Functional analysis of 110 phosphorylation sites on the circadian clock protein FRQ identifies clusters determining period length and temperature compensation

All vegetative cultures were maintained on complete medium slants bearing 1 × Vogel's, 1.6% glycerol, 0.025% casein hydrolysate, 0.5% yeast extract, 0.5% malt extract, and 1.5% agar (Vogel 1956). Sexual crosses were performed on Westergaard's agar plates containing 1 × Westergaard's salts, 2% sucrose, 50 ng/ml biotin, and 1.5% agar (Westergaard and Mitchell 1947). Liquid culture medium (LCM) contains 1 × Vogel's, 0.5% arginine, 50 ng/ml biotin, and 2% glucose.

To lower the cost of making a large number of frq mutants, a method described in Baker et al. (2009) was modified to use yeast homologous recombination-based integration of PCR fragments (Wang et al. 2014) bearing FRQ point mutations to restriction-digested pCB05 in place of the QuickChange II Site-directed Mutagenesis Kit (Stratagene). Four primer sets were used as flanks to facilitate homologous recombination in a yeast strain (FY834) by which point mutations of frq were introduced from PCR primers. To introduce mutations to aa 1–214 of FRQ, two PCR reactions were performed: one with a forward primer “frq segment 1F” (5′-GAACCAGAACGTAGCAGTGTG-3′) and a reverse primer “#pA R” bearing a point mutation(s) to FRQ and the other using a forward primer “#pA F” which is reverse and complementary to “#pA R” and a reverse primer “frq segment 1R” (5′-GACGATGACGACGAATCGTG-3′), and then the two PCR products were co-transformed into yeast along with pCB05 (Baker et al. 2009) digested with BstXI and XhoI to create a circular construct. Similarly, to introduce mutations falling in aa 215–437 of FRQ, primers “frq segment 2F” (5′-GTGAGTTGGAGGCAACGCTC-3′) and “frq segment 2R” (5′-GTCCATATTCTCGGATGGTA-3′ were used for PCRs in combination with pCB05 digested with XhoI to NruI; “frq segment 3F” (5′-GTCGCACTGGTAACAACACCTC-3′) and “frq segment 3R” (5′-CAGCACATGT TCAACTTCAT CAC-3′) were designed for pCB05 digested with NruI and FseI (FRQ aa 438–675), and “frq segment 4F” (5′-CACCGATCTTTCAGGAGACCCTG-3′) and “frq segment 4R” (5′-CACTCAGGTC TCAATGGTGA TG-3′) work for pCB05 digested with FseI and MluI (FRQ aa 676–989). If multiple phosphosites span two or more PCR segments mentioned above, corresponding restriction enzymes and primers encompassing the region were chosen and combined for recombination in yeast. All mutations were verified by cycle sequencing at the Dartmouth Core facility. The open reading frame of frq bearing 84 phosphomutations (frq^84A) from (Baker et al. 2009) was custom-synthesized and purchased from Genscript, and to frq^84A, additional 26 phosphosites identified in (Tang et al. 2009) were further mutated to Ala by PCR reactions using primer pairs bearing mutations to create frq^110pA. All frq mutant constructs were targeted by homologous recombination to its native locus. Plasmids verified by cycling sequencing were linearized with AseI and SspI and PCR-purified for Neurospora transformation. Neurospora transformation was performed as previously reported (Colot et al. 2006). The recipient strain used in transforming frq mutants is Δfrq::hph; Δmus-52::hph; ras-1^bd; C-box luc at his-3, and all frq mutants made in this study were in the ras-1^bd genetic background (Belden et al. 2007) and bear a V5H6 tag at their C-termini and frq-C-box-driven codon-optimized firefly luciferase gene at the his-3 locus (Gooch et al. 2008), except for the strains in Fig. 5, all of which bear frq-C-box-driven luciferase at the csr-1 locus rather than his-3. These strains were constructed by crossing phosphomutants from Baker et al. (2009) to frq-C-box-luc at csr-1.

