Cross-Talk between the Cellular Redox State and the Circadian System in Neurospora

The circadian system is composed of a number of feedback loops, and multiple feedback loops in the form of oscillators help to maintain stable rhythms. The filamentous fungus Neurospora crassa exhibits a circadian rhythm during asexual spore formation (conidiation banding) and has a major feedback loop that includes the FREQUENCY (FRQ)/WHITE COLLAR (WC) -1 and -2 oscillator (FWO). A mutation in superoxide dismutase (sod)-1, an antioxidant gene, causes a robust and stable circadian rhythm compared with that of wild-type (Wt). However, the mechanisms underlying the functions of reactive oxygen species (ROS) remain unknown. Here, we show that cellular ROS concentrations change in a circadian manner (ROS oscillation), and the amplitudes of ROS oscillation increase with each cycle and then become steady (ROS homeostasis). The ROS oscillation and homeostasis are produced by the ROS-destroying catalases (CATs) and ROS-generating NADPH oxidase (NOX). cat-1 is also induced by illumination, and it reduces ROS levels. Although ROS oscillation persists in the absence of frq, wc-1 or wc-2, its homeostasis is altered. Furthermore, genetic and biochemical evidence reveals that ROS concentration regulates the transcriptional function of WCC and a higher ROS concentration enhances conidiation banding. These findings suggest that the circadian system engages in cross-talk with the cellular redox state via ROS-regulatory factors.


Introduction
Biological processes, such as development and physiology are regulated via underlying circadian clock mechanisms that are synchronized with the daily light cycle of the Earth. The circadian rhythm (''circa'': around; ''diem: day) is composed of selfsustaining oscillations with a period of approximately 24 hr [1]. The circadian clock is composed of three elements: an input pathway, an oscillator and an output pathway. The main oscillator in circadian clocks is likely a transcription-translation feedback loop (TTFL) that is present in a diverse range of organisms from bacteria to humans [2]. A portion of the positive feedback mechanism in eukaryotes consists of transcription factors that form heterodimers via PER, ARNT and SIM (PAS) domains [3]. The input pathway resets the oscillator based on daily changes to external stimuli (entrainments), such as light and temperature; the oscillator can be entrained to periods around 24 hr. The oscillator then controls circadian rhythmic behaviors (various biological activities and aspects of physiology) via the output pathway [2]. Multiple feedback loops, not a single isolated loop, must be linked together to sustain a stable rhythm [4].
The filamentous fungus Neurospora crassa exhibits a circadian rhythm during asexual spore formation (conidiation). The rhythm has a period of approximately 22 hr in continuous darkness and this model organism is commonly used for studying circadian clock biology. Neurospora's oscillator is composed of the FRQ (FRE-QUENCY) and WC (WHITE COLLAR)-1 and WC-2 complex (WCC) and called FRQ/WCC oscillator (FWO). WC-1 and WC-2 formed a heterodimer via the PAS domain. The mechanisms underlying the function of FWO are well understood and comprise a TTFL. WCC, a transcription factor, binds to the Clock box (Cbox) in the frq promoter region and promotes frq expression [5,6]. The frq transcript oscillates in a circadian manner, and the levels reach a maximum at circadian time (CT) 4 and trough a minimum at CT 16 [7]. However, many studies have shown that frq-null mutants exhibit rhythmic conidiation under particular conditions such as the addition of farnesol and menadione to the medium as well as in chain elongation (cel), vivid (vvd) or sod-1 double mutants [8][9][10][11]. These data indicate the existence of other oscillators, such as FRQ-less oscillator (FLO) and WC-FLO [12][13][14][15]. Although the molecular mechanisms underlying the additional oscillators and their relationships with each other remain unknown, several observations have suggested the presence of a larger and more complex regulatory system [14]. These oscillators regulate the expression of clock-controlled genes (ccgs) via output pathways, including the osmotically sensitive (OS) MAPK pathway, to facilitate rhythmic conidiation [16].
Although the free-running rhythm in Neurospora has a period length of approximately 22 hr, the rhythm can be entrained to a daily (24 hr) cycle via light and temperature. WC-1 possesses a light-oxygen-voltage (LOV) domain, which is a type of the PAS domain and can bind flavin adenine dinucleotide (FAD) via LOV.
