CFTR variants are associated with chronic bronchitis in smokers

Abstract

Introduction:

Loss of function variants in both copies of the cystic fibrosis transmembrane conductance regulator (CFTR) gene cause cystic fibrosis (CF); however, there is evidence that reduction in CFTR function due to the presence of one deleterious variant can have clinical consequences. Here, we hypothesize that CFTR variants in individuals with a history of smoking are associated with COPD and related phenotypes.

Methods:

Whole genome sequencing was performed through the NHLBI TOPMed program in 8597 subjects from the COPDGene study, an observational study of current and former smokers. We extracted clinically annotated CFTR variants and performed single variant and variant-set testing for COPD and related phenotypes. Replication was performed in 2,118 subjects from the Evaluation of COPD Longitudinally to Identify Predictive Surrogate Endpoints (ECLIPSE) study.

Results:

We identified 301 coding variants within the CFTR gene boundary: 147 of these have been reported in individuals with CF, including 36 CF-causing variants. We found that CF causing variants were associated with chronic bronchitis in variant-set testing in COPDGene (one sided p-value=0.0025, OR =1.53) and in meta-analysis of COPDGene and ECLIPSE (one sided p-value=0.0060, OR =1.52). Single variant testing revealed that the F508del variant was associated with chronic bronchitis in COPDGene (one sided p-value=0.015, OR=1.47). In addition, we identified 32 subjects with two or more CFTR variants on separate alleles, and these subjects were enriched for COPD cases (p=0.010).

Conclusions:

Cigarette smokers who carry one deleterious CFTR variant have higher rates of chronic bronchitis, while presence of two CFTR variants may be associated with COPD. These results indicate that genetically-mediated reduction in CFTR function contributes to COPD related phenotypes, in particular chronic bronchitis.

INTRODUCTION

Chronic obstructive pulmonary disease (COPD) is a complex disease typically caused by cigarette smoke and influenced by genetic factors. COPD is phenotypically heterogeneous, with varying manifestations of emphysema, chronic bronchitis, airway wall thickening and bronchiectasis despite similar degrees of lung function impairment. This variability likely reflects the contribution of multiple pathologic mechanisms. Chronic bronchitis is a particularly problematic phenotype in COPD as it is associated with pulmonary exacerbations and has few treatment options (1, 2). Since chronic bronchitis shares some clinical and pathological features with cystic fibrosis (CF), it has been proposed that there may be common mechanisms involved.

CF is the most common lethal autosomal recessive disorder in populations of European descent, and one in thirty-five Americans is a carrier of a loss of function variant in the Cystic Fibrosis Transmembrane Conductance Regulator (CFTR). In addition to CF, several disorders have been associated with variants in CFTR, such as idiopathic pancreatitis (3, 4), congenital bilateral absence of the vas deferens (5), and allergic bronchopulmonary aspergillosis (6). Furthermore, there is evidence that cigarette smoking can lead to acquired CFTR dysfunction (7–10). Cigarette smokers and COPD patients have reduced function of CFTR in the upper and lower airways in addition to chronic bronchitis. CFTR dysfunction has been shown to reduce airway surface liquid and decrease mucociliary transport (7, 10, 11). Therefore, it is possible that acquired CFTR dysfunction through cigarette smoking may contribute to COPD, and this effect may be compounded by genetic variation in CFTR.

CFTR potentiators are a new class of CF medications, which function by directly correcting underlying gating defects in mutant CFTR (7). In vitro studies have demonstrated that the CFTR potentiator ivacaftor can improve CFTR protein function in epithelial cells exposed to cigarette smoke, and this is reflected in measures of epithelial function including mucociliary transport, airway surface liquid depth and ciliary beating (7, 12). In addition, a pilot study of ivacaftor in patients with COPD and chronic bronchitis demonstrated the potential for increased CFTR activity and respiratory symptoms (13). Furthermore, there is evidence that the CFTR potentiator icenticaftor can increase FEV₁, as well as reduce systemic inflammation and sputum colonization in COPD patients (14). Collectively, these data indicate that improvement of CFTR function using existing drugs could improve lung function in COPD patients. However, the question remains as to which patients would most benefit from this treatment.

