FRQ-Interacting RNA Helicase Mediates Negative and Positive Feedback in the Neurospora Circadian Clock

The Neurospora circadian oscillator comprises FREQUENCY (FRQ) and its transcription activator, the White Collar Complex (WCC). Repression of WCC's transcriptional activity by FRQ via negative feedback is indispensable for clock function. An unbiased genetic screen that targeted mutants with defects in negative feedback regulation yielded a fully viable arrhythmic strain bearing a novel allele of FRQ-interacting RNA helicase (frh), an essential gene that encodes a putative exosome component protein. In the allele, frhR806H, clock function is completely disturbed, while roles of FRQ-interacting RNA helicase (FRH) essential for viability are left intact. FRHR806H still interacts with FRQ, but interaction between the FRQ–FRHR806H complex (FFC) and WCC is severely affected. Phosphorylation of WC-1 is reduced in the mutant leading to constantly elevated WCC activity, which breaks the negative feedback loop. WCC levels are considerably reduced in the mutant, especially those of WC-1, consistent both with loss of positive feedback (FRQ-dependent WC-1 stabilization) and with a reduced level of the FRQ-mediated WCC phosphorylation that leads to high WCC activity accompanied by rapid transcription-associated turnover. FRH overexpression promotes WC-1 accumulation, confirming that FRH together with FRQ plays a role in WC-1 stabilization. Identification of a viable allele of frh, displaying virtually complete loss of both negative and positive circadian feedback, positions FRH as a core component of the central oscillator that is permissive for rhythmicity but appears not to modulate periodicity. Moreover, the results suggest that there are clock-specific roles for FRH that are distinct from the predicted essential exosome-associated functions for the protein.

C IRCADIAN clocks are an evolutionarily conserved means through which metabolic, physiological, and developmental processes are controlled (Bell-Pedersen et al. 2005;Gatfield and Schibler 2007;Brunner and Kaldi 2008). A circa 24-hr clock allows organisms to predict day-night changes thus providing them benefits for survival (Ouyang et al. 1998;Dodd et al. 2005). Our understanding of the circadian clock has advanced dramatically since clock genes period and frq were cloned in Drosophila and Neurospora, respectively (Young and Kay 2001;Dunlap 2008). Since then, research on many model organisms has revealed a conserved aspect of regulation in which negative feedback plays an essential role in maintaining a self-sustained circadian clock (Dunlap 1999;Heintzen and Liu 2007;Mackey 2007). The circadian oscillator in fungi and animals, for instance, comprises a negative element(s) whose gene product(s) inhibits the activity of its own transcriptional activators (positive elements), a heterodimer of proteins interacting via PAS domains (Young and Kay 2001). The inhibition combined with delayed turnover of the negative elements leads to oscillatory transcription factor activity. Since those positive elements also control expression of other genes (named clock-controlled genes, ccgs), the negative feedback loop gives rise to rhythmic expression of many ccgs that are involved in processes of metabolism, development, stress responses, etc.
In the Neurospora clock, components of the FRE-QUENCY (FRQ)-White Collar Complex (WCC)-based oscillator have been identified by forward genetics, and negative feedback was proven by showing that FRQ overexpression would stop the clock (McClung et al. 1989;Aronson et al. 1994). Subsequent identification of the heteromeric WCC as the activator of frq expression provided the minimal set of elements for the feedback loop, and the model of the core oscillator as a single negative feedback loop (Crosthwaite et al. 1997). Later work established that FRQ actually binds to and inhibits the activity of its own transcription activators, the WCC (Cheng et al. 2001a;Denault et al. 2001;Froehlich et al. 2003), and this negative feedback gives rise to rhythmic expression of frq and other ccgs (Correa et al. 2003). In the current view of the feedback loop, newly synthesized FRQ enters the nucleus, where it brings casein kinase (CK I) and CK II to phosphorylate the WCC and inhibits its activity (Schafmeier et al. 2005;He et al. 2006). In the process, appreciable levels of the FRQ-WCC complex are seen in the nucleus, and the feedback loop is finally closed as FRQ-mediated phosphorylation of the WCC impedes DNA binding, WCC is stabilized (Schafmeier et al. 2008), and the complex is exported from the nucleus (Hong et al. 2008).
An important addition to the understanding of the cycle came when a putative RNA helicase found to copurify with FRQ was shown to be essential for the clock (Cheng et al. 2005). The FRQ-interacting RNA helicase (FRH), encoded by an essential gene, frh, shows significant homology to mtr4p, a cofactor of the Saccharomyces exosome complex involved in mRNA export and degradation (Mitchell and Tollervey 2000).
Recently it was shown that FRH interacts, albeit weakly, with exosome components and strains with reduced FRH have reduced frq mRNA stability (Guo et al. 2009).
To further understand the mechanism of negative feedback, we utilized a genetic screen aimed at uncovering components required for feedback. A strain, Mut10, was isolated, which showed essentially normal growth and viability but an arrhythmic phenotype. Genetic mapping revealed a chromosomal translocation in the strain tightly linked to a missense mutation (R806H) in the frh locus. Those two mutations were separated and the clock phenotype was shown to be due to the R806H change in FRH, which leads to a weakened interaction between FRQ and the WCC and a defect in WCC phosphorylation. While confirming involvement of FRH in negative feedback, this result strongly suggests that the essential functions of FRH do not constitute important elements of the clock cycle, since the frh mutant strain shows normal growth and development while having completely lost circadian regulation. Interestingly, WC-1 and WC-2 are both downregulated in the mutant strain, suggesting that FRH, like FRQ, participates in the interlocked positive feedback loop that maintains WCC stability (Lee et al. 2000;Cheng et al. 2001b;Schafmeier et al. 2006Schafmeier et al. , 2008.

