|Home | About | Journals | Submit | Contact Us | Français|
Most organisms employ biochemical clocks to keep daily time (Dunlap et al., 2004; Pittendrigh, 1993). Circadian clocks are oscillators defined by three key characteristics: they must free run with an approximately 24 h period, are entrainable by relevant environmental stimuli, e.g., light; and exhibit temperature compensation (TC), the relative invariance of period lengths over a physiologically relevant range of temperatures. Among these defining characteristics, TC is the least well understood.
Temperature affects the clock in three ways. First, discrete temperature steps up or down can reset the clock (e.g., Glaser and Stanewsky, 2007; Liu et al., 1998). Second, temperature can impose physiological limits on rhythmicity (Kalmus, 1934; Liu et al., 1997; Njus et al., 1977). Finally, while the phase of the clock responds to discrete temperature changes, paradoxically, its period is buffered against changes in temperature. Although this relative independence of period length had been noted (Kalmus, 1935; Pittendrigh, 1954; Wahl, 1932), Hastings and Sweeney (1957) first recognized that this buffering was not due to temperature independence but rather reflected a TC mechanism. Because both the magnitude and sign of the period change with temperature varied across the physiological range, they reasoned that TC could be a network property perhaps arising from multiple reactions having opposing contributions to period length, but unlikely to be based on a single temperature-independent process as had been supposed. Notably, even homeotherms have temperature-compensated clocks (e.g. Barrett and Takahashi, 1995; Tosini and Menaker, 1998), suggesting that TC may not simply be a property appended to a temperature-dependent clock but may instead be a property of the regulatory architecture.
Information obtained from Neurospora, Drosophila, and cultured mammalian cells is consistent with a negative feedback loop of conserved architecture at the core of the clock. In this loop two proteins that interact via PAS domains comprise a transcriptional activator that drives expression of genes encoding protein(s) that, in turn, feed back to repress the activator (see Allada et al., 2001; Schibler and Sassone-Corsi, 2002; Wijnen and Young, 2006). The Neurospora clock provides an example of such a circuit. Briefly, WHITE COLLAR-1 (WC-1) and WC-2 interact via PAS domains to make a transcriptional activator, the White Collar Complex (WCC), which activates transcription of the frequency (frq) gene. FRQ protein in turn, with the help of FRQ-interacting RNA helicase (FRH), depresses the activity of the WCC, in part by promoting its phosphorylation. Over time, FRQ is progressively phosphorylated and consequently targeted by the F-box and WD40 repeat-containing protein-1 (FWD-1) for degradation by the proteasome (reviewed in, Brunner and Schafmeier, 2006; Dunlap and Loros, 2006; Liu and Bell-Pedersen, 2006).
frq and FRQ are implicated in a number of temperature-dependent responses (Nowrousian et al., 2003). Temperature resetting has been explained by shifts in the level of FRQ without underlying changes in frq expression (Liu et al., 1998), reflecting the fact that the steady state level of FRQ rises as a function of constant temperature (Diernfellner et al., 2005; Liu et al., 1997). Moreover, frq pre-mRNA is alternatively spliced to direct production of both long and short FRQ, each of which can support compensated clocks (Diernfellner et al., 2007; Dunlap et al., 2007b), and this splicing is temperature dependent (Colot et al., 2005; Diernfellner et al., 2005). Augmented levels of long FRQ at high temperatures and increased amounts of short FRQ at low temperatures extend the physiological range of rhythmic banding (Liu et al., 1997). However, changes in the ratio of long to short FRQ do not lead to altered TC, nor is this ratio relevant to TC as a general property (Diernfellner et al., 2007). These data implicate frq/FRQ in a number of temperature responses; however, we emphasize that not all effects of temperature on the clock contribute to TC processes.
