Genetic Modifiers of Lung Disease in Cystic Fibrosis
http://www.100md.com
《新英格兰医药杂志》
ABSTRACT
Background Polymorphisms in genes other than the cystic fibrosis transmembrane conductance regulator (CFTR) gene may modify the severity of pulmonary disease in patients with cystic fibrosis.
Methods We performed two studies with different patient samples. We first tested 808 patients who were homozygous for the F508 mutation and were classified as having either severe or mild lung disease, as defined by the lowest or highest quartile of forced expiratory volume in one second (FEV1), respectively, for age. We genotyped 16 polymorphisms in 10 genes reported by others as modifiers of disease severity in cystic fibrosis and tested for an association in patients with severe disease (263 patients) or mild disease (545). In the replication (second) study, we tested 498 patients, with various CFTR genotypes and a range of FEV1 values, for an association of the TGF1 codon 10 CC genotype with low FEV1.
Results In the initial study, significant allelic and genotypic associations with phenotype were seen only for TGF1 (the gene encoding transforming growth factor 1), particularly the –509 and codon 10 polymorphisms (with P values obtained with the use of Fisher's exact test and logistic regression ranging from 0.006 to 0.0002). The odds ratio was about 2.2 for the highest-risk TGF1 genotype (codon 10 CC) in association with the phenotype for severe lung disease. The replication study confirmed the association of the TGF1 codon 10 CC genotype with more severe lung disease in comparisons with the use of dichotomized FEV1 for severity status (P=0.0002) and FEV1 values directly (P=0.02).
Conclusions Genetic variation in the 5' end of TGF1 or a nearby upstream region modifies disease severity in cystic fibrosis.
Cystic fibrosis is a recessive genetic disorder that reflects mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) gene.1 Classic cystic fibrosis reflects two loss-of-function alleles, whereas nonclassic cystic fibrosis is characterized by at least one mutant CFTR allele that confers partial function and, in most cases, better survival.
There is great variability of pulmonary phenotype and survival in cystic fibrosis, even among patients who are homozygous for the most prevalent mutation, F508.1,2 Although environmental influences may modify clinical disease, there is probably additional genetic variation (i.e., the presence of "modifier" genes3) that contributes to the expression of the final phenotype.
To assess the effect of non-CFTR genetic polymorphisms on the clinical phenotype, we studied variants of 10 genes previously reported as modifiers in cystic fibrosis, including genes encoding 1-antiprotease (1AP)4,5,6,7,8; angiotensin-converting enzyme (ACE)9; 2-adrenergic receptor (ADR2)10; two glutathione S-transferases (GSTM1 and GSTP1)11,12,13; interleukin-10 (IL10)9; mannose-binding lectin 2 (MBL2)14,15,16,17,18; nitric oxide synthase 3 (NOS3)19; transforming growth factor 1 (TGF1)9,20; and tumor necrosis factor (TNF).9,12 Chloride conductance that is linked to genetic background may modulate the clinical phenotype in cystic fibrosis, but the specific loci and genes have not been identified.21
In the initial study, we tested patients who were homozygous for the F508 mutation and had one of the two extremes of phenotypes associated with lung function (i.e., severe or mild impairment) on the basis of the hypothesis that adverse and beneficial genetic variants would be enriched, respectively, in these two groups of disease severity. The classification of pulmonary function as severe or mild (i.e., the lowest or highest quartile of lung function for age) was confirmed by estimating final forced expiratory volume in one second (FEV1) for each patient on the basis of multiple spirometric measurements during the five years before enrollment (see the Supplementary Appendix, available with the full text of this article at www.nejm.org). Genotypes of the modifier variants were compared in the groups of patients with the severe or mild phenotype, in a manner similar to a case–control design. We replicated our findings in a different population of patients with cystic fibrosis.
Methods
Patients
For the initial study, the sample consisted of patients with cystic fibrosis, enrolled from 44 sites, who were homozygous for the F508 mutation. The 840 patients who were initially enrolled were chosen because their FEV1 measurements were in the lowest quartile or highest quartile for age among F508 homozygotes. The lung function in 275 of these patients was classified as severely impaired (lowest quartile), and that of 565 patients as mildly impaired (highest quartile). A total of 32 patients were excluded because they had inadequate spirograms (2 patients), were not homozygous for the F508 mutation (8), or did not have more than a 90 percent probability of having lung function congruent with that of others in the severe or mild category (22) (see Validation of Subjects in the Supplementary Appendix). There were 808 patients in the final data set. No patient was excluded because of race or ethnic background; 96.7 percent of the patients were self-identified as white.
In the replication study, we tested 498 patients whose sputum cultures were negative for Burkholderia cepacia and who had CFTR genotypes associated with pancreatic exocrine insufficiency; 70.5 percent were F508 homozygotes. As in the initial study, no patient was excluded because of race or ethnic background. The majority of the patients in the replication study were from a research study at the University of North Carolina at Chapel Hill, Case Western Reserve University in Cleveland, or the Hospital for Sick Children in Toronto (Supplementary Appendix). The study was approved by the biomedical institutional review board of the University of North Carolina and the institutional review board of each participating institution. Patients and parents of minors provided written informed consent.
Data Collection
For the initial study, each patient received a unique code that was used to allow data processing while maintaining anonymity. Key data were obtained from source documents, including pulmonary-function reports from the previous five years, which provided measurements of height and weight to calculate body-mass index as an index of nutrition, and sputum microbiologic testing during the previous three years. Other data were obtained on case-report forms. For the replication study, we used FEV1 data that were available at the sites (Supplementary Appendix).
Genotyping
For the initial study, genetic testing was performed by sequencing 1AP-Z and MBL2 B, C, and D "null" alleles; by single-nucleotide polymorphism (SNP) technology (BeadArray, Illumina) for 1AP-S and G1237A alleles, GSTP1, IL10, MBL2 promoter variants X and Y, NOS3, and TGF1; or by published methods for ACE, ADR2, GSTM1, and TNF (Supplementary Appendix). MBL2 structural (null) variants (B, C, and D) were combined to construct the O/O genotype. Other possible modifier genes for lung disease and their SNPs (65 and 135, respectively) were genotyped by Illumina but were not examined for the initial study. At least 798 patients were successfully genotyped for most genetic variants, except for four alleles: 1AP-Z (781 patients), two ADR2 alleles (741 and 743 patients, respectively), and TNF (743 patients). For regions flanking TGF1, we tested 31 SNPs at the Genome Analysis Facility at the University of North Carolina at Chapel Hill (TaqMan SNP Genotyping Assay by ABI-7900HT, Applied Biosystems) (Supplementary Appendix). After correction for multiple testing, none of the SNPs in the initial study showed significant overall departure from Hardy–Weinberg equilibrium.22 For the replication study, TGF1 codon 10 genotypes were determined by sequencing (at the University of North Carolina and Case Western Reserve University) and by allele-specific oligonucleotide testing (at the Hospital for Sick Children) (Supplementary Appendix).
Statistical Analysis
For the initial study, the association between polymorphisms and the phenotype for the severity of impairment of lung function was assessed with the use of Fisher's exact tests of genotype and allele frequencies. All tests were two-sided, with an alpha level of 0.05 considered to indicate statistical significance. Unadjusted P values are reported, whereas multiple-comparison corrections were performed for the 16 polymorphisms by recomputing the tests for 10,000 random permutations of severity-of-impairment status. The adjusted P value was based on the permutation distribution of the smallest P value among the 16.
For TGF1 polymorphisms, estimated haplotype reconstructions23 were compared with severity status to compute standard contingency table (chi-square) statistics. P values for these haplotypes were assessed with the use of 10,000 random permutations of severity status. Logistic regression was used to estimate the effect of TGF1 genotypes on the odds of a patient's having the phenotype for severe impairment under three genetic models (recessive, codominant, and dominant). Additional multivariate logistic-regression analyses included the covariates of the presence or absence of diabetes mellitus, Pseudomonas aeruginosa airway infection, meconium ileus, and a physician's diagnosis of asthma.24 Linkage disequilibrium patterns among TGF1 polymorphisms and 31 flanking SNPs were analyzed with the use of E-M estimation of haplotype frequencies, as implemented with the LDMAX program in the GOLD software package.25,26
For the replication study, we used results from the initial study to guide the design and to establish primary and secondary analytic approaches (Supplementary Appendix). Analysis of the initial study data suggested that the use of a dichotomized phenotype and a recessive model for an increased risk of severity of impairment due to the codon 10 C allele would provide the greatest power. The primary analysis for the replication study was performed with Fisher's exact test of association between the TGF1 codon 10 CC genotype and FEV1 status below or above a defined threshold (an FEV1 of 68 percent of the predicted value, estimated for the age of 20 years), with the use of a mixed linear regression model (Supplementary Appendix). In the secondary analysis in the replication study, we used a Wilcoxon test to compare FEV1 values for patients with the codon 10 CC genotype with those for patients with other (TC/TT) genotypes (Supplementary Appendix). As an additional conservative approach, two-sided P values were calculated for both primary and secondary analyses of the replication data. To reduce multiple comparisons, association tests were performed only on the entire sample in the replication study. Tests of Hardy–Weinberg equilibrium were also performed within groups that were classified according to the severity of impairment of lung function,22 following predictions that departures from Hardy–Weinberg equilibrium can arise with associated genes in case–control studies27 (Supplementary Appendix).
