The De-Ubiquitinylating Enzyme, USP2, Is Associated with the Circadian Clockwork and Regulates Its Sensitivity to Light

Previous studies [15] have identified two isoforms of Usp2 that are produced by alternative splicing; a 69-kDa protein (Ubp69 or USP2a) and a 45-kDa protein (Ubp45 or USP2b). Although studies have reported that Usp2 exhibits a highly robust circadian expression pattern in liver and SCN [13], [14], those studies did not directly compare the two transcripts. Our data show that both forms exhibit rhythmic expression profiles at both the mRNA and protein level. However in its temporal expression, Usp2b is the dominant circadian form, exhibiting a peak at ZT8 in liver and ZT12 in SCN and retina (Fig. 1A). Both USP2a and USP2b protein expression changes over time in SCN, but USP2b appears to be more abundant (Fig. 1B). Expression of both USP2a and USP2b is higher at night and early morning (ZT16-ZT0) yet each is detectable in the SCN throughout the day (Fig. 1B).

We hypothesized that USP2 plays a central role in the circadian clock mechanism. To directly test that idea, we targeted the deletion of exons 3 and 4, which encode a portion of the catalytic region of the enzyme common to both isoforms, and inserted a stop codon (Fig. 2A). The deletion was confirmed in stem cells (Fig. 2B) and in germ line mice (Fig. 2C) by southern blotting. Western blotting from USP2^−/− total retina lysate using an antibody generated against the C-terminal region of USP2 showed that USP2a was undetectable. However, we did detect low levels of immunoreactivity at about 45 kDa in USP2^−/− retina (Fig. 2D). This could reflect a cross-reactivity or a modified USP2 protein. We used RT-PCR with Usp2a- and Usp2b-specific primers to determine if any mutant transcripts were produced in retina, liver, and SCN. We failed to detect a transcript using primers for Usp2a, but the Usp2b primers did reveal a low-abundance transcript. This transcript was cloned and cDNA sequencing not only confirmed the effectiveness of our deletion strategy (Fig.2a) but also showed that Exon 1 of the wildtype Usp2b transcript (Ensembl Gene Identifier ENSMUST00000065461) was alternatively spliced in frame to Exon 5 (see Fig. 2A). These data confirm that a transcript lacking the deleted Exons but encoding the C-terminal antibody-binding site is present at low levels in USP2^−/− mice.

To understand whether USP2 is required for normal circadian function, we used cages equipped with running wheels to determine free-running behavior of USP2^+/+ and USP2^−/− mice in constant darkness (DD) following a period of 12-hour light/dark (LD). Free-running periods measured in USP2^−/− mice were indistinguishable from USP2^+/+ mice (Fig. 3A–B). Although period differences were not found, USP2^−/− mice were significantly more active during subjective night when maintained in DD (Fig. 3C).

Ubiquitinylation by the ubiquitin ligase JETLAG in Drosophila plays a role during light resetting [12]. We therefore asked whether mice deficient in de-ubiquitinylation by USP2 might also exhibit altered circadian sensitivity to light. First, USP2^+/+ and USP2^−/− mice housed in 12-hour LD were placed in constant light (LL) of low irradiance (∼5–20 µW⋅cm⁻²) and every 10 days the irradiance was increased (Fig. 4A, arrows). As expected, LL increased the period of the activity rhythm and higher irradiances decreased rhythmic activity in both USP2^+/+ and USP2^−/− animals. However, USP2^−/− mice were strikingly different than wild type. Their period length increased slowly in response the lower irradiance levels, and, more strikingly, they appeared to exhibit a dramatic delay in activity onset at the lowest irradiance tested. This suggested that the lowest irradiance (∼5–20 µW⋅cm⁻²) was sufficient to cause a phase delay but not sufficient to increase the period length as is expected in LL conditions and as is shown by the control animals.