Theory and experiment have suggested mechanisms underlying TC. The simplest model invokes a pair of opposing biochemical reactions, each with normal and equal Q10s having opposing effects on progress of the clock cycle (Hastings and Sweeney, 1957). As temperature changes, cycle duration is kept constant as the rates of opposing reactions change by the same magnitude. A similar solution proposed for Arabidopsis envisions reactions in two clock-associated feedback loops with opposing responses to temperature changes so that the composite is balanced (Gould et al., 2006). An attractive mathematical formulation for the idea of opposing reactions, the balance equation (Ruoff et al., 2005; Ruoff et al., 1997), models the rate of the circadian cycle as the sum of many reactions whose contributions to period must equal 0. A mechanistically unspecified amplitude model (Lakin-Thomas et al., 1991) having some empirical support (Lahiri et al., 2005; Liu et al., 1998) achieves TC by having the size of the circadian limit cycle increase with temperature, thereby compensating for increases in angular velocity. The possibility of competing intra-and intermolecular reactions involving Drosophila PER (Huang et al., 1995; Price, 1997; Sawyer et al., 1997) led to a model in which opposing effects of dimerization and nuclear entry lead to TC (Hong et al., 2007; Hong and Tyson, 1997). This and additional mathematical models (Akman et al., 2008; Kurosawa and Iwasa, 2005; Leloup and Goldbeter, 1997; Takeuchi et al., 2007) await experimental testing. Genetic screens for TC mutants have suffered from the fact that circadian period is often dependent on dosage of clock components (e.g., Smith and Konopka, 1982); thus, temperature-sensitive alleles whose effective doses drop with increasing temperature can be mistaken for TC mutants. In general, although TC is a universal aspect of circadian rhythms, mechanisms that have been generalizable (e.g., Ruoff et al., 1997) have not been reducible to mechanistic specifics, and mechanisms that are mechanistically specified (e.g., Huang et al., 1995) have not been generalizable.
Two mutant strains, however, chrono (chr) and period-3 (prd-3), merited further study because they displayed the unusual characteristic of enhanced or extended TC (Gardner and Feldman, 1981). The period length of the Neurospora clock is slightly undercompensated between 18°C and 30°C beyond which the period shortens and TC is lost. In chr mutants, however, TC extends beyond 30°C, and prd-3 strains are overcompensated such that the period increases as a function of temperature (Gardner and Feldman, 1981). Reasoning that these types of TC might be informative, we determined the molecular bases of the mutations in chr and prd-3 and studied their roles in TC. The result is that these two independently derived period mutants, later found to be defective in TC, identify separate subunits of the same enzyme, casein kinase 2 (CK2). Furthermore, genetic manipulation of CK2 activity, unique among kinases and phosphatases examined, affects both period length and TC in a predictable manner. These and additional data are consistent with a role for CK2 in the mechanism underlying TC.
First, we confirmed the phenotypes of chr and prd-3 and established an epistasis relationship between these mutants. Consistent with previous data (Gardner and Feldman, 1981), chr shows an extended range of TC (Fig 1A, black squares vs. white circles) and prd-3 overcompensates compared to WT. In our hands, however, prd-3 shows poor rhythmic banding at 30°C and above (Fig 1A, white triangles). Interestingly, prd-3; chr double mutants have longer periods than either single mutant but overcompensate and band poorly above 30°C (Fig 1A, black triangles), demonstrating additivity of period but clear epistasis of prd-3 over chr with respect to TC.
chr and prd-3 were cloned using single nucleotide polymorphism (SNP) mapping. Genetic mapping had placed chr on the right arm of LG VI and prd-3 near the centromere of LG I (Gardner and Feldman, 1981) and these locations were refined by crossing onto a WT Mauriceville background and following segregation of SNPs (Dunlap et al., 2007a; Lambreghts et al., 2009). This localized chr (Fig 1B) and prd-3 (Fig 1C) on the physical map and suggested candidate genes for both mutants in different subunits of CK2 (arrows, Fig 1B and 1C).
CK2, a multifunctional kinase (Litchfield, 2003), is a tetramer composed of two identical catalytic α subunits and two, possibly different, regulatory β subunits (Sup Fig 1). Sequencing of NCU05485.3, encoding the β1 subunit of CK2 (encoded by ckb-1) in the chr background, revealed a C→T transition, resulting in an R265C change within an extended C-terminal tail region of CK2β1 that is conserved among fungi (Sup Fig 1, top and middle). Since the β subunit of CK2 mediates contacts between the catalytic α subunits and substrate (Litchfield, 2003), we speculate that such functions may be perturbed in chr. Functional complementation of prd-3 (Fig 1C, bottom) identified NCU03124.3 on pLORIST H031 B12 as encoding the α subunit of CK2. Sequencing revealed a T→C transition resulting in a T43H mutation near the highly conserved phosphate anchor (Taylor et al., 1992) of the CK2 subdomain I (Sup Fig 1, bottom) of CK2α (encoded by cka). Genetic rescue confirmed that these mutations in ckb-1 and cka conferred the period length defects in chr and prd-3, respectively (see Sup Table 1).