MATERIALS AND METHODS
Plasmids, strains, media, and Neurospora methods: All strains used in this research included in their genetic background the ras-1 bd allele that serves to clarify the clock-controlled conidiation pattern in race tube assays (Belden et al. 2007). Because ras-1 bd does not affect the rhythm but only clarifies its expression, strains bearing only ras-1 bd are referred to here as ''wild type'' (WT). A construct pYL51A was made by Y. Liu, containing the hph gene with frq 39-UTR under control of the frq promoter and the frq ORF under control of the qa-2 promoter. pYL51A was targeted to the his-3 locus and the strain containing the reporter gene, 197-9 [ras-1 bd , his-3T(frqp-hph-frq39UTR, qa-2pfrq)], was subjected to EMS mutagenesis as described below. The knock-in (KI) constructs frh R806H-KI , qa-frh, qa-wc-1, and qa-wc-2 were transformed into the ras-1 bd ; mus-52 strain to avoid ectopic integration via nonhomologous recombination (Ninomiya et al. 2004;Colot et al. 2006). The transformants were backcrossed to WT to obtain homokaryon knock-in strains.
Genetic screen for negative feedback loss mutants: A liquid suspension of conidia from strain 197-9 (2 3 10 7 /ml) was treated with 100 mm EMS in 0.067 m potassium phosphate buffer (pH 7.0) for 5 hr. The treated conidia, which displayed an $25% survival rate, were washed three times with buffer and diluted 10-fold in the final wash. Conidial suspension ($10 7 cells) was plated onto 13 Vogel's plus 13 FIGS plates with 400 mg/ml hygromycin. The plates were cultured in constant dark and in total, 83 surviving colonies were picked after 7 days of culture.
DNA and RNA analyses: Neurospora genomic DNA was prepared from liquid cultures using the Puregene DNA isolation kit from Gentra Systems. Neurospora RNA was prepared as previously described (Aronson et al. 1994). In Southern analysis, 0.5 mg restriction digested genomic DNA was resolved on a 1% agarose gel and the digoxigenin (DIG) system from Roche was used for probe labeling and detection. In Northern analysis, 10 mg total RNA was resolved on a 1.2% agarose gel containing 2% formaldehyde, and the DIG system was used for probing and detection. Allele-specific PCR was used to differentiate SNP genotypes of mapping progeny or frh alleles of mutant and knock-in strains. Allele specific DNA oligos were designed with the online SNAPER program (http:/ /ausubellab.mgh.harvard.edu/). Oligo MS266 (AAGC CATTGTTATCGCCGTTGA) and MS254 (GAACGATGTTC GGCTGTAGGTA) flanked the translocation site and the PCR product was used in Southern analysis for chromosomal translocation.
SNP marker identification and SNP mapping: Conidia of a Mauriceville isolate (FGSC no. 2225) were cultured at 30°in constant light (LL) for 4 hr. A cDNA library was made from the extracted mRNA using a Stratagene Uni-Zap kit. The cDNA clones were sequenced and SNP markers were identified by comparing the Mauriceville sequences with the genomic sequence of the reference Oak Ridge isolate (Lambreghts et al. 2009). SNP6 and SNP72 on LGI were named after the contig numbers where they are localized (http:/ /www.broad. mit.edu/annotation/genome/neurospora/Home.html) and selected for mapping purposes. Additional SNP markers between those two markers were found by sequencing part of the Mauriceville genomic DNA in this region. Progeny from a cross between WT in Mauriceville and Mut10 in Oak Ridge were subject to SNP mapping. The rhythmicity phenotype was scored using the race tube assay and the SNP genotype was scored by allele-specific PCR.
Sequence analysis of the chromosome translocation: KpnI digests of Mut10 genomic DNA were self-ligated and subjected to inverse PCR with DNA oligos near the mapping locus on LGI. The PCR products were cloned and sequenced to obtain the region of chromosome translocation.
Forced heterokaryon assay: Mut10 was individually crossed with inl and pan-2. Progeny were screened for Mut10; inl and Mut10; pan-2 strains. A forced heterokaryon strain of Mut10 and WT was obtained by culturing Mut10; inl and pan-2 together on minimal medium, as well as the reciprocal heterokaryon of Mut10; pan-2 and inl.
Real-time PCR analyses: cDNA samples were prepared with the Invitrogen Superscript III RT reaction kit. Real-time PCR analyses were performed using the power SYBR Green system from ABI.

