2. Mammalian Circadian Rhythms and Ubiquitin Ligases
2.1. Overview of Circadian Rhythms in Mammals
2.1.1. The Suprachiasmatic Nucleus
The SCN is a bilateral structure situated in the anterior hypothalamus, dorsal to the optic chiasm
[27]. It consists of approximately 20,000 neurons, each functioning as autonomous circadian clocks
[28]. The SCN is subdivided into ventrolateral and dorsomedial regions, otherwise referred to as the core and shell SCN, respectively
[27]. The intrinsic period of individual SCN neurons can vary from ~22 h to 30 h
[28][29]. However, intercellular coupling between SCN neurons results in their oscillating in synchrony with a significantly narrower period range
[28][30]. Neurons in the ventrolateral SCN receive direct inputs from the retina; they also synthesize and secrete gastrin-releasing peptide (GRP) and vasoactive intestinal polypeptide (VIP)
[27]. VIP is a neuropeptide vital to interneuronal coupling within the SCN, and by extension to the robustness of the central pacemaker
[31]. Once released, both VIP and GRP signal to cells in the core and shell SCN
[31][32]. Dorsomedial SCN neurons utilize a different neuropeptide, arginine vasopressin (AVP), to communicate with and couple to other clock neurons
[31][32]. The inhibitory neurotransmitter, gamma-aminobutyric acid (GABA), is expressed by nearly all SCN neurons and has been shown to contribute to oscillator coupling and the refinement of circadian outputs
[27][33].
As the intrinsic period of the SCN deviates slightly from 24 h, photic entrainment is required to maintain synchrony between the central pacemaker and environmental cycles. This involves the daily resetting of the clock by light, which synchronizes the SCN’s phase with the solar cycle
[6][34]. In mammals, photic entrainment is fully dependent on the retina, and relies on the detection of changes in environmental light intensity
[34][35]. As the master pacemaker, the SCN receives and integrates environmental photic signals from the retina, using them to synchronize its own neuronal clocks, after which the SCN can convey temporal information to peripheral clocks in the brain and body through synaptic and humoral mechanisms
[35]. A direct projection from the retina to the SCN, known as the retinohypothalamic tract (RHT), transmits photic signals to the central clock via the actions of secreted glutamate and pituitary adenylate cyclase-activating peptide (PACAP)
[36]. Within the retina, the non-image forming, intrinsically photoreceptive retinal ganglion cells (ipRGCs) are the main players in photic entrainment, using the photopigment melanopsin to detect light in the blue wavelength range, and transmitting the photic signal to the SCN via the RHT
[37][38][39].
2.1.2. The Mammalian Core Clock Machinery
Circadian TTFLs operate on the principle of negative feedback inhibition, whereby elements in the positive limb of the feedback loop drive the expression of elements within the negative limb, eventually shutting off their own expression until the cycle begins anew ~24 h later
[40]. Within the primary TTFL of mammals, the core clock genes,
Clock and
Bmal1 (officially known as
Arntl), represent the positive elements, whereas
Period1 (Per1),
Period2 (Per2),
Cryptochrome1 (Cry1), and
Cryptochrome2 (Cry2) represent the negative elements (
Figure 2)
[41][42][43][44][45]. The positive limb is most active during the subjective day, when the helix-loop-helix transcription factors, CLOCK and BMAL1, heterodimerize and bind to the E-box
cis-regulatory elements of the
Per and
Cry gene promoters, activating their transcription
[2][7]. Following their translation and accumulation in the cytoplasm, PER and CRY proteins heterodimerize and translocate to the nucleus
[2][7]. This initiates the negative limb, as PER and CRY repress their own transcription, either by binding to CLOCK/BMAL1 and blocking E-box-mediated transactivation or by displacing CLOCK/BMAL1 from E-boxes
[46]. The two modes of repression are differentially mediated by PER and CRY
[46][47][48]. In the “blocking”-style repression, CRY1 binds to CLOCK-BMAL1-E-box complexes independently of PER to inhibit transactivation: the repression occurs even as CLOCK and BMAL1 remain bound to the E-box
[46][47][48]. In the “displacement”-style repression, PER2, in the presence of CRY, displaces CLOCK/BMAL1 from the E-boxes
[46][47][48]. PER proteins on their own have no repressive activity towards CLOCK/BMAL1
[46]. Eventually, PER and CRY are degraded by the proteasomal pathway, allowing CLOCK/BMAL1 to once again occupy E-box elements and activate gene expression
[2][7].
