Introduction: The genetic architecture of sleep disordered breathing (SDB) traits in humans remains largely unknown. Transcription factors such as HIF1A regulate gene expression, facilitate transcriptional programs, and respond to environmental signals. Identifying how variations in transcription factor binding sites (TFBS) across the genome associate with SDB phenotypes may help identify common mechanisms underlying the pathogenesis of SDB and its cardiometabolic consequences. As part of our ongoing genetic analysis of SDB traits at sequence-level resolution, we examined TFBS network mutations in individuals participating in the NHLBI TOPMed Consortium.

Materials and Methods: We analyzed 3,525 individuals from 5 NHLBI research cohort studies (892 African-, 206 Asian-, 2013 European-, and 414 Hispanic/Latino-Americans) with overnight objective recordings of average and minimum oxyhemoglobin saturation (SpO2) during sleep, percent sleep with SpO2 <90% (Per90), and the apnea-hypopnea index (AHI) and deep sequencing coverage. Mixed-effect models were used to adjust for age, sex, BMI, cohort, and self-identified ancestry with empirical genetic relatedness matrices to account for population structure and relatedness. For each of 581 transcription factors, TFBS variants with predicted functional consequences (fathmm-MKL score > 0.5) across the genome were tested together in set-based analyses using MM-SKAT.

Results: The lead transcription factor associations were with CTCF (average SpO2 and Per90 SKAT p = 6.5 × 10-15 and 2.3 × 10-8 respectively), ERG (minimum SpO2 p = 7.8 × 10-22), and TAL1 (AHI p = 5.9 × 10-22). CTCF, FOXA1 and MYC were among the top 5 transcription factors in 3 analyses (average and minimum SpO2 and Per90), while PPARG, RELA, and TAL1 were among the top 5 transcription factors in 2 analyses.

Conclusions: Important regulators of cell cycles (FOXA1, MYC), transcriptional insulation (CTCF), inflammation (RELA, PPARG), lipid metabolism (PPARG), and hematopoesis (ERG, TAL1) are associated with SDB traits through altered transcription factor binding sites throughout the genome. These results provide further insight into the genetic regulation of traits associated with SDB.

Support: American Thoracic Society Foundation, K01 HL135405, R35 HL135818, R01 HL113338.

Introduction: Sleep-disordered breathing (SDB) is associated with increased risk of cardiovascular disease and mortality. There is a limited understanding of the bases for differences in patterns of overnight hypoxemia across individuals. The genetic architecture of SDB in humans is largely unknown, despite significant heritability. Further understanding may help elucidate mechanisms influencing obstructive sleep apnea and oxygenation as well as identify individuals at increased morbidity risk.
Materials and Methods: We performed a Whole-Genome Sequence Analysis (WGSA) in the TOPMed Consortium using the Apnea Hypopnea Index (AHI) and complementary measures of overnight oxyhemoglobin saturation (SpO2): average and minimum SpO2 and the percentage of the sleep episode with SpO2 <90% (Per90). WGSA single-variant and gene-based analyses included pooled samples with deep (>30×) sequence coverage of 71.7 million variants from the Cleveland Family Study (507 African-Americans, 486 European-Americans) and the Framingham Heart Study (459 European-Americans). We used mixed effect models adjusting for age, sex, BMI, cohort, and self-identified ancestry with empirical kinship to account for relatedness and population structure. Gene-based test variants were comprised of adverse coding mutations (e.g. stop loss) with strong predicted functional consequences (fathmm-MKL score >0.9).
Results: Heritability based on genetic variants at all frequencies was estimated at 0.29 (AHI), 0.44 (average SpO2), 0.20 (minimum SpO2), and 0.24 (Per90). AHI was associated near DIP2C and CLRN1-AS1 (rs75429350 p = 4.13 × 10-8; 3:150545337_T/G p = 4.87 × 10-8). Average SpO2 was associated near SNX3 (6:108585281_A/G p = 1.12 × 10-9) and C2orf73 (rs73931770 p = 4.03 × 10-8). Minimum SpO2 associations included regions near ADAM3A (rs150052514 p = 8.27 × 10-10) and rs148168881 (chr13q33 intergenic, p = 6.62 × 10-9). Per90 was associated with the AK5 region (rs201266006 p = 1.12 × 10-8). Secondary population- and sex-specific single variant analyses included an association between AHI and 12:8091955_T/C in females (SLC2A3 region, p = 7.74 × 10-11). PCSK9 and CHAF1B were significantly associated with minimum SpO2 in pooled gene-based analyses (Proprotein convertase subtilisin/kexin type 9, p = 1.19 × 10-6; Chromatin assembly factor I, subunit B, p = 2.92 × 10-6). ESRRG was associated with the AHI (Estrogen related receptor gamma, p = 1.21 x 10-6). Secondary gene analyses included an AHI association with KLHL1 (kelch like family member 1; European-American males p = 1.7 × 10-10).
Conclusions: These are the first WGSA analyses performed on sleep traits. Preliminary results suggest that novel loci are associated with SDB traits. Work is ongoing to characterize non-coding functional regions and extend these analyses using four additional studies with objective sleep recordings in TOPMed Phases 2 and 3 (n > 10,000).

A disease trait often can be characterized by multiple phenotypic measurements that can provide complementary information on disease etiology, physiology, or clinical manifestations. We previously performed analysis on the principal components of the phenotypes (PCPs) and principal components of the heritability (PCHs) for six obstructive sleep apnea hypopnea syndrome (OSAHS) related phenotypes using the data from the Cleveland Family Study, which include both African-American and European-American families. We demonstrates that principal components generally result in higher heritability and linkage evidence than individual traits and the PCHs can be transferred across populations. We identified the largest linkage peak for the third PCH (PCH3) on chromosome 8 with a maximum LOD score of 3.7 in CFS European Americans. In this study, 487 European Americans and 507 African Americans from CFS have been whole-genome sequenced (WGS) by the NHLBI Trans-Omics for Precision Medicine (TOPMed) Program. There are 144,565 variants identified in the 20Mb linkage region in CFS European Americans. We hypothesized that low frequency and rare variants contribute to the observed linkage evidence. We divided the low frequency and rare variants (defined as MAF<5%) into three groups: 1) functional coding variants, 2) functional non-coding variants and 3) the rest of variants. We observed 88 and 258 genes with at least 3 functional coding and non-coding low frequency or rare variants, respectively. We observed 1.5 to 2.3 times enrichment of genes associated with PCH3 using gene-based SKAT or burden analysis with or without adjusting for BMI. By restricting to low frequency or rare functional coding variants segregating within the families with family-specific LOD score >0.1, we identified 17 genes with at least 3 low frequency or rare variants. The enrichment was further improved to 7 times and we identified 6 out of the 17 genes with SKAT P<0.05. Three (DLGAP2, CSMD1 and ERI1) of the 6 genes were further replicated in CFS African Americans (SKAT P<0.05). Further replication using additional TOPMed cohorts is underway. Our study suggests analyzing multiple traits and combining linkage evidence in searching for low frequency and rare variants using WGS data can be fruitful.