Circadian Clock Control by SUMOylation of BMAL1

See allHide authors and affiliations

Science  26 Aug 2005:
Vol. 309, Issue 5739, pp. 1390-1394
DOI: 10.1126/science.1110689


The molecular machinery that governs circadian rhythmicity is based on clock proteins organized in regulatory feedback loops. Although posttranslational modification of clock proteins is likely to finely control their circadian functions, only limited information is available to date. Here, we show that BMAL1, an essential transcription factor component of the clock mechanism, is SUMOylated on a highly conserved lysine residue (Lys259) in vivo. BMAL1 shows a circadian pattern of SUMOylation that parallels its activation in the mouse liver. SUMOylation of BMAL1 requires and is induced by CLOCK, the heterodimerization partner of BMAL1. Ectopic expression of a SUMO-deficient BMAL1 demonstrates that SUMOylation plays an important role in BMAL1 circadian expression and clock rhythmicity. This reveals an additional level of regulation within the core mechanism of the circadian clock.

SUMOylation—the covalent linking of small ubiquitin-related modifier protein (SUMO) to lysine residues—is a reversible posttranslational modification controlled by an enzymatic pathway analogous to the ubiquitin pathway (13). The addition of SUMO on target proteins has been implicated in transcriptional regulation by a number of mechanisms (36). A hallmark of SUMOylation is that, for most substrates, only a small fraction of intracellular substrate molecules are modified at any given time. SUMO modification is rendered reversible by SUMO-specific proteases, such as the yeast ubiquitin-like protease type 1 (ULP1) or the mammalian sentrin-specific protease type 1 (SENP1) (7, 8).

BMAL1 is a member of the basic helix-loop-helix (bHLH)/Per–aryl hydrocarbon receptor nuclear transporter (ARNT)–Sim (PAS) domain family of transcription factors which, together with its heterodimerization partner CLOCK, drives the circadian clock mechanism (911). Between the two PAS domains, which mediate CLOCK:BMAL1 dimerization (12), there is a linker region of undetermined function. Some specific lysine residues in this region are well conserved across species and among different isoforms of the BMAL1 family (Fig. 1A). Because lysines are targeted by multiple posttranslational modifications (13), we hypothesized that these residues could constitute regulatory sites. Computer modeling analysis has revealed that the best match for K223, K229, K259, and K272 (14) in mouse BMAL1 (mBMAL1) corresponds to the SUMOylation consensus motif ψKxE/D (3), where ψ is a hydrophobic residue and x may be any amino acid (Fig. 1A).

Fig. 1.

mBMAL1 is SUMO1 modified in vivo at a highly conserved lysine residue. (A) Schematic representation of mBMAL1 and lysine conservation within the linker region between the PAS domains. Conserved lysine residues are boxed. Potential SUMO consensus motif s are highlighted. dCYCLE is the Drosophila homolog of BMAL1. mARNT is the mouse ARNT, member of the bHLH/PAS family. (B) COS1 cells were transfected with expression vectors for Myc-BMAL1, His-SUMO1, GFP-SUMO1, and His-SUMO1(G97A). Cell extracts were immunoblotted (IB) with an antibody (α) to Myc. (C) (Left) Lysates from cells expressing Myc-BMAL1 and His-SUMO1 were immunoprecipitated with anti-Myc or control immunoglobulin G (IgG) and then revealed by Western analysis with an antibody to SUMO1. (Right) The same membrane from the left panel was stripped and probed with anti-Myc. (D) COS1 cells were transfected with expression vectors for Myc-BMAL1 (0.25 μg), His-SUMO1 (+, 0.5 μg; ++, 1 μg), Flag-UBC9 (1 μg), and Flag-SENP1 (1 μg) and protein extracts were immunoblotted with antibody to Myc. The lower panel shows an anti–α-tubulin immunoblot from the lower part of the same membrane. (E) Immunoblot with an antibody to Myc of protein extracts from COS1 cells expressing BMAL1 wild type or Lys→Arg mutants and His-SUMO1 as indicated. K259 is the major SUMOylation site in BMAL1.

