Clock gene expression is altered in veterans with sleep apnea

Abstract

Clock gene dysregulation has been shown to underlie various sleep disorders and may lead to negative cardio-metabolic outcomes. However, the association between sleep apnea (SA) and core clock gene expression is unclear. We performed a cross-sectional analysis of 49 Veterans enrolled in a study of SA outcomes in veterans with chronic kidney disease, not selected for SA or sleep complaints. All participants underwent full polysomnography and next morning whole blood collection for clock gene expression. We defined SA as an apnea-hypopnea index ≥15 events/h; nocturnal hypoxemia(NH) was defined as ≥10% of total sleep time spent at <90% oxygen saturation. We used quantitative real-time PCR to compare the relative gene expression of clock genes between those with and without SA or NH. Clock genes studied were Bmal1, Ck1δ, Ck1ε, Clock, Cry1, Cry2, NPAS2, Per1, Per2, Per3, Rev-Erb-α, RORα, and Timeless. Our cohort was 90% male, mean age was 71 yr (SD 11), mean body mass index was 30 kg/m² (SD 5); 41% had SA, and 27% had NH. Compared with those without SA, Per3 expression was reduced by 35% in SA (P = 0.027). Compared with those without NH, NPAS2, Per1, and Rev-Erb-α expression was reduced in NH (50.4%, P = 0.027; 28.7%, P = 0.014; 31%, P = 0.040, respectively). There was no statistical difference in expression of the remaining clock genes by SA or NH status. Our findings suggest that SA or related NH and clock gene expression may be interrelated. Future study of 24 h clock gene expression in SA is needed to establish the role of clock gene regulation on the pathway between SA and cardio-metabolic outcomes.

INTRODUCTION

Sleep apnea (SA) is a disorder characterized by dysregulated breathing during sleep resulting in recurrent hypoxia and arousal (1). SA is common in the United States with one in four men and one in 10 women having mild or worse SA (18). Though older age, obesity, and male sex are risk factors, SA, independently of these factors, has been shown to lead to metabolic dysregulation and increased risk of hypertension, cardiovascular disease and death (9, 17, 18). The mechanisms for such metabolic dysregulation and cardiovascular risk in SA are complex but may, in part, be due to altered circadian regulation of normal homeostatic mechanisms resulting from SA (12, 13, 16).

Human biological functions are rhythmically regulated by a genetically encoded molecular clock located in nearly every cell in the body and regulated by a central or “master” regulator located in the suprachiasmatic nucleus of the brain (11). This system maintains temporal homeostasis of bodily function via modulation of gene expression at the molecular level in peripheral tissues (11). In recent years, investigators have identified the key clock genes at work centrally and peripherally and have gained insight into the control mechanisms for the expression and suppression of these genes in a cyclical pattern to generate the 24 h circadian cycle (6, 11). This ability to synchronize with and anticipate regular environmental changes allows the body to adapt in ways that influence survival and health (11). Conversely, polymorphisms and altered expression of clock genes have been associated with negative health outcomes including hypertension and cardiovascular disease (8, 16).

While several sleep disorders have been associated with clock gene dysregulation (as cause or consequence), the association of SA and clock gene expression has not been well studied (2). Literature to date has been limited to very small sample sizes and selected cohorts and have produced discordant findings (3, 10). In addition, studies to date have focused on only one or a few components of the array of clock genes that are fundamental to generation of the 24 h circadian loop (3, 10). Therefore, in a cohort of veterans unselected for sleep disorders, we sought to compare the relative expression of several key clock genes among those with and without SA or related nocturnal hypoxemia (NH).

METHODS

For this analysis, we performed a cross-sectional study of a subset of enrollees in the Sleep and Nephrology Outcomes Research (SNORE) Study. The SNORE Study is an ongoing prospective cohort study of sleep in veterans with chronic kidney disease (CKD). Our protocol was approved by the University of Florida Institutional Review Board and by the Malcom Randall Veterans Affairs Medical Center (VAMC) Research and Development Committee.

