H3K23/H3K36 hypoacetylation and HDAC1 up-regulation are associated with adverse consequences in obstructive sleep apnea patients

The aim of this study is to determine the roles of global histone acetylation (Ac)/methylation (me), their modifying enzymes, and gene-specific histone enrichment in obstructive sleep apnea (OSA). Global histone modifications, and their modifying enzyme expressions were assessed in peripheral blood mononuclear cells from 56 patients with OSA and 16 matched subjects with primary snoring (PS). HIF-1α gene promoter-specific H3K36Ac enrichment was assessed in another cohort (28 OSA, 8 PS). Both global histone H3K23Ac and H3K36Ac expressions were decreased in OSA patients versus PS subjects. H3K23Ac expressions were further decreased in OSA patients with prevalent hypertension. HDAC1 expressions were higher in OSA patients, especially in those with excessive daytime sleepiness, and reduced after more than 6 months of continuous positive airway pressure treatment. H3K79me3 expression was increased in those with high C-reactive protein levels. Decreased KDM6B protein expressions were noted in those with a high hypoxic load, and associated with a higher risk for incident cardiovascular events or hypertension. HIF-1α gene promoter-specific H3K36Ac enrichment was decreased in OSA patients versus PS subjects. In vitro intermittent hypoxia with re-oxygenation stimuli resulted in HDAC1 over-expression and HIF-1α gene promoter-specific H3K36Ac under-expression, while HDAC1 inhibitor, SAHA, reversed oxidative stress through inhibiting NOX1. In conclusions, H3K23/H3K36 hypoacetylation is associated with the development of hypertension and disease severity in sleep-disordered breathing patients, probably through up-regulation of HDAC1, while H3K79 hypermethylation is associated with higher risk of cardiovascular diseases, probably through down-regulation of KDM6B.

Gene and protein expression changes of specific histone modifying enzymes responsible for the global histone methylation/acetylation patterns. To  www.nature.com/scientificreports/  www.nature.com/scientificreports/ in patients with OSA, we next investigated whether such hypoacetylation occurred specifically at the promoter regions of the HIF-1α and HIF-2α genes. We examined this possibility by ChIP method in PBMC lysates from another cohort (cohort 2) of 28 treatment-naïve OSA patients, and 8 matched PS subjects (Table 1). H3K36Ac enrichment over the HIF-1α (0.52 ± 0.92 versus 1.23 ± 1.25 fold change, adjusted p = 0.006; Fig. 2H) gene promoter region was decreased in the treatment-naïve OSA patients as compared with that in the PS subjects. This specific histone modification was negatively correlated with both AHI (r = − 0.548, p = 0.001, Fig. 2I) and ODI (r = − 0.449, p = 0.006, Fig. 2J), and positively with minimum SaO2 (r = 0.434, p = 0.008, Fig. 2K). These correlations remained statistically significant in multivariate linear regression model 1 or model 2 (Supplementary  Tables S5, S6, S7), but did not reach statistical significance in the analysis of separate PS and OSA groups. Additionally, H3K36 Ac enrichment over the HIF-2α gene promoter region (0.49 ± 0.3 versus 1.03 ± 1.06 fold change, adjusted p = 0.024, Supplementary Fig. S4) was decreased in the treatment-naïve OSA patients versus the PS subjects. However, this difference was not normally distributed and did not reach statistical significance in the non-parametric analysis. Moreover, no significant difference in either HIF-1α or HIF-2α gene expression level was noted between case and control groups ( Supplementary Fig. S4).