Binding to the C-box by WCC is enhanced by illumination and induces frq expression [5,17]. VIVID (VVD) is also a flavinbinding photo-receptor with a LOV domain, similar to WC-1 [18,19]. VVD is considered a negative regulator of WCC, preventing the transcriptional activation of WCC [20]. WCC also functions as a photoreceptor in the circadian clock and during light-inducible carotenoid synthesis [17,21]. Previously, we demonstrated the function of WCC accelerated light-induced carotenoid synthesis in both sod-1-deficient and ROS-treated cells [22,23].
ROS include singlet oxygen, superoxide (O 2 2 ), hydrogen peroxide (H 2 O 2 ) and hydroxyl radicals. Cellular ROS are passively generated by the mitochondria via respiration, as well as metabolism, and are actively generated by NADPH oxidase (NOX) [24]. Excess ROS are rapidly detoxified by ROS scavengers. In many organisms, singlet oxygen is removed by antioxidants such as carotenoids; O 2 2 is reduced to H 2 O 2 by SOD; and H 2 O 2 is reduced to water by catalase (CAT). Our observations suggested that cellular ROS affect WCC-mediated carotenogenesis. Because WCC is a photoreceptor, transcription factor, and a positive element in the circadian clock, we examined whether ROS affect circadian rhythms. Neurospora with sod-1 mutations demonstrate robust conidial banding compared with the wild type (Wt) and are hyper-sensitive to light entrainment [10,25]. Moreover, rhythmic conidiation is observed in Neurospora with both sod-1 and frq mutations [10], which suggests that cellular ROS modulate circadian output via FWO and FLO and stabilize conidial banding. In this study, we measured cellular ROS levels during conidiation banding to examine the behavior of cellular ROS, and we observed circadian oscillations of intracellular ROS during growth. We then investigated the role of cellular ROS in the circadian system and attempted to identify the mechanisms that manage the cellular ROS level.

Cellular ROS profiles in Neurospora
To examine the cellular ROS behavior during the daily cycle, we directly harvested the growth front in the race tube and measured the level of intracellular ROS because cellular ROS concentrations depend on the culture conditions (see Materials and Methods, Fig. S1). The liquid culture assay is the conventional method used to analyze molecular profiles in Neurospora clock studies. We checked frq rhythmic expression as methodological control. To measure cellular ROS levels, we used a lucigenin chemiluminescence assay, the results of which were supported using additional assays. The lucigenin chemiluminescence assay was primarily used to detect ROS generation. In general, this assay detects superoxide radicals in a cell [26]. However, H 2 O 2 can also be detected depending on the conditions (Fig. S2) [27]. Hence, the lucigenin assay detects superoxide radicals and H 2 O 2 . We used a band (bd) mutant that exhibits clear conidial banding and has been used as a wild type in Neurospora circadian clock studies [28] to first measure ROS levels. The bd mutation is a point mutation in ras-1, and we indicate the bd mutants as ras-1 bd in this paper [25]. The cellular ROS concentration in ras-1 bd mutants displayed a circadian oscillation with a period length of approximately 22 hr; the amplitude of this oscillation gradually increased with each cycle and stabilized after 3 cycles (Fig. 1A, B and Fig. S3, S4, S5). The peak corresponded to , and the trough corresponded to CT 6.
To confirm higher ROS levels in the sod-1 mutants during conidiation banding, we assessed cellular ROS levels at the minimum (CT 6) and maximum (CT 18) points during the ROS oscillation. During the initial growth, both the sod-1 and the ras-1 bd mutants displayed ROS oscillations with a higher amplitude than those of the Wt (Fig. 1C). However, each maximum level in the sod-1 and ras-1 bd mutants converged to a constant level after 3 cycles (Fig. 1D). The results obtained from ras-1 bd suggest that RAS signaling is involved in cellular ROS generation. A standard Wt strain was used as a background strain in subsequent experiments to minimize the complexity of the results. We also examined temporal ROS levels in the frq 10 , Dwc-1 and Dwc-2 mutants, which are arrhythmic mutants, to determine whether ROS oscillation requires a circadian clock. Although frq 10 , Dwc-1 and Dwc-2 mutants demonstrated ROS oscillation, the ROS levels in the frq 10 at CT 6 and 18 were half of those of the Wt, and the amplitudes in Dwc-1 and Dwc-2 were slightly weaker than in the Wt (Fig. 1E).
