Death of a Protein: The Role of E3 Ubiquitin Ligases in Circadian Rhythms of Mice and Flies

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].

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].

Alongside the primary TTFL there exist ancillary feedback loops that stabilize and fine-tune the clock, connecting it to other cellular pathways [2,45]. An important secondary TTFL found in mammals is the one involving REV-ERB nuclear receptors and retinoic acid receptor-related orphan receptors (RORs), both acting on ROR elements (ROREs) in gene promoters [2,45,49,50]. The Bmal1 promoter contains two ROREs, which lead to transcriptional activation when bound by RORα, RORβ, and RORγ, and transcriptional repression upon the binding of Rev-Erbα and Rev-Erbβ [2,49,50,51]. BMAL1 controls the expression of its own activators and repressors through E-box-mediated transactivation of Rorα/β/γ and Rev-Erbα/β, driving their rhythmic expression [2]. Yet another ancillary TTFL involves the three PAR domain-basic leucine zipper (PAR bZIP) transcription factors, albumin D site-binding protein (DBP), thyrotroph embryonic factor (TEF), and hepatic leukemia factor (HLF), and the bZIP transcription factor, nuclear factor interleukin-3 regulated (NFIL3/E4BP4), acting at D-box elements [52]. D-box sequences are present in potentially thousands of genes including Per1 and Per2 [52,53,54]. The binding of DBP triggers D-box-mediated transactivation, while NFIL3/E4BP4 binding, in association with another circadian transcription factor, basic helix-loop-helix family member e41 (BHLHE41/DEC2), results in repression [55]. The Dbp, Nfil3/E4bp4, and Dec1/2 genes are themselves under E-box regulation [56,57].

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 [58]. F-box proteins are the substrate recognition subunits of the SCF (Skp1-Cullin 1-F-box protein) family of E3 ubiquitin ligases [58]. Casein kinase 1 (CK1)-mediated phosphorylation of PER2 was shown to trigger an association between β-TrCP and PER2 [59]. The interaction between β-TrCP and PER2 is also facilitated by the SUMOylation of PER2 by SUMO2 [60]. The expression of mutant β-TrCP that lacks an F-box inhibits the degradation of phosphorylated PER2 [59]. In another study, β-TrCP1 and β-TrCP2 were identified as binding partners of PER1 [61]. These interactions are dependent on the phosphorylation of PER1 by CK1ε [61]. Furthermore, the silencing of β-TrCP1 stabilizes PER1 and inhibits CK1ε-induced PER1 degradation [61]. Cell-free assays showed that SCF complexes containing β-TrCP are capable of ubiquitinating PER1 [61]. In NIH-3T3 fibroblasts, the knockdown of β-TrCP1 or the expression of dominant-negative β-TrCP1 elicits a lengthening of molecular oscillations [62]. PER2 mutants that are unable to interact physically with β-TrCP1/2 exhibit severely disrupted or damped rhythms in fibroblasts [62]. 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 [63]. 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 [64]. 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) [65]. 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 [65]. β-TrCP1/2 may also influence the circadian clock by ubiquitinating other clock proteins such as DEC1 and targeting them for proteasomal degradation [66].

MDM2 is a RING finger type E3 ligase that serves as a scaffold, bringing E2 enzymes to protein substrates for ubiquitination [67]. PER2 can form trimeric complexes with MDM2 and its substrate, p53 [68]. However, PER2 and MDM2 can directly associate with each other independently of p53 [68]. Furthermore, MDM2 binds to PER2 at the latter’s PAS domain and an inner region that undergoes extensive phosphorylation [68]. CK1δ/ε-mediated phosphorylation of PER2 is not necessary for either MDM2 binding or MDM2-dependent ubiquitination of PER2 [68]. 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 [68]. The knockdown of mdm2 extends the circadian period in murine embryonic fibroblasts (MEFs), whereas the overexpression of mdm2 shortens it [68].