IP was performed as previously described (Wang et al. 2016). Briefly, 2 mg of total protein was incubated with 20 μl of V5 agarose (Sigma-Aldrich, #7345) as indicated rotating at 4°C for 2 hours. The agarose beads were then washed twice with the protein extraction buffer (50 mM HEPES [pH 7.4], 137 mM NaCl, 10% glycerol, 0.4% NP-40) and eluted with 100 µl of 5 × SDS sample buffer heated at 99°C for 5 minutes.

V5H6-tagged FRQ was immunoprecipitated with 20 μl of V5 agarose (Sigma-Aldrich, Catalog #7345) from 2 mg of centrifugation-cleared lysate, FRQ-bound V5 agarose was thoroughly washed twice using the protein extraction buffer, and all supernatant was carefully removed by pipetting. To make a total reaction volume of 52 µl, 40 μl of H₂O, 5 μl of 10×NEBuffer for Protein MetalloPhosphatases (PMP), 5 μl of 10 mM MnCl₂, and 2 μl of lambda protein phosphatase (NEB, Catalog #P0753S) were added to the washed FRQ-coupled V5 resin. The mixture was incubated at 30°C for 30 minutes, and then 50 µl of 5 × SDS sample buffer was added and heated at 99°C for 5 minutes (Zhou et al. 2018).

For WB, equal amounts (15 μg) of cleared protein lysate were loaded per lane in an SDS-PAGE gel. FRQ, FRH, WC-1, and WC-2 antibodies were previously described (Garceau et al. 1997; Denault et al. 2001; Froehlich et al. 2002). Antibody against V5 (Thermo Pierce) was used at 1:5,000 dilution as the first antibody in WB (Wang et al. 2021).

To better resolve FRQ phosphorylation events, Phos-tag chemical purchased from ApexBio was added at the final concentration of 20 μM to the 6.5% SDS-PAGE Tris-Glycine gel with a ratio of 149:1 acrylamide/bisacrylamide (Wang et al. 2019).

Luciferase assays were performed as previously described (Larrondo et al. 2012). 96-well plates with each well containing 0.8 ml of the luciferase assay medium were inoculated with conidial suspension and unless otherwise specified, strains in luciferase assays were cultured at 25°C and in constant light for 16–24 hours and then transferred to the dark at the same temperature for recording light signals. Bioluminescence signals were recorded with a CCD camera every hour, data were obtained with ImageJ and a custom macro, and period lengths of the strains were manually calculated. Raw data from three replicates are shown, and time (in hours) is on the x-axis while arbitrary units of the signal intensity are on the y-axis. In Fig. 4, the strains were synchronized at 20, 25, or 30°C plus light overnight and then transferred to darkness at the same temperature used in synchronization to monitor light production by a CCD camera. Strains in Fig. 5 were entrained at 25°C for two days on a 12/12 light/dark cycle before transferring to the dark at either 20, 25, or 30°C to monitor light production by a CCD camera. Luciferase assay medium contains 1 × Vogel's salts, 0.17% arginine, 1.5% bacto-agar, 50 ng/ml biotin, and 0.1% glucose. Except for Fig. 5 (see Fig. 5 legend for controls used), WT used in the luciferase assays was 661–4a (ras-1^bd, A) that contains the frq-C-box fused to the codon-optimized firefly luciferase gene (transcriptional fusion) at the his-3 locus.

A total of 110 phosphosites on FRQ (Fig. 1a) have been identified by mass spectrometry (Baker et al. 2009; Tang et al. 2009), but mutagenetic analyses have been conducted covering only some of these phosphosites. In this study, to screen phosphosites on FRQ impacting the pace of the circadian oscillator, we adopted a strategy successfully employed in a recent publication by which a small group of phosphoresidues from over 95 sites on WCC was identified for determining the repression of WCC and thereby the closure of the feedback loop (Wang et al. 2019). To this end, we engineered a series of frq mutants (replacing Ser/Thr with Ala) covering all the 110 phosphosites in a group manner (Fig. 1a and 1b) and then assayed the roles of these phosphoevents in period determination by tracking bioluminescence signals in real-time.