Thus, these decades-old mutants, unique among the dozens of known clock gene alleles as having extended-or overcompensation, identify separate subunits of the same holo-enzyme. Together, the simplest interpretation of the data is that CK2 plays a role in facilitating TC. Although not previously implicated in TC, CK2 has established roles in circadian clocks of Arabidopsis (Sugano et al., 1998), Neurospora (Yang et al., 2003; Yang et al., 2002) and Drosophila (Akten et al., 2003; Lin et al., 2002; Lin et al., 2005; Nawathean and Rosbash, 2004) that are consistent with the increased period lengths we see in the CK2 mutants. Since a possibility was that these mutations were hypomorphic, we pursued the effect of CK2 dosage on TC by altering dosage of both subunits to see whether the independent identification of two CK2 subunits was more than a remarkable coincidence.
Phosphorylation-mediated turnover of FRQ is a major determinant of period length, and reducing the activity of kinases acting on FRQ results in increased circadian period length (Liu et al., 2000). Thus, a trivial explanation for the chr and prd-3 mutant phenotypes would be that period length increases with temperature simply because the mutant alleles chr and prd-3 encode temperature-sensitive proteins. A more interesting result, however, would be if chr and prd-3 were genuine hypomorphs having lower levels of activity but normal temperature dependence; in this case reductions of the WT allele dosage should mimic the chr and prd-3 phenotypes. To achieve various gene dosages, we generated heterokaryons of either chr or prd-3 bearing different proportions of knockout versus WT nuclei. In terms of protein expression these approximate an allelic series. We confirmed this directly for CK2β1 by using an antibody (Ab) against the CKB-1 protein and showed, while WT and chr strains have comparable levels of CKB-1, Δckb-1 heterokaryons have lower levels of protein (Sup Fig 2).
Knockout heterokaryons bearing reduced dosage of cka (cka + Δcka, referred to as Δcka) or ckb-1 (ckb-1 + Δckb-1, referred to as Δckb-1) mostly recapitulate the original homokaryotic prd-3 and chr phenotypes, respectively. Like prd-3, many Δcka heterokaryons overcompensate between 20°C and 25°C and all show increased periods (Fig 2A). These properties are consistent with prd-3’s being a hypomorph. Although prd-3 does not consistently express an overt rhythm at 30°C, the knockdown strains show reduced period lengths at this temperature (Fig 2A). Heterokaryon Δckb-1 knockdowns resemble the chr mutant, having increased period lengths with extended-or overcompensation at the highest temperature (Fig 2B).
Heterokaryon knockdowns were crossed to WT to generate homokaryotic Δckb-1 knockout strains. These strains appear overtly arrhythmic (Fig 2C) and provide a context in which to test the hypothesis that CK2 plays an essential role in TC.
Because we suspected that chr might be hypomorphic, we hypothesized that as the level of CK2 is reduced, the TC profile of the clock should move from WT slightly undercompensated (negative slope), to extended TC (flat slope), and finally to overcompensated (positive slope). We tested this by generating a heterokaryotic strain in which the sole source of ckb-1 mRNA is driven by the quinic acid-2 promoter (qa-2, hereafter called qa) (Aronson et al., 1994; Dunlap and Loros, 2005) in proportion to the concentration of quinic acid (QA) in the medium (Fig 3A, top panel); the strain bears the genotype (his-3 + his-3+qa-ckb-1; Δckb-1) and is referred to as qa-ckb-1. We confirmed that addition of QA yields increased levels of CKB-1 in a dose-dependent manner (Fig 3A, bottom panel). QA itself does not affect the TC curve of the WT (data not shown).
Inducing ckb-1 gene expression in Δckb-1 has dramatic effects on period length and TC. Higher levels of CKB-1 result in shorter periods matching WT period lengths with typical WT undercompensation (Fig 3B, squares). By contrast, reduced levels of QA, yielding lower levels of CKB-1, result in steadily longer periods progressing to overcompensation (positive slope) (Fig 3B, circles). In confirmation of our working model, we see a consistent relationship between ckb-1 dosage and the resulting slope of the TC profile.
