Applying Functional Genomics to Chronic Obstructive Pulmonary Disease

Abstract

Genome-wide association studies (GWAS) have identified more than 20 genomic regions associated with chronic obstructive pulmonary disease (COPD) susceptibility. However, the functional genetic variants within these COPD GWAS loci remain largely unidentified, thus limiting translation of these GWAS discoveries to new disease insights. Whole-exome and whole-genome sequencing studies have the potential to identify rare genetic determinants of COPD. Efforts to understand the biological effects of novel COPD genetic loci include gene-targeted murine models, integration of additional omics data (including transcriptomics and epigenetics), and functional variant identification. COPD genetic determinants likely act through biological networks, and a variety of network-based approaches have been used to gain insights into COPD susceptibility and heterogeneity.

Although chronic obstructive pulmonary disease (COPD) is strongly influenced by cigarette smoking, several lines of evidence suggest that genetic susceptibility predisposes to COPD risk. Among smokers, COPD clusters in families (1). Alpha-1 antitrypsin (AAT) deficiency and cutis laxa are monogenic syndromes that often include emphysema as part of their syndrome constellation. Over the past 30 years, COPD genetic studies have evolved from linkage analysis to candidate gene association studies to genome-wide assessments. We review the current evidence for specific genetic determinants of COPD as well as functional studies to identify key genes and functional variants related to COPD susceptibility.

Genome-Wide Association Studies of COPD and Lung Function Levels

Genome-wide association studies (GWAS), which test commercialized panels of hundreds of thousands of single-nucleotide polymorphisms (SNPs) in large samples of cases/control subjects, population cohort members, or family units, have been successfully applied in COPD (2). With the large number of statistical tests performed in GWAS, stringent adjustment for multiple statistical testing is required; P values less than 5 × 10⁻⁸ are typically designated as genome-wide significant. A recent report from the International COPD Genetics Consortium identified 22 genome-wide significant regions associated with spirometrically defined COPD (3). Of note, nine of these regions had been previously associated with COPD in GWAS, and nine other regions had been previously associated with lung function levels but not COPD. It has been reassuring that a number of these COPD GWAS loci have been replicated in additional study populations by other investigators, suggesting that these associations are valid. In parallel with these GWAS of COPD cases and control subjects, large general population GWAS of spirometric measures have identified an even larger number of genetic determinants of forced expiratory volume in 1 second (FEV₁) and FEV₁/forced vital capacity (4). The significant overlap between COPD and lung function determinants has been both unsurprising—COPD was defined by lung function, after all—and remarkable, because it is not apparent how genetic determinants of small effect on lung function could lead to the profound reductions in lung function and pathological abnormalities (e.g., emphysema, small airway destruction and fibrosis) seen in advanced COPD. Although the impact of individual COPD GWAS determinants on COPD risk is modest, genetic risk scores hold promise for providing a more profound predictive capacity for the presence and severity of COPD (5).

Rare Genetic Determinants of COPD

With the substantial insights into COPD pathobiology resulting from AAT deficiency, other rare genetic determinants of COPD were searched for using whole-exome and whole-genome sequencing (WGS) studies. Stanley and colleagues identified rare variants in the TERT gene, previously associated with idiopathic pulmonary fibrosis, to be associated with emphysema (6). Of interest, their functional studies of specific genetic variants were essential; without the ability to separate three nonsynonymous variants that functionally influenced telomere length in COPD cases from one nonsynonymous variant in control subjects that did not impact telomere length, this relationship would likely not have been identified. Similar detailed functional insights of specific genetic variants are rarely available for most genes implicated through DNA sequencing studies.

Qiao and colleagues analyzed whole-exome sequencing in extended pedigrees ascertained through severe, early-onset COPD cases without AAT deficiency (7). They found 69 genes that had potentially functional variants segregating with COPD in at least two families but no genes that had variants segregating in more than three families. These results suggested that rare coding genetic determinants of COPD are likely quite genetically heterogeneous.