RESULTS
A novel genetic screen for negative feedback mutants: The concept underlying circadian negative feedback is that the protein product of the frq locus acts to turn down its own expression. Therefore, to develop a screen for mutations affecting negative feedback, we engineered a strain bearing (1) a selectable marker under control of the frq promoter, and (2) a means through which FRQ could be overexpressed. In this strain, when FRQ is overexpressed, the selectable marker should be turned off unless the feedback loop connecting FRQ to repression is blocked by mutation. Specifically, we constructed strain 197-9 [ras-1 bd , his-3T(frqp-hph-frq39UTR, qa-2p-frq)], which contains (1) the selectable marker gene hph under control of the frq promoter and thereby activated by the transcription factor WCC and (2) the frq ORF under control of the QA inducible qa-2 promoter ( Figure 1A). Expression of hph encodes hygromycin phosphotransferase, which confers cell resistance to hygromycin. Since FRQ represses WCC activity in constant darkness (DD), growth of the strain was inhibited in darkness under hygromycin selection. However, since frq is expressed in the light regardless of the degree of FRQ overexpression (Crosthwaite et al. 1995), the strain was hyg resistant in the light. After EMS mutagenesis, survivors were plated and kept in the dark under hygromycin, which killed most of cells. Out of $10 7 mutated conidia, 83 mutants grew on hygromycin, therefore signaling a potential loss of negative feedback. Among those mutant strains, 21 were arrhythmic, 6 showed period lengths longer than WT, and 13 showed shorter period lengths, and the remaining mutants showed no obvious defects in clock oscillation, suggesting that they developed alternative means of resistance. These data are consistent with the concept that negative feedback is essential for the circadian clock and that defects in this regulation might lead to clock phenotypes.
Since FRQ is a negative element in the oscillator (Aronson et al. 1994), we assumed that some mutants represented frq alleles. To identify the frq mutants, arrhythmic strains were crossed with 732-1 (frq 7 , un-10; ras-1 bd ; a). The frq 7 allele confers a long period length ($29 hr) and un-10, which is tightly linked to frq, is temperature sensitive for growth. So in these crosses, novel alleles of frq arising from the EMS mutagenesis and screen would fail to segregate progeny with wildtype rhythms and virtually all un-10 strains would have a 29-hr period length. In this way, 6 of the 21 arrhythmic mutants were localized to frq, and the recessive nature of these mutations was confirmed by a failure to grow on Figure 1.-A genetic screen for mutants defective in circadian negative feedback. (A) Scheme of the genetic screen. Expression of the selectable marker, hph, is under control of the frq promoter ( frqP); therefore it is repressed by negative feedback in the dark via FRQ inhibiting the WCC transcriptional activity. Mutants with defects in negative feedback were obtained after EMS mutagenesis through selection for resistance to hygromycin. Overexpression of FRQ by QA induction driven through the qa-2 promoter (qa-2P) was subsequently used to confirm that there was a break in the negative feedback loop connecting FRQ to the WCC. (B) Race tube assay with hygromycin selection. The mycelial growth front was marked with a thin black line each day. The thick solid black lines on race tubes represent light (LL) to dark (DD) transfer. The dots mark growth fronts 1 day after LD transfer. The original strain 197-9 that was subjected to EMS mutagenesis grew normally in light, reflecting light-induced WCC activity that activates the frqP. After LD transfer, the growth rate decreased dramatically, verifying that the WCC activity was inhibited by FRQ. In comparison with 197-9, the growth rate of Mut10 was faster in DD, suggesting that negative feedback was severely affected. Consistent with this, QA-induced FRQ could not restore the loss of negative feedback in Mut10, confirming that the strain contains a bona fide feedback mutation not at the frq locus. (C) Race tube assay in free running conditions. WT and Mut10 sibling strains were cultured in light and their clocks were assayed after transfer to DD. Mut10 is arrhythmic under free running conditions, although its growth rate is comparable to the WT sibling strain. Period length is shown from a regression analysis of n ¼ 6 bands with a standard deviation (SD). AR, arrhythmic.
plates containing hygromycin and QA. These novel frq alleles were set aside and attention focused on the mutations identifying potential new clock genes.
A novel mutation conferring a loss of negative feedback and an arrhythmic phenotype: Among the strains corresponding to the remaining 15 arrhythmic mutants was Mut10. On race tubes with hygromycin selection, both Mut10 and the control strain (197-9) grew normally in light, confirming that WCC activity is not repressed by FRQ in light (Crosthwaite et al. 1995). After L-D (light to dark) transfer, growth of 197-9 dropped dramatically, reflecting the repression of WCC activity by negative feedback. In comparison, the growth rate of Mut10 slowed but then rebounded, suggesting that negative feedback is severely affected. In this strain, overexpression of FRQ by QA could not restore negative feedback, confirming that Mut10 is not a frq allele but a bona fide negative feedback mutant ( Figure 1B).
The Mut10 strain is profoundly arrhythmic on race tubes under free running conditions in the dark with regular medium, suggesting that the loss of negative feedback indeed disrupts clock function ( Figure 1C). While the conidiation rhythm as seen on race tubes is an output of the circadian clock, it can mask the underlying clock rhythm in certain conditions . To confirm the circadian phenotype at the molecular level, the time course of frq mRNA expression was followed in the mutant strain with Northern analysis. Consistent with previous reports (Aronson et al. 1994), frq mRNA in the WT strain oscillated in a circadian manner with two peaks around time points DD12 and DD36 ( Figure  2A). In the Mut10 strain, overall frq message level was distinctly elevated compared with WT and the frq expression pattern showed a random fluctuation presumably reflecting the loss of negative feedback regulation ( Figure 2A). As expected (Garceau et al. 1997), FRQ expression and protein phosphorylation oscillated in the WT strain ( Figure 2B). In comparison, FRQ expression in Mut10 was constitutively high with all phospho-isoforms present at all time points ( Figure  2B), as if there were overexpression of FRQ resulting from abundant frq transcription, but the newly synthesized FRQ could not shut down frq transcription effectively due to defects in negative feedback.