While several small studies have investigated association of CFTR variants with the deleterious effects of cigarette smoke on CFTR function, results have been mixed (15–22). Other larger studies have been limited by including non-smokers in addition to smokers (23, 24). To address this question with greater power, a large sample size of smokers with and without COPD along with CFTR gene sequencing data is required to ascertain whether CFTR variants, together with cigarette smoke, contribute to reduced lung function in smokers with COPD. Here, we perform the largest investigation of CFTR variants in COPD to date, including subjects with whole genome sequencing (WGS) data from two large cohorts to test the hypothesis that deleterious variants in CFTR are associated with COPD and related phenotypes.

METHODS

Study Populations

The Genetic Epidemiology of COPD (COPDGene) and Evaluation of COPD Longitudinally to Identify Predictive Surrogate Endpoints (ECLIPSE) studies have been described previously (25, 26). Briefly, COPDGene enrolled 10,192 non-Hispanic white and African American subjects with a minimum of 10 pack-years lifetime smoking history. Subjects with diagnosed lung diseases other than COPD or asthma were excluded. The ECLIPSE study is a multicenter multinational 3-year longitudinal study that enrolled 3,291 subjects of GOLD stage 2–4. In COPDGene, COPD was defined by a postbronchodilator ratio of forced expiratory volume in 1 second (FEV₁) to forced vital capacity (FVC) < 0.7 (Global Initiative for Chronic Obstructive Lung Disease (GOLD) 1–4); severe COPD was defined as GOLD stages 3–4. In ECLIPSE, only subjects with GOLD stage 2–4 were included. Chronic bronchitis was defined using the classical definition of self-reported chronic cough and phlegm for ≥ 3 months per year over the past two years. Bronchodilator response was defined as the % change in pre/post bronchodilator FEV₁. Visual scoring of bronchiectasis was performed using CT scans for 1,372 COPDGene subjects with WGS data(27). Subjects who were found to have diffuse bronchiectasis on chest CT scan were excluded from COPDGene.

Institutional review boards approved the studies at all participating institutions and all participants provided written, informed consent per study protocols.

Whole Genome Sequencing

Whole genome sequencing data was generated through the NHLBI TOPMed consortium to a mean depth of 30X using DNA from blood, PCR-free library construction and Illumina HiSeq X technology(28). For COPDGene, Freeze 5b WGS data was used which includes 8,598 subjects including 5,773 non-Hispanic white (NHW) and 2825 African American (AA). For replication in ECLIPSE, Freeze 8 WGS data was used which included 2345 subjects, and a subset of 2212 were included in this analysis. Reads were mapped to human genome assembly version GRCh38 and computational phasing was performed using Eagle 2.4 (Dec 13, 2017).

Identification and annotation of CFTR variants

All variants within the CFTR gene boundary (chr7:117,465,784–117,715,971, GRCh38) were extracted from WGS data using bcftools (29). The WGS annotator pipeline (30) was used to characterize all variants. Coding variants were identified as variants classified as in frame deletion, frameshift, missense, splice acceptor, splice donor, splice region, stop gained or synonymous variants according to the Ensembl Variant Effect Predictor (VEP) consequence. Annotation of known CF-causing variants was downloaded from the CFTR2 consortium website (https://cftr2.org/) (accessed on May 4, 2021). These variants are categorized as CF-causing, varying-clinical significance, non-CF causing and unknown significance. For variants that were not reported in the CFTR2 database, SNPEff functional effect predictions were used to identify variants with likely functional impact. Phased sequencing data from subjects with two or more known CF-causing variants was visually inspected to determine whether these subjects are compound heterozygotes with pathogenic variants on both chromosomes. These subjects are of interest as loss of function of both copies of CFTR would likely have a greater clinical consequence. We hypothesized a priori that heterozygous CFTR variants would have a deleterious effect in smokers due to a decrease in CFTR function, therefore we expected that the minor allele (i.e. less common allele) of CFTR variants would be associated with increased chronic bronchitis, increased severe COPD, increased risk of severe exacerbations, decreased BMI, decreased FEV₁ percent predicted, decreased percent emphysema and increased airway wall thickness. We used one-sided p-values for these tests, while we used two-sided p-values for associations with bronchodilator response (as a percent of predicted FEV₁) as we did not have a prediction regarding direction of effect.

