Circadian Clock Control of Translation Initiation Factor eIF2α Activity Requires eIF2γ-Dependent Recruitment of Rhythmic PPP-1 Phosphatase in Neurospora crassa

ABSTRACT The circadian clock controls the phosphorylation and activity of eukaryotic translation initiation factor 2α (eIF2α). In Neurospora crassa, the clock drives a daytime peak in the activity of the eIF2α kinase CPC-3, the homolog of yeast and mammalian GCN2 kinase. This leads to increased levels of phosphorylated eIF2α (P-eIF2α) and reduced mRNA translation initiation during the day. We hypothesized that rhythmic eIF2α activity also requires dephosphorylation of P-eIF2α at night by phosphatases. In support of this hypothesis, we show that mutation of N. crassa PPP-1, a homolog of the yeast eIF2α phosphatase GLC7, leads to high and arrhythmic P-eIF2α levels, while maintaining core circadian oscillator function. PPP-1 levels are clock-controlled, peaking in the early evening, and rhythmic PPP-1 levels are necessary for rhythmic P-eIF2α accumulation. Deletion of the N terminus of N. crassa eIF2γ, the region necessary for eIF2γ interaction with GLC7 in yeast, led to high and arrhythmic P-eIF2α levels. These data supported that N. crassa eIF2γ functions to recruit PPP-1 to dephosphorylate eIF2α at night. Thus, in addition to the activity of CPC-3 kinase, circadian clock regulation of eIF2α activity requires dephosphorylation by PPP-1 phosphatase at night. These data show how the circadian clock controls the activity a central regulator of translation, critical for cellular metabolism and growth control, through the temporal coordination of phosphorylation and dephosphorylation events.

transcription, with up to 50% of the eukaryotic genome regulated by clock at the transcriptional level (9,(11)(12)(13)(14)(15)(16)(17). Furthermore, several transcript-modifying processes (including mRNA capping, splicing, polyadenylation, and deadenylation) are under clock control (11,(18)(19)(20)(21). While mRNA rhythms contribute to the generation of rhythmic protein abundance, approximately 40 to 50% of rhythmic proteins in both mouse liver and in the fungus N. crassa derive from mRNAs that are not rhythmic (22)(23)(24)(25), suggesting circadian regulation of protein stability and/or mRNA translation. In support of clock control of mRNA translation, the expression and/or phosphorylation of several translation factors are rhythmic in eukaryotic cells (22,(26)(27)(28), including rhythms in the phosphorylation and activity of the highly conserved translation initiation factor eIF2a (29)(30)(31). Interestingly, many of the proteins in this class (rhythmic protein, arrhythmic mRNA) revealed a metabolic time of day partitioning, with daytime peaks in proteins involved in catabolism or energy utilization and nighttime peaks in anabolism or energy storage (25). These findings support that clock control of translation impacts the metabolic state of the cell. Thus, understanding the connection between the clock and control of the energetically expensive process of translation is crucial for a complete understanding of cellular growth control.
How the clock controls translation is just beginning to be unraveled, with recent studies revealing a conserved role for the clock in control of translation initiation (29)(30)(31). Translation initiation starts with the formation of the ternary complex, which contains initiation factor eIF2, composed of a, b and g subunits, Met-tRNA i Met and GTP (32,33). There are multiple steps in the process, but initiation ends and elongation begins when eIF5 mediates the hydrolysis of GTP-eIF2a to GDP-eIF2a, which along with the other initiation factors, dissociates and allows 40S-and 60S-ribosomal subunit joining to create the translation-competent 80S ribosome (34,35). To initiate another round of translation, the guanine nucleotide exchange factor (GEF) eIF2B must charge GDP-eIF2a with GTP in a recycling step that is critical for controlling overall translation rates (32,33). Ser51 phosphorylated eIF2a (P-eIF2a) inhibits eIF2B GEF activity by competitively binding to the limiting eIF2B (33), thus leading to reduced translation initiation of many mRNAs (32,33), while also promoting translation of mRNAs with special motifs, including upstream ORFs (uORFs) (36). The levels of P-eIF2a have been correlated with cell growth, cancer, memory and learning and are stimulated by the integrated stress response (ISR) and the mammalian target of rapamycin (mTOR) pathways (37)(38)(39). In mammals (29,30) and N. crassa (31), the circadian clock controls rhythms in P-eIF2a abundance. Thus, studies examining the interplay between the clock and the known input pathways will help reveal the full range of translational regulation by eIF2a phosphorylation.
