Long-Term Night Shift Disrupts Daily Body Rhythms and Metabolism More Than Constant Nighttime Light in Female Day-Active Rodents

The study was performed on females of the diurnal Sudanian grass rat (Arvicanthis ansorgei) born and bred at the Chronobiotron (UAR 3415, CNRS and University of Strasbourg, France). Animals were 9‐month‐old at the start of the experiment. Each Arvicanthis was individually housed in a type III cage enriched with a gnawing wooden stick, nesting frill, and ad libitum access to water and food (Rodent chow, SAFE 105 diet, SAFE, Augy, France). Room temperature was maintained at 23°C ± 1°C, and the standard light condition (150 lx) was defined as a light–dark (LD) 12 h:12 h cycle with lights on at 07:00 (Zeitgeber (ZT)0) and off at 19:00 (ZT12).

All experiments were performed at the Chronobiotron (UAR3415) in accordance with the Guide for the Care and Use of Laboratory Animals and the French National Law (implementing the European Union Directive 2010/63/EU), and they were approved by the Regional Ethical Committee of Strasbourg for Animal Experimentation (CREMEAS) and authorized by the French Minister of Higher Education, Research and Innovation (APAFIS #34226‐2021120314452587 v4). Every effort was made to minimize pain and discomfort and reduce the number of animals used.

Three lighting protocols were tested during these experiments (Figure): first, control (i.e., standard light exposure (LD 12:12); second, Chronic Night Shift (CNS)) during which standard light exposure was modified with a 10‐h delay of the light phase (150 lx), on four consecutive days per week (i.e., from Monday to Thursday), the last 3 days of the week (i.e., from Friday to Sunday) being back to the standard condition following a 10‐h light phase advance; and third, chronic Light At Night (LAN) during which the standard light exposure was modified with the addition of 6‐h of 150‐lx light (between ZT15 and ZT21) during the night phase on four consecutive days per week (i.e., from Monday to Thursday) while on the last 3 days of the week, (i.e., from Friday to Sunday), light exposure was normalized, returning to the standard condition (LD 12:12). These weekly protocols were repeated over 10 weeks. S1A

At the end of the 10‐week lighting protocols, Arvicanthis were sacrificed at four time points every 6 h over 24 h: ZT2, ZT8, ZT14 and ZT20. Animals (n = 6 for each ZT per group, except n = 7 for the control group at ZT2 and ZT20, and the CNS group at ZT14, and n = 5 for the CNS group at ZT20; n total = 74) were deeply anesthetized by intraperitoneal injection of a mixture of 50 mg/kg zolazepam (Zoletil, Virbac, Carros, France) and 10 mg/kg xylazine (Paxman, Virbac). The depth of anesthesia was verified by the absence of paw and eyelid withdrawal reflex following strong pressure. After decapitation, approximately 2 mL of blood were collected in a tube containing 50 μL of 4% EDTA (ethylenediaminetetraacetic acid, EU0007‐B, Euromedex, Souffelweyersheim, France; used as anticoagulant), kept on ice, then subsequently centrifuged at 2500 rpm for 30 min at 4°C to recover the plasma. Two samples were taken from the right lobe of the liver, rinsed in 0.9% NaCl solution, and collected in a tube that was immediately frozen in liquid nitrogen. The brain was dissected and frozen at −30°C in isopentane (2‐methylbutane). All samples were then stored at −80°C.

Plasma samples were analyzed with colorimetric kits assessing the concentration of various plasma metabolites: total cholesterol (80106, Biolabo, Maizy, France), High Density Lipoproteins (HDL) cholesterol (90206, Biolabo), Low Density Lipoproteins (LDL) cholesterol (90816, Biolabo), non‐esterified fatty acids (NEFA, 434‐91795, Fujifilm Wako, Neuss, Germany), triglycerides (87319, Biolabo), glucose (87409, Biolabo) and albumin (80002, Biolabo). Plasma insulin was measured by Enzyme‐linked immunosorbent assay (Rat/Mouse insulin ELISA EZRMI‐13K, Merck Millipore, Molsheim, France). For each circulating molecule analyzed, the standard curve was determined from standard solutions whose concentrations were supplied by the assay kit, quality controls were used to verify calibration, and samples were deposited in duplicate and averaged. These assays were analyzed using a spectrophotometer (Multiskan FC, ThermoFisher).

