Daily liver rhythms: Coupling morphological and molecular oscillations

Significance

The liver plays an essential role in the coordination of daytime-dependent metabolism, and it is arguably the peripheral organ with the most dramatic circadian oscillations. Thus, it expresses a large number of clock-controlled genes, and its mass and macromolecular content fluctuate by more than 30% every day. These alterations are accompanied by fluctuations in hepatocyte cell size, actin polymerization, global protein synthesis, and ribosome accumulation. Here, we summarize some of our work on circadian rhythms by focusing on the liver.

Abstract

In mammals, a hierarchically organized circadian timing system orchestrates daily rhythms of nearly all physiology. A master pacemaker in the brain’s suprachiasmatic nucleus (SCN) synchronizes subsidiary clocks in most peripheral organs. By driving anabolic and catabolic cycles of proteins, lipids, and carbohydrates and by detoxifying endo- and xenobiotic components, the liver plays an important role in adapting the metabolic needs to rest-activity rhythms. In keeping with these functions, the liver expresses many clock-controlled genes that are required for these processes. Remarkably, however, this organ also fluctuates in size and morphological parameters. In mice, the mass of the liver increases and decreases by 30 to 40% during the 24-h day. The oscillation in liver mass is accompanied by daily rhythms of similar amplitudes in hepatocyte cell size and global RNA and protein accumulation. The number of ribosomes, which parallels the ups and downs of liver size, appears to be the rate-limiting factor in driving the diurnal rhythms of overall protein synthesis. Obviously, the rapid increase in hepatocyte size within the liver engenders mechanical stress, which must be dealt with by increasing the physical robustness of cells. Indeed, the actin cytoskeleton of hepatocytes undergoes dramatic polymerization cycles. Thus, massive intracellular and subcortical F-actin bundles are assembled during the night, at which the liver reaches its maximal size. In turn, the oscillation in actin polymerization elicits rhythms in myocardin-related transcription factors-serum response factor signaling, which participate in the circadian transcription of the core clock gene Per2 and thereby contribute to the synchronization of hepatocyte clocks.

Most light-sensitive organisms from cyanobacteria to mammals are equipped with temperature-compensated circadian clocks, allowing them to adapt their physiological needs to daytime in an anticipatory manner. The mammalian circadian timing system displays a hierarchical architecture. It is composed of a master pacemaker located in the suprachiasmatic nucleus (SCN) of the ventral hypothalamus and self-sustained and cell autonomous clocks that are operative in nearly all cells of the body (reviewed in refs. 1 and 2). The SCN, in conjunction with a sleep homeostat, drives rest-activity cycles. These free-run under constant conditions (i.e., constant darkness and temperature) with a period of approximately 24 h and are synchronized by daily light–dark cycles to exactly 24 h. In addition, the SCN synchronizes circadian oscillators in peripheral tissues. The molecular makeup of circadian oscillators is virtually identical in SCN neurons and peripheral cell types and relies on two coupled negative transcriptional–translational feedback loops (TTFLs) in clock gene expression. In the primary TTFL the heterodimeric transcription factors CLOCK and BMAL1 activate the transcription of genes encoding their own repressors. The latter include the Period genes Per1, Per2, and Per3 and the Cryptochrome genes Cry1 and Cry2, which assemble into large macromolecular complexes (1, 3, 4). Once PER-CRY complexes reach a critical nuclear concentration and/or activity, they bind to CLOCK-BMAL1 complexes and suppress their transactivation potential. In a secondary TTFL, members of the nuclear orphan receptor families Retinoic acid receptor-related orphan receptors (ROR) and REV-ERB (officially known as Nuclear Receptor Subfamily 1 Group D, NR1D) control the circadian transcription of Bmal1 and Clock. Since Rev-erbα (Nr1d1) and Rev-erbβ (Nr1d2) are CLOCK-BMAL1 target genes, the primary and secondary TTFLs are tightly coupled, which increases the robustness of the clockwork circuitry (5, 6).