The mechanism of clock control of eIF2a phosphorylation is currently best understood in the fungus N. crassa where the activity of the ISR responsive kinase CPC-3 (the homolog of yeast/mammalian GCN2) is thought to be modulated at different times of day via GCN1-dependent delivery of rhythmic levels of uncharged tRNAs (31). CPC-3 is required for Ser51 phosphorylation of eIF2a (P-eIF2a), and hyperactivation of CPC-3 kinase activity, either by pharmacological induction (3-AT) or by a constitutively active mutation (cpc-3 c ), abolished P-eIF2a rhythms. However, it is not known whether CPC-3 is sufficient to drive rhythms in P-eIF2a accumulation (31). In particular, we were interested in learning how P-eIF2a is converted back to the initiation competent dephosphorylated eIF2a and whether a phosphatase might also contribute to the daily rhythms in eIF2a activity.
Protein phosphatase 1 (PP1) dephosphorylates eIF2a in yeast (40) and mammalian (41) cells. This activity requires the catalytic subunit GLC7 in yeast, as well as PP1a, PP1b, or PP1g isoforms in mammalian cells, and also requires one or more noncatalytic regulatory subunits to target PP1 to P-eIF2a (42). In mammalian cells, the RVxF motif present on GADD34 (PPP1R15A) and CReP (PPP1R15B) recruits PP1 to dephosphorylate Ser51 on eIF2a (43,44). GADD34 and/or CReP homologs are present in chickens, frogs, and zebrafish, and a degenerate ortholog was identified in Drosophila (45). In yeast cells, however, there is no GADD34 or CReP homolog, but instead, an N-terminal extension of eIF2g contains an RVxF motif that recruits GLC7 to dephosphorylate Ser51 of eIF2a (45). N. crassa PPP-1 (NCU00043) is the homolog of yeast GLC7 and is essential for survival (46). PPP-1 was previously shown to dephosphorylate FRQ protein to regulate the pace of the circadian clock (47). However, it was not known whether PPP-1 also functions to dephosphorylate P-eIF2a and control rhythmic eIF2a activity in N. crassa. In this study, we show that dephosphorylation of eIF2a in vitro required PPP-1, that PPP-1 levels are clock-controlled with a peak during the subjective night, and that the rhythm in PPP-1 accumulation is necessary for cycling P-eIF2a levels. Our study further revealed that the N terminus of eIF2g, which lacks a consensus RVxF motif, is required to recruit PPP-1 to dephosphorylate eIF2a and maintain robust P-eIF2a rhythmicity but is not required to maintain circadian clock function.

RESULTS
PPP-1 phosphatase reduces P-eIF2a levels. To determine whether phosphatase PPP-1 regulates the levels of P-eIF2a in N. crassa, the levels and phosphorylation status of eIF2a were examined in ppp-1 RIP mutant cells (47) from cultures grown in constant dark (DD) and harvested at 28 h (subjective night), which represents the low point of P-eIF2a abundance in wild-type (WT) cells (31) (Fig. 1A). P-eIF2a levels were significantly higher in ppp-1 RIP mutant cells compared to WT cells harvested at 28 h, as well as in cells harvested in the subjective morning (DD40) (see Fig. S1A in the supplemental material) or grown in constant light (LL) (see Fig. S1B). The ppp-1 RIP mutant was previously shown in vitro to reduce PPP-1 activity on P-phosphorylase by ;70% (47). P-eIF2a abundance was ;2-fold higher in ppp-1 RIP cells compared to WT cells, suggesting that PPP-1 promotes the dephosphorylation of P-eIF2a. The abundance of total eIF2a was not altered in the ppp-1 RIP cells (Fig. 1A). Complementation of ppp-1 RIP cells with a WT copy of ppp-1 inserted into the csr-1 locus (ppp-1 RIP ; csr-1::ppp-1) reduced P-eIF2a levels to back to WT levels (Fig. 1A). These data support a role for PPP-1 in maintaining the low levels of P-eIF2a present at subjective night.
