Genome-Wide Association Analysis of Blood Biomarkers in Chronic Obstructive Pulmonary Disease

Abstract

Rationale: A genome-wide association study (GWAS) for circulating chronic obstructive pulmonary disease (COPD) biomarkers could identify genetic determinants of biomarker levels and COPD susceptibility.

Objectives: To identify genetic variants of circulating protein biomarkers and novel genetic determinants of COPD.

Methods: GWAS was performed for two pneumoproteins, Clara cell secretory protein (CC16) and surfactant protein D (SP-D), and five systemic inflammatory markers (C-reactive protein, fibrinogen, IL-6, IL-8, and tumor necrosis factor-α) in 1,951 subjects with COPD. For genome-wide significant single nucleotide polymorphisms (SNPs) (P < 1 × 10⁻⁸), association with COPD susceptibility was tested in 2,939 cases with COPD and 1,380 smoking control subjects. The association of candidate SNPs with mRNA expression in induced sputum was also elucidated.

Measurements and Main Results: Genome-wide significant susceptibility loci affecting biomarker levels were found only for the two pneumoproteins. Two discrete loci affecting CC16, one region near the CC16 coding gene (SCGB1A1) on chromosome 11 and another locus approximately 25 Mb away from SCGB1A1, were identified, whereas multiple SNPs on chromosomes 6 and 16, in addition to SNPs near SFTPD, had genome-wide significant associations with SP-D levels. Several SNPs affecting circulating CC16 levels were significantly associated with sputum mRNA expression of SCGB1A1 (P = 0.009–0.03). Several SNPs highly associated with CC16 or SP-D levels were nominally associated with COPD in a collaborative GWAS (P = 0.001–0.049), although these COPD associations were not replicated in two additional cohorts.

Conclusions: Distant genetic loci and biomarker-coding genes affect circulating levels of COPD-related pneumoproteins. A subset of these protein quantitative trait loci may influence their gene expression in the lung and/or COPD susceptibility.

Clinical trial registered with www.clinicaltrials.gov (NCT 00292552).

At a Glance Commentary

Scientific Knowledge on the Subject

Studies of circulating biomarkers in multiple diseases have provided valuable insights into disease pathophysiology and treatment strategies. In chronic obstructive pulmonary disease (COPD), pneumoproteins including Clara cell secretory protein and surfactant protein D and systemic inflammatory markers including C-reactive protein, fibrinogen, IL-6, IL-8, and tumor necrosis factor-α have been reported to be associated with COPD risk, COPD mortality, COPD exacerbations, and lung function decline. However, the association between genetic variants and blood biomarker levels has been seldom investigated in COPD.

What This Study Adds to the Field

Genome-wide significant associations of several single nucleotide polymorphisms and circulating levels of Clara cell secretory protein and surfactant protein D were identified, whereas genome-wide association analysis of fibrinogen, IL-6, IL-8, tumor necrosis factor-α, and C-reactive protein did not show any significant associations. Remote genetic loci and biomarker-coding genes were associated with the blood levels of several protein biomarkers. A subset of these protein quantitative trait loci may influence mRNA expression in sputum and COPD susceptibility. Thus, genome-wide association analysis of biomarkers may be a useful approach to search for new genetic determinants of complex diseases.

Studies of biomarkers in multiple diseases have provided valuable insights into disease pathophysiology (1) and treatment strategies (2). In this context, peripheral blood biomarkers are most frequently studied because of their accessibility. However, blood biomarkers have limitations. First, there can be ambiguity in temporal relationships (i.e., was the biomarker a cause or effect of disease) (3). Second, biomarkers may be affected by many environmental and metabolic confounders, as well as genetic determinants, which may or may not be related to the disease of interest. Importantly, genome-wide association studies (GWAS) have identified significant associations between genetic variants and circulating biomarkers for multiple diseases (4–7), but the relationship of these genetic associations to disease susceptibility has been variable (3).

Chronic obstructive pulmonary disease (COPD) is a worldwide disease with increasing morbidity and mortality (8, 9). Many biomarkers that may reflect the underlying pathophysiology of COPD have been studied for decades but remain of uncertain use in tracking COPD outcomes. COPD may result from localized inflammation of the respiratory system and from systemic inflammatory insults. Clara cell secretory protein (CC16) and surfactant protein D (SP-D) are produced predominantly in the respiratory system (pneumoproteins) and have recently been related to COPD susceptibility and COPD-related clinical phenotypes (10–14). Other systemic inflammatory markers including C-reactive protein (CRP), fibrinogen, IL-6, IL-8, and tumor necrosis factor (TNF)-α have also been reported to be associated with COPD risk, COPD mortality, COPD exacerbations, or lung function decline (15–18). However, exploration for the association between genetic variants and blood biomarker levels has been seldom investigated in COPD, although in a candidate gene study our group recently reported strong associations between variants in the gene encoding SP-D (SFTPD) and both SP-D levels and COPD risk (19).

We hypothesized that GWAS would identify genetic variants associated with the levels of circulating protein biomarkers related to COPD and that genetic association studies of these COPD biomarkers could lead to the identification of genetic determinants of COPD that did not reach genome-wide significance in previous GWAS of COPD. We selected seven biomarkers for this genome-wide association analysis, which were classified as either lung-specific markers (pneumoproteins) or as systemic inflammatory markers that have been extensively studied in COPD.