Single nucleotide polymorphism mapping of Mut10 identified a chromosomal translocation and a missense mutation in frh: Initial genetic mapping revealed little segregation between the arrhythmic phenotype of Mut10 and two genetic markers on linkage group I (LGI) near the centromere, mat (10.9 cM) and his-3 (7.3 cM) ( Figure 3A). SNP mapping Lambreghts et al. 2009) using Mut10 (in the Oak Ridge background) and WT (in the Mauriceville geographical variant background of Neurospora) (Metzenberg et al. 1985) was used to achieve a finer genetic resolution. Consistent with the initial mapping data, the mutation was mapped to a region of $5 MU between markers SNP6 and SNP72 near centromere I (CENI) ( Figure  3A). Unexpectedly however, 2 of 10 recombinant progeny showed recombination with both SNP markers, suggesting either an exceptional frequency of double recombination, gene conversion, or the presence of a chromosomal rearrangement tightly linked to the mutation conferring arrhythmicity. To resolve this, additional SNP markers were identified by sequencing Mauriceville genomic DNA in the area between SNP6 and SNP72 and were scored among those recombinant progeny. Using these additional markers, Mut10 was mapped in a 20-kb region; the two ''aberrant'' progeny also showed recombination with all newly identified markers, supporting the idea of a chromosomal translocation tightly associated with arrhythmicity. To confirm this and to localize the breakpoint on LGI, primer pairs were chosen in the 20-kb region from WT DNA and used for PCR using Mut10 genomic DNA as a template. All pairs except for one yielded the expected products, and the one that did not yield any product indicated the site of a chromosomal translocation breakpoint, later confirmed by Southern blot ( Figure 3B).
The translocation junction fragment was cloned using self-ligation of mutant genomic DNA digest and inverse PCR. Sequencing analysis identified a reciprocal chromosomal translocation between LGI L and LGII L (Figure 3, B and C; supporting information, Figure S1 A). We hypothesized that arrhythmicity could derive from effects of the translocation on nearby gene(s), but knockouts (Colot et al. 2006) of 10 ORFs on LGI near the junction showed no clock phenotype (data not shown). However, ORFs on LGII in proximity to the breakpoint included frh only about 100 kb away (,5 cM) from the junction site, and sequencing of Mut10 DNA identified a G-to-A transition leading to an arginine-tohistidine missense mutation as FRH R806H ( Figure 3D). FRH is highly homologous to mtr4p, a component of the Saccharomyces exosomal complex. The frh gene is essential, and even knockdowns are barely viable (Cheng et al. 2005), in contrast to the Mut10 strains, which are fully viable and show robust growth.
Identification of this mutation in frh suggested that the translocation, although tightly linked, was irrelevant to the clock defect, so SNP markers were used to score additional progeny of a Mut10 with WT cross to identify strains bearing each mutation separately ( Figure S1, A and B). The strain with only the chromosomal translocation between LGI L and LGII L displayed a wild-type clock phenotype and a normal growth rate (data not shown). This strain designated as T(IL;IIL)JD10 may serve as a useful tool for genetic study in Neurospora (Perkins 1997). FRH expression appears normal in frh R806H (Figure S1 C), suggesting the mutation does not affect gene expression level or protein stability, but only the function of the protein.
Codominance of frh R806H with respect to rhythm expression: Heterokaryons (HETs) were generated between Mut10 and WT strains using auxotrophic markers pan-2 and inl. All heterokaryon strains show weak conidiation rhythms with approximately normal period lengths, an intermediate phenotype between Mut10 and WT ( Figure 4A), indicating that frh R806H is codominant to WT. Given this result, and to confirm that no additional mutations in the Mut10 background contributed to the arrhythmic phenotypes, a phenocopy approach was used. A construct bearing the frh R806H fragment linked to the bar gene (which confers resistance to ignite) was targeted by transformation to the frh locus in WT (Figure S2 A). The transformants were Genetic mapping analyzing progeny from a cross between Mut10 and WT showed linkage to traditional markers (mat and his-3) and newly identified SNP markers near the centromere region on linkage group I (LGI). The genetic distance is reported as the number of recombinants over the number of total progeny analyzed. SNP markers were named after their contig numbers in Neurospora genome (release 7) (http:/ /www.broad.mit.edu/annotation/ genome/neurospora). (B) Chromosomal translocation in Mut10. A diagram of the chromosomes in WT and Mut10 is shown (left). The red solid line and the blue dashed line represent chromosomes of LGI and LGII, respectively, from WT (top left) and Mut10 (bottom left). The thick black bar indicates a fragment amplified by oligos MS266 and MS254 (see materials and methods) from WT genomic DNA. The fragment flanks a BglII site (arrowhead) and was used as a probe in the Southern analysis (right); other BglII sites are also marked by arrows. The breakpoint of the chromosomal translocation is within the probe region, therefore the probe hybridizes to DNA from two chromosomes in Mut10, resulting in a different digestion pattern from WT as shown in the Southern analysis. See text and Figure S1 for details. (C) The translocation junction fragment was cloned by inverse PCR and sequencing of the junction region identified the exact translocation point between LGI (sequence in red) and LGII (sequence in blue). A thymidine nucleotide was lost as a result of the translocation event. (D) Sequencing of the frh locus in Mut10 revealed a G-to-A mutation leading to an arginine-to-histidine change (R806H). crossed with WT to obtain homokaryotic KI progeny. Genotyping of the frh allele in the knock-in strains was achieved by Southern analysis and allelic-specific PCR ( Figure S2, B and C). The control strain in which a WT sequence copy of frh linked with bar was knocked into the locus, frh WT-KI , showed rhythmic conidiation on a race tube assay. In comparison, frh R806H-KI strains were arrhythmic ( Figure 4B). These data confirm that the FRH R806H missense mutation breaks the clock, probably by effectively competing with WT FRH, without adversely affecting cell viability or growth. Given the myriad essential functions of the Saccharomyces homolog of FRH, including mRNA processing, transport and stabilization, the identification of an allele that specifically abrogates the clock function without impairing essential functions is noteworthy.