In the 1980s, the Schibler lab developed an in vitro transcription system with nuclear extracts from several rat tissues that mimicked liver-specific transcription of the serum albumin (Alb) gene. These cell-free assays facilitated the characterization of transcription factors occupying elements of the Alb promoter. Studies on one of these transcription factors resulted in contradictory results obtained by Chris Mueller and Jérôme Wuarin, two lab mates. Chris, a Canadian postdoc, discovered a new transcription factor binding to the D-site of the Alb promoter, and he hence dubbed it DBP (for D-site binding protein). His screening of a bacteriophage λ GT11 expression library with a ³²P-radiolabeled oligonucleotide encompassing the D-site resulted in the isolation of a full-length DBP complementary DNA (cDNA). By overexpressing this cDNA in Escherichia coli, he could produce large quantities of recombinant DBP, which in turn allowed him to raise DBP antibodies. So, he possessed all the tools necessary for analyzing the expression of this transcription factor on the messenger RNA (mRNA) and protein level. At the time only a few mammalian transcription factors were known, and the discovery of every new transcription factor was readily published in a prestigious journal (7). Chris’ Cell paper on DBP secured him an assistant professorship at Queen’s University in Kingston, Canada. Jérôme Wuarin, a new PhD student in Schibler’s lab, decided to pursue the studies on DBP, after Chris went back to Canada. However, after 3 mos of hard work Jérôme was still unable to detect either DBP mRNA or protein in liver extracts by using Chris’ DBP cDNA or antibodies, respectively. He thus urged his professor to either retract the Cell paper or roll his sleeves up and reproduce Chris’ results with his own hands. Schibler was then still capable of using scientific equipment other than the telephone, and he thus opted for the second alternative. Fortunately, he could detect both DBP mRNA and protein, and a retraction of Chris’ Cell paper could thus be avoided. Recalling the work on circadian rhythms by his close friend Michael Rosbash at Brandeis University, Schibler considered that circadian expression might account for the discrepant results obtained by Jérôme and Chris or himself. This was indeed a viable hypothesis, since Jérôme, a farmer’s son, was used to wake up early in the morning, and he sacrificed the rats between 8 am and 10 am. In contrast, Chris, a night owl, and Schibler, who was usually busy with administrational work in the morning, did so in the late afternoon. Toward testing the “circadian conjecture,” Jérôme prepared rat liver nuclear extracts at 4-h intervals around the clock, and his results were quite spectacular. His immunoblot experiments unveiled that DBP accumulation oscillated with an about 300-fold circadian amplitude (i.e., zenith/nadir levels), and that DBP was barely detectable in the morning (8). This serendipitous finding motivated the Schibler group to focus its research efforts on circadian, rather than tissue-specific gene expression.

Results and Discussion

Circadian Clocks Everywhere.

Soon after the discovery of DBP, two highly related leucine zipper transcription factors, TEF and HLF, were identified in the laboratories of Michael Rosenfeld (9) and Thomas Look (10), respectively. Michael and Thomas graciously provided the cDNA clones for TEF and HLF, and we could thus demonstrate that the Tef and Hlf genes are also expressed in a strongly circadian fashion (11, 12). This motivated us to conduct genetic loss-of-function experiments for the three PAR-domain bZip transcription factors DBP, TEF, and HLF. These experiments revealed unexpected and interesting functions of PAR bZip transcription factors in neurotransmitter homeostasis in the brain (13) and endo- and xenobiotic detoxification in the liver and the small intestine (14). Somewhat disappointingly, however, they also demonstrated that PAR bZip proteins are not essential core clock constituents. Thus, although PAR bZip proteins are expressed in the SCN, triple knockout mice displayed normal circadian locomotor activity (15). Rather, their cyclic expression is governed by the TTFLs described above, and PAR bZip transcription factors thus belong to the class of clock output regulators. Clock (for Circadian locomotor output cycles kaput) was identified as the first mammalian core clock gene by a heroic forward genetics screen by Takahashi et al. (16) and isolated by positional cloning (17). Soon thereafter, three isoforms of mammalian Period genes (mPer1, mPer2, mPer3) were discovered owing to their sequence homology to their Drosophila counterpart per (18–21). Somewhat later, two Cryptochrome genes (mCry1, mCry2) were recognized to be essential for circadian rhythm generation (22). The discovery of mammalian clock genes has been reviewed in detail by Turek and Kolker (23). With the availability of molecular probes for the mRNA and protein products of mammalian core clock genes, their spatial and temporal expression patterns could be examined. Several studies reported that clock genes were not only expressed in the SCN, but also in peripheral organs and even in cells cultured in vitro (23). Aurelio Balsalobre, a postdoc in Schibler’s lab provided evidence that cultured Rat-1 fibroblasts indeed possessed functional circadian clocks whose phases could be synchronized by a serum shock (24). Moreover, the laboratories of Menaker, Tei, and Takahashi demonstrated that slices of many peripheral tissues from rats and mice expressing luciferase from Period gene promoters displayed circadian bioluminescence cycles (25, 26). The daily oscillations of peripheral organ slices and cultured fibroblasts showed damping after a few cycles. Nonetheless, Emi Nagoshi, a postdoc in Schibler’s lab, and Welsh et al. demonstrated by single cell fluorescence and luminescence recordings, respectively, that the oscillators of cultured mouse fibroblasts are self-sustained and cell-autonomous (25, 27). Thus, the gradual damping observed in cell populations was due to a loss of phase coherence between individual cells rather than to the damping of oscillators in individual cells.

The Actin Cytoskeleton of Hepatocytes Undergoes Daily Oscillations.