Environmental cues, such as light and temperature, reset the circadian rhythm to a period of approximately 24 hr. In fish and flies, the cellular ROS level depends on the presence of light and varies with the light/dark (LD) cycle [29,30]. In Neurospora, the cellular ROS levels in the mycelium were reduced by light at CT 18, but not by temperature changes (Fig. S6). The ROS levels transiently increased after 5-10 min of illumination and then decreased after 60 min of illumination (Fig. 1F). The transient induction of ROS may be generated from intracellular photosensitizers, such as flavin [31]. Under photoperiodic conditions, continuous light attenuated the increase in ROS levels. ROS levels increased again after approximately 10 hr of illumination but then underwent a phase shift when the sample was placed in the dark (Fig. 1G). These results suggest that cellular ROS levels are controlled by circadian clocks and illumination.

Cellular ROS levels are managed via circadian clocks using catalase
To understand how cellular ROS levels are managed, we next examined the cellular antioxidants present in cells exposed to light. In Neurospora, of the four CAT genes (CAT-1, CAT-2, CAT-3 and CCT-1/CAT-4), CAT-1 is the only isoform that is light-inducible [32,33]. Cellular CAT activity increased slightly after illumination ( Fig. S7), especially CAT-1 activity, and cat-1 transcripts were significantly induced at CT 18 when the mycelia were exposed to light ( Fig. 2A & B). Furthermore, we speculated that the ROS oscillation observed under free-running conditions might also be produced by CAT-1. To test this, we evaluated cat-1 mRNA levels using a quantitative real-time PCR (qRT-PCR) assay. cat-1 transcripts from Wt cells oscillated with a peak at CT 24/0; the amplitude of the cat-1 oscillation in the frq 10 mutants was higher than that in the Wt (Fig. 2C). Furthermore, the cat-1 transcripts from the Dwc-1 and Dwc-2 mutants were arrhythmic and much lower compared with the Wt (Fig. 2D). These results suggest that wc-1 and wc-2 up-regulate the expression of cat-1. Cellular CAT activity also oscillated slightly with a peak at CT 12 ( Fig. 2E). Interestingly, the phase for cellular CAT activity was out of sync with the phase for the cat-1 transcripts. This result suggests that CAT family members other than CAT-1 may mask CAT-1 activity after cell lysis. Cellular CAT activity levels at CT 12, the maximum during the CAT activity oscillation, decreased with each cycle and then stabilized at lower levels (Fig. 2F). The increase and ensuing stability of the maximum ROS levels during each cycle may result from a decrease in cellular CAT during each cycle. In frq 10 mutants, the CAT activity levels were higher than in the Wt, which is consistent with the changes in ROS levels and cat-1 expression (Fig. 2G). However, the cat-1 levels in the Dwc-1 and Dwc-2 mutants were not consistent with the results from testing cellular catalase activity and cellular ROS levels (Fig. 2G). These observations suggest that additional CATs may supplement the cellular peroxidative capacity when cat-1 transcripts become low in the absence of wc-1 and wc-2.
To examine whether the ROS-destroying function of CAT affects cellular ROS levels and the circadian rhythm, we inhibited all CATs using 3-AT (3-amino-1,2,4-triazole). Wild-type samples that were treated with 3-AT displayed a robust conidial banding, similar to that of the sod-1 mutant, and high ROS levels (Fig. S8).
