Administration of Exogenous Melatonin Improves the Diurnal Rhythms of the Gut Microbiota in Mice Fed a High-Fat Diet

Body weights were recorded in the present study, and the results showed an increase in final body weight after 2 weeks of HFD feeding (P < 0.001) (Fig. 1A and B). Our previous study confirmed that administration of exogenous melatonin improved subcutaneous adipose accumulation in HFD-fed mice (17), and the relative weight of subcutaneous adipose (P > 0.05) tended to be low in the HFD plus melatonin (MelHF) group. The amount of visceral adipose tissue (P < 0.05) was markedly reduced in the MelHF group in this study (Fig. 1C and D).

The circadian clock and metabolism are generally impaired in HFD-fed mice (27, 32). Thus, we further analyzed the diurnal variation in circadian clock genes (Clock, Cry1, Cry2, Per1, and Per2) in response to HFD and administration of exogenous melatonin (Fig. 2A; Table 1). Interestingly, Clock mRNA showed significant rhythmicity in the livers of control (P < 0.01) and MelHF (P < 0.05) mice, but not in the HFD group (P > 0.05), whereas the expression of Cry1, Cry2, Per1, and Per2 in the liver showed a significant daily rhythm in all groups (P < 0.05).

Next, we determined the diurnal patterns of serum lipids and glucose in the three experimental groups (Fig. 2B; Table 2). Serum triglycerides (TG) exhibited significant rhythmicity in control and MelHF mice (P < 0.01) but not in HFD mice (P > 0.05). A significant diurnal rhythm of low-density lipoprotein (LDL) was observed in only control mice (P < 0.01). Serum glucose exhibited rhythmicity in all three groups (P < 0.01). No daily rhythms were observed in the levels of serum cholesterol (CHOL) and high-density lipoprotein (HDL) (P > 0.05). Despite the rhythmicity, overall lipid indexes were very high in HFD-fed mice, while the trends in the MelHF group were similar to those of control subjects, and the values were much lower than those for the HFD-fed mice at specific time points, as previously shown (17).

To determine whether serum lipid rhythmicity was associated with the liver expression of clock genes, we performed Pearson correlation analysis among serum lipid indexes and circadian clock genes (Clock, Cry1, Cry2, Per1, and Per2) (Fig. 2C). Surprisingly, serum TG concentration was positively correlated with Clock expression but exhibited a negative correlation with the mRNA levels of Cry2 and Per1 (P < 0.001). Together, the rhythmicity of lipid indexes, especially TG concentration, was widely observed in the blood and was markedly associated with clock gene expression. The daily rhythm of TG was impaired in the HFD-fed mice, which was markedly improved by administration of exogenous melatonin.

The gut microbiota has been identified as a key element involved in host circadian rhythms and itself also undergoes circadian oscillation, which is disturbed in HFD-fed mice or obesity models (27, 28, 33). Our previous study demonstrated that melatonin treatment improved lipid metabolism by reprogramming the gut microbiota in HFD-fed mice (17); thus, we hypothesized that administration of exogenous melatonin would improve the daily rhythm of the gut microbiota.

Mice were sacrificed every 4 h within a 24-h period, and metagenomic DNA was extracted from the cecal contents. The gut microbiota was tested by 16S rRNA gene sequencing, and the compositions were similar to those observed in our previous study (17), that is, the most abundant phylum, Bacteroidetes, was decreased in HFD-fed mice, and the abundance of Firmicutes increased; melatonin reversed these alterations (see Fig. S1 in the supplemental material). Firmicutes exhibited significant rhythmicity in control and MelHF mice (P < 0.05) but not in HFD mice (P > 0.05), while Bacteroidetes exhibited rhythmicity in only the control and HFD groups (P < 0.05) (Table 3). The relative abundance of Firmicutes peaked at 4:00 in the HFD group but at 8:00 in the control and MelHF groups (Fig. S1). However, Proteobacteria and Actinobacteria failed to show a diurnal variation at the phylum level (Table 3).

