic diversity, that sets them apart from other groups in the area.
Second, Waters suggests that our comparison of the Mlabri with hill tribes from a different geographic region (Chiang Rai and Mae Hong Son provinces of Thailand) leads to our conclusion that “the Mlabri were isolated from these groups,” and that had we examined neighboring groups of the Mlabri, we might have reached a different conclusion. These statements misrepresent our work; in particular, we found that the Mlabri were not genetically distinct from other hill tribes for which we had data, as the mtDNA sequence, Y-STR alleles, and autosomal STR alleles of the Mlabri are all found in other groups. Moreover, this sharing pattern is in stark contrast to African foraging groups, such as the !Kung and Pygmies, who are genetically distinct from their horticultural neighbors. It is precisely this sharing of alleles between the Mlabri and other groups that is the basis for our suggestion that the Mlabri may have reverted to their current exclusively foraging lifestyle from a previous horticultural lifestyle, rather than having always been foragers.
Finally, Waters states that we claimed that our data “solidly” support the scenario of an extreme founder event from a horticultural group, followed by reversion to a foraging lifestyle, for the origin of the Mlabri. This is not true; we were careful to state that our data only suggest such a scenario. We agree with Waters that genetic analysis of neighboring groups of the Mlabri (in particular, the Tin Prai) would be useful to further evaluate the scenario we proposed for the origin of the Mlabri. And we clearly agree with Water's concluding statement concerning the importance of interactions between horticultural and foraging groups, as we make exactly that point in the penultimate sentence of our paper.
Citation: Stoneking M, Pakendorf B, Oota H (2005) Author's reply. PLoS Biol 3(8): e270.
==== Refs
References
Waters T Comment on “Recent origin and cultural reversion of a hunter-gatherer group” PLoS Biol 2005 3 e269 10.1371/journal.pbio.0030269 16089504
Oota H Pakendorf B Weiss G von Haeseler A Pookajorn S Recent origin and cultural reversion of a hunter–gatherer group PLoS Biol 2005 3 e71 10.1371/journal.pbio.0030071 15736978
|
0
|
PMC1187861
|
CC BY
|
2021-01-05 08:21:26
|
no
|
PLoS Biol. 2005 Aug 16; 3(8):e270
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030270
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 1611034210.1371/journal.pcbi.001002505-PLCB-RA-0084R2plcb-01-03-01Research ArticleBiochemistryBioinformatics - Computational BiologyEvolutionMicrobiologySystems BiologyEubacteriaArchaeaPrediction of Transcriptional Terminators in Bacillus subtilis and Related Species Prediction of Transcriptional Terminatorsde Hoon Michiel J. L. *¤aMakita Yuko ¤bNakai Kenta Miyano Satoru Human Genome Center, Institute of Medical Science, University of Tokyo, Minato-ku, Tokyo, JapanMurray Diana EditorWeill Medical College of Cornell University, United States of America* To whom correspondence should be addressed. E-mail: [email protected]¤a Current address: Center for Computational Biology and Bioinformatics, Columbia University, New York, New York, United States of America
¤b Current address: Unit of Genetics of Bacterial Genomes, Institut Pasteur, Paris, France
8 2005 12 8 2005 13 7 2005 1 3 e2522 4 2005 1 7 2005 Copyright: © 2005 De Hoon 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 prokaryotes, genes belonging to the same operon are transcribed in a single mRNA molecule. Transcription starts as the RNA polymerase binds to the promoter and continues until it reaches a transcriptional terminator. Some terminators rely on the presence of the Rho protein, whereas others function independently of Rho. Such Rho-independent terminators consist of an inverted repeat followed by a stretch of thymine residues, allowing us to predict their presence directly from the DNA sequence. Unlike in Escherichia coli, the Rho protein is dispensable in Bacillus subtilis, suggesting a limited role for Rho-dependent termination in this organism and possibly in other Firmicutes. We analyzed 463 experimentally known terminating sequences in B. subtilis and found a decision rule to distinguish Rho-independent transcriptional terminators from non-terminating sequences. The decision rule allowed us to find the boundaries of operons in B. subtilis with a sensitivity and specificity of about 94%. Using the same decision rule, we found an average sensitivity of 94% for 57 bacteria belonging to the Firmicutes phylum, and a considerably lower sensitivity for other bacteria. Our analysis shows that Rho-independent termination is dominant for Firmicutes in general, and that the properties of the transcriptional terminators are conserved. Terminator prediction can be used to reliably predict the operon structure in these organisms, even in the absence of experimentally known operons. Genome-wide predictions of Rho-independent terminators for the 57 Firmicutes are available in the Supporting Information section.
Synopsis
In prokaryotes, genes belonging to the same operon are transcribed in a single mRNA molecule. Transcription starts as the RNA polymerase binds to the promoter and continues until it reaches a transcriptional terminator. To understand the gene regulatory network of transcription in bacteria, it is important as a first step to determine the operon structure. In this paper, the authors show that (unlike in Escherichia coli) most terminators in Bacillus subtilis function independently of the terminator protein Rho. As these Rho-independent terminators consist of an inverted repeat followed by a stretch of thymine residues, their presence can be predicted directly from the DNA sequence. The authors derived a decision rule by analyzing experimentally known terminating sequences in B. subtilis, and show that the operon boundaries can be found with a high accuracy (about 94%) in B. subtilis and other Firmicutes, even in the absence of experimentally known operons in the given organism. The properties of the transcriptional terminators are shown to be conserved within the Firmicutes phylum. For bacteria other than Firmicutes, the prediction accuracy is considerably lower, suggesting that Rho-dependent or possibly currently unknown termination mechanisms are important in these organisms.
Citation:De Hoon MJL, Makita Y, Nakai K, Miyano S (2005) Prediction of transcriptional terminators in Bacillus subtilis and related species. PLoS Comp Biol 1(3): e25.
==== Body
Introduction
Since the sequencing of the first bacterial genome, the proteobacterium Haemophilus influenzae [1], the complete genomes of 213 microbial organisms have been sequenced, while the sequencing of many more microbial genomes is under way. The availability of these microbial genomes allows us to predict the function of genes in less-characterized bacteria based on their homology to well-studied organisms, such as Escherichia coli and Bacillus subtilis. Similarly, one may attempt to predict the transcriptional regulation of genes in less-characterized organisms using existing knowledge of gene regulation in E. coli and B. subtilis.
Operons, a group of adjacent genes on the same strand of DNA that are transcribed into a single mRNA molecule, form the basic unit of transcription in prokaryotes. Transcription starts from a promoter upstream of the first gene and continues until the RNA polymerase reaches a terminator structure downstream of the last gene in the operon.
Since genes on the same strand belonging to different operons are separated by a terminator followed by a promoter, the operon structure of a bacterial genome can be predicted by the space in base pairs between the genes [2,3]. However, an analysis of the DBTBS database of transcriptional regulation in B. subtilis [4] revealed that more than 20% of its genes in known polycistronic operons are transcribed from more than one promoter. These additional promoters are often located downstream of the first gene, such that only part of the operon is transcribed from the internal promoter. Similarly, we found that almost 6% of the known polycistronic operons contain an internal read-through terminator, at which partial continuation of transcription occurs. The existence of such internal promoters and terminators complicates the definition of an operon.
Transcriptional units can be defined more precisely by the location of the promoters and the transcriptional terminators. Previously, a prediction of transcriptional units of E. coli by searching for promoters and terminators using hidden Markov models yielded an accuracy of about 60% [5]. In this paper, we consider the prediction of transcriptional terminators in B. subtilis and related species, in particular those of the Firmicutes phylum, to which B. subtilis belongs. This phylum consists of a heterogeneous group of mostly Gram-positive bacteria whose genomes have a low G+C content. Several important disease-causing organisms belong to the Firmicutes phylum, such as Clostridia, Streptococci, Staphylococci, and Mycoplasmas, as well as important industrial microbes such as the Lactobacilli.
In the Gram-negative E. coli, belonging to the phylum of Proteobacteria, transcriptional termination is achieved by Rho-independent terminators, which can function in vitro, and Rho-dependent terminators, which need the protein Rho to be present to be functional [6]. Most bacteria contain a protein homologous to E. coli's Rho; notable exceptions are the Firmicutes
Mycoplasma genitalium and M. pneumoniae,
Streptococcus pneumoniae and S. pyogenes, and Ureaplasma urealyticum, and the cyanobacterium Synechocystis sp. strain PCC6803 [7–9]. However, the relative importance of Rho-dependent termination varies between bacteria. In the proteobacteria E. coli and Rhodobacter sphaeroides, as well as the actinobacterium Micrococcus luteus, Rho is essential [10–12]; in the proteobacterium Caulobacter crescentus, Rho is required for oxidative stress survival [13]. Also, an analysis of the average RNA folding energy near stop codons suggested that no stem-loops are formed in the Firmicutes
Mycoplasma genitalium and M. pneumoniae, the actinobacterium Mycobacterium pneumoniae, the cyanobacterium Synechocystis sp. PCC 6803, the proteobacterium Helicobacter pylori, the spirochetes Treponema pallidum and Borrelia burgdorferi, the aquifica Aquifex aeolicus, and the euryarchaeota Methanococcus jannaschii,
Methanobacterium thermoautotrophicum,
Archaeoglobus fulgidus, and Pyrococcus horikoshii, implying that Rho-independent termination does not play a significant role in these organisms [14]. On the other hand, in the Gram-positive B. subtilis the Rho protein is dispensable [15], suggesting that Rho-independent termination is dominant in this organism. Indeed, the only known case of Rho-dependent termination in B. subtilis is the rho gene itself. Similarly, Rho is not essential for viability or virulence in the Gram-positive bacterium Staphylococcus aureus [9]. Furthermore, Gram-positive bacteria (except for Micrococcus luteus [12]) being resistant to the Rho-inhibiting antibiotic bicyclomycin previously led to the suggestion that Rho is usually dispensable in these bacteria [9].
Rho-independent terminators consist of an inverted repeat in the primary DNA sequence, followed by a short stretch of thymine residues. The inverted repeat gives rise to a stem-loop structure in the transcribed mRNA molecule, which halts the RNA polymerase complex. The decreased binding of the uridine stretch of the nascent RNA polymerase complex to the corresponding adenine stretch in the DNA causes the polymerase to dissociate from the DNA, terminating transcription. Experimentally, the presence of a transcriptional terminator is often established by measuring the mRNA length in a Northern blotting experiment, which is usually not precise enough to determine the exact termination site. In experiments in which the termination site was determined in a primer extension experiment of the 3′ end of the mRNA, termination was shown to occur at or near the T-stretch following the stem-loop. Rho-independent terminators in E. coli can be distinguished reliably from intracistronic sequences and random sequences from the Gibbs free energy of stem-loop formation and the properties of the T-stretch [16,17]. The properties of Rho-dependent terminators are less well-studied.
The apparent dominance of Rho-independent termination in B. subtilis and Staphylococcus aureus and the feasibility of distinguishing Rho-independent terminators from random and intracistronic sequences suggests that prediction of transcriptional terminators may be a reliable method to predict operons in these and related organisms. However, our current knowledge of transcriptional terminators does not suffice for such a prediction. First, it is unclear if the properties of transcriptional terminators as found in E. coli are conserved in other prokaryotes, in particular since E. coli and B. subtilis are evolutionarily distant. Indeed, the few experimentally known terminators in the Gram-positive Streptomyces lividans suggest that there is no need for the stem-loop of a transcriptional terminator to be followed by a T-stretch in this organism [18,19]. Second, we need to make sure that transcriptional terminators can be distinguished from other intercistronic sequences, rather than intracistronic or random sequences, in particular since stem-loop structures in intercistronic sequences may serve other biological functions such as mRNA processing or transcription factor binding. Third, the operon prediction will be reliable only if the fraction of Rho-terminated operons is sufficiently small. We note that previous attempts at terminator prediction in B. subtilis [20] and Synechococcus sp. WH8102 [21,22] for the purpose of operon prediction were unsuccessful.
Here, we analyzed 463 experimentally known terminating sequences in B. subtilis in order to discover their deciding properties. Using a set of 567 experimentally known non-terminating sequences, occurring between genes in the same operon, we derived a decision rule to distinguish between terminating and non-terminating sequences in B. subtilis. We show that this decision rule is also valid for other Firmicutes, which allowed us to reliably predict their operon structure from the DNA sequence, even in the absence of experimentally known operons in these organisms.
Results
Statistical Properties of Rho-Independent Terminators in B. subtilis
We created a set of 463 known terminating sequences and 567 known non-terminating sequences in B. subtilis by collecting experimentally identified operons from the literature. A stem-loop structure followed by a T-stretch, indicative of a Rho-independent terminator, was found in 425 of the 463 terminating sequences. Here, we analyze these transcriptional terminators and compare their statistical properties with those for the evolutionarily distant E. coli, for which 148 proposed and experimentally identified terminators were analyzed previously [16]. The experimentally known operons in B. subtilis, the corresponding terminator sequences, and the supporting experimental evidence are available from the DBTBS database [4], as well as in Datasets S1–S3.
Figure 1 shows the distribution of the Gibbs free energy ΔG of stem-loop formation at 25 °C. The distribution shows a peak around −16 kcal/mole, compared to −14 kcal/mole for E. coli [16]; the extent of the distribution is comparable to E. coli. The distribution of the length of the stem, shown in Figure 2, reveals that the transcriptional terminators in B. subtilis tend to have relatively long stems: 75% of the stems have a length of 9 ± 2 in B. subtilis, whereas in E. coli 75% have a length of 7 ± 2. As a result, the density of the Gibbs free energy, calculated by dividing the free energy by the number of nucleotides in the stem-loop structure, is somewhat lower in B. subtilis than in E. coli (Figure 3). As in E. coli, the high density of the Gibbs free energy is made possible by a high G+C content of 62.4% of the stem (78.2% in E. coli), compared to an average 36.3% G+C fraction in B. subtilis non-coding regions.
Figure 1 Distribution of the Gibbs Free Energy of Stem-Loop Formation
The distribution is calculated from 425 experimentally identified transcriptional terminators in B. subtilis. The dotted curve shows the distribution for E. coli, as calculated from 147 previously collected Rho-independent terminator sequences in this organism [16].
Figure 2 Distribution of the Length of the Stem in Nucleotides
The distribution is calculated from 425 transcriptional terminators in B. subtilis and 147 previously published Rho-independent terminators in E. coli [16].
Figure 3 Distribution of Gibbs Free Energy of Stem-Loop Formation, Divided by the Length of the Stem-Loop Structure in Nucleotides
The distribution is calculated from 425 transcriptional terminators in B. subtilis, and 147 previously published Rho-independent terminators in E. coli [16].
Figure 4 shows the distribution of the number of thymine residues in a 15−base pair T-stretch following the stem-loop. This distribution, with a median of nine thymine residues, is similar to the distribution previously found for E. coli [16], for which the median is equal to ten.
Figure 4 Distribution of the Number of Thymine Residues in the 15 Base Pair T-Stretch following the Stem Loop
The distribution is calculated from 425 transcriptional terminators in B. subtilis, and 147 previously published Rho-independent terminators in E. coli [16].
To characterize the stem-loops of B. subtilis Rho-independent terminators in more detail, we consider the characteristics of the stem and the loop separately. As shown in Figure 5, about 70% of the loops consist of 4 ± 1 nucleotides, compared to 85% for E. coli. Whereas tetranucleotides are most abundant (28% of the total), they are not as ubiquitous as in E. coli, where they represent 55% of the total.
Figure 5 Distribution of the Number of Residues in the Loop of Rho-Independent Terminators
The distribution is calculated from 425 transcriptional terminators in B. subtilis and 147 previously collected Rho-independent terminator sequences in E. coli [16].
The tetranucleotides GAAA and TTCG were especially prominent among E. coli loops [16]. We did not find this tendency for B. subtilis, where TTT, AAT, TGA, and AAAA occurred most often. Also, we did not find evidence that the words GCGGG, GCGGGG, and GGCCC appear most often in the 3′ arm of the stem-loop, as was found for E. coli [16]. Instead, we found the words GGCAG (19 times) and GCAGG and TCCGG (17 times each); GCGGG appeared 11 times, GGCCC once, and GCGGGG not at all. The loop is usually closed by a 5′-C-G-3′ pair (35.5% of the loops in B. subtilis Rho-independent terminators), although not as often as in E. coli, where they constitute 59% of the loop closing pairs.
We note that 82 out of 148 previously analyzed Rho-independent terminators in E. coli were proposed in the literature, but were not experimentally verified [16]. Hence, the set of E. coli terminators may be biased towards more typical cases.
In many terminators, the T-stretch following the stem-loop structure can base-pair to a complementary sequence in front of the stem-loop, suggesting that the T-stretch may form part of the stem-loop structure. This is particularly evident in stem-loop structures that act as transcriptional terminators on both strands of the DNA. Here, the T-stretch can base-pair to an A-stretch in front of the stem-loop, which acts as a T-stretch for transcription in the opposite direction. In the derivation of the decision rule below, we found that allowing the T-stretch to base-pair to the sequence upstream of the stem-loop did not improve the prediction accuracy. In this paper, we therefore limited the stem-loop structure to the sequence upstream of the T-stretch.
Position of Transcriptional Terminators in B. subtilis
As shown in Figure 6, transcriptional terminators are typically located closely downstream of the stop codon of the last gene in the operon and often even partially overlap the gene. Out of 425 experimentally known Rho-independent terminators, 395 are located within 100 base pairs downstream of the stop codon. Of the remaining 30 genes, 14 are immediately followed by a convergently transcribed gene on the opposite strand, such that their 3′ ends are very close to each other or even overlap, leaving little space to fit a Rho-independent transcriptional terminator. As it is difficult to reconcile the requirements of the terminator and the coding region of the downstream gene, the terminator may be located much further downstream in such cases. For example, a Northern blotting experiment [23,24] showed that the terminator of the gene yxlG is located 1,136 base pairs downstream of the stop codon, inside the coding region of the convergently transcribed yxlH gene, whose 3′ end overlaps for eight base pairs with that of yxlG.
Figure 6 Distribution of the Position of B. subtilis Rho-Independent Terminators with Respect to the Stop Codon of the Last Gene in the Operon
The distance between the first nucleotide of the stem-loop and the last nucleotide of the stop codon is shown.
In the genome-wide search for transcriptional terminators in B. subtilis, described below, we found 466 putative terminators located more than 100 base pairs downstream of the stop codon. Of these, 54 were followed by a convergently transcribed gene with less than 20 base pairs between the stop codons.
A large number of base pairs between the stop codon and the terminator may suggest the presence of a currently unidentified open reading frame. Coding regions highly homologous to known or hypothetical proteins in other organisms were found between the stop codon and the putative terminator of the B. subtilis genes yxiT, fliT, ypbS, metA, and ypfD. These coding regions may correspond to currently unidentified genes in B. subtilis, or to genes recently discarded from the B. subtilis genome. In the latter case, the position of the terminator may not have stabilized yet.
Prediction of Transcriptional Terminators in B. subtilis
For E. coli, the following decision rule was derived previously [16]
where T is the score for the thymine stretch, ΔG is the Gibbs free energy of stem-loop formation in kcal/mole, and n
SL is the number of nucleotides in the entire stem-loop structure. The numerical values for the coefficients were found by fitting this equation to maximize the difference between transcriptional terminators and intracistronic sequences.
The T-stretch score T was calculated as follows
where x
0 = 0.9, and xi = 0.9 · xi
−1 if the i
th nucleotide is a thymine, and xi = 0.6 · xi
−1 otherwise [16]. To avoid the usage of ad-hoc parameters, instead we use an exponentially decaying function for the T-stretch
where δi is one if the i
th nucleotide is a thymine, and zero otherwise. The parameter λ is fitted from the experimentally known transcriptional terminators.
Using a logistic regression model to fit this formula to the 463 known terminating sequences and 567 non-terminating sequences in B. subtilis, we found the decision rule
with λ = 0.144. This decision rule resulted in a sensitivity of 93.95% and a specificity of 94.36% in predicting transcriptional terminators in B. subtilis. The previously proposed T-stretch scoring function (Equation 2) resulted in a slightly lower sensitivity and specificity. As in the case of E. coli [16], we found that dividing the Gibbs free energy of stem-loop formation by the length of the stem-loop structure is slightly more accurate than using the Gibbs free energy directly.
As our prediction rule considers only Rho-independent terminators while the training set contains all experimentally known terminating sequences, the high prediction accuracy of about 94% suggests that Rho-dependent terminators account for about 6% or less in B. subtilis. Whereas false negatives may also be due to inaccuracies in the numerical values of the parameters in the decision rule (Equation 4), this is unlikely to play a major role. First, the scores d of the known terminators follow a bell-shaped distribution (Figure S1) with a tail for negative d, such that an imprecision in the numerical values of the decision rule will not strongly affect the accuracy. Second, for most of the false negative predictions we were not able to find any stem-loop structures near the discriminant line d = 0 that might conceivably function as a Rho-independent terminator.
A more likely cause of false negative predictions is the presence of Rho-dependent terminators, as well as imperfections in the list of experimentally known operons. For example, some of the false-positive predictions show very clear terminator structures, which may represent read-through terminators or terminator/anti-terminator structures that have not yet been identified experimentally. Partial continuation of transcription at read-though terminators may be regulated or depend on cellular conditions, and cannot always be detected easily in a given experiment. Furthermore, it is sometimes difficult to determine if the 3′ end of an mRNA molecule is produced by transcriptional termination, by mRNA degradation, or by an mRNA processing event, in particular because mRNA processing sites may also be characterized by stem-loops [25,26].
In some cases, the Gibbs free energy, together with the properties of the T-stretch, does not suffice to predict if a stem-loop terminates transcription. Figure 7 shows the example of the yqfSU operon in B. subtilis. This operon is unusual, as its two genes are separated by a gene (yqfT) that is transcribed in the opposite direction. As yqfT is followed by two genes (yqfS and yqfR) on the opposite strand, we expect yqfT to be transcribed monocistronically. Indeed, we find a strong Rho-independent terminator immediately downstream of yqfT. Except for the loop sequence, the complementary sequence on the opposite strand is identical to the yqfT terminator. However, a Northern blotting experiment [27] revealed that yqfS and yqfU are transcribed together. Hence, the stem-loop structure terminates transcription on the forward strand, but not on the reverse strand, in spite of the similarity of the stem-loop structure and T-stretch on the two strands. Intervening genes on the opposite strand of DNA were also found for the experimentally known operons yflMK (with intervening gene yflL), yfhQ-fabL-sspE (with intervening gene yfhS), and yqxD-dnaG-sigA (with antE overlaying dnaG on the opposite strand).
Figure 7 The yqfSU Operon in B. subtilis Consists of the Two Genes yqfS and yqfU, Separated by the Intervening Gene yqfT, Located on the Opposite Strand
The terminator sequence downstream of yqfT is virtually identical to the complementary sequence on the opposite strand. However, a Northern blotting experiment [27] revealed that the complementary sequence does not act as a transcriptional terminator. Arrows indicate transcription start sites; stem-loops represent transcriptional terminators.
Prediction of Transcriptional Terminators in Other Bacteria
As the decision rule (Equation 4) is quite accurate in predicting transcriptional terminators, and hence transcriptional units, in B. subtilis, the question arises if the same decision rule can be applied to other organisms related to B. subtilis. Very few transcriptional terminators have been identified experimentally in other prokaryotes (except for E. coli), making it difficult to verify their conservation in general, or to assess the accuracy of the decision rule when applied to other organisms. However, genes followed by downstream genes on the opposite strand are very likely the last gene in the (mono- or polycistronic) transcription unit, and must therefore be followed by a transcriptional terminator. Hence, we can create a positive set of transcriptional terminators, even in the absence of experimentally known operons, by collecting all genes in the genome that are followed by genes on the opposite strand. As a few examples exist in which a single gene is located between the genes of an operon on the opposite strand, as shown above, we require that a gene is followed by at least two genes on the opposite strand for inclusion in the positive sample set.
The construction of the positive set of transcriptional terminators is based on the assumption that their properties do not depend on whether the downstream gene is on the same strand or the opposite strand of DNA. Creating such a set of positive examples is more difficult for operon prediction based on the intergenic distance [2,3], which is likely to depend on whether two neighboring genes are on the same strand of DNA.
To assess the sensitivity of terminator prediction in other organisms, we apply the decision rule (Equation 4) to the downstream sequence of all genes in the positive set, and count how often it can detect the presence of a transcriptional terminator. The validity of the decision rule can be verified further by analyzing the properties of the predicted transcriptional terminators.
The value for the sensitivity calculated in this manner depends on both the effectiveness of the decision rule in detecting Rho-independent terminators, and the relative importance of Rho-independent terminators with respect to other (possibly Rho-dependent) mechanisms of transcriptional termination. A high sensitivity indicates that Rho-independent termination is dominant in the organism, and that the decision rule effectively finds the Rho-independent terminators. A low sensitivity can arise if the organism has a large number of transcriptional terminators that are not Rho-independent, or if the decision rule is not a valid description of the Rho-independent terminators in that organism.
The sensitivity of the decision rule (Equation 4) was assessed in the complete genomes of 57 Gram-positive and Gram-negative organisms belonging to the Firmicutes phylum, and 19 Gram-positive and 10 Gram-negative bacteria outside of the Firmicutes phylum. Figure 8 shows that the prediction rule finds more than 90% of the transcriptional terminators for most Firmicutes; on average, the sensitivity is 94.4%. A lower prediction sensitivity, between 80% and 90%, was found for the Bacillaceae Bacillus halodurans, B. clausii, Oceanobacillus iheyensis, Thermoanaerobacter tengcongensis, and Geobacillus kaustophilus. For organisms outside of the Firmicutes phylum, the prediction accuracy is considerably lower.
Figure 8 Sensitivity of Predicting Transcriptional Terminators, Evaluated for 57 Firmicutes and 29 Other Bacterial Species
Firmicutes are shown in dark gray; other bacterial species are shown in light gray. (+), (−), or (0) in front of the organism name denotes that the organism is Gram-positive, Gram-negative, or lacks a cell wall, respectively.
For B. subtilis, the decision rule found a Rho-independent terminator downstream of 97.2% of genes followed by at least two genes on the opposite strand. This indicates an even lower bound on the fraction of genes in B. subtilis whose transcription is terminated by Rho, and suggests that perhaps half of the 6% false positives found above are due to inaccuracies in the set of collected terminating and non-terminating sequences.
Tables S1–S3 show the statistical properties of the predicted terminator sequences in more detail. The transcriptional terminators of Bacillus clausii and Geobacillus kaustophilus are characterized by a slightly lower number of thymine residues in the T-stretch. The other three Firmicutes with lower prediction accuracies have a slightly lower Gibbs free energy density ΔG/L
SL of around −0.54 kcal/mole/nucleotide due to a longer average stem length, particularly for Oceanobacillus iheyensis with an average stem length of 11.5 nucleotides. In comparison, the density of the Gibbs free energy in B. subtilis is −0.644 kcal/mole, with an average stem length of 9.1 nucleotides.
The transcriptional terminators of the Staphylococci, Clostridia, Lactobacilli, Lactococci, and Streptococci also have slightly lower values for the Gibbs free energy density compared to B. subtilis. For Lactococcus lactis subsp. lactis, this is compensated for by a larger number of thymine residues in the T-stretch. Enterococcus faecalis has an average Gibbs free energy density almost equal to that of B. subtilis, in spite of a relatively long stem of 11.9 nucleotides on average.
Modest prediction sensitivities of 90.20% and 82.89% were found for Mycoplasma genitalium and Mycoplasma pneumoniae, respectively, while higher sensitivities were obtained for other Mollicutes, with Mesoplasma florum L1 and Mycoplasma pulmonis UAB CTIP reaching 100%. The transcriptional terminators are characterized by a low Gibbs free energy density of stem-loop formation, ranging from −0.276 kcal/mole/nucleotide for Mycoplasma hyopneumoniae to −0.497 kcal/mole/nucleotide for Mycoplasma gallisepticum, which explains the previous result (based on the RNA folding energy) that Rho-independent termination plays an insignificant role in these organisms [14]. However, as shown in Table S3, a larger number of thymine residues in the T-stretch ensures that the sensitivity of terminator prediction is high in spite of the lower Gibbs free energy density in Mollicutes. In contrast to previous work [14], our results therefore suggest that Rho-independent termination represents the main mode of transcriptional termination in Mollicutes. We must mention, however, that based on the predicted transcriptional terminators we found an unusually high number of genes per operon of 3.8 for Mycoplasma genitalium, compared to around two for other Firmicutes, suggesting that some terminators were missed in the prediction. However, since no Rho protein has been identified in Mycoplasma genitalium [8], these cannot be Rho-dependent terminators, unless perhaps Rho proteins are imported from the host organism (in which Mycoplasmas live) into the bacterial cell.
The high prediction sensitivities found in the Firmicutes suggest that the mechanism of transcriptional termination is conserved in this phylum, both in terms of the validity of the decision rule (Equation 4) to detect Rho-independent terminators, as well as the predominance of Rho-independent transcriptional termination compared to Rho-dependent termination. This allows us to perform an accurate genome-wide prediction of transcriptional terminators, and hence transcriptional units, of the 57 fully sequenced bacterial species of the Firmicutes phylum. The full set of transcriptional terminators predicted in a genome-wide search in these 57 Firmicutes is available in Dataset S4.
As the separation between Gram-positive and Gram-negative bacteria is one of the most basic divisions in the phylogeny of prokaryotes, one may expect that the decision rule (Equation 4) applies to all Gram-positive organisms. However, as shown in Figure 8, the decision rule detects considerably fewer terminators for Gram-positive bacteria outside of the Firmicutes phylum, with sensitivities around 50% for most organisms and as low as 11.8% for Mycobacterium avium subsp. paratuberculosis. As shown in Table S2, the statistical properties of the predicted terminators tend to deviate more than within the Firmicutes group, with longer stems and fewer thymine residues in the T-stretch. The decision rule may therefore be inappropriate for organisms other than Firmicutes.
For comparison, Figure 8 also shows the prediction sensitivity for some well-studied Gram-negative organisms. Except for Helicobacter pylori, with a sensitivity of about 85%, these sensitivities are generally lower than in Firmicutes, ranging from 32% in Caulobacter crescentus to 76% for Aquifex aeolicus. For E. coli, a sensitivity of 67% was found, suggesting a considerable role for Rho-dependent transcriptional termination in this organism.
Perhaps surprisingly, these prediction accuracies suggest that the mechanism of transcriptional termination is conserved in the Firmicutes phylum, but not for Gram-positive bacteria in general. To study the conservation of the properties of Rho-independent transcriptional terminators, we calculated the average Gibbs free energy and the average number of thymine residues in the T-stretch from the predicted terminators in these organisms. Figure 9 shows the position for each organism in this two-dimensional space. The Firmicutes appear in two conserved groups, one consisting of the class of Mollicutes (including the Mycoplasmas) and one consisting of the other classes of the Firmicutes phylum, with on average a higher Gibbs free energy of stem-loop formation. As shown in Figure 9, the transcriptional terminators of the bacterial species other than Firmicutes display a much larger variation in the properties of their stem-loops and T-stretches. Note, however, that these species may contain a large number of transcriptional terminators that are not Rho-dependent, which affects the accuracy with which the properties of the Rho-independent terminators can be calculated.
Figure 9 Average Gibbs Free Energy of Stem-Loop Formation and the Average Number of Thymine Residues in the T-Stretch
These are calculated from the predicted Rho-independent terminators in the 82 bacterial species we consider. Circles represent organisms belonging to the Firmicutes phylum; crosses represent other bacterial species.
Discussion
As the properties of Rho-independent terminators and the predominance of Rho-independent termination are well conserved within the Firmicutes phylum, computationally locating transcriptional terminators from the DNA sequence can be an accurate method to predict the operon structure in these organisms. This is particularly useful for organisms with few experimentally known operons, for which more traditional learning methods cannot be used. We predicted the transcriptional terminators, and hence the operon structure, for 57 prokaryotes belonging to the Firmicutes phylum, with an expected sensitivity and specificity of at least 94%.
The comparison of the terminator structures predicted in Firmicutes shows that they are conserved as two nearby groups, one corresponding to the Mollicutes class (lacking a cell wall), and one corresponding to the other (Gram-positive) classes of Firmicutes. From our terminator prediction results in prokaryotes other than Firmicutes, we find that the properties of Rho-independent transcriptional terminators are not conserved in bacteria in general, or even in Gram-positive bacteria. In addition, the low prediction sensitivities suggest that other (possibly Rho-dependent) modes of transcriptional termination also play an important role in prokaryotes outside of the Firmicutes phylum. However, if Rho is not essential in Gram-positive bacteria in general, as suggested previously [9], then a currently unknown mechanism of transcriptional termination must be present in Gram-positive bacteria other than Firmicutes. Also for the cyanobacterium Synechocystis sp. strain PCC6803, which apparently lacks the Rho protein altogether [7], we expect the existence of alternative termination mechanisms, as the prediction of Rho-independent transcriptional terminators showed a sensitivity of only 61%.
While Rho-independent termination is dominant in B. subtilis, we were not able to find a clear transcriptional terminator for 28 experimentally known operons as well as 15 other genes followed by two or more genes on the opposite strand. These operons may therefore be followed by a Rho-dependent terminator. The list of experimentally known operons in B. subtilis together with the set of predicted transcriptional terminators, both of which are available in the Supporting Information, may therefore contribute to further research of Rho-dependent and other mechanisms of transcriptional termination.
Materials and Methods
Calculation of the RNA secondary structure was done with Mfold [28,29]. To calculate the Gibbs free energy of formation for the terminator stem-loop structure, we first searched the DNA sequence downstream of a gene to find candidate thymine stretches with a length of 15 base pairs. We required the T-stretch to start with at least two thymine residues. Relaxing this condition did not improve the prediction accuracy, but it significantly increased the running time, which can be more than a day for a complete genome. T-stretches with less than three thymine residues, as well as T-stretches with a T-score of less than 2.5, were ignored, as such weak T-stretches are unlikely to function as a terminator even if preceded by a very strong stem-loop. We then let Mfold calculate the RNA secondary structure of the upstream 75 base pairs at a temperature of 25 °C. Usually, Mfold finds more than one stem-loop or other secondary structures in this sequence. If so, we analyzed the output of Mfold to find the first paired nucleotide, and removed it together with the nucleotides further upstream. We then recalculated the secondary structures in this shortened RNA stretch, and repeated until a single stem-loop structure was found. We allowed up to three consecutive nucleotides in the stem-loop to be mismatched or unmatched, as well as a gap of at most three nucleotides between the last base-paired nucleotide in the stem-loop and the start of the T-stretch.
As soon as the sequence upstream of the T-stretch folded as a single stem-loop structure, we found its Gibbs free energy −ΔG and applied the decision rule (Equation 4) to determine if it qualified as a Rho-independent transcriptional terminator. If so, the terminator structure and the corresponding Gibbs free energy were saved. If not, we continued with the next candidate T-stretch.
The sequence to be searched for the existence of a transcriptional terminator depends on the position of neighboring genes. We started the search at 25 base pairs upstream of the stop codon, as the stem-loop often partially overlaps the coding region of the gene. The search continued until 500 base pairs downstream of the stop codon. However, if the gene was followed by another gene on the same DNA strand, the search sequence ended at its start codon. The search sequence we chose was considerably larger than in previous studies, as we found that transcriptional terminators located far downstream of the stop codon are not uncommon, while the false-positive rate of terminator prediction is quite low.
To find the decision rule (Equation 4), we searched the downstream sequence of 463 known terminating sequences and 567 known non-terminating sequences, retaining all terminator-like structures for each sequence and the corresponding Gibbs free energy ΔG. We used the Newton-Raphson method to iteratively update the decision rule, for each sequence choosing the highest-scoring terminator-like structure among the candidates during the iteration, until no further improvement could be achieved. The iteration started from a random initial score for the numerical parameters in Equation 4; the final values for the parameters were found to be independent of the initial random choice. The parameter λ, representing the fall-off in the scoring function for the T-stretch, was determined from the positive examples only.
Finding the appropriate sections of the DNA sequence, running the Mfold program, analyzing its results, and running the Newton-Raphson iteration was automated using the Python scripting language [30]. The Python script is available upon request from the author.
The analysis of transcriptional terminators in E. coli is based on a previously collected set of 148 Rho-independent terminators in that organism [16]. As one of the terminator sequences appears to be missing from the published list, the analysis presented here is based on 147 terminator sequences only.
Supporting Information
Dataset S1 List of Experimentally Identified Operons in B. subtilis, Annotated Manually
Each row describes one operon, with the following columns:
The genes the operon consists of
The operon name
The position of the Rho-independent transcriptional terminator, if present
The Gibbs free energy of stem-loop formation, in kcal/mole, evaluated at 25 °C using Mfold
The RNA sequence of the terminator structure, with <left>, </left> and <right>, </right> delimiting the left and right arm of the stem-loop
The experimental evidence for the operon
A list of literature references with PubMed numbers
Optionally, some additional comments on the operon structure or the experimental evidence
This information is also available through the DBTBS database of transcriptional regulation in B. subtilis (http://dbtbs.hgc.jp).
(254 KB TXT)
Click here for additional data file.
Dataset S2 List of Genes and Experimentally Known Operons in B. subtilis, in a Machine-Readable Form
(220 KB TXT)
Click here for additional data file.
Dataset S3 List of 425 Rho-Independent Terminators, Deduced from the List of Experimentally Known Operons in B. subtilis
Each row in this list contains the following columns:
The operon name
Last gene in the operon
Position of the first nucleotide in the stem-loop structure with respect to the stop codon of the last gene in the operon
The Gibbs free energy of stem-loop formation, in kcal/mole, evaluated at 25 °C using Mfold
The RNA sequence of the terminator structure, with <left>, </left> and <right>, </right> delimiting the left and right arm of the stem-loop
(39 KB TXT)
Click here for additional data file.
Dataset S4 The Predicted Transcriptional Terminators in 57 Bacterial Organisms of the Firmicutes Phylum
(2.4 MB TGZ)
Click here for additional data file.
Table S1 Terminator Prediction in Bacterial Species in the Firmicutes Phylum
(5 KB TXT)
Click here for additional data file.
Table S2 Terminator Prediction in Gram-Positive Bacteria other than Firmicutes
(2 KB TXT)
Click here for additional data file.
Table S3 Terminator Prediction in Gram-Negative Bacteria outside the Firmicutes Phylum
(1 KB TXT)
Click here for additional data file.
Figure S1 Distribution of the Discriminant Scores d as Calculated from the Decision Rule for Known Terminators in B. subtilis
(197 KB EPS)
Click here for additional data file.
This work was supported by the Bioinformatics Joint Project “Education and Research Organization for Genome Information Science” with support from MEXT (Ministry of Education, Culture, Sports, Science and Technology of Japan), and a Grant-in-Aid for Scientific Research on Priority Areas (C) “Genome Information Science” by MEXT (Ministry of Education, Culture, Sports, Science and Technology of Japan). YM was supported by a fellowship of the Japan Society for the Promotion of Science.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. MJLdH and YM analyzed the data. MJLdH and KN wrote the paper.
A previous version of the article appeared as an Early Online Release on July 13, 2005 (DOI: 10.1371/journal.pcbi.0010025.eor).
==== Refs
References
Fleischmann RD Adams MD White O Clayton RA Kirkness EF 1995 Whole-genome random sequencing and assembly of Haemophilus influenzae Rd Science 269 496 512 7542800
Price MN Huang KH Alm EJ Arkin AP 2005 A novel method for accurate operon predictions in all sequenced prokaryotes Nucleic Acids Res 33 880 892 15701760
Moreno-Hagelsieb G Collado-Vides J 2002 A powerful non-homology method for the prediction of operons in prokaryotes Bioinformatics 18 S329 S336 12169563
Makita Y Nakao M Ogasawara N Nakai K 2004 DBTBS: Database of transcriptional regulation in Bacillus subtilis and its contribution to comparative genomics Nucleic Acids Res 32 D75 D77 Available: http://dbtbs.hgc.jp . Accessed 12 July 2005. 14681362
Yada T Nakao M Totoki Y Nakai K 1999 Modeling and predicting transcriptional units of Escherichia coli genes using hidden Markov models Bioinformatics 15 987 993 10745988
Reynolds R Bermúdez-Cruz RM Chamberlin MJ 1992 Parameters affecting transcription termination by Escherichia coli RNA polymerase J Mol Biol 224 31 51 1372365
Opperman T Richardson JP 1994 Phylogenetic analysis of sequences from diverse bacteria with homology to the Escherichia coli
rho gene J Bacteriol 176 5033 5043 8051015
Fraser CM Gocayne JD White O Adams MD Clayton RA 1995 The minimal gene complement of Mycoplasma genitalium
Science 270 397 403 7569993
Washburn RS Marra A Bryant AP Rosenberg M Gentry DR 2001
rho is not essential for viability or virulence in Staphylococcus aureus
Antimicrob Agents Chemother 45 1099 1103 11257021
Das A Court D Adhya S 1976 Isolation and characterization of conditional lethal mutants of Escherichia coli defective in transcription termination factor Rho Proc Natl Acad Sci U S A 73 1959 1963 132662
Gomelsky M Kaplan S 1996 The Rhodobacter sphaeroides 2.4.1 rho gene: Expression and genetic analysis of structure and function J Bacteriol 178 1946 1954 8606169
Nowatzke WL Keller E Koch G Richardson JP 1997 Transcription termination factor Rho is essential for Micrococcus luteus
J Bacteriol 179 5238 5240 9260971
Italiani VC Zuleta LF Marques MV 2002 The transcription termination factor Rho is required for oxidative stress survival in Caulobacter crescentus
Mol Microbiol 44 181 194 11967078
Washio T Sasayama J Tomita M 1998 Analysis of complete genomes suggests that many prokaryotes do not rely on hairpin formation in transcription termination Nucleic Acids Res 26 5456 5463 9826772
Ingham CJ Dennis J Furneaux PA 1999 Autogenous regulation of transcription termination factor Rho and the requirement for Nus factors in Bacillus subtilis
Mol Microbiol 31 651 663 10027981
d'Aubenton Carafa Y, Brody E, Thermes C 1990 Prediction of Rho-independent Escherichia coli transcription terminators J Mol Biol 216 835 858 1702475
Ermolaeva MD Khalak HG White O Smith HO Salzberg SL 2000 Prediction of transcription terminators in bacterial genomes J Mol Biol 301 27 33 10926490
Deng ZX Kieser T Hopwood DA 1987 Activity of a Streptomyces transcriptional terminator in Escherichia coli
Nucleic Acids Res 15 2665 2675 3031607
Neal RJ Chater KF 1991 Bidirectional promoter and terminator regions bracket mmr , a resistance gene embedded in the Streptomyces coelicolor A3(2) gene cluster encoding methylenomycin production Gene 100 75 83 2055482
De Hoon MJL Imoto S Kobayashi K Ogasawara N Miyano S 2004 Predicting the operon structure of Bacillus subtilis using operon length, intergene distance, and gene expression information Proc Pac Symp Biocomput 9 276 287
Chen X, Su Z, Dam P, Palenik B, Xu Y, et al 2004 Operon prediction by comparative genomics: An application to the Synechococcus sp. WH8102 genome Nucleic Acids Res 32 2147 2157 15096577
Chen X Su Z Xu Y Jiang T 2004 Computational prediction of operons in Synechococcus sp. WH8102 Genome Inform Ser Workshop Genome Inform 15 211 212
Tojo S Matsunaga M Matsumoto T Kang CM Yamaguchi H 2003 Organization and expression of the Bacillus subtilis
sigY operon J Biochem (Tokyo) 134 935 946 14769884
Yoshida K Ishio I Nagakawa E Yamamoto Y Yamamoto M 2000 Systematic study of gene expression and transcription organization in the gntZ-ywaA region of the Bacillus subtilis genome Microbiology 146 573 579 10746760
Meinken C Blencke HM Ludwig H Stülke J 2003 Expression of the glycolytic gapA operon in Bacillus subtilis : Differential syntheses of proteins encoded by the operon Microbiology 149 751 761 12634343
Mäder U Hennig S Hecker M Homuth G 2004 Transcriptional organization and posttranscriptional regulation of the Bacillus subtilis branched-chain amino acid biosynthesis genes J Bacteriol 186 2240 2252 15060025
Urtiz-Estrada N Salas-Pacheco JM Yasbin RE Pedraza-Reyes M 2003 Forespore-specific expression of Bacillus subtilis
yqfS , which encodes Type IV apurinic/apyrimidinic endonuclease, a component of the base excision repair pathway J Bacteriol 185 340 348 12486072
Zuker M 2003 Mfold web server for nucleic acid folding and hybridization prediction Nucleic Acids Res 31 3406 3415 12824337
Mathews DH Sabina J Zuker M Turner DH 1999 Expanded sequence dependence of thermodynamic parameters improves prediction of RNA secondary structure J Mol Biol 288 911 940 10329189
Van Rossum G Drake FL Jr. 2003 An introduction to Python Bristol (United Kingdom) Network Theory Ltd 120 p.
|
16110342
|
PMC1187862
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Aug 12; 1(3):e25
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010025
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 1611034310.1371/journal.pcbi.001002605-PLCB-RA-0003R3plcb-01-03-02Research ArticleBioinformatics - Computational BiologyCell BiologyEvolutionSystems BiologyYeast and FungiComparative Genomics and Disorder Prediction Identify Biologically Relevant SH3 Protein Interactions Biologically Relevant SH3 Protein InteractionsBeltrao Pedro *Serrano Luis EMBL Structural and Computational Biology, Heidelberg, GermanyGerstein Mark EditorYale University, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 12 8 2005 13 7 2005 1 3 e2611 1 2005 5 7 2005 Copyright: © 2005 Beltrao and Serrano.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.Protein interaction networks are an important part of the post-genomic effort to integrate a part-list view of the cell into system-level understanding. Using a set of 11 yeast genomes we show that combining comparative genomics and secondary structure information greatly increases consensus-based prediction of SH3 targets. Benchmarking of our method against positive and negative standards gave 83% accuracy with 26% coverage. The concept of an optimal divergence time for effective comparative genomics studies was analyzed, demonstrating that genomes of species that diverged very recently from Saccharomyces cerevisiae
(S. mikatae, S. bayanus, and S. paradoxus), or a long time ago (Neurospora crassa and Schizosaccharomyces pombe), contain less information for accurate prediction of SH3 targets than species within the optimal divergence time proposed. We also show here that intrinsically disordered SH3 domain targets are more probable sites of interaction than equivalent sites within ordered regions. Our findings highlight several novel S. cerevisiae SH3 protein interactions, the value of selection of optimal divergence times in comparative genomics studies, and the importance of intrinsic disorder for protein interactions. Based on our results we propose novel roles for the S. cerevisiae proteins Abp1p in endocytosis and Hse1p in endosome protein sorting.
Synopsis
How can we tackle the complexity of a living cell? It is commonly said that living organisms are complex and display “emergent” properties. Emergence is perceived in this context as behaviors that appear at the system level but are not observable at the level of the system's components. In the cell this would be equivalent to saying that the cellular complexity could be explained if we could understand the interplay between the cellular components: that is, not just describe the “parts” that make up a cell but understand how they interact with each other to perform the necessary tasks.
A big step on the road to understanding cellular complexity will be a complete list of all relevant interactions between the cellular components. Although a lot of progress as been made in this direction, we are often dependent on experimental methods that are costly and time consuming. It's a big challenge for computational biology to process the current available knowledge and to propose new ways of predicting the interactions between cellular components.
Here the researchers studied protein interactions that are mediated by small linear peptide motifs,specifically interactions between a protein's SH3 domain and its targets, usually small peptide stretches containing a PXXP motif (where P is proline and X is any amino acid). The results showed that the putative target motifs that are conserved in ortholog proteins and are within regions that do not have a defined secondary structure are more likely to be relevant binding sites. Besides proposing a way to combine secondary structure information with comparative genomics to predict protein–protein interactions, the researchers highlight a possible role of intrinsically disordered proteins in SH3 protein interactions. The results also show that when looking for conservation of these motifs, it is important to carefully select the species used in the study: comparisons between species that have diverged to a certain extent—not too little and not too much—are the most informative.
Citation:Beltrao P, Serrano L (2005) Comparative genomics and disorder prediction identify biologically relevant SH3 protein interactions. PLoS Comp Biol 1(3): e26.
==== Body
Introduction
Important advances have been made in using computational methods to mine the ever-growing quantity of experimental results in order to derive predictions of protein–protein interactions. For such interactions there are methods that explore sequence and structure analysis, like gene fusion [1,2], gene order [3], phylogenetic profiling [4–7], correlated mutations [8,9] and multimeric threading [10,11]. It as also been shown that it is possible to combine different experimental and functional data to predict protein interactions, especially when weighted using Bayesian networks [12]. The accumulation of validated interactions can also be mined by interlog mapping in order to transfer protein interaction annotations across species [13,14].
The work described here deals with the prediction of protein interactions mediated by recognition modules that target small linear motifs [15,16] and more specifically interactions involving SH3 domains. This type of asymmetric binding between globular domains and linear peptides was first reported in the work on Src kinase [17–20], and many other domains have now been shown to have similar properties [15,16]. In a previous study [21], knowledge from phage display experiments was used to derive a position-specific scoring matrix (PSSM) for particular SH3 domains, which was then used to predict putative target ligands. Later, Tong et al. devised a strategy where two-hybrid screening and PSSM were combined to derive a high-confidence network [22]. It was reasoned that an interaction identified by two-hybrid screening was more likely to be biologically relevant if the target protein had a high-scoring linear peptide according to the PSSM of the bait SH3 domain.
In this work we set out to obtain a high-confidence, biologically relevant protein interaction network, starting from the consensus information and using computational methods. The study showed that it is possible to greatly increase the accuracy of consensus-based predictions of protein–linear sequence interactions by taking into consideration the fact that biologically relevant target ligands of SH3 domains are more likely to be within disordered regions and conserved in orthologs. The method's performance was improved by selection of species within an optimal divergence time from the species of interest.
It has been proposed that intrinsic disorder may play a role in protein interactions [23–26], and there are documented cases where binding is coupled to folding [27,28] (reviewed in [29]). It has also been observed that small linear motifs tend to accumulate in protein regions predicted to be intrinsically disordered [30] and that proline-rich regions are usually devoid of secondary structure [31]. In most structures that we are aware of, the SH3 domain is in complex only with short target peptides, and not with full proteins. In all cases the ligands adopt a nonregular secondary structure, but there is little information one can take from these, in respect to the order/disorder of target sites in the context of the whole target protein. Although there is currently no experimental evidence to support that the SH3 domains preferentially bind to intrinsically disordered regions, the results presented here show that binding motifs within disordered protein regions are more likely to be biologically relevant binding sites than equivalent sites within ordered regions. We use the method developed to suggest novel SH3 interactions for Saccharomyces cerevisiae and provide information about the binding sites within the target proteins.
Results/Discussion
Identification and Conservation of SH3 Domains and Selection of Genomes
Using profile hidden Markov models (see Materials and Methods; Figure 1), all putative SH3 domains, and their key binding positions (see Materials and Methods) were determined in S. cerevisiae and in a set of thirteen yeast species: Candida glabrata, Debaryomyces hansenii, Kluyveromyces lactis, Yarrowia lipolytica [32], C. albicans [33], S. paradoxus, S. bayanus, S. mikatae [34], S. castellii, S. kudriavzevii,
S. kluyveri [35], Neurospora crassa [36], and Schizosaccharomyces pombe [37].
Figure 1 Conservation Study of the SH3 Domains of S. cerevisiae in Ten Other Yeast Genomes
CD, conserved domain (the SH3-containing protein has an ortholog and the ortholog SH3 domain is possibly conserved, i.e., less than three conservative changes and no nonconservative changes in the binding positions); DD, divergent domain (SH3-containing protein has an ortholog in this genome but the domain is not on the same branch of the phylogenetic tree); NO, no ortholog (no ortholog found for SH3-containing protein in a particular genome); PD, possibly divergent (SH3-containing protein has an ortholog in this genome but the ortholog SH3 domain has at least one nonconservative change in the binding positions or more than two conservative changes in the binding positions).
In S. castellii,
S. kluyveri, and S. kudriavzevii, no orthologs for the majority of the S. cerevisiae SH3 domains could be identified (results not shown). However, these genomes had only been sequenced with a 2- to 3-fold coverage [35], which may have led to some genomic regions being poorly sequenced. As a result of this, these three genomes were not included in our work.
The ortholog SH3 domains were split into three groups: conserved domain, possibly divergent (if the putative ortholog SH3 domain was in the same branch of the phylogenetic tree and had more than two conservative changes in the binding positions; see Materials and Methods), or divergent domain (if the putative ortholog SH3 domain was not in the same branch of the phylogenetic tree) (Figure 1). As expected, the percentage of conserved domains was higher in genomes of species that had diverged recently from S. cerevisiae.
Intuitively we can expect that there will be an optimal divergence time for the species used in a particular comparative genomic study. In recently diverged species, most protein sequence is conserved and the statistical power for comparative genomics of biological features is therefore smaller. Interspecies conservation becomes less meaningful in a background of low evolutionary divergence. On the other hand, finding a conserved consensus in a very divergent genome might be more significant but only if there was no major change in the specificity of the domain. This change will be more probable the more divergent the species is from the species of interest.
To test the improvement of consensus-based predictions with a comparative genomics approach, an initial set of genomes was chosen based on the conservation analysis of the SH3 domains across the different yeast species (Figure 1). N. crassa and Sch. pombe were excluded because the SH3 domains in these two species might be too divergent to observe conservation of the S. cerevisiae motifs. Very close relatives (S. paradoxus, S. bayanus, and S. mikatae) were excluded as these species would have lower statistical power. Therefore, the first group analyzed consisted of five yeast genomes that broadly covered the hemiascomycete phylum, containing the four recently reported genomes of C. glabrata, D. hansenii, K. lactis, and Y. lipolytica that we grouped with the C. albicans genome.
Evaluation of the “Conservation” Approach
To evaluate the predictive power of our method, two positive datasets, containing experimentally verified SH3–linear peptide interactions, and one negative dataset, containing noninteracting protein pairs were defined (see Materials and Methods). The binding motifs of the SH3 domains of S. cerevisiae included in the two sets of positive standards (15 SH3 domains in the gold set and ten in the platinum set) were taken from the data published by Tong et al. [22]. Table 1 shows the consensus sequence used in the study and also, for each SH3 domain, the total number of peptides found matching this sequence in the S. cerevisiae proteome. From this a measure of accuracy and coverage (see Materials and Methods) based on the positive and negative datasets was calculated. For simple pattern matching of consensus sequence, the accuracy (defined as TP/[TP + FP], where TP indicates true positives and FP indicates false positives) for predicting protein interaction was 12% and the coverage (defined as TP/P, where P indicates all positives) was 92% when using the gold positives set (see Figure 2A).
Table 1 SH3 Consensus Sequence Information
From the SH3 domains in [22], we obtained the consensus sequences from the phage display data, and counted the number of pattern matches found in S. cerevisiae proteins with at least one putative ortholog in the other ten yeast genomes considered in our study.
Figure 2 Size of Probing Window When Looking for Conservation of the Consensus Sequence in Orthologs of the Putative Target Protein
We defined the conservation score as simply the number of species where the consensus sequence is conserved. With this information the accuracy and coverage were calculated, with the gold (A) and platinum (B) positive sets, for consensus sequence conserved in different numbers of species and for different sizes of the probing region.
Using T-Coffee [38], an alignment of all putative orthologs (obtained using the BLAST reciprocal best hit method [39]) of S. cerevisiae proteins containing sequences matching a consensus sequence for an SH3 domain was carried out. This alignment was then used to determine the level of conservation of putative target ligand sites by searching for sequences matching the same consensus sequence in the orthologs. We did not search for conservation of putative target motifs in genomes without an ortholog for the domain under consideration. If there is no ortholog SH3 domain in the comparing species then the conservation of the motif in the ortholog of the putative target is not biologically relevant and should not be counted to increase our confidence in the putative interaction. Having said this, it should be noted that there could be several technical reasons why the ortholog of an SH3 domain could be missed in a genome. There might be errors in the genome assembly, genome annotations, domain annotation, or ortholog assignment. Thus, we also tried to calculate conservation scores without disregarding genomes with no ortholog for the domain under consideration. While this did not change the results significantly (data not shown), we felt that the first approach was more stringent.
In the orthologs, the search was restricted to a window surrounding the putative target ligand in the S. cerevisiae sequence, and we called this the probing region. In Figure 2, accuracy versus coverage for increasing probing regions is plotted, and it can be seen that by searching in a wider region of the alignment both coverage and accuracy are increased, especially for higher conservation scores (the complete analysis with the number of hits and false and true positives for each positive set is given in Table S1). Optimal results were obtained using a probing region of 210 alignment positions. It is important to emphasize that these were not necessarily amino acids, but 100 gaps or amino acids on each side of a motif of ten amino acids. This result could be due to poor alignment of some proteins, especially those rich in proline sequences. In fact most of the gain in coverage was due to interactions with proline-rich proteins that were difficult to align and had multiple gaps (i.e., Las17p, App1p, and Vrp1p). Also, these data may suggest that these small target ligands may be easily moved in primary sequence space during evolution, owing to compensatory mutations in proteins that are already proline-rich in nature. For both sets of positives a big improvement in accuracy was observed when we selected for consensus sequence conserved in the five genomes used (3.8-fold increase with the gold positives and 3.3-fold increase with the platinum positives). There was, however, a similar fold reduction in the coverage, 3-fold for the gold and 4.3-fold for the platinum set.
Since most known target proteins in the SH3 interaction network are proline-rich and a large probing window was used, it is possible that the hits found in orthologs were due to chance and lacked biological meaning. To eliminate this possibility two “decoy” proline-rich patterns were analyzed: PXXXPXXXP and EXXPXXP (where X is any amino acid), different from the consensus sequences. Both patterns were found with high frequency (>400 hits) on S. cerevisiae proteins. Using these two patterns, a loss in accuracy and coverage was observed (an average of 1.4 times less accuracy and 1.2 times less coverage for the PXXXPXXXP motif and an average of 3.4 times less accuracy and 2.5 times less coverage for the EXXPXXP motif). Thus, we can rule out the possibility that the results were generated by chance and can confirm that the observed phenomenon was the conservation of specific SH3 binding motifs and not of proline-rich tracks.
However, the accuracy obtained with conservation alone was still poor (using the gold set, accuracy = 46% and coverage = 31%, and using the platinum set, accuracy = 30% and coverage = 16%). A hypergeometric test allowed us to say that that the improvement in both positive sets and for all conservation scores was significant (p < 0.05) and not due to random sampling.
Combining Comparative Genomics and Disorder Prediction
Since SH3 domains generally bind linear amino acid stretches, we tried to improve the accuracy of our consensus-based method by extracting secondary structure information about the sequences containing the target motifs. It has been argued that there might be biological advantages in presenting binding sites within unstructured regions [23–26]. It has also been observed that small linear motifs tend to accumulate in protein regions predicted to be intrinsically disordered [30] and that proline-rich regions are usually devoid of secondary structure [31]. To our knowledge there is no clear experimental evidence to support that SH3 domain target sites are generally unstructured before binding, but since SH3 domains bind small linear peptide motifs that are proline-rich, we hypothesized that SH3 domain targets might be mainly found in unstructured regions of the polypeptide chain. Therefore we used GlobPlot [30] in combination with coil-region predictions [40] to identify and study all consensus sequences found within disordered protein regions.
Combining disorder prediction with comparative genomics resulted in a significant (p < 0.01, using a hypergeometric test) increase in the accuracy of protein target prediction (there was a 2-fold average increase in both sets) (Figure 3). The decrease in coverage was 1.4-fold for the gold and 1.1-fold for the platinum set. For consensus sequence conserved in five or more genomes, we obtained 94% accuracy with 28% coverage for the gold set. For consensus sequence conserved in four or more genomes, we obtained 83% accuracy with 26% recovery for the platinum set. These results argue that intrinsic disorder plays an important role in SH3 protein interactions; however, further experimental work is needed to verify this observation.
Figure 3 Combining Conservation and Secondary Structure Prediction
We calculated, with the gold (A) and platinum (B) positive sets, the accuracy and coverage for target prediction when including or excluding secondary structure information. We used a probing region of 210 alignment positions in this analysis.
Since the platinum positive set was independent (see Materials and Methods), the values obtained with this set may be used as a score for the performance of our method compared to others. Higher values of coverage and accuracy with the gold positive set were observed when using our method, but it should be noted that this could be due to a possible bias (see Materials and Methods). A detailed record of the number of hits and false and true positives for each conservation level in both positive sets can be found in Table S2.
Using the methods described in this work, we show proof of concept on how to integrate secondary structure prediction with comparative genomics to increase the accuracy of consensus-based prediction of peptide recognition modules. However, the method employed involves a clear trade-off between accuracy and coverage.
Of the 59 interactions in the final high-confidence interaction presented by Tong et al. [22], the method was able to predict 20 interactions when restricting for consensus sequences within disorder and found in four of the five genomes used. We tried to look for distinguishing features within these 20 interactions, compared to the remaining 39 that the method did not predict. There were no statistical differences in the average size of protein targets (p = 0.32 with a t-test), average proline content of protein targets (p = 0.12 with a t-test), usage of Class II motif (p = 0.21 with a hypergeometric distribution test), or conservation of SH3 domain (p = 0.82 with a t-test). There was a statistically significant difference in the average conservation of the target proteins (p = 0.03 with a t-test). The protein targets the method was able to predict were on average conserved in 8.7 of the ten species, while the targets not recovered were conserved in 7.6 species. This small but significant difference highlights the bias this method has for conserved interactions. A higher level of confidence can be placed in any putative target motif found conserved in most yeast species analyzed, but this level of conservation will only happen for essential interactions. It is important to note that for this reason this method will always miss species-specific protein interactions. However, adding more genomes of species within an appropriate divergence time should alleviate this problem, a concept discussed in more detail below. Another possible cause of loss in coverage could be interactions that are mediated by currently uncharacterized motifs or through noncanonical SH3 binding (i.e., through globular regions of the target protein).
As shown by other authors (reviewed in [41]), it should be possible to further improve the reliably of a protein interaction network, and therefore our method, by adding information from other sources of data (i.e., RNA expression, and essentiality and function information). This is especially true if the information is efficiently combined, e.g., employing a Bayesian network [12]. It was our intention to develop a method that could be used in species where these sources of information were not available, but in the future we will try to develop weighting schemes to include such sources for prediction of interactions mediated by small linear motifs.
Determining an Optimal Divergence Time for the Genomes Used When Searching for Conservation of Target Ligands of SH3 Domains
Included in our initial hypothesis was the notion that there might be an optimal time of divergence to efficiently use the comparative genomics approach. To test this, phylogenetic data [32,42,43] with approximate values for the divergence times of the yeast species from S. cerevisiae (see Materials and Methods) were used to create seven groups of four genomes with increasing average divergence time from S. cerevisiae. Using the gold positives, the highest accuracy obtained for a small range of coverage values was determined for each of these groups. For different coverage ranges the highest accuracy was generally obtained with groups of genomes that had diverged from S. cerevisiae on average around 400–950 million years (My) (Figure 4).
Figure 4 Optimal Divergence Time to Search for Conservation of Target Motif of SH3 Domains
We designated seven groups of species with an increasing average divergence time from S. cerevisiae and calculated for each group the highest accuracy obtained for restricted windows of coverage. We used the gold positive and the negative set to calculate the accuracy and coverage (see Materials and Methods). The seven groups of species are as follows: (1) S. bayanus, S. paradoxus, S. mikatae, and C. glabrata (average divergence of 112.5 My from S. cerevisiae); (2) S. paradoxus, S. mikatae, C. glabrata, and K. lactis (average divergence of 200 My from S. cerevisiae); (3) S. mikatae, C. glabrata, K. lactis, and C. albicans (average divergence of 387.5 My from S. cerevisiae); (4) C. glabrata, K. lactis, C. albicans, and D. hansenii (average divergence of 575 My from S. cerevisiae); (5) K. lactis, C. albicans, D. hansenii, and Y. lipolytica (average divergence of 725 My from S. cerevisiae); (6) C. albicans, D. hansenii, Y. lipolytica, and N. crassa (average divergence of 875 My from S. cerevisiae); and (7) D. hansenii, Y. lipolytica, N. crassa, and Sch. pombe (average divergence of 950 My from S. cerevisiae). The individual values for the divergence time from S. cerevisiae were taken from the literature [32,42,43]. Although we tried to create groups that would not have genomes of species with very different separation dates from S. cerevisiae, it should be noted that because of the small number of available genomes, the groups are not homogenous. Also, the values of the divergence time of each species were not always obtained with the same method. Therefore, this range of values should be viewed critically.
To explore this issue further we tried to find out which genomes might be more or less informative for our consensus-based predictions. For each possible combination of two or more genomes we calculated the highest accuracy obtained for 11 small windows of coverage (with intervals of 5% of coverage from 15% to 70%). Figure 5 shows the average of the individual genome representations in all possible groups, in the groups scoring in the highest 20% accuracies and in the groups scoring within the lowest 20% accuracies, over all the coverage windows studied. For each species, a t-test determination was carried out to see whether the average frequencies within the highest and lowest combinations were significantly different from the frequency in all possible combinations. From the analysis of the results the more informative genomes are C. albicans, D. hansenii, C. glabrata,
K. lactis, and Y. lypolytica. We can also see that N. crassa and Sch. pombe are not over-represented in the highest scoring groups, suggesting that they have less informative genomes. More importantly, it is clear that including the genomes of S. bayanus,
S. mikatae, or S. paradoxus leads to a decrease in the accuracy of predictions. These observations correlate well with the degree of divergence observed for the SH3 domains (see Figure 1) and with our proposed range for optimal divergence time.