These results indicated that cellular CATs are involved in the conidial banding and cellular ROS levels. We also isolated a cat-1 loss-of-function mutant (cat-1 RIP ) to examine the specific effects of CAT-1 [34]. The transient increase in ROS after illumination was similar in Wt and cat-1 RIP cells, although the ROS levels in cat-1 RIP were slightly reduced compared with those of the Wt and decreased more slowly (Fig. 3A). In addition to CAT-1, mycelial carotenoids are induced by light and may eliminate cellular ROS [35]. Thus, the induction of carotenoids rather than CAT-1 could reduce the ROS levels. ROS levels were significantly higher in the sod-1 mutants (Fig. 3A). With respect to the cellular ROS oscillation, the levels in cat-1 RIP at CT 6 were higher than those in the Wt, and the Cellular ROS levels in the growth fronts from the race tube were measured using a lucigenin chemiluminescence assay. frq and ß-tub transcripts were detected using semi-quantitative RT-PCR and were used as experimental controls. (B) Cellular ROS generation at CT 18 in subsequent cycles for ras-1 bd mutants. Mycelia were harvested at CT 18 during each cycle (from DD 27.5 to 225.5). (C) ROS generation in mycelia at the minimum (CT 6) and maximum levels (CT 18) for the Wt as well as the ras-1 bd and sod-1 mutants. (D) Cellular ROS generation in mycelia at CT 18 in subsequent cycles for the Wt as well as the ras-1 bd and sod-1 mutants. (E) Cellular ROS generation in mycelia for the Wt and the clock mutants (frq 10 , Dwc-1 and Dwc-2) at CT 6 and 18. (F) Effect of brief light exposure on cellular ROS levels. Mycelia in the race tube were exposed to light at CT 18 (DD 27.5) for 5, 10, 15, 30, 45, 60 and 90 min. (G) Effects of the LD cycle on cellular ROS levels. Mycelia in the race tube were exposed to light for 12 hr at CT 24 (DD 11) and then transferred to dark conditions. Cellular ROS levels were measured in 3-h intervals. CT = circadian time, ZT = Zeitgeber time. All values are shown as mean 6 standard error (SEM). doi:10.1371/journal.pone.0028227.g001 oscillations were moderate compared with those of the Wt (Fig. 3B). The maximum ROS levels at CT 18 for each subsequent cycle were not significantly different between Wt and cat-1 RIP cells (Fig. S9). This result suggested that cat-1 is involved in ROS oscillation but not in homeostasis. Conidial banding and frq expression in cat-1 RIP was normal in DD after light-resetting and light-entrainment (LD) (Fig.  S10, S11, S12). These results indicate that the lack of ROS oscillation does not affect the circadian rhythms.

Cellular ROS are mainly produced by NADPH oxidase
We examined NADPH oxidase (NOX), an active ROS generator. In Neurospora, two NOXs (NOX-1 and NOX-2) have been isolated [24]. Dnox-1 mutants had a greater reduction in cellular NOX activity, low ROS levels and a loss of cellular ROS oscillation (Fig. 4A, B and Fig. S13). For conidiation banding, Dnox-1 mutants displayed an arrhythmic phenotype under freerunning conditions and indeterminate conidiation in an LD cycle compared to the Wt (Fig. 4C, Fig. S14 and S15). The expression of light-inducible genes, such as frq and albinos (als), was elevated in sod-1 mutants compared with in the Wt [10,22]. In contrast, lower ROS levels may reduce the level of light-induced frq transcripts because cellular ROS levels in the Dnox-1 mutants were low during the transient ROS increase (Fig. 4D). As we expected, the lightinduced frq level in the Dnox-1 mutants was lower than in the Wt Catalase activity during illumination, as measured using an in-gel assay. Mycelia were exposed to light at CT 18 (DD 49.5). CAT-1 was detected as CAT-1a (reduced form) and CAT-1c (oxidized form) because of the oxidative conditions in catalases. (C) Temporal cat-1 expression in the race tube growth fronts from the Wt and frq 10 mutants as determined by qRT-PCR. (D) cat-1 mRNA accumulation in the race tube growth fronts from the Wt, frq 10 , Dwc-1 and Dwc-2 mutants at CT 12 (minimum) and 24 (maximum), as assayed by northern blot analysis. (E) Temporal cellular catalase activity in Wt cells. Cellular catalase activity was determined using a spectrometric assay. (F) Cellular catalase activity at CT 12 (maximum) in subsequent cycles for the Wt, as measured by a spectrometric assay. (G) Cellular catalase activity at CT 12 (maximum) and 24 (minimum) in the Wt and frq 10 , as well as the Dwc-1 and Dwc-2 mutants, as measured by a spectrometric assay. All values, more than three replicates, are shown as mean 6 standard error (SEM). doi:10.1371/journal.pone.0028227.g002 (Fig. 4E). Cellular NOX activity showed no circadian rhythmicity: it slightly decreased during the initial growth and then subsequently stabilized (Fig. S16). NOX activity was slightly altered in the frq 10 , Dwc-1 and wc-2 mutants (Fig. 4F). These results suggest that NOX can be regulated by the circadian system to generate cellular ROS.