To test these hypotheses, we conducted association analyses in three steps (Figure 1). First, GWAS was conducted to localize the genetic loci associated with the circulating levels of two pneumoproteins (CC16 and SP-D) and five inflammatory biomarkers (fibrinogen, CRP, IL-6, IL-8, and TNF- α) in subjects with COPD from the Evaluation of COPD Longitudinally to Identify Predictive Surrogate Endpoints (ECLIPSE) cohort. Then, for the significantly associated single nucleotide polymorphisms (SNPs) in biomarker GWAS, we studied their association with mRNA expression levels of each biomarker in induced sputum samples and susceptibility to develop COPD in a collaborative mega-analysis of 2,939 cases of COPD and 1,380 smokers with normal lung function. These genetic association results were tested in two additional cohorts.

Figure 1.

A schematic overview of the genetic analyses in this study. COPD = chronic obstructive pulmonary disease; GWAS = genome-wide association study; ICGN = International COPD Genetics Network; SNPs = single nucleotide polymorphisms.

Methods

Study Populations

GWAS for blood biomarker levels were performed in patients with COPD recruited from The ECLIPSE study (ClinicalTrials.gov identifier NCT 00292552; GlaxoSmithKline study code SCO104960). The details of the ECLIPSE cohort have been reported elsewhere (19–21). Briefly, cases of COPD aged 40–75 years old and with greater than or equal to 10 pack-years of smoking at the time of enrollment were included; COPD was defined by post-bronchodilator spirometric criteria (post-bronchodilator FEV₁/FVC ratio <0.7 and a post-bronchodilator FEV₁ <80% predicted). Among 1,981 genotyped cases of COPD with self-reported white ethnicity, 1,951 subjects were measured for at least one of the seven biomarkers.

Gene expression mRNA levels of biomarker genes were assayed from induced sputum in a subset of ECLIPSE subjects (141 subjects) (22). The sputum samples were collected at the same time as blood sample collection for biomarker measurements.

The association of genome-wide significant SNPs with susceptibility to COPD was tested in a previously described combined case-control population (21), which included subjects recruited from the ECLIPSE study, Norway GenKOLS study, and the National Emphysema Treatment Trial (NETT)–Normative Aging Study (NAS). One subject was subsequently found to be a pipe smoker only and was excluded from this analysis.

Two additional cohorts were used to attempt to validate the SNPs associated with susceptibility to COPD in GWAS. From the International COPD Genetics Network (ICGN), which has been described elsewhere (23–25), a total of 983 probands and 1,876 siblings (all whites) were genotyped for family-based association analysis (21). The other validation cohort is the COPDGene Study (www.COPDGene.org) (26). For this study, the first 999 subjects (498 cases and 501 control subjects) who are all non-Hispanic whites were used for association analysis. Details on each cohort are available in the online supplement.

Genotyping, Quality Control, and Population Stratification

Genotyping was performed using Illumina platforms (HumanHap 550 V3 for the ECLIPSE cohort and the HumanHap 550 [V1, V3, and Duo] for the Norway GenKOLS cohort; Illumina, Inc., San Diego, CA). The Illumina Quad 610 (Illumina) was used for genotyping of the NETT-NAS cohort. Details on quality control and adjustment for population stratification using principal component analysis in these cohorts were published elsewhere (20, 21). Additional genotyping for candidate SNPs highly associated with circulating biomarkers was performed in the ICGN subjects using Sequenom iPLEX SNP genotyping protocol (San Diego, CA). The genome-wide SNP genotyping performed in the first 999 COPDGene subjects with the Illumina Omni1 Quad platform (Illumina) was also used for replication of COPD genetic associations (27).

Sputum Induction, RNA Isolation, and Microarray Analysis

Details on sputum induction and quantification, isolation, and amplification of RNA, and subsequent gene expression profiling were described elsewhere (22, 28, 29) and a brief summary is also available in the online supplement.

Measurement of Biomarkers

CC16, SP-D, fibrinogen, CRP, IL-6, IL-8, and TNF-α were measured in plasma of ECLIPSE subjects; study visits were scheduled at least 4 weeks after any recent COPD exacerbations. The methods for measuring these seven biomarkers have been described elsewhere (10, 14, 28, 30, 31) and details are available in the online supplement.

Statistical Analysis

For GWA analysis, plasma levels of biomarkers except fibrinogen were transformed to a natural log scale to approximate a normal distribution. When the level of a biomarker was below the lower limit of quantification, it was assigned half the value of lower limit of quantification. The proportion of cases below the lower limit of quantification was very low in ECLIPSE cases of COPD (0.05% for CC16; 0% for SP-D, hs-CRP, and fibrinogen; and 2.13% for IL-8) except for IL-6 (23.8%) and TNF-α (70.9%). GWAS were done in PLINK version 1.05 (pngu.mgh.harvard.edu/∼purcell/plink/) using linear and logistic regression under an additive model. Genome-wide significant SNPs were determined by a conservative P value less than 1 × 10⁻⁸. Markers were excluded in PLINK analysis when their minor allele frequency (MAF) was less than 1% or if they had extreme Hardy-Weinberg deviation (P value < 10⁻⁸).