In summary, as CKB-1 dose increases the TC profile changes from overcompensation to slight undercompensation. It is important to note here that although it was the characterization of TC mutants that initiated these studies, the experiment described in Fig 3B involves only WT proteins whose level of expression is altered. By controlling the amount of WT CKB-1 we can dictate the TC profile. Taken together, the data support a model in which CK2 modulates TC under physiological conditions.
The phosphorylation state of FRQ and of all clock-associated proteins is affected by other kinases and phosphatases. Given this, a trivial explanation of the preceding results would be that any mutation affecting the activity of enzymes acting on clock components would affect TC. We tested this by manipulating the levels of two kinases, casein kinase 1 (CK-1a) (Gorl et al., 2001; He et al., 2006) and protein kinase A (PKAC-1) (Huang et al., 2007), as well as two phosphatases, protein phosphatase 1 (PPP-1) and protein phosphatase 2A (PP2A) (Yang et al., 2004).
We were particularly interested in whether changes in CK1 dosage might affect TC. Since ck-1a (NCU00685.3) is an essential gene in Neurospora (Gorl et al., 2001), we used an inducible-promoter knock-in approach (L. Larrondo et al., manuscript in preparation). Briefly, we replaced ~0.6 kb of sequence upstream of the predicted CK-1a ORF with the QA-inducible promoter similar to that described above (Fig 4A, top left). This allowed us to control and reduce CK1 levels without completely eliminating the protein, thereby allowing cell survival. We verified induction of message (data not shown) and protein expression using α-CK-1a antibodies (Fig 4A, bottom left).
Like CK2, CK1 determines period length. Unlike CK2, however, changes in CK1 levels do not significantly affect TC. In the absence of QA, the reduced dose of ck-1a in heterokaryons bearing the knock-in construct increases period by approximately 3 h (Fig 4B left, top); this phenotype is rescued upon addition of QA (Fig 4B left, bottom), consistent with previous findings (Gorl et al., 2001; He et al., 2006). To complete the analysis, a heterokaryon bearing the qa-ck-1a knock-in was crossed to obtain a homokaryon in which the only source of CK1 is the qa-driven construct. When the dosage of CK1 in this strain is restricted by limiting QA, the effect on TC does not approach that seen with changing CK2 (Fig 4B, right).
The potential roles of PKA and the phosphatases were also examined by controlling their doses. Endogenous promoters were replaced with ~1.1 kb of the qa promoter for both pkac-1 (NCU06240) and the catalytic subunit of PP2A (pph-1, NCU06630) and crossed to generate homokaryons. Changes in steady state expression were measured by quantitative RT-PCR (Fig 4A, bottom right) on samples grown in the absence or presence of 10-2 M QA. Less pronounced changes in period lengths were observed in homokaryons bearing qa-pkac-1 (Fig 4C, left) and qa-pph-1 (Fig 4C, right) than were seen for CK1, but no significant differences were seen in the TC profiles of these strains. Finally, heterokaryon knockouts of ppp-1 showed no significant period reduction, consistent with RIP mutants (Yang et al., 2004), and showed no effect on TC (Sup Fig 3).
Together, these data indicate that not all regulators of clock protein phosphorylation influence TC. Rather, they suggest that CK2 plays a special role in TC.
Since strains with low levels of CK2β1 show increasing periods as a function of temperature, a trivial explanation would be that CK2 activity might be compromised with increasing temperature, but only when CK2β1 levels are reduced to below normal physiological limits. To address this, we assessed endogenous CK2 activity on synthetic peptides (Kuenzel and Krebs, 1985) across a range of CK2β1 levels. Regardless of the amount of CK2β1, CK2 activity approximately doubles over the indicated ten degree temperature (representative experiment, Fig 5A); the trivial explanation does not hold. Additionally, CK2β1 levels do not vary, as assessed by Western blot, between 20°C and 30°C (data not shown).