Figure 5 Most Informative Genomes in the Search for Conservation of Target Motif of SH3 Domains
We created all possible combinations of two or more genomes of our set of ten genomes. For each combination we calculated the highest accuracy obtained for 11 windows of coverage from 15% to 70% at intervals of 5%. We then calculated the average frequency, over all coverage windows, of each individual species in all groups of genomes, in the combinations of genomes scoring within the 20% highest accuracy values and in the combinations scoring in the lowest 20% values of accuracy. We then used a t-test to determine, for each species, whether the average frequencies within the highest and lowest combinations were significantly different from the frequency in all possible combinations. *, p < 0.05; **, p < 0.001.
In a recent report Eddy [44] used a theoretical model to study the statistical power of comparative genome sequence analysis. The model showed that, at close evolutionary distances, the number of comparative genomes needed to obtain the same statistical power increases. The model also suggests that the decline in statistical power for divergence times above optimal is smaller than for divergence times below optimal. In general our results support some of the proposals made by this model. According to the model it should be possible to obtain a high accuracy with closely divergent species but it would be necessary to use considerably more genomes at that distance. The author suggests that, for example, for human/baboon distances it would be necessary to use about seven times more genomes than at human/mouse distance to obtain the same statistical strength. For future work, we are therefore considering extending our method to include a weighing scheme based on the evolutionary distance between the comparing species and the target species. We think this could be achieved using an adaptation of the theoretical model proposed by Eddy [44].
It would be also interesting to study how many genomes would suffice to accurately predict an SH3 target interaction. Since the decrease in statistical power for Sch. pombe and N. crassa is small compared to species closely related to S. cerevisiae, we calculated the accuracy and coverage after addition of one or two of these species, to the five species selected previously, for different conservation scores. In general, an increase in coverage with little or no decrease in accuracy was observed (see Table S3). Addition of any of the closely related species, instead of N. crassa or Sch. pombe, resulted in a large loss of accuracy with moderate gain in coverage (results not shown). We believe that the improvement gained by adding species within the optimal divergence time would be better than that observed with N. crassa and Sch. pombe. The result generated with the latter two species suggests only that a sufficient number of genomes was not reached, since addition of more genomes still improved our scores. However, at present there are not enough genomes available to empirically tackle this question of a sufficient number of genomes for SH3 target prediction.
We believe the main factor determining the optimum divergence time is the conservation level of the biological feature. A biological feature that has higher conservation will require genomes of more divergent species to be accurately identified. Interaction types that are equally conserved should be accurately predicted with genomes of species at the same divergence times. This might mean that the same genomes could be used to predict interactions for other protein domains that bind small linear peptides (i.e., PDZ, WW, SH2, 14–3–3). Other interaction types that are mediated by larger interaction surfaces are probably more conserved and therefore might require genomes from more divergent species.
Although some results [34,35] have shown the importance of having genomes of recently divergent species in the study of DNA regulatory regions, recent findings [45] have shown that regulatory systems can be conserved over hundreds of millions of years. We argue that the concept of optimal divergence time presented should also be taken into consideration for protein–DNA interactions.
In this paper we show that for the study of SH3 protein interactions the genomes with more relevant information are from species that diverged around 400–950My ago from the species of interest. As was suggested by Eddy [44], this optimum might be specific for the particular interaction type being analyzed. Nevertheless, we believe that our results should be taken into consideration when identifying other biological features using comparative genome sequence analysis.
Predictions of Novel SH3–Linear Peptide Interactions
We used the method described above and the genomes of C. glabrata, K. lactis, C. albicans, D. hansenii, Y. lipolytica, N. crassa, and Sch. pombe to predict a set of 69 interactions regarding consensus sequence conserved in four of the seven genomes used (see Figure 6 and Table S4 for a complete list of the predicted interactions). Genomes of species that were over-represented in groups of genomes scoring within the 20% highest accuracies or under-represented in groups of genomes scoring within the 20% lowest accuracies were used. Some experimental evidence was found to support 37 of these interactions, all of which occurred between proteins labeled as belonging to the same compartments. Of the 32 remaining predictions, eight might not be possible since the putative interaction partners are annotated as having different cellular compartments, although in some cases a link between the two compartments could be possible (see below for some examples). Benchmarking with the gold positive and negative sets resulted in an accuracy of 73% and coverage of 37%. The level of conservation was chosen to allow for higher coverage, but it is important to note that higher accuracy for particular interactions can depend on the degree of conservation observed. We have included information about this in Table S4.
Figure 6 Predictions of S. cerevisiae SH3 Interactions
We considered that a potential target consensus sequence, found by pattern matching, in an S. cerevisiae protein would be biologically relevant if it was within an unstructured region of the S. cerevisiae protein and also conserved in four of the seven comparison genomes used. (C. glabrata, K. lactis, C. albicans, D. hansenii, Y. lipolytica, N. crassa, and Sch. pombe). Red lines indicate the interactions for which we found some experimental evidence in protein interaction databases [59–61]; thin black lines indicate interactions between proteins that are labeled as locating to different compartments; thick black lines indicate interactions for which we found no evidence. There were two S. cerevisiae SH3 domains for which we could not predict any interaction because of the stringency applied. A complete list of the interactions with function, localization, and binding positions is given in Table S4.
As expected we obtained a highly interconnected network with a very significant over-representation of proteins participating in processes typically associated with SH3 domains in S. cerevisiae. GO::TermFinder [46] was used to find significantly shared GO terms within the list of targets of the predictions. Amongst the most significant process associations found were cytoskeleton organization and biogenesis (p = 3.67 × 10−15), morphogenesis (p = 7.62 × 10−12), establishment of cell polarity (p = 1.19 × 10−11), actin cortical patch assembly (p = 5.09 × 10−9), and bud site selection (p = 1.28 × 10−8).
Some of the proposed interactions were further explored taking into account which S. cerevisiae biological processes these proteins were involved in. An interesting example is the proposed interaction between Abp1p with the P-type ATPases Dnf1p and Dnf2p. These proteins are required for phospholipid translocation and they mainly localize to the plasma membrane and intercellular compartments. The regulation of the lipid bilayer arrangement by Dnf1p and Dnf2p was demonstrated to be critical for budding endocytic vesicles [47]. It is also known that Abp1p is one of the activators of the Arp2/3 complex and is important in coupling the actin and membrane dynamics during endocytosis [48]. Following from the proposed interaction seen using our method, we suggest that Abp1p might target Dnf1p and Dnf2p to sites of endocytosis to play a role in endocytic vesicle formation or maintenance.
In order to calculate accuracy and coverage scores, we initially considered as “negative” interactions between proteins that did not share the same cellular compartment. After having obtained our list of predicted interactions, we decided to investigate them without disregarding these “negative” interactions. This decision was made because the negative set is based in part on high-throughput measurements that do not take into account the dynamics of cellular localization. Two proteins might not share a compartment in a given cellular condition, but this might change in different cellular states (examples in S. cerevisiae include cell cycle, pheromone response, and filamentous growth). This reasoning actually leads us to think that the localization data on proteins are underevaluated and, if anything, will result in an underestimation of our accuracy scores.
Within our set of final predictions, Hse1p-mediated interactions are examples of those occurring between proteins marked as belonging to different compartments. According to our results the SH3 domain of Hse1p has a high probability of binding to proline-rich regions of Ste20p, Bck1p, and Las17p. Hse1p was recently reported to be part of a complex that binds ubiquitin and is important in sorting proteins in the endosome [49,50]. Knowing that both Ste20p and Bck1p are involved in the response to mating and that Hse1p is involved in the trafficking/sorting of the alpha-factor pheromone receptor, these SH3 domain interactions might be part of the sorting mechanism of the alpha receptor in the multivesicular bodies. Activated alpha-factor pheromone receptors recruit Ste20p by the dissociation of Gβγ subunits (reviewed in [51]). There is some evidence that Ste20p activation can lead to the phosphorylation of Bck1p in the mating response [52]. Activated mating receptors are internalized after phosphorylation and ubiquitination of their carboxy-terminal tails and are targeted to the vacuole for degradation [53]. We propose that these internalized vesicles are decorated with complexes containing Ste20p, Bck1p, and Las17p and that the interaction of the SH3 domains of Hse1p with these proteins might be important in the sorting of internalized mating receptors.
Conclusion
We present here a method to predict biologically relevant protein interactions mediated by peptide recognition modules. Conservation of target linear peptides and analysis of protein disorder can be effectively combined to screen for biologically relevant interactions that are predicted from binding matrixes obtained from experimental data. However, the method has a small coverage and still relies on experimental determination of the SH3 target consensus sequence. In the future it should be possible to predict the target motifs using available structural data and homology modeling [54,55].
This study provides some evidence for the importance of intrinsic disorder in the context of protein interactions. Specifically, binding motifs within disordered protein regions are more likely to be biologically relevant binding sites than equivalent sites within ordered regions. To our knowledge there is no experimental evidence currently available to support the idea that in general SH3 domains bind within unstructured regions; therefore, particular cases should be investigated carefully. Nevertheless, we hope our observations will contribute to discussion of the role of intrinsically disordered protein regions.
The analysis carried out demonstrated that there is an optimal divergence time for the species to be included in comparative genomics when looking for the conservation of binding sites of peptide recognition modules. For SH3 domains in yeast, this interval is between 400 and 950 My, and although these divergence times may be specific to SH3 domains and to yeast evolution, the concept should be taken into consideration for future comparative studies.
Finally we have used this method to predict novel SH3–linear peptide interactions for S. cerevisiae. The interaction map obtained contains information on the binding regions of both interaction partners and should allow experimentalists to devise effective and precise system perturbations by targeting a particular interaction.
Materials and Methods
SH3 domain conservation.
We created a phylogenetic tree (see Dataset S1) produced by the neighbor-joining method from a ClustalW alignment [56] of the SH3 domains of the 13 yeast species in our set. The SH3 domains were identified using SMART [57]. Putative orthologs for all S. cerevisiae proteins were determined by the BLAST reciprocal best hit method [39]. We considered that a putative ortholog of a S. cerevisiae SH3 domain was not conserved if the two domains were not in the same branch of the phylogenetic tree.
After eliminating these “divergent” domains, we did multiple sequence alignments of the groups of orthologous domains. To determine the binding positions, we included in the alignments the SH3 domain of Fyn. From visual inspection of crystal structures of complexes of SH3 domains with ligands, we decided to analyze the positions Tyr91, Tyr93, Arg96, Thr97, Asp99, Asp100, Asp118, Trp119, Tyr132, Pro134, and Tyr137 of Fyn that we considered might influence binding specificity. By manual inspection of the alignments we extracted the positions of all domains corresponding to the positions of the Fyn SH3 domain that are important for binding specificity and determined their conservation. Any substitution that scored a non-negative value in the blosum62 matrix that would not result in a reversal of charge was considered to be conserved.
Positive and negative datasets.
We considered a positive set of 59 interactions (containing 15 different SH3 domains from 15 different proteins) defined by Tong et al. [22]; this we called the gold set. Tong et al. obtained the final set of interactions by the overlap of two sets of interactions obtained with two different methods. They used phage display data to create a PSSM and used it to scan the S. cerevisiae proteome. Using a threshold on the PSSM they selected the first set of interactions, then they created a second interaction network by yeast two-hybrid screening and obtained the final network (our gold set) by the overlap of the two.
We considered a second positive standard, which we called the platinum set, of higher confidence, with 19 interactions (containing ten different SH3 domains from ten different proteins) derived from the overlap of the two-hybrid assays, obtained from Tong et al. [22], with the MIPS complexes dataset [58].
The two positive datasets overlap only partially (ten interactions from the platinum set are also in the gold set).
To build our negative dataset we assumed that two proteins that do not share the same subcellular compartment according to MIPS localization data [58] cannot interact, and we compiled a list of all S. cerevisiae proteins pairs that do not share at least one subcellular compartment.
Since we also used the phage display data from Tong et al. [22] to derive the consensus sequences recognized by the yeast SH3s used in this study, the gold set might be biased. We would like to stress that we did not use a PSSM as in the Tong et al. paper and therefore even our initial motif-based predictions without any filtering are not the same as the network obtained by Tong et al. with the phage display data.
We did not merge the two positive datasets, thus keeping the platinum one as a truly independent positive dataset. We decided to also use the gold set because although it is not appropriate to use the absolute performance value calculated with this set to compare our method with others, it still served as a check for the relative performance of different filters of our method.
Accuracy and coverage determination.
The ratio between true positives (TP) and the sum of true positives plus false positives (FP) was used as a measure of accuracy. True positives were the number of predicted interactions within a positive set. False positives were the number of predicted interactions found within the negative set. To measure the coverage of the methods, we tracked the ratio TP/P, where P is the total number of positives in the positive set.
Estimated divergence time from S. cerevisiae.
The estimated divergence times of the other yeast species from S. cerevisiae were as follows: C. glabrata, 300 My; D. hansenii, 800 My; K. lactis, 400 My; Y. lipolytica, 900 My; C. albicans, 800 My; S. paradoxus, 50 My; S. bayanus, 50 My; S. mikatae, 50 My; N. crassa, 1,000 My; and Sch. pombe, 1,100 My. These values were based on phylogenetic studies found in the literature [32,42,43].
Supporting Information
Dataset S1 Phylogenetic Tree of the SH3 Domains in the Study
The phylogenetic tree of all the SH3 domains of the yeast species in our study.
(14 KB DND)
Click here for additional data file.
Table S1 Detailed Analysis of the Conservation of Target Consensus Sequence in Putative Targets of S. cerevisiae SH3 Domains
(248 KB PDF)
Click here for additional data file.
Table S2 Detailed Analysis of the Conservation of Target Consensus Sequence in Putative Targets of S. cerevisiae SH3 Domains within Unstructured Regions of Proteins
(248 KB PDF)
Click here for additional data file.
Table S3 Effect of Addition of More Informative Genomes on Accuracy and Coverage Scores
(17 KB PDF)
Click here for additional data file.
Table S4 List of Predicted Interactions
(52 KB PDF)
Click here for additional data file.
We are grateful to G. Cesarenni, Phil Irving, Caroline Lemerle, and Ignacio Enrique Sanchez for useful criticism and discussion. This work was partly funded by EU grant QLRT 2000–01663. PB is supported by a grant from Fundação para a Ciência e Tecnologia through the Graduate Program in Areas of Basic and Applied Biology.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. LS and PB conceived and designed the experiments, analyzed the data, and wrote the paper. PB wrote the scripts.
A previous version of the article appeared as an Early Online Release on July 13, 2005 (DOI: 10.1371/journal.pcbi.0010026.eor).
Abbreviations
Mymillion years
PSSMposition-specific scoring matrix
==== Refs
References
Enright AJ Ouzounis CA 2001 Functional associations of proteins in entire genomes by means of exhaustive detection of gene fusions Genome Biol 2 RESEARCH0034 11820254
Marcotte EM Pellegrini M Ng HL Rice DW Yeates TO 1999 Detecting protein function and protein–protein interactions from genome sequences Science 285 751 753 10427000
Dandekar T Snel B Huynen M Bork P 1998 Conservation of gene order: A fingerprint of proteins that physically interact Trends Biochem Sci 23 324 328 9787636
Pellegrini M Marcotte EM Thompson MJ Eisenberg D Yeates TO 1999 Assigning protein functions by comparative genome analysis: Protein phylogenetic profiles Proc Natl Acad Sci U S A 96 4285 4288 10200254
Gaasterland T Ragan MA 1998 Microbial genescapes: phyletic and functional patterns of ORF distribution among prokaryotes Microb Comp Genomics 3 199 217 10027190
Goh CS Bogan AA Joachimiak M Walther D Cohen FE 2000 Co-evolution of proteins with their interaction partners J Mol Biol 299 283 293 10860738
Pazos F Valencia A 2001 Similarity of phylogenetic trees as indicator of protein–protein interaction Protein Eng 14 609 614 11707606
Gobel U Sander C Schneider R Valencia A 1994 Correlated mutations and residue contacts in proteins Proteins 18 309 317 8208723
Pazos F Valencia A 2002 In silico two-hybrid system for the selection of physically interacting protein pairs Proteins 47 219 227 11933068
Lu L Arakaki AK Lu H Skolnick J 2003 Multimeric threading-based prediction of protein–protein interactions on a genomic scale: Application to the Saccharomyces cerevisiae proteome Genome Res 13 1146 1154 12799350
Lu L Lu H Skolnick J 2002 MULTIPROSPECTOR: An algorithm for the prediction of protein–protein interactions by multimeric threading Proteins 49 350 364 12360525
Jansen R Yu H Greenbaum D Kluger Y Krogan NJ 2003 A Bayesian networks approach for predicting protein–protein interactions from genomic data Science 302 449 453 14564010
Yu H Luscombe NM Lu HX Zhu X Xia Y 2004 Annotation transfer between genomes: protein–protein interologs and protein–DNA regulogs Genome Res 14 1107 1118 15173116
Walhout AJ Sordella R Lu X Hartley JL Temple GF 2000 Protein interaction mapping in C. elegans using proteins involved in vulval development Science 287 116 122 10615043
Kuriyan J Cowburn D 1997 Modular peptide recognition domains in eukaryotic signaling Annu Rev Biophys Biomol Struct 26 259 288 9241420
Castagnoli L Costantini A Dall'Armi C Gonfloni S Montecchi-Palazzi L 2004 Selectivity and promiscuity in the interaction network mediated by protein recognition modules FEBS Lett 567 74 79 15165896
Sadowski I Stone JC Pawson T 1986 A noncatalytic domain conserved among cytoplasmic protein-tyrosine kinases modifies the kinase function and transforming activity of Fujinami sarcoma virus P130gag-fps Mol Cell Biol 6 4396 4408 3025655
Mayer BJ Hamaguchi M Hanafusa H 1988 A novel viral oncogene with structural similarity to phospholipase C Nature 332 272 275 2450282
Cicchetti P Mayer BJ Thiel G Baltimore D 1992 Identification of a protein that binds to the SH3 region of Abl and is similar to Bcr and GAP-rho Science 257 803 806 1379745
Ren R Mayer BJ Cicchetti P Baltimore D 1993 Identification of a ten-amino acid proline-rich SH3 binding site Science 259 1157 1161 8438166
Brannetti B Via A Cestra G Cesareni G Helmer-Citterich M 2000 SH3-SPOT: An algorithm to predict preferred ligands to different members of the SH3 gene family J Mol Biol 298 313 328 10764600
Tong AH Drees B Nardelli G Bader GD Brannetti B 2002 A combined experimental and computational strategy to define protein interaction networks for peptide recognition modules Science 295 321 324 11743162
Dyson HJ Wright PE 2005 Intrinsically unstructured proteins and their functions Nat Rev Mol Cell Biol 6 197 208 15738986
Dunker AK Brown CJ Lawson JD Iakoucheva LM Obradovic Z 2002 Intrinsic disorder and protein function Biochemistry 41 6573 6582 12022860
Tompa P 2002 Intrinsically unstructured proteins Trends Biochem Sci 27 527 533 12368089
Dafforn TR Smith CJ 2004 Natively unfolded domains in endocytosis: Hooks, lines and linkers EMBO Rep 5 1046 1052 15520805
Radhakrishnan I Perez-Alvarado GC Parker D Dyson HJ Montminy MR 1997 Solution structure of the KIX domain of CBP bound to the transactivation domain of CREB: A model for activator:coactivator interactions Cell 91 741 752 9413984
Longhi S Receveur-Brechot V Karlin D Johansson K Darbon H 2003 The C-terminal domain of the measles virus nucleoprotein is intrinsically disordered and folds upon binding to the C-terminal moiety of the phosphoprotein J Biol Chem 278 18638 18648 12621042
Dyson HJ Wright PE 2002 Coupling of folding and binding for unstructured proteins Curr Opin Struct Biol 12 54 60 11839490
Linding R Russell RB Neduva V Gibson TJ 2003 GlobPlot: Exploring protein sequences for globularity and disorder Nucleic Acids Res 31 3701 3708 12824398
Mayer BJ Saksela K 2005 SH3 domains Cesarenni G Gimona M Sudol M Yaffe M Modular protein domains Weinheim (Germany) Wiley-VCH 46 55
Dujon B Sherman D Fischer G Durrens P Casaregola S 2004 Genome evolution in yeasts Nature 430 35 44 15229592
Jones T Federspiel NA Chibana H Dungan J Kalman S 2004 The diploid genome sequence of Candida albicans
Proc Natl Acad Sci U S A 101 7329 7334 15123810
Kellis M Patterson N Endrizzi M Birren B Lander ES 2003 Sequencing and comparison of yeast species to identify genes and regulatory elements Nature 423 241 254 12748633
Cliften P Sudarsanam P Desikan A Fulton L Fulton B 2003 Finding functional features in Saccharomyces genomes by phylogenetic footprinting Science 301 71 76 12775844
Galagan JE Calvo SE Borkovich KA Selker EU Read ND 2003 The genome sequence of the filamentous fungus Neurospora crassa
Nature 422 859 868 12712197
Wood V Gwilliam R Rajandream MA Lyne M Lyne R 2002 The genome sequence of Schizosaccharomyces pombe
Nature 415 871 880 11859360
Notredame C Higgins DG Heringa J 2000 T-Coffee: A novel method for fast and accurate multiple sequence alignment J Mol Biol 302 205 217 10964570
Tatusov RL Koonin EV Lipman DJ 1997 A genomic perspective on protein families Science 278 631 637 9381173
Frishman D Argos P 1997 Seventy-five percent accuracy in protein secondary structure prediction Proteins 27 329 335 9094735
Xia Y Yu H Jansen R Seringhaus M Baxter S 2004 Analyzing cellular biochemistry in terms of molecular networks Annu Rev Biochem 73 1051 1087 15189167
Keogh RS Seoighe C Wolfe KH 1998 Evolution of gene order and chromosome number in Saccharomyces, Kluyveromyces and related fungi Yeast 14 443 457 9559552
Hedges SB 2002 The origin and evolution of model organisms Nat Rev Genet 3 838 849 12415314
Eddy SR 2005 A model of the statistical power of comparative genome sequence analysis PLoS Biol 3 e10. DOI: 10.1371/journal.pbio.0030010 15660152
Gasch AP Moses AM Chiang DY Fraser HB Berardini M 2004 Conservation and evolution of cis- regulatory systems in ascomycete fungi PLoS Biol 2 e398. DOI: 10.1371/journal.pbio.0020398 15534694
Boyle EI Weng S Gollub J Jin H Botstein D 2004 GO::TermFinder—Open source software for accessing Gene Ontology information and finding significantly enriched Gene Ontology terms associated with a list of genes Bioinformatics 20 3710 3715 15297299
Pomorski T Lombardi R Riezman H Devaux PF van Meer G 2003 Drs2p-related P-type ATPases Dnf1p and Dnf2p are required for phospholipid translocation across the yeast plasma membrane and serve a role in endocytosis Mol Biol Cell 14 1240 1254 12631737
Schafer DA 2002 Coupling actin dynamics and membrane dynamics during endocytosis Curr Opin Cell Biol 14 76 81 11792548
Bilodeau PS Winistorfer SC Kearney WR Robertson AD Piper RC 2003 Vps27–Hse1 and ESCRT-I complexes cooperate to increase efficiency of sorting ubiquitinated proteins at the endosome J Cell Biol 163 237 243 14581452
Bilodeau PS Urbanowski JL Winistorfer SC Piper RC 2002 The Vps27p Hse1p complex binds ubiquitin and mediates endosomal protein sorting Nat Cell Biol 4 534 539 12055639
Elion EA 2000 Pheromone response, mating and cell biology Curr Opin Microbiol 3 573 581 11121776
Zarzov P Mazzoni C Mann C 1996 The SLT2(MPK1) MAP kinase is activated during periods of polarized cell growth in yeast EMBO J 15 83 91 8598209
Hicke L 1999 Gettin' down with ubiquitin: Turning off cell-surface receptors, transporters and channels Trends Cell Biol 9 107 112 10201076
Villanueva J Fernandez-Ballester G Querol E Aviles FX Serrano L 2003 Ligand screening by exoproteolysis and mass spectrometry in combination with computer modelling J Mol Biol 330 1039 1048 12860126
Kiel C Wohlgemuth S Rousseau F Schymkowitz J Ferkinghoff-Borg J 2005 Recognizing and defining true Ras binding domains II: In silico prediction based on homology modelling and energy calculations J Mol Biol 348 759 775 15826669
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
Letunic I Copley RR Schmidt S Ciccarelli FD Doerks T 2004 SMART 4.0: Towards genomic data integration Nucleic Acids Res 32 D142 D144 14681379
Mewes HW Frishman D Gruber C Geier B Haase D 2000 MIPS: A database for genomes and protein sequences Nucleic Acids Res 28 37 40 10592176
Xenarios I Rice DW Salwinski L Baron MK Marcotte EM 2000 DIP: The database of interacting proteins Nucleic Acids Res 28 289 291 10592249
Zanzoni A Montecchi-Palazzi L Quondam M Ausiello G Helmer-Citterich M 2002 MINT: A Molecular INTeraction database FEBS Lett 513 135 140 11911893
Bader GD Donaldson I Wolting C Ouellette BF Pawson T 2001 BIND—The Biomolecular Interaction Network Database Nucleic Acids Res 29 242 245 11125103
|
16110343
|
PMC1187863
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Aug 12; 1(3):e26
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010026
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 1611034410.1371/journal.pcbi.001002805-PLCB-RA-0051R3plcb-01-03-04Research ArticleBiochemistryBioinformatics - Computational BiologyMolecular Biology - Structural BiologyEubacteriaNoneEvidence of a Double-Lid Movement in Pseudomonas aeruginosa Lipase: Insights from Molecular Dynamics Simulations MD Simulations of a LipaseCherukuvada Subbulakshmi Latha 1Seshasayee Aswin Sai Narain 1Raghunathan Krishnan 1Anishetty Sharmila 1Pennathur Gautam 12*1 Centre for Biotechnology, Anna University, Chennai, India
2 AU-KBC Research Centre, Madras Institute of Technology, Chennai, India
Murray Diana EditorCornell University Weill Medical College, United States of America* To whom correspondence should be addressed. E-mail: [email protected] 2005 12 8 2005 14 7 2005 1 3 e289 3 2005 11 7 2005 Copyright: © 2005 Cherukuvada 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.
Pseudomonas aeruginosa lipase is a 29-kDa protein that, following the determination of its crystal structure, was postulated to have a lid that stretched between residues 125 and 148. In this paper, using molecular dynamics simulations, we propose that there exists, in addition to the above-mentioned lid, a novel second lid in this lipase. We further show that the second lid, covering residues 210–222, acts as a triggering lid for the movement of the first. We also investigate the role of hydrophobicity in the movement of the lids and show that two residues, Phe214 and Ala217, play important roles in lid movement. To our knowledge, this is the first time that a double-lid movement of the type described in our manuscript has been presented to the scientific community. This work also elucidates the interplay of hydrophobic interactions in the dynamics, and hence the function, of an enzyme.
Synopsis
Lipases hydrolyse long-chain fatty acid esters at water-oil interfaces through the mechanism of interfacial activation mediated by the movement of a lid subdomain that covers the active site. Studying lid movement is an area of active research in the field of protein dynamics. The lipase from Pseudomonas aeruginosa is a 29-kDa protein that was previously crystallized in the open conformation, and as expected, an approximately 20-residue lid subdomain was identified. In the present study, the authors report extensive molecular dynamics simulations of the P. aeruginosa lipase. They show that this protein has two lids covering the substrate-binding pocket. The first lid is the one proposed from the known crystal structure. The second lid, a much shorter one, lies over the binding pocket facing the first lid. Furthermore, using position-restrained simulations, these authors show that movement of the second lid may actually be a trigger for the movement of the first, and that this triggering action is driven by hydrophobic contacts between the two lids. This computational study paves a way for experimentalists to study the structure and dynamics of this protein in greater detail in order to understand coupled subdomain movements in a comprehensive fashion.
Citation:Cherukuvada SL, Seshasayee ASN, Raghunathan K, Anishetty S, Pennathur G (2005) Evidence of a double-lid movement in Pseudomonas aeruginosa lipase: Insights from molecular dynamics simulations. PLoS Comp Biol 1(3): e28.
==== Body
Introduction
Lipases are enzymes that catalyze both the formation and cleavage of long-chain acylglycerols [1]. In recent years, crystal structures of several lipases—bacterial, fungal and mammalian—have been determined [2–8]. Industrially, lipases—especially those of bacterial and fungal origins—are seemingly attractive as biocatalysts due to their cofactor independence, broad substrate specificity, enantioselectivity, and stability in organic solvents [9]. In addition, there have been a number of reports implicating lipases in pathogenicity and biofilm formation [10,11].
All lipases, despite extensive sequence divergence, have a common fold known as the alpha/beta hydrolase fold [12]. The canonical alpha/beta hydrolase fold has eight parallel beta strands, with the second one anti-parallel. Strands 3–8 are connected by alpha helices, which pack on either side of the beta sheet. The active site of lipases is the classical catalytic triad composed of a serine, an aspartate/glutamate, and a histidine [13]. An interesting characteristic of lipases is their interfacial activation. One of the common modes by which this activation occurs is through a lid, also called flap: a stretch of amino acids that moves towards closing the hydrophobic binding pocket in an aqueous environment and opens in an oil-water interface [14]. For further details, see Protocol S1.
The lipase in this study—that from Pseudomonas aeruginosa—is a 29-kDa protein. This lipase differs from the canonical alpha/beta hydrolase fold in that the first two beta strands and one alpha helix are missing. The active site residues in this lipase are Ser82, Asp229, and His251. The loop containing the active-site histidine is stabilized by an octahedrally coordinated calcium ion. On top of the active site, a lid subdomain, comprising the alpha helix residues 125–148 and its surrounding loops, is present in an open conformation in the crystal structure [15].
Molecular dynamics (MD) studies have been used extensively in the study of protein folding [16] and enzyme catalysis [17]. MD simulations have been used to show that there are three mobile regions in the lipase from Rhizomucor miehei, one of which corresponds to the active-site lid [18]. The role of lids and protein orientation with respect to the lipid-water interface in lipase catalysis has been investigated using MD simulations [19,20]. Dependence of enantioselectivity of Candida antarctica lipase on temperature has also been studied using MD [21]. Our group has recently used MD simulations to explain pH-dependent enantioselectivity of C. rugosa lipase and has elucidated the role of the lid in the same [22]. A short introduction to MD simulations is provided in Protocols S2, S3, and S4.
Lipases were traditionally believed to have a single lid. By carrying out extensive molecular dynamics simulations, we infer the presence of a second lid in P. aeruginosa lipase. (The lid proposed from the crystal structure is henceforth called lid1, and the second lid as seen from our results, lid2.) We show that the movement of lid2 acts as a trigger for the movement of lid1. Using similar MD simulations of rational in-silico mutants of this lipase, we elucidate the importance of hydrophobic interactions in movement of the lids. Thus we provide novel insights into interfacial activation in P. aeruginosa lipase.
Results/Discussion
Comparisons of the crystal structures of the open and the closed forms of lipases have been used to conclusively identify mobile loops. However, the crystal structure of only the open form of the lipase from P. aeruginosa is available. Under these circumstances, it is imperative to use other techniques, such as MD simulations, to identify such mobile loops that are essential for the enzyme's activity. Simulations of the open and the closed form of Humicola lanuginose lipase, a fungal lipase, in water for 12 ns using a standard procedure has shown that the open form tends to move towards the closed conformation, while not closing entirely, whereas the closed form remains as such [19]. Ours is, to our knowledge, the first comprehensive study on the dynamics of P. aeruginosa lipase.
The open structure of P. aeruginosa lipase was obtained from the Protein Data Bank (PDB). This was solvated in explicit water and energy minimized. The energy-minimized structure was subjected to MD simulation for 20 ns. The backbone root mean square deviation (RMSD) as calculated after least square fit increased steadily up to about 2.5 ns and reached a plateau. The RMSD value started increasing again after about 4.5 ns and stabilized at around 7 ns. General simulation parameters are shown in Protocol S5.
The First Lid
Since the starting structure is in the open, active conformation, and lipases are known to be activated only at lipid-water interfaces, lid1 can be expected to close in aqueous environments. When the simulated structure (represented here by the 10-ns snapshot) is superimposed on the energy-minimized starting structure, and lid1 highlighted, the movement of this lid towards a typical closed conformation becomes apparent. It can also be seen that the mobile region consists of a helix-loop-helix motif. In the crystal structure, the first helix covers residues 113–117 while the second helix encompasses residues 126–146. In this helix-loop-helix motif, residues 116–138 show high RMS deviations. The C-terminal portion (139–146) of the second helix remains largely rigid. The loop region following this motif—residues 150–156—appears to be highly flexible. This lid has moved by a distance of 1.25 nm, as measured by the distance between the Ca atoms of V130 in the starting structure and in the simulated structure. The RMSD of the lid after least-square fit of the protein backbones is approximately 0.9 nm (Figure 1).
Figure 1 Lid Movement in P. aeruginosa Lipase: Quantitative Measures
Here we show the time variation of the distance between the lids and the RMSD of the two flaps as a function of time. The RMSD of lid1 (red) after least-square fit of the protein backbones is approximately 0.9 nm. The RMSD of the helix in lid2 (blue) is also shown. This is corroborated by a plot of distance between the two lids against time (black). There is an appreciable decrease in this distance after 6 ns. The lid1-lid2 distance stabilizes at about 8 ns, after which there is little relative movement between the two lids.
The Second Lid
On careful analysis we find two regions, besides lid1, that show high mobility. This includes a stretch of amino acids from residues 20 to 30 and another from 203 to 228. Of the two, the latter is found to move towards lid1, thereby augmenting the effect of the movement of lid1. While most residues in the 203–228 region belong to loops, residues 210–222 compose an alpha helix, whose movement towards the opposite lid is significant. When we refer to lid2, we point to the above-mentioned helical region. A spacefill model of both the starting and the simulated structures (Figure 2A) clearly reflects the effect of the movement of the lids in reducing the solvent accessibility of the binding pocket. Quantitatively, the exposed area of the side chains forming the pocket HA decreased from 98.4 Å2 to 0.86 Å2, which is almost a 100% decrease. This has led us to the conclusion that there is a second lid, which, together with the first lid, acts as a double door in protecting the hydrophobic binding site from the polar solvent. The structures of the lipase obtained at different time points of the simulation are shown in Figure 2B.
Figure 2 Lid Movement in P. aeruginosa Lipase: Qualitative Representation
(A) This figure represents the simulation results. The structure on the left (i) is the open form of the lipase, which reveals the active site cavity. Binding pocket residues are coloured yellow, and lid residues red (lid1) and blue (lid2). There appears to be nearly a 100% decrease in the solvent-accessible area when the lids close (structure on right [ii]).
(B) Time evolution of the structure backbone is shown. Blue corresponds to the open crystal structure of the lipase, while red indicates the position of the lids after 3 ns (left [i]), 7 ns (centre [ii]), and 10 ns (right [iii]).
The 10 ns structure does not differ much from the final 20-ns structure (lid1 and lid2 RMSDs are less than 0.1 nm) and hence was taken as the representative structure on the basis of which simulations involving the closed form were performed. Also, the third highly mobile region, residues 20–30, does not cover the binding pocket and hence is not included in further discussion. This movement is touched upon in Protocol S6. Moreover, restraining the movement of this third loop does not seem to affect lid movement in any manner.
A “back” simulation was performed using the closed structure in a water-octane interface. We observed that both the lids actually opened out, and the structure thus formed was nearly the same as the initial crystal structure. The RMSD variation of the lids during the course of the simulation is shown in Figure 3A. The open form obtained after 20 ns and the crystal structure are superimposed to a C-alpha RMSD of 0.08 nm and is shown in Figure 3B.
Figure 3 Simulation Results of Closed Form in Octane-Water Interface
This model reiterates our hypothesis of the second lid in the lipase.
(A) This shows the variation of the distance (black) and RMSD of the two lids (red for lid1 and blue for lid2) with the time course of simulation. As the lid opens, there is a corresponding increase in distance between the two lids. This movement stabilizes from 8 ns until the end of simulation.
(B) Here the final structure of the simulation (after 20 ns; red) is superimposed on the crystal structure (blue) to a C-alpha RMSD of 0.08 nm.
Lid2 Movement Triggers Movement of Lid1
In an earlier paper, Gunasekaran and coworkers [23] demonstrated the effect of triggering loops in enzyme catalysis using restrained MD. To investigate the interdependence of the movements of the two lids, we have applied the same approach. For this we carried out two separate simulations, one with lid1 restrained and the other with lid2 restrained. We observe that while each lid, with the other lid restrained, moves initially, subsequent and significant movement of lid1 is seen only when lid2 is also mobile. RMSD of lid1 increased to 0.4 nm in about 2 ns in both the unrestrained and lid2-restrained systems. The second RMSD increase at about 5 ns, from 0.4 to 0.9 nm, seen in the unrestrained system, is absent in the lid2-restrained system (Figure 4A). The extent of movement of lid2 appears to be independent of that of lid1. In both the unrestrained and the lid1-restrained systems, the RMSD of lid2 in the stable state is approximately 0.4 nm. However, the nature of the trajectory is different in the two cases. In the restrained case, lid2 moves to an RMSD of 0.4 nm within 6 ns, while this is achieved only after 8 ns in the case of the unrestrained molecule (Figure 4B). However, this minor difference can be attributed to the chaotic nature of MD simulations.
Figure 4 Lid1 Movement is Triggered by Lid2, but Lid2 Movement Is Independent of Lid1
(A) Red represents the movement of lid1 when lid2 is restrained, while black is that in the unrestrained system. The movement of lid1 is significantly hindered when lid2 is restrained.
(B) Blue represents the movement of lid2 when lid1 is restrained, and black represents the movement in the unrestrained system. The movement of lid2 is the same in both cases.
Lid Closure is Driven by Hydrophobic Interactions
A helical wheel representation of the two lids (Figure 5) in the Pseudomonas lipase shows that the helices involved are amphipathic, with hydrophobic side chains protruding towards the binding site and towards the opposite lid. The presence of this hydrophobic surface facing the binding site is an ideal construction that allows the binding cleft and the lids to come together through hydrophobic interactions. We hypothesize that significant movement of lid1 corresponding to the 5 ns region in the unrestrained system may require the presentation of hydrophobic residues by lid2 in near proximity to lid1, which does not happen when lid2 is restrained. In fact, close (less than 0.1 nm between contacting surfaces) contacts between Ala126, Val130, and Leu131 on lid1, and Pro210, Phe214, and Ala217 on lid2, are seen in the final stable structure obtained from the unrestrained simulation. This led us to look at the role of hydrophobic interactions in the lid movement. To get further insights into the movement of the lids, we performed further MD with rational in silico mutants.
Figure 5 Helical Wheel Diagrams for Lid1 and Lid2
Shown are the long helix in lid1 (A) and the helix in lid2 (B). Hydrophobic residues are shaded in grey; those in bold face the binding pocket and opposite lid. Letters stand for amino acids.
The Role of Hydrophobic Interactions in Interfacial Activation
We performed molecular dynamics on different in silico mutants of the two residues in lid2, Ala217 and Phe214, so as to implicate the role of hydrophobicity in the movement of the lids. These residues were chosen because they make contacts with residues on lid1 in the closed conformation. Pro210, which also contacts residues on lid1 in the closed conformation, was left untouched because of the possibility of profound local impacts that a proline can impart to a structure. This Pro acts as the N-terminal cap to the lid2 helix and may also be a hinge residue. The aim of this study is to assess the importance of hydrophobic residues in the triggering lid on the movement of the major lid1, and hence only residues from lid2 were mutated.
At the first level, we carried out MD studies on single mutants: the Phe214 and Ala126 were independently mutated to Ser in two different simulations. This was done so as to investigate the role of individual residues in domain movement. These mutations do not seem to affect the lid movement, which is almost the same as in the wild type. The RMSDs of the lids are similar in both the cases (Figure 6; for qualitative information, see Protocol S7).
Figure 6 Simulations with Lipase Single Mutants
(A) This figure corresponds to the simulation of mutant Ala217Ser. The movement of the lids very similar to those in the wild-type lipase. Thus, despite the mutation, the lids are able to close in an aqueous environment.
(B) Here, the simulated structure is a mutant in Phe214Ser.
Red, lid1 RMSD; blue, lid2 RMSD; black, lid1-lid2 distance.
Subsequently we performed a double mutation, Phe214Ser/Ala217Ser. After 20 ns of simulation, the lids did not close. The RMSD plot of the simulation is highly fluctuating and there is nearly no change in the distance between the lids (Figure 7). Thus the double mutation affected the lid movement to a great extent, establishing the robustness that multiple hydrophobic contacts confer to the function of a protein.
Figure 7 Simulation with Lipase Double Mutants
Both the alanine at 217 and the phenylalanine at 214 were mutated to serines, and the simulation was carried out in an aqueous medium. The lids do not close in an aqueous environment. Thus, by mutating the key contact residues to serines, the hydrophobicity of lid2 is lost, and it is unable to function as a trigger for the lids to converge and close.
Red, lid1 RMSD; blue, lid2 RMSD; black, lid1-lid2 distance.
We have also looked at the importance of hydrophobicity in maintaining the closed structure. The closed form of the lipase was mutated at both residues Phe214 and Ala217 to Ser. This structure was simulated in both water and an octane-water interface. In both cases, the lids failed to open. This behaviour in water, where the wild-type lipase is expected to be in a closed form, shows that these hydrophobic contacts are essential only to drive the lid-closure process, but they do not affect the stability of the closed conformation. That the mutant is unable to open up in a water-octane interface, contrary to the wild type (Figure 8), shows that hydrophobicity appears to be essential to lid opening and hence interfacial activation as well. Possibly there is a greater hydrophobic affinity between octane and lid2 than between the lids; thus there appears to be an intricate role for hydrophobicity in the domain movement of lipase.
Figure 8 Simulation of Closed-Structure Double Mutant in Octane-Water Interface
The closed-structure mutant Phe214Ser/Ala217Ser was simulated in an oil-water interface. The lids do not open to expose the lid to the octane-water interface. Thus, hydrophobicity appears to be essential for the lids to open. Hydrophobic affinity between the lid and the octane may be higher than that between the lids, allowing the lids to move apart. In the absence of such hydrophobic groups in lid2, this movement does not take place.
Red, lid1 RMSD; blue, lid2 RMSD; black, lid1-lid2 distance.
Conclusions and Future Directions
We have performed restrained and unrestrained MD simulations of P. aeruginosa lipase in an aqueous environment. This is a lipase whose structure has not been studied as extensively as fungal lipases, leading to paucity of data. We show that this lipase has two lids covering the binding cleft in its putative closed conformation, and that the movement of each lid is not entirely independent of that of the other. We also investigated the role played by an intricate hydrophobic interaction network in the dynamics and, hence, the function of the enzyme. While MD techniques, given the highly developed nature of the force fields used, are fairly accurate, the results that we show amount to a strong hypothesis. However, our theories are amenable to experimental verification. A crystal structure of the closed form of the lipase would allow comparison of experimentally determined structures of the two conformations of the lipase and hence identify mobile loops. On the other hand, fluorescence resonance energy transfer experiments, coupled with site-directed mutagenesis, could be carried out to elucidate the role of hydrophobic contacts in driving lid closure.
Materials and Methods
Software and infrastructure.
The simulations were carried out using GROMACS 3.2 running on single IBM 2.8 GHz systems loaded with Fedora Core 2 Linux at the BTIS facility, Centre for Biotechnology, Anna University. Visualizations were done on an SGI Fuel workstation. GROMACS [24,25] was downloaded from http://www.GROMACS.org.
Simulation.
The GROMACS force field was used for the simulations. The calcium that stabilizes the active-site histidine-containing loop was treated using nonbonded potentials, and it was found that this loop is rigid. The protein was solvated in a box that left a space of 0.2 nm around the solute. SPC water molecules (approximately 3,000) were added. All nonbonded potentials were truncated at 1 nm. For energy minimization, the steepest descents method was used. During the MD, grid-type neighbour searching was done. Long range electrostatics was handled using PME [26]. The system was weakly coupled to an external bath using Berendsen's method [27]. The reference temperature was fixed at 300 K. All bonds were constrained with LINCS [28,29]. A time step of 2 fs was used. Structures were written every 0.5 ps to the trajectory. Each nanosecond took approximately 18 h of CPU time.
The above standard procedure has been used by Jensen et al. [19] to simulate the pair of open and closed forms of H. lanuginose lipase in aqueous conditions. They show that, in 12 ns, the open form tends to move towards a closed conformation, while the closed conformation remains as such.
In order to generate the water-octane interface, a water box (~3,000 water molecules), similar to the one used for the aqueous simulations, was built around the protein. A second box containing ~200 molecules of octane was laid over the water box. Following this, standard simulation protocols were used. The protein molecule, with its lid facing the octane box, moved to the interface within the first nanosecond. All box dimensions are given in Protocol S8.
Analysis.
All visualizations were done using Rasmol or the Swiss PDB Viewer. Trajectory analyses were carried out using tools built inside the GROMACS package. RMSDs were calculated for the protein/protein fragment backbone using least squares fit. The radius of gyration was also calculated for the backbone. Solvent-accessible surface area [30] was calculated using Whatif [31] for the side chains of the residues making up the acyl-binding pocket (HA). Lid1-lid2 distance was calculated between the respective centres of mass. For computing hydrogen bonds, cut-off donor-acceptor distance was 0.35 nm.
Supporting Information
Protocol S1 Lipases
(154 KB PDF).
Click here for additional data file.
Protocol S2 MD Simulations: A Short Perspective
(17 KB PDF)
Click here for additional data file.
Protocol S3 MD Simulations: A Model Protocol
(97 KB PDF)
Click here for additional data file.
Protocol S4 MD Simulations: Position Restraints
(14 KB PDF)
Click here for additional data file.
Protocol S5 Simulation Parameters
(71 KB PDF)
Click here for additional data file.
Protocol S6 The Third Mobile Loop
(97 KB PDF)
Click here for additional data file.
Protocol S7 Single Mutants
(29 KB PDF)
Click here for additional data file.
Protocol S8 Box Sizes
(7 KB PDF)
Click here for additional data file.
Video S1 Main Simulation Movie
(2.1 MB WMV)
Click here for additional data file.
Accession Numbers
The P. aeruginosa lipase PDB (http://www.rcsb.org/pdb/) accession number is 1EX9, and the Uniprot (http://www.ebi.ac.uk/swissprot/access.html) accession number is P26876.
In appreciation of Prof. Varadachari Krishnan, JNCASR, Bangalore, India. We thank the Department of Biotechnology Government of India through the BioTechnology Information Services programme and the Department of Science and Technology (SP/SO/D-18/1999) for providing funds. One of us, SLC, thanks the Council of Scientific and Industrial Research for a studentship. We also thank the developers of GROMACS for making the software available under the GNU license. We thank the anonymous reviewers for their valuable comments and suggestions.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. ASNS and GP conceived and designed the experiments. ASNS and KR performed the experiments. SLC, ASNS, KR, SA, and GP analyzed the data. KR wrote the paper.
A previous version of the article appeared as an Early Online Release on July 14, 2005 (DOI: 10.1371/journal.pcbi.0010028.eor).
Abbreviations
MDmolecular dynamics
PDBProtein Data Bank
RMSDroot mean square deviation
==== Refs
References
Jaeger KE Dijkstra BW Reetz MT 1999 Bacterial biocatalysis: Molecular biology, three-dimensional structures and biotechnological applications of lipases Annu Rev Microbiol 53 315 351 10547694
Grochulski P Li Y Schrag JD Bouthillier P Smith P 1993 Insights into interfacial activation from an open structure of Candida rugosa lipase J Biol Chem 268 12843 12847 8509417
Brady L Brzozowski AM Derewends ZS Dodson G Tolley S 1990 A serine protease triad forms the catalytic centre of a triacylglycerol lipase Nature 343 767 770 2304552
Brzozowski AM Savage H Verma CS Turkenburg JP Lawson DM 2000 Structural origins of interfacial activation in Thermomyces (Humicola) lanuginosa lipase Biochemistry 39 15071 15082 11106485
Schrag JD Cygler M 1993 1.8A refined structure of the lipase from Geotrichum candidum
J Mol Biol 230 575 591 8464065
Carriere F Thirstrup K Hjprth S Ferrato F Nielson PF 1997 Pancreatic lipase structure-function relationships by domain exchange Biochemistry 36 239 248 8993339
Schrag JD Li Y Cygler M Lang D Burgdorf T (1997 ) The open conformation of a Pseudomonas lipase Structure 5 187 202 9032074
Derewenda ZS Derewenda U Dodson CG 1992 The crystal and molecular structure of Rhizomucor miehei triacylglyceride lipase at 1.9 A resolutions J Mol Biol 227 818 839 1404390
Jaeger KE Reetz MT 1998 Microbial lipases form versatile tools for biotechnology Trends Biotechnol 16 396 403 9744114
Berto P Commenil P Belingheri L Dehorter B 1999 Occurrence of a lipase in spores of Alternaria brassicicola with a crucial role in the infection of cauliflower leaves FEMS Microbiol Lett 180 183 189 10556710
Stehr F Kretschmar M Kroger C Hube B Schafer W 2003 Microbial lipases as virulence factors J Mol Catalysis 22 347 355
Schrag JD Cygler M 1997 Lipases and alpha/beta hydrolase fold Methods Enzymol 284 85 107 9379946
Ollis DL Cheah E Cygler M Dijkstra B Frolow F 1992 The alpha/beta hydrolase fold Protein Eng 5 197 211 1409539
Brzozowski AM Derewenda U Derewenda ZS Dodson GG Lawson DM 1991 A model for interfacial activation in lipases from the structure of a fungal lipase-inhibitor complex Nature 351 491 494 2046751
Nardini M Lang DA Liebeton K Jaeger KE Dijkstra BW 2000 Crystal structure of Pseudomonas aeruginosa lipase in the open conformation. The prototype for family I.1 of bacterial lipases J Biol Chem 275 31219 31225 10893416
Fersht AR Dagget V 2002 Protein folding and unfolding at atomic resolution Cell 108 573 582 11909527
Warshel A 2003 Computer simulations of enzyme catalysis: Methods, progress and insights Annu Rev Biophys Biomol Struct 32 425 443 12574064
Peters GH Bywater RP 1999 Computational analysis of chain flexibility and fluctuations in Rhizomucor miehei lipase Protein Eng 12 747 754 10506284
Jensen M Jensen TR Kjaer K Bjornholm T Mouritsen OG 2002 Orientation and conformation of a lipase at an interface studied by MD simulations Biophys J 83 98 111 12080103
Peters GH Bywater RP 2001 Influence of a lipd interface on protein dynamics in a fungal lipase Biophys J 81 3052 3065 11720974
Ottoson J Fransson L Hult K 2002 Substrate entropy in enzyme enantioselectivity: An experimental and molecular modeling study of a lipase Protein Sci 11 1462 1471 12021445
James JJ Lakshmi BB Raviprasad V Ananth MJ Kanguene P 2003 Insights into pH-dependent enantioselective hydrolysis of ibuprofen esters by Candida rugosa lipase Protein Eng 16 1017 1024 14983082
Gunasekaran K Ma B Nussinov R 2003 Triggering loops and enzyme function: Identification of loops that trigger and modulate movements J Mol Biol 332 143 159 12946353
Lindahl E Hess B van der Spoel D 2001 GROMACS 3.0: A package for molecular simulation and trajectory analysis J Mol Mod 7 306 317
Berendsen HJC van der Spoel D van Drunen R 1995 GROMACS: A message-passing parallel MD implementation Comp Phys Comm 91 43 56
Essman U Perela L Berkowitz ML Darden T Lee H 1995 A smooth particle mesh Ewald method J Chem Phys 103 8577 8592
Berendsen HJC Postma JPM Dinola A Haak JR 1984 MD with coupling to an external bath J Chem Phys 81 3684 3690
Miyamoto S Kollman PA 1992 SETTLE: An analytical version of the SHAKE and RATTLE algorithms for rigid water models J Comp Chem 13 952 962
Hess B Bekker H Berendsen HJC Fraaije JGEM 1997 LINCS: A linear constraint solver for molecular simulations J Comp Chem 18 1463 1472
Eisenhaber F Lijnzaad Argos P Sander C Scharf M 1995 The double cube lattice method: Efficient approaches to numerical integration of surface area and volume and to dot surface contouring of molecular assemblies J Comp Chem 16 273 284
Vriend G 1990 WHATIF: A molecular modeling and drug design program J Mol Graph 8 52 56 2268628
|
16110344
|
PMC1187864
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Aug 12; 1(3):e28
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010028
|
oa_comm
|
==== Front
Aust New Zealand Health PolicyAustralia and New Zealand Health Policy1743-8462BioMed Central London 1743-8462-2-151601416510.1186/1743-8462-2-15ResearchEngaging with holism in Australian Aboriginal health policy – a review Lutschini Mark [email protected] Centre for the Study of Health and Society & VicHealth Koori Health Research and Community Development Unit, Department of Public Health, University of Melbourne, Melbourne, Australia.2005 13 7 2005 2 15 15 28 4 2005 13 7 2005 Copyright © 2005 Lutschini; licensee BioMed Central Ltd.2005Lutschini; 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 ideal concept of Aboriginal holistic health is centrally placed in Australian Aboriginal health policies and strategies. Its effective uptake promises, as advocates suggest, reorienting the complex Australian health system to enable health improvements. However, continual reminders assail us that Aboriginal health is shocking, appalling, disastrous, disgraceful and damning. Could incapacity to engage effectively with the concept undermine health system improvements? The aim of this review of Australian literature was to identify the range of meanings attached to Aboriginal holistic health and engage with their implications for the health system.
Results
In terms of literature synthesis I found that policy makers cannot rely on this approach to provide coherent arguments for meaningful engagement with the concept because authors in general: are uncritical and un-reflexive in the use and interpretation of the concept; often provide no reference for their understandings; tend to alter the concept's definition and constituent elements without justification; ignore the wide range of mainstream literature about holism and health; and fail to acknowledge and examine the range of Aboriginal concepts of health. I used the ten themes from this literature to highlight implications for the health system, and found that a most profound contradiction exists in the acceptance of the English language concept 'holistic' as immutably Aboriginal. Additionally, a range of contradictions and mixed messages within the themes challenge the validity of the concept. Significantly, with the boundary of the concept constructed as diffuse and ethereal, the diverse and uncritical literature, and mixed thematic meanings, it is possible to justify any claim about the health system as holistic.
Conclusion
It seems not so much incapacity to engage, but incapacity to coherently articulate Aboriginal concepts of health, which prevents advisory bodies such as the National Indigenous Council to imbue whole-of-government approaches in accordance with Aboriginal values.
==== Body
Background
The seemingly entrenched poor health status of Australian Aboriginal peoples is for the Commonwealth Government 'a glaring and intractable problem' [1:90]. Since the 1970s, the concept of holism is advocated as providing 'a new way of thinking in Aboriginal health' [2]. The concept is often acknowledged as the Aboriginal definition of health (hereafter the Definition):
Health is not just the physical well-being of the individual, but the social, emotional, and cultural well-being of the whole community. This is a whole-of-life view and it also includes the cyclical concept of life-death-life [3:ix]
However, beyond the Definition there is little detail as to the general and specific effects for the health system that acceptance of the concept implies. Therefore, in order to develop an improved knowledge base, a standard approach for policy makers is to perform or commission a literature review on a topic, and then distil the key themes and related meanings. Thus, a greater understanding may be achieved to enable more effective transfer of meanings into the complex Australian health policy environment. The genesis of this study began with a question often faced by policy makers: what is the holistic concept of Aboriginal health and how does it affect health policies, strategies, and programs? From a review of Australian Aboriginal health literature emerged ten themes that I use as focal points to highlight a number of conflicts and tensions complicating policy makers' effective engagement of Aboriginal holistic health.
Results
Literature Characteristics
Table 1 shows the diversity of the 153 publications detected (structure), with journal articles (28%) and reports (31%) constituting the bulk of publications (59%). In terms of publication origin, a total of 39% came from governments compared to 16% from Aboriginal organisations, followed by 15% from 'other' – publications with no explicit organisational affiliation. In terms of year of publication, the release of the National Aboriginal Health Strategy in 1989 is followed by a steady and rapid increase to peak at 74 publications in the period 2000–2002. The citation source is the explicit reference by the authors of their source of holistic, with 24% referring to the NAHS, 12% to 'other' sources such as Aboriginal reports and research, and 64% providing no source. The Definition was directly cited as the source of holistic in 12% of the publications.
Table 1 Characteristics of the Health Professional Literature*
Structure n % Origin n %
Book 6 4 Professional Association 5 3
Policy 9 6 Research Centre 10 7
Monograph 10 7 Multi-institutional 15 10
Inquiry Submission 10 7 University 16 10
Health Strategy 18 12 Other 22 15
Journal Article 43 28 Aboriginal Organisation 24 16
Report 48 31 State Government 25 16
Commonwealth Government 35 23
Total 153 100 Total 153 100
Year n % Citation Source n %
1988 – 1990 1 1 NAHS 36 24
1991 – 1993 6 4 Other 19 12
1994 – 1996 19 12 None 98 64
1997 – 1999 36 24 Total 153 100
2000 – 2002 74 48
2003 – 2004 14 9 NAHS Definition 18 12
Unknown 3 2
Total 153 100
*n = number; % = percent of total; citation source refers to the referenced source of holistic.
Content and Thematic Analysis
Table 2 shows the themes of Aboriginal holistic health. The number of times a theme occurred in the literature, as evident in the examples, are counted (n) and presented in ascending order.
Table 2 Thematic Boundaries Constructing Aboriginal Holistic Health*
Theme n
Problems with the Aboriginal holism
Examples: difficult to define; holistic health care compounds unrealistic expectations; holistic concept is used to distract health services from their core business; and data definitions and standards not adequately developed to encompass holistic view. 9
Concept Confusion
Examples: ecological model, WHO definition of health, primary health care, ethnomedicine, and social medicine. 17
Consistent with Comprehensive Primary Health Care
Examples: holistic comprehensive primary health care; CPHC is holistic; supports provision of CPHC; and holistic CPHC services. 26
Essential to Improved Health Status
Examples: it must be understood that when the harmony of these interrelations is disrupted, Aboriginal ill health will persist; improvement of Aboriginal health depends upon more holistic systems; and a holistic approach to the delivery of services is essential to the improvement of Aboriginal health. 27
Opposite of the Western, Biomedical Approach
Examples: holistic lifestyle opposite of European lifestyle; not built around specialities or body parts; in contrast to mind/body dichotomy of biomedicine; and body parts programs conflict with principle of holistic health. 29
Exemplified in Aboriginal Community Controlled Health Services
Examples: Indigenous services insist on an holistic understanding; Aboriginal medical services incorporate an holistic approach; Aboriginal community controlled health services take account of the holistic context of service delivery; and they deliver holistic primary health care. 38
Mainstream Health System Failure
Examples: fragmentation of roles; lack of coordination; areas that affect health outside the health portfolio; and vertical and inflexible programs. 39
Broad View of Health
Examples: broader context of health; whole of life cycle; multi-faceted view of health; and encompass all aspects of life. 42
Embodied by Aboriginal People
Examples: the holistic view of health traditionally held by Indigenous people; Aboriginal concepts of health are holistic; acceptance of Aboriginal peoples' holistic view of health; and a holistic Aboriginal concept of health. 59
Underpinning Philosophy of Health
Examples: Aboriginal holism should be an underlying principle and philosophy of policy, program development, service delivery, strategies, and practice. 75
*n = number of instances a theme was evident in statements
Elements of Aboriginal Holistic Health
Table 3 shows the number of times (n) that authors stated constituent elements of the concept. The elements from the Definition were overall the most commonly stated (italicised), with other elements included (non-italicised) at a lower number.
Table 3 Inter-related Elements of Aboriginal Holistic Health*
Element n Element n
Ideological 2 Economic 9
Lifestyle 2 Mental 9
Nutrition 2 Physical Environment/Infrastructure 11
Service Environment/access 2 Individual 25
Education 3 Spiritual 29
Governance 3 Well being 29
Identity 3 Community (development, capacity, leadership) 32
Life-death-life 5 Emotional 33
Land 6 Physical 33
Political Environment 6 Cultural 35
Whole-of-life-view 6 Social 38
Family 7
*n = number of instances an element was stated
Discussion
There is no definitive source providing a comprehensive grounding framework to enable effective engagement with the concept of Aboriginal holistic health. Therefore, policy makers have to navigate and interpret a diverse health literature and assemble disparate messages into saleable policy options. In this discussion, I attempt to show how an examination of this literature raises a number of tensions underlying attempts to meaningfully transfer Aboriginal cultural concepts into health policy. This was a text-based study as access to and use of easily accessible published literature is a prime tool in a policy maker's kit. However, it is important to note that the study findings would be improved through interviews and the inclusion of unpublished literature and transcripts of speeches and presentations. Nevertheless, the outcomes of literature reviews are often the first point in establishing the conceptual framework of health policy, strategies and programs.
Immutably Aboriginal?
The literature conveys a strong sentiment that holism is embodied by Aboriginal people (Table 2), and therefore it was necessary to find and understand its original source. The Definition was apparently first written in 1974 by the National Aboriginal and Islander Health Organisation (NAIHO, now the National Aboriginal Community Controlled Health Organisation, NACCHO), however I could not obtain Beaton's citation [5]. The 1989 National Aboriginal Health Strategy (NAHS) appeared as the next likely source. Since the release of the NAHS the explicit use of 'Aboriginal holistic health' in publications increased from 6 references in the period 1991–1993 to peak at 74 references during 2000–2002 (see citation source, Table 1).
In 24% of these publications, authors explicitly cite the NAHS as their source of holistic health (Table 1), in which it is mentioned twice – first on page 60 and the second much deeper in the document:
The Working Party has endorsed the need for a wholistic approach to improving Aboriginal health. This approach will encompass social, cultural, political, economic, environmental and physical factors, not all of which are easy to quantify. [3:219]
However, the NAHS did not provide a source for 'holistic', a practice repeated in 64% of the publications (Table 1). Additionally, the NAHS does not explicitly link 'holistic' and the Definition (which occurs on page x), although 12% of publications did (Table 1). Therefore, within the Aboriginal health literature it is unclear from where 'holism' originated.
Furthermore, no publication in the Aboriginal health literature referred to the extensive non-Aboriginal literature on holistic health. Holism occurs in mainstream health documents such as the National Health Priority Action Areas [6], and the Victorian government's Municipal Public Health Planning Framework [7]. It is a popular keyword in the Australian Journal of Holistic Nursing; is a philosophy extending into the realm of medical doctors as 'alternative' or 'complementary' medicine [8-11]; and is advocated as underlying a 'new kind of GP' [12]. Neither these strategies nor two reviews of holism and health in the non-Aboriginal literature referenced or discussed Aboriginal perspectives [8,13].
Among the five 'reviews' about Aboriginal understandings of health, when using 'holistic' none investigated the root of the word [14-18]. This uncritical acceptance is mirrored in the only research project based on the concept [19,20]. However, the Oxford English Dictionary and the Barnhart Dictionary of Etymology stated that the terms holism and holistic were coined in 1926 by the biologist and former South African Prime Minister Jan Christiaan Smuts [21-23]. Their definitions have obvious connections with the Definition, where both invoke the interconnectedness or 'whole' aspects of holism:
'parts of a whole are in intimate interconnection, such that they cannot exist independently of the whole, or cannot be understood without reference to the whole, which is thus regarded as greater than the sum of its parts' [22:828]
'to designate the tendency in nature to produce wholes (i.e. bodies or organisms) from the ordered grouping of unit structures' [23:307]
It could be surmised that the appropriation of an English language construct was necessary as it is claimed that:
'... while Aboriginal languages do accommodate the complex inter-related constructs involved, Western languages cannot and nor can the relevant Aboriginal constructs be translated' [24:90]
Perhaps this partly explains why the concepts of mwarre, punyu, and wankaru [15,25,26] do not receive any in-depth attention. Additionally, the cross-cutting rivalries and intra-cultural conflict within Aboriginal Australia could be another reason, as the advocacy of one term from one Aboriginal group is seen as an offence to other groups. It also seems difficult to believe that after thirty years of existence of the concept, Aboriginal health leaders and health professionals could not better articulate its meanings, or the meanings of other concepts. An understanding of the meanings attached to these concepts would prove valuable to policy makers who are advised to consider the heterogeneity of Aboriginal cultures [27]. This has significant implications for consultation mechanisms, as outline further below.
Finally, a profound contradiction exists in that 'holism' in text originates from a Western source, but the NAHS states that 'Aboriginal culture is the very antithesis of Western ideology' [3:ix]. Furthermore, although exclusively written in the English language the NAHS is referred to as an 'Aboriginal document' [28]. If culturally based concepts allude to a range of meanings, and if 'such meanings delineate the conceptual field within which action develops' [26:65], then whose cultural base is it?
Essential to Improve Health?
