Decorin and TGF-β1 polymorphisms and development of COPD in a general population

Background Decorin, an extracellular matrix (ECM) proteoglycan, and TGF-β1 are both involved in lung ECM turnover. Decorin and TGF-β1 expression are decreased respectively increased in COPD lung tissue. Interestingly, they act as each other's feedback regulator. We investigated whether single nucleotide polymorphisms (SNPs) in decorin and TGF-β1 underlie accelerated decline in FEV1 and development of COPD in the general population. Methods We genotyped 1390 subjects from the Vlagtwedde/Vlaardingen cohort. Lung function was measured every 3 years for a period of 25 years. We tested whether five SNPs in decorin (3'UTR and four intron SNPs) and three SNPs in TGF-β1 (3'UTR rs6957, C-509T rs1800469 and Leu10Pro rs1982073), and their haplotypes, were associated with COPD (last survey GOLD stage = II). Linear mixed effects models were used to analyze genotype associations with FEV1 decline. Results We found a significantly higher prevalence of carriers of the minor allele of the TGF-β1 rs6957 SNP (p = 0.001) in subjects with COPD. Additionally, we found a significantly lower prevalence of the haplotype with the major allele of rs6957 and minor alleles for rs1800469 and rs1982073 SNPs in TGF-β1 in subjects with COPD (p = 0.030), indicating that this association is due to the rs6957 SNP. TGF-β1 SNPs were not associated with FEV1 decline. SNPs in decorin, and haplotypes constructed of both TGF-β1 and decorin SNPs were not associated with development of COPD or with FEV1 decline. Conclusion Our study shows for the first time that SNPs in decorin on its own or in interaction with SNPs in TGF-β1 do not underlie the disturbed balance in expression between these genes in COPD. TGF-β1 SNPs are associated with COPD, yet not with accelerated FEV1 decline in the general population.


Background
Chronic obstructive pulmonary disease (COPD) is characterized by irreversible airway obstruction and persistent airway inflammation. Transforming growth factor-β 1 (TGF-β 1 ) is one of the important cytokines involved in this inflammatory process, which has been associated with cell proliferation and differentiation. It is furthermore involved in repair of the extracellular matrix (ECM) after inflammation and tissue injury amongst others by promoting synthesis of elastin and collagen. Studies have shown that TGF-β 1 expression is increased in the airways of COPD patients [1,2] In contrast, a recent article from Pons et al showed that alveolar macrophages from COPD patients release less TGF-β 1 in response to lipopolysaccharide than smokers with normal lung function and nonsmokers [3] This may lead to a reduced anti-inflammatory and anti-elastolytic response in COPD patients, subsequently contributing to progressive ECM destruction.
Decorin is a component of the ECM that regulates collagen fibrillogenesis. [4][5][6] In addition, it can interact with a wide variety of growth factors, cytokines and adhesion molecules through its extensive binding area, thereby not only playing a role in ECM assembly but also in control of cell proliferation and tissue morphogenesis. [7]TGF-β 1 has been shown to downregulate synthesis of decorin in fibroblasts and decorin can in turn inhibit TGF-β 1 . [8] Decorin may thus act as a negative feedback regulator of TGF-β 1 mediated repair responses. Conversely, TGF-β 1 can downregulate expression of decorin in fibroblasts from emphysema patients. [9] We have shown previously that decorin expression is diminished in the peribronchiolar area of lung tissue from patients with severe emphysema, while TGF-β 1 production from fibroblasts of these patients is increased. [10] Noordhoek et al showed that TGF-β 1 and basic fibroblast growth factor give a stronger reduction of decorin production in the culture supernatant of fibroblasts from patients with severe emphysema than from patients with mild emphysema. [9] It thus appears that the regulation of decorin production is disturbed in lung tissue from patients with severe emphysema. This will lead to diminished binding and neutralization of TGF-β 1 by decorin followed by higher TGF-β 1 concentrations and activity with lower decorin production as a result.
We hypothesized that the reciprocal regulation of the TGF-β 1 and decorin genes is disturbed in COPD due to a genetic mutation in one or both of these genes. We have tested this hypothesis by investigating three single nucleotide polymorphisms (SNPs) in TGF-β 1 and five SNPs in decorin on the development of COPD and on lung function decline in a large cohort derived from the general population (the Vlagtwedde/Vlaardingen cohort).