To determine if loss of USP2 indeed altered the minimum irradiance needed to induce a phase delay at the light-dark transition USP2^+/+, USP2^+/− and USP2^−/− mice were entrained to LD conditions and at the beginning of expected light offset (ZT12) exposed to 4-hours of low irradiance light (ZT12–16) followed by 7 days in darkness (DD) (Fig. 4B–C, arrows). They were then re-entrained to the 12-hour LD cycle for a week and the experiment was repeated with light of higher irradiance. USP2^−/− mice displayed enhanced sensitivity to light (Fig. 4C) at low irradiance levels (Log irradiances 2.5 and 3 which correspond to ∼0.05 µW⋅cm⁻² and ∼0.1 µW⋅cm⁻²) and heterozygotes exhibited intermediate phase delays between those of WT and USP2^−/− animals (Fig. 4C). Although this trend was apparent across the full range of light levels tested, statistically significant differences were detected only at the lower two irradiance levels in which wild type mice either failed to exhibit any phase shift (Log irradiance 2.5) or exhibited moderate phase shifts (Log irradiance 3). To rule out an influence of genotypic differences in phase angle of entrainment on measured phase shifts this parameter was determined in mice entrained to an LD cycle as above and then released into DD, and differences were not found. Furthermore, the increased sensitivity to light was specific to the paradigm in which light was extended by 4 hours at the time of light off-set (ZT12). Phase shifting was not significantly different in USP2^+/+ and USP2^−/− mice in response to 4 hours of light beginning at either ZT14 or ZT21 (Fig. 4D).

We next asked whether the expression of core elements of the molecular clock were altered in the SCN or liver of USP2^−/− mice. The rationale was that such disruption would be expected if USP2 functions within the molecular clock. We began with a quantitative PCR analysis of multiple clock gene mRNAs at different times in an LD cycle (Fig. 5). Using two-way ANOVA for comparison of the patterns seen in USP2⁺⁺ and USP2^−/− mice we found that the phase of the peak of Per1 and Cry1 expression is 4 hours earlier in USP2^−/− SCN. On the other hand, the peak expression of both Bmal1 and Dbp was 4 hours later in USP2^−/− liver. Hence, the phasing of expression of some clock components is altered in different ways in USP2^−/− SCN and liver. Although these differences suggest that loss of USP2 alters clock activity, it provides no insight into how those changes are mediated.

USP2 encodes de-ubiquitinylating enzymes that stabilize target proteins by removing the ubiquitins of K48-linked chains that flag proteins for proteasomal degradation [16]. Therefore, reduced or altered expression of proteins in USP2^−/− mice would provide clues to understanding the molecular targets of USP2 within the clockwork. We began by determining relative PER1 and BMAL1 protein abundance by western blotting in extracts of micro-dissected SCN over the course of a day from mice housed in LD (Fig. 6A and B). First, BMAL1 appeared to be down-regulated in USP2^−/− mice at all time points except ZT20. However, the reduction was statistically significant only at ZT8 and 12 (Fig. 6A, right). PER1 was also significantly reduced at ZT8 and ZT12 and its peak abundance was delayed to ZT16 (Fig. 6B, right). These data indicate that both PER1 and BMAL1 are reduced in the SCN at ZT12 when USP2^−/− mice are highly sensitive to light-induced phase delays.

Decreased PER1 and BMAL1 expression in the SCN of USP2^−/− mice suggests that USP2 might directly interact with protein complexes containing BMAL1 and PER1. To test this we used co-immunoprecipitation to look for protein complexes in wild type SCN protein extracts containing both clock proteins and USP2b, the more abundant USP isoform. We found that anti-BMAL1 and anti-USP2 both co-precipitated small amounts of USP2b along with CRY1, BMAL1 and CK1ε (Fig. 6C). In addition, anti-PER1 precipitated complexes with increased relative amounts of BMAL1 and USP2b (Fig. 6C), while CRY1 and CK1ε antibodies were less effective at bringing down USP2b (Fig. 6C). These data indicate that SCN extracts contain native protein complexes containing USP2b, BMAL1, PER1 and other clock proteins. Interestingly, anti-ubiquitin antibodies co-precipitate CRY1 with larger amounts of BMAL1, CK1ε and USP2b and each protein migrates at its normal molecular weight. This implies that additional ubiquitinylated proteins or ubiquitin chains coexist in complexes containing native, non-ubiquitinylated clock-proteins.