Single-Variant Association Testing

The workflow for genetic variant testing is described in Figure 1. Testing of each individual variant for phenotype association was performed using linear regression for quantitative traits and logistic regression for binary outcomes using R base functions. For single variant testing, only variants with a minor allele count ≥ 10 were included. Analyses were adjusted for age, sex, pack-years of smoking, current smoking status, and principal components of genetic ancestry (PC). Calculation of PCs has been previously described (31, 32). Analyses in COPDGene were performed in NHW and AA individuals combined, using 3 PCs of genetic ancestry. For ECLIPSE, 10 PCs of genetic ancestry were used. For each single-variant analysis, we also performed permutation analysis by permuting the variant/non-variant carrier status among all subjects 20,000 times, then computing the p-value using the number of permutations in which the test statistic is more extreme than the observed test statistic. As described above, we used one-sided p-values for association testing with all phenotypes except bronchodilator response.

Figure 1: Workflow for CFTR genetic variant testing.

COPDGene served as the discovery cohort and significant findings were replicated in ECLIPSE. Four groups of variants were tested for association with 10 phenotypes in COPDGene using the (Sequence) Kernel Association Test (SKAT) and burden testing. Only variant groups and phenotypes with significant associations in grouped variant testing were included in single variant testing. Furthermore, single variant testing was only performed for variants with a minor allele count (MAC) greater than 10.

Gene-based Association Testing

Gene-based testing of rare variants (< 5% minor allele frequency) was performed using burden tests in which we collapsed rare CFTR variants into a single burden variable and tested for association with phenotype using linear and logistic regression. In addition we used SNP-set (Sequence) Kernel Association Test (SKAT)-O (33) as an additional method for gene-based association testing. All CFTR variants were tested in a combined analysis, in addition to testing subsets of variants grouped according to known pathogenicity using annotations from the CFTR2 database. SKAT-O tests were performed both with weighting by percent pancreatic insufficiency (obtained from the from the CFTR2 consortium website) as a measure of variant severity, and with no weighting.

RESULTS:

Identification of CFTR variants in COPDGene participants

After quality control measures (28), a total 8595 subjects including 3848 COPDGene cases and 4691 smoking controls were available for analysis (Table 1). In these subjects, we identified 11,567 variants within the gene boundary of CTFR as defined by Ensembl (chr7:117,465,784–117,715,971) which includes 14,241 bp upstream and 47,306 bp downstream of the coding region of transcript NM_000492.3. Of these variants, 10,577 are single nucleotide variants (SNV) and 990 are insertion-deletion polymorphisms (indel). Of these, there were 301 variants that are located within the coding region of the RefSeq Select transcript (NM_000492.3) (Supplementary Table 1). Using the CFTR2 database, we found that 147 variants have been reported in CF patients; 36 are CF-causing variants, 25 are variants of varying clinical consequence (may cause CF in some individuals but not others), 18 are non-CF causing variants (may cause CFTR dysfunction but not sufficient to cause CF) and 68 variants have not been evaluated or are of unknown significance. Four variants with high minor allele frequency (>0.05) were excluded from further analysis; three of these are synonymous variants (rs1800136 [legacy 4521G/A], rs1800130 [P1290P], and chr7:117595001:T:G), while rs213950 (lV470M) is a missense variant known to be non-CF causing. After these, the most frequent variants were chr7:117509093:G:A (R75Q) with 459 counts and chr7:117559655:G:A (1716G/A) with 290 counts, both of which are non-CF causing missense variants. We additionally identified 177 subjects that are heterozygous for the common p.Phe508del (legacy F508del) variant (rs199826652). We discovered 154 variants that have not been previously described in the CFTR2 database, including 1 stop-gain and 89 missense variants which are predicted to have moderate to high impact on CFTR through SNPEff functional impact prediction.

Table 1:

Description of study subjects in COPDGene

	COPDGene		ECLIPSE
	COPD Cases	Smoking Controls	COPD Cases	Smoking Controls
Number of subjects	3848	4691	1953	165
% Male	56.31	51.18	34.46	43.64
Age	62.91 (8.68)	56.72 (8.42)	63.36 (7.12)	56.30 (9.63)
Race	77.36	59.24	98.16	96.36
% Non-Hispanic White
% African American	22.64	40.76
% Current Smokers	55.85	39.91	61.90	63.03
Smoking history, pack-Years	51.62 (27.40)	38.44 (21.29)	48.94 (27.44)	30.02 (20.30)

Variant-set testing for association with COPD and related phenotypes.