The uncertainty about holism's 'Aboriginality' disempowers the idea that it is essential to improve health (Table 2). Adding to this is the variation of wording and constituent elements of the Definition. For example, NACCHO use this version in their 2003–2006 business plan:
Aboriginal health is holistic, encompassing mental health and physical, cultural and spiritual health. Land is central to well being. Crucially, it must be understood that when the harmony of these interrelations is disrupted, Aboriginal ill health will persist. [29:5]
This significantly diverges from the Definition – where did the social and community go and how did land become central? To what specific interrelations do they refer? In trying to clarify the interrelated elements, in the literature most of the time authors replicated elements stated in the Definition (Table 3), while other elements are added, subtracted or modified without a justification for doing so. The Definition is frequently inserted into documents in cut-down, re-worded and re-phrased versions. This is from the National Aboriginal and Torres Strait Islander Health Strategy: Framework for Action by Governments (hereafter the Framework):
A holistic approach: recognising that the improvement of Aboriginal and Torres Strait Islander health status must include attention to physical, spiritual, cultural, emotional and social well-being, community capacity and governance. [27:2]
The Framework was developed by the National Aboriginal and Torres Strait Islander Health Council, agreed to by all governments, and critically, it claims Aboriginal ownership through extensive consultations. This places the concept in a central strategic position to frame actions in the health system. However, how can it be given due merit considering the implications about its cultural validity in combination with a selective use of constituent elements and shifting definitional boundaries?
Mixed Messages
Furthermore, in considering the integrity of the publications, they come from a wide variety of authors from different disciplines, writing in different styles of publication with various organisational affiliations (Table 1). Their understanding, interpretation and application of holism when they write about Aboriginal health are questionable. This is evident in the mixed messages of confusion, opposition, connection and alignment between Aboriginal and Western ideas of health.
There is conceptual confusion (Table 2) where some authors explicitly link holism to the World Health Organization's definition of health, an ecological or ecosystems approach, a new public health approach, and a systems model of thinking [30-37]. The working party of the NAHS quoted the WHO definition of primary health care in full [3:x], without questioning its cultural origins. Additionally, the Framework states that:
Within the health system, the crucial mechanism for improving Aboriginal and Torres Strait Islander health is the availability of comprehensive primary health care services. [27:1]
The acceptance and uptake of this Western approach appears unproblematic in the NAHS and the literature where Aboriginal holistic health is consistent with comprehensive primary health care (Table 2). Further confusion ensues through the conceptual overlap with primary care, comprehensive primary health care [38,39], and vertical primary health care, which 'forms only a part of comprehensive primary health care, which is a broader, holistic approach to health problems' [40:146]. These differences appear trivial, but their philosophical underpinnings have important implications for operationalisation into health system [38,39].
Both the mainstream and Aboriginal health literature position holism as opposite, counter, or antithetical to the Western biomedical model of health [8,13,26,41-45]. Three senses to this discourse emerge from the literature, the most prominent being the health professions past ethos of colonialism – 'the domination of Aboriginal health care by the medical model approach fitted well with other assimilation policies of the period' [3:59]. In a second sense biomedicine, in contrast with holism, is said to devalue or disregard personal and societal social, economic, environmental and cultural factors on health [33,46]. A third sense relates to biomedicine's Enlightenment rationalist heritage which tends to emphasise a mechanistic view of the body, reducing 'health' to an absence of biochemical and physiological symptoms [33,47].
However, there seems to be a degree of connection between biomedicine and holism because the phrase 'not just the physical well-being of the individual' of the Definition seems to connect with biomedical constructions. Additionally, both the NAHS and the Framework heavily depend on biomedical indices (statistics of illness) in the construction of health need. Therefore, there are points of connection between Aboriginal concepts of health or well-being and biomedical knowledge, as Anderson has noted [26].
Furthermore, there seems to be some alignment between Western and Aboriginal views as both accept the need for a broad view of health. The 1948 WHO definition of health underpins the 1978 Declaration of Alma Ata as accepted in Australian through the 1979 'Health for All' policy [15]. These developments precede the NAHS (1989), and there are similarities between the Definition and the WHO version of health. The Australian health system is also more accepting of other world views in what Heather Eastwood (2000) calls a 'postmodernist movement' that challenges the superiority of the positivist, modern world view. Also, policy makers are more aware of the need to consider ethnicity and culture in health policy [48].
These confusing messages detract from the clarity needed to support the theme of Aboriginal holism as an underpinning philosophy of health (Table 2). What they also signal is that both Western and Aboriginal societies are heterogeneous and contain a multiplicity of health concepts [26,49]. Moving the discussion from a conceptual level to the service delivery level, the literature contains further contradictory meanings.
Which Health System Failure?
Aboriginal people access both Aboriginal-specific and mainstream health services [38,50]. This contrasts with a theme suggesting the failure of the mainstream health system to address Aboriginal health (Table 2), principally because health 'is holistic, a concept that many Western models of healthcare delivery fail to identify and therefore accommodate' [42:222]. Nevertheless, some advocates are fervent that mainstream services should [51], and have [52] become more holistic. Therefore, it is no surprise that the Commonwealth Health portfolio:
is pursuing a two pronged approach, which aims to both improve accessibility and responsiveness of the mainstream health system and to provide complementary action through Indigenous specific health programs. [53:161]
The efficacy of this policy seems justifiable as large potential gains in Aboriginal health could be realised through reorienting Australia's $66.6 billion health system [54]. Advocates readily support work to improve mainstream system access following from many initiatives developed in the NAHS and the Framework. Additionally, governments fund and support more than 100 Aboriginal community controlled health services (ACCHS) initiated since 1971 [24,50] that are said to exemplify holism (Table 2). Published literature from ACCHS would assist policy makers in clarifying service delivery implications of holism, however only 16% of the publications came from Aboriginal organisations (Table 1).
Some descriptive literature is available [55,56] but none based on evaluations of their relative effectiveness [57], although community controlled service delivery is often uncritically proclaimed as 'an uncontested good' [57:4]. Furthermore there is an array of different types of Aboriginal service delivery models covered under the banner of holistic services, ranging from dedicated Aboriginal medical services to multi-purpose cooperatives with health as part of their overall function, to separate substance misuse services [56]. This lack of critical evaluation research is a major impediment to understanding the efficacy of ACCHS, and how the lessons of their operations could influence mainstream health system reforms.
This adds to the theme where some authors note problems with the concept (see Table 2), which include that relying on it may divert attention and resources away from the core business of health services, and lead to difficulties in establishing effective measures and indicators [49,58,59]. However, they fail to elucidate specific criteria and evidence for their assertions. These factors combine with the findings outlined above mean any argument for success of failure of the health system has limited empirical basis for justification. As such, the Commonwealth could justify saying that:
The health needs of Indigenous Australians are largely met through the funding and delivery of mainstream health services, with services specially targeting Aboriginal and Torres Strait Islander people complementing these mainstream services. [60:v]
In a broader sense, often accused of poor coordination Commonwealth Government agencies are beginning 'to engage in holistic thinking and think beyond the boundaries, conceptualise broad outcomes, and understand areas of commonality' [61:53]. However, government officials face a maze of uncoordinated Indigenous structures to navigate [62]. Of the thousands of Aboriginal organisations the potential exists for an 'Indigenous order of Australian government' [63], but attempts such as coalitions of Aboriginal organisations [64,65] fail to present coherent policy positions.
Health Professional Separatism?
In terms of health service delivery, authors are quite willing to claim that Aboriginal people embody holism (Table 2). For example, within ACCHS 'the Health Worker role is driven by a holistic approach' [66:xiii]. Furthermore, the Queensland Department of Health ' understands that recruiting Indigenous peoples will better position Queensland Health to... develop a holistic approach to health' [67:3]. These meanings imply a strict separatism between Aboriginal people and ACCHS and non-Aboriginal people and mainstream services.
An implication of this is seen in the definition of community control where 'Aboriginal people must determine and control the pace, shape and manner of change and decision-making at local, regional, state and national levels' [24:77]. Seemingly in agreement, the Australian Medical Association (AMA) supports Aboriginal holism [68]. However, they staunchly defend placing doctors at the centre of primary health care: 'the AMA strongly recommends that all PHC should be delivered through general practice' [69:4] and emphasise 'The unique clinical skills of GPs in providing holistic/social care' [69:2]. Additionally, in spite of the assimilation label attached to the medical profession in the NAHS, the NAHS Working Party accepts the WHO primary health care model developed under the conceptual umbrella of the WHO definition of health. This was developed by experts (professors and medical doctors) with a background in social medicine, one of whom became the first Director-General of the WHO [49].
In practice, mainstream doctors question their own scientific training in providing health care [8], and there is an increased number of both doctors working in Aboriginal health services, and of Aboriginal doctors [50]. Medical doctors receive particular criticism in Aboriginal health, but they have a significant role in the advocacy for Aboriginal rights and in working with Aboriginal organisations [70]. Medical educators have access to medical curricula improvements to incorporate Aboriginal health [71,72]. There also exist a wide range of internet resources specifically targeting the health professional [73-75]. The proposition of separatism between Aboriginal and non-Aboriginal people, concepts of health and antithetical, unchanging cultures appears undercut by the willingness of health professionals to challenge culturally established norms, and to be culturally sensitive in their approach.
Implications for Health Policy and Strategy Development Processes
I posed the question earlier: could incapacity to engage effectively with the concept undermine health system improvements? I have shown that through the medium of literature synthesis that effective engagement is not possible which is a significant barrier for systemic reform because health strategy construction relies heavily on published literature. This partly undermines the theme of an underpinning philosophy of health (Table 2) as accepted as an explicit policy principle by the Victoria, Northern Territory, New South Wales, Queensland, and Western Australia Governments, and as advocated for as a policy principle by professional associations and a range of non-Government organisations.
However, arguably more meaningful engagement occurs through oral communication. Consultation meetings include a wide range of Aboriginal and non-Aboriginal people across the country. This multitude of voices project preferences through ideological, institutional, professional, and cultural perspectives. While advocates may agree with the concept in principle, each could be tapping into their particular view of holism. For example, some medical literature suggests doctors view 'holistic treatments' as alternative or complementary medicine such as acupuncture, relaxation, massage, and hypnotism [8,41,43].
Additionally, strategies receive ministerial endorsement after processing through a hierarchy of advisory and consultative committees. For example, in the development of the National Aboriginal and Torres Strait Islander Nutrition Strategy and Action Plan, the majority-Aboriginal working party was in effect a sub-sub-sub-committee of the endorsing committee: the Australian Health Ministers Advisory Council [76]. This and other endorsing councils do not include an Aboriginal person to explain the complexities and nuances of Aboriginal cultural concepts, and thus revert to understandings from the literature or from their own perspective, clearly as many authors do in the literature.
It could also be argued that one Aboriginal person on endorsing committees would be insufficient, given the hundreds of Aboriginal groups in Australia. This extends to the validity of the majority Aboriginal membership mandate of steering committees, as each may draw on distinct cultural concepts, such as mwarre, punyu, and wankaru. The in-group rivalries are well noted between different Aboriginal tribes as well as between urban, rural and remote communities [77,78]. Finally, the insignificant number of Aboriginal people in all the health system's structures – in 2001 only 0.9% of health care providers were Aboriginal [79:18] – precludes occupying enough positions to give adequate force to oral transfer.
It is perhaps for these reasons that the new reporting framework for the Council of Australian Governments (COAG), set in the Overcoming Indigenous Disadvantage: Key Indicators report [80], has a complete absence of references to any Aboriginal concept of health. Significantly, the COAG reporting framework effectively orients all commonwealth ministers to produce efforts consistent with this framework. It is therefore questionable whether Aboriginal concepts and knowledge could receive full argumentation and explanation in these processes, because policy 'is the outcome of the competition between ideas, interests, and ideologies' [81:3]. As such, given the heterogeneity of Aboriginal cultures, ineffective transfer of meanings through written and oral communication, undercut by rivalries and intra-cultural conflict preventing national Aboriginal unification, then the approach of the Howard government of dealing directly with individual Aboriginal communities could well be justified [82].
Conclusion
I conclude that literature synthesis cannot provide an answer to the question: what is the holistic concept of Aboriginal health? Rather, policy makers can justify any answer based on the diversity of the literature, subsequent themes and range of meanings. Furthermore, I question the cultural basis advocates attribute to the concept as immutably Aboriginal, which undermines the credibility of its use. In terms of the range of Aboriginal concepts available, the effective transformation of their meanings into the health system could be enhanced through coherent and articulate written and oral communication. The lack of knowledge about them is serious barrier that should be addressed through more research. This would, in part, enhance the capacity of policy makers to engage meaningfully and confidently with Aboriginal concepts of health. However, further attention needs to be given to the processes of health policy and strategy development so that these meanings receive adequate consideration and argumentation in written and oral forms. Otherwise, health outcomes for Aboriginal people may continue along an appalling path.
Methods
Literature Detection
The Australian Aboriginal health literature was defined as Australian publications on Aboriginal affairs with a health specific focus or section and which explicitly stated the terms 'holistic' or 'wholistic'. Publication detection occurred using the search string: (Aborig* OR Indigenous) AND (holistic OR wholistic) AND health, applied to the databases of Informit: APAIS (health, public affairs), ATSIhealth, CINAHL, Health and Society, Australian Medical Index, Rural; as well as MEDLINE and PubMed. Results from each database were cross-referenced and duplicate items removed, with periodic repeat searches ensuring new publication capture. Additionally, snowball searching of journals enabled capture of older publications and those not listed in the databases. Finally, Aboriginal affairs reports were detected through relevant government and organisational internet sites and searched for key terms. The publication characteristics are shown in Table 1.
Content and Thematic Analysis
Through content analysis, the various statements and phrases explicitly associated with the key terms were entered into a spreadsheet and grouped into common categories to 'count' (Table 2 – n) the number of instances a theme occurred. Content analysis is not concerned with the subject categories' production, reception or with any effects [4], and the count number should not be taken as the sole indicator of the importance or otherwise of a theme. The themes emerged from the overt and implied meanings of statements and evolved with multiple readings of the literature, until a clear list emerged, and at times more than one theme was evident within a statement (Table 2).
List of Abbreviations
ACCHS – Aboriginal Community Controlled Health Services
AMA – Australian Medical Association
COAG – Council of Australian Governments
GP – General Practitioner
NACCHO – National Aboriginal Community Controlled Health Organisation
NAHS – 1989 National Aboriginal Health Strategy
NAIHO – National Aboriginal and Islander Health Organisation
PHC – Primary Health Care
WHO – World Health Organization
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
ML is solely responsible for the study.
Acknowledgements
I thank Professor Ian Anderson and Dr Jeannie Devitt for their supervision and support during the study. Thank you to Yin Paradies, Emma Kowal, and Jane Lloydfor constructively critical comments on drafts of this paper. A scholarship from the then Cooperative Research Centre for Aboriginal and Tropical Health provided a stipend for six months to complete the study, which was a component of a Master of Public Health degree delivered through the Menzies School of Health Research.
==== Refs
Hilless M Healy J Health Care Systems in Transition: Australia 2001 2001 Copenhagen: European Observatory on Health Care Systems, World Health Organisation
Hunter P Searching for a New Way of Thinking in Aboriginal Health The Australian Health Consumer 1999 2 16 17
National Aboriginal Health Strategy Working Party A National Aboriginal Health Strategy 1989 Canberra: Department of Aboriginal Affairs
Chapman S Kerr C, Taylor R, Heard G Content Analysis Handbook of Public Health Methods 1998 Sydney: McGraw-Hill Companies, Inc 483 490
Beaton N The Role of the General Practitioner in Aboriginal Health Australian Family Physician 1994 23 11 13 8141678
Commonwealth Department of Health and Aged Care and the Australian Institute of Health and Welfare National Health Priority Areas Report – Diabetes Mellitus – 1998 1999 Canberra: Commonwealth of Australia
Victorian Department of Human Services Municipal Public Health Planning Framework 2001 Melbourne: Victorian State Government
Eastwood H Why Are Australian GPs Using Alternative Medicine? Postmodernisation, Consumerism and the Shift Towards Holistic Health Journal of Sociology 2000 36 133 156
Zollman C Vickers A What is Complementary Medicine? British Medical Journal 1999 319 693 696 10480829
Baer HA Hays J McClendon N McGoldrick N Vespucci R The Holistic Health Movement in the San Francisco Bay Area: Some Preliminary Observations Social Science and Medicine 1998 47 1495 1501 9823045 10.1016/S0277-9536(98)00238-X
Barrett B Marchand L Scheder J Plane MB Maberry R Appelbaum D Rakel D Rabago D Themes of Holism, Empowerment, Access, and Legitimacy Define Complementary, Alternative, and Integrative Medicine in Relation to Conventaionl Biomedicine The Journal of Alternative and Complementary Medicine 2003 9 937 947 14736364 10.1089/107555303771952271
Kamien M The Tasks of General Practice Australian Family Physician 2002 31 1 3
McKee J Holistic Health and the Critique of Western Medicine Social Science and Medicine 1988 26 775 784 3287633 10.1016/0277-9536(88)90171-2
Burden J Bourke C, Bourke E, Edwards B Health: A Holistic Approach Aboriginal Australia: An Introductory Reader in Aboriginal Studies 1994 St. Lucia: University of Queensland Press 157 178
Brady M Robinson G WHO Defines Health?: Implications of Differing Definitions on Discourse and Practice in Aboriginal Health Aboriginal Health: Social and Cultural Transitions 1995 Darwin: NTU Press 187 192
Maher P A Review of 'Traditional' Aboriginal Health Beliefs Australian Journal of Rural Health 1999 7 229 236 10732513 10.1046/j.1440-1584.1999.00264.x
Morgan D Slade M Morgan C Aboriginal Philosophy and its Impact on Health Care Outcomes Australian and New Zealand Journal of Public Health 1997 21 597 601 9470265
Dillon A Total Health Aboriginal and Islander Health Worker Journal 1999 23 3 4
Houston ES The Past, the Present, the Future of Aboriginal Health Policy. Doctoral Thesis 2003 Division of Health Sciences, Curtin University of Technology
Western Australian Office of Aboriginal Health Aboriginal Definition of Health Need Report on Indigenous Funding – Volume 3: Consultant's Reports 2001 3 Canberra: Commonwealth Grants Commission 247 423
Barnhart RK Ed The Barnhart Dictionary of Etymology 1988 H.W. Wilson Company
Soanes C Stevenson A Eds Oxford Dictionary of English: Second Edition 2003 London: Oxford University Press
Simpson JA Weiner ESC Eds The Oxford English Dictionary (2nd ed) Volume VII Hat-Intervacuum 1989 London: Clarendon Press
Bell K Couzos S Daniels J Hunter P Mayers N Murray R Aboriginal Community Controlled Health Services General Practice in Australia: 2000 2000 Canberra: Commonwealth Department of Health and Aged Care 74 103
Devanesen D Traditional Aboriginal Medicine Practice in the Northern Territory Quarterly 2000 33 10 13
Anderson I Grbich C Aboriginal Well-being Health In Australia: Sociological Concepts and Issues (2nd ed) 1999 2 Sydney: Longman 53 73
National Aboriginal and Torres Strait Islander Health Council National Strategic Framework for Aboriginal and Torres Strait Islander Health: Framework for Action by Governments 2003 Canberra: NATSIHC
Murray R Bell K Couzos S Grant M Wronski I Couzos S Aboriginal Health and the Policy Process Aboriginal Primary Health Care: An Evidenced Based Approach (2nd edition) 2003 Melbourne: Oxford University Press 1 36
National Aboriginal Community Controlled Health Organisation National Aboriginal Community Controlled Health Organisation Business Plan 2003–2006 2003 Canberra: NACCHO
Hetzel B Communication and Health – Health as an Ecosystem Medical Journal of Australia 1990 153 548 551 2233479
Barlett B An Aboriginal Health Worker's Guide to Family, Community and Public Health 1995 Alice Springs: Central Australian Aboriginal Congress
Khoury P Germov J Aboriginal Health as a Social Product Second Opinion: An Introduction to Health Sociology 1998 Melbourne: Oxford University Press 57 74
Knight J Germov J Models of Health Second Opinion: An Introduction to Health Sociology 1998 Melbourne: Oxford University Press 136 155
Hetzel B Historical Perspectives on Indigenous Health in Australia Asia Pacific Journal of Clinical Nutrition 2000 9 157 163 10.1046/j.1440-6047.2000.00139.x
Devitt J Hall G Tsey K An Introduction to the Social Determinants of Health in Relation to the Northern Territory Indigenous Population Occasional Paper Series No 6 2001 Darwin: Cooperative Research Centre for Aboriginal and Tropical health
Devitt J Hall G Tsey K Underlying Causes The Health and Welfare of Territorians 2001 Darwin: Northern Territory Government 9 18
Mathews S Jenkin R Frommer M Tjhin M Rubin G When Research Reports and Academic Journals are Clearly Not Enough Strengthening the Links between Aboriginal Health Research and Health Outcomes Occasional Papers Series Issue No 4 2001 Darwin: Cooperative Research Centre for Aboriginal and Tropical Health
Anderson I Young H Markovic M Manderson L Koori Primary Health Care in Victoria: Developments in Service Planning Australian Journal of Primary Health: Interchange 2000 6 24 35
Kelleher H Why Primary Health Care Offers a More Comprehensive Approach for Tackling Health Inequalities than Primary Care Australian Journal of Primary Health 2001 7 57 61
Tilton E Aboriginal Health – An Historical Approach The Health and Welfare of Territorians 2001 Darwin: Northern Territory Government 143 150
Davis A George J States of Health: Health and Illness in Australia 1988 Artarmon: Harper Educational
Brown N Reflections on the Health Care of Australia's Indigenous People Journal of Quality in Clinical Practice 1999 19 211 222 10619148 10.1046/j.1440-1762.1999.00333.x
Wearing M Grbich C Medical Dominance and the Division of Labour in the Health Professions Health In Australia: Sociological Concepts and Issues 1999 2 Sydney: Longman 197 216
Dodson M Aboriginal and Torres Strait Islander Social Justice Commissioner – Second Report 1994 1995 Canberra: Commonwealth of Australia
McDermott R Beaver C Models for Horizontal Equity in Resource Allocation in Aboriginal Health Australian and New Zealand Journal of Public Health 1996 20 13 15 8799059
Lewis J Geelong LH A Dialectical Perspective on the Environment and Health Analysing Health Policy 1997 Chapter 8 Allen & Unwin 1 9
Hahn R Kleinman A Biomedical Practice and Anthropological Theory: Frameworks and Directions Annual Review of Anthropology 1983 12 305 333 10.1146/annurev.an.12.100183.001513
Kimberley Aboriginal Medical Services Council, Effective Healthcare Australia, James Cook University Guidelines for the Development, Implementation and Evaluation of National Public Health Strategies in Relation to Aboriginal and Torres Strait Islander Peoples: Approaches and Recommendations 2002 Melbourne: National Public Health Partnership
Brady M Kunitz S Nash D WHO's Definition?: Australian Aborigines, Conceptualisations of Health and the World Health Organization Migrants, Minorities and Health: Historical and Contemporary Studies 1997 272 290
Dwyer J Silburn K Wilson G Consultant Report No 1 National Strategies for Improving Indigenous Health and Health Care 2004 Canberra: Commonwealth of Australia
Social Health Reference Group Consultation Paper for the Development of the Aboriginal and Torres Strait Islander National Strategic Framework for Mental Health and Social and Emotional Well Being 2004–2009 2003 Canberra: Commonwealth Department of Health and Ageing
Urbis Keys Young Evaluation of the Emotional and Social Well Being (Mental Health) Action Plan 2001 Canberra: Commonwealth of Australia
Commonwealth Department of Health and Aged Care 2002–2003 Portfolio Budget Statements Outcome 7: Aboriginal and Torres Strait Islander Health 2001 Canberra: Commonwealth of Australia
Australian Institute of Health and Welfare Australia's Health 2004 AIHW Cat No AUS 44 2004 Canberra: Australian Institute of Health and Welfare
Office for Aboriginal and Torres Strait Islander Health Better Health Care – Studies in the Successful Delivery of Primary Health Care Services for Aboriginal and Torres Strait Islander Australians 2001 Canberra: Commonwealth of Australia
Shannon C Longbottom H Consultant Report No 4 Capacity Development in Aboriginal and Torres Strait Islander Health Service Delivery – Case Studies 2004 Canberra: Commonwealth of Australia
National Centre for Epidemiology and Population Health Commonwealth Grants Commission Indigenous Funding Inquiry Submission IfI/SUB/0060 2000 Canberra: Commonwealth Grants Commission
Australian Bureau of Statistics The Health and Welfare of Australia's Aboriginal and Torres Strait Islander Peoples 2001 2001 Canberra: Commonwealth of Australia
Wilson J Remote Area Aboriginal Health Services for Managers: Key Practice Challenges Australian Journal of Rural Health 2001 9 138 140 11421966 10.1046/j.1440-1584.2001.00382.x
Commonwealth Department of Health and Aged Care Health and Aged Care Portfolio Submission to the Commonwealth Grants Commission's Inquiry into Indigenous Funding 2000 Canberra: Commonwealth Grants Commission
Management Advisory Committee Connecting Government: Whole of Government Responses to Australia's Priority Challenges 2004 Canberra: Australian Public Service
House of Representatives Standing Committee on Aboriginal and Torres Strait Islander Affairs Many Ways Forward: Report of the Inquiry into Capacity Building and Service Delivery in Indigenous Communities 2004 Canberra: The Parliament of the Commonwealth of Australia
Sanders W Towards an Indigenous Australian Order of Australian Government: Rethinking Self-determination as Indigenous Affairs Policy 2002 Canberra: Centre for Aboriginal Economic Policy Research, Australian National University
From Dispossession to Reconciliation
Media Release – New Coalition of Aboriginal Organisations
Curtin Indigenous Research Centre, Centre for Educational Research and Evaluation Consortium, Jojara & Associaties Training Re-visions A National Review of Aboriginal and Torres Strait Islander Health Worker Training 2001 Perth: Curtin Indigenous Research Centre
Queensland Health Queensland Indigenous Health Workforce Strategy 1999 Brisbane: Queensland Government
Australian Medical Association Preventable Chronic Disease Strategies in Aboriginal and Torres Strait Islander Peoples 2001 Canberra: AMA
AMA Position Paper on Primary Health Care
Media Release – NACCHO and AMA Announce Major Aboriginal Health Report
Phillips G the Committee of Deans of Australian Medical Schools CDAMS Indigenous Health Curriculum Development Project: National Audit and Consultations Report 2004 Melbourne: Centre for the Study of Health and Society, University of Melbourne
Yaxley L Aboriginal Health: An Evaluation of the Implementation of Curriculum, Stage One 2001 Melbourne: Royal Australian College of General Practitioners
Victorian Aboriginal Community Controlled Health Organisation, VicHealth Koori Health Research and Community Development, Unit Teaching Koori Issues to Health Professionals and Health Students – A Community Report 2001 Melbourne: VicHealth Koori Health Research and Community Development Unit
Kokotinna – A Staff Development Program for Aboriginal and Torres Strait Islander Cultures and Health
Aboriginal and Torres Strait Islander Cultural Awareness Training Program
National Aboriginal and Torres Strait Islander Nutrition Working Party National Aboriginal and Torres Strait Islander Nutrition Strategy and Action Plan 2000–2010 2001 Melbourne: National Public Health Partnership
Weaver S Australian Aboriginal Policy: Aboriginal Pressure Groups or Government Advisory Bodies? Part 1 Oceania 1983 54 1 22
Peters-Little F The Community Game: Aboriginal Self-Definition at the Local Community Level 1999 Canberra: Australian Institute of Aboriginal and Torres Strait Islander Studies
Australian Institute of Health and Welfare Health and Community Services Labour Force 2001 2003 Canberra: AIHW
Steering Committee for the Review of Government Service Provision Overcoming Indigenous Disadvantage: Key Indicators 2003 2003 Canberra: Productivity Commission
Bridgman P Davis G The Australian Policy Handbook 2000 2 St Leonards: Allen & Unwin
Office of Indigenous Policy Coordination New Arrangements in Indigenous Affairs 2005 Canberra: Commonwealth Department of Immigration and Multicultural and Indigenous Affairs, Australian Government
|
16014165
|
PMC1187865
|
CC BY
|
2021-01-04 16:38:28
|
no
|
Aust New Zealand Health Policy. 2005 Jul 13; 2:15
|
utf-8
|
Aust New Zealand Health Policy
| 2,005 |
10.1186/1743-8462-2-15
|
oa_comm
|
==== Front
Biomed Digit LibrBiomedical Digital Libraries1742-5581BioMed Central London 1742-5581-2-41598241510.1186/1742-5581-2-4ResearchPMD2HD – A web tool aligning a PubMed search results page with the local German Cancer Research Centre library collection Bohne-Lang Andreas [email protected] Elke [email protected] Anke [email protected] German Cancer Research Centre Heidelberg, Central Spectroscopy – Molecular Modeling, Im Neuenheimer Feld 280, D-69120 Heidelberg, Germany2 University of Applied Sciences Darmstadt, Information and Knowledge Management, Campus Dieburg, Max-Planck-Strasse 2, D-64807 Dieburg, Germany2005 27 6 2005 2 4 4 15 1 2005 27 6 2005 Copyright © 2005 Bohne-Lang et al; licensee BioMed Central Ltd.2005Bohne-Lang 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
Web-based searching is the accepted contemporary mode of retrieving relevant literature, and retrieving as many full text articles as possible is a typical prerequisite for research success. In most cases only a proportion of references will be directly accessible as digital reprints through displayed links. A large number of references, however, have to be verified in library catalogues and, depending on their availability, are accessible as print holdings or by interlibrary loan request.
Methods
The problem of verifying local print holdings from an initial retrieval set of citations can be solved using Z39.50, an ANSI protocol for interactively querying library information systems. Numerous systems include Z39.50 interfaces and therefore can process Z39.50 interactive requests. However, the programmed query interaction command structure is non-intuitive and inaccessible to the average biomedical researcher. For the typical user, it is necessary to implement the protocol within a tool that hides and handles Z39.50 syntax, presenting a comfortable user interface.
Results
PMD2HD is a web tool implementing Z39.50 to provide an appropriately functional and usable interface to integrate into the typical workflow that follows an initial PubMed literature search, providing users with an immediate asset to assist in the most tedious step in literature retrieval, checking for subscription holdings against a local online catalogue.
Conclusion
PMD2HD can facilitate literature access considerably with respect to the time and cost of manual comparisons of search results with local catalogue holdings. The example presented in this article is related to the library system and collections of the German Cancer Research Centre. However, the PMD2HD software architecture and use of common Z39.50 protocol commands allow for transfer to a broad range of scientific libraries using Z39.50-compatible library information systems.
==== Body
Background
The mainstream realm of biosciences research is comprised, from the bibliographic point of view, within the accumulated indexing efforts of the United States National Library of Medicine (NLM). The PubMed [1] database currently maintained by NLM indexes 4571 journals with over 13 million articles [2] (verified on 11-Jan-2005). PubMed's web-based interface delivers comprehensive and usable access to nearly all relevant publications and even provides the option to find articles related to the initial search result. PubMed can be searched using author name(s), keywords, and other criteria as search topics. The results page displays brief information about the retrieved articles (authors, title, journal, PubMed ID) in list form. Individual resulting articles can be marked, and an individual results page can be displayed with selected articles. Opening an author link displays an abstract and typically displays the availability of an electronic full text. Electronic article availability is dependent on the searcher's institutional affiliation as expressed in the computer's Internet protocol (IP) identity. Systems like LinkOut [3] and PubMed's Outside tool [4] can provide electronic access to subscribed articles on opening the respective links article by article. Normally this selected display of relevant articles completes the task of searching PubMed.
A subsequent research working step is acquisition of relevant article full texts that lack direct linking. The individual and institutional subscription situation will provide several common alternatives, such as a local library print or e-journal holding confirmed through the local catalogue. A final resort may be a time-consuming and costly loan request.
Methods
The web tool PMD2HD [5] has been developed to close a knowledge inequity: how can articles without full-text links embedded in PubMed be automatically interpreted for local holding status? The case for a local holding discovery tool was refined using the German Cancer Research Centre library collection. However, the tool can be adapted to use in different library systems or even for searching several collections within a network of federated libraries.
In the scenario envisioned for the use of PMD2HD, a researcher will can carry out a PubMed search as usual, finishing by marking (select all) the complete results web page with the shortcut command control-a (Windows™) and copying it with control-c into the clipboard (Figure 1). (This way copies the complete list to the clipboard. If the user selects only a few articles by mouse click and control-c; it works in the same way.) In the next step the complete clipboard content is pasted (control-v) into the text box of the PMD2HD tool web page (Figure 2). By pressing the [Check!] button the users trigger further data processing (Figure 3). A fundamental premise during design and development was that the tool should be as easy to use for scientists as possible, integrated into their standard workflow.
Figure 1 PubMed results page. In PubMed selected articles marked with control-a.
Figure 2 PMD2HD web tool. The user can paste the content from the clipboard directly into the form.
Figure 3 Data processing. The figure shows the order of the different steps. First the user selects his articles and then the data processing is triggered. The web tool asks the PubMed utilities (Entrez Programming Utilities, ) to retrieve the ISSN. With this item the Horizon system is queried. If the journal is available, the system is called again to find out if an ejournal entry is set. With all this information a results page is generated in a last step.
PMD2HD is written in PHP [6], a very common script language for web-based applications. The program starts with extracting all PubMed IDs (formed like: PMID: 15584372) from the entered text and arranges them in a list. Using this list the program queries the PubMed E-Utilities to retrieve a structured record in XML format [7] for each entry. Table 1 displays an example XML record. The key information of this record is the ISSN. The ISS number serves as the search argument for querying the local library system via the Z39.50 interface (at the German Cancer Research Centre, the Horizon system from Ameritech Library Services is used).
Table 1 XML record from PubMed. XML entry for this publication: Ravindranath MH, Muthugounder S, Presser N, Viswanathan S.Anticancer therapeutic potential of soy isoflavone, genistein. Adv Exp Med Biol. 2004; 546:121-65. PMID: 15584372
<?xml version="1.0"?>
<!DOCTYPE PubmedArticleSet PUBLIC"-//NLM//DTD PubMedArticle, 1st November 2004//EN"">
<PubmedArticleSet>
<PubmedArticle>
<MedlineCitation Owner="NLM" Status="In-Process">
<PMID>15584372</PMID>
<DateCreated>
<Year>2004</Year>
<Month>12</Month>
<Day>08</Day>
</DateCreated>
<Article PubModel="Print">
<Journal>
<ISSN>0065-2598</ISSN>
<JournalIssue>
<Volume>546</Volume>
<PubDate>
<Year>2004</Year>
</PubDate>
</JournalIssue>
</Journal>
<ArticleTitle>Anticancer therapeutic potential of soy isoflavone, genistein.
</ArticleTitle>
<Pagination>
<MedlinePgn>121-65</MedlinePgn>
</Pagination>
<Abstract>
<AbstractText>Genistein (4'5, 7-trihydroxyisoflavone) occurs as a glycoside...
</AbstractText>
</Abstract>
<Affiliation>Laboratory of Glycoimmunotherapy, John Wayne Cancer Institute, 2200 Santa
Monica Blvd., Santa Monica, CA 90404-2302, USA. [email protected]</Affiliation>
<AuthorList CompleteYN="Y">
<Author ValidYN="Y">
<LastName>Ravindranath</LastName>
<ForeName>Mepur H</ForeName>
<Initials>MH</Initials>
</Author>
<Author ValidYN="Y">
<LastName>Muthugounder</LastName>
<ForeName>Sakunthala</ForeName>
<Initials>S</Initials>
</Author>
<Author ValidYN="Y">
<LastName>Presser</LastName>
<ForeName>Naftali</ForeName>
<Initials>N</Initials>
</Author>
<Author ValidYN="Y">
<LastName>Viswanathan</LastName>
<ForeName>Subramanian</ForeName>
<Initials>S</Initials>
</Author>
</AuthorList>
<Language>eng</Language>
<PublicationTypeList>
<PublicationType>Journal Article</PublicationType>
</PublicationTypeList>
</Article>
<MedlineJournalInfo>
<Country>United States</Country>
<MedlineTA>Adv Exp Med Biol</MedlineTA>
<NlmUniqueID>0121103</NlmUniqueID>
</MedlineJournalInfo>
<CitationSubset>IM</CitationSubset>
</MedlineCitation>
<PubmedData>
<History>
<PubMedPubDate PubStatus="pubmed">
<Year>2004</Year>
<Month>12</Month>
<Day>9</Day>
<Hour>9</Hour>
<Minute>0</Minute>
</PubMedPubDate>
<PubMedPubDate PubStatus="medline">
<Year>2004</Year>
<Month>12</Month>
<Day>9</Day>
<Hour>9</Hour>
<Minute>0</Minute>
</PubMedPubDate>
</History>
<PublicationStatus>ppublish</PublicationStatus>
<ArticleIdList>
<ArticleId IdType="pubmed">15584372</ArticleId>
<ArticleId IdType="medline">101832566</ArticleId>
</ArticleIdList>
</PubmedData>
</PubmedArticle>
</PubmedArticleSet
Z39.50 [8,9] is an ANSI standard. Dating from 1996, this protocol has been specified for communication between library database systems and enables searching in heterogeneous databases. The Z39.50 protocol ensures independence from the queried catalogue database, query syntax and operating system.
If the catalogue system finds an entry for a requested ISSN, it returns all information about this entry as a record that is structured according to USMARC. One of the record items is the holding record of the journal subscription. Next, the web tool checks if the publication year of the article is in the Library's journal holding range. If so, the catalogue database is queried again via Z39.50 to get all information for this journal.
One drawback in universal applicability of this Z39.50 method is that some ejournals do not conform to an ISSN search, and occasionally a journal name lookup in the online catalogue is necessary. However, a Z39.50 journal name query is not performed by a string matching search. Instead, all names a journal has ever had, including name changes, are returned to the web tool and have to be parsed. Finally, a results page is generated showing all the data that have been collected in the two preceding steps. As PMD2HD prepares a unified list of articles with respect to both electronic and print holdings of local libraries, it can therefore be regarded as efficient tool that reduces the number of manual user actions for list collection and holding verification.
Results
The PMD2HD results page (Figure 5) combines the main article data in a table, with a second column that describes holding and location information. Five possible actions are presented:
1. If the PubMed entry contains a DOI [10], a direct link to the full text of the article is provided. The full text can be accessed if the organization (or user) is a subscriber of this journal. This means that access is not always possible despite having found the URL of the full text.
2. The holding record of the respective print medium provided from the local library system is provided.
3. A link to the related ejournal is provided.
4. An indication 'Online?' is shown when the web tool cannot fetch an ejournal address from the online catalogue. This advises the user to check the bibliographic data manually when the stored names for the print journal and the ejournal are not exactly the same.
5. A link to the loan request web form. If the holding record of a print journal or an ejournal link is found, a link to the loan request page is not necessary.
For the link to the loan request page, the article data is filled in automatically from the retrieved data fields, and the users only have to enter their names, email addresses and local account number (Figure 5). The ISSN number automatically inserted in the comment field helps the library staff dealing with the loan request. (An in-house electronic document delivery service is planned in the future.)
Discussion
PMD2HD is an easy-to-use web-based tool offering a usable interface and direct integration into the workflow of a typical PubMed search task. PMD2HD has also directed the Z39.50 search protocol for citation verification purposes directly at the library catalogue. PMD2HD unifies two different technical principles of data transmission, the connection-oriented Z39.50 and the non-connection client-server hypertext transmission (HTTP) data exchange. PMD2HD provides data storage that connectionless protocols cannot in terms of storing session parameters or preliminary results. Using the dedicated protocol Z39.50 overcomes the problem of inconsistencies in query languages, data format, and display styles that would ordinarily hinder a web tool's interactivity with a local library management system.
PMD2HD, with small modifications, could be further adapted for library systems with support for the Z39.50 interface, particularly focused scientific libraries and collections. In fact, the PMD2HD application could expand beyond local application to perform a cascading sequence of subsequent transactions, scalable either to a formal network of libraries or the individual needs of scientists who have access to several libraries. The common denominator requirement for alternative implementation is Z.39.50 support. Preliminary investigation in Germany demonstrated that there are two large library networks and four important libraries, including 'Die Deutsche Bibliothek', offering access via Z39.50. Examples all over the world include BIBSYS, the Library of Congress and numerous university libraries, and CURL (Consortium of University Research Libraries) in UK [11]. A test of some of the mentioned libraries showed that access was possible. An additional advantage of Z39.50 is that it does not require any commercial software investment, other than the library system already in use. Development and testing of Z39.50-based routines can be performed using YAZ, a Z39.50 client that is freely available from Index Data, Copenhagen [12].
How does PMD2HD [5] compare with existing tools such as LinkOut [3] or Outside Tool [4]? As a free PubMed service, LinkOut requires that a library uploads a list of the subscribed journal titles. The user of a certain library selects in his myNCBI the library (by using a so-called filter). Even more than one library can be selected. Now the user can perform a search, and on displaying the abstract page the icons of subscribed journals appear. PubMed's Outside Tool, in contrast to LinkOut, shows an icon that provides a link to an external program to which the PubMed ID is submitted. The program can fetch information about the article by using the PubMed ID. The service of keeping holdings up-to-date with Outside Tool is normally offered by a company, and the library has to maintain its subscription list to the company. The Outside Tool service of PubMed is free, but the service of the companies is not. Both services do not interactively check the local library system for holdings but instead rely on an uploaded list of local holdings. The PMD2HD service is independent from PubMed, yet dependent on the immediacy that a z39.50 interface to the library system provides. PMD2HD's greatest advantage is the one-page condensed listing and the inherent simplicity of that format. In contrast, both the LinkOut and Outside Tool services do not integrate the holding result of multiple records on to a single result page. The PMD2HD integrated holdings output list can be printed as a to-do list for when a library visit is necessary.
Conclusion
Optimal benefit from PubMed literature retrieval can only be achieved when as many relevant articles as possible can be acquired in full text form as quickly as possible. PMD2HD has been created as a Web-accessible tool to perform this task on the library collection of the German Cancer Research Centre. PMD2HD helps scientists to obtain literature as fast and effortlessly as possible, and besides it saves time and money for unnecessary interlibrary loan transactions. PMD2HD can be implemented with other integrated library systems or used for retrieval within a network of federated libraries, as its underlying protocol Z39.50 is very widespread among scientific libraries. Use of Z39.50 prevents retrieval failure due to incompatibilities of query syntax and result display forms.
List of abbreviations used
ANSI American National Standards Institute
BIBSYS A consortium for all Norwegian University Libraries, the National Library and a number of college and research libraries.
CURL Consortium of University Research Libraries
DOI Digital Object Identifier
HTTP Hypertext Transmission Protocol
ID Identification
ISSN International Standard Serial Number
IP Internet Protocol
NLM National Library of Medicine
PMID PubMed Identification
USMARC US Machine-Readable Cataloguing
XML Extensible Markup Language
Availability and requirements
• Project name: PMD2HD (A web tool aligning a PubMed search results page with the DKFZ library collection.)
• Project home page:
• Operating system(s): Platform independent
• Type of software: Web-service
• Programming language at server side: PHP 4 [6]
• Other requirements at server side: Apache web server [11], PHP-Yaz module [12]
• Requirements at client side: standard web-browser
• License: free access
• Any restrictions: The access in the presented form is only useful for users of the local library.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AL developed the web tool, EL participated in its design, co-ordination and in writing the manuscript, AT programmed the Z39.50 module with a configuration to the local library system.
Figure 4 Results page of the web tool. The results page lists next to the article data the information on how to access it.
Figure 5 Web form for the loan request. The article data are filled automatically into the different form fields. The user only needs to enter his personal data like name or email.
Figure 6 Icon gallery at PubMed abstract site with an activated LinkOut and OutSide Tool.
Figure 7 PubMed PubMed LinkOut Dataflow.
Figure 8 PubMed OutSide Tool Dataflow.
==== Refs
PubMed
PubMed articles
PubMed's LinkOut Service
PubMed's Outside Tool
PMD2HD web tool
PHP
XML
Z39.50 ANSI
Z39.50 and Internet
Example for a DOI link
Apache web server
Yaz PHP module
CURL Z39.50 feasibility study
|
15982415
|
PMC1187866
|
CC BY
|
2021-01-04 16:38:25
|
no
|
Biomed Digit Libr. 2005 Jun 27; 2:4
|
utf-8
|
Biomed Digit Libr
| 2,005 |
10.1186/1742-5581-2-4
|
oa_comm
|
==== Front
Biomed Digit LibrBiomedical Digital Libraries1742-5581BioMed Central London 1742-5581-2-51598752710.1186/1742-5581-2-5ReviewReview of Doody's Core Titles in the Health Sciences 2004 (DCT 2004) Spasser Mark A [email protected] Robert B. Greenblatt, MD Library, Medical College of Georgia, 1459 Laney Walker Blvd., Augusta, GA 30912, USA2005 29 6 2005 2 5 5 21 6 2005 29 6 2005 Copyright © 2005 Spasser; licensee BioMed Central Ltd.2005Spasser; 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.
==== Body
Overview
For the last forty years, the Brandon / Hill Selected Lists of medical and nursing and allied health books have served as indispensable collection development tools in the medical library community. Following the announcement in April 2004 that the Brandon / Hill Selected Lists would no longer be updated, Doody Enterprises, Inc. decided to develop and publish a list of core titles that would help medical librarians in their collection development decisions. Doody's Core Titles in the Health Sciences 2004 (DCT 2004) is the result of the collaborative effort of approximately 200 content specialists, medical library collection development experts, medical book wholesalers, and the publisher's staff.
History
The "Brandon List," a labor of love for creators Al Brandon and Dorothy Hill, librarians at the Levy Library at Mt. Sinai School of Medicine in New York City, debuted in July 1965 and quickly became vital for medical libraries making collection development decisions. Ultimately split into two lists – one addressing medicine and the other covering allied health and nursing specialties – the Brandon/Hill Lists were updated every two years.
Their initial objective was to develop a list of selected titles for hospital libraries. However, the lists had great reach and influence, and medical libraries of all sizes and types relied on their guidance. Although the librarians had help and support from colleagues, publishers, and medical book wholesalers, they essentially reviewed and made the selections themselves, and they never maintained that their selection process was anything but subjective.
Each new list was published in the J Med Libr Assoc, and then the Medical Library Association (MLA) sold reprints of the lists to medical book wholesalers, who, in turn, distributed the lists for free throughout the medical library community. In 2001, the Brandon / Hill Selected Lists became available online () [1,2].
Brandon's death was followed several years later by Hill's retirement, and in April of 2004, the staff of the library at Mt. Sinai announced that they would not be updating the Brandon / Hill Selected Lists in medicine, nursing and allied health. The discontinuation of the series left a void in the collection development literature, and the profession was rightly concerned.
Doody Enterprises as a Solution
At the time publication of the Brandon/Hill Lists ceased, Doody Enterprises already had a solid reputation in publishing collection development tools. Its first venture, a bimonthly print journal called Doody's Health Sciences Book Review Journal (Doody's Journal) debuted in 1993, and was conceived as the health sciences equivalent to ALA's journal Choice. Developed in close consultation with the company's Library Board of Advisors (LBA), Doody's Journal was designed by medical librarians for medical librarians. From 1993 – 1998, the MLA endorsed Doody's Journal and its electronic version, Doody's Electronic Journal, which first appeared in May, 1995, as a "valuable collection development, cataloging, and reference tool" for its members.
Because of Doody's track record of providing timely and authoritative reviews of newly published books from most of the English-language medical publishers in the world, the company was encouraged to take up the Brandon/Hill mantle. After consultation with their LBA and extensive market research, Doody announced its plans to introduce the inaugural edition of a new Web-based annual publication called Doody's Core Titles in the Health Sciences by the 4th quarter of 2004.
Purpose and Character of the List
The purpose of DCT 2004 is to provide a comprehensive, timely, and authoritative list of book and software titles that represent essential knowledge for professionals or students, and that are highly recommended for libraries that serves some segment of the health sciences community. The DCT 2004's scope is comprehensive, with each edition covering titles in 119 specialties in clinical medicine, basic sciences, nursing, allied health and associated health professions (e.g., dentistry, chiropractic, veterinary medicine, history of medicine, medical ethics, etc.).
The Web-based review process allows for timely publication, with availability following the selection process by mere weeks. The list is updated on a weekly basis with new pricing and edition information. Finally, the authoritativeness of the list derives from the individuals involved in selection and the selection process itself.
The Selection Process
The original list for each of the specialties represented in DCT 2004 is selected by Content Specialists, academically-affiliated health sciences faculty who, in most cases, serve (or have served) as Editorial Review Group Chairs for Doody's Book Review Service™. Each list of core book and software titles is then reviewed by a panel of up to three Library Selectors, (collection development medical librarians) who add titles, as they deem necessary, then score each title based on five criteria essential for responsible collection development.
Doody provides both Content Specialists and Library Selectors with Web-based tools that allow access to the titles that appeared in the final issues of the Brandon / Hill Selected Lists; a searchable database of all in-print book and software titles in Doody's Book Review Service™; an aggregated database consisting of all information from Doody's database (including reviews and ratings; tables of contents licensed from Majors; and bibliographic data on all in-print book and software titles from the web sites of the three major medical book distributors.
According to Doody's description of its methodology, "use of the same five scoring tools and criteria for each title and selector – and averaging the scores across multiple selectors – yields a more "objective" measure than simply asking 'is this a core title?"' In addition, the greater the number of selectors scoring titles in a given specialty, the greater the degree of "objectivity" achieved. Doody states that "relying upon scoring by multiple individuals to provide objectivity is a time-tested approach," and is part of their effort to bring evidence-based methodology to collection development.
The five criteria for collection development are: Authoritativeness of Author(s) and Publisher; Scope and Coverage of the Subject Matter; Quality of Content; Usefulness and Purpose; and Value for the Money.
Each title is graded on a scale of 0 – 3. If a librarian feels that for a given criterion a title should not be considered a core title, they assign a score of 0. If a librarian feels that a title does not belong on the list, the selector scores the title a "0" for each of the criteria. For the other scoring options, a "1" means that the librarian judges the book to be "good" in that particular criterion, a "2" translates to "very good," and a score of "3" in any particular criterion means that the library selector judges the title as "excellent" in that aspect. If a library selector cannot grade a title in one or more criteria because of lack of familiarity with the title, the selector gives the title an "NS" (or "not able to score") designation. If a title receives "NS" scores across all criteria from all selectors, its final score reads "Title Not Scored," meaning that the librarian selectors were unable to rate this title based on the collection development criteria and their first-hand knowledge of the title.
Interpreting the Scores
DCT 2004 offers scores for 1,901 titles in 119 specialties. Titles receiving a score of 3.0, which represents 18% of the titles on the DCT 2004 list, are visually designated with a symbol and are referred to as "essential core titles." DCT 2004 contains 343 "essential core titles". The cost of these titles, based on the retail prices at publication of DCT 2004 amounts to $39,292.43.
Titles scoring between 2.6 and 2.9 represent 37% of the titles of the DCT 2004 list. The 702 titles falling into this scoring range are designated as "key core titles" and visually represented with a different symbol. These "key core titles" represent an investment of $84,087.47, based on retail list prices as of the time of publication of DCT 2004. Estimated costs are calculated based on the current list price of each title on the unique list of titles falling into a given scoring range.
Notes on How Titles are Listed
Titles are listed by:
Edition In nearly all cases, the latest edition is listed, provided it was available for review. In a few instances, the most recent edition is listed, though not yet available, or a link is provided. This irregularity is due to efforts to publish the inaugural edition in time to be of use for library collection budgets; it should disappear with the next edition of DCT 2004.
Volume or Cover Type The goal of the DCT 2004 is to offer information about the single volume of a title that is offered in single and multivolume packages; about the hard cover version of a book offered in both hard and soft cover; and to give the most up-to-date retail pricing available.
Publisher Mergers and acquisitions have resulted in core titles changing publishers. Although the DCT 2004 strives to list the publisher which currently owns the right to the title, the way titles are listed in the two databases that supplied the information for DCT 2004 has made that difficult to do
Scoring A unique attribute of DCT 2004 is the score assigned to each title. See the section above for a complete description of the process.
Display options:
When selecting titles, features include View Mode (condensed and expanded) and various Sort By selections: Author's Last Name, Title, Price, Score, and Copyright Year. Identifying icons are Essential Core and Key Core titles. Titles can be chosen and viewed in the following ways:
• List Overview and Analysis
• Titles By Specialty
• Unique Title List
• Printable List
Doody's DCT 2004 – Psychiatry, Pharmacy/Pharmacology, & Nursing Theory
To illustrate both the strengths and weaknesses of the DCT 2004, psychiatry from the clinical sciences; pharmacy for the associated health professions and pharmacology from the basic sciences; and nursing theory are used.
In terms of psychiatry, it is clear that most core titles are represented. These include (but are certainly not limited to):
American Psychiatric Association's Diagnostic and Statistical Manual of Mental Disorders: Text Revision, 4th Edition, 2000;
Davis's Neuropsychopharmacology: The Fifth Generation of Progress: An Official Publication of the American College of Neuropsychopharmacology, 5th Edition, 2002;
Janicak's Principles and Practice of Psychopharmacotherapy, 3rd Edition, 2001;
Sadock, Kaplan and Sadock's Comprehensive Textbook of Psychiatry – 2 Volume Set, 7th Edition, 2000
Schatzberg's Manual of Clinical Psychopharmacology, 4th Edition, 2003, and The American Psychiatric Publishing Textbook of Psychopharmacology, 3rd Edition, 2004;
Stahl's Essential Psychopharmacology: Neuroscientific Basis and Practical Applications, 2nd Edition, 2000; and
Yudofsky's The American Psychiatric Publishing Textbook of Neuropsychiatry and Clinical Neurosciences, 4th Edition, 2002.
In terms of pharmacy from the associated health sciences and pharmacology from the basic sciences, core titles listed include:
ASHP's AHFS Drug Handbook, 2nd Edition, 2003;
Allen's Ansel's pharmaceutical dosage forms and drug delivery systems, 8th ed., 2005;
Avery's drug treatment : principles and practice of clinical pharmacology and therapeutics, 3rd ed., 1987;
Cooper, The Biochemical Basis of Neuropharmacology, 8th Edition, 2003;
DiPiro's Pharmacotherapy: A Pathophysiologic Approach, 5th Edition, 2002; Evans's Trease and Evans pharmacognosy, 15th ed., 2002.
Gahart's 2005 Intravenous Medications: A Handbook for Nurses and Allied Health Professionals, 2005;
Hardman's Goodman & Gilman's the pharmacological basis of therapeutics, 10th ed., 2001;
Katzung's Basic and Clinical Pharmacology, 9th Edition, 2004
Gennaro's Remington's pharmaceutical sciences, 20th ed.;
Shargel's Applied biopharmaceutics & pharmacokinetics, 2005; and
Tietze's Clinical Skills for Pharmacists: A Patient-Focused Approach, 2nd ed., 2004.
It is clear that some fundamental texts are missing from the pharmacy/pharmacology lists. For example, Block, Wilson and Gisvold's Textbook of Organic Medicinal and Pharmaceutical Chemistry, 11th Edition, 2004; Goldfrank's toxicologic emergencies, 7th ed., 2002; and Koda-Kimble's Applied therapeutics : the clinical use of drugs, 8th ed. 2005.
With regard to nursing theory, many core titles are listed:
Anderson's Community as Partner: Theory and Practice in Nursing, 4th Edition, 2004;
Andrews's Transcultural Concepts in Nursing Care, 3rd Edition, 1999;
Cherry's Contemporary Nursing: Issues, Trends and Management, 2nd Edition, 2002;
Chinn's Integrated Knowledge Development in Nursing, 6th Edition, 2004;
Chitty's Professional Nursing: Concepts and Challenges, 4th edition, 2004;
Fawcett's Relationship of Theory and Research, 3rd edition, 1999;
Kim's Nursing Theories: Conceptual and Philosophical Foundations, 1999;
Orem's Nursing: Concepts of Practice, 6th Edition, 2001; and
Tomey's Nursing Theorists and Their Work, 5th Edition, 2002.
Missing, however, are: Fawcett's Contemporary nursing knowledge :analysis and evaluation of nursing models and theories, 2005; and some version of Nightingale's Notes on Nursing.
Thus, the strengths and weaknesses of Doody's DCT 2004 are illustrated by an analysis of the selection and rating of items in the psychiatry, pharmacology, and pharmacy sections. In the three disciplines discussed above, the DCT 2004 is more or less complete, but the rating system has too little variation (1–3) to measure meaningful differences between items. For instance, what is the true difference between mean scores of 2.5 and 2.6? or between 2.9 and 3.0? This overly restrictive range results in a system as necessarily idiosyncratic as the Brandon-Hill lists it de facto replaces. Take, for instance, the ratings of such foundational texts as the American Psychiatric Association's Diagnostic and Statistical Manual of Mental Disorders or Sadock, Kaplan and Sadock's Comprehensive Textbook of Psychiatry: the former is rated 3.0 and the latter 2.9. It is hard to understand how these classics are not rated the same. Not to put to fine a point on it, but if 3.0 is the highest rating of the most important psychiatric texts (the two Essential Core texts are each given a 3.0), while the mean score for Key Core titles is about 2.75, then is this mean difference between "Essential" and "Key" a distinction without a difference?
Restriction of range problem could result in spurious relationships or the attenuation of significant relationships; that is, a scale with restricted range is simply less sensitive to meaningful differences [3,4]. For example, when the titles are distributed by score, we see that 189 titles are scored at 2.9, 158 at 2.8, 206 at 2.7, and 183 and 2.6. It's difficult, if not impossible, to decipher how these numbers are representing quality differences. As with many poorly explained phenomena modeled as bell curves, most items fall into the upper middle quartile.
Moreover, the owners and designers of DCT 2004 have a different interpretation of "objectivity" than what it commonly refers to, at least in terms of scale construction. For example, as mentioned earlier in the review, they state that "Relying upon scoring by multiple individuals to provide objectivity is a time-tested approach." In terms of psychometric test construction, this refers to reliability not to objectivity. Again from earlier in the review, "Use of the same five scoring tools and criteria for each title and selector – and averaging the scores across multiple selectors – yields a more objective measure than simply asking 'is this a core title."' This is typically a method to establish a scale's construct validity, not its objectivity. In fact, I think it's safe to say that objectivity is not considered a scale property.
Price seems to have little or no association with rating. The average cost of an Essential Core title is $114.56 and $119.78 for a Key Core title. The selection of key texts seems to be adequate, at least for the four disciplines discussed in this review. For nursing, Doody's DCT 2004 is certainly not as comprehensive as the Brandon-Hill in Nursing (for example, the 2002 edition of the Print Nursing Books and Journals 2002 contains 370 nursing books), but given its far more complex selection and elaborate rating systems, it may strive for comprehensiveness, but achieve selectivity. Perhaps, a way to resolve this seeming paradox is to acknowledge that while the DCT 2004 is inclusively comprehensive by covering most of the health sciences, it is more selective in terms of what comprises any given discipline.
Conclusion
While the construction of the DCT 2004 is founded on some shaky methods and assumptions, it is still a valuable and welcome addition collection development tool. As a tool for health sciences collection development, the DCT 2004 lists and rates the preponderance of important texts in the basic, clinical, and associated health sciences.
While not entirely successful, the developers of the list have given its conception and realization very careful consideration, and while not more objective than the Brandon Hill it purports to replace, the DCT 2004 is certainly more thoughtful in terms of it's selection and rating criteria. Notwithstanding its faults, the DCT 2004 is an important resource for health sciences librarians who are responsible for developing and maintaining monographic collections.
List of Abbreviations Used
LBA – Library Board of Advisors
DCT 2004 – Doody's Core Titles
Competing interests
I was a Library Selector for
• Radiologic Technology
• Respiratory Therapy
• Case Management
• Research
• Theory
in the DCT 2004.
Acknowledgements
I want to gratefully thank Dan Doody not only for allowing me to participate in the development of the DCT 2004, but for access to it so that I could write this review.
==== Refs
Hill DR Stickell HN Crow S Print Books and Journals for the Small Medical Library
Hill DR Stickell HN Print Nursing Books and Journals
Gati I Person-environment fit research: problems and prospects J Vocat Behav 1989 35 181 193 10.1016/0001-8791(89)90039-0
Johns G Difference score measures of organizational behavior variables: a critique Organ Behav Hum Perform 1981 27 443 463 10.1016/0030-5073(81)90033-7
|
15987527
|
PMC1187867
|
CC BY
|
2021-01-04 16:38:25
|
no
|
Biomed Digit Libr. 2005 Jun 29; 2:5
|
utf-8
|
Biomed Digit Libr
| 2,005 |
10.1186/1742-5581-2-5
|
oa_comm
|
==== Front
BMC Complement Altern MedBMC Complementary and Alternative Medicine1472-6882BioMed Central London 1472-6882-5-161603365110.1186/1472-6882-5-16Research ArticleAn experimental study of sexual function improving effect of Myristica fragrans Houtt. (nutmeg) Tajuddin [email protected] Shamshad [email protected] Abdul [email protected] Iqbal Ahmad [email protected] Kunwar Mohammad Yusuf [email protected] Department of Ilmul Advia (Unani Pharmacology), Faculty of Unani Medicine, Aligarh Muslim University, Aligarh-202002, India2005 20 7 2005 5 16 16 16 1 2005 20 7 2005 Copyright © 2005 Tajuddin et al; licensee BioMed Central Ltd.2005Tajuddin 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
Myristica fragrans Houtt. (nutmeg) has been mentioned in Unani medicine to be of value in the management of male sexual disorders. The present study was undertaken to evaluate the aphrodisiac effect of 50% ethanolic extract of nutmeg along with its likely adverse effects and acute toxicity using various animal models.
Methods
The suspension of the extract was administered (100, 250 and 500 mg/kg, p.o.) to different groups of male rats daily for seven days. The female rats involved in mating were made receptive by hormonal treatment. The general mating behaviour, libido and potency were studied and compared with the standard reference drug sildenafil citrate. Likely adverse effects and acute toxicity of the extract were also evaluated.
Results
Oral administration of the extract at the dose of 500 mg/kg, produced significant augmentation of sexual activity in male rats. It significantly increased the Mounting Frequency, Intromission Frequency, Intromission Latency and caused significant reduction in the Mounting Latency and Post Ejaculatory Interval. It also significantly increased Mounting Frequency with penile anaesthetisation as well as Erections, Quick Flips, Long Flips and the aggregate of penile reflexes with penile stimulation. The extract was also observed to be devoid of any adverse effects and acute toxicity.
Conclusion
The resultant significant and sustained increase in the sexual activity of normal male rats without any conspicuous adverse effects indicates that the 50% ethanolic extract of nutmeg possesses aphrodisiac activity, increasing both libido and potency, which might be attributed to its nervous stimulating property. The present study thus provides a scientific rationale for the traditional use of nutmeg in the management of male sexual disorders.
==== Body
Background
Nutmeg is the dried kernel of broadly ovoid seed of Myristica fragrans Houtt. (Myristicaceae), a bushy evergreen tree 10–20 m high, indigenous to India, Indonesia and Srilanka [1,2]. It is now cultivated in many tropical countries of both the hemispheres [3]. M. fragrans commonly known as nutmeg is widely used as spice and in alternative medicine it has been reported to have aphrodisiac [4,5], stomachic, carminative [6], tonic [7], nervous stimulant [8], aromatic, narcotic, astringent, hypolipidemic, antithrombotic, antifungal, antidysentric and anti inflammatory properties [9]. It is reported to be useful in paralysis [10] and increases blood circulation [11]. It has also been reported to have antioxidant property [12]. Phytochemical studies indicate that nutmeg contains a volatile oil, a fixed oil, proteins, fats, starch and mucilage. The fixed oil contains myristin and myristic acid. Nutmeg yields 5–15% of volatile oil, which contains pinene, sabinene, camphene, myristicin, elemicin, isoelemicin, eugenol, isoeugenol, methoxyeugenol, safrole, dimeric phenylpropanoids, lignas and neolignas [1,13,14] and by expression it yields a product known as nutmeg butter. Eugenol the major constituent inhibits lipid peroxidation and maintains activities of enzymes like superoxide dismutase, catalase, glutathione peroxidase, glutamine transferase and glucose-6-phosphate dehydrogenase [15], and has also been reported to have vasodilatory [16] and smooth muscle relaxant property [17]. There is no information regarding the chemical constituents in 50% ethanolic extract of nutmeg, but their amount in the extract seems to be lower than in nutmeg. However, its preliminary phytochemical studies that were performed as described by Jenkins et al. [18] indicate the presence of sterols, phenols, alkaloids and. Amino acids. The chloroform extract of nutmeg showed analgesic and anti-inflammatory activity in rodents [19], and also exhibited antidiarrheal activity by increasing tissue contents of Na+ and Cl- ions [20]. The petroleum ether extract of M. fragrans fruits possesses anti-diarrheal property [21], and its n-hexane extract has been reported to have memory enhancing effect in mice [22]. The ethanolic extract (50%) of nutmeg exhibited significant aphrodisiac effect in male mice in our earlier study, in which only mounting frequency and mating performance were considered as the markers for sexual function [23].
Considering above information, nutmeg-extract (50% ethanolic) was subjected to a detailed screening for sexual function improving activity using various animal models. The standard drug sildenafil citrate was used as a referent for quantitative comparison. Likely treatment-related adverse effects and acute toxicity were also evaluated. The doses used in the study were selected according to Freirich[24], multiplying the Unani clinical doses [10,25,26] by the conversion factor of 7.
Methods
Plant material
Dried kernel of M. fragrans (nutmeg) was procured from the market (Delhi, India). The plant material was authenticated by Prof. S. H. Afaq, In-charge, Pharmacognosy Section, Department of Ilmul Advia, Faculty of Unani Medicine, Aligarh Muslim University, Aligarh, India. A voucher (M144) sample was deposited for further reference.
