In vitro studies.

NCI-H716 cells (from a male subject; American Type Culture Collection, Manassas, VA) are a well-validated model of the enteroendocrine L cell (3, 36). The cells were propagated in RPMI-1640 with 10% FBS, 100 IU/ml penicillin, and 100 μg/ml streptomycin. For gene expression, cells were synchronized 2 days after plating on Matrigel by overnight starvation in RPMI with 0.5% FBS, followed by a 1-h shock with 10 μM forskolin in RPMI with 10% FBS. After the shock, the cells were washed and incubated in RPMI with 10% FBS; this was considered time 0. Total RNA was then extracted every 6 h for 42 h using the RNeasy Plus Mini kit with Qiashredder (Qiagen, Germantown, MD). Clock gene, glucose-dependent insulinotropic peptide (GIP) receptor, and melatonin receptor mRNA levels were determined by qRT-PCR, using Hist1H3a as the internal control, as described (18). The following primers (Invitrogen, Burlington, ON, Canada) were used: Arntl, Hs00154147_m1; Per2, Hs00256143_m1; Hist1H3a, Hs00543854_s1; GIPR, Hs00609210_m1; Mtrn1a, Hs00195567_m1; and Mtrn1b, Hs00173794_m1. For secretion studies, cells were plated, synchronized as above, and at 0, 24, 36, and 48 h after synchronization were incubated for 2 h in RPMI with 0.5% FBS alone or containing 10⁻³ M bethanechol (Sigma-Aldrich) or 10⁻⁷ M GIP (Bachem Americas, Torrance, CA). Total media and cell GLP-1 content were determined by RIA (Millipore, Billerica, MA), as described (18).

Subjects and experimental design of the human study.

The protocol was approved by the Research Ethics Board of Mount Sinai Hospital, Toronto, ON, Canada. Eight healthy male volunteers (18–26 yr old) were recruited by public advertisement at the University of Toronto; all gave written, informed consent and self-reported that they were nonsmokers, were on no medications, and had no known medical conditions or sleep disorders. Volunteers were admitted to the Toronto Centre for Advanced Reproductive Technology (TCART) on four consecutive Fridays at 1700 for 21 continuous hours. Participants were asked to not drink alcohol within 24 h or eat within 3 h of admission and to follow a regular sleep schedule (1100–0800 ± 1 h) during the preceding 5 days. The subjects' anthropometric characteristics are shown in Table 1. Four different conditions were tested with subjects being randomized between two test (T) conditions each week: T1 or T2 in weeks 2 and 4 and T3 and T4 in weeks 1 and 3 (Fig. 2).

In T1 (normal light-dark cycle with regular sleep), volunteers were maintained under standard bright light (600 lux) until 2200, when the lights were switched off, and subjects remained under dim-red light (dark, <2 lux) until 0800, when the regular bright lights were turned on until the end of the study. During the periods of wakefulness, all the volunteers remained in the same room, seated around a circular table, while playing board games and/or maintaining lively conversations with each other. At least two researchers were in the room at all times to collect blood samples and ensure compliance with the wakefulness requirements. At midnight, volunteers were moved to another room and allowed to sleep in a bed, wherein they fell asleep within 20 min and slept continuously until 0600, as visually monitored by two investigators who remained in the room. All the beds were located in the same room but were separated individually by a black curtain to grant some privacy to the volunteers. Participants were never wakened during blood sampling.

In T2 (sleep deprivation in the dark), participants were kept under the same light conditions as above but remained under wakefulness conditions throughout, as described above.

In T3 (sleep deprivation with nocturnal light exposure) and T4 (sleep deprivation with nocturnal filtered-light exposure), participants were kept under standard bright light and remained awake throughout, as described above. At 2200, the volunteers donned optical filters, which were not removed until 0800, to mimic the timing of lights-off and -on in T1 and T2. In T3, the optical filters were placebo and allowed the transmission of wavelengths from the entire light spectrum; in T4, the glasses completely filtered wavelengths <480 nm (≈0% transmission).

Of the eight volunteers that began the study, two withdrew after the first day and were replaced; thus these replacement subjects did not complete one study each (T3 and T4, respectively). One of the original participants also missed one of the studies (T4), and one of the substitutes was later excluded from the study due to noncompliance with meal requirements. As a result, six subjects were included in the statistical analysis of T1–T3 whereas, to reduce the risk for type 2 errors in the analysis, the data from T4 were not included due to the failure of one remaining subject to complete this study.