Glycated hemoglobin (HbA1c) and glycated albumin are both used for monitoring chronic hyperglycaemia in humans, notably for diabetic patients [19, 20]. HbA1c and glycated albumin, with their different half‐lives, reflect glycaemia over the past 2 to 3 months, and 2 to 3 weeks, respectively [21]. The proportions of native, glycated and cysteinylated forms of albumin and hemoglobin were evaluated by liquid chromatography‐mass spectrometry (LC–MS). After the samples were placed in denaturing conditions (acidification), glycated and cysteinylated proteoforms can indeed be identified by MS as a glycation event causes a 162 Da increase in the mass of a native protein, and a cysteinylation event causes a 119 Da increase. For hemoglobin, 199 μL of water containing 0.1% of formic acid were used to dilute 1 μL of red blood cells. For albumin, 3 μL of plasma were diluted in 22 μL of acidic water (0.1% formic acid). For each type of sample, 5 μL were injected onto a column of reverse‐phase chromatography (Vydac 208TP C8; i.d. 2.1 × 250 mm, 300 Å, 5 μm particle size; Grace, Columbia, MD, USA) using an Agilent 1200 Series HPLC system (Agilent Technologies, Palo Alto, CA, USA) coupled to a quadrupole‐time‐of‐flight (Q‐TOF) mass spectrometer (maXis II, Bruker Daltonik GmbH, Bremen, Germany). Detailed information about our nanoLC‐MS/MS method can be found in a previous work [22]. Areas under the curves of the extracted ion chromatograms of the 10 most intense charge states were used to calculate the relative abundance of native, glycated, and cysteinylated albumin forms. Results for each of the three forms are presented as a percentage of the sum of all forms.

Data are expressed as means ± standard error of the mean (SEM). Statistical analysis was performed using SigmaPlot 13.0 (SigmaPlot software, Jandel Scientific, Chicago, IL, USA). The normal distribution of the data was assessed by a D'Agostino & Pearson or Shapiro–Wilk normality test, depending on the sample size. Parameters (body mass, frequency and number of meals, quantity of food intake, locomotor activity) were analyzed by two‐way analyses of variance (ANOVA) with repeated measures (RM), followed by a Holm‐Šidák multiple comparison test where appropriate. Plasma and qPCR analyses were compared using 2‐way ANOVA followed by a Holm‐Šidák multiple comparison test where appropriate. Several parameters (body mass gain, blood glucose) required 1‐way ANOVA with or without RM (fasting and end of IPGTT blood glucose in the CNS group). When the normality test failed for 1‐way ANOVA, we used Kruskal–Wallis 1‐way ANOVA on ranks, followed by a post hoc Dunn's test. When the normality test failed for 1‐way RM ANOVA, we used Friedman RM ANOVA on ranks, followed by a post hoc Dunn's test. Daily rhythmicity of food intake, plasma parameters and gene expression was analyzed using cosinor regression with the following equation:Significant mean levels (A), amplitudes (B) and acrophases (C) of these rhythms were compared using 1‐way ANOVA followed by a post hoc test, as above. Daily rhythmicity was considered significant only when the three parameters (mesor, amplitude, and acrophase) exhibited significant differences at a threshold set at p ≤ 0.05.

Actimetry analysis (Figure 1A–C) shows that CNS led to a disruption in locomotor activity rhythm, which can drift with a period larger than 24 h (Figure 1B), while LAN just triggered a bout of activity during nocturnal light exposure (Figure 1C). On average, the period of rest‐activity rhythm was lengthened by CNS, while it remained very close to 24 h in both LAN and control animals (respectively, 24.31 ± 0.25 vs. 23.98 ± 0.01 and 24.02 ± 0.02 h, Kruskal‐Wallis 1‐way ANOVA on ranks, H(2) = 8.85, p = 0.01). Interdaily stability was drastically reduced during the first week of CNS exposure, and this persisted after 1 month with no further deterioration. Interdaily stability was also significantly impaired from the first week (and after 1 month) of LAN exposure, although this decrease was more moderate than that with CNS (2‐way RM ANOVA, F(4, 42) = 8.41, p < 0.001, Figure 1D). Total activity over 24 h did not vary between groups, nor over the course of the experiment. However, the percentage of daytime locomotor activity (last day of the week) was lower after 1 month of CNS than after 1 month of LAN (respectively 56% ± 5% vs. 78% ± 5%, 2‐way RM ANOVA, F(2, 42) = 4.73, p = 0.02, Figure 1E).

