Research ArticleCell Biology

Distinct control of PERIOD2 degradation and circadian rhythms by the oncoprotein and ubiquitin ligase MDM2

See allHide authors and affiliations

Science Signaling  13 Nov 2018:
Vol. 11, Issue 556, eaau0715
DOI: 10.1126/scisignal.aau0715

Multiple routes to PER2 degradation

The circadian clock is formed by a 24-hour transcriptional-translational feedback loop, wherein the abundance of its core components—CLOCK, BMAL, PER, and CRY—influences cellular biochemistry and responses to stimuli at different times of the day. The E3 ubiquitin ligase β-TrCP promotes the degradation of phosphorylated PER2. Liu et al. found that the E3 ligase MDM2 also ubiquitylated and promoted the degradation of PER2 but did so in a manner that was independent of PER2 phosphorylation. The E3 ligase activity of MDM2 was required for maintaining the proper length of the circadian cycle. In addition to identifying a previously unknown mechanism that contributes to PER2 stability, these findings may have implications in cancer, because MDM2 is an oncoprotein that affects the cellular response to genotoxic stress, and the loss of PER2 makes cells more susceptible to genotoxic stress and animals more prone to develop cancer.


The circadian clock relies on posttranslational modifications to set the timing for degradation of core regulatory components, which drives clock progression. Ubiquitin-modifying enzymes that target clock components for degradation mainly recognize phosphorylated substrates. Degradation of the circadian clock component PERIOD 2 (PER2) is mediated by its phospho-specific recognition by β-transducin repeat–containing proteins (β-TrCPs), which are F-box–containing proteins that function as substrate recognition subunits of the SCFβ-TRCP ubiquitin ligase complex. However, this mode of regulating PER2 stability falls short of explaining the persistent oscillatory phenotypes reported in biological systems lacking functional elements of the phospho-dependent PER2 degradation machinery. We identified PER2 as a previously uncharacterized substrate for the ubiquitin ligase mouse double minute 2 homolog (MDM2) and found that MDM2 targeted PER2 for degradation in a manner independent of PER2 phosphorylation. Deregulation of MDM2 plays a major role in oncogenesis by contributing to the accumulation of genomic and epigenomic alterations that favor tumor development. MDM2-mediated PER2 turnover was important for defining the circadian period length in mammalian cells, a finding that emphasizes the connection between the circadian clock and cancer. Our results not only broaden the range of specific substrates of MDM2 beyond the cell cycle to include circadian components but also identify a previously unknown regulator of the clock as a druggable node that is often found to be deregulated during tumorigenesis.


Circadian rhythms are endogenously generated 24-hour oscillations of biochemical, physiological, and behavioral processes that allow organisms to adapt to external environmental conditions. The coupling of the mammalian circadian system to various cellular processes provides a means to understand the timing of when events take place in normal proliferative cells, a phenomenon believed to reflect evolutionary adaptation [for review, see (1)]. As a result, identifying shared regulatory elements that operate under normal conditions and are relevant to the timely execution of cellular events might help establish nodes that, when deregulated, could be relevant to human pathologies.

From a molecular standpoint, the circadian clock is formed by a transcriptional-translational feedback loop, wherein the expression of genes encoding the core components of the clock drives the different phases of the daily cycle and the protein products of those genes influence the cell’s biochemistry (2). In mammalian cells, the clock is driven by the heterodimer formed by the transcription factors circadian locomotor output cycles kaput (CLOCK) and the brain and muscle Arnt-like protein-1 (BMAL1), which initiates the transcription of PERIOD (PER) and CRYTOCHROME (CRY) genes (PER1, PER2, PER3, CRY1, and CRY2) as well as other clock-controlled genes (ccgs), several of which encode cell cycle regulators. Heterodimerization of PER and CRY feeds back to negatively regulate the clock because nuclear translocation of the PER:CRY complex inhibits further CLOCK:BMAL1 transcriptional activity [for review, see (3)]. Thus, the stability of PER and CRY determines the timing with which the repression phase ends and a new round of transcription begins. This process is mediated by distinct phosphorylation events in PER and CRY that precede their E3 ligase–mediated ubiquitylation and proteasomal degradation [for review, see (3) and references within].

PER2 is a large protein with a well-defined N-terminal region containing two tandemly organized PER-ARNT-SIM (PAS) domains (known as PAS-A and PAS-B) (4) followed by a PAS-associated C-terminal motif (PAC), which contributes to PAS domain folding (5). Structural and functional studies have established that the PAS domains of PER proteins participate in critical intermolecular interactions and are responsible for multiple homo- and heterodimeric protein interactions as well as PER’s stability, cellular localization, and inhibitory activity toward CLOCK:BMAL (510). In addition, PER2 contains motifs and domains that play critical functional roles in its cellular localization (nuclear localization and export signal motifs), stability (binding domain for E3 ligase), and targeted posttranslational modifications [phosphorylation sites for casein kinases 1 ε and δ (CK1ε and CK1δ) and glycogen synthase kinase 3β (GSK3β)] that influence the periodic accumulation and distribution of PER2 in the cell [for review, see (11)]. Furthermore, the stability of PER2, which seems to depend on its phosphorylation status (12, 13), is influenced by environmental stimuli and homeostatic cellular conditions (1416) and is a critical determinant of the period length and phase of circadian rhythms (12, 13, 17). As a result, PER2 acts as a cellular rheostat that integrates signals and helps to robustly compensate for profound changes in environmental conditions that would otherwise disrupt the circadian clock.

Phosphorylation of PER2 by CK1ε or CK1δ (CK1ε/δ) can either stabilize or destabilize PER2 depending on which site in PER2 is modified (15, 17, 18). Accordingly, PER2S662G, a PER2 variant linked to familial advanced sleep phase syndrome (19), contains a missense mutation that prevents priming-dependent phosphorylation of flanking sites by CK1ε/δ, thus stabilizing PER2 independent of its cellular location (20). Conversely, priming-independent clusters located in the C-terminal portion of PER2’s PAS domains are targeted by CK1ε/δ and are required for ubiquitin ligase–mediated degradation of PER2 (21). Presently, our understanding of the molecular players involved in PER2 degradation is limited to the role of β-transducin repeat–containing protein (β-TrCP), an F-box/WD40 repeat–containing substrate recognition subunit of the ubiquitin ligase complex SCF (Skp1–Cul1–F-box), which mediates phosphorylation-dependent degradation of proteins (21, 22). The mammalian β-TrCP E3 ligase subfamily includes β-TrCP1 and β-TrCP2 (β-TrCP1/2), which are closely related in sequence and indistinguishable in function but encoded by different genes (23). Biochemical evidence points to direct interactions between β-TrCP1/2 and PER1, but β-TrCP1 appears to be the sole form implicated in the binding of PER2 in vitro (21, 24). Regardless of these findings, there is no clear answer as to whether β-TrCP isoform selectivity of the PER substrate actually occurs in vivo, although β-TrCP–mediated degradation contributes to generating cyclic quantities of PER proteins relevant to the function of the clock (22). As has been noted, endogenous β-TrCP activities depend on both their localization and abundance in cells, with β-TrCP1 being predominantly located in the nucleus and β-TrCP2 being predominantly located in the cytoplasmic compartment; cytoplasmic β-TrCP2 is the most unstable form of either E3 ligase (23).

Overexpression of both dominant-negative forms of β-TrCP in cells neither increases PER2 stability nor causes phosphorylated PER2 to accumulate; instead, it results in rapid degradation of PER2 by a yet unknown mechanism (22). Similarly, expression of the dominant-negative form of CK1ε in a CK1δ−/− background perturbs, but does not abrogate, circadian rhythms (25), a result that mimics those obtained using pharmacological inhibitors of CK1ε/δ (26). Furthermore, Zhou et al. (15) have defined a phosphoswitch in PER2 that controls a three-stage kinetic degradation process. This mechanism allows fine-tuning of the fraction of PER2 that is stable, thereby adjusting the length of the circadian period in response to diverse environmental stimuli (15). Whereas the rapid initial decay in PER2 abundance is phosphorylation dependent and mediated by the activity of β-TrCP, the second “plateau” stage results from priming phosphorylation and accumulation of PER2 (15). Triggering PER2’s degradation in the third kinetic stage and during the falling phase of the circadian cycle is hardly affected by CK1 inhibition and is likely independent of both phosphorylation and β-TrCP activity (15). Moreover, the aberrant sleep-wake behavioral phenotype of mice bearing loss-of-function mutations in the genes encoding β-TrCP1 and β-TrCP2 is molecularly linked to atypical kinetics of PER2 degradation (27). In β-TrCP1/2 knockout cells, PER proteins were still ubiquitylated and degraded, although at a slower rate than in wild-type controls (27). Thus, other E3 ligase(s) in addition to β-TrCP1/2 are likely to target PER proteins and contribute to overall PER stability.

We previously reported that PER2 forms a stable complex with the tumor suppressor and checkpoint protein p53 (28). The PER2:p53 complex undergoes time-of-day–dependent nuclear-cytoplasmic shuttling, thus generating an asymmetric distribution of each protein in different cellular compartments (29). In unstressed cells, PER2 mediates p53’s stability by binding to its C-terminal domain and preventing p53 from being ubiquitylated at sites targeted by the RING finger–containing E3 ligase mouse double minute 2 homolog (MDM2) (28). We found that PER2, p53, and MDM2 coexist as a trimeric and stable complex in the nuclear compartment, although p53 is released from the complex to become transcriptionally active after cells experience genotoxic stimuli (30). As a result, we asked whether PER2 could also act as a bona fide substrate for the E3 ligase activity of MDM2 in the absence of p53. Unlike β-TrCP, MDM2 acts as a scaffold protein to facilitate catalysis by bringing the E2 ubiquitin–conjugating enzyme and substrate together in a phosphorylation-independent manner (23, 31). In this study, we show that whereas an N-terminal epitope in MDM2 was critical for PER2:MDM2 complex stability, other epitopes located near MDM2’s C terminus, where MDM2’s E3-RING activity resides, were dispensable for this interaction. Consequently, PER2 was efficiently ubiquitylated in vitro and in cells at numerous sites by MDM2 in a process that was preferentially mediated by UbcH5a, a robust E2 ubiquitin–conjugating enzyme with innate preference for various polyubiquitin chain linkages (32). Furthermore, MDM2-mediated ubiquitylation of PER2 was independent of PER2 phosphorylation. Accordingly, PER2’s half-life was critically influenced by the abundance and enzymatic activity of MDM2, as shown in cells in which MDM2 expression was either enhanced or silenced and its catalytic activity was pharmacologically inhibited. As a consequence, direct manipulation of MDM2 expression influenced period length by reducing PER2 stability. Therefore, our results provide evidence that phosphorylation-independent degradation of PER2 also contributes to tight control of PER2 turnover in cells, contrary to the prevalent phosphorylation-centric view of its degradation.