As this analysis was performed on a synthetic substrate, we wondered how CK2 might phosphorylate a bona fide substrate as a function of temperature. While CK2 has many targets in Neurospora, one potential target likely to affect TC is the central clock component FRQ. FRQ has a number of putative CK2 phosphosites and calmodulin kinase-free, partially purified lysate fractions containing CK2 subunits phosphorylate FRQ in vitro while ckaRIP and ckb-1RIP mutant extracts show reduced levels of FRQ phosphorylation (Yang et al., 2002).
To confirm direct activity of CK2 on FRQ, we performed in vitro kinase assays with full-length GST-FRQ. CK2 is unique among kinases in that it can efficiently use GTP as a co-substrate in phosphotransfer reactions; thus, phosphorylation by GTP is considered strong evidence that a substrate is a direct CK2 target (Sugano et al., 1998; Yde et al., 2005). WT lysates incubated with full-length GST-FRQ and GTP yielded efficient phosphorylation (Fig 5B), indicating that CK2 can directly phosphorylate FRQ. However, since in vitro kinase assays can be promiscuous, we also assayed for phosphorylation of endogenous levels of FRQ by CK2. Briefly, whole cell extracts from cells grown at 30°C were incubated at 30°C with radioactive GTP and were subsequently immunoprecipitated with antiserum against FRH, a protein that interacts strongly and stoichiometrically with FRQ (Cheng et al., 2005). Immunoprecipitates revealed bands the size of FRQ in WT, but not in frq10 or in qa-ckb-1 under non-inducing conditions (Fig 5B, lower panels). FRQ is directly phosphorylated in a time-dependent manner; however, as lysates contain both kinases and phosphatases and only a limited pool of radioisotope, phosphorylation peaks at 30 min and then declines, presumably reflecting phosphatase activity and depletion of radioisotope. In this assay, there appears to be no detectable phosphorylation of FRQ when CK2β1 levels are low, so we could not assess Q10.
Since the rate of FRQ degradation is a key factor in determining period length (Liu et al., 2000; Ruoff et al., 2005), we assessed the rate of FRQ degradation as a function of reduced CK2β1 activity and temperature. We found FRQ to be degraded slightly more slowly in chr than in WT at 25°C (Fig 5C, top panel), an effect that is exacerbated at higher temperature (Fig 5C, bottom panel). An artificially high temperature of 34°C was chosen to exaggerate the subtle difference in FRQ stability seen between WT and chr. FRQ is similarly phosphorylated more slowly in a chr ; frq7 background compared to frq7 (data not shown). On one level, these data are simply consistent with expectations based on prior work, in that increased periods are due to increased FRQ stability. However, Fig 5A shows that even with low CK2β1 activity, consistent with chr, the rate of in vitro kinase activity appeared to increase with temperature. Thus, WT levels of CK2 appear critical for modulating FRQ stability specifically at higher temperatures.
To directly assess this, we tested FRQ stability in a strain expressing low levels of CK2β1. For this, we used a standard light to dark transfer assay: steady state levels of frq/FRQ are high in light, frq mRNA is rapidly degraded after the transfer to darkness, approximating the chase of a pulse-chase. In WT, FRQ degradation does not vary between the two temperatures (Fig 5D). In contrast, FRQ is roughly twice as stable at higher temperature relative to lower temperature in the qa-ckb-1 strain in the absence of QA (Fig 5D, right). Taken together with the chr data, it appears that FRQ is degraded more slowly as temperature increases when CK2β1 activity is reduced. This is consistent with a model in which over-and extended TC in strains containing hypomorphic CK2 or low dosage of WT CK2 results in a temperature-dependent misregulation of FRQ turnover. As noted above, other kinases known to phosphorylate FRQ, including CK1, do not show this effect. These data implicate CK2 in a special and perhaps unique role in maintaining the rate of FRQ turnover specifically at higher temperatures and thereby sculpting the TC profile.