Three series of recording in calorimetric cages (last day of baseline week, first week of light exposure, after 1 month of light exposure) were compared. Food intake followed a daily rhythm (Figure 2A–C), with an acrophase at ZT7 for the three groups during the baseline period (cosinor regression, p < 0.05). This rhythmicity was preserved in the control group and after LAN exposure, but was lost after 1 month of CNS exposure (cosinor regression p > 0.05, Figure 2B). The acrophases of the feeding rhythm are detailed in Table 1. While the food intake schedule remained stable in control and LAN‐exposed individuals, a 4‐h phase delay was detected following the first week of exposure to CNS (1‐way ANOVA, F(2, 21) = 90.94, p < 0.001), with a daily peak of food intake close to ZT11. The acrophase of the CNS group could not be analyzed after 1‐month exposure, because food intake was no longer rhythmic. The amplitude of the rhythm, compared with its own baseline value, was slightly reduced in the control group during the second passage in the Phenomaster cage (i.e., first experimental week; 1‐way ANOVA F(2, 21) = 4.35, p < 0.03), whereas it increased drastically after exposure to LAN (1‐way ANOVA F(2, 21) = 44.15, p < 0.001). In contrast, rhythm amplitude in the CNS group remained stable between baseline and the first week of CNS.

Analysis of the quantity of food ingested over 24 h and the quantity ingested during daytime showed no difference between the groups, nor over time. However, the ratio of daytime intake to total 24‐h intake decreased in CNS‐exposed animals (2‐way RM ANOVA, F(4, 42) = 2.68, p = 0.04, Figure 2D). At the end of the first week of light exposure, individuals in the CNS group had a 10% reduction in this ratio compared with the control and LAN groups. This reduction became larger after 1 month, more than half of the food (60%) was consumed at night by the CNS‐exposed animals.

The number of meals per 24 h remained stable and consistent among the different groups and over time. However, the ratio of the number of daytime meals to the total number of meals over 24 h decreased after CNS exposure from the first week to 1 month of light exposure (2‐way RM ANOVA, F(4, 42) = 3.32, p = 0.02, Figure 2E).

All these findings indicate the strong impact of the CNS‐induced circadian desynchronization leading to a flattened daily rhythm in food intake, with more food taken at night rather than during daytime. In contrast, exposure to LAN did not disrupt the feeding/fasting rhythm.

Analyses of feeding data were detailed day by day to assess possible adaptation over the experimental week. During the baseline period, daytime meal ratios (between ZT0 and ZT12) for all three groups were stable over the week, with no change along the week days (p > 0.05, Figure S2A). During the first week of CNS exposure, animals showed a decrease in daytime meal ratio between day 2 and day 4, and this decrease persisted on days 6 and 7, during which animals were back to standard light–dark cycle (Figure S2B). This indicates that there was no resynchronization of the daily rhythm in food intake after CNS exposure. Also, during the first week of LAN exposure, the ratio decreased until day 4 (lowest ratio), but then increased to day 7 (last experimental day back to standard light–dark cycle), showing a rapid resynchronization of the daily rhythm in food intake in LAN‐ exposed individuals (2‐way RM ANOVA, F(10, 105) = 2.93, p = 0.004, Figure S2B). After 1 month of experimental conditions, the ratio of the number of daytime meals was lower over the whole week after CNS exposure compared with control or LAN exposure (respectively 0.42 ± 0.025 vs. 0.62 ± 0.025, 0.65 ± 0.025, 2‐way RM ANOVA, F(2,105) = 24.21, p < 0.001, Figure S2C). These detailed analyses along the week days demonstrate the inability of CNS‐exposed individuals to resynchronize quickly, but also in the long term, since their daily ratios after 1 month of light treatment remained lower than in the 2 other (i.e., LAN‐exposed and control) groups. By contrast, LAN‐exposed individuals were able to resynchronize to the LD cycle at the end of each experimental week.

During the 10 weeks of monitoring, individuals from all groups exhibited a progressive body mass gain (2‐way RM ANOVA, F(9,419) = 120.09, p < 0.001), but differently according to the lighting exposures (2‐way RM ANOVA (interaction), F(18, 419) = 2.97, p < 0.001, Figure 3A). Compared to the control group, both CNS‐ and LAN‐exposed Arvicanthis gained more body mass, although there was no difference between them (12.9% ± 1.0% and 13.3% ± 1.4% vs. 8.5% ± 1.4% respectively, 1‐way ANOVA, F(2, 47) = 4.55, p = 0.02, Figure 3A). By contrast, adiposity estimated by the ratio of perigonadal fat mass to body mass assessed in individuals of the actimetry cohort was not significantly affected between lighting groups (2.63% ± 0.39% in n = 7 CNS, 2.22% ± 0.20% in n = 8 LAN, and 2.26% ± 0.22% in n = 9 Control; 1‐way ANOVA, F(2, 21) = 0.64, p > 0.1).