The oncogenic E3 ligase MDM2 interacts with PER2 in the absence of p53

In vertebrates, phosphorylation-dependent β-TrCP–mediated ubiquitylation and proteasomal degradation provides a means of regulating endogenous amounts of PER2 in the cell and, thus, its daily accumulation [for review, see (14)]. A large body of evidence supports a more central role for PER2 as the integrator of intracellular signals and as a sensor of environmental conditions. Thus, much effort has been devoted to understanding how various phosphorylation events determine PER2’s degradation rate (1416, 33, 34), whereas phosphorylation-independent mechanisms of degradation have remained largely unexplored. Building on our previous finding that PER2 forms a stable complex with the tumor suppressor p53 and the RING-type E3 ligase MDM2 (28, 30), we evaluated whether MDM2 provides an alternative route for degradation of PER2 that is independent of phosphorylation and, at the same time, influences the circadian period.

Using a bacterial two-hybrid system, we identified human MDM2 as a direct interactor of PER2. We tested for protein-protein interactions using bacteria coexpressing a pTRG human liver complementary DNA (cDNA) library and pBT-PER2 bait constructs (28). Each of the three bait constructs [full-length PER2, the N-terminal region of PER2 (residues 1 to 821), and the C-terminal region of PER2 (residues 822 to 1255)] were screened independently. About 4 × 106 clones were screened, and putative positive interactors were preserved in nonselective medium in the presence of antibiotics, verified in selective screening medium containing 3-amino-1,2,4-triazole (3-AT), and confirmed by patching onto a dual-selective medium containing 3-AT and streptomycin (Fig. 1A). Cotransformations of the reporter strains with pBT-PER2/pTRG-CRY and pBT-LGF2/pTRG-Gal11p were used as positive interacting controls (Fig. 1A). False-positive interactions in two-hybrid screenings were initially ruled out by patching in dual-selective medium and by including cotransformation controls with empty vectors (Fig. 1A). Results from these screenings suggested that, in addition to the already identified PER2:p53:MDM2 complex (30), PER2:MDM2 might exist as its own entity and that this association might be independent of p53 binding (Fig. 1A).

Fig. 1 PER2 directly interacts with MDM2.

(A) Putative interactors of PER2 were identified using a two-hybrid screening method in which bacteria cells were cotransformed with a plasmid containing PER2 cDNA as bait (pBT-PER2) and a human cDNA liver library in pTRG. Cotransformed cells were maintained in tetracycline and chloramphenicol medium (LB Tet/Cam) to select for the presence of the two plasmids and selected for interaction by growing in His-dropout selective medium (MM) containing either 3-AT or 3-AT + streptomycin (Strep). Cells cotransformed with λcI-Gal4 (pBT-LGF2) and α-Gal11P (pTRG-Gal11P) as well as pBT-PER2 and pTRG-CRY were used as positive interacting controls. Cotransformants containing the empty pTRG vector were used as negative controls. Results with controls and pTRG-MDM2 are shown. n = 6 independent experiments. (B) Representative immunoblot showing PER2 and MDM2 in immunoprecipitates (IP) from isogenic p53+/+ or p53−/− HCT116 cells treated with the MG132 proteasome inhibitor and expressing Myc-tagged MDM2 (Myc-MDM2) or an empty vector. Extracts were immunoprecipitated with antibodies recognizing endogenous PER2. An IgG antibody (Ab) was used as negative control. n = 3 independent experiments. (C) The indicated in vitro transcribed and translated tagged proteins (lanes 1 to 4) were mixed and incubated to allow complexes to form before immunoprecipitation with antibodies recognizing the FLAG tag and immunoblotting for Myc and FLAG. n = 3 independent experiments. (D) Competition experiments were carried out by preincubating in vitro transcribed and translated FLAG-MDM2(C470A) with the 4B2, SMP14, and 4B11 antibodies that recognize residues 19 to 59, 154 to 167, and 383 to 491 of MDM2, respectively, before adding recombinant Myc-PER2. Protein binding to FLAG antibody–conjugated beads was detected by immunoblotting for Myc and FLAG. IgG was used as negative control. n = 2 independent experiments. Molecular weight markers (in kilodaltons) are indicated to the left of each blot.

We next evaluated the presence of the PER2:MDM2 complex in cells lacking endogenous expression of p53. Initial experiments were carried out using colorectal HCT116 cells (TP53+/+, PER2+/+; referred to as HCT116p53+/+ hereafter) and, to avoid confounding variables, its null-isogenic HCT116 cell variant lacking p53 expression [TP53−/−, PER2+/+; referred to as HCT116p53−/− hereafter] (Fig. 1B) (35). An antibody recognizing PER2, but not an immunoglobulin G (IgG) control antibody, pulled down endogenous PER2-associated MDM2 in both HCT116p53+/+ and HCT116p53−/− cell extracts, further supporting their p53-independent interaction (Fig. 1B). The p53 and MDM2 proteins form a regulatory feedback loop in which p53 transcriptionally increases MDM2 expression, and cells lacking p53 usually exhibit a constitutively low amount of MDM2, which is enhanced by MDM2’s self-ubiquitylation activity and increased turnover [for review, see (36)]. Accordingly, detection of the endogenous PER2:MDM2 complex in HCT116p53−/− cells was facilitated by the addition of the proteasome inhibitor MG132 to the reaction (Fig. 1B). In addition, Myc-tagged MDM2 coimmunoprecipitated with endogenous PER2 in transfected HCT116p53−/− cells (Fig. 1B). Similar results were obtained using a human non–small cell lung carcinoma line (H1299) that has a homozygous partial deletion of the TP53 gene. In this case, complexes were detected in cells cotransfected with Myc-PER2 and FLAG-tagged MDM2 (FLAG-MDM2), the ubiquitin ligase activity–deficient mutant FLAG-MDM2(C470A) (37), or the E3 ligase β-TrCP1 (fig. S1A). Last, to further dissect molecular interactions in the PER2:MDM2 complex, we used in vitro transcribed and translated tagged proteins to develop a cell-free system that recapitulated the interaction between PER2 and MDM2 that was observed in cells (Fig. 1C).

To better understand the interaction between these two molecules, we mapped the regions of each protein involved in PER2-MDM2 binding. Epitope mapping was carried out by preincubating recombinant FLAG-MDM2(C470A) with various epitope-specific antibodies that recognize native conformations of MDM2 [4B2 (residues 19 to 59), SMP14 (residues 154 to 167), and 4B11 (residues 383 to 491, the RING domain) (38)] before adding recombinant Myc-PER2 (fig. S1B). Preincubation with SMP14 completely abolished PER2 binding (Fig. 1D), suggesting that the epitope comprising residues 154 to 167 of MDM2 is critical for the stability of the PER2:MDM2 interaction. Accordingly, immunoprecipitation of various FLAG-MDM2 deletion mutants confirmed that the N-terminal hydrophobic pocket of MDM2 (residues 1 to ~110) was dispensable for PER2 recognition, although it contacted PER2 (fig. S1C). Conversely, transfection of HCT116p53−/− cells with various recombinant FLAG and GST (glutathione S-transferase) dual-tagged constructs of PER2 [FLAG-GST-PER2(182–475), FLAG-GST-PER2(1–185), FLAG-GST-PER2(186–404), FLAG-GST-PER2(575–682), FLAG-GST-PER2(683–872), FLAG-GST-PER2(873–1135), FLAG-GST-PER2(1121–1255), and FLAG-GST-PER2(1–1255)] showed that MDM2 binding primarily occurred within the PER2 PAS-containing domain [residues 182 to 475 (5)] and in an inner region (residues 683 to 872) that is heavily posttranslationally modified (fig. S1D). Overall, our findings establish the existence of a PER2:MDM2 complex that forms independently of the presence of p53.

PER2 is a bona fide substrate of MDM2

Next, we asked whether MDM2 could ubiquitylate PER2. To test this possibility, we evaluated PER2 ubiquitylation in a cell-free system enriched in E1 and E2 enzymes containing in vitro transcribed and translated FLAG-MDM2 or FLAG-MDM2(C470A) and multi-tagged Myc-DO-PER2 proteins. Our data showed that Myc-DO-PER2 was ubiquitylated, as evidenced by the formation of a high–molecular weight smear, when it was incubated in the presence of wild-type MDM2, but not the ligase-deficient mutant form (Fig. 2A). Similarly, ubiquitylated forms of Myc-PER2 immunoprecipitated from in vitro reactions performed in the presence of FLAG-ubiquitin (fig. S2A) and from lysates of cotransfected cells were maintained in the presence of the proteasome inhibitor MG132 (fig. S2B).

Fig. 2 The E3 ligase MDM2 targets PER2 for ubiquitylation.

(A) In vitro ubiquitylation reactions were carried out using recombinant transcribed and translated Myc-DO-PER2, FLAG-MDM2, or MDM2(C470A) plus E1 and E2 enzymes. Samples were immunoblotted with antibodies recognizing the DO and FLAG epitopes. (B) Fragments of PER2 comprising residues 356 to 872 and 873 to 1255 were expressed using the TNT system as Myc-tagged recombinant proteins, and purified proteins were used for in vitro ubiquitylation assays and immunoblotted for Myc and FLAG. (C) In vitro ubiquitylation reactions were carried out using in vitro transcribed and translated proteins at the indicated enzyme/substrate ratios and immunoblotted for ubiquitin, Myc, and FLAG. Polyubiquitylated forms of Myc-PER2(683–872) are indicated with a bracket. For all panels, n = 3 independent experiments.

