Circadian clock-mediated control of stem cell division and differentiation: beyond night and day

Exhaustive research over the past few decades has begun to elucidate the full range of human physiology that is regulated in synchrony with the solar day. With regard to neuronal function, this includes not only the control of sleep and wakefulness, but also modulation of mood, cognition, sensory acuity, breathing rate and body temperature (Schmidt et al., 2007). Nearly all aspects of digestion and detoxication – from gastric emptying time to fat processing and xenobiotic degradation by the liver – are under circadian control (Dallmann et al., 2014). Many aspects of the circulatory and immune systems, including heartbeat and blood pressure, vascular leakage and even plasma composition, are also regulated daily (Dallmann et al., 2012; Scheiermann et al., 2013).

Underlying this panorama of circadian physiology are circadian clocks that regulate cellular and molecular processes at all levels: in each tissue examined, the expression of one in ten genes is regulated in a circadian fashion, either through circadian initiation of transcription or through circadian control of post-transcriptional processes, such as elongation and message stability (Lim and Allada, 2013). Mitochondrial activity is also regulated by the circadian clock (Peek et al., 2013), along with a variety of intracellular signaling cascades (Robles et al., 2014). Supporting this widespread control, the circadian clock mechanism itself is cell-autonomous; most cells of the human body possess the same molecular clockwork. These clocks are then synchronized to one another via redundant systemic cues to ensure optimum correspondence with the environment (Brown and Azzi, 2013). For the most part, these cues originate from the suprachiasmatic nuclei (SCN) of the hypothalamus: the ‘master clock’ tissue in mammals. The SCN directs the timing of body clocks under most circumstances via autonomous nervous control of hormones such as glucocorticoids, or through direct innervation of other brain regions to send indirect signals via body temperature and rhythmic food intake (Dibner et al., 2010). Because the SCN is itself synchronized to light via the retinohypothalamic tract, the result is a flexible system of clocks, each with an intrinsic period of about one day, that is constantly readjusted to the timing of environmental light. At the same time, because of this redundant and partly autonomous hierarchical organization, considerable flexibility is possible. For example, repeated feeding signals at uncharacteristic times of the day in rodents result in the phase-shifting of clocks in peripheral tissues, such as the intestine, liver and heart, as well as in food anticipatory behavior, even as the SCN continues to track the daily light cycle (Patton and Mistlberger, 2013).

For most aspects of circadian-controlled physiology, one can easily imagine why such regulation exists: in a world divided into day and night, the rhythmic control of digestive function and detoxication can be synchronized with rhythmic food intake that is directed by rhythmic sleep and wakefulness. Likewise, diurnal cardiac function parallels the expected changes in energy needs on a systemic level, and circadian regulation of mitochondrial energy production could reflect the same optimization on a cellular level. In large part, teleological arguments for the evolutionary necessity of circadian clocks propose that the processes that clocks control run most optimally when correctly coordinated with the environment.

However, increasing evidence suggests that the circadian clock controls a much greater proportion of rhythmic physiology than can be easily explained in this fashion. Nowhere is this conundrum more apparent than in the case of stem cells, development and tissue regeneration. In plants, circadian control of development is well established (Nagel and Kay, 2012; Thines and Harmon, 2011), but recent studies suggest that circadian clocks could play a more important role in animals than previously suspected, even in developmental and regenerative processes that are considerably faster or slower than a biological day. In this Review, I first provide an overview of the molecular mechanisms involved in circadian clocks and then discuss how such mechanisms can influence stem cell biology and hence tissue development, homeostasis and regeneration.

The molecular basis of known circadian clocks involves feedback loops of transcription and translation (Fig. 1). Therein, dedicated clock proteins [e.g. the heterodimeric transcription factors CLOCK and BMAL1 (ARNTL – Mouse Genome Informatics) and their homologs] activate the transcription of other clock genes (e.g. period genes Per1-Per3 and the cryptochromes Cry1 and Cry2), the protein products of which repress their own transcription by blocking CLOCK:BMAL1-mediated activation (Brown et al., 2012). Other interlocked feedback loops provide essential robustness within this simple mechanism. For example, CLOCK:BMAL1 promotes transcription of the Rev-Erbα (Nr1d1 – Mouse Genome Informatics) nuclear receptor, which itself represses Bmal1 expression (Preitner et al., 2002), and oscillatory cAMP-dependent signaling within the SCN also reinforces circadian transcriptional oscillations (O'Neill et al., 2008). A host of other proteins – kinases and phosphatases (Reischl and Kramer, 2011), chromatin modifiers (Sahar and Sassone-Corsi, 2013) and RNA-binding factors (Kowalska et al., 2012) – contribute necessary roles to the precise regulation of clock RNA and protein levels (Fig. 2). In turn, the cellular regulation of clock-controlled processes can be achieved by the same cis-acting elements that direct clock gene expression, by circadian cascades of downstream transcription factors or by systemic regulation via hormones, metabolic products and body temperature (Brown and Azzi, 2013). Recently, transcription-independent oscillation of protein oxidation has also been reported, but its mechanism remains as yet unknown (O'Neill and Reddy, 2011).