At the genus level, 8 genera were significantly altered, and most of them exhibited a marked daily rhythmicity, except for Bacteroides, Desulfovibrio, and Clostridiales (P > 0.05) (Fig. 3; Table 3; see also Fig. S2). Parasutterella (P < 0.05), Alloprevotella (P < 0.01), Parabacteroides (P < 0.01), and Alistipes (P < 0.01) were only rhythmic in the control group. Intestinimonas exhibited a significant rhythm in only HFD-fed mice (P < 0.01). Ruminococcaceae (P < 0.05), Helicobacter (P < 0.05), and Roseburia (P < 0.01) showed a daily rhythm in only the melatonin-treated mice. Oscillibacter, Rikenella, and Lachnoclostridium exhibited marked cycles in the control and HFD groups (P < 0.05) but not in the MelHF group (P > 0.05). Anaerotruncus showed a diurnal pattern in the control and MelHF groups (P < 0.01) but not in HFD-fed mice (P > 0.05). We also noticed that Lachnospiraceae was rhythmic in the HFD and MelHF groups (P < 0.05) but not in the control group (P > 0.05). In addition, Lactobacillus and Ruminiclostridium exhibited rhythmicity regardless of HFD and melatonin challenges (P < 0.05).

Collectively, our data showed that most of the microbiota exhibited daily variation and that the diurnal network of some gut microbiota was affected by HFD and reversed, at least in part, by administration of exogenous melatonin (). Fig. S2

The gut microbiota has a widespread and modifiable effect on host gene regulation (34); thus, metabolism, genetic information, environmental information, cellular processes, human diseases, and organismal system pathways were further annotated according to the microbiota compositions by Tax4Fun analysis (see Fig. S4A). Our data show that short-term HFD feeding markedly affected cell growth and death, endocrine and metabolic diseases, the endocrine system, the nervous system, the immune system, and environmental adaptation (P < 0.05), while administration of exogenous melatonin influenced lipid metabolism, terpenoids, and polyketides (P < 0.05). We then further analyzed lipid metabolism (Fig. S4B) and identified eight pathways that mainly contributed lipid metabolism-annotated genes, namely, lipid biosynthesis, fatty acid biosynthesis, glycerophospholipid metabolism, glycerolipid metabolism, sphingolipid metabolism, fatty acid degradation, biosynthesis of unsaturated fatty acids, and synthesis and degradation of ketone bodies (Fig. S4C).

We then investigated whether the gut microbiota (top 50) also showed an association with clock gene expression and serum lipid levels by Spearman’s test (Fig. 4A). The relative abundances of Rikenella, Alistipes, and Enterorhabdus were positively correlated with Clock mRNA (P < 0.05) (Fig. 4). Ten genera (i.e., Helicobacter, unidentified Lachnospiraceae, Intestinimonas, Ruminiclostridium, Oscillibacter, Rikenella, Blautia, Negativibacillus, Harryflintia, and Caproiciproducens) showed a positive association with Cry1 mRNA (P < 0.05), while the correlation was negative between Cry2 mRNA and most genera, such as Helicobacter, unidentified Lachnospiraceae, Intestinimonas, Roseburia, Oscillibacter, Anaerotruncus, Mucispirillum, Butyricicoccus, Angelakisella, Tyzzerella, Streptococcus, Caproiciproducens, and Peptococcus. The expressions of Per1 and Per2 shared the markedly correlation to the relative abundances of Roseburia, Phyllobacterium, Anaerotruncus, Butyricicoccus, and Butyricimonas. Together, 29 genera were found to be correlated with clock gene expression; these correlations were mostly positive with Clock, Cry1, and Per2 mRNA and negative with Cry2 and Per1 mRNA.

A correlation analysis between serum lipid indexes and the gut microbiota was further conducted, and 17 genera (34% of top 50) were observed to be markedly correlated with TG concentrations (Fig. 4), including Lactobacillus, Bacteroides, Helicobacter, Parabacteroides, Ruminiclostridium, Oscillibacter, Rikenella, Alistipes, Anaerotruncus, Mucispirillum, Butyricicoccus, Enterorhabdus, Negativibacillus, Acinetobacter, Streptococcus, Caproiciproducens, and Peptococcus. The relative abundances of Dechloromonas, “Candidatus Arthromitus,” and Streptococcus showed significant correlations with both TG and HDL levels, while negative correlations were noticed between LDL and unidentified Clostridiales, Alistipes, and Butyricimonas.

