Shared Genetic Basis for Obstructive Sleep Apnea and Adiposity Measures

Abstract

Introduction

Obesity and obstructive sleep apnea each have a substantial genetic basis and commonly co-exist in individuals. The degree to which the genetic underpinnings for these disorders overlap has not been previously quantified.

Methods

A total of 1802 individuals from 310 families in the Cleveland Family Study underwent home sleep studies as well as standardized assessment of body mass index and circumferences at the waist, hip, and neck. In 713 participants with laboratory sleep studies, fasting blood samples were assayed for leptin, adiponectin, and resistin. Variance component models were used to estimate heritability and genetic correlations.

Results

The heritability of the apnea hypopnea index was 0.37 ± 0.04 and 0.33 ± 0.07 for home and laboratory sleep studies respectively. The genetic correlations between apnea hypopnea index and anthropomorphic adiposity measures ranged from 0.57 to 0.61 suggesting obesity can explain nearly 40% of the genetic variance in sleep apnea. The magnitude of the genetic correlations between apnea severity and adipokine levels was substantially less than those with anthropomorphic measures, ranging from 0.11–0.46. After adjusting for body mass index, no significant genetic correlation with apnea severity was observed for any of the other adiposity measures.

Conclusions

Substantial but not complete overlap in genetic bases exist between sleep apnea and anthropomorphic indices of adiposity, and this overlap accounts for more than one third of the genetic variance in apnea severity. These findings suggest that genetic polymorphisms exist that importantly influence sleep apnea susceptibility through both obesity-dependent and obesity-independent pathways.

Introduction

Obstructive sleep apnea (OSA) is an extremely common disease associated with substantial co-morbidity ¹. Obesity is one of the strongest risk factors for OSA such that a 10% weight gain is associated with a 6-fold increase in OSA risk ². Proposed mechanisms by which obesity might predispose to OSA include mechanical effects on both lung volume and upper airway caliber as well as neurohormonal effects on ventilatory drive ³.

Growing evidence has demonstrated familial clustering of OSA ^4–8. These studies have supported the contention that an underlying genetic susceptibility to OSA exists. Extensive research has also suggested that obesity has a strong genetic basis ^{9, 10}. Given the close relationship between these two disorders, it would not be surprising that many of their genetic risk factors would overlap. However, it is not known to what degree the genetic bases for these two disorders are shared. Prior analyses restricting to individuals with a body mass index (BMI) below a certain threshold or adjusting for BMI in multivariate analyses have suggested that underlying causative loci for OSA exist that are independent of BMI ^4–6. However, these analytic approaches do not permit quantification of the amount of sharing in genetic susceptibility for related phenotypes, a question of critical importance in addressing the genetic basis for complex traits. In this study, we sought to define the proportion of the genetic variance in OSA that is explained by obesity using multiple methods to quantify adiposity. These analyses also allowed us to compare the strength of the genetic correlation between OSA severity and a general measure of adiposity such as BMI to measures of regional fat deposition. Given evidence suggesting that fat deposition around the neck or waist may be more important in causing OSA than global levels of obesity ^11–13, and that fat distribution patterns are heritable independent of overall adiposity ^11–13, it is plausible that genes exist which predispose to OSA via effects on regional fat distribution without affecting overall adiposity. Therefore, we hypothesized that the genetic correlation to apnea severity would be greater for relevant regional adiposity measures than for BMI. In addition, since growing research indicates that the health outcomes of obesity (which overlap the health outcomes of OSA) are mediated in part by the action of adipose derived hormones, we also quantified the genetic correlation between OSA severity and adipokine levels.

Methods

Subjects

The Cleveland Family Study is a longitudinal family-based epidemiological cohort designed to study the genetics of obstructive sleep apnea (OSA). Details on recruitment of this cohort have been previously described ^{4, 14}. Briefly, index probands with a laboratory confirmed diagnosis of OSA and at least two first-degree relatives available to be studied, were recruited along with family members. A total of 2284 individuals from 361 families were studied on up to four occasions over 15 years.

Institutional Review Board approval was obtained from University Hospitals Case Medical Center and written informed consent was obtained from each participant. We certify that all applicable institutional and governmental regulations concerning the ethical use of human volunteers were followed during this research.