One example of a basic cell biological process that is modulated by the clock is the cell division cycle in mammals, which can run at widely varying speeds but nevertheless shows stage-dependent circadian gating. The cell cycle itself can be loosely divided into stages labeled G1, S, G2 and M, in which S indicates DNA synthesis/replication and M indicates mitosis. Both stages are flanked by intermediary periods (G1 and G2, respectively), terminating in kinase-controlled ‘checkpoints’. On a molecular level, circadian control of the cell cycle has been documented at multiple levels and in multiple different scenarios (Fig. 3). For example, an initial report of cell division in the regenerating mouse liver documented circadian transcription of the Wee1 checkpoint gene, suggesting control at the G2/M checkpoint. Consistent with this idea, CDC2, the target of the WEE1 kinase, shows circadian phosphorylation in the liver (Matsuo et al., 2003). A G1/S checkpoint regulator in hepatocytes, the p21-waf1 (Cdkn1a – Mouse Genome Informatics) gene, is also clock-regulated via REV-ERBα response elements (RREs) within its promoter (Grechez-Cassiau et al., 2008). Further studies in cell culture have illuminated an even more complicated picture. Cultured mouse fibroblasts show a complex clock gating pattern that suggests multiple control points (Feillet et al., 2014; Nagoshi et al., 2004). Other subsequent studies have highlighted potential control via the CHK1/2 (CHEK1/2 – Mouse Genome Informatics) proteins binding to the clock-associated TIM protein (Unsal-Kacmaz et al., 2005; Yang et al., 2010) or via transcriptional control of the p16-Ink4A (Cdkn2a – Mouse Genome Informatics) locus by the clock protein NONO, which acts as a partner for PER proteins. Such regulations imply circadian G2/M checkpoint control and, consistent with the importance of these regulations, elimination of NONO has been shown to be sufficient to eliminate circadian cell cycle gating in fibroblasts (Kowalska et al., 2013). Recent cell-based studies have also provided strong evidence in reverse, for modulation of the circadian cycle by cell division (Bieler et al., 2014). Therefore, the ultimate picture that emerges is one of complex multilayer control, the cell- and tissue specificity of which still remains to be established.

The observations that the circadian clock regulates cell division in cultured mammalian cells (Nagoshi et al., 2004) and in adult animal tissues (Matsuo et al., 2003) can be understood in terms of diurnal optimization. As the oxidation reactions inherent to metabolic processes result in the production of mutation-causing free radicals, circadian control of the cell cycle to segregate DNA replication away from periods of maximum respiration would optimize the time available for the DNA repair process, without relying upon independent checkpoint mechanisms to pause the cell cycle during DNA repair (Chen and McKnight, 2007). The observation that multiple types of DNA damage repair are themselves coordinately regulated by the circadian clock supports this hypothesis (Sancar et al., 2010). Similarly, clock-deficient mouse strains (notably Per2-deficient mice) have been documented to have increased spontaneous cancer rates in some, but not all studies (Antoch et al., 2013; Fu et al., 2002). This reflects possible increases in DNA damage and supports a role for clock genes as tumor suppressors (Cao et al., 2009; Yang et al., 2009). However, these observations are a simplification: other clock-deficient mice do not show increased cancer rates (Antoch et al., 2008; Gauger and Sancar, 2005), and cell lines from multiple clock-deficient strains possess normal DNA repair properties (Gaddameedhi et al., 2012). Therefore, other benefits of circadian regulation of the cell cycle might also exist, as we explore below.

The direct consequences of circadian control of the cell division throughout the body have been established in numerous recent studies. For example, elimination of the SCN ‘master clock’ results in tumor growth two to three times faster than in controls (Filipski et al., 2002). Consistent with the possible importance of this regulation, clock protein misexpression and/or a lack of circadian control has been documented in multiple tumor types (Hwang-Verslues et al., 2013; Luo et al., 2012; Zhao et al., 2013) and immortalized cell lines (Yeom et al., 2010). In normal physiology, circadian cell division has been documented in adult hippocampal neurogenesis (Bouchard-Cannon et al., 2013), in intestinal and skin epithelial cell division (Geyfman et al., 2012; Janich et al., 2013; Karpowicz et al., 2013), and in multiple immune cell populations (Fortier et al., 2011; Keller et al., 2009) – essentially anywhere that cell division occurs in adult animals.

