In this study, we tested five candidate genes for association with three COPD-related phenotypes. We found associations in EPHX1 and SERPINE2 for Pao2, and in SFTPB for PASP. To confirm the associations with Pao2, we tested the same SNPs from EPHX1 and SERPINE2 in a different population and demonstrated gene-level, but not SNP-level, association with a closely related phenotype.
This is the first study of the relationship between genetic polymorphisms and hypoxemia in subjects with COPD. While there have been no prior studies directly addressing genetic determinants of hypoxemia in COPD, there have been previous candidate gene studies of pulmonary hypertension in COPD patients. In a cohort of 103 individuals with COPD, Eddahibi et al33
reported an association between pulmonary hypertension and a variant in the Serotonin transporten (SLC6A4) gene. Yildiz et al34
reported an association between endothelial nitric oxide synthase (NOS3) variants and pulmonary hypertension in a sample of 40 cases; there have been similarly sized studies of the relationship between the angiotensin-converting enzyme insertion-deletion polymorphism and pulmonary hypertension in COPD patients, but these studies have had conflicting results.34,35,36
Regarding potential biological explanations for our observed genetic associations with hypoxemia, EPHX1 and SERPINE2 are both excellent candidate genes for COPD susceptibility based on the known detoxifying function of EPHX1 and the gene expression and human linkage data available for SERPINE2. Although we continued to observe significant associations with hypoxemia after adjustment for both FEV1 level and a quantitative measure of overall emphysema severity, it is possible that the associations of SNPs in EPHX1 and SERPINE2 with hypoxemia may be due to the effects of these SNPs on emphysema susceptibility or COPD severity. It is possible that our statistical adjustment for emphysema does not capture important features of emphysema extent and distribution that could impact levels of hypoxemia, or that small airway disease is influencing the development of hypoxemia. It is also possible that the observed associations with hypoxemia may represent a pathophysiologic process separate from the development of COPD, although this is less likely based on our method of candidate gene selection.
At the SNP level, while no individual SNP was significantly associated with hypoxemia, the associated SNPs did come from similar genomic regions based on the LD patterns present in our samples. In EPHX1, a genomic region encompassing the 5′ promoter region and exon 1 was significantly associated with hypoxemia in our analyses of both NETT and Boston Early-Onset COPD Study participants. This is an area of alternative splicing of the RNA transcript.37
One exon 1 splice variant is expressed predominantly in the liver, whereas another variant is present in other body tissues, including the lung. Each splice variant has its own 5′ promoter region, suggesting that this region may play a tissue-specific role in regulating EPHX1 messenger RNA levels.
The EPHX1 SNP most strongly associated with hypoxemia in NETT participants, rs1051741, is a synonymous coding SNP in exon 8 that has previously been associated38
with emphysema distribution in the NETT population. While this would not change the amino acid structure of the resulting protein, synonymous amino acid substitutions can affect messenger RNA levels through differences in transfer RNA availability.39
Alternatively, this associated SNP may be in LD with a nearby functional variant.
In SERPINE2, the most highly associated SNPs from the NETT and Boston Early-Onset COPD Study populations fall within a 15-kb region spanning the 3′ end of SERPINE2. There is no known function for these intronic SNPs; however, in both study populations, this region was in moderate LD, suggesting that these signals could be due to a single, untyped locus. Two of the SNPs significantly associated in our study, rs975278 and rs729631, have been associated with COPD or FEV1
/FVC ratio in previous studies.6,9
Regarding our findings of an association between SFTPB variants and PASP in the NETT subjects, the two associated SNPs in our studies are intronic SNPs in tight LD, spanning a region around exon 8, which has not previously been associated with disease.
The principal strength of our study is that we used two well-phenotyped cohorts to test for genetic associations with three, clinically important, COPD-related phenotypes. To our knowledge, the NETT Genetics Ancillary Study cohort is the largest population of COPD patients in which arterial blood gas values, pulmonary artery pressures, and genetic polymorphisms have been characterized. In addition, we were able to test our genetic associations while adjusting for potential confounding factors such as the level of FEV1, smoking history, and Pao2 level (in the PASP analyses).
Our study has the following limitations:
- Both study populations were selected for the presence of severe COPD, so our findings may not be generalizable to patients with mild-to-moderate COPD.
- We attempted to replicate our associations with Pao2 using a similar, but not identical, phenotype. The phenotype used in the Boston Early-Onset COPD Study population is the self-reported use of continuous supplemental oxygen, which is a less accurate measure of hypoxemia than Pao2. However, the use of continuous supplemental oxygen is tightly regulated in the United States by the Center for Medicare and Medicaid Services and private payers, with the principle criteria for the coverage of continuous oxygen therapy being documented room air oxygen saturation ≤ 88% or Pao2 ≤ 55 mm Hg.40
- Our replication population has a limited sample size. Given the sample size of our replication population, our study is at risk of false-negative failure to identify small-but-real SNP effects as well as false-positive association results.
- No individual SNP was significantly associated with hypoxemia in both study populations. However, there are reasons to anticipate cases of gene-level, but not SNP-level, replication in complex disease genetics. In monogenic disorders, such as cystic fibrosis and α1-antitrypsin deficiency, numerous disease-causing mutations have been elucidated in the cystic fibrosis transmembrane regulator (or CFTR) and SERPINA1 genes. Some authors41 have suggested that gene-level analyses are more appropriate in complex-disease genetics than SNP-level analyses due to the presence of multiple important variations within a single gene. Nonetheless, until specific functional variants are identified in EPHX1 and SERPINE2 and are associated with hypoxia in multiple populations, we present these results as suggestive, not definitive, evidence of a causal relationship between these genes and hypoxemia in patients with severe COPD.
- We were unable to attempt replication of our association between SFTPB and PASP as we did not have a suitable replication population with pulmonary artery pressure measurements. This association should be tested in a well-powered replication cohort when such data become available.
In summary, our study used two extensively phenotyped cohorts of subjects with severe COPD to identify genetic associations with three important COPD-related traits. Our findings suggest that in subjects with severe COPD, polymorphisms in EPHX1 and SERPINE2 contribute to the development of hypoxemia, and that polymorphisms in SFTPB contribute to pulmonary arterial hypertension. This provides additional information regarding the complex network of genes underlying the variety of clinical COPD-related phenotypes.