To minimize the effects of confounders, we adjusted biomarker GWAS and COPD genetic association analyses for some covariates. To select the most proper covariates across all biomarkers, multivariate analyses were done for each biomarker adjusting for well-known confounders in COPD (see Table E1 in the online supplement), and age, sex, amount of smoking in pack-years, and current smoking status were the most consistently significant variables associated with each biomarker. Biomarker GWAS were adjusted for covariates including the previously mentioned four variables and principal components for genetic ancestry produced by a modified EIGENSTRAT method (32). To estimate the effect sizes of SNPs identified by the GWAS on the plasma levels of CC16 and SP-D, we generated a pruned subset of relatively independent SNPs based on lack of strong linkage disequilibrium (threshold r² of 0.5). The impact of the individual and combined SNPs on plasma protein levels was determined by the change of adjusted r² values in regression analyses.

In the ECLIPSE cohort, seven principal components for genetic ancestry were adjusted. Principal components for genetic ancestry and age and pack-years of smoking (since NAS subjects were all male and NETT subjects were all ex-smokers) were adjusted in the collaborative COPD GWAS populations.

In the COPDGene replication cohort, principal components for genetic ancestry were also adjusted to minimize the confounding effect caused by any population substructure that may exist. In the International COPD Genetics Network, family-based association analysis was performed with the PBAT program. Details on covariates in each analysis are described in the online supplement. Quantile-quantile (Q-q) plots, Manhattan plots, and SNP annotation were performed using the WGAViewer (http://compute1.lsrc.duke.edu/softwares/WGAViewer/, version 1.26G, Duke Institute for Genome Sciences & Policy, Durham, NC) (33). Regional association results of genome-wide significant SNPs were plotted using LocusZoom (http://csg.sph.umich.edu/locuszoom, version 1.1, University of Michigan, Ann Arbor, MI) (34).

Because different genotyping chips from Illumina were used in ECLIPSE and COPDGene, genotype imputation was performed in the COPDGene population. The details of this imputation process were described previously (27) and a brief summary is also available in the online supplement. Additional analytical details are also provided in the online supplement.

Results

Characteristics of Study Population and Blood Biomarker Levels in COPD

Table 1 summarizes the main demographic and clinical characteristics of the 1,951 ECLIPSE subjects with COPD studied here (GWAS). Most of them were males older than 60 years of age with significant cumulative smoking exposure (pack-years) and 36% of subjects were still smoking. Airflow limitation was moderate to very severe. Table 1 also shows the mean (± SD) of the circulating levels of the seven biomarkers studied. CRP levels were significantly correlated with those of fibrinogen, IL-6, and IL-8, whereas CC16 was weakly correlated with the other pneumoprotein, SP-D (Table 2).

TABLE 1.

CHARACTERISTICS OF ECLIPSE STUDY POPULATION

Variable	ECLIPSE Subjects
N	1,951
Male, %	1,288 (66)
Age	63.6 ± 7
Smoking (current), %	703 (36)
Pack-years	49.1 ± 27.3
FEV1, % predicted	44.1 ± 15
CC16, ng/ml	5.6 ± 3.3
SP-D, ng/ml	136.5 ± 75.8
Fibrinogen, mg/dl	457.1 ± 99
IL-6, pg/ml	5.2 ± 20.3
IL-8, pg/ml	14.4 ± 33.1
TNF-α, ng/ml	65.9 ± 994.6
hsCRP, mg/L	6.9 ± 12.2

Definition of abbreviations: CC16 = Clara cell secretory protein; ECLIPSE = Evaluation of COPD Longitudinally to Identify Predictive Surrogate Endpoints; hsCRP = C-reactive protein; SP-D = surfactant protein D; TNF = tumor necrosis factor.

N (%) or mean ± standard deviation.

TABLE 2.

CORRELATION BETWEEN BIOMARKERS IN SUBJECTS WITH COPD^*

Correlation	Log SP-D (n = 1,899)	Fibrinogen (n = 1,540)	Log IL-6 (n = 1,881)	Log IL-8 (n = 1,880)	Log TNF-α (n = 1,889)	Log CRP (n = 1,632)
Log CC16	0.16 (<0.0001)	−0.002 (0.93)	0.09 (0.0001)	0.06 (0.02)	−0.02 (0.32)	0.04 (0.09)
Log SP-D		0.03 (0.28)	0.03 (0.20)	0.03 (0.23)	0.02 (0.29)	0.01 (0.78)
Fibrinogen			0.10 (<0.0001)	0.02 (0.43)	0.03 (0.22)	0.38 (<0.0001)
Log IL-6				0.06 (0.01)	−0.05 (0.04)	0.23 (<0.0001)
Log IL-8					−0.03 (0.17)	0.06 (0.01)
Log TNF-α						0.04 (0.11)

Definition of abbreviations: CC16 = Clara cell secretory protein; COPD = chronic obstructive pulmonary disease; CRP = C-reactive protein; SP-D = surfactant protein D; TNF = tumor necrosis factor.

^*Pearson correlation coefficient (P value).