Phenotype collection

Subjects participated in phenotyping at up to four separate time points each occurring about 4 years apart. For the first three visits, data, including an overnight sleep study, were performed in the participants’ homes while the last visit occurred in a general clinical research center (GCRC). In all examinations, height was measured to the nearest 0.1 cm and weight measured to the nearest 0.1 kg with a digital scale. BMI was computed as the ratio of weight to height squared. Waist and hip circumferences were measured with the subject standing to the nearest 0.1 cm. Neck circumference was measured to the nearest 0.5 cm just below the thyroid prominence with the head in the Frankfort horizontal plane. All measurements were made in duplicate and averaged.

Overnight in-home sleep monitoring was performed with an Edentrace I or II monitor (Eden Prairie, MN) measuring airflow, chest wall impedance, pulse oximetry, and heart rate. Respiratory events were defined as cessations (apneas) or discrete reductions (hypopneas) in airflow or chest impedance, lasting at least 10 seconds and associated with at least a 3% fall in oxygen saturation. Sleep time was estimated from visual inspection of the sleep record, correlated with a sleep diary completed by the subject. The apnea hypopnea index (AHI) was computed by dividing the number of respiratory events by the estimated hours of sleep time. The AHI measured in this manner has been previously shown in our cohort to correlate closely with AHI obtained from full laboratory polysomnography (intraclass correlation = 0.83) ¹⁵.

In the most recent round of data collection, a subset of 725 individuals was selected based on expected genetic informativity by choosing pedigrees where siblings had extremes (either high or low) of AHI. A more detailed explanation of the selection scheme has been previously published ¹⁶. Selected individuals underwent more detailed phenotyping in the GCRC including laboratory polysomnography (Compumedics, Abbotsford, AU) followed by blood sampling between 7 and 8 a.m. after an overnight fast. Secondary analysis was performed in this group using the laboratory-derived AHI with adipokine levels as biomarkers of obesity and obesity-related metabolic pathways. Leptin was assayed using an ELISA (R&D Systems, Minneapolis, MN). Adiponectin and resistin were assayed using a Luminex immunoassay (Linco Research, St. Charles, MO).

Statistical analysis

In order to reduce potential age effects, the visit at which the subject was closest to the median age of the cohort (35 yrs) was utilized for the primary analysis evaluating the association between the AHI and anthropometric values. All traits were log transformed (after adding a constant of 1 when necessary) prior to analysis in order to approximate a normal distribution. Narrow sense heritability of the AHI and obesity-related traits as well as bivariate genetic correlations between AHI and obesity traits were conducted using the maximum likelihood estimate based variance component approach implemented in the statistical genetics software package, SOLAR version 4.0.7 ¹⁷.

Adjustment for ascertainment was performed by conditioning on probands. The variance component analysis partitions the variance of a phenotype into components due to measured covariates, additive genetic effects, and random environmental effects. Age, age squared, gender, age by gender interaction and height were included as covariates in all models. Narrow sense heritability (h²) is computed as the ratio of the additive genetic variance to the sum of genetic and environmental variances and represents the proportion of the total variance in a trait (after adjusting for covariate effects) that is explained by additive genetic effects. Bivariate models were also performed to simultaneously decompose the variances in two phenotypes. In this extension of the modeling, the covariance between two phenotypes can also be partitioned into genetic and environmental components. The genetic (ρ_g) and environmental (ρ_e) correlations can be related to the overall correlation (ρ_t) between the two traits by the equation: ${\rho }_t={\rho }_g\times \sqrt{h_1^2}\times \sqrt{h_2^2}+{\rho }_e\times \sqrt{1-h_1^2}\times \sqrt{1-h_2^2}$, where $h_1^2$ and $h_2^2$ correspond to the heritabilities of the two traits. Using this nomenclature, the proportion of the total genetic variance that is due to shared genetic effects is estimated by the square of the genetic correlation. Likelihood ratio tests were used to separately test the hypotheses that the two traits share no common genetic basis (ρ_g = 0) and that the two traits have the identical genetic basis (|ρ_g| = 1). Because results from preliminary analyses stratified by race demonstrated no differences in heritability or correlation estimates between Caucasians and African-Americans, data were pooled across races for this report.

Results

Of the 2284 individuals participating in the Cleveland Family Study with at least one sleep study, 480 were excluded for missing anthropomorphic data. An additional two probands were excluded as they had no biologic relatives who participated (due to non-paternity) leaving 1802 individuals from 310 families for the primary analysis with an average of 5.8 phenotyped individuals per family. For the analysis of the sample studied in the GCRC, 12 of the 725 subjects did not have complete adipokine profiles leaving 713 individuals from 139 families (5.1 individuals per family).

