The circadian clock has roles in mesenchymal stem cell fate decision

RUNX2 is the ‘master switch’ of osteogenic differentiation, with a highly conserved DNA-binding runt domain. Runx −/− mice show a lack of osteogenic markers and complete blockage of intramembranous and endochondral ossification [61]. Cleidocranial dysplasia resulting from the mutation of RUNX2 is characterised by skeletal dysplasia, such as clavicle hypoplasia and abnormal clavicle [62]. Surprisingly, the overexpression of RUNX2 also shows bone defects in mice, including a dramatic reduction in osteoblast numbers with immature phenotypes. Therefore, RUNX2 plays a major role in MSCs’ early lineage commitment to osteoblasts and maintains the latter at an earlier stage to provide a large number of immature osteoblasts.

RUNX2 shows expression rhythmicity in various tissues, including the SCN and bone [63], and is involved in connecting the circadian clock and MSC fate decision. RUNX2 influences enamel formation in MSCs under the control of circadian pacemaker [64]. Some researchers suppose that RUNX2 relies on the activity of BMAL1, as the rhythm of RUNX2 mRNA disappears in mouse SCN without Bmal1 [63]. Still, some believe REV-ERBα is the upstream regulator of RUNX2 and disturbs the recruitment of RUNX2 to the targeted promoter. Knockdown of REV-ERBα enhances the differentiation and function of pre-osteoblasts [65].

miRNA is also one of the primary mechanisms underlying the rhythm of RUNX2. miR-7-5p disentangles CRY2’s inhibition on the BMAL1-CLOCK complex. The released BMAL1-CLOCK stimulates P300 transcription, promoting histone H3 acetylation and forming a transcription complex with RUNX2 to enhance bone formation (Fig. 2) [66]. Another miRNA, miR-433, shows robust circadian rhythmicity with a peak after dark and targets RUNX2 3’-UTR causing its degradation [67]. In miR-433-inhibited mice skulls, the mRNA levels of RUNX2 and its downstream osteocalcin increased significantly [68].

Besides acting as a transcription factor, RUNX2 also indirectly regulates the transcription of clock genes and/or clock-controlled genes (CCGs) by working as a scaffold to recruit growth factors and cytokines [63, 69], including P300, CCAAT/enhancer-binding protein, and histone deacetylase [70], with its protein–protein interacting domains.

To our knowledge, Osterix, cAMP-dependent transcription factor 4 (ATF4), Forkhead box O (FOXO), and other transcription factors in osteogenic differentiation have rarely been described in circadian rhythm studies. The role of these transcription factors in circadian rhythm remains to be identified.

Peroxisome proliferator-activated receptor γ (PPARγ) belongs to the nuclear hormone receptor superfamily with the highest expression in adipose tissue. It plays a vital role in the balance between osteogenic and adipogenic differentiation of MSCs [71–73]. PPARγ up-regulation promotes adipogenic differentiation and inhibits the generation of osteoblasts and chondrocytes [74], with its down-regulation having opposite effects [75]. Besides, the PPARγ agonist thiazolidinedione (TZD) is associated with bone loss and fracture [76]. Clinical experience and animal studies suggest that the fate change of MSCs associated with PPARγ may be the main reason for this.

PPARγ and its cofactor PGC-1α display a clear circadian expression pattern in many tissues, and PPARγ even acts as a crucial peripheral clock activator of the cardiovascular system and metabolism [77, 78]. Multiple randomised clinical trials have shown that the clock hormone melatonin up-regulates PPARγ [79, 80]. And the inhibition of Clock or Per2 changes the expression of osteocalcin in MSCs, in association with extraordinary increases in PPARγ [31]. BMAL1 and REV-ERBα are induced by PPARγ and PGC-1α [77, 81, 82]. PPARγ is deeply involved in the regulatory networks between circadian rhythms and MSC fate decisions (Fig. 2), with the mechanisms described hereafter.