GWAS of Circulating Biomarkers in COPD

A total of 588,352 genotyped SNPs were included in the analysis after applying quality control filters in ECLIPSE data. Q-q plots of the observed against expected P value distributions revealed markedly different patterns in pneumoproteins and systemic inflammatory biomarkers. Q-q plots of CC16 and SP-D revealed prominent deviations of allelic-association P values beyond expected P values (Q-q plots are shown in Figure E1; lambda values are 0.994 for CC16 and 1.006 for SP-D), suggesting significant associations between SNPs and circulating levels of these biomarkers, whereas the other Q-q plots did not show an excess of low P values. The top 10 SNPs associated with the inflammatory biomarkers including fibrinogen, IL-6, IL-8, TNF-α, and CRP, none of which reached genome-wide statistical significance, are listed in Table E2. Based on these results, further evaluation focused only on CC16 and SP-D.

Eleven SNPs showed genome-wide significant associations with circulating levels of CC16, and the SNP with the highest rank was rs3741240 (P = 1.42 × 10⁻²⁶; MAF 0.36), which is located in the 5′ UTR of the CC16 coding gene (SCGB1A1) (Table 3). SNPs close to AHNAK and ASRGL1 were also highly associated with CC16 levels. Although all of these genome-wide significant SNPs were located on the chromosome containing SCGB1A1 (chromosome 11), we found two discrete foci of associated SNPs: one region is located near SCGB1A1, and the other region is approximately 25 Mb away from SCGB1A1 across the centromere.

TABLE 3.

SNPs HIGHLY ASSOCIATED WITH CIRCULATING CC16 LEVELS AND SP-D LEVELS OF SUBJECTS WITH COPD

	SNP	Rank	P	CHR	Coordinate	Type	Closest Gene	Distance to Gene	Minor Allele	MAF	HWE
CC16	rs3741240	1	1.42 × 10−26	11	62186542	5′ UTR	SCGB1A1	0	A	0.36	0.05
	rs2077224	2	1.67 × 10−14	11	62197427	Downstream	AHNAK	3,589	A	0.38	0.63
	rs2509956	3	1.03 × 10−13	11	62196723	Downstream	AHNAK	4,293	C	0.37	0.96
	rs10836312	4	6.72 × 10−11	11	34810443	Intergenic	APIP	93,402	C	0.41	0.93
	rs2463822	5	1.02 × 10−10	11	62103420	Upstream	ASRGL1	−1,354	T	0.12	1
	rs906902	6	2.42 × 10−10	11	34780278	Intergenic	EHF	97,197	A	0.40	0.82
	rs10466455	7	2.63 × 10−10	11	34780936	Intergenic	EHF	97,855	C	0.40	0.85
	rs2463836	8	3.11 × 10−10	11	62126494	Intronic	ASRGL1	0	A	0.12	1
	rs2246920	9	3.48 × 10−10	11	62104937	Regulatory region; 5′ UTR	ASRGL1	0	T	0.12	1
	rs17157266	10	1.29 × 10−09	11	62199817	Downstream	AHNAK	1,199	C	0.18	0.07
	rs7929679	11	7.26 × 10−09	11	34805849	Intergenic	APIP	97,996	G	0.49	0.72
SP-D	rs1885551	1	1.16 × 10−39	10	81712353	Intronic	SFTPD	0	G	0.10	0.39
	rs2146192	2	1.16 × 10−39	10	81715738	Intronic	SFTPD	0	C	0.10	0.39
	rs7084667	3	1.14 × 10−29	10	81748708	Intergenic	SFTPD	−5,712	T	0.34	0.72
	rs3923564	4	1.76 × 10−27	10	81735981	Intronic	SFTPD	0	G	0.04	1
	rs1885553	5	5.33 × 10−21	10	81711711	Intronic	SFTPD	0	C	0.33	0.69
	rs1932571	6	5.37 × 10−20	10	81761489	Intergenic	RP11-369J21.6	15,944	T	0.18	0.22
	rs7905482	7	8.37 × 10−15	10	81863964	Intergenic	RP11-369J21.2	11,651	T	0.36	0.30
	rs911887	8	7.15 × 10−14	10	81701523	Intronic	SFTPD	0	C	0.42	0.75
	rs12778754	9	2.19 × 10−13	10	81869642	Intergenic	RP11-369J21.4	−16,055	A	0.31	0.27
	rs8048576	10	8.96 × 10−13	16	84423034	Intronic	ATP2C2	0	A	0.12	0.92
	rs728616	11	2.02 × 10−12	10	81847914	Intronic	RP11-369J21.2	0	T	0.08	0.75
	rs3851050	12	1.27 × 10−11	10	81882370	Upstream	RP11-369J21.4	−3,327	C	0.41	0.43
	rs1265110	13	2.76 × 10−11	6	31119422	Intronic	CCHCR1	0	T	0.11	0.49
	rs12220777	14	6.69 × 10−11	10	81788104	Upstream	RP11-369J21.5	−3,189	C	0.08	7.3 × 10−6
	rs6585424	15	1.38 × 10−10	10	81933748	Intronic	ANXA11	0	G	0.13	0.48
	rs6585454	16	1.38 × 10−10	10	81940701	5′ UTR	ANXA11	0	G	0.13	0.48
	rs8063863	17	1.56 × 10−10	16	84400850	Upstream	ATP2C2	−1,283	T	0.15	0.41
	rs2074488	18	2.28 × 10−10	6	31240431	Upstream	HLA-C	−524	T	0.13	0.85
	rs9266629	19	3.99 × 10−10	6	31346822	Upstream	FGFR3P	1,026	C	0.21	0.11
	rs11201799	20	1.44 × 10−09	10	81881954	Upstream	RP11-369J21.4	−3,743	A	0.09	0.40
	rs7078012	21	4.95 × 10−09	10	81705433	Intronic	SFTPD	0	T	0.13	0.06
	rs1923539	22	5.03 × 10−09	10	81694950	Within noncoding gene	RP11-479O17.4	0	A	0.25	0.60
	rs1265093	23	5.78 × 10−09	6	31107187	Intronic	PSORS1C1	0	A	0.26	0.27
	rs3130559	24	8.34 × 10−09	6	31097301	Intronic	PSORS1C1	0	T	0.20	0.63

