Introduction

Chronic obstructive pulmonary disease (COPD) is a major cause of morbidity and mortality worldwide1, and it is estimated to affect nearly 8.2% of the adult Chinese population2. COPD is a heterogeneous disease. In addition to its effects in the lung, COPD is also fraught with a considerable amount of extrapulmonary complications and comorbidities, such as osteoporosis, anemia, weight loss, depression, cardiovascular disease, and cancer3. Although lung function, as measured by forced expiratory volume in 1 second (FEV1), is a useful indicator of poor health outcomes in COPD, it is limited in that the extrapulmonary dimensions of COPD cannot be easily captured with this one measurement. In response, there are large, ongoing initiatives to identify biomarkers that can complement FEV1 in accurately identifying those patients with COPD that are at risk for extrapulmonary complications. Recent work on peripheral blood biomarkers has suggested that there may be biomarker signatures in blood that are associated with COPD phenotypes4,5. Serum adiponectin levels are a valuable diagnostic and prognostic marker of COPD.

Adiponectin is a secretory 30 kDa protein synthesized by adipocytes in healthy subjects. Adiponectin is highly abundant in plasma and is a key cytokine in energy homeostasis, regulating both glucose and lipid metabolism. In humans, down regulation of adiponectin and its receptors is associated with systemic inflammation and extrapulmonary effects, such as weight loss, osteoporosis, adipose atrophy, and skeletal muscle atrophy associated with age6. Recent human studies have demonstrated a significant increase in serum adiponectin levels in COPD and a direct correlation of these levels to the severity of the disease7,8. Plasma adiponectin is increased in underweight patients with COPD and is inversely associated with body mass index, degree of airflow limitation, and adverse cardiovascular outcomes. Elevation of plasma adiponectin has been associated with an accelerated decline in lung function9,10,11. Adiponectin also contributes to the development of emphysema7. Moreover, genetic variants in the adiponectin gene have been associated with plasma adiponectin levels and COPD risk in a Chinese population12.

Twin and family studies have demonstrated an estimated 30%–70% heritability for circulating adiponectin levels13. Recently, genome-wide association study (GWAS) methodology has been widely used to identify candidate genes that influence plasma adiponectin levels, and CDH13 encoding T-cadherin is one of the identified genes. Several single nucleotide polymorphisms (SNPs) in the CDH13 gene have been shown to be determinants of blood adiponectin levels in multi-ethnic populations14,15,16,17,18. T-cadherin has been reported to be an adiponectin receptor and was discovered after adipoR1 and adipoR2; it mainly binds to hexameric and high molecular weight (HMW) isoforms of adiponectin19, which are the most abundant isoforms in the lung lining fluid20. The aim of the current study was to evaluate the association of variants in the CDH13 gene with COPD in a Han Chinese population.

Materials and methods

Subjects

The subjects in this study were recruited for our previous studies12,21,22. Briefly, we investigated 646 unrelated ethnic Han Chinese individuals including 279 COPD patients and 367 age-matched control individuals. COPD was diagnosed according to the Global Initiative for Chronic Obstructive Lung Disease (GOLD) criteria23. Patients were excluded if they had any other comorbidities or significant respiratory diseases, such as bronchial asthma, bronchiectasis, lung cancer, or pulmonary tuberculosis. Among the age-matched non-COPD control individuals, none had a history of chronic lung disease, atopy, acute pulmonary infection in the 4 weeks before assessment for this study, or a family history of COPD.

The use of human tissue and the protocol in this study strictly conformed to the principles expressed in the Declaration of Helsinki and were approved by the Ethical Committee of the West China Hospital of Sichuan University. Written informed consent was obtained from all subjects before their participation in the study.

Biochemical measurements

Blood samples were collected at baseline from patients and controls after an overnight fast. Separated plasma was used for lipid, glucose and adiponectin analyses. The plasma levels of total cholesterol, triglycerides and glucose were determined with an enzymatic kit (Boehringer Mannheim, Mannheim, Germany). Circulating total adiponectin levels were measured by the enzyme-linked immunosorbent assay method (Quantikine, R&D Systems, Minneapolis, MN, USA).

