SDB is known to aggregate strongly within families. In the Cleveland Family Study, individuals with an affected relative have been shown to have a 1.5-fold greater risk of having OSA themselves (2
). The heritability for AHI is 32–36% in both European Americans and African Americans (4
). In this study, we sought to evaluate the role of polymorphisms in 52 candidate genes in explaining the familial aggregation of this disorder. The set of candidate genes was selected based on biologic knowledge of relevant pathways, similarity in phenotype to monogenic diseases, and linkage data from our cohort. Our results support a potential pathogenic role for polymorphisms in GDNF and CRP in European Americans and for a polymorphism in HTR2A in African Americans. The persistence of associations between SDB with CRP and GDNF after BMI adjustment suggests that these genetic variants influence SDB susceptibility through obesity-independent pathways. In contrast, the attenuation of the association between HTR2A and SDB suggests that this association is likely through the influence of HTR2A on weight.
The finding that GDNF variants are associated with SDB phenotypes is especially interesting in light of the likely influences of ventilatory control abnormalities in the pathogenesis of SDB. Ventilatory control abnormalities, such as those that influence the sensing of oxygen or CO2
or ventilation at sleep state transitions, may predispose to OSA by promoting ventilatory instability (and periodic breathing) (12
), impairing the arousal response to airway obstruction (13
), or contributing to imbalanced activation of upper airway muscles compared with chest wall muscles (14
). GDNF plays a particularly important role in the development of neural pathways vital for normal respiration, specifically influencing the development and differentiation of noradrenergic neurons, including those in the A5 nucleus of the ventrolateral pons, which plays a critical role in respiratory pattern generation (15
). The knockout of the GDNF gene results in abnormal central respiratory output. In addition, GDNF seems to play a trophic role for sensory afferent neurons in the carotid body (16
), and thus may be important in the development of hypoxic responses. The importance of GDNF in ventilatory control is highlighted by the fact that severe mutations in GDNF are associated with the central congenital hypoventilation syndrome (17
). Although reductions in respiratory responses to hypoxia and to inspiratory load challenges among family members of OSA probands have been demonstrated (18
), this is the first evidence that variants in a gene in a ventilatory control pathway are associated with SDB.
The finding that variants in CRP were associated with SDB provides further evidence that pathways that mediate inflammation may be important in the pathogenesis of SDB. Although most research of inflammation and OSA has considered inflammation as a response to OSA-related stresses (21
), it is also plausible that abnormalities in inflammatory pathways may contribute to OSA severity by influencing pharyngeal patency through effects on mucosal edema, by contributing to pharyngeal neuropathic changes (24
), or possibly through effects on central ventilatory control. Prior small candidate genes studies have suggested that SDB is associated with other inflammatory cytokines, such as tumor necrosis factor-α (25
) and IL-6 (27
). If replicated, such findings may suggest a role for antiinflammatory therapy in the management of OSA.
Because serotonin plays an important role not only in sleep-wake and appetite regulation but also in upper airway dilator muscle activity through modulation of hypoglossal motor output, serotonergic receptors have gained consideration as candidate genes for SDB. In particular, the serotonin 2A receptor has been found to be the predominant excitatory serotonin receptor subtype at hypoglossal motor neurons (28
), and administration of serotonin 2A receptor agonist improves upper airway stability in animal models (29
). Although HTR2A may influence SDB through pleiotropic pathways (influencing both airway function and obesity), the association we observed was attenuated with BMI adjustment. It was also stronger in our African American sample, which tended to be more obese. Thus, further disentangling the influence of this genetic variant on intermediate pathways may require a larger sample with a greater BMI range. In both Chinese and Turkish populations, the A allele in a SNP in the HTR2A promoter region (A-1438G; rs2070040) has been associated with greater OSA severity (30
). This SNP is in weak LD (D′ = 0.12; r2
= 0.01) with the strongest OSA SNP identified for the African American cohort in this study.
