Obstructive sleep apnea and hypertriglyceridaemia share common genetic background: Results of a twin study

Obstructive sleep apnea is associated with an increased risk of hypertension, diabetes and dyslipidaemia. Both obstructive sleep apnea and its comorbidities are at least partly heritable, suggesting a common genetic background. Our aim was to analyse the heritability of the relationship between obstructive sleep apnea and its comorbidities using a twin study. Forty‐seven monozygotic and 22 dizygotic adult twin pairs recruited from the Hungarian Twin Registry (mean age 51 ± 15 years) attended an overnight diagnostic sleep study. A medical history was taken, blood pressure was measured, and blood samples were taken for fasting glucose, total cholesterol, triglyceride, high‐density lipoprotein cholesterol, low‐density lipoprotein cholesterol and lipoprotein (a). To evaluate the heritability of obstructive sleep apnea and its comorbidities bivariate analysis was performed with an adjustment for age, gender, body mass index (BMI) and smoking after false discovery rate correction and following exclusion of patients on lipid‐lowering and antidiabetic medications. There was a significant correlation between indices of obstructive sleep apnea severity, such as the apnea–hypopnea index, oxygen desaturation index and percentage of sleep time spent with oxygen saturation below 90%, as well as blood pressure, serum triglyceride, lipoprotein (a) and glucose levels (all p < .05). The bivariate analysis revealed a common genetic background for the correlations between serum triglyceride and the oxygen desaturation index (r = .63, p = .03), as well as percentage of sleep time spent with oxygen saturation below 90% (r = .58, p = .03). None of the other correlations were significantly genetically or environmentally determined. This twin study demonstrates that the co‐occurrence of obstructive sleep apnea with hypertriglyceridaemia has a genetic influence and heritable factors play an important role in the pathogenesis of dyslipidaemia in obstructive sleep apnea.


| INTRODUC TI ON
Obstructive sleep apnea (OSA) is the most common sleep-related breathing disorder and is characterized by repetitive episodes of complete or partial collapse of the upper airways. Recurrent obstructive respiratory events cause sleep fragmentation, chronic intermittent hypoxaemia (CIH) and sympathetic bursts, which play a cardinal role in the development of cardiovascular and metabolic comorbidities, such as hypertension, diabetes and dyslipidaemia (Framnes & Arble, 2018).
Apart from the direct causal relationship, co-occurrence of diseases such as OSA and its comorbidities could be due to common genetic loci, non-overlapping genetic factors (i.e., simultaneous occurrence of two highly heritable treats) or a common environmental driving factor (i.e., diet). Twin studies can explore the possible role of heritable and environmental determinants of diseases. Such studies have demonstrated that heritable factors determine blood pressure (Cui, Hopper, & Harrap, 2002;Jermendy et al., 2011), fasting serum glucose (Jermendy et al., 2011) and lipid levels (Jermendy et al., 2011). There is growing evidence that OSA and sleep architecture are heritable as well (Guilleminault, Partinen, Hollman, Powell, & Stoohs, 1995;Redline et al., 1995;Szily et al., 2019). Of note, obesity, which is a common aetiological factor in hypertension, diabetes, dyslipidaemia and OSA, is itself genetically determined with an almost 40%-80% heritability (Tarnoki et al., 2014).
We hypothesized that the relationship between OSA and its comorbidities is at least partly determined by genetic factors. We analysed the results of a Hungarian twin study with an aim to determine the influence of heritable and environmental factors on this relationship. We were also able to distinguish overlapping (common genes) and non-overlapping (unshared genes) genetic correlations between the two traits.