How does a serum shock synchronize circadian oscillators of cultured cells, and is this finding relevant for understanding how peripheral circadian clocks are phase-entrained in vivo? Answering these questions was far from trivial, because studies by us and others revealed that numerous signaling pathways could synchronize the timekeepers of cultured cells (reviewed in ref. 28). Since serum contains multiple signals found to be potent synchronizers by candidate approaches, we wished to identify signaling pathways involving a rhythmically accumulating blood-borne cue. To this end, Alan Gerber, a PhD student in Schibler’s lab, designed an unbiased screening approach for immediate early transcription factors, dubbed Synthetic TAndem Repeat PROMoter (STAR-PROM) screening, and used it to examine human and rat serum samples harvested at regular intervals around the clock (29). Pleasingly, the STAR-PROM-screening revealed that the stimulation of the myocardin-related transcription factors-serum response factor (MRTF-SRF) signaling pathway depended on the time of day at which the blood samples were harvested. As expected, the maximal stimulation of this pathway was different for temporally staged rat and human samples, presumably because rats and humans are nocturnal and diurnal, respectively. The rhythmically active signaling component in serum was identified as a protein, but its precise identity has not yet been uncovered. The MRTF-SRF pathway, discovered by Treisman et al. is activated through signals promoting actin polymerization (30). Thus, when the transcriptional coactivator MRTF is bound to globular actin (G-actin), it is unable to associate with chromatin-bound SRF and hence incapable of stimulating the transcription of MRTF-SRF target genes. Upon polymerization of G-actin to filamentous actin (F-actin), MRTF is released and becomes available for the activation of MRTF-SRF target genes, which, as revealed by ChIP-seq experiments, include mPer2 (31). If this pathway indeed played a role in the synchronization of peripheral clocks in vivo, actin polymerization would be expected to oscillate in the liver and other peripheral organs. The biochemical separation of G-actin and F-actin in liver homogenates indeed confirmed this assumption (Fig. 1A). Thus, F-actin and G-actin accumulated according to antiphasic daily cycles, with zenith values being reached at about ZT00 and ZT12, respectively (29). Moreover, the staining of histological tissue sections with fluorescent phalloidin, an F-actin-specific dye, revealed massive intracellular and subcortical actin bundles in liver and spleen harvested at ZT00 (ZT stands for Zeitgeber time, where ZT00 and ZT12 correspond to the times when the lights are switched on and off, respectively). In contrast, tissue sections of liver and spleen harvested at ZT12 only displayed traces of subcortical actin bundles and diffuse, weakly stained intracellular F-actin (29) (Fig. 1B). Altogether, these findings indicated that the MRTF-SRF signaling pathway contributes to the synchronization of circadian gene expression in at least some peripheral organs. However, the findings described below suggest that the actin polymerization rhythms in mouse liver may serve yet another purpose, namely, to increase the mechanical robustness of hepatocytes during the night.

Fig. 1.

Daily oscillations of actin polymerization in the liver. (A) Total actin, F-actin, and G-actin were prepared and analyzed from two mice sacrificed at 4-h intervals around the clock according to ref. 27. For the immunoblot experiments shown in the Top panels, the preparations from the two mice sacrificed at each time point were pooled. The actin content in the F- and G-actin fractions and in the unfractionated liver homogenates were quantified from the immunoblots. In the Lower panel, relative fractions of F- and G-actin (depicted as dashed gray and solid black lines, respectively) are expressed as percentages of total actin. ZT = Zeitgeber Time, where ZT00 and ZT12 are the times when the lights were switched on and off, respectively. (B) Fluorescence microscopy images of paraformaldehyde-fixed liver sections stained with phalloidin for F-actin and DAPI (4′,6-diamidino-2-phenylindole) for DNA (nuclei). Representative images of two mice sacrificed at either ZT0 or at ZT12 are shown. (Scale bar, 20 µm.) The panels (A and B) are adapted from ref. 29, with copyright permission from Elsevier.

The Volume, Mass, and Macromolecular Content of the Liver Oscillate during the Day.

The inspection of histological liver sections (Fig. 1B) revealed another difference between livers harvested at ZT00 and ZT12, in that the size of hepatocytes appeared to be larger at the former as compared to the latter time. If the number of hepatocytes per liver remained constant throughout the day, the entire liver would be expected to oscillate in a diurnal manner. As shown in Fig. 2 A and B this turned out to be the case (32). Mice are nocturnal animals, and they thus consume about 80% of their food during the night. When mice were fed ad libitum (AL) or exclusively during the night (NF), their liver mass fluctuated by 34% and 43%, respectively, over the course of the day. In contrast, no statistically different liver masses were detected in mice fed exclusively during the day (DF), although all three groups of animals ingested nearly identical quantities of food. Hence, feeding-fasting and light–dark cycles must be coupled to induce daily fluctuations in liver mass. The diurnal oscillations in liver mass and volume did not just reflect fluctuations in water uptake, as they were accompanied by similarly impressive fluctuations in macromolecular content. Thus, in night-fed mice the contents of total RNA (Fig. 2C) and soluble protein (Fig. 2D) increased by 44% and 65%, respectively, from ZT12 and ZT00, when normalized to DNA. Given that the hepatic content of DNA, an indicator for cell number, remained virtually constant throughout the day (32), this indicated that liver cells accumulated considerably more proteins and RNA during the night than during the day. The daily fluctuations in liver size and actin polymerization are schematically illustrated in Fig. 2E. We wish to emphasize that the daily fluctuation in liver size is not peculiar to mice, as oscillations of similar amplitudes have also been reported for birds (15, 33) and humans (34). What is the physiological significance of the diurnal fluctuations of liver size? The liver is arguably the most metabolically active organ, and several metabolic processes conducted by hepatocytes generate reactive oxygen species (ROS). ROS not only damage DNA, but also ribosomes and proteins (35, 36), and the inhibition of protein synthesis is one of the consequences of oxidative stress (36, 37). We speculate that the oscillation of ribosome and protein accumulation in hepatocytes ensures that every day a fraction of damaged ribosomes and functionally important proteins are replaced with new ones. Through osmosis, the rise in the concentrations of proteins and, probably, glycogen is expected to be compensated by an increase in water and salt uptake, which in turn enlarges the cellular volume.