Fasting blood glucose levels increased continuously following CNS exposure. Indeed, after only 1 week, fasting blood glucose was increased, and this increase was amplified after 1‐month CNS exposure (1‐way RM ANOVA, F(2, 32) = 7.33, p = 0.002). On the other hand, individuals in the control and LAN groups experienced no change in fasting blood glucose levels, so that after 1 month of light treatment, their fasting blood glucose levels were lower than those of the CNS‐exposed group (Figures 3B,C and S3A).

During the baseline period, individuals of the 3 groups showed similar responses to the glucose tolerance test (p > 0.05, Figure 3B). After 1 month under control or LAN conditions, this response remained similar to their respective baseline responses, with no difference between these two groups. By contrast, individuals exposed for 1 month to CNS had poorer glucose tolerance, and their blood glucose levels were significantly higher at t30, t60 and t120 compared with control and LAN conditions (2‐way RM ANOVA (interaction), F(10, 217) = 2.20, p = 0.02, Figure 3C). Blood glucose levels at the end of the test (t180) were stabilized for both control and LAN groups, while in individuals exposed to CNS for 1 month, blood glucose levels at t180 were higher than during baseline (115.0 ± 11.3 vs. 74.6 ± 6.3 mg/dL, respectively; Friedman test, Q(2) = 11.4, p = 0.003, Figure S3B).

In the present study, no HbA1c was detected in any of the three groups. Conversely, the proportion of glycated albumin could be measured in the different groups, revealing significantly higher rates in the LAN group (21.9% ± 0.6%) compared with both the control group (17.0% ± 0.6%) and the CNS group (17.9% ± 0.6%; (2‐way ANOVA, F(2, 61) = 20.44, p < 0.001), Figure 5C). By contrast, there was no difference between the 3 lighting groups in daily levels of plasma glucose or insulin in fed Arvicanthis (2‐way ANOVA, F(2, 55) = 1.60, p = 0.21 and F(2, 53) = 0.12, p = 0.89, respectively; Figure S3C–E).

All results below for the daily rhythms in the plasma, brain and liver were obtained on samples after 10 weeks of exposure to LAN, CNS or regular light–dark cycle. Total cholesterol levels were higher in CNS‐ and LAN‐exposed individuals than in controls (2‐way ANOVA, F(2, 57) = 10.50, p < 0.001), but only in LAN‐exposed individuals plasma total cholesterol levels showed a daily rhythm with a peak at ZT20.5 ± 1.2 h (cosinor regression, Figure 4A). There was no difference in total HDL cholesterol levels between the different groups (2‐way ANOVA, F(2, 57) = 1.19, p = 0.31), but plasma HDL cholesterol was rhythmic only in control and LAN‐exposed animals, with a nocturnal acrophase at ZT18.4 ± 1.5 h and ZT21.3 ± 1.7 h, respectively (cosinor regression, Figure 4B). Mean levels in plasma LDL cholesterol were higher in individuals exposed to CNS than in controls (0.62 ± 0.04 vs. 0. 50 ± 0.03 mmol/L, respectively; 2‐way ANOVA, F(2, 55) =3.35, p = 0.04) while levels being intermediate in LAN‐exposed animals (0.57 ± 0.03 mmol/L, Figure 4C). Furthermore, plasma LDL cholesterol did not exhibit daily rhythmicity in any of the three groups. Plasma triglyceride levels showed a daily rhythm with a peak at ZT18.5 ± 0.9 h only in CNS‐exposed individuals (cosinor regression, Figure 4D), but no difference in triglyceride levels was detected between the different groups (Kruskal‐Wallis 1‐way ANOVA on ranks, H(2) = 1.48, p = 0.48). Plasma NEFA did not show daily rhythmicity in any of the three groups, or difference in mean levels (Kruskal–Wallis 1‐way ANOVA on ranks, H(2) = 3.43, p = 0.18, Figure 4E).