Results
Initial Study
Characteristics of Patients
The pulmonary and nutritional characteristics of 263 patients with the phenotype for severe impairment of lung function were distinctly different from those of 545 patients with the phenotype for mild impairment (Table 1). Even though patients in the group with the mild phenotype were approximately 12 years older than patients in the group with the severe phenotype, the average FEV1 at enrollment of patients with mild impairment was higher and the yearly decline in FEV1 was approximately one third of that in patients with severe impairment. The group with the mild phenotype had relatively preserved nutrition (on the basis of the body-mass index), whereas the group with the severe phenotype was malnourished. The two groups were similar with respect to the sex ratio; the presence or absence of P. aeruginosa in sputum; diabetes mellitus (when adjusted for age) or asthma; and enrollment site (86.4 percent of patients with the severe phenotype were matched by enrollment of one or more patients with the mild phenotype from the same site). When we divided the mild group into two age groups according to predetermined criteria (Supplementary Appendix), there were even more striking differences between the severe group and the younger mild group with respect to FEV1 and its rate of decline. The patients in the older mild group had FEV1 values similar to those of the severe group, despite being approximately 22 years older.
Table 1. Characteristics of 808 Patients with Cystic Fibrosis Homozygous for the F508 Mutation, According to the Phenotype for Severe or Mild Impairment of Lung Function and Age (Initial Study).
Genotype and Allelic Associations
In the analysis of previously reported genes (Table 2), significant associations with phenotype were seen only for TGF1 variants; the codon 10 variant had a multiple-comparison corrected P value of 0.01 (the most significant association among the 16 in Table 2). The frequencies of the genotypes and "minor" alleles for genetic variants tested were similar to those previously reported among white subjects (Table 2).4,5,6,7,8,9,10,11,12,13,14,15,19,20,29,30,31,32 In further analyses of subgroups, there were four significant or suggestive P values (between 0.01 and 0.10) for variants 1AP-Z and GSTM1 (Table 1 in the Supplementary Appendix); none of the P values were significant after multiple-testing corrections.
Table 2. Prevalence of Polymorphic Genotypes According to the Severe Phenotype (N=263) or the Mild Phenotype (N=545) (Initial Study).
The TGF1 variant genotypes were strongly associated with phenotype (Table 2). A statistical association was also seen in a permutation test of reconstructed haplotype frequencies for the three TGF1 SNPs (P=0.007). In tests of direct allelic association, only two TGF1 SNPs (–509 and codon 10) were significant (P=0.009 and P=0.001, respectively). The prevalence of two TGF1 genotypes in the severe and mild groups is shown in Figure 1. Patients with the phenotype for severe impairment of lung function were twice as likely to be homozygous for the TT genotype of –509 and the CC genotype of codon 10.
Figure 1. Prevalence and Association of TGF1 Genotypes in Groups Categorized According to Lung Impairment (Initial Study).
In Panel A, patients with the pulmonary phenotype for severe impairment of lung function (dark blue bars, representing 263 patients) were twice as likely to be homozygous for the TT genotype of –509 and the CC genotype of codon 10 as were patients with mild impairment (medium blue bars, representing 299 patients between the ages of 15 and 28 years, and light blue bars, representing 246 patients who were 29 years of age or more). P values, calculated with Fisher's exact test, are for the comparison between genotypes of patients with severe impairment of lung function and those of patients with mild impairment. Panel B shows P values (–log10 scale) according to genomic position of the three TGF1 variants that were initially tested and 31 single-nucleotide polymorphisms (SNPs) in flanking regions. Five additional genes in the region with described functions in the Swiss-Prot/TrEMBL protein-sequence database are shown for reference.
Further analyses were undertaken with the use of logistic regression for the two most significant TGF1 variants, with the use of three genetic models (recessive, codominant, or dominant) to test for the effect of the higher-risk genotype on the odds of having the severe phenotype (Table 3). The recessive and codominant models were most highly significant; for the recessive model, the odds ratios for genotype effects were 2.18 and 2.25 for the overall analysis, and they ranged from 1.58 to 3.16 for subgroup analyses by sex. Similar results were obtained after adjustment for four covariates: the presence or absence of diabetes mellitus, meconium ileus at birth, P. aeruginosa infection, and asthma (Table 2 in the Supplementary Appendix).
Table 3. Logistic-Regression Analysis of TGF1 Genetic Variants Associated with Severe Lung Disease among Patients with Severe Impairment (N = 260) and Mild Impairment (N = 544) (Initial Study).
SNPs Flanking TGF1 and Linkage Disequilibrium Patterns
We reasoned that testing additional SNPs around TGF1 would further elucidate the genetic association and linkage disequilibrium patterns (Figure 1 in the Supplementary Appendix). A total of 31 SNPs surrounding TGF1 were genotyped; Figure 1 displays the P values for the association of SNP genotypes with the severity of disease (–log10 scale), shown according to genomic position. The TGF1 variant codon 10 retained the strongest evidence of an association with severity. Of the flanking SNPs, only the two directly 5' of TGF1 (7381 and 7045, GenBank accession number NT_011109 ) were significant, and those SNPs were in strong linkage disequilibrium with –509 and codon 10 (Figure 1 of the Supplementary Appendix). The results indicate that these four SNPs belong to a haplotype "block" and are congruent with lower-resolution data from the International HapMap Consortium.33 The Hardy–Weinberg equilibrium tests within severity groups provided further evidence of the recessive action of the codon 10 CC genotype (Results section of the Supplementary Appendix). These data are consistent with an association between disease severity and one or more founding polymorphisms in the 5' end of TGF1 or immediately upstream.
Replication Study
Among patients in the replication study, the distribution of sex, age, and FEV1 were representative of a mixed pediatric and adult population of patients with cystic fibrosis (Table 4). Primary genetic analysis showed a strong association of the TGF1 codon 10 CC genotype with lower FEV1 values when patients were divided into two groups according to the FEV1 value with the use of criteria derived from the initial study (FEV1 at age 20, <68 percent of the predicted value or 68 percent) (Table 4 and Supplementary Appendix). The prevalence of the CC genotype was two times as great among patients with an FEV1 of less than 68 percent as among those with an FEV1 of 68 percent or more (P=0.0002). The result was not highly sensitive to the dichotomization threshold, since P values of less than 0.01 were achieved for FEV1 thresholds ranging from 65 to 74 percent of the predicted value. The secondary analysis also showed an association between codon 10 genotypes and FEV1 (P=0.02); specifically, the FEV1 was lower in patients with the CC genotype than in those with the TC/TT genotypes (62.8 percent vs. 68.2 percent of the predicted value). Similar results were seen among patients who were stratified according to whether they were homozygous for the F508 mutation (Table 4).
Table 4. Characteristics of Patients and TGF1 Codon 10 Genotypes Stratified According to FEV1 and Mean FEV1 According to Codon 10 Genotype (Replication Study).
Discussion
Studies have shown that multiple genetic polymorphisms act as modifiers of lung disease in cystic fibrosis, but these studies involved small numbers of patients, patients with a broad range of CFTR genotypes, or limited clinical phenotyping that did not address long-term outcome (survival).4,5,6,7,8,9,10,11,12,13,14,15,16,17,19,20 To increase the likelihood of identifying genetic modifiers that are relevant to the clinical outcome in cystic fibrosis, we performed two sequential studies in different samples of patients. The initial study tested 10 candidate modifier genes, and the replication study was performed to confirm TGF1 as an important modifier in cystic fibrosis.