Down-regulation of PER1 and BMAL1 in USP2^−/− mice and the presence of USP2b in native clock protein complexes containing PER1 and BMAL1 suggests that USP2b may de-ubiquitinylate and stabilize either PER1 or BMAL1. If USP2b targets PER1 or BMAL1 one would predict that increasing expression of USP2b would increase PER1 or BMAL1 stability and abundance. To test this, increasing concentrations of a FLAG-USP2b construct was transiently transfected into HEK293 cells along with a constant amount of a bicistronic construct expressing FLAG-GFP and MYC-BMAL1. We used one-way ANOVA to ask whether target protein increased significantly with increasing FLAG-USP2b. We found that MYC-BMAL1 rose in parallel with USP2b (Fig.7A), and that the abundance of FLAG-GFP as a control remained nearly constant (Fig. 7A). In contrast, co-expression of increasing amounts FLAG-USP2b with MYC-CLOCK or MYC-PER1 had no significant effect on their protein abundance (Fig. 7B and C). These data indicate that USP2b can stabilize BMAL1, but not PER1 or CLOCK.

To verify these results we used a second approach in which turnover of target, MYC-tagged proteins was determined in the presence or absence of USP2b (Fig. 7D-F). Each MYC-tagged protein was expressed in HEK293 cells with either FLAG-USP2b or FLAG-GFP and after 24 hours protein synthesis was inhibited with cycloheximide and abundance of the MYC-tagged protein was measured over time. Consistent with the effect of increasing FLAG-USP2b (Fig. 7A), the turnover of BMAL1 was delayed in the presence of FLAG-USP2b (Fig. 7D). Using regression analysis we estimated that the half-life of MYC-BMAL1 was about 3.9 hours and significantly (p<0.05) increased to greater than 8 hours in the presence of USP2b. In contrast, USP2b did not significantly (p>0.05) delay turnover of MYC-CLOCK or MYC-PER1 (Fig. 7E and F). Given that BMAL1 is down-regulated in USP2^−/− SCN (Fig. 6A) and that its stability is altered by over-expression of USP2b in both assays (Fig. 7A and D) our data suggest that BMAL1 is a likely protein target of USP2b.

To determine if USP2b alters the ubiquitinylation state of proteins in complex with BMAL1, we co-expressed HA-ubiquitin and the bicistronic construct expressing FLAG-USP2b and MYC-BMAL1; as a control we used the FLAG-GFP and MYC-BMAL1 bicistronic construct. We then immunoprecipitated using FLAG or MYC antibodies and analyzed for the presence of BMAL1 (MYC), GFP or USP2b (FLAG), and ubiquitin (HA) (Fig. 8A–C). We first determined that transfection of the FLAG-GFP and MYC-BMAL1 or the FLAG-USP2b and MYC-BMAL1 initially resulted in similar expression levels of MYC-BMAL1 (Fig. 8A). In the control immunoprecipitations (Fig. 8B, left) the FLAG antibody did not pull down MYC-BMAL or HA-Ubiquitin, but the MYC antibody did pull down MYC-BMAL1. In the experimental group (Fig. 8B, right) both FLAG and MYC antibodies co-precipitated MYC-BMAL1 and FLAG-USP2b as expected if USP2b and BMAL1 were present in the same protein complex. In addition, the HA antibody immunoprecipitated FLAG-USP2b, which is consistent with the fact that USP2b can bind to ubiquitin chains. Finally, the MYC-BMAL1 blot in Figure 8B was re-probed with the HA antibody. This generated a smear above the BMAL1 bands in control IPs, but not when BMAL1 was co-expressed with USP2b (Fig. 8C). Thus, the ubiquitinylation of proteins in complex with MYC-BMAL1 was reduced in the presence of USP2b. Taken together with the altered stability and abundance of BMAL1 in the presence of USP2b, this suggests that BMAL1 is a target of USP2b.

