APOM and High-Density Lipoprotein are associated with Lung Function and Percent Emphysema

Abstract

Chronic obstructive pulmonary disease (COPD) is linked to cardiovascular disease; however, there are few studies on the associations of cardiovascular genes with COPD.

We assessed the association of lung function with 2,100 genes selected for cardiovascular diseases among 20,077 European-Americans and 6,900 African-Americans. We performed replication of significant loci in the other racial group and an independent consortium of Europeans, tested the associations of significant loci with percent emphysema, and examined gene expression in an independent sample. We then tested the association of a related lipid biomarker with FEV₁/FVC and percent emphysema.

We identified one new polymorphism for FEV₁/FVC (rs805301) in European-Americans (p=1.3×10⁻⁶) and a second (rs707974) in the combined European-American and African-American analysis (p=1.38×10⁻⁷). Both SNPs flank the gene for apolipoprotein M (apoM), a component of HDL. Both replicated in an independent cohort. SNPs in a second gene related to apoM and HDL, PCSK9, were associated with FEV₁/FVC among African-Americans. rs707974 was associated with percent emphysema among European-Americans and African-Americans, and APOM expression was related to FEV₁/FVC and percent emphysema. Higher HDL levels were associated with lower FEV₁/FVC and greater percent emphysema.

These findings suggest a novel role for the APOM/HDL pathway in the pathogenesis of COPD and emphysema.

INTRODUCTION

Chronic obstructive pulmonary disease (COPD) is a leading cause of death globally[1] and is characterized by persistent airflow obstruction.[2,3] Emphysema is defined anatomically by permanent enlargement of airspaces distal to terminal bronchioles with destruction of alveolar walls.[4]

Familial studies suggest a genetic influence on COPD.[5-7] Recent genome-wide association studies (GWAS) have identified loci associated with the ratio of forced expiratory volume in one second to forced vital capacity (FEV₁/FVC) among participants of European ancestry.[8-13] Many of these genes have been shown to influence susceptibility to COPD;[14-16] however, they explained little more than 3% of the variance in lung function.

Emphysema also has a familial predisposition.[17] However, understanding of the genetic basis for emphysema, beyond alpha1-antirypsin deficiency, is more limited. A GWAS identified one genetic locus for radiologist-defined emphysema on computed tomography (CT) but none for quantitatively assessed emphysema.[18] Candidate gene association studies have identified additional genes for emphysema.[19-24]

Complimentary genotyping strategies to better delineate the genetic basis of COPD and emphysema are therefore warranted. One such strategy is a “gene-centric” genotyping chip, which includes a large panel of candidate genes and often better gene coverage than GWAS chips. No such chips have been designed specifically for lung disease; however, the ITMAT/Broad/CARe (IBC) chip[25] includes 2,100 candidate genes primarily selected for cardiovascular disease.

Respiratory and cardiac function are tightly linked at cellular,[26,27] physiologic,[28] structural,[29] and anatomic levels. For example, endothelial dysfunction is implicated in the pathogenesis of atherosclerosis[30] and emphysema in animal models[31-34] and humans,[35] the later via ceramide-mediated endothelial cell apoptosis.[36-38] High-density lipoprotein (HDL) may also be relevant to COPD and emphysema, as HDL increases in vitro ceramide levels.[39] HDL levels and function are affected by apolipoprotein M (apoM).[40-42]

We examined associations of FEV₁/FVC on the IBC chip in European-American and African-American participants in the Candidate-gene Association Resource (CARe) consortium.[43] Findings were replicated in the SpiroMeta consortium.[11] We performed additional analyses of identified genes with the percentage of emphysema-like lung (percent emphysema), of gene expression, and of HDL with lung function and percent emphysema in the Multi-Ethnic Study of Atherosclerosis (MESA) SNP Health Association Resource (SHARe) and MESA COPD Study.

METHODS

Study Samples

Analyses of Lung Function

The association of genes and lung function were assessed in the seven CARe cohorts that measured spirometry: Atherosclerosis Risk in Communities (ARIC), Coronary Artery Risk Development in young Adults (CARDIA), Cleveland Family Study (CFS), Cardiovascular Health Study (CHS), Framingham Heart Study (FHS), Jackson Heart Study (JHS) and the subset of MESA with spirometry. These cohorts have been previously described[44-53] and are summarized in the supplement. Exclusion criteria were lack of valid spirometric or genetic data, age less than 23 years and a restrictive pattern of spirometry, defined as FVC less than the lower limit of normal[54] and FEV₁/FVC of greater than 0.70.

Replication of Lung Function SNPs

Replication for FEV₁/FVC was performed in the SpiroMeta consortium, a large independent sample of 14 GWAS studies.[11] Replication in airflow obstruction was performed using publically available data from the SpiroMeta and CHARGE consortia,[9] which partly overlaps with European-American participants in the CARe consortium. Details are provided in the supplement.