Variants were grouped according to pathogenicity. Four groupings were tested: 1) CF-causing variants; 2) CF-causing variants and variants of varying clinical consequence; 3) CF-causing, varying clinical consequence, and variants that have not been reported that in CFTR2 that may have a functional effect (moderate or high impact in SnpEff); and 4) All coding variants. The only association that reached the threshold for significance after correction for multiple comparison (p<0.05/10 or 0.005) was the association of CF-causing variants with chronic bronchitis: 68 subjects out of 248 with CF-causing variants have chronic bronchitis (27.4%) while 1,597 out of 8,345 subjects without CF-causing variants have chronic bronchitis (19.1%)(p=0.0025, OR=1.53)(Table 2). We hypothesized that variants associated with a larger percentage of patients having pancreatic insufficiency reflected a greater impact of the variant on CFTR function. Therefore, SKAT-O variant-set testing was performed with and without weighting for % pancreatic insufficiency as a measure of variant severity. This analysis confirmed that CF-causing variants are associated with chronic bronchitis, although there was no difference in the weighted and unweighted analysis, and the associations were not significant after correction for multiple comparisons (Supplementary Table 2). Since we hypothesized that the combination of cigarette smoke and heterozygous CFTR variants would result in greater reduction of CFTR function, we performed a stratified analysis of current vs former smokers, where we found that 39.5% of currently smoking subjects with CF-causing variants had chronic bronchitis, compared to 23.9% of currently smoking subjects without CFTR variants, however the p-value did not reach the stringent threshold for significance after correction for multiple comparison (p=0.0082, OR=1.62)(Supplementary Table 3). In contrast, in former smokers we found that 17.2% of subjects with CF-causing variants had chronic bronchitis, compared to 13.5% of subjects without CFTR variants (p=0.082). We additionally found that in an analysis of COPD cases alone, there was a significant enrichment of chronic bronchitis in subjects with CF-causing variants (38.1%) compared to subjects without CFTR variants (25.5%)(p=0.0022, OR=1.72)(Supplementary Table 4). Finally, we found an association of borderline significance between all coding variants and severe COPD (p=0.0063, OR=1.14) (Table 2). Bronchiectasis was visually scored using CT scans for 1,372 subjects, however there was no association between the presence of bronchiectasis and CFTR variants (Table 2).

Table 2: Burden testing in COPDGene.

P-values and effect sizes for variant-set testing of CFTR variants with COPD and related phenotypes.

			CF-causing	CF-causing + Varying clinical consequence	CF-causing + Varying clinical consequence + Predicted functional	All coding variants	Controls
# of Variants			36	61	206	2971
# of Subjects 3			248	455	732	2309	6281
Chronic	Cases		68 (27.4 %)	109 (24.0 %)	169 (23.1 %)	463 (20.1 %)	1202 (19.1 %)
Bronchitis	Controls		180	346	563	1844	5079
	p-value (OR)		0.0025*(OR=1.53)	0.033 (OR=1.19)	0.0089 (OR=1.20)	0.41
COPD	Cases		134 (54.0 %)	213 (47.0 %)	337 (46.4 %)	1095 (47.8 %)	2751 (44.1 %)
	Controls		114	240	389	1197	3489
	p-value (OR)		0.039*(OR=1.28)	0.28	0.084	0.021 (OR=1.09)
Severe COPD	Cases		50 (20.2 %)	81 (17.9 %)	131(18.0 %)	437 (19.1 %)	1031 (16.5 %)
	Controls		198	372	595	1855	5209
	p-value (OR)		0.26	0.29	0.072	0.0063 (OR=1.14)
Severe	Yes		24 (9.7 %)	45 (9.9 %)	79 (10.8 %)	264 (11.4 %)	761 (12.1 %)
Exacerbations	No		224	410	653	2043	5520
	p-value (OR)		0.19	0.20	0.19	0.26
BMI	Mean (SD)		29.0 (6.1)	28.8 (6.0)	28.8 (6.1)	28.8 (6.1)	28.9 (6.3)
	p-value		0.36	0.48	0.24	0.27
FEV1 percent	Mean (SD		73.4 (25.5)	76.0 (25.4)	75.8 (25.4)	75.4 (25.4)	76.5 (25.2)
predicted	p-value		0.16	0.49	0.20	0.16
Percent	Mean (SD)		7.0 (9.9)	6.3 (10.0)	6.4 (10.0)	6.5 (10.0)	6.1 (9.4)
Emphysema	p-value		0.39	0.39	0.11	0.090
Airway Wall	Mean (SD		1.1 (0.2)	1.1 (0.2)	1.1 (0.2)	1.1 (0.2)	1.1 (0.2)
Thickness	p-value		0.22	0.17	0.35	0.33
Bronchodilator		Mean (SD	7.7 (9.4)	6.3 (10.2)	5.9 (9.4)	6.2 (9.4)	5.7 (10.4)
Response % FEV1 2		p-value	0.021 (beta=1.54)	0.85	0.85	0.27
Bronchiectasis	Yes		16 (30.2 %)	23 (28.8 %)	36 (28.1 %)	117 (30.6 %)	312 (31.5 %)
	No		37	57	92	265	678
	p-value		0.31	0.33	0.34	0.28

