This multicohort study is the largest reported genetic association analysis of radiographic emphysema in COPD to date, and the first genome-wide association study for this disease phenotype. Our most significant finding was at the BICD1 locus with the presence of emphysema as a radiologist-assessed qualitative trait; this locus has not been previously implicated in the pathogenesis of emphysema.
The GWAS results showed little overlap between the extent of emphysema assessed using quantitative assessment of low attenuation areas and radiologist qualitative scores. For the top 254 SNPs with meta-analysis P
values less than 5 × 10−4
from association analysis of qualitative emphysema, only 4 and 5 of them have meta-analysis P
values less than 5 × 10−4
for association with quantitative emphysema with and without adjustment for BMI, respectively. These differences are likely due to several factors. First, although the correlation between qualitative (radiologist-derived) and quantitative (low attenuation areas) analysis is statistically significant, the magnitude of this correlation suggests that there are still substantial differences in the extent of emphysema determined by these two methods (27
); in the present study, 11.5 and 2.3% of subjects in the ECLIPSE and Norway cohorts, respectively, who had absent or trivial emphysema by radiologist score, had −950 HU values above the bottom quartile of quantitative emphysema measurements based on subjects with severe emphysema by radiologist score. Second, the NETT study was included in the quantitative, but not the qualitative, analysis; although theoretically inclusion of this study should increase power, differences in the study population and the measurement of emphysema could have contributed to increased heterogeneity and thus reduced the power of our analysis. Third, it is known that BMI, radiation dose, and slice thickness affect the apparent X-ray attenuation values in the CT scan and, therefore, quantitative analysis of low-attenuation areas (39
The locus most highly associated with qualitative emphysema is in a region that covers the second exon of the gene BICD1. BICD1
is one of two human homologs of Drosophila
). The function of BICD protein has been well characterized—it forms a complex with dynein–dynactin and plays a critical role in mediating dynein function. Dynein is involved in many cellular processes, including mitosis, nuclear migration and mRNA transport, and transport of a variety of axonal and dendritic vesicles. The second exon of BICD1
encodes the sequence located in the coiled-coil domain at the N terminus of the protein, which directly interacts with dynein (41
). Mangino and colleagues, following up a linkage study, identified SNPs in the first intron of BICD1
gene associated with telomere length in leukocytes (44
); the variants were associated with BICD1
mRNA levels and the estimated effect of the variants was equivalent to between 15 and 20 years of age-related attrition in telomere length (44
). Although the identified SNPs in our study were located in a different linkage disequilibrium block, their findings suggest that the effect of BICD1
in emphysema could be related to shortened telomeres, which has been previously observed in patients with COPD (46
). Directly studying the association among variants in BICD1
, telomere length, and emphysema is a reasonable next step for investigation.
Our second-ranked locus in the qualitative emphysema analysis did not reach genome-wide significance; this locus is at 8p23, a genomic region showing linkage with postbronchodilator FEV1
in families ascertained through early-onset COPD (47
) and 55 kb downstream from CSMD1
, a tumor suppressor gene known to be important in lung cancer (48
). In our analysis of quantitative emphysema, the most significant association with correction for BMI was in chromosome 10, approximately 170 kb downstream of PARD3
[par-3 partitioning defective 3 homolog (C. elegans
)], a gene important in maintaining tight junctions, with homozygous deletions reported in lung cancer cell lines (49
). Without correction for BMI (Table E2), our most significant association was at rs1348350, 115 kb downstream from MBL2
[mannose-binding lectin (protein C) 2], a pattern recognition receptor important for innate immunity, genetic variants of which have been associated with hospitalization for COPD (50
), and with age of first Pseudomonas aeruginosa
infection and speed of pulmonary function decline in children with cystic fibrosis (51
). Another potentially interesting signal was at rs4905179, 6 kb upstream from SERPINA6
[serpin peptidase inhibitor, clade A (α1
-antiproteinase, antitrypsin), member 6], and 37 and 49 kb downstream from SERPINA2
(the gene encoding α1
-antitrypsin), respectively; SERPINA6
is also known as corticosteroid-binding globulin, and differential expression of the mouse orthologous gene has been reported between strains differentially susceptible to cigarette smoke (52
). Whether these associations are indeed true susceptibility loci for emphysema remains to be determined.
COPD is a complex disease, and the chronic airflow limitation that is characteristic of COPD results from a combination of pathological processes, including emphysema. The contribution that each of these factors makes in an individual subject cannot be determined by lung function measurements alone. In addition, assessment of emphysema severity may have important prognostic and therapeutic implications. Chest CT scans provide a sensitive and noninvasive method to identify emphysema in patients with COPD. Because phenotypic heterogeneity is one of the key issues in identifying consistent genetic association results in COPD, objective classification of patients with COPD based on emphysema measurements using CT may help reduce this heterogeneity. In addition to being the largest reported study in which CT analysis was conducted to objectively define emphysema, our subjects represented a range of severity as assessed by lung function.
Our study also had several limitations. The CT scans performed in each cohort did not use identical CT acquisition protocols or scanner types, and these differences might have led to differences in the measurement of the extent of emphysema; however, protocols were standardized within each study. Moreover, our strongest signal is from radiologist-defined emphysema, which is arguably less affected by the previously described factors. Although we obtained more significant results by radiologist scoring, our findings do not necessarily indicate the superiority of one method over another for future analyses. Radiologist scoring and density mask analysis methods each have their advantages and disadvantages, and although our study is large for a CT-based study, it is relatively small and underpowered to detect effect sizes typically found by genome-wide association studies.
In conclusion, we report the first GWAS analysis of CT-defined emphysema in patients with COPD and identified BICD1 as a potential susceptibility gene. BICD1 has been suggested to be involved in regulating telomere length, and it has been previously suggested that telomere length is altered in cigarette smoke–induced emphysema. Several loci associated with both qualitative and quantitative emphysema were also identified that did not reach genome-wide significance levels. Additional studies are needed to evaluate the role of these genes in emphysema.