Definition of abbreviations: CC16 = Clara cell secretory protein; CHR = chromosome; COPD = chronic obstructive pulmonary disease; HWE = P value for deviation from Hardy-Weinberg equilibrium; MAF = minor allele frequency; SNP = single nucleotide polymorphisms; SP-D = surfactant protein D.

Twenty-four genome-wide significant SNPs were associated with circulating levels of SP-D in cases of COPD (Table 3). SFTPD, the SP-D coding gene, is located on chromosome 10, and the highest ranked associations were with two SNPs upstream from the transcription start site for SFTPD (P = 1.16 × 10⁻³⁹; MAF 0.10) (Table 3). Some of these SFTPD SNPs have been previously reported to be associated with SP-D level in a candidate gene analysis by our study group (35). Our genome-wide analysis also revealed that loci encompassing PSORS1C1 and HLA-C were associated with SP-D blood levels. SNPs near those genes were previously reported to be associated with COPD susceptibility in an analysis of sputum gene expression quantitative trait loci in the ECLIPSE Study (22). In addition to these associations on chromosome 10, genome-wide significant associations for SP-D levels were also found on chromosomes 6 and 16. The multiple genomic regions of association to CC16 and SP-D levels are depicted in Figure 2 and Figures E2 and E3.

Figure 2.

Manhattan plots for (A) Clara cell secretory protein and (B) surfactant protein D. Genome-wide significant single nucleotide polymorphisms are noted in red and closest genes to them are also listed at the bottom of plots. GWAS = genome-wide association study.

To identify the contribution of relatively independent SNPs to the plasma protein levels of CC16 and SP-D, we removed SNPs in strong linkage disequilibrium (r² > 0.5), resulting in 5 SNPs for CC16 and 12 SNPs for SP-D. The differences in circulating levels of CC16 by genotypes of the independent SNPs are presented in Figure 3. One SNP (rs2463822) near ASRGL1 showed the most prominent contribution by genotype to circulating CC16 levels. Additionally, each of these SNPs individually and in combination contributed to the increased variance explained in regression models. The difference in adjusted r² by the addition of all independent SNPs in the regression model was 0.06 for CC16 and 0.167 for SP-D (see Table E3), suggesting that the identified SNPs contribute to the overall level of variation in plasma CC16 or SP-D.

Figure 3.

Differential circulating levels of Clara cell secretory protein (CC16) stratified by genotype of top genome-wide association study significant associations for CC16. Top single nucleotide polymorphisms were classified by the closest genes. Mean (column) and standard deviation (error bar) of CC16 level are shown.

mRNA Expression of CC16 and SP-D Coding Genes in Sputum

Circulating levels of CC16 had a weak tendency to be correlated with SCGB1A1 gene expression in sputum (Pearson correlation coefficient ρ = 0.17; P = 0.06), and SFTPD gene expression in sputum was more significantly associated with blood levels of SP-D (ρ = 0.23; P = 0.009). The correlations between circulating biomarkers and mRNA expression of biomarker coding genes are summarized in Table E4.

Five of 11 SNPs associated with CC16 levels at genome-wide significance (rs10466455, rs10836312, rs906902, rs3741240, and rs2509956) were significantly associated with gene expression of SCGB1A1 in sputum (Table 4). These five SNPs are located in two discrete foci. Three of them (rs10466455, rs10836312, and rs906902) are in the short arm of chromosome 11 and the others (rs3741240 and rs2509956) are located in the long arm of chromosome 11. The presence of the minor allele in each of these SNPs was negatively associated with the level of SCGB1A1 mRNA expression in sputum, and it was in the same direction as the association of the minor allele with the circulating levels of CC16 (see Table E4). In terms of SFTPD expression, none of 24 genome-wide significant SNPs related to the levels of circulating SP-D was significantly associated with SFTPD expression in sputum, but three of them (rs1923539, rs1885551, and rs2146192) showed marginal P values (Table 5).

TABLE 4.