Clock Gene Sub-study Participants and Recruitment

SNORE Study enrollees were recruited from the North Florida/South Georgia Veterans Health System. In brief, participants were selected based upon age 18–89 yr and a Modification of Diet in Renal Disease (MDRD) estimated glomerular filtration rate (eGFR) between 15 and 44 ml·min⁻¹·1.73 m⁻² within 1 yr before the query date. Veterans were excluded if they were already on treatment for SA or on nocturnal oxygen therapy, had active cancer (except for prostate cancer or nonmelanoma skin cancer), were on dialysis or with kidney or other solid organ transplant, or had life expectancy <3 yr. Eligible veterans were randomly ordered and recruited via opt-out letter mailings in conjunction with primary care provider approval for contact. Enrollment for the SNORE Study occurred over a 2 yr period between March 2013 and April 2015. For the Clock Gene Sub-study, consent for whole blood samples was obtained starting with consecutive enrollments beginning in December 2015, and only those SNORE participants who had morning whole blood samples obtained were included in this analysis (n = 49).

Overview of Study Visit Procedures

Participants presented to the Malcom Randall VAMC on the afternoon of their study visit day. After informed consent was obtained, the study visit included measurement of blood pressure, heart rate, and anthropometrics. Thereafter, participants completed questionnaires about their medical history, daytime sleepiness, and quality of life. Next, participants underwent sleep study set-up and proceeded to a nearby hotel (5 min shuttle ride to our facility) to undergo unattended polysomnography. The next morning, they returned to the Malcom Randall VAMC to return equipment and undergo collection of serum and urine, and, for those in the Clock Gene Sub-study, whole blood for gene expression. This concluded their study visit.

Data Collected

Polysomnography and definition of key sleep variables.

Trained study personnel attached leads for unattended 25-channel polysomnography recording system to monitor sleep stage, breathing, body movements, oxygen saturation, electrocardiography, leg movements, and snoring (AURA PSG System; Grass Technologies, West Warwick, RI). Sleep studies were staged and respiratory events scored by a single American Board of Sleep Medicine-certified, registered sleep technician. Scoring rules for respiratory events were according to the American Academy of Sleep Medicine (AASM) Scoring Manual v 2.0, 2012. Apneas were scored if there was cessation of airflow lasting >10 s. Hypopneas were scored if there was a clear reduction in airflow for >10 s accompanied by ≥4% oxygen desaturation. The apnea-hypopnea index (AHI) was defined as the average number of apneas and hypopneas per hour of sleep. We categorized severity of sleep apnea according to AASM guidelines with normal: 0–4.9 events/h, mild: 5–14.9 events/h, moderate: 15–29.9 events/h, and severe: ≥30 events/h (1). NH was defined as spending ≥10% of total sleep time at < 90% oxygen saturation per previously published literature in older men (5).

Clock gene expression.

Studies have suggested that clock gene expression in human peripheral whole blood monocytes can be used as indicators of central circadian rhythm changes (4, 15). We collected early morning samples of whole peripheral blood at the Malcom Randall VAMC Clinical Laboratory in PAXGene Blood RNA tubes (2.5 ml). Samples were immediately transported to the University of Florida where they were processed, and RNA was isolated with the PAXGene Blood RNA kit per manufacturer’s instructions (Qiagen). RNA was then used as a template for reverse transcription and converted to cDNA using the High-Capacity cDNA Reverse Transfection kit (Applied Biosystems, Foster City, CA). This cDNA was used for relative quantification of clock gene expression by using quantitative real-time PCR (RT-PCR) according to the 2^-ΔΔCT method of Livak and Schmittgen (7). In brief, this method utilizes RT-PCR to determine fold change in mRNA expression of the target clock gene relative to a within-sample reference (β-actin in our study) (7). For each RT-PCR reaction, we used 10 µl TaqMan Fast Advanced Master Mix, 1 µl respective TaqMan Assay, 6 µl nuclease-free water, and 3 µl cDNA (reagents from ThermoFisher Scientific, Waltham, MA). All reactions were performed in duplicate, and we used average transcript expression values (C_T) for analyses. Target clock genes for this analysis were Bmal1, Ck1δ, Ck1ε, Clock, Cry1, Cry2, NPAS2, Per1, Per2, Per3, Rev-Erb-α, RORα, and Timeless based upon what genes have been elucidated to play a fundamental role in circadian control of key metabolic processes as per human and animal studies (11).