SNP selection and genotyping

CDH13 is located on chromosome 16q and spans ∼1.17 megabases (CDH13: ENSG00000140945) (http://www.ensembl.org/index.html). Genotype data of the Chinese population for the CDH13 region were obtained from the HapMap website (http://www.hapmap.org/), and ten tag single-nucleotide polymorphisms (SNPs) in intron 1 ranging from 81 218 244 bp to 81 449 467 bp (NCBI build 37.3) were selected using the Tagger software implemented in the Haploview software24, with an r2 threshold of 0.8 and minor allele frequencies of 0.1 (supplementary Table S1).

Genomic DNA was extracted from peripheral blood leukocytes using a commercial extraction kit (Bioteke Corporation, Beijing, China) according to the manufacturer's instructions. SNPs were genotyped using the ABI SNaPshot method (Applied Biosystems, Foster City, CA, USA) as described previously12,22. Briefly, the PCR products (see supplementary Table S1 for primers) were purified by incubating with shrimp alkaline phosphatase and exo-nuclease I. Then, the purified PCR products were used as the templates for a SNaPshot reaction using specific SNaPshot primers (supplementary Table S1). The SNaPshot reaction products were analyzed on an ABI 3130 Genetic Analyzer (Applied Biosystems, Carlsbad, CA, USA). The data were analyzed by GeneMapper 4.0 software (Applied Biosystems, Foster City, CA, USA).

Statistical analysis

Statistical analyses were performed in SPSS version 17.0 (SPSS Inc, Chicago, IL, USA) and Microsoft Excel (Microsoft Corporation, Redmond, WA, USA). The demographic and clinical data of the COPD patients and the control subjects were compared using the χ2 test and Student's t-test. A multiple logistic regression analysis using BMI and glucose as covariates was performed to correct the significant P-value of adiponectin. A two-sided significance level of P<0.05 was used for all significant tests.

The Hardy-Weinberg equilibrium (HWE) test using two-sided χ2 analysis was performed for each SNP among COPD cases and controls. The frequency of each allele at each SNP locus was obtained by adding the frequency of its homozygote to half of the frequency of the heterozygote in which it appears. Differences in the distribution of genotypes or alleles under different genetic models (including dominant, recessive and additive models) between the COPD patients and the controls were estimated by using the χ2 test, and the best genetic model for each SNP was determined using Akaike's information criterion (AIC). The model with the smallest AIC value corresponds to the minimal expected entropy. Odds ratios (ORs) and 95% CIs were calculated by unconditional logistic regression analyses. Linear regression was used to study the association of SNPs with two COPD phenotypes (FEV1 and FEV1/FVC ratio). Multiple testing corrections were carried out using a false discovery rate (FDR)25. Q-values (=corrected P-values) were adjusted with subject demographics, such as gender, age, BMI and smoking history. The significance level was set at Q<0.05. FDR calculations were performed using the R language package fdrtool26.

Pairwise linkage disequilibrium (LD) estimation and haplotype reconstruction were performed using SHEsis (http://analysis.bio-x.cn). For haplotype analysis, only haplotypes with a frequency >3% in at least one group were tested. We also used Haploview 4.224 to estimate LD.

Results

General characteristics of the subjects

The baseline characteristics, the biochemical features and the results of the pulmonary function tests of the subjects in this study have been reported in our previous articles12,21,22 and are presented in supplementary Table S2. All patients had FEV1 values <80% of predicted, and thus were diagnosed with moderate to severe COPD according to the Global Initiative for Chronic Obstructive Lung Disease23 (classification of severity: mild=FEV1≥80% of predicted; moderate=FEV1≥50% to <80% of predicted; severe=FEV1≥30% to <50% of predicted; and very severe=FEV1<30% of predicted). The COPD cases and control subjects did not significantly differ in sex, age or smoking history (P>0.05). Compared to control individuals, the COPD patients had worse pulmonary function with significantly decreased FEV1, FEV1/predicted and FEV1/FVC ratios (P<0.01). In addition, the COPD patients had statistically higher glucose concentrations (5.86±0.16 mmol/L vs 5.13±0.12 mmol/L, P<0.01), higher adiponectin levels (8.54±0.66 μg/mL vs 6.12±0.57 μg/mL, P<0.01), and a significantly lower BMI (22.03±2.27 kg/m2 vs 23.91±2.48 kg/m2, P<0.01).