Analyses of Percent Emphysema

Percent emphysema was examined among all participants in MESA SHARe, which comprises all participants who consented to genetic analyses in MESA,[44] MESA Family[55] and MESA Air Pollution[56] studies. Spirometry was not required.

Gene Expression Analyses

mRNA expression was examined in peripheral blood mononuclear cells in MESA COPD Study, an independent sample described in the supplement.

Appropriate Institutional Review Boards approved study protocols and written informed consent was obtained from all participants.

Phenotypic Measures

Spirometry

Pre-bronchodilator spirometry was performed by trained and certified spirometry technicians in accordance with the American Thoracic Society guidelines. Spirometry methods and equipment were highly standardized and in some cases identical across cohorts, as described in the supplement.

Percent emphysema

Percent emphysema was assessed in MESA SHARe on lung fields of cardiac CT scans, which image approximately 70% of lung volume from the carina to the lung bases, at a single center by trained readers, as previously described and validated compared to full-lung scans.[57] Percent emphysema was defined as percentage of total voxels in the lung less than −950 Hounsfield Units (HU). The MESA COPD Study used the same approach on full-lung scans using Apollo (Vida Diagnostics) software.

HDL

HDL was measured in EDTA plasma using the cholesterol oxidase method (Roche Diagnostics Corporation, Indianapolis, IN) after precipitation of non-HDL with magnesium/dextran.[58]

Genotyping

All CARe participants were genotyped with the IBC Illumina iSELECT array, a 50,000 gene-centric SNP array.[25] All genotyping was performed at a single center. Quality control methods are described in the supplement.

MESA SHARe participants were genotyped using the Affymetrix Genome-Wide Human SNP Array 6.0 platform at a single center.

Statistical Analyses

Analyses of candidate genes with FEV₁/FVC employed linear regression, stratified by race and adjusted for age, age², height, height², sex, smoking status, pack-years, pack-years², site (if applicable), and the first 10 principal components (PCs) for ancestry. Association testing of rank-normalized residuals was performed under an additive genetic model.[59-61]

Cohort-specific association results were meta-analyzed, again stratified by race, using inverse variance weighting in METAL[62] with cohort-specific and overall genomic control. A priori, we planned to replicate our top loci identified among European-Americans in the African-American cohorts and vice-versa as distinct cohorts. Race-specific results were then meta-analyzed in METAL[62] for the combined European-American and African-American analyses. The Bonferroni-adjusted thresholds for statistical significance in CARe were 1.31×10⁻⁶ in European-Americans and 1.13×10⁻⁶ in African-Americans and combined analyses, which are exceedingly conservative for the IBC chip.

Analyses for log-transformed percent emphysema used the same analytical approach supplemented with a linear mixed effects model for family-based-data,[60] and adjustment for age, sex, site, scanner, height, weight, tube current, cigarettes per day, pack-years, asthma and PCs.

Analysis details for gene expression and HDL association studies with lung function and percent emphysema are provided in the supplement.

To address multiple comparisons, we considered analysis of percent emphysema to be analogous to a modified Holm’s procedure[63] on the pathway of apoM. We hypothesized that SNPs rs805301 and rs707974 are in linkage disequilibrium (LD) with causative APOM variant that affects percent emphysema, which affects FEV₁/FVC, thus the Holm-Bonferonni corrected threshold for statistical significance for subsequent analyses was set at 0.025.

RESULTS

The mean age of the 26,977 CARe participants with spirometry was 54+/−13 years, 52% ever-smoked, with median pack-years of 20. Additional characteristics of the 20,077 European-American and 6,900 African-American participants are shown by cohort and race in table 1.

Table 1.

Characteristics of European-American and African-American CARe participants with spirometry in the Candidate-gene Association Resource (CARe) cohorts