Other measurements.

At baseline, participants underwent measurement in triplicate of blood pressure, heart rate, height, weight, and waist, neck, and hip circumference. Subjects also completed a medical history questionnaire in which they were queried about self-reported race, sex, age, marital status, income, education, employment, tobacco use, alcohol use, physical activity, sleep disorders, medication use including dosages and comorbid conditions. Comorbid conditions queried and defined by self-report include cardiovascular disease, congestive heart failure, and peripheral vascular disease. Hypertension and diabetes mellitus were more specifically defined as self-report and/or use of respective related medications (antihypertensive or antidiabetic). Each participant also completed the Kidney Disease Quality of Life-Short Form to assess quality of life. We assessed daytime sleepiness with the Epworth Sleepiness Scale; a score ≥10 suggests excessive daytime sleepiness. We also estimated renal function with serum creatinine measured from freshly collected serum samples using the Cobas analyzer, enzymatic methods (Roche Diagnostics, Indianapolis, IN). This assay was calibrated to be traceable to the primary reference material at the National Institute of Standards (IDMS-traceable). We estimated glomerular filtration rate with the MDRD equation. Urinary albumin excretion was measured using the spot urine microalbumin and the spot urine creatinine. Urine microalbumin was measured using the Cobas analyzer (both analyzer by Roche Diagnostics). Urine creatinine was measured using the same assay as for the serum creatinine. We defined albuminuria as urinary albumin-to-creatinine ratio of ≥30 mg/g Cr.

Statistical Analysis

We compared baseline characteristics by category of SA by ANOVA or χ²-tests for continuous and categorical variables, respectively. We utilized Kruskal-Wallis test for skewed continuous data or Fisher’s exact test for categorical variables with small cell sizes. For each sample of whole blood, we examined the C_T data for outliers of our reference, β-actin, and excluded those data points where β-actin was ≥23, indicating unreliable cDNA (n = 7). Furthermore, within each clock gene analysis, we examined C_T data for outliers of each clock gene and excluded outliers by Grubb’s outlier test (if n < 25) or Rosner’s extreme studentized deviate test (if n > 25) using P < 0.05 to identify outliers.

We analyzed data by comparing the mean fold change of transcript between disease (i.e., SA) and comparison group (i.e., no SA) using the t-test, where SA was a priori defined as AHI ≥ 15. We then repeated these analysis alternatively defining SA as AHI ≥ 5 and AHI ≥ 30. Also, for clock genes exhibiting statistically significant changes (P < 0.05) in expression in the presence of SA, we further examined expression by severity of SA, defining SA in four categories of none (AHI 0–4.9), mild (AHI 5–14.9), moderate (AHI 15–29.9), and severe (AHI ≥ 30) to query a dose response with test for linear trend with ANOVA. We considered renal function (MDRD eGFR or urinary albumin excretion as continuous variable) and any comorbid conditions listed in Table 1 that was associated with SA or NH (i.e., differentially distributed by our predictor) and clock gene expression for adjustment as a confounder of this association. We then repeated the above analysis comparing relative gene expression between those with and without NH (% total sleep time < 90% ≥ 10%). Finally, in secondary analyses, we examined the correlation between other sleep parameters that may be related to circadian expression of clock genes and individual clock gene expression. Specifically, we determined the Pearson correlation between each of total sleep time (minutes), total sleep time spent in REM sleep (minutes), time spent in slow wave sleep (minutes), and sleep efficiency (total sleep time divided by total bed time, %), and clock gene expression normalized to β-actin expression with adjustment for multiple comparisons.

Table 1.