Discussion
Epigenetic changes are involved in the switching ' on' and ' off ' of many genes, and histone modifications has been regarded as epigenetic gene targets for the regulation of disease-associated cellular changes in hypoxic microenvironment. Little is known about the role of histone modifications in OSA, which is characterized by chronic IHR rather than persistent hypoxia. Therefore, in this study, we sought to identify specific histone modification patterns associated with OSA and its clinical phenotypes. We have proved, for the first time, the relationship between global H3K23/H3K36 hypoacetylation in association with HDAC1 up-regulation and the development of EDS/hypertension phenotype in OSA patients. Specifically, HIF-1α gene promoter-specific H3K36 enrichment was decreased in OSA patients and in response to in vitro IHR stimuli. In contrast, global H3K79 hypermethylation along with KDM6B down-regulation was found to be related to higher incident cardiovascular risk. Additionally, we found that SAHA, a HDAC inhibitor, could reverse IHR-induced ROS over-production through NOX1 in vitro. H3K36Ac is localized predominantly to the promoters of RNA polymerase II-transcribed genes, a pattern inversely related to that of H3K36 me 16 . H3K36 acetylation has been shown to increase chromatin accessibility, and promotes DNA double-strand break repair. Furthermore, H3K36 modification is cell cycle regulated with chromatin-bound H3K36me3 peaking in G1 where non-homologous end joining occurs, while H3K36Ac peaking in S/G2 phase when homologous recombination predominates 17 . The H3K36 hypoacetylation in OSA patients and in response to in vitro short-term IHR stimuli suggest that DNA double-strand break repair may be dampened by chronic IHR in OSA through H3K36 hypoacetylation. The decreased enrichment of H3K36 acetylation over the HIF-1α gene promoter region in OSA patients and in response to IHR stimuli suggest that H3K36Ac may play a specific role in regulating HIF-1α-mediated adaptive responses to hypoxia. However, separate cohorts were used for the global histone and HIF-1α promoter-specific histone enrichment analyses, and this may limit the cause-effect interpretation of the experiments. Moreover, correlations between global H3K23Ac/H3K36Ac expression and metrics such as AHI, arousal index, mean SaO2 were quite weak, so further studies are needed. Finally, many environmental exposures, such as cigarette smoking, may cause specific histone modification change. In the lung tissues of smokers with COPD, the expressions of HDAC2, HDAC5, HDAC8, HDAC7, and HDAC10 were decreased significantly, while H3K23Ac or H3K36Ac has not yet been linked to smoking exposure 18 . However, we applied multivariate linear regression models to adjust for all possible confounding factors.
In line with our findings that global H3K23 was hypoacetylated in OSA patients, H3K23 acetylation has been reported to be reduced in diabetic mice, and its hypo-acetylation contributed to Drosophila learning defect and depression-like behaviors in rats [19][20][21] . Since the co-existence of both H4K16ac and H3K23ac was correlated with higher gene expression levels in a plant study, and lipopolysaccharide can trigger IL-6 production via enhancing recruitment of H3K23 acetylation to IL-6 promoter region, we speculated that the simultaneous hypoacetylation of H3K36 and H3K23 in OSA patients may lead to impaired DNA repair ability under chronic IHR circumstances, and affect specific adaptive gene responses through regulating HIF-1α 22,23 .
Up-regulation of HDAC1 has been shown to be associated with adverse prognosis of various cancers, and plays critical roles in cellular senescence, aging, myelination, and adult neurogenesis 24 . In line with our findings that HDAC1 was up-regulated in OSA patients and in response to IHR stimuli, HDAC1 inhibition was reported to protect against hypoxia-induced swelling in cardiomyocytes through enhancing cell stiffness 25  www.nature.com/scientificreports/ conditions through binding to the oxygen-dependent degradation domain of HIF-1α 26 . Although HDAC1 expressions were reduced after CPAP treatment in a small group of OSA patients, the lack of a time control group or sham CPAP device and the lack of CPAP adherence data limit the interpretation of these data. In line with our findings that HDAC3 was down-regulated in OSA patients with high hypoxic load, HDAC3 has been reported to mediate cardioprotection of remifentanil post-conditioning by targeting GSK-3beta in cardiomyocytes in IHR injury 27,28 . In contrast, HDAC3 could contribute to IHR-induced cell apoptosis of myocardioblasts 28 . However, the reversion of IHR-induced ROS over-production with SAHA through inhibition of the NOX1 gene indicates that HDAC1/3 inhibition may be a new therapy to reduce oxidative stress in OSA. Several recent studies have shown that HDAC inhibition with SAHA, trichostatin A, or valproic acid, reduced ROS production in experimental diabetes, pulmonary hypertension, and atherosclerosis, through inhibitions of NOX1/2/4/5 expression via epigenetically regulating chromatin accessibility [29][30][31][32] . HDAC inhibitors have displayed clinical efficacy in treating specific tumors 33 . SAHA inhibits both class I and class II HDACs, but selectively alters transcription of as few as 2-5% of genes 34 . Further research is required to determine whether HDAC1 inhibition could be used to reduce cardiovascular risk in OSA patients. H3K79 methylation has been shown to be implicated in several processes of DNA replication and repair of DNA