Distribution of the SNPs in the CDH13 gene between COPD patients and controls

Frequencies of alleles and genotypes in the population are fundamental quantities in human statistical genetics, which form the basis of association studies between SNPs and common diseases. Table 1 summarizes the genotype and allele frequencies of each SNP in both COPD patients and controls. All genotype distributions of the tested SNPs were in Hardy-Weinberg equilibrium (HWE) in patients and control subjects (all P>0.05), illustrating that our subjects presented the source population well. SNPs rs4783244 and rs12922394 exhibited significant differences in allele or genotype frequencies between COPD patients and control subjects, whereas other SNPs did not. For rs4783244, compared to the allele G, the allele T was associated with a significantly decreased risk of COPD (OR=0.70, 95% CI: 0.56–0.91, P=0.007, FDR Q=0.022). For rs12922394, compared to the allele C, the allele T is associated with a significantly decreased risk of COPD (OR=0.72, 95% CI: 0.56–0.93, P=0.013, FDR Q=0.031). The genotypes of both rs4783244 and rs12922394 were distributed significantly differently between COPD patients and control subjects (P<0.05).

Table 1 Distributions of the CDH13 SNPs in COPD patients and controls.
Full size table

Association of genotypes with COPD under different genetic models

The maximum power in genetic association studies is reached when concordance is observed between the “true” model of inheritance of disease susceptibility loci and the genetic model used in the analysis. We compared the genotype frequencies of every polymorphism between groups under the dominant, recessive and additive genetic models. The best genetic model for each SNP was determined using Akaike's information criterion. For each SNP, if one allele's frequency was relatively lower than another's, it was recognized as the minor allele (Table 1), which was assumed to be a risk allele compared to a wild type allele. As shown in Table 2, the minor allele T at rs4783244 was associated with a decreased COPD risk under a recessive model (TT vs GT+GG: OR=0.42, 95% CI: 0.19–0.91, P=0.023, FDR Q=0.042), whereas the minor allele T at rs12922394 was associated with a decreased COPD risk under a dominant model (CT+TT vs CC: OR=0.70, 95% CI: 0.51–0.95, P=0.022, FDR Q=0.039).

Table 2 Association between CDH13 SNPs and COPD risk under different genetic models.
Full size table

Association between SNPs rs4783244 and rs12922394 and lung function

To determine whether SNPs rs4783244 and rs12922394 influence COPD phenotypes, we assessed their associations with the FEV1 and FEV1/FVC ratio in the entire population, including 279 COPD patients and 367 controls (Table 3). The T allele at rs4783244 showed significant associations with both increased FEV1 (P=0.017) and FEV1/FVC ratio (P=0.024) under a recessive model (TT vs GT+GG), and the association remained significant even after correction for multiple testing (FDR Q=0.031 and 0.042 for FEV1 and FEV1/FVC ratio, respectively). The minor allele T at rs12922394 was again associated with both increased FEV1 (P=0.028) and FEV1/FVC ratio (P=0.030) under a dominant model (CT+TT vs CC), although this did not meet the threshold for multiple testing correction (FDR Q>0.05).

Table 3 Associations between polymorphisms rs4783244 and rs12922394 and COPD phenotypes.
Full size table

Association between SNPs rs4783244 and rs12922394 and plasma adiponectin levels

The relationships between SNPs rs4783244 and rs12922394 and plasma adiponectin levels were examined in both COPD patients and control subjects (Figure 1). Consistently with findings in a previous study17,18,27, after adjusting for BMI and glucose levels, the subjects with homozygous TT at either rs4783244 (5.62±0.49 μg/mL in controls and 7.73±0.41 μg/mL in COPD patients) or rs12922394 (5.57±0.52 μg/mL in controls and 7.62±0.47 μg/mL in COPD patients) exhibited a significant decrease in plasma adiponectin levels when compared with the subjects with homozygous GG at rs4783244 (6.63±0.41 μg/mL in controls and 8.91±0.44 μg/mL in COPD patients) or CC at rs12922394 (6.39±0.42 μg/mL in controls and 8.72±0.43 μg/mL in COPD patients), respectively (all P<0.05).

Figure 1
figure 1

The allele effects of rs4783244 and rs12922394 on plasma adiponectin levels. Mean±SD. bP<0.05 vs subjects with genotypes GG at rs4783244 and CC at rs12922394.