Since USP2 expression is under circadian control [14] downstream of CLOCK/BMAL1, our work identifies a novel feedback loop at the protein level that would be expected to stabilize BMAL1. However, since USP2 is likely to have multiple targets the relationship between USP2 and BMAL1 does not necessarily explain the enhanced phase delays observed in USP2^−/− mice. As a preliminary test of the hypothesis that altered BMAL1 abundance or turnover in USP2^−/− mice contributes to the enhanced phase delays at low irradiance, we conducted phase shift experiments like those in Figure 4B using BMAL1^+/− and BMAL1^+/−/USP2^+/− mice (Fig. 9). BMAL1 heterozygotes express BMAL1 at about half the level of wildtype mice [17], which is comparable to the reduction seen in SCN of USP2^−/− mice at ZT8 and ZT12 (Fig 6A, right). These mice also exhibited significantly enhanced phase delays compared to wildtype in response to light exposure beginning at ZT12 (Fig. 9A–C). However, this was not a precise phenocopy of the USP2^−/− effect because the enhanced response was modest at low irradiance (Log irradiance 2.5 and 3) and was more pronounced at higher irradiance (Log irradiance 4 and 4.5).

In contrast, USP2/BMAL1 double heterozygotes were significantly more sensitive to low irradiance light (Log irradiance 2.5 and 3) compared to wildtype, and as in the case of USP2^−/− mice, the difference between WT and double heterozygotes was not significant at higher irradiance (Fig. 9C). In response to the two lowest light levels double heterozygotes showed an average phase delay of 38 and 12 minutes less, respectively, than USP2^−/− mice and the pattern of phase delays resembled those of USP2^−/− mice at low irradiance (compare Fig. 9C and Fig. 4C). The enhanced phase shifts in both BMAL1^+/− and BMAL1^+/−/USP2^+/− mice suggest that BMAL1 in USP2^−/− mice is at least partially responsible for the enhanced phase delays.

How could increased BMAL1 turnover in USP^−/− mice contribute to the light sensitivity phenotype that we have described? An important clue comes from the finding that it is the ubiquitinylated form of BMAL1 that is transcriptionally active [10] and that transcriptional activity is closely coupled to BMAL1 turnover [9], [10]. Thus, increased ubiquitinylation and turnover of BMAL1 at the light to dark transition could lead to the increased transcription of CLOCK/BMAL1 regulated genes such as Per1 that are known to be directly involved in delay phase shifts specifically in the lighting paradigm used in our studies. To test this we measured the pre-mRNA expression levels of the CLOCK/BMAL1 controlled genes (Per1, Dbp and Rev-erbα) in USP2^+/+ and USP2^−/− SCN during exposure to early evening low irradiance light (Fig. 10A). We used RT-qPCR with primers designed for un-spliced mRNA rather than processed mRNA because those transcripts are closer to the immediate event of CLOCK/BMAL1 induced transcription. We used c-Fos as a control because it is induced by light in the SCN through a pathway that does not involve CLOCK/BMAL1 [18], [19]. Per1, Rev-erbα, and Dbp pre-mRNA were all significantly increased (2–4 fold) in the SCN of USP2^−/− mice compared to controls (USP2^+/+) within 2 hours of low irradiance light (Log irradiance 3) initiated at ZT12, while the pre-mRNA of the immediate early gene, c-Fos, was unaffected (Fig. 10A). These data suggest that at the molecular level, the transcriptional activity of CLOCK/BMAL1 in the SCN is enhanced specifically in USP2^−/− mice during the low irradiance light treatment.