Total plasma albumin was rhythmic in control Arvicanthis, with nocturnal acrophase at ZT17.1 ± 1.5 h, and arrhythmic in animals exposed to 10 weeks of CNS or LAN, while mean levels of plasma albumin did not differ between the three groups (2‐way ANOVA, F(2, 57) = 0.30, p = 0.74, Figure 5A). Subsequently, an analysis of post‐translational modifications of plasma albumin was performed by mass spectrometry on each sample (Figure 5B–D). In the control group, in contrast to the higher nocturnal concentrations of total albumin, the percentages of native albumin showed a daily rhythm with higher daytime values, the acrophase being at ZT5.5 ± 0.5 h. Glycated albumin and cysteinylated albumin percentages also showed daily rhythms in the control group, with higher values at night (acrophase at ZT17.0 ± 1.1 h and ZT18.2 ± 1.1 h, respectively), in other words, out of phase with the rhythmicity of native protein and in phase with total albumin levels. In animals exposed to CNS or LAN, as for plasma albumin, there was no daily organization of post‐translational variations in both native, glycated and cysteinylated forms. Furthermore, the distribution of the three forms of albumin differed between lighting groups, with a marked effect of LAN. Native albumin percentages in LAN‐exposed Arvicanthis (40.4% ± 0.9%) were reduced compared to control and CNS‐exposed animals (46.2% ± 0.8% and 47.4% ± 0.9%, respectively; 2‐way ANOVA, F(2, 61) = 17.75, p < 0.001). Conversely, glycated albumin percentages in LAN (21.9% ± 0.6%) were increased compared to control and CNS groups (17.0% ± 0.6% and 17.9% ± 0.6%, respectively; 2‐way ANOVA, F(2, 61) = 20.44, p < 0.001). Cysteinylated albumin percentages in the LAN group (37.6% ± 0.7%) were also larger than in the CNS group (34.8% ± 0.7%), but not different from the control group (36.7% ± 0.7%; 2‐way ANOVA, F(2, 61) = 4.46, p < 0.02).

Previously, we examined the effects of CNS exposure on rest‐activity rhythm and metabolic function in young male Arvicanthis [9], allowing assessment of sex differences in the metabolic and circadian effects of CNS exposure in Arvicanthis. Thus, deeper metabolic alterations (i.e., increased body mass gain, increased fasting glucose) were observed in females compared to male Arvicanthis. It should be noted, however, that the female Arvicanthis studied here were 9 month‐old and exposed to CNS for 10 weeks, while younger males (2–3 months) were exposed to CNS during 12 weeks in the previous experiment [9]. Another study in Arvicanthis reported that following acute exposure to blue light at night, only males displayed impaired glucose response [30]. In nocturnal mice, exposure to a CNS differentially impacts females and males. Clock gene expression in liver was weaker in females and had an inverted daily pattern in males, while decreased glucose tolerance was observed only in males [31]. Studies in humans have also reported differential metabolic effects according to sex after acute circadian misalignment due to a single 12‐h jet‐lag protocol lasting 3 days. Among others, energy expenditure and lipid oxidation rate during simulated jet‐lag were increased in women, but remained unchanged in men. Fullness feeling during 12‐h jet‐lag was reduced in women, but not in men. By contrast, total sleep duration during 12‐h jet‐lag was reduced similarly in men and women [3]. Epidemiological studies reveal higher risk of lipid and glucose metabolism disorders in response to shift work in women than in men [32, 33, 34]. The comparison of the results in male and female Arvicanthis exposed to CNS are consistent with that conclusion, thus supporting the relevance of this diurnal animal model to assess the deleterious effects of shift work.

The rest‐activity rhythm, an indicator of the sleep–wake cycle, was affected differently by chronic LAN and CNS in Arvicanthis. Locomotor activity rhythm was drastically disrupted after CNS in females as in males [9]. During weeks or months of exposure to CNS, the light stimuli can occur at inappropriate times in relation to internal clocks and misalign the rest‐activity rhythm [35]. This disturbance was much more moderate in response to chronic LAN, although interdaily stability was also reduced. The present results in diurnal Arvicanthis show that the direct effects of light at night did not lead to desynchronization of locomotor activity, but rather to temporary activation during the light exposure. In the LAN‐exposed group, this light‐induced nocturnal activity disappeared during the three experimental days when the standard light/dark was restored, reinforcing the idea that LAN‐induced nocturnal activity represents a direct behavioral activation to light (so‐called positive masking) in diurnal Arvicanthis [35]. By contrast, in male nocturnal rats, acute LAN induced a transient reduction in locomotor activity (i.e., negative masking) [36], while chronic exposure of female nocturnal rats to dim light at night led to a loss in day‐night variations of locomotor activity [37].