The initial study used an extreme phenotype design (essentially case–control) with five key features. First, to reduce genetic heterogeneity, we studied only F508 homozygotes. Second, on the basis of the hypothesis that adverse (and beneficial) genetic variants would segregate with severe (and mild) lung disease, we studied patients at the extremes of lung function (lowest 25th percentile and the highest 25th percentile for age). Third, the severity of lung function was determined with the use of FEV1, the functional measure that best correlates with clinical status and outcome.34,35 Fourth, the classification of pulmonary status (severe vs. mild) was confirmed by estimating final FEV1 for each patient on the basis of previous measurements of FEV1 during the five years before enrollment. Finally, a large number of patients were enrolled in order to improve statistical power. The sample was genetically homogeneous for F508, and 96.7 percent of the patients were white; thus, population stratification is unlikely to have caused spurious association with the phenotype.36
The sample population and design of the replication study differed from the initial study in the following ways: most of the patients were from three sites (the University of North Carolina at Chapel Hill, Case Western Reserve University, or the Hospital for Sick Children, Toronto), there was a broad range of spirometric values (i.e., patients were not selected solely from the extremes of phenotype), and the CFTR genotypes were mixed "pancreatic insufficient" mutations (approximately 70 percent were homozygous for the F508 mutation). The primary analysis for the replication study was performed on the basis of the initial study, in which the greatest association of the CC genotype with phenotype was seen when a comparison was made between patients whose FEV1 values were below 68 percent of the predicted value and those whose values were 68 percent of the predicted value or more. Despite the differences in study design, sample population, and analytic approach, the replication study clearly confirmed that the TGF1 codon 10 CC genotype acted as an adverse modifier of lung disease in cystic fibrosis. Moreover, the magnitude of the effect (odds ratio, approximately 2.2) was similar to that in the initial study. The adverse TGF1 genotypes that we report differ from those found in two previous smaller studies in cystic fibrosis.9,20
In addition to the biologic plausibility of TGF1 as a modifier of lung disease (see below), recent association studies have linked TGF1 polymorphisms to asthma and chronic obstructive pulmonary disease.29,30,31,32 However, these studies involved relatively few polymorphisms, leaving open the possibility that the associations may be due to nearby causative genes in linkage disequilibrium with TGF1. Our inclusion of 31 SNPs flanking TGF1 greatly narrows the possibility of a modifier gene in the region (Figure 1). The association evidence includes the 5' end of TGF1 and only one other gene (MGC4093) of unknown structure and function, which lies between TGF1 (–509) and SNP number 7045. Transcripts of MGC4093 have been reported in UniGene libraries of lung tissue, but Celedon et al.32 argued that MGC4093 could be ruled out as a modifier of chronic obstructive pulmonary disease on the basis of associations observed between TGF1 and the disease in two samples of white patients. One sample showed the greatest evidence among 3' TGF1 SNPs, and the other sample showed the greatest evidence among the 5' SNPs, –509 and codon 10. The 3' SNPs exhibit little linkage disequilibrium with the 5' SNPs,33 and Celedon et al. concluded that their data implicated TGF1 alone. Our genetic associations and patterns of linkage disequilibrium are consistent with the report of Celedon et al., and we propose TGF1 as a likely modifier in another disease of the lung (i.e., cystic fibrosis).
The gene most likely to modify the clinical phenotype in our study, TGF1, has multiple functions related to growth and differentiation, immune responses, proinflammatory and antiinflammatory effects, and extracellular matrix production.37,38 TGF1 has been implicated in the pathogenesis of lung disease in animal models and humans, including disease progression in idiopathic pulmonary fibrosis, and in association with chronic obstructive pulmonary disease and asthma.29,30,31,32,39,40,41 The association of the TGF1 –509 T allele with asthma is postulated to involve increased TGF1 activity, reflecting increased Yin Yang 1 binding and promoter function and accompanied by higher circulating levels of TGF-1.30,31,42 The TGF1 polymorphisms in our study that are associated with the phenotype for severe impairment of lung function in cystic fibrosis are compatible with the hypothesis in asthma. Specifically, the –509 TT and the codon 10 CC genotypes correlate with an increase in gene expression, TGF-1 secretion, and circulating levels of TGF-1.43,44,45 The specific cellular- and organ-level pathophysiological mechanisms of increased TGF-1 activity in cystic fibrosis remain to be defined. However, this association will guide mechanistic studies and future strategies for therapeutic intervention.
In striking contrast, these same TGF1 alleles (–509 T and codon 10 C) are protective against chronic obstructive pulmonary disease induced by smoking.29,32 The magnitude of the increased prevalence of TGF1 genotypes as adverse modifiers in our studies of cystic fibrosis and in studies of asthma,30,31 and as protective modifiers in chronic obstructive pulmonary disease,29,32 is consistent — in other words, 12 to 14 percent prevalence for the –509 TT genotype and 19 to 22 percent for the codon 10 CC genotype. The contrast of TGF1 polymorphisms as adverse modifiers in our studies of cystic fibrosis and in asthma, as compared with the protective effect seen in smokers, indicates that the same genetic polymorphisms may be protective or adverse, depending on environmental and other genetic factors.
In summary, we used two study designs and enrolled a large number of patients to test genes previously implicated as modifiers in cystic fibrosis. Of these candidates, only TGF1 variants were strongly associated with pulmonary phenotypes that are predictive of the long-term outcome. These TGF1 polymorphisms are common in cystic fibrosis, and the odds ratios for an association of the higher-risk alleles with severe disease are relatively high (>2.0) for a contribution of genetic modifiers to a mendelian disorder. Thus, genetic variation in TGF1 or the immediate upstream region is an important genetic mechanism that modifies disease severity and clinical outcome in cystic fibrosis.
Supported by grants (CFF KNOWLE00A0 and CFF DRUMM00A0) from the Cystic Fibrosis Foundation, grants (HL68890, RR00046, and RR00059) from the National Institutes of Health, by Genome Canada through the Ontario Genomics Institute, by the Lloyd Carr-Harris Foundation, and by the Canadian Cystic Fibrosis Foundation.
We are indebted to Colette Bucur for assistance in coordination of the study; to Pam Davis for her thoughtful discussions of study design; to Bonnie Ramsey and members of the Cystic Fibrosis Foundation–sponsored Therapeutics Development Network for review of the original protocol and useful suggestions; to Larry Silverman for suggestions about genetic study design and protocols; to Wanda O'Neal for useful discussions and review of the manuscript; to Zaoqing Zhou for assisting with genotyping protocols; to Ken Friedman for the selection of SNPs; to the Program in Molecular Biology (and its director, Bill Marzluff) for developmental support; to Kellie Buchanan, Brett Buno, Blair Easton, Vonnie Mrva, and Joe Robinson for technical assistance; to Sherry Beecher for cell culture; to Kathy Hohneker and Joe Robinson for their assistance in patient enrollment; to the University of North Carolina facility (and its director, Steve Oglesbee) for lymphocyte transformation; to the University of North Carolina at Chapel Hill Genome Analysis Facility (and its director, Laura Livingston) for assistance in several aspects of genotyping; to the University of North Carolina Center for Bioinformatics (and David Fenstermacher, and Hemant Kelkar, director) for their assistance with design of the database and data management; to Jianhua Hu for statistical assistance; to the University of North Carolina Hospitals Molecular Genetics Laboratory (and its director, Jessica Booker) for assistance with DNA extraction and CFTR assays; to members of the Advisory Committee for the Gene Modifiers Study Group (including Larry Brody, Gary Cutting, Chris Penland, and Ben Wilfond); to Lap-Chee Tsui, Peter Paré, and Yves Berthiaume (coinvestigators on Genome Canada), and Roxanne Rousseau, Jennifer Breaton, and Mary Christofi (Canadian coordinators), for support and assistance in patient enrollment for the replication study; to Lisa Brown for editorial assistance; and to Beth Godwin for administrative support.
* Gene Modifier Study Group investigators are listed in the Appendix.
Source Information
From the Departments of Pediatrics (M.L.D., M.W.K., M.D.S., J.M.D., R.J.D.), Genetics (M.L.D.), and Epidemiology and Biostatistics (K.G.), Case Western Reserve University, Cleveland; the Cystic Fibrosis–Pulmonary Research and Treatment Center, School of Medicine (A.H., R.P., M.Z., J.R.Y., M.R.K.), the Department of Biostatistics, School of Public Health (F.Z., F.A.W.), and the Molecular Biology–Biotechnology Center for Bioinformatics (D.F., A.X.), University of North Carolina at Chapel Hill, Chapel Hill; the Program in Integrative Biology (P.D.), Program in Genetics and Genomic Biology (R.D., J.Z.), and Population Health Sciences (M.C.), Hospital for Sick Children, Toronto; and the James Hogg iCAPTURE Center for Cardiovascular and Pulmonary Research, University of British Columbia, Vancouver, B.C., Canada (A.J.S.).
Address reprint requests to Dr. Knowles at the Cystic Fibrosis–Pulmonary Research and Treatment Center, 7019 Thurston-Bowles Bldg., CB# 7248, University of North Carolina, Chapel Hill, NC 27599, or at knowles@med.unc.edu.