THE ASSOCIATION OF SNPs AFFECTING BLOOD CC16 LEVELS WITH THE LEVELS OF mRNA EXPRESSION OF SCGB1A1 IN SPUTUM

Genome-Wide Significant SNPs Affecting Blood CC16 Levels					Association with mRNA Expression of SCGB1A1 in Sputum
CHR	SNP	BP	Minor Allele	Nearest Gene	Beta	95% CI	P
11	rs10466455	34737512	C	EHF	−0.65	−1.12 to −0.17	0.009
11	rs10836312	34767019	C	APIP	−0.65	−1.12 to −0.17	0.009
11	rs906902	34736854	A	EHF	−0.65	−1.12 to −0.17	0.009
11	rs3741240	61943118	A	SCGB1A1	−0.52	−0.96 to −0.09	0.020
11	rs2509956	61953299	C	AHNAK	−0.55	−1.05 to −0.05	0.034

Definition of abbreviations: Beta = regression coefficient; BP = physical position (base pair); CC16 = Clara cell secretory protein; CHR = chromosome; CI = confidence interval (lower, upper); SNP = single nucleotide polymorphism.

The linear models were adjusted for age, sex, pack-years, current smoking status, and principal components for genetic ancestry.

Only SNPs with P < 0.05 are shown.

TABLE 5.

THE ASSOCIATION OF SNPs AFFECTING BLOOD SP-D LEVELS WITH mRNA EXPRESSION OF SFTPD IN SPUTUM

Genome-Wide Significant SNPs Affecting Blood SP-D Levels					Association with mRNA Expression of SFTPD in Sputum
CHR	SNP	BP	Minor Allele	Nearest Gene	Beta	95% CI	P
10	rs1923539	81684930	A	RP11-479O17.4	0.09	−0.001 to 0.17	0.055
10	rs1885551	81702333	G	SFTPD	0.10	−0.005 to 0.20	0.065
10	rs2146192	81705718	C	SFTPD	0.10	−0.005 to 0.20	0.065

Definition of abbreviations: Beta = regression coefficient; BP = physical position (base-pair); CHR = chromosome; CI = confidence interval (lower, upper); SNP = single nucleotide polymorphism; SP-D = surfactant protein D.

The linear models were adjusted with age, sex, pack-years, current smoking status, and principal components for genetic ancestry.

No SNPs with P < 0.05 were observed; SNPs close to P = 0.05 are shown even though they are not statistically significant.

SNPs Highly Associated with Circulating Levels of Biomarkers and COPD Susceptibility

In ECLIPSE, 11 SNPs showed genome-wide significant associations with circulating levels of CC16 and 24 genome-wide significant SNPs were associated with circulating levels of SP-D; these 35 SNPs were tested for association with COPD affection status in the COPD GWAS collaborative cohorts (in which all cases were combined in one group and all control subjects were combined in another group) and in two validation cohorts. In the COPD GWAS collaborative mega-analysis of 2,939 patients with COPD compared with 1,380 smokers with normal lung function, one SNP affecting CC16 levels and five SNPs affecting SP-D levels were nominally associated with the presence of the disease (P < 0.05) (Table 6). One of five SNPs affecting SP-D levels (rs7078012) was reported to be associated with COPD in our previous report (19). In separate analyses of the NETT-NAS and Norway GenKOLS study populations within these COPD GWAS collaborative study populations, significant associations were not found with these SNPs even though the trends for COPD susceptibility were consistent (see Table E6). The CC16 SNP (rs17157266) and four SFTPD SNPs (all except rs1885553, which had inconsistent direction of association in the ECLIPSE cohort compared with NETT-NAS and Norway) were tested for association in the COPDGene (first 999 subjects) and ICGN studies. None of these five SNPs were significantly associated with COPD susceptibility in the ICGN or COPDGene cohorts.

TABLE 6.

THE ASSOCIATION OF TOP BIOMARKER GWAS SNPs WITH RISK OF COPD IN THE COMBINED COLLABORATIVE COPD GWAS MEGA-ANALYSIS

Protein	CHR	SNP	BP	Nearest Gene	Minor Allele	Test	OR	95% CI	P*
CC16	11	rs17157266	61956393	AHNAK	C	ADD	1.20	1.05 to 1.37	0.008
Surfactant protein D	16	rs8063863	82958351	ATP2C2	T	ADD	0.80	0.70 to 0.92	0.001
	16	rs8048576	82980535	ATP2C2	A	ADD	0.82	0.71 to 0.95	0.008
	10	rs7078012	81695413	SFTPD	T	ADD	0.84	0.73 to 0.97	0.017
	10	rs1885553	81701691	SFTPD	C	ADD	1.11	1.00 to 1.24	0.049
	10	rs1923539	81684930	RP11-479O17.4	A	ADD	1.13	1.00 to 1.27	0.043

Definition of abbreviations: BP = physical position (base-pair); CC16 = Clara cell secretory protein; CHR = chromosome; CI = confidence interval (lower, upper); COPD = chronic obstructive pulmonary disease; GWAS = genome-wide association study; OR = odds ratio; SNP = single nucleotide polymorphism.

Only SNPs with P < 0.05 are shown.

^*P values adjusted for age, pack-years of smoking, and 16 principal components for genetic ancestry.

