Obstructive sleep apnea (OSA) is a common disorder associated with increased risk for cardiovascular disease, diabetes, and premature mortality. There is strong clinical and epidemiologic evidence supporting the importance of genetic factors influencing OSA, but limited data implicating specific genes. Disease severity is most commonly measured by the apnea-hypopnea index (AHI), which quantifies the number of complete and partial episodes of upper airway obstruction per hour of sleep.
We report genetic associations of AHI using high depth genomic sequencing data from the NHLBI Trans-Omics for Precision Medicine (TOPMed) project and focus on the Cleveland Family Study (CFS), a cohort with enrichment for OSA. We identified an associated 19 Mb linkage region on chromosome 7 in European Americans and examined rare variants that contributed to the observed linkage evidence in gene-based analysis. We identified 18 candidate genes (P < 0.05) associated with AHI under the linkage peak in CFS. Follow-up analysis of these variants in the 18 candidate genes in 4 independent TOPMed studies identified 21 non-coding rare variants in Caveolin-1 (CAV1) significantly associated with AHI (P = 5.4×10-7) in 3,259 European Americans. These non-coding variants significantly contribute to the CAV1 expression level in skeletal muscle (P = 3.8×10-4) in GTEx data. Caveolin-1 is a membrane scaffolding protein that is essential in the formation of plasma membrane lipid rafts and mediates cholesterol trafficking and regulates several signaling molecules including TGF-beta, TLR4 and endothelial nitric oxide synthase (eNOS). It is implicated in disorders associated with OSA, including metabolic and endothelial dysfunction, immune regulation and inflammation and lung remodeling. Common CAV1 variants were associated with severe OSA in a small case-control study. Our results suggest that caveolin-1 may play a significant role in OSA pathogenesis.

Background: Sleep duration is heritable, and short and long sleep durations are each associated with numerous adverse health outcomes. We performed a whole genome sequence (WGS) association study using multi-ethnic data from the Trans-Omics for Precision Medicine (TOPMed) program, freeze5b.

Methods: We harmonized self-reported usual sleep duration across eight cohorts. We designated subjects reporting usual sleep duration 5hrs as short sleepers, 6-8hrs as normal sleepers, and hrs as long sleepers. For binary traits, we removed cohort-ethnicity subgroups with less than two cases, and individuals with ambiguous sleep timing categories (e.g. “usual sleep between 5-6 hours”). This resulted in sample sizes of 22304, 20021, and 19486, for the quantitative-, short-, and long- sleep analyses, respectively.
We applied linear (quantitative duration) and logistic (short and long sleep) mixed-models, with random effects modeled via a genetic relatedness matrix to account for population structure. All models were adjusted for age, sex, and cohort-ethnicity subgroup. For continuous sleep duration, residuals were rank-normalized and then used in the association analysis. We tested the associations of single variants with minor allele count using the Score test for continuous sleep duration, and the SAIGE/SPA for short and long sleep. We also tested rare variants in aggregation units defined by genes and functional categories (high confidence loss of function, missense, protein altering indels, and synonymous), and by sliding window tests across the genome, testing all variants with a fathmm-xf score greater than 0.5 and maximum alternate allele frequency of 0.01, in non-overlapping windows of size 3000bp. For aggregate tests we used both the SKAT test with weights of Beta(1, 25), and unweighted Burden test. We required at least 2 variants in each set for SKAT, and at least 11 alternate alleles for the burden test.

Results: Rs141617869, located 27kb downstream of BMP15 in the Xp11.22 region, was associated with continuous sleep duration (p = 2.7e-08); and a rare variant in the 2p22.3 region was associated with short sleep (p=2.6e-08). The top gene set association was TNFRSF10D for short sleep (high confidence loss of function; SKAT p=3.4e-09), IL31RA for long sleep (high confidence loss of function; SKAT p=1.0e-08), and XRCC1 for continuous sleep duration (missense; burden p= 9.9e-07). There were additional significant sliding window associations (p<1e-8).