Subjects
We used data from 2467 subjects of the Vlagtwedde/ Vlaardingen cohort participating in the last survey in 1989/1990. This general population-based cohort of Caucasians of Dutch descent started in 1965. Surveys, during which pulmonary function measurements were performed, were held every three years. The selection of the cohort has been described previously. [11][12][13] Surveys were performed every 3 years during which information was collected on respiratory symptoms, smoking status, age and gender by the Dutch version of the British Medical Council standardized questionnaire. A blood sample was taken and spirometry was performed. Details on pulmonary function measurements are provided in the additional file 1. The methodology for standardization and equipment used for lung function measurements was the same throughout the study. In 1989/1990 neutrophil depot of centrifuged blood was collected and stored at -20°C. In 2003/2004 DNA was extracted from these samples with the QiaAmp ® DNA Blood Mini Kit and checked for purity and concentration with the NanoDrop ® ND-1000 UV-Vis Spectrophotometer. The study protocol was approved by the local university hospital's medical ethics committee and participants gave written informed consent.

Genotyping
We genotyped DNA of those subjects with more than 1500 ng isolated DNA available (N = 1390). Three SNPs, previously associated with COPD or level of lung function were genotyped in TGF-β 1 : rs6957 in the 3'UTR, rs1800469 in the promoter region (C-509T) and a coding SNP rs1982073 (Leu10Pro, G/T). [14][15][16] Coding SNPs in decorin have been identified in the NCBI and Celera databases, but are only prevalent in African populations (frequency 0.05-0.12) but not in Caucasian populations (frequency 0.00). According to the HapMap database there are two large LD blocks in the decorin gene, and a region including the 3'UTR that forms no LD block. [17]. There are 4 haplotype tagging SNPs located in introns, resulting in 3 major haplotypes, which cover the information of the gene. Therefore, we genotyped one SNP in the 3'UTR (rs1803343), and the 4 haplotype-tagging SNPs: rs11106030, rs741212, rs566806, rs516115 and rs3138241. The genotyping protocol is described in the additional file 1; the characteristics of the genotyped SNPs in additional file 2. To determine whether the SNPs were in Hardy Weinberg equilibrium and whether they were in linkage disequilibrium, tests were performed with the statistical package R (version 1.9.1).

Statistics
We identified subjects with COPD using the GOLD criteria (GOLD stage II or higher, i.e. FEV 1 /VC< 70% and FEV 1 <80% predicted) at the last survey [18] Characteristics of subjects with and without COPD at the last survey are presented in table 1. Differences in allele frequencies and haplotype frequencies between subjects with and without COPD were tested using Chi-square tests. We used ANOVA and linear regression models to study the effect of SNPs on first and last available FEV 1 and FEV 1 /VC (adjusted for gender, age, pack-years, and height in regression models).
Linear Mixed Effect (LME) models were used to investigate the effect of SNPs in TGF-β 1 and decorin on annual FEV 1 decline in the general population, like published previously. [19,20] Time was defined as time in years relative to the first FEV 1 , starting from the age of 30. [21] Variables included in the model were age at entry, gender, pack-years, the first FEV 1 after age 30, and their interaction with time. Since including the level of the first FEV 1 after age 30 and its interaction with time could introduce bias due to regression to the mean, these variables were also included in the model as random effect variables. The results of these analyses showed no change in estimates of the variables in the model or a better fit of the model, which indicates that there was no bias due to regressionto-the-mean. Therefore, the results are presented without these random effects. To test whether SNPs were associated with FEV 1 decline within subjects with COPD, we performed LME analyses on these subjects only. Since Celedón et al found stronger linkage results of TGF-β 1 SNPs and lung function in smokers only, we additionally performed LME models stratified to smoking status. [14] We also included interaction terms of TGF-β 1 SNPs and decorin SNPs to test for genetic interaction of these SNPs.
Instead of performing pre-or post-hoc power analysis and correction for multiple testing, we performed permutation tests to assess whether our results might have been found due to chance. Genotypes were randomly shuffled among individuals to produce 3000 datasets. The LME models were rerun on each of these datasets to generate a distribution of the beta estimates for additional FEV 1 decline of the homozygous minor allele genotype compared to FEV 1 decline of the homozygous wild type allele genotype under the null hypothesis, being no association of the SNPs under study and FEV 1 decline. If the observed beta estimate from the true data is found in the lower 5% percentile of the empiric cumulative distribution (p < 0.05), one can assume that the observed beta estimate is not found due to chance.
We also estimated TGF-β 1 haplotype frequencies in the whole population and in subjects with a COPD phenotype. Estimated haplotype frequencies for TGF-β 1 higher than 1% in the general population were used to construct phased multi-locus genotypes of TGF-β 1 . For decorin, we constructed the phased multi-locus genotypes as known from the HapMap database. With Chi-square tests we determined for each haplotype whether there was a difference in prevalence of carriers between subjects with and without COPD. Also, the excess decline in FEV 1 in the whole population was determined for each phased multilocus genotype in the LME.
Statistical analyses were performed using SPSS (version 12.0.1 for Windows), the statistical package R (version 1.9.1) and Arlequin [22].