Figure 2. Schematic of mammalian circadian clock transcription–translation feedback loops (TTFLs) (A) and the relationships of known E3 ligases to core clock proteins (B). (A) In the primary TTFL of mammals, transcription factors CLOCK and BMAL1 heterodimerize and activate the transcription of target genes including Cry1/2, Per1/2, Rev-Erbα/β, Rorα/β/γ, Dbp, and Dec1/2 by binding to the E-box elements in their promoters. Once translated, PER and CRY proteins accumulate in the cytoplasm before heterodimerizing and translocating to the nucleus, where they ultimately repress E-box-dependent transcription. In a secondary TTFL, RORα/β/γ binds to ROR elements (RORE) in the promoters of Bmal1 and other genes to activate their transcription. In contrast, REV-ERBα/β binds to ROREs to repress transcription. Another auxiliary feedback loop involves DBP, TEF, and HLF, which bind to the D-box elements of Per1/2 genes (among others), initiating D-box-mediated transactivation. Binding of E4BP4 at D-box elements results in transcriptional repression. (B) All mammalian E3 ubiquitin ligases that have been demonstrated to target a core clock protein for degradation are indicated in black. E3 ligases whose targets are either unknown or not core clock proteins are indicated in the chart in the bottom right corner. Except for FBXL3 and FBXL21, the site of action (nucleus vs. cytoplasm) for all other E3 ligases is unknown. For illustrative purposes, these E3 ligases are shown as acting in the nucleus.
2.2. Mammalian Ubiquitin Ligases
2.2.1. β-TrCP1 (FBXW1) and β-TrCP2 (FBXW11)
Beta-transducin repeat-containing proteins (β-TrCP) were the first E3 ubiquitin ligases to be implicated in the regulation of the clock machinery. β-TrCP is an F-box protein of the Fbws class, characterized by the presence of an F-box motif and tandem WD40 repeats
[49]. F-box proteins are the substrate recognition subunits of the SCF (Skp1-Cullin 1-F-box protein) family of E3 ubiquitin ligases
[49]. Casein kinase 1 (CK1)-mediated phosphorylation of PER2 was shown to trigger an association between β-TrCP and PER2
[50]. The interaction between β-TrCP and PER2 is also facilitated by the SUMOylation of PER2 by SUMO2
[51]. The expression of mutant β-TrCP that lacks an F-box inhibits the degradation of phosphorylated PER2
[50]. In another study, β-TrCP1 and β-TrCP2 were identified as binding partners of PER1
[52]. These interactions are dependent on the phosphorylation of PER1 by CK1ε
[52]. Furthermore, the silencing of β-TrCP1 stabilizes PER1 and inhibits CK1ε-induced PER1 degradation
[52]. Cell-free assays showed that SCF complexes containing β-TrCP are capable of ubiquitinating PER1
[52]. In NIH-3T3 fibroblasts, the knockdown of
β-TrCP1 or the expression of dominant-negative β-TrCP1 elicits a lengthening of molecular oscillations
[53]. PER2 mutants that are unable to interact physically with β-TrCP1/2 exhibit severely disrupted or damped rhythms in fibroblasts
[53]. However, β-TrCP1 appears to be dispensable for circadian rhythms at the behavioural level, as β-TrCP1 knockout mice are phenotypically normal with respect to period length and light-induced phase shifts
[54]. On the other hand, introducing the PER2 S478A mutation, which can no longer be phosphorylated by CK1δ/ε and thus cannot recruit β-TrCP1/2, results in a dramatic lengthening of the behavioural period in knock-in mice and the accumulation of PER2 protein in the nucleus and cytoplasm of the liver, suggesting that β-TrCP2 may compensate for the loss of β-TrCP1
[55]. In line with this, inducible β-TrCP2 knockout mice exhibit a dramatic circadian phenotype characterized by unstable behavioural rhythms and period variability under constant darkness (DD)
[56]. The fact that the ubiquitination of PER2 is still observed in the absence of β-TrCP1 and β-TrCP2 indicates that PER2 is a substrate of other E3 ligases
[56]. β-TrCP1/2 may also influence the circadian clock by ubiquitinating other clock proteins such as DEC1 and targeting them for proteasomal degradation
[57].
