Genetic Underpinnings of Obstructive Sleep Apnea: Are We Making Progress?

Sleep disorders are common and affect all age groups with significant public health implications. While substantial progress has been made in identifying the genetic basis of some forms of sleep disorder such as restless leg syndrome^1,2 and narcolepsy,³ the genetic basis of OSA remains to be determined. Sleep apnea is a highly heterogeneous condition defined by recurrent reductions (hypopneas) or cessations (apneas) in breathing during sleep as a result of narrowing or collapse of the pharyngeal airway. Obstructive sleep apnea (OSA) results from reduction in airflow in the presence of respiratory effort while central apnea results from the absence of respiratory effort.⁴ Apart from apnea of prematurity and newborns, OSA is significantly more common than central apnea, with a prevalence of approximately 2-4% of the adult US population.⁵ The key pathophysiological feature in OSA is the occurrence of upper airway obstruction during sleep that does not occur during wakefulness resulting in recurrent arousal from sleep, excessive daytime sleepiness, a substantial increased risk of road traffic accidents and cognitive impairment.⁶ OSA has also been reported as an independent risk factor for diurnal hypertension, stoke and increased adverse cardio-vascular events.^7,8

The etiology of OSA is complex; it is not restricted to airway muscle dysfunction alone but rather arises as a consequence of a number of interrelated pathologies and risk factors. The strongest risk factors for OSA are obesity, age and gender.^9,10 Approximately 60–90% of patients with OSA can be classified as morbidly obese, however, not all obese individuals are affected with OSA.¹¹ The prevalence of OSA is 2-3 times higher in males than females, risk in females increases after menopause and with increased BMI.¹² Other risk factors include, craniofacial abnormalities, race and certain congenital conditions, such as Marfan's syndrome, Down syndrome and the Pierre Robin sequence as well as certain acquired conditions, such as acromegaly, hypothyroidism, and menopause.

Environmental factors that are known to exacerbate OSA include alcohol ingestion which reduces the activity of the genioglossus muscle,¹³ sedative use, sleep deprivation, tobacco use,¹⁴ and reduced nasal patency due to congestion and respiratory allergies.¹⁵ Finally, age of onset also masks significant heterogeneity in the etiology of OSA. While the incidence of pediatric onset OSA and its sequelae are similar to adult onset disease,¹⁶ most cases are attributed to enlarged tonsils, adenoids and soft palates resulting in narrowing of the upper airways,¹⁷ which can be alleviated by tonsillectomy.¹⁸ Genetic factors clearly play a role in OSA. Familial aggregation of OSA has now been described for over 30 yr,¹⁹ with over 43% of children with OSAS in one study having at least one relative with OSAS symptoms.²⁰ Overall, the genetic component of OSA has been estimated at 40%, with the rest attributable to environmental factors.²¹

OSA has been the subject of numerous genetic studies, including four genome-wide linkage analysis studies in European-Americans,^22,23 African-Americans,^23,24 and Filipinos.²⁵ OSA was phenotyped on the basis of apnea/hypopnea index (AHI) or respiratory disturbance index (RDI) measured using either overnight, in-home, portable monitors or by specialist-attended polysomnography at dedicated sleep facilities. Multipoint variance component linkage analysis was performed for the OSA-associated quantitative phenotypes of AHI and BMI and bivariate multipoint analysis was performed on RDI and metabolic syndrome (MeS) components. All four studies identified several regions of the genome that showed suggestive, but not conclusive, linkage to OSA.

In addition to the linkage studies over 30 candidate gene association studies have been reported in OSA²⁶ using various definitions of the syndrome and in a variety of different populations. Commonly studied candidate genes have included the angiotensin converting enzyme (ACE), adrenergenic receptors (ADRB2/3), serotonin receptors and transporters (HTR2A, HTR2C, SERT), endothelins (EDN1), leptin (LEP, LEPR) and apolipoproteins (APOE4). Several positive associations have been reported, however, none have been consistently replicated. As with most candidate gene studies, most were either underpowered, poorly controlled, or poorly phenotyped.