	European-American Participants							African-American Participants							Combined
	ARIC	FHS	CHS	CARDIA	MESA	CFS	All	ARIC	JHS	CHS	CARDIA	MESA	CFS	All
Number	8,825	5,316	3,479	1,228	1,049	180	20,077	2,740	1,597	503	1,052	799	209	6,900	26,977
Age (years) mean±SD	54±6	48±13	73±6	36±3	65±10	51±15	55±13	53±6	50±12	73±6	34±4	65±10	48±14	52±13	54±13
Male Sex(%)	47	46	44	47	50	47	46	37	38	35	40	47	41	39	44
Height(cm) mean±SD	169±9	170±9	165±9	172±9	169±10	170±10	169±10	168±9	169±9	164±9	170±10	168±10	169±10	168±9	169±10
BMI# mean±SD	27±5	27±5	26±4	26±5	28±5	32±8	27±5	30±6	32±7	28±5	29±7	30±6	34±8	30±7	28±6
Cigarette Smoking(%)  Never  Former  Current	41 36 23	52 35 13	47 43 10	61 20 19	40 51 9	50 21 29	46 36 18	49 23 28	69 17 14	51 35	62 11 27	40 45 15	40 35 25	55 23 22	48 33 19
Pack years¶ median(IQR)	26 (12-40)	16 (7-30)	30 (13-50)	10 (4-19)	19 (8-35)	14 (5-32)	23 (10-38)	17 (8-30)	14 (7-26)	21 (10-38)	6 (3-11)	15 (6-28)	11 (5-24)	14 (6-27)	20 (9-36)
ppFEV1 mean±SD	94±16	98±14	89±22	99±11	92±16	96±18	94±17	97±17	95±15	92±24	100±14	94±19	92±19	96±17	95±17
FEV1/FVC mean±SD	0.74 ±0.08	0.76 ±0.07	0.69 ±0.10	0.79 ±0.06	0.73 ±0.09	0.77 ±0.07	0.74 ±0.09	0.76 ±0.08	0.80 ±0.09	0.70 ±0.11	0.81 ±0.06	0.75 ±0.10	0.78 ±0.08	0.77 ±0.09	0.75±0.09

Abbreviations: ARIC = Atherosclerosis Risk in Communities, CARDIA = Coronary Artery Risk Development in Young Adults, CFS = Cleveland Family Study, CHS = Cardiovascular Health Study, FHS = Framingham Heart Study, MESA = Multi-Ethnic Study of Atherosclerosis, JSH = Jackson Heart Study, SD = Standard Deviation, cm= centimeters, BMI = Body Mass Index, IQR = Inter Quartile Range, ppFEV₁ = percent predicted forced expiratory volume in one second, FEV₁/FVC = ratio of forced expiratory volume in one second over forced vital capacity.

^#BMI is the weight in kilograms divided by the square of the height in meters

^¶In ever smokers

Association Study of 2,100 Candidate Genes with Lung Function in CARe

Among European-Americans, we identified one new SNP (rs805301) for FEV₁/FVC (figure 1a). Among African-Americans, no SNPs were significantly associated with FEV₁/FVC using the Bonferroni cutoff; however, three SNPs were significant with the less conservative cutoff (p<x10⁻⁵; figure 1b). In the combined European-American and African-American analysis, we identified a second new SNP (rs707974) for FEV₁/FVC (figure 1c; table 2).

Figure 1.

Manhattan Plots of association results for FEV₁/FVC stratified by race and combined.

Manhattan Plots ordered by chromosome position of association results for FEV₁/FVC. Top 3 loci are labeled with arrows. (A) Meta-analysis of 38,294 SNPs among 20,077 European-American participants. The solid black line represents 1 × 10⁻⁶. (B) Meta-analysis of 44,416 SNPs among 6,900 African-American participants. The solid black line represents 1 × 10⁻⁶. (C) Meta-analysis of SNPs among combined European-American & African-American participants. The solid black line represents 1 × 10⁻⁶.

Abbreviations: FEV₁/FVC = ratio of forced expiratory volume in one second over forced vital capacity.

Table 2.

Top five SNPS associated with the FEV₁/FVC ratio among European-American and African-American participants in the Candidate-gene Association Resource (CARe) cohorts

European-American Participants (n=20,077)
SNP ID (function#)	Chr.	Gene¶	Coded allele	Allele freq.+	β§ (SE)	p-value
rs2070600 (ns)	6	AGER/RNF5	T	0.05	0.162 (0.026)	2.19 × 10−10
rs805301 (intron)	6	BAG6/APOM	C	0.37	− 0.054 (0.011)	1.32 × 10−6
rs1286664 (intron)	3	RARB	T	0.17	0.067 (0.014)	2.07 × 10−6
rs6941112 (intron)	6	STK19/C4B	A	0.33	0.053 (0.011)	2.90 × 10−6
rs3117582 (upstream)	6	BAG6	T	0.89	0.081 (0.018)	5.23 × 10−6

African-American Participants (n=6,900)
SNP ID (function#)	Chr.	Gene¶	Coded allele	Allele freq.+	β§ (SE)	p-value
rs1951269 (unknown)	14	NG/ NFKBIA	A	0.78	−0.097 (0.021)	4.08 × 10−6
rs707974 (3′UTR)	6	GPANK1/APOM	G	0.02	0.291 (0.064)	5.17 × 10−6
rs565436 (intron)	1	PCSK9	A	0.60	0.08 (0.018)	9.32 × 10−6
rs533375 (intron)	1	PCSK9	A	0.25	−0.083 (0.021)	5.29 × 10−5
rs7156874 (intron)	14	PSMA6	A	0.02	0.271 (0.068)	7.33 × 10−5

Combined European-American and African-American Participants (n=26,977)
SNP ID (function#)	Chr.	Gene¶	Coded allele	Allele freq.+	β§ (SE)	p-value
rs2070600 (ns)	6	AGER/RNF5	T	0.04	0.168 (0.025)	8.37 × 10−12
rs1286664 (intron)	3	RARB	T	0.18	0.067 (0.014)	1.60 × 10−8
rs805301 (intron)	6	BAG6/ APOM	C	0.43	−0.050 (0.009)	1.26 × 10−7
rs707974 (3′UTR)	6	GPANK1/ APOM	G	0.10	0.088 (0.016)	1.38 × 10−7
rs6941112 (intron)	6	STK19/C4B	A	0.30	0.052 (0.011)	1.22 × 10−6

The European-Americans SNPs represent five loci (r² range; 0.005-0.20).

