Modest sleep restriction does not influence steps, physical activity intensity, or glucose tolerance in obese adults

Abstract

Sleep restriction (<6 hrs) and PA are risk factors for obesity, but little work has examined the inter-related influences of both risk factors. In a free-living environment, thirteen overweight/obese adults were sleep-restricted for five nights to six-hour time-in-bed each night, with and without regular exercise (45 minutes/65% VO₂max) (counterbalanced design). Two days of recovery sleep followed sleep restriction. Subjects were measured during a mixed meal tolerance test (MMT), resting metabolic rate, cognitive testing and fat biopsy (n=8). Sleep restriction increased peak glucose response (+7.3 mg/dL, p=0.04), elevated fasting non-esterified fatty acid concentrations (NEFA, +0.1 mmol/L, p=0.001), and enhanced fat oxidation (p<0.001) without modifying step counts or PA intensities. Inclusion of daily exercise increased step count (+4,700 steps/day, p<0.001), decreased the insulin response to a meal (p=0.01) but did not prevent the increased peak glucose response or elevated NEFA levels. The weekend recovery period improved fasting glucose (p=0.02), insulin (p=0.02), NEFA concentrations (p=0.001) and HOMA-IR (p<0.01) despite reduced steps (p<0.01) and increased sedentary time (p<0.01). Abdominal AT (AT) samples, obtained after baseline, sleep restriction, and exercise, did not differ in lipolytic capacity following sleep restriction. Fatty acid synthase protein content tended to increase following sleep restriction (p=0.07), but not following exercise. In a free-living setting, sleep restriction adversely affected circulating NEFAs, fuel oxidation, and peak glucose response but did not directly affect glucose tolerance or AT lipolysis. Sleep restriction-associated metabolic impairments were not mitigated by exercise, yet recovery sleep completely rescued its adverse effects on glucose metabolism.

Introduction

Increases in the prevalence of obesity and insulin resistance have paralleled lifestyle changes such as insufficient sleep. Approximately 33–45% of American adults do not obtain the recommended 7–9 hours of sleep/night, and about 40% of adults choose to sleep longer on weekends to “catch-up” for the shorter weekday sleep schedules (Depner et al., 2019). Sleep restriction (SR) strongly associates with systemic insulin resistance (Depner et al., 2019; Nedeltcheva, Kessler, Imperial, & Penev, 2009; Schmid et al., 2009), and therefore obese/overweight adults may be particularly susceptible to its adverse effects. Interestingly, 2-days of catch-up sleep have been shown to normalize insulin sensitivity in healthy men after 5-days of restricted sleep (Broussard, Wroblewski, Kilkus, & Tasali, 2016). Despite extensive work in healthy adults investigating workweek SR and catch-up weekend sleep, the impact on glucose tolerance in overweight and obese adults is not well-known.

Compounding the effects of SR on metabolism is the possibility of SR modifying PA patterns. Studies utilizing at-home SR have yielded varying observations on free-living PA, from increased activity (Brondel, Romer, Nougues, Touyarou, & Davenne, 2010) to decreased activity (Schmid et al., 2009). Investigations of healthy males who were sedentary but underwent exercise-training before or during a single night of sleep deprivation (24 h or 60 h), were protected against SR-induced reductions in insulin sensitivity (VanHelder, Symons, & Radomski, 1993), blunted glycemic responses to glucose ingestion (de Souza, Dattilo, de Mello, Tufik, & Antunes, 2017), and elevated fasting NEFA concentrations (de Souza et al., 2017), which can be predictive of insulin resistance or type 2 diabetes.

Although SR induces insulin resistance (Buxton et al., 2010), the underlying mechanisms are unclear. Many studies (Broussard et al., 2015; Donga et al., 2010; Schmid et al., 2011) report elevated circulating NEFA concentrations in sleep-restricted adults suggesting that disrupted AT insulin sensitivity may be a potential mechanism of sleep-restriction mediated insulin resistance. Indeed, one study showed a 30% reduction in AT insulin signaling capacity after 5-days of SR (Broussard, Ehrmann, Van Cauter, Tasali, & Brady, 2012), and another study in healthy males showed a blunting of skeletal muscle insulin signaling after 2-days of SR (Sweeney, Jeromson, Hamilton, Brooks, & Walshe, 2017). These limited findings suggest that sleep-restriction induced NEFA released from AT may impair insulin signaling in peripheral tissues, as has been repeatedly shown in studies not focusing on sleep. Exercise not only increases AT insulin signaling (Thyfault & Bergouignan, 2020) but also improves skeletal muscle fatty acid handling and thus, it may confer functional benefits in overweight/obese adults whose adipose and muscle metabolism is negatively affected by SR.

The purpose of this study was three-fold: to 1) determine the glucoregulatory effects of moderate SR (~6 h of sleep/night) in a free-living environment in overweight/obese individuals, 2) determine the effects of SR on PA patterns and dietary composition, and 3) determine the impact of SR on AT insulin sensitivity. We hypothesized that in a free-living setting SR would decrease PA, resulting in reduced glucose tolerance and AT insulin sensitivity, and speculated that the addition of regular exercise during SR would ameliorate the detrimental impact on whole body glucose and fatty acid metabolism as well as improvement in AT insulin sensitivity.