Next, we asked whether the expression or abundance of BMAL1 or PER1 protein was altered in response to the same light exposure. Western blotting for BMAL1 from protein lysates isolated from USP2^+/+ and USP2^−/− SCN showed no statistically significant effect of low irradiance light but did confirm our earlier data (Figure 6A) that BMAL1 is less abundant in USP^−/− mice (Fig. 10B, graph). The striking finding in these experiments was that after 2 hours of low irradiance light PER1 protein was nearly undetectable in the SCN of wildtype mice (Fig. 10B, left). This unexpected and previously un-reported finding contrasts sharply with the finding that in USP2^−/− mice PER1 protein was on the rise over this time course in either light or darkness (Fig. 10B, right). The decline in PER1 in low irradiance light in wildtype mice is not consistent with a direct effect of USP2 on PER1 since USP2 would be expected to have a stabilizing effect. It is generally thought that increasing PER1 is required for light-induced phase shifts. Our data suggest that at low light levels, which normally have little or no effect on phase in wild type animals, PER1 expression is actually declining. In contrast, the increased Per1 pre-mRNA (Fig. 10A) and rising PER1 protein (Fig. 10B) in USP2^−/− animals suggests that expression of PER1 is enhanced during the low light signal, which in this case leads to a phase delay. This suggests a scenario in USP2^−/− mice in which increased turnover of BMAL1 coincides with increased CLOCK/BMAL1-driven gene expression in the USP2^−/− SCN.

Although we often correlate protein abundance and stability with functional activity, our data on BMAL1 are consistent with other work indicating that it is the ubiquitinylated form of BMAL1 that is transcriptionally active [10] and that transcriptional activity is closely coupled to BMAL1 turnover (Fig. 10C). [9], [10]. Hence, we suggest that increasing expression of Per1 and other CLOCK/BMAL1 transcripts in USP2^−/− mice compared to wild type following low light might account for the enhanced phase delays observed in USP2^−/− mice.

Construction of the Usp2 targeting vector and the general strategy for introducing LoxP sites upstream and downstream of Exons 3–4 as well as upstream of the thymidine kinase (TK) neomycin (NEO) cassette (Fig. 2) was based on a previously published targeting vector strategy [39]. The resulting targeting construct was electroporated into R1 ES cells and after selection of G418 resistant ES cell clones with correct targeting, the clones were transiently transfected with a plasmid encoding Cre recombinase [39]. This resulted in clones carrying a conventional KO allele with a single LoxP site (Fig. 2A) and a “floxed” allele with LoxP sites in intron 2–3 and 4–5 with deletion of the TK-Neo cassette as verified by Southern blotting. These clones were injected into blastocysts in the Medical College of Wisconsin transgenic facility. Germ line transmission was obtained only for the conventional KO allele (Fig. 2A–C).

All mice were maintained according to a protocol approved by Institutional Animal Care and Use Committee at the Medical College of Wisconsin (Protocol Number AUA00000032). Male USP2^+/+ or USP2^−/− mice were crossed with C57BL/6 mice and were housed in LD conditions (ZT0 = 0600 hrs; ZT12 = 1800 hrs). Male BMAL1^+/− mice from Jackson Laboratories were crossed with USP2^−/− females. Mice were euthanized at indicated times by CO₂ (1L/min) followed by cervical dislocation.

Wheel running experiments involved males (1–11 months old); in each experiment the mice were age-matched. Mice were housed individually in running wheel cages (Actimetics Inc, Evanston, IL) on a 12-hour LD schedule with ad libitum feeding. Room fluorescent lighting produced light levels of 60–120 µW/cm² depending on cage position, as measured with a radiometer (International Light, Newberryport, MA) from within the cage. In the constant light (LL) paradigm fluorescent light was attenuated to 5–20 µW/cm² and every 10 days this was increased sequentially to 30–60, 50–100, 60–120, and 100–160 µW/cm². In phase shift experiments each cage was equipped with a white LED that produced an un-attenuated irradiance (I) at the bottom, center of the cage of ∼9 µW⋅cm⁻² (i.e., Log I 5.0) and neutral density filters (Kodak) were used to attenuate irradiance to ∼4.5 (Log I 4.5), ∼0.9 (Log I 4.0), ∼0.1 (Log I 3.0) and ∼0.05 µW⋅cm⁻² (Log I 2.5).

Wheel running activity was recorded and analyzed using ClockLab™ software (Actimetrics, Evanston, Il). Phase shifts were calculated as the difference between the average onsets of activity 5 days following the experimental light treatment and the average onset of activity for the preceding 4 days. Mice were re-entrained to 12-hour LD for 7 days before each subsequent experimental light treatment.