2.2.2. Mouse Double Minute 2 Homolog (MDM2)
MDM2 is a RING finger type E3 ligase that serves as a scaffold, bringing E2 enzymes to protein substrates for ubiquitination
[58]. PER2 can form trimeric complexes with MDM2 and its substrate, p53
[59]. However, PER2 and MDM2 can directly associate with each other independently of p53
[59]. Furthermore, MDM2 binds to PER2 at the latter’s PAS domain and an inner region that undergoes extensive phosphorylation
[59]. CK1δ/ε-mediated phosphorylation of PER2 is not necessary for either MDM2 binding or MDM2-dependent ubiquitination of PER2
[59]. MDM2 destabilizes PER2 and appears to work cooperatively with β-TrCP to control the abundance of PER2 during the rising and falling phases of the protein’s circadian cycle
[59]. The knockdown of
mdm2 extends the circadian period in murine embryonic fibroblasts (MEFs), whereas the overexpression of
mdm2 shortens it
[59].
2.2.3. FBXL3
FBXL3 was first reported in three sister studies in 2007 as an E3 ubiquitin ligase for CRY proteins
[60][61][62]. By chemical-induced mutagenesis, two mutations in FBXL3 were identified that dramatically lengthen the circadian period: the C358S substitution termed
afterhours (Afh) and the I364T mutation termed
overtime (Ovtm) [61][62]. In both
Afh and
Ovtm mutant mice, CRY1/2 protein levels are not appreciably altered but PER1/2 abundance is strongly suppressed, suggesting that the stabilization of CRY1/2 in the presence of lower E3 ligase activity is compensated for by a reduction in E-box-dependent transcription of
Cry1/2 genes
[61][62]. FBXL3 binds specifically to CRY1 and CRY2 but not to other core clock proteins and promotes their degradation by proteasomes
[60][62]. FBXL3 was also shown to promote CRY2 ubiquitination in an F-box-dependent manner, suggesting a direct effect of FBXL3 on CRY2
[60]. Both the
Afh and
Ovtm mutations reduce the rate of CRY1/2 degradation; in the case of the
Afh mutation, this was attributed to the reduced binding of FBXL3(
Afh) to CRY proteins and reduced catalytic activity
[60][62]. As a consequence of their effects on CRY stability, these mutations dampen the amplitude of circadian gene expression
[61][62].
Crystal structure analysis revealed a bipartite interaction between FBXL3 and CRY2, in which the C-terminal tail of FBXL3 occupies the FAD-binding pocket of CRY2, and the leucine-rich repeat (LRR) domain of FBXL3 is a secondary site of contact for CRY2 at three key structural motifs
[63]. As these interactions occur in the absence of FAD or PER binding to CRY2, either of these factors can disrupt the FBXL3–CRY2 complex or prevent its formation
[63]. The formation of FBXL3–CRY complexes is required for the recruitment of SKP1 and CUL1, thereby forming the fully functional SCF complex
[64]. This substrate-dependent formation of SCF complexes appears to be specific for FBXL3 and is not observed with FBXL21
[64].
CRY2 binding may also recruit FBXL3 to other substrates such as c-MYC and TLK2 to induce their ubiquitination and subsequent degradation, thereby linking the clock to proteolysis in other physiological systems
[65]. In addition to CRY1/2, FBXL3 has also been shown to physically associate with REV-ERBα in mouse livers
[66]. REV-ERBα recruits FBXL3 to RORE sites, where it derepresses gene expression by inhibiting the actions of REV-ERBα:HDAC3 complexes
[66].