References
Welsh MJ, Ramsey BW, Accurso FJ, Cutting GR. Cystic fibrosis. In: Scriver CR, Beaudet AL, Sly WS, Valle D, eds. The metabolic and molecular bases of inherited disease. 8th ed. New York: McGraw-Hill, 2001:5121-88.
Kerem E, Corey M, Kerem B, et al. The relation between genotype and phenotype in cystic fibrosis -- analysis of the most common mutation (F508). N Engl J Med 1990;323:1517-1522.
Mekus F, Laabs U, Veeze H, Tummler B. Genes in the vicinity of CFTR modulate the cystic fibrosis phenotype in highly concordant or discordant F508del homozygous sib pairs. Hum Genet 2003;112:1-11.
Doring G, Krogh-Johansen H, Weidinger S, Hoiby N. Allotypes of alpha 1-antitrypsin in patients with cystic fibrosis, homozygous and heterozygous for deltaF508. Pediatr Pulmonol 1994;18:3-7.
Mahadeva R, Westerbeek RC, Perry DJ, et al. Alpha1-antitrypsin deficiency alleles and the Taq-I GA allele in cystic fibrosis lung disease. Eur Respir J 1998;11:873-879.
Mahadeva R, Stewart S, Bilton D, Lomas DA. Alpha-1 antitrypsin deficiency alleles and severe cystic fibrosis lung disease. Thorax 1998;53:1022-1024.
Frangolias DD, Ruan J, Wilcox PJ, et al. Alpha 1-antitrypsin deficiency alleles in cystic fibrosis lung disease. Am J Respir Cell Mol Biol 2003;29:390-396.
Henry MT, Cave S, Rendall J, et al. An alpha1-antitrypsin enhancer polymorphism is a genetic modifier of pulmonary outcome in cystic fibrosis. Eur J Hum Genet 2001;9:273-278.
Arkwright PD, Pravica V, Geraghty PJ, et al. End-organ dysfunction in cystic fibrosis: association with angiotensin I converting enzyme and cytokine gene polymorphisms. Am J Respir Crit Care Med 2003;167:384-389.
Buscher R, Eilmes KJ, Grasemann H, et al. 2 adrenoceptor gene polymorphisms in cystic fibrosis lung disease. Pharmacogenetics 2002;12:347-353.
Baranov VS, Ivaschenko T, Bakay B, et al. Proportion of the GSTM1 0/0 genotype in some Slavic populations and its correlation with cystic fibrosis and some multifactorial diseases. Hum Genet 1996;97:516-520.
Hull J, Thomson AH. Contribution of genetic factors other than CFTR to disease severity in cystic fibrosis. Thorax 1998;53:1018-1021.
Flamant C, Henrion-Caude A, Boelle PY, et al. Glutathione-S-transferase M1, M3, P1 and T1 polymorphisms and severity of lung disease in children with cystic fibrosis. Pharmacogenetics 2004;14:295-301.
Garred P, Pressler T, Madsen HO, et al. Association of mannose-binding lectin gene heterogeneity with severity of lung disease and survival in cystic fibrosis. J Clin Invest 1999;104:431-437.
Gabolde M, Guilloud-Bataille M, Feingold J, Besmond C. Association of variant alleles of mannose binding lectin with severity of pulmonary disease in cystic fibrosis: cohort study. BMJ 1999;319:1166-1167.
Davies JC, Turner MW, Klein N. Impaired pulmonary status in cystic fibrosis adults with two mutated MBL-2 alleles. Eur Respir J 2004;24:798-804.
Carlsson M, Sjoholm AG, Eriksson L, et al. Deficiency of the mannan-binding lectin pathway of complement and poor outcome in cystic fibrosis: bacterial colonization may be decisive for a relationship. Clin Exp Immunol 2005;139:306-313.
Yarden J, Radojkovic D, De Boeck K, et al. Polymorphisms in the mannose binding lectin gene affect the cystic fibrosis pulmonary phenotype. J Med Genet 2004;41:629-633.
Grasemann H, van's Gravesande KS, Buscher R, et al. Endothelial nitric oxide synthase variants in cystic fibrosis lung disease. Am J Respir Crit Care Med 2003;167:390-394.
Arkwright PD, Laurie S, Super M, et al. TGF-beta(1) genotype and accelerated decline in lung function of patients with cystic fibrosis. Thorax 2000;55:459-462.
Bronsveld I, Mekus F, Bijman J, et al. Chloride conductance and genetic background modulate the cystic fibrosis phenotype of Delta F508 homozygous twins and siblings. J Clin Invest 2001;108:1705-1715.
Weir BS. Genetic data analysis II: methods for discrete population genetic data. Sunderland, Mass.: Sinauer Associates, 1996.
Qin ZS, Niu T, Liu JS. Partition-ligation-expectation-maximization algorithm for haplotype inference with single-nucleotide polymorphisms. Am J Hum Genet 2002;71:1242-1247.
Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. New York: Wiley, 2000.
Excoffier L, Slatkin M. Maximum-likelihood estimation of molecular haplotype frequencies in a diploid population. Mol Biol Evol 1995;12:921-927.
Abecasis GR, Cookson WO. GOLD -- graphical overview of linkage disequilibrium. Bioinformatics 2000;16:182-183.
Nielsen DM, Ehm MG, Weir BS. Detecting marker-disease association by testing for Hardy-Weinberg disequilibrium at a marker locus. Am J Hum Genet 1998;63:1531-1540.
Schluchter MD, Konstan MW, Davis PB. Jointly modeling the relationship between survival and pulmonary function in cystic fibrosis patients. Stat Med 2002;21:1271-1287.
Wu L, Chau J, Young RP, et al. Transforming growth factor-beta1 genotype and susceptibility to chronic obstructive pulmonary disease. Thorax 2004;59:126-129.
Pulleyn LJ, Newton R, Adcock IM, Barnes PJ. TGFbeta1 allele association with asthma severity. Hum Genet 2001;109:623-627.
Silverman ES, Palmer LJ, Subramaniam V, et al. Transforming growth factor-beta1 promoter polymorphism C-509T is associated with asthma. Am J Respir Crit Care Med 2004;169:214-219.
Celedon JC, Lange C, Raby BA, et al. The transforming growth factor-beta1 (TGFB1) gene is associated with chronic obstructive pulmonary disease (COPD). Hum Mol Genet 2004;13:1649-1656.
The International HapMap Consortium. The International HapMap Project. Nature 2003;426:789-796.
Kerem E, Reisman J, Corey M, Canny GJ, Levison H. Prediction of mortality in patients with cystic fibrosis. N Engl J Med 1992;326:1187-1191.
Corey M, Edwards L, Levison H, Knowles M. Longitudinal analysis of pulmonary function decline in patients with cystic fibrosis. J Pediatr 1997;131:809-814.
Cardon LR, Palmer LJ. Population stratification and spurious allelic association. Lancet 2003;361:598-604.
Blobe GC, Schiemann WP, Lodish HF. Role of transforming growth factor beta in human disease. N Engl J Med 2000;342:1350-1358.
Akhurst RJ. TGF beta signaling in health and disease. Nat Genet 2004;36:790-792.
Xaubet A, Marin-Arguedas A, Lario S, et al. Transforming growth factor-beta1 gene polymorphisms are associated with disease progression in idiopathic pulmonary fibrosis. Am J Respir Crit Care Med 2003;168:431-435.
Morris DG, Huang X, Kaminski N, et al. Loss of integrin alpha(v)beta6-mediated TGF-beta activation causes MMP12-dependent emphysema. Nature 2003;422:169-173.
Sterner-Kock A, Thorey IS, Koli K, et al. Disruption of the gene encoding the latent transforming growth factor-beta binding protein 4 (LTBP-4) causes abnormal lung development, cardiomyopathy, and colorectal cancer. Genes Dev 2002;16:2264-2273.
Grainger DJ, Heathcote K, Chiano M, et al. Genetic control of the circulating concentration of transforming growth factor type beta1. Hum Mol Genet 1999;8:93-97.
Yamada Y, Miyauchi A, Goto J, et al. Association of a polymorphism of the transforming growth factor-beta1 gene with genetic susceptibility to osteoporosis in postmenopausal Japanese women. J Bone Miner Res 1998;13:1569-1576.
Suthanthiran M, Li B, Song JO, et al. Transforming growth factor-beta 1 hyperexpression in African-American hypertensives: a novel mediator of hypertension and/or target organ damage. Proc Natl Acad Sci U S A 2000;97:3479-3484.