Methods

This study was approved by the University of Missouri’s Health Science Institutional Review Board (Protocol #2011452; clincialtrials.gov#NCT03556410), and all participants provided written informed consent. Participants (seven males and six females) were regular sleepers as defined by 7–9 hours/weeknight, overweight/obese (BMI 25 – 40 kg/m²), and ages 21–40 years (Table 1). Subjects were excluded if they did excessive driving or operated heavy machinery, had sleep apnea, or taking any glucose/lipid altering medications. To control for effects of the menstrual cycle, female participants initiated testing 1–7 days following the start of menstruation.

Table 1.

Baseline Characteristics

Characteristics	Overall Mean ± SEM	Males	Females
Age (years)	28.8 ± 1. 2	27.3 ± 1.6	30.5 ± 1.7
Body Mass Index (kg/m2)	31.5 ± 1.0	31.2 ± 1.4	31.8 ± 1.6
Body Fat (%)	35.3 ± 1.6	32.6 ± 1.9	38.5 ±2.1
Lean mass (kg)	57.4 ± 2.2	61.8 ± 2.4	52.2 ± 2.6@
Fat mass (kg)	32.8 ± 2.2	31.0 ± 3.0	34.9 ± 3.3
Bone Mineral Density	0.5 ± 0.4	0.0 ± 0.4	1.1 ± 0.4
Waist (cm)	102 ± 3	103.7 ± 3.1	99.0 ± 5.7
Hip (cm)	115 ± 2	111.9 ± 2.1	117.2 ± 3.9
VO2max (mL/kg/min)	34.3 ± 1.9	36.3 ± 2.6	32.0 ± 2.8
Blood Pressure (mmHg)	119±2 / 69±2	121±3 / 69±2	117±3 / 69±3
Pittsburg sleep quality index	2.8 ± 0.3	3.0 ± 0.4	2.5 ± 0.4
Epworth sleepiness scale	5.7 ± 1.2	5.1 ± 1.7	6.5 ± 1.8
Berlin Score	0.7 ± 0.2	1.0 ± 0.2	0.3 ± 0.2
O2 Adjusted Index (events/hour)	2.7 ± 0.6	3.3 ± 0.8	2.0 ± 0.9
Morningness-Eveningness Questionnaire	53.8 ± 2.5	50.0 ± 3.1	58.3 ± 3.3
Fasting Circulatory Markers
Triglycerides (mg/dL)	115.1 ± 20.6	123.0 ± 24.3	91.8 ± 32.1
Cholesterol (mg/dL)	164.2 ± 7.0	168.1 ± 9.8	159.7 ± 10.6
Low Density Lipoproteins (mg/dL)	108.8 ± 7.5	110.2 ± 10.2	106.8 ± 12.5
High Density Lipoproteins (mg/dL)	43.3 ± 4.0	35.0 ± 4.3	53.0 ± 4.6@

^@p<0.05 vs. male participants

Experimental Design

Two, two-week experimental SR protocols were utilized with a crossover, counterbalanced design (Figure 1). A baseline week preceded both conditions and sleep was restricted during the second week (Monday–Friday) to six hours of time-in-bed (Condition 1–SR) and repeated with a daily exercise regimen (Condition 2– SREX). Mixed meal tests (MMT; Ensure Original Vanilla – 440 kcal (64g carbohydrate, 12g fat, 18g protein), Abbott Laboratories, USA) were administered following each week of baseline monitoring, following both five-day intervention periods (sleep with or without exercise), and the subsequent two-day weekends of recovery sleep. Following a 10 h overnight fast, the MMT was administered in the morning (0600–0700 h) with the same start time for each subject. Blood samples were collected at fasting and throughout the three-hour MMT. In a subset of subjects (n=8), abdominal AT biopsies were collected on one baseline study day and after both condition’s five sleep restricted weeknights. There was a minimum of a two-week washout period between study conditions; sleep was not monitored during this period.

Figure 1.

Experimental design.

Screening

Questionnaires:

Screening questionnaires were completed providing a health inventory, and PA and food habits. Sleep habits were assessed with the Epworth Sleepiness Scale (ESS) (Johns, 1991), Pittsburg Sleep Quality Index (PSQI score <5) (Buysse, Reynolds, Monk, Berman, & Kupfer, 1989), Berlin Questionnaire (Netzer, Stoohs, Netzer, Clark, & Strohl, 1999) (one risk factor in addition to obesity), and Morningness-Eveningness Questionnaire (MEQ) (Horne & Ostberg, 1976) and used in combination to exclude subjects from the study.

Sleep Apnea Screening:

Subject were initially screened using the Berlin questionnaire. Additionally, a wrist-based pulse oximeter (WristOX2, Model 3150, Nonin Medical, Inc., MN, USA) for one night to assess pulse and oxygen for sleep apnea screening. We excluded based on the oxygen desaturation index, a 4% reduction in SpO₂ (Dempsey, Veasey, Morgan, & O’Donnell, 2010.)

Body Composition and Aerobic Capacity:

Dual x-ray absorptiometry (DXA) was used to determine body composition. Participants also performed a peak aerobic capacity test (VO₂peak) using a continuous treadmill protocol (supplement 1).

Testing Procedures

Sleep Restriction and Monitoring:

Both conditions of the experimental design incorporated SR periods. Participants who habitually slept 7–9 hours/night restricted sleep to ~six hours for 5-days (6 h time-in-bed). Shortened sleep was accomplished by subjects delaying bedtime while maintaining habitual wake time. Adherence was monitored through an Actiwatch Spectrum Plus (Philips Respironics, OR, USA) for two weeks for both conditions. Self-report sleep diaries of time-in-bed/time-out-of-bed were kept and were used to verify Actiwatch data, a standardized protocol (McGovney, Curtis, McCann, & McCrae, 2019). Sleep interval detection algorithms (Actiware-version 6.0.9) scored epochs using default settings, requiring 10 minutes of immobile/activity for sleep onset/termination of sleep, respectively, and determining sleep onset latency, wake-after-sleep onset, and sleep efficiency (supplement 1).

