This study represents one of the first systematic evaluations of a large number of genetic loci identified as relevant to heart, lung, blood, and sleep phenotypes in relationship to sleep apnea. OSA represents one of the most common sleep disorders with a prevalence of 2 to 4% in children and greater than 10% in adults 
. The substantial neurocognitive and cardiovascular morbidity attributed to OSA as well as the likely rising prevalence of this condition with the growing obesity epidemic make understanding the genetic basis for this disease an important priority. Using genetic studies to identify novel molecular pathways involved in OSA pathogenesis may allow for the development of new treatments for this disorder – an important goal given the relatively poor tolerance of current treatment options.
Using data collected from three large cohorts and a genotyping platform that selected for candidate genes relevant to cardiovascular, respiratory and sleep physiology, we identified several SNPs associated with OSA related phenotypes.
Among AAs, the lysophosphatidic acid receptor 1 (LPAR1) gene was identified as a potential susceptibility locus. The rs7030789 SNP in an intronic region of LPAR1 was significantly associated with AHI in the discovery sample, demonstrated nominal association with the OSA phenotype in the CARe dataset as well as in the replication analysis comparing clinical OSA cases to controls. In addition, this locus showed evidence for replication in EAs strengthening the evidence base for a true susceptibility locus near this SNP. A locus close to LPAR1 has been shown to increase LPAR1 expression and subsequently increase circulating monocyte numbers 
, suggesting a pro-inflammatory role for this gene. In addition, LPAR1 is expressed in proliferating regions of the embryonic cerebral cortex 
. An LPAR1 knockout mouse model has been found to have abnormal innate behaviors as well as craniofacial abnormalities 
, suggesting potential neural and/or skeletal mechanisms for OSA pathogenesis.
The other loci identified in the CARe cohorts included one in plekstrin (PLEK), which is a substrate for protein kinase C in platelets and a wide range of leukocytes including monocytes and macrophages. This gene plays a role in actin assembly; knock out of PLEK in mouse models results in defective exocytosis 
. In addition, there was suggestive evidence for an association with a locus in the gene for the inositol triphosphate receptor 2 (ITPR2). The ITPR2 gene plays a central role in intra-cellular calcium regulation, which is vital to cellular stability, cellular adhesion, and second messenger activity. Prior candidate gene studies have associated SNPs in ITPR2 with markers of inflammation and endothelial dysfunction as well as blood pressure phenotypes 
. Together with the LPAR1 findings, these results suggest alterations in inflammatory pathways including monocyte/macrophage function may be particularly important in OSA pathogenesis among AAs as well as the possibility of overlapping pathways that may mediate both hypertension and OSA. However, this connection between loci associated with inflammation and OSA cannot distinguish between the possibilities that inflammation is a primary causal factor for OSA versus that inflammation plays a secondary role in increasing OSA severity among those with an underlying predisposition.
In EAs, a polymorphism in PTGER3, a prostaglandin E2 receptor, was significantly associated with OSA. In a replication analysis using data from a clinic-based cohort which had a skewed distribution of AHI due to high representation of patients with sleep apnea or sleep apnea symptoms, this SNP was associated with an incremental increase in disease severity as measured by the AHI level. Although association with the dichotomous trait OSA was not replicated, this may be due to limited power for this trait since nearly 80% of subjects in the WASHS cohort met OSA criteria. The PTGER3 gene is expressed in neuronal tissue and modulates neurotransmitter release in both central and peripheral neurons. A haplotype analysis of PTGER3 recently suggested this gene may represent a risk factor for hypertension 
. Unfortunately, sleep apnea was not characterized in that study.
It is important to note that our findings were not replicated in the Cardiovascular Health Study. This cohort is substantially older than the cohorts from which the primary results are based and the mechanisms underlying OSA in the elderly may differ substantially from OSA in middle aged populations. For example, obesity plays a much weaker role in the elderly while issues related to ventilatory control are much more prominent as evidenced by the greater number of central events in older individuals. Only one of the associations (rs7030789 in LPAR1) could be replicated across both EAs and AAs. This may be due to differences in allele frequencies impacting power (as for rs1409986 in PTGER3), differences in linkage disequilibrium structure or other ancestry related effects.