Dunning AM, Ellis PD, McBride S, et al. A transforming growth factor beta1 signal peptide variant increases secretion in vitro and is associated with increased incidence of invasive breast cancer. Cancer Res 2003;63:2610-2615.(Mitchell L. Drumm, Ph.D.,)
Background Polymorphisms in genes other than the cystic fibrosis transmembrane conductance regulator (CFTR) gene may modify the severity of pulmonary disease in patients with cystic fibrosis.
Methods We performed two studies with different patient samples. We first tested 808 patients who were homozygous for the F508 mutation and were classified as having either severe or mild lung disease, as defined by the lowest or highest quartile of forced expiratory volume in one second (FEV1), respectively, for age. We genotyped 16 polymorphisms in 10 genes reported by others as modifiers of disease severity in cystic fibrosis and tested for an association in patients with severe disease (263 patients) or mild disease (545). In the replication (second) study, we tested 498 patients, with various CFTR genotypes and a range of FEV1 values, for an association of the TGF1 codon 10 CC genotype with low FEV1.
Results In the initial study, significant allelic and genotypic associations with phenotype were seen only for TGF1 (the gene encoding transforming growth factor 1), particularly the –509 and codon 10 polymorphisms (with P values obtained with the use of Fisher's exact test and logistic regression ranging from 0.006 to 0.0002). The odds ratio was about 2.2 for the highest-risk TGF1 genotype (codon 10 CC) in association with the phenotype for severe lung disease. The replication study confirmed the association of the TGF1 codon 10 CC genotype with more severe lung disease in comparisons with the use of dichotomized FEV1 for severity status (P=0.0002) and FEV1 values directly (P=0.02).
Conclusions Genetic variation in the 5' end of TGF1 or a nearby upstream region modifies disease severity in cystic fibrosis.
Cystic fibrosis is a recessive genetic disorder that reflects mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) gene.1 Classic cystic fibrosis reflects two loss-of-function alleles, whereas nonclassic cystic fibrosis is characterized by at least one mutant CFTR allele that confers partial function and, in most cases, better survival.
There is great variability of pulmonary phenotype and survival in cystic fibrosis, even among patients who are homozygous for the most prevalent mutation, F508.1,2 Although environmental influences may modify clinical disease, there is probably additional genetic variation (i.e., the presence of "modifier" genes3) that contributes to the expression of the final phenotype.
To assess the effect of non-CFTR genetic polymorphisms on the clinical phenotype, we studied variants of 10 genes previously reported as modifiers in cystic fibrosis, including genes encoding 1-antiprotease (1AP)4,5,6,7,8; angiotensin-converting enzyme (ACE)9; 2-adrenergic receptor (ADR2)10; two glutathione S-transferases (GSTM1 and GSTP1)11,12,13; interleukin-10 (IL10)9; mannose-binding lectin 2 (MBL2)14,15,16,17,18; nitric oxide synthase 3 (NOS3)19; transforming growth factor 1 (TGF1)9,20; and tumor necrosis factor (TNF).9,12 Chloride conductance that is linked to genetic background may modulate the clinical phenotype in cystic fibrosis, but the specific loci and genes have not been identified.21
In the initial study, we tested patients who were homozygous for the F508 mutation and had one of the two extremes of phenotypes associated with lung function (i.e., severe or mild impairment) on the basis of the hypothesis that adverse and beneficial genetic variants would be enriched, respectively, in these two groups of disease severity. The classification of pulmonary function as severe or mild (i.e., the lowest or highest quartile of lung function for age) was confirmed by estimating final forced expiratory volume in one second (FEV1) for each patient on the basis of multiple spirometric measurements during the five years before enrollment (see the Supplementary Appendix, available with the full text of this article at www.nejm.org). Genotypes of the modifier variants were compared in the groups of patients with the severe or mild phenotype, in a manner similar to a case–control design. We replicated our findings in a different population of patients with cystic fibrosis.
Methods
Patients
For the initial study, the sample consisted of patients with cystic fibrosis, enrolled from 44 sites, who were homozygous for the F508 mutation. The 840 patients who were initially enrolled were chosen because their FEV1 measurements were in the lowest quartile or highest quartile for age among F508 homozygotes. The lung function in 275 of these patients was classified as severely impaired (lowest quartile), and that of 565 patients as mildly impaired (highest quartile). A total of 32 patients were excluded because they had inadequate spirograms (2 patients), were not homozygous for the F508 mutation (8), or did not have more than a 90 percent probability of having lung function congruent with that of others in the severe or mild category (22) (see Validation of Subjects in the Supplementary Appendix). There were 808 patients in the final data set. No patient was excluded because of race or ethnic background; 96.7 percent of the patients were self-identified as white.
In the replication study, we tested 498 patients whose sputum cultures were negative for Burkholderia cepacia and who had CFTR genotypes associated with pancreatic exocrine insufficiency; 70.5 percent were F508 homozygotes. As in the initial study, no patient was excluded because of race or ethnic background. The majority of the patients in the replication study were from a research study at the University of North Carolina at Chapel Hill, Case Western Reserve University in Cleveland, or the Hospital for Sick Children in Toronto (Supplementary Appendix). The study was approved by the biomedical institutional review board of the University of North Carolina and the institutional review board of each participating institution. Patients and parents of minors provided written informed consent.
Data Collection
For the initial study, each patient received a unique code that was used to allow data processing while maintaining anonymity. Key data were obtained from source documents, including pulmonary-function reports from the previous five years, which provided measurements of height and weight to calculate body-mass index as an index of nutrition, and sputum microbiologic testing during the previous three years. Other data were obtained on case-report forms. For the replication study, we used FEV1 data that were available at the sites (Supplementary Appendix).
Genotyping
For the initial study, genetic testing was performed by sequencing 1AP-Z and MBL2 B, C, and D "null" alleles; by single-nucleotide polymorphism (SNP) technology (BeadArray, Illumina) for 1AP-S and G1237A alleles, GSTP1, IL10, MBL2 promoter variants X and Y, NOS3, and TGF1; or by published methods for ACE, ADR2, GSTM1, and TNF (Supplementary Appendix). MBL2 structural (null) variants (B, C, and D) were combined to construct the O/O genotype. Other possible modifier genes for lung disease and their SNPs (65 and 135, respectively) were genotyped by Illumina but were not examined for the initial study. At least 798 patients were successfully genotyped for most genetic variants, except for four alleles: 1AP-Z (781 patients), two ADR2 alleles (741 and 743 patients, respectively), and TNF (743 patients). For regions flanking TGF1, we tested 31 SNPs at the Genome Analysis Facility at the University of North Carolina at Chapel Hill (TaqMan SNP Genotyping Assay by ABI-7900HT, Applied Biosystems) (Supplementary Appendix). After correction for multiple testing, none of the SNPs in the initial study showed significant overall departure from Hardy–Weinberg equilibrium.22 For the replication study, TGF1 codon 10 genotypes were determined by sequencing (at the University of North Carolina and Case Western Reserve University) and by allele-specific oligonucleotide testing (at the Hospital for Sick Children) (Supplementary Appendix).
Statistical Analysis
For the initial study, the association between polymorphisms and the phenotype for the severity of impairment of lung function was assessed with the use of Fisher's exact tests of genotype and allele frequencies. All tests were two-sided, with an alpha level of 0.05 considered to indicate statistical significance. Unadjusted P values are reported, whereas multiple-comparison corrections were performed for the 16 polymorphisms by recomputing the tests for 10,000 random permutations of severity-of-impairment status. The adjusted P value was based on the permutation distribution of the smallest P value among the 16.
For TGF1 polymorphisms, estimated haplotype reconstructions23 were compared with severity status to compute standard contingency table (chi-square) statistics. P values for these haplotypes were assessed with the use of 10,000 random permutations of severity status. Logistic regression was used to estimate the effect of TGF1 genotypes on the odds of a patient's having the phenotype for severe impairment under three genetic models (recessive, codominant, and dominant). Additional multivariate logistic-regression analyses included the covariates of the presence or absence of diabetes mellitus, Pseudomonas aeruginosa airway infection, meconium ileus, and a physician's diagnosis of asthma.24 Linkage disequilibrium patterns among TGF1 polymorphisms and 31 flanking SNPs were analyzed with the use of E-M estimation of haplotype frequencies, as implemented with the LDMAX program in the GOLD software package.25,26
For the replication study, we used results from the initial study to guide the design and to establish primary and secondary analytic approaches (Supplementary Appendix). Analysis of the initial study data suggested that the use of a dichotomized phenotype and a recessive model for an increased risk of severity of impairment due to the codon 10 C allele would provide the greatest power. The primary analysis for the replication study was performed with Fisher's exact test of association between the TGF1 codon 10 CC genotype and FEV1 status below or above a defined threshold (an FEV1 of 68 percent of the predicted value, estimated for the age of 20 years), with the use of a mixed linear regression model (Supplementary Appendix). In the secondary analysis in the replication study, we used a Wilcoxon test to compare FEV1 values for patients with the codon 10 CC genotype with those for patients with other (TC/TT) genotypes (Supplementary Appendix). As an additional conservative approach, two-sided P values were calculated for both primary and secondary analyses of the replication data. To reduce multiple comparisons, association tests were performed only on the entire sample in the replication study. Tests of Hardy–Weinberg equilibrium were also performed within groups that were classified according to the severity of impairment of lung function,22 following predictions that departures from Hardy–Weinberg equilibrium can arise with associated genes in case–control studies27 (Supplementary Appendix).