Although over 100 phosphosites have been reported on FRQ, it is unknown whether they represent the entirety of the phosphoevents on the protein. To this end, we first engineered two frq mutants, frq^84pA and frq^110pA in which the 84 phosphosites (Baker et al. 2009) and all the 110 phosphosites (Baker et al. 2009; Tang et al. 2009), respectively, were mutated to Ala. The circadian clock was assayed in a strain bearing a codon-optimized firefly luciferase gene driven by the frq-C-box at the his-3 locus (Larrondo et al., 2015) in which the endogenous wild-type (WT) frq gene was replaced by the engineered frq mutants. Compared with WT, both frq^84pA and frq^110pA become arrhythmic with a high amplitude of the luciferase signal (Fig. 2a), suggesting an impaired feedback loop lacking repression of frq expression caused by these mutations. The level of FRQ in frq^84pA became extremely low but was detectable compared to that in WT (Fig. 2b). FRQ phosphorylation in frq^84pA was analyzed using a modified Phos-tag system by which single phosphorylation events on WC-1 and WC-2 could be unambiguously resolved (Wang et al. 2019). To our surprise, despite elimination of all the 84 phosphorylation sites, robust FRQ phosphorylation in frq^84pA was still detected reproducibly by the Phos-tag assay especially when compared to a lambda phosphatase-treated sample (Fig. 2c), meaning that the 84 phosphosites do not include all major phosphoevents on FRQ. Similar to frq^84pA, the level of FRQ in frq^110pA is dramatically reduced but its phosphorylation totally disappeared, reflected by the same migration pattern of FRQ bands from samples treated with or without phosphatase (Fig. 2d); these data suggest that all major phosphoevents on FRQ that occur under these growth conditions have been directly or indirectly eliminated by the 110 mutations introduced. It is worth noting that FRQ stability is known to increase in mutants disrupting phosphorylation in the N-terminal and middle parts of the protein (Baker et al. 2009; Tang et al. 2009), so the extremely reduced FRQ abundance in frq^84pA and frq^110pA suggests an undesirable side effect caused by the large quantity of mutations that have been introduced, rather than through the elimination of phosphorylation per se.

To directly examine the overall effect of FRQ phosphorylation on period length, we first made two mutants, frq^57pA and frq^27pA, together encompassing all the 84 phosphoresidues (Baker et al. 2009) mutated to Ala–frq^57pA encompasses 57 phosphosites falling in amino acids (aas) 1 to 682 of FRQ were mutated to Ala altogether, and frq^27pA bears Ala mutations to the 27 phosphosites in aa 683–989 of FRQ. Consistent with the arrhythmicity observed in frq^84pA and frq^110pA (Fig. 2a), frq^57pA does not develop an oscillating clock while frq^27pA displays a robust rhythm with an extremely decreased period, 14.1 hours (Fig. 3a), shorter than any other frq mutants bearing point mutations or deletions to the same region of FRQ, such as frq^S900A (19.5 hours) (Baker et al. 2009), frq^Δ899–989 (18.7 hours) (Baker et al. 2009), frq mutants (M14 [21.1 hours], M17 [20.9 hours], M18 [19.9 hours], and M19 [21.0 hours]) (Tang et al. 2009), or frq^C23A (18.97 hours) (Cha et al. 2011); this suggests an additive effect contributed cooperatively by multiple phosphoevents at the C-terminus of FRQ in controlling the period length. To more specifically elucidate roles of phosphorylations in smaller regions of FRQ, four additional frq mutants derived from frq^110pA were generated, each of them containing Ala mutations to phosphosites spanning ∼200–300 amino acids (Fig. 1b). In frq^1–259pA, all phosphorylatable residues between aa 1 and 259 of FRQ were changed to Ala, while keeping the remaining aa 260–989 WT and therefore potentially phosphorylatable; in frq^260–471pA, phosphosites between 260 and 471 were changed to Ala; in frq^472–708pA phosphosites between aa 472 and 708 were changed to Ala; and in frq^709–989pA, phosphosites between aa 709 and 989 were changed to Ala. Luciferase analysis showed that frq^1–259pA and frq^472–708pA exhibit a loss of rhythmicity; frq^260–471pA has an increased period length (29.4 hours), while frq^708–989pA displays a decreased period length (14.9 hours) (Fig. 3b), consistent with the circadian phenotype of frq^27pA (Fig. 3a). frq^1–259pA bears mutations in and near to the coiled-coil domain that is required for FRQ to interact with itself and other core clock components (Cheng et al. 2001) as well as mutations near but not within the nuclear localization signal (NLS) (Luo et al. 1998), which would seem to explain the lost rhythmicity seen in the mutant. However, that is not the case (see below: frq^115–193pA and frq^194–220pA). Phosphorylation surrounding the coiled coil (CC) and NLS was eliminated in frq^115–193pA and frq^194–220pA, respectively, which showed periods of 20.7 and 26 hours, respectively (Fig. 6a and 6b), suggesting that abolishing phosphorylation within or near to these domains does not completely eliminate FRQ function, and arrhythmicity in frq^1–259pA is not entirely the result of eliminating phosphorylation within and close to CC and NLS. Because L-FRQ alone is sufficient for maintaining a clock at 25°C (Liu et al. 1997), the arrhythmicity of frq^1–259pA should not result from disruption of S-FRQ expression, which is also supported by the robust rhythmicity noted in frq^1–114pA, albeit with a longer period (see below).