Loss of positive feedback in Mut10 suggests that FRH regulates WC-1 post-translationally: As a part of the circadian negative feedback loop, FRQ inhibits DNA binding by the WCC by mediating WCC phosphorylation (Froehlich et al. 2003;Schafmeier et al. 2005). As a result of the loss of negative feedback, WC-1 and possibly WC-2 were hypophosphorylated in Mut10 and frh R806H strains ( Figure 5, A and B and Figure S1 C). This led to an increase of WCC activity and the dramatic increase of frq mRNA and FRQ expression (Figure 2). Intriguingly however, WC-1 and WC-2 expression levels were very low in frh R806H , while FRQ levels were extremely high ( Figure 5 and Figure S1 C). This result seems at first paradoxical given the other role of FRQ in the posttranscriptional positive feedback loop in which, when FRQ is induced, WC-1 is upregulated within hours (Lee et al. 2000;Cheng et al. 2001b;Schafmeier et al. 2006). The failure to induce or maintain WCC expression in frh R806H while FRQ is overwhelmingly abundant suggested that the positive loop is disrupted in the mutant as well. To evaluate this, we examined the expression of the wc genes in more detail. The reduced level of WC-1 and WC-2 seen in frh R806H could be due to reduced transcript levels and/or reduced protein stability. Transcript levels were followed by RT-PCR, revealing that both wc-1 and wc-2 mRNA levels were mildly reduced in frh R806H ( Figure 5C). The reduction in wc-2 mRNA levels (approximately two-to threefold) was similar to the change in WC-2 protein and may explain the reduction in the WC-2 levels. However, the approximately twofold reduction in wc-1 mRNA level did not seem sufficient to account for the six-to eightfold reduction in the WC-1 levels. Given this, another possibility is that FRH influences the stability of the WCC.
To assess stability, wc-1 and wc-2 were placed under control of the qa-2 promoter and overexpressed in the frh R806H background ( Figure S3 A and Figure 6A). Subsequent transfer to low QA and high glucose medium repressed additional WC synthesis so WC-1 and WC-2 stability could be examined. WC-2 protein was stable in the frh R806H background (data not shown), consistent with the possibility that downregulation of WC-2 in the mutant was at the transcriptional level. In contrast, WC-1 protein was markedly less stable in frh R806H in comparison with frh WT , suggesting that FRH also regulates WC-1 in a post-translational manner ( Figure 6B). Recent data suggest that FRQ stabilizes WC-1 by promoting its inactivation via phosphorylation (Schafmeier et al. 2008), and indeed Western analysis shows WC-1 to be more hypophosphorylated in the Mut10 strain ( Figure 5, A and B). Since WC-1 levels are low in the mutant strains, we induced WC-1 in WT and frh R806H respectively to compare phosphorylation ( Figure 6C). WC-1 becomes hyperphosphorylated in WT apparently due to FRQdependent phosphorylation (Schafmeier et al. 2005) whereas WC-1 is hypophosphorylated in frh R806H ( Figure  6C). Therefore, an interpretation consistent with the data is that FRH R806H blocks or fails to promote this WC-1 phosphorylation, and the resulting hyperactive and hypophosphorylated WC-1 is unstable (Schafmeier et al. 2008).
These results from mutant strains suggest that FRH plays its role in positive feedback through effects on WC-1 regulation. To confirm this hypothesis, we constructed qa-frh strains ( Figure S4, A and B) in which FRH dosage is controlled by the inducer QA (Aronson et al. 1994). Western analyses showed that FRH can be well induced by QA, and induction of FRH increased both FRQ and WC-1 levels ( Figure 6D). All FRQ protein is bound by FRH, and FRH stabilizes FRQ (Cheng et al. 2005). FRH stabilizes WC-1 by promoting FRQ-dependent WC-1 phosphorylation. In summary, these results suggest that it is most likely the interaction of both FRQ and FRH as the FFC with the WCC that is essential to promote WC-1 phosphorylation and stability. In this model FRH, the pan-2 and inl) and HET2 (heterokaryon of Mut10; inl and pan-2) showed weak rhythmic conidiation on race tubes with a strong conidiation background. The phenotype is between wild type and Mut10, suggesting that Mut10 is codominant with WT. Period is shown by regression analysis of n ¼ 6-7 conidial bands on the race tube. (B) Race tube assay of frh knock-in strains (see Figure S2 for details). The frh WT-KI (SM52) strain, bearing a normal copy of frh with bar as a selectable marker, showed a normal circadian rhythm in free running conditions. In comparison, two separate frh R806H-KI strains (SM53 and SM54) showed arrhythmicity, which phenocopied Mut10 and frh R806H . constant partner of FRQ, assists FRQ both in repressing WCC activity and in maintaining WC-1 stability, and FRH, as a component in the oscillator, has dual roles in both negative and positive feedback.