Nocturnin−/− mice show deregulated circadian patterns in PPARγ gene expression and increased bone mass [83]. Nocturnin adapts to daily environmental changes by changing the organism’s metabolic state [84, 85]. It works as a pivotal mediator of mitochondrial function by dephosphorylating nicotinamide adenine dinucleotide phosphate (NADP+/NADPH) [86]. Nocturnin is highly expressed in MSCs, and its overexpression enhances the lipogenesis of preadipocytes 3T3-L1. It enhances PPARγ transcription activities and guides PPARγ nuclear translocation.

Per1 and Per2 have antagonistic influences on PPARγ activity. PER1 enhances PPARγ activity operating as its natural modulator, while PER2 directly and specifically inhibits the recruitment of PPARγ to the target promoter [87]. However, the PER2 regulation of adipogenesis has significant cell/tissue specificity [87], as it is regulated by a complicated network and works with other core clock products, such as PER1, BMAL1, and REV-ERBα. For example, a substantial reduction in PPARγ is found in per1/2−/− mice [88]. However, Per3 is believed to be a nonessential clock component with modest effects on circadian behaviours [89, 90], and it does inhibit adipogenesis in vivo in a clock manner [91]. The deletion of Per3 promotes adipogenesis by clock outputs [92]. PER3 binds to PPARγ and inhibits its activity [91, 93]. Increasing PPAR is observed in Per3−/− mice, which means that the differentiation arrest resulting from Per3 happens during the early stages.

D-box binding protein (DBP) and E4 promoter-binding protein 4 (E4BP4), contributors to circadian regulatory loops, interact with the active site of the PPARγ promoter to regulate its rhythmic expression and are enhanced by histone deacetylase inhibitor (HDACi) [94]. Decreased H3K9 acetylation of the DBP promoter leads to a lower level of DBP and a diminished binding between DBP and the PPARγ promoter, which gives rise to a lower level of PPARγ [95].

The circadian clock is critical to early chondrogenesis and endochondral ossification [96, 97]. Indian Hedgehog (Ihh) is key in governing chondrogenic differentiation, and its activity is regulated by PER1 and BMAL1 in the growth plate [97]. Bmal1 deficiency leads to the deregulated rhythmic expression of Ihh and Per [98]. Our previous studies show that Prg4 and related chondrogenic genes are downstream targets of Bmal1 in cartilage formation and chondrogenic differentiation [99]. Cartilage tissue engineering, which combines MSCs with other factors, such as collagen, is a promising therapy for degenerated temporomandibular joint and osteoarthritis. Sry-related HMG box 9 (SOX9), a key transcription regulator of chondrocyte differentiation [100], controls the synthesis of collagen type II, the hallmark of chondrogenic commitment [101]. Impaired circadian rhythm induces osteoarthritis resulting from SOX9 inhibition in articular chondrocytes, and the deletion of Bmal1 also leads to reduced SOX9 [102] (Table 1). Two additional SOX family members, SOX6 and SOX5, are also involved in chondrogenesis regulation. SOX6 is negatively regulated by its transactivator PER1 (Fig. 2). It shows a long-term decrease after exposure to 10 nM PTH for 1 h in chondrocyte-like ATDC5 cells with no distinct changes in level SOX5 and RUNX2 levels [103].

Genome-wide expression analysis reveals the rhythm of global histone labels that, depending on an intact clock, are under tight regulation by enzymatic feedback loops [109]. The clock recruits epigenetic enzymes to the promoter of important transcription factors and key signalling molecules involved in differentiation. Besides, CLOCK, functioning as an epigenetic modifier, was subsequently found to have a histone acetyltransferase (HAT) activity greatly amplified by its combination with BMAL1 [116]. Additionally, clock-controlled SIRTs and EZH2 provide possibility in elucidating how the clock is linked to epigenetic mechanism in MSCs differentiation.