2.2.4. FBXL21
The F-box protein FBXL21 is the closest homologue of FBXL3
[49]. Initial studies revealed that FBXL21 is expressed in the brain and neuroendocrine tissues of sheep, and physically associates with ovine CRY1
[67].
Fbxl21 harbours functional E- and D-box elements within its promoter, resulting in high and rhythmic expression of the gene in the ovine and murine SCN
[67]. Furthermore, FBXL21 overexpression abrogates the repressive effects of CRY1 on CLOCK/BMAL1-mediated transcription
[67]. A subsequent study showed that the
past-time (
Psttm) mutation, which is a missense mutation (G149E) in the
Fbxl21 gene, significantly shortens the circadian period and antagonizes the period-lengthening effects of FBXL3(
Ovtm)
[68]. FBXL21 binds to CRY1/2 with a higher affinity than FBXL3, effectively outcompeting FBXL3
[68]. In the nucleus, where both FBXL21 and FBXL3 are present, FBXL21 sequesters CRY proteins from FBXL3 and protects them from FBXL3-induced proteasomal degradation
[68]. In the cytosol, where FBXL3 is absent, FBXL21 triggers the slow degradation of CRY1/2
[68]. It was further shown that FBXL21 preferentially forms SCF complexes in the cytoplasm but not in the nucleus
[68]. These collective observations are consistent with the different potencies of the two FBXL homologs in CRY destabilization, where FBXL21 appears to be a weaker E3 ligase for CRY than FBXL3
[68]. The effects of FBXL21 on the stability of nuclear and cytoplasmic CRY are such that
Psttm mutant mice have altered core clock gene expression that is characterized by a higher expression of E-box-regulated genes
[68]. Although a different group confirmed many of these findings, including the antagonism between FBXL3 and FBXL21, the stabilization of CRY by FBXL21, and the preferential localization of FBXL21 in the cytoplasm, in their case,
Fbxl21-deficient mice did not show a circadian period phenotype, unlike the
Psttm mutants
[69].
2.2.5. DDB1–CUL4A and DDB1–CUL4A–CDT2
The DDB1–CUL4A–CDT2 E3 ubiquitin ligase complex has been shown to target CRY1 for degradation
[70][71]. In vitro assays revealed that DDB1–CUL4A–CDT2 directly ubiquitinates CRY1 at Lys-585, marking the protein for proteasomal degradation
[71]. CRY1 physically binds to CDT2 and the silencing of
Cdt2 prevents complex formation between CRY1 and DDB1–CUL4A
[71]. In mouse hepatoma cells, the knockdown of
Ddb1 or overexpression of the ubiquitination-defective mutant CRY1 K585A enhances CRY1 stability and increases the amplitude of circadian oscillations as measured by a
Bmal1-Luc reporter
[71]. In a subsequent study, liver-specific
Ddb1 knockout mice were shown to have impairments in hepatic gluconeogenesis but were protected from high-fat-diet-induced hyperglycemia
[70]. These effects are due to elevated levels of CRY1, which binds to the FOXO1 transcription factor and promotes its ubiquitination and degradation
[70]. In turn, the lower abundance of FOXO1 in
Ddb1 knockouts suppresses gluconeogenic gene expression
[70]. Besides CRY1, DDB1–CUL4A has been shown to interact with CLOCK-BMAL1
[72]. These circadian transcription factors bind to an adaptor protein of the DDB1–CUL4A complex, WD repeat-containing protein 76 (WDR76)
[72]. Through this interaction, CLOCK-BMAL1 recruits DDB1–CUL4A to E-boxes of the
Per1 and
Per2 genes as well as other circadian genes
[72]. DDB1–CUL4A enhances the monoubiquitination of histone H2B at E-box sites, which subsequently inhibits CLOCK-BMAL1 binding while promoting the association with PER complexes
[72].