To separately follow the impact of these phosphoevents, the 110 phosphorylation sites on FRQ were further divided into eight additional frq segments (Fig. 1b), which were mutated and analyzed by real-time luciferase assays as above. Consistent with the phenotypes of frq^1–259pA and frq^472–708pA, frq^115–259pA and frq^472–615pA are arrhythmic. frq^1–114pA, frq^260–383pA, and frq^384–471pA show increased period lengths compared to WT with period lengths of 26.7, 25.8, and 25.4 hours, respectively. frq^616–708pA and frq^709–865pA showed ∼WT period lengths. The period length of frq^866–989pA is 15 hours (Fig. 3c), mostly recapitulating the short period observed in frq^27pA (Fig. 3a) and frq^709–989pA (Fig. 3b) and indicating that phosphorylation of the C-terminal tail of FRQ contributes tremendously to period length determination. Expression of FRQ, FRH, WC-1, and WC-2 in all these eight frq mutants (Fig. 3c) is comparable to that in WT (Supplementary Fig. 1). Except for frq^616–708pA, the other seven mutants have normal FRQ–FRH interaction (Supplementary Fig. 1). Interaction between FRQ/FRH and WC-1/WC-2 is decreased in frq^384–471pA, frq^472–615pA, and frq^616–708pA, and it becomes undetectable in frq^115–259pA (Supplementary Fig. 1), consistent with the lost rhythmicity seen in the strain. These data indicate that ablation of certain phosphorylations in the N-terminal and middle regions of FRQ causes period-lengthening effects; conversely, removal of phosphorylations within the FRQ C-terminus results in an extremely shortened period, suggesting an autoinhibitory role for this C-terminal domain. In agreement with the period changes of the frq phosphomutants in Fig. 3c, canonical frq alleles except for frq¹ at the N-terminus of FRQ display a lengthened period, while frq² (bearing the same mutation as frq⁴ and frq⁶ at Ala 895) shows a decreased period (Feldman 1982; Aronson et al. 1994a), suggesting that these mutations may impact phosphorylation of other residues, leading to period changes, although they are not phosphorylatable per se or conversely, neighboring phosphorylation events might modulate period lengths via impacting these nonphosphorylatable but functionally crucial residues.