Abbreviations: SNP= single nucleotide polymorphism, FEV₁/FVC= ratio of forced expiratory volume in one second over forced vital capacity, Chr = chromosome, Ref = reference, β = effect estimate, SE = standard error, ns=non-synonymous coding SNP, UTR = untranslated.

Gene abbreviations: AGER = advanced glycosylation end product-specific receptor (also known as RAGE), RNF5 = ring finger protein 5, BAG6= BCL2-associated athanogene 6 (also known as BAT3), APOM = apolipoprotein M, RARB = retinoic acid receptor, beta, STK19 = serine/threonine kinase 19, C4B = complement component 4B (Chido blood group), NG = near gene, NFKBIA = nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, alpha, GPANK1 = G patch domain and ankyrin repeats 1, PCSK9 = proprotein convertase subtilisin/kexin type 9, PSMA6 = proteasome (prosome, macropain) subunit, alpha type, 6.

^#Function reported is reported dbSNP Genome Build 37.3 (http://www.ncbi.nlm.nih.gov/projects/SNP/)

^¶If two genes are listed for the SNP, the first gene listed was annotated using dbSNP (Genome Build 37.3) and the second gene was annotated using the IBC chip (annotation from Genome Build 36). If there is only one gene listed, there was no discrepancy between Genome Builds 37.3 and Build 36.

⁺Frequency of allele labeled in table as “coded allele”

^§rank-normalized residuals of FEV₁/FVC adjusted for age, age², height, height², sex, smoking status, pack-years, pack-years² the first 10 principal components (PCs) for ancestry, genomic control at the cohort and meta-analysis level. A cohort-specific site covariate was included in the regression model for cohorts with multiple sites (ARIC, CARDIA, CHS, and MESA).

The new SNP (rs805301) identified in European-Americans was selected for the IBC chip as a variant in APOM based upon Genome Build 36 and is annotated in BAG6 on Genome Build 37.3, which is the upstream flanking gene of APOM. It replicated among African-Americans (p=0.036) and remained significant in the combined meta-analysis (table 2). The risk allele (C) was associated with a decrease in FEV₁/FVC in both racial groups.

The new SNP identified in the combined European-American and African-American analysis, rs707974, was the second most significant SNP in African-Americans (table 2) and would have been significant with less stringent Bonferroni cutoff. It was also selected as an APOM variant and is now annotated in GPANK1, the downstream flanking gene of APOM separated by an open-reading frame, C6orf47. It was not significant for FEV₁/FVC in European-Americans (p=2.84×10⁻⁵).

SNPs rs805301 and rs707974 were not in high LD in European-Americans or African-Americans (r²=0.07 and r²=0.03, respectively), suggesting that they are separate loci (Figure 2). They were also not in high LD with the previously described AGER SNP rs2070600 in European-Americans[11,12] (r²=0.035 and r²=0.37, respectively). In addition, rs805301 remained associated with FEV₁/FVC after adjustment for rs2070600 (p=6.82×10⁻⁴) and rs2070600 was only nominally associated with FEV₁/FVC among African-Americans (p=0.009). These findings suggest that associations of rs805301and rs707974 with lung function are unrelated to AGER.

Figure 2.

Regional Association Plots of top SNPs flanking APOM (rs805301, rs707974) and PCSK9 SNPs for FEV₁/FVC

The selected SNPs with the lowest p value are illustrated by the purple diamond. The correlations (r²) of surrounding SNPs in the region are indicated by the colors shown on the graph. For the SNPs flanking APOM(rs805301 and rs707974), a 600kb flanking size was selected to include the AGER SNP on the plot whereas 500kb flanking size was selected for the PCSK9 SNP. Plots were generated using LocusZoom.[97] The Genome builds/LD populations implemented were hg 18/HapMap Phase II CEU and hg/19 1000 Genomes Nov 2010 AFR for European-American and for African-American participants, respectively.

Abbreviations: SNPs = single nucleotide polymorphisms, APOM = apolipoprotein M, PCSK9 = proprotein convertase subtilisin/kexin type 9, FEV₁/FVC = ratio of forced expiratory volume in one second over forced vital capacity.