2.2.6. FBXW7
FBXW7 is an F-box protein of the Fbws class
[49]. Several studies have implicated the involvement of FBXW7 in the regulation of circadian rhythms by targeting different proteins. In mice injected with renal carcinoma cells, the expression of FBXW7 exhibits circadian oscillations in the tumours, driven by DBP, which binds to and transactivates
Fbxw7 in a rhythmic fashion
[73]. The mammalian target of rapamycin (mTOR), a key protein in cell growth and the circadian control of translation, oscillates in anti-phase to FBXW7 protein
[73]. A prior study demonstrated that FBXW7 ubiquitinates mTOR and targets it for proteasomal degradation
[74]. A separate study revealed that FBXW7 binds to CRY2 in colorectal cancer cells, potentially through a direct interaction between the degron motif of CRY2 and the narrow face of the WD40 domain of FBXW7
[75]. It was further shown that FBXW7 destabilizes CRY2 by promoting its ubiquitination and subsequent proteasomal degradation
[75]. REV-ERBα is another identified target of FBXW7
[76]. FBXW7 was demonstrated to physically interact with REV-ERBα, enhancing its ubiquitination and destabilizing it
[76]. Cyclin-dependent kinase 1 (CDK1)-mediated phosphorylation of REV-ERBα at Thr-275 is required for its recognition by FBXW7
[76]. The amplitude of circadian gene expression is suppressed when
Fbxw7 is silenced in cultured cells or ablated in mouse livers
[76]. The deletion of
Fbxw7 specifically in the liver alters the hepatic circadian transcriptome and disrupts whole-body lipid and glucose metabolism
[76].
2.2.7. TNF Receptor-Associated Factor 2 (TRAF2)
The ubiquitin ligase TRAF2 was initially identified as a CRY1-binding protein in a high-throughput yeast two-hybrid screen
[77]. However, a subsequent study showed that TRAF2 overexpression does not alter CRY1 abundance, suggesting that the interaction between TRAF2 and CRY1 does not lead to the degradation of the latter
[78]. The same study also revealed that TRAF2 physically binds to BMAL1 and reduces its abundance
[78]. The physical association is mediated by the zinc finger domain of TRAF2 and not by the TRAF domain, the canonical substrate recognition site
[78]. Furthermore, the deletion of the RING domain of TRAF2 stabilizes BMAL1 protein, indicating that the effects of TRAF2 are dependent on its ubiquitin ligase activity
[78]. Consistent with these results, the overexpression of TRAF2 promotes the ubiquitination of BMAL1 and its degradation by proteasomes
[78]. The TRAF2-dependent reduction in BMAL1 abundance ultimately attenuates E-box-mediated transcription and dampens
Per1 oscillations
[78].
2.2.8. STIP1 Homology and U-Box-Containing Protein 1 (STUB1)
STUB1 was identified in a mass spectrometric analysis of BMAL1-binding partners
[79]. This interaction is selective for BMAL1 and is not observed with CLOCK
[79]. The overexpression of wild-type STUB1, but not of a catalytically inactive form, reduces the abundance of BMAL1 protein in HEK293T cells, indicating that STUB1 affects BMAL1 stability through its ubiquitin ligase activity
[79]. Along these lines, STUB1 catalyzes the K48-linked polyubiquitination of BMAL1, which is associated with proteasomal degradation
[79]. STUB1 is primarily localized to the cytosol, but upon oxidative stress, it translocates to the nucleus where it can destabilize BMAL1 to attenuate cellular senescence
[79].
2.2.9. UBE3A
UBE3A is a HECT-domain-containing ubiquitin ligase expressed in multiple tissues, including the SCN
[80]. It is also the causative gene for the neurodevelopmental disorder Angelman Syndrome (AS), in which UBE3A expression is absent due to the loss of the maternal allele amid the normally silenced paternal allele
[81]. There is evidence for and against paternal imprinting of the
Ube3a gene in SCN neurons
[82][83]. Sleep disturbances are one of the symptoms of AS, which include delayed development, intellectual disability, impaired speech, and motor dysfunction
[84]. Ablating the maternal copy of
Ube3a in mice consistently disrupts sleep homeostasis
[80][83][85]. However, one study found that this mutation also lengthens the circadian period under constant darkness (DD), accelerates recovery from jetlag (i.e., mice re-entrain more rapidly to an advanced light–dark schedule), and suppresses locomotor activity under constant light (LL), whereas another study found no effect on circadian period
[80][86]. At the molecular level, the activation of UBE3A in mouse embryonic fibroblasts (MEFs) by the viral oncogenes E6/E7 triggers the ubiquitination of BMAL1 and a reduction in its protein abundance through proteasomal degradation, ultimately leading to a loss of circadian rhythms
[87]. These effects are direct, as UBE3A physically binds to, polyubiquitinates, and destabilizes BMAL1
[86][87]. Consistent with these observations, the loss of the maternal allele of
Ube3a elevates BMAL1 abundance in the murine hypothalamus
[86]. Lastly, even in the absence of E6/E7-mediated transformation, the endogenous activity of UBE3A is required for maintaining robust rhythms of
Per2 expression in MEFs
[87].