In addition to our primary findings (q values ≤ 0.10), we observed suggestive associations between the SDB traits with genetic variants in the leptin receptor (LEPR) in African Americans, hypocretin 2 receptor (HCRTR2) in African Americans, and endothelin-1 (EDN1) in European Americans (lowest q-values of 16, 16, and 12%, respectively) highlighted in the online supplement. Leptin plays a key role in weight homeostasis and has been implicated in influencing hypercapnic ventilatory drive (32
). Hypocretin/orexin is a neurotransmitter known to have important effects on sleep/wake regulation, with more recent work implicating hypocretins in influencing upper airway neuromuscular activity (33
). Circulating hypocretin levels have been reported to be inversely correlated with OSA severity (34
). Prior analysis from European Americans in the CFS found evidence for linkage to AHI at 6p12, the location of the hypocretin receptor 2 (HCRTR2) (3
). Endothelin 1 (EDN1), a potent vasoconstrictor implicated in hypertension, is associated with respiratory failure at birth in EDN1 knockout mice and hypoventilation and blunted ventilatory response to hypoxemia and hypercapnia in mice with one functional EDN1 gene (35
). EDN1 knockout mice also show marked craniofacial morphologic abnormalities characteristic of disturbed pharyngeal arch development, as seen in humans with OSA caused by Treacher Collins syndrome and Pierre Robin syndrome (36
). In a European population, evidence for an increase in the AHI was observed for homozygote carriers of the minor allele for a missense coding SNP (rs5370) in EDN1 (37
). The same SNP allele showed an increased risk of AHI greater than or equal to 15 in the current sample of European Americans and can be considered replication at a nominal 0.05 threshold.
It is of interest that modeling the dichotomous and continuously measured traits revealed both overlapping and unique associations. Whereas genetic variants for the OSA trait are most likely to identify susceptibility loci influencing clinical disease, those that are associated with AHI are likely variants that are associated with incremental increases in the AHI across a wide range.
As has been reported for other traits, many of the identified genetic regions associated with OSA differed between the two races. This may reflect differences in allele frequencies for the causal variants or differing linkage disequilibrium patterns. In addition, this may reflect a difference in the relevant mechanisms for OSA pathogenesis between the two groups. For example, the importance of various craniofacial morphologies on OSA risk differs by race (38
). Whether relative differences exist in other intermediate phenotypes, such as ventilatory control, is unknown. The fewer number of positive SNPs identified for the African American cohort may also reflect the stricter threshold for significance used in this group. Because of the greater genetic complexity in this population, nearly double the numbers of SNPs were required for genotyping to obtain the same level of coverage for the genes of interest, and so double the number of comparisons for which to account.
Two areas of overlap should be noted across the two races. First, rs2808630 found in the 3′ untranslated region of CRP was associated with both OSA and AHI in European Americans after adjustment for multiple comparisons and also met nominal significance criteria for association to OSA in African Americans. In addition, two separate SNPs in EDN1 were suggestive of an association with OSA in European Americans (rs2071943) and African Americans (rs9296344).
Given that the genes studied were selected based on biologic plausibility, we highlighted findings that met a threshold FDR of 10%. This threshold corresponds to a nominal P
values of 0.0002 and 0.00009 in the European American and African American cohorts, respectively, and is much stricter than prior genetic work in OSA and more conservative than a suggested nominal P
value of 0.00005 for candidate gene studies (39
). However, even with this approach, the q-value results demonstrate that three of the four primary findings would meet a stricter 5% FDR threshold.
To our knowledge, this is the largest candidate gene study for sleep apnea to date. We were able to assess simultaneously the association between SDB phenotypes and 52 genes that have strong a priori
evidence for being involved in sleep apnea pathogenesis, and were able to examine associations in two racial groups. Of note, this study includes the only African American cohort to date to have undergone a genetic association study for SDB, despite the fact that this group is at risk for OSA at an earlier age (40
). Strengths include the relatively large sample size compared with prior studies in this field and the extensive coverage obtained for each of the genes considered. Although genome-wide association has gained extensive popularity for identifying causal variants for complex diseases, because of the large number of hypotheses being simultaneously tested in such a study design, the sample size required to obtain statistical significance with genome-wide association is substantially larger than with candidate gene studies. This is a particular problem with regards to SDB research given the substantial expense of overnight sleep studies required for phenotyping.
The findings in this study await replication in independent cohorts to confirm the generalizability of the identified associations. Furthermore, more detailed sequencing is needed along with functional assays to identify the causal variants, if any, at each of the regions associated with the SDB phenotypes. Nevertheless, this study identified genetic regions in both European American and African American populations worthy of further investigation as containing risk alleles. Given the increasingly recognized adverse impact of SDB on diabetes, heart disease, and stroke risk, these variants may prove also to be important in determining cardiovascular risk.