Sensitivity analyses restricted to participants free of clinical cardiovascular disease, age 55 years or less, and free of asthma yielded similar results, as did analyses additionally adjusted for diabetes, hypertension and asthma (supplement). Analyses stratified by smoking status yielded similar results (supplement).

The other top loci in African-Americans were in NFKBIA and PCSK9 (table 2). PCSK9 is related to apoM[64] and six of the top 30 SNPs for FEV₁/FVC in African-Americans were in PCSK9 (figure 2). SNPs in neither gene replicated in European-Americans. Regional association plots for additional loci are displayed in supplementary figure 3.

Results for the FEV₁ are displayed in supplementary figures 4, 5 and 6. Top SNPs associated with FEV₁/FVC and FEV₁ are presented in supplementary tables 1 and 2.

Replication of SNPs Flanking APOM in SpiroMeta

Both rs805301 and rs707974 replicated for FEV₁/FVC in 20,288 European participants in the SpiroMeta consortium in a consistent direction (β=−0.03, p=0.02 and β=0.05, p=0.02, respectively).

We reviewed publically available results from the SpiroMeta-CHARGE GWAS meta-analysis of airflow obstruction.[9] SNP rs805301 was associated with airflow obstruction (p=0.004) and rs707974 was nominally associated with airflow obstruction in individuals without asthma (p=0.026; supplement).

Association of SNPs Flanking APOM with Percent Emphysema in MESA

SNP rs707974 was significantly associated with percent emphysema among 2,551 European-Americans and 2,457 African-Americans (p=4.74×10⁻⁴ and p=0.009, respectively) and in combined analyses (p=1.67×10⁻⁵; table 3) in MESA. The characteristics of these participants are shown in supplementary table 3. The direction of the association of rs707974 with percent emphysema and lung function was consistent: risk allele (A) was associated with greater percent emphysema and a lower FEV₁/FVC.

Table 3.

Association of SNPs flanking APOM with percent emphysema among European-American and African-American participants in MESA.

European-American Participants (n = 2,552)
SNP	Ref. allele	Allele freq.#	β¶ (SE)	p-value
rs805301	C	0.37	+0.019 (0.018)	0.29
rs707974	G	0.10	−0.098 (0.028)	4.74 × 10−4

African-American Participants (n = 2,483)
SNP	Ref. allele	Allele freq.#	β¶ (SE)	p-value
rs805301	C	0.58	+ 0.013 (0.018)	0.46
rs707974	G	0.02	−0.160 (0.061)	0.009

Combined European-American and African-American Participants (n = 5,035)
SNP	Ref. allele	Allele freq.+	β¶ (SE)	p-value
rs805301	C		+0.015 (0.001)	0.14
rs707974	G		−0.094 (0.022)	1.67 × 10−5

Abbreviations: SNP= single nucleotide polymorphism, No = number of participants, β = effect estimate, SE = standard error, HDL = high density lipoprotein cholesterol, PCs = principal components for ancestry

^#Reference allele frequency

^¶Log transformed percent emphysema-₉₅₀ adjusted for adjusted for age, sex, site, CT scanner, height, weight, weight greater than 220 pounds, cigarettes per day, pack-years, asthma, and PCs.

⁺Not reported, combined across races

We performed association testing under an additive genetic model stratified by race and subsequently in the combined population controlling for race/ethnicity for each quantitative phenotype. Genome-wide significant set at 10×⁻⁶ using Bonferonni correction. The SNPs are not in tight LD (r² = 0.059)

The association with percent emphysema persisted in an independent sample of 1,138 European-American and 1,563 African-American MESA participants who did not have spirometry measures and who were therefore excluded from the lung function analysis (p=0.02 and p=0.003; respectively). Additional adjustment for socioeconomic status yielded similar results whereas restriction to 418 European-Americans and 209 African-Americans with FEV₁/FVC < 0.70 yielded non-significant results; however, the effect size was greater in African-Americans and similar in European-Americans in these groups compared to the overall MESA sample (supplement). SNP rs805301 was not significantly associated with percent emphysema.

PCSK9 was nominally associated with percent emphysema in European-Americans (p=0.04) but not African-Americans. AGER SNP rs2070600 was significantly associated with percent emphysema among European-Americans and African-Americans (p=2.54×10⁻⁴ and p=0.001, respectively).

Gene Expression of SNPs Flanking APOM in MESA COPD

APOM expression was significantly, inversely associated with FEV₁/FVC (table 4) in an independent sample of 101 participants in the MESA COPD Study, the characteristics of which is described in the supplement.. We secondarily examined expression of GPANK1, BAG6 and PCSK9. GPANK1 expression was associated with FEV₁/FVC (β=−0.096; 95%CI −0.175,−0.017; p=0.02) whereas BAG6 and PCSK9 expression were not associated with FEV₁/FVC.