We then asked whether ubiquitin modifications of PER2 were confined to specific domains in the protein. We focused on three relevant regions in PER2: (i) the N-terminal region containing the PAS-A and PAS-B domains and the PAC domain (residues 1 to 682), which are engaged in homo- and heterodimeric protein-protein interactions (5, 6, 810); (ii) a middle region that is heavily posttranslationally modified and mainly involved in protein-protein interactions and in cellular shuttling of PER2 (residues 356 to 872); and (iii) a C-terminal fragment (873 to 1255) that directly interacts with various ligands and with PER2’s binding partner CRY (Fig. 2B and fig. S2C) (39, 40). To this end, we reconstituted a functional E1:E2:MDM2 or MDM2(C470A) ubiquitin ligase in vitro system and used various recombinant tagged fragments of PER2 as substrates. The results showed that PER2(1–682), PER2(356–872), and PER2(873–1255) incorporated multiple ubiquitin moieties when incubated with wild-type MDM2, but not when incubated with MDM2(C470A), as shown by the presence of a smear in immunoblots (Fig. 2B and fig. S2C).

We previously established that residues within the center region of PER2 are at the interface with p53 and facilitate the formation of a stable complex where circadian and checkpoint signals converge (28, 30). Thus, we turned our attention to defining ubiquitylation events taking place, specifically, within this region (residues 683 to 872) of PER2. As shown for p53, whereas ubiquitylation by MDM2 is influenced by various factors including the enzyme/substrate ratio, the incorporation of diverse ubiquitin chains in the substrate results from the multivalent nature of linkage-specific conjugations (41, 42). Therefore, we initially optimized the reaction conditions to ensure that targeting of PER2(683–872) by MDM2 was taking place within the initial velocity region and the ubiquitin conjugation was biochemically defined (fig. S3). Initially, we tested scenarios in which we varied the amounts of adenosine 5′-triphosphate (ATP)–Mg2+ chelate (fig. S3A), ubiquitin cosubstrate (fig. S3A), and time during which products accumulated (fig. S3B). Next, because MDM2 cooperates with various E2 ubiquitin–conjugating enzymes from the UbcH5 family (UbcH5a, UbcH5b, and UbcH5c) to mediate either specific or promiscuous ubiquitin linkages on different substrates (42), we tested the relevance of UbcH5 specificity for MDM2-mediated ubiquitylation of PER2(683–872) (fig. S3C). In vitro ubiquitylation reactions were performed using recombinant enzymes (E1 and UbcH5a, UbcH5b, or UbcH5c) and tagged MDM2, MDM2(C470A), and PER2(683–872) substrates. The p53 protein was used as a positive control because it is efficiently modified by MDM2 in the presence of ubiquitin-conjugated UbcH5a, UbcH5b, or UbcH5c (fig. S3C). All three UbcH5 enzymes promoted the addition of at least a single ubiquitin molecule to PER2(683–872), but UbcH5a appeared to be more effective in catalyzing the incorporation of at least a second molecule (fig. S3C). Among the eight possible types of ubiquitin linkages, UbcH5a exhibits selectivity for Lys11, Lys48, and Lys63 linkages (43). Of these, Lys11 and the canonical Lys48 linkages are involved in the formation of polyubiquitin chains and proteasome-mediated turnover, whereas the Lys63 linkage plays a nondegradative role and is usually involved in protein recruitment and localization (43). Last, we observed a dose-dependent effect of MDM2 in the accumulation of slower-migrating polyubiquitylated forms of PER2(683–872) that were undetectable when the reaction took place in the presence of increasing concentrations of MDM2(C470A) (Fig. 2C and fig. S4). Thus, our results indicate the existence of early ubiquitylation events within PER2 upon which polyubiquitylation chains can be built.

MDM2 ubiquitylates PER2 at conserved Lys residues flanking the CK1 site

To gain further insight into the role that MDM2 plays in PER2 function, we focused our efforts on identifying relevant lysine residues in PER2 that could be targeted for modification. A highly conserved cluster of lysine residues (Lys789, Lys790, Lys793, Lys796, Lys798, Lys800, and Lys803) mapping within PER2(683–872) was targeted for mutagenesis (fig. S5, A and B), and recombinant proteins were subjected to in vitro ubiquitylation with MDM2 or MDM2(C470A) (Fig. 3). PER2(683–872) was efficiently ubiquitylated by MDM2; however, a form of PER2(683–872), in which all seven conserved lysine residues were replaced with alanine [PER2(683–872)-KA] (fig. S5B), was not targeted for modification (Fig. 3A). These results indicate that one or more of the Lys residues in PER2(683–872) is a putative target for MDM2-mediated ubiquitylation in cells (Fig. 3).

Fig. 3 Residues within the central domain of PER2 are ubiquitylated by MDM2.

(A) In vitro ubiquitylation reactions were carried out using purified recombinant UbcH5a, either FLAG-MDM2 or the catalytically inactive form FLAG-MDM2(C470A), plus various Myc-tagged forms of PER2(683–872): Myc-PER2(683–872), wild-type (WT); Myc-PER2(683–872)-WT-KA, containing the three substitution mutations K798A, K800A, and K803A; Myc-PER2(683–872)-KA-WT, containing the four substitution mutations K789A, K790A, K793A, and K796A; and Myc-PER2(683–872)-KA, in which all seven lysine residues were mutated to alanine (K789A, K790A, K793A, K796A, K798A, K800A, and K803A). The reactions were immunoblotted for Myc and FLAG. n = 3 independent experiments. (B) In vitro ubiquitylation reactions were carried out using recombinant forms of Myc-PER2(683–872), in which each lysine residue was replaced with alanine individually. Reactions were performed in the presence of UbcH5a and either FLAG-MDM2 or FLAG-MDM2(C470A). The reactions were immunoblotted for Myc and FLAG. n = 2 independent experiments. Molecular weight markers (in kilodaltons) are indicated to the left of each blot, and modified forms of the protein substrate are indicated by arrows on the right. Representative results are shown in all panels.

To gain further insight into the relevance of the lysine residues in this cluster for PER2 ubiquitylation, we generated two additional PER2(683–872) mutants, with each mutant protein having either the first four or the last three lysine residues substituted with alanine. The first four lysine residues were replaced with alanine (K789A, K790A, K793A, and K796A) in PER2(683–872)-KA-WT, whereas the last three lysine residues were replaced with alanine (K798A, K800A, and K803A) in PER2(683–872)-WT-KA (fig. S5B). Ubiquitylation assays showed that, whereas the addition of ubiquitin moieties was reduced in PER2(683–872)-WT-KA, ubiquitin incorporation was completely abrogated in PER2(683–872)-KA-WT (Fig. 3A). Although this result might imply that ubiquitylation events occur primarily within the upstream mutated lysine residues in PER2(683–872)-KA-WT, we cannot rule out other scenarios including orderly addition of ubiquitin moieties and structural rearrangements, phenotypes that might be disrupted as a result of a single mutation.

To investigate the role of specific lysine residues in PER2 ubiquitylation, we first generated single substitution mutations in the lysine cluster using site-directed mutagenesis and evaluated their performance in an in vitro ubiquitylation assay. Overall ubiquitylation was reduced when unique lysine residues, instead of clusters, were replaced by alanine, in particular, Lys789, Lys790, Lys796, and Lys800 (Fig. 3B and fig. S5E). Unfortunately, there is limited structural information currently available on PER2 that could aid in interpreting the aforementioned results (5, 44, 45), especially with respect to the 683 to 872 region. Therefore, although they are theoretical in nature, we used protein-protein docking simulations to help us predict binding interfaces for ubiquitin molecules in each PER2 fragment that was experimentally assessed in Fig. 3A (fig. S5, C and D). Accordingly, we generated molecular models of wild-type PER2(683–872), PER2(683–872)-WT-KA, and PER2(683–872)-KA-WT and used quality assessment tools to estimate their reliability before carrying out unbiased protein-protein docking simulations (figs. S5C and S6, A to D). Unbiased protein-protein docking results strongly favored the placement of a ubiquitin molecule within the Lys798-Lys803 domain, with the C-terminal end of ubiquitin making direct contact with Lys800 (fig. S5C). Modeling and docking predictions suggested that the mutations in PER2(683–872)-WT-KA would cause a conformational change that would result in none of the remaining lysine residues in the fragment, being accessible for ubiquitylation, except for Lys750, which is outside of the two clusters (fig. S5C). These findings arise from the analysis of major cluster hits for each protein-ubiquitin docking simulation, the identification of the two most dominant ubiquitin interface poses (fig. S5C), and the comparison of conformational states that show differences between PER2(683–872) and the PER2(683–872)-WT-KA or PER2(683–872)-KA-WT interfaces, although rigorous validation of modeling predictions inevitably rests upon experimental data.

Modeling results also predicted that it would be unlikely for PER2(683–872)-KA-WT to be ubiquitylated at any lysine residue due to a large conformational change that would preclude the access of ubiquitin to a reactive lysine residue (Fig. 3A and fig. S5C). In agreement with the predicted models and protein-ubiquitin docking results, overall ubiquitylation was reduced when single Lys residues, instead of clusters, were replaced by alanine (Fig. 3B and fig. S5E). Ubiquitylation experiments carried out using PER2(683–872)-K750A and PER2(683–872)-WT-KA-K750A, both of which contain the K750A mutation, supported our molecular model (fig. S5D). Whereas PER2(683–872)-K750A showed reduction ubiquitylation compared to wild-type PER2(683–872), posttranslational modification was, as predicted, completely abrogated in PER2(683–872)-WT-KA-K750A (fig. S5D). Overall, our results show that a conserved cluster of Lys residues comprising residues 789 to 803 is directly targeted for ubiquitylation, albeit to a different extent, and that this phenomenon may reflect conformational changes and substrate accessibility occurring in the protein as result of ubiquitylation modifications.