Results
Initial Study
Characteristics of Patients
The pulmonary and nutritional characteristics of 263 patients with the phenotype for severe impairment of lung function were distinctly different from those of 545 patients with the phenotype for mild impairment (Table 1). Even though patients in the group with the mild phenotype were approximately 12 years older than patients in the group with the severe phenotype, the average FEV1 at enrollment of patients with mild impairment was higher and the yearly decline in FEV1 was approximately one third of that in patients with severe impairment. The group with the mild phenotype had relatively preserved nutrition (on the basis of the body-mass index), whereas the group with the severe phenotype was malnourished. The two groups were similar with respect to the sex ratio; the presence or absence of P. aeruginosa in sputum; diabetes mellitus (when adjusted for age) or asthma; and enrollment site (86.4 percent of patients with the severe phenotype were matched by enrollment of one or more patients with the mild phenotype from the same site). When we divided the mild group into two age groups according to predetermined criteria (Supplementary Appendix), there were even more striking differences between the severe group and the younger mild group with respect to FEV1 and its rate of decline. The patients in the older mild group had FEV1 values similar to those of the severe group, despite being approximately 22 years older.
Table 1. Characteristics of 808 Patients with Cystic Fibrosis Homozygous for the F508 Mutation, According to the Phenotype for Severe or Mild Impairment of Lung Function and Age (Initial Study).
Genotype and Allelic Associations
In the analysis of previously reported genes (Table 2), significant associations with phenotype were seen only for TGF1 variants; the codon 10 variant had a multiple-comparison corrected P value of 0.01 (the most significant association among the 16 in Table 2). The frequencies of the genotypes and "minor" alleles for genetic variants tested were similar to those previously reported among white subjects (Table 2).4,5,6,7,8,9,10,11,12,13,14,15,19,20,29,30,31,32 In further analyses of subgroups, there were four significant or suggestive P values (between 0.01 and 0.10) for variants 1AP-Z and GSTM1 (Table 1 in the Supplementary Appendix); none of the P values were significant after multiple-testing corrections.
Table 2. Prevalence of Polymorphic Genotypes According to the Severe Phenotype (N=263) or the Mild Phenotype (N=545) (Initial Study).
The TGF1 variant genotypes were strongly associated with phenotype (Table 2). A statistical association was also seen in a permutation test of reconstructed haplotype frequencies for the three TGF1 SNPs (P=0.007). In tests of direct allelic association, only two TGF1 SNPs (–509 and codon 10) were significant (P=0.009 and P=0.001, respectively). The prevalence of two TGF1 genotypes in the severe and mild groups is shown in Figure 1. Patients with the phenotype for severe impairment of lung function were twice as likely to be homozygous for the TT genotype of –509 and the CC genotype of codon 10.
Figure 1. Prevalence and Association of TGF1 Genotypes in Groups Categorized According to Lung Impairment (Initial Study).
In Panel A, patients with the pulmonary phenotype for severe impairment of lung function (dark blue bars, representing 263 patients) were twice as likely to be homozygous for the TT genotype of –509 and the CC genotype of codon 10 as were patients with mild impairment (medium blue bars, representing 299 patients between the ages of 15 and 28 years, and light blue bars, representing 246 patients who were 29 years of age or more). P values, calculated with Fisher's exact test, are for the comparison between genotypes of patients with severe impairment of lung function and those of patients with mild impairment. Panel B shows P values (–log10 scale) according to genomic position of the three TGF1 variants that were initially tested and 31 single-nucleotide polymorphisms (SNPs) in flanking regions. Five additional genes in the region with described functions in the Swiss-Prot/TrEMBL protein-sequence database are shown for reference.
Further analyses were undertaken with the use of logistic regression for the two most significant TGF1 variants, with the use of three genetic models (recessive, codominant, or dominant) to test for the effect of the higher-risk genotype on the odds of having the severe phenotype (Table 3). The recessive and codominant models were most highly significant; for the recessive model, the odds ratios for genotype effects were 2.18 and 2.25 for the overall analysis, and they ranged from 1.58 to 3.16 for subgroup analyses by sex. Similar results were obtained after adjustment for four covariates: the presence or absence of diabetes mellitus, meconium ileus at birth, P. aeruginosa infection, and asthma (Table 2 in the Supplementary Appendix).
Table 3. Logistic-Regression Analysis of TGF1 Genetic Variants Associated with Severe Lung Disease among Patients with Severe Impairment (N = 260) and Mild Impairment (N = 544) (Initial Study).
SNPs Flanking TGF1 and Linkage Disequilibrium Patterns
We reasoned that testing additional SNPs around TGF1 would further elucidate the genetic association and linkage disequilibrium patterns (Figure 1 in the Supplementary Appendix). A total of 31 SNPs surrounding TGF1 were genotyped; Figure 1 displays the P values for the association of SNP genotypes with the severity of disease (–log10 scale), shown according to genomic position. The TGF1 variant codon 10 retained the strongest evidence of an association with severity. Of the flanking SNPs, only the two directly 5' of TGF1 (7381 and 7045, GenBank accession number NT_011109 ) were significant, and those SNPs were in strong linkage disequilibrium with –509 and codon 10 (Figure 1 of the Supplementary Appendix). The results indicate that these four SNPs belong to a haplotype "block" and are congruent with lower-resolution data from the International HapMap Consortium.33 The Hardy–Weinberg equilibrium tests within severity groups provided further evidence of the recessive action of the codon 10 CC genotype (Results section of the Supplementary Appendix). These data are consistent with an association between disease severity and one or more founding polymorphisms in the 5' end of TGF1 or immediately upstream.
Replication Study
Among patients in the replication study, the distribution of sex, age, and FEV1 were representative of a mixed pediatric and adult population of patients with cystic fibrosis (Table 4). Primary genetic analysis showed a strong association of the TGF1 codon 10 CC genotype with lower FEV1 values when patients were divided into two groups according to the FEV1 value with the use of criteria derived from the initial study (FEV1 at age 20, <68 percent of the predicted value or 68 percent) (Table 4 and Supplementary Appendix). The prevalence of the CC genotype was two times as great among patients with an FEV1 of less than 68 percent as among those with an FEV1 of 68 percent or more (P=0.0002). The result was not highly sensitive to the dichotomization threshold, since P values of less than 0.01 were achieved for FEV1 thresholds ranging from 65 to 74 percent of the predicted value. The secondary analysis also showed an association between codon 10 genotypes and FEV1 (P=0.02); specifically, the FEV1 was lower in patients with the CC genotype than in those with the TC/TT genotypes (62.8 percent vs. 68.2 percent of the predicted value). Similar results were seen among patients who were stratified according to whether they were homozygous for the F508 mutation (Table 4).
Table 4. Characteristics of Patients and TGF1 Codon 10 Genotypes Stratified According to FEV1 and Mean FEV1 According to Codon 10 Genotype (Replication Study).
Discussion
Studies have shown that multiple genetic polymorphisms act as modifiers of lung disease in cystic fibrosis, but these studies involved small numbers of patients, patients with a broad range of CFTR genotypes, or limited clinical phenotyping that did not address long-term outcome (survival).4,5,6,7,8,9,10,11,12,13,14,15,16,17,19,20 To increase the likelihood of identifying genetic modifiers that are relevant to the clinical outcome in cystic fibrosis, we performed two sequential studies in different samples of patients. The initial study tested 10 candidate modifier genes, and the replication study was performed to confirm TGF1 as an important modifier in cystic fibrosis.