Large-scale WGS analyses of COPD are currently being performed through the U.S. National Heart, Lung, and Blood Institute TOPMed (Trans-Omics in Precision Medicine) program. These WGS efforts are unlikely to find many more novel common variant associations for COPD and COPD-related phenotypes than GWAS, except possibly in African American subjects, because the widely used genome-wide SNP genotyping panels for GWAS provide less-adequate coverage for populations of African ancestry. These WGS efforts will be very helpful for statistical fine mapping of previously identified GWAS regions for COPD and COPD-related phenotypes (8), and they will also enable identification of rare coding and noncoding variants associated with COPD and COPD-related phenotypes. In addition, WGS of COPD will provide insights into genetic determinants of omics data types, enable studies of resistant smokers in TOPMed, and empower better genotype imputation in studies without WGS data.

Murine Models of COPD Susceptibility Genes

Although GWAS have identified multiple genomic regions that contain COPD susceptibility determinants, the key gene (or genes) in those regions is typically unknown; multiple genes can be located near GWAS signals, and functional variants often do not influence the nearest gene. One approach to implicate the key gene within a COPD GWAS region is to determine whether orthogonal evidence exists to support a particular gene in COPD pathogenesis from animal models. As shown in Table 1, there are currently seven genes that are located within (or near) genome-wide significant association regions for COPD and that have been implicated in emphysema pathogenesis on the basis of a knock-out or transgenic murine model. These seven genes (HHIP [9], FAM13A [10], IREB2 [11], AGER [12], MMP1 [13], MMP12 [14], and SFTPD [15]) point to potentially relevant biological processes involved in COPD pathogenesis. Some of these biological processes, such as protease–antiprotease balance (related to MMP1 and MMP12), have been known for decades; however, other biological processes, such as mitochondrial iron (related to IREB2) have not been widely studied before the GWAS era.

Table 1.

Functional validation of chronic obstructive pulmonary disease genome-wide association study genes in murine models

Gene	Reference	Model	Phenotype	Postulated Biological Effect/Pathway
MMP1	D’Armiento et al., 1992 (13)	Transgenic	Increased emphysema	Collagenase activity
MMP12	Hautamaki et al., 1997 (14)	Knock-out	Decreased emphysema	Metalloelastase activity
SFTPD	Wert et al., 2000 (15)	Knock-out	Increased emphysema	Matrix metalloproteinase activity
HHIP	Lao et al., 2015 (9)	Heterozygous knock-out	Increased emphysema	Lymphocyte activation
AGER	Sambamurthy et al., 2015 (12)	Knock-out	Decreased emphysema	Neutrophil recruitment
IREB2	Cloonan et al., 2016 (11)	Knock-out	Decreased emphysema and airway disease	Mitochondrial iron
FAM13A	Jiang et al., 2016 (10)	Knock-out	Decreased emphysema	Wnt/β catenin

Omics Approaches to Identify Key Genes and Functional Variants within COPD GWAS Loci

As shown in Figure 1, GWAS have been extremely successful in identifying genomic regions that are clearly statistically associated with complex diseases and other phenotypes; more than 60,000 genome-wide significant SNP–phenotype associations have been reported. However, the functional variants responsible for those associations have only been identified in a tiny fraction of associated regions. Integration of GWAS results with omics data, including transcriptomics, epigenetics, proteomics, and metabolomics, can help to identify the key gene and the functional variants driving those GWAS associations.

Figure 1.

Disconnect between GWAS associations and functional variant identification in complex diseases and related phenotypes. GWAS = genome-wide association study. Notes: 1) For more details, see Reference 21. 2) GWAS Associations from GRASP at https://grasp.nhlbi.nih.gov/Search.aspx on June 3, 2018.

Lamontagne and colleagues integrated expression quantitative trait locus (eQTL) data from 1,038 lung tissue samples with International COPD Genetics Consortium GWAS results (16). They assessed transcriptome-wide eQTLs, Bayesian colocalization, and Mendelian randomization evidence for shared effects of association signals on eQTL and COPD GWAS. They identified the statistically most likely causal gene in 32% of loci that included multiple genes, including DSP. However, not every GWAS SNP will be an eQTL in lung tissue samples.