| Subjects and study design
We studied the results of 69 Caucasian twin pairs (51.3 ± 15.1 years, 29% male) of the Hungarian Twin Registry (Littvay, Metneki, Tarnoki, & Tarnoki, 2013) who participated in the Hungarian Twin Sleep Study, which had a primary objective to investigate the heritability of OSA using objective tests (Szily et al., 2019).
During the study, twins attended the Sleep Unit of the Department of Pulmonology, Semmelweis University. In the evening, a medical history was taken and patients filled out the Epworth Sleepiness Scale (ESS), which was followed by an inpatient polysomnography on the same night for both twins. As optional tests in the morning, blood pressure was measured in 57 twin pairs and fasting serum glucose (n = 48 pairs) and lipid profile (n = 64 pairs), including total cholesterol, triglyceride, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C) and lipoprotein (a), were also determined. Data on diet were available in 86 subjects and alcohol consumption was known in 78 volunteers. Fifty-three participants regularly consumed meat, 77 regularly ate vegetables and 48 had regular alcohol consumption (n = 7, daily alcohol intake). A ketogenic diet was not reported by any volunteer.

The study has been approved by the Semmelweis University
Ethics Committee (TUKEB 30/2014). All participants gave their informed consent.

| Sleep studies
Inpatient sleep studies were performed with the Somnoscreen Plus Tele PSG device (Somnomedics GmBH) between 22:00 and 06:00 hours. Electroencephalogram, electrooculogram, electrocardiogram, electromyogram, oxygen saturation, nasal flow, thoracic and abdominal movements and body position were recorded. Sleep stages, arousals and respiratory events were manually scored according to the American Academy of Sleep Medicine recommendations (Berry et al., 2012). Apnea was defined as a 90% airflow decrease, which lasted for more than 10 s, and hypopnea was defined as at least 30% airflow decrease lasting for at least 10 s, which related to a ≥3% oxygen desaturation or an arousal. The apnea-hypopnea index (AHI), the oxygen desaturation index (ODI) and percentage of sleep time spent with oxygen saturation below 90% (TST90%) were calculated. OSA was defined by an AHI ≥5/hr.

| Statistical analysis
The normality of data has been assessed with the Kolmogorov-Smirnov test. Monozygotic (MZ) and dizygotic (DZ) twins were compared with the t test, Mann-Whitney test and chi-squared test.
Indices of interest, including AHI, ODI and TST90%, as well as ESS, were log-transformed and linear regression analysis was carried out with serum glucose, serum lipids and lipoproteins and morning blood pressures. In subsequent steps, the heritability was estimated only for the associations with a p < .1.
A descriptive estimate of the genetic influence was calculated using the bivariate co-twin correlation (Pearson's test) in MZ (rMZ) and DZ (rDZ) pairs. If the within-pair similarity for a phenotype is greater in MZ than DZ pairs this provides evidence for a genetic influence. In contrast, higher rDZ than rMZ demonstrates an environmental influence.
A bivariate Cholesky decomposition was carried out to derive the magnitude of covariation between the investigated phenotypes of interest and to estimate what proportion of this correlation is attributable to common underlying genetic and environmental factors (Neale & Maes, 2004). In order to estimate the amount of overlap between genes or the environment that influences the two parameters, genetic and environmental twin correlations between those phenotypes were calculated using SOLAR Eclipse version 8.1.1 (http://www.solar-eclip se-genet ics.org/). For RhoG (genetic correlation), two separate hypotheses were tested: (a) RhoG = 0 (the test for overlapping genetic correlations) or (b) RhoG = 1 or −1 (the test for non-overlapping positive or negative genetic correlations). If a p value for RhoG = 0 is significant (p < .05), then there are shared loci between the two traits. Significance of RhoG = 1 or −1 (without RhoG = 0 being significant) is suggestive of a high heritability for each of the two traits but determined by different genes. The RhoE (environmental correlation) and RhoG values were computed and reported along with their estimated standard errors.
The -testrhoE option tested the significance of the rhoE difference from zero. The -testrhoG option tested the significance of the rhoG parameter from zero and also from either 1 or −1 (depending on whether rhoG is negative or positive, and not exactly 1 or −1 already). The latter test is a test for pleiotropy. These analyses were adjusted for age, gender, BMI and smoking status (ever versus never smokers). To eliminate the modifying effects of medications on glucose and lipid levels we performed the analyses also after excluding subjects taking these drugs. To adjust for multiple comparisons, a false discovery rate (FDR) correction was performed. A p value <.05 was considered significant.