Fig. 2.

Diurnal changes in size and macromolecular content of the liver. (A) Liver masses determined at Zeitgeber Times ZT0 and ZT12 as percentages of body mass in mice fed ad libitum (AL, N = 4), during the night (NF, N = 9), and during the day (DF, N = 9). Data are represented as the means ± SD; ***P < 0.001; ****P < 10⁻⁶; two-sided Student’s t test; n.s. means “not statistically significant.” (B) Liver masses determined as percentages of body mass at 4-h intervals around the clock in mice fed exclusively during the night or during the day. The data are represented as the means ± SD for 9 mice per time point. (C) Diurnal RNA accumulation in mice fed either exclusively during the night or during the day. The values of whole cell RNA were normalized to the DNA determined in the same extracts. DNA amounts did not change during the day and serve as proxies for cell numbers. Since ribosomal RNA accounts for >80% of whole cell RNA, these data also reflect the diurnal accumulation of ribosomes. The data are represented as the means ± SD for 4 to 8 mice per time point. The RNA/DNA ratio was found to be 1.44 ±SD-fold higher at ZT0 than at ZT12 (N = 8). (D) Accumulation of soluble protein at ZT0 and ZT12 in night-fed and day-fed mice. The values of soluble proteins (100,000 × g supernatants) were normalized to the DNA determined in the same liver homogenates. DNA amounts per whole liver did not change during the day and serve as proxies for cell numbers. The data are represented as the means ± SD for 8 mice per time point. (E) Schematic representation of the diurnal oscillations in liver volume, hepatocyte size, and actin cytoskeleton. Fibrillar actin (F-actin) bundles are shown in green. The items in the scheme are not drawn to scale. The data and figures (A–D) and the cartoon (E) are adapted from ref. 32, with copyright permission from Elsevier.

The Number of Ribosomes May Be Rate-Limiting for the Nocturnal Increase in Protein Synthesis.

As ribosomal RNA (rRNA) accounts for about 80% of whole-cell RNA in the liver and other organs, the increase of RNA during the night indicated a nocturnal surge of ribosome accumulation in hepatocytes. Moreover, the inspection of polyribosome profiles prepared from livers harvested either during the day or during the night revealed two striking differences (32). First, the polyribosomes were heavier at night than during the day, suggesting higher protein synthesis rates during the nocturnal phase. Second, in contrast to diurnal polyribosome profiles, nocturnal polyribosome profiles displayed barely detectable amounts of 40S and 60S free ribosomal subunits. Hence, as soon as these subunits are recycled after the termination of translation, they re-engage in new initiation events. These observations suggested that during the nocturnal activity phase ribosomes become rate-limiting for global protein synthesis. Flore Sinturel, a postdoc in Schibler’s group, and Alan Gerber were interested in examining the molecular mechanisms underlying the daily cycles of ribosome accumulation (32). Mammalian ribosomes consist of four rRNAs (28S, 5.8S, 18S, and 5S) and eighty ribosomal proteins. Three of the four rRNAs (18S, 5.8S, 28S) are processed from a 47S/45S pre-rRNA that is transcribed in the nucleolus by RNA polymerase I. In the liver, the 47S/45S pre-rRNA is synthesized at constant rates and in stoichiometric excess over ribosomal proteins throughout the day. In contrast, ribosomal profiling (Ribo-Seq) experiments revealed that the translation efficiency of mRNAs encoding ribosomal proteins strongly oscillates during the day, reaching the highest efficiency during the night (38, 39). During the day, nucleolar rRNAs are in large excess over ribosomal proteins, and only a fraction of them can therefore be incorporated into complete and functional ribosomal subunits. Superfluous 28S and 18S rRNAs are polyadenylated by PAPD5, and the poly(A)-tails on rRNAs that are not assembled into complete ribosomal subunits mark them for degradation by the nuclear exosome. The cartoon of Fig. 3 displays the putative mechanisms accounting for the synthesis of ribosomes during the day and the night.

Fig. 3.