At the stage of transcriptional activation, the BMAL1-CLOCK complex acetylates the RUNX2 promoter by Clock itself and by recruiting acetyltransferases P300/CBP, thus promoting osteogenesis in MSCs [66]. At the repression stage, CRY (a constitutive component of the PER complex) occupies the C-terminal transactivation domain of BMAL1, leaving P300/CBP dissociated from its binding site [117]. Once CRY is loaded to the BMAL1-CLOCK complex, it directly recruits the SIN3 complex that comprises HDAC1 and HDAC2 to remove the acetylation tag of promoters (Fig. 3). Thus, the PER-SIN-HDACs pathway targeting histones H3 and H4 controls gene expression precisely and dynamically [118]. In addition, REV-ERBα and RORα exert downstream effects on the expression of NF-kB (which inhibits osteoblast differentiation) by recruiting HDACs [104]. REV-ERBα calls up co-repressors NCOR1/SMRT to ROR element. HDAC3, serving as the catalytic component of NCOR1/SMRT, rhythmically modulates histone deacetylation of the ROR element [119, 120].

Sirtuins (SIRTs) are class III NAD + dependent deacetylases that participate in many biological processes, including the circadian clock and cell differentiation [121–123]. This family has seven members (SIRT1–7) distributed across the cytoplasm, nucleus, and mitochondria. SIRT1 binds BMAL1-CLOCK in a circadian manner and counteracts the HAT activity of CLOCK [124]. Its deacetylation ability extends to BMAL1 and PER2 and relieves BMAL1-CLOCK from CRY and promotes PER2 degradation [125, 126]. SIRT1 requires the cofactors NAD + and NAMPT to function, which are both under tight circadian control [86, 127]. Thus, the deacetylase activity of SIRT1 exhibits a circadian pattern.

The MSC-specific overexpression of SIRT1 increased alveolar bone mass in vivo and promoted MSCs’ proliferation and osteogenic differentiation [128]. Aged mice with a MSC-specific knockout of SIRT1 show reduced subcutaneous fat and a progressive loss of trabecular and cortical bone volume [129]. SIRT1 exerts its effects mainly by deacetylating important transcription factors and signalling pathways, influencing their activity and localisation. For example, SIRT1 directly or indirectly deacetylates RUNX2and PPARγ [130, 131] and deacetylates β-catenin, facilitating its accumulation in the nucleus [132]. Along these lines, SIRT1, serving as a circadian hallmarker, causes dynamic changes in the epigenome to generate transcriptional responses controlling MSC differentiation.

Methylation is also a widespread histone modification in mammalian cells and is linked to transcriptional repression. EZH2, mediating gene silencing via trimethylation of H3K27, is directly regulated by the circadian clock via the E-BOX and RORE motifs in fish species [133]. EZH2 is a repressor of differentiation-associated genes in vitro. It inhibits osteogenesis by inhibiting BMP2 and inhibits chondrogenic differentiation by inhibiting ATOH8 [106, 107] (Table 1). However, in in-vivo animal studies, the additive effects of VDR‐mediated EZH2 transcriptional increase show increased osteogenesis, as EZH2 also inhibits MSC senescence by repressing p16 transcription [134]. MSCs are cells of mesodermal origin which derived from ESCs. The level of EZH2 in MSCs is significantly lower than that in ESCs, and a specific EZH2 inhibitor, GSK126, dramatically drives ESCs differentiation into MSCs by reducing H3K27me3 [135], suggesting that EZH2 levels in specific cell types are crucial for the balance of pluripotency maintenance, cell differentiation, and cell senescence.

Tissue-specific DNA methylation patterns play central roles in establishing cell identity during differentiation. Generally, DNA methylation is considered one of the hallmark mechanisms of gene silencing. For example, DNMT inhibitors, such as 5-aza-2'-deoxycytidine, induce gene activation and cell differentiation [136]. However, the genome-wide DNA methylation profiling of MSCs shows an increasing methylation level of the global genome and CpG sites in the process of osteogenic differentiation [137]. These results appear to be incompatible with the gene-silencing role of DNA methylation. Recent studies suggest that the function of DNA methylation varies depending on the methylation location. The methylation of the gene body, and not the promoter, promotes genome stability and even transcription [138], suggesting that promoter segments may not be the DNA methyltransferase target sites during MSCs differentiation.