MDM2 binding to and ubiquitylation of PER2 is independent of PER2 phosphorylation status

Whereas phosphorylation of mouse Per2 on Ser478 (Ser480 in human PER2) by CK1ε/δ is a prerequisite for β-TrCP binding and subsequent ubiquitylation (21, 24), priming phosphorylation at Ser659 (Ser662 in human PER2) by CK1ε/δ, or at downstream sites is not directly involved in β-TrCP–mediated degradation (15, 20, 46, 47). Therefore, we asked whether phosphorylation of PER2 was required for MDM2 to bind to or ubiquitylate PER2, or both.

To determine whether MDM2 binding to PER2 depended on PER2 phosphorylation, we incubated Myc-MDM2 with FLAG-PER2 that had been pretreated with either λPPase, a promiscuous phosphatase with activity toward phosphorylated serine, threonine, and tyrosine residues, or CK1ε(1–305), a constitutively active form of CK1ε (Fig. 4A and fig. S7A). CK1ε(1–305) lacks the autoinhibitory C-terminal domain and phosphorylates substrates to the same extent as its full-length counterpart [fig. S7, B and C, and (48)]. Pretreatment of PER2 with λPPase did not abrogate binding to either wild-type MDM2 or catalytically inactive MDM2(C470A); thus, neither phosphorylation of PER2 nor MDM2 E3 ligase activity was a prerequisite for formation of the PER2:MDM2 complex (Fig. 4A). Both MDM2 and MDM2(C470A) were also able to bind to recombinant PER2 that had been subjected to CK1ε(1–305)-mediated phosphorylation and purified by affinity chromatography (Fig. 4A). These in vitro assays establish that CK1ε-mediated phosphorylation of PER2 does not abrogate the formation of the PER2:MDM2 complex. Notably, treatment of PER2 with either CK1ε(1–305) or λPPase resulted in a conspicuous accumulation of PER2 (Fig. 4A). Although addressing this particular finding is beyond the scope of the present work, we speculate that changes in PER2 abundance may result from its binding to CK1ε(1–305) and posttranslational events associated with folding that are more noticeable in the context of an in vitro assay, as biochemical studies have previously suggested (15). This speculation would benefit from more structural information about PER2.

Fig. 4 Binding and ubiquitylation of PER2 by MDM2 is independent of PER2 phosphorylation status.

(A) In vitro transcribed and translated FLAG-PER2 was treated (+) or not (−) with either λ phosphatase (λPPase) or Myc-CK1ε(1–305), a constitutively active form of CK1ε. The dephosphorylated (λPPase-treated) or phosphorylated (CK1ε-treated) FLAG-PER2 was then incubated with either Myc-MDM2 or catalytically inactive Myc-MDM2(C470A) in binding buffer. The reactions were immunoprecipitated with an antibody recognizing the FLAG epitope and immunoblotted for FLAG and Myc. Lanes 7 to 9 are negative controls. Blots are representative of n = 3 independent experiments. (B) Equal amounts of extracts from p53+/+ HCT116 cells that had been treated (or not, control) with the CK1ε/δ inhibitor PF-670462 (PF670) overnight were immunoprecipitated with an antibody that recognizes endogenous PER2 (or IgG, control) and protein A–conjugated beads that bind immunoglobulins before immunoblotting for PER2 and MDM2. Inputs correspond to aliquots (~100 μg) of total extracts. The asterisk indicates a nonspecific band. Immunoblots are representative of n = 3 independent experiments. (C) p53+/+ HCT116 cells were cotransfected with either an empty vector (−) or a plasmid encoding Myc-CK1ε plus a plasmid encoding FLAG-PER2, FLAG-PER2(S480A), FLAG-PER2(S662A), or FLAG-PER2(S662D). Cells were treated with the proteasome inhibitor MG132 (+MG) or vehicle (−MG) for 4 hours before harvesting. Protein complexes were immunoprecipitated using antibodies recognizing FLAG and protein A–conjugated beads and immunoblotted for FLAG, MDM2, Myc, and β-TrCP1. IgG was used as a negative control. Immunoblots are representative of n = 3 independent experiments. (D) In vitro transcribed and translated Myc-PER2(356-872) and Myc-p53 were treated with λPPase before carrying out ubiquitylation reactions in the presence or absence of FLAG-MDM2. FLAG-p53 was used as a positive control. The circled “+” symbols in (A) and (D) indicate that Myc-MDM2 or FLAG-MDM2 was the last protein added to the reactions. Molecular weight markers (in kilodaltons) are indicated on the left, and modified forms of the protein substrate are indicated by arrows on the right. Immunoblots are representative of n = 2 independent experiments. Representative results are shown in all panels.

We next tested whether endogenous PER2 was able to bind MDM2 despite the inhibition of CK1δ/ε activity in cultured cells (Fig. 4B). Endogenous PER2:MDM2 complexes were immunoprecipitated from HCT116 p53+/+ cells treated with PF-670462 (49), a specific CK1ε/δ inhibitor with a proven effect on circadian rhythms (15, 50). As expected, and based on the role of CK1ε/δ in promoting PER2 instability, treatment of cells with PF-670462 resulted in an apparent overall increase in endogenous PER2 abundance in comparable amounts of extracts from vehicle-treated cells (Fig. 4B). Immunoprecipitation of PER2 from extracts of PF-670462–treated cells showed binding of endogenous MDM2 to PER2, a qualitative result that reflects the presence of a stable PER2:MDM2 complex despite the absence CK1ε/δ kinase activity. Further speculation on the role of CK1δ/ε in PER2-MDM2 binding should be avoided because of differences in the amount of immunoprecipitated PER2 between samples. Overall, our data favor a model in which the formation of the endogenous PER2:MDM2 complex occurs regardless of the activity of CK1ε/δ toward the PER2 substrate.

We then asked whether modification in the critical PER2 Ser662 priming site plays a role in PER2 binding to MDM2 or β-TrCP1. Both wild-type PER2 and the PER2(S662A) mutant bound MDM2 in cotransfected H1299 cells (fig. S8A), indicating that MDM2 binding to PER2 is independent of priming modifications at Ser662. Immunoprecipitation experiments in which in vitro transcribed and translated proteins were allowed to form complexes before immunoprecipitation also indicated that the S662A mutation did not compromise β-TrCP1 binding to PER2 (Fig. 1C).

We next evaluated whether the interplay between phosphorylation at the Ser480 and Ser662 sites would affect the distribution of MDM2 bound to PER2, as was previously reported to be the case for β-TrCP1 bound to PER2 (15). We cotransfected HCT116p53+/+ cells with constructs encoding Myc-CK1ε plus FLAG-PER2, FLAG-PER2(S480A), PER2(S662A), or PER2(S662D), which mimics phosphorylation at Ser662, and expressed the recombinant proteins in the presence or absence of the proteasome inhibitor MG132 (Fig. 4C and fig. S8B). In all cases, endogenous MDM2 and β-TrCP1 coimmunoprecipitated with FLAG-tagged PER2 (Fig. 4C). As expected, slow-migrating forms of PER2 were detected when PER2 was coexpressed with Myc-CK1ε, indicative of phosphorylation (fig. S8B). In agreement with the previous experiments (Fig. 4, A and B), endogenous MDM2, but not endogenous β-TrCP1, bound to PER2 regardless of CK1ε expression (Fig. 4C). As expected, β-TrCP1 only bound to PER2 when CK1ε was expressed and the proteasome was inhibited, whereas its binding was compromised when Ser480 of PER2 was mutated (Fig. 4C). The fact that both MDM2 and β-TrCP1 immunoprecipitated with FLAG-PER2 in this assay (Fig. 4C) indicates that PER2 in cells may be associated with one or the other E3 ligase but does not imply the existence of a trimeric complex that includes PER2 and both E3 ligases. Our data also show that MDM2 bound PER2(S662A) and PER2(S662D) to approximately the same extent, confirming that PER2 priming is dispensable for recognition by MDM2 (Fig. 4C). As is the case for β-TrCP1, mutation of Ser662 favors a greater amount of MDM2 protein associated with PER2 (Fig. 4C). In summary, our results point to a model in which recognition of PER2 by MDM2 is indifferent to phosphorylation in general, as shown using λPPase, and to phosphorylation by CK1ε in particular. Neither Ser480 nor Ser662 plays a direct role in the recognition of PER2 by MDM2. Because CK1ε/δ binding to PER2 affects PER2’s stability [Fig. 4 and (15)], other roles for CKε/δ may exist and influence the formation of the PER2:MDM2 complex.

To test whether CK1ε-mediated phosphorylation was a prerequisite for MDM2-mediated PER2 ubiquitylation, we performed a two-step in vitro ubiquitylation reaction in which Myc-PER2(356–872) was first incubated with FLAG-CK1ε to allow for phosphorylation and then purified before the ubiquitylation reaction was carried out with FLAG-MDM2. Ubiquitylation of Myc-PER2(356–872) by MDM2 was abrogated neither by CK1ε-mediated phosphorylation nor by CK1ε binding to PER2 (fig. S8C). To rule out the contribution of phosphorylation events other than those mediated by CK1ε to MDM2-mediated ubiquitylation of PER2, we performed in vitro ubiquitylation reactions following treatment of recombinant PER2 fragment proteins with λPPase (Fig. 4D). As shown for PER2(356–872), none of the treatments caused a discernible effect in the ubiquitylation activity of MDM2 toward its substrate (Fig. 4D). Thus, phosphorylation and MDM2-mediated ubiquitylation of PER2 seem to follow parallel posttranslational paths during PER2’s accumulation in the nucleus.