Upon admission to TCART, an intravenous catheter was inserted in the forearm by a certified phlebotomist. Basal blood samples (4 ml) were taken every 2–3 h into heparin-coated tubes for melatonin and cortisol measurements. Two identical standard meals (each consisting of a combination of two convenience food dishes; Table 2) containing a total of 850 kcal (53.7% carbohydrate, 26.5% fat, and 19.8% protein) were offered at 2300 and 1100 to be consumed within 10 min [meal tolerance test (MTT)]. Blood samples (2 ml) were collected into 200 μl of 0.1 mM diprotin A (Sigma-Aldrich, Oakville, ON, Canada), 5,000 KIU/ml aprotinin (Sigma-Aldrich), and 0.03 M EDTA immediately before and every 10–30 min for 180 min after the meal. Plasma was used for insulin, GLP-1, and glucose measurements. In addition, orange-flavored glucose drinks (300 kcal) were provided at 1800 and 0600 to fulfill the daily energy requirements of the participants (2,300 kcal/24 h) and to preclude any confounder effects of different fasting periods prior to the meal tolerance tests (4). Blood samples were also taken before and after consumption of the glucose drinks.

Assays and measurements.

Plasma glucose was determined by enzymatic assay (Analox Instruments, Lunenburg, MA). Plasma insulin and total GLP-1, melatonin, and cortisol were measured using kits from MesoScale Discovery (Gaithersburg, MD), Rocky Mountain Diagnostics (Colorado Springs, CO), and DRG International (Springfield, NJ), respectively.

Calculations and statistical analysis.

Data are expressed as means ± SE. Areas under the curves (AUC) were calculated by the trapezoid rule. Insulin resistance was estimated by two different surrogate measuresments: the homeostatic model assessment (HOMA-IR) index, calculated from fasting plasma insulin (FPI) and fasting plasma glucose (FPG) concentrations [FPI (mU/l) × FPG (mmol/l)/22.5], and Matsuda's composite index, calculated using both fasting and postprandial concentrations of insulin and glucose [10,000/(FPI × FPG × mean insulin levels during MTT × mean glucose levels during MTT)1/2]. Because the relationship between insulin and glucose concentrations during the fasting state reflects mainly the balance between hepatic glucose output and insulin secretion, it is generally accepted that the HOMA-IR is an indicator of hepatic insulin resistance (28, 29, 39, 46). However, during the postabsorptive state, insulin sensitivity equally reflects both suppression of hepatic glucose production and glucose disposal by all tissues in the body, and therefore, Matsuda's composite index is generally considered an indicator of whole body insulin sensitivity (28, 39). Accordingly, whole body insulin resistance is estimated by calculating the inverse number of Matsuda's composite index. To compare the different groups in the volunteers, either one-way or two-way ANOVA with repeated measurements were conducted, as appropriate; the in vitro data were analyzed by one-way ANOVA. Homoscedasticity of the data was checked by the Bartlett's test, and a Holm-Sidak multiple comparison test was run for post hoc analysis. Significance was set at P < 0.05. GraphPad Prism 6 software was used to perform the statistical analyses.

Rhythmic activity of the human enteroendocrine L cell in vitro.

The existence of a cell-autonomous independent clock in the human L cell was first established in the NCI-H716 cell line, which demonstrated rhythmic oscillations in the mRNA levels of two canonical clock genes, Bmal-1 (Fig. 1A) and Per-2 (Fig. 1B), over a 42-h period. These patterns were more robust during the first 24 h and waned over time, which was likely due to loss of synchrony, as reported for other endocrine cells (27). A variation in the GLP-1 secretory response to two well-established L cell secretagogues, bethanechol and GIP (3, 40), was also observed in these cells such that the GLP-1 response to identical bethanechol doses was significantly higher at times 24 and 48 h after synchronization than at times 0 and 36 h (P < 0.01; Fig. 1C). Similarly, GLP-1 secretion induced by the same concentrations of GIP was also significantly higher at times 24 and 48 h than at 36 h after synchronization (P < 0.05-0.01; Fig. 1D). Finally, to determine whether GLP-1 secretion might be modulated by melatonin, a key circulating hormone that follows a circadian rhythm, the presence of mRNA transcripts for the receptors Mel1A and Mel1B was examined in the NCI-H716 cells. Compared with the GIP receptor, which was readily detectable, no amplification was detected for either of the melatonin receptors (Fig. 1E). Together, these findings suggest that the existence of a cell-autonomous clock in the human L cell that may govern rhythmic secretory activity independently of melatonin signaling.

Differential effects of acute sleep deprivation with or without nocturnal light exposure on the daily profile of melatonin and cortisol levels.

As expected, control subjects maintained under a normal light-dark cycle (T1; Fig. 2) demonstrated a bell-shaped curve in circulating melatonin levels, with levels rising to a plateau between 0200 and 0800 and quickly returning to baseline once the lights were turned on. No difference in the melatonin profile was observed between these individuals and those who were sleep deprived in the dark (T2). In contrast, nocturnal light exposure (T3) significantly reduced (P < 0.01–0.05) the nighttime rise (from 2300 to 0600) in melatonin levels (Fig. 3A).