FBXL3 was first reported in three sister studies in 2007 as an E3 ubiquitin ligase for CRY proteins [69,70,71]. 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) [70,71]. 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 [70,71]. FBXL3 binds specifically to CRY1 and CRY2 but not to other core clock proteins and promotes their degradation by proteasomes [69,71]. FBXL3 was also shown to promote CRY2 ubiquitination in an F-box-dependent manner, suggesting a direct effect of FBXL3 on CRY2 [69]. 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 [69,71]. As a consequence of their effects on CRY stability, these mutations dampen the amplitude of circadian gene expression [70,71].

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 [72]. 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 [72]. The formation of FBXL3–CRY complexes is required for the recruitment of SKP1 and CUL1, thereby forming the fully functional SCF complex [73]. This substrate-dependent formation of SCF complexes appears to be specific for FBXL3 and is not observed with FBXL21 [73].

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 [74]. In addition to CRY1/2, FBXL3 has also been shown to physically associate with REV-ERBα in mouse livers [75]. REV-ERBα recruits FBXL3 to RORE sites, where it derepresses gene expression by inhibiting the actions of REV-ERBα:HDAC3 complexes [75].

The F-box protein FBXL21 is the closest homologue of FBXL3 [58]. Initial studies revealed that FBXL21 is expressed in the brain and neuroendocrine tissues of sheep, and physically associates with ovine CRY1 [76]. 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 [76]. Furthermore, FBXL21 overexpression abrogates the repressive effects of CRY1 on CLOCK/BMAL1-mediated transcription [76]. 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) [77]. FBXL21 binds to CRY1/2 with a higher affinity than FBXL3, effectively outcompeting FBXL3 [77]. In the nucleus, where both FBXL21 and FBXL3 are present, FBXL21 sequesters CRY proteins from FBXL3 and protects them from FBXL3-induced proteasomal degradation [77]. In the cytosol, where FBXL3 is absent, FBXL21 triggers the slow degradation of CRY1/2 [77]. It was further shown that FBXL21 preferentially forms SCF complexes in the cytoplasm but not in the nucleus [77]. 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 [77]. 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 [77]. 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 [78].

The DDB1–CUL4A–CDT2 E3 ubiquitin ligase complex has been shown to target CRY1 for degradation [79,80]. In vitro assays revealed that DDB1–CUL4A–CDT2 directly ubiquitinates CRY1 at Lys-585, marking the protein for proteasomal degradation [80]. CRY1 physically binds to CDT2 and the silencing of Cdt2 prevents complex formation between CRY1 and DDB1–CUL4A [80]. 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 [80]. 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 [79]. These effects are due to elevated levels of CRY1, which binds to the FOXO1 transcription factor and promotes its ubiquitination and degradation [79]. In turn, the lower abundance of FOXO1 in Ddb1 knockouts suppresses gluconeogenic gene expression [79]. Besides CRY1, DDB1–CUL4A has been shown to interact with CLOCK-BMAL1 [81]. These circadian transcription factors bind to an adaptor protein of the DDB1–CUL4A complex, WD repeat-containing protein 76 (WDR76) [81]. Through this interaction, CLOCK-BMAL1 recruits DDB1–CUL4A to E-boxes of the Per1 and Per2 genes as well as other circadian genes [81]. 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 [81].

FBXW7 is an F-box protein of the Fbws class [58]. 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 [82]. 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 [82]. A prior study demonstrated that FBXW7 ubiquitinates mTOR and targets it for proteasomal degradation [83]. 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 [84]. It was further shown that FBXW7 destabilizes CRY2 by promoting its ubiquitination and subsequent proteasomal degradation [84]. REV-ERBα is another identified target of FBXW7 [85]. FBXW7 was demonstrated to physically interact with REV-ERBα, enhancing its ubiquitination and destabilizing it [85]. Cyclin-dependent kinase 1 (CDK1)-mediated phosphorylation of REV-ERBα at Thr-275 is required for its recognition by FBXW7 [85]. The amplitude of circadian gene expression is suppressed when Fbxw7 is silenced in cultured cells or ablated in mouse livers [85]. The deletion of Fbxw7 specifically in the liver alters the hepatic circadian transcriptome and disrupts whole-body lipid and glucose metabolism [85].