Extraction procedure
Nutmeg was crushed to coarse powder and sieved through No. 20–40 mesh sieve and refluxed by mixing the powdered nutmeg with 1:3 w/v in 50% ethanol v/v by Soxhlet apparatus for 6 h. The extract was filtred and the solvent from the filtrate was removed by rotary evaporator under reduced pressure and low temperature. The yield of extract was 21.20% w/w in terms of dried starting material. The extract was preserved in a refrigerator.
Chemicals
Sildenafil citrate was obtained from Zydus Cadila (Ahmadabad, India). Ethinyl oestradiol was a gift from Infar Limited (Kolkata, India). Progestereon was procured from Sun Pharmaceutical Industries Limited (Mumbai, India) and 5% xylocaine ointment was obtained from Astra IDL Limited (Bangalore, India).
Animals
Three month old male and female albino rats of Wistar strain weighing 350–400 g and 225–275 g, respectively, were used for aphrodisiac study. Adult albino mice of either sex weighing 25–35 g were used for acute toxicity testing. All the animals were housed singly in separate standard propylene cages and maintained under standard laboratory conditions (temperature 24–28°C, relative humidity 60–70%, 12 h light-dark cycle) with water and food (Gold Mohar, Lipton-India) ad libitum. The Departmental ethical committee for animal care and use approved the experimental design.
Drug preparation
Nutmeg in Unani medicine is orally administered, therefore the extract was suspended in distilled water using Tween-80 (1%) for oral administration. Similarly sildenafil citrate and ethinyl oestradiol were also suspended in distilled water using Tween -80 (1%) separately, for oral use. Progesterone was dissolved in olive oil for subcutaneous injection. All the drug solutions were prepared just before administration.
Mating behaviour test
The effect of the test drug on mating behaviour was studied according to the methods described by Dewsbury and Davis Jr. [27] and Szechtman et al. [28], modified by us. Healthy and sexually experienced male rats were selected for the study. They were divided in to five groups each consisting of six rats and placed individually in separate propylene cages during the experiment. Group 1 served as control group and received 10 ml/kg of distilled water orally, daily for 7 days at 18:00 h. Groups 2–4 received suspension of the extract orally at the doses of 100, 250 and 500 mg/kg, respectively, once a day for 7 days at 18:00 h. Group 5 served as standard group and given suspension of the standard drug 1 h before the commencement of the experiment. Since the male animals should not be tested in unfamiliar circumstances the animals were brought to the laboratory and exposed to dim light (in 1 w fluorescent tube in a laboratory of 14' × 14') at the stipulated time of testing daily for 6 days before the experiment.
The female rats allow mating only during the estrus phase. Thus, they were artificially brought into oestrus (heat) by the method of Szechtman et al [28]. They were administered suspension of ethinyl oestradiol orally at the dose of 100 μg/animal 48 h prior to the pairing plus progesterone injected subcutaneously, at the dose of 1 mg/animal 6 h before the experiment. The receptivity of the female animals was confirmed before the test by exposing them to male animals, other than the control, test and standard animals. The most receptive females were selected for the study. The experiment was carried out on the 7th day after commencement of the treatment of the male animals. The experiment was conducted at 20:00 h in the same laboratory and under the light of same intensity. The receptive female animals were introduced into the cages of male animals with 1 female to 1 male. The observation for mating behaviour was immediately commenced and continued for first 2 mating series. The test was terminated if the male failed to evince sexual interest. If the female did not show receptivity it was replaced by another artificially warmed female. The occurrence of events and phases of mating were called out to be recorded on an audio-cassette as soon as they appeared. Their disappearance was also called out and recorded. Later, the frequencies and phases were determined from cassette transcriptions: number of mounts before ejaculation or Mounting Frequency (MF), number of intromission before ejaculation or Intromission Frequency (IF), time from the introduction of female into the cage of the male upto the first mount or Mounting Latency (ML), time from the introduction of the female up to the first intromission by the male or Intromission Latency (IL), time from the first intromission of a series upto the ejaculation or Ejaculatory Latency (EL), and time from the first ejaculation upto the next intromission by the male or Post Ejaculatory Interval (PEI). In the second mating series only the EL was recorded. The values for the observed parameters of the control, test and standard animals were statistically analysed by using one-way analysis of variance (ANOVA) method.
Test for libido
The test was carried out by the method of Davidson [29], modified by us. Sexually experienced male rats were divided into five groups each consisting of six rats and kept singly in separate propylene cages during the experiment. Group 1 represented the control group, which received 10 ml/kg of distilled water orally, once a day for 7 days at 18:00 h. Group 2–4 received suspension of the extract orally at the doses of 100, 250 and 500 mg/kg, respectively, daily for 7 days at 18:00 h. Group 5 served as standard group and given suspension of the standard drug orally at the dose of 5 mg/kg, 1 h prior to the commencement of the experiment. The female rats were made receptive by hormonal treatment and all the animals were accustomed to the testing condition as previously mentioned in mating behaviour test. The animals were observed for Mounting Frequency (MF) on the evening of 7th day at 20:00 h. The penis was exposed by retracting the sheath and 5% xylocaine ointment was applied 30, 15 and 5 min before starting observations. Each animal was placed individually in a cage and the receptive female rat was placed in the same cage. The number of mountings was noted. The animals were also observed for intromission and ejaculation. The MF in control, test and standard animals was statistically analysed by employing one-way analysis of variance (ANOVA) method.
Test for potency
The test was carried out by the methods of Hart and Haugen [30] and Hart [31], modified by us. The male rats were divided in to five groups each consisting of six rats and placed individually in separate propylene cages during the experiment. Group 1 represented the control group, which received 10 ml/kg of distilled water orally daily for 7 days. Group 2–4 received suspension of the test drug orally at the dose of 100, 250 and 500 mg/kg, respectively, daily for 7 days. Group 5 served as standard group and received suspension of the standard drug orally at the dose of 5 mg/kg, 1 h prior to the commencement of the experiment. On the 8th day, the test for penile reflexes was carried out by placing the rat on its back, in a glass cylinder for partial restraint. The preputial sheath was pushed behind the glans by means of thumb and index finger and held in this manner for a period of 15 min. Such stimulation elicits a cluster of genital reflexes. The following components were recorded: Erections (E), Quick Flips (QF) and Long Flips (LF). The frequency of these parameters observed in control, test and standard groups was statistically analysed by using one-way analysis of variance (ANOVA) method.
Adverse effects
All treated rats were observed at least once daily for any overt sign of toxicity (salivation, rhinorrhoea, lachrymation, ptosis, writhing, convulsions and tremors), stress (erection of fur and exophthalmia) and changes in behaviour (such as spontaneous movement in the cage, climbing, cleaning of face). In addition, food and water intake was noted.
Acute toxicity testing
The acute toxicity of the extract was studied in adult albino mice of either sex. They were divided into five groups each consisting of six mice. The suspension of the extract was administered orally at four different doses of 500, 1000, 2000 and 4000 mg/kg, respectively, to different groups of mice separately. Control animals received 10 ml/kg of distilled water orally. The animals were observed continuously for the initial 4 h for behavioural changes and mortality and intermittently for the next 6 h and then again at 24 h and 48 h after dosing. The behaviour parameters observed were convulsion, hyperactivity, sedation, grooming, loss of righting reflex and increased respiration.
Statistical analysis
The significance of difference between the means was determined by one-way analysis of variance (ANOVA) with post-hoc't' test. P value <0.05 was considered as significant.
Results
Effect of the extract on mating behaviour
The results of mating behaviour test show that the extract at the dose of 500 mg/kg, significantly increased the Mounting Frequency (MF) (P < 0.001), Intromission Frequency (IF) (P < 0.001), Ejaculatory Latency in first series (EL1) (P < 0.001), Ejaculatory Latency in second series (EL2) (P < 0.001), and caused significant reduction in the Mounting Latency (ML) (P < 0.001), Intromission Latency (IL) (P < 0.01) and Post Ejaculatory Interval (PEI) (P < 0.001), as compared to control group. The dose of 250 mg/kg of the extract significantly increased the MF(P < 0.001), IF (P < 0.01) and significantly decreased the ML (P < 0.01), PEI (P < 0.001), IL (P < 0.05), EL1 (P < 0.05) and did not significantly alter the EL2, in comparison with the control group. Whereas, the extract at the dose of 100 mg/kg, significantly increased the MF (P < 0.01) and PEI(P < 0.05), but did not affect the IF, ML, IL, EL1 and EL2 in a significant manner as compared to control group. However, the standard drug increased the MF (P < 0.001), IF (P < 0.001), EL1 (P < 0.001), EL2 (P < 0.001) and PEI (P < 0.001) as well as decreased the ML (P < 0.001) and IL (P < 0.001) in a highly significant manner as compared to control animals (Table 1).
Table 1 Effect of 50% ethanolic extract of M. fragrans (nutmeg) on mating behaviour in male rats
Parameters Mean ± SEM
Control (10 ml/kg) Nutmeg (100 mg/kg) Nutmeg (250 mg/kg) Nutmeg (500 mg/kg) Sildenafil citrate (5 mg/kg)
Mounting Frequency (MF) 11.50 ± 1.22 14.3 ± 0.49** 23.70 ± 1.47*** 43.80 ± 0.94*** 48.70 ± 2.34***
Intromission Frequency (IF) 5.50 ± 1.22 5.50 ± 0.42 NS 8.17 ± 0.60** 12.50 ± 0.76*** 24.70 ± 0.81***
Mounting Latency (ML, in sec) 35.30 ± 1.51 35.80 ± 1.46 NS 28.50 ± 0.34** 22.80 ± 0.87*** 11.70 ± 1.37***
Intromission Latency (IL, in sec) 40.00 ± 5.29 37.70 ± 1.97 NS 34.00 ± 1.65* 27.50 ± 1.80** 15.00 ± 0.89***
Ejaculatory Latency in first series (EL1, in sec) 198.00 ± 0.98 211.00 ± 6.69 NS 218.00 ± 2.13* 235.00 ± 6.65*** 344.50 ± 12.00***
Ejaculatory Latency in second series (EL2, in sec) 297.33 ± 8.10 300.00 ± 4.98 NS 319.00 ± 6.98 NS 358.00 ± 7.22*** 398.16 ± 13.50***
Post Ejaculatory Interval (PEI, in sec) 364.00 ± 12.22 336.00 ± 7.92* 301.00 ± 6.89*** 224.00 ± 4.69*** 99.00 ± 5.68***
Tabular values are mean ± SEM, n = 6 (number of animals in each group); significant difference from control, NS: Not significant. *P < 0.05, **P < 0.01; ***P < 0.001.
Effect of the extract on libido
The results obtained with the test for libido show that the extract at the dose of 500,250 and 100 mg/kg, significantly increased the Mounting Frequency (MF) (P < 0.001, P < 0.01 and P < 0.05, respectively) as compared to control group. The standard drug also significantly increased the MF (P < 0.001) as compared to control animals. Intromission and Ejaculation were absent in control, test and standard groups (Table 2).
Table 2 Effect of 50% ethanolic extract of M. fragrans (nutmeg) on mounting frequency (test for libido) in male rats
Parameters Mean Frequency ± SEM
Control (10 ml/kg) Nutmeg (100 mg/kg) Nutmeg (250 mg/kg) Nutmeg (500 mg/kg) Sildenafil citrate (5 mg/kg)
Mounting Frequency (MF) 6.17 ± 0.98 7.83 ± 0.47* 8.50 ± 0.56** 14.50 ± 4.43*** 23.00 ± 2.17***
Intromission Frequency (IF) Nil Nil Nil Nil Nil
Ejaculation (EJ) Absent Absent Absent Absent Absent
Tabular values are mean ± SEM, n = 6 (number of animals in each group); significant difference from control, *P < 0.05, **P < 0.01; ***P < 0.001.
Effect of the extract on potency
The test for potency revealed that the extract at the dose of 500 mg/kg, significantly increased the frequency of Erections (E) (P < 0.001), Quick Flips (QF) (P < 0.001) and Long Flips (LF) (P < 0.001) as well as the aggregate of these penile reflexes (TPR) (P < 0.001) in comparison with the control group. The test drug at the dose of 250 mg/kg, significantly increased the E (P < 0.05), LF (P < 0.01) and TPR (P < 0.05) but did not significantly affect the QF. Whereas, the extract at the dose of 100 mg/kg, did not alter the E, QF, LF and TPR. The standard drug also significantly increased the E (P < 0.001), QF (P < 0.001), LF (P < 0.001) and TPR (P < 0.001) with respect to the control animals (Table 3).
Table 3 Effect of 50% ethanolic extract of M. fragrans (nutmeg) on penile reflexes (test for potency) in male rats>
Parameters Mean Frequency ± SEM
Control (10 ml/kg) Nutmeg (100 mg/kg) Nutmeg (250 mg/kg) Nutmeg (500 mg/kg) Sildenafil citrate (5 mg/kg)
Erections (E) 7.67 ± 1.63 7.50 ± 0.42 NS 8.00 ± 0.36* 12.66 ± 0.75*** 19.00 ± 2.64***
Quick Flips (QF) 5.17 ± 0.75 5.50 ± 0.49 NS 5.83 ± 0.56 NS 8.66 ± 0.36*** 17.30 ± 4.13***
Long Flips (LF) 2.17 ± 1.17 3.33 ± 0.30 NS 4.50 ± 0.42** 8.50 ± 0.42*** 12.00 ± 2.26***
Total Penile Reflexes (TPR) 15.01 ± 3.55 16.33 ± 1.21 NS 18.33 ± 1.34* 29.82 ± 1.53*** 48.30 ± 9.03***
Tabular values are mean ± SEM, n = 6 (number of animals in each group); significant difference from control, NS: Not significant. *P < 0.05, **P < 0.01; ***P < 0.001.
Adverse effects
No treatment-related overt signs of toxicity, stress and changes in behaviour were observed. The food and water intake of all the treated animals remained similar to those of the control group.
Acute toxicity studies
No mortality and changes in the behaviour were observed in all the treated and control groups of mice up to a dose of 4000 mg/kg.
Discussion
The present study was aimed to investigate the aphrodisiac effect of nutmeg extract (50% ethanolic) along with its acute toxicity using various animal models. The study exhibits a marked change in sexual behaviour of male rats. The results of the present investigations show that the test drug significantly increased the Mounting Frequency (MF) and Intromission Frequency (IF) as compared to control group. However, the standard drug produced a greater increase in thse parameters. The MF and IF are considered the indices of both libido and potency. Thus, the increase in the MF and IF, indicates that nutmeg, along with increasing libido, probably also increases the potency. The significant increase in the Ejaculatory Latency (EL) suggests that the extract and standard drug prolonged the duration of coitus. The significant increase in the EL in both first and second series as well as the decrease in the Post Ejaculatory Interval (PEI), i.e. the refractory period between first and second series of mating, suggest that the test drug intensified sexual activity in a sustained manner. The test drug also caused a significant reduction in the Mounting Latency (ML) and Intromission Latency (IL) as compared to control animals, while a highly significant decrease was observed in the ML of animals treated with the referent drug. This also provides an evidence for aphrodisiac effect of the test drug. These findings show that the test drug produces a striking enhancement of over-all sexual performance of normal animals.
MF after penile anesthetization of rats is a reliable index of 'pure' libido and the penile reflexes of the rats are a good model of 'pure' potency [29]. Therefore, in the present study the extract was also studied for effect on these components of sexual behaviour.
The effect of the test drug on libido was studied by assessing the MF after genital anaesthetization which does away with the reinforcing effect of genital sensation thus affording the study of pure libido or intrinsic sexual desire. During the experiment the test drug produced a significant increase in the MF of sexually normal male rats. Whereas, the MF was much reduced in control, test and standard animals in comparison with the MF of corresponding groups in mating behaviour test where the penis had not been anaesthetized. However, the test for libido revealed that Intromission and Ejaculation were absent in all groups of animals, as the genital sensations which are absent due to penile anaesthetization are necessary for the development of these two events. Thus, it may be inferred that the test drug produced a striking increase in 'pure' libido.
The test for potency exhibited that the extract significantly increased the frequency of all the components of penile reflexes: Erections (E), Quick Flips (QF) and Long Flips (LF) as compared to control group, but comparatively less than the standard drug. The aggregate of these penile reflexes (TPR) was also significantly increased in both test and standard animals. This indicates that the test drug increases 'pure'potency also.
Although the effect of the extract on 'pure' libido and 'pure' potency was evaluated by using two different methods, a rough comparison of the results indicates that the test drug augmented both libido and to an equal extent potency. The positive inferences from the specific tests for libido and potency substantiate the indications of the mating behaviour test to show in a rather conclusive manner that the test drug enhances both the libido and potency in normal male animals. These conclusions are further supported by an earlier study reporting libido and potency increasing effect of nutmeg in mice[23].
In addition, nutmeg, a well know spice and a herbal drug is widely used in Unani medicine without any known or recorded toxicity in the management of male sexual disorders. Such herbal drugs may be directly used, without any toxicity testing. However, when an extract or active fraction of such drug is used it is better to evaluate possible toxicity. Although it is the normal practice to determine the LD50 value, now it is acceptable to limit the study to an acute toxicity test using multiple doses including reasonably high doses of the drug [32].
In this connection the test drug was also subjected to an acute toxicity testing and it was tested up to a high concentration of 4000 mg/kg, orally (eight times more than the aphrodisiac dose, evaluated in the present study). Even at this dose the extract did not produce signs of toxicity or treatment- related adverse effects in the tests for aphrodisiac activity. This suggests that its short term use for this purpose is apparently safe.
With regard to the mechanism of the test drug, it is difficult to explain the exact mechanism responsible for improving sexual function. The drug induced changes in neurotransmitter level or their action at cellular level could change sexual behaviour [33]. Nutmeg is mentioned in ethnomedical literature as nervous stimulant [8]. The extract (50% ethanolic) of nutmeg is also reported to have nervous stimulant action in male albino rats [34]. Thus, the resultant aphrodisiac activity of the test drug might be attributed to its nervous stimulating property. Preliminary phytochemical studies indicate the presence of sterols, phenols, alkaloids and amino acids in the extract. Hence, the sexual function improving effect of the test drug might be due to the presence of such compounds. Moreover, nutmeg, merits further studies for detailed sexual function improving activities, especially at higher doses. In addition, further research is also needed for the identification of its active constituent (s) responsible for sexual function improving activities and the mechanism by which it augments sexual function.
Conclusion
The resultant significant and sustained increase in the sexual activity of male rats, with out any conspicuous adverse effects and toxicity, suggests that nutmeg possesses clinically applicable aphrodisiac activity, and also lends support to the claims for its traditional usage as sexual function enhancing medicine. Further, the study also indicates that the aphrodisiac effects of the test drug may be due to its nervous stimulating property. Thus, it may prove to be an effective and safe alternative remedy in sexual disorders.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
T-Supervised the design and coordination of the study.
SA- Practically conducted the design of the study.
AL- Participated and performed the statistical analysis.
IA- Participated and performed the statistical analysis.
KA-Supervised the design and coordination of the study.
All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors are thankful to the Department of Ilmul Advia of Aligarh Muslim University, Aligarh, India for providing all the facilities to carry out the study. They are also grateful to Professor S. H. Afaq (In-charge, Pharmacognosy Section) for his valuable suggestions.
==== Refs
Council of Scientific and Industrial Research The Wealth of India (Raw Materials), 1962 5 New Delhi: CSIR Publication 474 479
Gils CV Cox PA Ethnobotany of nutmeg in the spice islands J Ethnopharmacol 1994 42 117 124 8072304 10.1016/0378-8741(94)90105-8
Evans WC Trease and Evans' Pharmacognosy 2003 15 Philadelphia: Elsevier Science Limited 269 270
Burkill IH Dictionary of the Economic Products of the Malay Peninsula 2 2 London: I-Z, Crown Agents 1554 1556
Ghani N Khazeenatul Advia 1921 2 1 Lucknow: Matba Munshi Nawal Kishore 241 242
Khory RN Katrak NN Materia Medica of India and their therapeutics 1985 Delhi: Neeraj Publishing House 524 532
Chopra RN Chopra IC Handa KL Kapur LD Indigenous Drug of India 1958 Kolkata: UN Dhur and Sons Pvt. Ltd 201
Ainslie W Materia Indica 1979 1 Delhi: Neeraj Publishing House 249 252
Nadkarni AK Dr. K.M.Nadkarni's Indian Materia Medica 2002 1 3 Mumbai: Popular Prakashan Pvt. Ltd 830 834 12455700
Antaki DZ Tazkirah-ulil-Albab 1930 2 Cairo: Matba Aamirah Sharfiyah 103
Lindley J Flora Medica 1981 New Delhi: Ajay Book Service 21 25
Murcia MA Egea I Romojaro F Parras P Jimenez AM Martinez-Tome M Antioxidant evaluation in dessert spices compared with common food additives. Influence of irradiation procedure J Agric Food Chem 2004 52 1872 1881 15053523 10.1021/jf0303114
Isogai A Suzuki A Tamura S Structure of dimeric phenoxypropanoids from Myristica fragrans Agar Biol Chem 1973 37 193 194
Janssen J Laeckman GM Nutmeg oil: Identification and quantification of its most active of platelet constituents as inhibitors aggregation J Ethnopharmacol 1990 29 179 188 2115612 10.1016/0378-8741(90)90054-W
Kumaravelu Subramanyam S Dakshinmurthy DP Devraj NS The antioxidant effect of eugenol on carbon-tetrachloride-induced erythrocyte damage in rats J Nut Biochem 1996 7 23 28 10.1016/0955-2863(95)00162-X
Criddle DN Madeira SV Soares de Murta M independent Endothelium-dependent and -vasodilator effects of eugenol in the rat mesenteric vascular bed J Pharm Pharmacol 2003 55 359 365 12724042 10.1211/002235702694
Damiani CE Rossoni LV Vasallo DV Vasorelaxant effect of eugenol on rat thorasic aorta Vascular Pharmacol 2003 40 59 66 10.1016/S1537-1891(02)00311-7
Jenkins GL Kneval AM Digangi FE Quantitative Pharmaceutical Chemistry 1967 New York: MacGraw Hill Book Company
Olajide OA Ajayi FF Ekhelar AL Awe SO Makinde JM Alada ARA Biological effect of Myristica fragrans (nutmeg) extract Phytother Res 1999 13 344 345 10404545 10.1002/(SICI)1099-1573(199906)13:4<344::AID-PTR436>3.0.CO;2-E
Wessinger J Effect of nutmeg, aspirin, chlorpromazine and lithium on normal intestinal transport Proc West Pharmacol Soc 1985 28 267 273
Grover JK Khandkar S Vats V Dhunnoo Y Das D Pharmacological studies on Myristica fragrans -antidiarrheal, hypnotic, analgesic and hemodynamic (blood pressure) parameters Methods Find Exp Clin Pharmacol 2002 24 675 680 12616960 10.1358/mf.2002.24.10.802317
Parle M Dhingra D Kulkarni SK Improvement of mouse memory by Myrisatica fragrans seeds J Med Food 2004 7 157 161 15298762 10.1089/1096620041224193
Tajuddin Ahmad S Latif A Qasmi IA Aphrodisiac activity of 50% ethanolic extracts of Myristica fragrans Houtt. (nutmeg) and Syzygium aromaticum (L.) Merr. & Perry. (clove) in male mice: a comparative study BMC Complement Altern Med 2003 3 6 14567759 10.1186/1472-6882-3-6
Freirich EJ Quantitative comparision of toxicity of anti-cancer agents in mous, rat, dog, monkey and man Cancer Chemotherapy Report 1968 50 219 244
Attar HZ Ikhtiyarat – e – Badiyee 1638 109 1 Lucknow : Matba Munshi Nawal Kishore
Khan M A Moheet – e – Azam 1893 2 1 Kanpur: Matba Nizami 104 105
Dewsbury DA Davis HN Jr Effect of reserpine on the copulatory behaviour of Male rats Physiol Behav 1970 5 1331 1333 5524518 10.1016/0031-9384(70)90050-8
Szechtman H Moshe H Rabi S Sexual behaviour Pain sensitivity and stimulates endogenous opioid in male rats Eur J Pharmacol 1981 70 279 285 6262094 10.1016/0014-2999(81)90161-8
Davidson JM Zewi H Sexology: Sexual biology, behaviour and therapy selected papers of Fifth World Congress of Sexology: 1981; Jerusalem 1982 Excerpta Medica, Amesterdem-Princeton-Oxford 42 47
Hart BL Haugen CM Activation of sexual reflexes in male rats by spinal implementation of testosterone Physiol Behav 1968 3 735 738 10.1016/0031-9384(68)90144-3
Hart BL Activation of sexual reflexes in male rats by dihydrotestosterone but not estrogen Physiol Behav 1979 23 107 109 515197 10.1016/0031-9384(79)90129-X
Babu V Gangudevi T Subramanium A Antidiabetic activity of ethanol extract of Cassia kleinii leaf in streptozotocin -induced diabetic rats and isolation of an active fraction and toxicity evaluation of the extract Ind J Pharmacol 2003 35 290 296
Suresh Kumar PK Subramoniam A Pushpangadan P Aphrodisiacs activity of Vanda tessellate (Roxb.) Hook ex don. extract in male mice Ind J Pharmacol 2000 32 300 304
Ahmad S A scientific study of some unani aphrodisiacs MD Thesis 2001 Aligarh: Faculty of Unani medicine, Aligarh Muslim University
|
16033651
|
PMC1187868
|
CC BY
|
2021-01-04 16:31:46
|
no
|
BMC Complement Altern Med. 2005 Jul 20; 5:16
|
utf-8
|
BMC Complement Altern Med
| 2,005 |
10.1186/1472-6882-5-16
|
oa_comm
|
==== Front
BMC AnesthesiolBMC Anesthesiology1471-2253BioMed Central London 1471-2253-5-101600017310.1186/1471-2253-5-10Research ArticleA randomized, controlled trial of spinal endoscopic adhesiolysis in chronic refractory low back and lower extremity pain [ISRCTN 16558617] Manchikanti Laxmaiah [email protected] Mark V [email protected] Jose J [email protected] Vidya Sagar [email protected] Kim S [email protected] Carla D [email protected] Doris E [email protected] Sue R [email protected] Pain Management Center of Paducah, 2831 Lone Oak Road, Paducah, Kentucky, 42003, USA2 Case University School of Medicine, University Hospitals of Cleveland, 11100 Euclid Avenue, Cleveland, OH, 44106, USA2005 6 7 2005 5 10 10 8 11 2004 6 7 2005 Copyright © 2005 Manchikanti et al; licensee BioMed Central Ltd.2005Manchikanti 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
Postoperative epidural fibrosis may contribute to between 5% to 60% of the poor surgical outcomes following decompressive surgery. Correlations have been reported between epidural scarring and radicular pain, poor surgical outcomes, and a lack of any form of surgical treatment. The use of spinal endoscopic adhesiolysis in recent years in the management of chronic refractory low back and lower extremity pain has been described.
Methods
A prospective, randomized, double-blind trial was conducted to determine the outcome of spinal endoscopic adhesiolysis to reduce pain and improve function and psychological status in patients with chronic refractory low back and lower extremity pain. A total of 83 patients were evaluated, with 33 patients in Group I and 50 patients in Group II. Group I served as the control, with endoscopy into the sacral level without adhesiolysis, followed by injection of local anesthetic and steroid. Group II received spinal endoscopic adhesiolysis, followed by injection of local anesthetic and steroid.
Results
Among the 50 patients in the treatment group receiving spinal endoscopic adhesiolysis, significant improvement without adverse effects was shown in 80% at 3 months, 56% at 6 months, and 48% at 12 months. The control group showed improvement in 33% of the patients at one month and none thereafter. Based on the definition that less than 6 months of relief is considered short-term and longer than 6 months of relief is considered long-term, a significant number of patients obtained long-term relief with improvement in pain, functional status, and psychological status.
Conclusion
Spinal endoscopic adhesiolysis with targeted delivery of local anesthetic and steroid is an effective treatment in a significant number of patients with chronic low back and lower extremity pain without major adverse effects.
==== Body
Background
Postoperative epidural fibrosis, the formation of dense scar tissue adjacent to the dura mater following surgical laminectomy, may play a role in up to 60% of the poor surgical outcomes following decompressive surgery [1-3]. A correlation between peridural scarring and radicular pain [4-7], and poor clinical outcomes [8,9] has been reported by some, while others [10-13] have questioned the role of epidural fibrosis as a causative factor. Increased complication rates have been reported with revision spine surgery with increased occurrence of dural tears, nerve root injury, and bleeding [14,15]. Phillips and Cunningham [16] reported that no form of surgical treatment or adhesion lysis procedure was safe or effective for post-lumbar laminectomy syndrome.
Epidural fibrosis results from the invasion of postoperative hematoma by dense fibrous tissue originating from the periosteum and within the deep surface of the paravertebral musculature [17,18]. Epidural fibrosis may extend into the neural canal and adhere to the dura mater and nerve roots, with mechanical tethering of nerve roots or dura by adhesions, which may in turn contribute to persistent back and leg pain following lumbar laminectomy. However, epidural fibrosis also may develop without surgical intervention, secondary to annular tear, hematoma, infection, or intrathecal contrast media [18-20]. Perineural fibrosis can render nerve roots hyperesthetic and hypersensitive to compression forces by interfering with cerebrospinal fluid-mediated nutrition [6] or by making the nerve susceptible to injury [7].
A moderate proportion of patients show improvement in pain and functional level with interventional pain management procedures, including fluoroscopically-directed epidural steroid injections and percutaneous adhesiolysis utilizing a special catheter [21,22]. In addition, initial clinical studies of spinal endoscopic adhesiolysis [23-29] and a preliminary report of a randomized controlled trial [30] showed improved clinical outcomes. However, a recent prospective, randomized, double-blind trial [31] comparing caudal epidural with targeted steroid placement on affected nerve roots during spinal endoscopy for chronic sciatica, concluded that targeted placement of steroid did not significantly reduce pain intensity and anxiety and depression compared with caudal epidural steroid injection.
Most studies utilized post lumbar laminectomy syndrome or epidural fibrosis as inclusion criteria, whereas one study [28] included patients with lumbar spinal stenosis, and another [31] exclusively studied patients without history of surgery. In addition, some studies [26,27,30] specifically described inclusion criteria as patients without long-term improvement following fluoroscopically directed epidural steroid injections and one-day percutaneous adhesiolysis. Consequently, these studies represent heterogenous populations. Even though retrospective evaluations [26,27] have shown the effectiveness of spinal endoscopic adhesiolysis in patients after lack of long-term effect following fluoroscopically-directed caudal epidural steroid injections and one-day percutaneous adhesiolysis, the effectiveness has not been demonstrated in controlled trials.
Spinal endoscopy also has been utilized for diagnostic purposes. Even though multiple authors have described various types of findings, including the identification of inflammation with an endoscope, neither the reliability nor the clinical utility of spinal endoscope as a diagnostic tool has been established [23-25,28,29]. Consequently, no attempt was made to evaluate the diagnostic utility of spinal endoscopy.
This randomized, double-blind, controlled trial of spinal endoscopic adhesiolysis and targeted delivery of steroids was designed to evaluate their effectiveness in patients with chronic low back and lower extremity pain who lacked significant response to fluoroscopically-directed epidural steroid injections and one-day percutaneous adhesiolysis with hypertonic saline neurolysis, as well as to other conservative modalities of treatment.
Methods
This study was designed to evaluate the effectiveness of spinal endoscopic adhesiolysis in chronic, refractory low back and lower extremity pain. The study was undertaken in an interventional pain management practice (a specialty referral center) in a private practice setting, in accordance with the guidelines for randomized controlled trials [32,33], and the quality checklists of systematic reviews [32,34-39]. The trial also was designed to meet criteria of a pragmatic or a practical clinical trial [30,40,41]. The protocol was approved by the Institutional Review Board of the Ambulatory Surgery Center where the study was conducted. The objective was to evaluate the effectiveness of spinal endoscopic adhesiolysis compared to caudal epidural steroid injection. The design consisted of a control group and an intervention group. Group I (control group) was treated with introduction of the spinal endoscope up to the S3 level, followed by injection of a local anesthetic and steroid. Group II (intervention group) was treated with appropriate spinal endoscopic adhesiolysis at L4, L5 or S1 level(s) unilaterally or bilaterally based on symptomatology, followed by a targeted injection of local anesthetic and steroid.
Inclusion and exclusion criteria
The majority of participants in this study were identified from existing patients of the interventional pain management practice. Eligible new patients were screened and identified as candidates for the program.
Inclusion criteria
patients between 18 and 65 years of age with a history of chronic low back and lower extremity pain of at least two year's duration, without facet joint pain based on controlled comparative local anesthetic blocks, and without significant improvement with conservative treatment including fluoroscopically-directed epidural injections and one-day percutaneous adhesiolysis with hypertonic saline neurolysis, and willingness to participate in the clinical trial were enrolled.
Criteria for lack of significant response to caudal epidural steroid injections was pain relief (≥ 50%) for one week or less following a second epidural steroid injection, and relief of four weeks or less following a third epidural steroid injection or any of subsequent epidural steroid injections. Criteria for lack of response to one-day percutaneous adhesiolysis with hypertonic saline neurolysis was considered as lack of response to the first adhesiolysis procedure, less than two months of pain relief (≥ 50%) following the second or subsequent procedures.
Exclusion criteria
patients with cauda equina syndrome, compressive radiculopathy, surgical intervention in the previous six months, opioid abuse and dependency evaluated by adherence monitoring, including random drug testing and opioid use of no greater than hydrocodone 100 mg per day, methadone 60 mg or morphine 100 mg or equivalent doses of other drugs; uncontrolled major depression or psychiatric disorders; uncontrolled or acute medical illnesses including severe cardiac, pulmonary, or other disorders; chronic conditions that could interfere with the interpretations of the outcome assessments such as severe hip or knee arthritis, neuropathy, or other disorders; pregnant or lactating women; history of adverse reaction to local anesthetic or steroids; inability to understand the informed consent and protocol; or inability to be positioned in the prone position during the procedure [30].
Evaluation
All patients were provided with the protocol and the informed consent document approved by the Institutional Review Board for this study. The informed consent document described the details of the trial.
The screening evaluation consisted of demographic data, medical/surgical history with co-existing diseases, radiographic investigations, physical examination, psychological evaluation with Pain Patient Profile (P-3®), Visual Analogue Scale (VAS) pain scores, work status, Oswestry Disability Index 2.0, and lumbar spine range of motion with ARCON ROM computerized dual inclinometer system, based on AMA "Guides to the Evaluation of Permanent Impairment" validity criterion utilizing three consecutive measurements with ± 5° or ± 10% of mean value.
Interventions
All patients in both groups were provided identical preparation. All procedures were performed using fluoroscopy in an ambulatory surgery center in sterile operating rooms by one physician (LM).
Procedure
The procedure included appropriate preparation with intravenous access, pre-procedure antibiotic administration, sterile preparation, and appropriate sedation with fentanyl and midazolam. Access to the epidural space was obtained with a RK® needle. An epidurogram was obtained which identified filling defects and/or epidural fibrosis. Adhesiolysis was carried out in the intervention group utilizing the myeloscope® spinal endoscopic video-guided catheter system and introducer system, with final positioning of the fiberoptic endoscope on the side and level of the defect and the source of pain with an additional injection of contrast to identify adhesiolysis, followed by targeted injection of local anesthetic and steroid.
Following initial epidurography in the control group, a 0.9 mm guidewire was inserted through the needle, which was advanced under fluoroscopic guidance to S3 level. Then, a 2-mm × 17.8-cm dilator with catheter (sheath) was passed over the guidewire again up to S3. At that time, a 0.8-mm fiberoptic spinal endoscopic video-guided system was introduced into the catheter through the valve and was advanced until the tip was positioned at the distal end of the catheter through the valve, as determined by video and fluoroscopic images not to exceed S3. Following this, 10 mL of 1% lidocaine and 6 mg to 12 mg of betamethasone or 40 mg to 80 mg of methylprednisolone were injected through the epiduroscope. Following the completion of the procedure, the endoscope was removed and appropriate sterile Bioclusive dressing was applied.
In the intervention group, following initial epidurography, a 0.9-mm guidewire was inserted through the needle (occasionally facilitated by a small incision with a #11 straight blade), which was advanced under fluoroscopic guidance to the level of suspected pathology. Following this, a 2-mm × 17.8-cm dilator with catheter (sheath) was passed over the guidewire. Once the catheter was advanced to the tip of the guidewire, the wire was removed. A 0.8-mm fiberoptic spinal endoscopic video-guided system was introduced into the catheter through the valve and advanced until the tip was positioned at the distal end of the catheter, as determined by video and fluoroscopic images. In conjunction with gentle irrigation using normal saline, the catheter and fiberoptic myeloscope were manipulated and rotated in multiple directions, with visualization of the nerve roots at various levels. Painful nerve root was confirmed by endoscopic manipulation, based on pre-procedural clinical and radiographic evaluation. Adhesiolysis and decompression were carried out by distension of the epidural space with normal saline and by mechanical means utilizing the fiberoptic endoscope. Adhesiolysis was confirmed by injection of non-ionic contrast material (Omnipaque 240®) and an epidurogram was performed on at least two occasions. The volume of sodium chloride solution utilized for irrigation was closely monitored. The protocol limited the total volume of contrast and sodium chloride solution not to exceed 100 mL. Adhesiolysis was limited to L4, L5 or S1 levels, either unilaterally or bilaterally. Following completion of the procedure, 4 mL to 8 mL of lidocaine 1%, preservative free, mixed with either 6 mg or 12 mg of betamethasone or 40 mg or 80 mg of methylprednisolone was injected after assuring that there was no evidence of subarachnoid leakage of contrast. The injection of betamethasone or methylprednisolone was based on its availability in the market. Methylprednisolone was utilized if betamethasone was not available. If pathology was identified at multiple levels, the procedure was carried out at those levels, and the injectate was given in divided doses. If there was a question of subarachnoid leakage of the contrast, a Racz catheter® was passed into the epidural space, and a mixture of local anesthetic was injected very slowly in incremental doses, followed by injection of the steroid if satisfactory follow-up was obtained without any subarachnoid blockade. Following the injection of local anesthetic and steroid, the scope was removed and appropriate sterile Bioclusive dressing was applied.
Co-interventions
No specific co-interventions were offered. Baseline drug therapy was allowed to be continued with no changes being made towards increasing opioids until after the unblinding and/or documented failure of intervention. However, opioid decreases were implemented based on improvement in functional status and reduction in pain following the interventions. Self-directed exercises as tolerated were also prescribed.
Outcomes assessment
Outcomes were assessed at 3-month, 6-month, and 12-month intervals post-treatment with the Visual Analogue Scale (VAS) pain scale, Oswestry Disability Index 2.0, work status, opioid intake, range of motion measurement by ARCON ROM computerized evaluation, and psychological evaluation by P-3. They were compared to baseline within both groups and with each other at various time intervals. Duration of relief was judged to be short-term, if relief was less than 6 months. If relief lasted for at least 6 months, it was considered long-term. Significant relief was defined as pain relief of 50% or greater.
VAS was measured on a 10 cm scale. P-3 psychological evaluation [42] and Oswestry Disability Index 2.0 [43] were assessed by administration of appropriate questionnaires. Range of motion was evaluated by a certified physical therapist, blinded to the type of treatment. Based on P-3, scores of 55 or higher were considered positive for a diagnosis of depression, whereas, scores of 56 or higher were considered to provide the diagnosis of anxiety or somatization.
Opioid intake was determined as none, mild, moderate, or heavy based on the dosage, frequency and schedule of the drug as follows: Considered as mild was an intake of Schedule IV opioids, i.e., propoxyphene napsylate, pentazocine hydrochloride, tramadol hydrochloride up to a maximum of four times, or hydrocodone less than 40 mg per day; considered moderate was an intake of Schedule III opioids, i.e., hydrocodone, up to 40 mg per day; and considered heavy was an intake of Schedule II opioids, i.e., oxycodone, morphine, meperidine, transdermal fentanyl, and methadone, in any dosage.
Employment and work status (employed, unemployed, housewife, disabled, and retired) were determined from the pre-treatment and post-treatment work status. Only employed and unemployed patients were considered to be eligible for employment, whereas disabled patients and retired patients were considered not employable. Patients in the "housewife" category were considered neither employable nor unemployable.
Statistical methods
Study design
Randomization was 2:3 with two patients randomized to the control group (Group I), for every three patients randomized to the spinal endoscopic adhesiolysis group (Group II). Randomization was performed by the statistician using a computer-generated random allocation sequence in blocks of 15 patients.
The random allocation was concealed from the physician doing the procedure and personnel in the operating room until the intervention. Randomization was not revealed to the personnel in the recovery room or to the reviewing physician. After treatment, the patient was never in contact with anyone with knowledge to the randomization assignment, until after the unblinding was performed.
Unblinding was considered at a patient's request and/or treatment was considered a failure at 3 months or later. All other patients were unblinded at 12 months. Patients were also given an option to discontinue or to withdraw from the study for any and all reasons. They were considered withdrawn if follow-up was lost.
Intent-to-treat analysis
An intent-to-treat analysis was performed by including all subjects by carrying forward the last observation.
Statistical analysis
Demographic data were analyzed by means of the Student's t test and the chi-squared test. Fischer's exact test was used wherever the expected value was less than five. For analyzing Outcome measurement based on Visual Analogue Scale (VAS) Report and Oswestry Disability Index, range of motion (ROM), depression, anxiety and somatization scores, student's t test (parametric) and The Mann-Whitney U test (non-parametric) were used to test mean differences between groups. A paired t test and Wilcoxon signed-rank test were used to compare pre-and post-treatment results for individual patients. When results from both parametric and non-parametric tests were similar, P values from the parametric tests were reported in the tables and text. Results were considered statistically significant if the P value was less than 0.05.
Results
The study was conducted from January 2002 through December 2003. As per the protocol, initial results were published in 2003; this preliminary report included a 6 month follow up with 16 patients in caudal epidural steroid injection group, and 23 patients in spinal endoscopy group [30]. To facilitate this publication, results of two patients in the control group and 12 patients in the intervention group were unblinded by the statistician for purposes of the evaluation and preliminary publication. This unblinding was not revealed to investigators, other staff, and the participants of the study. Consequently, 39 patients reported at 6 months were also included in the present report.
A diagram illustrating flow of the trial is depicted as Figure 1. In the control group, one patient was lost to follow-up after 3 months. In the intervention group, two patients withdrew from the study. One patient experienced no improvement, withdrew from the study, and underwent further surgical intervention. The second patient in the intervention group reported no significant relief and withdrew from the study and refused further follow-up. Intent to treat analysis was performed by using baseline or last follow-up data in both groups. All patients received only one treatment during the study period. Patients were considered withdrawn if they received any other interventional techniques. Last follow-up was utilized for analysis with 3-month data at 6 months and 12 months in 15 patients, and 6-month data at 12 months in two patients in the control group. In the intervention group, baseline data was utilized in two patients. In the intervention group, of the 11 patients unblinded at 3 months, five of them participated in outcomes assessment at 6 months and 12 months while six of them participated only at 6 months assessment. Thus, 3 month assessment results were utilized at 6 months and 12 months, whereas the results of 6 month assessments were utilized at 12 months in eight patients (six patients from 3 month unblinding and two patients from 6 month unblinding).
Figure 1 Schematic depiction of patient flow during the trial.
Demographic characteristics
Table 1 illustrates the demographic characteristics. All the patients presented with back and lower extremity pain. Most patients had pain with unilateral symptoms, while bilateral symptoms were seen in 12% of the patients in each group.
Table 1 Demographic characteristics
Group I Group II
Number of patients 33 50
Age (Years) Mean ± SD 47 ± 9.4 50 ± 9.0
Gender Male 54% (18) 36% (18)
Female 46% (15) 64% (32)
Height (Inches) Mean ± SD 66 ± 3.6 66 ± 3.5
Weight (Lbs) Mean ± SD 181 ± 42.4 174 ± 36.8
Duration of pain (years) Mean ± SD 12.4 ± 5.9 11.8 ± 6.5
Mode of onset of the pain Traumatic 39% (13) 46% (23)
Non-traumatic 61% (20) 54% (27)
Back and lower extremity pain 100% (33) 100% (50)
Bilateral pain 12% (4) 12% (6)
History of previous surgery 73% (24) 84% (42)
Epidural fibrosis on MRI 73% (24) 84% (42)
Disc herniation on MRI 12% (4) 10% (5)
Procedural characteristics
In the control group, of the four patients (12%) with bilateral back and lower extremity pain, two patients received 12 mg of betamethasone and two patients received 80 mg of methylprednisolone. Of the remaining 29 patients, 16 patients received 6 mg of betamethasone and 13 patients received 40 mg of methylprednisolone. The volume of contrast was 8.6 ± 1.25 mL with a range of 8 to 12 mL (Table 2).
Table 2 Description of procedural characteristics
Group I 33 Group II 50
Betamethasone
12 mg 6% (2) 4% (2)
6 mg 48% (16) 56% (28)
Total 55% (18) 60% (30)
Methylprednisolone
80 mg 6% (2) 8% (4)
40 mg 39% (13) 32% (16)
Total 45% (15) 40% (20)
Contrast in mL Mean ± SD 8.6 ± 1.25 11.2 ± 2.74
Range 8 – 12 8 – 16
Sodium chloride solution for irrigation in mL Mean ± SD None 55.0 ± 11.07
Range 35 – 70
() indicates number of patients
In the intervention group, adhesiolysis was performed bilaterally in six patients (12%). Adhesiolysis was performed at one level in two patients, at two levels in 47 patients, and at four levels in one patient. Unilateral adhesiolysis was performed in only one patient at L4 level. No bilateral adhesiolysis was performed at L4. Most commonly, adhesiolysis was performed at L5 and S1. The volume of sodium chloride solution injected was 55.0 ± 11.07 mL with a range of 35 to 70 mL. The volume of contrast was 11.2 ± 2.74 mL with volumes ranging from 8 to 16 mL. There were no cases of subarachnoid blockade identified prior to injection of local anesthetic and steroid. Thus, although it was available as part of the protocol, a Racz catheter was not used for any patient procedure in the study. Twelve milligrams of betamethasone in 8 mL of 1% lidocaine was injected in two patients and 80 mg of methylprednisolone in 8 mL of 1% lidocaine was injected in four patients, 6 mg of betamethasone in 4 mL of 1% lidocaine was injected in 28 patients, and 40 mg of methylprednisolone in 4 mL of 1% lidocaine was injected in 16 patients (Table 2).
Outcome measures
A significant proportion of patients in the spinal endoscopic adhesiolysis group showed pain relief compared to the control group, as well as compared to the baseline findings (Fig 2).
Figure 2 Outcome measurement based on Visual Analogue Scale report.
Significant pain relief (≥ 50%) in months was calculated for both groups. Calculations for all patients showed that significant relief was seen for 0.7 ± 0.73 months in the control group, whereas, 7.6 ± 4.7 months of relief was noted for the intervention group. Significant pain relief was longer in the intervention group. Duration of significant relief (≥ 50%) (mean ± SD) was 9.3 ± 3.6 months in patients considered as successful (40 of 50).
In the control group, the proportion of patients with significant relief greater than 50% at 1 month was 33%, and at 3 months, 6 months, and 12 months was 0. By contrast, in the intervention group relief was 90% at 1 month, 80% at 3 months, 56% at 6 months, and 48% at 12 months (Fig. 3).
Figure 3 Proportion of patients with significant relief (≥ 50%) at 1 month, 3 months, 6 months and 12 months.
Functional outcome measurement was carried out based on Oswestry Disability Index 2.0. Significant improvements were seen in the intervention group compared to baseline in the same group, as well as compared to the control group (Fig. 4).
Figure 4 The Outcome Measurement Based on Oswestry Disability Index.
Analysis of range of motion evaluations showed significant improvements in the intervention group compared to the baseline, as well as the control group at intervals of 3 months, 6 months, and 12 months (Table 3).
Table 3 Analysis of range of motion evaluation
Baseline 3 months 6 months 12 months
Group I Group II Group I Group II Group I Group II Group I Group II
33 50 33 50 33 50 33 50
Flexion (Normal 60°) Mean ± SD 25.4 ± 10.0 25.9 ± 11.4 26.6 ± 10.3 35.8*# ± 11.7 25.8 ± 10.4 36.7*# ± 13.7 25.6 ± 10.3 35.7*# ± 14.4
Extension (Normal 25°) Mean ± SD 9.7 ± 3.9 9.0 ± 3.3 10.9 ± 5.1 14.7*# ± 5.3 10.5 ± 5.1 15.8*# ± 6.5 10.9 ± 5.3 16.3*# ± 7.0
Lateral Flexion (Normal 25°) Mean ± SD 8.1 ± 2.9 8.4 ± 2.8 7.9 ± 3.0 14.0*# ± 5.3 7.8 ± 3.0 14.6*# ± 6.1 7.7 ± 2.8 15.1*# ± 6.8
* Indicates significant difference with Group I (P = 0.002)
# Indicates significant difference within the Group compared to baseline (P = 0.001)
Table 4 illustrates psychological outcomes of depression, anxiety, and somatization derived from P-3 scores. Significant improvement was noted in psychological parameters in the intervention group compared to the control group, as well as to baseline status in the treatment group.
Table 4 Analysis of psychological status
Baseline 12 months
Group I Group II Group I Group II
33 50 33 50
Depression Diagnosis (≥ 55) 61% (20) 68% (34) 58% (19) 34%*(17)
Score Mean ± SD 56.9 ± 8.8 57.0 ± 9.9 55.5 ± 10.6 47.8*# ± 10.4
Anxiety Diagnosis (≥ 56) 58% (19) 62% (31) 55% (18) 28%*(14)
Score Mean ± SD 55.6 ± 10.6 55.9 ± 11.9 54.9 ± 9.9 46.8*# ± 12.1
Somatization Diagnosis (≥ 56) 58% (19) 74% (34) 52% (17) 30% (18)
Score Mean ± SD 55.4 ± 8.9 56.6 ± 11.4 55.9 ± 10.4 47.8*# ± 12.3
* Indicates significant difference with Group I (P = < 0.05)
# Indicates significant difference with Baseline values within the Group (P = < 0.001)
Patients were evaluated for opioid intake, which was rated from none to significant as described in the methods section. Significant opioid intake was 40% in Group II at the end of 12 months, compared to 74% at baseline. For Group I, significant opioid usage was 55% at 12 months, compared to 61% at baseline.
Evaluation of employment status showed that in the intervention group employment increased to 32% at 12 months from 2% at baseline as compared to 6% at baseline and at 12 months in the control group. As illustrated in Table 5, almost all the patients deemed employable in the intervention group were employed at 12 months, in contrast to no change noted in the control group. In addition, in the intervention group, eight patients disabled at baseline were also employed at 12 months. There were no patients in this study with active workers' compensation injury cases or litigation.
Table 5 Change in proportion of patients with employment status
Employment Status Group I Group II
Baseline At 12 months Baseline At 12 months
Employed 2 (6%) 2 (6%) 1 (2%) 16 (32%)*
Unemployed 2 (6%) 2 (6%) 8 (16%) 1 (2%)
Housewife 2 (6%) 2 (6%) 1 (2%) 1 (2%)
Disabled 26 (79%) 26 (79%) 38 (76%) 30 (60%)
Over 65 (yrs) 1 (3%) 1 (3%) 2 (4%) 2 (4%)
Total 33 50 33 50
*Indicates significant difference (P < 0.01)
Blinding
The blinding was judged to be satisfactory. Following the treatment, and within one hour prior to discharge, patients were asked what treatment they believed they had received. Twenty-six of 33 patients in Group I and 42 of 50 patients in Group II believed they had received spinal endoscopy. Two patients in Group I and two patients in Group II were unable to theorize as to which procedure they may have received. The remaining patients guessed the wrong treatment. There was no significant difference among the groups as to whether they believed they had received the endoscopy or epidural steroid injection.
Adverse events
There was one case of subarachnoid block in Group II, identified after completion of the procedure and injection of local anesthetic and steroid. No adverse effects were noted in this patient. There were no other adverse events noted.
Discussion
This randomized, double-blind, controlled evaluation demonstrated that following spinal endoscopic adhesiolysis a significant proportion of patients with chronic, refractory low back and lower extremity pain experienced significant pain relief (≥ 50%) at 3 months (80%), 6 months (56%), and at 12 months (48%), compared to the control group with only 33% of patients showing improvement at 1 month, and none thereafter. Associated improvements in VAS scores, Oswestry Disability Index, range of motion, and psychological status were also noted as compared to baseline measurements and results of the control group. These results are important in that the patients in this study represented a subset of patients who have not only failed multiple conservative modalities of management but also lacked significant, or long-term, response to fluoroscopically-directed epidural steroid injections and one-day percutaneous adhesiolysis.
Numerous studies have evaluated the effectiveness of spinal endoscopic adhesiolysis [24-29,31]; however, these studies utilized heterogenous inclusion criteria. Igarashi et al [29] evaluated patients with degenerative lumbar spinal stenosis. Manchikanti et al [26] evaluated only patients with a history of previous surgical intervention. In another study, Manchikanti et al [27] included patients who had not previously undergone surgery (a total of 16%). Dashfield et al [31] included only non-surgical patients. Geurts et al [24] reported results of spinal endoscopic adhesiolysis in 20 patients suffering with chronic low back pain. They reported > 50% reduction in pain in 40% of the patients at 3 months, and 35% at 6, 9, and 12 months. Richardson et al [25] reported results in 38 patients, with 19 of those patients identified with failed back surgery syndrome. They reported significant improvement based on Visual Analogue Scale and functional abilities. However, they have not reported data with regards to the proportion of patients with sustained relief at various time periods. Manchikanti et al [26,27] in two different studies, reported 75% relief at 3 months, 40% at 6 months, and 22% at 12 months in post-lumbar laminectomy patients; and, in a heterogenous group of patients including both post laminectomy and non-surgical patients, 52% of the patients at 3 months, 21% of the patients at 6 months, and 7% of the patients after 12 months. Igarashi et al [29] evaluated 58 patients with degenerative lumbar spinal stenosis, dividing them into two groups based on the presenting symptoms of either a monosegmental group (n = 34) or a multisegmental group (n = 24). They showed that relief of low back pain was observed up to 12 months after epiduroscopy in both groups, whereas relief of leg pain was evident up to 12 months after epiduroscopy in the monosegmental group, and up to 3 months after epiduroscopy in the multisegmental group. Dashfield et al [31] compared caudal epidural steroid injections with targeted steroid placement during spinal endoscopy for chronic sciatica in a prospective, randomized, double-blind trial. They randomized 60 patients with a 6-to 18-month history of sciatica to either a targeted epidural local anesthetic and steroid placement with a spinal endoscope, or caudal epidural local anesthetic and steroid placement. They defined sciatica as pain in the distribution of lumbar nerve root, accompanied by neurosensory and motor deficits, with or without back pain. They excluded patients with history of previous spinal surgery, coagulopathy, progressive motor neuron disorders or peripheral vascular disease, and patients receiving epidural corticosteroid injections within 3 months. No significant differences were found between the groups for any of the measures at any time. However, there were significant differences within both groups compared with pre-treatment values. The results of the present evaluation may not be compared to either the studies by Igarashi et al [29] or by Dashfield et al [31].
Igarashi et al [29] evaluated patients only with spinal stenosis. One-day percutaneous adhesiolysis also was shown to be effective in refractory spinal stenosis [44]. However, Igarashi et al [29] did not treat their patients with spinal stenosis with percutaneous adhesiolysis, which is considered as a safer and more effective procedure. The study by Dashfield et al [31] was performed in patients who were not expected to have epidural fibrosis and who had not been treated with either epidural steroid injections or with 1-day or 3-day percutaneous adhesiolysis. Consequently, there was no significant difference noted between caudal epidural steroid injections and targeted steroid placement with spinal endoscopy. In clinical practice in the United States, invasive intervention with spinal endoscopy as an initial treatment is not widely accepted.
The most common and worrisome complications of spinal endoscopy with adhesiolysis and injection of corticosteroids are related to dural puncture, spinal cord compression, intravascular injection, vascular injury, cerebral vascular or pulmonary embolus, infection, steroids, instrumentation with endoscope, and administration of high volumes of fluids potentially resulting in excessive epidural hydrostatic pressures, resulting in blindness, neurapraxia, numbness, intravascular injections, brain damage, and death [23-30,45-58]. Even though no major complications have been noted in this study, it is recommended that all precautions be undertaken, along with exhaustion of other modalities of treatments prior to embarking on spinal endoscopic adhesiolysis considering the safety and cost. The cost of the endoscope and the procedure are higher than either caudal epidural steroid injections or 1-day or 3-day percutaneous adhesiolysis procedure. The safety and effectiveness of 1-day and 3-day percutaneous adhesiolysis has been demonstrated [21,22].
The present evaluation utilized early unblinding in some patients, did not include a placebo group, and adapted a randomization ratio of 2:3 instead of 1:1. Considering the difficulties of recruiting patients to a double-blind trial, the authors considered the best way to recruit patients and give them a reasonable level of comfort was to offer additional treatment(s) if they failed the study, rather than allowing them to suffer for a year. Based on this allocation, the authors managed to include an acceptable number of patients. The control group for the study was not a true placebo group since interventions of caudal epidural steroid injections were used; nevertheless, the injections were ineffective in these patients. One of the objectives of the study was to demonstrate whether epidural steroid injections administered after adhesiolysis are effective as opposed to traditional or fluoroscopically-directed epidural steroid injections. In addition, this also served to provide a level of comfort to patients enrolled in the study since they knew they would receive some type of active treatment rather than a placebo. The authors believed that this type of randomization with a control group receiving standard treatment to be more effective and provide optimal results. A randomization process with a 2:3 ratio was selected to convince patients to enroll in the study, as they would have a higher chance of being included in a treatment group rather than a control group. The statistical validity was maintained throughout the study and an intent-to-treat analysis was incorporated in the study.
Trials of healthcare interventions are often described as either explanatory or pragmatic [40,41]. Explanatory trials generally measure efficacy – the benefit a treatment produces under ideal conditions. Consequently, explanatory trials often use carefully-defined subjects in a well-controlled research setting. By contrast, pragmatic trials, also known as practical clinical trials, measure effectiveness – benefit the treatment produces in routine clinical practice. Tunis et al [40] commented on the prevalence and significance of gaps in knowledge about clinical effectiveness of interventions and suboptimal evidence available to answer the critical questions. Most systematic reviews performed in interventional pain management include studies providing data not applicable to patients treated in typical practice settings. Consequently, limited quantity and quality of available scientific information impedes the efforts of public and private health insurers in developing evidence-based coverage policies for many new and existing technologies [59,60].
The substantial differences between explanatory and pragmatic trials illustrate a paradigm shift to clinical practice. Patient selection in an explanatory approach is based on the principles of homogenous population, primarily aiming to further scientific knowledge. However, in a pragmatic or practical clinical trial, the design reflects variations between patients that occur in real life clinical settings, and aims to inform choices between treatments. The authors consider this trial to be close to pragmatic or practical rather than explanatory. Even with appropriate randomization, the major focus of clinical research in the modern era of medicine, multiple other sources of bias may affect results. In this study, independent assessment was utilized. However, without a placebo treatment, in pragmatic approaches, the treatment response is the total difference between two treatments, including both treatment and associated placebo effects, as this will best reflect the likely clinical response in practice [22,30,58,61-65]. Practical clinical trials are expected to best address questions about the risks, benefits, and cost of an intervention as they would occur in routine clinical practice [41]. Thus, the most distinctive features of practical clinical trials are that they select patients from practices, either simulating actual practices or actual clinical practices. In addition, practical clinical trials often are designed to compare viable alternative clinical strategies. This study achieves both the distinctive features of practical clinical trials by selecting the population from an actual clinical practice and also by comparing viable alternative clinical strategies.
This procedure may be considered as a replacement for large bore catheter for percutaneous adhesiolysis, as we have not derived any diagnostic information. However, visualization of the scar tissue and freeing of the scar tissue from the nerve root may provide some additional benefit. The present day available catheters for percutaneous adhesiolysis are smaller bore. Consequently, spinal endoscopy with larger bore and improved flexibility appears to have a role. Percutaneous adhesiolysis has been shown to be an effective and safe procedure [21,22,45,46]. However this study went beyond percutaneous adhesiolysis and selected the patients after insufficient response after 1-day percutaneous adhesiolysis.
The issue remains for the patients who have had successful relief for 6 months or 12 months with regards to further treatment when pain returns and functional status deteriorates. Based on the present literature, with proper indications and precautions, the procedure may be repeated after approximately 6 months. In addition, the authors believe that even if patients have not responded previously to these procedures, if they have responded initially to spinal endoscopy, they may respond to 1-day or 3-day percutaneous adhesiolysis or even caudal epidural steroid injections. However, published data is not available at present to support this assumption and controlled trials are recommended to evaluate this postulate. Additional effect from spinal endoscopic adhesiolysis may be dependent on hydrostatic pressures created by the administration of sodium chloride solution which is not the case with lesser volume percutaneous adhesiolysis. However, percutaneous adhesiolysis also may be modified to accommodate this feature.
Conclusion
This controlled trial demonstrates that spinal endoscopic adhesiolysis reduces pain and improves functional and psychological status without adverse effects up to 12 months.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
LM conceived and designed the study, processed the data and wrote the manuscript.
MVB participated in the study's design and revised the manuscript.
JJR participated in its design and collected the clinical data.
VSP performed the statistical analysis.
KSD collected the clinical data.
CDM collected the clinical data.
DEB participated in the procedure.
SRW participated in the procedure.
