Circadian rhythms in the absence of the clock gene Bmal1

See allHide authors and affiliations

Science  14 Feb 2020:
Vol. 367, Issue 6479, pp. 800-806
DOI: 10.1126/science.aaw7365

Redundancy in circadian clocks?

The transcription factor BMAL1 is a core component of the mammalian circadian clock; without it, circadian behaviors are abolished. However, Ray et al. found that in animals lacking BMAL1, peripheral tissues synchronized with a brief pulse of the glucocorticoid hormone dexamethasone appear to retain a 24-hour pacemaker that sustains rhythmic gene expression, protein abundance, and protein phosphorylation in excised liver cells and fibroblasts (see the Perspective by Brown and Sato). These oscillations persisted in the absence of cues from changes in light or temperature. The results raise intriguing questions about the possible nature of the oscillator that maintains the observed rhythms.

Science, this issue p. 800; see also p. 740


Circadian (~24 hour) clocks have a fundamental role in regulating daily physiology. The transcription factor BMAL1 is a principal driver of a molecular clock in mammals. Bmal1 deletion abolishes 24-hour activity patterning, one measure of clock output. We determined whether Bmal1 function is necessary for daily molecular oscillations in skin fibroblasts and liver slices. Unexpectedly, in Bmal1 knockout mice, both tissues exhibited 24-hour oscillations of the transcriptome, proteome, and phosphoproteome over 2 to 3 days in the absence of any exogenous drivers such as daily light or temperature cycles. This demonstrates a competent 24-hour molecular pacemaker in Bmal1 knockouts. We suggest that such oscillations might be underpinned by transcriptional regulation by the recruitment of ETS family transcription factors, and nontranscriptionally by co-opting redox oscillations.

The primary regulator of circadian rhythmicity in mammals is thought to comprise transcriptional-translational feedback loops (TTFLs) that drive periodic expression of clock gene products (1, 2). In this scheme, BMAL1 (also known as MOP3 or ARNTL) is believed to serve as an indispensable component of the system (3), acting as a transcription factor that heterodimerizes with CLOCK (4) to activate circadian gene expression. Disruption of Bmal1 in mammals leads to a range of physiological abnormalities, including the abolition of circadian behavior (3, 5), aberrations in the sleep-wake cycle (6, 7), abnormal retinal function (8), neurodegeneration (9), and shorter life span (10). Deletion of Bmal1 disrupts robust oscillations of core clock components (11). However, Bmal1 may not be essential for all molecular oscillations beyond the canonical circadian circuit (8), particularly at the whole-genome or proteome scale, and TTFL models may not provide a comprehensive representation of all molecular circadian clocks (1216).

We explored whether 24-hour transcriptional oscillations are possible in Bmal1−/− mice under physiological conditions. To do this, we analyzed a liver RNA-sequencing (RNA-Seq) dataset in which Bmal1−/− mice [conventional Bmal1 knockout (KO)] had been entrained to a standard 12-hour light: 12-hour dark (LD) cycle for several days (17). Under these conditions, Bmal1−/− mice exhibit 24-hour locomotor activity rhythms (18), which are not observed under constant conditions (continuous environmental darkness) (3). We found that 8002 genes displayed 24-hour rhythms at a false discovery rate (FDR) < 0.05. These were detected by the RAIN (rhythmicity analysis incorporating nonparametric methods) algorithm, which detects both symmetric and nonsymmetric waveforms in time series data (19). This demonstrates that many liver transcripts are rhythmic under synchronized conditions (i.e., in an LD cycle).

In mammals, tissue clocks, such as those in the skin and liver, exist in a hierarchy and are synchronized by a central pacemaker residing in the suprachiasmatic nucleus (SCN) of the brain through a range of mechanisms including endocrine, autonomic, temperature, and feeding cues. This synchronization occurs such that organs assume relative phases to each other and the SCN, but also within each tissue so that individual cells are in phase with each other (20). To avoid the effects of such synchronization (which may convey a desynchronized signal to tissues in vivo) and test whether Bmal1−/− mice might retain an intrinsic timekeeping function, we assayed skin fibroblasts and liver tissues from these animals outside the body.