The two light protocols used in the present study also resulted in different effects on feeding‐fasting rhythm of diurnal Arvicanthis. The daily rhythm of food intake was not altered after 5 weeks of LAN, whereas females exposed to CNS no longer showed any feeding rhythmicity. In terms of daily changes in the day to night ratio of meals, LAN‐exposed animals were able to resynchronize food intake rhythm to LD cycle before the end of each experimental week, while CNS exposure, which induced circadian misalignment, did not allow resynchronization of feeding‐fasting rhythm during the last 3 days of each experimental week. After 1 month of treatment, the ratio of number of daytime meals was stable in all three groups on the different days of the week. However, this ratio remained lower every day in animals exposed to CNS. Thus, exposure to CNS disrupts food intake rhythm in the short and long term in diurnal female Arvicanthis. In nocturnal female mice, chronic exposure to dim light at night led to reduced day‐night variations of feeding behavior [38]. Furthermore, in male mice exposed to CNS for 50 days or more, feeding activity was distributed similarly between dark and light hours, also highlighting a chronic circadian misalignment of feeding‐fasting rhythm [39, 40].

As expected in a diurnal species, the respective phases of oscillations of the clock genes Bmal1, Per1 and Per2 were very close between the SCN and liver in Arvicanthis, in agreement with data obtained in these same structures in the baboon, a diurnal monkey [41]. Furthermore, the molecular clock in the SCN was similarly phased in both diurnal and nocturnal rodents according to the light–dark cycle [42]. When compared between diurnal Arvicanthis and nocturnal mice, the rhythms of clock gene expression in the liver were shifted by about 6 h for Bmal1, Per1 and Per2 genes (around ZT15, ZT6, ZT6 vs. ZT20, ZT12, ZT13, respectively) [31, 43, 44]. Similar phase‐shifts have already been described for hepatic expression of clock genes between the diurnal baboons and nocturnal mice [41]. Consistently, phase‐shifts in the rhythmic expression of liver metabolic genes such as Pepck, Glut2, and G6pc are also found between diurnal Arvicanthis and nocturnal mice [31, 44, 45].

In the Arvicanthis SCN, chronic LAN had moderate long‐lasting effects on clock gene expression (including Per2, Bmal1, and Ciart), except for Per1 that lost its diurnal rhythmicity. In rat SCN, chronic LAN also affected Per1 expression by dampening its day‐night variations [46]. In sharp contrast, Arvicanthis exposed to CNS displayed altered SCN clock, with 4 clock genes no longer showing significant rhythmicity. Even though some of these effects may partly be due to the sampling procedure at fixed ZTs in desynchronized animals, potentially at different circadian times from each other, our data demonstrate that CNS strongly disrupts synchronizing conditions and misaligns the circadian timing system to local time. SYT10, which is highly expressed in the SCN, may play a role in regulating neuronal communication [23]. The absence of Syt10 gene rhythmicity following exposure to CNS and chronic LAN may impair intracellular SCN functioning.

The liver contains a powerful circadian clock, which is under the timing control of the SCN and is also synchronized by timing of food intake [47, 48]. In the liver of CNS‐exposed Arvicanthis, Per1, Per2 and Bmal1 genes display rhythmic expression in antiphase, while mean Per2 levels are lower and the amplitude of Per2 and Bmal1 rhythms is decreased. In nocturnal mice exposed to CNS, the hepatic expression of clock genes can also be inverted, or lose their daily rhythmicity [31, 43, 44]. Conversely, daily rhythmicity of clock genes after exposing Arvicanthis to LAN is preserved as in control animals. Previous studies have reported immediate effects of LAN (i.e., one single exposure to 1‐ or 2‐h light at night) on clock gene expression in the liver of nocturnal rats and diurnal Arvicanthis [49, 50], but these works are not directly comparable to the present study, which analyzed effects after chronic (10‐weeks) LAN conditions (i.e., 6‐h light at night for 4 days a week) and after 3 days back to standard light–dark cycle before tissue sampling.

Together, these data indicate that circadian misalignment after CNS results mainly from repeated LD shifts that compromise temporal communication between the SCN and peripheral clocks, leading ultimately to disturbed behavioral and metabolic rhythms. By contrast, LAN‐exposed Arvicanthis exhibited almost unaltered circadian clocks in the SCN and liver, which is consistent with their daily rhythms of locomotor activity and food intake remaining stable on the long term.