Results
Allelic frequencies for the minor alleles of the TGF-β 1 and decorin SNPs in this population were comparable to those reported in the Celera and/or in the NCBI dbSNP database: TGF-β1 rs6957 0.18, rs1800469 0.28, rs1982073 0.38, decorin rs1803343 0.02, rs11106030 0.06, rs741212 0.12, rs566806 0.26, rs516115 0.22 and rs3138241 0.06. All SNPs were in Hardy Weinberg equilibrium. The TGFβ 1 rs1800469 SNP was in significant LD with rs1982073 and rs6957. Rs6957 was in almost significant LD with rs1982072 (p = 0.06). The decorin SNPs were in significant LD. Graphs of the LD patterns with D', r and P-values in both genes are presented in the additional file 3.

Prevalence of SNPs and haplotypes in TGF-β 1 and decorin in COPD and control subjects
The distribution of the TGF-β1 rs6957 genotypes was significantly different between subjects with and without COPD (p = 0.001, table 2). The other TGF-β 1 SNPs were not associated with COPD. We also found no association of SNPs in decorin with the prevalence of COPD. We used estimated haplotype frequencies higher than 0.01 to construct phased multi-locus genotypes for TGFβ 1 . The haplotype consisting of the minor allele for TGFβ 1 rs6957 and the wild type alleles for TGF-β 1 rs1800469 and rs1982073 was more prevalent in subjects with COPD (p = 0.014). Because the prevalence of carriers of other haplotypes containing the minor allele at TGF-β 1 rs6957 was also increased in subjects with COPD, this finding only reflects the individual association of the TGFβ 1 rs6957 SNP with COPD. Carriers of at least one haplotype with the minor alleles for TGF-β 1 rs1800469 and rs1982073 and the wild-type allele for rs6957 were less prevalent in COPD (p = 0.030). We found no significant associations of phased multi-locus genotypes in decorin with the prevalence of COPD (table 3). We also did not find associations of haplotypes containing SNPs of both TGF-β 1 and decorin with COPD (data not shown).

Lung function
We found no significant associations (i.e. cross-sectional) between the SNPs tested and FEV 1 and FEV 1 /VC at the first or at the last survey in linear regression models (data not shown). The mean adjusted annual decline in lung function (expressed as decrease in FEV 1 in ml/yr) was determined for subjects with the wild-type genotype for the SNPs in TGF-β 1 and decorin using LME models. The outcome of the mean annual decline concerns females with age 30 when entered in the LME, a mean first FEV 1 of the population, and zero pack-years. The mean of these adjusted annual declines was 19.2 ml/yr (range 18.7-19.6). We did not find any significant association of SNPs in either TGF-β 1 or decorin with accelerated lung function decline (table 4). We added interaction terms of TGF-β 1 en decorin SNPs in the model, but found no significant inter-actions. In addition, we did not find any significant association of haplotypes of either TGF-β 1 or decorin with accelerated lung function decline (results not shown). We also tested whether SNPs were associated with lung function decline within subjects with COPD or within smokers, but found no significant associations (table 4 and additional file 4). To test whether results were not missed due to chance, we performed permutation tests. We ran 3000 permutations on our sample of 1390 subjects and performed LME analyses on each of these permutations. The lack of associations with lung function decline was confirmed in these analyses.

Discussion
Decorin and TGF-β 1 can act as each other's feed back regulators in ECM turnover and their expression is respectively decreased and increased in lung tissue of COPD patients. We assessed whether polymorphisms in decorin and TGF-β 1 are associated with the development of COPD and accelerated lung function decline in the general population. This is the first study assessing SNPs in decorin and we did not find any association with COPD or lung function loss. Contrary to our hypothesis, the observed disturbed balance between decorin and TGB-β 1 in COPD is not caused by a combination of SNPs in their genes, since we found no significant interaction terms of decorin and TGF-β 1 SNPs with respect to FEV 1 decline. Moreover, we found no associations of phased multi-locus genotypes containing SNPs of both TGF-β 1 and decorin with the presence of GOLD stage II and III COPD in our population. This disturbed balance may be affected by SNPs in TGF-β 1 alone since the 3'UTR SNP in TGF-β 1 is predictive of COPD (stage GOLD II). We found, however, no association of SNPs in TGF-β 1 with longitudinal decline in lung  function. In addition, no associations were observed of SNPs in TGF-β 1 with level of FEV 1 or FEV 1 /VC cross-sectionally.