This approach enabled us to synchronize tissues and then allow them to free-run under constant conditions, whereby they might reveal endogenous rhythmicity. To synchronize liver tissues from Bmal1−/− and Bmal1+/+ mice, we treated them with a 15-min pulse of the glucocorticoid hormone dexamethasone (DEX), a standard and potent synchronizer of the molecular circadian clockwork in peripheral tissues (21). Forty-eight hours after synchronization, we collected samples every 3 hours for 3 days (Fig. 1A) and subjected these to RNA-Seq to quantify gene expression. Similar to what we saw in mice under entrained LD conditions, a large number of transcripts [5790 with p value < 0.05] oscillated in Bmal1−/− liver slices under constant conditions (Fig. 1B and fig. S1, C and D).

Fig. 1 Rhythmic transcriptome of Bmal1−/− mouse liver tissues and skin fibroblasts.

(A) Schematic representation of the experimental strategy used in this study. Cells and tissues were cultured outside the body (ex vivo) and synchronized by a single DEX pulse to evaluate their rhythmicity under constant conditions after this treatment [constant darkness (DD); gray and black bars show subjective external day and night, respectively] without any masking signal from the SCN. (B) Twenty-four-hour oscillating transcripts identified at different stringency levels (with RAIN) in wild-type and Bmal1−/− liver tissues. (C) Heatmap representation of the rhythmic transcripts (FDR < 0.1 in RAIN) in Bmal1+/+ and Bmal1−/− liver tissues. Corresponding abundance profiles for the rhythmic candidates identified in each genotype are displayed in the same order. (D) Venn diagram showing the overlap between the rhythmic genes (FDR < 0.05) identified in Bmal1-KO mice in LD cycle (Gene Expression Omnibus accession: GSE70499) (17), and in Bmal1−/− liver tissues under constant conditions (DD) as obtained in our study. (E) Twenty-four-hour oscillating transcripts identified at different stringency levels (with RAIN) in wild-type and Bmal1−/− MSFs. (F) Heatmap representation of the rhythmic transcripts (FDR < 0.1 in RAIN) in Bmal1+/+ and Bmal1−/− MSFs. Corresponding abundance profiles for the rhythmic candidates identified in each genotype are displayed in the same order.

Even after applying more stringent FDR, we identified 5098 transcripts at FDR < 0.1 or 3822 at FDR < 0.05 in Bmal1−/− liver tissue (Fig. 1B). With few exceptions, if a transcript oscillated in Bmal1−/− liver, it did not do so in Bmal1+/+ tissue and vice versa—that is, the sets of oscillating transcripts were almost mutually exclusive (Fig. 1C), with slightly different phase distribution patterns (fig. S1E). We tested the overlap between these rhythmic transcripts and those that we quantified as rhythmic in vivo. At FDR < 0.05, there was a highly significant overlap (2034 genes, Fisher’s exact test; p < 0.0001) between the rhythmic genes identified in both the datasets (Fig. 1D). We also determined which genes were synchronized by DEX by performing a pulse-chase experiment (fig. S2A). We found that 15.5% (788 out of 5098) of rhythmically expressed genes in Bmal1−/− tissues responded to glucocorticoid synchronization (fig. S2, B to F).