The ubiquitin ligase TRAF2 was initially identified as a CRY1-binding protein in a high-throughput yeast two-hybrid screen [86]. 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 [87]. The same study also revealed that TRAF2 physically binds to BMAL1 and reduces its abundance [87]. The physical association is mediated by the zinc finger domain of TRAF2 and not by the TRAF domain, the canonical substrate recognition site [87]. 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 [87]. Consistent with these results, the overexpression of TRAF2 promotes the ubiquitination of BMAL1 and its degradation by proteasomes [87]. The TRAF2-dependent reduction in BMAL1 abundance ultimately attenuates E-box-mediated transcription and dampens Per1 oscillations [87].

STUB1 was identified in a mass spectrometric analysis of BMAL1-binding partners [88]. This interaction is selective for BMAL1 and is not observed with CLOCK [88]. 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 [88]. Along these lines, STUB1 catalyzes the K48-linked polyubiquitination of BMAL1, which is associated with proteasomal degradation [88]. 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 [88].

UBE3A is a HECT-domain-containing ubiquitin ligase expressed in multiple tissues, including the SCN [89]. 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 [90]. There is evidence for and against paternal imprinting of the Ube3a gene in SCN neurons [91,92]. Sleep disturbances are one of the symptoms of AS, which include delayed development, intellectual disability, impaired speech, and motor dysfunction [93]. Ablating the maternal copy of Ube3a in mice consistently disrupts sleep homeostasis [89,92,94]. 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 [89,95]. 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 [96]. These effects are direct, as UBE3A physically binds to, polyubiquitinates, and destabilizes BMAL1 [95,96]. Consistent with these observations, the loss of the maternal allele of Ube3a elevates BMAL1 abundance in the murine hypothalamus [95]. 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 [96].

UBE2O is a ubiquitin-conjugating enzyme with hybrid E2/E3 activity [97]. 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 [98]. The overexpression of UBE2O reduces the levels of endogenous BMAL1 in HEK293T cells in a dose-dependent fashion, whereas its knockdown elevates BMAL1 abundance [98]. These effects are specific to BMAL1 and are not observed with CLOCK [98]. UBE2O was further shown to ubiquitinate BMAL1 and dramatically reduce its half-life [98]. The conserved region 2 (CR2) domain of UBE2O is essential for BMAL1 binding and ubiquitination [98]. 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 [98].

SIAH2 was identified in a functional screen for REV-ERBα-directed ubiquitin ligases [99]. The overexpression of SIAH2 selectively destabilizes REV-ERBα and REV-ERBβ, but not the other proteins tested, including PER1, PER2, CRY1, and CLOCK [99]. In contrast, the SIAH2 paralog, SIAH1, does not influence REV-ERB stability [99]. More importantly, SIAH2 was shown to physically interact with REV-ERBα/β and promote its ubiquitination [99]. 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 [99]. 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 [99]. 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 [100]. In a surprising twist, Siah2 deficiency disrupts the circadian hepatic transcriptome only in female mice [100]. 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 [100]. The underlying cause for this sexual dimorphism remains unknown.

ARF-BP1 (also known as HUWE1) and PAM are HECT and RING finger type E3 ligases, respectively [101,102]. 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 [103]. Co-overexpression of ARF-BP1 and PAM strongly suppresses REV-ERBα protein levels, whereas their simultaneous depletion stabilizes REV-ERBα [103]. Furthermore, ARF-BP1 and PAM exclusively promote the K48-linked polyubiquitination of REV-ERBα, targeting the protein for degradation [103]. 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 [103].