Although the level of FRQ is under circadian regulation, the amount of FRH was invariant through time (Cheng et al. 2005) (Figure S1 C), indicating that rhythmic production of FRH is not a part of the clock cycle. Given this, the influence of FRH dosage on period length could be assessed in a straightforward manner by using qa-frh strains. At low QA concentration, qa-frh yielded low FRH and showed a slow growth rate with arrhythmicity on race tubes. This is consistent with a previous study using RNAi to reduce frh expression (Cheng et al. 2005), and confirms that FRH plays essential roles in development and in the clock. However, the change from arrhythmicity to rhythmicity is abrupt; when FRH levels pass a certain threshold, negative feedback is recovered and the clock operates with a normal period (Figure S4 C). This result suggests that FRH functions more like a permissive ''switch'' for rhythmicity than a period ''fine tuner.'' FRH R806H affects interactions among FRQ, FRH, and the WCC: FRH R806H is associated with reduced WC-1 phosphorylation, and yet Q-PCR analysis confirmed that there is no difference in CKI or CKII levels between WT and frh R806H (data not shown). Given this, two hypotheses to explain the nature of the defect in FRH R806H are that the mutation might affect the interaction of FRH with FRQ or the FRH/FRQ complex with the WCC, in either case breaking the feedback loop. The first of these is not the case: Co-immunoprecipitation using a-FRH clearly showed that FRH R806H still forms a complex with FRQ ( Figure 7A), a finding that is consistent with codominance of the mutant. In contrast to this, the interaction between FRQ-FRH R806H and the WCC is disrupted in Mut10 as indicated by reciprocal co-IP data using a-FRQ or a-WC-2 in which FRQ no longer associates with WC-2 in the Mut10 background ( Figure 7B). Because a caveat of the co-IP analysis is that WCC level is low in Mut10, we (A) Circadian time course samples of WT and Mut10 were subjected to Western analysis and blotted with FRQ, WC-1, and WC-2 antisera. The progeny 248-9 and 353-9 are from two successive generations of backcrosses between Mut10 and WT; both strains are arrhythmic. While FRQ levels were elevated in Mut10, WC-1 and WC-2 levels in such strains were much lower than in WT, suggesting that the high FRQ level could not maintain the WCC level, because the positive feedback loop of FRQ promoting WCC expression was disrupted in the mutant (see text for details). Arrowheads point to bands corresponding to hyperphosphorylated WC-1 or WC-2, representing inactive WCC (Schafmeier et al. 2005). Both White Collar proteins were hypophosphorylated in Mut10, consistent with loss of negative feedback that leads to upregulated WCC activity. (B) frh R806H-KI strains phenocopy Mut10 at the molecular level, showing defects in negative feedback leading to high expression of FRQ and defects in positive feedback leading to reduced expression of the WCC. Arrowheads point at hyperphosphorylated WCC. (C) Real-time PCR analysis of mRNA levels in samples collected over 1 day in darkness or after 24 hr in light. wc-1 (top) and wc-2 (bottom) mRNA levels were in general reduced in frh R806H compared to WT. (Error bar 6 SD, n ¼ 3 PCR repeats.) performed co-IP in strains in which WC-1 and WC-2 expression was driven by the qa-2 promoter and could be artificially induced ( Figure 6A). Here, despite the high level of qa-2-driven gene expression, the mutant strains showed low levels of WCC interacting with FRQ, especially WC-1 ( Figure 7C) and were arrhythmic on race tubes (Figure S3 B). Thus, frh R806H appears to be a clock-specific allele that affects the interaction of the FFC with the WCC, leading to a defect in the negative feedback loop.