SCN regions were removed by placing the brain in a matrix device (Zivic Instruments, Pittsburgh, PA). 1 mm coronal sections of the hypothalamus were removed and 0.5 mm×0.5 mm explants representing the SCN were micro-dissected and either immediately frozen in liquid Nitrogen for protein analysis or in RNA Later (Qiagen) for RNA analysis. Five SCN regions were pooled for western blotting and one SCN region was used for each RT-qPCR run. Retinal explants were obtained by micro-dissection and immediately frozen in liquid nitrogen for western blotting (10 retinas/sample) or stored in RNA Later for RNA analysis (2 retinas/per sample). Liver was explanted (1 mm×1 mm), stored in RNA later, and used for RT-qPCR. RT-qPCR using Qiagen reagents and a Bio-Rad iCycler PCR unit was conducted according to detailed methods and with gene specific primers described below.

Ten mg of tissue in RNA later (Qiagen) was transferred to 500 µl of lysis buffer (RNeasy kit, Qiagen) and disrupted with a Pellet Pestle homogenizer (Kontes) for 1 minute followed trituration with an 18-gauge needle for 1 minute. The homogenate was transferred to a Qiashredder spin column (Qiagen) and spun at maximum speed for 2 minutes. RNA was isolated with an RNeasy kit (Qiagen) according to the product protocol. Total RNA was measured by spectrophotometry and diluted with ultra pure H₂O to a final concentration of 0.05 µg/µl. Resultant cDNA using iScript cDNA synthesis kit (Bio-Rad) was diluted 1:5 in ultra pure H₂O and 5 µl was used per 25 µl total RT-qPCR reaction (Bio-Rad iCycler). Each reaction was performed at 95° for 3 minutes, followed by 38 cycles of 30 seconds at 95°, annealing temperature for 30 seconds, and 20 seconds at 72°. The optimized annealing temperature for each primer pair was calculated from the mean melting temperature of the primer pair. Forward and reverse primers for Usp2 were:

Usp2a forward5′-GTCCCCGTCCCTGCTGCTCTCCAC-3′,

Usp2a reverse 5′- CATCTGTCGCCCCTTTTCTTCATCA-3′

Usp2b forward 5′- GGCCGCCCCCTGCTGAGAT-3′,

Usp2b reverse 5′- GAACGGCTGGCTGCTTGGTAGAGG-3′.

Primers were optimized and the slopes of the standard curves were ∼3.2 ± 0.5 with correlation coefficients equaling 0.99.

The remaining PCR primer sequences were obtained from previously published work: Clock[32], Bmal1 [33], Per1 [34], Per2[35], Cry1, Rev-erbα, Dbp[36], pre-Per1 (courtesy of Charles Weitz, Harvard Medical School), pre-Fos[37], pre-Dbp, pre- Rev-erbα[38], and RNA Polymerase II[39]. RT(−) controls in each PCR sample used were negative. RT-qPCR products were quantified using the comparative ΔCT method described previously [40].

Tissue samples were homogenized (1 min), triturated with an 18-gauge needle (1 min) and sonicated (2 sec) on ice in lysis buffer containing 75 mM Hepes (pH 7.4), 1.5 mM EGTA, 150 mM KCl, 1.5 mM MgCl₂, 15% glycerol, 0.05% NP-40, and EDTA-free complete protease inhibitor cocktail (Roche). This was repeated twice and samples were then spun at 10,000 x g for 10 minutes at 4°C. The supernatant (lysate) was used for either western blotting or immunoprecipitation (IP). For IP 200–300 µg of protein was incubated on ice for 30 minutes with 50 µl of protein G beads or protein A beads conjugated to the appropriate antibody DMP (Sigma). After spinning beads and protein lysates at 10,000 x g (10 min) at 4°C, the lysate was incubated on ice with the appropriate IP antibody (30 min) while the beads were washed. The cleared lysates, beads and antibodies were incubated, rotating, overnight at 4°C, spun at 10,000 x g (30 sec) and washed 4 times with PBS. IPed complexes were eluded by boiling (10 min) in 50 µl of 6X sample buffer followed by a spin (5 min) at 10,000 x g. Twenty µg of total protein or 10–15 µl of immunoprecipitated protein sample was loaded onto pre-cast 10% denaturing gels (Bio-Rad) for all western blots. Transfers were with a small Genie electrophoretic device (Idea Scientific, Minneapolis, MN) and proteins were detected by ECL or by irradiance detection using an Odyssey imaging system (LI-COR Biosciences).