2.2.10. UBE2O
UBE2O is a ubiquitin-conjugating enzyme with hybrid E2/E3 activity
[88]. Initially identified from a mass spectrometry screen for BMAL1-binding proteins, UBE2O was shown to physically associate with BMAL1 in mouse Neuro2a cells and whole brain tissue
[89]. The overexpression of UBE2O reduces the levels of endogenous BMAL1 in HEK293T cells in a dose-dependent fashion, whereas its knockdown elevates BMAL1 abundance
[89]. These effects are specific to BMAL1 and are not observed with CLOCK
[89]. UBE2O was further shown to ubiquitinate BMAL1 and dramatically reduce its half-life
[89]. The conserved region 2 (CR2) domain of UBE2O is essential for BMAL1 binding and ubiquitination
[89]. The effects of UBE2O on BMAL1 stability lead to attenuated BMAL1 transcriptional activity when UBE2O is overexpressed and a higher amplitude of
Per2 rhythms when it is silenced
[89].
2.2.11. Seven in Absentia 2 (SIAH2)
SIAH2 was identified in a functional screen for REV-ERBα-directed ubiquitin ligases
[90]. The overexpression of SIAH2 selectively destabilizes REV-ERBα and REV-ERBβ, but not the other proteins tested, including PER1, PER2, CRY1, and CLOCK
[90]. In contrast, the SIAH2 paralog, SIAH1, does not influence REV-ERB stability
[90]. More importantly, SIAH2 was shown to physically interact with REV-ERBα/β and promote its ubiquitination
[90]. Ablating the RING domain of SIAH2 interferes with its ability to destabilize REV-ERB, indicating the importance of its catalytic function as an E3 ubiquitin ligase
[90]. In synchronized U2OS cells, the knockdown of
Siah2 slows the turnover of REV-ERBα, thereby affecting its rhythms and the expression of its target genes, as well as lengthening the circadian period
[90]. However, in mouse models, the absence of
Siah2 does not impact REV-ERBα protein rhythms, likely as a result of compensation by other E3 ligases, although other clock genes such as
Bmal1 and
Per2 are moderately affected at the transcript level
[91]. In a surprising twist,
Siah2 deficiency disrupts the circadian hepatic transcriptome only in female mice
[91]. Global circadian gene expression in the female liver is phase-advanced by ~9 h such that genes that are normally expressed during the night now peak in the daytime
[91]. The underlying cause for this sexual dimorphism remains unknown.
2.2.12. ARF-BP1 and PAM (Myc-BP2)
ARF-BP1 (also known as HUWE1) and PAM are HECT and RING finger type E3 ligases, respectively
[92][93]. Both were co-purified with REV-ERBα in the presence of lithium chloride, an inhibitor of glycogen synthase kinase 3 beta (GSK3β) and a known inducer of REV-ERBα degradation, and subsequently identified by mass spectrometry
[94]. Co-overexpression of ARF-BP1 and PAM strongly suppresses REV-ERBα protein levels, whereas their simultaneous depletion stabilizes REV-ERBα
[94]. Furthermore, ARF-BP1 and PAM exclusively promote the K48-linked polyubiquitination of REV-ERBα, targeting the protein for degradation
[94]. The knockdown of both ARF-BP1 and PAM in hepatoma cells results in a higher amplitude of REV-ERBα protein rhythms and consequently a lower expression and oscillatory amplitude of
Bmal1 [94].