Extended TC may arise at least in part due to misregulated CK2-mediated phosphorylation of FRQ at high temperatures. Extended TC might thus also result from eliminating those CK2 target sites on FRQ less effectively phosphorylated at high temperature by low CK2 doses. Hence, we analyzed FRQ strains bearing mutations in multiple putative CK2 phosphosites, which increased period length (m3, m4 and m5 from Yang et al., 2003, generous gift of Y. Liu). However, these strains did not show TC defects (data not shown). We then constructed a set of S/T→A mutations in FRQ using three criteria to identify candidate sites; sites chosen were predicted to be CK2 targets (score > 0.5 as assessed by NetPhosK 1.0 http:/www.cbs.dtu.dk/services/NetPhosK/), conserved in at least 3/5 related fungal species, and known to be phosphorylated in vivo as assessed by mass spectrometry (C. Baker, J. Loros, J.C. Dunlap, unpublished data). We made multiple rather than individual mutations as CK2 hypomorphs likely result in the coordinated hypophosphorylation of several sites. Thus, we settled on eight possible CK2 sites and three cluster mutants were made, frqQ1, frqQ2 and frqQ12 as well as a knock-in version of the WT, frqWT-KI (Fig 6A).
To assess the effects of these mutant frq alleles on TC, we used each to replace the endogenous locus (Fig 6B). Homokaryotic strains bearing frqQ1 and frqQ12, but not frqwt-KI or frqQ2, show defects in TC (Fig 6C). frqQ1 strains have a longer period, which persists as temperature increases. Moreover, FRQQ1 has a longer half-life than FRQWT-KI in vivo at 28°C (Fig 6D). This pattern is reminiscent of the chr mutant, albeit with a longer period. By contrast, frqQ2 strains show no significant difference from WT. Thus, we predicted and found phosphorylation sites on FRQ that, when mutated, phenocopy the extended TC characteristic of chr. The enhanced compensation characteristic of these strains is unique among existing frq missense mutants.
Finally, we wanted to ensure that CK2 was capable of directly phosphorylating those residues that we had mutated in frqQ1. Thus, we made GST and GST-fusions of peptides bearing the S/Ts of interest (Sup Fig 4A). These were purified (Sup Fig 4B) and radioactive kinase assays were performed with either recombinant human CK2α/β (Sup Fig 4C, left side) or recombinant N. crassa CKA (Sup Fig 4C, right side). All GST-peptide fusions were phosphorylated to varying degrees (see bands at the level of the asterisks) while GST alone was not. The reduction in phosphorylation intensity between GST-538 and GST-538A indicates that serine 538 is indeed phosphorylated, while the other serines appear also to be phosphorylated. In the lane with GST-554/558 either or both S/Ts are phosphorylated and we cannot distinguish between these possibilities. Under these conditions, bands corresponding to the sizes of various CK2 proteins (arrows) are seen indicating the anticipated autophosphorylation of this kinase (Pagano et al., 2005). Thus, this together with our previous data strongly suggests CK2 directly phosphorylates these residues in vivo.
We suggest a mechanistic entrée to the cryptic processes underlying TC, a canonical characteristic of circadian rhythms. Two classically isolated mutants were chosen because they showed unusual characteristics, extended TC and overcompensation. Analysis of these mutants identified CK2 as a key regulator of TC. The identities of chr and prd-3 are consistent with mounting evidence for the role of CK2 in circadian clocks of many model organisms (for a recent review see Mizoguchi et al., 2006). Moreover, the epistatic relationship we observed between chr and prd-3 is consistent with their identities: any activity of the β1 kinase regulatory subunit requires a functioning α catalytic subunit. The perhaps remarkable coincidence that two strains selected for defects in period length, and only later shown to have defects in TC, both bear mutations in subunits of the same kinase emphasizes that CK2 plays a key role in establishing TC.
Importantly, these analyses have defined a role for CK2 in TC without solely relying on the use of missense mutant alleles. While many mutants affecting TC exist (e.g., frq7, perS, perL etc.), it could not be excluded that their phenotypes were due to temperature-sensitive defects, irrelevant to their WT function in TC control. The use here of controlled-expression strains with only WT proteins, has ruled out this caveat. Moreover, these results demonstrate that not all mutations that change period lead to defects in TC, consistent with other alleles e.g., timelessUltraLong (Rothenfluh et al., 2000) and frq1 and frq2 (Gardner and Feldman, 1981).