To determine whether changes in PPP-1 protein abundance can control P-eIF2a levels, ppp-1 was put under the control of the copper regulatable Ptcu-1 promoter (48), and PPP-1 levels were detected using a PPP-1-specific antibody (see Fig. S2A). Consistent with the idea that PPP-1 controls P-eIF2a levels in vivo, copper sulfate (Cu) repression of Ptcu-1::ppp-1 led to low PPP-1 protein expression and high P-eIF2a levels. Conversely, addition of the copper chelator bathocuproinedisulfonic acid (BCS), led to high PPP-1 protein expression and low P-eIF2a levels compared to the control (U, untreated) (Fig. 1B). No significant changes were observed in eIF2a levels in any of the conditions used. Together, these data support the idea that PPP-1 either directly and/ or indirectly reduces P-eIF2a levels in N. crassa.
To determine whether PPP-1 is responsible for dephosphorylating P-eIF2a, the eIF2 complex was purified from N. crassa cells containing a C-terminal V5-tagged eIF2g (eIF2g::v5) by coimmunoprecipitation with anti-V5 antibody (see Fig. S2B). To test whether there is a stable association of PPP-1 phosphatase with the eIF2 complex, we first examined P-eIF2a levels over time from the immunoprecipitated complex without addition of cell extract (Mock). We found that P-eIF2a levels in the mock treatments were unchanged over time, suggesting that PPP-1, or other phosphatases, were not copurified in the eIF2 complex (see Fig. S2C). Consistent with these data, PPP-1 was not detected in the eIF2 complex using anti-PPP-1 antibody in Western blots. To examine whether addition of PPP-1 can dephosphorylate P-eIF2a in the eIF2 complex, total protein extracts containing endogenous PPP-1 from subjective evening (DD28) cells were added. To avoid potential rephosphorylation by the CPC-3 kinase, we utilized Dcpc-3 extracts (Fig. 1C). Cell extracts deficient in PPP-1 (ppp-1 RIP ; Dcpc-3) were also examined (Fig. 1C). Extracts from Dcpc-3 cells led to an ;50% reduction of P-eIF2a levels after 120 min, while no significant dephosphorylation of eIF2a was detected using extracts from ppp-1 RIP ; Dcpc-3 cells despite the ppp-1 RIP mutant retaining some activity.
These data suggested that the residual PPP-1 activity in the ppp-1 RIP mutant is not sufficient to dephosphorylate P-eIF2a at a level that is detectable in in vitro assays. Taken together, these data support that in vitro dephosphorylation of P-eIF2a depends on the transient presence of PPP-1.
In S. cerevisiae, activation of the eIF2a kinase GCN2 in vivo requires its association with ribosomes (49). Uncharged tRNAs are transferred from the ribosome to GCN2 by GCN1 to activate GCN2 (50-53). We found that N. crassa PPP-1 associates with ribosomes (see Fig. S2D), suggesting the possibility that this interaction may facilitate direct access to its substrate P-eIF2a. Taken together, these results support the idea that PPP-1 promotes P-eIF2a dephosphorylation and are consistent with PPP-1 directly dephosphorylating eIF2a.
PPP-1 phosphatase is required for clock control of P-eIF2a levels. To determine whether PPP-1 phosphatase controls rhythmic P-eIF2a levels in N. crassa, the phosphorylation status of eIF2a was examined in WT and ppp-1 RIP cells grown in a circadian time course (Fig. 2). In WT cells, P-eIF2a, but not total eIF2a levels, were rhythmic, with FIG 1 PPP-1 phosphatase reduces P-eIF2a levels and is critical for dephosphorylation of P-eIF2a in vitro. (A) Western blot of protein extracted from WT, ppp-1 RIP mutant, and ppp-1 RIP ; csr1::ppp-1 complemented strains during the subjective night (DD28) and probed with anti-P-eIF2a and total eIF2a antibodies. The P-eIF2a/total eIF2a signal is plotted below for each strain (mean 6 the SEM, n = 3; *, P , 0.05 [Student's t-test]). (B) Western blot of protein from Ptcu1-ppp-1 cells grown in the presence of copper sulfate (Cu), BCS, or untreated (U); harvested at DD28; and probed with anti-PPP-1, anti-P-eIF2a, and anti-eIF2a antibodies. The graph below shows the average signal of P-eIF2a/total eIF2a (mean 6 the SEM, n = 3; *, P , 0.05 [Student's t-test]). (C) In vitro dephosphorylation assay using cell extracts from Dcpc-3 and ppp-1 RIP ; Dcpc-3 cells incubated with P-eIF2a from eIF2g::v5 cells for 0, 30, 60, or 120 min. P-eIF2a and total eIF2a levels were examined by Western blotting. The graph below shows the average signal of P-eIF2a normalized to total protein for each time point and normalized to the value at time zero (mean 6 the SEM, n = 4; *, P , 0.05 [Student's t-test compared to time zero]). In panels A to C, membranes were stained with amido black as a protein loading control.
