Proteins in the Neurospora Circadian Clockworks*

We, and every other organism in the solar system, live in a profoundly rhythmic environment. The frequencies of these environmental cycles on Earth range fromminutes to years, but for most organisms the dominant periodicity is that of the earth’s rotation. A wide variety of organisms, including most eukaryotes and nearly all higher organisms, express circadian biological clocks that enable them to anticipate this particular environmental rhythmicity, and the output of these clocks are circadian rhythms. In fact, these cell-based clocks are so useful and their regulation so pervasive that for most living things the time of day that cells experience has almost everything to do with the subjective time dictated by the cell’s internal clock and relatively little to do with whether it happens to be relatively warmer or cooler, or whether it is, just then, dark or light outside. For these clocks to be truly useful all year long in a variable and systematically changing environment, theymust be robust, must be resettable by environmental cues, and must keep time equally well even though the ambient temperature or nutrition changes. True circadian clocks reflect the following necessities. The period length of the cycles organized by a circadian clock is close to but not equal to 24 h (circa dian); the cycles persist for many days in the absence of environmental cues; the phase of the rhythm (i.e. when, in real hours, the peaks and troughs in the cycle occur) is set by the last seen environmental time cues (lights on or off, or for most organisms a major temperature transition); the cycles persist through a normally noisy environment. Likewise, the period lengths are approximately the same if the average ambient temperature is at thewarmor cool end of the physiological range, a property known as temperature compensation. Rhythms with these characteristics, circadian rhythms, are widely believed to share a common molecular basis across wide phylogenetic groups of organisms. Long period rhythms (12 to 100 h) lacking one or more of these characteristics are not circadian rhythms; such rhythms are not infrequently found, particularly in systemswhere circadian regulation has been impaired, but they generally reflect organismspecific biology and are not thought to be informative regarding the common molecular bases of circadian rhythms. The circadian oscillator is viewed as the biochemical feedback loop(s) underlying circadian rhythmicity per se, and the circadian system is taken to include the oscillator and the aspects of cell and organism physiology (including driven circuits and feedback loops that acquire daily periodicity via output from the circadian oscillator) that together produce the overt metabolic and behavioral rhythms in the living organism. The biochemical oscillators underlying these rhythms are based at the level of the cell in all organisms exhibiting circadian rhythms, and at least conceptually, the problem of understanding biological timekeeping can be subdivided into three general questions. What are the molecular components of the core feedback loop(s) and how do they work together to keep time? How is such a circadian system entrained to light and temperature cycles in the real world? Given an entrainable clock mechanism, how and through what pathways does it act to regulate the metabolism of the cell in which it operates, and thereby the physiology and behavior of the whole organism? The filamentous fungus Neurospora crassa rose to prominence as a circadian model system because of the genetic andmolecular tractability of its clock, as well as for its overall similarity to the animal circadian systems (chiefly Drosophila) being dissected at the same time (1). It has served as one of the most significant and durable model systems in this field, playing a central part in both framing the questions and providing the answers to the central questions of chronobiology. The Neurospora genome of 10,620 genes is documented by a 16-fold sequence coverage supported by sophisticated bioinformatics tools; targeted gene replacements are routine with systematic efforts to delete all genes well under way (2); availability of inexpensive whole genome microarrays is fostering transcriptional profiling; and a detailed single nucleotide polymorphism map combined with routine transformation using regulatable promoters has greatly smoothed the path of cloning and analysis of genes arising from screens.

We, and every other organism in the solar system, live in a profoundly rhythmic environment. The frequencies of these environmental cycles on Earth range from minutes to years, but for most organisms the dominant periodicity is that of the earth's rotation. A wide variety of organisms, including most eukaryotes and nearly all higher organisms, express circadian biological clocks that enable them to anticipate this particular environmental rhythmicity, and the output of these clocks are circadian rhythms. In fact, these cell-based clocks are so useful and their regulation so pervasive that for most living things the time of day that cells experience has almost everything to do with the subjective time dictated by the cell's internal clock and relatively little to do with whether it happens to be relatively warmer or cooler, or whether it is, just then, dark or light outside.