Table 4. Association of APOM gene expression with the FEV₁/FVC and CT percent: emphysema in 101 participants in the MESA COPD Study.

	Mean difference in FEV1/FVC per increase in APOM gene expression# (95% CI)	p-value	Mean difference in Percent Emphysema¶ per increase in APOM gene expression# (95% CI)	p-value
Model 1	−0.091 (−0.178, −0.003)	0.04	1.05 (0.10, 2.0)	0.03
Model 2	−0.092 (−0.178, −0.003)	0.04	1.04 (0.13, 1.95)	0.03
Model 3			0.76 (−0.14, 1.67)	0.10
Model 4			1.14 (0.20, 2.09)	0.02

Abbreviations: FEV₁/FVC= ratio of forced expiratory volume in one second over forced vital capacity, APOM = apolipoprotein M

Model 1: Adjusted for age, gender, cohort and race/ethnicity

Model 2: Additionally adjusted for smoking status, and pack-years

Model 3: Additionally adjusted for height, and weight

Model 4: Additionally adjusted BAG6 (probe-set 210208)

^#APOM probe-set 214910_s_at

^¶Log-transformed CT percent emphysema below −950 HU

We tested the association of gene expression with FEV₁/FVC and with percent emphysema in the MESA COPD Study implementing linear regression models weighted to account for the sampling schema as described in the supplement.

APOM expression was positively associated with percent emphysema in minimally adjusted models and after adjustment for BAG6 (table 4). BAG6 expression was not associated with percent emphysema except after adjustment for APOM (p=0.01). PCSK9 was significantly associated with percent emphysema (β=1.150; 95%CI 1.0, 1.32; p=0.016).

Association of HDL with Lung Function and Percent Emphysema in MESA

Among 3,044 participants with spirometry, higher HDL levels were independently associated with a lower FEV₁/FVC (−0.24% per 10 mg/dl HDL; 95%CI:−0.45, −0.03; p=0.027).

Among 8,367 participants with percent emphysema, higher HDL levels were independently associated with greater percent emphysema (0.53% increase in percent emphysema per 10 mg/dl HDL; 95% CI: 0.34, 0.73; p<0.001). Figure 3 shows the multivariate relationship of HDL to percent emphysema, which was non-linear (p<0.001) with a plateau at HDL levels greater than 60 mg/dL.

Figure 3.

Multivariate association between HDL and percent emphysema

Results of multivariate analyses of the relationship between percent emphysema and plasma HDL among 8,367 MESA SHARe participants are shown. The solid line indicates smoothed regression line adjusted for age, sex, race/ethnicity, height, weight, educational attainment, scanner, tube current, total cholesterol, exercise, pack-years, cigarettes per day, alcohol use, inhaled steroids and use of statins. Figure and p value was produced using a loess smoothing function in a generalized additive model in R/GAM (R version 2.13.0).Dashed lines indicated 95% confidence intervals.

To assess for potential survival bias among older participants, we repeated the HDL-emphysema analysis among 5,241 participants 45-65 years old and found consistent results. Findings were also consistent within strata of gender, race and smoking history (supplement).

DISCUSSION

This large, biracial study identified two new SNPs for FEV₁/FVC, one in European-Americans (rs805301) and one in the combined European-American and African-American analysis (rs707974). Both SNPs were originally selected as APOM polymorphisms and are now annotated in genes flanking APOM. Both replicated in an independent sample. In addition, rs707974 was significantly associated with percent emphysema in both European-Americans and African-Americans, APOM gene expression was associated with FEV₁/FVC and percent emphysema and HDL was associated with FEV₁/FVC and percent emphysema.

The identified SNPs flanking APOM are unlikely to be causative variants but might be linked with a functional APOM variant. Consistent with this thinking, the APOM promoter SNP rs805297 alters APOM expression[65] and is in weak linkage disequilibrium with rs707974 among African-Americans and European-Americans (r²=0.36 and r²=0.32, respectively) and rs805301 among European-Americans (r²=0.23; supplementary figure 7).

APOM encodes apoM, a lipoprotein-associated plasma protein.[66] The majority of apoM is found in HDL.[67] In murine models, modifying APOM gene expression changes apoM plasma concentration, which affects HDL levels, pre-β-HDL formation, reverse cholesterol transport and remodels plasma HDL.[40,67] Hence APOM gene expression alters the function and quality of HDL.