Ding et al. ® a peak in the subjective late morning ( Fig. 2A and B), consistent with our previous studies (31). While P-eIF2a and total eIF2a levels fluctuated in ppp-1 RIP cells, P-eIF2a rhythms were abolished ( Fig. 2C and D). Because the circadian clock was previously shown to be functional in ppp-1 RIP cells (47), it seemed unlikely that P-eIF2a rhythms were abolished due to a clock defect in these cells. However, to confirm clock function in the mutant, FRQ::LUC protein rhythms were examined in WT and ppp-1 RIP cells. Consistent with published data (47), FRQ levels oscillated robustly in ppp-1 RIP cells, but with an ;2 h shorter period compared to WT cells (Fig. 2E). Taken together, these data support the idea that the loss of P-eIF2a rhythms in ppp-1 RIP cells is not due to loss of rhythmicity of the core oscillator, but instead results from disruption of downstream circadian regulation of P-eIF2a levels.
Deletion of the N terminus of eIF2calters eIF2a phosphorylation levels and the dephosphorylation rate of eIF2a in vitro. The N terminus of N. crassa eIF2g (NCU02810) resembles the N terminus of the S. cerevisiae eIF2g in that it has an 80amino-acid extension compared to eIF2g homologs in higher eukaryotes. In S. cerevisiae this region is required to recruit PPP-1 to eIF2a (45) (see Fig. S3A). We predicted that if the N terminus of N. crassa eIF2g functions analogously, the levels of P-eIF2a would be high in strains that have an N-terminal eIF2g deletion. To test this prediction, residues 2 to 62 were deleted from the endogenous eIF2g gene (here referred to as eIF2g D2-62 ) (see Fig. S3A), and P-eIF2a levels were examined in a circadian time course (Fig. 3). As predicted, removal of this putative phosphatase-recruiting domain resulted in significantly higher P-eIF2a levels in eIF2g D2-62 compared to WT cells. Furthermore, P-eIF2a levels in eIF2g D2-62 cells were not significantly different than the high levels observed in ppp-1 RIP cells (Fig. 3A). These results support a role for the N-terminal region of N. crassa eIF2g in recruiting PPP-1 phosphatase to P-eIF2a in vivo.
Deletion of the N terminus of eIF2cdisrupts P-eIF2a level rhythms. To determine whether the N-terminal extension of eIF2g is essential for circadian clock control of P-eIF2a, the levels of P-eIF2a were examined over a circadian time course in eIF2g D2-62 cells. The levels of P-eIF2a increased over time, and P-eIF2a rhythms were severely dampened in eIF2g D2-62 cells (Fig. 3C). When the data were detrended to account for the increasing levels of P-eIF2a over time, a rhythm with significantly reduced amplitude and period was detected (see Fig. S4). This dampened P-eIF2a rhythm observed in eIF2g D2-62 cells in vivo is consistent with residual PPP-1 phosphatase activity observed in vitro in eIF2g D2-62 ; Dcpc-3 extracts (Fig. 3B). Total eIF2a levels in eIF2g D2-62 cells were arrhythmic ( Fig. 3D; see also Fig. S4 in the supplemental material). Unlike the short period FRQ::LUC rhythm observed in ppp-1 RIP cells (Fig. 2E), the period of FRQ::LUC reporter rhythms was not significantly altered in eIF2g D2-62 cells compared to WT cells (Fig. 3E). Therefore, it is likely that a different regulator is used to target PPP-1 to dephosphorylate FRQ. Also, PPP-1 protein levels were still rhythmic in eIF2g D2-62 cells, suggesting the mutation did not impact PPP-1 protein expression (see Fig. S4D). Taken together, these data support a role for the N terminus of eIF2g in recruiting PPP-1 to P-eIF2a and promoting circadian clock control of P-eIF2a levels.