PowerPoint slide

Full size image

Linkage disequilibrium (LD) between SNPs and haplotype analysis

In HapMap populations, there is a substantive difference in LD structure at the CDH13 locus between European and Asian subjects (supplementary Figure S1). We then assessed the LD in pairwise combinations of alleles in different SNPs by means of maximum likelihood from the genotype frequency in the COPD and control groups. Pairwise LD between SNPs is shown in supplementary Table S3 and Figure 2. Based on LD determinations, three blocks with moderate LD were detected: block 1, composed of rs4783244 and rs12922394; block 2, composed of rs11646011 and rs11640875, and block 3, composed of rs4783266, rs11640522, rs11646849, and rs11860282.

Figure 2
figure 2

Linkage disequilibrium (LD) plots for CDH13. The LD plots were generated by Haploview 4.2. Polymorphisms are identified by their dbSNP rs numbers, and their relative positions are marked by vertical lines within the white horizontal bar. The numbers within squares indicate the D' value, expressed as a percentile.

PowerPoint slide

Full size image

We estimated the frequencies of haplotypes constructed from phased multi-locus genotypes in CDH13. The haplotypes with a frequency higher than 3% in at least one group were used in the haplotype analysis. As shown in Table 4, haplotype GC in block 1 was associated with an increased risk of COPD (OR=2.21, 95% CI=1.44–3.37, P=1.92×10−4), whereas haplotype TT was protective against COPD (OR=0.79, 95% CI=0.63–0.99, P=0.041). In block 3, two haplotypes were associated with an increased risk of COPD (GTAC: OR=2.68, 95% CI=1.45–4.95, P=0.001; ATGT: OR=2.05, 95% CI=1.38–3.03, P=3.09×10−4), whereas haplotype ATGC was protective against COPD (OR=0.70, 95% CI=0.55–0.89, P=0.003).

Table 4 Frequencies of pairwise haplotype constructed by SNPs in CDH13.
Full size table

Discussion

Adiponectin, although a secreted protein of fat cells, is also expressed in the airway epithelium and BAL fluid in the emphysematous form of COPD28. Adiponectin has important roles in insulin sensitization, cardioprotection, and anti-inflammatory processes. There have been several clinical studies reporting on the relationship between circulating adiponectin and COPD, and plasma adiponectin level is elevated in patients with stable and acute exacerbation of COPD7,10,11. The most bioactive form of adiponectin is a 400 kDa HMW complex, which is also dramatically increased in COPD patients29. The CDH13 gene encodes the T-cadherin protein, which is a receptor for hexameric and HMW forms of adiponectin19.

In this case-control study in a Han Chinese population, we carried out the first investigation of a possible association between SNPs in the CDH13 gene and COPD risk. Our current findings suggested that rs4783244 and rs12922394 were associated with a risk of COPD. In comparison with allele G at rs4783244 or allele C at rs12922394, the minor allele T at either site was associated with a decreased COPD risk. Moreover, the genotypes at rs4783244 and rs12922394 were also found to affect the plasma adiponectin levels, consistently with findings from previous studies17,18,27. However, given the high frequencies of the G allele at rs4783244 (70.7%) and the C allele at rs12922394 (73.4%) related to the elevated levels of adiponectin in control subjects, the effects of the genotypes on the susceptibility of COPD were limited. The frequency of the T allele at rs4783244 was 29.3% in our control subjects, which appears to be lower than the 36% in non-diabetic Han Chinese subjects30. This might be attributable to the differences between the two studies in subjects' age, gender, and smoking history. Compared to the non-diabetic subjects in Li's study30, our control subjects are older (65±8 years vs 50±11 years), more likely to be male (88% vs 63%), and heavier smokers. Indeed, T allele frequency is approximately 29% in Han Chinese in HapMap project data, which is comparable to our results. In addition to the genotype analysis, our study also adopted a haplotype-based approach. We observed that some haplotypes with a low frequency dramatically changed the risk of COPD in the opposite direction (Table 4), which indicates the complexity of the CDH13 gene in the development of COPD.