For these clocks to be truly useful all year long in a variable and systematically changing environment, they must be robust, must be resettable by environmental cues, and must keep time equally well even though the ambient temperature or nutrition changes. True circadian clocks reflect the following necessities. The period length of the cycles organized by a circadian clock is close to but not equal to 24 h (circa dian); the cycles persist for many days in the absence of environmental cues; the phase of the rhythm (i.e. when, in real hours, the peaks and troughs in the cycle occur) is set by the last seen environmental time cues (lights on or off, or for most organisms a major temperature transition); the cycles persist through a normally noisy environment. Likewise, the period lengths are approximately the same if the average ambient temperature is at the warm or cool end of the physiological range, a property known as temperature compensation. Rhythms with these characteristics, circadian rhythms, are widely believed to share a common molecular basis across wide phylogenetic groups of organisms. Long period rhythms (12 to ϳ100 h) lacking one or more of these characteristics are not circadian rhythms; such rhythms are not infrequently found, particularly in systems where circadian regulation has been impaired, but they generally reflect organismspecific biology and are not thought to be informative regarding the common molecular bases of circadian rhythms.
The circadian oscillator is viewed as the biochemical feedback loop(s) underlying circadian rhythmicity per se, and the circadian system is taken to include the oscillator and the aspects of cell and organism physiology (including driven circuits and feedback loops that acquire daily periodicity via output from the circadian oscillator) that together produce the overt metabolic and behavioral rhythms in the living organism. The biochemical oscillators underlying these rhythms are based at the level of the cell in all organisms exhibiting circadian rhythms, and at least conceptually, the problem of understanding biological timekeeping can be subdivided into three general questions. What are the molecular components of the core feedback loop(s) and how do they work together to keep time? How is such a circadian system entrained to light and temperature cycles in the real world? Given an entrainable clock mechanism, how and through what pathways does it act to regulate the metabolism of the cell in which it operates, and thereby the physiology and behavior of the whole organism?
The filamentous fungus Neurospora crassa rose to prominence as a circadian model system because of the genetic and molecular tractability of its clock, as well as for its overall similarity to the animal circadian systems (chiefly Drosophila) being dissected at the same time (1). It has served as one of the most significant and durable model systems in this field, playing a central part in both framing the questions and providing the answers to the central questions of chronobiology. The Neurospora genome of 10,620 genes is documented by a Ͼ16-fold sequence coverage supported by sophisticated bioinformatics tools; targeted gene replacements are routine with systematic efforts to delete all genes well under way (2); availability of inexpensive whole genome microarrays is fostering transcriptional profiling; and a detailed single nucleotide polymorphism map combined with routine transformation using regulatable promoters has greatly smoothed the path of cloning and analysis of genes arising from screens.

Known Components of Neurospora Circadian Feedback Loop
Transcription and translation-based negative feedback loops are central elements in eukaryotic circadian clocks (1,(3)(4)(5)(6)(7). In animals and fungi these feedback loops have a single transcription step in which a heterodimeric transcriptional activator (a positive element) promotes expression of a gene encoding a protein (the negative element in the loop) that depresses the activity of the activator. In all circadian systems there are additional feedback loops that are nested around the core, often connecting output to input or providing additional levels of control that can fine tune the oscillator. When this central feedback loop is first disabled by mutation and then genetic background, growth conditions, and nutrition are carefully . adjusted, noncircadian rhythms in growth can usually be detected (e.g. Ref. 8). Although there is continued interest in whether these cryptic rhythms may inform research on the circadian system, most are poorly characterized molecularly (an exception being Ref. 9), and readers are referred elsewhere for a fuller description of them (6, 10 -15).