Physical Activity Monitoring:

An Actigraph (ActiGraph Corp., FL, USA) was worn on the dominant hip for the two-week period. Exercise performed was also recorded and detailed within the sleep diary. The cut off points for sedentary, light and moderate to vigorous PA were according to Freedson et al (Freedson, Melanson, & Sirard, 1998).

Resting Metabolic Rate (RMR) and Exercise Regimen:

RMR (Parvo Medics TrueOne 2400, UT, USA) were collected in the fasting state prior to each of the MMTs with a 20-minute resting period and 15-minute data collection period.

During the SREX condition, participants completed 45 minutes of supervised walking at 65% VO₂peak each day of SR (Figure 1). The exercise was conducted at the same time for each subject with most subjects exercising either in the morning before work or at noon.

Dietary Records:

Participants were instructed to record all food/beverage consumed and time of consumption for 4-days during each baseline and intervention weeks, Thursday-Sunday. Caffeine was not limited throughout the conditions other than abstaining for 10 hours prior to the MMT. Dietary records were analyzed with Nutritionist Pro (Axxya Systems LLC, WA, USA).

Stress Evaluation:

Participants completed the Depression Anxiety Stress Scale (DASS) questionnaire (Brown, Chorpita, Korotitsch, & Barlow, 1997). Questionnaire were initiated after the one-hour blood draw of MMT.

Blood and Tissue Samples

Blood Samples:

Samples collected in EDTA tubes with DPP-IV inhibitor and aprotinin added. Samples were immediately tested for glucose concentrations (YSI 2300 STAT PLUS, YSI Incorporated, OH, USA) and were measured with Cholestech LDX Analyzer (Abbott Laboratories/Alere, CA, USA). The sample was frozen and later tested for TNFα and IL6 concentrations (MilliporeSigma, MA, USA).

AT Biopsy:

Biopsy collections were conducted after baseline, SR, and SREX. Biopsies were taken following the overnight fast, the RMR measurement and prior to the MMT. The biopsy was taken from either the left or right anterior axillary line slightly above or below the umbilicus following procedures previously published (Santosa, Swain, Tchkonia, Kirkland, & Jensen, 2015) (supplement 1).

AT Lipolysis Assay:

The tissue aliquot was immediately held in low glucose DMEM (and 2% bovine serum albumin (BSA;A8806–5G, MilliporeSigma, MA, USA) on ice using procedures reported previously (Winn et al., 2019). Stimulation media was analyzed for secretion of glycerol (glycerol assay kit - ab133130, Abcam Inc., USA) and NEFA (NEFA-HR(2), Wako Diagnostics, CA, USA) (supplement 1).

Western Blots:

Triton XS-100 tissue lysates were used to produce Western blot ready Laemmli samples from ISO stimulated AT (Porter et al., 2017). Protein samples were probed with primary antibodies (hormone sensitive lipase (HSL; 4107S, Cell Signaling Technology, MA, USA), phospho- (p) HSL(Serine 660, PA5–64494, Invitrogen, CA, USA), and fatty acid synthase (FAS, C2065, 4107S, Cell Signaling Technology, MA, USA)) (Porter et al., 2017).

Calculations

Insulin resistance and insulin sensitivity were calculated using the homeostatic model assessment of insulin resistance (HOMA-IR) (Matsuda & DeFronzo, 1999), Matsuda index of insulin sensitivity (Matsuda & DeFronzo, 1999), and AT insulin resistance (Adipo-IR) (Porter et al., 2017). Absolute area under the curve (AUC) was calculated by trapezoidal rule on GraphPad Prism version 8.3 (GraphPad Software, LLC, CA, USA) using all time points.

Statistical Analysis

The statistical analysis was performed using SPSS statistical software, version 26 (IBM, Inc., NY, USA), with a linear mixed model created for a 2×2 ANOVA with repeated measures. Measures were run as baseline vs. intervention, SR vs. SREX, intervention vs. recovery, and baseline vs. recovery. Additionally, the intensity of PA was analyzed with the same repeated-measures model, with the addition of baseline weekdays vs. baseline weekend, and weekend vs. recovery. Glucose and insulin curves during the MMT were modeled by conditions, study days, and time (11 time points, 0–180 minutes). To evaluate the ex vivo AT response to adrenergic stimulation after the intervention period, repeated measures ANOVA tested baseline vs. SR vs. SREX. Statistical significance was p≤0.05. All data are shown as mean±SEM.

Results

Subject characteristics

Thirteen overweight/obese adults (seven males, six females) completed both study conditions, and eight subjects (four males, four females) completed three AT biopsies each. Baseline concentrations of cholesterol, triglycerides, low-density lipoproteins, and high-density lipoproteins were all in normal ranges (Table 1). Subjects had a mean BMI of 31.5±1.0 kg/m² and percent body fat of 35.3±1.6%. A global PSQI score of 2.8±0.3 was indicative that all subjects were “good” sleepers. The ESS score of 5.7±1.2 indicated that two subjects had “mild” daytime sleepiness. The Berlin questionnaire score was 0.7±0.2, while the pulse oximetry screening for sleep apnea indicated no subjects exhibited sleep apnea. Additional sex differences were explored for baseline characteristics. Male subjects had more lean mass compared to females (61.8 ± 2.4 kg vs. 52.2 ± 2.6; p=0.018) but body fat percentage only tended (p=0.06) to be elevated in females (32.6 ± 1.9 % vs. 38.5 ±2.1). Unsurprisingly, female participants had elevated HDL compared to males (35.0 ± 4.3 mg/dL vs. 53.0 ± 4.6; p=0.015).