The initial study used an extreme phenotype design (essentially case–control) with five key features. First, to reduce genetic heterogeneity, we studied only F508 homozygotes. Second, on the basis of the hypothesis that adverse (and beneficial) genetic variants would segregate with severe (and mild) lung disease, we studied patients at the extremes of lung function (lowest 25th percentile and the highest 25th percentile for age). Third, the severity of lung function was determined with the use of FEV1, the functional measure that best correlates with clinical status and outcome.34,35 Fourth, the classification of pulmonary status (severe vs. mild) was confirmed by estimating final FEV1 for each patient on the basis of previous measurements of FEV1 during the five years before enrollment. Finally, a large number of patients were enrolled in order to improve statistical power. The sample was genetically homogeneous for F508, and 96.7 percent of the patients were white; thus, population stratification is unlikely to have caused spurious association with the phenotype.36
The sample population and design of the replication study differed from the initial study in the following ways: most of the patients were from three sites (the University of North Carolina at Chapel Hill, Case Western Reserve University, or the Hospital for Sick Children, Toronto), there was a broad range of spirometric values (i.e., patients were not selected solely from the extremes of phenotype), and the CFTR genotypes were mixed "pancreatic insufficient" mutations (approximately 70 percent were homozygous for the F508 mutation). The primary analysis for the replication study was performed on the basis of the initial study, in which the greatest association of the CC genotype with phenotype was seen when a comparison was made between patients whose FEV1 values were below 68 percent of the predicted value and those whose values were 68 percent of the predicted value or more. Despite the differences in study design, sample population, and analytic approach, the replication study clearly confirmed that the TGF1 codon 10 CC genotype acted as an adverse modifier of lung disease in cystic fibrosis. Moreover, the magnitude of the effect (odds ratio, approximately 2.2) was similar to that in the initial study. The adverse TGF1 genotypes that we report differ from those found in two previous smaller studies in cystic fibrosis.9,20
In addition to the biologic plausibility of TGF1 as a modifier of lung disease (see below), recent association studies have linked TGF1 polymorphisms to asthma and chronic obstructive pulmonary disease.29,30,31,32 However, these studies involved relatively few polymorphisms, leaving open the possibility that the associations may be due to nearby causative genes in linkage disequilibrium with TGF1. Our inclusion of 31 SNPs flanking TGF1 greatly narrows the possibility of a modifier gene in the region (Figure 1). The association evidence includes the 5' end of TGF1 and only one other gene (MGC4093) of unknown structure and function, which lies between TGF1 (–509) and SNP number 7045. Transcripts of MGC4093 have been reported in UniGene libraries of lung tissue, but Celedon et al.32 argued that MGC4093 could be ruled out as a modifier of chronic obstructive pulmonary disease on the basis of associations observed between TGF1 and the disease in two samples of white patients. One sample showed the greatest evidence among 3' TGF1 SNPs, and the other sample showed the greatest evidence among the 5' SNPs, –509 and codon 10. The 3' SNPs exhibit little linkage disequilibrium with the 5' SNPs,33 and Celedon et al. concluded that their data implicated TGF1 alone. Our genetic associations and patterns of linkage disequilibrium are consistent with the report of Celedon et al., and we propose TGF1 as a likely modifier in another disease of the lung (i.e., cystic fibrosis).
The gene most likely to modify the clinical phenotype in our study, TGF1, has multiple functions related to growth and differentiation, immune responses, proinflammatory and antiinflammatory effects, and extracellular matrix production.37,38 TGF1 has been implicated in the pathogenesis of lung disease in animal models and humans, including disease progression in idiopathic pulmonary fibrosis, and in association with chronic obstructive pulmonary disease and asthma.29,30,31,32,39,40,41 The association of the TGF1 –509 T allele with asthma is postulated to involve increased TGF1 activity, reflecting increased Yin Yang 1 binding and promoter function and accompanied by higher circulating levels of TGF-1.30,31,42 The TGF1 polymorphisms in our study that are associated with the phenotype for severe impairment of lung function in cystic fibrosis are compatible with the hypothesis in asthma. Specifically, the –509 TT and the codon 10 CC genotypes correlate with an increase in gene expression, TGF-1 secretion, and circulating levels of TGF-1.43,44,45 The specific cellular- and organ-level pathophysiological mechanisms of increased TGF-1 activity in cystic fibrosis remain to be defined. However, this association will guide mechanistic studies and future strategies for therapeutic intervention.
In striking contrast, these same TGF1 alleles (–509 T and codon 10 C) are protective against chronic obstructive pulmonary disease induced by smoking.29,32 The magnitude of the increased prevalence of TGF1 genotypes as adverse modifiers in our studies of cystic fibrosis and in studies of asthma,30,31 and as protective modifiers in chronic obstructive pulmonary disease,29,32 is consistent — in other words, 12 to 14 percent prevalence for the –509 TT genotype and 19 to 22 percent for the codon 10 CC genotype. The contrast of TGF1 polymorphisms as adverse modifiers in our studies of cystic fibrosis and in asthma, as compared with the protective effect seen in smokers, indicates that the same genetic polymorphisms may be protective or adverse, depending on environmental and other genetic factors.
In summary, we used two study designs and enrolled a large number of patients to test genes previously implicated as modifiers in cystic fibrosis. Of these candidates, only TGF1 variants were strongly associated with pulmonary phenotypes that are predictive of the long-term outcome. These TGF1 polymorphisms are common in cystic fibrosis, and the odds ratios for an association of the higher-risk alleles with severe disease are relatively high (>2.0) for a contribution of genetic modifiers to a mendelian disorder. Thus, genetic variation in TGF1 or the immediate upstream region is an important genetic mechanism that modifies disease severity and clinical outcome in cystic fibrosis.
Supported by grants (CFF KNOWLE00A0 and CFF DRUMM00A0) from the Cystic Fibrosis Foundation, grants (HL68890, RR00046, and RR00059) from the National Institutes of Health, by Genome Canada through the Ontario Genomics Institute, by the Lloyd Carr-Harris Foundation, and by the Canadian Cystic Fibrosis Foundation.
We are indebted to Colette Bucur for assistance in coordination of the study; to Pam Davis for her thoughtful discussions of study design; to Bonnie Ramsey and members of the Cystic Fibrosis Foundation–sponsored Therapeutics Development Network for review of the original protocol and useful suggestions; to Larry Silverman for suggestions about genetic study design and protocols; to Wanda O'Neal for useful discussions and review of the manuscript; to Zaoqing Zhou for assisting with genotyping protocols; to Ken Friedman for the selection of SNPs; to the Program in Molecular Biology (and its director, Bill Marzluff) for developmental support; to Kellie Buchanan, Brett Buno, Blair Easton, Vonnie Mrva, and Joe Robinson for technical assistance; to Sherry Beecher for cell culture; to Kathy Hohneker and Joe Robinson for their assistance in patient enrollment; to the University of North Carolina facility (and its director, Steve Oglesbee) for lymphocyte transformation; to the University of North Carolina at Chapel Hill Genome Analysis Facility (and its director, Laura Livingston) for assistance in several aspects of genotyping; to the University of North Carolina Center for Bioinformatics (and David Fenstermacher, and Hemant Kelkar, director) for their assistance with design of the database and data management; to Jianhua Hu for statistical assistance; to the University of North Carolina Hospitals Molecular Genetics Laboratory (and its director, Jessica Booker) for assistance with DNA extraction and CFTR assays; to members of the Advisory Committee for the Gene Modifiers Study Group (including Larry Brody, Gary Cutting, Chris Penland, and Ben Wilfond); to Lap-Chee Tsui, Peter Paré, and Yves Berthiaume (coinvestigators on Genome Canada), and Roxanne Rousseau, Jennifer Breaton, and Mary Christofi (Canadian coordinators), for support and assistance in patient enrollment for the replication study; to Lisa Brown for editorial assistance; and to Beth Godwin for administrative support.
* Gene Modifier Study Group investigators are listed in the Appendix.
Source Information
From the Departments of Pediatrics (M.L.D., M.W.K., M.D.S., J.M.D., R.J.D.), Genetics (M.L.D.), and Epidemiology and Biostatistics (K.G.), Case Western Reserve University, Cleveland; the Cystic Fibrosis–Pulmonary Research and Treatment Center, School of Medicine (A.H., R.P., M.Z., J.R.Y., M.R.K.), the Department of Biostatistics, School of Public Health (F.Z., F.A.W.), and the Molecular Biology–Biotechnology Center for Bioinformatics (D.F., A.X.), University of North Carolina at Chapel Hill, Chapel Hill; the Program in Integrative Biology (P.D.), Program in Genetics and Genomic Biology (R.D., J.Z.), and Population Health Sciences (M.C.), Hospital for Sick Children, Toronto; and the James Hogg iCAPTURE Center for Cardiovascular and Pulmonary Research, University of British Columbia, Vancouver, B.C., Canada (A.J.S.).
Address reprint requests to Dr. Knowles at the Cystic Fibrosis–Pulmonary Research and Treatment Center, 7019 Thurston-Bowles Bldg., CB# 7248, University of North Carolina, Chapel Hill, NC 27599, or at knowles@med.unc.edu.