Descriptive information about participants is presented in Table 1. The population was approximately evenly split between men and women and between African-Americans and Caucasians. Using thresholds of 5, 15, and 30 apneas plus hypopneas per hour, the prevalence of OSA was 48.7%, 26.8%, and 15.9% respectively in the primary cohort. The laboratory cohort, reflecting the most recent exam cycle, was, as expected, somewhat older and heavier than the overall cohort.

Table 1.

Clinical characteristics of study subjects

	Home sample (N=1802)	GCRC sample (N=713)
Age (yr)	35.3 ± 19.4	41.3 ± 19.3
Male (%)	44.7	44.0
African-American (%)	48.6	58.6
AHI	14.8 ± 23.7	17.0 ± 25.0
AHI > 5 (%)	48.7	51.5
AHI > 15 (%)	26.8	32.7
AHI > 30 (%)	15.9	19.1
BMI (kg/m2)	29.5 ± 9.5	32.2 ± 9.3
BMI ≥ 30 kg/m2 (%)	42.2	54.8
Waist circumference (cm)	91.1 ± 22.9	97.1 ± 21.3
Hip circumference (cm)	105.6 ± 21.3	111.5 ± 18.3
Neck circumference (cm)	36.5 ± 5.8	38.5 ± 5.2
Leptin (ng/ml)		41.1 ± 49.6
Adiponectin (pg/ml)		19.1 ± 14.8
Resistin (pg/ml)		12.4 ± 7.0

Data presented as mean ± SD or as percentages.

AHI: apnea hypopnea index, BMI: body mass index, GCRC: general clinical research center.

The heritability estimates for AHI were similar in the two samples despite use of different sleep monitoring approaches: 0.37 (SE 0.04) in data derived from home studies and 0.33 (SE 0.07) when using data collected in the GCRC. The heritabilities for the anthropomorphic measures of adiposity were somewhat greater than for AHI ranging from 0.49–0.56 (Table 2). Among the biochemical measures, leptin demonstrated the lowest heritability while adiponectin and resistin had greater heritabilities than the anthropomorphic measures.

Table 2.

Heritability of apnea hypopnea index and adiposity traits

	Heritability
	Home sample	GCRC sample	GCRC sample with BMI adjustment
Apnea Hypopnea Index	0.37 (0.04)	0.33 (0.07)	0.27 (0.07)
Body Mass Index	0.55 (0.04)	0.56 (0.07)	---
Waist	0.50 (0.04)	0.49 (0.08)	0.39 (0.08)
Hip	0.49 (0.04)	0.50 (0.07)	0.35 (0.08)
Neck	0.55 (0.04)	0.51 (0.07)	0.43 (0.07)
Leptin		0.26 (0.08)	0.09 (0.07)
Adiponectin		0.62 (0.08)	0.62 (0.08)
Resistin		0.78 (0.07)	0.78 (0.07)

Maximum likelihood estimates displayed with standard errors in parentheses.

All models include age, age², sex, age*sex, and height as covariates.

BMI: body mass index, GCRC: general clinical research center.

With adjustment for BMI, the heritability of AHI was only modestly attenuated to 0.27 (SE 0.07) (Table 2, last column), supporting the notion that obesity cannot explain the majority of the genetic variance in AHI. Similarly, except for leptin, each of the other adiposity measures continued to have substantial heritability after BMI adjustment, suggesting each of these traits has genetic determinants that are unique from those that determine BMI.

The genetic correlation between AHI and each of the four anthropomorphic measures ranged from 0.56–0.67 (Table 3), suggesting each of these traits could explain similar degrees of the genetic variance in AHI. Correlations also were similar for data collected in the home and GCRC visits, suggesting that these associations did not differ according to the study visit type. Testing the null hypothesis that the genetic correlation between AHI and each of these anthropomorphic measures was zero (i.e., that there is no shared genetic basis between AHI and obesity) resulted in highly significant p-values rejecting this hypothesis. A second null hypothesis that the genetic correlation between these traits is one (i.e., that the genetic bases for AHI and obesity are identical) was also rejected as very unlikely. The magnitude of the genetic correlations between AHI and the three biochemical adiposity measures were weaker than for the anthropomorphic measures, with values ranging from 0.11 for resistin to 0.46 for leptin. The correlation with adiponectin was negative as expected given the known reduction in adiponectin levels associated with obesity ¹⁸.