damage 35 . Moreover, H3K79 methylation, shows a high positive correlation with transcriptional activity 36,37 . In this study, the global H3K79 hypermethylation in OSA patients with high hsCRP, which has been linked with a high cardiovascular risk 15 , suggest that H3K79me3 may serve as an endogenous protective mechanism against chronic IHR-induced DNA damage in OSA. Most KDM enzymes are structurally similar to the HIF hydroxylase factor inhibiting HIF-1, suggesting that KDM enzymes may act as molecular oxygen sensors in the cell 38 . Hypoxia-inducible KDMs have been shown to support the adaptive and adverse gene transcription induced by HIF-1α, and they can also control genome-wide chromatin landscape, especially KDMs which demethylate H3K9 and H3K27 sites 39 . The down-regulation of KDM6B in OSA patients with high hypoxic load and incident cardiovascular events in this study suggest that long-term oxidative stress-related insults may impair the maintenance of chromatin landscape and provoke cellular senescence and endothelial dysfunction associated with chronic IHR circumstances in OSA. Although previous studies showed that hypoxia either inhibits or induces the enzymatic activity of KMD6B depending on HIF-2α, KDM6B protein levels remained unchanged under hypoxic conditions and decreased upon re-oxygenation [40][41][42] . Further study is required to determine whether KDM6B supplementation can reverse H3K79 hypermethylation and ameliorate adverse consequences in severe OSA. One limitation is that this study was conducted using PBMCs; thus, the findings associating KDM6B to future cardiovascular events may not extend to other cell types. Given that a recent U.S. healthcare system-based analysis of mortality and markers of severe morbidity identify sleep apnea as a risk factor for COVID-19 mortality, our results may provide particular insight into immune function with OSA and discover mechanistic pathways underlying COVID-19 morbidity 43 . Another limitation is that we did not check some histone modifying enzymes known to regulate histone methylation, such as other JmjC-KDM demethylases and G9a methylase 44,45 .
In conclusions, H3K23/H3K36 hypoacetylation is associated with the development of prevalent hypertension or disease severity in patients with sleep-disordered breathing, probably through up-regulation of HDAC1, while H3K79 hypermethylation is associated with higher risk of incident cardiovascular diseases, probably through down-regulation of KDM6B. HIF-1α and HIF-2α gene promoter-specific H3K36 hypoacetylation may play a role in regulating adaptive gene responses to chronic IHR injury of OSA. HDAC1 inhibition can reduce oxidative stress induced by IHR through inhibition of NOX1. These findings provide further insights into novel epigenetic mechanisms by which chronic IHR leads to adverse consequences in OSA, and opens the possibility of using HDAC1 inhibitor in preventing the morbidities and mortality.  (E) OSA patients with low KDM6B protein expression had a higher risk for recent-onset cardiovascular events or hypertension in the following 3 years than those with high KDM6B protein expression. (F) HDAC3 protein expression was decreased in OSA patients with high hypoxic load. (G) HDAC1 gene expression was increased in OSA patients with severe excessive daytime sleepiness (EDS). (H) H3K36Ac enrichment over the HIF-1α gene promoter region was decreased in OSA patients. This histone marker was negatively correlated with (I) apnea hypopnea index (AHI) and (J) oxygen desaturation index (ODI), and positively with (K) minimum oxygen saturation. *adjust for age, body mass index, gender, smoking history, alcoholism history, and co-morbidities (diabetes mellitus, hypertension, stroke, cardiac disease, chronic kidney disease) by multivariate linear regression analysis. #compared between primary snoring (PS) subjects and obstructive sleep apnea (OSA) patients with a particular phenotype (a high hypoxic load or severe EDS), adjusted by multivariate linear regression analysis. ##compared between OSA patients with a particular phenotype (a high hypoxic load or severe EDS) and those without the phenotype, adjusted by multivariate linear regression analysis. Hollow circle indicates PS subjects; solid star indicates OSA patients. PBMC = peripheral blood mononuclear cell; cum = cumulative; CI = confidence interval. www.nature.com/scientificreports/ www.nature.com/scientificreports/ was obtained from each subject participating in the study aged 20 years or older. All the methods were carried out in accordance with relevant guidelines and regulations. The exclusion criteria included ongoing infections, autoimmune disease, use of immunosuppressive agent in the past 6 months, narcolepsy, severe obesity (body mass index, BMI, ≧35 kg/m 2 ), old age (> 65 year-old), and those with a BMI < 21 kg/m 2 . OSA (apnea hypopnea index, AHI, ≧5 events/hour) and PS (AHI < 5 events/hour) were diagnosed by full night polysomnography examination at the sleep center of Kaohsiung Chang Gung Memorial Hospital 46 . The Epworth Sleepiness Scale (ESS) recorded at the examination was used to measure sleep propensity in every study subject 47 . Prevalent hypertension was defined as systolic blood pressure ≧140 mmHg and/or diastolic blood pressure ≧90 mmHg or medication use at diagnosis of OSA. Incident cardiovascular events (coronary artery disease: angioplasty, coronary artery bypass grafting, myocardial infarction, or angina) or hypertension were defined as those occurred in the following 3 years after the first blood sampling at diagnosis 48 .