References
Welsh MJ, Ramsey BW, Accurso FJ, Cutting GR. Cystic fibrosis. In: Scriver CR, Beaudet AL, Sly WS, Valle D, eds. The metabolic and molecular bases of inherited disease. 8th ed. New York: McGraw-Hill, 2001:5121-88.
Kerem E, Corey M, Kerem B, et al. The relation between genotype and phenotype in cystic fibrosis -- analysis of the most common mutation (F508). N Engl J Med 1990;323:1517-1522.
Mekus F, Laabs U, Veeze H, Tummler B. Genes in the vicinity of CFTR modulate the cystic fibrosis phenotype in highly concordant or discordant F508del homozygous sib pairs. Hum Genet 2003;112:1-11.
Doring G, Krogh-Johansen H, Weidinger S, Hoiby N. Allotypes of alpha 1-antitrypsin in patients with cystic fibrosis, homozygous and heterozygous for deltaF508. Pediatr Pulmonol 1994;18:3-7.
Mahadeva R, Westerbeek RC, Perry DJ, et al. Alpha1-antitrypsin deficiency alleles and the Taq-I GA allele in cystic fibrosis lung disease. Eur Respir J 1998;11:873-879.
Mahadeva R, Stewart S, Bilton D, Lomas DA. Alpha-1 antitrypsin deficiency alleles and severe cystic fibrosis lung disease. Thorax 1998;53:1022-1024.
Frangolias DD, Ruan J, Wilcox PJ, et al. Alpha 1-antitrypsin deficiency alleles in cystic fibrosis lung disease. Am J Respir Cell Mol Biol 2003;29:390-396.
Henry MT, Cave S, Rendall J, et al. An alpha1-antitrypsin enhancer polymorphism is a genetic modifier of pulmonary outcome in cystic fibrosis. Eur J Hum Genet 2001;9:273-278.
Arkwright PD, Pravica V, Geraghty PJ, et al. End-organ dysfunction in cystic fibrosis: association with angiotensin I converting enzyme and cytokine gene polymorphisms. Am J Respir Crit Care Med 2003;167:384-389.
Buscher R, Eilmes KJ, Grasemann H, et al. 2 adrenoceptor gene polymorphisms in cystic fibrosis lung disease. Pharmacogenetics 2002;12:347-353.
Baranov VS, Ivaschenko T, Bakay B, et al. Proportion of the GSTM1 0/0 genotype in some Slavic populations and its correlation with cystic fibrosis and some multifactorial diseases. Hum Genet 1996;97:516-520.
Hull J, Thomson AH. Contribution of genetic factors other than CFTR to disease severity in cystic fibrosis. Thorax 1998;53:1018-1021.
Flamant C, Henrion-Caude A, Boelle PY, et al. Glutathione-S-transferase M1, M3, P1 and T1 polymorphisms and severity of lung disease in children with cystic fibrosis. Pharmacogenetics 2004;14:295-301.
Garred P, Pressler T, Madsen HO, et al. Association of mannose-binding lectin gene heterogeneity with severity of lung disease and survival in cystic fibrosis. J Clin Invest 1999;104:431-437.
Gabolde M, Guilloud-Bataille M, Feingold J, Besmond C. Association of variant alleles of mannose binding lectin with severity of pulmonary disease in cystic fibrosis: cohort study. BMJ 1999;319:1166-1167.
Davies JC, Turner MW, Klein N. Impaired pulmonary status in cystic fibrosis adults with two mutated MBL-2 alleles. Eur Respir J 2004;24:798-804.
Carlsson M, Sjoholm AG, Eriksson L, et al. Deficiency of the mannan-binding lectin pathway of complement and poor outcome in cystic fibrosis: bacterial colonization may be decisive for a relationship. Clin Exp Immunol 2005;139:306-313.
Yarden J, Radojkovic D, De Boeck K, et al. Polymorphisms in the mannose binding lectin gene affect the cystic fibrosis pulmonary phenotype. J Med Genet 2004;41:629-633.
Grasemann H, van's Gravesande KS, Buscher R, et al. Endothelial nitric oxide synthase variants in cystic fibrosis lung disease. Am J Respir Crit Care Med 2003;167:390-394.
Arkwright PD, Laurie S, Super M, et al. TGF-beta(1) genotype and accelerated decline in lung function of patients with cystic fibrosis. Thorax 2000;55:459-462.
Bronsveld I, Mekus F, Bijman J, et al. Chloride conductance and genetic background modulate the cystic fibrosis phenotype of Delta F508 homozygous twins and siblings. J Clin Invest 2001;108:1705-1715.
Weir BS. Genetic data analysis II: methods for discrete population genetic data. Sunderland, Mass.: Sinauer Associates, 1996.
Qin ZS, Niu T, Liu JS. Partition-ligation-expectation-maximization algorithm for haplotype inference with single-nucleotide polymorphisms. Am J Hum Genet 2002;71:1242-1247.
Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. New York: Wiley, 2000.
Excoffier L, Slatkin M. Maximum-likelihood estimation of molecular haplotype frequencies in a diploid population. Mol Biol Evol 1995;12:921-927.
Abecasis GR, Cookson WO. GOLD -- graphical overview of linkage disequilibrium. Bioinformatics 2000;16:182-183.
Nielsen DM, Ehm MG, Weir BS. Detecting marker-disease association by testing for Hardy-Weinberg disequilibrium at a marker locus. Am J Hum Genet 1998;63:1531-1540.
Schluchter MD, Konstan MW, Davis PB. Jointly modeling the relationship between survival and pulmonary function in cystic fibrosis patients. Stat Med 2002;21:1271-1287.
Wu L, Chau J, Young RP, et al. Transforming growth factor-beta1 genotype and susceptibility to chronic obstructive pulmonary disease. Thorax 2004;59:126-129.
Pulleyn LJ, Newton R, Adcock IM, Barnes PJ. TGFbeta1 allele association with asthma severity. Hum Genet 2001;109:623-627.
Silverman ES, Palmer LJ, Subramaniam V, et al. Transforming growth factor-beta1 promoter polymorphism C-509T is associated with asthma. Am J Respir Crit Care Med 2004;169:214-219.
Celedon JC, Lange C, Raby BA, et al. The transforming growth factor-beta1 (TGFB1) gene is associated with chronic obstructive pulmonary disease (COPD). Hum Mol Genet 2004;13:1649-1656.
The International HapMap Consortium. The International HapMap Project. Nature 2003;426:789-796.
Kerem E, Reisman J, Corey M, Canny GJ, Levison H. Prediction of mortality in patients with cystic fibrosis. N Engl J Med 1992;326:1187-1191.
Corey M, Edwards L, Levison H, Knowles M. Longitudinal analysis of pulmonary function decline in patients with cystic fibrosis. J Pediatr 1997;131:809-814.
Cardon LR, Palmer LJ. Population stratification and spurious allelic association. Lancet 2003;361:598-604.
Blobe GC, Schiemann WP, Lodish HF. Role of transforming growth factor beta in human disease. N Engl J Med 2000;342:1350-1358.
Akhurst RJ. TGF beta signaling in health and disease. Nat Genet 2004;36:790-792.
Xaubet A, Marin-Arguedas A, Lario S, et al. Transforming growth factor-beta1 gene polymorphisms are associated with disease progression in idiopathic pulmonary fibrosis. Am J Respir Crit Care Med 2003;168:431-435.
Morris DG, Huang X, Kaminski N, et al. Loss of integrin alpha(v)beta6-mediated TGF-beta activation causes MMP12-dependent emphysema. Nature 2003;422:169-173.
Sterner-Kock A, Thorey IS, Koli K, et al. Disruption of the gene encoding the latent transforming growth factor-beta binding protein 4 (LTBP-4) causes abnormal lung development, cardiomyopathy, and colorectal cancer. Genes Dev 2002;16:2264-2273.
Grainger DJ, Heathcote K, Chiano M, et al. Genetic control of the circulating concentration of transforming growth factor type beta1. Hum Mol Genet 1999;8:93-97.
Yamada Y, Miyauchi A, Goto J, et al. Association of a polymorphism of the transforming growth factor-beta1 gene with genetic susceptibility to osteoporosis in postmenopausal Japanese women. J Bone Miner Res 1998;13:1569-1576.
Suthanthiran M, Li B, Song JO, et al. Transforming growth factor-beta 1 hyperexpression in African-American hypertensives: a novel mediator of hypertension and/or target organ damage. Proc Natl Acad Sci U S A 2000;97:3479-3484.
Dunning AM, Ellis PD, McBride S, et al. A transforming growth factor beta1 signal peptide variant increases secretion in vitro and is associated with increased incidence of invasive breast cancer. Cancer Res 2003;63:2610-2615.(Mitchell L. Drumm, Ph.D.,)