The kinases involved in phosphorylation of FRQ, especially CK1 and CK2, have been implicated in controlling temperature compensation of the core oscillator (Liu et al. 1997; Mehra et al. 2009; Hu et al. 2021) in which the circadian period length is only slightly altered across a range of physiological temperatures. Compensation is a conserved characteristic observed across diverse circadian systems. To explore whether the phosphorylation clusters on FRQ regulate the core clock at other temperatures, the eight frq phosphomutants in Fig. 3c were further examined at 20, 25, and 30°C: frq^260–383pA and frq^384–471pA show a period trend similar to that seen in WT; frq^1–114pA and frq^709–865pA display constant period lengths across temperatures even more so than WT; frq^115–259pA and frq^472–615pA remain arrhythmic, and frq^616–708pA showed a decreased period at higher temperatures, indicating this strain has an undercompensated clock (Fig. 4 and Supplementary Fig. 2). Interestingly, frq^866–989pA bearing Ala mutations at amino acids 900, 904, 915, 917, 923, 929, 931, 950, 956, 967, and 968 of FRQ demonstrates enhanced period lengths at higher temperatures and therefore has an overcompensated clock (Fig. 4, bottom left), indicating that phosphorylation of the C-terminal tail of FRQ is involved in maintaining period lengths at enhanced temperatures. This result is consistent with a recent publication showing that mutation of three CK2 in vitro-phosphorylated sites not covered in this study, S980, S981, and S982, also result in an increased period at an elevated temperature (Hu et al. 2021). Alternatively, these 11 sites are located close to the PEST-2 domain of FRQ (Gorl et al. 2001), so their phosphorylation may indirectly impact its function leading to the period adjustment. It is worth noting that the number of mutations introduced to FRQ does not always correlate with the severity of the period alteration. For example, frq^866–989pA bearing 11 mutations displays a dramatically shortened period at 25°C and an overcompensated clock across the three temperatures (Fig. 3c and 4), while frq^616–708pA with 12 mutations still exhibits a WT period at 25°C and an undercompensated oscillator at higher temperatures, while frq^709–865pA carrying nine mutations maintains a WT period at 20, 25, and 30°C (Fig. 3c and 4). frq^866–989pA shows a much stronger period phenotype at the higher temperature than the frq^Q2 mutant which bears Ala mutations to four phosphosites 685, 800, 915, and 929 but retains normal temperature compensation (Mehra et al. 2009), suggesting that FRQ C-terminal phosphorylations contribute collaboratively to maintaining the period length across temperatures.

Given that our mutational analysis of FRQ phosphosites revealed specific domains involved in temperature compensation, we investigated at a more detailed level the involvement of single, double, or triple phosphosites on FRQ in temperature compensation. A subset of the FRQ phosphosite mutants constructed in Baker et al. (2009) were crossed to the C-box-luciferase reporter targeted to the csr-1 locus, and two siblings from each cross were screened at 20, 25, and 30°C (n = 3 at each temperature) (Supplementary Table 1). The negative control, ras-1^bd (clock WT), had normal temperature compensation, and the positive control, ras-1^bd, prd-3 (Mehra et al. 2009) was overcompensated as expected. Most FRQ phosphosites, when mutated, did not perturb temperature compensation, even when period length was changed (Fig. 5a shows representative examples; Supplementary Table 1 contains period length data at all temperatures for all of the strains tested). However, mutation of S538A & S540A or of S538A & S540A & S548A on FRQ resulted in extreme overcompensation in which period length increased as temperature increased (Fig. 5b). Compared to S538A & S540A, the additional mutation of S548 to Ala increased the period length dramatically and also caused arrhythmicity at 30°C, suggesting that this site acts synergistically with the others in this cluster. Mutation of S573A & S574A caused modest undercompensation (Fig. 5c). Statistical differences between period lengths at low vs high temperatures determined using Student's t-test (Fig. 5d) indicate that of these mutants that were examined, no single phosphosite alone is responsible for period modulation with temperature. Rather, only mutation of a combination of several key phosphosites perturbs temperature compensation, and it appears that undercompensation or overcompensation phenotypes are determined by distinct phosphosites on FRQ.