Therefore, beyond a context in DNA repair and cancer prevention, circadian control could serve as a metronome to coordinate complex processes, whether or not they are related to the 24 h day. For example, in the case of dermal wound repair, disruption of circadian gating by eliminating PERIOD clock repressor proteins resulted in fibroblast and keratinocyte hyperproliferation, and in collagen undersecretion; eliminating the BMAL1 clock activator protein resulted in underproliferation of the same cells, whereas collagen was overproduced (Kowalska et al., 2013). One possible explanation for this phenomenon could be a reciprocal regulation of matrix production and cell division by fibroblasts, mediated via the circadian clock. Such an interpretation is supported by similar studies of regeneration in the Drosophila intestine after chemically induced damage. Here, not only is the division of intestinal stem cells (ISCs) regulated in circadian fashion, but so too is the expression of hundreds of other genes in the healing intestine, regulating everything from the stress response to cell polarity (Karpowicz et al., 2013). In mammals, similar widespread circadian control of epithelial cell processes has also been documented and is necessary for gut homeostasis (Mukherji et al., 2013).

Analogous arguments can be drawn for most other instances of circadian cell division. The observed circadian regulation of adult hippocampal neurogenesis (Bouchard-Cannon et al., 2013) could complement circadian regulation of dendritic spine formation and stabilization (Liston et al., 2013), allowing for more efficient reintegration of new neurons into the adult brain. A circadian clock is also present in tooth ameloblasts, where it controls antiphase rhythms of enamel matrix endocytosis and secretion, as well as ameloblast maturation (Lacruz et al., 2012; Zheng et al., 2013). Circadian oscillations in the release of hematopoietic stem cells (HSCs), coupled with antiphase circadian expression of the chemokine Cxcl12 and coordinated with GSK3β-dependent changes in HSC migratory properties (Lapid et al., 2013), could result in the coordinated release of HSCs and repopulation of the bone marrow stem cell niche (Méndez-Ferrer et al., 2009, 2008). Analogous rhythmic division and tissue colonization has also been documented for monocytes (Nguyen et al., 2013).

Because circadian clocks control the expression of cell cycle regulatory genes such as p16-Ink4A (Kowalska et al., 2013) and pathways, such as the NOTCH, WNT and HIPPO signaling pathways (Karpowicz et al., 2013), an immediate and obvious prediction of this widespread control is that not only cell division but also cell differentiation, e.g. that of stem cell populations, might be under circadian influence. As with other cell types, stem cells follow the ordered stages of the cell cycle, but they may also enter into a dormant G0 phase (Fig. 3), during which time circadian heterogeneity could play an important role in driving cell fate decisions, as I describe below.

Studies have shown that clock genes can indeed directly influence stem and progenitor cell fate. For example, the manipulation of Per3 expression influences adipocyte fate, probably by regulating the expression of peroxisome proliferator-activated receptor γ (Pparg – Mouse Genome Informatics), which plays a key role in adipose tissue development (Costa et al., 2011). A similar role has been demonstrated for the circadian deadenylase nocturnin during adipogenesis (Kawai et al., 2010). Disruption of the clock gene Bmal1 also led to increased adipogenesis and, correspondingly, the attenuation of Bmal1 expression in pre-adipocytes in vitro led to downregulation of the WNT signaling pathway – probably through transcriptional control of multiple pathway members – and increased adipogenesis (Guo et al., 2012). The converse occurred upon Bmal1 overexpression. An inversed role has been ascribed to Bmal1 in the case of myogenesis: here, knockout of Bmal1 led to reduced muscle mass and blunted expression of key myogenic regulators, presumably due to circadian control of WNT pathway activity (Chatterjee et al., 2013). Bmal1 deficiency is also linked to reduced osteoblast proliferation in adult mice, although a circadian role here has not been proven (Chen et al., 2012). Furthermore, in the complex differentiation processes leading to the development of the immune system, T_h17 T helper cell development is regulated by the circadian factors REV-ERBα and RORγ (RORC – Mouse Genome Informatics) (Yu et al., 2013), and circadian cytokine release in vivo and in vitro is likely to influence other immune cell subpopulations (Fortier et al., 2011; Keller et al., 2009; Wang et al., 2011). Finally, in addition to regulating metabolism, multiple circadian clock genes directly regulate the size and differentiation of pancreatic islets (Marcheva et al., 2010).