^1.Four variants with allele frequency > 5% (881 counts) were excluded from analysis

^2.All -p-values are one-sided except for BDR which is two sided

^3.Chronic bronchitis, severe exacerbation, and BMI data were unavailable for 2 subjects; COPD and severe COPD data were unavailable for 58 subjects; FEV1 percent predicted data was unavailable for 58 subjects; percent emphysema data was unavailable for 618 subjects; airway wall thickness data was unavailable for 619 subjects; bronchodilator response data was unavailable for 169 subjects; and bronchiectasis data was unavailable for 7209 subjects.

^*Indicates p-values that are significant after correction for multiple comparisons (p<0.05/10 or 0.005). Odds ratios or beta coefficients are shown for all nominally significant associations (p<0.05).

Single-Variant Testing for Association with COPD and related phenotypes

For phenotypes in which there was a significant association using variant-set testing, we performed single variant testing for all variants within the group with minor allele count of at least 10. This resulted in one CF-causing variant (F508del) tested for association with chronic bronchitis (Table 3), and 36 variants tested for association with severe COPD (Supplementary Table 5). We found that F508del was significantly associated with chronic bronchitis (one sided p-value=0.016, OR=1.47). While R75Q was nominally associated (p<0.05) with severe COPD after performing permutation analysis (p=0.02); no associations with severe COPD met the threshold for significance after correction for multiple comparisons (p<0.05/36 or 0.0014).

Table 3:

Single Variant testing of F508del for association with chronic bronchitis in COPDGene and ECLIPSE.

	COPDGene	ECLIPSE	Meta-analysis
Allele Counts	177	57
One-sided p-value from logistic regression	0.016	0.055	0.081
One-sided p-value from Firth regression	0.016	-	-
One-sided p-value with permutation	0.028	0.061	-
Odds ratio	1.47	1.67	1.52

Compound heterozygotes in COPDGene

We next searched for subjects who may be compound heterozygotes, meaning that these subjects have two different CFTR variants on opposite chromosomes. There were no subjects with two CF-causing variants. We identified 32 subjects that were either heterozygous for F508del in addition to carrying another CFTR variant or were heterozygous for two CFTR variants that have varying clinical consequence (Supplementary Table 6). We found that compound heterozygous subjects were enriched for COPD: out of the 32 compound heterozygotes, 21 were COPD cases while 11 were controls, whereas in non-compound heterozygous individuals there were 3827 COPD cases and 4680 controls (p=0.010)(Table 4). There was no enrichment of chronic bronchitis or bronchiectasis in compound heterozygotes (Table 4).

Table 4: Compound heterozygotes in COPDGene.

Numbers of subjects identified who are compound heterozygotes for CFTR variants.

	Clinically significant or predicted to be functional1	Clinically significant or predicted to be functional + Varying clinical consequence2	All Compound Heterozygotes	Controls	One sided p-value for all compound heterozygotes3
Total number of subjects	8	14	32	8565
COPD					0.010*
Cases	5	8	21	3827
Controls	3	6	11	4680
Chronic bronchitis					0.13
Yes	3	3	9	1656
No	5	11	23	6907
Bronchiectasis					0.090
Yes	0	1	3	426
No	1	1	2	941

¹These 8 subjects all carry one copy of the F508del variant and one variant of unknown function according to CFTR2 that is predicted to have moderate effect according to SNPeff

²This group includes the 8 subjects from the first group, one subject that carries one F508del variant and one variant of varying clinical consequence, and 5 subjects that carry two variants of varying clinical consequence

³p-values were computed using Fishers exact test to test whether COPD, chronic bronchitis and bronchiectasis cases were enriched in all compound heterozygotes compared to controls. Statistical testing was not performed for the other two groups due to the small sample sizes.