Central components of the transcription-translation-based circadian negative feedback loop in Neurospora include two negative elements, the proteins FREQUENCY (FRQ) (1,10,16,17) and FRQ-RELATED HELICASE (FRH) (18), and two positive elements, WHITE COLLAR-1 (WC-1) and WHITE COLLAR-2 (WC-2) (17). FRQ and FRH thus perform a function like that of the PERs and CRYs in the mammalian clock, and WC-1 and WC-2 form a unit and function like the BMAL1 and CLOCK/NPAS2 heterodimers in mammals, but more of this later. In addition to these core players, there are a number of proteins with supporting roles, the various kinases and phosphatases that modulate activities and interactions among the core as well as the elements of the protein degradation pathway that are required for the daily turnover of FRQ. Inherent to the oscillator is the daily self-sustained rhythm in the synthesis and turnover of frq mRNA and FRQ proteins, so progress of the Neurospora oscillator, like other eukaryotic clocks, is characterized by cycles in the levels of transcript and protein. Loss-of-function mutations in frq (10), wc-1 (17), or wc-2 (19) result in loss of circadian rhythmicity; partial loss-of-function mutations in wc-2 (19) or frq (10) can result in significant period length changes (yielding periods from 16 to 35 h), in addition to partial loss of temperature and nutritional compensation of the clock. Environmental signals such as changes in ambient light and temperature reset the Neurospora clock by changing the levels of frq mRNA and FRQ protein (20,21) as further outlined below.
frq Transcripts and FRQ Proteins, Negative Elements in the Cycle frq was identified by genetics screens in the early 1970s (22), and molecular dissection of the Neurospora circadian oscillator can be traced to the positional cloning of frq in the mid-1980s. Subsequent progress on the molecular basis of rhythmicity has largely been tied to identification and analysis of the genes and proteins that regulate its synthesis, action, and destruction. Early work on the frq transcripts based on the sequence of cDNA fragments (10) were consistent with a simple transcription unit, but inconsistencies in these data foreshadowed identification of the frq antisense message (23). This nearly 5-kb antisense transcript (qrf, for frq backwards) arises from the frq locus, beginning from a promoter 3Ј to the region necessary for complementation of frq null mutations. qrf appears not to encode a protein but is rhythmically expressed at low levels with a peak phase opposite to that of frq, is light-induced, and appears to play a role in ensuring precise entrainment to light/dark cues (23). frq also has one of the most complex transcription units seen in a lower eukaryote; frq transcripts arise from multiple promoters, and each frq transcript can be spliced in a complex manner that is influenced by ambient temperature (Fig. 1). In this way, temperature-influenced splicing determines whether long (989 amino acids) or short (889 amino acids) FRQ proteins are synthesized (24). Both forms are needed for robust rhythmicity. At temperatures in the low end of the physiological range (Ͻ22°C) predominantly short FRQ is used and less overall FRQ is needed, whereas at higher temperatures (Ͼ26°C) the large form and higher overall levels are used (25); strains expressing only one form show small but statistically significant differences in period length (25), although the biological significance of these differences is clouded by accompanying changes in dosage. Despite these distinguishable physiological functions, half-lives of the two FRQ isoforms are not different and no distinctions in molecular activities are known. The temperature-modulated translation is governed by upstream open reading frames in the 5Ј-untranslated region (UTR) 2 of frq (26,27). Although this complicated temperature regulation of forms and amounts does not play a role in temperature compensation, it does seem to help in keeping the phase of the rhythm steady across a temperature range. 3 Temperature compensation appears to derive from a balancing of synthesis and turnover of components, especially FRQ (28).