The data can be thought about in terms of a simple model (Fig 7). CK2 and CK1 are essential for FRQ phosphorylation and period length determination; however, CK1 dose has little effect on TC. We show that FRQ is a direct target of CK2 and that FRQ is poorly phosphorylated by CK2 when CK2β1 is low. FRQ is degraded more slowly as temperature increases when CK2β1 is at low levels. This is consistent with mutants in which putative CK2 sites are deleted (frqQ1, frqQ12), thereby phenocopying CK2 hypomorphs. Thus, it appears that certain CK2 target sites are needed on FRQ for efficient FRQ degradation preferentially (though not exclusively) at high temperature. Since FRQ steady state levels increase with temperature (Liu et al., 1997), either CK2’s differential phosphorylation of these sites on FRQ at high temperature, or differing effects on degradation (e.g., FRQ might bind FWD-1 more efficiently at high temperature when these sites are phosphorylated) would counteract the accumulation of excess FRQ as temperature increases. In this way, we imagine that the TC curve is sculpted, or to reiterate the conclusion of Hastings and Sweeney (1957), that the effect of ambient temperature on period length is actively managed by the cell. By contrast, CK1 phosphorylation of FRQ is not required differentially at different temperatures, as if CK1 is necessary but not sufficient for timely FRQ turnover at elevated temperatures.
Finally, the data suggest that TC is an independently evolved layer that is not intrinsic to a core oscillator. In this scenario, through evolution, proto-oscillators that may not have been initially compensated may have co-opted the functions of utility kinases (such as CK2) to effect TC, and perhaps for other advantages as well (see` Allada and Meissner, 2005). Since the Q10 of CK2 activity seems to be within a thermodynamically predicted range, it seems that evolution has not tinkered much with the catalytic arm of this enzyme to achieve TC. This is reasonable given the many roles of CK2 in the cell; rather, evolution may have used the necessities arising from the global involvement of CK2 to an advantage in developing a capacity for TC. Interestingly, an observation in Arabidopsis is consistent with this view (Edwards et al., 2005). It seems plausible that effective CK2 activity is held within limits to service the many roles of the enzyme and that this homeostasis aids TC. We also note that TC is robust as regards CK2 levels: frank overcompensation was seen only at inducer levels four log orders below those that gave WT period lengths, levels that drove biochemically undetectable amounts of CK2β1 synthesis. It is apparent that in this system or in similarly constructed oscillators, two or even four-fold changes (as from loss of an allele in a diploid) would not affect TC. Instead, we suspect that subtle changes in target clock proteins—e.g., in a constellation of CK2 sites in FRQ—may allow the clock to fine-tune TC. This is akin to the idea of a region in Drosophila PER that is polymorphic across latitudinal clines (Sawyer et al., 1997). A definitive catalogue and time-history of FRQ phosphorylation would provide further insight into this problem.
WT Neurospora crassa strains used include 328-4 (ras-1bd A) and 87-74 (his-3; ras-1bd a). Knockouts (Δcka, Δckb-1 and Δppp-1) were generated by using hph to replace the entire ORF using the Neurospora Knockout Consortium’s methods (Colot et al., 2006). Culture conditions, including race tubes and handling of Neurospora were as previously described (Dunlap and Loros, 2005).
For α-CKB-1, a polyclonal Ab was raised in rabbits against two synthetic peptides, MDTLKNAANYVGDKVC-amide and Ac-ISDKVSENKHDAKAC-amide, by 21st Century Biochemicals, LLC (Marlboro, MA). Westerns were performed as previously described (Garceau et al., 1997). The α-FRH Ab was made as previously described (Cheng et al., 2005). The α-CK-1a Ab was a gift of M. Brunner.
Kinase assays with the synthetic peptide were performed with a modified protocol using a commercial CK2 Assay Kit (Upstate, Lake Placid, NY).
FRQ degradation was assessed by Western blot after transfer from light to dark and FRQ half-lives were calculated by the following equation: t1/2 = ln(2) / m where m is determined from an exponential fit in the form y = A e-mx. Half-lives were averaged from three independent curves and standard deviations were determined.
For more details see Supplementary Experimental Procedures.
L. Larrondo kindly provided strains 1061-1 and 1066-6. We thank C. Hammell for technical assistance with GST-FRQ purification, P. Ruoff for helpful discussions and R. Lambreghts and H. Benchabane for critical reading of the manuscript. This work was supported by a Canadian Institutes of Health Research Postdoctoral Fellowship to A.M. and by grants from the National Institutes of Health to J.C.D. (RO1 GM34985 and PO1 GM68087) and to J.J.L. and J.C.D. (RO1 GM083336).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.