Rhythmic phosphorylation of eIF2a requires rhythmic PPP-1 levels. PPP-1 phosphatase and the N terminus of eIF2g are necessary for circadian rhythms of P-eIF2a levels but not for core clock function. Thus, rhythmic control of eIF2a activity may be through clock control of the levels and/or activities of PPP-1 phosphatase and/or eIF2g. Prior mass spectrometry proteomic studies suggested that PPP-1 protein, but not eIF2g, could be rhythmic (25). To determine whether the circadian clock controls the levels of PPP-1 phosphatase and/or eIF2g, PPP-1::luciferase (PPP-1::LUC) and eIF2g::V5 C-terminal translational fusion constructs were generated and used to replace the corresponding endogenous loci. No change in P-eIF2a levels was observed in cells containing the V5-tagged version of eIF2g::V5 compared to WT cells, indicating the tag does not alter the function of eIF2g (see Fig. S5). PPP-1::LUC protein accumulated rhythmically in WT cells but not in control clock mutant Dfrq cells, (Fig. 4A), demonstrating that PPP-1 protein levels are clock-controlled. Consistent with PPP-1 functioning as an eIF2a phosphatase, the early evening peak (with phase CT [circadian time] 14, which corresponds to DD24) in PPP-1::LUC levels correlated with the trough of P-eIF2a levels (see Fig. 2A, DD24). Alternatively, eIF2g::V5 levels did not cycle in WT cells (Fig. 4B). In addition, PPP-1::LUC rhythmicity was not altered in Dcpc-3 cells that are unable to phosphorylate eIF2a (31) (see Fig. S6), indicating that PPP-1 protein level rhythms arise from mechanisms that are independent of rhythmic eIF2a activity. Together, these data suggested the possibility that the nighttime peak in PPP-1 levels may be critical for P-eIF2a rhythms. protein extracted from the indicated strains harvested at DD28 were probed with anti-P-eIF2a or total eIF2a antibodies. P-eIF2a/total eIF2a signals are plotted below (mean 6 the SEM, n = 3; *, P , 0.05 [Student's t-test]). (B) In vitro dephosphorylation assay using cell extracts from Dcpc-3 and eIF2g D2-62 ; Dcpc-3 cells incubated with pulled down P-eIF2a from eIF2g::v5 and eIF2g D2-62 ::v5 cells, respectively, for 0, 30, 60, and 120 min. P-eIF2a and total eIF2a levels were examined by Western blotting. The graph below shows the average signal of P-eIF2a normalized to total protein for each time point and normalized to the value at time zero (mean 6 the SEM, n = 5; *, P , 0.05 [Student's t-test compared with time zero]). (C and D) Western blots of protein from eIF2g D2-62 cells grown in a circadian time course, harvested at the indicated times in DD (Hrs DD), and probed with anti-P-eIF2a (C) or anti-total eIF2a (D) antibody. Plots of the data (mean 6 the SEM, n = 5) below display the average P-eIF2a (C) or eIF2a (D) signal normalized to total protein (solid line). Both P-eIF2a and total eIF2a in eIF2g D2-62 cells were arrhythmic determined by F tests of the fit to a line (dotted lines). Membranes were stained with amido black as a protein loading control. (E) Luciferase activity from a FRQ::LUC translational fusion expressed in WT (black line) and eIF2g D2-62 (gray line) cells grown in DD and recorded every 90 min over 6 days (Hrs DD). The average normalized bioluminescence signal is plotted (mean 6 the SEM, n = 24 for WT and n = 32 for eIF2g D2-62 ). The period (h) (mean 6 the SEM) is shown on the right. Clock Control of eIF2a Activity Requires PPP-1 ® To determine whether rhythmic accumulation of PPP-1 is necessary for rhythms in P-eIF2a levels, protein from strains containing Ptcu-1::ppp-1 (Fig. 1B) grown in a circadian time course were isolated and examined by Western blotting with anti-PPP-1 antibody. In WT cells, PPP-1 protein levels were rhythmic, peaking in the subjective early evening (DD24), consistent with the PPP-1::LUC rhythms (Fig. 5A). In Ptcu1::ppp-1 cells grown in the presence of the activating chelator BCS, PPP-1 levels were high and noncycling ( Fig. 1B and 5B), and P-eIF2a levels were low and arrhythmic (Fig. 5C). In Ptcu1:: ppp-1 cells grown in the presence of the repressive copper ion (Cu), PPP-1 protein levels were low (Fig. 1B), and P-eIF2a levels were high and arrhythmic (Fig. 5D). Thus, nonrhythmic PPP-1 expression at either low or high levels abolished P-eIF2a rhythms. These data demonstrated that the rhythmic accumulation of PPP-1 protein is necessary for circadian rhythms in P-eIF2a levels.