Both genetic and environmental factors contribute to the pathogenesis of COPD. Cigarette smoking is a major risk factor for COPD, and genetic and smoking interactions have been associated with lung function in COPD31,32. It is possible that the effects of CDH13 variants on COPD risk may be even more significant in individuals with low exposure to cigarette smoke than in moderate and heavy smokers, as recruited in the present study. Because the dose of toxic substrates was so high in heavy smokers, it overwhelmed the effects of the CDH13 genotype. In moderate smokers, the relative effect of the CDH13 genotype may be uncovered.

The association between CDH13 variants, plasma adiponectin, and the risk of COPD may be explained by the function of the T-cadherin receptor that CDH13 encodes. The amino acid motif of T-cadherin is well conserved in higher eukaryotes compared to that of E-cadherin, suggesting some biological significance. T-cadherin may act as a co-receptor along with other signaling molecules, but its physiological roles are largely unknown. rs4783244 or rs12922394 might be associated with an increased baseline level of T-cadherin in tissues. The increased amount of T-cadherin may capture the free adiponectin molecules in the plasma, resulting in a lowering of plasma HMW adiponectin levels, while simultaneously increasing the adiponectin signals in the targeting cells, augmenting the effect of adiponectin per cell. Consistently with this explanation, ablation of the T-cadherin receptor increased plasma adiponectin levels in mice20,33. Therefore, CDH13 genotypes and adiponectin genotypes may need to be taken into account if plasma adiponectin levels are used as a marker for COPD risk.

The statistically significant associated SNPs (rs4783244 and rs12922394) are in the first intron of the CDH13 gene, which is usually removed during the gene-splicing process. Although they cause no apparent functional change, intronic SNPs may modify gene function by affecting the regulation of gene expression34. In addition, the SNPs associated with the statistical signal may play a role as a surrogate marker for a causal functional SNP or SNPs because the surrounding nucleotide sequence did not match the known transcription factor binding site or miRNA targeted sequence. rs4783244 or rs12922394 could be in LD with another polymorphism of the gene that may affect the CDH13 expression level. However, it is also likely that the causal sequence change or changes in this region have yet to be identified, as suggested by the analysis of the significant haplotypes. For example, SNP rs4783266 was significantly associated with COPD risk as part of a haplotype, but not individually (Table 4). We selected SNPs with minor allele frequencies of >10% in the Han Chinese population using HapMap project data, but this is not suited for situations where genetic architecture is such that multiple rare disease-causing variants contribute significantly to disease risk. Recent studies have demonstrated that the identification of rare variants may lead to critically important insights about disease etiology through the implication of new genes and/or pathways35,36. The rare variants of the CDH13 gene should be investigated to clarify their role in susceptibility to the development of COPD.

The significant results in the current study could prove to be false positives because of the relatively small sample size and the lack of replication evidence. However, chromosome 16q, where CDH13 is located, has been associated with the pulmonary function that is used in the diagnosis of COPD37. Further studies using larger populations are needed. However, even with a larger sample, the functional and biological impacts of the described polymorphisms would require further study. Given the number of processes and pathways in which CDH13 functions and the significance of association between CDH13 and COPD, it is unlikely that CDH13 alone contributes to the sequelae in COPD; instead, genes regulated by CDH13, or those that regulate CDH13, may influence the disease. Therefore, the investigation of CDH13-related pathways and gene networks may lead to a better understanding of the pathophysiology of COPD.

In conclusion, our comprehensive analysis of SNPs in the CDH13 gene suggests that CDH13 genotypes and haplotypes may play a role in COPD development. CDH13 encodes one of the receptors for adiponectin, and plasma adiponectin level is elevated in patients with a stable and acute exacerbation of COPD. Therefore, we may need to take into account, both CDH13 genotypes and adiponectin genotypes, if plasma adiponectin levels are used as a marker for risk of COPD. Identifying the genetic background of patients with COPD allows us to identify risk groups in the community and develop new individual-based targets for therapy. Our results represent a key step in the development of genetic susceptible biomarkers for the detection of COPD. Because this is the first case-control study investigating the association of CDH13 with the risk of COPD, additional studies with large patient populations are required to confirm our findings.

Author contribution

Yi-ming YUAN, Jin-long ZHANG, Si-cheng XU, Ren-song YE, You ZHANG, Yan-Jie ZHANG, Yu-long CHEN, Dan XU, and Yu-lan LIU contributed to the acquisition and analysis of data; Zhi-guang SU contributed to the study design and data analysis and wrote the manuscript.