A prior analysis of CFS data has also utilized a candidate gene approach 
. In that study, 53 candidate genes for OSA were genotyped and three loci were identified meeting chip-wide significance criteria, 2 in EAs (loci in C-reactive protein - CRP and glial cell derived neurotrophic factor - GDNF) and 1 in AAs (locus in serotonin 2A receptor - HTR2A). None of the top loci from this study were included as candidate genes in the prior report and one of the top loci from that report (GDNF) was not genotyped on the IBC platform used here. For the two overlapping loci (CRP and HTR2A), associations with apnea phenotypes were again confirmed among CFS EAs and AAs, although the level of statistical significance did not meet the stricter threshold used in this paper due to the greater number of SNPs interrogated. Of note, the CRP locus did not replicate among EAs from ARIC and FHS, demonstrating the importance of replication cohorts.
Of interest, although the African American sample was relatively small, it was derived from a single cohort that was assembled explicitly to enhance the power to detect genetic associations through recruitment of affected probands and multiplex families. OSA is common in minority populations and prior segregation analyses found stronger transmission of one or several alleles in AAs compared to EA families 
. Although genetic studies in AAs may be limited by the coverage patterns of existing chips, the genetic diversity of this admixed population also provides opportunities to identify novel variants and to perform in silico “fine mapping” 
. Despite the sample size, we identified two loci significantly associated with sleep apnea phenotypes in AAs and a third nearly significant association. Furthermore, we were able to replicate the strongest finding in an independent sample.
Strengths of this study include its use of standardization in both phenotyping and genotyping across several cohorts. For the primary analyses, sleep apnea was objectively measured from overnight polysomnography by a central laboratory minimizing measurement error and all genotyping was conducted by a single laboratory. The use of a broadly defined candidate gene approach (i.e., with genes selected for heart, lung, blood and sleep traits) provides balance in examining relevant but not narrowly defined susceptibility loci without the statistical penalties incurred by using a genome-wide association approach. This is particularly relevant for conditions such as OSA that are expensive to phenotype. On the other hand, this approach is not amenable to identifying completely novel loci.
We also conducted analyses in parallel to identify loci important in both African-Americans and those of European ancestry. Demonstrating a similar pattern of association between apnea phenotypes and rs7030789 in both groups provides additional evidence as to the presence of a true association. Another strength of the current study is the use of both principal component adjustment and genomic control correction to prevent inflation of the false positive rate due to cryptic population stratification or other sources of bias. Nevertheless, additional analyses from independent cohorts are needed to replicate the findings presented in this study.
Limitations of the work include the relatively small number of cohorts included and modest sample size compared to other cross-cohort genetic analyses. Unfortunately, few cohorts with detailed sleep apnea phenotyping as well as genotyping currently exist, particularly in minority populations. This limited our opportunity to conduct replication studies which will be needed to definitively confirm our findings. Furthermore, it should be noted that no functional data are yet available to demonstrate a direct effect of any of the variants identified with pathophysiological changes relevant to OSA.
Overall, the relatively small number of loci identified and replicated in this study presumably reflects a combination of the limited power to detect variants with low prevalence or small effect sizes and that many of the genetic variants contributing to OSA risk are either rare, have fairly modest effects, or both. This is in line with current views that OSA represents a complex disease having many genetic and environmental contributors each of which explains a relatively small proportion of the total attributable risk. The limited success in identifying genetic determinants of OSA might also reflect heterogeneity in genetic etiology across the various cohorts studied which differ in mean age and BMI. In fact, the association between OSA and rs7030789 appeared to be much stronger among leaner individuals. Another possible explanation is that causal variants with relatively large effects may exist in regions of the genome that were not interrogated in this candidate gene approach. Genome wide association (GWA) study designs will be required to assess whether such variants exist.
In summary, our results identify several loci, each of which has been implicated in inflammation, associated with a sleep apnea phenotype. While much work has focused on the potential pro-inflammatory effect of OSA, inflammation may play a causal role in OSA pathogenesis as well. Inflammation is prominent in both the mucosal and muscular layers of upper airway tissues of patients with OSA 
. In addition, inflammation may also impact ventilatory or sleep-wake control mechanisms relevant to OSA pathogenesis.
Of note, though all of the cohorts studied had a high prevalence of obesity, the identified associations were independent of obesity supporting the importance of identifying pathways leading to OSA within an obese population. Additional studies are needed to better understand the molecular pathways that increase susceptibility to this highly prevalent disorder, the overlap of such pathways with ones that also mediate cardiopulmonary disorders, and the extent to which the associations are moderated by ancestral background.