Sleep

Habitual sleep duration at home during the baseline weekdays was 8.0±0.1 h/night and was not different between SR and SREX (Table 2). Subjects slept 5.9±0.1 during SR and 5.9±0.0 h/night during SREX; both significantly lower than baseline sleep duration (p<0.001). Independent of exercise, sleep efficiency during SR was increased by 2.2±0.9% compared to baseline weekdays (p<0.001).

Table 2.

Actigraphy data for the study conditions.

	Condition	Baseline	Baseline Weekend	Intervention	Recovery
Duration (hours)	SR	7.9±0.2	7.7±0.2	5.9±0.1 *†	7.4±0.3
	SREX	8.1±0.2	7.1±0.2a d	5.9±0.0 *†	8.2±0.4
Efficiency (%)	SR	88.5±1.6	88.2±1.5	91.1±1.3 *	89.7±1.3
	SREX	89.4±1.0	89±1.4	91.8±0.6 *	90.2±1.3
WASO (min)	SR	41±5.6	39.2±4.4	24.6±3.8 *†	35.7±3.1
	SREX	37.2±3.1	28.1±2.4	22.5±2.4 *†	39.3±5.9
Bed Time (hh:mm)	SR	23:19	22:58	0:36 *	23:43
	SREX	23:23	23:44	0:21 *	23:08
Wake Time (hh:mm)	SR	07:26	06:58	6:36	7:18
	SREX	07:41	07:06	6:25	7:24

Data are mean ± SEM. SR (sleep restriction); SREX; WASO, wake after sleep onset.

Main effect of study day:

^*p<0.05 vs. baseline;

^†p<0.05 vs. recovery.

Post hoc:

^ap<0.05 vs. baseline (within condition);

^dp<0.05 vs. recovery (within condition);

^#p<0.05 vs. condition 1 (within study day).

Sleep duration on the baseline weekend was 0.6 h less (p=0.01) than baseline weekdays. Specifically, sleep duration on the weekend preceding SREX was 1.0 h less than during baseline weekdays (p=0.03). Following 5-days of restricted sleep, independent of exercise, subjects extended sleep during the recovery weekend period (~7.8±0.2 h), similar to baseline weekdays and weekend. During the recovery weekend, independent of the preceding intervention, subjects spent more time awake (+14 min) during the night compared to sleep restriction, independent of exercise (p<0.001).

Energy metabolism

Fasting glucose concentrations were similar between the baseline and intervention periods (Table 3). The pattern of glucose responses to the MMT was different between baseline and SR±EX (condition-by-study day interaction, p<0.01; Figure 2A,B), such that independent of exercise, SR caused higher glucose concentrations throughout the first hour and lower glucose concentrations in the third hour. Peak glucose concentrations were also higher following sleep restriction, independent of exercise, compared to baseline (p=0.04). When analyzed as glucose AUC over 3 h, there was no difference in total glucose concentrations (Figure 2C).

Table 3.

Fasting blood markers and insulin sensitivity indices.

	Condition	Baseline	Intervention	Recovery
Fasting Blood Markers
Glucose (mg/dL)	SR	84.8±1.9	82.5±2.3	82.2±1.6 *
	SREX	84.1±2.2	83.0±2.1	80.9±1.4 *
Insulin (μIU/mL)	SR	12.7±1.7	14.4±3.5	11.2±1.8 *
	SREX	13.4±2.0	9.9±1.3	10.9±1.5 *
C-peptide (mg/dL)	SR	1,175±71	1,194±140	1,237±168
	SREX	1,338±177	952±82	991±83
NEFA (mmol/L)	SR	0.29±0.03	0.42±0.05 *†	0.28±0.03
	SREX	0.33±0.05	0.40±0.05 *†	0.30±0.03
TNFα (pg/mL)	SR	2.9±0.3	2.8±0.4	2.9±0.3
	SREX	2.5±0.2	2.6±0.2	2.3±0.3
IL-6 (pg/mL)	SR	3.1±0.5	3.6±0.9	4.0±1.1
	SREX	2.6±0.7	3.0±0.8	2.7±0.8
Mixed Meal Test AUCs
C-peptide (mg/min/dL)	SR	485,242±32,811	511,181±37,813	469,063±27,041
	SREX	511,595±38,647	475,664±35,990	483,768±33,819
Insulin Sensitivity Indices
HOMA-IR	SR	2.7±0.4	3.0±0.7	2.3±0.4 *
	SREX	2.8±0.5	2.0±0.3	2.2±0.3 *
Matsuda	SR	5.3±0.9	5.2±0.9d	6.2±1.0
	SREX	5.3±1.1	5.8±0.9	5.6±0.8

Data are mean ± SEM. GLP-1, glucagon like peptide-1; GIP, gastric inhibitory polypeptide; NEFA, nonesterified fatty acid; TNFα, tumor necrosis factor alpha; IL-6, interleukin 6; HOMA-IR; homeostatic model of assessment of insulin resistance.