These results, taken in the context of the forced heterokaryon data ( Figure 4A), also suggest that frq R806H acts as a dominant negative allele. We know that FRH is much more abundant than FRQ in the cell (Cheng et al. 2005). In heterokaryons of WT and frh R806H , FRH R806H may compete with FRH WT for FRQ WT interaction, partially blocking negative feedback and obscuring rhythmic conidiation on race tubes. DISCUSSION We have identified a dominant negative allele of frh with a discrete loss of the circadian clock-specific role(s) for the FRH protein. FRH had been previously identified as a protein copurifying with FRQ whose knockdown resulted in arrhythmicity (Cheng et al. 2005), and it was therefore proposed as a component of the core clock. However, because frh appears to be the ortholog of an essential gene in Saccharomyces, mtr4 (dob1), that encodes a DEAD/DEAH box RNA helicase component of the exosome complex, and the rhythm defects were exposed only in a narrow range of expression allowing extremely weak growth and aberrant development, it was difficult to completely rule out pleiotropic effects as underlying the rhythm defect. Identification of FRH R806H obviates this concern; strains bearing this mutant protein are normally robust, displaying essentially wild-type growth and development, and yet are profoundly arrhythmic. Although other explanations are possible, the simplest interpretation of the data is that all of the essential functions performed by FRH are maintained normally in FRH R806H , and the corollary of this is that those essential functions may not be required for the clock-specific roles of FRH. The qa-wc-1; qa-wc-2; frh R806H strain was cultured in 2% glucose (ÀQA condition) and then moved to 0.1% glucose with 0.01 m QA (1QA condition) and grown for 8 hr. Extracts were made and subjected to Western analysis and the result showed that WC-1 and WC-2 could be induced by QA. (B) Reduced WC-1 stability in frh R806H . WC-1 and WC-2 were induced in WT and frh R806H by QA for 12 hr before QA was washed away and CHX was added. Time course samples collected after QA release were subjected to Western analysis and the change in WC-1 levels as determined by densitometry is plotted as a function of time. WC-1 was less stable in frh R806H than in WT, suggesting that FRH R806H had a defect in maintaining WC-1 stability and FRH has a role in post-translation regulation of WC-1. WC-2 was stable in frh R806H . (C) Defects in WC-1 phosphorylation in frh R806H . (Left) WCC was induced in WT and frh R806H , and samples were cultured in dark and extracted at subjective dusk time. WC-1 in frh R806H shows faster mobility, consistent with hypophosphorylation. (Right) WC-1 is normally phosphorylated under these conditions and can be dephosphorylated by l-PPase and this dephosphorylation can be inhibited by sodium vanadate. Samples from CT15 (DD4) were analyzed: N signifies no addition to the reaction mix; PPase, addition of phosphatase; PPase1V, addition of both PPase and the PPase-inhibitor vanadate. (D) FRH regulates FRQ and WC-1 expression. qa-frh KI strains were initially cultured in 2% glucose media for 48 hr (0 QA) and then moved into 0.1% glucose media with 0.01 m QA for induction. Time course samples (6 hr and 24 hr) were collected and extracts were subjected to Western analysis. FRH is inducible by QA in qa-frh KI . Induction of FRH leads to increases in FRQ and WC-1 expression. The experiment shown is representative; variability in the kinetics of FRH, FRQ, and WC-1 induction was noted, but not in the presence of induction.
In the 1107 amino acid FRH protein, the DEAD/ DEAH domain is near the N terminus, a helicase domain is in the middle of the protein, and the DSHCT (DOB1/SKY12/helY-like helicase C-terminal) domain is near to the C terminus. The R806H mutation is near to but not within the C-terminal DSHCT domain, and since all domains essential for RNA helicase function appear intact in frh R806H which, further, does not affect the interaction with FRQ, it seems likely that this mutation is in the region specific for binding with WCC. Consistent with this, salient aspects of circadian function previously ascribed to FRQ are compromised in FRH R806H . The original and principal role of FRQ in the oscillator is to mediate negative feedback: Soon after its synthesis, hypophosphorylated FRQ goes to the nucleus (Luo et al. 1998) and interacts with the WCC (Cheng et al. 2001b;Denault et al. 2001), resulting in a decrease in the ability of the WCC to bind to DNA (Froehlich et al. 2003) probably as a result of its increased phosphorylation (Schafmeier et al. 2005;He et al. 2006). FRQ then ushers the WCC out of the nucleus closing the feedback loop (Hong et al. 2008).
The data presented here show that FRH is clearly required for completion of the negative feedback loop because this cycle is blocked at this first step, the interaction between FRQ and the WCC, with the result that FRQ expression remains constitutively high. These data are also wholly consistent with the recent observation (Schafmeier et al. 2008) that transcriptionally active WC-1 is unstable and that positive feedback (Lee et al. 2000) is a delayed result of the FRQ-mediated WCC phosphorylation that initiates negative feedback. It is clear that FRH is also required for this, because in frh R806H , despite the presence of elevated levels of FRQ, WC-1 and WC-2 fail to accumulate and WC-1 remains hypophosphorylated ( Figure 5B). By promoting the phosphorylation of WC-1 that reduces its DNA binding and both inactivates and stabilizes it, FRH supports the positive feedback loop that allows WCC accumulation (Lee et al. 2000;Schafmeier et al. 2006). The slight decline in wc-1 and wc-2 mRNA levels ( Figure 5C) may reflect additional subtle interconnections among the feedback loops, but because the oscillator is known to be insensitive to such changes in the levels of WCC expression (Cheng et al. 2001b), these minor changes in the frh R806H strain are not likely to underlie the complete loss of rhythmicity. Rather it is the loss of Figure 7.-The point mutation in FRH R806H has no effect on FRQ and FRH interaction, but abrogates interaction between FRQ, FRH, and the White Collar Complex. (A) Co-immunoprecipitation assay. FRQ and FRH form the FRQ-FRH complex (FFC) in WT, shown by reciprocal co-IP using a-FRQ and a-FRH. Similarly, the complex formation is intact in Mut10 as well, where a-FRH co-immunoprecipitates FRQ from extracts made from Mut10 strains. PI, preimmune serum. (B) Reciprocal co-immunoprecipitation assay using a-WC-2 and a-FRQ shows that the FFC and the WCC interact in extracts from WT but not Mut10 cultures. (Top left) IP with a-WC-2 pulls down WC-2 and FRQ. (Bottom left) WC-1 is detected in the immunoprecipitates from WT extracts treated with a-FRQ or a-WC-2. In contrast, in extracts made from Mut10, a-WC-2 did not pull down FRQ (top right), neither did a-FRQ pull down WC-1 (bottom right). However in Mut10 extracts, a-WC-2 does pull down WC-1 (bottom right), indicating that the WCC is intact in Mut10. (C) Coimmunoprecipitation assay in strains where WCC is induced shows that interaction between the FRQ/FRH complex and the WCC is reduced in frh R806H . (Left) In WT, IP with a-WC-1 or a-WC-2 pulls down FRQ. (Right) In frh R806H , a-WC-1 or a-WC-2 pulls down some FRQ protein, but the level is reduced in comparison with WT, arguing the interaction between FFC and WCC is reduced in mutant. negative feedback, and the data show that FRH is required with FRQ to execute this function.