All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Alkalay RN Kim DH Urry DW Xu J Parker TM Glazer PA Prevention of postlaminectomy epidural fibrosis using bioelastic materials Spine 2003 28 1659 1665 12897488 10.1097/00007632-200308010-00006
Gil K Frymoyer JW Frymoyer JW The management of treatment failure after decompressive surgery, The Adult Spine: Principles and Practice 1991 New York: Lippincott-Raven Publishers 849 870
Fritsch EW Heisel J Rupp S The failed back surgery syndrome. Reasons, intraoperative findings, and long-term results: A report of 182 operative treatments Spine 1996 21 626 633 8852320 10.1097/00007632-199603010-00017
Ross JS Robertson JT Frederickson RC Petrie JL Obuchowski N Modic MT deTribolet N Association between peridural scar and recurrent radicular pain after lumbar discectomy: magnetic resonance evaluation Neurosurgery 1996 38 855 861 8692415 10.1097/00006123-199604000-00053
Hoyland JA Freemont AJ Jayson MI Intervertebral foramen venous obstruction. A cause of periradicular fibrosis? Spine 1989 14 558 568 2749370
Rydevik BL The effects of compression on the physiology of nerve roots J Manipulative Physiol Ther 1992 15 62 66 1740654
Songer M Ghosh L Spencer D Effects of sodium hyaluronate on peridural fibrosis after lumbar laminectomy and discectomy Spine 1990 15 550 554 2402695
North RB Campbell JN James CS Conover-Walker MK Wang H Piantadosi S Rybock JD Long DM Failed back surgery syndrome: 5-year follow-up in 102 patients undergoing repeated operation Neurosurgery 1991 28 685 690 1831546 10.1097/00006123-199105000-00008
Dullerud R Graver V Haakonsen M Haaland AK Loeb M Magnaes B Influence of fibrinolytic factors on scar formation after lumbar discectomy. A magnetic resonance imaging follow-up study with clinical correlation performed 7 years after surgery Spine 1998 23 1464 1469 9670398 10.1097/00007632-199807010-00007
Pawl RP Arachnoiditis and epidural fibrosis: The relationship to chronic pain Curr Rev Pain 1998 2 93 99
Annertz M Jönsson B Stromquist B Holtas S No relationship between epidural fibrosis and sciatica in the lumbar postdiscectomy syndrome: A study with contrast-enhancement magnetic resonance imagery in symptomatic and asymptomatic patients Spine 1995 20 449 453 7747228
Cervellini P Curri D Volpin L Bernardi L Pinna V Benedetti A Computed tomography of epidural fibrosis after discectomy. A comparison between symptomatic and asymptomatic patients Neurosurgery 1988 6 710 713 3216968
Coskun E Suzer T Topuz O Zencir M Pakdemirli E Tahta K Relationships between epidural fibrosis, pain, disability, and psychological factors after lumbar disc surgery Eur Spine J 2000 9 218 223 10905440 10.1007/s005860000144
Benoist M Ficat C Baraf P Cauchoix J Postoperative lumbar epiduro-arachnoiditis: Diagnosis and therapeutic aspects Spine 1980 5 432 436 6450453
Cauchoix J Ficat C Girard B Repeat surgery after disc excision Spine 1978 3 256 259 152469
Phillips FM Cunningham B Managing chronic pain of spinal origin after lumbar surgery Spine 2002 27 2547 2553 12435991 10.1097/00007632-200211150-00029
Larocca H MacNab I The laminectomy membrane J Bone Joint Surg Br 1974 56B 545 550 4421702
McCarron RF Wimpee MW Hudkins PG Laros GS The inflammatory effects of nucleus pulposus: A possible element in the pathogenesis of low back pain Spine 1987 12 760 764 2961088
Cooper RG Freemont AJ Hoyland JA Jenkins JP West CG Illingworth KJ Jayson MI Herniated intervertebral disc-associated periradicular fibrosis and vascular abnormalities occur without inflammatory cell infiltration Spine 1995 20 591 598 7604329
Parke WW Watanabe R Adhesions of the ventral lumbar dura. Adjunct source of discogenic pain? Spine 1990 15 300 303 2141188
Heavner JE Racz GB Raj P Percutaneous epidural neuroplasty. Prospective evaluation of 0.9% NaCl versus 10% NaCl with or without hyaluronidase Reg Anesth Pain Med 1999 24 202 207 10338168 10.1016/S1098-7339(99)90128-1
Manchikanti L Rivera JJ Pampati V Damron KS MCManus CD Brandon DE Wilson SR One-day lumbar epidural adhesiolysis and hypertonic saline neurolysis in treatment of chronic low back pain: A randomized, double-blind trial Pain Physician 2004 7 177 186 16868590
Manchikanti L Singh V Epidural lysis of adhesions and myeloscopy Curr Pain Headache Rep 2002 6 427 435 12413401
Geurts JW Kallewaard JW Richardson J Groen GJ Targeted methylprednisolone acetate/hyaluronidase/clonidine injection after diagnostic epiduroscopy for chronic sciatica: A prospective, 1-year follow-up study Reg Anesth Pain Med 2002 27 343 352 12132057 10.1053/rapm.2002.27175
Richardson J McGurgan P Cheema S Prasad R Gupta S Spinal endoscopy in chronic low back pain with radiculopathy: A prospective case series Anaesthesia 2001 56 454 460 11350333 10.1046/j.1365-2044.2001.01524-3.x
Manchikanti L Pampati V Bakhit CE Pakanati RR Non-endoscopic and endoscopic adhesiolysis in post lumbar laminectomy syndrome. A one-year outcome study and cost effective analysis Pain Physician 1999 2 52 58 16906216
Manchikanti L The value and safety of epidural endoscopic adhesiolysis Amer J Anesthesiol 2000 27 275 279
Krasuski P Poniecka AW Gal E Wali A Truong A Hart AM Epiduroscopy: Review of techniques and results Pain Clinic 2001 13 71 76 10.1163/15685690152385772
Igarashi T Hirabayashi Y Seo N Saitoh K Fukuda H Suzuki H Lysis of adhesions and epidural injection of steroid/local anesthetic during epiduroscopy potentially alleviate low back leg pain in elderly patients with lumbar spinal stenosis Br J Anesth 2004 93 181 187 10.1093/bja/aeh201
Manchikanti L Rivera J Pampati VS Damron KS Beyer CD Brandon DE Wilson SR Spinal endoscopic adhesiolysis in the management of chronic low back pain: A preliminary report of a randomized, double-blind trial Pain Physician 2003 6 259 268 16880869
Dashfield AK Taylor MB Cleaver JS Farrow D Comparison of caudal steroid epidural with targeted steroid placement during spinal endoscopy for chronic sciatica: a prospective, randomized, double-blind trial Br J Anaesth 2005 94 514 519 15695544 10.1093/bja/aei084
Systems to rate the strength of scientific evidence Evidence Report/Technology Assessment No 47 University of North Carolina; Agency for Healthcare Research and Quality AHRQ Publication No 02-E016 2002
The Standards of Reporting Trials Group A proposal for structured reporting of randomized controlled trials JAMA 1994 272 1926 1931 7990245 10.1001/jama.272.24.1926
van Tulder M Furlan A Bombardier C Bouter L Editorial Board of the Cochrane Collaboration Back Review Group. Updated method guidelines for systematic reviews in the Cochrane collaboration back review group Spine 2003 28 1290 1299 12811274 10.1097/00007632-200306150-00014
Nelemans PJ deBie RA deVet HCW Injection therapy for subacute and chronic benign low back pain Spine 2001 26 501 515 11242378 10.1097/00007632-200103010-00014
Niemisto L Kalso E Malmivaara A Seitsalo S Hurri H Cochrane Collaboration Back Review Group Radiofrequency denervation for neck and back pain: a systematic review within the framework of the Cochrane collaboration back review group Spine 2003 28 1877 1888 12923479 10.1097/01.BRS.0000084682.02898.72
Koes BW Bouter LM van der Heijden GJMG Methodological quality of randomized clinical trials on treatment efficacy in low back pain Spine 1995 20 228 235 7716630
Koes BW Scholten RJPM Mens JMA Bouter LM Epidural steroid injections for low back pain and sciatica. An updated systematic review of randomized clinical trials Pain Digest 1999 9 241 247
van Tulder MW Koes BW Bouter LM Conservative treatment of acute and chronic nonspecific low back pain: A systematic review of randomized controlled trials of the most common interventions Spine 1997 22 2128 2156 9322325 10.1097/00007632-199709150-00012
Tunis SR Stryer DB Clancy CM Practical Clinical Trials. Increasing the value of clinical research for decision making in clinical and health policy JAMA 2003 290 1624 1632 14506122 10.1001/jama.290.12.1624
Roland M Torgerson DJ What are pragmatic trials? BMJ 1998 316 285 9472515
Tollison CD Langely JC Pain Patient Profile (P-3®) Manual 1995 National Computer Systems, Minneapolis
Fairbank JC Pynsent PB The Oswestry Disability Index Spine 2000 25 2940 2953 11074683 10.1097/00007632-200011150-00017
Manchikanti L Pampati VS Rivera JJ Fellows B Beyer CD Damron KS Cash KA Effectiveness of percutaneous adhesiolysis and hypertonic saline neurolysis in refractory spinal stenosis Pain Physician 2001 4 366 373 16902683
Manchikanti L Bakhit CE Percutaneous lysis of epidural adhesions Pain Physician 2000 3 46 64 16906207
Viesca C Racz G Day M Spinal techniques in pain management: lysis of adhesions Anesthesiol Clin North Am 2003 21 745 766 10.1016/S0889-8537(03)00088-9
Saberski L Brull S Spinal and epidural endoscopy: A historical review Yale J Bio Med 1995 68 7 15 8748461
Bromage RP Benumof JL Paraplegia following intracord injection during attempted epidural anesthesia under general anesthesia Reg Anesth Pain Med 1998 23 104 107 9552788 10.1016/S1098-7339(98)90120-1
MacLean CA Bachman DT Documented arterial gas embolism after spinal epidural injection Ann Emerg Med 2001 38 592 595 11679875 10.1067/mem.2001.118008
Mateo E Lopez-Alarcon MD Moliner S Calabuig E Vivo M De Andres J Grau F Epidural and subarachnoid pneumocephalus after epidural technique Eur J Anesthesiol 1999 16 413 417 10.1046/j.1365-2346.1999.00495.x
Katz JA Lukin R Bridenbaugh PO Gunzenhauser L Subdural intracranial air: An unusual cause of headache after epidural steroid injection Anesthesiology 1991 74 615 1825771
Knight JW Cordingley JJ Palazzo MG Epidural abscess following epidural steroid and local anesthetic injection Anaesthesia 1997 52 576 578 9203886 10.1111/j.1365-2044.1997.156-az0161.x
Tabandeh H Intraocular hemorrhages associated with endoscopic spinal surgery Am J Ophthalmol 2000 129 688 690 10844075 10.1016/S0002-9394(99)00470-5
Kusher FH Olson JC Retinal hemorrhage as a consequence of epidural steroid injection Arch Opthalmol 1995 113 309 313
Purdy EP Gurjit SA Vision loss after lumbar epidural steroid injection Anesth Analg 1998 86 119 122 9428864 10.1097/00000539-199801000-00024
Sandberg DI Lavyne MH Symptomatic spinal epidural lipomatosis after local epidural corticosteroid injections: Case report Neurosurgery 1999 45 162 165 10414580 10.1097/00006123-199907000-00037
Manchikanti L Role of neuraxial steroids in interventional pain management Pain Physician 2002 5 182 199 16902669
Manchikanti L Staats PS Singh V Schultz DM Vilims BD Jasper JF Kloth DS Trescot AM Hansen HC Falasca TD Racz GB Deer T Burton AW Helm S Lou L Bakhit CE Dunbar EE Atluri SL Calodney AK Hassenbusch S Feler CA Evidence-based practice guidelines for interventional techniques in the management of chronic spinal pain Pain Physician 2003 6 3 80 16878163
Garber AM Evidence-based coverage policy Health Aff (Milwood) 2001 20 62 82 10.1377/hlthaff.20.5.62
Tunis R Kang JL Improvements in Medicare coverage of new technology Health Aff (Milwood) 2001 20 83 85 10.1377/hlthaff.20.5.83
Manchikanti L Manchikanti KN Damron KS Pampati VS Effectiveness of cervical medial branch blocks in chronic neck pain: A prospective outcome study Pain Physician 2004 7 195 202 16868592
Manchikanti L Pampati VS Bakhit CE Rivera JJ Beyer CD Damron KS Barnhill RC Effectiveness of lumbar facet joint nerve blocks in chronic low back pain: A randomized clinical trial Pain Physician 2001 4 101 117 16906173
Manchikanti L Cash KA Moss TL Rivera J Pampati VS Risk of whole body radiation exposure and protective measures in fluoroscopically guided interventional techniques: A prospective evaluation BMC Anesthesiol 2003 3 2 12904269 10.1186/1471-2253-3-2
Manchikanti L Pampati VS Rivera JJ Beyer CD Damron KS Barnhill RC Caudal epidural injections with Sarapin or steroids in chronic low back pain Pain Physician 2001 4 322 335 16902678
Bogduk N Karasek M Two-year follow-up of a controlled trial of intradiscal electrothermal annuloplasty for chronic low back pain resulting from internal disc disruption Spine J 2002 2 343 350 14589465 10.1016/S1529-9430(02)00409-6
|
16000173
|
PMC1187869
|
CC BY
|
2021-01-04 16:28:05
|
no
|
BMC Anesthesiol. 2005 Jul 6; 5:10
|
utf-8
|
BMC Anesthesiol
| 2,005 |
10.1186/1471-2253-5-10
|
oa_comm
|
==== Front
BMC AnesthesiolBMC Anesthesiology1471-2253BioMed Central London 1471-2253-5-91598241410.1186/1471-2253-5-9Debate"Anxiebo", placebo, and postoperative pain Svedman Paul [email protected] Martin [email protected] Torsten [email protected] Department of Plastic and Reconstructive Surgery, Malmö University Hospital, Malmö, Lund University, Sweden2 Department of Clinical Neuroscience, Cognitive Neurophysiology Research Group, Karolinska Institute, Stockholm, Sweden3 Department of Anaesthesiology and Intensive Care and Multidisciplinary Pain Center, Uppsala University Hospital, Uppsala University, Uppsala, Sweden2005 27 6 2005 5 9 9 24 3 2004 27 6 2005 Copyright © 2005 Svedman et al; licensee BioMed Central Ltd.2005Svedman 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
Surgical treatment and its consequences expose patients to stress, and here we investigated the importance of the psychological component of postoperative pain based on reports in the clinical literature.
Discussion
Postoperative pain remains a significant clinical problem. Increased pain intensity with increased demand for opioid medication, and/or a relative unresponsiveness to pain treatment was reported both when the analgesia was administered by means of conventional nurse injection regimes and patient-controlled analgesia (PCA). Both the quality of the analgesia, and the sensitivity of postoperative models for assessing analgesic efficacy could be significantly influenced.
The findings could be explained by increased penetration of an algesic anxiety-related nocebo influence (which we chose to call "anxiebo") relative to its analgesic placebo counterpart. To counteract this influence, the importance of psychological effects must be acknowledged, and doctors and attending nurses should focus on maintaining trustful therapist-patient relationships throughout the treatment period. The physical mechanism of anxiebo should be further explored, and those at risk for anxiebo better characterized. In addition, future systemic analgesic therapies should be directed towards being prophylactic and continuous to eliminate surgical pain as it appears in order to prevent the anxiebo effect.
Addressing anxiebo is the key to developing reproducible models for measuring pain in the postoperative setting, and to improving the accuracy of measurements of the minimum effective analgesic concentration.
Summary
Anxiebo and placebo act as counterparts postoperatively. The anxiebo state may impair clinical analgesia and reduce the sensitivity of analgesic trials. Ways to minimize anxiebo are discussed.
==== Body
Background
In spite of considerable progress in postoperative analgesia, recent studies show that adequate pain relief remains elusive for a significant fraction of hospitalized surgical patients [1-3]. Continuous awareness of this problem and further efforts to improve treatment are required despite the availability of acute pain services [3]. Surgical treatment and its consequences cause psychological stress, and pain intensity after surgery is influenced by both analgesic drug effects and psychological factors. These factors comprise on the one hand an analgesic placebo influence induced by the therapeutic situation per se, and on the other hand its clinically less studied algesic anxiety-related nocebo influence counterpart [4-6] (in this article the term "anxiebo" is used to better focus on how to deal with this connection clinically).
The clinical effects of anxiebo have been only incompletely evaluated. In this work, relevant reports in the literature were assessed with regard to evidence of anxiebo, and its clinical expressions are described here. Implications of the findings on clinical practice, development of analgesic drugs, and assay sensitivity of clinical trials are discussed.
Discussion
The anxiebo-placebo relationship
The levels of perceived self-control (self-efficacy) and anticipatory anxiety are important factors in determining whether somebody will be a placebo or anxiebo responder [7-9], and may be influenced by the presence or absence of supportive social interaction. In this respect, the central role of the therapist/patient relationship should be recognized. Anxiebo may result from either a lack of supportive social interaction or inherently weak self-control in connection with exposure to painful stimuli (Figure 1). The lower the overall experience of control, the stronger the reported anticipatory anxiety, the pain experience expressed as pain-intensity ratings, and the autonomic activation.
Figure 1 Factors influencing postoperative anxiebo or placebo.
The effects of anxiebo and placebo can be clearly visualised on PET scans of the brain [9]. During placebo analgesia the activity patterns in the brain, brainstem and descending antinociceptive systems demonstrate that the endogenous opioid system is activated, and there are indications that the placebo state is sensitive to the effects of morphine [9,10]. These findings suggest that a placebo patient under opioid treatment should be able to closely approach a pain-free state. In anxiebo, on the other hand, scan findings indicate that the sensitivity to opioids is reduced but not abolished [9,11].
Reviews of analgesic trial outcomes have indicated that the degree of pain relief obtained may range between 0 and 100 per cent both for placebo and active drug patients [12,13]. A plausible explanation for this is that placebo and anxiebo act reciprocally and that the degree of shift in either direction is determined by how psychological elements of the analgesia are handled.
Evidence of anxiebo in postoperative analgesia
A combination of nonopioid analgesics and opioids is most often used to achieve postoperative analgesia, and the opioids are frequently administered by nurse-administered injections. Delay between need and injection constitutes the major problem with this approach [2]. Classic references indicate that preoperative encouragement and preparation of patients receiving nurse injections may decrease their opioid dosage by 50% [14,15], and directed patient information and psychological support are advocated. There is a moderate correlation between variable degrees of anxiety on the one hand, and pain intensity and requirement for opioid analgesics on the other, and these effects appear not only when the opioid injections are administered by an attendant nurse but also in patient-controlled analgesia (PCA) [16], where the access to intermittent doses of opioid should not be a limitation. While morphine at low doses is used as an anxiolytic, a deeper analysis of this effect, for instance with regard to the dose-response relationship and individual variation, appears to be lacking. Other approaches to reducing anxiety include the preoperative use of benzodiazepine [17] or distracting the patient, for instance by exposure to music [18]. Furthermore, a study of dental out-patients has suggested that a postoperative telephone call from the therapist demonstrating care and reassurance may improve the analgesia [19].
Several studies indicate that a significant fraction of PCA patients experiences moderate or severe pain [1,3]. Regional nerve blocks – which also require more doctor-patient interaction – on average produce clearly better results than systemic opioid techniques [2]. Even considering pain components that are known to be opioid resistant, the only way of explaining why timely treatment with systemic opioids should leave a proportion of patients with these levels of residual pain is by implicating a degree of anxiebo. This link is further confirmed by a controlled study that showed that PCA in the early phase after knee surgery (performed under general anesthesia) was associated with low oxygen tension in the subcutis, while a normal tension was observed after more-effective regional analgesia [20]. These findings indicate different levels of psychological stress and sympathetic tone in the two patient groups, and the oxygen tension in the group that received only the less-effective opioid analgesia was so low that it predisposed to surgical wound infection.
Studies into the use of PCA lend themselves also to further evaluation because of the considerable use of PCA as a research technique for assessing analgesic efficacy. There are many indications in the literature that PCA functions well as an analgesic method and that the patient-control concept is accepted by the patients. However, the question of control may be difficult to assess in a clinical situation involving an element of dependence. Recent reports show that many PCA patients do not recognize themselves as being in control of their treatment, and they may experience side effects and fear on utilization that limit their ability to control pain [21,22]. It is well known also from other studies that several psychological parameters are predictors of pain and/or opioid use (i.e. the propensity to press the controlling button) in PCA [16], of which anxiety-related factors are common. Thus, the connection between anxiebo and opioid use may explain why 20% of patients using PCA were found to press the analgesic-delivery button at an unchanged rate when the injected opioid dose was decreased [23]. Against this background, the findings in 18 trials assessing the efficacy of continuous basal or therapeutic infusions of opioid in connection with PCA are of interest. In the two largest studies [24,25], the addition of infusions at different rates did not reduce pain intensity, while the injected PCA dosages were clearly reduced in one study but not in the other. The remaining (smaller) studies demonstrated neither clinically convincing reductions in PCA dosage nor reductions in pain intensity that might be of general importance. Assuming full sensitivity to the continuously administered opioid, one should have expected the pain intensity to approach zero with increasing dosage, and as a result reproducible decreases in PCA analgesic dosage. The findings indicate that the assumption that the patients titrate to a common pain intensity level, and adapt their injection rates to an added analgesic stimulus, may have important limitations. It is therefore reasonable to implicate anxiebo in these examples of remarkable and long-lasting unresponsiveness to a continuous dosage of opioid. The underlying causes of these indications of poor sensitivity to the effect of opioids have, to our knowledge, not been further explored in the clinical literature.
In PCA, it is logical that the combination of an anxiebo-induced decrease in opioid sensitivity and an increase in the rate of self-administration as part of a coping reaction to anxiebo may in certain patients lead to pronounced overuse of morphine that is unrelated to the need for analgesia.
Implications for therapy, drug development, and research
The observations above indicate that both nurse injection regimes and PCA may produce signs of anxiebo characterized by increased pain intensity with increased demand for opioid medication, and/or a relative unresponsiveness to pain treatment, which may last for days (Figure 2). Avoiding anxiebo and thereby promoting placebo thus appears to be an important component of effective analgesic treatment. In order to accomplish this, the central role of a trustful therapist-patient relationship (doctor and nurse vs patient) throughout the treatment period should be recognized. Assuming that anxiebo could be eliminated clinically, the placebo state would be more prevalent, although its maximal overall influence in terms of improving the outcome of opioid treatment cannot yet be judged.
Figure 2 Reciprocal effects of anxiebo and placebo in postoperative patients. The propensity for a fraction of patients with patient-controlled analgesia (PCA) to press the analgesic-delivery button for reasons unrelated to pain may reflect a coping mechanism.
It is likely that the direct caregiver-patient relationship in nurse injection regimes can be used to better advantage, and minimizing stressful break-through pain should be emphasized. With regard to PCA, we do not doubt its role when applied in a suitable, psychologically supportive environment, but the method is more complex than generally thought. For those patients who do not adapt to the control concept, and respond with signs of anxiebo, extra support may be required or alternative treatment should be sought.
Conceptually, it would appear advantageous to direct future systemic analgesic therapies towards being prophylactic and continuous (partly to eliminate surgical pain as it appears, and partly to prevent the anxiebo effect) rather than reactive and intermittent. By analogy, the use of continuously acting psychotropic drugs that specifically counteract elements of the anxiety state may be effective in patients who respond poorly to analgesia.
The mechanisms underlying anxiebo should be further explored and other measures aimed at reducing anxiebo should be examined. In particular, further efforts should be devoted to defining patient groups and individuals that are especially at risk. This may be done by developing easily applicable scales for identifying at-risk patients preoperatively, and by preparing these patients with suitable information. Techniques should be sought which allow postoperative diagnosis of anxiebo. We suggest that perceived expectancy of future pain as opposed to subjective estimates of current pain may be a parameter of interest for better predicting a patient's risk for the anxiebo effect.
While the drug placebo effect is regularly accounted for in clinical trials, the effect of its anxiebo counterpart is not. It appears likely that trial assay sensitivity is reduced with variable penetration depending on the specific therapeutic circumstances of the individual study. Anxiebo may result in false-negative or borderline results, as indicated above for PCA models. Such systemic errors are not removed by simply increasing the size of the trial. The key to a reproducible, sensitive model is the proper handling of individual factors interacting within the psychological context, with respect for the therapist-patient relationship. Education and support to personnel handling the patients postoperatively seems to be the critical factor. Optimally, the anxiebo effect should be accounted for, perhaps by conducting a preliminary sensitivity test using a drug (such as morphine) with known analgesic effects.
Since patients or patient groups receiving PCA may not titrate their opioid dosage (Figure 2), it follows that the general usability of the clinical concept of minimum effective analgesic concentration as a guide for dosage can be questioned in particular studies.
Summary
The importance of psychological reactions are commonly acknowledged in postoperative analgesia, but the way these reactions express themselves and the degree of disturbance they may cause by producing anxiebo rather than placebo states are at present incompletely considered. Maintaining trustful therapist-patient relationships throughout the treatment period is very important. Future systemic analgesic therapies should be directed towards being prophylactic and continuous to eliminate surgical pain as it appears so as to prevent the anxiebo effect. The physical mechanism of anxiebo should be further explored, and more effort made to define patient groups and individual patients especially at risk. Addressing anxiebo is of importance also in the development of reproducible models for assessing analgesic efficacy in the postoperative setting, and may improve the accuracy of measurements of the minimally effective analgesic concentration.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
All three authors participated in the design and preparation of the manuscript, and all approved the final draft.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Dolin SJ Cashman JN Bland JM Effectiveness of acute postoperative pain management: I. Evidence from published data Br J Anaesth 2002 89 409 423 12402719 10.1093/bja/aef207
Svensson I Sjostrom B Haljamae H Assessment of pain experiences after elective surgery J Pain Symptom Manage 2000 20 193 201 11018337 10.1016/S0885-3924(00)00174-3
Werner MU Soholm L Rotboll-Nielsen P Kehlet H Does an acute pain service improve postoperative outcome? Anesth Analg 2002 95 1361 1372 12401627 10.1097/00000539-200211000-00049
Benedetti F Amanzio M Casadio C Oliaro A Maggi G Blockade of nocebo hyperalgesia by the cholecystokinin antagonist proglumide Pain 1997 71 135 140 9211474 10.1016/S0304-3959(97)03346-0
Staats P Hekmat H Staats A Suggestion/placebo effects on pain: negative as well as positive J Pain Symptom Manage 1998 15 235 243 9601159 10.1016/S0885-3924(97)00363-1
Benedetti F Pollo A Lopiano L Lanotte M Vighetti S Rainero I Conscious expectation and unconscious conditioning in analgesic, motor, and hormonal placebo/nocebo responses J Neurosci 2003 23 4315 4323 12764120
Bandura A O'Leary A Taylor CB Gauthier J Gossard D Perceived self-efficacy and pain control: opioid and non-opioid mechanisms J Pers Soc Psychol 1987 53 563 571 2821217 10.1037//0022-3514.53.3.563
Glass DC Singer JE Leonard HS Krantz D Cohen S Cummings H Perceived control of aversive stimulation and the reduction of stress responses J Pers 1973 41 577 595 4761393
Petrovic P Kalso E Petersson KM Ingvar M Placebo and opioid analgesia – imaging a shared neuronal network Science 2002 295 1737 1740 11834781 10.1126/science.1067176
Pollo A Amanzio M Arslanian A Casadio C Maggi G Benedetti F Response expectancies in placebo analgesia and their clinical relevance Pain 2001 93 77 84 11406341 10.1016/S0304-3959(01)00296-2
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 10.1016/S0304-3959(00)00486-3
McQuay H Carroll D Moore A Variation in the placebo effect in randomised controlled trials of analgesics: all is as blind as it seems Pain 1996 64 331 335 8740611 10.1016/0304-3959(95)00116-6
Moore RA Gavaghan D Tramer MR Collins SL McQuay HJ Size is everything- large amounts of information are needed to overcome random effects in estimating direction and magnitude of treatment effects Pain 1998 78 209 216 9870574 10.1016/S0304-3959(98)00140-7
Egbert LD Battit GE Welch CE Bartlett MK Reduction of postoperative pain by encouragement and instruction of patients. A study of doctor-patient rapport New Engl J Med 1964 270 825 827 14108087
Hayward J Information – a prescription against pain The study of Nursing Care Project Reports, Series 2 1975 No 5. London, Royal College of Nursing
Gil KM Ginsberg B Muir M Sykes D Williams DA Patient-controlled analgesia in postoperative pain: The relation of psychological factors to pain and analgesic use Clin J Pain 1990 6 137 142 2135004
Kain ZN Sevarino F Pincus S Alexander GM Wang SM Ayoub C Kosarussavadi B Attenuation of the preoperative stress response with midazolam: effects on postoperative outcomes Anesthesiology 2000 93 141 147 10861157 10.1097/00000542-200007000-00024
Wang SM Kulkarni L Dolev J Kain ZN Music and preoperative anxiety: a randomized, controlled study Anesth Analg 2002 94 1489 1494 12032013 10.1097/00000539-200206000-00021
Barlett B Firestone AR Beck FM Vig KWL Marucha PT Effect of a telephone call on orthodontic pain and anxiety Proceedings of the 82nd International Association of Dental Research/American Association of Dental Research/Canadian Association of Dental Research Conference: March 10–13 2004, Honululu
Akça O Melischek M Scheck T Hellwagner K Arkiliç CF Kurz A Postoperative pain and subcutaneous oxygen tension Lancet 1999 354 41 42 10406365 10.1016/S0140-6736(99)00874-0
Taylor N Hall GM Salmon P Is patient-controlled analgesia controlled by the patient? Soc Sci Med 1996 43 1137 1143 8890414 10.1016/0277-9536(96)00056-1
Chumbley GM Hall GM Salmon P Patient-controlled analgesia: an assessment by 200 patients Anaesthesia 1998 53 216 221 9613264 10.1046/j.1365-2044.1998.00314.x
Keeri-Szanto M Drugs or drums: what relieves postoperative pain? Pain 1979 6 217 230 88710 10.1016/0304-3959(79)90129-5
Parker PK Holtmann B White PF Patient-controlled analgesia. Does a concurrent opioid infusion improve pain management after surgery? JAMA 1991 266 1947 1952 1895471 10.1001/jama.266.14.1947
Rayburn WF Smith CV Woods MP Geranis BJ Combined continuous and demand narcotic dosing for patient-controlled analgesia after cesarean section Anesthesiol Rev 1990 17 58 62 10149050
|
15982414
|
PMC1187870
|
CC BY
|
2021-01-04 16:28:05
|
no
|
BMC Anesthesiol. 2005 Jun 27; 5:9
|
utf-8
|
BMC Anesthesiol
| 2,005 |
10.1186/1471-2253-5-9
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1711601179710.1186/1471-2105-6-171Research ArticleSystematic determination of the mosaic structure of bacterial genomes: species backbone versus strain-specific loops Chiapello H [email protected] I [email protected] F [email protected] G [email protected] A [email protected] M-A [email protected] Karoui M [email protected] Mathématique, Informatique & Génome, INRA Domaine de Vilvert, 78352 Jouy-en-Josas cedex, France2 Unité de Recherches Laitières et Génétique Appliquée, INRA Domaine de Vilvert, 78352 Jouy-en-Josas cedex, France2005 12 7 2005 6 171 171 5 4 2005 12 7 2005 Copyright © 2005 Chiapello et al; licensee BioMed Central Ltd.2005Chiapello 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
Public databases now contain multitude of complete bacterial genomes, including several genomes of the same species. The available data offers new opportunities to address questions about bacterial genome evolution, a task that requires reliable fine comparison data of closely related genomes. Recent analyses have shown, using pairwise whole genome alignments, that it is possible to segment bacterial genomes into a common conserved backbone and strain-specific sequences called loops.
Results
Here, we generalize this approach and propose a strategy that allows systematic and non-biased genome segmentation based on multiple genome alignments. Segmentation analyses, as applied to 13 different bacterial species, confirmed the feasibility of our approach to discern the 'mosaic' organization of bacterial genomes. Segmentation results are available through a Web interface permitting functional analysis, extraction and visualization of the backbone/loops structure of documented genomes. To illustrate the potential of this approach, we performed a precise analysis of the mosaic organization of three E. coli strains and functional characterization of the loops.
Conclusion
The segmentation results including the backbone/loops structure of 13 bacterial species genomes are new and available for use by the scientific community at the URL: .
==== Body
Background
Systematic genome comparisons play an increasingly important role in genome analysis and annotation. There are mainly two kinds of approaches used for whole genome comparisons: whole proteome comparison studies and whole genomic sequence alignment studies. Both approaches are powerful tools to study genome organization and evolution rules with different time scale considerations. These approaches have been employed with success in a recent study comparing the genome of yeast S. cerevisiae to three related yeast species genomes [1,2]. Genomewide comparative analysis of the yeast chromosomes has considerably improved gene annotation and has permitted the prediction of new motifs conserved in intergenic regions that act potentially as regulatory elements of gene expression [1].
Whole genome-alignments tools have shown important developments in the last years. It is now possible to align rapidly two or more long genomic DNA sequences with several tools like MultiPipMaker [3], Vista [4], Mummer [5-7] and MGA [8]. Some of them include graphical interfaces to display and browse genome alignments [3,4,7]. Other resources provide precomputed alignments for genome of related species, such as EnteriX or Colibase for enterobacteria [9,10].
Here we focus on whole genomic sequence alignments in the particular case of strains of single bacterial species. Since the publication of a second strain of Helicobacter pylori in 1999 [11], sequence data on closely related bacterial genomes has rapidly accumulated in public databases. The availability of complete genome sequences for multiple strains of numerous species opens up new perspectives for studying short term evolutionary processes. For example pairwise alignment of the complete genomes of the enterohemorrhagic Escherichia coli 0157:H7 strains (Sakaï or EDL 933) with the E. coli K-12 laboratory strain, allowed the definition of a 4.1 Mb sequence that was highly conserved between the two strains [12,13]. It was proposed that this common sequence corresponds to the conserved backbone of the E. coli chromosome, which is interrupted by numerous DNA segments called strain-specific loops, distributed throughout the backbone [12].
Examination of "mosaic" structures of backbones and loops offers a potential approach to trace the dynamics of genome evolution at the bacterial species level. The backbone, conserved in all aligned genomes of the species, probably corresponds in large part to the common ancestral strain and is the part of the genome under vertical selective pressure. As such, the backbone is also likely to be the most adapted part of the genome, which could be relevant when studying essential functional elements of the cell (such as genes, motifs or signals). Loops differ among strains. Some may correspond to mobile elements, like prophage [14] and insertion sequences [15], and may be associated with strain-specific pathogenicity. However little is known about functional elements associated with small loops.
Up to now, no systematic strategy for backbone/loop segmentation has been proposed for closely related bacterial genomes. The existing studies are either limited to pairwise comparisons or choose a reference genome which is compared to several related genomes. Precomputed alignments are often limited to a subgroup of species and use different softwares and parameters, making results generally non-reproducible or non-comparable.
In this paper we address the problem of defining a strategy to obtain a backbone/loop segmentation of bacterial genomes at the intraspecies level. This approach is based on two recent genome aligners: Mummer3 [7] and MGA [8]. Using a validated benchmark dataset, we developed a simple treatment of alignment results which permits a robust definition of the mosaic structure. Our approach does not take any genome as a reference and has no restriction for the number of genomes to align. We used our method to define this segmentation for 13 bacterial species. Validated backbone/loop segmentation results are stored in the MOSAIC database and are freely accessible through a user friendly Web interface. The backbone/loop segmentation determined using three E. coli genomes illustrates important properties of this structure, and indicates that intraspecies segmentation is a useful mean of enhancing bacterial genome annotation.
Results
The global strategy of genome segmentation and database integration used in this study is outlined Figure 1. A reference set, consisting in a manually verified genome alignment, was used to set appropriate segmentation parameters. Using this strategy, alignments and segmentations were performed systematically for 13 bacterial species, for which at least two genomes have been sequenced. Loop and backbone coordinates were then integrated in the MOSAIC database together with NCBI genome annotations.
Figure 1 Flow diagram of bacterial genomes segmentation in MOSAIC. The bacterial genomes segmentation includes four main steps in MOSAIC: NCBI bacterial genomes selection using Mummer and MGA, processing of genome alignments using MGA, backbone/loops segmentation and database integration using Perl scripts.
Validation of segmentation parameters
The loop coordinates of the E. coli K-12 and O157:H7 Sakai genomes validated by Hayashi et al. [16] (see Methods) were used as a basis to define an alignment strategy and to develop a treatment of alignment results adapted to bacterial backbone/loop segmentations. These strains are known to belong to distantly related E. coli lineages [17] and their genomes are more distantly related between each other compared with genomes within other species [18]. Parameters allowing a pertinent alignment of such different genomes were therefore expected to produce reliable results for more closely related strains. The K-12/Sakai comparisons were performed using different parameters of the MGA software, and those leading to the best results, as compared to coordinates obtained by Hayashi et al., were chosen. This set was used to produce alignments for all species so that results may be compared.
MGA software provides three types of results: matches (anchored MEMs of a minimal given length), aligned gaps (segments between anchored MEMs, shorter than a user-defined size and aligned with ClustalW) and unaligned gaps. Matches were computed using an iterative process on MEM size: MEM of at least 50 bp were used for the first MGA step and MEM of at least 20 bp were computed in the second recursive step. These two kinds of MEM were included into the backbone of the segmented genomes. The gaps were then treated as follows : gaps longer than 3000 bp (unaligned MGA gaps) were considered as loops, and gaps shorter than 3000 bp were aligned with ClustalW. Aligned gaps with more than 76 % identity were considered as backbone, others, as loops. Minimal size of loops and backbone segments was set to 20 nucleotides each. This strategy generated a backbone/loop profile of the K-12/Sakai alignment that differed by around 2400 nt (0,1 %) from that validated by Hayashi et al. [16].
Genome selection for the backbone/loop segmentation
In order to select a subset of genomes for which backbone/loop segmentation makes sense, an analysis using the Mummer package was performed (see Methods). Three categories of results we obtained. The first category includes 33 genomes for which MGA alignments and backbone/loop segmentations are feasible, as they have not been submitted to numerous and important rearrangements. The second category includes genomes that can be aligned after minor correction of their sequences (Reverse complement and Translation operators, see Material & Methods section). This second category concerns 5 genomes (4 species). The last category corresponds to 17 genomes belonging to 8 species: Neisseria meningitis, Prochlorococcus marinus, Salmonella enterica, Shigella flexneri, Streptococcus pyogenes, Tropheryma whipplei, Xylella fastidiosa and Yersinia pestis. These genomes were excluded because Mummerplot results revealed rearrangements covering a large part of the genome.
Genome alignments and backbone/loop segmentation
Twenty four genome alignments were generated and treated for backbone/loop segmentation using MGA and our defined set of parameters. These included two quadruple alignments (E. coli, C. pneumoniae), four triple alignments (C. pneumoniae, E. coli, S. aureus, S. pyogenes) and eigthteen pairwise alignments. For one species, Buchnera aphidicola, segmentation results were not exploited due to too low coverage (this value estimates the percentage of total genome length included in the backbone, in this case 40 %, see Discussion).
Validated segmentation results including backbone size, loop size, loop number and genome coverage are described in Table 1. The coverage ranged from 68 % for E. coli quadruple alignment to 99 % for C. pneumoniae strains. Species comparisons giving high coverage values may be a consequence of to the choice of closely related strains for sequencing, but may also indicate that overall horizontal transfer is less important in some species than in others.
Table 1 Segmentation results obtained from MGA alignments and included in the MOSAIC database. For each segmentation result, the first column describes the species and genomes used for segmentation analyses; the number of compared strains is indicated between parentheses. Total loop sizes and loop number of each genome are entered in the same order as strain names, and separated by '+'. Coverage corresponds to the ratio between backbone size and total genome size of a strain; here the mean value for all compared strains is given in percents.
Compared genomes (numbers of strains) Backbone size (Mb) Cumulative loop size (kb) [Loop number] Coverage (mean)
Agrobacterium tumefaciens
C58 Cereon circ X C58 Univ. Wash circ (2) 2.09 751[24]+751[25] 74 %
C58 Cereon lin RC X C58 Univ. Wash lin (2) 1.82 252[13]+253[13] 88 %
Bacillus anthracis
Ames X Ames 'Ancestor' (2) 3.93 528[26]+528[24] 90 %
Bacillus cereus
ATCC14579 X ATCC10987 (2) 4.02 1390[2878]+1203[2872] 76 %
Chlamydophila pneumoniae
AR39 RC+TR X CWL029 X J138 X TW183 (4) 1.22 10[13]+10[13]+7[11]+6[12] 99%
CWL029 X J138 X TW183 (3) 1.21 15[14]+11[12]+10[13] 99 %
CWL029 X J138 (2) 1.21 21[15]+17[14] 99 %
J138 X TW183 (2) 1.22 9[9]+8[10] 99 %
CWL029 X TW183 (2) 1.22 13[6]+9[6] 99 %
AR39 RC+TR X CWL029 (2) 1.22 8[7]+8[7] 99%
Escherichia coli
K-12 X Sakai X EDL933 X CFT073 (4) 3.52 1119[848]+1978[830]+2008[830]+1711[811] 68 %
K-12 X Sakai X CFT073 (3) 3.73 904[827]+1763[795]+1496[770] 73 %
Helicobacter pylori
26695 X J99 (2) 1.24 428[957]+403[967] 75 %
Listeria monocytogenes
EGD X 4b F2365 (2) 2.67 270[644]+230[638] 92 %
Mycobacterium tuberculosis
CDC1551 X H37Rv (2) 4.19 217[164]+225[162] 95 %
Staphylococcus aureus
MW2 X MU50 X N315 (3) 2.59 226[388]+283[382]+220[388] 92 %
Streptococcus agalactiae
2603V/R X NEM316 (2) 1.88 276[135]+327[132] 86 %
Streptococcus pneumoniae
R6 X TIGR4 (2) 1.91 128[282]+250[294] 91 %
Streptococcus pyogenes
M1GAS X MGAS315 X MGAS8232 (3) 1.62 235[275]+283[282]+277[282] 86 %
M1GAS X MGAS315 (2) 1.64 210[191]+258[192] 88 %
M1GAS X MGAS8232 (2) 1.65 206[225]+249[231] 88 %
Vibrio vulnificus
YJ016 K2 X CMCP6 K2 TR (2) 1.63 222[198]+210[199] 89 %
YJ016 K1 RC X CMCP6 K1 TR (2) 2.73 628[340]+555[338] 82%
The number of loops in a segmented genome appeared to be also highly variable among bacterial species, ranging from 6 (Chlamydophila pneumoniae strain CWL29 compared to strain TW183) to 2878 (Bacillus cereus, strain ATCC14579 compared to ARCC10987). Results of table 1 revealed two extremes situations. Some species have few very long loops, as Agrobacterium tumefaciens (24/25 loops for the circular chromosome, mean length of loops around 28 kilobases). Others (Bacillus cereus) contain a large number of short loops (mean length around 400 bases). These differences will need to be further examined in details, in relation with genome annotations.
Database integration and web interface
Alignments were integrated into the MOSAIC database and are accessible through the Web interface. Access to the mosaic structure of genomes is made by species selection or gene name selection. For each segmented genome, a local view of the physical map of the segmented genome is available, using MuGeN software [19] (see Figure 2). This graphical visualization of loop and backbone structures is associated with Genbank/NCBI genome annotations. In addition, an overall graphical view of backbone and loops structure is presented using EMBOSS cirdna program. Finally, lists of loop and backbone segments can also be inspected and downloaded according to different criteria like size, genome position or functional characterization.
Figure 2 Graphical visualization of the backbone/loop structure available through the Web interface of Mosaic. 'Physical map' mode of MOSAIC corresponding to the graphical visualization of a 15 kb portion of the E. coli K-12, O157:H7 Sakai and CFT073 segmented genomes (data correspond to the comparison of three E. coli strains described in results). Genbank annotations are indicated with coloured arrows. Supplementary annotations are indicated as red boxes. Backbone is indicated in grey whereas loops appear in green.
In depth analysis of the backbone/loop structure of three E. coli genomes
A more precise analysis of the segmentation results from the comparison of E. coli strains K-12 [20], O157:H7 Sakai [12] (named Sakai below) and CFT073 [21] (named CFT below) was performed. A 3.73 Mb length backbone (exhibiting more than 97 % identity between the three strains) and three sets of strain-specific loops (of very different total length) were identified. The K-12 genome included 827 K-12 loops (total length 0.9 Mb, 20 % of the K-12 genome), the Sakai genome, 795 Sakai loops (total length 1.8 Mb, 33 % of the Sakai genome) and the CFT genome, 770 CFT loops (total length 1.5 Mb, 29 % of the CFT genome). The differences in total loop sizes are in keeping with the different total genome size of the three strains (K-12: 4.6 Mb ; Sakai: 5.5 Mb ; CFT: 5.2 Mb).
A large proportion of short loops in the E. coli genome
Basic statistics concerning the size distribution of the three loop sets are described in Table 2 and Figure 3. Results of Table 2 show a remarkable number of short loops for the three species: three quarters of the K-12 loops are shorter than 486 nucleotides (respectively 863 and 314 for Sakai and CFT). The histogram of loop size distribution for the three E. coli strains (Figure 3) reveals that the loop population is heterogeneous. Interestingly it appears that the loop populations exhibit roughly the same profile in the three strains, which may comprise three sub-populations: numerous very short loops (length around 100 bp), medium-size loops (length around 1–2 kb) and a few very long loops (length > 10 kb). This may reflect a wide diversity of functional properties conferred by loops: the longer loops probably encode several genes (and correspond for example to bacteriophage or pathogenicity islands). The shorter ones might have regulatory roles or affect gene expression.
Table 2 Size distribution of loops (in bp) obtained from segmentation of the E. coli genomes K-12, O157:H7 Sakai (SAK) and CFT073 (CFT). Minimal size (Min), Mean size, Maximal size (Max), First Quartile (1st Qu.), Median size, and Third Quartile (3rd Qu.) are shown.
K-12 loops SAK loops CFT loops
Min 20 20 20
Mean 1093 2217 1942
Max 40120 96682 150690
1st Qu. 34 32.5 31
Median 113 109 77
3rd Qu. 486 863 314
Figure 3 Distribution of the loop sizes of three E. coli genomes (K-12, SAKAI and CFT073). Loop sizes range from 20 bp to 40 120 to 151 690 bp. Log10 scale is used on the x-axis.
Functional elements associated with backbone and loops
The distribution of functional elements in the backbone/loop structure was analyzed. Functions identified by classical annotations of bacterial genomes i.e. genes, tRNA, rRNA, phages and Insertion Sequences (IS) were first considered. As expected tRNA and rRNA were mainly present in the backbone. One exception concerns a rather large proportion of tRNA present in the Sakai loops (27 %) compared to K-12 (9%) and CFT (13 %) loops. The Sakai strain contains 18 specific tRNAs not present in K-12 [12], Hayashi et al. observed that these tRNAs recognize codons which are used with an increased frequency in Sakai loops. Not surprisingly, we observed that phages and IS are quasi-totally included in loops (>98%). The ten longest loops of K-12 correspond systematically to known phages, or phage remnants of the E. coli genome.
One hundred E. coli K-12 loops are associated with BIME
To refine functional categorization of smaller loops we examined the correspondence between loops and Bacterial Interpersed Mosaic Element (BIME). BIME are short palindromic repetitive DNA elements found in the genomes of E. coli and other enterobacteria [22], and are present exclusively in intergenic regions. BIME are composed of three types of palindromic units (Y, Z1 and Z2). Three sub-families of BIME have been described: BIME-1, which are composed of one Y and one Z1; BIME-2 ("composite" BIME) which contain two to twelve Y and Z2; and a third category ("atypical BIME"), which refers to all other palindromic units associations. BIME sizes range from 140 bp for BIME-1 to several kb for BIME-2 or atypical BIME. BIME are reported to have several functions: mRNA stabilization, transcription termination, translational control and genomic rearrangements [22]. Using the MOSAIC database we identified 100 loops associated with BIME. BIME coordinates were obtained from the "short repeated palindromes in enterobacteria" Web site [23]. They are distributed as follows: in 31 cases, a BIME was present within the loop. In 29 cases, the BIME covered the entire loop region and extended into flanking backbone sequences. In 40 cases, the BIME accounted for more than 50% of the loop length and extended over to one side of the backbone. Results concerning BIME distribution in backbone and loops on the K-12 genome (Table 3) clearly indicate that loops are enriched in BIME elements: in particular 2/3 of the DNA regions associated to BIME are located on the loops. This tendency is particularly striking for BIME-1, for which 71 % of the cumulated length is associated with loops. Interestingly, BIME-2 are quasi equally distributed between backbone and loops. This result is a generalization to all K-12 BIMEs of a result observed in a previous work [24]. PCR analysis of 3 BIME-1 and 3 BIME-2 loci in 51 E. coli and Shigella isolates showed that BIME-1 are either present or absent among isolates whereas BIME-2 are generally present in the same set of isolates but exhibit a high level of length polymorphism [24]. Figure 2 illustrates two examples of loops associated with BIME-1 and BIME-2, as visible with MuGeN through the MOSAIC Web interface. This association of loops with BIME is a first clue in characterizing the functionality of short loops in E. coli strains.
Table 3 Distribution of BIME (in percent of length) in backbone and loops regions of the E. coli K-12 genome, as determined from the triple K-12, Sakai and CFT073 alignment.
K-12
Backbone Loops
BIME 38 % 62 %
BIME 1 29 % 71 %
BIME 2 47 % 53 %
atypical BIME 37 % 63 %
Discussion
Backbone/loop segmentation as a step towers analysis of genome evolution
Studying backbone and loops of bacterial genomes is an efficient way to distinguish the two major modes of evolution acting on bacterial genomes. The backbone may be considered as the part of the genome susceptible to vertical long-term evolution. Backbones are very similar for closely related strains and variability comes mainly from punctual mutations or insertions/deletions of oligonucleotides. The loop population (defined in MOSAIC as variable regions of 21 bp or more) is more heterogeneous : the number of loops and the average loops length varies greatly from one species to another (Table 1). Loops can be viewed as elements issued from short-term evolution processes. One such process is horizontal transfer. For example acquisition/loss of distinct prophage sets seems to be a rapid process, which can be observed between closely related strains [25]. Significantly, for some genomes, phages are the major contributors to loop length [14]. Eleven loops of E. coli K-12 are associated with phages and constitute 24 % of the total E. coli K-12 loop length. A contrasting example is found in H. pylori: this species does not contain prophage, although it contains large loops that may be associated with pathogenicity islands [11]. Our results indicate that medium-size loops (scale of the gene size) are constituted, at least in part, from known variable elements of bacterial genomes like Insertion Sequences. The relatively large number of short loops found in some species (E. coli, B. cereus) is quite surprising. Such small loops may be due to replication errors ('copy-choice' of DNA polymerase, slippage mechanism), which can generate small insertions or deletions [26] or may correspond to highly polymorphic regions. As opposed to large or medium size loops it is likely that these shorter loops arose from non-horizontal transfer events.
Consequences of segmentation from multiple alignments
Alignments including more than two genomes generally yield a more robust but smaller backbone than pairwise alignments. This is due to the fact that a larger set of genome variations is taken into account. In the future, about ten or more genomes will be available for some species. One possible consequence is that the backbone length will shrink steadily with new strain genomes. In that case, the backbone may rather be redefined as, for example, the subset of chromosomal regions present in at least half of the strains. Alternatively, the backbone size may decrease but rapidly reach a minimal size, which will be stable even when new strain genomes will be added for alignment.
As a consequence of multiple comparisons, loop populations are greater and more heterogeneous. They include for example elements present in only one genome (which may correspond to acquisition of a very specific characteristic by one strain), elements present in a subset of strains or elements present in all genomes but one (which may correspond to a deletion in one strain). It will be important to systematically classify loops obtained from multiple comparisons in order to facilitate their identification through the MOSAIC interface.
To estimate the importance of loops corresponding to DNA present in the common ancestor but lost in one of the compared strains a preliminary study was performed: all sequences present in the K-12 loops (from the triple alignment) were blasted against the Salmonella typhimurium genome (considered as the outgroup), and matching sequences present in the same genome environment were considered as "ancestral loops". Ten loops, corresponding to a total length of 3658 bp, matched this criterium. This suggests that only a minor subset of the loops correspond to deletions that occurred in either E. coli Sakai or CFT genomes.
Backbone/loops segmentation for divergent or rearranged genomes
Some genomes of species like Buchnera aphidicola were too divergent to be segmented with our procedure. In fact, these genomes are clearly atypical in terms of evolutionary distance within a species: despite complete colinearity of their genomes, B. aphidicola Sg and Ap genomes display a high degree of divergence at the nucleotide level, making them as different as E. coli/S. typhimurium genomes [27]. Comparison of Sg and Ap genomes is thus almost the same situation as comparing different species, but would be possible by adapting the alignment parameters. This raises the question of bacterial species definition: the evolutionary distances within a species and between species are very heterogeneous. For example, it has recently been confirmed that Shigella is phylogenetically indistinguishable from E. coli [28]. Our method will also be easily extended to bacterial species where numerous chromosomal rearrangements have occurred, using recently developed genome aligners such as MAUVE [29]. Intra-species comparison of divergent and/or rearranged genomes will open the way to segmentation of genomes from different, but closely related species.
A new category of genome annotation
To our knowledge, this work is the first study allowing systematic mosaic genome segmentation of all available strains (ranging from two to four) in 13 bacterial species. Examination of the backbone of a bacterial species should greatly facilitate refinement of gene annotation and prediction of conserved sites with potential regulatory roles. Examination of the gene content in loops is important for identification of putative horizontally transferred genes. Genes adapted to specific ecological environments or involved in pathogenicity of a specific strain should also be found in the strain-specific loops. Indeed, the ASAP database (A Systematic Annotation Package for community analysis of genomes) [30] recently added the features type 'island' and 'conserved_segments' in order to provide lists of regions that are specific or common to the two E. coli K-12 and O157:H7 genomes.
Conclusion
Genome aligners were used to build a robust strategy for bacterial genome segmentation. Backbone/loops structures were systematically determined for 38 bacterial genomes. The MOSAIC resource makes it easy to visualize, annotate, and analyse loops and backbone segments of these genomes. First analyses reveal a surprising diversity in the number of loops from one species to another. In addition some species accumulate a large number of short loops, unsuspected previously.
Methods
Species selection
Complete bacterial genomes were downloaded from the NCBI microbial genome database: , version of 06/24/2004. Twenty one species (55 genomes) for which genome sequences of at least two different strains are available were selected for analysis: Agrobacterium tumefaciens, Bacillus anthracis, Bacillus cereus, Buchnera aphidicola, Chlamydophila pneumoniae, Escherichia coli, Helicobacter pylori, Listeria monocytogenes, Mycobacterium tuberculosis, Neisseria meningitidis, Prochlorococcus marinus, Salmonella enterica, Shigella flexneri, Staphylococcus aureus, Streptococcus agalactiae, Streptococcus pneumoniae, Streptococcus pyogenes, Tropheryma whipplei, Vibrio vulnificus, Xylella fastidiosa, and Yersinia pestis [see Additional file 1].
Segmentation strategy
Backbone/loop segmentations were determined using a simple procedure based on Mummer3 [7] and MGA [8] results (see figure 1).
Selection of genomes without rearrangement
In the first step, the subset of genomes for which it is possible to define a reliable backbone was identified using mummer and mummerplot scripts of the Mummer3 package. First, all Maximal Exact Matches (MEM, not necessarily unique) of at least 20 bp in both forward and reverse strands of the compared genomes were computed using the mummer program. Visualization of results between each pair of sequences was then performed using the mummerplot program. This graphical visualization was used to decide whether a common backbone could be defined for the considered genomes. In several cases, this step led us to adjust one of the genomes before the segmentation step. Two operators were defined: the reverse complement operator, named RC, and the translation operator, named TRx, where x indicates that bases from position 1 to x were transferred to the end of the genome. This number x of bases shifted to the end of the genome was determined by the position of the first aligned MEM detected by MGA between the two genomes. These operators allowed us to assign the same strand and the same start position to all compared genomes. They were applied to a subset of genomes before alignment with MGA software. Genomes where rearrangements covering more than half of the total length were detected by mummerplot and excluded at this step. They can not be handled properly by MGA and would therefore lead to inaccurate segmentation.
Backbone/loop segmentation
The second step was to use the MGA software to perform whole genome alignments on the subset of selected genomes and to define backbone and loops. MGA is a powerful multiple genome aligner which presents two major advantages. First, it performs simultaneous multiple alignments based on MEM (Maximal Exact Matches present in all aligned genomes) selection, without considering any genome as the reference. Second, a consistent and robust backbone for the aligned genomes can be defined using its MEM anchoring algorithm followed by treatment of gaps (i.e. regions between the anchored MEM). Parameters used in MGA were adjusted by comparison with a manually curated reference set of loops of two E. coli strains: K-12 and O157:H7 Sakai [16]. After Mummer 1 alignment, backbone/loop junctions were extracted and systematically aligned using the fasta3 algorithm. Each alignment was checked by eye inspection and in many cases, the backbone sequence was extended by a few to several base pair [Pr. T. Hayashi, personal communication]. Further analysis using whole genome PCR scanning confirmed that the loops longer than 500 pb are indeed variable elements [31]. A simple treatment of MGA alignment results was developed to define the boundaries of loops and to enhance their concordance with this manually determined pairwise reference dataset [see 'Results' section, 'Validation of segmentation parameters' subsection]
Coverage calculation and database integration
Results of MGA alignments were generated in XML format. Backbone/loop segmentations were processed with a Perl script using the SAX module for XML parsing. For each aligned genome, backbone and loop coordinates were computed and coverage (length of the backbone divided by total length of the genome) was calculated. Results were then integrated into the MOSAIC relational database. The database was implemented using the PostgreSQL relational database system. The MOSAIC relational model is generic and not dedicated to any alignment tool or genome species. The Web interface was also written in Perl language using standard modules for database access (DBI module for DataBase Interface) and dynamic pages (CGI module for Common Gateway Interface). Different graphical visualizations of the backbone/loop structure were developed using the MuGeN software [19] and the cirdna program which is part of the EMBOSS package [32].
Authors' contributions
H. Chiapello performed segmentation results, conceived the MOSAIC application and drafted the manuscript. I. Bourgait, and A. Gendrault-Jacquemard participated in the database design and segmentation results integration. G. Heuclin performed loop analysis. F. Sourivong built the Web interface. M-A. Petit participated in data analysis and helped to draft the manuscript with M. El Karoui who supervised the study.
Supplementary Material
Additional File 1
The 55 bacterial genomes for which at least two strains have been sequenced. For each species and each strain, NCBI accession number and genome length are indicated. The 'MGA aligt.' column indicates if the genomes have been included in an MGA alignment. Genomes corrections are indicated in the 'Correction column' as follows : '-', no correction, 'RC', Reverse Complement strand, and 'TR+x' means that segment in position 1 to x of the genome has been shifted at the end of the genome. A brief comment is given for genomes excluded from MGA alignments.
Click here for file
Acknowledgements
We thank C. Hennequet-Antier for her assistance in using R software and Dr A. Gruss for many helpful discussions. We are indebted to Professor T. Hayashi and Dr K. Kurokawa for sharing data before publication and helpful discussions. This work is supported in part by the "ACI IMPbio" grant from the French Ministry of Research.
==== Refs
Kellis M Patterson N Endrizzi M Birren B Lander ES Sequencing and comparison of yeast species to identify genes and regulatory elements Nature 2003 423 241 54 12748633 10.1038/nature01644
Kellis M Patterson N Birren B Berger B Lander ES Methods in comparative genomics: genome correspondence, gene identification and regulatory motif discovery J Comput Biol 2004 11 319 355 15285895 10.1089/1066527041410319
Schwartz S Elnitski L Li M Weirauch M Riemer C Smit A Green ED Hardison RC Miller W MultiPipMaker and supporting tools: Alignments and analysis of multiple genomic DNA sequences Nucleic Acids Res 2003 31 3518 24 2003 Jul 1 12824357 10.1093/nar/gkg579
Frazer KA Pachter L Poliakov A Rubin EM Dubchak I VISTA: computational tools for comparative genomics Nucleic Acids Res 2004 32 W273 9 2004 Jul 1 15215394
Delcher AL Kasif S Fleischmann RD Peterson J White O Salzberg SL Alignment of whole genomes Nucleic Acids Res 1999 27 2369 76 10325427 10.1093/nar/27.11.2369
Delcher AL Phillippy A Carlton J Salzberg SL Fast algorithms for large-scale genome alignment and comparison Nucleic Acids Res 2002 30 2478 83 12034836 10.1093/nar/30.11.2478
Kurtz S Phillippy A Delcher AL Smoot M Shumway M Antonescu C Salzberg SL Versatile and open software for comparing large genomes Genome Biol 2004 5 R12 14759262 10.1186/gb-2004-5-2-r12
Höhl M Kurtz S Ohlebusch E Efficient multiple genome alignment Bioinformatics 2002 18 S312 20 12169561
Florea L McClelland M Riemer C Schwartz S Miller W EnteriX 2003: Visualization tools for genome alignments of Enterobacteriaceae Nucleic Acids Res 2003 31 3527 32 2003 Jul 1 12824359 10.1093/nar/gkg551
Chaudhuri RR Khan AM Pallen MJ ColiBASE: an online database for Escherichia coli, Shigella and Salmonella comparative genomics Nucleic Acids Res 2004 32 D296 9 2004 Jan 1 14681417 10.1093/nar/gkh031
Alm RA Ling LS Moir DT King BL Brown ED Doig PC Smith DR Noonan B Guild BC deJonge BL Carmel G Tummino PJ Caruso A Uria-Nickelsen M Mills DM Ives C Gibson R Merberg D Mills SD Jiang Q Taylor DE Vovis GF Trust TJ Genomic-sequence comparison of two unrelated isolates of the human gastric pathogen Helicobacter pylori Nature 1999 397 176 80 9923682 10.1038/16495
Hayashi T Makino K Ohnishi M Kurokawa K Ishii K Yokoyama K Han CG Ohtsubo E Nakayama K Murata T Tanaka M Tobe T Iida T Takami H Honda T Sasakawa C Ogasawara N Yasunaga T Kuhara S Shiba T Hattori M Shinagawa H Complete genome sequence of enterohemorrhagic Escherichia coli O157:H7 and genomic comparison with a laboratory strain K-12 DNA Res 2001 8 11 22 11258796
Perna NT Plunkett G 3rdBurland V Mau B Glasner JD Rose DJ Mayhew GF Evans PS Gregor J Kirkpatrick HA Posfai G Hackett J Klink S Boutin A Shao Y Miller L Grotbeck EJ Davis NW Lim A Dimalanta ET Potamousis KD Apodaca J Anantharaman TS Lin J Yen G Schwartz DC Welch RA Blattner FR Genome sequence of enterohaemorrhagic Escherichia coli O157:H7 Nature 2001 409 529 33 11206551 10.1038/35054089
Canchaya C Fournous G Brussow H The impact of prophages on bacterial chromosomes Mol Microbiol 2004 53 9 18 15225299 10.1111/j.1365-2958.2004.04113.x
Schneider D Duperchy E Depeyrot J Coursange E Lenski R Blot M Genomic comparisons among Escherichia coli strains B, K-12, and O157:H7 using IS elements as molecular markers BMC Microbiol 2002 2 18 12106505 10.1186/1471-2180-2-18
E. coli O157:H7 Sakai Genome Information, K12 and Sakai loop coordinates 2002
Reid SD Herbelin CJ Bumbaugh AC Selander RK Whittam TS Parallel evolution of virulence in pathogenic Escherichia coli Nature 2000 406 64 7 10894541 10.1038/35017546
Konstantinidis KT Tiedje JM Genomic insights that advance the species definition for prokaryotes Proc Natl Acad Sci U S A 2005 102 2567 72 2005 Feb 15 15701695 10.1073/pnas.0409727102
Hoebeke M Nicolas P Bessieres P MuGeN: simultaneous exploration of multiple genomes and computer analysis results Bioinformatics 2003 19 859 64 12724296 10.1093/bioinformatics/btg101
Blattner FR Plunkett G 3rdBloch CA Perna NT Burland V Riley M Collado-Vides J Glasner JD Rode CK Mayhew GF Gregor J Davis NW Kirkpatrick HA Goeden MA Rose DJ Mau B Shao Y The complete genome sequence of Escherichia coli K-12 Science 1997 277 1453 74 9278503 10.1126/science.277.5331.1453
Welch RA Burland V Plunkett G 3rdRedford P Roesch P Rasko D Buckles EL Liou SR Boutin A Hackett J Stroud D Mayhew GF Rose DJ Zhou S Schwartz DC Perna NT Mobley HL Donnenberg MS Blattner FR Extensive mosaic structure revealed by the complete genome sequence of uropathogenic Escherichia coli Proc Natl Acad Sci USA 2002 99 17020 4 12471157 10.1073/pnas.252529799
Bachellier S Clement JM Hofnung M Short palindromic repetitive DNA elements in enterobacteria: a survey Res Microbiol 1999 150 627 639 10673002 10.1016/S0923-2508(99)00128-X
E. coli K12 BIMES 1999
Bachellier S Clement JM Hofnung M Gilson E Bacterial interspersed mosaic elements (BIMEs) are a major source of sequence polymorphism in Escherichia coli intergenic regions including specific associations with a new insertion sequence Genetics 1997 145 551 62 9055066
Brussow H Canchaya C Hardt WD Phages and the evolution of bacterial pathogens: from genomic rearrangements to lysogenic conversion Microbiol Mol Biol Rev 2004 68 560 602 15353570 10.1128/MMBR.68.3.560-602.2004
Viguera E Canceill D Ehrlich SD Replication slippage involves DNA polymerase pausing and dissociation EMBO J 2001 20 2587 95 11350948 10.1093/emboj/20.10.2587
Tamas I Klasson L Canback B Naslund AK Eriksson AS Wernegreen JJ Sandstrom JP Moran NA Andersson SG 50 million years of genomic stasis in endosymbiotic bacteria Science 2002 296 2376 9 12089438 10.1126/science.1071278
Wei J Goldberg MB Burland V Venkatesan MM Deng W Fournier G Mayhew GF Plunkett G 3rdRose DJ Darling A Mau B Perna NT Payne SM Runyen-Janecky LJ Zhou S Schwartz DC Blattner FR Complete genome sequence and comparative genomics of Shigella flexneri serotype 2a strain 2457T Infect Immun 2003 71 2775 86 12704152 10.1128/IAI.71.5.2775-2786.2003
Darling AC Mau B Blattner FR Perna NT Mauve: multiple alignment of conserved genomic sequence with rearrangements Genome Res 2004 15 184 194 15590941 15590941
Glasner JD Liss P Plunkett G 3rdDarling A Prasad T Rusch M Byrnes A Gilson M Biehl B Blattner FR Perna NT ASAP, a systematic annotation package for community analysis of genomes Nucleic Acids Res 2003 31 147 5114 1394-403 12519969 10.1093/nar/gkg125
Ohnishi M Terajima J Kurokawa K Nakayama K Murata T Tamura K Ogura Y Watanabe H Hayashi T Genomic diversity of enterohemorrhagic Escherichia coli O157 revealed by whole genome PCR scanning Proc Natl Acad Sci U S A 2002 99 17043 8 2002 Dec 24; Epub 2002 Dec 12 12481030 10.1073/pnas.262441699
Emboss
|
16011797
|
PMC1187871
|
CC BY
|
2021-01-04 16:27:23
|
no
|
BMC Bioinformatics. 2005 Jul 12; 6:171
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-171
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1751601416810.1186/1471-2105-6-175SoftwareMPrime: efficient large scale multiple primer and oligonucleotide design for customized gene microarrays Rouchka Eric C [email protected] Abdelnaby [email protected] Nigel GF [email protected] Department of Computer Engineering and Computer Science, Speed School of Engineering, University of Louisville, Louisville, Kentucky, USA2 Department of Anatomical Sciences and Neurobiology, University of Louisville School of Medicine, Louisville, Kentucky, USA3 Bioinformatics Research Group, University of Louisville, Louisville, Kentucky, USA2005 13 7 2005 6 175 175 2 2 2005 13 7 2005 Copyright © 2005 Rouchka et al; licensee BioMed Central Ltd.2005Rouchka 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
Enhancements in sequencing technology have recently yielded assemblies of large genomes including rat, mouse, human, fruit fly, and zebrafish. The availability of large-scale genomic and genic sequence data coupled with advances in microarray technology have made it possible to study the expression of large numbers of sequence products under several different conditions in days where traditional molecular biology techniques might have taken months, or even years. Therefore, to efficiently study a number of gene products associated with a disease, pathway, or other biological process, it is necessary to be able to design primer pairs or oligonucleotides en masse rather than using a time consuming and laborious gene-by-gene method.
Results
We have developed an integrated system, MPrime, in order to efficiently calculate primer pairs or specific oligonucleotides for multiple genic regions based on a keyword, gene name, accession number, or sequence fasta format within the rat, mouse, human, fruit fly, and zebrafish genomes. A set of products created for mouse housekeeping genes from MPrime-designed primer pairs has been validated using both PCR-amplification and DNA sequencing.
Conclusion
These results indicate MPrime accurately incorporates standard PCR primer design characteristics to produce high scoring primer pairs for genes of interest. In addition, sequence similarity for a set of oligonucleotides constructed for the same set of genes indicates high specificity in oligo design.
