Calorie restriction regulates circadian clock gene expression through BMAL1 dependent and independent mechanisms

30% CR is one of the most commonly used CR for C57B6 mice21,22, therefore, we selected this feeding paradigm to perform the gene expression analysis. To assay the effects of CR on the molecular circadian clockworks, we compared expression of the core circadian genes across the day in the liver of mice that were subjected to different feeding paradigms. Ad libitum (AL) fed mice had unlimited access to food throughout the day. CR mice received 30% less of their daily AL food intake for two months. In this group food was provided two hours after the light was turned off (ZT14); indeed, mice are nocturnal animals, so the nighttime feeding is most physiologically relevant. In the third, time restricted (TR) feeding group, animals were provided with the 100% of their average daily AL food intake at ZT14 (the same time as the CR group) for two weeks; this time period is sufficient to reset circadian clocks in peripheral tissues23,24. We noticed that animals in both CR and TR groups consumed all the food within the first 3–5 hours, in agreement with our previous report on TR feeding25; therefore, the TR group represents an appropriate control for the CR group. Indeed, animals consume the food at the same time as the CR group, but there is no reduction in the amount of consumed calories. The fourth group of animals has also been on the TR regimen for two weeks, but the animals did not receive food on the day of tissue collection. Thus, this fasting (F) group experiences food deprivation for 24 to 48 hours, and thus represents another control for the CR group: the control for sharp reduction in the calorie intake without any metabolic adaptation, in contrast to CR.

As expected, the expression of core clock genes exhibited oscillatory pattern of expression in the liver of AL mice (Fig. 1) in agreement with previously reported data. We observed that 30% CR dramatically affected the expression of circadian clock genes. According to the two way ANOVA analysis, mRNA expression for Bmal1, Per1, Per2, Cry2, (Fig. 1) and Rev Erb β, Per3 and Ror γ (Supplementary Fig. S1) were significantly affected by CR. Cosinor wave analysis of circadian rhythms in the expression of the circadian clock genes revealed that most feeding regimens did not significantly affect rhythmicity or acrophase for most of the genes, however, CR and fasting disrupted circadian rhythms in the expression of Per1 and CR induced circadian rhythms in the expression of Clock gene. CR did not affect Cry1, Rev Erb α and Ror α mRNA expression (Fig.1 and Supplementary Fig. S1) and even statistically significantly reduced the expression of the Clock gene. TR had a similar effect as CR on the expression of Rev Erb β and Clock genes. Thus, for Per1, Per2, Per3, Bmal1 and Cry2 genes the effect on the expression was CR specific and different from the effects of TR and Fasting.

Several clock-controlled transcription factors, such as Hlf, Tef, Dbp, E4bp4, Dec1 and Dec2, have been shown to be the transcriptional targets of CLOCK/BMAL1 complex, and reported to oscillate robustly with different phases in the liver26,27. Previous reports have shown feeding/fasting cycle to have variable effects on the mRNA expression of some of these transcription factors28,29. We assayed the mRNA expression of these clock-controlled genes in the liver of 30% CR, TR, Fasting and AL mice by qPCR. CR significantly induced the expression of Hlf and Tef (Fig. 2a, Supplementary Fig. S1f) at several time points; TR also affected the expression of these genes. CR up regulated the expression of the D box transcription factor (Dbp,) at ZT6 and ZT10, whereas Fasting reduced Dbp expression (Fig. 2b), in agreement with earlier report in fasting animals30. E4bp4 mRNA levels were not affected by CR or TR, but were up regulated by Fasting later at night time points compared to AL (Fig. 2c). There was statistically significant reduction in Dec1 expression at ZT22 in CR, it was increased in TR group and, as previously demonstrated, it was down regulated at several time points during the day in the Fasting group (Fig. 2d). Dec2 mRNA expression was significantly down regulated in CR animals and it was upregulated in TR and Fasting at several time points (Fig. 2f). Variable results have been observed in Pparα mRNA expression upon CR31,32. We showed that the effect of 30%CR on Pparα mRNA expression was not statistically significant (Fig. 2e). These results indicate that CR affects the expression of clock-controlled transcription factors; the effect of CR on Dbp, Dec1 and Dec2 is statistically significantly different from TR or Fasting.

Next we assayed the protein levels of the core clock genes (BMAL1, CLOCK, PER1, PER2 and CRY1) in the liver of mice from different feeding groups (30% CR, TR, Fasting and AL). Rhythmic oscillations at the protein level in our AL group were similar to that demonstrated in previous reports33,34. BMAL1, PER1 and PER2 have shown circadian rhythms, while daily changes in CLOCK and CRY1 were not circadian according to the cosinor wave analysis33,34. Based on the two way ANOVA, CR significantly reduced the protein level of core clock gene, CRY1, while the expression of CLOCK, BMAL1, PER1 and PER2 were downregulated but the effect was not statistically significant as shown in Fig. 3. CLOCK protein levels was down regulated at one time point upon Fasting (Fig. 3 and Supplementary Fig. S2). PER1 and PER2 showed temporal changes in AL and TR as expected, having a peak around ZT 14, whereas in Fasting maximum abundance was around ZT 22-2 for Per2 and around ZT 2 for Per1. The CRY1 protein level was dramatically reduced in both CR and Fasting groups compared to AL and TR (Fig. 3 and Supplementary Fig. S2) at all tested time points. Thus, both Fasting and CR led to significant down regulation of CRY1 on protein level at every time point, at the same time the effect of Fasting on the expression of CLOCK was time of the day dependent. TR effect on the expression for all tested clock genes was not statistically different from AL.