MDM2 targets PER2 for degradation in cells

Because polyubiquitylation of PER2 is likely a signal for proteasomal degradation, we asked whether MDM2-mediated activity toward PER2 affects PER2’s half-life. HCT116p53+/+ cells were transfected with FLAG-MDM2 and harvested at different times after being treated with cycloheximide (CHX), an inhibitor of protein biosynthesis commonly used to estimate the half-life of other core clock proteins (29, 51). Analysis of cell lysates showed that the amount of endogenous PER2 greatly decreased shortly after CHX addition in samples overexpressing MDM2 compared to control cells not overexpressing MDM2, shortening PER2’s half-life to about twofold (Fig. 5A). We speculated that the converse, a decrease in endogenous MDM2 abundance, would favor PER2 stability and prolong its half-life, mirroring the effect of β-TrCP1 down-regulation on PER2 abundance. To test this hypothesis, we transfected HCT116p53+/+ cells, which produce both MDM2 and β-TrCP1 endogenously, with small interfering RNA (siRNA) targeting one or the other E3 ligase. Lysates were collected at various times after CHX addition and analyzed to show endogenous amounts of PER2, MDM2, and β-TrCP1 by immunoblotting (Fig. 5B). The results unmasked two distinct, yet related, features associated with PER2 abundance. Knocking down MDM2 had a greater influence than knocking down β-TrCP1 on the accumulation of PER2 in untreated cells (fig. S9), but knockdown of either E3 ligase affected the degradation of PER2 similarly following the inhibition of translation (Fig. 5B, lanes 7 to 12 versus 13 to 18). First, overall endogenous amounts of PER2 increased after knocking down either E3 ligase before CHX addition (Fig. 5B and fig. S9). The amount of PER2 increased by about threefold in MDM2 knockdown cells and by about twofold in β-TrCP1 knockdown cells compared to cells treated with a scrambled siRNA (fig. S9). Second, depletion of either MDM2 or β-TrCP1 stabilized PER2 to a similar extent compared to cells treated with the scrambled siRNA (Fig. 5B). Overall, our results emphasize the existence of an alternative mode of regulation of PER2 stability. These results imply that although both MDM2 and β-TrCP1 promote PER2 degradation, MDM2 exerts a greater influence on PER2 stability than does β-TrCP1.

Fig. 5 MDM2 targets PER2 for degradation.

(A) p53+/+ HCT116 cells were transfected with either empty vector or FLAG-MDM2 and allowed to express the constructs for 24 hours before adding CHX (t = 0 hours). Cells were harvested at different times after CHX addition, and extracts were immunoblotted for endogenous PER2 and MDM2, and FLAG-MDM2 using antibodies recognizing PER2, MDM2, and FLAG. Tubulin was used as a loading control. The abundance of PER2 was quantified using ImageJ software v1.45, and values were normalized to tubulin. Immunostaining intensity was plotted as the mean ± SD from three independent experiments. The curve was fitted using Microsoft Excel. n = 3 independent experiments. a.u., arbitrary units. (B) HCT116 cells were transfected with siRNAs targeting MDM2 (siMDM2), β-TrCP1 (siβ-TrCP1), or a scrambled siRNA for 48 hours before CHX addition (t = 0). Cells were harvested at different times after CHX treatment, and extracts were immunoblotted for endogenous PER2 (short and long exposures of blot are shown), MDM2, and β-TrCP1. Tubulin was used as a loading control. n = 3 independent experiments. (C) Real-time bioluminescence recordings were carried out in circadian-synchronized MEFmPer2::LUC cells maintained for 24 hours (t1, cells in the rising phase of the clock) or 33 hours (t2, cells in the falling phase of the clock) in luciferin-containing medium before adding CHX plus vehicle [dimethyl sulfoxide (DMSO)], sempervirine nitrate (SN), PF-670462 (PF670), or both SN and PF-670462. Treatments were performed in triplicate, and recordings were continued for ~30 hours after the addition of inhibitors. Data were normalized to PER2::LUC abundance immediately before drug addition. Mean PER2 half-life is shown ± SD. Significance was determined by Student’s t test between the indicated pairs of groups (***P < 0.0005, **P < 0.005, and *P < 0.05; n.s., not significant). Results were calculated from three biological replicates conducted in triplicate. (D) Summary of PER2 protein half-life values obtained under various conditions and their statistical significance as determined by Student’s t test as described in (C).

MDM2 cooperates with β-TrCP to promote PER2 degradation during the circadian cycle

We next asked whether the activity of MDM2 is relevant to PER2 stability during each rising and falling phase of the circadian cycle. We measured real-time circadian rhythms in mouse embryonic fibroblast (MEF) cells that exhibit circadian-dependent bioluminescence due to the endogenous Per2 coding region being replaced with sequences encoding a Per2-luciferase fusion protein (mPer3::LUC) by knock-in (MEFmPer2::LUC) (52). MEFmPer2::LUC cells maintain robust rhythms in luciferase activity for several days, and the mPer2::LUC fusion protein exhibited rhythms of accumulation and posttranslational modifications that mirror those described in vivo (10, 52). We monitored degradation of mPER2::LUC by bioluminescence recordings in cells treated with CHX and sempervirine (SN), a compound that specifically inhibits the ubiquitin ligase activity of MDM2 (53, 54) during either rising (t1) or falling (t2) phases of the circadian cycle (Fig. 5, C and D, and fig. S10, A and B). Inhibition of MDM2 activity with SN increased the half-life of PER2 compared to vehicle-treated cells in both rising and falling circadian phases [t1(DMSO versus SN, in hours): 2.47 ± 0.211 versus 4.20 ± 0.209 and t2(DMSO versus SN, in hours): 2.25 ± 0.080 versus 3.39 ± 0.209], suggesting a role for the E3 ligase in modulating PER2 stability (Fig. 5, C and D). As expected, a PER2’s half-life also increased when cells were treated with the specific CK1ε/δ inhibitor PF-670462, which prevents phosphorylation of PER2 at Ser480 and subsequent recognition of PER2 by β-TrCP1 [t1(DMSO versus PF-670462, in hours): 2.47 ± 0.211 versus 3.47 ± 0.177 and t2(DMSO versus PF-670462, in hours): 2.25 ± 0.080 versus 3.14 ± 0.209 and (Fig. 5, C and D) (21)]. These results establish that MDM2 negatively regulates PER2 stability during the accumulation and degradation phases of the circadian cycle.

We next evaluated how SN and PF-670462 treatments compared to each other in terms of mPER2::LUC stability in both rising and falling phases (Fig. 5C). Results showed a marginal, but consistent, significant increase in mPER2::LUC half-life in SN-treated cells as compared to PF-670462–treated cells only during the rising phase [t1(PF-670462 versus SN, in hours): 3.47 ± 0.177 versus 4.20 ± 0.209 and t2(PF-670462 versus SN, in hours): 3.14 ± 0.209 versus 3.39 ± 0.209] (Fig. 5, C and D). These results suggest that MDM2 and β-TrCP are both needed during the circadian accumulation phase of PER2 but might have redundant roles during the falling phase during which no statistically significant difference in PER2 half-life was observed between treatments. Consequently, it seemed relevant to explore the contribution of CK1ε/δ for PER2 accumulation in the context of MDM2 activity.

We incubated MEFmPer2::LUC cells with a combination of SN and PF-670462 inhibitors in the presence of CHX. The simultaneous application of both inhibitors had a synergistic effect on the stability of mPER2::LUC compared to the addition of PF-670462 alone, but not to SN alone, in either phase [t1(PF-670462 versus PF-670462 + SN, in hours): 3.47 ± 0.177 versus 4.45 ± 0.254 and t2(PF-670462 versus PF-670462 + SN, in hours): 3.14 ± 0.209 versus 3.86 ± 0.209] (Fig. 5, C and D). Overall, our data support a model in which PER2 stability depends, a priori, on the interplay between both E3 ligases.

MDM2’s E3 ligase activity is required to maintain circadian period

Maintenance of circadian oscillations relies on the production of the rate-limiting component PER2 for the formation of a functional PER2:CRY inhibitory complex (10). Therefore, we hypothesized that alterations in PER2 stability that result from tuning MDM2’s abundance or activity should affect the length of the circadian period. Our initial studies focused on measuring the period length of the circadian oscillator in MEFmPer2::LUC cells in which the amount of MDM2 in cells was either augmented by its overexpression or silenced by siRNA targeting (Fig. 6, A and B, and fig. S11, A and D). As expected from our biochemical findings (Fig. 5), synchronized MEFmPer2::LUC cells overexpressing MDM2 exhibited a shorter period length (25.20 ± 0.100 hours versus 24.53 ± 0.050 hours) compared to empty vector-transfected cells (Fig. 6A and fig. S11, A and B). Furthermore, transfecting MEFmPer2::LUC with increasing amounts of Myc-MDM2 plasmid resulted in a dose-dependent shortening of circadian period length by up to ~1.5 hours even at low plasmid concentrations (fig. S11C), suggesting that a tight regulation of MDM2 needs to be maintained under physiological conditions to ensure proper oscillation. Next, we challenged the model by hypothesizing that siRNA-mediated knockdown of MDM2 in MEFmPer2::LUC cells should result in the converse phenotype, a lengthened period (Fig. 6B and fig. S11D). Our data showed that knocking down MDM2 resulted in significant lengthening of the circadian period (25.50 ± 0.141 hours versus 26.75 ± 0.480 hours; fig. S11B), confirming the requirement of MDM2 for normal circadian oscillations.

Fig. 6 The activity of MDM2 influences the length of the circadian period.

(A) MEFmPer2::LUC cells were transfected with either pCS2+ myc (empty vector) or pCS2+ myc-MDM2 and allowed to express the constructs for 24 hours before cells were circadian synchronized. The abundance of PER2::LUC was monitored by bioluminescence for 7 days. (B) MEFmPer2::LUC cells were transfected with scrambled siRNA or an siRNA targeting MDM2 (siMDM2) before synchronization and bioluminescence monitoring. (C) MEFmPer2::LUC cells were treated with DMSO (vehicle) SN before synchronization and bioluminescence monitoring. (D) MEFmPer2::LUC cells were synchronized and incubated with DMSO (vehicle), HLI373, PF-670462 (PF670), or HLI373 plus PF-670462 before bioluminescence monitoring. Cells were maintained with the inhibitors at all times during data collection. For all panels, n = 3 independent biological replicates repeated in triplicate. The bar graphs indicate the length of the circadian period calculated using the LumiCycle analysis software (Actimetrics). Values represent the mean ± SD from three independent experiments. Significance was determined by Student’s t test between the indicated pairs of groups (***P < 0.0005, **P < 0.005, and *P < 0.05).