Control subjects exhibited the expected cortisol pattern, with the lowest levels detected at the beginning of the night and a peak in the early morning, followed by a rapid decline (Fig. 3B). The morning peak was significantly reduced (P < 0.01 at 0600) following sleep deprivation, irrespective of the light conditions, and cortisol levels remained steadily elevated afterward, resulting in significantly augmented circulating levels at 1400 (P < 0.05). Subjects under continuous light exposure also showed significantly (P < 0.05) elevated cortisol levels during the early night (0200) compared with the normal sleep group, as reported previously (35).

Differential effects of acute sleep deprivation with or without nocturnal light exposure on basal insulin, glucose, and GLP-1 levels.

Basal levels of all parameters were determined at 2300, 0600, and 1100; these time points were selected to allow ≥5 h after the last caloric ingestion in order for insulin and GLP-1 levels to return to baseline (14, 24, 45). When the participants were maintained under the regular light-dark cycle and allowed to sleep, a diurnal variation in insulin basal levels was found (Fig. 4A), with concentrations peaking at 0600 (P < 0.001). No significant differences in basal glucose levels were observed (Fig. 4B), although significantly elevated HOMA index values were detected at 0600 (P < 0.001; Fig. 4C), consistent with the elevated insulin levels. Sleep deprivation in the dark significantly reduced insulin concentrations (P < 0.01) and HOMA values (P < 0.05) at 0600, and the diurnal variation in these parameters was suppressed. The profiles of basal insulin levels and HOMA index values exhibited by the sleep-deprived group under nocturnal light exposure were comparable with those observed in the sleep-deprived subjects in the dark, with only small increments that did not reach significance. Interestingly, and consistent with the profile in insulin basal levels, a peak in GLP-1 concentrations was also found at 0600 (P < 0.001; Fig. 4D). Although this peak was consistently observed under all different conditions, the magnitude was reduced significantly (P < 0.05) in the subjects exposed to continuous light compared with the control group.

Differential effects of acute sleep deprivation with or without nocturnal light exposure on the insulin, glucose, and GLP-1 responses to meal ingestion.

The time course profiles of insulin (Fig. 5, A and B), glucose (Fig. 6, A and B), and GLP-1 (Fig. 7, A and B) were examined in response to identical meals at 2300 (under dark conditions in both the control and sleep-deprived in the dark groups but under light conditions in the nocturnal light group) and 1100 (under bright light conditions in all groups). When the volunteers slept normally in between the meals, the expected rhythm in insulin and glucose responses was observed (17, 21, 47). Hence, the first phase (30 min) of the insulin response was significantly enhanced (by 54%) in the morning (P < 0.05; Fig. 5C), in association with significantly better glucose tolerance (by 14%, P < 0.05; Fig. 6C), compared with the night. In contrast, the diurnal variation in the first phase of insulin response was not detected in subjects who were sleep deprived in the dark (Fig. 5C), although the variation in glucose tolerance was maintained in these participants (13% difference, P < 0.05; Fig. 6C). Furthermore, although sleep deprivation per se did not affect the integrated (120 min) insulin response at any time, it was significantly affected by light (Fig. 5D) such that, at 2300, when the subjects had been exposed to only 1 h of nocturnal light, the AUC of the insulin response was significantly increased by 68% (P < 0.05) compared with the same meal served under dark conditions. This potentiation of the integrated insulin response was further magnified when the meal was served at 1100, after an entire night under continuous light exposure, such that the insulin response was enhanced by 92% (P < 0.01). However, this enhancement in the insulin response did not result in any improvement in glucose tolerance at 2300, and indeed, glucose levels were not appropriately decreased at 1100 (Fig. 6C), which was likely due to significantly elevated insulin resistance at both time points (P < 0.01–0.001; Fig. 6D).

A variation in the first phase of postprandial GLP-1 responses (30 min) was also observed in the control subjects, although surprisingly, the changes were opposite with respect to the pattern observed in the insulin response. Thus, the AUC of the GLP-1 response was significantly higher (by 23%, P < 0.05) at 2300 compared with 1100 (Fig. 7C). This variation was maintained in the subjects with sleep deprivation under dark conditions (28% difference, P < 0.01) but was lost when the participants were exposed to nocturnal light. Consistent with the potentiation of insulin responses induced by nighttime light, the integrated (120 min) GLP-1 postprandial responses were also augmented in response to nocturnal light, reaching significance at 1100 (by 25%, P < 0.05; Fig. 7D).

Finally, the insulin, glucose, and GLP-1 responses to an oral glucose tolerance test (OGTT) administered at 0600 were also studied in control and sleep-deprived subjects under both dark and nocturnal light conditions; however, no significant differences in any of these parameters were found between the different conditions tested (Table 3).