Baseline characteristics by presence or absence of disease (sleep apnea or nocturnal hypoxemia)

	No Sleep Apnea (n = 29)	Sleep Apnea (n = 20)	P	No Hypoxemia (n = 36)	Hypoxemia (n = 13)	P
Demographics and Renal Function
Age, yr, mean (SD)	70.1 (10.4)	72.2 (11.6)	0.5	70.6 (11.2)	71.8 (10.0)	0.73
Male sex, n (%)	25 (86)	19 (95)	0.64	32 (89)	12 (92)	1.0
Caucasian, n (%)	23 (79)	12 (60)	0.15	23 (64)	12 (92)	0.03
Body mass index, kg/m2, mean (SD)	29.9 (4.9)	30.2 (4.6)	0.86	29.9 (4.8)	30.5 (4.5)	0.71
eGFR, ml·min−1·1.73 m−2, mean (SD)	38.1 (9.3)	38.3 (7.6)	0.95	38.0 (8.7)	38.6 (8.6)	0.81
Albuminuria, n (%)	10 (34)	13 (65)	0.05	19 (53)	4 (31)	0.17
Sleep Study Data
Apnea-hypopnea index, median [IQR]	5.8 [2.0–8.4]	32.2 [17.7–48.6]	<0.01	7.6 [2.5–15.4]	31.9 [14.4–46.2]	0.003
No SA (AHI < 5), n (%)	14 (48)		N/A	13 (36)	1 (8)	0.03
Mild SA (AHI 5–14.9), n (%)	15 (52)			12 (33)	3 (23)
Moderate SA (AHI 15–29.9), n (%)		8 (40)		6 (17)	2 (15)
Severe SA (AHI ≥ 30), n (%)		12 (60)		5 (14)	7 (54)
% Total sleep time $S{a}_{O_2}$ < 90%, median, [IQR]	0.7 [2.0–8.4]	7.7 [2.7–31.4]	<0.01	1.3 [0–2.8]	25.4 [16.7–40.0]	<0.001
≥10% Total sleep time $S{a}_{O_2}$ < 90% n (%)	4 (14)	9 (45)	0.02			N/A
Total sleep time, minutes, mean (SD)	356.5 (122.4)	360.8 (87.5)	0.89	359.4 (110.8)	355.1 (342.0)	0.90
Total bed time, minutes, mean (SD)	511.6 (109.6)	531.1 (92.1)	0.52	521.9 (99.4)	512.9 (113.7)	0.79
Sleep efficiency (%), mean (SD)	69.4 (17.3)	68.3 (14.8)	0.82	68.8 (16.8)	69.5 (14.8)	0.89
Lights out time, median [IQR]	21:41 [21:00–22:06]	21:05 [19:55–22:11]	0.25	21:25 [20:57–22:06]	20:40 [19:34–22:26]	0.44
Lights on time, median [IQR]	05:51 [04:57–07:00]	06:26 [05:35–07:15]	0.52	05:57 [05:12–07:05]	06:30 [05:04–07:04]	0.87
Comorbidity Data
Self-reported health status, excellent or very good, n (%)	7 (24)	3 (15)	0.44	8 (22)	2 (15)	0.60
Excessive daytime sleepiness,* n (%)	7 (24)	8 (40)	0.24	12 (33)	3 (2)	0.49
Diabetes mellitus, n (%)	12 (41)	13 (65)	0.15	18 (50)	7 (54)	0.81
Hypertension, n (%)	27 (93)	19 (95)	1.00	33 (92)	13 (100)	0.56
Peripheral vascular disease, n (%)	3 (10)	1 (5)	0.64	3 (8)	1 (8)	1.00
Congestive heart failure, n (%)	3 (10)	2 (10)	1.00	2 (6)	3 (23)	0.11
Cardiovascular Disease, n (%)	9 (31)	4 (20)	0.52	8 (22)	5 (39)	0.29

eGFR, estimated glomerular filtration rate; SA, sleep apnea; AHI, apnea-hypopnea index; $S{a}_{O_2}$, oxygen saturation; IQR, interquartile range.

^*Defined as Epworth Sleepiness Scale Score >10.