We then treated MEFmPer2::LUC cells with the cell-permeable MDM2 inhibitor SN or the inhibitor of p53 degradation HLI373, which inhibits ubiquitylation of p53 by the E3 ligase Hdm2, at a dose that (i) prevented MDM2 autoubiquitylation and degradation (5456) and (ii) did not affect cell viability (figs. S11, E and F). Synchronized MEFmPer2::LUC cells were maintained in the presence of the inhibitor throughout the time course during bioluminescence recording (Fig. 6C and fig. S11, B and G). Average bioluminescence rhythms of SN-treated cells showed a significant lengthening of the circadian period of ~2 hours compared to controls (25.35 ± 0.311 hours versus 27.33 ± 0.340 hours; Fig. 6C and fig. S11B), which closely resembled the result in siMDM2-treated cells (Fig. 6B and fig. S11B), suggesting that control over MDM2’s activity is a major point of circadian regulation. Similarly, we also observed an increase in period length when MEFmPer2::LUC cells were treated with HLI373 (vehicle: 24.97 ± 0.208 hours, HLI373: 26.10 ± 0.283 hours; fig. S11G). Therefore, despite the fact that changes in MDM2 abundance influence circadian oscillations, MDM2’s E3 ligase activity is actually the chief contributor to the observed phenotype.

Last, we evaluated whether the combined effect of PF-670462 and HLI373 on PER2 stability results in a synergistic change in circadian lengthening (Fig. 6D). We maintained synchronized MEFmPer2::LUC cells with either each inhibitor alone (PF-670462 or HLI373) or a combination of both (PF670 + HLI373) and recorded the long-term effect on bioluminescence rhythms for 5 days. In agreement with previous results [Fig. 5C and (15)] for PF-670462, treatment of MEFmPer2::LUC cells with HLI373 resulted in a significant increase in circadian period length even at low concentrations (vehicle: 25.47 ± 0.208 hours; HLI373: 26.10 ± 0.200 hours; PF670: 27.53 ± 0.404 hours; PF670 + HLI373: 29.50 ± 0.173 hours; fig. S11B). Similarly, treatment of MEFmPer2::LUC cells with SN alone or in combination with PF-670462 also caused period lengthening, further supporting the specific involvement of MDM2 activity in circadian oscillation. Our findings showed that, although statistically significant, the effect of both inhibitors is not additive but synergistically affects the cell’s period length. Overall, our data suggest a model in which both events, ubiquitylation of PER2 by MDM2 and PER2 phosphorylation by CK1ε/δ, are relevant to determining the circadian period length despite the appearance that both events seem, a priori, to take place independently of each other.


Timely degradation of regulatory proteins is essential for most aspects of cellular homeostasis and is relevant to signaling processes involved in cell growth, proliferation, and survival. It is, therefore, not surprising that malfunctioning of any aspect of the protein degradation process results in a wide spectrum of diseases and disorders [for review, see (57)]. Because mammalian circadian rhythm also relies on a continuous cycle of protein synthesis and degradation, it is not exempted from problems associated with protein turnover dysregulation. Mice bearing loss-of-function mutations in or expressing knockdown constructs targeting ubiquitin-modifying enzymes involved in clock regulation (for example, FBXL3, FBXL21, FBW1A, HUWE1, PAM, UBE3A, and SIAH2) exhibit a phenotype wherein the free-running period of locomotor activity is longer, shorter, or dampened [for review, see (58)]. Because clock components control the expression of an array of genes involved in multiple cellular processes, it is reasonable to expect that alteration in their expression and accumulation is linked to various human diseases (59).

This brings into consideration the relevance of PER2, a clock component that functions at the intersection of the circadian cycle and the cellular response to DNA damage (30), and its turnover depends on its phosphorylation by CK1ε/δ and binding to β-TrCP1/2, followed by ubiquitylation and proteasomal degradation [see (14) and references within]. Whereas substantive research has made a compelling case for how PER2 accumulates and how its abundance modulates the function of the clock, it poses the question of whether PER2 turnover relies exclusively on β-TrCP1/2. Although the answer could have certainly been affirmative, several observations suggested to us that alternative mechanisms controlling PER2 stability could exist. Such observations include the counterintuitive finding that PER2’s half-life was reported to be shorter in cultured cells coexpressing dominant negative forms of β-TrCP1 and β-TrCP2 (22), the likelihood for phosphorylation-independent mechanisms of PER2 degradation explaining its three-stage kinetics of degradation (15), and the presence of ubiquitylated forms of PER2 in a biological system in which β-TrCP1 and β-TrCP2 are knocked down (27). As a result, we turned to our findings that established that PER2 is able to form a trimeric complex with MDM2 and p53 (28) and asked whether MDM2 might play a role in PER2 stability.

Our results show that PER2 binds to MDM2 (PER2:MDM2) in vitro in manner that is independent of p53 and the phosphorylation status of PER2 and exists as a readily detectable endogenous complex in cells in various experimental settings (Fig. 1 and fig. S1). This is a nontrivial finding because, to the best our knowledge, all well-established E3 ubiquitin ligases acting on clock components only recognize phosphorylated substrates. This includes, in addition to β-TrCPs, the E3 ligases FBXL3 and FBXL21, which act on CRY1/2 that has been phosphorylated by adenosine monophosphate (AMP)–activated protein kinase (AMPK) (60), FBXW7, which acts on REV-ERBα phosphorylated by cyclin-dependent kinase 1 (61), and a yet uncharacterized E3 ligase that targets BMAL1 phosphorylated by GSK3β (62). The challenge of identifying previously unreported E3 ubiquitin ligases targeting clock components has led to the development of screens that revolve around identifying enzyme-substrate binding or functional interactions (63, 64). Of these, the recent identification of the ubiquitin ligase Siah2, which promotes REV-ERVα turnover, has been the most promising finding, yet it also belongs to the group of ligases that recognize phosphorylated substrates (64).

Binding of PER2 to MDM2 involves a region of MDM2 distinct from those identified for p53 binding and E3 ligase activity (Figs. 1 and 2), a result that is in agreement with the existence of the PER2:MDM2:p53 complex (28). We established PER2 as a previously unknown substrate of MDM2 and, conversely, MDM2 as a previously uncharacterized E3 ligase responsible for PER2 ubiquitylation (Fig. 2). The relevance of these initial findings lies in the existence of an alternative mechanism to recognize and target PER2 for degradation that is independent of phosphorylation (Figs. 2 to 4).

The binding specificity of MDM2 for E2 enzymes of the UbcH5 family defined the intrinsic preference for Lys11, Lys48, and Lys63 ubiquitin linkages in MDM2-catalyzed PER2 ubiquitylation, resulting in the incorporation of ubiquitin molecules at multiple sites on PER2 (Figs. 2 and 3). Docking simulations of Lys11, Lys48, and Lys63 ubiquitin conformations on PER2 favored elongation and, thus, suggest the formation of polyubiquitylinated chains (43). Although the addition of multiple ubiquitin units to PER2 by MDM2 represents a previously unknown finding among circadian proteins, posttranslational modifications of this nature by MDM2 are not uncommon as shown, for example, in the case of p53 and FOXO4 (41, 65).

We found that the accumulation and half-life of endogenous PER2 varied in scenarios in which MDM2 abundance or activity was manipulated, which also altered the cell’s circadian period length (Figs. 5 and 6). These findings raise the question of how MDM2’s role as a PER2 regulator would fit into the functioning of the actual mammalian clock mechanism when acting under normal physiological conditions. This is certainly a difficult question to address, especially considering that MDM2 distribution in normal cells is largely nuclear, that MDM2 could promote either mono- or polyubiquitylation of substrates depending on its abundance, and that rhythmic changes in the abundance of MDM2 protein and transcripts are largely absent in unstressed cells [for review, see (29, 66)]. Whereas these well-established premises create constraints around the possible function of MDM2 within the clock molecular mechanism, we propose a few scenarios for further consideration. For example, it is possible that, under physiological conditions, translocation of PER2 to the nucleus would initially result in time-of-day accumulation of CK1ε/δ-dependent phosphorylation events in PER2 that may serve to prime the substrate first for β-TrCP1/2–mediated degradation and later for MDM2 targeting. Furthermore, it is not uncommon to find that the generation of polyubiquitylation substrates for targeted proteasomal degradation requires both priming of monoubiquitylated substrates and intrinsic E3 ligase activity of more than one enzyme, as has been shown, for example, in the case of p53 (67, 68).

Phosphorylation of PER2 by CK1ε/δ either stabilizes or destabilizes the circadian factor depending on the phosphocluster targeted in PER2 and thus adjusts the length of the circadian period to diverse environmental stimuli such as temperature or metabolic signals (15, 17). Our findings open the possibility of PER2’s stability being modulated by signals that converge on MDM2, for example, those generated in response to genotoxic or cytotoxic cellular stress, and for which a change in period length might provide a fitness advantage. MDM2’s activity can be modulated by posttranslational modifications, stability, localization, or binding and can be exquisitely tuned by, for example, alteration in oxygen availability, exposure to low-dose radiation, or even slight changes in growth factor concentrations (66). Certainly, phase resetting of the mammalian circadian clock has been shown to occur in response to DNA damage and metabolic stress in both cell culture and animal models, a phenotype that is increasingly associated with the existence of cross-talk mechanisms between clock proteins and checkpoint components (30, 6971).

At this point, the role of MDM2:PER2 interaction in the mammalian system and within any of the scenarios described above remains largely within the domain of speculation and represents an area of active research in our laboratory. We expect that mounting biochemical, molecular, and genetic evidence will provide a conceptual framework within which we can understand how cells relate and respond to environmental perturbations, no longer in isolation, but in the context of multicellular systems.