Putative model of daytime-dependent ribosome synthesis in the liver. The synthesis of 47/45S pre-ribosomal RNA (pre-rRNA) and mRNAs encoding ribosomal proteins (RP mRNAs) is constant throughout the day. However, the translation efficiency of RP mRNAs oscillates during the day, and newly synthesized RPs are thus assumed to be rate-limiting for the assembly of complete, functional ribosomal subunits. Surplus rRNAs not incorporated into complete ribosomal particles are polyadenylated by PAPD5 (a subunit of the TRAMP complex), and polyadenylated rRNAs are degraded by the nuclear exosome. RP synthesis is maximal during the night, and the proportion of rRNAs assembled into functional ribosomal subunits is thus higher during the nocturnal activity phase than during the diurnal resting phase. This leads to a daily oscillation of ribosome accumulation in the cytoplasm. Since ribosomes appear to be rate-limiting for overall translation in hepatocytes, global protein synthesis is higher during the night than during the day. RNA polymerase I complexes are symbolized by light brown hexagons, and RPs associated with small and large subunits are represented as green and light brown ovals, respectively. The cartoon is taken from ref. 32, with copyright permission from Elsevier.

According to the most parsimonious scenario, the ribosomal protein whose synthesis rate is the lowest, can be expected to be rate-limiting for the synthesis of functional ribosomes overall. Among all ribosomal proteins, Rps15a is among the components with the lowest translational output, both in ad libitum–fed and night-fed mice. As shown in Fig. 4A and SI Appendix, Fig. S1A, the peak-to-trough amplitudes of RPS15A synthesis are 4.37 in mice fed during the night and 2.63 in mice fed ad libitum. These values are nearly identical to the median amplitudes determined across all ribosomal proteins in mice under nocturnal (4.1, Fig. 4 B and C) or ad libitum (SI Appendix, Fig. S1 B and C) feeding. The diurnal amplitude distributions of all ribosomal proteins are also compared in SI Appendix, Fig. S2.

Fig. 4.

Ribosomal protein synthesis and ribosome accumulation in the livers of night-fed mice. (A) Relative diurnal translation of Rps15a mRNA in night-fed mice, as revealed by ribosome profiling. RPF = ribosome-protected mRNA fragments; amp = amplitude, expressed as fold change. The results from four independent experiments are shown. The light-blue curve represents a fit to the diurnal translation profile. The data are taken from ref. 38. (B) Diurnal amplitudes (fold changes) of relative translation output measured for all ribosomal proteins of mice fed during the night. Note that the median value and the value obtained for Rps15a (panel A) are very similar. (C) Distribution of the amplitudes (fold changes) of the diurnal translation rates determined for all ribosomal proteins as a function of relative mRNA abundance. (D) Top panel: Cosine-profile for the temporal pattern of ribosomal protein translation rates with an amplitude (fold change) of 4.1. Bottom panel: Temporal rRNA accumulation pattern (assumed to mimic ribosome accumulation) as a function of ribosomal protein translation rates (Top panel, assumed to mimic ribosome synthesis) for an rRNA half-live (hl) of 8.5 h. The mathematical model assumes that ribosomes decay according to first-order kinetics.

If the temporal synthesis profiles of ribosomes paralleled the temporal pattern of ribosomal protein production (Fig. 4 D, Top), and if all ribosomes decayed with the same statistical probability according to first-order kinetics, ribosome half-lives of 8 h and 12 h would fit the simulated diurnal accumulation profiles of rRNA in mice fed during the night (Fig. 2C, NF conditions, and Fig. 4 D, Bottom) and fed ad libitum (SI Appendix, Fig. S1D), respectively (see Methods for mathematical simulation). For night-fed mice the relationship between the temporal patterns of synthesis and accumulation as a function of half-lives is depicted in the simulation of SI Appendix, Fig. S3. Our estimated ribosome half-lives are much shorter than the half-lives of 4 to 5 d determined previously by kinetic labeling approaches for mouse and rat liver (40, 41). Theoretically, a large fraction of radiolabeled nucleotides could be recycled and reused for RNA synthesis. If so, the kinetic labeling experiments actually tracked the specific radioactivity of nucleotide precursor pools, rather than that of rRNAs. However, as outlined below this scenario is virtually impossible. Indeed, experiments with radioactive orotic acid or methionine, which label pyrimidines and posttranscriptionally added methyl groups in rRNA, respectively, revealed the same long half-lives (40). Moreover, double-labeling experiments by Hirsch and Hiatt demonstrated that ribosomal proteins labeled with ¹⁴C-arginine and ribosomal RNAs labeled with ³H-orotic acid decayed with the same half-life of ~5 d in the rat liver (42). Hence, in the radiolabeled tracer reutilization scenario mentioned above, radiolabeled pyrimidines, S-adenosylmethionine (the universal methyl donor synthesized from methionine), and arginine would have to be recycled at exactly the same rate. We consider this to be highly unlikely and therefore propose a different hypothesis to reconcile the diurnal accumulation of ribosomes with the long half-lives reported previously. This speculative model is schematically illustrated in SI Appendix, Fig. S4. Aging ribosomes may accumulate modifications in rRNAs and/or RPs that eventually render them dysfunctional. Subsequently, disabled ribosomes are recognized and marked for degradation by ribophagy (43, 44) and/or other mechanisms. At least in rat liver the time for newly synthesized ribosomes to accumulate critical levels of degradation may last several days, as no loss of radiolabeled rRNA and RPs was observed during the first 2 d following the intraperitoneal injection of the radioactive tracers (see figure 1 in ref. 42). However, once marked for degradation, dysfunctional ribosomes may decay with very short half-lives of a few hours, and they are replaced with new, functional ones by daily surges of ribosome synthesis (SI Appendix, Fig. S4). It is interesting to note in this context that the mechanistic Target of Rapamycin (mTOR) pathway may be implicated in the regulation of both the diurnal synthesis and the diurnal decay of ribosomes. As aforementioned, the synthesis of RPs appears to be rate-limiting for the synthesis of ribosomes. The diurnal translation of RP-encoding mRNAs is known to be regulated by mTOR, whose activity is controlled by feeding-fasting cycles (45). During the fasting period mTOR activity is low, and this may promote the degradation of disabled ribosomes through recognition by the ribophagy receptor NUFIP1 (44).