Because CR had a significant effect on clock genes mRNA expression and protein levels, our next question was if the effect of CR was through the regulation of BMAL1 transcription activity. We analyzed the mRNA expression of clock genes in whole body Bmal1−/− mice subjected to 30% CR for 2 months. Similar to others, our results also demonstrated that expressions of Per1, Per2 and Cry1 were arrhythmic in AL fed Bmal1−/− mice (Fig. 4 and cosinor analysis). According to three way ANOVA, the effect of CR was statistically significant for all three tested genes in wild type but in Bmal1−/− mice the effect of CR was significant only for Cry1. Thus, BMAL1 is required for the effect of CR on Per1 and Per2 but not for Cry1 mRNA expression.

Effects of CR on the expression of Per1 and Per2 were previously reported: Per1 and Per2 genes were identified, using multiple microarray data analysis, among the group of genes with differential expression in the multiple tissues such as liver, brain, skeletal muscles of mice subjected to lifespan extending diets such as CR35. These genes were proposed as longevity-associated candidate genes and later for some of them their differential expressions were confirmed in tissues of long lived dwarf mice35. However, one important caveat in these studies is that the time of tissue collection was not defined, making the comparison between samples complicated. Indeed, microarray analysis of the expression upon CR was performed in multiple groups independently and for every selected gene, including Per1 and Per2, the effects of CR on the expression was not confirmed in every study35,36. We hypothesized that the observed discrepancy in results between different studies is due to different time of sample collections and the expression of longevity candidates genes might significantly change across the day. Indeed the effects of CR on Per1 and Per2 expressions in our study were significant at some time and not significant at another time of the day (Fig. 1). We selected ten longevity candidate genes: the expressions of Fmo3, Cyp4a14, Parp16 and Igfals genes were reported to be up regulated upon CR while the expressions of Cyp4a12b, Mup4, Hes6, Hsd3β5, Serpina12 and Alas2 were reported to be down regulated upon CR. We assayed the mRNA expression of these genes across different time of the day in the liver of mice maintained on different diets. Interestingly, some of the genes tested such as Hsd3β5 and Serpina12 exhibited rhythmic expression around the day. We did not find statistically significant effect of CR on the expressions of Igfals and Serpina12. The expressions of 8 candidate genes was statistically significant either up or down regulated upon 30% CR (Fig. 5 and Supplementary Fig. S3) at least at some time points, in agreement with published microarray data; however, only for a subset of longevity candidate genes the effect was CR-specific. Expression of flavin monoxygenase (Fmo3) (Fig. 5a) was up regulated significantly with high amplitude rhythms upon CR, but not upon TR or Fasting. Expression of Cyp450 gene, Cyp4a12b, (Fig. 5b) was down regulated upon CR, while there was no effect of TR or Fasting. Expression of Parp16 (Supplementary Fig. S3a) was statistically significantly up regulated, and expression of, Mup4 and Alas2 (Fig. 5f and Supplementary Fig. S3c) was statistically significantly down regulated upon CR at several time points, while at other time points there was no difference between CR and AL, TR or Fasting samples. Effects on the expression of other tested target genes were not CR-specific: for example, while the expression of Cyp450 gene, Cyp4a14 (Fig. 5c) was dramatically up regulated upon CR, and expressions of Hes6 (Fig. S3b) and Hsd3β5 (Supplementary Fig. S3d) were down regulated upon CR (in agreement with the published microarray data), similar up or down regulations were detected upon Fasting too. These results emphasize the importance of the circadian approach for the study of the effect of CR on transcriptome. The effect of CR on Fmo3, Cyp4a14a, Cyp4a12b, Mup4, Hes6, Parp16, and Hsd3B5 was statistically significantly different from effect of TR and Fasting.

All the animal studies were performed with approval from the Institutional Animal Care and Use Committee (IACUC) of Cleveland State University (Protocol No. 21124-KON-S). The care and use of mice were carried out in accordance with the guidelines of the Institutional Animal Care and Use Committee (IACUC) of the Cleveland State University.