| Study participants
Characteristics of the participants are summarized in Table 1.
Comparing MZ to DZ twins, significant differences were observed in age, gender, BMI, morning systolic and diastolic blood pressure, lipoprotein (a) levels, TST90% and ESS (all p < .05). There was no difference in the prevalence of smoking, OSA, hypertension or dyslipidaemia; however, the prevalence of diabetes tended to be higher in DZ twins (p = .05). OSA was diagnosed in 58 participants (42 mild, AHI 5-14.9/hr; 12 moderate, AHI 15-29.9/hr; four severe, AHI ≥30/hr).

| Bivariate co-twin correlation
Bivariate co-twin correlation of blood pressure, glucose, lipid and sleep parameters was higher among MZ than DZ twins. In MZ twins, all correlations were statistically significant (p < .01) and nominally higher than in DZ twins. Among DZ twins there were significant correlations in systolic blood pressure (SBP) (p < .01), LDL-C (p = .02) and triglyceride (p = .01) levels (Table 2).

| Heritability of associations between indices of OSA severity, blood pressure, glucose and lipids
The significant associations of the bivariate analysis in all subjects adjusted for age, gender and BMI are shown in Table 3.
Genetic correlations for the associations between AHI and triglyceride, ODI and triglyceride, TST90% and triglyceride were significant (both p-values for RhoG were <.05). Significant RhoG = 1 or −1 with non-significant RhoG = 0 means that although both traits were significantly heritable, they did not share common genetic loci. None of the relationships were environmentally determined (p > .05). However, when the results were adjusted for multiple comparisons (FDR correction) only the genetic correlation between TST90% and serum triglyceride levels remained significant. Adjustment for smoking did not alter significance of this relationship (both p-values for RhoG were <.05).
After taking into consideration the effects of antidiabetic and lipid-lowering drugs, we found that heritable relationships between AHI and triglyceride, ODI and triglyceride and TST90% and triglyceride remained significant. The relationships between TST90% and triglyceride (p = .04), as well as TST90% and LDL-C (p = .01), were influenced significantly by the environment (Table 4). When performing FDR correction, the heritable relationships between ODI and triglyceride levels, as well as TST90% and triglyceride levels, were still significant (both p-values for RhoG were <.05). However, the genetic association between AHI, triglyceride and the environmental correlations became insignificant (p > .05). When this model has been adjusted for smoking, associations for non-overlapping genetic correlations became insignificant between serum triglyceride and ODI (p = .09 for RhoG different from 1 or −1), as well as TST90% (p = .08). p-values for RhoG different from 0 remained significant (both p = .02), suggestive of common genetic loci between markers of overnight hypoxaemia and serum triglyceride levels. Moderate OSA (%) 9 9 9 Severe OSA (%) 3 3 2 Note: Data are shown as mean ± standard deviation or median/25%-75% percentile/. MZ and DZ groups were compared with the t test and Mann-Whitney and chi-squared tests. Significant differences between the MZ and DZ groups are indicated in bold.
TA B L E 1 Comparison of monozygotic (MZ) and dizygotic (DZ) twins