Plasmid constructs

The human PER2 (NM_022817), p53 (NM_000546), MDM2 (NM_002392), β-TrCP1 (NP_003930), and CK1ε (BC006490, Addgene) full-length cDNAs were cloned downstream from a tag-encoding sequence into pCS2+ 3×FLAG- and (Myc)6-tag vectors (FLAG-PER2, FLAG-p53, FLAG-MDM2, FLAG–β-TrCP, Myc-CK1ε, Myc-PER2, Myc-p53, and Myc-Mdm2) modified for ligation-independent cloning (Novagen). Various mutants of MDM2 [MDM2(C470A)] and PER2 [PER2(S662A), PER2(683–872)-WT-KA, PER2(683–872)-KA-WT, PER2(683–872)-KA, PER2(683–872)-K789A, PER2(683–872)-K790A, PER2(683–872)-K793A, PER2(683–872)-K796A, PER2(683–872)-K798A, PER2(683–872)-K800A, PER2(683–872)-K803A, PER2(683–872)-KA, PER2(683–872)-K750A, PER2(683–872)-WT-KA-K750A, PER2(S480A), and PER2(S662D)] and the constitutively active form of CK1ε [residues 1 to 305 (48)] were generated from the FLAG- and Myc-tagged templates, respectively, using QuikChange II site-directed mutagenesis and following the manufacturer’s instructions (Agilent). Deletion constructs of MDM2 [MDM2(1–117), MDM2(117–497), MDM2(1–230), MDM2(230–497), MDM2(1–434), MDM2(434–497), MDM2(1–434), and MDM2(434–497)] and PER2 [PER2(1–682), PER2(356–872), PER2(683–872), and PER2(873–1255)] were obtained by polymerase chain reaction amplification and subcloning in either pCS2+ 3×FLAG- or (Myc)6-tag vectors. Various lengths of cDNA fragments of PER2 were cloned into the Sal I/Not I sites of pGEX-4T-3 and then amplified, and the products were subcloned in pCS2+ FLAG. Therefore, fragments of PER2 comprising residues 1 to 185, 182 to 475, 186 to 404, 575 to 682, 683 to 872, 873 to 1135, 1121 to 1255, and 1 to 1255 are referred to in the text as FLAG-GST-PER2(1–185), FLAG-GST-PER2(182–475), FLAG-GST-PER2(186–404), FLAG-GST-PER2(575–682), FLAG-GST-PER2(683–872), FLAG-GST-PER2(873–1135), FLAG-GST-PER2(1121–1255), and FLAG-GST-PER2(1–1255), respectively.

Bacterial two-hybrid screening

The two-hybrid interaction screening was performed using the BacterioMatch II system (primary size: 6.9 × 106; Agilent Technologies) and the pBT (bait)/pTRG (target) interacting system following the manufacturer’s instructions. Briefly, detection of protein-protein interactions is based on transcriptional activation of the HIS3 reporter gene, which allows cells to grow in the presence of 3-AT (Sigma). A specific set of PER2 baits (only results using PER2 full length as bait are shown in Fig. 1A, pBT-PER2, 50 ng) and target plasmid pair from a commercially available human liver library (pTRG cDNA library, Agilent Technologies) were cotransformed using BacterioMatch II validation reporter competent cells (XL1-Blue MRF’ Kan strain, 100 μl). These cells were grown in SOC medium (Thermo Fisher Scientific) and plated in nonselective M9+ His-dropout broth [1× M9 medium (Sigma), chloramphenicol (25 μg/ml), and tetracycline (12.5 μg/ml)] at 37°C for 24 hours. Selection was performed by replicate plating of cotransformants in selective screening medium containing 3-AT [1× M9 medium (Sigma), chloramphenicol (25 μg/ml), tetracycline (12.5 μg/ml), and 5 mM 3-AT]. Per the manufacturer’s instructions, colonies were grown in nonselective screening medium, not in selective screening medium (5 mM 3-AT). Cotransformation was achieved, but the protein pair did not demonstrate a detectable interaction using the two-hybrid assay. Putative positive clones were then patched onto BL-tetracycline/chloramphenicol agar plates.

Activation of a second reporter cassette, addA (provides resistance to streptomycin), was used to verify protein interaction between the bait and the target protein and to eliminate weak and possibly false-positive interactions. Accordingly, putative positive colonies were patched from a selective screening medium (5 mM 3-AT) onto a dual selective screening medium [5 mM 3-AT and streptomycin (12.5 μg/ml)]. The pBT-LFG2/pTRG-Gal11P cotransformant was used as a positive control, whereas cotransformation of pBT-PER2 with either empty pTRG or pTRG-Gal11P vectors as well as the pBT/pTRG cotransformation pair of empty vectors were used as negative controls. Isolated positive cotransformants were grown in LB broth supplemented with tetracycline and chloramphenicol, and cDNA clones were purified and sequenced. Some of the positive interactors were already reported, and their interaction was functionally verified (28).

Cell culture, transient transfections, and treatments

Human colorectal carcinoma HCT116 [TP53(+/+), PER2(+/+)] and human non–small cell lung carcinoma H1299 cell lines were purchased from the American Type Culture Collection and propagated according to the manufacturer’s recommendations. The H1299 cells contained a homozygous partial deletion of the TP53 gene that results in the absence of p53 expression. The HCT116 null-isogenic clone [TP53(−/−), PER2(+/+)] was purchased from Genetic Resources Core Facility Biorepository and Cell Center (Johns Hopkins School of Medicine) and maintained in McCoy’s 5A modified medium containing 10% fetal bovine serum (FBS), penicillin (50 U/ml), and streptomycin (50 μg/ml). The MEFmPer2::LUC cells (gift of S. Kojima, Virginia Tech) were cultured in Dulbecco’s modified Eagle’s medium (DMEM; glucose, 4.5 g/liter) supplemented with 10% FBS, penicillin (50 U/ml), and streptomycin (50 μg/ml) and maintained at 37°C and 5% CO2.

Plasmid transfections were performed at 50 to 80% cell confluency and optimized using Lipofectamine LTX (Invitrogen) and HyClone HyQ-RS reduced serum medium (GE Healthcare) following the manufacturer’s instructions. Proteins were allowed to express for several hours before cells were either harvested or circadian synchronized. Synchronization was by serum shock (72) or dexamethasone treatment (73). Lysates were from cells collected at the indicated times, with t = 0 occurring just before CHX (100 μg/ml) addition.

For siRNA transfections, HCT116 p53+/+ cells were grown in McCoy’s 5A medium containing 10% FBS, penicillin (50 U/ml), and streptomycin (50 μg/ml) until reaching 60 to 80% confluency. Knockdown was optimized using DharmaFECT 2 reagent (GE Dharmacon) to deliver siRNAs targeting either MDM2 (5′-GAGATTTGTTTGGCGTGCCAAGCTT-3′) or β-TrCP1 (5′-CGGAAACTCTCAGCAAGCTATGAAA-3′) following the manufacturer’s instructions. A scramble siRNA sequence with no homology to any known mammalian gene served as control. Forty-eight hours after transfection, cells were serum shocked for 2 hours, after which the medium was replaced with a serum-free version and CHX was added. Samples were collected at different times after treatment, and extracts were prepared in NP-40 lysis buffer containing 10 mM tris-HCl (pH 7.5), 137 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% NP-40, 80 mM β-glycerophosphate, 1 mM Na3VO4, 10 mM NaF, and protease inhibitors (10 μM leupeptin, 1 μM aprotinin A, and 0.4 μM pepstatin).

Last, endogenous amounts of PER2 were monitored in HCT116 cells treated with CHX and incubated with SN (1 μg/ml, ChromaDex Inc.), PF-670462 (0.1 or 1 μM, Cayman Chemical Co.), or a combination of both inhibitors throughout the time course analyzed. The vehicle (DMSO) was used as control.

Immunoprecipitation and immunoblot assays

Immunoprecipitation of protein complexes was performed using either extracts from transfected cells or in vitro binding reactions. Unless indicated, proteins were in NP-40 lysis buffer, and extracts (0.5 to 1 mg) were incubated by rotation with either M2 agarose beads conjugated to an antibody that recognizes FLAG (Sigma-Aldrich) or Myc (9E10) (Santa Cruz Biotechnology) for either 2 hours or overnight at 4°C, respectively. In other cases, immunoprecipitations were carried out in a two-step procedure, with extracts first being incubated with an uncoupled antibody recognizing FLAG, Myc, or PER2 overnight at 4°C before the addition of protein A beads (50% slurry; Sigma-Aldrich). Bound beads were washed four times with wash buffer [20 mM tris-HCl (pH 7.5), 100 mM NaCl, 5 mM EDTA, 0.1% Triton X-100, and 0.5 mM phenylmethylsulfonyl fluoride] before the addition of Laemmli buffer. Complexes were resolved by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting using the specific antibodies indicated in each case. Primary antibodies recognized FLAG (Sigma-Aldrich), Myc (Santa Cruz Biotechnology), PER2 (Sigma-Aldrich), MDM2 (Santa Cruz Biotechnology), β-TrCP1 (Cell Signaling Technology), p53 (DO1 clone, Santa Cruz Biotechnology), or ubiquitin (Enzo Life Sciences). Secondary antibodies were horseradish peroxidase–conjugated α-rabbit or α-mouse IgGs (Invitrogen), and chemiluminescence reactions were performed using the SuperSignal West Pico Substrate (Pierce).

In vitro binding and epitope blocking assays

In vitro transcription and translation of either pCS2+ Myc-MDM2 or pCS2+-FLAG-PER2, β-TrCP1MDM2, MDM2(C470A), and p53 were carried out using the SP6 high-yield TNT system (Promega) following the manufacturer’s instructions. As indicated in each case, aliquots (1 to 4 μl) of the indicated recombinant proteins were preincubated for 15 min at room temperature to allow complex formation before adding NP-40 lysis buffer. Epitope blocking was performed by preincubating in vitro the transcribed and translated FLAG-MDM2(C470A) with α-4B11, α-4B2, or α-SMP14 antibody (0.1 mg/ml each; Calbiochem) for 2 hours at 4°C before adding recombinant Myc-PER2. Binding reactions were allowed to proceed overnight at 4°C with rotation. In other experiments, binding of Myc-MDM2 or Myc-MDM2(C470A) proteins to FLAG-PER2 was evaluated in λPPase or CK1ε-treated samples as described in the section below. Reactions were diluted in NP-40 lysis buffer, and complexes were immunoprecipitated using antibodies recognizing the FLAG epitope (Sigma) and protein A beads (50% slurry) as described earlier.