Because eliminating phosphorylation in aa 115–259 or 472–615 resulted in arrhythmicity (Fig. 3c), additional frq mutant strains bearing fewer, more select mutations were generated to these and their neighboring regions (Fig. 6a) in order to understand the roles of these phosphoevents in period manipulation. frq^1–65pA carrying nine mutations displayed a WT period length, while frq^66–114pA with eight point mutations showed a long period length similar to that in frq^1–114pA, suggesting that the effect of phosphorylations in aa 1–114 on period length is mainly caused by those in aa 66–114 (Fig. 6a). The period of frq^115–193pA was only slightly shorter than WT, while frq^194–259pA remained arrhythmic, similar to frq^115–259pA (Fig. 6a), indicating the arrhythmicity in frq^115–259pA is due mainly to the loss phosphosites in aa 194–259. It seems that phosphorylation may not impact FRQ dimerization, because the period length of frq^115–193pA remains ∼WT although it bears mutations close to and within the CC domain (aa 143–176) (Cheng et al. 2001). Although frq^472–615pA is arrhythmic (Fig. 3c), frq^472–570pA shows a long period of 46.3 hours, which, to our knowledge, is the longest period seen in frq phosphomutants to date, and frq^571–615pA shows 26.4 hours (Fig. 6a). frq^616–680pA displays a long period, 26.1 hours, and frq^681–708pA is only slightly shorter (Fig. 6a). frq^616–708pA shows an intermediate period between frq^616–680pA and frq^681–708pA, which suggests an averaging effect of two neighboring phosphorylation clusters on period length. Bearing mutations near the FFC domain, frq^616–708pA has less FRH and WCC complexed with FRQ (Supplementary Fig. 1) but it still maintains a ∼WT period (Fig. 3c), consistent with the evidence that the amount of FRH (Hurley et al. 2013) or WCC (Liu et al. 2019) in the FFC-WCC is not a determinant of the period length, even though the feedback loop relies on their presence in the complex.

To elucidate why loss of phosphorylation between aa 194 and 259 causes arrhythmicity (Fig. 6a), two additional mutants, frq^194–220pA and frq^221–259pA, were generated and assayed by luciferase analyses. frq^194–220pA has mutations to phosphosites near the NLS (aa 194–199) but is robustly rhythmic, albeit with a longer period length (Fig. 6b), suggesting that phosphorylation does not control the nuclear localization of FRQ required for a functional clock (Luo et al. 1998). This is consistent with a prior report that FRQ phosphorylation does not significantly impact its subcellular localization (Cha et al. 2011). The arrhythmicity seen in frq^221–259pA (Fig. 6b) may be caused by elimination of sites near FCD1 (Fig. 1a), a domain required for CK1 interaction and phosphorylation of the N-terminus of FRQ (Querfurth et al. 2011). frq^472–536pA and frq^537–570pA are 6 and 13 hours longer than WT, respectively, but frq^472–570pA is ∼24 hours longer (Fig. 6b), which is significantly longer than the additive lengthening of 19 hours (6 + 13 hours), suggesting that the cumulative effect of phosphorylations on period length can be stronger than the additive effect from constituent parts. frq^472–536pA contains three mutations in and close to one of the only two regions of FRQ predicted to have secondary structures (Fig. 1b). This is also a region that comprises the FCD2, so the lengthened periods of the two mutants (frq^472–536pA and frq^537–570pA) may be due to reduced CK1 interaction, consistent with an observation that the period length is determined by FRQ-CK1 interaction (Liu et al. 2019).

An intermediate period length has been observed when different FRQ mutations (at nonphosphorylatable residues) were combined intramolecularly; examples include frq³ and frq⁷ (Aronson et al. 1994a), frq^{S548A, S900A} (Baker et al. 2009), and frq^{M9 + 18} (Tang et al. 2009). To check whether this is true for mutants at phosphoresidues in different regions, some of previously reported FRQ phosphomutations were combined together. For instance, both frq^{S238A, S240A} and frq^{S238A, S240A, S390A, S392, S394A} each display a period length ∼2 hours longer than WT (Baker et al. 2009). The combination of S238A, S240A, S390A, S392A, and S394A would be predicted to be ∼26 hours, close to what we observed experimentally in frq^{S238A, S240A, S390A, S392A, S394A} (27.1 hours, Fig. 7a). Similarly, mutations in frq^{S538A, S540A} (5 hours longer than WT [+5 hours]), frq^{S541A, S545A} (+3 hours), and frq^{S632A, S643A} (+3 hours) (Baker et al. 2009) were all introduced to a single frq mutant together, frq^{S538A, S540A, S541A, S545A, S632A, S634A}, which exhibits a rhythm of 32.2 hours (Fig. 7a), exactly what is anticipated from an additive effect of the three original mutants. Taken together, these data indicate an additive effect of certain FRQ phosphomutations on period length, although this may not be extended to any combinations of FRQ phosphorylations.