| D ISCUSS I ON
This is the first study estimating the contribution of heritable and environmental factors to the associations between indices of OSA severity and measures of common comorbidities associated with OSA.
We found that common genetic factors significantly determined the relationship between indices of chronic intermittent hypoxia and serum triglyceride levels.
Genetic susceptibility is known to contribute to the development of OSA (Guilleminault et al., 1995;Palmer et al., 2003;Redline et al., 1995), hypertension (Cui et al., 2002;Jermendy et al., 2011), diabetes (Lehtovirta et al., 2010;Newman et al., 1987) and dyslipidaemia (Jermendy et al., 2011). Not surprisingly, the relationship within the monozygotic twin pairs for respective measures was stronger than in dizygotic twins in our study. Previously, we published the results of an ACE model dividing common genetic, shared and unshared environmental components and reported around 70% heritability of OSA parameters in the same cohort (Szily et al., 2019). In the current manuscript, we expanded the data on blood pressure, serum glucose and lipid measurements, reporting on their heritability; however, due to a lower number of subjects, no ACE modelling has been performed. High heritability of AHI, ODI, TST90%, blood pressures and metabolic parameters are the reason why RhoG = 1 or −1 was significant for some associations (Tables   3 and 4). Previous studies estimated that heritability contributes to development of hypertension in 46% (Cui et al., 2002), of T2DM in 50%-92% (Lehtovirta et al., 2010;Newman et al., 1987) and of dyslipidaemia in 25%-60% (Jermendy et al., 2011).
As expected, the indices of OSA severity correlated with blood pressure values, serum glucose and triglyceride levels. Interestingly, the correlation with cholesterol levels was significant only for ODI and not AHI. This is in line with a previous study that the stratification of OSA severity based on ODI is more strongly associated with lipid abnormalities than the AHI-based grouping (Tisko et al., 2014). Supporting this, after adjustment for multiple comparison (FDR correction) only overlapping genetic correlations between markers of CIH and serum triglyceride levels remained significant.
It is known that fragmented sleep, chronic intermittent hypoxia and increased sympathetic tone can all contribute to dyslipidaemia.
Potential mechanisms include increased food and calorie intake due to hormonal changes, increased lipolysis and increased hepatic produc- Several genetic loci have been identified that may play an important role in the pathogenesis of OSA and related dyslipidaemia.
Transcriptional factor PPAR-gamma has antioxidant and anti-inflammatory effects and plays a key role in glucose and lipid metabolism (Janani & Ranjitha Kumari, 2015), contributing to the development of metabolic syndrome. In high-fat diet-induced obese mice exposed