A novel and potentially very powerful genetic screen for strains with defects in circadian negative feedback has also been presented, and the resolution of the defect in Mut10 presents several lessons. First it underscores the value of unbiased genetic screens, which here uncovered a very discrete mutation in an essential gene that allows a distinction to be made between the essential functions of FRH and its role as a part of the FFC in the clock. Second, the co-occurrence of the T(IL; IIL) translocation tightly linked to frh R806H in Mut10 may serve as a cautionary note. Because the strain was fully viable and the rhythm defect was linked to the translocation that involved LGI, traditional forward genetic mapping of the arrhythmic mutant with respect to widely spaced conventional markers on LGI always failed to correctly localize frh on LGII. Progeny obtained by independent assortment, or through the crossover of a translocation chromosome with centromere II (bearing the mutant frh) and the genetically marked normal sequence chromosome bearing LGI, would be lacking all genes distal to the translocation breakpoint on LGII; these would not survive. It was only through the use of SNP markers, 6 years after the strain was isolated, that frh R806H could be identified.
While identification of FRH R806H allows separation of clock-specific functions of FRH from those functions required for cell viability, we cannot categorically exclude the possibility that some of the essential functions of FRH may also impact the clock. For example, the relatively minor downregulation of wc-1 and wc-2 mRNA levels in frh R806H might signal a role for FRH in RNA stability. In any case, continued study of FRH, the FFC, and the circadian FFC-WCC complex of FRQ, FRH, WC-1, WC-2, and its associated kinases is likely to be informative. FIGURE S1.-Genetic and molecular resolution of the mutations found in the Mut10 strain. (A) Shown is an alternative view of the homology based chromosomal pairing at meiosis in a cross between Mut10 (Oak Ridge) and WT (Mauriceville). Since Mut10 contains a chromosomal translocation, pairing can only be achieved through formation of a cruciform structure during meiosis. Progeny generated from a crossover between the frh locus and the translocation site would separate the translocation and the frh missense mutation. By scoring the genetic markers shown, we obtained progeny bearing only the translocation (containing markers SNP6O and SNP72O) and progeny bearing frh R806H but not the translocation (containing SNP6M and SNP72M). (B) Race tube assay shows that frh R806H , like Mut10, is arrhythmic under free running conditions. (C) Western analysis of clock proteins in frh R806H freed from the chromosomal translocation background showed elevated constitutive FRQ expression, indicating loss of negative feedback, and reduced WCC expression level, indicating defects in positive feedback. FRH level in frh R806H was invariant and similar to WT, suggesting that the mutation affects the protein function but not FRH gene expression or stability.  Figure 4. (A) Constructs bearing frh WT or frh R806H flanked by the selectable bar marker at the 3' end were transformed into a mus-52 (ku80) strain chosen to foster a high rate of homologous recombination (COLOT et al. 2006;NINOMIYA et al. 2004). The thick bar region represents a probe used in the Southern analysis below. The double crossovers point to the approximate positions where the recombination would occur to generate frh knock-in strains. Arrow heads mark the positions of PCR primers used in (C). (N: NruI site, B: BstEII site) (B) Genomic DNA samples from WT, frh WT-KI and frh R806H-KI were subjected to enzyme digestion and Southern analysis. Since knock-in strains contained the selectable marker bar, a size shift was expected and also observed on the Southern gel confirming the strain identity. (C) A G-A missense mutation was found by sequencing in the frh R806H allele. Based on this, allele specific oligos (see materials and methods) were used in PCR analyses to differentiate frh R806H-KI and frh WT-KI . Diagram of the knock-in strategy used to replace the wc-1/wc-2 loci with QA inducible wc-1/wc-2 constructs. qa-2 P: qa-2 promoter region. Those two constructs were transformed into frh R806H respectively, and a cross between qa-wc-1; frh R806H and qa-wc-2; frh R806H was made to obtain a qa-wc-1; qa-wc-2; frh R806H strain in the progeny. (B) Race tube assay of strains overexpressing WC-1 and WC-2 in frh R806H . Restoration of the positive feedback loop was mimicked through WCC induction by QA. However, the strain was still arrhythmic on race tubes, suggesting that loss of negative feedback in frh R806H leads to clock malfunction and that restoring the positive feedback loop alone cannot rescue the clock.