In Drosophila, the central pacemaker function is mediated by a coupled network of ~150 clock neurons [104]. These neurons form seven anatomical clusters: the four large and the four small ventrolateral neurons (l-LN_vs and s-LN_vs, respectively), the lateral posterior neurons (LPNs), the dorsolateral neurons (LN_dS), and three classes of dorsal neurons (DN1, DN2, and DN3) [104]. Each cluster expresses a distinct combination of molecular markers and neuropeptides and influences different facets of clock-controlled behaviour [105]. The l-LN_vs and s-LN_vs are the only neurons within the network to express the neuropeptide, pigment-dispersing factor (PDF), which is homologous to mammalian VIP [106]. There is a lone, fifth s-LN_v that is devoid of PDF and is frequently considered with the LN_ds [104]. The s-LN_vs are recognized as the master pacemaker neurons, as they are essential for the generation of behavioural rhythms under DD [107,108,109]. In addition to driving DD rhythms, the s-LN_vs in Drosophila are responsible for the early morning activity displayed by this crepuscular organism, earning them the moniker of morning cells (M-cells) [110,111]. Time-of-day information is conveyed by s-LN_vs to the rest of the network through the rhythmic release of PDF from their dorsal projections [106,112]. In flies, there are three separate pathways that mediate circadian photoreception [113]. The most studied mechanism involves CRYPTOCHROME, which functions as a blue light-sensitive photoreceptor in Drosophila [114]. The translucent nature of the fly’s body and the fact that CRY is present in all clock cells mean that CRY-mediated photoentrainment can occur in a cell-autonomous manner in both central and peripheral clocks [115]. The Hofbauer–Buchner eyelets, as well as the compound eye and ocelli, constitute the remaining pathways for circadian photoreception [116,117,118].

The architecture of the molecular clock machinery in Drosophila is broadly similar to that in mammals, with some notable exceptions. In the primary feedback loop, CYCLE (CYC), the homolog of mammalian BMAL1, partners with CLOCK (CLK) to activate the transcription of genes containing E-box elements (Figure 3) [119,120,121]. Two such genes are period (per) and timeless (tim), whose protein products heterodimerize and accumulate in the cytoplasm [122,123,124]. Following a timed delay, PER:TIM complexes translocate to the nucleus, where they associate with CLK:CYC to repress E-box-dependent transcription, thereby closing the feedback loop [125,126]. The regulated degradation of PER and TIM leads to the derepression of CLK:CYC and the initiation of the next round of E-box-mediated transcription [120]. Unlike mammalian CRY, which couples with PER to form the negative elements of the TTFL, Drosophila CRY functions as a blue light-responsive photoreceptor whose primary role is in light-induced clock resetting [114]. Upon light activation, CRY interacts physically with TIM and promotes its degradation [127,128].

Additional feedback loops exist to fine-tune the pacemaking of the primary TTFL. One such loop regulates the rhythmic transcription of clock with a pair of opposing bZIP transcription factors: PDP1ε activates clk transcription, whereas VRILLE inhibits it [129,130]. Even though vrille and pdp1ε are both E-box-regulated genes, their protein products peak at different phases, with VRILLE peaking prior to PDP1ε; this allows their opposing actions on clk to be temporally segregated [129]. In a third feedback loop, the product of the E-box-dependent gene clockwork orange (cwo) acts antagonistically to CLK:CYC by competing for E-box binding and repressing transcription [131,132].

As is the case in mammals, the molecular clock machinery in Drosophila is regulated by the ubiquitin–proteasome pathway [120]. Regulated proteolysis has been noted for CLK, PER, TIM, and CRY [120]. A recent study also hinted at potential, non-proteolytic functions of ubiquitination [133]. In the next section, we explore the functions of the ubiquitin ligases that have been implicated thus far in circadian rhythms of fruit flies.

SLIMB is the Drosophila homolog of β-TrCP and the first E3 ubiquitin ligase to be implicated in the regulation of the circadian clock machinery in flies. Slimb mutants display complete arrhythmicity under DD whereas the restoration of slimb expression specifically in the LN_vs rescues the phenotype [134]. In contrast, the LN_v-targeted overexpression of slimb on a wild-type background results in period lengthening [134]. Another group found that the ubiquitous overexpression of slimb using the tubulin promoter renders the flies arrhythmic under DD, as does the overexpression of a slimb mutant lacking the F-box in tim-expressing clock cells [135]. Slimb-deficient flies are characterized by low-amplitude PER and TIM protein rhythms in head extracts as well as the persistence of the hyperphosphorylated forms of both proteins throughout the entire circadian cycle, suggesting a defect in their rhythmic turnover [134]. SLIMB preferentially associates with PER following the phosphorylation of the latter protein by DOUBLETIME (DBT), the fly homolog of CK1ε, and triggers PER degradation [135]. The silencing of slimb in S2 cells promotes the accumulation of hyperphosphorylated PER but only in the presence of DBT [135]. Thus, DBT-mediated phosphorylation of PER likely creates a signal that leads to the recruitment and binding of SLIMB. It has also been suggested that clock- and light-dependent degradation of PER and TIM may occur by different mechanisms, as slimb deficiency does not affect the rhythmic cycling of either protein in fly head extracts [134].