Subjects. This study was approved by the Institutional Review
Isolation of PBMC RNA and protein from whole blood samples. PBMCs were isolated from heparinized blood of all study subjects using a two-layer Ficoll-Histopaque density gradient centrifugation (His- www.nature.com/scientificreports/ hexamers) were designed and purchased from Roche according to the company's protocols (www. roche-appli ed-scien ce. com), and their sequences are given in  52 . The PCR primers for the two selected gene promoters located from − 300 to + 30 bp of their transcription start site were listed in Table 2. The conditions for the reaction were: 50 °C, 2 min; 94 °C, 10 min, then 95 °C, 20 s; 60 °C, 30 s for 50 cycles. Resulting PCR products were measured and elaborated by the sequence detector ABI GeneAmp PCR System 9700 (PerkinElmer Biosystems). Histone enrichment was compared by fold enrichment method, in which the ChIP signals are divided by the no-antibody signals, representing the ChIP signal as the fold increase in signal relative to the background signal 53,54 . In vitro human monocyte cell line culture under IHR stimuli. The human monocytic leukemia cell line THP-1 cells obtained from ATCC (1 × 10 6 cells/ml) were re-suspended in a 5 cm culture dish containing 5 ml RPMI 1640 medium, and then exposed to normoxia or IHR in a custom-designed, incubation chambers which are attached to an external O2-CO2-N2 hand-driven controller as previously described 55 . Air-phase set point for IHR consisted of a 25-min hypoxic period (0% O2 and 5% CO2), followed by 35 min of re-oxygenation (21% O2 and 5% CO2), 7 h each day for 9 days 56 . Control cells were maintained in NOX condition for the same durations. HDAC inhibitor, SAHA (Sigma-Aldrich, Saint Luis, MO, USA), and HAT inhibitor, Garcinol (Sigma-Aldrich, Saint Luis, MO, USA), were used to inhibit or augment HDAC expression, respectively. H2DCFDA (catalog no. D6883; Sigma, USA) was used to measure THP-1 intracellular ROS. Cell-associated mean fluorescent intensity was measured by flow cytometry in FL1 channel at excitation and emission wavelengths of 488 and 535 nm, respectively, using the CytomicsTM FC500 (Beckman Coulter ; USA ). Gene expression levels of the NADPH oxidase NOX1/2/4/5 genes were determined using quantitative RT-PCR method.

Statistical analysis.
Continuous values were expressed as mean ± standard deviation (SD). The differences between two groups were analyzed using the Student's t-test, Mann-Whitney U-test, or X 2 -test, as appropriate. Subgroup comparisons of continuous variables in the validation cohort were performed using the one-way ANOVA with LSD method for post-hoc tests, followed by multivariate linear regression analysis to adjust for all potential confounding factors (age, BMI, gender, smoking, alcoholism, and co-morbidities), and to obtain adjusted p values. The null hypothesis was rejected at p < 0.05. All analyses were performed using SPSS software version 20.0 (SPSS Corp., Chicago).

Ethics approval and consent to participate. This study was approved by the Institutional Review
Board of Chang Gung Memorial Hospital, Taiwan (Certificate number: 102-3887B). Written informed consent was obtained from each subject participating in the study.

Data availability
The data that support the findings of this study are available in supplementary information online and from the corresponding author upon reasonable request.