RESULTS

Cohort Characteristics

The final 49 consecutive SNORE Study enrollments comprised the Clock Gene Sub-study Cohort, which was similar to the overall SNORE cohort (data not shown). Mean age of our cohort was 70.9 yr (SD 10.8), 90% were male, and 71% identified themselves as Caucasian. The mean body mass index of our cohort was 30.0 kg/m² (SD 4.7). The average MDRD eGFR was 38.2 ml·min⁻¹·1.73 m⁻² (SD 8.6), and 47% had albuminuria. The prevalence of SA defined as AHI ≥15 was 41%, and 29, 31, 16, and 25% had none, mild, moderate, and severe SA, respectively. Twenty-seven percent of our enrollees had NH. Participants underwent AM blood draw for clock gene expression at median time [interquartile range] of 09:02 [08:04–09:41]. Relative to lights on time, AM blood draw was a median [interquartile range] of 2.75 [2.16–3.58] hours from waking. Those with and without SA were similar across demographics, renal function, and comorbid conditions except for a greater prevalence of albuminuria and NH among those with SA (Table 1). Other trends observed that did not reach statistical significance but that may be clinically important were older age, greater prevalence of non-Caucasian race, and diabetes among those with SA. Notably total sleep time and lights out/on times were similar between those with and without SA. Subjects with NH were more likely to be Caucasian and had a higher AHI and greater severity of SA with a trend toward higher prevalence of congestive heart failure but were otherwise similar across baseline characteristics and comorbid conditions (Table 1).

Clock Gene Expression by SA or NH Status

Figure 1 depicts the percent change in relative clock gene expression for each of the target genes we studied. Of these, Per3 expression was statistically significantly decreased by 35% among those with SA compared with those without SA (P = 0.02). While potentially clinically important changes in the expression of other clock genes compared with control (no SA) were observed, none of these reached statistical significance (Fig. 1). When we defined SA as AHI ≥ 5, we found similar results overall and specifically, Per3 expression reduced by 35% (P = 0.01, data not shown). When SA was defined as severe (AHI ≥ 30), Per3 expression was persistently decreased but the point estimate was lower (17%) and not statistically significant (data not shown). We further observed a trend toward downregulation of Per3 with increasing severity of SA in four categories (Fig. 2, P trend 0.012). Next, compared with those without NH, Per1, NPAS2, and Rev-erbα were downregulated by 29, 50, and 31%, respectively, among those participants with NH (Fig. 3). Because there was no association between comorbid conditions or renal function and our predictors, SA or NH (Table 1), we did not perform any further adjustments in our analysis. Finally, we found no correlation between total sleep time, total REM sleep time, slow wave sleep, or sleep efficiency and expression of any of the clock genes we studied (data not shown).

Fig. 1.

Percent difference in clock gene expression for those with sleep apnea (SA) (compared with those without sleep apnea). Data table below shows degree of percent difference for each clock gene, P value for statistical significance of that difference from the no SA group, and the sample size for each analysis. *P < 0.05.

Fig. 2.

Relative gene expression for sleep apnea in 4 categories of increasing severity.

Fig. 3.

Percent difference in clock gene expression for those with nocturnal hypoxemia (NH) (compared with those without NH). Data table below shows degree of percent difference for each clock gene, P value for statistical significance of that difference from the no NH group, and the sample size for each analysis. *P < 0.05.

DISCUSSION

We found that among veterans with CKD, SA is associated with decreased morning expression of clock genes (as assessed from whole blood) that are key to circadian regulation of daily homeostatic processes. Per3 expression was reduced in SA, and Per1, NPAS2 and Rev-erbα expression was reduced among those with SA-related NH. Our findings suggest that there may be an important interrelationship between SA and clock gene regulation that warrants further investigation.