Synchronization of Hepatic and Other Peripheral Oscillators.

After the discovery that serum can synchronize the circadian oscillators of cultured fibroblasts (24), many signals were found to reset the phase of peripheral clocks in vitro. These include glucocorticoid hormones, retinoic acids, fibroblast growth factor (FGF), endothelin, prostaglandins, forskolin, tumor promoters like TGFβ, Ca²⁺, glucose, synthetic ligands of the orphan receptor REV-ERBα, and body temperature rhythms (reviewed in ref. 46). In mice and rats, feeding-fasting cycles appear to be the dominant Zeitgebers for circadian clocks of many, if not most peripheral tissues (47, 48), and it has been postulated that signaling through Insulin/IGF-1, mTOR, and ERK kinases plays an important role in this process (49, 50). The levels of glucocorticoid hormones in blood oscillate with a large amplitude in laboratory rodents and humans, and they indeed participate in the synchronization of circadian clocks in liver and other peripheral organs (51–53). As mentioned above, diurnal rhythms of actin polymerization drive the cyclic activity of MRTF-SRF, which helps synchronizing liver clocks through the rhythmic expression of PER2 (29). Perhaps most surprisingly, simulated body temperature rhythms imposed on cultured cells phase-entrain their circadian clocks (54, 55), and experiments with mice exposed to daily temperature cycles indicate that body temperature rhythms also participate in the synchronization of peripheral oscillators in vivo (54). As shown by genetic and pharmaceutical loss-of-function experiments, Heat Shock transcription Factor 1 (HSF1) is required for the temperature-dependent synchronization of peripheral oscillators (55, 56). In the liver, HSF1 is translocated from the cytoplasm to the nucleus during the night, when body temperature reaches zenith values (57). Moreover, the Per2 promoter encompasses five phylogenetically conserved heat shock elements (HSEs) (58), which serve as DNA recognition sequences for HSF1. Therefore, HSF1 may synchronize circadian clocks of peripheral organs by contributing to the cyclic transcription of Per2. Feeding fasting cycles probably contribute to the oscillation of body temperature through postprandial thermogenesis. To what extent other signaling pathways involved in the synchronization of peripheral clocks are interconnected remains an open question.

In cultured cells and explants of peripheral tissues, the transcription factors BMAL1 and CLOCK are required to keep the circadian clock ticking (59, 60). Yet, the circadian expression of PERs and CRYs, the core clock components of the negative limb of the TTFL, is similar in wild-type mice and mice devoid of BMAL1 or CLOCK. Therefore, the circadian transcription in peripheral tissues of intact animals must be driven by immediate early transcription factors sensing rhythmic systemic signals, rather than by core clock transcription factors. The latter are, however, required to drive the transcription of clock-controlled genes that assume important roles in daytime-dependent liver functions (58, 61). The observation that intact circadian oscillators in hepatocytes are dispensable for the regulation of diurnal liver physiology begs the question of why peripheral clocks are self-sustained and cell-autonomous. The answer may be rather trivial: Once a self-sustained and cell-autonomous clockwork circuitry was established during phylogeny, its function was maintained in all cells of the body.

Remarkably, the SCN master timekeeper does not respond to the signaling pathways it orchestrates directly or indirectly to phase-entrain peripheral circadian clocks. Thus, its phase is not affected by feeding-fasting rhythms, glucocorticoid hormones, insulin receptor signaling, or body temperature rhythms. Moreover, circadian gene expression with high amplitudes is maintained for many weeks in SCN slice cultures, and a change of culture medium, corresponding to a serum shock, does not significantly affect its circadian phase (62, 63). Hence, the SCN appears to be refractory to the blood-borne cues which rapidly synchronize the oscillators of peripheral cells or tissues kept in culture. Why is the SCN resilient to the signals it uses to synchronize peripheral clocks? Conceivably, this property allows the SCN to rapidly reset the phase of peripheral clocks as soon as the signals disappear, which uncouple the phase of peripheral oscillators from that imposed by the master pacemaker.