==== Body
Background
Recent advances in DNA sequencing technologies have resulted in the availability of the assembly of a number of genomes at various stages. As of December 24, 2004, the Genomes OnLine Database [1] lists 1,245 on-going or completed genome projects. Included are the genomes of human [2,3], mouse [4], Norwegian brown rat [5], fruit fly [6], fugu [7], chimpanzee, and zebrafish [8]. In addition, various projects have been concerned with the curation of the genic regions of these genomes, most notably RefSeq [9].
Advances in microarray technology [10] have made it possible to simultaneously study the expression levels of thousands of genes under a given condition. Since a molecular biologist may be interested in studying a particular subset of genes in a given genome using microarray experimentation, it becomes necessary for the scientist to design complementary sequences to uniquely identify the genes of interest for either a cDNA product [10] or synthetic oligonucleotide [11-13] approach.
Custom microarray experimentation involves several steps, each of which must be performed in a timely manner in order to avoid becoming a bottleneck in the system. The steps involved are: determination of target genes; designing primers or oligonucleotides to create products uniquely identifying the target genes; spotting of the products on the microarray slide; detection of the presence of the gene under a given condition; and data analysis. MPrime focuses on the primer and oligonucleotide design stage in the microarray experimentation process.
Comparison to existing primer and oligo design software
Within the past few years, several algorithms for primer design have become available [14-21] including some specifically targeting whole genome analysis for microarrays [22-24]. In addition, more recent publications have become available discussing methods for oligo design [25-30], due to the emergence of custom oligonucleotide arrays. However, many of the primer design programs do not consider product similarity for cross hybridization, in part due to the lack of available genomic sequence at the time of their publication (see table 1).
Table 1 Comparison of properties of MPrime with various primer and oligonucleotide design programs.
Program Reference Primers Oligos Large Scale Design Sequence Similarity Avoidance
Primer Master Proutski and Holmes (1996) [19] X
PRIMO Li, et al. (1997) [17] X
GenomePRIDE Haas, et al. (1998) [14] X X X X
Primer3 Rozen and Skaletsky (2000) [21] X X
DOPRIMER Kampke et al., (2001) [15] X X
PrimeArray Raddatz et al. (2001) [20] X X
GST-PRIME Varotto et al. (2001) [23] X X
ProbeSelect Li and Stormo (2001) [25] X X X
PROBE Pozhitkov and Tautz (2002) [27] X X X
PRIMEGENS Xu et al. (2002) [24] X X X
OligoWiz Nielsen et al. (2003) [26] X X X
OligoArray 2.0 Rouillard et al. (2003) [29] X X X
OligoPicker Wang and Seed (2003) [30] X X X
ROSO Reymond, et al. (2004) [28] X X X
MPrime 2005 X X X X
There are a number of limitations to these approaches. Some of the algorithms find a primer or oligo from a single sequence at a time. Such an approach is time consuming and prone to human error, since the results must be transferred one at a time. Other algorithms may consider multiple sequences, yet fail to consider a sequence similarity score, resulting in products which do not uniquely identify genes. Some approaches pre-compute a set of products for a given genome, which works well if the complete genome is available, but behaves poorly in an adaptive environment. Very few of these approaches allow the user to specify whether they would like to design either primers or oligos, limiting them to one or the other.
MPrime attempts to overcome each of these limitations by allowing for efficient large scale multiple product design. MPrime approaches primer and oligo design by looking at a user-defined subset of genes of interest identified by either a GenBank [31] or RefSeq [9] accession number, by a gene name, or by the raw fasta sequence data. When the user specifies a particular genome of interest, MPrime will search through that genome to ensure that the products chosen will uniquely identify the gene the user is trying to identify.
Due to the methods employed, MPrime can be useful in large-scale primer and oligo design. Two software packages, GenomePRIDE [14] and GST-PRIME [23], have been described as offering large-scale, genome-wide design of oligomers and primer pairs for the construction of custom microarray products. MPrime provides advantages to each of these approaches.
GST-PRIME allows for an automated, genome-wide design of primer pairs in a similar fashion to MPrime. MPrime improves upon the functionality of GST-PRIME by allowing for the construction of gene-specific oligos for use with oligo-based custom microarrays. In addition, while GST-PRIME allows for the detection of primer pairs for a cDNA based chip on a whole genome, users of the system are limited to only choosing genes associated with a protein GenBank Identifier (GI) number. This inherently imposes limitations for the user. For instance, computationally predicted genes, genes supported by biological evidence (ESTs for example), or pre-published gene sequences cannot be used. MPrime avoids these complications by allowing the user to obtain the gene information directly from the GI number, or optionally from a raw fasta formatted sequence.
GenomePRIDE is the only known software package that incorporates high throughput primer pair and oligo design. MPrime distinguishes itself from GenomePRIDE by taking into account several factors, including GC content, secondary structure loop formation, and the presence of a GC clamp. While omitting these calculations makes the detection of primers and oligos much more efficient, it can also potentially lead to problems with both PCR product amplification and probe to product binding affinities. In addition, MPrime is freely available via a web interface, whereas GenomePRIDE has an associated cost of 150 euros for academic institutions.
Implementation
Selection of genic regions
The first step in primer design is to decide the regions to study. The interests of our collaborating lab lie in collating sets of mammalian programmed cell death genes related to neurological diseases such as Alzheimer's, Parkinson's, and Huntington's for use in microarray analyses. Therefore, an interface has been incorporated into MPrime in order to allow the user to retrieve genic regions by either a GenBank accession, gene name, or by a keyword. Retrieval by a keyword allows the user to have a general idea of the region of interest without having to spend time searching for an exact gene name or GenBank accession. Potential candidates are presented to the user. MPrime allows the user to refine their results by deciding on which candidate regions will be included before proceeding to primer or oligo design.
Primer design calculations
Although there are potentially a number of applications for the use of MPrime, it was originally designed with the intent for the construction of primers for custom cDNA microarray chips. In order to obtain the primer products with a high throughput, multiplex PCR can be implemented as the underlying biological experimentation. To ensure that the multiplex PCR experiments function properly and the microarray experimentation produces results that can be meaningfully analyzed the primer products should adhere to the properties of uniformity, sensitivity, and specificity.
Product uniformity
Multiplex PCR [32] is a technique allowing for the amplification of two or more targets in an organism of interest by incorporating multiple sets of PCR primers into a single reaction. Since each set of primers is subject to the exact same reaction conditions, they must be defined uniformly in order to react to the thermo cycling in a similar fashion. The properties which need to be relatively uniform include primer melting temperature, primer G+C concentration, and primer length.
MPrime requires the user to input a range of primer melting temperatures along with an optimal melting temperature. Any primers designed are required to fall inside of this range. There are a number of different approaches to calculating the melting temperature (Tm) of an oligonucleotide sequence, including an arbitrary method based on the formula Tm = 2 * (A + T) + 4 * (G + C) [33]; a long probe method based on the length and GC concentration [34]; and a nearest neighbor method based on dinucleotide interactions [35,36]. MPrime incorporates the nearest neighbor approach using the formula of Rychlik, et al. [36] where the Tm is calculated as follows:
Where ΔH is the enthalpy for helix formation, ΔS is the entropy for helix formation, c is the molar concentration of the primer (set at 250 pM); M is the molar concentration of Na+ (set at 50 mM) and R is the gas constant (1.987 cal/degree * mol).
Calculation of the primer G+C concentration is a straight-forward calculation of the percentage of the primer oligo containing G's or C's. MPrime requires the user to input the minimum, maximum, and optimal G+C concentration for the primers. In addition, the user can specify a GC clamp length, or the number of G's or C's required at the 3' end of the primer. The GC clamp helps to promote primer end stability, resulting in more efficient and specific bonding of the primer to the amplicon template.
In addition, MPrime requires the user to enter information concerning the primer length. The minimum, maximum, and optimal primer lengths are required. The primer length uniformity along with the uniform G+C content helps to maintain a uniform primer melting temperature.
Product sensitivity
PCR amplification can potentially fail due to either self or paired annealing. There are three main cases to be considered. In the first case, if either of the individual primers have a large number of consecutive complementary bases within their sequences they can potentially self-anneal to form a secondary structure such as a stem loop (see figure 1A). In the second case, complementary bases in paired primers can interact in their mid-regions to form partial double stranded structures (see figure 1B). The final case occurs when the primers interact with each other at the 3' end of either primer to produce primer-dimers (see figure 1C).
Figure 1 Primer secondary structure formation. Shown in figure 1A) is the secondary structure stem loop formation for the primer ACATACTGTGAGAAACACAGTATGT. Figure 1B) illustrates an example double-stranded structure formation between the primers ACTAGTACGTAGATCATTCG and GGATGCATACACGGAGAGAT. Note the run of six straight matches between the two sequences. Shown in figure 1C) is the primer-dimer formation between the primers ACTAGTACGTAGATCATTCG and GCATCTACCAGCGATAGCTA. Note the 3' end of the second primer has a run of seven straight matches to the first. The scores for each of these is based on the formula 4 * |G + C| + 2 * |A+T|. For figure 1A) there are 6 A-T base pairs, and 4 G-C base pairs, yielding a score of 28. For figure 1B), there are 6 A-T base pairs, and 3 G-C base pairs, yielding a score of 24. For figure 1C), there are 4 A-T base pairs and 3 G-C base pairs on the 3' end of the second sequence, yielding a score of 20.
One problem that can occur in product design is that either the products or the corresponding probe sequences could form secondary or tertiary structures, thus causing them to fail to interact, resulting in an apparent low signal. In order to overcome these problems of unwanted annealing between primers and products, MPrime incorporates a scoring scheme for paired-end and self-end annealing to reduce primer dimers; paired annealing to reduce partial double stranded structures; and self annealing to reduce secondary structure formation of single stranded sequences within both products and individual primers. These scoring schemes are adapted from Kampke, et al. [15] where each G-C base pair is assigned a score of 4 and each A-T base pairing is assigned a score of 2.
Product specificity
It is important that each primer set results in only the product of interest, and that each product uniquely identifies the gene being studied. In order to build in these constraints, MPrime searches the primers against a RefSeq [9] database for the organism being studied using wublastn [37]. The highest scoring sequences should be the genic region of interest; all other similar, suboptimal alignments will result from spurious hits. In order to assure that the product designed is specific to the gene of interest, the blast scores for all database matches are considered, and a specificity score is computed as the percent identity of the match multiplied by the percent of the product covered by the match. The blast results are then resorted based on this specificity score. The native locus should have a score of 1 (100% identity * 100% coverage). The next highest scoring segment is then taken to give an idea of the specificity of the designed product. For instance, if the product is completely unique, the second highest score will be 0. If the next highest scoring segment has an 80% identity of 60% of the sequence, the score would be 0.48. The product specificity score is weighted and calculated into the overall primer score. Note that in some instances, a gene will have duplicate entries in RefSeq. This may result in the product specificity score of 1 for all possible products. In such a case, the best scoring primer set is reported along with the product specificity score. In these cases, it is up to the user to determine whether or not such a primer set should be used.
MPrime incorporates an overall scoring scheme for each pair of primers. The basis behind this scheme is to find the best scoring primer pair with the smallest deviation from the overall optimal values. The smaller the score is, the more likely it is that the primer pair will function as desired. Each of the parameters entered is weighted evenly, with the exception that a larger weight is associated to product specificity. Similar scoring schemes have been proposed in Kampke, et al. [15] and are used in Primer3 [21].
Oligonucleotide calculations
MPrime also has the capability of calculating oligonucleotides for sequences of interest. The scoring scheme for determining the optimal oligonucleotides is very similar with the exception that there are no primer-primer interactions that can occur. With oligonucleotides, the similarity scores are weighted slightly higher, and are based on the actual length of the oligos desired. Typically, the 3' end of genes tends to be more variable and therefore more unique in their sequence. In addition, MPrime has the added feature of allowing the user to specify a region in which the oligo must be located.
Sequence input into MPrime
Sequences can be input into MPrime in one of four ways: by gene name, by GenBank accession number, by keyword, or by the raw fasta sequence data. The user can specify whether they are interested in sequences from the human, rat, mouse, fruit fly, or zebrafish genome. MPrime will then search for the sequence of interest, and return sequence records from RefSeq and the appropriate GenBank databases matching the search string. If the GenBank accession is given, MPrime first checks to see if there is a gene matching exactly.
Once the regions of interest are decided upon, the sequences are retrieved from local copies of either RefSeq or GenBank. For cDNAs, these sequences are then searched in order to retrieve primer pairs adhering to a set of guidelines. The primer length, G+C content, melting temperature, self-hybridization characteristics, and primer-primer hybridization characteristics must be taken into account in order to allow the primers to effectively be used in PCR experiments [15]. In addition, information on the sequence product is taken in as well, including the product length, G+C content, and melting temperature. The MPrime software was developed using both Perl and C++.
Since the end goal is to create products for incorporation into gene microarray chips, it is extremely important to provide relatively uniform product lengths in order to provide equal concentrations of sequence products and to be able to distinguish between the expression levels of different genes within a single experiment. The MPrime interface allows the user to enter in a range of values along with an optimal product length.
After potential primer pairs are created, they are searched against the appropriate database of genomic sequence in order to ensure the primer pairs and subsequent products are unique. If the products on the microarray chip are to uniquely identify genes, then it is necessary to check that the resulting products do not represent domains repeated in several genes.
Validation of MPrime
Sequence products produced by primer pairs selected from MPrime were validated using two independent techniques: (i) polymerase chain reaction and (ii) sequencing using the Beckman Coulter CEQ8000 Genetic Analysis System. Oligonucleotides constructed from MPrime were checked for their specificity by a nucleotide sequence search against Mus musculus sequences.
Polymerase chain reaction (PCR)
After primer pairs are determined, the sequence products are amplified using a 96 well plate DNA Engine Tetrad Thermal Cycler (MJ Research, Inc., Boston, MA). A MultiPROBE II Liquid Handler (Packard Instruments, Boston, MA) is used to help automate the process of creating the 96 well trays used in PCR amplification. The resulting products are then spotted on customized microarrays using a BioChip Arrayer (Packard Instruments).
PCR was performed using mouse brain total RNA (Clontech, Palo Alto, CA) and reverse transcribed into cDNA. RT-PCR was performed for a set of housekeeping genes and the primers used for RT-PCR were the same as those designed by MPrime program. Briefly, 5 μl of total RNA was used reverse transcribed using SuperScript II (Invitrogen, Carlsbad, CA) as described by the manufacture protocols. PCR reaction including 10 μl of cDNA template in a 100 μl reaction volume was amplified using DNA Engine Tetrad (MJ Research, Inc., Boston, MA). The thermal cycling profile consisted of 95°C for 3 min, 94°C for 30 s, 60°C for 30 s, 72°C for 30 s and 72°C for 7 min. Cycling kinetics were performed using 30 cycles. Amplified PCR reactions were separated on 2% agarose gel in the presence of ethidium bromide for visualization.
Sequencing
Purified mouse housekeeping genes products were sequenced using the CEQ 8000 Genetic Analysis System (Beckman Coulter Inc., Fullerton, CA). PCR was carried out using left primer and target cDNA from mouse brain as described above for 25 cycles. Sequences were then compared using a blast search against the corresponding GenBank and/or RefSeq entry for verification.
Oligonucleotide specificity
MPrime-computed oligonucleotides were compared against sequences from the organism Mus musculus using NCBI Blast [38]. The nucleotide-nucleotide Blast program blastn was used with default parameters. The advanced search option limiting to the Mus musculus dataset was used. Similar sequences reporting a bit score of 99.7 or higher (five mismatches in 70 bases) were reported as significant alignments for the purpose of oligonucleotide specificity.
Results
MPrime has allowed us to create multiple primers for related genic regions in a short span of time which are reported back to the user in the form of a web interface that can be stored as a tab delimited file that can be read into an Excel spreadsheet (figure 2). In our hands, detection of primers on a set of 138 rat actin, myosin and various muscle related genes with an average size of 2.7 KB finished in approximately 80 seconds on a dual processor AMD Athelon system with 1 GB of memory (results not shown). Analysis of the primers designed using the MPrime strategy compare well to the widely used Primer3 [21] software (results not shown).
Figure 2 Web interface and example of primer pair output. The initial inputs were for the rat genome with GenBank accession NM_053986, gene ACTB, and keyword "cancer". The default values for the primer size, GC%, melting temperature, product size, and GC clamp length were used. The user was able to refine these choices to include the four genic regions indicated in the results. A similar screen is produced when oligonucleotides are chosen as well (not shown).
A set of eleven randomly chosen mouse housekeeping genes (GAPDH, Actb, Sdf4, Ubb, Ywhaz, Hprt, Tubd1, Tuba1, Ppicap, Hspca, and Htf9c) was chosen for study. The resulting MPrime primer pairs detected for each of these genes is located in Table 2. The results of the PCR amplification showing products on the order of 400 bases for each of the primer pairs for these genes are shown (Figure 3). Sequencing of these products and subsequent similarity searches led to the identification of uniquely identifiable regions for each of these eleven product pairs. Sample results for the stromal cell derived factor 4 (Sdf4) are shown in terms of both the sequence (Figure 4) and the resulting sequence similarity results (Figure 5). The remaining ten sequences show similar results in terms of both their sequence composition and sequence similarity, validating that the sample set of primers produces the desired products. Thus, MPrime provides an effective alternative to finding primers one gene product at a time.
Table 2 Sample set of genes and resulting primer pairs using MPrime1.0. The top sequence in each primer pair is the forward primer, the second sequence is the reverse primer.
Gene Name Product Size Primer Pairs
GAPDH 400 AATGTGTCCGTCGTGGATCT
GGGTCTGGGATGGAAATTGT
Actb 400 GGCGCTTTTGACTCAGGATT
AGTTGGGGGGACAAAAAAAA
Sdf4 400 AGACCTGCCAACCACTCATC
TGCTTGCCAAAAACTTCACT
Ubb 400 TTCTGTGAGGGTGTGAGGGT
TTTATCCTGGATCTTGGCCT
Ywhaz 400 ACAATGTTCTTGGCCCATGT
AGGAAGAGGAGGAGGAAGGA
Hprt 400 TTATGCCGAGGATTTGGAAA
AACCTTAACCATTTTGGGGC
Tubd1 400 TAAGATGCTGGGTGTCCGTA
TAAGAGCTGGCTGTTGCTGA
Tuba1 400 GACCCTCGCCATGGTAAATA
AATCCACACCAACCTCCTCA
Ppicap 400 ACTCCCTCCCTCTTTCCCTG
CAGCAGAGAAAAGCTCCACC
Hspca 400 AGGAAACCCAGACCCAAGAC
ACACCAAACTGGCCAATCAT
Htf9c 400 TGTGAATTCCTGGTCGGAGT
TCTTTCTCTGTCCCTCCTCC
Figure 3 PCR amplification results for eleven mouse housekeeping genes. The resulting bands indicate single products with product sizes near 400 bases in length.
Figure 4 Sequencing results for the primer product for mouse stromal cell derived factor 4 (Sdf4) sequence.
Figure 5 Comparison of the PCR product for the mouse stromal cell derived factor 4 (Sdf4) obtained from the MPrime PCR primers (Figure 4) to the RefSeq sequence NM_011341.3 [GenBank:46195814].
MPrime-calculated oligonucleotides of length 70 for the set of eleven housekeeping genes are shown (Table 3). Each oligo was designed to occur in the last 500 bases on the 3' end of the sequences. The oligos were searched against the Mus musculus sequences in NCBI's database using NCBI Blast [38] with the default parameters. Alignments resulting in five or fewer matches over the entire length of the oligo were reported and summarized (Table 4). The resulting sequence similarities and bit scores are given. A perfect bit score for an oligo of length 70 using the default blast parameters is 139. The results show that ten of the eleven housekeeping genes map uniquely to a single RefSeq gene sequence, with the exception being GAPDH. Nine of these ten also map to a single genomic locus, with the exception of Ubb, which maps to three separate loci (including the native locus). Ubb, Hprt, Tubd1, Tuba1, Hspca, and Htf9c also map to additional sequences not shown in the table. In each case, these additional sequences represented either redundant mRNA or cDNA sequences for the corresponding gene, or their corresponding chromosomal location. For these ten sequences, the results suggest MPrime produces highly specific oligonucleotide sequences.
Table 3 Computed 70-mer oligos for eleven mouse housekeeping genes using MPrime 1.3 default parameters.
Gene name 70-mer oligonucleotide sequence
GAPDH ACTTTGTCAAGCTCATTTCCTGGTATGACAATGAATACGGCTACAGCAACAGGGTGGTGGACCTCATGGC
Actb AATAGTCATTCCAAGTATCCATGAAATAAGTGGTTACAGGAAGTCCCTCACCCTCCCAAAAGCCACCCCC
Sdf4 TGTTAAAAGAAAACATGAAGAGAGCTGTGGCTCTAGCTCAGTGGTCGAACGCTGCCCAGCAAGTAAAACG
Ubb CCTCCGTCTGAGGGGTGGCTATTAATTATTCGGTCTGCATTCCCAGTGGGCAGTGATGGCATTACTCTGC
Ywhaz CTGTCACCGTCTCCCTTTAAAATCCTTCCTCCTCCTCTTCCTCCTCCTCCTCCTCCTCACATAATGATGG
Hprt TTTTAGAAATGTCAGTTGCTGCGTCCCCAGACTTTTGATTTGCACTATGAGCCTATAGGCCAGCCTACCC
Tubd1 TTTGGACAGCTTTGCATTGTTGGAGCAAGTTGTTGCCAGTTATGGTAGTCTTGGACCCTAAGCCAAGAGG
Tuba1 GCTTCCACAGGGATGTTTATTGTGTTCCAACACAGAAAGTTGTGGTCTGATCAGTTAATTTCTATGTGGC
Ppicap ATTGTATTCAAATGAAAATTTACTAGAAGGTTTCAGCCAGCACTCACTCCAGGACTGAGAGTCCCAGGGC
Hspca TTAAAACAACCTGACAGGAATTCCCCAAGTGGCTTGTTTTCCAAAGTCCCGAGAACAACCCTAAGTTTCC
Htf9c CAGACTCCGCACTGTGAGATGCTTATCCTGTTTGAGAGGATGCAACAACACCCCAATGGCATAGAAGCCC
Table 4 Blastn similarities to MPrime computed oligos. 1mRNA sequence from which the RefSeq sequence was derived.
Gene name GenBank identifier Score (bits) Notes
GAPDH 47607489 139 RefSeq entry for GAPDH
Actb 6671508 139 RefSeq entry for Actb
498651 139 mRNA for beta actin
26104752 139 beta actin cDNA
191660 139 Mouse A-X beta actin mRNA
22316184 99.6 Chromosome X genomic sequence
Sdf4 46195814 139 RefSeq entry for Sdf4
475687541 139 mRNA for Sdf4
21953005 139 Chromosome 4 genomic sequence
26097254 131 Unknown EST
Ubb 6755918 139 RefSeq entry for Ubb
551771 139 mRNA for Ubb
Ywhaz 31981422 139 RefSeq entry for Ywhaz
297480011 139 mRNA for Ywhaz
1304165 139 mRNA for phosholipase A2
Hprt 7305154 139 RefSeq entry for Hprt
1939841 139 mRNA for Hprt
38348687 139 Chromosome X genomic sequence
Tubd1 9790052 139 RefSeq entry for Tubd1
58137751 139 mRNA for Tubd1
17221230 139 Chromosome 11 genomic sequence
Tuba1 6755900 139 RefSeq entry for Tuba1
2022071 139 mRNA for Tuba1
202222 139 Tuba1 gene, 3' end
Ppicap 6755143 139 RefSeq entry for Lgals3bp (Ppicap pseudonym)
2970321 139 mRNA for mama
17017768 139 Chromosome 11 genomic sequence
26348284 139 Ppicap cDNA
397799 139 CyCAP mRNA
Hspca 42476088 139 RefSeq entry for Hspca
128359861 139 Hspca mRNA
Htf9c 6680314 139 RefSeq entry for Htf9c
3189761 139 Htf9c mRNA
3818382 139 Chromosome 16 genomic sequence
For GAPDH, there were not any oligonucleotides found in the 500 bases on the 3' end of the RefSeq sequence of length 70 that were unique to GAPDH. The highest scoring oligo resulted in 37 matches with a perfect bit score of 139. This is reflected in the high sequence similarity score reported by MPrime. For these 37 matches, fourteen were to predicted similar mRNAs; five were similar to GAPDH; and four were GAPDH mRNA. The remaining fourteen sequences were genomic matches to chromosome 2 (2 sequences) chromosome 3 (3), chromosome 4 (2), chromosome 7 (2), chromosome 8 (3), chromosome 10 (1), and chromosome 15 (1).
MPrime uses RefSeq as the default database for searching for gene sequences. In order to understand more fully what happens with GAPDH, RefSeq was examined more closely. A search of RefSeq with the constraint "Mus musculus similar to glyceraldehyde-3" "Mus musculus" [ORGN] results in 118 different sequences, many of which completely cover the GAPDH sequence used by MPrime. Since the GAPDH sequence is not unique within the set of RefSeq sequences, a unique 70-mer could not be detected.
Discussion
In addition to the detection of primer products and single, long oligomers (on the order of 40–70 bases) that uniquely identify a gene of interest, individual researchers may want to design multiple probes for a single gene. Such an approach can help to limit the effects of cross-hybridization, the presence of multiple gene isoforms, and single nucleotide polymorphisms. An approach to design multiple short oligonucleotide probes (on the order of 25 bases long) has previously been presented [39]. MPrime presents the user the option of reporting back more than one resulting primer product or oligonucleotide for each gene. These results are reported in order based on their optimal score, with a score of 0 having the least deviation from the optimal user-selected conditions.
Researchers may additionally be interested in looking at multiple overlapping probes for each gene to produce tiled arrays that can help to detect the presence of different isoforms or single nucleotide polymorphisms present in an organism [40]. This option is not currently available in MPrime, but is a planned addition in a future release.
Conclusion
MPrime has proven to be an efficient and effective tool for both primer and oligonucleotide design. The methodology behind MPrime allows molecular biologists to construct a large number of primers and/or oligonucleotides for genes that need to have consistent properties. In addition, multiple sub-optimal results can be reported and tested. Preliminary tests on a set of housekeeping genes indicate that primer products obtained by MPrime-designed primer pairs produce uniquely identifying regions. Oligonucleotides designed by MPrime also appear to produce highly specific segments as indicated by similarity searches to a number of databases. As more organisms become sequenced and customized gene microarrays become cheaper, the availability of interactive tools such as MPrime to design sequences for custom use will become even more important.
Availability and requirements
Project name: MPrime: multiple primer design
Project home page:
Operating system: Platform independent (developed using Perl and C++)
Other requirements: MPrime is a freely available to both academic and commercial users as a web-based form.
List of abbreviations used
Actb: Beta actin
BLAST: Basic Local Alignment Search Tool
bp: Base pair
BRG: University of Louisville Bioinformatics Research Group
cDNA: Complementary DNA
DNA: deoxyribonucleic acid
GAPDH: Glyceraldehyde-3-phosphate dehydrogenase
GB: One billion bytes
Hprt: Hypoxanthine phosphoribosyltransferase
Hspca: Heat shock protein 1, alpha
Htf9c: HpaII tiny fragments locus 9C
KB: One thousand nucleotide bases
NCBI: National Center for Biotechnology Information
oligo: oligonucleotide
PCR: Polymerase chain reaction
Ppicap: Peptidylprolyl isomerase C-associated protein
RT-PCR: Reverse transcription polymerase chain reaction
Sdf4: Stromal cell derived factor 4
Tuba1: Tubulin, alpha 1
Tubd1: Tubulin, delta 1
Ubb: Ubiquitin B
Ywhaz: Tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, zeta polypeptide
Authors' contributions
ER coded the solution, and was involved in determining how to design primer pairs and oligonucleotides. In addition, ER was responsible for drafting the manuscript. AK participated in the design of the primer pair and oligonucleotide construction. In addition, he carried out the PCR amplifications and the DNA sequencing. AK was involved in testing and validation of MPrime. NC participated in the design of the primer pair in oligonucleotide construction as well as testing and validation of MPrime.
Acknowledgements
The authors wish to thank members of the University of Louisville Bioinformatics Research Group (BRG) for all of their support and important feedback. Support from NIH:NCRR grant P20 RR16481 (NC) is gratefully acknowledged.
==== Refs
Bernal A Ear U Kyrpides N Genomes OnLine Database (GOLD): a monitor of genome projects world-wide Nucleic Acids Res 2001 29 126 127 11125068 10.1093/nar/29.1.126
Lander ES Linton LM Birren B Nusbaum C Zody MC Baldwin J Devon K Dewar K Doyle M FitzHugh W Funke R Gage D Harris K Heaford A Howland J Kann L Lehoczky J Levine R McEwan P McKernan K Meldrim J Mesirov JP Miranda C Morris W Naylor J Raymond C Rosetti M Santos R Sheridan A Sougnez C Stange-Thomann N Stojanovic N Subramanian A Wyman D Rogers J Sulston J Ainscough R Beck S Bentley D Burton J Clee C Carter N Coulson A Deadman R Deloukas P Dunham A Dunham I Durbin R French L Grafham D Gregory S Hubbard T Humphray S Hunt A Jones M Lloyd C McMurray A Matthews L Mercer S Milne S Mullikin JC Mungall A Plumb R Ross M Shownkeen R Sims S Waterston RH Wilson RK Hillier LW McPherson JD Marra MA Mardis ER Fulton LA Chinwalla AT Pepin KH Gish WR Chissoe SL Wendl MC Delehaunty KD Miner TL Delehaunty A Kramer JB Cook LL Fulton RS Johnson DL Minx PJ Clifton SW Hawkins T Branscomb E Predki P Richardson P Wenning S Slezak T Doggett N Cheng JF Olsen A Lucas S Elkin C Uberbacher E Frazier M Gibbs RA Muzny DM Scherer SE Bouck JB Sodergren EJ Worley KC Rives CM Gorrell JH Metzker ML Naylor SL Kucherlapati RS Nelson DL Weinstock GM Sakaki Y Fujiyama A Hattori M Yada T Toyoda A Itoh T Kawagoe C Watanabe H Totoki Y Taylor T Weissenbach J Heilig R Saurin W Artiguenave F Brottier P Bruls T Pelletier E Robert C Wincker P Smith DR Doucette-Stamm L Rubenfield M Weinstock K Lee HM Dubois J Rosenthal A Platzer M Nyakatura G Taudien S Rump A Yang H Yu J Wang J Huang G Gu J Hood L Rowen L Madan A Qin S Davis RW Federspiel NA Abola AP Proctor MJ Myers RM Schmutz J Dickson M Grimwood J Cox DR Olson MV Kaul R Raymond C Shimizu N Kawasaki K Minoshima S Evans GA Athanasiou M Schultz R Roe BA Chen F Pan H Ramser J Lehrach H Reinhardt R McCombie WR de la BM Dedhia N Blocker H Hornischer K Nordsiek G Agarwala R Aravind L Bailey JA Bateman A Batzoglou S Birney E Bork P Brown DG Burge CB Cerutti L Chen HC Church D Clamp M Copley RR Doerks T Eddy SR Eichler EE Furey TS Galagan J Gilbert JG Harmon C Hayashizaki Y Haussler D Hermjakob H Hokamp K Jang W Johnson LS Jones TA Kasif S Kaspryzk A Kennedy S Kent WJ Kitts P Koonin EV Korf I Kulp D Lancet D Lowe TM McLysaght A Mikkelsen T Moran JV Mulder N Pollara VJ Ponting CP Schuler G Schultz J Slater G Smit AF Stupka E Szustakowski J Thierry-Mieg D Thierry-Mieg J Wagner L Wallis J Wheeler R Williams A Wolf YI Wolfe KH Yang SP Yeh RF Collins F Guyer MS Peterson J Felsenfeld A Wetterstrand KA Patrinos A Morgan MJ de JP Catanese JJ Osoegawa K Shizuya H Choi S Chen YJ Initial sequencing and analysis of the human genome Nature 2001 409 860 921 11237011 10.1038/35057062
Venter JC Adams MD Myers EW Li PW Mural RJ Sutton GG Smith HO Yandell M Evans CA Holt RA Gocayne JD Amanatides P Ballew RM Huson DH Wortman JR Zhang Q Kodira CD Zheng XH Chen L Skupski M Subramanian G Thomas PD Zhang J Gabor Miklos GL Nelson C Broder S Clark AG Nadeau J McKusick VA Zinder N Levine AJ Roberts RJ Simon M Slayman C Hunkapiller M Bolanos R Delcher A Dew I Fasulo D Flanigan M Florea L Halpern A Hannenhalli S Kravitz S Levy S Mobarry C Reinert K Remington K bu-Threideh J Beasley E Biddick K Bonazzi V Brandon R Cargill M Chandramouliswaran I Charlab R Chaturvedi K Deng Z Di FV Dunn P Eilbeck K Evangelista C Gabrielian AE Gan W Ge W Gong F Gu Z Guan P Heiman TJ Higgins ME Ji RR Ke Z Ketchum KA Lai Z Lei Y Li Z Li J Liang Y Lin X Lu F Merkulov GV Milshina N Moore HM Naik AK Narayan VA Neelam B Nusskern D Rusch DB Salzberg S Shao W Shue B Sun J Wang Z Wang A Wang X Wang J Wei M Wides R Xiao C Yan C Yao A Ye J Zhan M Zhang W Zhang H Zhao Q Zheng L Zhong F Zhong W Zhu S Zhao S Gilbert D Baumhueter S Spier G Carter C Cravchik A Woodage T Ali F An H Awe A Baldwin D Baden H Barnstead M Barrow I Beeson K Busam D Carver A Center A Cheng ML Curry L Danaher S Davenport L Desilets R Dietz S Dodson K Doup L Ferriera S Garg N Gluecksmann A Hart B Haynes J Haynes C Heiner C Hladun S Hostin D Houck J Howland T Ibegwam C Johnson J Kalush F Kline L Koduru S Love A Mann F May D McCawley S McIntosh T McMullen I Moy M Moy L Murphy B Nelson K Pfannkoch C Pratts E Puri V Qureshi H Reardon M Rodriguez R Rogers YH Romblad D Ruhfel B Scott R Sitter C Smallwood M Stewart E Strong R Suh E Thomas R Tint NN Tse S Vech C Wang G Wetter J Williams S Williams M Windsor S Winn-Deen E Wolfe K Zaveri J Zaveri K Abril JF Guigo R Campbell MJ Sjolander KV Karlak B Kejariwal A Mi H Lazareva B Hatton T Narechania A Diemer K Muruganujan A Guo N Sato S Bafna V Istrail S Lippert R Schwartz R Walenz B Yooseph S Allen D Basu A Baxendale J Blick L Caminha M Carnes-Stine J Caulk P Chiang YH Coyne M Dahlke C Mays A Dombroski M Donnelly M Ely D Esparham S Fosler C Gire H Glanowski S Glasser K Glodek A Gorokhov M Graham K Gropman B Harris M Heil J Henderson S Hoover J Jennings D Jordan C Jordan J Kasha J Kagan L Kraft C Levitsky A Lewis M Liu X Lopez J Ma D Majoros W McDaniel J Murphy S Newman M Nguyen T Nguyen N Nodell M The sequence of the human genome Science 2001 291 1304 1351 11181995 10.1126/science.1058040
Waterston RH Lindblad-Toh K Birney E Rogers J Abril JF Agarwal P Agarwala R Ainscough R Alexandersson M An P Antonarakis SE Attwood J Baertsch R Bailey J Barlow K Beck S Berry E Birren B Bloom T Bork P Botcherby M Bray N Brent MR Brown DG Brown SD Bult C Burton J Butler J Campbell RD Carninci P Cawley S Chiaromonte F Chinwalla AT Church DM Clamp M Clee C Collins FS Cook LL Copley RR Coulson A Couronne O Cuff J Curwen V Cutts T Daly M David R Davies J Delehaunty KD Deri J Dermitzakis ET Dewey C Dickens NJ Diekhans M Dodge S Dubchak I Dunn DM Eddy SR Elnitski L Emes RD Eswara P Eyras E Felsenfeld A Fewell GA Flicek P Foley K Frankel WN Fulton LA Fulton RS Furey TS Gage D Gibbs RA Glusman G Gnerre S Goldman N Goodstadt L Grafham D Graves TA Green ED Gregory S Guigo R Guyer M Hardison RC Haussler D Hayashizaki Y Hillier LW Hinrichs A Hlavina W Holzer T Hsu F Hua A Hubbard T Hunt A Jackson I Jaffe DB Johnson LS Jones M Jones TA Joy A Kamal M Karlsson EK Karolchik D Kasprzyk A Kawai J Keibler E Kells C Kent WJ Kirby A Kolbe DL Korf I Kucherlapati RS Kulbokas EJ Kulp D Landers T Leger JP Leonard S Letunic I Levine R Li J Li M Lloyd C Lucas S Ma B Maglott DR Mardis ER Matthews L Mauceli E Mayer JH McCarthy M McCombie WR McLaren S McLay K McPherson JD Meldrim J Meredith B Mesirov JP Miller W Miner TL Mongin E Montgomery KT Morgan M Mott R Mullikin JC Muzny DM Nash WE Nelson JO Nhan MN Nicol R Ning Z Nusbaum C O'Connor MJ Okazaki Y Oliver K Overton-Larty E Pachter L Parra G Pepin KH Peterson J Pevzner P Plumb R Pohl CS Poliakov A Ponce TC Ponting CP Potter S Quail M Reymond A Roe BA Roskin KM Rubin EM Rust AG Santos R Sapojnikov V Schultz B Schultz J Schwartz MS Schwartz S Scott C Seaman S Searle S Sharpe T Sheridan A Shownkeen R Sims S Singer JB Slater G Smit A Smith DR Spencer B Stabenau A Stange-Thomann N Sugnet C Suyama M Tesler G Thompson J Torrents D Trevaskis E Tromp J Ucla C Ureta-Vidal A Vinson JP Von Niederhausern AC Wade CM Wall M Weber RJ Weiss RB Wendl MC West AP Wetterstrand K Wheeler R Whelan S Wierzbowski J Willey D Williams S Wilson RK Winter E Worley KC Wyman D Yang S Yang SP Zdobnov EM Zody MC Lander ES Initial sequencing and comparative analysis of the mouse genome Nature 2002 420 520 562 12466850 10.1038/nature01262
Gibbs RA Weinstock GM Metzker ML Muzny DM Sodergren EJ Scherer S Scott G Steffen D Worley KC Burch PE Okwuonu G Hines S Lewis L DeRamo C Delgado O Dugan-Rocha S Miner G Morgan M Hawes A Gill R Celera Holt RA Adams MD Amanatides PG Baden-Tillson H Barnstead M Chin S Evans CA Ferriera S Fosler C Glodek A Gu Z Jennings D Kraft CL Nguyen T Pfannkoch CM Sitter C Sutton GG Venter JC Woodage T Smith D Lee HM Gustafson E Cahill P Kana A Doucette-Stamm L Weinstock K Fechtel K Weiss RB Dunn DM Green ED Blakesley RW Bouffard GG de Jong PJ Osoegawa K Zhu B Marra M Schein J Bosdet I Fjell C Jones S Krzywinski M Mathewson C Siddiqui A Wye N McPherson J Zhao S Fraser CM Shetty J Shatsman S Geer K Chen Y Abramzon S Nierman WC Havlak PH Chen R Durbin KJ Egan A Ren Y Song XZ Li B Liu Y Qin X Cawley S Worley KC Cooney AJ D'Souza LM Martin K Wu JQ Gonzalez-Garay ML Jackson AR Kalafus KJ McLeod MP Milosavljevic A Virk D Volkov A Wheeler DA Zhang Z Bailey JA Eichler EE Tuzun E Birney E Mongin E Ureta-Vidal A Woodwark C Zdobnov E Bork P Suyama M Torrents D Alexandersson M Trask BJ Young JM Huang H Wang H Xing H Daniels S Gietzen D Schmidt J Stevens K Vitt U Wingrove J Camara F Mar AM Abril JF Guigo R Smit A Dubchak I Rubin EM Couronne O Poliakov A Hubner N Ganten D Goesele C Hummel O Kreitler T Lee YA Monti J Schulz H Zimdahl H Himmelbauer H Lehrach H Jacob HJ Bromberg S Gullings-Handley J Jensen-Seaman MI Kwitek AE Lazar J Pasko D Tonellato PJ Twigger S Ponting CP Duarte JM Rice S Goodstadt L Beatson SA Emes RD Winter EE Webber C Brandt P Nyakatura G Adetobi M Chiaromonte F Elnitski L Eswara P Hardison RC Hou M Kolbe D Makova K Miller W Nekrutenko A Riemer C Schwartz S Taylor J Yang S Zhang Y Lindpaintner K Andrews TD Caccamo M Clamp M Clarke L Curwen V Durbin R Eyras E Searle SM Cooper GM Batzoglou S Brudno M Sidow A Stone EA Venter JC Payseur BA Bourque G Lopez-Otin C Puente XS Chakrabarti K Chatterji S Dewey C Pachter L Bray N Yap VB Caspi A Tesler G Pevzner PA Haussler D Roskin KM Baertsch R Clawson H Furey TS Hinrichs AS Karolchik D Kent WJ Rosenbloom KR Trumbower H Weirauch M Cooper DN Stenson PD Ma B Brent M Arumugam M Shteynberg D Copley RR Taylor MS Riethman H Mudunuri U Peterson J Guyer M Felsenfeld A Old S Mockrin S Collins F Genome sequence of the Brown Norway rat yields insights into mammalian evolution Nature 2004 428 493 521 15057822 10.1038/nature02426
Adams MD Celniker SE Holt RA Evans CA Gocayne JD Amanatides PG Scherer SE Li PW Hoskins RA Galle RF George RA Lewis SE Richards S Ashburner M Henderson SN Sutton GG Wortman JR Yandell MD Zhang Q Chen LX Brandon RC Rogers YH Blazej RG Champe M Pfeiffer BD Wan KH Doyle C Baxter EG Helt G Nelson CR Gabor GL Abril JF Agbayani A An HJ ndrews-Pfannkoch C Baldwin D Ballew RM Basu A Baxendale J Bayraktaroglu L Beasley EM Beeson KY Benos PV Berman BP Bhandari D Bolshakov S Borkova D Botchan MR Bouck J Brokstein P Brottier P Burtis KC Busam DA Butler H Cadieu E Center A Chandra I Cherry JM Cawley S Dahlke C Davenport LB Davies P de PB Delcher A Deng Z Mays AD Dew I Dietz SM Dodson K Doup LE Downes M Dugan-Rocha S Dunkov BC Dunn P Durbin KJ Evangelista CC Ferraz C Ferriera S Fleischmann W Fosler C Gabrielian AE Garg NS Gelbart WM Glasser K Glodek A Gong F Gorrell JH Gu Z Guan P Harris M Harris NL Harvey D Heiman TJ Hernandez JR Houck J Hostin D Houston KA Howland TJ Wei MH Ibegwam C Jalali M Kalush F Karpen GH Ke Z Kennison JA Ketchum KA Kimmel BE Kodira CD Kraft C Kravitz S Kulp D Lai Z Lasko P Lei Y Levitsky AA Li J Li Z Liang Y Lin X Liu X Mattei B McIntosh TC McLeod MP McPherson D Merkulov G Milshina NV Mobarry C Morris J Moshrefi A Mount SM Moy M Murphy B Murphy L Muzny DM Nelson DL Nelson DR Nelson KA Nixon K Nusskern DR Pacleb JM Palazzolo M Pittman GS Pan S Pollard J Puri V Reese MG Reinert K Remington K Saunders RD Scheeler F Shen H Shue BC Siden-Kiamos I Simpson M Skupski MP Smith T Spier E Spradling AC Stapleton M Strong R Sun E Svirskas R Tector C Turner R Venter E Wang AH Wang X Wang ZY Wassarman DA Weinstock GM Weissenbach J Williams SM WoodageT Worley KC Wu D Yang S Yao QA Ye J Yeh RF Zaveri JS Zhan M Zhang G Zhao Q Zheng L Zheng XH Zhong FN Zhong W Zhou X Zhu S Zhu X Smith HO Gibbs RA Myers EW Rubin GM Venter JC The genome sequence of Drosophila melanogaster Science 2000 287 2185 2195 10731132 10.1126/science.287.5461.2185
Aparicio S Chapman J Stupka E Putnam N Chia JM Dehal P Christoffels A Rash S Hoon S Smit A Gelpke MD Roach J Oh T Ho IY Wong M Detter C Verhoef F Predki P Tay A Lucas S Richardson P Smith SF Clark MS Edwards YJ Doggett N Zharkikh A Tavtigian SV Pruss D Barnstead M Evans C Baden H Powell J Glusman G Rowen L Hood L Tan YH Elgar G Hawkins T Venkatesh B Rokhsar D Brenner S Whole-genome shotgun assembly and analysis of the genome of Fugu rubripes Science 2002 297 1301 1310 12142439 10.1126/science.1072104
UCSC Genome Browser 2005
Pruitt KD Tatusova T Maglott DR NCBI Reference Sequence (RefSeq): a curated non-redundant sequence database of genomes, transcripts and proteins Nucleic Acids Res 2005 33 D501 D504 15608248 10.1093/nar/gki025
Schena M Shalon D Davis RW Brown PO Quantitative monitoring of gene expression patterns with a complementary DNA microarray Science 1995 270 467 470 7569999
Gerhold D Rushmore T Caskey CT DNA chips: promising toys have become powerful tools Trends Biochem Sci 1999 24 168 173 10322428 10.1016/S0968-0004(99)01382-1
Lockhart DJ Dong H Byrne MC Follettie MT Gallo MV Chee MS Mittmann M Wang C Kobayashi M Horton H Brown EL Expression monitoring by hybridization to high-density oligonucleotide arrays Nat Biotechnol 1996 14 1675 1680 9634850 10.1038/nbt1296-1675
Pease AC Solas D Sullivan EJ Cronin MT Holmes CP Fodor SP Light-generated oligonucleotide arrays for rapid DNA sequence analysis Proc Natl Acad Sci U S A 1994 91 5022 5026 8197176
Haas S Vingron M Poustka A Wiemann S Primer design for large scale sequencing Nucleic Acids Res 1998 26 3006 3012 9611248 10.1093/nar/26.12.3006
Kampke T Kieninger M Mecklenburg M Efficient primer design algorithms Bioinformatics 2001 17 214 225 11294787 10.1093/bioinformatics/17.3.214
Li LC Dahiya R MethPrimer: designing primers for methylation PCRs Bioinformatics 2002 18 1427 1431 12424112 10.1093/bioinformatics/18.11.1427
Li P Kupfer KC Davies CJ Burbee D Evans GA Garner HR PRIMO: A primer design program that applies base quality statistics for automated large-scale DNA sequencing Genomics 1997 40 476 485 9073516 10.1006/geno.1996.4560
McKay SJ Jones SJ AcePrimer: automation of PCR primer design based on gene structure Bioinformatics 2002 18 1538 1539 12424127 10.1093/bioinformatics/18.11.1538
Proutski V Holmes EC Primer Master: a new program for the design and analysis of PCR primers Comput Appl Biosci 1996 12 253 255 8872397
Raddatz G Dehio M Meyer TF Dehio C PrimeArray: genome-scale primer design for DNA-microarray construction Bioinformatics 2001 17 98 99 11222267 10.1093/bioinformatics/17.1.98
Rozen S Skaletsky H Primer3 on the WWW for general users and for biologist programmers Methods Mol Biol 2000 132 365 386 10547847
Fernandes RJ Skiena SS Microarray synthesis through multiple-use PCR primer design Bioinformatics 2002 18 Suppl 1 S128 S135 12169540
Varotto C Richly E Salamini F Leister D GST-PRIME: a genome-wide primer design software for the generation of gene sequence tags Nucleic Acids Res 2001 29 4373 4377 11691924 10.1093/nar/29.21.4373
Xu D Li G Wu L Zhou J Xu Y PRIMEGENS: robust and efficient design of gene-specific probes for microarray analysis Bioinformatics 2002 18 1432 1437 12424113 10.1093/bioinformatics/18.11.1432
Li F Stormo GD Selection of optimal DNA oligos for gene expression arrays Bioinformatics 2001 17 1067 1076 11724738 10.1093/bioinformatics/17.11.1067
Nielsen HB Wernersson R Knudsen S Design of oligonucleotides for microarrays and perspectives for design of multi-transcriptome arrays Nucleic Acids Res 2003 31 3491 3496 12824351 10.1093/nar/gkg622
Pozhitkov AE Tautz D An algorithm and program for finding sequence specific oligonucleotide probes for species identification BMC Bioinformatics 2002 3 9 11882251 10.1186/1471-2105-3-9
Reymond N Charles H Duret L Calevro F Beslon G Fayard JM ROSO: optimizing oligonucleotide probes for microarrays Bioinformatics 2004 20 271 273 14734320 10.1093/bioinformatics/btg401
Rouillard JM Zuker M Gulari E OligoArray 2.0: design of oligonucleotide probes for DNA microarrays using a thermodynamic approach Nucleic Acids Res 2003 31 3057 3062 12799432 10.1093/nar/gkg426
Wang X Seed B Selection of oligonucleotide probes for protein coding sequences Bioinformatics 2003 19 796 802 12724288 10.1093/bioinformatics/btg086
Benson DA Karsch-Mizrachi I Lipman DJ Ostell J Wheeler DL GenBank Nucleic Acids Res 2005 33 D34 D38 15608212 10.1093/nar/gki063
Chamberlain JS Gibbs RA Ranier JE Nguyen PN Caskey CT Deletion screening of the Duchenne muscular dystrophy locus via multiplex DNA amplification Nucleic Acids Res 1988 16 11141 11156 3205741
Wallace RB Shaffer J Murphy RF Bonner J Hirose T Itakura K Hybridization of synthetic oligodeoxyribonucleotides to phi chi 174 DNA: the effect of single base pair mismatch Nucleic Acids Res 1979 6 3543 3557 158748
Meinkoth J Wahl G Hybridization of nucleic acids immobilized on solid supports Anal Biochem 1984 138 267 284 6204550 10.1016/0003-2697(84)90808-X
Breslauer KJ Frank R Blocker H Marky LA Predicting DNA duplex stability from the base sequence Proc Natl Acad Sci U S A 1986 83 3746 3750 3459152
Rychlik W Spencer WJ Rhoads RE Optimization of the annealing temperature for DNA amplification in vitro Nucleic Acids Res 1990 18 6409 6412 2243783
WU BLAST 2005
NCBI BLAST 2005
Rahmann S Fast large scale oligonucleotide selection using the longest common factor approach J Bioinform Comput Biol 2003 1 343 361 15290776 10.1142/S0219720003000125
Shoemaker DD Schadt EE Armour CD He YD Garrett-Engele P McDonagh PD Loerch PM Leonardson A Lum PY Cavet G Wu LF Altschuler SJ Edwards S King J Tsang JS Schimmack G Schelter JM Koch J Ziman M Marton MJ Li B Cundiff P Ward T Castle J Krolewski M Meyer MR Mao M Burchard J Kidd MJ Dai H Phillips JW Linsley PS Stoughton R Scherer S Boguski MS Experimental annotation of the human genome using microarray technology Nature 2001 409 922 927 11237012 10.1038/35057141
|
16014168
|
PMC1187872
|
CC BY
|
2021-01-04 16:27:25
|
no
|
BMC Bioinformatics. 2005 Jul 13; 6:175
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-175
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1791602661310.1186/1471-2105-6-179SoftwareProcessing methods for differential analysis of LC/MS profile data Katajamaa Mikko [email protected]šič Matej [email protected] Turku Centre for Biotechnology, Tykistökatu 6, FIN-20521, Turku, Finland2 VTT Biotechnology, Tietotie 2, P.O. Box 1500, FIN-02044 VTT, Espoo, Finland2005 18 7 2005 6 179 179 8 3 2005 18 7 2005 Copyright © 2005 Katajamaa and Orešič; licensee BioMed Central Ltd.2005Katajamaa and Orešič; 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
Liquid chromatography coupled to mass spectrometry (LC/MS) has been widely used in proteomics and metabolomics research. In this context, the technology has been increasingly used for differential profiling, i.e. broad screening of biomolecular components across multiple samples in order to elucidate the observed phenotypes and discover biomarkers. One of the major challenges in this domain remains development of better solutions for processing of LC/MS data.
Results
We present a software package MZmine that enables differential LC/MS analysis of metabolomics data. This software is a toolbox containing methods for all data processing stages preceding differential analysis: spectral filtering, peak detection, alignment and normalization. Specifically, we developed and implemented a new recursive peak search algorithm and a secondary peak picking method for improving already aligned results, as well as a normalization tool that uses multiple internal standards. Visualization tools enable comparative viewing of data across multiple samples. Peak lists can be exported into other data analysis programs. The toolbox has already been utilized in a wide range of applications. We demonstrate its utility on an example of metabolic profiling of Catharanthus roseus cell cultures.
Conclusion
The software is freely available under the GNU General Public License and it can be obtained from the project web page at: .
==== Body
Background
Liquid chromatography coupled to mass spectrometry [1,2] (LC/MS) has been widely used in proteomics [3] and metabolomics [4] research. In this context, the technology has been increasingly utilized for differential profiling, i.e. broad screening of biomolecular components across multiple samples (corresponding to different conditions, interventions, or time points) in order to elucidate the observed phenotypes or discover biomarkers [5,6].
Typical LC/MS experiments include several analytical stages, starting with sample pre-treatment which commonly includes sample cleanup and extraction methods. The sample can then be introduced to the LC column where the molecules separate based on their size (size exclusion chromatography), affinity to stationary phase (affinity chromatography), polarity (ion exchange chromatography), or hydrophobicity (reversed phase chromatography). The retention time measures the time between the sample injection and the appearance of the compound peak maximum after chromatographic separation. In analyses of complex mixtures, it is likely that many analytes elute at the same or similar time and individual compound peaks cannot be resolved by LC alone. Mass spectrometry (MS) can then be used to separate the co-elutants according to mass-to-charge ratio (m/z). The co-elutants enter the LC-MS interface where they are ionized and introduced into the mass spectrometer where m/z is measured. Several ionization methods exist, among the most commonly used are the soft ionization methods such as electrospray ionization (ESI) and atmospheric pressure – chemical ionization (APCI). The principles of mass detection can also vary, with the most common instruments being triple quadrupole, (quadrupole) ion trap, (quadrupole) time of flight mass spectrometers [2]. While discussion of the merits of each type of chromatography, ion source, and mass detector are beyond the scope of this paper, it is evident that many different types of applications can be developed with LC/MS. Due to such variety of possible applications and approaches it is also challenging to develop a generic solution for processing and analysis of LC/MS data. Additionally, the commercial software solutions provided by instrument vendors are limited to the instruments provided by the vendors. Although this may change in the future by adoption of mzData [7] data representation format, mzData does not represent the raw data and as such may have its limitations.
One increasingly utilized type of LC/MS application is differential profiling, where the extraction, LC methods, and MS instrument setup are set to provide a broad coverage of compounds, with the main aim to enable relative quantitative comparisons for individual compounds across multiple samples. The applications of such approach can be found in domains of systems biology, functional genomics, and biomarker discovery. While such approaches cannot match targeted analytical measurements in ability to accurately quantitate individual analytes, it is the role of data processing methods to enable comparative studies of analytes, even if they may be unknown [5]. The data processing for differential profiling usually proceeds through several stages. Spectral filtering stage aims at reducing the complexity of spectra and removing the noise. Peak detection finds the peaks corresponding to the compounds or fragments thereof. Alignment, data processing step specific to profiling experiments, aims at matching the corresponding peaks across multiple sample runs. The role of normalization is then to reduce the systematic error by adjusting the intensities within each sample run.
Few integrated solutions for differential analysis of LC/MS data have already been introduced for proteomics and metabolomics applications. MarkerLynx, the commercial package from Waters, Inc. is an add-on to MassLynx (Waters, Inc.) software. IMPRESS and WINLIN software packages (TNO Pharma, The Netherlands) perform the smoothing on each mass trace separately, followed by entropy based method to filter the traces [8]. The alignment is then performed using the partial linear fit method initially developed for aligning NMR spectra [9]. Proprietary MassView software [10] and a toolkit by Radulovic et al. [11] were developed upon the similar principles for proteomics applications, while approaching the peak detection in 2 dimensions (retention time and m/z). The Bioinformatics Toolbox in Matlab (Mathworks, Inc.) contains capabilities for preprocessing of mass spectra which can be utilized on MS data analysis applications of low memory and performance demand.
One of the challenges in algorithm and application development in domain of LC/MS data processing is that a solution for a particular stage of processing is of limited use if it is not embedded into the full data processing pipeline. Therefore, an integrated LC/MS software environment enabling easy integration of new methods would benefit both the algorithm developers and the end users. In this paper we report development of tools for differential profiling of LC/MS data, aiming primarily at metabolomics applications, as well as a new platform independent open source software package MZmine built to integrate these tools. We demonstrate its utility on an example of metabolic profiling of Catharanthus roseus cell cultures.
Implementation
MZmine is a collection of methods for data processing stages used in differential profiling of LC/MS data. Scope of the software is limited to data processing, and therefore other tools should be used for statistical analyses following the initial data processing.
Software design of the toolbox
Main goals in the design of the software have been good usability and expandability with new data processing methods. To facilitate good usability, we have developed a graphical user interface shown in Figure 1. The GUI allows user to experiment with different combinations of the data processing methods and parameter values for each of the steps and visually check quality of intermediate results. Some experimenting and visual validation is often required to find the best methods and parameter values for a new set of data.
Figure 1 MZmine graphical user interface: (A) List of imported raw data files. (B) Total ion chromatogram (TIC) for selected file. (C) Mass spectrum for selected retention time for the same file. (D) Peak list for the same file, with listed m/z values, retention times, and intensities. (E) 2d map of the same file, with retention time on x-axis and m/z on y-axis. (F) Zoomed-in spectra for a different file. (G) Peak alignment matrix for all files listed. (H) Available alignment results, e.g. for different normalizations. The spectra shown in the GUI are from lipidomic profiling of mouse white adipose tissue using Quattro Micro (Waters, Inc.) triple quadrupole mass spectrometer.
The toolbox is implemented as a stand-alone Java application. While using Java language means only slight performance degradation compared to C++ [12], it affords platform independence. The class model for the software contains interfaces for each of the data processing stages, and new data processing methods can be added to the toolbox by implementing a suitable interface.
Input data formats and conversion
The toolbox accepts input in NetCDF format. In order to implement the support for NetCDF files in the toolbox, we used NetCDF Java Library (Version 2) by Unidata community [13]. Most of the mass spectrometer vendors provide converters for translating raw data files from their proprietary format to this common presentation. We have tested the toolbox with NetCDF metabolomics or protein tryptic digest data created from the following instruments: Quattro Micro (Waters), QTof Premier (Waters), QSTAR Pulsar (Applied Biosystems), LTQ-FTMS (Thermo Finnigan), and LCQ (Thermo Finnigan). In the future, toolbox will also include support for upcoming new mass spectrometry data formats such as mzData [7] and mzXML [14].
Since raw LC/MS data files can be large relative to the available main memory, the core classes for representing raw data do not load all spectra into memory at once, but retrieve necessary parts from disk when requested. This makes it possible to visualise and work with several large raw data files at the same time.
Smoothing and peak detection
Smoothing aims to remove noise in the measured spectra, which facilitates further peak detection. Smoothing is an optional stage in data processing and can also be left out if the data is not noisy or if the input data is already available as centroids. For smoothing the spectra, toolbox offers implementations of a moving average filter and Savitzky-Golay filter.
After smoothing, peak detection is done to find the peaks in the measurement data. The toolbox contains two peak picking methods: local maximum method and recursive threshold method. Both of these methods work in similar steps, which are shown as a listing:
calc eXtracted Ion Chromatograms (XICs) for m/z bins
for each spectrum
find peaks in spectrum
filter out spectral peaks with the lowest intensities
connect new spectral peaks with previous ones
end
filter out too long and too short 2D-peaks
The extracted ion chromatogram (XIC) is a curve showing time vs. sum of intensities over a small m/z range. Only row in the listing where the two available peak detection methods differ is the step for finding one-dimensional spectral peaks, i.e. peaks found in mass spectra of each instrument scan. Local maximum method treats every local intensity maximum along the spectrum as a spectral peak, while recursive threshold method requires maximum to have a user-definable width that differentiates it from sharper noise peaks. Spectral peaks are filtered in both chromatographic and m/z directions to remove those with weakest intensities. To speed-up chromatographic filtering, a set of XICs is precalculated for m/z bins of user-definable width before looping through spectra. Chromatographic filtering is then done inside the loop by comparing spectral peak's intensity to intensities of a XIC curve that goes through the location of spectral peak.
During the loop through the spectra, one dimensional spectral peaks of current spectrum are connected with spectral peaks of the spectrum from the previous scan to form two-dimensional strings of spectral peaks. Joining occurs only between the peaks in successive spectra that have similar m/z values according to pre-set threshold, and form together a good shaped peak in the chromatographic direction. Figure 2 shows a simple example of the two-step process: first finding spectral peaks and then connecting them.
Figure 2 The two-step peak picking process used by the two available peak picking methods: (A) This plot is a zoom-in to a small part of a spectrum. In the first step, one-dimensional spectral peaks are detected in each spectrum alone. Green dots over the spectrum show the locations of detected spectral peaks. (B) This plot is a zoom-in to a small fragment of two-dimensional view of raw data. Black lines show two-dimensional peaks created by connecting successive spectral peaks. Peak height is calculated as the highest intensity among these data points, while the peak area corresponds to the sum of the intensities.
Choice of methods for smoothing and peak detection depends on the nature of input data. If data is already pre-processed and centroided, smoothing is not needed and the peak picking method based on searching for local maximums works the best. When working with spectral data acquired in continuous mode, recursive threshold peak picker gives better results.
Peak picking gives two measures for the size of the peak: peak height is defined as the maximum intensity of all datapoints forming the peak and peak area is measured as sum over intensities of all datapoints. It is user-selectable which of these two quantities is used in the further processing stages. Following peak detection, each LC/MS run s(s = 1...S) is represented by a peak list:
Ps = {pisc}; with i = 1...Ns and c = {mz, δmz, rt, δrt, height, area};, (1)
where Ns is a total number of peaks for run s and c is an index for parameters of each peak pis : mz is the mean m/z value for data points within the peak, δmz is standard deviation of m/z values within the peak, rt is retention time at the maximum intensity datapoint, δrt is the lengths of the peak in time, height is height of the peak and area is area of the peak calculated as described above.
Alignment
Alignment methods search for corresponding peaks across different LC/MS runs. Peaks from the same compound match usually closely in m/z values, but there can be variation in retention times between the runs. The former depends on mass accuracy and resolution of the mass spectrometer while the latter largely depends on the analytical method used.
The results of alignment are represented by a master peak list:
Q = {qjsc}; with j = 1...Npeaks, s = 1...S and c = {mz, δmz, rt, δrt, height, area};, (2)
where Npeaks is the number of rows in the master peak list matrix. Element qjs is set to empty value when no peak from the peak list s has been aligned to row j of master peak list.
The toolbox currently implements a simple alignment method utilizing the master peak list. This method takes one peak from a peak list at a time and aligns the peak to either the best matching existing row of the master peak list or appends a new row to the master peak list, if matching row is not found for the peak. The alignment process is described in the pseudocode:
for s = 1...S
for i = 1...Ns
alignToMaster(pis)
end
end
function alignToMaster(pis)
pick j such that
qjs does not contain p' i' s with dist(Qj, p'i's) <=dist(Qj, pis) and
minimizes dist(Qj, pis)
if (dist(Qj, pis)>user-defined threshold) or (could not pick j)
append pis to a new, empty row of Q
else
if qjs contains p'i's
assign qjs := pis
alignToMaster(p'i's)
else
assign qjs := pis
end
end
end
where the distance between a peak pis and a master peak list row Qj is calculated using function:
where pi,s ,mz and pi,s ,rt are the m/z ratio and retention time of a peak in an individual peak list and and are the average m/z ratio and retention time of peaks from all peak lists except s assigned to the same row Qj of the master peak list. k is an adjustable parameter that controls the balance between accuracy of m/z ratio and retention time values. Generally, k can be set to a larger number with increased mass accuracy and resolution of the mass detector.
After aligning peak lists as described above, it is likely that master peak list contains empty gaps, because not every peak is detected and aligned in every sample. Such missing values often complicate further statistical analyses, and for this reason we developed a secondary peak picking method for filling these gaps. This method uses and values for estimating location where a missing peak should be found. Search is then conducted to find the highest local maximum over a range around the expected location in the raw measurement signal. The size of the search range is a user-definable parameter in the gap-filling method. Intensity of the local maximum is then used as estimated peak height.
Normalization methods
Normalization is needed to reduce the systematic error in data. The toolbox implements two different approaches: a set of linear normalization methods and a new approach that utilizes multiple internal standard compounds injected to the samples.
Linear normalization methods divide all peak heights within a single peak list by the same number. Implementation of linear normalization method in the toolbox offers four different ways to calculate the normalization factor: average peak height, average squared peak height, maximum peak height and total raw signal.