SNPs Highly Associated with Circulating Levels of CC16 and SP-D and COPD-related Phenotypes

Because various COPD biomarkers have been reported to be associated with clinical phenotypes including body mass index and exacerbation frequency (14, 36, 37), we performed an association analysis of the six candidate SNPs associated with COPD susceptibility listed in Table 6 and several COPD clinical phenotypes in the ECLIPSE cohort. We found that SP-D-associated SNPs near the ATP2C2 and SFTPD genes showed significant associations with fat-free mass index, and that one of them was also associated with COPD exacerbation frequency (Table 7) at nominal levels of statistical significance.

TABLE 7.

THE ASSOCIATION OF TOP BIOMARKER GWAS SNPs WITH COPD-RELATED PHENOTYPES IN ECLIPSE^*

			FEV1		BMI		FFMI		% Emphysema		Exacerbation/2 Years
Biomarkers	SNP	Nearest Gene	Beta	P	Beta	P	Beta	P	Beta	P	Beta	P
CC16	rs17157266	AHNAK	−0.53	0.60	0.02	0.92	−0.06	0.58	−0.48	0.36	0.22	0.04
SP-D	rs8063863	ATP2C2	−0.79	0.24	−0.47	0.06	−0.30	0.02	−0.01	0.98	0.24	0.04
	rs8048576	ATP2C2	−0.68	0.35	−0.41	0.13	−0.32	0.02	−0.42	0.52	0.12	0.33
	rs7078012	SFTPD	0.09	0.89	0.73	0.005	0.33	0.01	−0.52	0.40	0.10	0.39
	rs1885553	SFTPD	0.30	0.55	0.39	0.04	0.11	0.26	−0.61	0.17	0.06	0.49
	rs1923539	RP11-479O17.4	−0.004	0.99	0.21	0.32	−0.02	0.83	−0.76	0.12	0.03	0.77

Definition of abbreviations: % Emphysema = % of low-attenuation area at −950 HU; BMI = body mass index; CC16 = Clara cell secretory protein; COPD = chronic obstructive pulmonary disease; ECLIPSE = Evaluation of COPD Longitudinally to Identify Predictive Surrogate Endpoints; Exacerbation/2 Years = reported frequency of COPD exacerbations during 2 years of follow-up; FFMI = fat-free mass index; GWAS = genome-wide association study; SNP = single nucleotide polymorphism; SP-D = surfactant protein D.

^*P values adjusted for age, sex, pack-years of smoking, current smoking status, and seven principal components for genetic ancestry.

Discussion

The results of this study: (1) identified genome-wide significant associations of several SNPs with circulating levels of CC16 and SP-D, whereas GWAS of fibrinogen, IL-6, IL-8, TNF-α, and CRP did not show any significant association; (2) revealed that remote genetic loci (i.e., loci that are located megabases away from the coding gene on the same chromosome or are located on a different chromosome) and biomarker-coding genes are associated with the blood levels of several protein biomarkers in COPD; and (3) showed that a subset of these protein quantitative trait loci may influence mRNA expression in sputum or COPD susceptibility.

Elevation of inflammatory biomarkers including fibrinogen, IL-6, IL-8, TNF-α, and CRP in COPD has been previously reported (17). Likewise, there have been previous studies of genetic associations of fibrinogen, CRP, and TNF-α levels in diseases other than COPD (6, 38, 39). To our knowledge, this study is not only the first GWAS of CC16, SP-D, IL-6, IL-8, and TNF-α in COPD, but also the first GWAS analysis integrating multiple biomarkers and assessing the association of the protein level, mRNA expression, and susceptibility to COPD.

In our study, we did not find a significant association of circulating inflammatory biomarkers with genetic variants including their respective coding genes, whereas the circulating pneumoproteins, CC16 and SP-D, were significantly associated with multiple genetic variants. One potential explanation for these different association results may be that genetic effects on circulating levels of SP-D and CC16 could have less complex genetic architecture, or that they may be less influenced by comorbid illnesses because systemic inflammatory markers could be affected by other inflammation-related morbidities including cardiovascular disease, cancer, obesity, and diabetes mellitus (31, 40). In addition, inflammatory cytokines, such as IL-6 and CRP, could be transcriptionally modulated during inflammation in hepatocytes (41), which might weaken associations of circulating biomarkers and their coding genes. This hypothesis receives some support by our finding that the top SNPs related to CRP level included the transcription factor hepatocyte nuclear factor (HNF1A); SNPs in this gene have been previously related to CRP and fibrinogen levels (see Table E2) (42, 43). As a result, genetic effects on circulating levels of systemic inflammatory biomarkers may be more difficult to detect than for pneumoproteins.

Our study showed that genetic effects on circulating biomarkers may be different by the type of biomarkers (i.e, pneumoprotein vs. systemic inflammatory biomarkers), and this finding may provide some insight into the source of systemic inflammatory markers in COPD. The origin of systemic inflammatory biomarkers in COPD has been debated; one common hypothesis is that these biomarkers spill over from pulmonary inflammation into the systemic circulation. In contrast to systemic biomarkers, CC16 and SP-D are known to be synthesized within respiratory tissues including endothelial cells, Clara cells, and type II pneumocytes. Therefore, the constitutional blood levels of pneumoproteins have been explained by leakage from the respiratory tract into the systemic circulation when the air-blood barrier integrity is damaged, reflecting intrapulmonary pathologic changes. In this respect, the different effects of genetic variants on systemic levels of biomarkers may be an indirect clue that the origin of systemic inflammatory biomarkers might be beyond the respiratory system.