ApoM and HDL are relevant to the pathogenesis of COPD, particularly emphysema, via three related pathways. First, HDL inhibits tumor necrosis factor-stimulated sphingosine kinase activity in human endothelial cells thereby increasing ceramide and decreasing sphingosine-1-phosphate (S1P) cellular levels.[39,68] Ceramide, a second messenger molecule, modulates endothelial cell apoptosis and is implicated in emphysema pathogenesis.[37,38]

Second, HDL-associated-apoM is the plasma carrier for S1P and this HDL-apoM subclass presents S1P to the S1P₁ endothelial cell receptor which is endothelium-protective.[42, 69] S1P has an essential role in maintaining endothelial barrier integrity in the lung and is implicated in emphysema pathogenesis.[42,70]

Third, HDL binds and incorporates alpha-1-antitrypsin. HDL-bound alpha-1-antitrypsin inhibits extracellular matrix degradation and apoptosis in vascular smooth muscle.[71,72]

The relevance of APOM and HDL to COPD pathogenesis is further reinforced by our findings that PCSK9 polymorphisms were associated with FEV₁/FVC and, nominally, percent emphysema and that PCSK9 gene expression was associated with percent emphysema. PCSK9 augments the degradation of low density lipoprotein receptors[73] and gain-of-function mutations in PCSK9 cause familial hypercholesterolemia[74]. HDL levels in patients with PCSK9 mutations are generally increased[75-77] and most placebo-controlled trials of PCSK9 inhibitors have shown modest increases in HDL levels.[78-82] Furthermore, plasma levels of PCSK9 are associated with plasma apoM levels.[64]

The association of apoM and HDL with FEV₁/FVC and emphysema are probably distinct from their relationships to cardiovascular disease and we speculate that the roles of HDL and apoM in the lungs are different from their roles in atherosclerosis. Although HDL has long been thought to be atheroprotective, the definitive epidemiologic study on HDL and cardiovascular disease suggested no benefit[83] and large-scale randomized clinical trials of cholesterol ester transfer proteins, which raise HDL levels, have yet to show a benefit on clinical cardiovascular events[84,85]. The literature on apoM in cardiovascular disease is relatively small and mixed, with animal studies suggesting atheroprotective effects [40,41]; however, in humans, plasma apoM levels were not associated with atherosclerotic disease[86].

Despite the strong mechanistic support implicating APOM in COPD, the latest genome build annotates the new SNPs in genes neighboring APOM, which raises the possibility that they are unrelated to APOM. GPANK1 may be involved in immunity[87] and BAG6 is implicated in apoptosis;[87] both are associated with lung cancer[88,89] and neither have been associated with cardiovascular disease. Given that genome builds change over time, the identification of the two new SNPs in the same region in separate racial groups, the gene expression findings, and the HDL associations all suggest that APOM rather than the other genes are implicated in COPD pathogenesis.

Prior studies of genetic risk for emphysema include one GWAS[18] and candidate gene association studies.[19-24] The GWAS identified BICD1 as associated with severe emphysema on radiologist interpretation but no loci for percent emphysema.[18] The candidate gene association studies did not include rs707974 or rs805301.

Two small studies found increased levels of HDL in severe COPD defined by spirometry.[90,91] Conversely, lower HDL was associated with lower FEV₁ in a population-based study (which did not report the association for FEV₁/FVC[92] and advanced COPD and emphysema patients from the Evaluation of COPD Longitudinally to Identify Predictive Surrogate End-points Study.[93]

The study has several potential limitations. The genomic inflation factor for the meta-analysis in European-Americans was 1.080, suggesting possible population stratification. However, we adjusted our analysis with 10 PCs and with cohort-specific and overall genomic control to address population stratification. Furthermore, we replicated the new APOM SNP identified in European-Americans (rs805301) in African-Americans and both SNPs replicated in an independent cohort and with gene expression, all of which makes population stratification less of a concern.

Although we analyzed two phenotypes of COPD in general population samples, these traits do not capture the entire phenotypic complexity of clinical COPD. Results for the two phenotypes, however, were consistent with each other, similar among patients with airflow limitation, and supported by gene expression in a study of clinical disease. Furthermore, multiple prior genes identified for lung function in population-based samples have been replicated in studies of clinical COPD.[14-16] Hence it is likely that the current findings apply to clinical COPD.

The association between HDL and percent emphysema may be subject to confounding and reverse causation, a small study suggested that HDL levels decrease in COPD patients undergoing lung transplantation.[94] However, we adjusted for multiple potential confounders in this well-phenotyped cohort and the genetic studies are unlikely to be subject to reverse causation.

Similar to other population-based GWAS, we used pre-bronchodilator spirometry for lung function measurement. Percent emphysema was measured on partial-lung CT scans however; we previously validated percent emphysema on partial-lung scans compared to full-lung scans in this cohort and have confirmed multiple prior hypotheses using them.[28,57] Percent emphysema, like lung function, is related to gender, body size, ancestry and socioeconomic status,[95] in addition to current smoking.[96] We adjusted, however, for all of these variables in the analyses.