To determine whether BMAL1 could be SUMO modified in vivo, we transiently expressed epitope (Myc)–tagged BMAL1 in COS1 cells (Fig. 1B). When Myc-BMAL1 was coexpressed with His-tagged SUMO1 or a fusion protein containing green fluorescent protein and SUMO (GFP-SUMO1), we observed Myc-BMAL1 species of higher molecular weight, proportional to the size of additional SUMO modification. No modification of Myc-BMAL1 was observed by coexpressing His-SUMO1[G97→A97 (G97A)], a single–amino acid mutant that is unable to be transferred onto target substrates (15). SUMOylation was further confirmed when immunoprecipitated Myc-BMAL1 reacted with antibodies to SUMO (Fig. 1C). BMAL1 modification was enhanced upon coexpression of ubiquitin-like conjugating enzyme 9 (UBC9), a SUMO conjugating type enzyme (E2) ligase that activates the SUMOylation pathway (3). Conversely, coexpression of the SUMO1-specific protease SENP1 (8) reduced BMAL1 SUMOylation (Fig. 1D and fig. S1A). Site-directed mutagenesis of the four consensus lysine residues in either Ala or Arg, alone or in combination, demonstrated that K259 is the major in vivo SUMOylation site (Fig. 1, A and E, and fig. S1B) (16). Importantly, K259 is placed within the larger SUMO consensus motif RxxVKVExK (Fig. 1A) (17). In vitro SUMOylation assays confirmed that BMAL1 is modified by SUMO1 specifically at the K259 residue (fig. S1C). BMAL1 can also be modified by SUMO2 in vivo at the K259 residue, comparable to modification by SUMO1 (fig. S2).

To establish the relevance of BMAL1 SUMOylation in circadian physiology, we analyzed the peripheral clock in the liver (10, 11, 18). Liver tissues from mice entrained at different zeitgeber times (ZT) were collected and protein extracts were prepared in the presence or absence of N-ethylmaleimide (NEM), an inhibitor of SUMO proteases that stabilizes SUMO-modified proteins (19) (Fig. 2). As previously reported (20, 21), BMAL1 shows rhythmicity in protein abundance and phosphorylation, with hyperphosphorylated forms at ZT9 and at ZT15 and lower protein levels at ZT21 (Fig. 2A, top). NEM stabilized a BMAL1 SUMOylated form (about 100 kD), the abundance of which oscillated in a circadian manner, with a peak at ZT9 (Fig. 2A, middle). No NEM-stabilized forms were detected at ZT15 and ZT21 (fig. S3), whereas no SUMOylated BMAL1 was detected in the absence of NEM (Fig. 2A, lanes 5 to 8). As a control, we examined the expression of the Period 1 (PER1) clock protein from the same extracts (Fig. 2A, bottom). We confirmed the identity of the 100-kD NEM-stabilized form by immunoprecipitation using an antibody to mBMAL1 (Fig. 2B). The effect of NEM was verified by analyzing the levels of SUMO-modified RanGAP1 in the same extracts (fig. S4). The 100-kD SUMO-BMAL1 from liver extracts was readily cleaved by the SUMO-specific protease ULP1 (7), in a NEM-sensitive manner (Fig. 2C, lanes 4 and 5). Similar results were obtained when the recombinant ULP1 used here was tested on SUMO-RanGAP1(fig. S5). In vivo SUMOylation was confirmed when the immunoprecipitated 100-kD BMAL1 form reacted with an antibody to SUMO1 (Fig. 2D). Finally, the rhythmic SUMOylation of BMAL1 is not likely due to oscillations in the Sumo genes (fig. S6). These findings indicate that BMAL1 undergoes rhythmic SUMOylation in vivo, with a timing that parallels BMAL1 circadian activation.

Fig. 2.

Circadian SUMOylation of BMAL1 in mouse liver. (A) (Top) Equal protein amounts from mouse liver extracts at the indicated zeitgeber time (ZT) prepared in radioimmunoprecipitation assay (RIPA1) buffer with or without NEM inhibitor were immunoblotted (IB) with an antibody (α) to mBMAL1 (32). The light-dark schedule entraining is indicated by the white and black bars. (Middle) A longer exposure of the above membrane shows the NEM-stabilized form of BMAL1 at about 100 kD (BMAL1-SUMO). (Bottom) Anti-mPER1 immunoblot on the same liver extracts. (B) Immunoprecipitates from NEM-supplemented liver extracts (ZT9) using an antibody to BMAL1 or control IgG, revealing native proteins using an antibody to BMAL1 for Western blot analysis. (C) The immune complexes obtained as in (B) were incubated with 2 units of recombinant ULP1 protease and analyzed by anti-BMAL1 immunoblot. Inhibition by NEM (5 mM) was obtained by preincubation of ULP1 before the cleavage assay. A lower exposure of this immunoblot shows that equivalent amounts of BMAL1 were immunoprecipitated from each sample (fig. S8). (D) Anti-BMAL1 and control immunoprecipitates obtained as in (B) were probed with an antibody to SUMO1 (left). The same membrane was stripped and analyzed by anti-BMAL1 immunoblot (right).

BMAL1 heterodimerizes with the transcription factor CLOCK to induce the transcription of target genes (911, 22). CLOCK also induces BMAL1 phosphorylation (23) by a yet unidentified mechanism. Here, we find that BMAL1 phosphorylation and SUMOylation increase in parallel at ZT9, concomitantly to the induction of Period gene expression (fig. S7). This may indicate that the CLOCK:BMAL1 interaction could trigger phosphorylation and SUMOylation of BMAL1, events possibly coupled to circadian gene activation.