We generated an antiserum against the C-terminal region of USP2 in guinea pig (Covance) that recognizes both USP2a and USP2b. Anti-Bmal1 antiserum (guinea pig) and PER1 (rabbit 1177) was kindly provided by Marina Antoch (Roswell Park Cancer Institute, Buffalo, NY) and David Weaver (UMass Medical Center, Worcester, MA) respectively. Other antibodies were obtained commercially as follows: anti-Bmal1 (rabbit, Abcam), anti-CK1ε (rabbit, Abcam), anti-Cry1 (mouse, Alpha Diagnostics International), anti-EGFR (rabbit, Abcam), anti-Per1 (rabbit, Pierce), anti-USP2 (rabbit, Abgent), γ-Tubulin (mouse, Abcam), anti-Ubiquitin (rabbit, Thermo Scientific), anti-HA (rabbit, Santa Cruz), anti-Flag (mouse, Sigma), and anti-c-MYC (mouse, Sigma).

A cDNA encoding mClock was provided by Charles Weitz (Harvard Medical School), and mBmal1 and mPer2 cDNAs were provided by Marina Antoch (Roswell Park Cancer Institute, Buffalo, NY). mPer1 and HA-Ubiquitin cDNAs were purchased from Addgene. The cDNA for mUsp2a and mUsp2b were cloned from mouse (MGI:1858178). The Gateway cloning system (Invitrogen) was used to clone the cDNAs into the pCMV-BICEP-4 bicistronic vector (Sigma), which was converted to a Gateway destination vector. All constructs were verified by sequencing and western blotting following transient transfection in NIH3T3 or HEK293 cells obtained from the private ATCC Biology Collection.

NIH3T3 or HEK293 cells were cultured in DMEM (Sigma) supplemented with 10% Cosmic serum, 4 mM L-glutamine, 1 mM sodium pyruvate, and 1% pen/strep were transfected at approximately 80% confluency with Lipofectamine reagent (Invitrogen) and Plus reagent (Invitrogen). For protein stability assays HEK293 cells were transfected with 1 µg of either FLAG-GFP/MYC-CLOCK, -BMAL1, or -PER1 and increasing amounts (from 0 µg to 1 µg) of FLAG-USP2b/MYC-GFP. Protein analysis was done 24 hours following transfection. For protein turnover assays HEK293 cells at about 80% confluence were transfected with a bicistronic vector expressing either FLAG-USP2b/MYC- BMAL1, MYC-CLOCK, or MYC-PER1. Control vectors included the same MYC-proteins with FLAG-GFP. Cycloheximide (20 µg/ml) treatment began 24 hours following transfection and extracts were western blotted 0, 1, 2, 4, and 8 hours later.

Usp2 rhythmicity at the level of the mRNA was tested using one-way ANOVA. Two-way ANOVA was used to test differences in phase shifts at various irradiances between wildtype and either USP2^−/−, USP2^+/−, BMAL1^+/−, or USP2^+/−/BMAL1^+/− mice and phase shifts measured in mice of different genotype at specific irradiance levels were compared to wildtype using an un-paired t-test. Period measurements and activity counts were compared between wildtype and USP2^−/− mice using an un-paired t-test. Two-way ANOVA was used to compare either mRNA or protein expression over time between wildtype and USP2^−/− mice. GraphPad Prism software (GraphPad Software, Inc) was used for both statistical analysis and plotting of quantitative data.