Main effect of study day:

^*p<0.05 vs. baseline;

^†p<0.05 vs. recovery.

Post hoc:

^ap<0.05 vs. recovery (within condition);

^dp<0.05 vs. recovery (within condition).

Figure 2.

Effects of sleep restriction on glucose and insulin responses with and without exercise. Glucose curves in response to an MMT for (A) SR and (B) SREX, and (C) glucose AUC. Insulin curves in response to an MMT for (D) SR and (E) SREX, and (F) insulin AUC. Data are mean ± SEM. Post hoc: ^dp<0.05 vs. recovery (within condition); ^#p<0.05 vs. SR (within study day).

Following weekend recovery, fasting glucose concentrations decreased slightly compared to baseline values (81.5±1.2 vs 84.4±1.6 mg/dL, respectively; p=0.02; Table 3). The pattern of glucose responses to the MMT after the recovery weekend demonstrated a study day-by-time interaction (p=0.05; Figure 2A,B) compared to the response to SR with or without exercise. The recovery glucose response was similar to baseline. Expressed as AUC, no differences between any of the MMT glucose responses were observed.

Fasting insulin concentrations were not different between baseline and the intervention (Table 3). The insulin patterns in response to MMT were not affected by SR with or without exercise (Figure 2D,E), but total insulin AUC was lower during SREX compared to SR alone (p=0.01). Surprisingly, following the recovery period, independent of exercise during SR, fasting insulin concentrations were lower than baseline (p=0.02). The peak insulin response to MMT was lower following SR recovery compared to SR (p=0.05), and SREX recovery was higher than SREX (p<0.01) (condition-by-study day interaction, p=0.001; Figure 2D,E). Additionally, insulin AUC after SR was elevated compared to the subsequent recovery (p=0.05) (condition-by-study day interaction, p=0.01; Figure 2F). HOMA-IR and Matsuda index were similar at baseline and not affected by SR with or without exercise (Table 3). Interestingly, compared to baseline, HOMA-IR was attenuated following recovery (p<0.01). Insulin sensitivity assessed via Matsuda index also improved after recovery following SR compared to SR (p=0.02) (condition-by-study day interaction, p=0.04; Table 3). Fasting c-peptide concentrations, which may be more reflective of insulin release than fasting insulin levels, tended to be decreased after sleep restriction, independent of exercise, compared to baseline (p=0.06), and were normalized after the recovery period. No differences were observed in AUC between any study day periods. In addition, we anticipated that SR would induce stress provoking an inflammatory response. Surprisingly, neither fasting TNFα and IL-6 concentrations changed throughout either condition (Table 3).

Physical Activity

Step counts during baseline weekdays were similar between conditions (Baseline_SR, 7,945±579 steps; Baseline_SREX, 8,137±709 steps) (Figure 3A). Step counts were maintained during SR (7,466±653 steps, p>0.05), and as expected steps increased with daily exercise (SREX, 12,877±670 steps, p<0.001 compared to both baseline and SR). Sedentary PA decreased during SREX compared to both baseline periods (−2.6%, p=0.05) and SR (−4.8%, p<0.001; Figure 3B). Light PA (LTPA) during the intervention was similar to PA during baseline weekdays. Daily exercise during SREX increased MVPA compared to SR (+4.4%, p<0.001) and B_SREX (+3.7%, condition-by-study day interaction p<0.001), while MVPA did not change with SR alone (Figure 3D).

Figure 3.

Effects of sleep restriction on PA with and without exercise.

Step counts (A), sedentary PA (PA) (B), light PA (C), and moderate-vigorous PA (MVPA) (D). Data are mean ± SEM. ****p<0.05 compared to all other study days. Post hoc: ^ap<0.05 vs. baseline (within condition); ^dp<0.05 vs. recovery (within condition); ^#p<0.05 vs. SR (within study day).

Step counts during the baseline weekend were reduced (~23%) compared to weekdays (Baseline, 8,040±471 steps; weekend, 6,164±494 steps; p<0.01). Although steps were reduced during the weekend, subjects decreased sedentary time (p=0.02; Figure 3B) and MVPA (p=0.01; Figure 3D), therefore, LTPA increased compared to baseline weekdays (p<0.001; Figure 3C).

During the recovery period, step counts decreased dramatically compared to the weekend (recovery, 4,639±412 steps; weekend, 6164±494 steps; p<0.01; Figure 3A). The recovery period had higher sedentary time compared to the weekend (+3.9%, p<0.01; Figure 3B). The increased LTPA observed during the weekend compared to baseline weekdays tended to be diminished following sleep restriction, independent of exercise (intervention vs. recovery, p=0.07; weekend vs. recovery, p<0.01). The decreased MVPA during the weekend compared to baseline weekdays was further exacerbated following the intervention with subjects performing 1.2% less MVPA during recovery than baseline weekend. (p=0.03).

Nonesterified fatty acids

Fasting NEFA concentrations were elevated after SR, independent of exercise (SR, 0.42±0.05 mmol/L; SREX, 0.40±0.05 mmol/L) compared to baseline (SR, 0.29±0.03 mmol/L; SREX, 0.33±0.05 mmol/L; p=0.001) and compared to recovery (SR, 0.28±0.03 mmol/L; SREX, 0.30±0.03 mmol/L; p=0.001; Table 3). The NEFA curves in response to MMT were not similar between the intervention and recovery (study day-by-time interaction, p=0.02; Figure 4A,B), due to the elevated fasting NEFA levels. Following the intervention, NEFA AUC during the MMT was elevated compared to baseline (p<0.05) and recovery (p<0.01; Figure 4C).