To investigate the role of CLOCK in the SUMOylation of BMAL1, Myc-BMAL1 or the Myc-BMAL1(K259R) mutant was coexpressed with increasing amounts of CLOCK. CLOCK induced BMAL1 SUMOylation in a dose-dependent manner. The Myc-BMAL1(K259R) mutant was refractory to the effect elicited by CLOCK (Fig. 3A and fig. S9).

Fig. 3.

CLOCK induces BMAL1 SUMOylation in vivo. (A) (Top) COS1 cells were transfected with expression vectors for Myc-BMAL1, Myc-BMAL(K259R) mutant, and Flag-mCLOCK, as indicated, and cell extracts were immunoblotted (IB) with an antibody (α) to Myc. (Bottom) Longer exposure of the same membrane. (B) (Top) Anti-mBMAL1 immunoblot of equal amount of total liver extracts derived from wild-type (wt), Cry1–/–/Cry2–/– (Crys–/–), Clock/Clock (Clk/Clk), and Per1–/– mice collected at indicated zeitgeber times (ZT) in RIPA1 buffer. As controls, protein extracts prepared without strong detergent and NEM (lane 1) or without NEM (lane 9). In lane 8, twice (2×) the amount of protein was loaded. (Bottom) The upper membrane was stripped and immunoblotted with an antibody to RanGAP1. Immunoprecipitations with an antibody to mBMAL1 confirm that SUMO-BMAL1 is not detected in liver extracts from Clock/Clock mice (fig. S11). (C) (Top) Western blot analysis with an antibody to BMAL1 of total protein extract prepared with or without NEM from MEFs derived from wild-type, Clk/Clk, and Crys–/– mice. (Bottom) Longer exposure of top panel reveals the SUMOylated BMAL1 form sensitive to NEM. (D) Western analysis of protein extracts from Clock/Clock MEFs with an antibody to Myc. MEFs were transiently transfected with expression vectors for Myc-BMAL1, His-SUMO1, and Flag-CLOCK. NS, nonspecific. The SUMOylated form of BMAL1 is indicated. Transient expression of CLOCK in Clock/Clock MEFs rescues BMAL1 SUMOylation.

The presence of SUMO-BMAL1 in the liver from wild-type mice was compared with that in mice carrying mutations in Clock proteins (Fig. 3B and fig. S10). A number of factors contribute to the autoregulatory loops that constitute the clock mechanism (22, 24). These include the products of the period genes, which are positively regulated by CLOCK:BMAL1 (25), and the Cryptochrome genes Cry1 and Cry2, the products of which act as potent repressors (26). Whereas neither the disruption of the Per1 gene (27) nor the combined mutation of the Cry1 and Cry2 genes (28) (Cry1–/–/Cry2–/–) generated reduction in SUMO-BMAL1 levels, there was a notable effect on BMAL1 SUMOylation in Clock/Clock mice (29). This effect was specific to BMAL1; there were no differences in the levels of SUMO-RanGAP1 in the same extracts (Fig. 3B, bottom). Moreover, no SUMO-BMAL1 was detected in mouse embryo fibroblasts (MEFs) derived from Clock/Clock mice (Fig. 3C), showing that the effect exerted by CLOCK is not restricted to the liver. SUMOylation of BMAL1 was rescued by transient expression of the wild-type Clock gene into MEFs from Clock/Clock mice (Fig. 3D, lane 3), confirming that a functional CLOCK protein is essential for SUMO modification of BMAL1. Interestingly, because the mutant CLOCK protein present in Clock/Clock mice is still able to heterodimerize with BMAL1 and bind DNA (29), SUMOylation may be an event downstream from transcriptional activation.

To determine the function of BMAL1 SUMOylation, we generated a retroviral expression system for Myc-BMAL1 or Myc-BMAL1(K259R) under the control of Bmal1 promoter. After infection, NIH3T3 cells were serum-shocked to analyze the rhythmic oscillation of BMAL1 expression. This approach recapitulates the circadian regulation of the clock in cell culture (30). Whereas wild-type BMAL1 expression peaked at 20 hours after serum shock, with a second peak of lower amplitude at 44 hours, the Myc-BMAL1(K259R) mutant displayed no circadian oscillation (Fig. 4A). These data support the notion that SUMO modification is required for BMAL1 rhythmicity. Whether under the regulation of the Bmal1 promoter or of a heterologous promoter, the Myc-BMAL1(K259R) protein was about twice as abundant as Myc-BMAL1 (fig. S1B), indicating that lack of SUMOylation may affect protein turnover. Protein half-life analysis by cycloheximide treatment experiments revealed that Myc-BMAL1(K259R) shows an average 50% increased stability compared with that of wild-type BMAL1 (Fig. 4B). Unlike BMAL1, the decay curve of BMAL1(K259R) is not exponential, suggesting a complex degradation mechanism for a protein that is not SUMOylated.