Table 3.

Genetic correlation (ρ_g) of adiposity traits with apnea hypopnea index

	Home sample			GCRC sample			GCRC sample with BMI adjustment
	ρg	H0: ρg=0	H0: |ρg|=1	ρg	H0: ρg=0	H0: |ρg|=1	ρg	H0: ρg=0	H0: |ρg|=1
Body Mass Index	0.59 (0.06)	p < 10−10	p < 10−10	0.64 (0.09)	p < 10−5	p < 10−5	---	---	---
Waist	0.61 (0.06)	p < 10−10	p < 10−10	0.67 (0.09)	p < 10−5	p < 10−4	0.22 (0.18)	p = 0.22	p < 10−4
Hip	0.57 (0.06)	p < 10−10	p < 10−10	0.61 (0.10)	p < 10−5	p < 10−5	−0.06 (0.19)	p = 0.77	p < 10−5
Neck	0.59 (0.06)	p < 10−10	p < 10−10	0.56 (0.10)	p < 10−4	p < 10−6	0.05 (0.16)	p = 0.73	p < 10−5
Leptin				0.46 (0.16)	p = 0.015	p < 10−3	−0.43 (0.36)	p = 0.20	p = 0.80
Adiponectin				−0.34 (0.13)	p = 0.012	p < 10−7	−0.15 (0.15)	p = 0.32	p < 10−5
Resistin				0.11 (0.13)	p = 0.41	p < 10−7	0.01 (0.14)	p = 0.92	p < 10−5

Maximum likelihood estimates displayed with standard errors in parentheses. ρ_g: genetic correlation.

All models include age, age², sex, age*sex, and height as covariates.

BMI: body mass index, GCRC: general clinical research center.

After adjustment for BMI, the magnitude of the genetic correlation between AHI with both regional body fat measurements and biochemical adiposity measures decreased substantially and in no case was the residual correlation significantly different from zero (Table 3, last column).

The genetic and total phenotypic correlations between the various adiposity measures were computed to provide a framework in which to interpret the correlations with AHI (Table 4). The correlations between the anthropomorphic adiposity measures ranged from 0.78–0.95 and were substantially greater than the correlation between AHI and any of these traits. While leptin has a high genetic correlation to the anthropomorphic measures, the genetic correlations for adiponectin and resistin to the other adiposity traits was substantially lower and of similar magnitude to the correlations these traits have with AHI.

Table 4.

Genetic correlations between adiposity traits

	Home sample				GCRC sample
	BMI	Waist	Hip	Neck	BMI	Waist	Hip	Neck	Leptin	Adiponectin	Resistin
BMI	1.00	0.95 (0.01)	0.95 (0.01)	0.84 (0.02)	1.00	0.95 (0.01)	0.97 (0.01)	0.85 (0.03)	0.93 (0.07)	−0.38 (0.10)	0.15 (0.10)
Waist		1.00	0.90 (0.02)	0.86 (0.02)		1.00	0.91 (0.03)	0.84 (0.04)	0.90 (0.07)	−0.39 (0.11)	0.18 (0.11)
Hip			1.00	0.78 (0.03)			1.00	0.79 (0.05)	0.89 (0.07)	−0.32 (0.11)	0.16 (0.11)
Neck				1.00				1.00	0.65 (0.11)	−0.44 (0.10)	0.05 (0.11)
Leptin									1.00	−0.32 (0.15)	0.26 (0.15)
Adiponectin										1.00	0.03 (0.11)
Resistin											1.00

Maximum likelihood estimates displayed with standard errors in parentheses. Analyses performed controlling for age, age², sex, age*sex, and height.

BMI: body mass index, GCRC: general clinical research center.

Discussion

In this work we have for the first time quantified the degree of shared genetic variance between OSA and several measures of adiposity. The heritability of AHI determined from in home sleep studies and polysomnography performed in a GCRC were 37% and 33% respectively, which is similar to estimates in prior work from the Cleveland Family Study as well as other cohorts ^{5, 16, 19}. The genetic correlation between AHI and obesity as assessed by four different anthropomorphic measures of adiposity were all consistently in the range of 0.6, suggesting nearly 40% of the total genetic variance in OSA is shared with obesity. While this is significantly greater than no sharing, it does suggest that the majority of the genetic variance in OSA is mediated via obesity independent mechanisms. In addition, the genetic correlations between AHI and all adiposity measures were all significantly less than 1 in statistical testing, and the observed genetic correlations were substantially less than those observed between different obesity-related measures. For example, the genetic correlations among anthropomorphic traits ranged from 0.75 to 0.95, similar to what has been reported in other studies ²⁰. Thus, although the strongest recognized risk factor for adult OSA is obesity, these analyses strongly suggest that the genetic basis for OSA likely involves additional etiological pathways, such as those related to craniofacial anatomy, ventilatory control, and sleep/wake regulation.