Another point to be emphasized in our study is that we found distant genetic loci affecting circulating levels of CC16 and SP-D, and one of the genetic loci we identified as associated with CC16 levels (rs10836312) is in linkage disequilibrium with SNPs associated with lung disease severity in cystic fibrosis. The effects of genetic variants in or near the CC16 and SP-D coding genes on circulating levels of their respective proteins have been previously reported (19, 44, 45) and the genetic variants in corresponding genes were also listed among the top SNPs in our study. However, there were remote genetic loci affecting the levels of CC16 and SP-D, which have no significant linkage disequilibrium with SNPs in the biomarker coding genes. The remote genetic loci affecting the level of CC16 are located on 11p13, near APIP and EHF, respectively. The different levels of circulating CC16 by genotype are shown in Figure 3. However, the associated SNPs are located near the APIP and EHF genes but they are not associated with variants within the coding regions of those genes. EHF belongs to an ETS transcription factor subfamily characterized by epithelial-specific expression. The encoded protein acts as a transcriptional repressor and has been associated with asthma and carcinogenesis. Even though a role of APIP or EHF in COPD has not been reported, recently SNPs at 11p13 were reported as lung disease severity modifying loci in patients with cystic fibrosis who were delta F508 homozygotes (46). The top SNP in that analysis at 11p13, rs12793173, is in weak LD (r² of 0.21, D’ of 0.98) with our top association in this region, rs10836312. Thus, it is possible that this genomic region plays a role in influencing CC16 levels and multiple obstructive lung diseases. Despite robust genome-wide significant genetic variations, further research is required to determine the biologic processes influenced by the remote genetic loci influencing circulating CC16 and SP-D levels.

We also found that circulating CC16 and SP-D protein levels showed some correlation with mRNA expression of the corresponding coding gene (SCGB1A1 for CC16 and SFTPD for SP-D) in sputum. Furthermore, some of the most significantly associated SNPs with CC16 protein levels were also significantly associated with CC16 mRNA levels, although SNPs associated with blood levels of SP-D failed to demonstrate statistically significant associations with SFTPD mRNA levels. These findings may be related to the anatomic distribution of cells producing these biomarkers. CC16 is not uniformly expressed in the respiratory tract but predominantly in trachea, bronchi, and terminal bronchioles (47, 48), and sputum may be a more suitable specimen for reflecting circulating CC16. Contrary to CC16, SP-D is sparsely expressed in conducting airways (11). Poor correlation of blood protein level with mRNA expression level for SP-D has been well known (49), potentially related to post-transcriptional modification and half-life of this protein biomarker. However, considering our results, the poor correlation of circulating biomarkers with mRNA expression might relate to selection of the target tissues used to measure gene expression. As a final analysis step, we showed that genetic variants derived from GWAS for circulating biomarkers in COPD could be a surrogate marker for assessing COPD susceptibility in a large combined population. It has been reported that some genetic determinants of circulating biomarkers are associated with blood protein levels and even disease susceptibility (6, 38, 39, 50, 51). Thus far, reports using genetic determinants of biomarkers as candidate loci for disease susceptibility have been rare; our results suggest that this may be a useful approach to assess the genetic risk factors of COPD and other complex diseases.

The results of this study revealed associations of SP-D blood levels near the promising loci encompassing PSORS1C1 and HLA-C for COPD susceptibility. In a previously reported analysis of sputum expression quantitative trait loci in the ECLIPSE Study, SNPs close to PSORS1C1 and HLA-C were associated with COPD susceptibility (22). Even though association with COPD susceptibility was not detected in the current study, loci neighboring PSORS1C1 and HLA-C were significantly associated with circulating levels of SP-D. These genes were reported to be associated with psoriasis susceptibility and an epidemiologic association between psoriasis and COPD has been also reported (52, 53). Therefore, further studies for the functional validation of these loci are necessary.

Our study provides novel findings but has some limitations that deserve comment. First, the association of genetic determinants of COPD biomarkers with COPD susceptibility was not replicated in two other populations (the ICGN and the first 999 subjects in the COPDGene study). This lack of replication may relate to the heterogeneity between cohorts, which used different inclusion criteria. In addition, small sample sizes of our replication cohorts and relatively weak associations of these SNPs with COPD in the ECLIPSE population also could have contributed to the failure of replication. Based on the nominal levels of association to COPD and the linkage disequilibrium between SNPs within regions of association to CC16 or SP-D, formal adjustment for multiple statistical testing was not performed. Further validation of our COPD associations in larger populations is required. Second, because the levels of circulating biomarkers could be affected by many environmental confounders, such as medication treatment and by disease subtype and severity, the optimal approach for covariate adjustment is uncertain; we used a standard set of covariates for all of the biomarkers that we studied. Finally, some of the biomarkers had measurements below the lower limit of quantification, and this likely limited our power to detect significant associations.

In conclusion, we performed GWAS of circulating COPD biomarkers and found some novel loci affecting the levels of plasma protein biomarkers. These biomarker loci may also influence COPD susceptibility, although further confirmation is required. Thus, GWAS of biomarkers may be a useful approach to search for new genetic determinants of complex diseases.