Wild type and whole body Bmal1−/− male mice were used for the experiments. Mice were of C57B6J background. Bmal1−/− mice were obtained from laboratory of Dr. Bradfield58. Animals were maintained on the 12:12 light:dark cycle with lights on at 7:00 am, and fed the 18% protein rodent diet (Harlan). The ad libitum (AL) group had unrestricted access to food. Calorie restriction (CR) was started at 3 months of age. For the first week animals had been on 10% restriction, for the second week on 20% restriction and on 30% restriction for the rest of the experiment. The CR group received their food once per day at ZT14 (two hours after lights were off). After two months of CR, tissues were collected at six different time points across the day. The time restricted (TR) feeding group started to receive 100% of their average daily intake as one meal at ZT14. TR started at 4.5 months of age, mice were on TR for 2 weeks before tissue collection. The fasting group (F) was on the same TR feeding regime for 2 weeks, but did not receive food on the day of tissue collection. All groups had unrestricted access to water. All tissue collection experiments were performed on 5 months old wild type (WT) and Bmal1−/− male mice. For all experiments three animals of each genotype, feeding regimen and time point were used.

For gene expression studies, liver tissues from 3 male mice on each diet (AL, CR, TR and Fasting) and for both genotype (WT and Bmal1−/−) were collected every four hours throughout the day, and stored at −80 °C. Total RNA was isolated using TriZol reagent (Invitrogen, Carlsbad, CA) as per the manufacturer’s protocol. Briefly, frozen liver piece was minced in 1 ml TriZol reagent with pestle on ice. Following chloroform extraction step, total RNA was precipitated with isopropanol by centrifugation and pellet obtained was washed with 70% Ethanol. RNA pellet was diluted in 30 μl of RNAse-free water and quantified on Nanodrop. RNA integrity was checked on 1% agarose gel run at 90 V for 30 minutes. 20 μl of RT mix was prepared using 1 μg of RNA, 50 ng of 50 uM random hexamer (N8080127, Invitrogen), 10 mM dNTP (DD0058, Biobasic), 0.1 M DTT and RNaseOUT™ Recombinant RNase Inhibitor (10777-019, 40 units/μl). It was then reverse transcribed by qPCR machine using 200 u/μl of SuperScript^® III Reverse Transcriptase (18080-044, Invitrogen) as per the manufacturer’s instructions. Incubation conditions used were: 65 °C for 10 minutes followed by incubation on ice for 1 minute; 25 °C for 5 minutes; 50 °C for 60 minutes; Inactivate the reaction by heating at 70 °C for 15 minutes. RNA quantification was performed using qPCR with Universal Syber Green mix (1725125, BioRad). The reaction was carried out in triplicates for the gene of interest and in duplicates for the normalizing control using CFX96 qPCR Detection System (BioRad) with 50 ng of cDNA. Thermal cycling conditions used were according to the instructions of SYBR Green mix protocol and are briefly described in and relative mRNA abundance was calculated using the comparative delta-Ct method with ribosomal 18S rRNA and Gapdh as reference genes as described in25. Water was used as the negative control for the qPCR analysis. Product specificity was confirmed by melting curve analysis while primer pair efficiency was calculated by generating standard curve using serial dilutions of standard. Primers used for the analysis of expression are listed in Supplementary materials.

For analysis of protein expression tissues from three male mice per time point were used for each feeding regimens and both genotypes (WT and Bmal1−/−). Figure 3 represents Western blotting when three liver samples from three different mice were pooled together at each time point for each diet. For quantitative data presented in Supplementary Fig. S3, liver samples from individual mice (N = 3) were run to estimate a variability between biological replicates and calculate means and errors. For lysates preparation, frozen liver pieces were lysed in cell signaling lysis buffer with Protease/Phosphatase Inhibitor Cocktail (Cell Signaling Technology, Beverly, MA, USA) using sonicator. Protein concentration was determined by Bradford protein assay kit according to manufacturer’s protocol using spectrophotometer and lysates were stored at −80C. 45 ug of protein was loaded on 3–8% tris-acetate and 4–12% bis-tris gels (Invitrogen). Protein was transferred on PVDF membrane at 110 mAmp. Equal loading of proteins was checked by Ponceau stain. Primary antibodies anti-CRY1 (SAB biosciences), anti-BMAL1 (Santacruz Biotechnology), anti-CLOCK (kindly provided by Dr Marina Antoch, Roswell Park Cancer Institute), anti-PER2 (Alpha Diagnostics), anti-PER1 (Thermoscientific) and anti-GAPDH (Cell Signaling) were used for Immunoblot analyses. Protein analysis and quantification was done using Scientific Imaging film and Odyssey FC imaging system (LI-COR).

For each feeding type and for each genotype, at least three animals for every time point were used for all experiments. Data are shown as average +/− S.D. To assay the effect of feeding and time of the day on mRNA and protein level we performed two-way ANOVA. If the effect of feeding and/or time of the day was found to be statistically significant, Bonferroni correction was used to calculate p value for pairwise comparison between each feeding regimen at every time of the day. To assay the effect of genotype, feeding and time of the day on mRNA levels first analysis was performed using three-way ANOVA, If the effect of feeding, genotype and/or time of the day was found to be statistically significant, Bonferroni correction was used to calculate p value for pairwise comparison between each feeding regimen at every time of the day. IBM SPSS Statistics 20 and GraphPad Prism Version 5.04 software packages were used for statistical analysis. P < 0.05 was considered as statistically significant difference.