We then investigated skin fibroblasts (MSFs) from Bmal1−/− mice. Confluent (nondividing) MSFs were synchronized with a pulse of DEX and then sampled under constant conditions 48 hours after synchronization. We detected rhythmic transcripts in Bmal1−/− MSFs (Fig. 1E) with negligible overlap with the rhythmic transcripts identified in wild-type cells (Fig. 1F). Rhythmic transcripts had similar amplitude distributions in Bmal1−/− and Bmal1+/+ liver tissues (fig. S3A). Moreover, the amplitude distributions observed in our study are comparable with those of earlier circadian transcriptome studies performed using wild-type mice (fig. S3B). As expected, the amplitude for the rhythmic transcripts was higher in liver tissues compared to fibroblasts in both the genotypes (fig. S3, A and C). Furthermore, period analysis in fibroblasts indicated a predominant period of 24 to 27 hours in both the genotypes, although in Bmal1+/+ liver tissue, we saw a greater number of transcripts that oscillated with a longer period (26 to 27 hours) (fig. S3, D and E). There were a few transcripts with a short period (18 to 21 hours), and harmonics of circadian period (8- to 12-hour ultradian rhythms) were negligible in all the datasets (table S1). Together, these results demonstrate circadian oscillations of gene expression in DEX-synchronized liver and fibroblasts of Bmal1 knockout mice.

A crucial feature of circadian clocks is that their free-running period remains ~24 hours throughout a broad range of physiological temperatures (1). To determine whether transcriptomic oscillations seen in the absence of Bmal1 exhibit this key characteristic of circadian rhythmicity, we synchronized fibroblasts by DEX treatment and maintained them at 27°, or 37°C for two complete 24-hour cycles (Fig. 2A). The temperature coefficient (Q10) for the rhythmic transcriptome was ~1 in both genotypes (Fig. 2B). This indicates that there is temperature compensation of genome-scale circadian oscillations in Bmal1−/− cells (Fig. 2, C and D). Circadian clocks can also be “entrained” by external cues (1). We synchronized fibroblasts from both genotypes with a DEX pulse 12 hours apart and then sampled them in free-running conditions at the same external (solar) time (Fig. 2E). Several transcripts in wild-type or Bmal1−/− fibroblasts had oppositely phased rhythms when in free-run (Fig. 2, F and G). This means that they retained their initial phases (i.e., antiphasic). If driven by an exogenous cue during free-run, the rhythms would instead be in an identical phase, which they are not. Taken together, these findings demonstrate the presence of free-running, temperature-compensated, and entrainable (i.e., circadian) rhythms in the absence of the core clock gene Bmal1.

Fig. 2 Temperature compensation of circadian transcriptional rhythms in the absence of BMAL1.

(A) Schematic representation of the experimental strategy used in temperature compensation analysis of the rhythmic transcriptome. MSFs (wild-type and Bmal1−/−) were synchronized by one DEX pulse and maintained at a constant temperature of 27°, 32° or 37°C for two complete circadian cycles and were subsequently sampled every 2 hours at these three different temperatures under constant conditions (DD) for RNA-Seq analysis. (B) Temperature independence of transcriptome-level circadian oscillations in wild-type and Bmal1−/− cells as temperature coefficient (Q10) for the rhythmic transcriptome was found to be almost 1 in both the genotypes. Data are represented as mean ± SD, n = number of cyclic genes (FDR < 0.1) in each condition (Bmal1+/+ 27°C = 113, 32°C = 577, 37°C = 1340; Bmal1−/− 27°C = 1397, 32°C = 1169, 37°C = 692), period = 18 to 30 hours. (C) Abundance profiles of temperature-compensated rhythmic transcripts identified at all three different temperatures (FDR < 0.1). Bmal1+/+ (n = 32) and Bmal1−/− (n = 140), period = 24 hours. Transcript abundances were calculated as FPKM (fragments per kilobase per million mapped reads) and represented on a log2 scale after z-score normalization. (D) Abundance profiles (log2-transformed FPKM) of six representative temperature-compensated rhythmic genes in Bmal1−/− cells. Samples from three biological replicates were pooled together for RNA-Seq analysis at each time point. (E) Schematic showing the experimental design for oppositely phased initial synchronization in circadian transcriptomics analysis. MSFs (wild-type and Bmal1−/−) were synchronized with single DEX pulses 12 hours apart and then sampled (3-hour resolution) in free-running conditions at the same external (solar) time. (F) Abundance profiles (log2 transformed FPKM) of representative clock genes in wild-type cells showing oppositely phased transcripts. (G) Oppositely phased abundance profiles (log2 transformed FPKM) of representative rhythmic genes in Bmal1−/− fibroblasts.