In recent years, growing evidence in chronobiology hints at a close connection between the circadian rhythm and DNA methylation. DNMTs oscillate in a clock pattern in the mouse liver and brain [139], and excessive exposure to artificial light leads to tissue-specific changes in DNMTs expression [140]. Taken together, this evidence suggests that DNMT may exert great influence during clock-controlled differentiation.

miRNAs are a class of short non-coding RNAs with a length of about 22 nucleotides. The circadian clock regulates miRNAs expression and activity, and the clock itself is regulated by miRNAs at post-transcriptional levels [141]. And miRNAs taking part in clock regulation have been extensively reviewed [142, 143]. In healthy young males, some circulating miRNAs display 24-h rhythms with two expression patterns: one peaks at night and the other peaks during the daytime [144, 145]. There are 104 miRNAs differently expressed in CLOCK-mutant mice compared with the controlled group, and they are linked to transcription regulation and PI3K-AKT and MAPK signalling pathways [146], which are closely related to MSC differentiation.

However, some researchers argue that mature miRNAs do not oscillate in a circadian pattern. Studies show at least 57 pri-miRNAs oscillated in the mouse liver, with only four mature miRNAs showing the same rhythm as their primary transcript [147]. Interestingly, the rhythmic expression of Dicer (the essential enzyme for miRNA maturation) is found in various tissues [148]. Studies are contradictory and may be model-dependent. The inconsistent findings among these studies can be partly explained by variations in the study samples, methods of assessment, and, most importantly, tissue specificity, which may hint at the role of circadian miRNA in establishing cell identity.

Mounting research has unveiled the importance of miRNAs in the regulation of MSC differentiation. For example, the expression profile in patients with fractures or osteoporosis differs from that in healthy individuals. Hackl et al. summarized the different expression of miRNAs in patients with bone disease and the control group. They rasied the possibilility that circulating miRNAs are used as biomarkers for bone disease [149]. Furthermore, many miRNAs are shared between the circadian clock and MSCs fate decisions. A candidate miRNA set potentially involved in circadian control has been established by various means, including microarrays [150], CHIP [142], and in silico prediction analysis of databases [151]. We found that some of them also function in the differentiation process of MSCs, as summarised in the Additional file 1.

Long non-coding RNAs (lncRNAs) are non-coding RNAs with 2000–100,000 nucleotides in length. A mouse transcriptome analysis indicated that more than 1000 lncRNAs show a diurnal rhythm [152]. The expression of a substantial portion of rhythmic lncRNAs is pronounced in regions of enhancer clusters and regulates gene expression, showing high similarity in rhythm phase and expression patterns with genes in their proximity. For instance, lnc-CROT is highly expressed at super-enhancer regions and modulates long-range circadian gene regulation [153]. Accumulating research has shown that lncRNAs, such as lnc-MEG3 and lnc-XIST, take part in osteogenic differentiation [154, 155]. Unfortunately, there are few reports on the role of lncRNAs in the clock-controlled differentiation process, and follow-up bioinformatics studies should be carried out.

Different kinds of epigenetic mechanisms show a pile-up effect on gene expression levels by influencing chromatin architecture and accessibility, which influence the binding of transcription factors to the genome. The molecular clock, working with various epigenetic regulators, contributes to spatiotemporal balance and the oscillation of chromatin accessibility [109, 156]. For example, RORα/β, BMAL1, and other circadian proteins promote global chromatin decondensation during the activation phase via the SWI/SNF complex (one of the epigenetic remodelling complexes). In turn, this promotes the load of REV-ERB, which rapidly responds to structure changes and inhibits circadian protein expression [157]. Furthermore, the chromatin landscape shaped by circadian regulation is critical for MSCs self-renewal and multilineage differentiation [158].