FRQ undergoes a daily cycle of phosphorylations that greatly influence its activities. As soon as FRQ appears, it is phosphorylated by one of several kinases including casein kinase I (CKI), CKII, and calcium/calmodulin-dependent kinase (29 -32), and this phosphorylation plays a central role in the feedback loop. Recent work has also shown that FRQ can associate with and be phosphorylated by the Neurospora checkpoint kinase 2 (PRD-4), an event that renders the circadian cycle sensitive to resetting by DNA-damaging agents (33). This suggests that phosphorylation of FRQ may be the common entry point for clock effects of a variety of environmental agents. The effects of phosphorylation are varied. It is apparent on reflection that for the feedback loop to cycle effectively, FRQ must turn over almost if it is not spliced, FRQ begins at AUG L , and if intron 2 is spliced, FRQ has a longer 5Ј-UTR and begins at AUG S . The end of the purple arrow marks the STOP codon for FRQ and is followed by a short 3Ј-UTR ending in the green arrow that marks the polyadenylation site. Just 3Ј to this are the transcription start sites for qrf, the antisense frq transcript (marked as a red line running right to left) whose point of termination within the promoter of sense frq is not known with certainty. A scale bar marks 1 kb. Adapted from Refs. 26 and 27. completely once per cycle, and the chief determinant of when this happens is its phosphorylation status (28,34). In this manner the kinases that act on FRQ play a major role in determining the period length of the clock, and alleles of FRQ that are more stable (such as FRQ7) (28,35) have long period lengths. Both kinase inhibitors and mutation of FRQ to alter normal sites of its phosphorylation can greatly increase the period of the cycles (32,34,36). Phosphorylation of FRQ can also influence its interactions with the white collar complex (WCC) (29). In general, the importance of FRQ phosphorylation to the operation of the clock has been established, but the phosphorylation pattern is quite complex and probably processive. There are a number of distinguishable phosphorylated forms of both FRQ proteins, and the determination of how each modification affects activities and interactions is an important goal of future work.

FRH, an RNA Helicase Essential for the Clock
The other required negative element, FRH, is likely to be an RNA helicase based on its sequence. This 1106-amino acid protein was identified as a protein that co-purifies with FRQ (18), and an RNA interference-mediated knockdown of its activity confirmed that it is essential for rhythmicity and also for life. A member of the SKI2 subfamily of RNA helicases, FRH contains a DEAD/DEAH box and helicase C domains, and its closest homolog is the Mtr4/Dob1 helicase from Saccharomyces, a component of the exosome complex of 3Ј-5Ј-exonucleases involved in RNA maturation and quality control (37,38). FRH is not a trivial or transient partner; nearly all FRQ in the cell is complexed with FRH, although much of the cell's FRH pool is not bound to FRQ. This FRQ-FRH complex (FFC) (7) is both nuclear and cytoplasmic and most FFC is not bound to the WCC. Next to nothing is known concerning the nature of the role of FRH, whether this entails RNA binding, or how it acts. So far the genetics of FRH is limited to the single knockdown strain, so further genetic analysis of this important protein, as well as the obvious experiments to see whether it is binding an RNA, will be of interest.

WC-1 and WC-2, Positive Elements in the Cycle
WC-1 and WC-2 are true transcription factors that contain acidic and polyglutamine transcriptional activation domains, GATA-type zinc finger DNA-binding domains, and the protein interaction domains known as PAS domains (39 -41). These proteins are found together as a complex consisting of one copy each of WC-1 and WC-2 (WCC) (41)(42)(43); WC-1 is predominantly nuclear but WC-2 is seen throughout the cell. Such heterodimeric complexes are typical of fungal and animal circadian feedback loops (1,17,44), and in fact the heteromeric complex in Neurospora was the first such to be described. The complex in mammals includes BMAL1 and CLOCK, and BMAL1, a sequence and functional homolog of WC-1 (39), plays an identical role in the cognate mammalian clock as noted above. The WCC can assemble and bind to the frq promoter in vitro (45)(46)(47) and is the chief regulator of frq expression; null mutants of either wc-1 or wc-2 display low levels of frq mRNA and FRQ (17,41,48,49). This heterodimer also functions as the principle photoreceptor in the cell by virtue of an FAD chromophore bound to WC-1 (45,49) and in that role, exclusive of any clock function, is responsible for light responses seen for several percent of the genome (50).