Larkin et al.,²⁷ have recently reported a large candidate gene study in approximately 1400 familial samples of African American and European American ancestry (n = 729 and 694, respectively). They selected fifty three candidate genes based on their putative function in craniofacial morphology, obesity and ventilator control, they also included several of the previously studied candidate genes listed above. Association testing was carried out against apnea-hypopnea index (AHI) as a quantitative trait and OSA as a dichotomous trait. The authors report association of three SNPs with OSA as a dichotomous trait in European Americans two in the glial-derived growth factor (GDNF) gene and one in the C-reactive protein gene (CRP), however, none of the associations remained significant after adjusting for BMI. The quantitative analysis of AHI also showed three associated SNPs in CRP and GDNF, two of which remained significant after correction for BMI, albeit at a false discovery rate (FDR) of 10%. No significant associations were reported in the African Americans with either OSA or AHI after adjusting for BMI. While the study covered a broader range of putative candidate genes compared with previous studies, the number of OSA cases was still relatively small (n = 190 African Americans and n = 144 European Americans) resulting in an underpowered study. Further studies will therefore be required to determine if the reported association between GDNF, CRP and AHI replicate following more stringent Bonferroni correction for multiple testing. More recently, variants in tumor necrosis factor (TNF) have been identified as important detriments of the variance in excessive daytime sleepiness associated with OSAS in children.²⁸ Association with the classic tumor necrosis factor (TNF) –308 SNP has also been recently described although this is a highly gene dense locus with very extensive LD patterns into neighboring MHC genes.²⁸ However, this association is more likely to modulate the degree of sleepiness that patients with OSAS experience rather than act as a bona fide risk for development of OSAS.²⁹

Recently, genome-wide association studies (GWAs) which interrogate upward of 600,000 SNPs across the genome have become feasible in large cohorts of patients and controls.³⁰ There are numerous advantages of GWAs approaches compared with candidate gene studies including: a) interrogation of the entire genome which allows for hypothesis-free testing of all genes compared with candidate gene approaches that rely on a highly subjective process of candidate selection b) dense coverage across genic regions which allow for a better assessment of association in relation to local genetic architecture compared to candidate gene studies that typically report on a few SNPs per gene c) genome-wide data can be used for quality control and estimation of population stratification d) GWAs allow for consistent replication of associations compared with candidate gene studies that frequently report on associations with different SNPs from the same gene e) well established statistical frameworks and benchmarks for statistical significance of discovery and replication signals.³¹ The result of which has been the discovery of upward of 1200 novel associations in over 200 complex and quantitative traits (http://www.genome.gov/gwastudies/) by GWAs that were previously intractable to candidate gene studies or linkage.

Although GWA studies have been reported for sleep related phenotypes such as duration and sleepiness, the only sleep disorder studies reported to date have been on narcolepsy and restless leg syndrome with no published GWAS either in adults or children for OSA.^{1–3,32–34} However, in this issue, Varvarigou and colleagues³⁵ performed a meta-analysis across a total of 31 population-based case-control studies reporting allele-frequency data on 48 polymorphism-OSA associations. As expected, sample sizes were generally small as were genetic effect sizes, rendering each study markedly underpowered to detect odds ratio as high as 2. Of multiple candidates examined, only the TNFA variant –308 A/G at rs1800629 was significantly associated with OSA under an additive model (3 studies, odds ratio = 1.82; 95% confidence interval = 1.26-2.61) and these results were robust to alternative genetic models. The authors concluded that the developing field of OSA genetics is currently dominated by small and underpowered investigations, and that the TNFA association of rs1800629 from their meta-analysis, needed to be replicated in larger studies using more comprehensive genotyping methods.