With respect to the physical association between SLIMB and PER, the first 100 amino acids of PER are required for SLIMB binding, as is the phosphorylation of PER at Ser-47 by DBT [136]. Phospho-Ser47 collaborates with other phosphorylation events of nearby residues to create a high-affinity SLIMB binding site, which is distinct from the major phospho-cluster spanning amino acids 583–596 of PER [136]. Both the phosphorylation of PER at Ser-47 and PER–SLIMB association occur near the end of the subjective night, indicating that SLIMB-directed degradation controls the downswing of nuclear PER levels at the end of the circadian cycle [136]. Mutations in per that abrogate (S47A) or mimic (S47D) Ser-47 phosphorylation lengthen or shorten, respectively, the behavioural period, suggesting that DBT-triggered degradation of PER by SLIMB is a key determinant of circadian clock speed [136].

The involvement of DUBE3A in circadian rhythms was first demonstrated in Drosophila [137]. A null mutation of the homologous gene dube3a results in weakly rhythmic or arrhythmic flies [137]. These phenotypes are apparent in older female flies (18–21 days) and in both young (4–7 days) and old male flies [137]. Arrhythmic behaviour under DD is recapitulated by the RNAi-mediated silencing of dube3a specifically in PDF neurons [96]. Interestingly, the overexpression of DUBE3A in these cells also results in arrhythmicity, suggesting that maintaining appropriate levels of DUBE3A is essential for the regulation of circadian rhythms [96]. DUBE3A is broadly expressed in the central nervous system of flies and is mainly localized to the cytoplasm [137]. The mechanistic underpinnings of DUBE3A’s effects on circadian rhythms remain to be elucidated.

CTRIP (short for circadian trip) is a HECT type E3 ligase in fruit flies with homology to mammalian trip12 [138]. In the fly brain, ctrip expression is strong and highly enriched in PDF neurons, but there is no evidence that it oscillates in a circadian fashion [139]. The deletion of the first five exons of ctrip disrupts the larval clock such that levels of PER, TIM, and CLK are elevated in the s-LN_vs and the period of their oscillations is lengthened [139]. The silencing of ctrip in tim-expressing clock cells produces similar effects in the s-LN_vs of adult flies, slowing the pace of behavioural rhythms [139]. Cell culture experiments revealed that a knockdown of ctrip attenuates CLK degradation, which is consistent with the notion that CLK is a substrate of CTRIP [139]. Furthermore, the phosphorylated forms of PER and TIM persist through the subjective day when ctrip is downregulated, suggesting that their degradation may also be impaired [139]. In a per-null background, ctrip silencing continues to enhance CLK levels whereas its effects on TIM are abolished [139]. It is therefore possible that CLK and PER, but not TIM, are both independent substrates of CTRIP.

JET was identified in a study that characterized fly mutants with strong rhythmic behaviours under LL [128]. A putative component of SCF complexes, JET has an N-terminal F-box and seven LRR domains important for protein–protein interactions [128]. Two mutations found within adjacent LRRs buffer the circadian clock against the disruptive effects of constant light, allowing for the persistence of behavioural rhythms [128]. In addition, jet mutants are less responsive to the phase-shifting effects of a nocturnal light pulse and require more days to re-entrain to a shifted light–dark schedule, effects that are rescued by over-expressing wild-type jet in clock cells [128]. These phenotypes are due to an impairment in light-induced TIM degradation in jet mutants [128]. JET-dependent TIM degradation requires the presence of CRY and is mediated by the proteasome [128]. Wild-type JET physically binds to TIM and ubiquitinates it, whereas mutant JET is less efficient catalytically [128].