DISCUSSION
In N. crassa and mice, circadian clock regulation of eIF2a phosphorylation controls rhythmic mRNA translation and protein accumulation (30,31). In N. crassa the eIF2a kinase, CPC-3, is necessary for the accumulation of P-eIF2a levels and a constitutively active allele causes arrhythmicity of P-eIF2a (31). Here, we show that protein phosphatase PPP-1, which peaks in levels during the subjective night, is also necessary for circadian rhythms in P-eIF2a levels. These data support a model whereby the circadian clock dynamically regulates both the phosphorylation, through the day-stimulated CPC-3 kinase, and dephosphorylation, by the night-peaking PPP-1 phosphatase, of eIF2a (Fig. 5E). The peak in activity of eIF2a at night, together with increased nighttime activity of translation elongation factor eEF-2 (28), provide a mechanism to explain increased rhythmic protein production at night in N. crassa (25).
While PPP-1 is necessary for rhythmic eIF2a activity, it is not sufficient to drive rhythms in P-eIF2a levels in strains with constitutively active CPC-3 (CPC-3 C ). In cpc-3 C cells, P-eIF2a levels are high and arrhythmic (31), despite normal rhythmic PPP-1 levels in the mutant (see Fig. S7 in the supplemental material). This may be due the levels or activity of PPP-1 not being sufficient to dephosphorylate the constantly high levels of P-eIF2a present in this mutant. While we showed that P-eIF2a levels are directly related to PPP-1 levels in a strain with WT CPC-3 activity (Fig. 1B), after 2 h in vitro only up to 50% of P-eIF2a was dephosphorylated by PPP-1 indicating that the dephosphorylation step may be kinetically unfavorable (Fig. 1C). These data are consistent with the slow in vitro dephosphorylation rate of eIF2a observed in yeast extracts (45). A second possibility for why PPP-1 rhythms are not sufficient to drive P-eIF2a rhythms in the cpc-3 C mutant is that PPP-1 may also regulate CPC-3 activity. This idea is supported by the presence of at least two phosphatases in S. cerevisiae known to target both P-eIF2a and P-GCN2. The 2A-related phosphatase SIT4, which responds to the Target of Rapamycin (TOR) pathway (54,55) and dephosphorylates eIF2a (56), also controls Ser577 phosphorylation and activity of GCN2. The phosphatase PPZ1 also impacts GCN2-dependent phosphorylation of eIF2a by an unknown mechanism (57,58). Thus, in addition to direct dephosphorylation of eIF2a, these data support a role for phosphatases controlling the activity of the eIF2a kinases. Experiments are under way to identify potentially rhythmic phosphorylation sites on CPC-3 that may be dephosphorylated by PPP-1. The presence of WT PPP-1 was necessary for dephosphorylation of P-eIF2a in vitro. Given the residual in vitro phosphatase activity in eIF2g D2-62 cells (Fig. 3B), we cannot absolutely rule out that other PPP-1-dependent phosphatases in the extracts perform the dephosphorylation of P-eIF2a. This residual activity may also explain why the rhythms of P-eIF2a levels are severely diminished, and not completely abolished, in eIF2g D2-62 cells (see Fig. S4). In any case, our data support that PPP-1 is recruited to P-eIF2a by the eIF2 subunit eIF2g to directly dephosphorylate P-eIF2a. Clock Control of eIF2a Activity Requires PPP-1 ® Kinases typically target specific substrates; however, phosphatases generally have a wide substrate range (45). In addition to dephosphorylation of eIF2a, S. cerevisiae Glc7, the catalytic subunit of PP1, dephosphorylates substrates that function in glycogen metabolism, glucose regulation, and cell division (59). Furthermore, PP1 requires