To identify the possible mechanisms that can drive or sustain molecular oscillations in the absence of the known core clock machinery, we tested whether BMAL2 (MOP9), which is a closely related paralog of BMAL1, could substitute for BMAL1 in its absence. However, two independent lines of evidence exclude BMAL2 as a driver of rhythms in the absence of BMAL1. First, deletion of Bmal1 alone leads to loss of Bmal1 and Bmal2 function because Bmal2 is entirely regulated by Bmal1 (18). Accordingly, we did not detect BMAL1 or BMAL2 (MOP9) by immunoblotting in Bmal1−/− liver tissue (fig. S4, A to D). Second, if BMAL2 was able to substitute for BMAL1’s function at the genomic scale, there should be a substantial overlap of downstream rhythmic genes in both Bmal1−/− and Bmal1+/+ MSFs and liver tissues. However, we did not observe this (see Fig. 1, C and F). Thus, we did not find evidence to suggest that BMAL2 (and by extension other related basic helix loop helix transcription factors) substitutes for BMAL1 function in Bmal1−/− tissue or cells.

What might, therefore, be the underlying molecular mechanism driving circadian rhythmicity in Bmal1−/− tissue? To establish this, we analyzed the promoter regions of rhythmically expressed transcripts (FDR < 0.1), focusing on the two principal phase peaks at subjective dawn and dusk (Fig. 3A). Unbiased motif analysis indicated enrichment [q value (Benjamini) < 0.05] of E26 transformation-specific (ETS) factors, for the dawn phase rhythmic transcripts in both Bmal1+/+ and Bmal1−/− mice. (Fig. 3A). We found rhythmic expression of multiple ETS transcription factors with comparable peak phases clustered around dawn (Fig. 3B). We observed rhythmic expression of 9 and 11 ETS transcription factors in Bmal1−/− and Bmal1+/+ liver tissues, respectively (FDR < 0.1), with 5 overlapping candidates (fig. S6A and table S3).

Fig. 3 Twenty-four-hour rhythmicity of ETS transcription factors and peroxiredoxin (PRDX) oxidation in Bmal1+/+ and Bmal1−/− tissues.

(A) Top sequence motifs (q < 0.05) of the circadian transcriptional regulators for the dawn phase and dusk phase rhythmic transcripts (FDR < 0.1) identified in Bmal1+/+ and Bmal1−/− liver tissues De novo sequence motif analysis was performed with +/− 300–base pair DNA sequence from the master peak binding sites by using HOMER. (B) Frequency distribution of the phases showing rhythmic expression of multiple ETS transcription factors with comparable peak phases with the cyclic dawn phase transcripts. (C) Rhythmic expression (q < 0.05) of three ETS transcription factors in organotypic liver culture in constant conditions (DD, left) and in Bmal1-KO mice in a light-dark cycle (LD, right; three biological replicates from a single cycle are concatenated to enable comparison with ex vivo liver data) (17). Samples from three biological replicates were pooled together for RNA-Seq analysis at each time point. (D) RNAi-mediated knockdown of the ETS transcription factors that are rhythmic in Bmal1+/+ and/or Bmal1−/− mice induce alteration in clock period length (fig. S6). Data analyzed from BioGPS circadian layout database (24). ***Indicates p < 0.0001, **indicates 0.0001 < p < 0.001, and *indicates 0.001 < p < 0.05 (t test). (E) Twenty-four-hour oscillation (RAIN, p < 0.05) of peroxiredoxin oxidation [oxidized/hyperoxidized peroxiredoxin (PRDX-SO2/3)] in Bmal1−/− and wild-type liver tissues detected by immunoblotting. Quantification of the immunoblots was done by densitometry, and data are represented as mean ± SEM (n = 3). β-actin was used as a loading control to normalize PRDX-SO2/3 monomer bands. Immunoblots for PRDX-SO2/3 and β-actin are provided in fig. S7.