WC-1 is 1167 amino acids long (39,42) and is thought to always act with WC-2 as a complex. Like FRQ, the expression of WC-1 is complex, with multiple promoters and splice forms (51) 4 ; it shows a low amplitude rhythm in expression levels, but unlike FRQ, WC-1 expression is regulated post-transcriptionally (probably at the level of WCC assembly or turnover) through a complex mechanism involving FRQ (39,52). As a central player in the clock, WC-1 is regulated by a host of factors including phosphorylation (39,52,53), protein-protein interactions (41,43), and environmentally induced changes in secondary structure (45,49). WC-1 is needed for FRQ and WC-2 to interact and also self-associates to some degree (54). In addition to the two PAS domains that play a role in proteinprotein interactions, WC-1 contains a LOV domain, a subclass of PAS domains associated with proteins that sense light, oxygen, and voltage, and this domain mediates the activation of WC-1 activity by light (45). This light-induced activation of WC-1, which results in rapid and massive transcription of frq, provides the mechanism for clock resetting by light (20), and this resetting mechanism (light-mediated transcriptional induction of a negative element in the clock loop) is also used in mammals (55).
WC-2 has a transcriptional activation domain, a single PAS domain, and a zinc finger DNA-binding domain. More abundant than WC-1, WC-2 is predominantly in the nucleus (43), and the amount of protein does not appear to be acutely regulated (43,56). WC-2, however, mediates the interactions between the regulated components in the clock cycle by complexing with WC-1 and with FRQ, consistent with the model in which FRQ acts to depress the level of its own transcript by interfering with the activation of the frq gene by the WCC (19,36,43,46). WC-1 and FRQ do not interact in the absence of WC-2 nor is DNA required for the interactions. Partially functional allelic versions of WC-2 have been useful in dissecting the differing roles of the protein in light and clock regulation. Some WC-2 variants clearly deficient in promoting light-induced gene expression show little if any defects in the light induction of frq (48) indicating that the light signal transduction pathway cannot be as simple as WCC activation of all lightinduced genes.

How It All Works: Molecular Events in Neurospora Circadian Cycle
For almost two decades research on the Neurospora clock contributed to and has been informed by the model of eukaryotic circadian clocks in which the daily translation of proteins is central to the oscillator (e.g. Ref. 35); work on Neurospora was the first to experimentally test this model and prove the importance of daily transcription in this cycle (16). Although in broad outline the model is still valid, recent work is beginning to flesh out the mechanism in pleasing detail. Fig. 2 serves as a guide for following the events within the Neurospora free running clock cycle starting from the left, the subjective morning.
When we enter the story at subjective dawn (left side), the WCC is already bound to the Clock Box (C-box) in the frq promoter, a region containing two GATG repeats (45,46), and is actively driving frq transcription. These messages are processed in the complex manner described in Fig. 1. FRQ is translated with little lag, dimerizes, and assembles with FRH into the FFC (7), and moves to the nucleus (24,57,58). FRQ, in the FFC, participates in several interactions here, and these determine the kinetics of the circadian cycle. The first and dominant action is that the FFC acts to reduce the activity of the WCC in driving frq expression. Many studies documented physical interactions between FRQ and the WCC (e.g. Refs. 43, 58, and 59), and early models predicted a simple reduction of activity based on this interaction removing the WCC from the frq promoter (46). Recently, more quantitative analyses have shown that there is less FRQ than WC-1 in the nucleus, prompting a model in which the FFC promotes the phosphorylation of one or both elements of the WCC, thereby rendering them transcriptionally less active (36,60,61). Indeed, phosphorylation of both WC-1 and WC-2 in the dark requires FRQ, and phosphorylation of the WCC reduces its ability to bind to DNA. These data are thus consistent with a model in which the FFC acts transiently to recruit one or more kinases (probably CKI and CKII) (61) to phosphorylate the WCC. The kinetic details of this (whether the FFC interacts with free WCC that is in equilibrium with DNA-bound WCC or the FFC inter-acts with the WCC on the FRQ promoter) are not yet known; for Fig. 2 I flipped a coin and represent the first possibility. From its first appearance in the cell though, so long as FRQ is present WCC activity is reduced and the only way for activity to recover is for FRQ to disappear.