We further determined the effect of melatonin treatment during daytime or nighttime on lipid metabolism and the gut microbiota. HFD-fed mice showed high relative weights of subcutaneous inguinal fat, periuterine fat, perirenal fat, and total fat (P < 0.001) (Fig. 5A to E). Administration of exogenous melatonin during daytime markedly reduced perirenal fat (P < 0.05) and total fat (P < 0.01) weights (Fig. 5D and E), but the trend was nonsignificant for the nighttime treatment compared with the control group (P > 0.05) (Fig. 5B to E). We also tested serum lipid indexes (Fig. 5F to J), and the results showed that serum TG and bile acid concentrations were markedly reduced in the daytime melatonin (MelD) group (P < 0.05) but not in the nighttime melatonin (MelN) group (P > 0.05). Taken together, we failed to notice any significant difference in host lipid metabolism between daytime and nighttime melatonin exposure.

We then investigated the gut microbiota compositions of the HFD, MelD, and MelN groups using 16S rRNA gene sequencing. At the phylum level, melatonin treatment during daytime or nighttime failed to alter the gut microbiota composition (Fig. 6A). Interestingly, administration of exogenous melatonin during nighttime significantly reduced the relative abundance of Firmicutes compared with that for the daytime treatment (P < 0.05). At the genus level, Lactobacillus, Intestinimonas, and Oscillibacter were significantly affected by melatonin treatment during the day or the night (P < 0.05) (Fig. 6B).

As gut microbiota correlated with serum lipids and both gut microbiota and serum lipid indexes exhibited a daily rhythmicity, which is highly driven by HFD feeding and melatonin drinking, we next performed fecal microbiota transplantation at two different time points (8:00 and 16:00) from the control, HFD, and MelHF groups into antibiotic-treated mice to investigate the response to HFD feeding. Body weights were recorded, and no significant difference was observed between the two time points (Fig. 7A to D). Interestingly, the relative weight of subcutaneous inguinal fat in the group receiving transplants from controls (MT-Cont group) was affected by the time at which the microbiota was transplanted (P < 0.05).

Similar to the results of our previous study (17), microbiota transplantation at 8:00 from the HFD group tended to enhance serum TG, CHOL, and HDL concentrations, which were slightly reversed in the group receiving transplants from MelHF mice (MT-MelHF group) (Fig. 7E to G). Conversely, serum TG concentration was reduced in the MT-HF group (P < 0.05) when microbiota transplantation was performed at 16:00 (Fig. 7E), and LDL was increased in the MT-MelHF group (P < 0.05) (Fig. 7H). Notably, microbiota transplantation from control subjects at 8:00 tended to enhance serum CHOL and HDL (P > 0.05) (Fig. 7F and G) and significantly increased TG concentrations (P < 0.05) (Fig. 7E) compared with those after microbiota transplantation at 16:00 (Fig. 7B). However, serum CHOL and HDL levels were lower at 16:00 than at 8:00 for HFD-derived microbiota transplantation (P < 0.05) (Fig. 7E to G). No difference was observed between the two time points in the MT-MelHF group.

In conclusion, our results show that most gut microbes exhibit a daily rhythm and are closely associated with clock gene expression and serum lipid levels. Melatonin improves the diurnal patterns of the gut microbiota in HFD-fed mice, which is further confirmed by microbiota transplantation. Microbiota transplantation early in the morning or in the late afternoon also lead to diverse responses to HFD. Taken together, we conclude that most gut microbiota cycles occur within a 24-h period, and the rhythm is disturbed by HFD feeding, while administration of exogenous melatonin improves diurnal patterns of some specific microbiota in HFD-fed mice. However, the detailed mechanisms behind melatonin mediated-gut microbiota and metabolic rhythmicity (directly targeted or indirect modulation of body weight) have not be fully resolved; thus, melatonin treatment in a healthy model and a 48-h rhythmic analysis are suggested to confirm the merit of melatonin in obesity.

ICR mice, a melatonin-deficient strain, were used in this study to eliminate the effect of endogenous melatonin production (SLAC Laboratory Animal Central, Changsha, China). As sex affects the melatonin profile (48), only female mice were used in this study to rule out this effect. All animals had free access to food and drinking water (temperature, 25 ± 2°C; relative humidity, 45% to 60%; lighting cycle, 12 h/day) during the experiment. The diets used in this study were as described in our previous study (17).