Furthermore, we found rhythmic expression of ETS transcription factors in Bmal1 knockout mice in vivo under LD cycles (17) (Fig. 3C). ETS binding sites are enriched in rhythmic enhancer RNAs in wild-type mice (22). Coexpression of the CLOCK-BMAL1-CRY1 complex has a strong inhibitory effect on the activity of other transcription factors, including ETS (23). This may explain why ETS transcription factors have less of a role in wild-type mice. We also investigated the effects of RNA interference (RNAi)–mediated inhibition of the ETS transcription factors that were identified as rhythmic in Bmal1+/+ and Bmal1−/− mice using BioGPS, a portal allowing access to circadian time courses from small interfering RNA (siRNA) knockdown of almost all genes (24). Depletion of many ETS transcription factors induced alteration in circadian period length in U2OS cells (Fig. 3D and fig. S6B). Thus, a range of ETS proteins could contribute to transcriptional oscillations in cells devoid of Bmal1.

An alternative mechanism that could generate molecular rhythms in Bmal1−/− is a nontranscriptional, biochemical oscillation (25). Oxidation-reduction state of peroxiredoxin (PRDX) proteins exhibit self-sustained oscillation in the absence of any TTFL mechanisms (12, 13, 26). Moreover, siRNA knockdown of PRDX proteins affects circadian rhythms in nucleated U2OS cells (12) (table S4). Consequently, we next determined whether similar oscillations of PRDX oxidation might be seen in DEX-synchronized Bmal1−/− liver. Lysates were immunoblotted by using an antiserum specific to overoxidized peroxiredoxin (PRDX-SO2/3) to monitor the redox status of PRDX. Statistically significant (RAIN, p < 0.05) cycling of PRDX-SO2/3 abundance was detected with a period ~24 hours in both Bmal1−/− and Bmal1+/+ liver tissues (Fig. 3E and fig. S7). We investigated the possible interactions among ETS transcription factors, PRDX proteins, and core clock components using the Search Tool for the Retrieval of Interacting Genes or Proteins (STRING) database. There were multiple interactions among ETS transcription factors, PRDX proteins, and clock components mediated through Trp53 and Sirt1 (fig. S8), which are important regulators of circadian clock gene expression (2729).

Next, we investigated whether such noncanonical rhythmicity is extended to the proteome and phosphoproteome levels in Bmal1−/− mice. Circadian proteome and phosphoproteome have been reported in wild-type mice (3032), but not in Bmal1−/− tissues. After synchronization with DEX, as described above, samples were harvested over two circadian cycles, labeled with 10-plex tandem mass tags (TMTs), and then analyzed by mass spectrometry to quantify the abundance of rhythmic proteins (Fig. 4A). A 24-hour oscillation in abundance of proteins and phosphosites was detected in both wild-type and Bmal1−/− MSFs and liver tissues (Fig. 4, B to D; fig. S9, A to D, and fig. S10, A to E). We detected 585 rhythmic proteins in wild-type and 364 in Bmal1−/− MSFs, with comparable peak phases of protein rhythms in both genotypes (Fig. 4C). Akin to the rhythmic transcriptome, we saw minimal overlap between the cyclic proteins and phosphoproteins identified in wild-type and Bmal1−/− cells (fig. S10, B to D). As expected, amplitudes for rhythmic phosphoproteins (phosphosites) were higher compared to proteins in both genotypes (fig. S10, F to H). The global rhythms of protein phosphorylation seen in Bmal1−/− implicates posttranslational processes as critical regulators of tissue rhythmicity, even in the absence of BMAL1.