But how do we know that the SCN indeed synchronizes subsidiary clocks in peripheral tissues? Although this assumption has been widely accepted in the field, it was not trivial to support it by compelling in vivo evidence. SCN-lesioned animals are arrhythmic, and the phase of circadian gene expression in peripheral organs can thus not be assessed by temporal series of biochemical experiments measuring mRNA or protein accumulation. The only reliable fashion of examining whether the phase coherence between peripheral clocks is lost in SCN-lesioned animals requires the recording of circadian gene expression in real time in individual animals. To this end, we engineered a device, dubbed RT-Biolumicorder, which allowed us to monitor bioluminescence emitted by peripheral tissues in real time in freely moving mice (Fig. 5A). Briefly, the apparatus consists of a lightproof cylindrical cage with reflecting walls that guide photons to the surface of a highly sensitive photomultiplier tube (PMT). The PMT records the bioluminescence produced by circadian luciferase reporter genes expressed in peripheral tissues of mice receiving luciferin through an implanted micro-osmotic pump or via the drinking water. Simultaneously, a passive infrared sensor (PIR) tracks the locomotor activity of the mouse placed into the RT-Biolumicorder (64). In addition, the food dispenser of this device can be automatically programmed to impose feeding-fasting cycles, if desired. Fig. 5B displays the PMT recordings of sham-operated (SCN proficient) and SCN-lesioned (SCN-deficient) PER2::LUC mice, which express a PER2-luciferase fusion protein in nearly all cells (65). As expected, the bioluminescence recordings of the sham-operated PER2::LUC mouse are rhythmic, with a period of approximately 24 h. In contrast, the amplitude of bioluminescence cycles monitored from the behaviorally arrhythmic SCN-lesioned mouse is strongly blunted. Hence, the phase coherence between peripheral organs is almost lost in the absence of the SCN master pacemaker (66). As depicted in Fig. 5B, imposed feeding-fasting rhythms rapidly synchronize circadian PER2::LUC expression in SCN-lesioned mice.

Fig. 5.

Realtime recording of circadian gene expression in freely moving mice. (A) The RT-Biolumicorder consists of a cylindrical cage with photon-reflecting walls equipped with a PMT, a programmable feeding machine, a water flask, and an infrared sensor recording the locomotor activity of mice. The device contains a large reflecting cone on top of the cage (external cone) that projects photons to the PMT and a small reflecting cone in the center of the floor (central cone) that projects photons to the reflecting walls. The schematic representation of the RT-Biolumicorder is taken from ref. 64, with copyright permission from Cold Spring Harbor Laboratory Press. (B) Whole body bioluminescence recordings of PER2::luciferase expression in SCN-proficient, sham-operated mice (Top) and SCN-lesioned mice (Bottom) kept in constant darkness. The feeding regimens are reported above the graphs. The black lines within the colored raw data (1 min photon counts) reflect 120-min time-binned traces. Note that 12-h/12-h feeding-fasting cycles establish the synchronization of circadian PER2::luciferase expression in SCN-lesioned mice (Bottom). The data are taken from ref. 66. (C) Bioluminescence recordings of sham-operated SCN-proficient mice (Top) and SCN-deficient mice (Bottom) transduced with an adenoviral vector harboring a Rev-erbα–luciferase reporter gene. As demonstrated in ref. 66, the bioluminescence of transduced mice is produced exclusively by the liver. (D) Cartoon schematically summarizing the data depicted in panels (B and C). Panels (B–D) were adapted from ref. 66, with permission license BY-NC/4.0/ from Cold Spring Harbor Laboratory Press.

Phase Coherence in the Liver.

By transducing a Rev-erbα-luciferase reporter gene into the liver via an adenoviral vector, we could also monitor circadian gene expression exclusively in the livers of sham-operated and SCN-lesioned mice. Surprisingly, the RT-Biolumicorder revealed circadian bioluminescence cycles with similarly high amplitudes in both SCN-proficient and SCN-deficient animals (Fig. 5C). Rhythmic liver Rev-erbα-luciferase expression persisted even in mice harboring functional oscillators exclusively in hepatocytes (66). We concluded therefore that hepatocyte oscillators must be coupled (schematically illustrated in Fig. 5D), although the mechanisms by which phase information is exchanged between these cells remain elusive. Conceivably, gap junctions connecting hepatocytes and pannexin hemichannels between hepatocytes and the extracellular space (67) are likely involved. In addition, or alternatively, paracrine mechanisms may participate in establishing phase coherence in the liver. One possible candidate for paracrine signal is the endocrine fibroblast growth factor FGF21. It is produced by hepatocytes in a circadian fashion (68) and induces immediate early gene expression in the liver (69). TGF-β, may be another candidate participating in the establishment of phase coherence among hepatocytes. This paracrine growth factor has been reported to contribute to the phase coupling of two- and three-dimensional cultures of U2OS cells (70), and TGF-B signaling follows a diurnal rhythm in the mouse liver (71). Given that phase coherence persists in mice that carry circadian clocks exclusively in hepatocytes, it is unlikely that signals generated by nonhepatic tissues are required for this process. Although the bioluminescence profiles of SCN-lesioned PER2::LUC mice recorded by the RT-Biolumicorder indicate that phase coherence is lost between organs, weak residual circadian peaks can still be detected. At least in part this may be due to the persistence of phase coherence in the liver even in the absence of a master pacemaker.