Fig. 4.

SUMOylation of BMAL1 controls the molecular clock. (A) Antibody (α) to Myc immunoblot (IB) of whole-cell lysates from NIH3T3 cells infected with retroviral vectors expressing either wild-type (top) or a K259R mutated Myc-mBMAL1 (bottom) under the control of the mBmal1 promoter (32). Equal infection efficiencies were obtained for both viruses, as revealed by quantitative analysis of neomycin-resistant (NEOr) gene expression (16). At time T = 0, serum-starved cells were shifted to medium containing 50% horse serum and collected at the indicated time points after serum shock. NI, not infected cells; NS, nonspecific band. Results were quantified and normalized with the anti–TATA-binding protein (TBP) immunoblot performed on the same membrane. Th e results are representative of three independent experiments, which gave equivalent results. (B) Myc-BMAL1(K259R) shows increased protein stability. NIH3T3 cells were transfected with equal amounts of Flag-CLOCK along with Myc-BMAL1 or Myc-BMAL1(K259R). At 36 hours after transfection, cells were treated with cycloheximide (CHX) at 50 μg/ml. At indicated times, cells were lysed and protein extracts were immunoblotted with antibody to Myc (top) and antibody to TBP (middle) as loading control. NT, not-transfected cells. Analogous results were obtained in COS1 cells (fig. S12). (Bottom) The immunoblots were quantified by densitometric analysis. The graph shows the percentage of protein amount relative to T = 0 (100%). The results are representative of three independent experiments, which gave analogous results. (C) Ectopic expression of BMAL1 (lanes 2 and 3) and BMAL1(K259R) mutant (lanes 8 to 10) in Bmal1–/– MEFs. Expression levels of BMAL1 protein were evaluated by anti-BMAL1 immunoblot at the indicated times after infection with retroviral vectors as in (A). Protein extracts from MEF cells infected with an empty vector (lanes 5 to 7), wild-type MEFs (lane 11), and Bmal1–/– MEFs (lane 1) were also analyzed as control. (D) Lack of SUMOylation alters the serum shock–dependent oscillation of Dbp, an E-box–controlled gene. Wild-type and Bmal1–/– MEFs were subjected to a serum shock 2 days after infection with the indicated retroviral vectors. At the indicated time, cells were harvested, and Dbp expression levels were estimated by quantitative real-time polymerase chain reaction. Values were normalized to the expression of Sumo-3, a nonoscillating gene (fig. S6), and plotted as relative fold of Dbp expression at ZT0 (set as 1) in wild-type MEFs. Mean values of four independent experiments are shown, with relative distance from average never above 5%.

To establish whether BMAL1 SUMOylation is involved in clock function, we undertook rescue experiments using MEFs generated from Bmal1–/– mice (31). These cells showed no BMAL1 expression (Fig. 4C, lane 1), compared with MEFs generated from wild-type mice (lane 11). Bmal1–/– MEFs were readily infected with the retrovirus vectors (Fig. 4A), expressing either BMAL1 (Fig. 4C, lanes 2 to 4) or the BMAL1(K259R) mutant (Fig. 4C, lanes 8 to 10). To study circadian rhythmicity, we scored for the serum shock–induced expression of endogenous dbp, an E-box controlled gene (30). Although wild-type MEFs showed consistent dbp circadian oscillation, this was altered in Bmal1–/– MEFs (Fig. 4D). Infection of Bmal1–/– MEFs with a virus expressing BMAL1 rescued circadian expression of dbp, whereas expression of the BMAL1(K259R) mutant protein generated a shorter period. This was likely due to an increased stability of a BMAL1 protein that cannot be SUMOylated, which in turn could influence the rhythmic expression of other clock proteins.

Our findings provide insight into the mechanisms that control the circadian levels of BMAL1 expression. Several mechanims could be proposed for such differences, including a direct interplay between SUMO modification and the yet unidentified BMAL1 degradation pathway or a SUMO-dependent interaction with partners that control BMAL1 stability (including CLOCK).

SUMOylation of BMAL1 constitutes another level of control within the core circadian clock. Other clock proteins may undergo SUMO-modification in domains distinct from the PAS linker region. Unique elements of the SUMO pathway may be selective for the circadian clock machinery. The recent discovery and characterization of E3 SUMO ligases (3) may provide key tools to address these questions.

Supporting Online Material

Materials and Methods

Figs. S1 to S12

References and Notes

Stay Connected to Science

Navigate This Article