On the other hand, the observed degree of genetic correlation between adiposity and OSA is greater than the correlations reported between obesity and other traits known to commonly co-exist with obesity. For example, the genetic correlations to adiposity traits are reported to be 0.22–0.28 for lipid phenotypes, and 0.13–0.16 for blood pressure phenotypes ²⁰. Thus, OSA appears to be much more closely related to obesity from a genetic standpoint than are components of the metabolic syndrome.

A further finding of this study was that there was no evidence that a larger proportion of the shared genetic variance between OSA and indices of adiposity existed for a general measure of body mass (BMI) as contrasted to measures of regional fat distribution (e.g., neck or waist circumferences). This suggests that, in our cohort, there are no important genes that influence OSA susceptibility through pathways that specifically influence body fat distribution (e.g., preferentially increasing neck circumference with little change in BMI or waist circumference). Rather it suggests obesity-related susceptibility genes for OSA increase apnea risk via global effects on obesity. This interpretation is further supported by the finding that after adjusting analyses for BMI, no significant genetic correlation persisted between AHI and the other adiposity measures, including measures of regional fat as well as adipokines. This would argue against an important genetic effect predisposing to OSA mediated through specific fat deposition patterns. However, since specific fat depots such as visceral fat or parapharyngeal fat were not directly measured, we cannot exclude the existence of apnea susceptibility polymorphisms that specifically influence one of these depots without affecting anthropomorphic measures.

It has been suggested that one mechanism by which obesity may influence OSA pathogenesis is through hormonal effects mediated by adipocytes. For example, leptin has been postulated to mediate OSA susceptibility via effects on ventilatory drive ²¹. Variability in adipokine levels has been found to closely aggregate within families and linkage studies have identified potential genetic risk factors regulating levels of each of the adipokines assessed in this study ^22–24. The significant estimates of heritability we calculated for levels of leptin, adiponectin and resistin also support an important genetic basis for these adipokines. We hypothesized that genetic susceptibility to OSA may be partly mediated through pathways that specifically influence levels of adipokines, such as leptin. However, our analyses did not support this hypothesis. In particular, the genetic correlations between adipokine levels and AHI were much smaller than those for anthropomorphic measures, and these correlations disappeared after adjustment for BMI. These results, therefore, do not support the presence of important genes influencing OSA susceptibility primarily via pathways that influence levels of these adipocyte-derived hormones independently of pathways regulating the overall level of BMI.

Study strengths include the availability of a number of indices of adiposity as well as measurement of OSA severity using standardized protocols in a large family-based cohort. Several limitations to this work should also be noted. Random errors due to imprecise measurement will inflate environmental variance and so lower observed heritabilities. Thus, differences in precision of the various adiposity measurements may explain some of the differences in the magnitude of heritability and genetic correlation to AHI. The only measure of OSA considered in this work was the AHI and it is possible that other metrics may have genetic bases that overlap with obesity in a different fashion. However, because AHI is the most commonly utilized measure of OSA in both clinical and research contexts and because our prior work has identified AHI as the OSA measure with the greatest heritability ²⁵, we focused our modeling on this metric. It should also be noted that the study population was selected for having a greater familial prevalence of OSA and so may potentially be unrepresentative of the general population. The fact that the heritabilities of AHI and obesity-related traits as well as the genetic correlations between the various adiposity measures are similar to values reported in other cohorts is reassuring in this regard.

In summary, our data confirm that OSA and obesity do share a substantial genetic basis but also underscore the potential importance of genetic factors unrelated to obesity that may determine OSA susceptibility. Our estimate that 35–40% of the genetic variance for OSA is shared with obesity suggests that in order to understand the genetic basis for OSA, it will be important to consider both obesity-related and obesity-independent genetic effects. Furthermore, our data also suggest the potential importance of including OSA-related traits in genetic epidemiological studies of obesity.