Probably the most-studied case of circadian regulation of stem cells so far is that of the hair follicle. Hair tissues proceed through alternate stages of hair production (anagen) and inactivity (telogen), in which spatially distinct niches harbor populations of dormant stem cells, dividing stem cells or mixtures of the two at different times. Different roles for the circadian clock have been proposed in each of these phases and their transitions. During anagen, proliferating stem cells of the hair follicle show marked circadian oscillations in cell division similar to those outlined above for other tissues. These oscillations in proliferation result in time-of-day-dependent hair growth, but circadian clock mutants surprisingly do not show altered hair dimensions overall. Instead, these animals lose a time-of-day-dependent genoprotective effect: whereas wild-type mice show time-of-day-dependent radiation toxicity, Clock mutants do not. Thus, a circadian clock has been suggested here to optimize DNA repair and replication cycles, primarily via CDC2/CYCLIN B-mediated synchronization of the G2/M checkpoint (Plikus et al., 2013).

During telogen and the transition to anagen, a very different picture has emerged. In the transition to anagen, a circadian pattern of cell division was again seen in dividing cells, but this time gating progression to anagen (Lin et al., 2009). In this study, the authors documented a circadian blockage of the G1 phase of the cell cycle, possibly due to clock-controlled elevation of p21. By contrast, the study of circadian function in dormant stem cells of the hair bulge (the permanent region of the hair follicle) during telogen resulted in a remarkable finding: although hair follicle ‘stemness’ genes – including those encoding members of the WNT, TGFβ, NOTCH and SHH pathways – in hair follicle stem cells were expressed in a circadian fashion dependent upon the expected clock genes, the phases of circadian oscillations in the hair follicle stem cell compartment were completely unsynchronized (Janich et al., 2011). In addition, genetic ablation of the circadian clock in these cells resulted in significant defects in the dynamics of hair follicle stem cell activation. With this finding, the authors suggested a possible second major role for circadian control of developmental processes: the generation of heterogeneity within the stem cell compartment (Janich et al., 2011). Given the importance of juxtacrine signaling during development, this heterogeneity might be important for optimizing cell fate decisions or could simply permit a broad population of cells expressing different receptors to respond to different cues at different times. In a subsequent study (Janich et al., 2013), the authors demonstrated the second possibility in vitro by showing that cultured and clock-synchronized keratinocyte stem cells responded better to certain differentiation cues at some times of day than at others. Consistent with their hypothesis, cells expressed different proliferation and differentiation pathway genes at different times of day. During times corresponding to early morning, differentiation pathways were high, whereas in the evening pathways corresponding to DNA replication and cell division predominated. Indeed, based on their gene expression data, the authors were able to distinguish ‘functional intervals’ during the stem cell differentiation process, in which activities of particular signaling pathways displayed maximum expression. In turn, clock genes themselves displayed ordered patterns of expression across these functional intervals, raising the possibility that circadian control might be necessary to achieve temporal orchestration of keratinocyte stem cell development within time windows even shorter than simply day and night (Janich et al., 2013). Further experiments using clock-deficient cells will no doubt shed further light upon this interesting hypothesis.

Given the circadian coordination of transcriptomic programs present in adult stem cells such as the keratinocyte stem cells described above, it is logical to wonder which types of stem cells show circadian oscillations, and when in development they do so. In vitro, multiple laboratories have demonstrated that, whereas pluripotent embryonic stem (ES) cells show no visible rhythmicity, their differentiated counterparts – even multipotent neural precursor cells (NPCs) – show robust circadian oscillations of gene expression (Kowalska et al., 2010; Yagita et al., 2010). In elegant experiments, it was shown that these rhythms are lost upon de-differentiation of these cells back to ES cells, and regained upon re-differentiation to NPCs (Yagita et al., 2010). Thus, cell-intrinsic rhythms of circadian gene expression are probably present at the very earliest stages of development. It has even been suggested that circadian rhythms of energy consumption in ES cells might briefly precede the emergence of canonical circadian transcriptional oscillations (Paulose et al., 2012).