To examine the overall effect of individual mutations that alter the period in the same direction, FRQ phosphomutations causing increased period lengths (Baker et al. 2009) were together introduced into a single frq mutant—frq^{S72A, S73A, S76A, S238A, S240A, S390A, S392A, S394A, S538A, S540A, S541A, S545A, S548A, S551A, S554A, S632A, S634A}—which, unexpectedly, displays a loss of rhythmicity (Fig. 7b). When this is combined with the mutations in frq^709–865pA (Figs. 3c and 4), the resultant mutant, frq^{S72A, S73A, S76A, S238A, S240A, S390A, S392A, S394A, S538A, S540A, S541A, S545A, S548A, S551A, S554A, S632A, S634A, 709–865pA}, surprisingly, fully restores rhythmicity to frq^{S72A, S73A, S76A, S238A, S240A, S390A, S392A, S394A, S538A, S540A, S541A, S545A, S548A, S551A, S554A, S632A, S634A} with a period length almost identical to that in frq^768–865pA (Fig. 7b). However, frq^{S72A, S73A, S76A, S238A, S240A, S390A, S392A, S394A, S538A, S540A, S541A, S545A, S548A, S551A, S554A, S632A, S634A, 866–989pA} still behaves arrhythmically as frq^{S72A, S73A, S76A, S238A, S240A, S390A, S392A, S394A, S538A, S540A, S541A, S545A, S548A, S551A, S554A, S632A, S634A} (Fig. 7b). In the second case, frq^709–989pA displays a circadian rhythm of ∼15 hours, which is the same as frq^866–989pA rather than frq^709–865pA (Fig. 3b and 3c), suggesting that the 11 phosphoevents occurring in aa 866–989 are epistatic to the nine found in aa 709–865. Collectively, these data suggest that the interplay between phosphogroups on FRQ can control rhythmicity and period length in diverse ways, including averaging, additive, or epistatic effects.

Phosphomimetics by amino acid substitutions like Asp (D) or Glu (E) are a widely used strategy to simulate phosphorylation by constitutively introducing a negative charge into a domain. In Neurospora, phosphomimetics have been successfully employed to study phosphorylation of the core clock components, WC-1, WC-2 and FRQ at certain sites, such as wc-1^S971D (Wang et al. 2019), wc-2^15pD (Wang et al. 2019), and frq^S548D (Baker et al. 2009), revealing interesting consequences caused by constant phosphorylation at these sites. To assess whether constitutive phosphorylation at certain sites impacts FRQ activity, several key phosphosites on the protein were mutated to Asp (D) or Glu (E) to mimic the negative charge of the phosphate group. The period length of frq^{S915A, S917A} and frq^S923A is ∼2 and 1 hour longer than WT, respectively (Baker et al. 2009), whereas frq^{S915D, S917D, S923D} and frq^{S915E, S917E, S923E} show a WT period (Fig. 8a); similarly, frq^S548A becomes 4-h longer, while frq^S548D maintains a WT period (Baker et al. 2009), suggesting that phosphorylation at these sites of FRQ contributes to maintaining the pace of the clock. However, unexpectedly, both frq^S900D and frq^S900E exhibit the same period length (18.4 and 18.9 hours, respectively) (Fig. 8b) as frq^S900A (∼18 hours) (Baker et al. 2009), suggesting that the structure of the phosphate group of pS900 plays a more important role than the negative charge that it carries in tuning the FRQ activity. Although the phosphate group and Asp/Glu are both negatively charged, their small structural distinctions may explain the failure of frq^S900D and frq^S900E as phosphorylation mimics and their behavior, instead, like phosphorylation eliminators.

The Neurospora strains generated in this study are available upon request. Supporting material is deposited at G3 online. All data used to draw conclusions of the article have been provided within the figures and tables.

available at G3 online. Supplemental material