Figure 4.

Effects of sleep restriction on non-esterified fatty acids (NEFA) and AT insulin resistance with and without exercise. NEFA responses to MMT in SR (A), and SREX (B) and NEFA AUC (C). Calculations for Adipo-IR (D) based on fasting insulin and NEFA concentrations. Data are mean ± SEM. AUC, area under the curve. ****p<0.05 compared to all other study days.

The recovery period effectively reduced fasting circulating NEFA concentrations to baseline levels. Consistent with HOMA-IR and the Matsuda index, the recovery sleep period tended to reduce Adipo-IR compared to SR, independent of exercise (p=0.06; Figure 4D).

AT samples were stimulated with a beta-adrenergic agonist isoproterenol and assessed for protein differences in key enzymes of fatty acid oxidation. The ratio of phospho-HSL to total HSL content was not different by any condition or unstimulated/stimulated scenarios (Figure 5A). Exploratory analysis revealed a trend of increased stimulated samples’ protein content of fatty acid synthase (FAS) after SR compared to baseline (p=0.07; an effect that was normalized by exercise; Figure 5B). There were no differences in levels of NEFA or glycerol between unstimulated/stimulated tissue samples for any of the three comparisons (Figure 5D,E). However, there was an increased change in FA released from unstimulated/stimulated AT samples in which there was a 1.58-fold increase in NEFA with SREX compared to baseline (p=0.04) and 1.75-fold increase compared to SR (p=0.03).

Figure 5.

AT adrenergic stimulation following sleep restriction with and without exercise. Baseline, SR, and SREX (A) ratio of phosphorylated HSL to total HSL and (B) FAS, with representative protein blots. Unstimulated and stimulated secretion of (D) NEFA and (E) glycerol.

Energy and nutrition

The subjects’ body weights and RMR did not change with either condition (Table 4). Total energy consumption was similar between weekdays but was greater during the baseline weekend preceding SREX compared to the baseline weekend of SR (p=0.03) and compared to the recovery period following SREX p<0.05). Energy expenditure (kcal/hour) from the accelerometer was elevated during SREX (p<0.05) compare to baseline weekdays, expenditure was not different from baseline during SR. Energy expenditure was reduced during the baseline weekend compared to baseline and further reduced during weekend recovery (p<0.05).

Table 4.

Energy expenditure and macronutrient composition of subjects’ diet during each study condition.

	Condition	Baseline	Baseline weekend	Intervention	Weekend recovery
Weight (kg)	SR	92.5±3.1		92.2±3.1	92.4±3.2
	SREX	92.3±3.5		92.4±3.3	92.5±3.3
REE (kcal)	SR	1740±66		1777±87	1726±57
	SREX	1727±70		1756±63	1705±50
RER	SR	0.89±0.01		0.87±0.02 *†	0.91±0.02
	SREX	0.90±0.01		0.85±0.01 *†	0.88±0.01
Energy Expenditure (kcal/hour)	SR	27.2±2.5	21.1±2.5 *†	26.2±2.5d	18.4±1.9
	SREX	28.0±2.8	23.6±3.0 *†	47.3±4.0ad	15.5±1.9
Energy Consumed (kcal/day)	SR	2264±1.87	2025±172	2182±172	2320±238
	SREX	2542±420	2512±246#	2153±214	2024±156b
%Protein	SR	15.96±1.06	18.36±1.50	16.78±1.41	17.02±1.32
	SREX	17.48±1.50	17.27±1.38	17.05±1.00	16.85±0.77
%CHO	SR	43.49±1.44	39.98±1.33†	44.34±2.54 *	45.57±1.79
	SREX	43.06±1.96	42.86±2.67†	48.42±1.86 *	45.44±2.25
%Fat	SR	36.37±1.73	39.45±1.61†	37.33±2.58	34.43±1.34
	SREX	35.38±2.57	36.27±2.56†	34.31±2.03	33.97±2.58
%Alcohol	SR	3.70±1.77	2.07±1.10	1.14±0.57	3.60±1.43
	SREX	3.19±1.53	4.24±1.97	0.43±0.40	2.57±1.34
Caffeine (mg)	SR	122.4±42.3	114.9±41.1	112.1±38.3	92.9±25.0
	SREX	127.5±35.0	107.4±31.9	125.7±37.1	71.6±28.1

Data are mean ± SEM. REE, resting energy expenditure; RER, respiratory exchange ratio; CHO, carbohydrate.

Main effect of study day:

^*p<0.05 vs. baseline;

^†p<0.05 vs. recovery.

Post hoc:

^ap<0.05 vs baseline,

^bp<0.05 vs. baseline weekend,

^dp<0.05 vs. recovery (within condition),

^#p<0.05 vs SR (within study day).

Dietary macronutrient composition was analyzed to determine changes in dietary composition that occurred with SR (Table 4). Subjects consumed more carbohydrates during sleep restriction, independent of exercise, than during baseline weekdays (46.4±1.9% vs. 43.3±1.3%, respectively; p=0.01), as well as more carbohydrates during weekend recovery compared to baseline weekend (p<0.01). Dietary fat intake decreased during weekend recovery compared to the baseline weekend period (p=0.04). Alcohol consumption tended to be higher during the baseline periods compared to both SR conditions (p=0.06), likely due to the required abstinence from alcohol on the Friday prior to the study day that followed the intervention. Subjects tended to consume more caffeine during the weekdays than weekend/recovery (p=0.06).