A number of metabolic genes, including Glut2, G6pc, Pepck, Gapdh, and Pparγ, lost their daily rhythmicity in the liver of Arvicanthis exposed to CNS. First of all, these data reveal an impact on multiple pathways involved in glucose and lipid metabolism. Second, they confirm earlier studies in nocturnal mice that also reported major disturbances in daily rhythmicity of metabolic genes in the liver after CNS [31, 44]. The hepatic rhythmicity of metabolic genes was also abolished in LAN‐exposed Arvicanthis. Thus, liver metabolic rhythms were altered whatever the disturbed light conditions, which contrasts with the differential effects of CNS and LAN on hepatic clock gene rhythms. This observation points to the direct deleterious effect of nocturnal light on liver metabolism. Both CNS and LAN disrupted daily rhythmicity of metabolic function in the liver, eventually bypassing the hepatic circadian clock, which remained essentially unaffected by LAN exposure. Desynchronizing light cues may alter liver metabolism via clock‐independent pathways, including first, autonomic afferent pathways; second, hormonal signals (e.g., glucocorticoids); and third, behavioral changes, such as arousal or feeding [49, 51, 52]. Although it does not provide clues to decipher the pathways involved, the present study demonstrates the long‐term loss of daily rhythmicity of metabolic genes in the liver after both CNS and LAN.

Daily patterns of circulating molecules reflect the deleterious effects of disturbed light conditions on hepatic function and its rhythmicity in Arvicanthis. Albumin, produced in the liver, is the most abundant protein in the blood. As illustrated in control Arvicanthis exposed to regular LD cycle, plasma concentrations of total albumin displayed daily variations, with higher levels during the resting (night) period. Interestingly, the daily rhythm of the native protein (peaking around midday) was phase‐shifted in comparison with total albumin rhythm (peaking in the middle of the night). By contrast, the daily rhythms of both glycated and cysteinylated forms peaked in the middle of the night, in phase with plasma variations of total albumin. Of note, the daily rhythms of total albumin and post‐translational forms were all lost after chronic LAN and CNS, revealing a major impact of disturbed lighting conditions on daily rhythmicity of circulating hepatic proteins. Interestingly, percentages of glycated albumin in LAN‐exposed Arvicanthis were increased compared to relative values in control and CNS groups, while percentages of cysteinylated albumin in LAN were also increased, but only compared to the CNS group. Glycated and cysteinylated albumin concentrations are significantly increased in diabetes and other oxidative stress‐related diseases [53, 54], their increased levels and/or daily arrhythmicity may participate to increased metabolic risk factors especially in LAN‐exposed Arvicanthis, and to a lower extent in CNS‐exposed animals.

Female Arvicanthis exposed to 10 weeks of CNS or chronic LAN showed larger weight gain than controls. This is in contrast to our previous study in male Arvicanthis who did not exhibit an increase in mass gain after 12 weeks of the same CNS protocol [9]. This result highlights a sex difference in the Arvicanthis metabolic response to circadian misalignment. In females, adiposity estimated by the ratio of perigonadal fat mass to body mass was not affected, suggesting that increased depots in other fat tissues, such as visceral or subcutaneous tissues, likely account for the larger mass gain. In nocturnal male mice, chronic exposure to dim light at night led to body mass gain [55]. Furthermore, CNS exposure in males of nocturnal mice and rats generally triggered body mass gain in the long term [39, 44, 56, 57]. These data suggest that the obesogenic effect of both LAN and CNS is not strictly dependent on the sex and temporal niche of locomotor activity pattern.

The Lipoprotein lipase (Lpl) gene, which encodes the rate‐limiting enzyme in triglyceride hydrolysis, is expressed in many peripheral tissues, including adipose tissue, skeletal muscle, and liver. Hepatic LPL plays an important role in plasma lipid homeostasis as its deletion in mice leads to increased plasma levels of triglycerides and cholesterol [58]. Daily rhythmicity of Lpl expression depends on the circadian clock [59]. In the present study, unexpectedly, the hepatic expression of the Lpl gene was rhythmic only in CNS‐exposed Arvicanthis, which could potentially be a response to regulate lipids and notably plasma triglycerides, albeit their levels were not specifically affected by CNS. In the brain, LPL plays a role in lipid sensing [60], and its expression in the SCN was rhythmic only in the control group. This suggests that daily variations of lipid sensing in the hypothalamus were altered in both LAN‐ and CNS‐exposed groups.