The toolbox also contains a new normalization method that utilizes information from multiple standard compounds, which are injected to each of the samples in known concentrations prior to LC/MS analysis. The standard compound peaks can be used to calculate a set of normalization factors, one for each standard compound. There are currently two different ways to use this information in normalization. One option is to search for a standard compound peak closest to the peak. The distance function is same as (3). A variation of this method is the method based on normalization using weighted contribution of each standard compound. In this method, the same distance metric as above (3) is used to calculate distance of a peak to each standard compound. Contribution of each standard to the final normalization factor is weighted using the inverse of distance between the peak and the standard.
where m is the number of injected standard compounds, nfl is the normalization factor calculated using lth standard compound and dist(p, ISl) is the distance between peak to be normalized and peak of the lth standard compound. Both methods reduce to the common single-standard calibration when m = 1, i.e. only a single internal standard is used.
Visualization methods
After processing, data is ready to be exported from the toolbox as a tab-delimited peak height and area matrix. This matrix can be then further processed with packages such as Matlab (MathWorks, Inc.) or R Statistical Language which already have a large collection of data analysis tools available for statistical analyses of multivariate data.
The toolbox also contains two visualization methods for quickly previewing the processed results. Both of these methods plot the peaks to a two-dimensional plot where x-axis is the retention time and y-axis m/z ratio. Logratio plot is useful for displaying differences in peak heights between two groups of samples (Figure 3). Differences are measured using logratio value, which compares average peak heights in two selected groups:
Figure 3 (A) Total ion chromatograms from two representative samples from Catharanthus roseus cell cultures. (B) Log-ratio plot view, comparing mean intensities of detected peaks between two selected groups of samples from Catharanthus roseus (10 elicited vs. 10 controls).
where and are average heights of peak pi, •, • in the first and second group of peak lists, respectively. In the logratio plot color coding is used for visualizing the logratio values: red shades for positive logratio values and green shades for negative logratio values.
Another visualization method is a coefficient of variation plot, which displays variation of peak heights within one group of samples:
where is the average peak height and σ(pi ,{s}, height) is the standard deviation of peak heights in the selected group of samples {s}. The coefficient of variation plot is drawn similarly as the logratio plot, but color coding is used for displaying the coefficient of variation for peak heights within a selected group of samples.
These visualization methods are particularly useful for quality control, because in the toolbox environment it is easy to go back to raw data and visually verify the findings.
Example: Metabolic profiling of plant secondary compounds in Catharanthus roseus
Studies of plant metabolites are a demanding area since plants produce large number of metabolites of high chemical diversity, many of which are unknown [15]. Plant secondary metabolites are produced as responses to changes in the environmental conditions. The biosynthetic pathways of secondary metabolites are largely unknown, and discovery driven 'omics' approaches promise to enhance our knowledge in this domain [16]. In order to illustrate the utility of the MZmine toolbox, we demonstrate it on metabolic profiling of cell cultures of the medicinal plant Catharanthus roseus. This plant has been extensively studied due to the presence of terpenoid indole alkaloids (TIA), several of which are in high demand for pharmaceutical use [17]. We focused on fraction containing most important secondary metabolites leading to TIA [see Additional file 1]. We profiled 20 samples, of which 10 were control strains and 10 were elicited strains. Elicitation induces the stress response and can therefore lead to production of secondary metabolites. The replicates are the same strain in parallel cultures corresponding to the same time point, so they can be considered as biological replicates. We also injected an internal standard compound vincamine (PubChem SID 390304).
Using MZmine toolbox with moving average filter (m/z = 0.3 window setting), recursive threshold peak detection (default settings), alignment (100s tolerance in retention time, otherwise default settings), gap-filling (60s tolerance in retention time), and normalization by total raw signal, we detected 2175 peaks. Representative total ion chromatograms from one elicited and one control sample are shown in Figure 3A. The log-ratio view for top 20% most intense peaks is shown in Figure 3B.
After exporting the processed data in tabular format, further analyses of the data matrix were performed in Matlab using PLS Toolbox (Eigenvector Research, Inc.) and with R Statistical Language. Principal components analysis [18] revealed clear differences between the elicited and control groups (Figure 4A). Using factor analysis (not shown), we found that the two of the main contributors to the clustering of the elicited group were ajmalicine (PubChem SID 153462) and tabersonine (PubChem SID 163306). The compounds were identified using our internal spectral library based on molecular weight and retention time. Their distribution within the elicited and control groups shows the compounds are significantly upregulated after elicitation (Figure 4B). Our findings are in line with recent report using the targeted approach [19].
Figure 4 Data analysis of Catharanthus roseus metabolite profile data, using the 2175 detected peaks as variables. (A) Principal components analysis shows differences between the elicited and control strains. Subsequent factor analysis revealed the clustering of the elicited group is largely due to the tabersonine and ajmalicine. (B) Comparison of intensity distribution between elicited and control groups for internal standard (Vincamine), Ajmalicine, and Tabersonine.
Discussion
System requirements and performance
MZmine is available under the GNU General Public License [see Additional file 2]. The toolbox needs Java Runtime Environment 5.0 or later (Sun Microsystems, Inc.) installed on the computer. Minimum system requirements for running the software are 2.4 GHz processor and 1 GB of memory. Also a high-resolution display mode or a dual monitor configuration is necessary for taking full advantage of the graphical user interface. As a reference, the total time from loading 50 NetCDF files (unprocessed continuous MS data acquired in full scan mode on Quattro Micro instrument) of size 100 Mbytes each to the export as a data matrix of approximately 10000 peaks took 40 minutes on a Dell Precision 650 workstation (Intel Xeon 3.06 GHz processor with 2 GB RAM), which is significantly less than the time of actual data acquisition by the LC/MS system (40 hours). The major performance bottleneck is the gap filling stage. Since many of the stages of the data processing can easily be parallelized, significant performance improvement could be gained by distributed computing. Future implementations of the software will include this capacity.
The software settings largely depend on the type of application, analytical method and instrument used for data collection. As part of the application development, it is advisable to experiment with different settings to optimize the performance.
Methods and applications
So far the MZmine has been primarily utilized in the domain of metabolomics, which included lipidomics and global metabolomics applications in biomedical domain, primary metabolite screening in microbes, and plant metabolomics. Specifically, the early prototype version of MZmine has been applied to lipidomic analyses in a recent study of PPARγ2 knock-out mouse model [20].
The methods currently available in the toolbox have been found sufficient in these applications. However, we will further focus especially on peak picking methods in the future, since this stage is the most crucial part of the data processing. In order to study the metabolic profiles with as many compounds as possible, every real peak should be found from raw LC/MS data files. On the other hand, false-positive peak detections complicate the statistical analyses and may prevent some interesting results to be found. In addition to these two issues, every single peak should be detected as exactly as possible in both m/z and retention time direction. This is necessary for determining peak area and height correctly, for successful alignment between samples, and for identification.
Currently it is difficult to perform comprehensive comparison of available methods for differential profiling of LC/MS data, since most of the methods are either proprietary or vendor-specific. However, repeatability studies on biological tissues have shown that the median CV is in the range of 18–23%, depending on normalization method (data not shown). This is consistent with published results [10,11]. The results on mixtures of internal standards were better, with CV <5%. The discrepancy in CV values is due to peak picking in complex biological mixtures, where many compounds are at the trace levels near the detection limit. This reaffirmed our belief that the stage of peak picking is the critical step in profiling of biological samples.
While the term quantitative analysis has recently been used to describe the methods of differential profiling [10,11], we believe true quantitative analysis would also require calculation of compound concentrations. None of the profiling toolboxes introduced so far have this ability. While ideally one would use isotope labelled standard and measure calibration curve for each compound, this is in practice impossible for complex biological mixtures where the compounds are of diverse chemical properties and many of them unknown. We believe our normalization method based on multiple internal standards is a step toward the ability to quantitate the compounds in the biological samples.
Current and future developments
We are currently developing a version enabling distributed computing and implementing a method for detection of natural isotope patterns. We are also going to extend the data import capability to mzXML and mzData formats and enable database connectivity.
On the algorithm side, in addition to improved peak picking, we are implementing two normalization methods, one based on multiplicative error model [6], and an enhanced version of multiple-standard method which takes into account information from compound identification. The initial application of the latter method will be developed for the lipid screening. One of the future goals is to enable automated handling of multiple spectra coming from single sample (i.e. MS and MS/MS or ESI+/MS and ESI-/MS). The latter, combined with database connectivity, will open the possibilities of automated identification of metabolites, as well as enable development of proteomics profiling applications utilizing MZmine.
Conclusion
We developed a platform independent software package for processing of LC/MS profile data. The software has already been tested and applied on a wide range of instruments and applications in domain metabolomics. Given its modular structure, the MZmine promises to be a powerful tool and test bench for development of new LC/MS data processing algorithms.
Availability and requirements
• Project name: MZmine LC/MS Toolbox
• Project home page:
• Operating system(s): Platform independent
• Programming language: Java
• Other requirements: Java Runtime Environment (JRE) 5.0 or higher
• Licence: GNU General Public License
List of abbreviations
MS: Mass spectrometry
XIC: Extracted ion chromatogram
TIC: Total ion chromatogram
LC/MS: Liquid chromatography – mass spectrometry
ESI(+/-): (Positive/negative) Electrospray ionization
APCI: Atmospheric pressure chemical ionization
QTof: Quadrupole – time of flight mass spectrometer
FTMS: Fourier transform mass spectrometer
CV: Coefficient of variance
m/z: Mass-to-charge ratio (m is molecular weight and z is charge of the ion)
GUI: Graphical user interface
API: Application programming interface
Authors' contributions
MK developed the software for the LC/MS Toolbox, developed the new algorithms for peak detection and alignment, normalization, and drafted the manuscript. MO initiated and supervised the study, designed the software toolbox and the experiment described in the paper and drafted the manuscript. Both authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Experimental methods for profiling Catharanthus roseus cell cultures. The file includes description of analytical methods for profiling of Catharanthus roseus cell cultures, which were used to demonstrate the utility of MZmine.
Click here for file
Additional File 2
MZmine Toolbox source code (version 0.42). The file includes the source code and license information of the MZmine Toolbox.
Click here for file
Acknowledgements
MK was funded by Academy of Finland SysBio Research Programme. MO's research was in part funded by Marie Curie International Reintegration Grant. We thank to Tuulikki-Seppänen Laakso for performing the LC/MS analyses used in this paper and to Heiko Rischer and Kirsi-Marja Oksman-Caldentey for providing the C. roseus material.
==== Refs
Ardrey RE Liquid chromatography - mass spectrometry: An introduction 2003 , John Wiley & Sons
de Hoffmann E Stroobant V Mass spectrometry: Principles and applications 2001 2. , John Wiley & Sons
Patterson SD Aebersold RH Proteomics: the first decade and beyond Nat Genet 2003 33 311 323 12610541 10.1038/ng1106
Goodacre R Vaidyanathan S Dunn WB Harrigan GG Kell DB Metabolomics by numbers: acquiring and understanding global metabolite data Trends Biotechnol 2004 22 245 252 15109811 10.1016/j.tibtech.2004.03.007
van der Greef J Davidov E Verheij E Vogels JTWE van der Heijden R Adourian AS Oresic M Marple EW Naylor S Harrigan GG and Goodacre R The role of metabolomics in systems biology: A new vision for drug discovery and development Metabolic profiling: Its role in biomarker discovery and gene function analysis 2003 Boston, Mass., Kluwer Academic Publishers 171 198
Oresic M Clish CB Davidov EJ Verheij E Vogels JTWE Havekes LM Neumann E Adourian A Naylor S Greef J Plasterer T Phenotype characterization using integrated gene transcript, protein and metabolite profiling Appl Bioinformatics 2004 3 205 217 15702951
Organization HP mzData
Davidov E Clish CB Oresic M Meys M Stochaj W Snell P Lavine G Londo TR Adourian A Zhang X Johnston M Morel N Marple EW Plasterer TN Neumann E Verheij E Vogels JTWE Havekes LM Greef J Naylor S Methods for the differential integrative omic analysis of plasma from a transgenic disease animal model OMICS A Journal of Integrative Biology 2004 8 267 288 15703476 10.1089/omi.2004.8.267
Vogels JTWE Tas AC Venekamp J Greef J Partial linear fit : a new NMR spectroscopy preprocessing tool for pattern recognition applications J Chemometrics 1996 10 425 438 10.1002/(SICI)1099-128X(199609)10:5/6<425::AID-CEM442>3.0.CO;2-S
Wang W Zhou H Lin H Roy S Shaler TA Hill LR Norton S Kumar P Anderle M Becker CH Quantification of proteins and metabolites by mass spectrometry without isotopic labeling or spiked standards Anal Chem 2003 75 4818 44826 14674459 10.1021/ac026468x
Radulovic D Jelveh S Ryu S Hamilton TG Foss E Mao Y Emili A Informatics platform for global proteomic profiling and biomarker discovery using liquid chromatography-tandem mass spectrometry Mol Cell Proteomics 2004 3 984 997 15269249 10.1074/mcp.M400061-MCP200
Lewis JP Neumann U Performance of Java versus C++
UniData NetCDF
Pedrioli PGA Eng JK Hubley R Vogelzang M Deutsch EW Raught B Pratt B Nilsson E Angeletti RH Apweiler R Cheung K Costello CE Hermjakob H Huang S Julian RK Kapp E McComb ME Oliver SG Omenn G Paton NW Simpson R Smith R Taylor CF Zhu W Aebersold R A common open representation of mass spectrometry data and its application to proteomics research Nat Biotechnol 2004 22 1459 1466 15529173 10.1038/nbt1031
Fiehn O Metabolomics - the link between genotyopes and phenotypes Plant Molecular Biology 2002 48 155 171 11860207 10.1023/A:1013713905833
Oksman-Caldentey KM Inze D Plant cell factories in the post-genomic era: new ways to produce designer secondary metabolites Trends Plant Sci 2004 9 433 440 15337493 10.1016/j.tplants.2004.07.006
Oresic M Rischer H Oksman-Caldentey KM Saito K, Willmitzer L and Dixon D Metabolomics of plant secondary compounds: profiling of Catharanthus cell cultures Plant metabolomics Biotechnology in Agriculture and Forrestry 2005 Heidelberg, Springer Verlag, in press
Jackson JE User's guide to principal components 1991 New York, NY, John Wiley & Sons
Lee-Parsons CW Ertürk S Tengtrakool J Enhancement of ajmalicine production in Catharanthus roseus cell cultures with methyl jasmonate is dependent on timing and dosage of elicitation Biotechnol Lett 2004 26 1595 1599 15604804 10.1023/B:BILE.0000045825.37395.94
Medina-Gomez G Virtue S Lelliott C Boiani R Campbell M Christodoulides C Perrin C Jimenez-Linan M Blount M Dixon J Zahn D Thresher RR Aparicio S Carlton M Colledge WH Kettunen MI Seppanen-Laakso T Sethi JK O'Rahilly S Brindle K Cinti S Oresic M Burcelin R Vidal-Puig A The link between nutritional status and insulin sensitivity is dependent on the adipocyte-specific Peroxisome Proliferator-Activated Receptor-{gamma}2 isoform Diabetes 2005 54 1706 1716 15919792
|
16026613
|
PMC1187873
|
CC BY
|
2021-01-04 16:27:26
|
no
|
BMC Bioinformatics. 2005 Jul 18; 6:179
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-179
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1801602661410.1186/1471-2105-6-180CommentaryCommunication and re-use of chemical information in bioscience Murray-Rust Peter [email protected] John BO [email protected] Henry S [email protected] Unilever Centre for Molecular Science Informatics, Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge. CB2 1EW, UK2 Department of Chemistry, Imperial College London, SW7 2AY, UK2005 18 7 2005 6 180 180 17 5 2005 18 7 2005 Copyright © 2005 Murray-Rust et al; licensee BioMed Central Ltd.2005Murray-Rust 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 current methods of publishing chemical information in bioscience articles are analysed. Using 3 papers as use-cases, it is shown that conventional methods using human procedures, including cut-and-paste are time-consuming and introduce errors. The meaning of chemical terms and the identity of compounds is often ambiguous. valuable experimental data such as spectra and computational results are almost always omitted. We describe an Open XML architecture at proof-of-concept which addresses these concerns. Compounds are identified through explicit connection tables or links to persistent Open resources such as PubChem. It is argued that if publishers adopt these tools and protocols, then the quality and quantity of chemical information available to bioscientists will increase and the authors, publishers and readers will find the process cost-effective.
==== Body
Introduction
In a previous article [1] we have argued the value of extracting the chemical information in bioscientific research, transforming it to XML and redisseminating it openly. The present article expands on the technical and cultural infrastructure required to support this. The technical aspects have been solved to proof-of-concept stage and we are starting to embark on experiments in the social domain. In this we thank BMC for inviting us to submit this and we present a model here which we believe could be attractive for bioscience publishers and their community.
We concentrate on the current publication of chemistry in bioscience. This includes:
1. mention of chemical compounds.
2. details of synthesis (in vivo and in vitro) of compounds.
3. proof of structure (spectra and analytical data).
4. Methods and reagents in bioscience bio-protocols
5. properties of compounds.
6. reactions and their properties, both in enzymes and enzyme-free systems.
This type of chemistry is very well understood and has a simple ontology which has not changed over decades[2]. Unlike much bioscience, where ontological tools are an essential part of reconciling the domain-dependent approaches, much chemistry has an implicitly agreed abstract description. The problems are primarily reconciling syntax and semantics. This is because chemists use abbreviated and lazy methods of communicating data, relying on trained readers to add information from the context. We have reviewed current problems of machine-understanding of chemistry[3] in a typical chemistry journal, many of which are perpetuated by the graphical orientation of conventional publishing houses. Here we take the view that a committed publishing house can create a cost-effective and human-tolerable system for authoring semantically correct chemistry in (bio)scientific documents.
We know from experience that Utopian visions do not sell themselves. The enormous and accepted value of the sequence and structures databases arose not from the demands of individual authors, but from wider communities of researchers, funders, and learned societies. Even now the deposition of protein structure data, without which journals will not generally accept a paper, is seen by some as a chore and at worst as the donation of information to competitors. Without that commitment and the resource, however, Structural Biology would not exist as a discipline. Here we present the following vision; that aggregated "small-molecule" chemical information, if deposited at publication, aggregated and disseminated, would be seen as worth paying the prices of inconvenience.
Generic infrastructure
For this proposal we make some assumptions about the evolving informatics environment:
• The costs of archiving and maintaining scientific information can be now very much lower than some of the more traditional approaches. There will always be areas (patents, safety, reference data) where intensive human effort is required in the curation of data and where comprehensiveness is critical. This argument will be strongly made by the current chemical secondary publishers who show no signs of changing their business model. However bioscience has shown that informatics research is willing to balance quantity and quality and accepts that data is always used under caveat emptor.
• Much data is now completely captured instrumentally and can, in principle, be transmitted without syntactic loss. Crystallography has shown that experimental data (in the CIF format) can be directly submitted to the publisher. Moreover with the development of expert programs it is possible to review the data by machine and that this leads to higher quality than before. The global aggregation of current small-molecule crystal structures, without any secondary curators or publishers, can now meet almost all the needs of the community.
• Most current publicly funded chemical data is never published; loss varies between 80% (crystallography) and 99.9%. Much of this is due to the lack of simple technical and cultural protocols, which we address later.
• The primary cost is human time. Storage and CPU costs are trivial (for our domain). We show how the measures here, if adopted, would save all members of the community considerable time. They would also lead to the creation of greatly enhanced information resources.
• A variety of repositories will become available. In some communities (e.g. Physics and Computer Science) self-archiving of (p)reprints is universal but in others it is rare. Early adopters of Institutional repositories (IRs) are starting to mandate that the output of publicly funded scholarship is reposited, and we infer that, perhaps with sharing schemes, this will become quasi-mandatory. There is potential conflict with publishers' licenses, which we address below.
• There will be sufficient publishers in bioscience who are attracted by approaches like ours, and that this will create a critical mass. The benefits will be interoperable approaches to authoring (at present the technical requirements of some publishers are grotesque, i.e. hardcopy, strange formats, etc.).
• Openness. Our approach does not require Open Access, but does require that chemical data are Open.
• Willingness for bioscientists to take a lead in chemical informatics. Chemical information sources are manually aggregated and curated secondary publications whose philosophy has barely altered over 120 years. There is a large quasi-monopoly of a small number of large organisations who have no interest or inclination in changing their business models or adopting the vision of the Semantic Web. These new technologies, however, are ideally suited to our model and require only modest investment.
• Open or cheap tools for publishing structured documents (in XML) which can be customised for different domains. As XML becomes the universal technology for publication and interoperability, the community has access to them and will become trained in their use. As Open source components become more widespread it becomes possible to envisage scientific authoring tools which meet many requirements of the publication process.
We look to bioscience to take a lead in helping realise the following vision. On the positive we now believe that there are already enough Open tools and Open resources which with communal will among bioscience authors and publishers can make the vision attractive and cost-effective.
Automatic capture of chemical information
Much chemical data is largely context-free in that it can be understood and recreated independently of the location or motivation. The primary data model is over 120 years old and was developed by Beilstein in the 19th century and comprises three components: the chemical compound, its properties and citations. A pure compound is described by an immutable structural formula and has precisely reproducible properties. There are qualifications (e.g. some properties may depend on the precise crystalline form) but it has served as the basis of a multimillion chemical information market, with the compound at the centre. Current thinking asserts that the biological action of a compound is, in principle, reproducible and predictable if the system is carefully enough replicated and the components understood. This is the central dogma of the chemically-based pharmaceutical industry.
Chemistry has a tradition of quality through properties and analysis, so every new compound (and many resynthesised ones) mentioned in the literature must be accompanied by measurements of properties to justify identity and purity. These facts are available, in text form, in the primary literature in which over a million new compounds are published annually. Because structure predicts properties, and because drug discovery is so difficult, the pharmaceutical industry tests many compounds for biological activity. It is therefore the primary financial engine for the chemical information industry.
The components
Techniques for managing items 1–5 listed above such as aggregating chemical compounds, properties and for searching the results, are very well understood and can be easily made nearly automatic. Most of the information of benefit to the community exists on the authors' computers in machine-processable form. It can be automatically converted into fine-grained XML[4] with almost no loss. The chemist has electronic copies of molecular structures, spectra and properties whose semantics are extremely well understood and where a simple technical protocol for conversion to XML and hence publication can be created. To support this part of the data publication process we have created the XML-based Chemical Markup Language (CML)[5]. The primary information components (all of which are common and well understood) are:
• Molecular structure
A compound is described by a compositional formula (e.g. CH3OH for methanol) and a graphical structural formula ("2D diagram"). These descriptions are required in bioscience publications for new compounds or where scientific arguments are based on details of chemical structure. There are a few widely used standalone tools (mainly commercial) for drawing structures and calculating certain properties. They output a variety of machine-processable formats (MDL Molfiles, ChemDraw CDX files, and increasingly CML). The main challenge is that the output is designed for the sighted human reader and that semantics may be implicit, discussed below. The Open Source community is creating tools (e.g. JChempaint[6]) that will be valuable in authoring publications.
• Chemical entities and names
Many compounds have no explicit structures and are mentioned by names or identifiers. Where these relate to specific compounds (rather than generic such as "phenols") it is valuable to link them to a precise identification.
• Spectra
Many traditional formats (JCAMP and some manufacturers) are satisfactorily machine-processable, and we expect the XML-based AnIML[7] to be widely adopted by manufacturers.
• Crystal Structures
Relatively few small-molecule crystal structures are reported in bioscience publications, but when they are we have a workflow-driven system that extracts the data automatically and reposits it
• Molecular properties
These are required as proof-of-synthesis and use a small dictionary of properties[8]. Their publication is highly ritualised and we expect that a publisher-wide template for the submission of this information would be straightforward to compile and welcomed by many authors.
Identifying compounds
The identification of chemical entities is the most valuable contribution that an author can make. In most cases s/he (as, say, the purchaser or creator of the materials) is the best judge of what was used. It is more considerably more difficult to identify compounds after publication as we show below.
We list possible methods of publishing the identity of compounds in machine-understandable form:
• Connection table
This is the most powerful method and we urge that every report of a chemical synthesis be accompanied by a connection table. It already exists in the authors' laboratory (in MOL, Chemdraw, SMILES, and increasingly as CML). It is rare that a pure molecule in the bioscience literature cannot be represented in this way. This is the single most important recommendation in this manuscript.
• Chemical structure diagram
This is a useful adjunct to a connection table (and some of the formats combine the two). Very occasionally (e.g. for catenanes, helicenes) a diagram is essential, but it should never be used instead of a connection table.
• InChI
For most compounds of bioscientific interest with known structure it is possible to generate a unique identifier using the new InChI (International Chemical Identifier)[9] from IUPAC. This has major advantages over non-semantic identifiers and Closed proprietary canonical identifiers such as SMILES[10]. In principle an InChI not only uniquely identifies the substance but also contains all the essential structural information. InChI in its current version (1.0) (and often other canonicalization schemes as well) has some significant limitations for systems with metal ions and ionic compounds, a situation which apparently will be addressed in a future InChI revision. Simple molecules such as e.g. cis-Platin are currently included in the handling, although the stereochemical aspects are not currently captured. Handling such aspects of e.g. metal-based drugs must clearly be a high priority in the development of InChI.
• Semantically free identifiers
These are provided by authorities (e.g. Chemical Abstracts, RTECS, PubChem, etc.). To be useful they should have an Open mechanism for their resolution (e.g. in PubChem), but this is often expressly forbidden. Thus Chemical Abstracts[11] forbids the public exposure of more than 0.1% of its content. Unless persistent Open machine-friendly resolution is available we deprecate the use of authority-controlled identifiers as unique IDs in primary publications. There are very few cases (e.g. zeolites) where identifiers are the best means of identification.
• Trivial ("Common") names
The structures of many compounds ("aspirin", "testosterone", "glycine"...) can only be found through lookup. In the past these names have been controlled in Closed collections but there are now an increasing number of Open lexicons of names:structures. The NCI led the way (220,000 names for ca 50,000 structures) and PubChem [12] has continued to develop this. If commercial suppliers make their catalogs Open then most of the common chemical names in scientific discourse can be automatically linked to connection tables.
• Systematic chemical names
Until now this has been a common means of transmitting chemical identity, but it now serves little purpose, although It may be required legally, e.g. for patents. Most chemists would prefer a structural diagram to a systematic name, and many regard name generation as a tedious chore. In principle IUPAC[2] chemical names obey a context-free grammar and there are complex rules for canonicalization. In practice most authors use a variety of shortcuts. This means that most compounds are reported with a variety of near synonyms (thus "2-hydroxy-toluene", "2-methyl-phenol" are semi-systematic variants for "1-hydroxy-2-methyl-benzene"). Free-text searching on chemical names has almost always low precision and often low recall. It is a common error to assume that deterministic grammars can parse any chemical name; in practice typographical errors, elisions and trivial fragments lower precision considerably. Some commercial tools are available but their algorithms are closed and little research has been done on their precision and recall. We suspect that they are composed of lexicons and heuristics but have no information on how they are maintained, especially in light of revisions of naming conventions.
Issues with chemical names
Chemical names can be used with more or less specificity. Thus "1,4-dichlorobenzene" is unambiguous in any context. However there are several areas where more generic language is used. This can arise because:
• The name refers to a class of compounds, whose members have similar structures and/or properties: "steroids", "amino acids", "monosaccharides", "polychlorinated biphenyls".
• The substance is a mixture of compounds: "60–80 petrol", "xylenes", "the phospho-inositols"
• The substance has not been fully identified: "the estradiol monobenzoate was ..." (there are two possibilities)
• The stereochemistry is ambiguous. The possibilities (in decreasing order of merit) include:
Stereochemistry is known and reported.
Stereochemistry is unknown and reported as such.
Stereochemistry is partially known and reported as such.
Stereochemistry is not reported but is unknown
Stereochemistry is not reported but is known
Stereochemistry is partially reported but is completely known
Stereochemistry is reported and is wrong
"Glutamic acid" is an example of ambiguity through unspecified stereochemistry. Thus PubChem lists the three isomers (Table 1).
Table 1 Isomers of Glutamic Acid
Name(s)
611 glutamic acid
33032 L-glutamic acid
23327 D-glutamic acid
A structure without any stereo information is more valuable than one with partial information of unknown quality. The InChI [9] is an extremely powerful tool here. We have recently shown that "staurosporine" reported in publications and suppliers catalogs contains many instances of incorrect stereochemistry, and some partially correct and incorrect. Given that this is a single substance, its structure and absolute configuration has been known for a considerable time there is no reason for using any structure other than PubChem CID: 44259
• Ionization
Protons are labile in aqueous systems and (for example) aminoacids can have several ionization states. The importance of ionization details varies;
"Acetic acid (0.1 M) was added..." [ionization state irrelevant]
Acetic acid forms a hydrogen-bonded dimer in the crystal. [single species, determinable by crystallography]
We computed the structure of glycine zwitterion acid (NH3+CH2CO2-) in the gas-phase [single species, distinct from NH2CH2CO2H]
"Glutamic acid is the most common excitatory neurotransmitter in the CNS" [irrelevant in macroscopic experiment, critical in modelling action at receptor]
• Tautomerism
Many neutral compounds, particularly with heteroatoms have mobile hydrogens in solution. Thus 2-hy-droxy pyridine (see Figure 1) exists as both forms with very rapid interchange. PubChem (as with many other systems) lists them as the same compound (CID8871) and gives the many synonyms including "2(1H)-Pyridinone" and "2-HYDROXY-PYRIDINE". InChI[9] has an extensive system for detecting tautomerism in compounds with heteroatoms, but does not yet address carbon compounds (e.g. CH2=CHOH as a tautomer of ethanal (acetaldehyde, CH3-CH=O).
Figure 1 Tautomers of Hydroxypyridine
• Imprecise or polysystemic use
This often occurs when a chemical entity is incorporated into a larger system
"This polysaccharide has a high mannose content" means "... contains many mannosyl fragments ..."
"HIV protease has a catalytic aspartic acid..." means "... an aspartyl residue ..."
The preceeding discussion shows how ambiguity and loss of information can occur if structured procedures are not followed. The examples in Table 2 show some suggested approaches to markup which can re-capture much of the information loss described above. The last example references a generic name, monosaccharide, in the IUPAC guide[2] to organic nomenclature with a suggested use of identifiers.
Table 2 Examples of approaches to chemical Identification.
Prose description More precise suggested naming using the CML[5] approach Type of information
Acetaldehyde has a general narcotic action <p><cml:molecule> <cml:identifier convention="iupac:inchi">1/C2H4O/c1-2-3/h2H,1H3</cml:identifier> <cml:identifier convention="pubchem:CID">177</cml:identifier> </cml:molecule> has a ...</p> precise, redundant
Benzo(a)pyrene is a potent mutagen and carcinogen <p><cml:molecule><cml:identifier convention="pubchem:CID">2336</cm l:identifier></cml:molecule> is a ...</p> precise
glycine (1 mmol) was added ... <p><cml:molecule title="glycine"><cml:identifier convention="iupac:inchi">1/C2H5NO 2/c3-1-2(4)5/h1,3H2,(H,4,5)</cml: identifier></cml:molecule> is a ...</p> hydrogens mobile
calculations on glycine zwitterion... <p><cml:molecule title="g><cml:identifier convention="pubchem:CID">InChI=1/C2H5NO2/c3-1-2(4)5/h1H3,3H2</cml: identifier></cml:molecule> is a ...</p> hydrogens precise
... a monosaccharide transporter... <p>a <cml:molecule title="monosaccharide"><cml:ident ifier convention="iupac:carbohydrate">2 -Carb-2</cml:identifier></cml:mol ecule> transporter ...</p> Data
Case studies
In this second section, we take 3 articles from BMC publications and show the success and problems of extracting chemistry in machine-understandable form. These have been randomly selected and do not necessarily reflect the average quality of BMC publications. We note that in our other studies of chemical text very few publications were error-free.
Case study 1: Identification of compounds in discourse and reagents in methods[13]
The abstract is typical of the discourse:
Background
Recent studies indicate that the G protein-coupled receptor (GPCR) signaling machinery can serve as a direct target of reactive oxygen species, including nitric oxide (NO) and S-nitrosothiols (RSNOs). To gain a broader view into the way that receptor-dependent G protein activation – an early step in signal transduction – might be affected by RSNOs, we have studied several receptors coupling to the Gi family of G proteins in their native cellular environment using the powerful functional approach of [35S]GTPgammaS autoradiography with brain cryostat sections in combination with classical G protein activation assays.
Results
We demonstrate that RSNOs, like S-nitrosoglutathione (GSNO) and S-nitrosocysteine (CysNO), can modulate GPCR signaling via reversible, thiol-sensitive mechanisms probably involving S-nitrosylation. RSNOs are capable of very targeted regulation, as they potentiate the signaling of some receptors (exemplified by the M2/M4 muscarinic cholinergic receptors), inhibit others (P2Y12 purinergic, LPA1lysophosphatidic acid, and cannabinoid CB1 receptors), but may only marginally affect signaling of others, such as adenosine A1, μ-opioid, and opiate related receptors. Amplification of M2/M4 muscarinic responses is explained by an accelerated rate of guanine nucleotide exchange, as well as an increased number of high-affinity [35S]GTP?S binding sites available for the agonist-activated receptor. GSNO amplified human M4 receptor signaling also under heterologous expression in CHO cells, but the effect diminished with increasing constitutive receptor activity. RSNOs markedly inhibited P2Y12 receptor signaling in native tissues (rat brain and human platelets), but failed to affect human P2Y12 receptor signaling under heterologous expression in CHO cells, indicating that the native cellular signaling partners, rather than the P2Y12 receptor protein, act as a molecular target for this action.
Conclusion
These in vitro studies show for the first time in a broader general context that RSNOs are capable of modulating GPCR signaling in a reversible and highly receptor-specific manner. Given that the enzymatic machinery responsible for endogenous NO production is located in close proximity with the GPCR signaling complex, especially with that for several receptors whose signaling is shown here to be modulated by exogenous RSNOs, our data suggest that GPCR signaling in vivo is likely to be subject to substantial, and highly receptor-specific modulation by NO-derived RSNOs.
The above contains reference to a considerable numbers of individual compounds. The authors helpfully publish a table of abbreviations to assist in the compound identification process (Figure 2). Using this as our data, we have attempted to identify (Table 3) the "small-molecules" mentioned in the discourse. Using PubChem and occasional suppliers catalogs, the elapsed real time was about 1 hour. It can be seen that of 19 molecules, 15 were identified without problems or error, 2 were not (CysNOGly and Glu-CysNO) and 2 required additional expertise by the reader. We estimate that it would take an author the same amount of time to add PubChem IDs for novel compounds and much less time if they were in common use in their laboratory.
Figure 2 Abbreviations used in reference 13.
Table 3 Identification of Small-molecules noted in Ref. 14
abbreviation author name PubChem ID Notes
Not found directly in PubChem. Located in supplier's catalog. Synonym from that found in PubChem
2MeSADP 2-methylthio-ADP [121990]
5-HT 5-hydroxytryptamine 5202
CCh carbacholine 521353
CP-55940 (-)-3-[2-hydroxy-4-(1,1-dimeth ylheptyl)-phenyl]-4-[3-hydroxyp ropyl]cyclohexan-1-ol 104895 IUPAC: 5-(1,1-dimethylheptyl)-2-[5-hydroxy-2-(3-hydroxyprop yl)cyclohexyl]-phenol
CysNO S-nitrosocysteine 39933
CysNOGly S-nitroso-cysteinyl-glycine Text search on PubChem found wrong compound. Not found in major supplier
DAMGO [D-Ala2, N-Me-Phe4, Gly5-ol]-enkephalin 104742
DPCPX 8-cyclopentyl-1,3-dipropylxanthi ne 1320
Glu-CysNO L-?-glutamyl-S-nitrosocysteine Identity unresolved
GSH glutathione 745
GSNO S-nitrosoglutathione 104858
LPA lysophosphatidic acid 3987
NA noradrenaline 951 PubChem CID covers both racemic and d-enantiomer
NO nitric oxide 84878 PubChem also lists 945 (with incorrect formula HNO) as nitric oxide
NOBF4 nitrosodium tetrafluoroborate 151929 Paper has a typographical error for "nitrosonium". Structure in PubChem is wrong (formula should be NO+BF4-, not H2NO+.BF4-)
SNAP S-nitroso-N-acetyl-D,L-penicill amine 5231 PubChem does not list stereochemistry
RSNO S-nitrosothiol Appears to be a generic compound (R-S-N=O)
SNP sodium nitroprusside 26256
It is only a little additional effort to convert each molecule to a more formal description expressed in e.g. CML[5] and which can carry not only an atom connection table and the corresponding InChI identifer, but also molecule "meta-data" describing the provenance of the information:
<cml:molecule xmlns:cml="" title="carbacholine">
<cml:metadataList title="generated automatically from Openbabel">
<cml:metadata name="dc:creator" content="OpenBabel version 1-100.1"/>
<cml:metadata name="dc:description" content="Conversion of legacy filetype to CML"/>
<cml:metadata name="dc:identifier" content="InChI"/>
<cml:metadata name="dc:content"/>
<cml:metadata name="dc:rights" content="open"/>
<cml:metadata name="dc:type" content="chemistry"/>
<cml:metadata name="dc:contributor" content="rzepa"/>
<cml:metadata name="dc:creator" content="Openbabel V1-100.1"/>
<cml:metadata name="dc:date" content="Tue May 17 12:02:50 BST 2005"/>
<cml:metadata name="cmlm:structure" content="yes"/>
</cml:metadataList>
<cml:identifier convention="iupac:inchi">InChI=1/C6H14N2O2.ClH/c1-8(2,3)4-5-10-6(7)9;/h4-5H2,1-3H3,(H-,7,9);1H</cml:identifier>
<cml:atomArray atomID="a1 a2 a3 a4 a5 a6 a7 a8 a9 a10 a11 a12 a13"
elementType="N C C O C O N C C C H H Cl"
formalCharge="1 0 0 0 0 0 0 0 0 0 0 0 -1"
x2="-1.892900 -1.178500 -0.464000 0.250500 0.964900 0.964900 1.761800
-2.305400 -2.476300 -1.480400 2.174300 2.476300 -1.921800"
y2="0.415300 0.827800 0.415300 0.827800 0.415300 -0.409700 0.628800 1.129800
-0.168000 -0.299200 1.343300 0.216300 -1.343300"/>
<cml:bondArray atomRef1="a1 a1 a1 a1 a2 a3 a4 a5 a5 a7 a7"
atomRef2="a2 a8 a9 a10 a3 a4 a5 a6 a7 a11 a12"
order="1 1 1 1 1 1 1 2 1 1 1"/>
</cml:molecule>
Such molecular datuments can be embedded in any XML-based document in a manner which can if needed survive e.g. journal production processes, and where the molecular information can be extracted and re-used at any stage.
Case study 2: Identity and properties of synthesised compounds[14]
Our critique of the chemistry requires context, given by the abstract:
Abstract background
Kynureninase is a key enzyme on the kynurenine pathway of tryptophan metabolism. One of the end products of the pathway is the neurotoxin quinolinic acid which appears to be responsible for neuronal cell death in a number of important neurological diseases. This makes kynureninase a possible therapeutic target for diseases such as Huntington's, Alzheimer's and AIDS related dementia, and the development of potent inhibitors an important research aim.
Results
Two new kynurenine analogues, 3-hydroxydesaminokynurenine and 3- methoxydesami-nokynurenine, were synthesised as inhibitors of kynureninase and tested on the tryptophan-induced bacterial enzyme from Pseudomonas fluorescens, the recombinant human enzyme and the rat hepatic enzyme. They were found to be mixed inhibitors of all three enzymes displaying both competitive and non competitive inhibition. The 3-hydroxy derivative gave low Ki values of 5, 40 and 100 nM respectively. [...]
Conclusion
For kynureninase from all three species the 2-amino group was found to be crucial for activity whilst the 3-hydroxyl group played a fundamental role in binding at the active site presumably via hydrogen bonding. The potency of the various inhibitors was found to be species specific. The 3-hydroxylated inhibitor had a greater affinity for the human enzyme, consistent with its specificity for 3-hydroxykynurenine as substrate, whilst the methoxylated version yielded no significant difference between bacterial and human kynureninase. [...]
We note that "quinolinic acid" has four mentions in the text, but its formula is not given. We took roughly three minutes to identify CID1066 in PubChem, with the additional useful information (from Medline/MeSH):
A metabolite of tryptophan with a possible role in neurodegenerative disorders. Elevated CSF levels of quinolinic acid are correlated with the severity of neuropsychological deficits in patients who have AIDS
The name "3-hydroxydesaminokynurenine" [the synthesized compound (4) presents a more serious problem. Although the structure is given in a diagram, the stereogenic centre is not marked. It would be a reasonable assumption that "kynurenine" refers to a natural product which is only found in one enantiomeric form and "desamino" was also chiral. Careful reading (requiring chemical expertise) showed that the authors had probably synthesised a racemic mixture, since they started with achiral compounds and did not report chiral reagents or a resolution step. The compound should have been reported as (R/S)-3-hydroxydesaminokynurenine or (much better) as the IUPAC-like name "IUPAC Name: (R/S) 2-amino-4-(3-hy-droxy-phenyl)-4-oxo-butanoic acid". Indeed many referees and editors would have insisted on this specification. In the event, as we show below, this is not the reported compound!
The tools we are proposing would immediately have queried both these concerns at time of authoring and, had they been available to the technical editor would have produced a more useful and more easily readable paper.
The publication of the synthesis or re-synthesis of compounds must be accompanied by analytical and property data to prove purity and identitity. The ritualistic presentation shown below (Figure 3) as copied from the manuscript is entirely typical of most chemical publications. Note that it is visually challenging to read and this is entirely due to the publisher's requirements of using a system designed to save paper rather than communicate useful information.
Figure 3 A linear text-based description of experimental detail and data taken from Ref. 14.
For each compound this compressed information is (manually) created from some or all of:
1. An elemental analysis (probably in machine-understandable form)
2. A calculated composition for the compound (machine understandable)
3. An infrared spectrum (machine understandable)
4. A 1H NMR spectrum (machine understandable)
5. A 13C NMR spectrum (machine understandable)
6. A low resolution mass spectrum (machine understandable)
7. A high resolution mass spectrum (machine understandable)
For the publication, the authors have to measure peak heights from the spectrum (possibly with a ruler), and transcribe them to a Word or PDF format, probably by typing the values or cut-n-pasting them. We have developed an Open Source robot (OSCAR)[8] which can understand this data if it is syntactically correct, and the result is shown in Figure 4:
Figure 4 OSCAR output from the text-based description in Ref 14.
The coloured parts are those that adhere to the publication guidelines. We found 7 changes that had to be made to the punctuation (missing punctuation, syntactic variation is common in many chemical papers). OSCAR can then understand and check the data. For compound [4] OSCAR produces the errror message:
There are fewer H atoms by NMR integration (7) than there are by elemental analysis (12)
This is acceptable because there are exchangeable groups. However OSCAR also gives the error flag:
There are more C-NMR environments (11) than there are C atoms from elemental analysis (10).
as it found the string "114.47 120.78". We also do not understand this and it may be an error (or it could be a solvent peak or other impurity). OSCAR also had problems interpreting the chemical formula: "C11H14NO4" which in fact turns out to be a charged species. In fact the compounds are poorly identified. They appear to be not the aminoacids "3-Hydroxydesami-nokynurenine (4)" and "3-Methoxydesaminokynurenine (5)" but their hydrochloride salts. This is not a trivial error; the melting points and infrared spectra of the parents and their salts will be significantly different and would cause errors if transcribed unthinkingly from the paper.
Even with OSCAR it took one of us ca 45 minutes to make sure that the above analysis was correct. From several anecdotal conversations with typical authors we estimate that it took about 2 hours to prepare this part of the submission; a thorough reviewer might take 0.5 hour to decipher it. All of this is unnecessary if the original connection tables, spectra and analytical data were made available in uncorrupted form. As it is, much of the original data is lost; using the reported peaks OSCAR does its best to recreate what the spectrum might have looked like (Figure 5). Precise peak shapes and traces of impurities are lost in this representation.
Figure 5 OSCAR generated spectrum of analytical information reported in Ref 14.
Case study 3. Identity of compounds and preservation of calculations[15]
Here too a number of small-molecules are reported without formulae;
Background [...] Phenols and anilines are generally recognized as substrates of the heme peroxidases (donor: H2O2 oxidoreductases EC 1.11.17). The peroxidases catalyze oxidation of the substrates by hydrogen peroxide or alkyl peroxides, usually but not always, via free-radical intermediates [1,2]. Nonphenolic compounds, such as indole-3-acetic acid, phenylenediamines, ferrocenes, phenothiaz-ines, phenoxazines, have also been investigated as peroxidase substrates [2-5]. Steady-state kinetics of peroxidase action has been described as a ping-pong scheme with compound I and compound II formation [1].
This paper also has issues with the identity of compounds (Figure 6). This is again a visually unacceptable format dictated by the prevailing business model of chemical publishing. Note "Napthyl" is misspelt, presumably because it has been (mis)typed by the authors, which would give unnecessary problems to chemical text-mining robots. Worse, the identity of AHA5 is genuinely unclear, in that the connection could be to either of the phenyl groups in the fragment: "Ph-C(O)N(-OH)-Ph". BHA (also described elsewhere by "benzhydroxamic acid") has no structural or compositional formula. Worse, BHA in the PDB ligand collection refers to 2-hydroxy-4-amino-benzoic acid (a completely different compound); "benzhydroxamic acid" has code BHO.
Figure 6 Structure diagram reported in Ref 15.
Another section of this article describes various computational modelling techniques applied to these molecules; here we can assume that the authors had precise coordinates for all the computed species available at the end of the research, although none of this data is actually made available via the final published article. Some of this data is used to drive a docking program, which itself implies a protocol used to specify various run-time parameters. Some of these are declared in the article, many probably default to values set internally within the program. There are also ambiguities in the declared computational protocol:
The optimized geometry of molecules was used for energies and charges calculations with a 6-31G basis set using RHF and B3PW91 (Density Functional Theory).
Here, the RHF and the B3PW91 protocols are mutually exclusive; either one or the other could have been used, but not in combination. Mapping either protocol to e.g. the appropriate input for the program package used can also be a challenge for anyone not totally familiar with the program; program manuals are still designed largely for human rather than machine use. Such ambiguities, and lack of data, make repetition of the modelling more difficult for others.
A proposed infrastructure
It should now be clear that the current system of communicating chemistry (which is common to all publishers and all disciplines) is inefficient, costly, lossy, and of questionable quality. We present a new XML-based approach which we show:
• takes less time
• conveys more information
• is easier to read
• allows published data to be aggregated and re-used
We note that when starting to draft a publication the author already has
• free text (A) (probably in handwritten form)
• properties (B) (probably handwritten form)
• spectra (C) (probably in digital form)
• molecules (D) (probably in MOL or ChemDraw files)
Electronic lab notebook technology is not well advanced in chemistry; our architecture would provide a good method for preserving conventional data. It looks as shown in Figure 7 (blue = XML):
Figure 7 Data-flow illustrating the use of XML.
The author would then use a tool which can manage structured XML documents and provide normal textual support (spellchecks, etc.). There are 4 additional tools required to support chemical information:
• A. Chemical lexical tool(AA) which can (a) parse free text(A) for possible compound names (b) look them up or (c) parse them to create connection table and (d) insert a reference (AX) to the lexicon in the text, e.g.:
... When foobarone is broken down, the presence of indole can be detected ...
might be marked up as
... When <cml:molecule name="foobarone" dictRef="natprod:foobarone"/> is broken down, the presence of <cml:molecule>
<identifier convention="iupac:inchi" title="indole">1/C8H7N/c1-2-4-8-7(3-1)5-6-9-8/h1-6H,9H</identifier>
</cml:molecule>
indole can be detected ...
• B. A controlled vocabulary (BB) of property types is used in a template to capture properties (B) and create a CML table (BX), e.g.
yield(93%), M.Pt. 273-275°C
becomes
<cml:list>
<cml:property dictRef="cml:yield">
<cml:scalar units="cml:percent">93%lt;/cml:scalar>
</cml:property>
<cml:property dictRef="cml:mpt">
<cml:scalar units="cml:celsius" minValue="273" maxvalue="275"/>
</cml:property>
</cml:list>
• C. Spectra in legacy format (C) are automatically converted to CMLSpect or AniML (CX).
• D. Molecules created in a conventional editor are either emitted in CML (DX) or automatically converted from legacy (D) .
The result is a single structured XML "datument"[16] containing fine-grained markup of facts (molecules, measurements, properties, chemical names). This datument can be used to create derivatives such as the "full-text" or the "supplemental data". The complete datument (if Open) or the "data" if not is then reposited (XX) where it can be harvested. New compounds with their names are fed back into the lexicon and all compound/property data is available for datamining and computational re-use (e.g. for further in silico prediction. A human or robot reader has access to the same lexicons and dictionaries as the author so that the semantics and ontology of authoring are the same as those of reading (and of preservation).
Metadata and rights
The social aspects of metadata and rights were addressed in (1). To meet these we place special emphasis on the XML and its metadata. Fine-grained XML (e.g. <scalar>...</scalar> or <molecule>...</molecule> represents facts which can be identified as Open and not the property of the publisher. Hyperlinks and structure for semantics (e.g. identification of compounds in PubChem) are also Open. Tools such as XSLT can then extract the factual, non-copyrightable information with little technical problem. Rights should be explicitly marked up. If the publisher supports Open Access [17] and also Open Data then it is valuable to label the appropriate components with Open licenses, such as the RDF metadata provided by Creative Commons. It is also possible to preserve authors' moral rights and provenance of data re-used within the paper (e.g. spectra of molecules or coordinates of protein structures).
Realising the vision
The transition to this architecture will have a cost, so short term-benefits are particularly attractive. Moreover most of the parties are not used to a communal approach (pressures are normally per-institution and per-publisher).
Costs
• Time lost in understanding and changing to a new system.
• New tools might cost money, or have to come from non-centralised budgets
• Only supported by a subset of publishers
• Communal deposition of data goes against the secretive culture
• Publishers have to invest in new system and react to community expectations
Benefits
• Open Access and Open Data[16]
• Greater quality in publications
• Data in theses and papers can be interchanged
• Greater readability, usability and innovation in publications
• Repository provides complete data record for laboratory, institution and world
• Modern informatics tools allow new types of search and aggregation
• Considerable time-savings during publication
• More efficient publishing reduces author frustration and time to publication
• and most importantly the arrival of the Scientific Semantic Web, allowing robots to read and take action on publications.
The benefits should also be clear for most individuals and organisations:
• funders can ensure a much higher of dissemination of funded data will be available
• institutions mandate a greater proportion of funded science published; better visibility and preservation
• researchers spend less time on unproductive operations
• reviewers have easier access to background ontology of data in documents
• editors get greater automation
• publishers are relieved of need to archive supplemental data
• readers have information prosthetics for easier and faster reading
• librarians develop one of the best early repository applications in the digital age
Potential
Because the chemical information is structured we now have a biocheminformatics cycle (this term – with spelling as here -is in modest use. We suggest its adoption to describe the management of chemical information in biosciences and not just in biochemistry) where, for the first time, large scale robotic data analysis can take place (Figure 8).
Figure 8 A Biocheminformatics Cycle.
The data in the research (laboratory, in silico, or both) are published in a lossless manner. Molecules and their properties have unique identifiers as described above and can be integrated into mainstream bioinformatics in the same manner as collections such as PubChem, MSDChem (at EBI), KEGG, etc. They will bring the added value of consistently captured property data and spectra. We also expect that many in silico properties will then be systematically added.
Compliance and adoption
The current dissemination of data through publishers is largely unsatisfactory. Some publishers, such as the International Union of Crystallography, see it as core business and others in the biosciences agree to link to international databanks. Many publishers allow the deposition of factual "supplementary data" but our experience with mainstream chemical publishers is that it is an unwelcome chore, poorly resourced and maintained, and with virtually no quality control or curation over the content. We believe that many publishers would welcome a model where they were no longer involved in data repositing.
The introduction of structured authoring tools will help this process considerably. Templates can be created for the chemical components described above and where the information exists in XML (connection tables, spectra, properties) it should be as easy as for committed authors as using a semantically void tool (e.g. Word). Where information needs to be converted from legacy formats we have created Open Web Services which publishers (and authors) may clone and customise. The main technical challenge will be the management of chemical names in free text.
Conclusions and the future
The analysis presented here introduces the basic concepts of chemistry in bioinformatics. Many areas remain to be addressed; we briefly describe two below which have immediate application.
Reactions
Chemical reactions are very patchily abstracted from the literature and the products are almost always closed. The motivation for the primary publication of reactions in bioscience includes:
1. record of synthesis of compound and proof thereof
2. record of an experimental protocol (e.g. biotinylation)
3. record of a biochemical reaction, including xenobiotic processes
4. description of systems biochemistry (coupled reaction pathways)
5. understanding of an enzyme mechanism
CMLReact (an extension of CML) has been created[18] to support these catagories of reaction. Items 1-2 require identical support as in mainstream chemistry (e.g. in journals supporting organic synthesis). Item 3 can be supported by CMLReact though there is little current experience. Item 4 is supported by SBML[19] and efforts such as BioPAX[20] (in which CML is a tool). Item 5 is particularly exciting and exemplified by our MACiE database where 150+ enzymes with 3D structures and proposed mechanisms have been collected[21]. Currently the abstraction is manual and expensive, but if the ideas in the current paper are implemented we shall present an extension whereby mechanisms can be relatively cheaply captured at source. This would be a major new resource in bioinformatics.
Evaluation metrics
The primary motivation for a publication, of course, is citability and the technology we describe raises the fear among chemists that the data in it might actually be read, analysed and re-used. However it also raises the vision of changing the "citation economy" (which values market perception) to a "reuse economy" where the data in an article (or as we prefer, a "datument") are valued by how often they are re-used.
Supplementary Material
Additional File 1
Click here for file
==== Refs
Murray-Rust P Mitchell JBO Rzepa HS BMC Bioinformatics 2005 6 141 15941476 10.1186/1471-2105-6-141
The International Union of Pure and Applied Chemistry
Murray-Rust P Rzepa HS Tyrell S Zhang Y Org Biomol Chem 2004 2 3192 3203 15534696 10.1039/b410732b
World Wide Web Consortium
Murray-Rust P Rzepa HS J Chem Inf Comp Sci 2003 43 757 772 10.1021/ci0256541
Krause S Willighagen E Steinbeck C Molecules 2000 2 93 98
Analytical Information Markup Language
Townsend J Adams S Waudby C de Souza V Goodman J Murray-Rust P Org Biomol Chem 2004 2 3294 3300 15534707 10.1039/b411033a
The IUPAC International Chemical Identifier
SMILES Home Page
CAS Information use Policies
Pubchem
Kokkola T Savinainen J Mšnkkšnen K Retamal M Laitinen J S-Nitrosothiols modulate G protein-coupled receptor signaling in a reversible and highly receptor-specific manner BMC Cell Biology 2005 6 21 15850493 10.1186/1471-2121-6-21
Walsh H O'Shea K Bottin N Comparative inhibition by substrate analogues 3-methoxy- and 3-hydroxydesa-minokynurenine and an improved 3 step purification of recombinant human kynureninase BMC Biochemistry 2003 4 13 14505498 10.1186/1471-2091-4-13
Kulys J Ziemys A A role of proton transfer in peroxidase-catalyzed process elucidated by substrates docking calculations BMC Structural Biology 2001 1 3 11545682 10.1186/1472-6807-1-3
Murray-Rust P Rzepa HS J Digital Inf 2004 5 248
Budapest Open Access Initiative
Holliday G Murray-Rust P Rzepa HS J Chem Inf Mod 2005
Systems Biology Markup Language
Biological Pathways Exchange
Holliday G Bartlett G Murray-Rust P Thornton J Mitchell J Abstracts of Papers, 226th ACS National Meeting, New York, NY, United States,, September 7–11, 2003, CINF-099
|
16026614
|
PMC1187874
|
CC BY
|
2021-01-04 16:27:25
|
no
|
BMC Bioinformatics. 2005 Jul 18; 6:180
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-180
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1851603365910.1186/1471-2105-6-185SoftwarePSI-BLAST-ISS: an intermediate sequence search tool for estimation of the position-specific alignment reliability Margelevičius Mindaugas [email protected] Česlovas [email protected] Institute of Biotechnology, Graičiūno 8, LT-02241 Vilnius, Lithuania2005 21 7 2005 6 185 185 17 3 2005 21 7 2005 Copyright © 2005 Margelevičius and Venclovas; licensee BioMed Central Ltd.2005Margelevičius and Venclovas; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Protein sequence alignments have become indispensable for virtually any evolutionary, structural or functional study involving proteins. Modern sequence search and comparison methods combined with rapidly increasing sequence data often can reliably match even distantly related proteins that share little sequence similarity. However, even highly significant matches generally may have incorrectly aligned regions. Therefore when exact residue correspondence is used to transfer biological information from one aligned sequence to another, it is critical to know which alignment regions are reliable and which may contain alignment errors.
Results
PSI-BLAST-ISS is a standalone Unix-based tool designed to delineate reliable regions of sequence alignments as well as to suggest potential variants in unreliable regions. The region-specific reliability is assessed by producing multiple sequence alignments in different sequence contexts followed by the analysis of the consistency of alignment variants. The PSI-BLAST-ISS output enables the user to simultaneously analyze alignment reliability between query and multiple homologous sequences. In addition, PSI-BLAST-ISS can be used to detect distantly related homologous proteins. The software is freely available at: .
Conclusion
PSI-BLAST-ISS is an effective reliability assessment tool that can be useful in applications such as comparative modelling or analysis of individual sequence regions. It favorably compares with the existing similar software both in the performance and functional features.
==== Body
Background
Protein sequence alignments are at the heart of many biological applications such as sequence database searches, annotation of new sequences, inference of functional regions, comparative protein modeling. Modern sequence comparison methods (e.g. PSI-BLAST [1]) often can reliably establish an evolutionary link between two proteins and align them even if they share little sequence similarity. However, the resulting significant match between these protein sequences may well include incorrectly aligned regions that are impossible to identify by straightforward inspection. Usually, the lower is the sequence similarity the more challenging is to distinguish alignment regions that can be trusted from those that may have errors. Yet, such a distinction is very important if the exact correspondence of residue positions in sequence alignments is used to extrapolate the biological information from one protein to another. Modeling protein structure by comparison (comparative modeling), identification of active site residues, selection of sites for point mutations are just a few examples where the reliability of aligned positions is critical.
The importance of delineating reliable alignment regions has been recognized more than a decade ago, however, earlier studies focused on pairwise alignments [2-5]. Currently, due to abundant sequence data, most protein sequence comparisons are performed within the context of multiple homologs, and the importance of pairwise alignments has diminished. By including multiple homologous sequences, methods such as PSI-BLAST are able to reliably detect more distant evolutionary links and also produce more accurate alignments. Unfortunately, even most advanced sequence alignment methods do make mistakes and the identification of reliable alignment regions remains an important problem. Estimation of position-specific alignment reliability is being addressed in some recent multiple sequence alignment methods [e.g. [6,7]]. However, in the multiple alignment case the position-specific reliability index estimates the overall proportion of correct pairwise matches in each alignment column without specifying the contribution of individual sequences. Yet in applications such as comparative modeling usually it is more important to know the position-specific alignment reliability for a given sequence pair than for the whole set of aligned sequences. Recently, a growing understanding of the importance of the problem led to several studies aiming at identification of reliable alignment regions for a pair of sequences within the context of multiple homologs. For example, one of these studies found that a substantial number of misaligned positions could be removed using the near-optimal alignment information [8]. Two other recent methods have been developed that predict reliable alignment regions either directly from a generated sequence profile [9,10] or using a consensus result of several alignment algorithms [11,12]. Both latter methods are implemented as web-based servers, which makes them easily accessible and simple to use, but not without certain limitations. For example, both servers require that one of the two sequences in the alignment would have a corresponding PDB structure, which in turn would have to be present in local databases used by these servers.
Here, we present the PSI-BLAST Intermediate Sequence Search tool (PSI-BLAST-ISS) that is primarily designed to help identify reliable regions of the alignment as well as suggest potential alignment variants in unreliable regions. In comparative modeling PSI-BLAST-ISS can also help identify best matching structural templates. In addition, PSI-BLAST-ISS can be used to detect remote homologs that cannot be identified by a straightforward single PSI-BLAST search. However, it should be noted that the detection of remote homologs, unlike in the original and subsequent implementations of the Intermediate Sequence Search (ISS) strategy [13-17], is not the main purpose of our tool.
Since PSI-BLAST-ISS might be most useful in comparative modeling we are going to refer to the sequence pair of interest as the target (query) and the template (reference) sequences throughout the article. However, it should be emphasized that the tool can be applied for any protein sequences that could be linked through common homologs, independently whether the three-dimensional structure for any of them is available or not.
The main idea of PSI-BLAST-ISS is to obtain a number of alignment variants for the sequence pair of interest (target and template) and analyze their consistency. This idea has stemmed from previous manual analysis of multiple PSI-BLAST alignment variants suggesting that regions where variants do agree are likely to be aligned correctly and display close structural similarity [18].
Implementation
The whole PSI-BLAST-ISS procedure may be described as the following steps: (1) identification of multiple sequences related both to the target and template sequences, (2) formation of a representative set from these sequences by filtering out close homologs, (3) generation of sequence profiles for each sequence from this representative set by searching a sequence database with PSI-BLAST, (4) using each of the generated profiles to search a second sequence database that includes sequences of both the target and the template, (5) retention of all the instances of significant matches between the target and the template, (6) merging all significant target-template alignments by taking the target sequence as a frame of reference and (7) reducing the multiple variants of aligned template into the consensus sequence. The latter option enables contrasting of the region-specific reliability for multiple target-template alignments simultaneously. All the seven main steps are illustrated in a sketch of the data flow (Fig. 1) and are described in more detail below.
Figure 1 Main steps in the PSI-BLAST-ISS execution. PSI-BLAST-ISS comprises seven main steps to produce consensus sequence alignment starting with the target sequence. The position-specific alignment reliability can be estimated either from individual target-template multiple alignments obtained in step 6 or from the combined alignment of consensus template sequences (step 7). For example, in this figure (step 7) only two template sequences are estimated to be reliably aligned with the last helix of the target, other templates lack consensus in this region.
As an input, PSI-BLAST-ISS takes the target sequence in FASTA format and a file containing a number of parameters that enable a user both to specify sequence databases and to control the execution of the whole ISS procedure at every step. The target sequence is initially searched against a sequence database to collect intermediate sequences (step 1). By default, the target is searched against the non-redundant sequence database. Intermediate sequences are collected from the user-specified PSI-BLAST iteration in the resulting output file using the expectation value (E-value) threshold provided as a parameter. The reduced representative sequence set is constructed by filtering the initial set to a user-defined percentage of sequence similarity with CD-HIT (Li et al., 2001), the sequence clusterization program (step 2). Optionally, a user may introduce a strict limit to the number of sequences to be included in the representative set or even supply independently pre-selected set of sequences. A PSI-BLAST-ISS user can also choose whether to collect intermediate sequences as complete protein sequences or just as sequence fragments matching the target sequence. In the case when the target sequence represents a domain that is also found in multidomain proteins the ability to select only homologous fragments of matching sequences may help to keep the ISS procedure from straying into the realm of unrelated sequences. Each of the intermediate sequences is used to generate a sequence profile in the form of the PSI-BLAST checkpoint file by running a user-defined number of PSI-BLAST iterations (step 3). The resulting checkpoint files are then used to restart PSI-BLAST searches in a second sequence database specified by the user (step 4). This database is expected to include sequences of both proteins of interest (target and template). In a common situation, when the template represents a structural template intended for use in comparative modeling, such a database may be derived by simply appending the target sequence to the PDB sequence database. In this case there is no need to define template(s) in advance since they are identified automatically. Searches against the second database generate corresponding multiple sequence alignments that contain a number of target-template alignment variants. The significance of the target-template alignment is then determined by counting the number of alignment variants that satisfy the expectation value threshold (step 5). Both parameters can be specified by the user. The significant target-template alignment variants are extracted and merged into a single multiple sequence alignment, where the target sequence is aligned with multiple instances of the template sequence according to different alignment variants (step 6). Such an alignment immediately reveals the regions where most (or all) alignment variants are identical and thus might be considered reliable as well as those regions where there is little agreement between alignment variants and therefore unreliable. Often it is useful to analyze position-specific reliability for target alignments with multiple templates. However, it may be inconvenient to contrast/compare at once many multiple sequence alignments obtained by PSI-BLAST-ISS. To make this task easier we introduced a step (step 7) that reduces template alignment variants into a consensus template sequence for each of the target-template alignments. The consensus sequence is generated by analyzing each column of the alignment. A residue is considered conserved in the consensus template sequence if its repetition count in the corresponding position exceeds the user-defined conservation threshold.
PSI-BLAST-ISS currently is implemented as a standalone UNIX-based tool meant to be installed and run locally. It consists of fairly independent modules linked together using Perl. Some of the sequence data processing tasks in PSI-BLAST-ISS are handled by a few modified SEALS scripts [19].
Results and Discussion
PSI-BLAST-ISS output
PSI-BLAST-ISS produces several types of results. Perhaps the most informative output file is the FASTA-formatted sequence alignment between the target and automatically detected multiple template sequences, each represented as a consensus sequence derived from multiple alignment variants. The definition line for each consensus template sequence indicates the strength of the consensus in the interval from 0 to 1 (0 – no consensus, 1 – complete agreement) and the number of significant target-template alignment variants that were used to produce the consensus. This output provides a possibility to simultaneously assess the alignment reliability between the target and multiple templates in a region-specific manner. In addition, the consensus strength and the number of significant target-template alignments may help in selecting templates that are structurally most consistent with the target. PSI-BLAST-ISS also produces individual FASTA-formatted multiple sequence alignment files for each target-template pair, where the target is aligned with multiple copies of the same template according to obtained multiple alignment variants. These alignment files provide a visual assessment of the region-specific alignment reliability as well as candidate alignment variants if further analysis of unreliably aligned stretches is needed. Finally, all the template sequences represented in the consensus alignment are collected together in a separate output file.