Redox state can affect the function of WCC
We next addressed the way that cellular ROS regulate the circadian clock. Because light-inducible genes, such as frq, are transcribed by WCC, ROS may regulate light-inducible gene expression via WCC and the transcriptional function of WCC in a free-running rhythm [10,22,23]. To examine this possibility, we measured frq expression under free-running conditions. The amplitude of the rhythmic frq expression in sod-1 mutants was enhanced compared with the Wt (Fig. 5A). Furthermore, the frq mRNA levels in a low ROS environment, such as in Dnox-1 mutants and in cells treated with the antioxidant N-acetyl-Lcysteine (NAC) were reduced compared with those in Wt mycelia grown on untreated medium (Fig. 5A). These results suggest that cellular ROS conditions affect the expression of frq transcripts through WCC.
WCC binds at the frq promoter region, which contains a C-box and promotes frq expression [5,6]. This binding is enhanced in the presence of light, both in vitro and in vivo [5]. Purified recombinant WC-1 and WC-2 proteins were tested using an electrophoretic mobility-shift assay (EMSA) to examine the effects of ROS on their binding activities. Because these recombinant proteins rapidly formed inclusion bodies after purification under normal conditions and could not be utilized for EMSAs, the purification was performed under dim red light and with the addition of the reducing reagent dithiothreitol (DTT) to prevent inactivation. DTT was removed using dialysis just before the EMSA was performed. Binding of WCC to the C-box was induced in the presence of H 2 O 2 despite the absence of light, although the levels were 2-fold less than under illumination ( Fig. 5B and Fig. S17). Flavoproteins such as VVD form flavin-cysteinyl adducts in response to light exposure and show light-dependent absorbance changes [18]. Recombinant WC-1 protein, isolated from E. coli, displayed light-induced absorbance changes (Fig. 5C). The lightinduced absorbance of WC-1 occurred in the presence of 1 mM H 2 O 2 despite the darkness, were reduced further in a concentration-dependent manner (Fig. 5C). Thus, ROS may promote flavin-cysteinyl adduct formation and affect the binding activity of WCC. ROS may be able to mimic the activating function of light on WC, even in the absence of light. However, the excess ROS prevented hetero-dimerization between WC-1 and WC-2 ( Fig. 5D and Fig. S17). These results suggest that increased ROS (0.5-1 mM H 2 O 2 ) may promote the function of WCC, but hyperoxidative (1-10 mM H 2 O 2 ) and low oxidative states (less than 0.5 mM H 2 O 2 ) can inhibit its function.

Discussion
In this study, under free-running conditions, we found two important characteristics of cellular ROS profiles. First, cellular ROS concentrations oscillate in a circadian manner (ROS oscillation). Second, the amplitude of the oscillation increases gradually with each subsequent cycle and then stabilizes (ROS homeostasis). Presumably, cellular ROS levels are maintained via a balance between the generation and the destruction of ROS. Our data indicate that most of the cellular ROS are constitutively supplied by NOX-1. With respect to ROS destruction, Neurospora constitutively may expresses sod-1 transcripts that can remove excess ROS at any time. In contrast, CAT-1 and carotenoids are conditionally regulated by light, osmotic stress and circadian clocks [33,36]. These findings suggest that SOD constitutively destroys NOX-generated ROS to a basal level, and that ROS are transiently eliminated by the rhythmic CAT-1. Furthermore, the amplitude of cellular CAT activity oscillation (including from CAT-2, CAT-3 and CCT-1/CAT-4, in addition to CAT-1) gradually decreases with each subsequent cycle. These decreases may gradually induce cellular ROS levels and maintain ROS homeostasis.