The respiratory exchange ratio (RER) shifted toward fatty acid oxidation in response to SR, independent of exercise (p<0.001; Table 4). The RER for recovery following SR tended to be higher than following SREX (p=0.06) (condition-by-study day interaction: baseline vs. recovery p=0.04).

Depression Anxiety and Stress Scale

Subjects reported higher scores on DASS42 during SR with or without exercise (intervention, 10.5±2.8), compared to baseline (4.4±1.1, p=0.02), and compared to recovery (2.8±0.9, p=0.01). Exercise did not alleviate the perceived stress of SR.

Discussion

This is the first study in a free-living setting to report the effects of modest sleep loss during a workweek on glucose tolerance, PA, and dietary composition in overweight men and women. Furthermore, the effects of prescribed daily exercise and a two-day sleep recovery period were assessed on those outcomes. Sleep restriction increased perceived stress, increased peak glucose levels during the MMT, elevated fasting NEFA concentrations, shifted macronutrient intake toward more carbohydrate consumption, and altered fuel oxidation toward enhanced fat oxidation. These outcomes were not normalized with prescribed exercise. Contrary to expectations, SR did not negatively affect overall glucose tolerance, HOMA-IR or the Matsuda index in overweight/obese adults. Importantly, the ad libitum sleep opportunity following SR, independent of exercise, improved fasting glucose, insulin, and NEFA concentrations, along with indexes of insulin resistance, despite reduced steps and increased sedentary time. These novel data highlight the robust negative effects of SR and that prescribed daily moderate-intensity exercise is not sufficient to counteract these adverse effects in obese individuals.

Most SR studies have implicated insulin resistance as the major negative health outcome (Buxton et al., 2010; Depner et al., 2019; Nedeltcheva et al., 2009; Wang, Greer, Porter, Kaur, & Youngstedt, 2016). Our data show that peak glucose concentrations were elevated (+7.3 mg/dL) by SR, independent of exercise, but this short-term modest SR in overweight adults did not increase HOMA-IR, fasting insulin, or fasting glucose concentrations. Discrepancy between our findings and those in healthy subjects may be due to a MMT being utilized as opposed to a glucose challenge, potentially making it more difficult to detect small changes in insulin sensitivity compared to other studies (Buxton et al., 2010; Rao et al., 2015; Robertson, Russell-Jones, Umpleby, & Dijk, 2013; Wilms et al., 2019) which used a hyperinsulinemic, euglycemic clamp. Furthermore, these changes in insulin sensitivity may be transient and normalized after longer periods of less drastic reductions in sleep duration (Robertson et al., 2013). Further the less drastic sleep loss in our study is more realistic to the sleep loss in free-living people (2h reduction) than what has been reported in nearly all the previous studies (4h or less) and thus may provide more translatable application to the general public. Longer duration free-living studies in overweight/obese adults would provide better insight into the consequences of moderate shortened sleep on insulin sensitivity.

Healthy men have been shown to reduce their spontaneous PA when sleep was restricted (four hours) (Schmid et al., 2009), with lower PA with late night sleep loss than early night sleep loss (Wilms et al., 2020). In contrast, Tajiri et al. (Tajiri, Yoshimura, Hatamoto, Tanaka, & Shimoda, 2018) reported that healthy free-living Japanese women increased step counts during SR, but with extra time awake accounted for, significance was lost. Surprisingly, we noted no changes in step counts or PA intensity during SR in a free-living setting during the workweek. Our subjects were relatively active (~8,000 steps/day) for an overweight/sedentary population. During the baseline weekend, these subjects were less active compared to the weekdays, and even more so during the recovery weekend from the sleep-restricted workweek. This low level of activity on the weekend aligns with previous work reporting lower weekend PA (Lakoski & Kozlitina, 2014), thus prospective work must consider which days of the week to SR subjects when studying PA behavior.

Weight gain has been linked with sleep loss. Previous studies showed that healthy adults experiencing insufficient sleep (Bosy-Westphal et al., 2008; Markwald et al., 2013) increased energy intake and weight with ad libitum food access. In contrast, our five days of SR had no impact on subject body weight, despite having unrestricted food access, and exercising throughout the sleep-restricted workweek did not influence energy consumption. Dietary records showed a higher percentage of carbohydrates (+3.1%) consumed during the SR workweek compared to baseline weekday, in agreement with some (Markwald et al., 2013), while others report increased fat consumption during SR (Brondel et al., 2010; Schmid et al., 2009). During the subsequent recovery weekend, subjects consumed a higher percentage of carbohydrates and lower fat compared to the baseline weekend. Similar to activity behaviors, dietary choices may be influenced in the days after SR and be a critical component of the risks for obesity and insulin resistance.

To date, few studies have monitored PA during periods of SR, and only one study has investigated exercise during a multi-night SR protocol. VanHelder et al. (VanHelder et al., 1993) reported that 60 hours of sleep deprivation, with concurrent exercise, was only partially protective against the increased insulin responses seen during total sleep deprivation. Recently, another study noted that high-intensity interval training for two weeks prior to 24-hour sleep deprivation attenuated increases in glucose, insulin, and NEFA concentrations (de Souza et al., 2017). While these studies examined extreme sleep loss, our study showed that in overweight adults performing moderate-intensity exercise while undergoing modest SR did not alter fasting glucose or insulin concentrations, but the exercise was effective at reducing the insulin AUC in response to MMT.