Performance of PSI-BLAST-ISS in the assessment of alignment reliability
Like for any method it is important to know how PSI-BLAST-ISS performs relative to other available methods. At the time of this study we have been aware of only two publicly available servers that estimate the position-specific reliability of sequence alignment using information from multiple sequences: the Consensus server [12] and SQUARE [9]. Of those, the performance of PSI-BLAST-ISS could be directly compared only with the Consensus server since SQUARE estimates reliability only for the supplied alignment and does not address the problem of alignment itself.
To compare PSI-BLAST-ISS and Consensus we chose protein sequences provided as prediction targets in the last round [20] of the community-wide protein structure prediction experiment known as CASP . These proteins represent a variety of different structural folds and different degree of similarity to known structures. We ran PSI-BLAST-ISS for all the target sequences assessed in CASP6, but only those, for which PSI-BLAST-ISS with default parameters generated at least ten significant alignment variants with a structural template, were further analysed. The "gold standard" in evaluating sequence alignments is to compare them with the alignments derived from protein structure superposition. For most targets PSI-BLAST-ISS detected multiple templates but for evaluating its performance we only considered a single template for each target. The DaliLite structure comparison program [21] was used both to select the template structurally closest to the target (the highest DaliLite Z-score) and to derive the "gold standard" alignment between the target and the template. The performance of PSI-BLAST-ISS was then assessed by checking to what extent alignment regions considered by PSI-BLAST-ISS to be reliable (consensus sequence assigned) agree with DaliLite structure-based alignments. In parallel, the same target-template sequence pairs were submitted to the Consensus server. The regions deemed by Consensus both structurally conserved and confidently aligned (indicated with 'S') were in turn contrasted with DaliLite structural alignments. Results obtained by PSI-BLAST-ISS and the Consensus server are presented in Table 1. In the case of PSI-BLAST-ISS, results for two consensus assignment thresholds (0.8 and 0.9) are provided.
Table 1 Comparison of PSI-BLAST-ISS and the Consensus server performance
Target Template Align length Rmsd, Å Seq id, % Consensus server PSI-BLAST-ISS (consensus, 0.8) PSI-BLAST-ISS (consensus, 0.9)
discrepancies d-len/cons-len discrepancies d-len/cons-len discrepancies d-len/cons-len
T0196 1jny 80 1.6 33 68–77 10/62 68–74, 77 8/56 69–72 4/50
T0200 1ush 210 2.7 16 71 1/45 - 0/58 - 0/50
T0202 1u0r 245 2.0 26 59–62, 94 5/146 59, 108 2/186 - 0/144
T0204 1gup 280 2.0 25 164 1/173 139, 141 2/17 - 0/14
T0208 1i60 254 3.1 11 258–262 5/102 292 1/90 - 0/85
T0211 1eut 126 1.7 22 - 0/13 - 0/65 - 0/55
T0216 1vpb 417 2.5 25 - 0/177 38–39, 71 3/238 - 0/107
T0222 1rzm 239 2.7 14 154 1/96 154, 246, 281–285 7/143 154, 246, 281–285 7/116
T0223 1vfr 123 2.5 11 - 0/22 113–119, 121–124, 128–132 16/39 113–116 4/24
T0228 1qpn 145 3.1 11 - 0/21 170–172, 174–181 11/57 - 0/39
T0229 1ml8 125 1.9 35 120–127 8/61 82–83, 120–127 10/114 120–127 8/84
T0231 1v6f 136 1.4 79 - 0/133 - 0/130 - 0/125
T0232 11gs 199 2.1 19 - 0/32 5–6, 40, 42–43, 66, 158 7/114 6, 40, 42–43, 66, 158 6/97
T0233 1kgz 319 1.8 36 136, 325–326 3/276 136, 245, 325–327 5/306 136, 245 2/279
T0234 1g76 118 2.8 14 16 1/59 11, 13–14, 16 4/67 11, 13–14, 16 4/58
T0235 1nb8 276 2.4 26 442–443, 478 3/44 207–208, 443, 478–479 5/104 207–208, 443 3/93
T0240 1lr0 70 2.5 17 25,33 2/39 10–11, 20–22, 24–26, 28–30, 65–66 13/51 20–22, 24–26 6/43
T0244 1iim 242 2.6 24 206, 229–234 7/157 231–234 4/153 - 0/115
T0246 1cnz 353 1.4 57 - 0/315 - 0/313 - 0/276
T0247 1pj5 338 2.1 25 76, 115–126, 128–129, 162–165, 261, 302 21/227 76, 162, 305 3/227 76 1/150
T0264 1vhv 244 2.1 34 60 1/120 17–19 3/92 17–19 3/81
T0265 1sfx 87 3.0 25 - 0/50 - 0/50 - 0/42
T0266 1dbu 145 1.8 25 60–65, 77 7/91 28, 77 2/132 - 0/95
T0267 1tiq 165 2.1 16 80–83 4/77 68, 80–81 3/102 68, 80–81 3/91
T0268 1m6y 277 1.6 49 121, 153–154 3/216 44, 126, 249 3/258 126 1/216
T0269 1qmv 182 2.1 35 89–92, 95, 119–120 7/125 119–120 2/128 - 0/99
T0274 1i0r 144 1.7 24 43,102–103 3/86 8, 43, 73 3/107 43 1/87
T0275 1mjh 125 2.0 30 26–30, 41–45 10/75 26–30, 41–42 7/107 29–30, 41 3/92
T0276 1sbq 153 1.7 26 92–94 3/56 18, 39, 150–151 4/122 - 0/84
T0279 1jr2 240 5.9 16 - 0/31 144–145, 202–204, 258 6/68 - 0/6
T0280 1o5o 144 3.0 19 170–175 6/55 119 1/51 - 0/46
T0282 1gq6 275 2.4 21 245 1/129 41–42, 208–210 5/202 42 1/138
Total: 113/3311 Total: 140/3947 Total: 57/3081
Average per target: 3.5/103.5 Average per target: 4.4/123.3 Average per target: 1.8/96.3
Fraction 3.4% Fraction 3.5% Fraction 1.9%
Target-template structure-based alignments that were used as reference are characterized by the number of superimposed residues (column Align length), root-mean-square deviation of their Cα atoms (Rmsd), and the sequence identity (Seq id). Differences between each structure-based alignment and alignments obtained either by the Consensus server or PSI-BLAST-ISS are reported in corresponding discrepancies columns. The discrepancies are reported as segments, and their begin-end positions are given with respect to the target sequence. Only consensus segments of at least 3 residues were considered. Columns d-len/cons-len provide ratios between the length of discrepancies (d-len) and the total length of the alignment considered to be reliable (cons-len) by the corresponding method.
The data in Table 1 indicate that using consensus assignment threshold of 0.8 PSI-BLAST-ISS produces more extensive coverage than the Consensus server at a slightly higher rate of discrepancies with DaliLite structure-based alignments. The visual inspection of the superimposed structures revealed that most of these alignment discrepancies are minor. Some of them occur simply due to a difference in a gap placement position when, for example, one of the structures in the pair has either single residue insertion or deletion. Some other discrepancies are short stretches at the transition of a conserved secondary structure into a non-conserved loop and also can hardly be considered alignment errors. Most of these minor discrepancies disappear once the consensus assignment stringency is increased to 0.9. While the coverage becomes only slightly less extensive than for the Consensus server, the discrepancy rate is almost two times lower. Thus the increase in the stringency of the PSI-BLAST-ISS consensus assignment lowers the chances of including both non-conserved structural motifs and alignment errors within regions assigned as reliable.
Utility of multiple alignment variants
A useful feature of PSI-BLAST-ISS is that it provides multiple alignment variants between the target and each template. Results in Table 1 show that regions where most alignment variants agree (consensus 0.8 or higher) usually represent reliably aligned structurally conserved stretches of protein chain. In contrast, the absence of a strong PSI-BLAST-ISS consensus indicates that any alignment variant in the corresponding region is not to be trusted. The unreliable alignment may point to a structural difference in the region such as in an example shown in Figure 2. Another possibility is that the structure of the region is conserved, however, because of the sequence dissimilarity or the variability of adjacent regions (insertions/deletions) sequence comparison programs fail to consistently come up with the same alignment variant. If related protein structures suggest that the considered region is indeed structurally conserved the correct alignment might be present among the variants generated by PSI-BLAST-ISS. For example, in the case of T0247 (Fig. 3), PSI-BLAST-ISS did not consider one of the structurally fairly conserved α-helices (115–132) reliably aligned with the corresponding region of the structural template (1pj5) and did not assign the consensus. Nevertheless, PSI-BLAST-ISS did suggest the correct alignment as one of the two major variants. In contrast, the Consensus server did supply a confident yet wrong alignment. It is easy to see that in this particular case an insertion on one side and a deletion on the other side of the otherwise conserved α-helix present a formidable problem for sequence-based methods. On the other hand, in cases like this, it might be possible to make a confident selection of the correct alignment by applying other methods that go beyond sequence comparison. In the homology modeling an assessment of different alignment variants within the context of the three-dimensional structure might be one of the potential solutions [e.g. [22]].
Figure 2 Lack of the alignment consensus may reflect a structural divergence of the motif. One of the α-helices (light color) displays a considerable difference in orientation in the two superimposed structures, target T0282 and the template 1gq6. All other regions of the structures are assigned the color gradient ranging from blue (N-termini) to red (C-termini). The lower part of the figure shows this α-helix and adjacent regions of T0282 aligned with 1gq6 according to both structural correspondence (dali) and a consensus alignment produced by PSI-BLAST-ISS (iss). The secondary structure of the target T0282 is shown above the sequence alignment.
Figure 3 Lack of the alignment consensus in a structurally conserved region due to variable adjacent regions. Structural superposition of T0247 with the template 1pj5. The considered T0247 α-helix (white) superimposes fairly closely with the corresponding α-helix (light yellow) in 1pj5, but has an insertion at one end and a deletion at the other end. The lower part of the figure shows the α-helix and adjacent regions of T0247 aligned with the corresponding fragment of the 1pj5 sequence. For the 1pj5 sequence the structure-based alignment (dali), the PSI-BLAST-ISS consensus alignment (iss), two individual PSI-BLAST-ISS alignment variants (iss_var1 and iss_var2) and the Consensus server alignment (cons_srv) are shown. The alignment obtained by the Consensus server includes only residues considered to be aligned confidently (residues assigned to 'S'). The secondary structure diagram for T0247 is also shown above the sequence alignment.
Selection of representative templates (homologs)
Often there is a need to choose a single or just a few best templates from a large number of distantly related target homologs. This becomes a challenge at low sequence similarity when the sequence signal is no longer a good indicator of structural relatedness (for example, see Fig. 1 in [23]). The number of significant target-template variants retained by PSI-BLAST-ISS for generation of consensus template sequence might guide such selection of the template(s). The higher is the number of target-template alignment variants that are accepted as significant, the closer structural relationship between them might be expected. This number is directly available from the file containing the alignment between the target and the individual template and is also reported within the definition line for each template in the consensus alignment file.
Detection of distant evolutionary relationships (homologous folds)
Multiple initiation points in the PSI-BLAST-ISS procedure ensure that the space of homologous sequences is explored more exhaustively than in the case of a single query-based search. Owing to that, PSI-BLAST-ISS may uncover distant evolutionary relationships, which are not seen if only a single query-initiated PSI-BLAST search is performed. In other words, PSI-BLAST-ISS may serve as a transitive PSI-BLAST tool for the detection of homologous folds. To test this PSI-BLAST-ISS capability we used CASP6 Homologous Fold Recognition targets (FR/H). These targets do have evolutionary related structures in the PDB database but these relationships could not be detected by PSI-BLAST searches initiated with the target sequence. For this test we required at least one significant match to a PDB structure (template) from all intermediate sequence searches. To make the comparison compatible with the CASP6 setting we only considered structural templates that were available from PDB at the time of the CASP6 experiment. We also excluded from consideration those FR/H CASP6 targets, for which at least one domain could be matched to a PDB structure using a straightforward PSI-BLAST search. As a result, out of fourteen considered FR/H targets, PSI-BLAST-ISS was able to identify related structures for four of them (1rxx for T0203, 1pk6 and several others for T0206, 1jx7 for T0224, 1qpn and other structures for T0228). An interesting case is T0228. While direct PSI-BLAST search initiated with the T0228 sequence failed to find any related structure, PSI-BLAST-ISS identified several structures producing over ten significant matches each (a default parameter). The latter result stresses the fact that sometimes the space of homologous sequences might be skewed in such a manner that a single sequence search may not be very effective in identifying important relationships.
Conclusion
We have described PSI-BLAST-ISS, a tool for delineating reliable alignment regions and suggesting possible alignment choices in unreliable yet structurally conserved regions. PSI-BLAST-ISS might be most useful in assessing target-template alignments in comparative modeling or judging whether the interpolation of biological information directly form alignments is feasible for individual sequence regions. Unlike two other recently published methods for predicting reliable alignment regions (SQUARE and the Consensus server) PSI-BLAST-ISS is not confined to reference (template) sequences with known three-dimensional structure. The performance of PSI-BLAST-ISS in alignment reliability estimation was directly compared with the Consensus server. We find that on a set of CASP6 targets PSI-BLAST-ISS on average is able to produce more extensive coverage of confident alignment or fewer errors, depending on the selected consensus stringency. The functionality of PSI-BLAST-ISS also extends into detection of non-apparent distant homologous relationships.
Availability and requirements
Project name: The PSI-BLAST intermediate sequence search tool (PSI-BLAST-ISS)
Project home page:
Operating systems: Unix-based platforms
Programming language: Perl
Other requirements: locally installed PSI-BLAST and CD-HIT (optional)
License: None
Any restriction to use by non-academics: None
Authors' contributions
MM carried out the software development, programming work and participated in manuscript preparation. ČV conceived of the study, participated in its design and coordination and drafted the manuscript. All authors read and approved the final manuscript.
Acknowledgements
This research project was supported in part by grants from Howard Hughes Medical Institute and the 6th European Community Framework Programme.
==== Refs
Altschul SF Madden TL Schäffer AA Zhang J Zhang Z Miller W Lipman DJ Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694 10.1093/nar/25.17.3389
Chao KM Hardison RC Miller W Locating well-conserved regions within a pairwise alignment Comput Appl Biosci 1993 9 387 396 8402204
Mevissen HT Vingron M Quantifying the local reliability of a sequence alignment Protein Eng 1996 9 127 132 9005433
Schlosshauer M Ohlsson M A novel approach to local reliability of sequence alignments Bioinformatics 2002 18 847 854 12075020 10.1093/bioinformatics/18.6.847
Vingron M Argos P Determination of reliable regions in protein sequence alignments Protein Eng 1990 3 565 569 2217130
Do CB Mahabhashyam MS Brudno M Batzoglou S ProbCons: Probabilistic consistency-based multiple sequence alignment Genome Res 2005 15 330 340 15687296 10.1101/gr.2821705
Poirot O O'Toole E Notredame C Tcoffee@igs: A web server for computing, evaluating and combining multiple sequence alignments Nucleic Acids Res 2003 31 3503 3506 12824354 10.1093/nar/gkg522
Cline M Hughey R Karplus K Predicting reliable regions in protein sequence alignments Bioinformatics 2002 18 306 314 11847078 10.1093/bioinformatics/18.2.306
Tress ML Grana O Valencia A SQUARE--determining reliable regions in sequence alignments Bioinformatics 2004 20 974 975 14764569 10.1093/bioinformatics/bth032
Tress ML Jones D Valencia A Predicting reliable regions in protein alignments from sequence profiles J Mol Biol 2003 330 705 718 12850141 10.1016/S0022-2836(03)00622-3
Prasad JC Comeau SR Vajda S Camacho CJ Consensus alignment for reliable framework prediction in homology modeling Bioinformatics 2003 19 1682 1691 12967965 10.1093/bioinformatics/btg211
Prasad JC Vajda S Camacho CJ Consensus alignment server for reliable comparative modeling with distant templates Nucleic Acids Res 2004 32 W50 4 15215349
Karplus K Barrett C Hughey R Hidden Markov models for detecting remote protein homologies Bioinformatics 1998 14 846 856 9927713 10.1093/bioinformatics/14.10.846
Li W Pio F Pawlowski K Godzik A Saturated BLAST: an automated multiple intermediate sequence search used to detect distant homology Bioinformatics 2000 16 1105 1110 11159329 10.1093/bioinformatics/16.12.1105
Park J Karplus K Barrett C Hughey R Haussler D Hubbard T Chothia C Sequence comparisons using multiple sequences detect three times as many remote homologues as pairwise methods J Mol Biol 1998 284 1201 1210 9837738 10.1006/jmbi.1998.2221
Park J Teichmann SA Hubbard T Chothia C Intermediate sequences increase the detection of homology between sequences J Mol Biol 1997 273 349 354 9367767 10.1006/jmbi.1997.1288
Salamov AA Suwa M Orengo CA Swindells MB Combining sensitive database searches with multiple intermediates to detect distant homologues Protein Eng 1999 12 95 100 10195280 10.1093/protein/12.2.95
Venclovas č Comparative modeling of CASP4 target proteins: combining results of sequence search with three-dimensional structure assessment Proteins 2001 Suppl 5 47 54 11835481 10.1002/prot.10008
Walker DR Koonin EV SEALS: a system for easy analysis of lots of sequences Proc Int Conf Intell Syst Mol Biol 1997 5 333 339 9322058
Cozzetto D Di Matteo A Tramontano A Ten years of predictions ... and counting Febs J 2005 272 881 882 15691322
Holm L Park J DaliLite workbench for protein structure comparison Bioinformatics 2000 16 566 567 10980157 10.1093/bioinformatics/16.6.566
Venclovas č Comparative modeling in CASP5: progress is evident, but alignment errors remain a significant hindrance Proteins 2003 53 Suppl 6 380 388 14579326 10.1002/prot.10591
Venclovas č Zemla A Fidelis K Moult J Assessment of progress over the CASP experiments Proteins 2003 53 Suppl 6 585 595 14579350 10.1002/prot.10530
|
16033659
|
PMC1187875
|
CC BY
|
2021-01-04 16:27:25
|
no
|
BMC Bioinformatics. 2005 Jul 21; 6:185
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-185
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1871604277910.1186/1471-2105-6-187Methodology ArticleRank-invariant resampling based estimation of false discovery rate for analysis of small sample microarray data Jain Nitin [email protected] HyungJun [email protected]'Connell Michael [email protected] Jae K [email protected] Division of Biostatistics and Epidemiology, Department of Health Evaluation Sciences, University of Virginia School of Medicine, Hospital West Complex, Room. 3181, P.O. Box 800717, Charlottesville, VA 22908-0717, USA2 Insightful Corporation, 2505 Meridian Parkway Suite 175, Durham, NC 27713, USA2005 22 7 2005 6 187 187 10 2 2005 22 7 2005 Copyright © 2005 Jain et al; licensee BioMed Central Ltd.2005Jain 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 evaluation of statistical significance has become a critical process in identifying differentially expressed genes in microarray studies. Classical p-value adjustment methods for multiple comparisons such as family-wise error rate (FWER) have been found to be too conservative in analyzing large-screening microarray data, and the False Discovery Rate (FDR), the expected proportion of false positives among all positives, has been recently suggested as an alternative for controlling false positives. Several statistical approaches have been used to estimate and control FDR, but these may not provide reliable FDR estimation when applied to microarray data sets with a small number of replicates.
Results
We propose a rank-invariant resampling (RIR) based approach to FDR evaluation. Our proposed method generates a biologically relevant null distribution, which maintains similar variability to observed microarray data. We compare the performance of our RIR-based FDR estimation with that of four other popular methods. Our approach outperforms the other methods both in simulated and real microarray data.
Conclusion
We found that the SAM's random shuffling and SPLOSH approaches were liberal and the other two theoretical methods were too conservative while our RIR approach provided more accurate FDR estimation than the other approaches.
==== Body
Background
In microarray data analysis, hypotheses relating to differential expression of many genes across the experimental conditions are tested simultaneously. Typical research questions examine the effects of disease status and drug response on the expression of each gene. An extremely large number of e.g. >40K genes can be currently represented on a microarray, so that its statistical results must be carefully analyzed taking a false positive error rate and multiple comparison issues into account. In order to control such a false-positive rate, traditional statistical methods often control the family-wise error rate (FWER), the probability of incorrectly accepting at least one false-positive hypothesis (or type-I error) among all hypotheses; for example, the commonly-used Bonferroni correction divides the type I error α by the total number of hypotheses for the test of each gene's differential expression, assuming the hypotheses under consideration are independent [1]. However, this independence assumption is unlikely to be true in microarray data, as functions of many genes are interrelated in varying degrees. Moreover, the methods controlling FWER are frequently found to be too conservative to identify many important genes in biological applications [2]. Several authors (e.g., Sidak, WestFall and Young) have developed step-down procedures that apply the severe Bonferroni correction only to the most extreme value of the test statistic, and step down the correction with the value of the test statistic. However, these methods still result in high false-negative error, likely missing many genes that are truly differentially expressed.
Benjamini and Hochberg (BH) [3] suggested that controlling false discovery rate (FDR), the expected proportion of false positives among all positive (or rejected) hypotheses, is more appropriate for large screening problems. Benjamini and Yekutieli (BY) [4] proposed a new FDR procedure considering a certain dependency structure among the test statistics. However, both the BH and BY procedures may still be too conservative when applied to real microarray data analysis [1]. This is mainly due to the fact that the independence or the artificial dependency assumptions made in these approaches may not be supported in real microarray data applications. Furthermore, microarray experiments are often conducted with a small number of replicates due to limited availability of RNA samples and/or budgetary constraints [2].
One of the key issues in estimating FDR is the assumption regarding the underlying null distribution. The Significance Analysis of Microarrays (SAM) method [5] uses a full permutation strategy, sampling across all genes and conditions to generate such a null distribution (mix-all). However, this strategy breaks many intrinsic correlation structures and does not generate a realistic or biologically-relevant null distribution for microarray data (see Figure 1; its detailed explanation in the Result section). Chip-by-chip permutation strategies [1], which randomly shuffle all the columns (chips) and preserve gene structure, are not applicable when the sample size is small because the number of independent permutations is too small to generate a null distribution with enough granularity to support desired significance calculations. In order to provide more stable estimation of such FDR values, a method based on the spacings LOESS histogram (SPLOSH) was also proposed based on a certain assumption about the p-value distribution [6].
Figure 1 Scatter plots of null data. (a) null data within the same condition from the resampling method; (b) null data between the different conditions from the resampling method; (c) null data within the same condition from the Mix-all method; (d) null data between the different conditions from the mix-all method;
In order to overcome these restrictions, we propose a rank-invariant resampling (RIR) approach to FDR estimation, especially for microarray data with a small number of replicates. In particular, we use the local pooled error (LPE) test [2], which has high statistical power in analyzing low-replicate microarray data, as a tool for discovery of differential expression. In brief, the LPE approach is based on a model for variance as a function of mean expression intensity, shrinking observed within-gene error estimates by pooling error information of other genes in local intensity ranges and characterizing the variance function by a non-parametric smoother in order to improve the accuracy of error estimation in small sample microarray data analysis. Consequently, the LPE approach provides a dramatically higher statistical power than other within-gene test methods, such as SAM and two-sample tests, for identifying differentially expressed genes in microarray data with limited replication. We compare the performance of our approach with that of four other approaches – BH, BY, mix-all, and SPLOSH, using both simulated and real microarray data sets.
Results
Simulation study
We first investigate whether the proposed resampling method provides a realistic null distribution. We generate a set of null data from a real array data set by the proposed resampling method and the mix-all method. Figure 1 displays array-by-array scatter plots of null data from both methods in the from of the so-called A (each gene's average intensity between two arrays) versus M (each gene's intensity difference between two arrays) transformation. First, the scatter plots (a) and (b) by our RIR algorithm show heterogeneous error variances on different intensity ranges assimilating those in the original microarray data quite well. On the other hand, the plots (c) and (d) show much bigger, yet homogeneous error variances regardless of the intensity levels, which are considerably different from those in the real data. For comparing the FDR estimation methods, we generate simulated data as follows. Instead of certain (parametric) distributional assumptions about microarray data, we use real microarray data to obtain such data. That is, let X1 and X2 be log2-transformed and normalized data from the replicated chips on the same experimental condition of a microarray study. We first compute M (=X1 - X2) and A (=(X1 + X2)/2), and then divide the intensity range of A into 100 intervals. Let be the maximum of (the absolute value of) M in each interval and à is the corresponding A. Then, for equivalently expressed genes, we use means and variances under two different experimental conditions at each interval for our generation of null data. For each of differentially expressed genes, we derive its two means (say μ1 and μ2) under two different conditions using equations: (μ1 + μ2)/2 = à and (μ1 - μ2) = δ, where δ is a factor determining the degree of differential expression. In this paper, we use it δ = 1.5; more discussion about this selection can be found in the Discussion section below. The corresponding variances are obtained from LPE baseline variance estimates. For our simulation study, we generate expression intensities of triplicate arrays with 10,000 genes under each of two conditions with 5%, 10%, 20%, or 50% differentially expressed genes. For example, the Bland-Altman plot (M versus A plot) of a simulated data set with 10% differentially expressed genes is displayed in Figure 2, in which differentially expressed genes are shown in the upper or lower boundaries as points marked with red x's. The above non-parametric, adaptive generation of simulated data has been found to provide the most realistic microarray data and differential expression pattern of many data generation methods and settings tried (data not shown). Note that since our simulated data were randomly generated with the same dynamic ranges and the same underlying resampling distribution, a normalization step was not additionally performed for these simulated data. However, IQR (inter-quartile-range) or non-parametric regression-based normalization (e.g., loess) is recommended prior to the application of the RIR algorithm in practice as in Dudoit et al. (2002).
Figure 2 M vs A plot of simulated data. The simulated data contains 10% significant genes (indicated by 'x'), and 90% insignificant genes.
We then apply LPE to the simulated data sets and estimate FDR by our RIR method, as well as BH, BY, mix-all, and SPLOSH. In brief, using the variance estimates and the LPE z-statistic is derived as , where and are the medians under two conditions and n1 and n2 are numbers of replicates in the two experimental conditions being compared; in our simulation study n1 = n2 = 3. Next, the FDR levels are estimated with the three FDR evaluation methods. The FDR levels of 0.2 or smaller have been examined because only such levels of FDR would be useful in practice. Figure 3 shows that BH and BY provide very conservative results while the mix-all approach gives somewhat liberal results, especially when a small (less than 10%) percentage of genes are differentially expressed. SPLOSH is conservative at very small FDRs, and then rapidly becomes very liberal. Our RIR method provides the most accurate FDR estimates compared to the others, especially in the cases with a small percentage of differentially expressed genes (5 or 10%).
Figure 3 Comparison of four FDR estimation methods. (a), (b), (c), and (d) are the plots between true and estimated FDR for simulated data with 5%, 10%, 20%, and 50% differentially expressed genes, respectively.
Application to the mouse immune response data
The microarray data of the immune response study is used to show performance with real data [2]. This study was performed with triplicate microarrays under each of Naive and 48 hour-activated cells, using Affymetrix MG-U74Av2 chips of 12488 probe sets. Table 1 displays the numbers of the selected differentially expressed genes at FDR 0.0001, 0.001, 0.01, or 0.05. The results again show that BH and BY are more conservative than others, whereas the SPLOSH and mix-all methods are more liberal than the others. Table 2 shows the minimum FDR (or q-value) estimates for the five well-known genes that were reported and confirmed in the original study [2]. The q-value estimates of several genes among them were greater than 0.01 by conservative BH and BY. One or more genes' q-value estimates were greater than 0.01 by SPLOSH and mix-all, whereas RIR identified all of these genes with q-value < 0.01.
Table 1 Numbers of differentially expressed genes discovered by five methods
FDR cutoff BY BH SPLOSH Mix-all RIR
0.0001 1397 1730 2876 2542 2074
0.001 1730 2162 3134 2958 2485
0.01 2160 2849 3467 3694 3382
0.05 2670 3661 5654 4594 4548
Table 2 Minimum FDR estimates of well-known genes found to be differentially regulated genes
Gene Symbol Gene Title BY BH SPLOSH Mix-all RIR
CD97 CD97 antigen 0.0230 0.0023 0.0489 <0.0001 0.0006
GATA3 GATA-binding protein-3 0.0208 0.0021 0.0489 <0.0001 0.0006
Clast3-pending CD40 ligand-activated specific transcript 0.1005 0.0103 <0.0001 0.0007 0.0034
GZMK Granzyme K 0.2768 0.0277 0.0524 0.0037 0.0091
FAF1 Fas-associated factor-1 1.0000 0.1100 <0.0001 0.0335 0.0038
Discussion and Conclusion
In this study we have demonstrated that our RIR-based FDR estimation method significantly outperforms the other popular approaches and provides very accurate FDR estimates, especially when a small percentage of genes are differentially expressed. Among the other FDR evaluation methods compared, the BH and BY methods were found to provide quite conservative results and failed to identify a number of truly differentially expressed genes in real microarray data, whereas the full-permutation (mix-all) approach appeared to yield false positives as significant genes.
In this study we found that one of the most critical steps in FDR evaluation is the generation of biologically-relevant null data. This step has failed and/or is difficult to assimilate in other theoretical and computational FDR estimation approaches. We believe that our heuristic, resampling-based approach provides a significant improvement on FDR estimation and a realistic and intuitive framework for understanding FDR in practice. Other approaches in use are based on quite restrictive mathematical assumptions and/or computational constraints, which result in a biologically unrealistic framework for statistical estimation and discovery. In particular, the simple, full permutation strategy produces both an inflated pooled variance and an inflated difference between the gene intensities, but results in a liberal testing framework because the inflation in the numerator of the test statistics (differential expression) is larger than that in the denominator (variance) in such a null distribution. On the other hand, the shuffling strategies across all conditions can not be applied to microarray data with a small sample size, as the number of independent permutations is too small to provide any meaningful results.
In many microarray studies under controlled experimental conditions, one may expect less than 10% of the genes to be differentially regulated, and thus removal of the top 10% genes from each local interval can be effective in generating a null-distribution excluding most of the differentially expressed genes. Our simulations show that removal of the top 5%, 10%, 20%, or even 50% genes does not affect the null distribution (data not shown), but we admit that these are yet subjective choices and may require a more extensive investigation. Our simulation studies have shown that removing the top 10% of genes produces results close to the true FDR among the four cases with 5% to 50% of differentially expressed genes. In Figure 3, we showed the comparison among the FDR evaluation methods for the simulated data with the proportions of differentially expressed genes varying between 5% and 50%. In many microarray studies, the proportion of differentially expressed genes would be lower than this. Thus, as somewhat expected, the mix-all approach, which is not sensitive to variability across different intensity ranges in microarray data, performs quite well if the proportion of differentially expressed genes is high and a large number of genes do not follow the baseline error distribution. Overall, the bigger such a proportion, the better the mix-all approach would perform. Note that with 5% and 10% of differentially expressed genes, the mix-all method performed poorer, with more liberal, underestimated FDR estimates, than our RIR approach. As Pounds and Cheng [6] reported, the FDR estimates of the mix-all approach are found to be somewhat unstable for low FDR, which may be a critical region in real data applications.
It has often been found that the results from simulation studies may be considerably affected by certain predefined parameters and settings, for example, δ for the differential expression magnitude and q for the estimation of null-gene proportion in our current study. As such we examined sensitivity of our results to these settings. First, we found that our results were not much different for different choices of q between 0.5 – 0.95 (data not shown). Also, although a more reasonable cross-validated approach is yet to be developed for choosing the δ value, our current parameter value was empirically chosen from an actual microarray data analysis. We then consistently used this value in our simulation study with varying proportion of differentially expressed genes up to 50% and found little effect of this setting on the resulting null distribution.
We note that our RIR-based FDR estimation is derived for each threshold value c of LPE z-score and that the ratio of V(c) and R(c) is then calculated only when R(c) > 0, so that this effectively provides an estimate of pFDR(Z > c), the q-value. Thus, the RIR-based FDR evaluation can be considered as a carefully designed resampling-based q-value estimation [7]. Note also that our RIR-based approach can be applied to microarray data analysis independent of different preprocessing methods.
In Table 2, several known genes' FDR estimates from the SPLOSH and mix-all approaches were larger than those of RIR. This is somewhat contrary with the observation that the SPLOSH and mix-all approaches were more liberal than the RIR as seen in Fig. 3 and Table 1. This may be due to the fact that these genes have relatively low variability, i.e., in high intensity regions, so that their significance is higher by considering such heterogeneous variability by RIR, but not by the others.
Methods
Generation of biologically relevant null distribution
It is critical to generate an underlying null distribution as close as possible to real microarray data because a gene's statistical significance can be dramatically different under different underlying null distributions. Therefore, our resampling strategy is designed to preserve the biological structure of each microarray data set as much as possible. Before describing our resampling strategy, we present an algorithm for constructing intervals, which is used in our resampling strategy. A naive approach for construction of intervals is to partition intensity ranges so that each interval has an equal number of genes. This approach may yield overly large test statistics in high intensity levels because intensities are very sparse in high levels and condense in the middle levels. In order to obtain the local intervals of the genes with homogeneous variances, we therefore construct adaptive intervals by the following algorithm.
Adaptive Interval (AI) algorithm
1. Estimate a baseline variance function for all data under consideration (within each experimental condition) by LPE
2. Obtain medians and variance estimates for each gene.
3. Order the medians and variances by the medians and denote the ordered medians and variances by ξ(i) and σ(i).
4. Obtain the first interval with threshold values ξ(1) and ξ(1) + σ(1).
5. Obtain the next interval with ξ(2) and ξ(2) + σ(2), where ξ(2) is the smallest median such that ξ(2) ≥ ξ(1) + σ(1).
6. Repeat step 5 to obtain the next intervals with ξ(i) and ξ(i) + σ(i), where i is the index of the smallest median such that ξ(i) ≥ ξ(i - 1) + σ(i - 1) until all the data are assigned to certain intervals.
Note that the number of genes in each interval is forced to be between given minimum and maximum numbers. In this paper, we used 10 and (1/100 of the total number of genes) for the minimum and maximum numbers, respectively. Note also that this AI algorithm is applied to the replicated array data under each experimental condition separately.
Our RIR procedure for generating null data is then as follows.
1. Calculate medians for each gene and obtain the ranks of these medians within each experimental condition.
2. Calculate rank differences between two conditions for each gene.
3. Construct the first intensity intervals using the AI algorithm above and retain rank-invariant genes by eliminating a certain percentage of genes with largest rank differences within each interval.
4. Construct the final intensity intervals of rank-invariant genes using the AI algorithm.
5. Obtain a set of null data by resampling intensities of rank-invariant genes within each interval.
6. Repeat the above step B times, e.g., 1,000, to obtain B independent sets of resampled null data.
In step 5 of the above procedure, a certain percentage of genes are eliminated to retain only rank-invariant expressed genes. In this current application, we remove 50% of all genes with largest rank differences; a discussion regarding other choices is presented later. Note that the AI algorithm is used twice in this RIR procedure; the first time to remove rank-variant genes evenly throughout the whole intensity ranges. Without this step, it was found that many genes in low intensity ranges were unproportionately removed due to the larger variability in those ranges (data not shown). This is a particularly important issue for Affymetrix data that have been summarized using the MAS5 procedure.
Estimation of FDR based on the RIR procedure
We calculate LPE Z-statistics Znull from null data as generated following the procedure described above. Generation of the null data is repeated many times independently. Let Zreal be a LPE Z-statistic computed from the real data. FDR at a threshold value c can be estimated as
where V(c) is the average number of Znull equal to or greater than c and R(c) is the number of Zreal equal to or greater than c. The proportion π0 of true null genes in real data can be estimated by the number of {Zreal ≤ λq} divided by the average number of {Znull ≤ λq}, where λq is the q-th quantile of Znull as suggested by Storey and Tibshirani (2003). In this paper, we use 0.9 for q; more discussion about this choice can be found in the Discussion section below. A gene's FDR value might be estimated as zero when no gene in the resampled null data exceeds its Zreal; in these cases we force the minimum estimate of FDR to be the reciprocal of the product between the numbers of genes and resampled null data sets, which is the finest resolution of our RIR FDR estimation. Note that the confidence bounds for at each threshold value c can also be obtained from the B resampled null data sets.
Other FDR estimation methods
SAM's full permutation (or mix-all) strategy randomly samples all intensity values across genes and conditions to generate null data, of which FDR estimation can be similarly performed as described above for our RIR approach. Benjamini and Hochberg (BH) [3] proposed the step-up procedure to control FDR. These approaches can be compared with our RIR approach based on the LPE statistics in the following manner. Let z(1) ≥ z(2) ≥ ... ≥ z(G) be LPE z-statistics for discovery of differential expression of G genes. Denote the corresponding ordered raw p-values as p(1) ≤ p(2) ≤ ... ≤ p(G). BH adjusted p-values are defined as = mink=i,...,G{min(p(k)G/k, 1))}. For control of FDR at level α, a gene i is claimed as significant if ≤ α. Thus, the BH estimate of FDR at a given critical value c can conservatively be defined as , where i* is min{i : z(i) ≥ c}. The adjusted p-values of Benjamini and Yekutieli (BY) [4] are defined as . Utilizing the information in both left-hand and right-hand sides of the p-value distribution, the SPLOSH FDR estimate is h(i) = mink≥i(r(k)), where r(k) is a conditional FDR (cFDR) estimate of gene k and cFDR is a FDR given the number of positives [8]. These four methods for FDR estimation are compared with our RIR method in the next section.
Authors' contributions
N.J. wrote the computer code and did the simulation work. All the authors contributed in developing the idea and wrote the manuscript.
Acknowledgements
This study was supported by the American Cancer Society grant RSG-02-182-01-MGO of J.K.L.
==== Refs
Dudoit S Yang Y Speed T Callow M Statistical methods for identifying differentially expressed genes in replicated cDNA microarray experiments Statistica Sinica 2002 12 111 139
Jain N Thatte J Braciale T Ley K O'Connell M Lee JK Local-pooled-error test for identifying differentially expressed genes with a small number of replicated microarrays Bioinformatics 2003 19 1945 1951 14555628 10.1093/bioinformatics/btg264
Benjamini Y Hochberg Y Controlling the false discovery rate: A practical and powerful approach to multiple testing Journal of the Royal Statistical Society, Series B 1995 57 289 300
Benjamini Y Yekutieli D The control of the false discovery rate in multiple testing under dependency Annals of Statistics 2001 29 1165 1188 10.1214/aos/1013699998
Tusher V Tibshirani R Chu C Significance analysis of microarrays applied to transcriptional responses to ionizing radiation Proceedings of the National Academy of Sciences 2001 98 5116 5121 10.1073/pnas.091062498
Pounds S Cheng C Improving false discovery rate estimation Bioinformatics 2004 20 1737 1745 14988112 10.1093/bioinformatics/bth160
Storey J Tibshirani R Parmigiani G, Garrett E, Irizarry R, Zeger S SAM thresholding and false discovery rates for detecting differential gene expression in DNA microarrays The Analysis of Gene Expression Data: Methods and Software 2003 New York: Springer-Verlag
Tsai C Hsueh H Chen J Estimation of false discovery rates in multiple testing: application to gene microarray data Biometrics 2003 59 1071 1081 14969487 10.1111/j.0006-341X.2003.00123.x
|
16042779
|
PMC1187876
|
CC BY
|
2021-01-04 16:27:23
|
no
|
BMC Bioinformatics. 2005 Jul 22; 6:187
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-187
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1921604864410.1186/1471-2105-6-192SoftwareStoring, linking, and mining microarray databases using SRS Veldhoven Antoine [email protected] Don de [email protected] Marcel [email protected] Jager Victor [email protected] Jan A [email protected] Guido [email protected] Department of Urology, Josephine Nefkens Institute, Erasmus MC, P.O. Box 1738, 3000 DR Rotterdam, The Netherlands2 Medical Oncology, Erasmus MC, Rotterdam, The Netherlands3 Bioinformatics, Erasmus MC, Rotterdam, The Netherlands4 Medical Informatics, Erasmus MC, Rotterdam, The Netherlands2005 27 7 2005 6 192 192 6 5 2005 27 7 2005 Copyright © 2005 Veldhoven et al; licensee BioMed Central Ltd.2005Veldhoven 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
SRS (Sequence Retrieval System) has proven to be a valuable platform for storing, linking, and querying biological databases. Due to the availability of a broad range of different scientific databases in SRS, it has become a useful platform to incorporate and mine microarray data to facilitate the analyses of biological questions and non-hypothesis driven quests. Here we report various solutions and tools for integrating and mining annotated expression data in SRS.
Results
We devised an Auto-Upload Tool by which microarray data can be automatically imported into SRS. The dataset can be linked to other databases and user access can be set. The linkage comprehensiveness of microarray platforms to other platforms and biological databases was examined in a network of scientific databases. The stored microarray data can also be made accessible to external programs for further processing. For example, we built an interface to a program called Venn Mapper, which collects its microarray data from SRS, processes the data by creating Venn diagrams, and saves the data for interpretation.
Conclusion
SRS is a useful database system to store, link and query various scientific datasets, including microarray data. The user-friendly Auto-Upload Tool makes SRS accessible to biologists for linking and mining user-owned databases.
==== Body
Background
The extraction of information from data generated by high-throughput experiments in genomics and proteomics has been likened to "attempting to drink from a fire hose". We are flooded with information on many levels such as whole genome DNA sequences, RNA expression, protein-protein interactions, protein modifications, and more. All this information is accessible in very different formats, ranging from well-organized curated gene sequences to unstructured free text in scientific literature. A system that can manage, link and query these heterogeneous types of datasets is therefore extremely valuable. The Sequence Retrieval System (SRS) is such a unified database system in which numerous different scientific databases have already been integrated [1].
Of special interest are data from high-throughput RNA expression microarrays [2,3]. Many of these datasets are freely available and, like information stored in other scientific databases, are from different platforms [4,5]. Integrating and mining these databases strongly facilitates the analysis of genes of interest but will also support discovery of disease markers, drug-targets and new knowledge in general [6-9]. One such platform is Oncomine, which has integrated many different microarray datasets, focussing on human cancer [10]. Additionally, standardized microarray depositories such as GEO (Gene Expression Omnibus) [11], ArrayExpress [12], and CIBEX [13] do or will soon provide options to browse and query the datasets [14-17]. No doubt, other platforms will be developed focussing on the integration of microarray data. If started from scratch, these initiatives will likely be limited in their direct linkage to other heterogeneous biological databases due to the laborious task of making those connections and programming the single and batch-wise query options. The universality and the availability of numerous scientific databases that have already been integrated in SRS make it a useful platform for integrating microarray databases. Although the SRS interface to query databases is quite user friendly, other aspects of working with SRS are not. These include (i) uploading microarray datasets, (ii) database security including setting user access, (iii) linking databases, (iv) generating standard views, and (v) communication with other programs such as statistical and clustering software. The current SRS interface has a major disadvantage in that it is not designed to perform complex calculations on the fly. This means that any microarray dataset to be uploaded must have all ratio and statistical calculations performed upfront. For example, once in SRS, one cannot change ratios from log10 to log2 or add an extra field per gene by dividing expression data of all "normal" by all "cancer" samples. However, software programs that perform calculations, statistical evaluations, clustering, protein domain predictions, homology searches, and more, can communicate with SRS. Interfaces can be generated that retrieve data from SRS, perform the required action and if desired, store the results in SRS. Alternatively, SRS allows direct integration of programs such as the BLAST and FASTA homology searches and the SRS-EMBOSS (European Molecular Biology Open Software Suite) tools [18,19].
Generating a database in which heterogeneous datasets are integrated is a challenge in itself. However, retrieving statistically meaningful data by comparing datasets from different sources, platforms and designs is particularly difficult [20]. There is a fast growing body of publications on microarray cross-platform comparisons, mainly showing how this can be achieved in very many different ways [8,21-26]. Statistical evaluations of data within a dataset of sufficient technical and biological replicates, are better defined and can be implemented per dataset within a database system [27,28]. The strategies and applications we discuss here to link, store and query scientific datasets in SRS, do not go beyond processed individual datasets and do not include cross-platform dataset integrations. We assume that each uploaded dataset consists of high-quality data and has been processed correctly.
In this paper we describe strategies to incorporate microarray databases into SRS and provide a database upload tool. Using the program Venn Mapper as an example, we show the possibility to automatically retrieve the stored microarray data from SRS for external statistical evaluation.
Implementation
Auto-Upload Tool
In order to import microarray databases into SRS (version 7.1.3), an Auto-Upload Tool was built (Figure 1). This PHP-written tool allows one to store databases of a predefined format into a user-owned and password protected directory on a local SRS server [see Additional file 1] [29]. In this directory, databases can be managed, viewed and uploaded into SRS. In the "edit-database" interface, links to other databases can be specified for each field. A standard view can be generated and the location of the dataset in the SRS directory determined. Finally, permissions can be set to control access to the various datasets in SRS. Upon uploading, the Auto-Upload Tool will generate the files required for SRS: (i) the SRS data file in which the spreadsheet input file is converted into a flat-file database, (ii) an Icarus syntax file (.is-file) which describes the layout of the flat-file database, (iii) a database index file (.i-file) which describes the way in which the different fields need to be indexed for the SRS system, (iv) a database view file (.view-file) in which a standard view is defined, and (v) an information file (.it-file) which can harbour a description of the dataset. These files are automatically placed into the SRS directory after which the Auto-Upload Tool updates the srsdb.i, user.i and site.i files. These files describe the name of the database, where the files are located (srsdb.i), user permissions (user.i), and configuration of the different database groups (site.i). The srssection command within the Auto-Upload Tool implements the changes in the configuration files after which srscheck and srsdo perform indexing of new databases and set links. Incorporation of a new dataset using this tool generally takes place within minutes. On our local SRS server, a DQS (Distributed Queuing System) batch-queue is installed to prevent data loss or corruption of datasets in case multiple users are editing datasets at the same time.
Figure 1 Screenshots of the database Auto-Upload Tool for SRS. Within the Auto-Upload Tool, a user can import a database and define links, SRS subdirectory, user access, a dataset description, and an SRS standard view.
External programs accessing SRS: Venn Mapper for SRS
An important feature of storing and linking microarray data in SRS is the accessibility of the datasets for other programs. As an example, we generated a PHP web interface for the Venn Mapper program that retrieves microarray data from SRS to calculate the statistical significance of the number of co-occurring differentially expressed genes in any combination of two experiments [26]. The functionality of the original Venn Mapper was enhanced by enabling the use of different ratio cut-offs for different microarray experiments. Upon login, the interface displays all microarray databases indexed in SRS to which the user has access. After selection of the datasets a second screen shows all fields (such as individual array experiments or averaged group ratios) of the selected databases. The microarray experiments of interest can be selected for Venn Mapper analysis after which the requested data is linked and exported from SRS into the Venn Mapper program. The output of the program is available for viewing and downloading. Information requests from the interface to SRS are made through the SRS getz command [30]. This powerful feature of SRS makes the integrated databases accessible to any external program.
Results and discussion
Preparing and linking of microarray databases in SRS
With respect to the microarray database set-up, there are two important considerations. First, in our experience, microarray data mining often starts with selecting genes based on their differential expression. Differential gene expression is best determined using statistical evaluation of the data based on sufficient technical and biological replicates [27,28]. Dependent on the statistical test and microarray platform, raw gene expression data and/or ratio calculations can be utilized. Second, making changes to a microarray dataset in SRS is impractical and datasets should be fully built before they are imported. This means that raw expression data and ratio calculations should be normalised and flagged and represented in a common format (such as log2). Importantly, statistical evaluation should be included. For simplicity in representation, datasets can be summarised in, for example, an average of all "normal" and "cancer" samples and additional fields of log2 "normal/cancer" can be included.
Linking of databases should be based on invariable and unique indexes. Links based on identifiers such as UniGene cluster identifiers that are regularly re-assigned, forces one to repeatedly update all databases that include such a denominator. Invariable links based on DNA sequence assignments such as GenBank and ProbeSet identifiers (IDs) are therefore more appropriate linking indexes. We recommend including only one of those invariable and unique linking fields in the microarray dataset to avoid the need for a regular update. Since many biological databases do not use these hard links, connecting microarray datasets to other databases can be achieved through a coupling file and/or by making use of the web of links provided by the biological databases (Figure 2). A coupling file can be minimal, containing for example only a RefSeq ID with the appropriate array spot number, or can contain a variety of indexes such as GenBank, RefSeq, SwissProt, OMIM, LocusLink, KEGG, GO, GeneCards, UniGene identifiers, that directly link to the different databases. Since some of these links are variable and biological databases keep growing with information, coupling files must be regularly renewed to update the various links.
Figure 2 SRS Universe example view of database linkage. The scientific database network consists of well-known interlinked biological databases. Microarray datasets are coupled to this network through a multi-linked Affymetrix coupling file, the single UniGene-linked NKB/NKI-coupling file, or directly to UniGene using GenBank accession numbers. For clarity, not all databases and links mentioned are represented in the scheme. Public SRS servers and indexed scientific databases can be viewed at [43].
Instead of or in combination with a coupling file, one can make use of the links provided in the various biological databases (Figure 2). For example, the GenBank accession code in the microarray dataset can be linked to the UniGene database. The LocusLink database is linked to the UniGene database through RefSeq accession numbers and also contains IDs that for example link to EMBL, OMIM, and SwissProt. In this way, almost all biological databases are linked through a network of direct and indirect connections. In case multiple roads lead to the same databases, SRS utilizes only one route. By assigning values to each link, the route taken is the one having the lowest sum of link values, even when this results in a lower number of connected fields (genes). Although SRS can be forced to take a specified path, one should be careful to be dependent on many different databases. Inconsistencies in and incompleteness of databases are accumulated when linking occurs in sequence.
Linking to the network of biological databases through a coupling file has various advantages. One can establish directly validated links to each database, including databases outside the network. In addition, errors in links can easily be corrected. Platform-specific or overall coupling files can be retrieved from the Affymetrix website [31] and from sources such as Resourcerer [32], KARMA [33], GeneHopper [34], ProbeMatchDB [35], DAVID [36], EnsMart [37], and Source [38]. Using these resources, microarray datasets from different platforms can be linked. This includes connecting cross-species datasets using ortholog converters such as HOMGL [39] and HOMOLOGENE [40].
Gene linking efficiency of different databases
A high accuracy and comprehensiveness of linking are essential for a successful comparison of microarray data from different platforms. The extent of linkage of various databases in SRS was examined (Figure 3). The percentage of fields of scientific databases and microarray datasets that are linked to other databases was assessed. As shown in Figure 2, most microarray databases are directly linked to the scientific database network via a single connection. The U133A and U95 Affymetrix coupling files contain direct links to UniGene, SwissProt, LocusLink, and OMIM. The linkage of the various microarray platforms to UniGene varies between 84% and 96%. On average, 60% of the genes of microarray datasets can be linked to each other via UniGene.
Figure 3 Linkage efficiency of databases within SRS. The percentage of records in a specific database (in rows) linked to others (columns) are shown. Numbers in brackets are the total number of fields in the particular database. The NKB/NKI Coupling [44], BC_NATURE_tVeer [45], Incyte Human UNIGEM V 1.0, PC_NATURE_Dhana. [46], Sigma/Compugen oligo array [47], and PC_PNAS_Lapointe [48], are linked to UniGene via a single accession code connection (see Figure 2). Direct links are depicted as grey cells. The Affymetrix U133A and U95 coupling files [31] are directly linked to different biological databases.
Auto-Upload Tool and external programs accessing SRS
The availability of many scientific databases in SRS, the universality of the system and its free access for academic use, make SRS an excellent mining system for heterogeneous microarray datasets. The Auto-Upload Tool facilitates the exchange of microarray datasets between separate SRS installations. Using a single data file and optional description file, any user can upload the identical data and customize it to their own SRS environment. We would urge researchers and microarray data repositories to make their data available in an SRS format. In addition, microarray software programmers could make their software available in an SRS compatible format or include SRS data export options. The commercial SRS GeneSpring® Connector and public EMBOSS are examples of such microarray-SRS integration ventures.
We plan to extend our efforts of integrating more microarray databases into SRS. In addition, software tools specific for microarray data analysis, such as Go Mapper and CoPub Mapper will be rewritten for SRS [26,41,42]. The CoPub Mapper literature mining program contains databases that store, for each gene, all MEDLINE records mentioning the gene. This directly links microarray expression data to the published literature and allows for co-publication research of gene-gene and gene-keyword combinations.
Conclusion
The Sequence Retrieval System is a versatile and useful database system to store, link and query various scientific databases, including microarray datasets. Fully processed datasets can be incorporated and linked to other datasets using the Auto-Upload Tool. This user-friendly program makes SRS accessible to users who can themselves add, link and mine databases within minutes. Datasets stored in SRS can be interrogated by external programs to perform virtually any computation.
Availability and requirements
Project Name: Auto-Upload Tool and Venn Mapper for SRS
Project home page:
Operating system: Platform independent
Programming language: PHP, JavaScript, Perl
Other requirements: Local SRS installation, DQS batch-queue, MySQL database server, PHP-enabled Webserver (like Apache)
License: SRS (Lion Bioscience)
Any restrictions to use by non academics: License needed
Authors' contributions
AV and DdL generated the Auto-Upload Tool. DdL, AV and MS generated the Venn Mapper for SRS program. AV and VdJ installed and managed the servers for the various tools. Funding for the project was obtained by JK and GJ. AV, JK and GJ contributed to the intellectual content and GJ supervised the project.
Supplementary Material
Additional File 1
describing how to use the Auto-Upload Tool program
Click here for file
Acknowledgements
We would like to thank EMBL/EBI and Lion Bioscience for making SRS available and Peter Hendriksen for careful reading of the manuscript. This work was supported by Erasmus MC Breedtestrategie and the Urologic Research Foundation (SUWO) Erasmus MC.
==== Refs
Zdobnov EM Lopez R Apweiler R Etzold T The EBI SRS server – recent developments Bioinformatics 2002 18 368 373 11847095 10.1093/bioinformatics/18.2.368
Brown PO Botstein D Exploring the new world of the genome with DNA microarrays Nat Genet 1999 21 33 37 9915498 10.1038/4462
Duggan DJ Bittner M Chen Y Meltzer P Trent JM Expression profiling using cDNA microarrays Nat Genet 1999 21 10 14 9915494 10.1038/4434
Heller MJ DNA microarray technology: devices, systems, and applications Annu Rev Biomed Eng 2002 4 129 153 12117754 10.1146/annurev.bioeng.4.020702.153438
Schena M Heller RA Theriault TP Konrad K Lachenmeier E Davis RW Microarrays: biotechnology's discovery platform for functional genomics Trends Biotechnol 1998 16 301 306 9675914 10.1016/S0167-7799(98)01219-0
Moreau Y Aerts S De Moor B De Strooper B Dabrowski M Comparison and meta-analysis of microarray data: from the bench to the computer desk Trends Genet 2003 19 570 577 14550631 10.1016/j.tig.2003.08.006
Rhodes DR Chinnaiyan AM Bioinformatics strategies for translating genome-wide expression analyses into clinically useful cancer markers Ann N Y Acad Sci 2004 1020 32 40 15208181 10.1196/annals.1310.005
Rhodes DR Yu J Shanker K Deshpande N Varambally R Ghosh D Barrette T Pandey A Chinnaiyan AM Large-scale meta-analysis of cancer microarray data identifies common transcriptional profiles of neoplastic transformation and progression Proc Natl Acad Sci U S A 2004 101 9309 9314 15184677 10.1073/pnas.0401994101
Welsh JB Sapinoso LM Kern SG Brown DA Liu T Bauskin AR Ward RL Hawkins NJ Quinn DI Russell PJ Sutherland RL Breit SN Moskaluk CA Frierson HF JrHampton GM Large-scale delineation of secreted protein biomarkers overexpressed in cancer tissue and serum Proc Natl Acad Sci U S A 2003 100 3410 3415 12624183 10.1073/pnas.0530278100
Rhodes DR Yu J Shanker K Deshpande N Varambally R Ghosh D Barrette T Pandey A Chinnaiyan AM ONCOMINE: a cancer microarray database and integrated data-mining platform Neoplasia 2004 6 1 6 15068665
Edgar R Domrachev M Lash AE Gene Expression Omnibus: NCBI gene expression and hybridization array data repository Nucleic Acids Res 2002 30 207 210 11752295 10.1093/nar/30.1.207
Brazma A Parkinson H Sarkans U Shojatalab M Vilo J Abeygunawardena N Holloway E Kapushesky M Kemmeren P Lara GG Oezcimen A Rocca-Serra P Sansone SA ArrayExpress – a public repository for microarray gene expression data at the EBI Nucleic Acids Res 2003 31 68 71 12519949 10.1093/nar/gkg091
Ikeo K Ishi-i J Tamura T Gojobori T Tateno Y CIBEX: center for information biology gene expression database C R Biol 2003 326 1079 1082 14744116
Stoeckert CJ JrCauston HC Ball CA Microarray databases: standards and ontologies Nat Genet 2002 32 469 473 12454640 10.1038/ng1028
Gardiner-Garden M Littlejohn TG A comparison of microarray databases Brief Bioinform 2001 2 143 158 11465732
Quackenbush J Data standards for 'omic' science Nat Biotechnol 2004 22 613 614 15122299 10.1038/nbt0504-613
Penkett CJ Bahler J Getting the most from public microarray data European Pharmaceutical Review 2004 9 8 17
Rice P Longden I Bleasby A EMBOSS: the European Molecular Biology Open Software Suite Trends Genet 2000 16 276 277 10827456 10.1016/S0168-9525(00)02024-2
Kulikova T Aldebert P Althorpe N Baker W Bates K Browne P van den BA Cochrane G Duggan K Eberhardt R Faruque N Garcia-Pastor M Harte N Kanz C Leinonen R Lin Q Lombard V Lopez R Mancuso R McHale M Nardone F Silventoinen V Stoehr P Stoesser G Tuli MA Tzouvara K Vaughan R Wu D Zhu W Apweiler R The EMBL Nucleotide Sequence Database Nucleic Acids Res 2004 32 D27 D30 14681351 10.1093/nar/gkh120
Marshall E Getting the noise out of gene arrays Science 2004 306 630 631 15499004 10.1126/science.306.5696.630
Zhou XJ Kao MC Huang H Wong A Nunez-Iglesias J Primig M Aparicio OM Finch CE Morgan TE Wong WH Functional annotation and network reconstruction through cross-platform integration of microarray data Nat Biotechnol 2005 23 238 243 15654329 10.1038/nbt1058
Mitchell SA Brown KM Henry MM Mintz M Catchpoole D LaFleur B Stephan DA Inter-platform comparability of microarrays in acute lymphoblastic leukemia BMC Genomics 2004 5 71 15387886 10.1186/1471-2164-5-71
Chiorino G Acquadro F Mello GM Viscomi S Segir R Gasparini M Dotto P Interpretation of expression-profiling results obtained from different platforms and tissue sources: examples using prostate cancer data Eur J Cancer 2004 40 2592 2603 15541960 10.1016/j.ejca.2004.07.029
Culhane AC Perriere G Higgins DG Cross-platform comparison and visualisation of gene expression data using co-inertia analysis BMC Bioinformatics 2003 4 59 14633289 10.1186/1471-2105-4-59
Shippy R Sendera TJ Lockner R Palaniappan C Kaysser-Kranich T Watts G Alsobrook J Performance evaluation of commercial short-oligonucleotide microarrays and the impact of noise in making cross-platform correlations BMC Genomics 2004 5 61 15345031 10.1186/1471-2164-5-61
Smid M Dorssers LC Jenster G Venn Mapping: clustering of heterologous microarray data based on the number of co-occurring differentially expressed genes Bioinformatics 2003 19 2065 2071 14594711 10.1093/bioinformatics/btg282
Cui X Churchill GA Statistical tests for differential expression in cDNA microarray experiments Genome Biol 2003 4 210 12702200 10.1186/gb-2003-4-4-210
Draghici S Statistical intelligence: effective analysis of high-density microarray data Drug Discov Today 2002 7 S55 S63 12047881 10.1016/S1359-6446(02)02292-4
Auto-Upload Tool Manual
Schaftenaar G Cuelenaere K Noordik JH Etzold T A Tcl-based SRS v. 4 interface Comput Appl Biosci 1996 12 151 155 8744778
Affymetrix
Tsai J Sultana R Lee Y Pertea G Karamycheva S Antonescu V Cho J Parvizi B Cheung F Quackenbush J RESOURCERER: a database for annotating and linking microarray resources within and across species Genome Biology 2001 2 software0002 16173164 10.1186/gb-2001-2-11-software0002
Cheung KH Hager J Pan D Srivastava R Mane S Li Y Miller P Williams KR KARMA: a web server application for comparing and annotating heterogeneous microarray platforms Nucleic Acids Res 2004 32 W441 W444 15215426 10.1093/nar/gkh661
Svensson BA Kreeft AJ van Ommen GJ den Dunnen JT Boer JM GeneHopper: a web-based search engine to link gene-expression platforms through GenBank accession numbers Genome Biol 2003 4 R35 12734015 10.1186/gb-2003-4-5-r35
Wang P Ding F Chiang H Thompson RC Watson SJ Meng F ProbeMatchDB – a web database for finding equivalent probes across microarray platforms and species Bioinformatics 2002 18 488 489 11934751 10.1093/bioinformatics/18.3.488
Dennis G JrSherman BT Hosack DA Yang J Gao W Lane HC Lempicki RA DAVID: Database for Annotation, Visualization, and Integrated Discovery Genome Biol 2003 4 P3 12734009 10.1186/gb-2003-4-5-p3
Kasprzyk A Keefe D Smedley D London D Spooner W Melsopp C Hammond M Rocca-Serra P Cox T Birney E EnsMart: a generic system for fast and flexible access to biological data Genome Res 2004 14 160 169 14707178 10.1101/gr.1645104
Diehn M Sherlock G Binkley G Jin H Matese JC Hernandez-Boussard T Rees CA Cherry JM Botstein D Brown PO Alizadeh AA SOURCE: a unified genomic resource of functional annotations, ontologies, and gene expression data Nucleic Acids Res 2003 31 219 223 12519986 10.1093/nar/gkg014
Bluthgen N Kielbasa SM Cajavec B Herzel H HOMGL-comparing genelists across species and with different accession numbers Bioinformatics 2004 20 125 126 14693820 10.1093/bioinformatics/btg379
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 32 D35 D40 14681353 10.1093/nar/gkh073
Alako BT Veldhoven A van Baal S Jelier R Verhoeven S Rullmann T Polman J Jenster G CoPub Mapper: mining MEDLINE based on search term co-publication BMC Bioinformatics 2005 6 51 15760478 10.1186/1471-2105-6-51
Smid M Dorssers LC GO-Mapper: functional analysis of gene expression data using the expression level as a score to evaluate Gene Ontology terms Bioinformatics 2004 20 2618 2625 15130934 10.1093/bioinformatics/bth293
Public SRS servers
NKI Central Microarray Facility
't Veer LJ Dai H van de Vijver MJ He YD Hart AA Mao M Peterse HL van der KK Marton MJ Witteveen AT Schreiber GJ Kerkhoven RM Roberts C Linsley PS Bernards R Friend SH Gene expression profiling predicts clinical outcome of breast cancer Nature 2002 415 530 536 11823860 10.1038/415530a
Dhanasekaran SM Barrette TR Ghosh D Shah R Varambally S Kurachi K Pienta KJ Rubin MA Chinnaiyan AM Delineation of prognostic biomarkers in prostate cancer Nature 2001 412 822 826 11518967 10.1038/35090585
Compugen Oligo Library
Lapointe J Li C Higgins JP Van de RM Bair E Montgomery K Ferrari M Egevad L Rayford W Bergerheim U Ekman P DeMarzo AM Tibshirani R Botstein D Brown PO Brooks JD Pollack JR Gene expression profiling identifies clinically relevant subtypes of prostate cancer Proc Natl Acad Sci U S A 2004 101 811 816 14711987 10.1073/pnas.0304146101
|
16048644
|
PMC1187877
|
CC BY
|
2021-01-04 16:27:25
|
no
|
BMC Bioinformatics. 2005 Jul 27; 6:192
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-192
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1931606096610.1186/1471-2105-6-193CorrectionAn empirical analysis of training protocols for probabilistic gene finders Majoros William H [email protected] Steven L [email protected] The Institute for Genomic Research, 9712 Medical Center Drive, Rockville, MD 20850, USA2005 1 8 2005 6 193 193 1 8 2005 1 8 2005 Copyright © 2005 Majoros and Salzberg; licensee BioMed Central Ltd.2005Majoros and Salzberg; 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.
==== Body
The summands in Equations (1)–(3) in the original paper [1] should be logarithmic. The corrected equations are given below.
Any references to these equations appearing in the text should be modified accordingly.
==== Refs
Majoros WM Salzberg SL An empirical analysis of training protocols for probabilistic gene finders BMC Bioinformatics 2004 5 206 15613242 10.1186/1471-2105-5-206
|
0
|
PMC1187878
|
CC BY
|
2021-01-04 16:27:25
|
no
|
BMC Bioinformatics. 2005 Aug 1; 6:193
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-193
|
oa_comm
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.