While there are no human studies of NH and clock gene expression, two studies have examined the association between SA and clock gene expression in whole blood (3, 10). Burioka et al. (3) used RT-PCR to determine 24 h Per1 mRNA expression (every 6 h) in eight healthy subjects and eight subjects with SA, matched for age and weight (mean age 43–45 yr). The investigators found that healthy subjects exhibited a sinusoidal diurnal pattern of Per1 expression, with a peak of expression at 6 AM and nadir at 2 AM (3). In contrast, those with SA had a blunted profile of expression without sinusoidal fluctuation (3). However, when SA subjects were treated with positive airway pressure (PAP) therapy, 24 h Per1 expression became sinusoidal with similar amplitude and phase as those without SA (3). In addition, measurement of markers of sympathetic nervous system (SNS) activity and inflammation using noradrenaline and interleukin-6/hs-CRP levels found that levels were elevated for those with SA compared with no SA and approached no SA levels after PAP therapy (3). Though this study was small, it suggests that SA is associated with dysregulation of Per1 gene expression and treatment of SA may restore 24 h Per1 expression and improve SNS activity and inflammation (3). This study did not study other clock genes or the role of SA-related NH in this pathway. Another study by Moreira et al. (10) compared the one-time morning expression of CLOCK, BMAL1, Cry1, Cry2, Per1, Per2, and Per3 in 13 male patients with severe SA and seven healthy male controls. They found that compared with healthy controls, CLOCK expression was decreased with a trend to decreased BMAL1 expression for those with SA (10). However, in the subset of SA patients who received PAP therapy, there was no restoration or change in clock gene expression, despite comparable duration and adequacy of PAP treatment (10). Together with our findings, it appears that SA is associated with some alteration in clock gene expression, but the target clock genes, the pattern of this alteration, and the responsiveness to PAP therapy are still to be elucidated. In addition, the impact of SA, independent of sleep disruption on clock gene expression, is unknown, though total sleep time was similar between those with and without SA in our cohort.

Altered clock gene expression associated with sleep disturbances such as sleep apnea may hypothetically contribute to some of the downstream negative cardiovascular consequences of SA. In rodent models, altered day/night rhythm leads to cardiovascular toxicity and concomitant alteration of clock gene expression; resynchronization restores clock gene expression patterns and mitigates observed cardiovascular effects (8). While this has not been studied in humans, the hypothesis that clock gene expression may be an intermediary pathway between SA and cardio-metabolic consequences is appealing. First, restoration of 24 h circadian clock gene expression could be a target or metric for treatment of SA. Second, 30–60% of SA patients do not tolerate standard treatment for SA with PAP therapy (14). As such, restoration of diurnal clock gene expression through alternative means may reduce SA-related cardio-metabolic toxicity by intervening upon this intermediary pathway. Nevertheless, the role of clock gene expression on the pathway between sleep disturbances and cardiovascular and metabolic outcomes remains to be elucidated.

To our knowledge, our analysis is the largest of its kind to examine clock gene expression in consecutively enrolled human subjects with and without SA, unselected for sleep complaints. This study has the advantage of have consistently timed morning measurements of clock gene expression. However, a cross-sectional study of this nature cannot establish causality or directionality of association. In addition, because we had one single time point measurement of clock gene expression, we are unable to comment on 24 h differences in clock gene expression. Finally, this was an older cohort with CKD. As such, our findings may not be generalizable to healthier cohorts. However, SA and CKD frequently coexist so our findings are relevant for this high-risk population.

In summary, our findings suggest that SA and SA-related NH may be associated with altered expression of clock genes key for regulation of daily bodily functions. This study lays the groundwork for more dedicated longitudinal studies of 24 h clock gene expression patterns in SA. A better understanding of the direction and magnitude of the association between SA and 24 h clock gene expression patterns may open the door for novel therapeutic targets to treat SA-related complications.

GRANTS

Veterans Affairs Clinical Science Research and Development Career Development Award CX000533-01A1 and University of Florida Department of Medicine, Division of Nephrology (to M. T. Canales); D. D. Dunlap Fund through the Fraternal Order of the Eagles (to M. Gumz); National Institute of Diabetes and Digestive and Kidney Diseases Grants R01-DK-045788 and R01-DK-107798 (to I. D. Weiner).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M.T.C., S.B., A.I., I.D.W., R.B.B., R.J.B., and M.L.G. conceived and designed research; M.T.C., M.R.H., and M.L.G. performed experiments; M.T.C., M.R.H., S.B., A.I., and M.L.G. analyzed data; M.T.C., M.R.H., S.B., A.I., I.D.W., R.B.B., R.J.B., and M.L.G. interpreted results of experiments; M.T.C. and S.B. prepared figures; M.T.C. drafted manuscript; M.T.C., M.R.H., S.B., A.I., I.D.W., R.B.B., R.J.B., and M.L.G. edited and revised manuscript; M.T.C., M.R.H., S.B., A.I., I.D.W., R.B.B., R.J.B., and M.L.G. approved final version of manuscript.