^*indicates p-values that are significant after correction for multiple comparisons (p<0.05/3)

Replication in ECLIPSE:

To attempt to replicate the results from COPDGene, we searched for CFTR variants in ECLIPSE. Whole genome sequencing and phenotyping data were available for 2212 subjects including 1953 cases and 165 controls. We identified 133 variants within the CFTR gene boundary including 19 CF-causing variants, 11 variants with varying clinical consequence, 13 variants that are not CF-causing and 32 variants that were not reported in CFTR2 or that have unknown significance (Supplementary Table 8). While the association of the 19 CF-causing variants with chronic bronchitis using burden testing did not reach statistical significance in ECLIPSE alone (one sided p=0.057), we found a significant association in meta-analysis of ECLIPSE and COPDGene (p=0.0060, OR=1.52)(Table 5). The only CF-causing variant in ECLIPSE with a minor allele count of greater than 10 was the F508del variant which was present in 57 subjects. Single variant testing revealed a suggestive association between F508del and chronic bronchitis in ECLIPSE(one sided p-value=0.055, OR=1.67)(Table 5), and in meta-analysis of COPDGene and ECLIPSE (one sided p-value=0.081, OR=1.52).

Table 5: Burden testing of association between CF-causing variants and chronic bronchitis in ECLIPSE.

P-values and effect sizes for association between CF-causing variants and chronic bronchitis in ECLIPSE and meta-analysis between ECLIPSE and COPDGene

	Number of variants	ECLIPSE p-value	ECLIPSE + COPDGene Meta-analysis p-value
All CF-causing variants in ECLIPSE	19	0.057	0.0060 (OR=1.52)
CF-causing variants in ECLIPSE also found in COPDGene	13	0.12	0.064

DISCUSSION:

This study is the largest to date characterizing the effect of CFTR variants in smokers with and without COPD. We found that CF-causing variants are associated with chronic bronchitis, and this is primarily driven by the most common CF-causing variant, F508del. We also found a suggestive association between all coding CFTR variants and severe COPD in the COPDGene study. Furthermore, we found that subjects that are compound heterozygotes for CFTR variants are at increased risk for COPD.

Several previous studies have shown that heterozygous CFTR variants can have a functional effect. For example, CFTR heterozygous variants are associated with idiopathic pancreatitis (3, 4), congenital bilateral absence of the vas deferens (5), bronchiectasis (34), and allergic bronchopulmonary aspergillosis (6). CF carriers may have an increased risk for developing airway obstruction and have been shown to have abnormalities in neutrophil function (35) and apoptosis (36) that may lead to a prolonged inflammatory state that could predispose to accelerated lung function decline. Furthermore, cigarette smoke is associated with decreased CFTR function in the upper and lower airways of both healthy smokers and smokers with COPD, and defective CFTR has been associated with symptoms of chronic bronchitis and dyspnea (7, 8). Therefore, it is possible that the presence of heterozygous genetic variants may increase the prevalence of chronic bronchitis or COPD in smokers. While several small studies have been conducted to test this hypothesis, results to date have been mixed. One study found that F508del variants were present at an increased frequency in subjects with chronic bronchitis and elevated sweat chloride levels (19). Several small studies have found modestly elevated CFTR variant frequencies in subjects with COPD or chronic bronchitis (17, 20, 22) (18). Most strikingly, a recent study including 108,035 Danish individuals identified 2858 F508del individuals and found that these individuals had an increased risk of bronchiectasis with an odds ratio of 1.31, as well as an increased risk of bronchiectasis with a hazard ratio of 1.88 (23). In addition, Miller et al. reported that CFTR variants were associated with an increase of chronic bronchitis with and odds ratio of 1.24 (24). However, other studies have failed to find that CFTR heterozygous variants have a functional effect. A study exposing CFTR heterozygous mice and cell lines to cigarette smoke found that CFTR heterozygosity did not have an impact on residual CFTR activity (21). In a study of obstructive pulmonary disease that included 250 F508del heterozygotes, COPD was not found to be increased, and measures of lung function were only lower in F508del heterozygotes who also had asthma (15, 16). Furthermore, genome-wide association studies (GWAS) of lung function, COPD, and emphysema have not identified CFTR as a susceptibility gene, though GWAS chips do not genotype the F508del variant, and this variant is typically not well imputed. Thus, the contribution of heterozygosity for CF variants to the etiology of COPD has been unclear, possibly due to the small sample size of studies to date, and the use of heterogeneous groups of patients, and the lack of gene sequencing to fully assess CFTR variants.