Fig. 4 Rhythmic proteome and phosphoproteome in the absence of BMAL1.

(A) Schematic representation showing selection of the time-point samples for “discovery” and “validation” TMT 10-plex quantitative proteomics experiments. Samples pooled from three biological replicates were analyzed in each experiment. (B) Heatmap representation of the rhythmic proteins (multiple testing adjusted p < 0.05) in Bmal1+/+ and Bmal1−/− MSFs. (C) Overlapped rose plots representing the frequency distribution of the phases of the cycling proteins in Bmal1+/+ and Bmal1−/− skin fibroblasts and liver tissues. (D) Heatmap representation of the rhythmic phosphosites (multiple testing adjusted p < 0.05) in Bmal1+/+ and Bmal1−/− MSFs. (E and F) Biological process (E) and cellular component (F) terms overexpressed for the rhythmic proteins identified in Bmal1−/− systems. Top 10 overexpressed Gene Ontology (GO) terms according to their fold-enrichment (Bonferroni corrected p < 0.05) are depicted. Mitochondrial GO terms highly overexpressed for the rhythmic transcripts and proteins identified in Bmal1−/− systems.

The functions of rhythmic transcripts and proteins in Bmal1 knockouts were diverse (fig. S11). Biological processes that appeared to be enriched significantly in Bmal1−/− compared to wild-type included metabolism, intracellular transport, and oxidation-reduction (Fig. 4E and figs. S11 and S12). Furthermore, specific cellular components were overrepresented, particularly the mitochondrion and its different subcompartments (Fig. 4F and fig. S13). In agreement with these observations, Bmal1 function has been reported to be important for mitochondrial fission and fusion dynamics (33), and for respiratory function in the liver (33, 34). We have thus demonstrated that although Bmal1 is necessary for the expression of 24-hour behavioral cyclicity, it is not required for 24-hour molecular rhythms at the transcriptional, translational, and posttranslational levels.


Materials and Methods

Figs. S1 to S13

Tables S1 to S4

References (3657)

References and Notes

Acknowledgments: We thank the Advanced Sequencing and Bioinformatics scientific technology platforms at the Francis Crick Institute for their support with next-generation sequencing. Funding: A.B.R. acknowledges funding from the Perelman School of Medicine, University of Pennsylvania, and the Institute for Translational Medicine and Therapeutics (ITMAT), Perelman School of Medicine, University of Pennsylvania. A.B.R. also acknowledges funding from the European Research Council (ERC Starting Grant no. 281348, MetaCLOCK), the EMBO Young Investigators Programme, and the Lister Institute of Preventive Medicine. A.B.R. was supported in part by a Wellcome Trust Senior Fellowship in Clinical Science (100333/Z/12/Z) at the University of Cambridge, and also in part by the Francis Crick Institute, which receives its core funding from Cancer Research UK (FC001534), the UK Medical Research Council (FC001534), and the Wellcome Trust (FC001534). Author contributions: S.R., U.K.V., and A.B.R. conceived and designed the experiments. S.R., U.K.V., A.S., and G.D. performed the MSFs and liver tissue time-course experiments. U.K.V. performed the RNA-Seq and quantitative real-time reverse-transcriptase polymerase chain reaction experiments and analyzed the data. S.R. performed the quantitative proteomics and phosphoproteomics and analyzed the data with support from A.P.S. and S.A.H. A.B.R. supervised the whole study. The manuscript was written by S.R., U.K.V., and A.B.R. All authors agreed on the interpretation of data and approved the final version of the manuscript. Competing interests: The authors declare no competing interests. Data and materials availability: The RNA-seq data have been deposited in the Gene Expression Omnibus (accession nos. GSE111696 and GSE134333). The mass spectrometry proteomics and phosphoproteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (35) partner repository with the dataset identifier PXD009243.

Stay Connected to Science


Navigate This Article