one or more noncatalytic regulatory subunits to target it to different cellular compartments and for substrate specificity. More than 180 PP1 regulatory subunits have been identified in mammalian cells (60), and 17 regulatory subunits were discovered in S. cerevisiae (61). Most, but not all, PP1 regulatory subunits contain a conserved RVxF motif, which is typically flanked by basic residues at the N terminus, and by acidic residues at the C terminus (62). Regulatory subunits that recruit PP1 to eIF2a in mammalian cells, GADD34 and CReP, contain an RVxF motif (43,44). In the PP1 regulatory subunit eIF2g in S. cerevisiae, the RVxF motif is present in an N-terminal domain that extends beyond homology to mammalian eIF2g (45), and deletion of the N terminus of eIF2g does not affect yeast cell growth, indicating that the eIF2 complex is functional in translation (63). Although N. crassa eIF2g lacks the conserved RVxF motif (see Fig. S3A), we show that the N terminus of eIF2g is important for P-eIF2a levels (Fig. 3A), in vitro dephosphorylation (Fig. 3B), and rhythmicity (Fig. 3C). Because the levels of eIF2g are not clock-controlled (Fig. 4B), we suggest that the interaction between the eIF2g and eIF2a in the eIF2 complex provides a platform for eIF2g to deliver PPP-1 at night, when it is at peak levels under the control of the clock (Fig. 4A). Furthermore, our data support the possibility that interactions between PPP-1 and eIF2, including eIF2g and eIF2a subunits, as well as CPC-3, may be localized to the ribosome (see Fig. S2C), although additional experiments are needed to confirm this possibility.
Disruption of P-eIF2a rhythms, either by deletion or mutation of CPC-3 kinase in N. crassa, impacts the rhythmic translation of alg-11, but not FRQ (31) or PPP-1 (see Fig. S6) protein rhythms, or overt developmental rhythms (31). These data support that under constant environmental conditions, circadian translational regulation by the rhythmic activity of eIF2a is gene specific, as opposed to a global translational response (31). In ppp-1 RIP cells, the period of FRQ::LUC accumulation rhythms is shorter compared to WT cells (47) (Fig. 2E). However, the short period FRQ::LUC rhythm in ppp-1 RIP is not due to loss of P-eIF2a rhythms in the mutant because disruption of P-eIF2a rhythms in eIF2g D2-62 cells did not significantly alter the period of FRQ::LUC rhythmicity (Fig. 3E). eIF2a phosphorylation regulates protein production to enable the organism to quickly respond to environmental stresses, including amino acid starvation. The circadian clock provides an additional layer of regulation of eIF2a activity to control the rhythmic translation of specific target genes. While the mechanisms underlying this specificity are not known, these data support the idea that temporal control of eIF2a activity provides organisms, from fungi to mammals, the ability to respond and adapt to internal and environmental stimuli (64). Because mRNA translation requires significant cellular energy, clock control of translation may provide a mechanism to coordinate energy metabolism with translation to partition translation to the times of day when energy levels are high.

MATERIALS AND METHODS
N. crassa strains and growth conditions. N. crassa vegetative growth conditions, transformation and crossing protocols were as described previously (65). Strains generated for use in this study are described in the supplemental materials and methods (see Text S1) and are listed in Table S1 in the supplemental material. The primers used in the generation and validation strains are listed in Table S2.