A total of 126 female mice (22.77 ± 0.10 g, approximately 4 weeks old) were randomly grouped into the control (Cont), HFD, and HFD plus melatonin (MelHF) groups (n = 42). Mice in the MelHF group received the HFD and melatonin-containing water (0.4 mg/ml melatonin [Meilun, Dalian, China], directly diluted in drinking water) (17). The melatonin solution was prepared daily and kept in a normal bottle with an aluminum foil cover to prevent light-induced degradation of melatonin. After 2 weeks of melatonin administration, 6 mice in each group were randomly killed at 0:00 (Zeitgeber time 16 [ZT16]), 4:00 (ZT20), 8:00 (ZT0, lights on), 12:00 (ZT4), 16:00 (ZT8), 20:00 (ZT12, lights off), and 24:00 (ZT16) (n = 6). Blood samples were collected by orbital bleeding. Liver, adipose tissue, and colonic digesta samples were weighed and collected.

Mice (26.89 ± 0.15 g) were randomly grouped into a control and three HFD groups (n = 12). One group of HFD mice received melatonin during daytime (8:00 to 16:00) and control water at night (16:00 to 8:00) (MelD), and another received melatonin during nighttime (16:00 to 8:00) and control water during daytime (8:00 to 16:00) (MelN). All mice were sacrificed at 8:00 a.m. after 2 weeks of feeding, and samples were collected for further analyses.

Mice were treated with antibiotics (1 g/liter streptomycin, 0.5 g/liter ampicillin, 1 g/liter gentamicin, and 0.5 g/liter vancomycin) to clear the gut microbiota (17). After 1 week of antibiotic treatment, the antibiotic-containing water was replaced with regular water, and the microbiota-depleted mice received transplants of the donor microbiota. Fecal supernatants from the control (MT-Cont), HFD (MT-HF), and MelHF (MT-MelHF) (treated for 14 days) mice were transplanted into the microbiota-depleted mice at 8:00 and 16:00 (for 5 days). Following transplantation, all mice further received HFD and regular water for an additional 14 days.

Serum samples were separated after centrifugation at 3,000 rpm for 10 min at 4°C. A Cobas c-311 Coulter chemistry analyzer was used to test serum biochemical parameters (17, 49), including triglycerides (TG), cholesterol (CHOL), high-density lipoprotein (HDL), low-density lipoprotein (LDL), glucose, and bile acid, as these indexes are commonly dysregulated in HFD-fed or obese subjects (50–53).

Liver samples were frozen in liquid nitrogen and ground, and total RNA was isolated by using TRIzol reagent (Invitrogen, USA) and then treated with DNase I (Invitrogen, USA). Reverse transcription was conducted at 37°C for 15 min at 95°C for 5 s. The primers used in this study were designed according to the mouse sequence (see Table S1 in the supplemental material). β-Actin was chosen as the housekeeping gene to normalize target gene levels. PCR cycling and relative expression determination were performed according to previous studies (54–61).

Total genome DNA from colonic samples was extracted for amplification using a specific primer with a barcode (16S V3+V4). Sequencing libraries were generated and analyzed according to our previous study (54, 62, 63). Operational taxonomic units (OTUs) were further used for genomic prediction of microbial communities by Tax4Fun analysis (64).

All statistical analyses were performed using one-way analysis of variance, and multiple comparisons were further conducted using Bonferroni analysis (SPSS 21 software). Spearman’s correlation analysis was conducted. The rhythmicity of clock genes, serum lipid indexes, and the gut microbiota was assessed by cosinor analysis using the nonlinear regression model within Sigmaplot V 10.0 (Systat Software, San Jose, CA, USA) (65). Multivariate analysis of variance for the time series was conducted by Duncan’s test, and values with different lowercase letters in the figure panels are significantly different. The data are expressed as the means ± standard errors of the means (SEMs). A P value of <0.05 was considered significant. All figures in this study were drawn by using GraphPad Prism 7.04.

Raw sequences are available in the NCBI Sequence Read Archive with accession numbers SAMN11246274↗, PRJNA528844↗, SAMN11245315↗, and PRJNA528812↗.