The COPD GWAS region on chromosome 15q25 has been intriguing but challenging to understand. That genomic region clearly contains genetic determinants related to smoking behavior and nicotine addiction (17), but it also appears to contain COPD genetic determinants unrelated to nicotine addiction (18)—potentially explained by IREB2 (11). Nedeljkovic and colleagues recently studied four COPD GWAS SNPs on chromosome 15q25 in the Rotterdam study (19). They examined eQTLs in 1,087 lung tissue samples and methylation QTLs in 1,489 blood samples. They found that all four SNPs were eQTLs and methylation QTLs for IREB2 (as well as for other genes in that region), and they also found differential methylation in COPD cases versus control subjects for CpG sites near IREB2. They concluded that DNA methylation may affect IREB2 expression to influence COPD risk; however, the functional variants for COPD pathogenesis on 15q25 are still unclear. Of interest, a trans-eQTL was also found for FAM13A in this 15q25 region, potentially suggesting a biological network connection between COPD GWAS loci on chromosomes 15q25 and 4q22.

Efforts to identify functional variants systematically within GWAS regions have traditionally been laborious and time consuming. Zhou and colleagues identified a likely functional variant in the HHIP COPD GWAS region by using chromatin conformation capture (3C) assays that demonstrated a localized segment of the GWAS region interacted with the HHIP promoter (20). Subsequently, they demonstrated enhancer activity within a 500-bp region and an SNP that altered binding to a transcription factor (SP3) that likely regulates HHIP (20). More recent efforts using higher-throughput approaches like massively parallel reporter assays hold promise for increasing the rate of functional variant identification in complex diseases (21).

Network-based Approaches to COPD Susceptibility and Heterogeneity

With a few exceptions (e.g., AGER [22]), COPD GWAS genetic determinants are primarily located in noncoding, regulatory regions of the genome. Although likely COPD susceptibility genes typically do not show substantial differences between COPD cases and control subjects for lung tissue gene expression levels, Morrow and colleagues noted that the genes that they interact with (based on protein–protein interactions, trans-eQTLs, etc.) are often differentially expressed in relation to COPD disease status or lung function (23). Thus, COPD susceptibility is determined by a network of interacting genes and proteins. Boyle and colleagues speculated that there are primary genetic drivers of a complex disease, and genes that are connected to these primary drivers will also influence disease susceptibility through their network connections to the primary drivers (24). This omnigenic model is concordant with network medicine views of complex diseases (25).

Multiple approaches can be used to identify the network components of a complex disease like COPD, including correlation-based networks, gene regulatory networks, protein–protein interaction networks, and Bayesian networks. Morrow and colleagues used weighted gene coexpression networks, a form of correlation-based networks, to identify a disease network module that was differentially expressed between COPD cases and controls; this network module was strongly related to B lymphocyte function (23). Network approaches can also be used to investigate COPD heterogeneity; Chang and colleagues applied network-based stratification to gene expression data, and they identified clusters of subjects with COPD on the basis of gene expression that were also related to clinical characteristics of COPD heterogeneity in both the ECLIPSE (Evaluation of COPD Longitudinally to Identify Predictive Surrogate End-points) and COPDGene (Genetic Epidemiology of COPD) study populations (26). Further research will be required to identify cell- and tissue-specific networks and to validate network relationships in the laboratory.

Unresolved Questions and Future Research Directions

Although substantial progress has been made in COPD genetics over the past 3 decades, many unresolved issues remain to be addressed. Only small percentages of the estimated genetic contributions to COPD susceptibility (27) and lung function levels (4) can be explained by known genetic loci. Larger genetic association studies will be required to create a more comprehensive set of COPD risk loci. The role of rare genetic variants in COPD genetic risk and heterogeneity should be investigated with whole-exome sequencing and/or WGS studies, because the relative impact of common versus rare genetic variants in COPD is still uncertain. It is unclear whether all genetic determinants of COPD will be primarily genetic determinants of lung function in general population samples or if a subset will be related directly to lung parenchymal and airway destruction without influencing spirometric values in nonsmoking general population samples. In addition to larger genetic association studies, the functional genetic variants in COPD GWAS loci, and the genes that they influence, need to be identified. The relationships between COPD GWAS genes need to be characterized, and the impact of the biological networks related to COPD susceptibility genes on COPD pathogenesis, heterogeneity, and progression needs to be understood. The epigenetic marks that influence COPD risk should be found, and the interactions between epigenetic and genetic risk factors should be interrogated. Finally, we need to determine how to integrate different omics data types most effectively to provide insights into COPD pathogenesis and heterogeneity. Progress in these areas could lead to improved understanding of COPD pathogenesis that will likely lead to more accurate diagnosis and prognosis as well as molecular targets for new treatment development.