In conclusion, we identified one new SNP related to FEV₁/FVC among European-Americans and a second new SNP in the combined European-American and African-American analysis that was also associated with percent emphysema. Both new SNPs flank APOM, and APOM expression was associated with FEV₁/FVC and percent emphysema. APOM encodes apoM, which is primarily bound to HDL, and higher levels of HDL were associated with lower FEV₁/FVC and greater percent emphysema. Together, these findings suggest a novel effect of the APOM/HDL-cholesterol pathway in the pathogenesis of COPD and emphysema. Further examination of this pathway is warranted to determine if it is targetable to treat or prevent COPD, and ongoing clinical trials of PCSK9-inhibitors[78-82] and other medications that raise HDL levels[85] may consider monitoring for pulmonary effects.

Supplementary Material

Supp Fig 1A

Supplementary Figure 1 Cohort and race specific quantile-quantile plots of association results for FEV₁/FVC.

Cohort and race specific quantile-quantile results for all SNPs associated with FEV₁/FVC. (A) Cohort specific genomic inflation factors calculated after controlling for 10 principal components analyses for ancestry for European-American participants. (B) Cohort specific genomic inflation factors calculated after controlling for 10 principal components analyses for ancestry for African-American participants.

Supp Fig 7

Supplementary Figure 7 Linkage Disequilibrium Plots based on population samples.

D’ pairwise linkage disequilibrium (LD) plots of population data from CEU-Utah residents with Northern and Western ancestry from CEPH collection (A) and ASW-African ancestry in Southwest USA (B). LD plots created with Haploview, HapMap 3 release 2. Red: D’ = 1 (LOD ≥ 2.0); blue: D’ = 1 (LOD < 2.0); pink: D’ < 1 (LOD ≥ 2.0); white: D’ < 1 (LOD < 2.0).

APOM SNP (rs805297), BAG6 SNP (rs805301) and GPANK1 SNP (rs707974) are circled in green. In the CEU population (a) Block 2 is 11kb in length and includes rs805301 and rs805297. Although rs707974 is in Block 3 (1kb), the D’=1 (LOD ≥ 2.0) for rs707974 and rs805297. In the ASW population (b) Block 3 is 11 kb in length and includes rs805297 and rs707974. SNP rs805301 is in Block 2 (2 kb) and the D’=0.78 (LOD<2.0) for rs805301 and rs805297.

APOM SNP (rs805297) is in LD with rs707974 among African-Americans and European-Americans (r²=0.36 and r²=0.32, respectively) and it is in LD with rs805301 among European-Americans (r²=0.23) but not African-Americans (r²=0.06) using HapMap3 release2.

LOD = logarithm of odds.

Supp Fig 2

Supplementary Figure 2 Quantile-quantile plots of meta-analysis association results for FEV₁/FVC stratified by race and across all races.

Quantile-quantile results are for meta-analysis of cohorts for SNPs associated with FEV1/FVC stratified by race and combined across all races. Genomic inflation factors calculated after controlling for 10 principal components for ancestry.

Supp Fig 3A-D

Supplementary Figure 3 Regional association plots of top SNPs associated with FEV₁/FVC stratified by race and combined across all races.

Regional association plots of top SNPs for FEV₁/FVC in European-American participants (A-D), African-American participants (E-H) and combined analysis of European-American and African-American participants (I-M). The selected SNPs with the lowest p value are illustrated by the purple diamond. The correlations (r²) of surrounding SNPs in the region are indicated by the colors shown on the graph. A 500Kb flanking size was selected for all plots. Plots were generated using LocusZoom. The Genome builds/LD populations implemented were hg 18/HapMap Phase II CEU for European-American and combined European-American and African-American analyses. The hg/19 1000 Genomes Nov 2010 AFR for African-American participants.

Abbreviations: SNPs = single nucleotide polymorphisms, FEV₁/FVC = ratio of forced expiratory volume in one second over forced vital capacity.

Supp Fig 5

Supplementary Figure 5 Quantile-quantile plots of meta-analysis association results for FEV₁ stratified by race and across all races.

Quantile-quantile results are for meta-analysis of cohorts for SNPs associated with FEV₁ stratified by race and combined across all races. Genomic inflation factors calculated after controlling for 10 principal components.

Supp Fig 6A-D

Supplementary Figure 6 Regional association plots of top SNPs associated with FEV₁ stratified by race and combined across all races.

Regional association plots of top SNPs for FEV₁ in European-American participants (A-E), African-American participants (F-I) and combined analysis of European-American and African-American participants (J-N). The selected SNPs with the lowest p value are illustrated by the purple diamond. The correlations (r²) of surrounding SNPs in the region are indicated by the colors shown on the graph. A 500Kb flanking size was selected for all plots. Plots were generated using LocusZoom. The Genome builds/LD populations implemented were hg 18/HapMap Phase II CEU for European-American and combined European-American and African-American analyses. The hg/19 1000 Genomes Nov 2010 AFR for African-American participants.

Abbreviations: SNPs = single nucleotide polymorphisms, FEV₁/FVC = ratio of forced expiratory volume in one second over forced vital capacity.