F I G U R E 1
Relationship between serum triglyceride levels and indices of obstructive sleep apnea (OSA) severity. There were significant associations between serum triglyceride levels and the apnea-hypopnea index (AHI; p < .001, r = .357, Panel a), the oxygen desaturation index (ODI; p < .001, r = .448, Panel b) and percentage of sleep time spent with oxygen saturation below 90% (TST90%; p < .001, r = .390, Panel c). Respiratory indices were log-transformed to chronic intermittent hypoxia, the expression of PPAR-gamma was found to be decreased (Zhang, Zhang, Li, & Hou, 2017). In humans, polymorphism Pro12Ala of the PPAR-gamma gene was associated with susceptibility to OSA (Sun et al., 2015) and increased risk of hyperlipidaemia in OSA patients in a Chinese population (Chen et al., 2017). Polymorphism of APOE E2 and E4 alleles are linked to higher LDL-cholesterol and lower HDL-cholesterol levels (Kataoka et al., 1996). These APOE alleles may be associated with OSA as well (Uyrum et al., 2015;Xu et al., 2015) and may predispose patients to OSA-related comorbidities . Intermittent hypoxia can activate the hepatic sterol regulatory element-binding protein-1 (SREBP-1) through the hypoxia inducible factor-1 (HIF-1), inducing triglyceride production, while not affecting the HMG CoAreductase, the enzyme essential for cholesterol synthesis (Li et al., 2006). Theoretically, genotypic variability of HIF-1 and SREBP-1 may therefore be related to the probability of dyslipidaemia. In line with this, different apolipoprotein E genotypes were associated with altered triglyceride and HDL-C response to the same level of overnight hypoxaemia in OSA (Tisko et al., 2014). This suggests that genetic susceptibility may determine whether people with OSA develop hypertriglyceridaemia.
In contrast to triglycerides, there was no genetic predisposition for the relationship between OSA parameters, blood pressures and serum glucose values. The studied participants had relatively mild (if any) OSA, and blood pressure values, as well as blood glucose levels, in patients with hypertension and diabetes were well controlled with medications. Of note, excluding patients on antidiabetic drugs did not alter the relationship. We did not exclude patients taking antihypertensive medications, as they comprised a significant (30%) proportion of the whole study population and analyses on a smaller subgroup would have compromised the robustness of our results. It is also possible that common shared (i.e., diet and exercise), rather than genetic, factors contribute to the relationship between OSA, hypertension and diabetes. Nevertheless, the heritability of the relationship between OSA, hypertension and diabetes must be investigated in a larger study enrolling medication-naïve participants.
OSA is commonly accompanied by metabolic syndrome; however, the association between OSA and its comorbidities is present irrespective of obesity (Gunduz et al., 2018;Lavie et al., 2000;Nagayoshi et al., 2016). Most importantly, the heritability of OSA (Palmer et al., 2003) and hypertension (Cui et al., 2002) was only mildly influenced by the adjustment for BMI in the previous twin studies. In addition, the association between OSA and metabolic syndrome is affected by age and gender (Sjostrom et al., 2002). The strength of our manuscript is that our analyses were adjusted for age, gender and BMI. However, it has been noted that only 29% of the population were male; therefore adjustment on gender has to be interpreted carefully. Although an adjustment has been made for BMI, central obesity (hip-waist ratio) may be a better predictor of metabolic outcomes. Unfortunately, these results were not recorded in this study.
Our study has limitations. The sample size was relatively low; therefore, the lack of associations needs to be interpreted carefully.
The study was originally powered to identify at least 40% heritability of obstructive respiratory events (Szily et al., 2019). We found an even higher additive genetic component (around 70%); however, the proportion of common and residual environmental factors cannot precisely be estimated due to the relatively low ratio of DZ twins. In bivariate twin studies investigating genetic correlation, such as in the current report, the power is primarily determined by the heritability of each analysed trait and the strength of the correlation (RhoG). As obstructive respiratory events (Szily et al., 2019) and triglycerides (Jermendy et al., 2011) are highly heritable (70% and 40%) and the RhoG is .6, the genetic correlation between markers of CIH and serum triglycerides can reliably be concluded (Verhulst, 2017). The additive genetic component for blood pressures (Jermendy et al., 2011) is even higher (around 60%) than for serum triglycerides. Therefore, it is unlikely that the lack of significant genetic correlation between hypertension and OSA is due to the low numbers. However, increasing the sample size could potentially reveal further genetic correlations for less heritable variables. As discussed above, the ratio of MZ:DZ twins could potentially contribute to underestimation of environmental effects. Therefore, lack of environmental correlations must be interpreted carefully. Moreover, most of the subjects had mild, if any, OSA, which could have hindered potential more complex relationships.
A wider range of AHI, ODI and TST90% values could potentially increase RhoG for some bivariate correlations, contributing to higher heritability. Notably, the studied population was unselected for OSA, and the prevalence of any OSA (42%) and moderate to severe OSA (12%) is within the range of the estimated prevalence of OSA in the general adult population. We found that smoking was a potential confounder for the genetic correlations.
However, this has to be interpreted carefully, as only 11% of the studied population were ever-smokers with a negligible smoking history. Finally, although fasting blood was taken for glucose and lipid measurements, we did not control for diet or regular exercise, which could have contributed to the results significantly. Although it is not routinely recommended to fast before lipid profile measurement, diet essentially contributes to the metabolic profile (Nordestgaard et al., 2016). The information on diet was retrospectively collected and was available in a limited number of subjects. Therefore, we decided against adjusting on this factor. Diet would most likely contribute to the environmental effect, which was underpowered, as discussed above. All these limitations need to be addressed in further studies. Nevertheless, we believe that our pilot data will still serve as an essential basis on which to design large twin studies analysing OSA and cardiometabolic disease.
In summary, our study suggests that OSA and hypertriglyceridaemia have a common genetic background. This further strengthens the concept of screening for OSA in those suffering from dyslipidaemia and screening for dyslipidaemia in patients with OSA.

CO N FLI C T O F I NTE R E S T
No conflicts of interest declared.

AUTH O R CO NTR I B UTI O N S
AB, ADT, DLT, JV, VM and LK conceived the study and participated in study design. DTK and BF organized the twin cohort. AB and LK performed sleep studies and analysed data. MM, JL and JS performed statistical analyses. The manuscript was drafted by MM and AB and was critically reviewed and approved by all authors.