SNP No COPD N (%) COPD N (%) P value df = 2 SNP No COPD N (%) COPD N (%) P value df = 2
It is puzzling that we observed that the TGF-β 1 rs6957 SNP and a haplotype in TGF-β 1 were associated with COPD, but not with excess decline in FEV 1 or with level of FEV 1 and FEV 1 /VC at the last survey. We have tested whether there were differences in first available FEV 1 (which might suggest a relation to maximal attained lung function level) between the genotypes that could explain the lack of association with FEV 1 decline but this was not the case. Another possibility would be that the FEV 1 decline is only affected by SNPs in certain subgroups, such as smokers.
Our stratified analyses showed no such effect.
Although the functionality of the TGF-β 1 rs6957 SNP is not known yet, it has previously been associated with lower pre-and post-bronchodilator FEV 1 and with lower FEV 1 /FVC. [14] Similarly, we have shown here that this SNP is associated with development of COPD. Various studies have indicated that the rs1800469 and rs1982073 SNPs are functional and result in higher levels of circulating TGF-β 1 . [23][24][25][26] Since TGF-β 1 has anti-inflammatory and pro-repair activities, these SNPs are thought to be protective against the development of COPD. Indeed, we and others have found that (carriers of haplotypes of) the minor alleles of these SNPs are significantly less prevalent in COPD patients compared to controls. [14,16]. Similar to Celedón et al, we found an association of a haplotype with at least one minor allele of the rs1800469 and rs1982073 TGF-β 1 SNPs and COPD, while they also found associations with these SNPs separately. [14,16] The differences in study populations may explain these dissimilarities, e.g. our subjects had milder COPD (FEV 1 <80% predicted) than the COPD patients in the Celedón study (FEV 1 <45% predicted). Despite the differences in associations, it is still conceivable that carrying both of the SNPs decreases the risk to develop COPD. The two other studies linking TGF-β 1 SNPs and COPD have also demonstrated that these SNPs are less prevalent in COPD, though these studies did not test haplotypes [15,16] Many SNPs have been described in the TGF-β 1 gene, but only a few have been intensively studied in genetic association studies. Cross-sectional studies have found associations of SNPs in TGF-β 1 with the presence of COPD, and with lower levels of FEV 1 and FEV 1 /FVC in several populations. [14][15][16] We did not analyze every SNP in the TGF-β 1 gene that was previously reported to be associated with COPD. However, since Celedón et al found strong LD (r 2 = 0.98) between promoter SNPs and 3'UTR SNPs in a Caucasian population, we are confident that any association that might exist would have been revealed by the SNPs or by their haplotypes. [14] This is the first study on SNPs in decorin in a general population or in COPD patients. We were interested in polymorphisms in this gene, since decorin expression in COPD patients is diminished. [9,10] Decorin plays a direct role in the repair processes after inflammation through its regulation of matrix metalloproteases and tissue inhibitors of metalloproteases. [27,28] Furthermore, decorin is the natural inhibitor of TGF-β 1 and may therefore influence the repair process in the lung indirectly. We hypothesized that these processes may be genetically influenced. Since the coding SNPs in decorin described in the NCBI and Celera databases were not prevalent in Caucasians (but only in African populations), we genotyped four tagging SNPs, located in introns, and additionally a 3'UTR SNP. Although we found no significant associations of these SNPs with COPD or lung function decline, we can not rule out completely that there is no genetic defect in decorin that increases the risk to develop COPD. However, since we selected tagging SNPs that cover the genetic information of the decorin gene according to Hap-Map and given the large population under study, we assume that we would have observed an association of SNPs or haplotypes in decorin if there existed one in this population.
The lack of a genetic association of SNPs in the decorin gene does not rule out an important role of the decorin protein in COPD development. Decorin is a member of the proteoglycan family, a family of macromolecules composed of a protein core with glycosaminoglycan side chains which are produced post-translationally. It is possible that the function or activation of decorin is disrupted through an altered posttranslational modification of this glycosaminoglycan chain. In this case, modifications in the protein core, which might be caused by SNPs, may not be important and will not be detected. Decorin can be expressed in six splice variants, but the function of these splice variants is not known yet. Nevertheless, a shift in prevalence of one of these splice variants may affect the biological role that decorin exerts in TGF-β 1 regulation, thereby influencing the pathology within the lung. Abbreviations: FEV 1 , forced expiratory volume in 1 second; TGF-β 1 , transforming growth factor-β 1 ; COPD, Chronic Obstructive Pulmonary Disease; WT, wild-type *decline in FEV 1 adjusted for gender, first FEV 1 after age 30 years, pack-years, and age; † P value indicates significance of the effect of the genotype on decline in FEV 1 compared to wild-type