The mechanisms for whole-body insulin resistance observed with SR are still unknown. Only a few investigators have explored the mechanisms in peripheral tissues. Broussard and colleagues (Broussard et al., 2012) have shown reduced Akt phosphorylation in adipocytes after 4-days of SR (4.5 hours of sleep), indicating a lack of inhibition on lipolysis. Likewise, Sweeney et al. (Sweeney et al., 2017) indicated a propensity to decrease Akt activity within the skeletal muscle after two nights of SR (50% habitual), demonstrating a reduction in insulin signaling capacity. Thus, it appears that insulin signaling in peripheral tissues may be disrupted, yet fatty acid metabolism has not been well explored.

Furthermore, studies have reported that NEFA are elevated with SR (Broussard et al., 2015; de Souza et al., 2017; Rao et al., 2015; Schmid et al., 2011). As glucose concentrations are not frequently elevated in fasting states in response to SR (Buxton et al., 2010; Wang et al., 2016; Wilms et al., 2019), AT may be utilizing circulating glucose as opposed to increased lipolysis of stored triglycerides in AT contributing to NEFA pool. In the present study, AT FAS protein content tended to be upregulated with SR, despite no adverse effects of SR on Adipo-IR. The increased FAS protein may be contributing to the development of insulin resistance as an excessive tissue FA accumulation is known to disrupt insulin signaling (Muoio & Neufer, 2012). Exercise mitigated this slight increase in FAS despite not effectively reducing the circulating NEFA concentrations. Further, increased NEFA concentrations and decreased RER data suggest a whole-body switch toward FA metabolism, similar to others (Hibi et al., 2017; Rao et al., 2015), and weekend sleep recovery restored fasting levels of NEFA demonstrating metabolic flexibility. Our data do not conclusively implicate AT lipolysis or fatty acid synthesis as the mechanisms for adipose dysfunction with SR, more mechanistic studies of both AT and skeletal muscle are needed to determine their role in sleep-restriction induced insulin resistance.

With the sedentary lifestyles and reduced sleep hours becoming an increasingly common lifestyle, few studies have pursued the lasting effects from the workweek on the subsequent weekend. Broussard et al. (Broussard et al., 2016) showed that two nights of recovery sleep following four nights of 4.5 hours sleep restored insulin sensitivity; whereas others (Ness et al., 2019) have noted that two nights of recovery sleep from five nights of 5 hours of sleep did not restore the loss in insulin sensitivity, but was effective in restoring lipemic responses. We observed that weekend recovery improved HOMA-IR compared to baseline, due to the reduced fasting glucose and insulin concentrations. Interestingly, the improvements in insulin resistance were independent of performing exercise during SR and occurred while the subjects simultaneously reduced step counts and increased sedentary time during the weekend. Weekend recovery sleep was also beneficial at restoring fasting NEFA concentrations. While these improvements suggest a strong influence of recovery sleep independent of exercise and weekend activity, Depner et al. (Depner et al., 2019) recently demonstrated that improvements during the weekend in insulin sensitivity were not protective against recurrent sleep loss the following week, highlighting the profound impact of sleep on glucose metabolism, and the need to investigate the long-term consequences of modest restricting sleep.

Additionally, inflammation may provide some underlying mechanisms to insulin resistance during SR (Wright et al., 2015). Reduced and fragmented sleep increases circulating proinflammatory cytokines, but discrepancies exist between the responses reported during sleep restriction (Haack, Sanchez, & Mullington, 2007; van Leeuwen et al., 2009; Vgontzas et al., 2004) and also with circadian misalignment (Wright et al., 2015). In the present study, no differences in fasting concentrations of TNFα or IL-6 concentrations were found with either study condition, as well as no correlation with indices of insulin sensitivity/resistance. This lack of inflammatory change may be due to the modest SR and duration utilized in this overweight population. Chronic and more severe sleep curtailment may be needed to influence inflammatory profiles.

Measures of depression, anxiety, and stress were elevated in response to a modest SR throughout a workweek. Despite PA being known to have positive effects on psychological health, moderate-intensity exercise throughout the sleep-restricted workweek was not effective at reducing the stress related to sleep loss. Additionally, stress scores were not correlated with any circulating markers or indexes of insulin sensitivity, suggesting that the subjects’ emotional state was not influencing insulin sensitivity.

Sleeping in a free-living environment can confound the data as it can be influenced by sleep/wake habits of others in the house, the environment is less controlled, especially with nighttime light exposure. Additionally, oral glucose tolerance tests, intravenous glucose tolerance tests, and insulin clamps are commonly used techniques to assess insulin sensitivity. The MMT, however, provides a physiological response that would be observed in the free-living situation.

Short-term modest SR does not overtly cause metabolic disturbances in overweight/obese adults, but SR increased perceived stress and increased fasting NEFA concentrations and increased fat oxidation. Ad libitum sleep during the recovery weekend effectively restored the metabolic fluctuations induced by SR despite increased sedentary time and reduced steps. Further the profound negative effects of sleep loss was not mitigate by daily moderate-intensity exercise in overweight/obese individuals.

Abbreviations:

SR

sleep restriction

SREX

sleep restriction with exercise

NEFA

nonesterified fatty acids

MMT

mixed-meal tests

PA

physical activity

MVPA

moderate-vigorous PA

FAS

fatty acid synthase

AT

adipose tissue