In spite of the evident potential for prenatal circadian oscillations that this research suggests, widespread and overt rhythms of behavior and physiology in mammals probably emerge much later in development, only after birth (Dolatshad et al., 2010). Circadian expression of certain clock genes in some tissues can be seen as early as embryonic day 20 (E20), but full adult-like oscillations develop as late as postnatal day 30 (P30), coincident with the emergence of circadian locomotor activity (Sladek et al., 2007). To date, no studies at cellular resolution have been performed in vivo in mammals to ascertain whether clocks in early developing tissues are absent or simply desynchronized, although in utero observations of rat embryos containing a circadian reporter also suggested that the first coherent circadian oscillations occur around or shortly before birth (Saxena et al., 2007). Nevertheless, it is likely that coherent circadian rhythms must exist in at least some tissues prior to birth, because phase at weaning is influenced directly by prenatal light timing, timed injection of a dopamine agonist or by exogenous melatonin (Davis, 1997; Davis and Gorski, 1985; Viswanathan et al., 1994).

A relatively late emergence of coherent circadian oscillations has also been documented in chicks, but evidence suggests that substantially earlier unsynchronized cell-autonomous oscillations exist (Gonçalves et al., 2012). In other vertebrates such as zebrafish, synchronized circadian oscillations of both the cell cycle (Dekens et al., 2003) and of cell cycle regulatory genes, including p20 and p21, during embryonic development are well documented (Laranjeiro et al., 2013). The evolutionary or developmental benefit of this regulation remains mostly unknown, but a recent study showed that genetic or environmental disruption of these circadian oscillations resulted in defects in angiogenesis (Jensen et al., 2012).

Consistent with the late emergence of coherent circadian rhythmicity during mammalian development, the important roles shown by many groups for circadian rhythms in stem cells of adult mammals appear to be unsubstantiated during embryogenesis. In mouse models, disruption of any of the canonical circadian clock genes still results in the birth of a basically normal mouse. In contrast to the severe defects observed in circadian clock-deficient adult mice during epidermal wound healing, the skin in newborns is normal in morphology (Kowalska et al., 2013). Furthermore, despite the alterations in hair follicle regeneration observed in Clock mutants, initial hair growth in pups is unchanged (Plikus et al., 2013). Only in some instances has the importance of non-circadian expression of Clock genes during mammalian development been postulated, e.g. in the case of the developing pancreas through regulation of WNT, NOTCH and cell division rates (Li et al., 2007). Thus, one is forced to assume that many of the important regulatory processes described above are substituted for or are unnecessary within the context of embryonic development, perhaps due to the rapid pace of cell division that makes circadian regulation suboptimal. Such ideas, however, are pure speculation.

It is clear that circadian clock-mediated regulation of both the division and differentiation of stem cells plays an important role in adult tissue renewal. The control of these processes by the circadian clock might not only be essential for correct healing and regeneration, but might also be of benefit during ageing. As already noted above, mice deficient in different circadian clock genes suffer from pathologies ranging from diabetes to arthritis and cancer (Antoch et al., 2008; Bunger et al., 2005; Fu et al., 2002; Marcheva et al., 2010).

In this Review, I have postulated that, broadly speaking, circadian clocks could serve two equally important roles in stem cell development and differentiation. On the one hand, circadian gating of multiple aspects of complex tissue homeostasis and regeneration could permit the optimal coordination of mutually beneficial or antagonistic processes. On the other hand, dephased oscillators could provide a source of heterogeneity for stem cells, allowing them to respond optimally to a variety of signals (Fig. 3). If the latter were true, maintenance of such heterogeneity would represent an issue important in stem cell biology. Normal peripheral circadian oscillators are entrained to a particular phase by a wealth of direct and indirect timing cues from the environment and from physiology (Saini et al., 2011). Escape from such entrainment could be envisioned in a variety of ways. Stem cells might fail to respond to normal cues, or they might themselves represent a heterogeneous population that responds to different cues. Finally, the epidermal stem cells in which clock heterogeneity has been convincingly documented are themselves in a unique environment of extreme temperature variation, and temperature represents an important entrainment signal to peripheral oscillators (Brown et al., 2002; Morf and Schibler, 2013). Therefore, the epidermal stem cells might simply be confused by conflicting environmental signals. Further experiments will no doubt soon shed light on this interesting issue.

Altogether, a wealth of research suggests that the circadian clock indeed controls more than just daily timekeeping. Understanding such regulation could hold clues to novel treatments for disease. Moreover, it could also help explain the mechanisms by which chronic ‘mis-phasing’ of clocks in a modern society of extended artificial light and widespread shiftwork might contribute to the increased prevalence of cancer, cardiac and metabolic diseases observed in recent times.

I thank Dr Ludmila Gaspar for critical comments and figure editing in this manuscript.