In addition to TIM, JET was later shown to induce CRY degradation [140]. Yeast 2-hybrid experiments revealed light-dependent, direct binding between JET and CRY [140]. In the presence of light, mutations or deficiency in jet results in a greater abundance of CRY in fly heads, a finding that is recapitulated in S2 cells [140]. TIM appears to be preferred over CRY as a substrate of JET, which is possibly rate-limiting; thus, in the presence of TIM, CRY is protected from light- and JET-dependent degradation [140]. The study suggests that JET mediates the degradation of both TIM and CRY in a sequential manner due to differences in their binding affinities [140]. Light-induced conformational changes in CRY increase its affinity for JET [141].

In an interesting twist, jet mutations also rescue the arrhythmic phenotypes of flies that are mutant for slowpoke (slo) and dyschromic (dysc), which encode a voltage-gated potassium channel and its regulator, respectively [142]. Both dysc and slo mutants also exhibit synaptic defects at the larval neuromuscular junction, which are rescued by mutating jet [142]. These findings raise the possibility that the influence of jet on circadian rhythms may extend beyond the molecular clock machinery.

BRWD3 functions as a substrate receptor for CUL4–RING (CRL4) E3 ubiquitin ligase complexes and was identified in an RNAi screen for E3 ligases that mediate light-induced CRY degradation [143]. Similar to JET, BRWD3 binds to CRY in a light-dependent manner [143]. However, unlike jet silencing, the knockdown of brwd3 attenuates light-evoked CRY degradation by proteasomes in S2 cells [143]. Further experiments showed that BRWD3 co-precipitates with other components of the CRL4 E3 ligase (namely, DDB1, CUL4A, CUL4B, and ROC1), and the BRWD3–CRL4 complex catalyzes the ubiquitination of CRY in vitro [143]. This contrasts with JET, which binds to but does not promote the ubiquitylation of CRY [143]. The study suggests that light during the day induces the formation of a protein complex consisting of TIM, CRY, BRWD3, and JET, causing BRWD3 and JET to ubiquitylate their respective substrates, CRY and TIM, and induce their degradation.

MORGUE is a combined E2/E3 protein that harbours an F-box as well as a variant E2 conjugase domain [144,145]. MORGUE has been shown to physically associate with SkpA, a subunit of the SCF complex, as well as K48-linked polyubiquitin chains [146]. The overexpression of morgue in tim- but not Pdf-positive cells protects flies from LL-induced behavioural arrhythmicity [147]. Under LL conditions, morgue overexpression drives robust PER rhythms only in a subset of DN1 neurons and not in other clock neurons including the LN_vs, indicating that morgue function in DN1s is critical for maintaining rhythmicity under LL [147]. MORGUE has been suggested to inhibit the CRY input pathway in DN1s, thus preventing the constant degradation of TIM and the consequent loss of rhythms under LL [147]. This potential relationship between MORGUE and the CRY input pathway is inferred from the observation that light-induced phase shifts are severely blunted in morgue-overexpressing flies, as they are in cry hypomorphs [147,148]. However, a molecular interaction between MORGUE and CRY has not been demonstrated thus far.