The regulation of cholesterol in Arvicanthis was altered by lighting exposure, especially with CNS. Plasma levels of LDL‐cholesterol were significantly increased by CNS exposure, while intermediate levels were found in the LAN‐exposed group as compared to control values. In male mice challenged with CNS, plasma concentrations of total cholesterol or LDL‐cholesterol were also increased [39, 57]. In a strain of mice prone to develop atherosclerosis (i.e., APOE*3‐Leiden.CETP strain), exposure to CNS accelerated hypercholesterolemia [61]. In humans, LAN has been associated with incident hypercholesterolemia in both men and women [62]. Together, these human and animal data highlight the deleterious effects of LAN and CNS on cholesterol metabolism, with a higher risk of atherosclerosis due to frequent increased concentrations of LDL‐cholesterol [28].

Acute exposure to white or green light at night in male rats impairs glucose response [36]. As aforementioned, Arvicanthis males, but not females, display impaired glucose responses following acute exposure to blue light at night [30]. The present data indicate that chronic LAN did not affect glucose tolerance or fasting blood glucose in female Arvicanthis. Of note, however, glycated albumin levels, which can be used as glycemic control markers [63], were higher in LAN than in the control and the CNS‐exposed groups. Glycated albumin levels therefore appear to be dissociated from fasting blood glucose values in LAN‐exposed Arvicanthis. The fact that total albumin concentrations were not notably different in the LAN group compared with the other groups suggests that altered albumin turnover is unlikely to account for this difference. In humans, reference values for glycated albumin range from 11% to 16%, with the diagnostic threshold for diabetes generally between 16% and 17% [63]. Although reference values are not available for Arvicanthis, these findings nonetheless appear to indicate poorer glycaemic control during the 2–3 weeks preceding sampling in LAN‐exposed animals. Although the IPGTT did not reveal altered glucose tolerance in these animals, it could be that the LAN condition induces an impaired regulation of postprandial glycaemia, resulting for instance in more frequent and higher postprandial glucose peaks.

In contrast to the LAN group, female Arvicanthis exposed to CNS develop impaired tolerance to glucose, indicative of a pre‐diabetic state, as already observed in male Arvicanthis [9] and male mice [39, 44]. Because IPGTTs were performed in early morning (ZT2), a time which corresponds to the lowest glucose tolerance in control Arvicanthis [9], it is very unlikely that the worst tolerance of CNS‐exposed animals could be attributed to different circadian timing compared to control Arvicanthis synchronized to the light–dark cycle. Furthermore, fasting blood glucose levels were higher in female Arvicanthis from the second week of CNS (this study), while they remain unaffected in males after 3 months of CNS [9]. Here again, the higher fasting glycemia in CNS‐exposed females might be explained by circadian variation rather than circadian disruption. In control Arvicanthis, the highest fasting glucose concentrations take place in late night (i.e., ZT19) [9]. Considering the drift in rest‐activity rhythm of CNS‐exposed Arvicanthis, we cannot fully exclude that a delayed rhythm of basal glucose levels in this group contributes to their higher fasting glycemia measured in early morning (ZT2). Nevertheless, our results show that diurnal rodents exposed to CNS exhibit impaired glucose tolerance in both sexes, but CNS‐exposed females also display fasting hyperglycemia, which is an additional metabolic risk for developing type 2 diabetes.

In conclusion, this research highlights the multiple deleterious effects of disturbed lighting conditions on circadian rhythmicity and metabolic health in a female diurnal rodent. The comparison with our previous work on males of this diurnal species [9] indicates both common and sex‐specific consequences of nocturnal light pollution. Although less harmful to metabolic health than CNS, chronic LAN in female Arvicanthis is still a health risk, because it is associated with body mass gain and abolition of liver and plasma metabolic rhythms. In addition, this study also highlights that only CNS triggers a prediabetic state, increases LDL‐cholesterol, and disturbs rest‐activity and feeding rhythms in female Arvicanthis. Thus, our results in this diurnal rodent are consistent with the role of the circadian system and its desynchronization (CNS) in the increased risks of developing diabetes, atherosclerosis, and obesity in shift workers.

In the context of a population widely exposed to nocturnal light, night work, shift work and chronic jet‐lag, it is important to take into account the present findings in a diurnal female animal model, in order to alert and prevent the deleterious effects of indoor light pollution, notably in shift workers. Timed feeding is one of the strategies currently being tested to mitigate the metabolic risks caused by circadian desynchronization [38, 61, 64, 65]. Further studies are needed to find the most effective chronotherapeutic countermeasures, possibly combining appropriate eating and sleeping schedules.