Ubiquitylation and degradation assays

Aliquots (1 to 4 μl) of in vitro transcribed and translated tagged proteins {FLAG-, Myc-, or Myc-DO [DO epitope sequence is EPPLSQETFSDLWKL, recognized by α-p53(DO-1) monoclonal antibody; Santa Cruz Biotechnology]}, or a combination of them, were allowed to bind before adding 1× ubiquitylation buffer (Enzo Life Sciences), 1 mM dithiothreitol (DTT), ubiquitin-aldehyde (20 μg/ml), ubiquitin (600 μg/ml), 1× ATP-energy regeneration system (5 mM ATP/Mg2+; Enzo Life Sciences), 40 μM MG132 (Cayman Chemical Co.), and HeLa S100 lysate fraction (1 mg/ml) (Enzo Life Sciences) or 1× E1 (ubiquitin-activating enzyme, Enzo Life Sciences) and 1× UbcH5a, UbcH5b, or UbcH5c (human ubiquitin-conjugating enzyme, Enzo Life Sciences) to a final volume of 10 to 15 μl. Reactions were then incubated for 1 hour at 37°C in a water bath, except for fig. S3B, where the incubation time varied as indicated in the figure. In other experiments, tagged PER2 recombinant proteins were pretreated with λPPase [200 U (New England Biolabs), 15 min at 25°C] before the ubiquitylation reaction was carried out in the presence of tagged MDM2 as aforementioned. Following, Laemmli sample buffer was added and ubiquitylated proteins were either resolved by SDS-PAGE and detected by immunoblotting or immunoprecipitated following the two-step protocol described in the section above.

Detection of ubiquitylated forms of PER2 in cells was carried out by cotransfecting HCT116 [TP53(+/+), PER2(+/+)] cells with pCS2+ FLAG-PER2 and either pCS2+ Myc-MDM2 or pCS2+ Myc-MDM2(C470A) plasmids. Cells were maintained in complete medium for 24 hours to allow for the recombinant proteins’ expression before adding MG132 (50 μM) and ubiquitin aldehyde (5 nM). Cells were harvested 4 hours later, and lysates were immunoprecipitated using α-FLAG antibody as described. Proteins were resolved by SDS-PAGE, and ubiquitylated forms of PER2 were detected by immunoblotting using an α-ubiquitin antibody.

Kinase and phosphatase assays

Complexes of PER2 and wild-type or constitutively active CK1ε were allowed to form for 30 min at 25°C before adding kinase buffer [30 mM Hepes (pH 7.5), 7 mM MgCl2, 0.5 mM DTT, 100 μM ATP, and protease inhibitor cocktail] in a final volume of 20 μl. Reactions were maintained for 30 min at 37°C and terminated by the addition of Laemmli buffer. When indicated, λPPase (New England Biolabs) was added directly to the substrate (PER2) and reactions were allowed to proceed for 30 min at 30°C before they were terminated. Depending on the experiment, proteins were resolved on 4% (for PER2) and 8% [for CK1ε and CK1ε(1−305)] SDS-PAGE or Phos-tag (WACO Chemicals, USA) gels and visualized by immunoblotting.

Homology model generation and protein-protein docking

The I-TASSER server (74) was used to create homology models of PER2(683–872) wild-type and mutant variants. Sequences were uploaded to the server in FASTA format. There were no restraints guiding modeling, homologous templates were not excluded, and secondary structures for specific residues were unbiased. The server uses templates from the Protein Data Bank (PDB) database to predict secondary structure of the query protein using LOMETS 4 (local meta-threading-server). Alternatively, the server uses ab initio modeling to assign secondary structures. Clustering was then performed to find the lowest free-energy model using SPICKER (75). The model with the highest C score value was then energy minimized using the Molecular Operating Environment with Amber12EHT parameters and subsequently assessed for the quality of the model using online servers including SWISS-MODEL (76, 77), ProSA (78), and VERIFY3D (79, 80). Models of all three constructs showed favorable energies relevant to their ANOLEA, PROCHECK, and z scores, as well as favorable three-dimensional structure and side-chain placements, and were deemed acceptable (fig. S5). Here, the three models were validated and used in confidence in further protein-ubiquitin docking experiments.

Protein binding interfaces were predicted by docking between ubiquitin [PDB: 1UBQ (81)] and each model of the PER2 fragments. The Schrödinger software suite (2017.2) and the BioLuminate interface, which uses the PIPER docking module, were used for interface determination. No biased or interfaced residue was set at the onset of docking. All PER2 models were treated equally in regard to how and where ubiquitin molecules were predicted to interact. Thirty structures for each PER2:ubiquitin docking pair were obtained and clustered using pairwise root mean square deviation, and key residues located at the interface were identified. Data files are available from the Virginia Tech Institutional Data Repository, VTechData, doi:10.7294/W4JW8C2R.

Analysis of protein half-life

Accumulation and half-life of endogenous proteins in HCT116 cell extracts (20 to 80 μg) was monitored by immunoblotting samples collected at different times after CHX addition, as indicated in the figure legends. Protein bands were quantified by immunoblot analysis using Bio-Rad ImageLab 5.1 software/Gel Doc XR+ system, and values were normalized to tubulin. Unless indicated, the percentage of the remaining protein was normalized to t = 0 and the data were fitted using Microsoft Excel.

In other experiments, the half-life of PER2 was measured in MEFmPer2::LUC cells by luminescence recording. Seeded cells were synchronized with dexamethasone (100 nM; 2 hours) and maintained in medium containing phenol red–free DMEM, 50 μM luciferin (Biosynth), 2% FBS, 1% penicillin/streptomycin, and 1% l-glutamine (Invitrogen) in a LumiCycle instrument (t = 0 hours). Addition of CHX (40 μg/ml) and DMSO (1% v/v), PF-670462 (1 μM), SN (1 μg/ml), or both inhibitors occurred during rising (t = 24 hours) or falling (t = 33 hours) phases. Three biological experiments were performed in parallel, with each treatment being plated in triplicate. Data were normalized to the PER2::LUC signal from untreated cells. The nonexponential decay of PER2 impedes the use of the traditional approach for calculating protein half-life, which is based on fitting the data to an exponential function; instead, PER2 half-life was determined at the time in which PER2::LUC signal was 50% of the initially detected amount following the Zhou et al. (15) method.

Real-time bioluminescence assays

MEFmPer2::LUC cells were seeded in 35-mm dishes and circadian synchronized by dexamethasone treatment (100 nM; 2 hours). Following medium replacement as described above, cells were allowed to stabilize in a LumiCycle 32-channel automated luminometer (Actimetrics) placed in a 37°C incubator for 24 hours before SN (1 μg/ml), PF-670462 (0.1 or 1 μM), HLI373 (5 μM), or a combination of inhibitors was added. In these assays, bioluminescence was continuously recorded for at least five additional days and data were analyzed using the LumiCycle analysis software (Actimetrics).

In other experiments, MEFmPer2::LUC cells were transiently transfected with either pCS2+ Myc-MDM2 or siRNA MDM2 for 24 or 48 hours, respectively, before dexamethasone synchronization. Following medium exchange, bioluminescence was recorded for at least five additional days. In each case, raw data were collected after dexamethasone synchronization (t = 0) and for the remainder of the experiment. Raw data beginning t = 24 hours after synchronization were considered when calculating the circadian period length. Period length was calculated using the LumiCycle data analysis software (Actimetrics). For all experiments, mean and errors were calculated on the basis of at least triplicates.


Fig. S1. In vitro binding studies with MDM2.

Fig. S2. MDM2 ubiquitylates PER2 in vitro and in cells.

Fig. S3. PER2 ubiquitylation is tightly regulated by MDM2.

Fig. S4. Controls for PER2 ubiquitylation assays.

Fig. S5. Identification of putative ubiquitylation sites within the PER2 central domain.

Fig. S6. Quality assurance metrics for the wild-type PER2(683–872) homology model.

Fig. S7. Detection of PER2 phosphorylation by CK1ε.

Fig. S8. Ubiquitylation of the central domain of PER2 is independent of phosphorylation.

Fig. S9. Quantification of endogenous PER2, MDM2, and β-TrCP1.

Fig. S10. Degradation curves for PER2 in response to inhibitor treatment.

Fig. S11. The abundance and activity of MDM2 influence circadian period length.


Acknowledgments: This article is dedicated by C.V.F. to the memory of Dr. James Maller, professor and Howard Hughes investigator, a pioneer in the field of cell cycle regulation, and a fine mentor. We thank D. G. S. Capelluto and J. Tyson for critical reading of the manuscript and all members of the Finkielstein laboratory for help and discussions. We would also like to thank S. Kojima for reagents and advice and J. Webster for comments and manuscript editing. Funding: This work was supported by the National Science Foundation MCB division (MCB-1517298) and the Fralin Life Science Institute to C.V.F. and the National Research Foundation of Korea (N01160447) to J.K.K. Author contributions: J.L. performed all experiments and statistical analyses discussed in this article except those specifically mentioned below. X.Z. carried out the experiments shown in Figs. 1C, 3B, 4, 5 (C and D), and 6D and figs. S3B, S5D, S10, and S11 (B, E, and G). T.G. contributed with Figs. 2 and 3A and figs. S2C, S3 (A and C), S5 (A and C), and S8C. E.L.W. contributed with figs. S4, S7, and S8B and L.J. with fig. S11F. A.M.B. performed and analyzed the modeling shown in figs. S5C and S6. J.L., X.Z., and J.K.K. analyzed the overall data and contributed to refining the hypothesis. C.V.F. and J.L. conceived this project. C.V.F. supervised and coordinated all investigators for the project and wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article