In this study, we sought to increase the power to detect the effect of rare CFTR variants by performing variant-set testing followed by individual testing of specific categories of variants. This allowed us to include ultra-rare variants, including variants only present in one subject in the dataset (singletons). We found that the combination of CF-causing variants was associated chronic bronchitis with statistical significance. The OR for the association in COPDGene was 1.53, and the OR in the meta analysis of COPDGene and ECLIPSE was 1.52. Similarly, the OR for the association of F508del with chronic bronchitis was 1.47 in COPDGene and 1.52 in the meta-analysis of COPDGene and ECLIPSE. This indicates that smokers with CF-causing variants are approximately 1.5 times more likely to have chronic bronchitis than subjects without CFTR variants, and the consistency of the OR across the two studies is an indicator of the validity of our findings. The finding that the OR is slightly higher in our study of only current or former smokers, compared to what has been reported in the literature (OR ranges 1.24–1.31), is consistent with the hypothesis that a history of cigarette smoking would result in a greater effect of CFTR variants. We also found suggestive evidence that variants with less established function (such as variants of varying clinical severity or predicted moderate impact) may be associated with chronic bronchitis. In addition, we found that the combination of all CFTR variants was nominally associated with severe COPD. This is of particular interest as it suggests that there could be a large number of COPD patients carrying CFTR variants that contribute to their disease severity and who could potentially benefit from treatment with CFTR modulators. Single variant testing of the association of all CFTR variants did not identify any associated variants that were significant after correction for multiple comparison, however the non-CF causing variant R75Q was nominally associated with severe COPD. R75Q is a relatively common missense variant which is not CF-causing but has been associated with pancreatitis (37), and increased frequency of R75Q has previously been found in patients with COPD (17).

We found that the only variant that was significantly associated with either chronic bronchitis or bronchodilator response using single variant testing was F508del. This was unsurprising given that F508del is the most common CF-causing variant identified in both COPDGene and ECLIPSE, as well as in the general population. Furthermore, F508del is a relatively severe class II variant, which produces a misfolded protein with little functional capacity. Therefore, it was one of the few variants for which we had sufficient power to detect associations with single variant testing. We identified 32 subjects that were compound heterozygotes for CFTR variants, meaning that they carry two copies of CFTR variants on separate chromosomes, and found that these subjects were enriched for COPD cases compared to non-compound heterozygotes. It is not possible to definitively conclude that these compound heterozygous subjects do not in fact have CF, due to the lack of CF diagnostic tests such as sweat chloride measurements in the COPDGene study. However, subjects with lung disease other than COPD or asthma, or with diffuse bronchiectasis on chest CT scans, were excluded. In the 32 compound heterozygotes identified here, only one subject reported a history of pneumonia, chronic bronchitis, or chronic cough or phlegm in early life (prior to age 15), suggesting that these subjects did not have history of early respiratory disease consistent with typical CF. We conclude that decreased CFTR activity due to two CFTR variants can result in COPD, based on the accepted GOLD definition (38).

While this study has several strengths, including being the largest study to characterize CFTR variants using whole genome sequencing in smokers with and without COPD and having replication in an independent cohort, there are also several limitations. Despite the large sample size, there were still small numbers of subjects with the less common CFTR variants, and therefore we are not able to determine whether these variants contribute to COPD. For example, the G551D variant is of particular interest since it can be corrected with ivacaftor, however we only identified 8 subjects that were heterozygous for this variant. The functional impact of most of the variants identified in our study are not known, and combining functional and non-functional variants reduces power for association studies. In addition, almost all subjects in both COPDGene and ECLIPSE have a history of smoking, and therefore we were not able to test if heterozygous CFTR variants have a function consequence in the absence of cigarette smoke. In summary, using unique analyses of CFTR variants in a cohort of smokers we found that CFTR variants, and particularly F508del are associated with chronic bronchitis.

Figure 2: Breakdown of CFTR variants.

A total of 301 coding SNVs were included in analysis.