CULLIN-3 is a RING finger type E3 ubiquitin ligase that forms SCF complexes distinct from those containing SLIMB or JET, which are CUL1-based [149]. The silencing of cul-3 in cry- or Pdf-expressing clock neurons leads to weakly rhythmic or arrhythmic behaviour [150]. The tim-specific overexpression of dominant-negative CUL3 mutants that cannot undergo neddylation also elicits arrhythmicity under DD [150]. A more recent study showed that CUL3 is required for light-induced clock resetting in CRY-deficient flies [151]. At the molecular level, cul-3 knockdown flies exhibit dampened PER and TIM oscillations in the s-LN_vs due to a lower peak expression of these proteins in the night [150]. TIM appears to be the major substrate of CUL3, as the dampening of TIM oscillations upon cul-3 silencing occurs at a much faster rate than that of PER [150]. In the absence of PER, CUL3 associates with the hypophosphorylated form of TIM [150]. This is different from TIM/SLIMB complexes, which are formed in the presence of PER and consist of phosphorylated TIM [134]. Neddylation-deficient CUL3 mutants exhibit an accumulation of hyperphosphorylated TIM in the cytoplasm, suggesting that CUL3 promotes the degradation of either phosphorylated TIM or its hypophosphorylated form, the latter of which undergoes further phosphorylation [150]. However, despite the accumulation of TIM in neddylation-deficient CUL3 mutants, a subsequent study did not find an expected decrease in the ubiquitylation levels of TIM in these animals [152]. Only upon the concurrent inhibition of CUL3 and SLIMB activity is there a reduction in TIM ubiquitylation, suggesting that these two E3 ligases are partly redundant or act cooperatively [152]. CUL3/TIM association at the day–night transition temporally coincides with the detection of K48-linked polyubiquitination of TIM, a signal for proteasomal degradation [152]. Interestingly, the K48-linked polyubiquitination of the signaling protein Ci (Cubitus interruptus) has previously been shown to be favoured by CUL3-based SCF complexes, whereas SLIMB-based SCF complexes preferentially catalyze K11 linkages [153]. It remains to be seen whether the same holds true for the ubiquitylation of TIM by CUL3 and SLIMB, and how potential differences in Ub linkage affect the fate of TIM. In addition to its effects on TIM, CUL3 influences the accumulation of PDF in the axonal terminals of s-LN_vs [133].

TANGO10 is an adaptor protein for CUL3 and contains both BTB (broad-complex, tramtrack, and bric à brac) and BACK (BTB and C-terminal kelch) domains, which are important for protein–protein interactions [133]. Global loss-of-function mutations in tango10 disrupt behavioural rhythmicity under DD, whereas the ubiquitous or pan-neuronal overexpression of wild-type tango10 rescues this phenotype [133]. Suppressing tango10 overexpression specifically in PDF neurons blocks the rescue, indicating that this gene is required in PDF neurons to maintain behavioural rhythms [133]. As expected, the silencing of endogenous tango10 in PDF neurons also leads to behavioural arrhythmicity [133]. In global loss-of-function tango10 mutants, the peak expression and oscillatory amplitude of PER and TIM are markedly reduced in the s-LN_vs but not in other groups of pacemaker neurons in the brain [133]. Intriguingly, PDF accumulates in the dorsal axonal terminals of the s-LN_vs of tango10 mutants to aberrantly high levels [133]. TANGO10 co-localizes with PDF in the dorsal terminals and fluctuates in expression in a circadian fashion [133]. In tango10 loss-of-function mutants, PDF neurons are hyperexcitable, potentially as a result of reductions in I_A potassium currents [133]. However, it is not known whether these changes in neuronal excitability are caused by, or the result of, perturbed molecular oscillations in the absence of tango10.

FBXL4 is an F-box protein of the Fbls class, and as such possesses tandem leucine-rich repeats that are important for protein–protein interactions [58]. Drosophila fbxl4 is expressed in the l-LN_vs in a rhythmic manner, peaking in the middle (mRNA) or the late (protein) night [154]. FBXL4 physically interacts with and promotes the ubiquitination of the GABA_A receptor, resistant to dieldrin (RDL) [154]. This ubiquitylation event targets RDL for degradation [154]. In the l-LN_vs, RDL abundance is rhythmic, reaching its peak in the late day and its nadir in the late night, anti-phasic to FBXL4 oscillations [154]. The daily fluctuation in RDL expression modulates the GABA sensitivity, and thus excitability, of l-LN_vs [154]. Although fbxl4 is dispensable for circadian rhythms at the behavioural level, as an output of the circadian clock (i.e., the fbxl4 gene promoter contains E-box elements and is directly regulated by CLOCK:CYCLE), it is critical for modulating the onset and duration of sleep through its effects on RDL in the l-LN_vs [154].