FRQ disappears through proteasomal degradation, and this too is triggered by its phosphorylation. FRQ seems like a kinase magnet: as soon as it appears, it begins to be phosphorylated by a host of kinases. Mutations in CKI (32), CKII (29,30), and checkpoint kinase 2 (33) result in substantial effects on period length, and FRQ has been shown to physically interact with CKI, PRD-4, and CAMK-1. It seems plausible that some kinases act processively, but which one(s) primes and which one(s) follows is not known; protein phosphatases also regulate the level of FRQ phosphorylation (62) although their effects on the clock may be more through their effects on WCC dephosphorylation. The kinetics of FRQ phosphorylation are perhaps the single most critical feature determining the period length of the clock; point mutations in single phosphorylated residues can shift the period from 22 to 35 h (30,34). Phosphorylation of FRQ reduces its affinity for the WCC (29) and also promotes its interaction with the WD-40 domain of FWD-1, the substrate recruiting subunit of an SCF-type ubiquitin (E3) ligase (63); FWD-1 is the ortholog of the Slimb protein that performs a similar function in the Drosophila clock (64). Eventually FWD-1 is recycled through the action of the COP9 signalosome (65). The end result of all this is a daily cycle in FRQ phosphorylation resulting in the precipitous turnover of FRQ around the middle of the night (24). When FRQ disappears, the phosphorylation-mediated inactivation of the WCC is reversed probably by protein phosphatase 2A (PP2A) (36) and the expression of frq (and the circadian cycle) begins anew.
Although the first and principal action of FRQ is its nuclear entry and inhibition of WCC activity, a third function of FRQ is in promoting the synthesis of its activators, WC-1 and WC-2 (39,41,46,59,66). A cytoplasmic reservoir of FRQ remains throughout the day eventually dwarfing the nuclear component, and within this fraction phosphorylation of FRQ at serines 885 and 887 allows FRQ to promote accumulation of WC-1, perhaps by fostering assembly of the WCC and thereby stabilizing WC-1 (52). The result is that despite constant expression of wc-1 mRNA there exists a low amplitude (and apparently dispensable) rhythm in WC-1 levels (39,59,66). A recent report notes, however, that the rhythm in WC-1 levels is FIGURE 2. Interlocked positive and negative feedback loops in the Neurospora clock. The top shows how the levels of frq mRNA, FRQ, and WC-1 proteins oscillate through time over one and a half cycles in constant darkness; subjective day (0 -12, light bar) and subjective night (12-24/0, dark bar) are shown. Below this, pertinent clock proteins and their helpers move between the cytoplasm and nucleus in the events of the circadian cycle. At time A in the morning, WC-1 and WC-2 are driving frq expression, and FRQ is actively translated, associates with FRH, moves to the nucleus, and associates with and leads to (around time B, afternoon) the phosphorylation of WC-1 and WC-2. This phosphorylation inactivates WC-1 and WC-2 so frq expression drops. Around the same time FRQ in the cytoplasm helps newly made WC-1 and WC-2 to associate. Moving toward time C, phosphorylated FRQ associates with FWD-1 and goes to proteasome (trash can) allowing old WC-1 and WC-2 to be dephosphorylated by PP2A and reactivated; new WC-1 and WC-2 move to the nucleus and join in reactivation of frq transcription in the late night. Several educated guesses are taken here; see text for details and caveats. Adapted from an idea of A. Claridge-Chang with permission. still seen in frq-null strains (9), so it seems likely that additional factors regulate WC-1. Expression of wc-2 mRNA is also promoted by nuclear FRQ (52,54,66), but WC-2 levels remain high and relatively constant (43) probably because of stability of the protein.