To summarize the genetic studies in OSA to date, the linkage studies reported modest non-overlapping linkage peaks without follow-up fine mapping. Thus, no gene variants have been identified to date using linkage. Most attention in the candidate gene studies has been focused on the role of APOE e4 isoform with both positive and negative associations being reported. A recent meta-analysis of these studies, however, showed lack of evidence for association.³⁶ A recent study examining associations of tagged SNPs in 52 candidate gene found associations with OSA in C-reactive protein (CRP) and glial cell line-derived neurotrophic factor (GDNF).²⁷ This candidate gene study used a 10% False Discovery Rate strategy and, as the authors acknowledge, the results have yet to be replicated. The meta-analysis by Varvarigou and colleagues lending support to association in the TNFA gene is similarly underpowered as the authors discuss.³⁵ Thus, the genetic underpinnings of OSA remain to be determined and as a complex genetic disorder the most appropriate approach would be genome wide association.³¹

Study design is the most important consideration for a successful GWAs particularly when applied to a heterogeneous complex disease such as OSA. Consideration should be given to age of onset of the cases and controls, the apparent etiological differences between pediatric and adult onset cases indicates they should be examined separately and compared with age-matched controls. Confounding variables are also likely to represent a significant challenge, particularly obesity amongst the adult cases where approximately 60–90% of adults with OSA are overweight, and the relative risk of sleep apnea from obesity (BMI > 29 kg/m²) is ≥ 10. An inadequately controlled study of adult OSA could therefore result in the identification of genes associated with obesity rather than genes associated with an increased risk OSA as occurred with type 2 diabetes.³⁷

The underlying heterogeneity in the etiology of OSA, which means that there is no single common phenotype but rather that it arises as a result of one or more of several contributory phenotypic factors acting alone or in concert, can also be exploited to identify the genetic basis of the disease. Intermediate phenotypes such as craniofacial morphology, upper airway control, ventilator control, and sleepiness can be studied individually. The advantages of using intermediate phenotypes over clinical diagnoses include reduced heterogeneity and increased robustness against confounders, the disadvantages include the difficulty in obtaining control individuals with the appropriate physiological metrics as these can be costly or invasive to obtain. Finally, OSA is global concern that affects individuals of all ancestries by using discovery or replication cohorts from multiple ancestries the possibility of false positives due to lifestyle-associated confounders can be reduced.

Knowledge of genetic risk factors is essential first step in developing predictive models that incorporate both genetic and phenotypic markers, thereby enabling early diagnosis and intervention.^17,33 Ultimately, these measures can result in reduction of morbidity and public health concerns associated with OSA in children and adults who are at increased risk, allowing for increased monitoring and reduction in personal risk by lifestyle modification (e.g., weight control). Therefore, the identification of genetic variants associated with increased risk for OSA could potentially translate into earlier recognition and treatment with reduced morbidity, and may also serve to identify potential targets for novel therapies,³⁸ and while linkage and association analyses have met with limited success, properly designed and well-powered GWA studies should begin to uncover the genetic basis of this serious and increasingly common disease. Well powered and quality controlled GWAs are therefore needed to determine if common variants play a role in OSA. While sequencing studies may soon be replacing standard GWAs, it is the opinion of these authors that GWAs has been highly successful in uncovering new associations and biology in complex diseases and should be performed in OSA prior to more expensive whole exome or whole genome sequencing studies, with targeted sequencing follow up at interesting GWAs loci in search for more rare causative variants.

Taken together, sleep disorders are prevalent conditions with significant public health implications. Significant progress has been made in identifying the genetic basis of sleep disorder such as restless leg syndrome (RLS) and narcolepsy, whereas the genetic basis of obstructive sleep apnea (OSA) remains to be determined. Several linkage and candidate gene studies have been conducted and while interesting genes have been reported, most have come up short in replication suggesting that their role in the pathogenesis of OSA remains unconfirmed. No GWAS has been reported yet in OSA, whereas GWAS on Narcolepsy and RLS have successfully identified genome-wide significant loci and replicated to standard criteria. While exome sequencing and high-density array-based approaches are likely to facilitate the discovery of novel OSA genes, GWAS remains untested and should be applied prior to significantly more costly approaches, such as whole-genome sequencing.

CITATION

Sleiman P, Hakonarson H. Genetic underpinnings of obstructive sleep apnea: are we making progress? SLEEP 2011;34(11):1449-1452

DISCLOSURE STATEMENT

The authors have indicated no financial conflicts of interest.