Our results obtained by RT-Biolumicorder recordings of mice expressing circadian luciferase reporter genes in all cells, or specifically in the liver, are consistent with the findings reported by Takahashi et al. (65). These authors monitored circadian PER2::LUC expression in explants of several tissues of intact and SCN-lesioned mice. Their results suggest that phase-coupling within organs is widespread, but that the synchrony between different peripheral organs gets lost in the absence of a master clock. Phase coherence between circadian oscillators has first been described for the SCN, in which paracrine and electrical signaling mechanisms account for the phase-coupling among neurons (72). Whereas the synchrony between individual cells persists in SCN preparations kept in vitro, it is rapidly lost in slice-cultures of liver and other peripheral tissues (65).

Concluding Remarks

Due to the immense complexity of living organisms, it is virtually impossible in biology to formulate a theoretical model capable of predicting the outcome of a nontrivial experiment. Therefore, research in life sciences is frequently guided by serendipity rather than analytical thinking. With some good luck, the answers revealed by experimentation may turn out to be more interesting than the questions addressed. As outlined in the Introduction section, we entered the field of circadian rhythms by resolving the discrepancy of results obtained by two collaborators studying the transcription factor DBP. Further studies in the lab revealed that self-sustained and cell-autonomous circadian oscillators are operative even in cultured fibroblasts. These findings and subsequent studies by the Menaker and Takahashi groups led to the currently held concept of a hierarchically organized timing system, in which a central pacemaker in the brain’s SCN synchronizes the clocks of peripheral tissues through a multitude of signaling pathways. Our studies on the liver revealed two observations that were again unpredictable. First, the entire organ oscillates in a daily fashion with regard to mass, volume, cell size, actin skeleton, and macromolecular content, and these fluctuations are driven by the integration of feeding-fasting rhythms and light–dark cycles. Second, intact hepatocyte clocks are not required for the orchestration of diurnal liver physiology. So why do nearly all mammalian cell types harbor self-sustained and cell-autonomous oscillators? When contemplating about how living organisms function, it helps to recall a quote of the grand late Sydney Brenner “The big lesson to learn here is that in science, only mathematics is the art of the perfect. Physics is the art of the optimal, and biology is the art of the satisfactory: If it works, you keep it; if it does not, you get rid of it.”

Methods

Actin Polymerization (Fig. 1).

The methods for the biochemical fractionation of G-actin and F-actin and the staining of F-actin in histological sections have been described in detail in ref. 29.

Liver Mass, Hepatocyte Size, and Macromolecular Contents of the Liver (Fig. 2).

The methods used to determine diurnal oscillations in liver mass, hepatocyte size, and RNA and protein accumulation have been described in detail in ref. 32.

Ribosome Profiling (Fig. 4 and SI Appendix, Figs. S1 and S2).

The techniques used to perform ribosomal profiling around the clock have been described in detail in refs. 38 and 39. The data reported in ref. 38 have been used for this study. The data have been deposited in the Gene Expression Omnibus database under accession GSE73554.

Bioluminescence Recording of Freely Moving Animals (Fig. 5).

The long-term monitoring of bioluminescence in freely moving mice using the RT-Biolumicorder has been described in detail in refs. 64 and 66.

Mathematical Simulations.

Estimation of ribosome half-lives (Fig. 4 and SI Appendix, Fig. S1).

In this simulation, it is assumed that all ribosomes decay with the same probability.

To link ribosomal protein synthesis, accumulation, and ribosome half live, we used a first-order kinetic model for the accumulation $a\left(t\right)$:

$$\frac{\mathit{da}}{\mathit{dt}}=s\left(t\right)-k·a(t),$$

with a 24 h periodic synthesis function $s\left(t\right)=s(1+ϵ·\text{cos}(\omega t))$ and a degradation rate $k=\frac{1}{\tau }$, which yields the accumulation $\begin{array}{c}a\left(t\right)=s·\tau (1+{ϵ}_a\text{cos}(\omega (t-t_d))\omega =\frac{2\pi }{24}\hfill \end{array}$ is the frequency and $ϵ$ in $[0,1]$ is the relative amplitude of the synthesis. Then

$${ϵ}_a=ϵ·\frac{1}{\sqrt{1+(\omega ·\tau {)}^2}},$$ [1]

is the relative amplitude of $a\left(t\right)$. Peak-to-trough ratios are related to relative amplitudes via $r=\frac{ϵ+1}{ϵ-1}$ or reciprocally $ϵ=\frac{r-1}{r+1}$. t_d is the delay ($\text{tan}\left(\omega ·t_d\right)=\omega ·\tau$) in the accumulation relative to the synthesis (not used here). Eq. 1 is used in Fig. 4D and SI Appendix, Figs. S1 and S3 to identify the half live ${\tau }_{1/2}=\ln (2)·\tau$ from the measured fold change amplitudes in the synthesis and accumulation.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The ribosome profiling data have been deposited in the Gene Expression Omnibus database under accession GSE73554 (73). All other data are included in the article and/or SI Appendix.