Expression and purification of PPP-1::His6 protein in E. coli. To validate the specificity of PPP-1 antibody, the ppp-1 ORF was amplified with the primers PPP-1::His6 F and PPP-1::His6 R containing restriction sites for NdeI and NotI using N. crassa cDNA as the template. The pET30b vector (Invitrogen) and PCR fragment were digested with NdeI and NotI restriction enzymes and then ligated with T7 ligase (NEB). The ligated plasmids were transformed to E. coli DH5a cells and screened by kanamycin resistance and restriction digestion to get an IPTG (isopropyl-b-D-thiogalactopyranoside)-inducible PPP-1::His6 fusion plasmid. The plasmid was transformed into E. coli BL21 cells and grown in 400 ml of Luria-Bertani medium at 37°C with shaking at 250 rpm to an optical density of 0.6. PPP-1::His6 expression was induced by adding 1 mM IPTG 1 h before protein extraction. PPP-1::His6 protein was purified with Ni-NTA column following published methods (67). PPP-1::His6 protein was visualized by Coomassie blue stain and Western blotting with PPP-1 antibody.
In vivo luciferase assays. Luciferase assays to examine bioluminescence rhythms arising from strains containing luciferase fusions were performed as previously described (28). Briefly, 5 ml of 1 Â 10 5 conidia/ml were inoculated into 96-well microtiter plates containing 150 ml of 1Â Vogel's salts, 0.01% glucose, 0.03% arginine, 0.1 M quinic acid, 1.5% agar, and 25 mM firefly luciferin (LUNCA-300; Gold Biotechnology, St. Louis, MO) (pH 6). After inoculation, the microtiter plate was incubated at 30°C in constant light (LL) for 24 h and transferred to DD 25°C to obtain bioluminescence recordings using an EnVision Xcite Multilabel Reader (Perkin-Elmer Life Science, Boston, MA), with recordings taken every 90 min over at least 5 continuous days. Raw luciferase activity data were analyzed for period using BioDARE (68). Raw reads were normalized to the mean to graph the data.
Statistical analysis. Circadian time course data were examined using F tests of the fit of the data to a sine wave or a line, as previously described (65,69). A Student's t-test was used to determine significance in changes in the levels of P-eIF2a and PPP-1. Error bars in all graphs represent the standard errors of the mean (SEM) from at least three independent experiments.
Sucrose gradient fractionation. Linear sucrose gradients (10 to 50% in 10 mM HEPES-KOH, 70 mM ammonium acetate, 5 mM magnesium acetate) were prepared in ultracentrifuge tubes by using a BIOCOMP gradient station (Fredericton, NB, Canada) and stored at 4°C before use. Extracts were prepared by adding polysome extraction buffer (100 mM KCl, 20 mM HEPES-KOH, 10 mM magnesium acetate, 15 mM b-mercaptoethanol, 100mg/ml cycloheximide) to ground tissues and centrifuging the solution to remove cellular debris and lipids. Next, 400 ml of the extract containing 100 A 260 units/ml (1 A 260 unit corresponds to an absorbance of 1.0 at 260 nm) was added onto the sucrose gradient and centrifuged at 41,000 rpm for 2h at 4°C. The samples were then divided into 14 fractions of approximately 1 ml each using the BIOCOMP. The absorbances at 260 nm were used as a proxy for RNA content and graphed against the fraction of the gradient. Disome, trisome, tetrasome, and pentosome fractions were pooled as the polysome fraction. Fractions representing the 40S (#4), 60S (#5), 80S (#6) ribosome and the pooled polysome fraction were boiled in SDS loading buffer (250 mM [pH 6.8] Tris-Cl, 8% SDS, 0.2% bromophenol blue, 40% glycerol, 20% b-mercaptoethanol), and 15 ml was separated on a 10% SDS-PAGE gel for Western blotting.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, DOCX file, 0.03 MB.

ACKNOWLEDGMENTS
We thank Rachel Stroh for technical assistance with strain construction, Yi Liu for the ppp-1 RIP strain, and Qun He for kindly providing PPP-1 antibodies for preliminary Western blot trials. We also thank Thomas Dever and Madhusudan Dey for advice on the in vitro dephosphorylation of eIF2a, and we thank Matthew Sachs and Cheng Wu and members of the Bell-Pedersen lab for helpful discussions.
This study was funded by NIH R01 GM058529 and R35 GM126966 to D.B.-P.