Our data suggest that cellular CAT and NOX enzymatic activity can be regulated by the circadian system. With respect to transcription levels, although cat-1 oscillation persists in the absence of frq, the amplitude is stronger than in the Wt. Interestingly, the rhythmic cat-1 expression disappears in the Dwc-1 and Dwc-2 mutants, unlike in the frq 10 mutants. A typical clock controlled gene, ccg-1, is regulated by FWO via an OS MAPK pathway [16]. cat-1 is also regulated by the OS MAPK cascade [36], and FWO may control cat-1 expression via this pathway. These results suggest that WCC may regulate the transcription of cat-1 via the OS MAPK pathway and produce rhythmic cat-1 expression.
The circadian system may directly or indirectly sense cellular ROS signals to control the cellular ROS levels. Genetic data suggest that increased ROS levels (as shown in sod-1 mutants) enhance conidiation banding and rhythmic frq expression, whereas lower ROS levels (as in response to antioxidant treatment and in Dnox-1 mutants) inhibit these effects (Fig. S18). Furthermore, in vitro analyses suggest that the transcriptional function of WCC depends on ROS levels. In these assays, H 2 O 2 was used as a ROS and 0.1-10 mM H 2 O 2 effectively modulated the function of WCC. Physiological H 2 O 2 concentrations in Neurospora were in the range of 0.5-3.5 mM during initial growth (Table S1). These results suggest that ROS, specifically H 2 O 2 , affects the function of WCC in cells. The PAS domains in the Aer receptor function as redox sensors in E. coli [37]. Biochemical analyses of VVD, a PAS-LOV protein in Neurospora, have suggested that the oxidation of VVD causes conformational changes required for its function [38]. WC-1 and WC-2, which possess PAS or LOV domains, may also be able to sense the cellular redox state via these domains. In Neurospora, increased FWO amplitudes lead to a robust circadian rhythm [39]. One possible reason for the robust conidiation observed in sod-1 mutants may be that the increased maximum ROS levels (CT 18) activate WCC and promote the accumulation of frq transcripts, which could lead to robust conidial banding. Meanwhile, excess ROS disassemble the hetero-complex of WC-1 and WC-2. This ROS-dependent WCC regulation may sustain ROS homeostasis. frq, wc-1 and wc-2 genetically down-regulate cellular CAT activity (Fig. 2G). The excess ROS-inhibited WCC could reduce the ROS levels via up-regulation of CATs and maintain the concentrations of ROS. As shown in the ras-1 bd and sod-1 mutants in Fig. 1D, the different ROS levels during the initial growth may stabilize to a single level via this mechanism. In general, excess ROS levels are toxic to the cell. However, redox signaling, in which ROS function as signaling molecules, has been studied in the cell from biological, physiological and pathological perspectives [40,41]. In Neurospora, ROS regulate morphogenesis and conidiation banding, including mycelial and conidial development, and function as cellular signal factors independent of the circadian clock [10,42,43]. Therefore, a balance in cellular ROS is important for cell development.
Genetic evidence suggests that WCC up-regulates the transcription of cat-1, and then cellular ROS control the functions of WCC. Based on these findings, in the absence of the frq gene, a feedback loop composed of WCC and CAT-1 via ROS oscillation may be a candidate for a WC-FLO. During ROS oscillation, the high ROS levels may promote cat-1 expression via WCC and reduce cellular ROS levels. Thereafter, the low ROS levels stop inducing cat-1 via WCC, and ROS levels rise again. However, in the presence of frq, WCC is negatively regulated by FRQ at CT 6, despite the high levels of ROS, and the cellular H 2 O 2 at CT 6 is 0.5 mM in Wt cells, which is sufficient to activate WCC in vitro (Table S1). Indeed, cat-1 RIP , which lacks ROS oscillation, demonstrates normal rhythmic frq expression. In the presence of frq, maximal ROS levels would be important for these circadian behaviors in WCC.
Is ROS elimination by CAT-1 useless for the circadian system? In the mycelium, cat-1 is induced by light and then rapidly reduces the ROS levels, counteracting the ROS generation. If ROS activates WCC, light-dependent ROS behavior as shown in Fig. 1G, may be explained because the activation of WCC should shift with photoperiodic conditions. However, the gradual decrease of light-dependent ROS levels in cat-1 RIP was insufficient to influence the conidiation rhythm and light-induced frq expression. Meanwhile, the sod-1 mutant, which has significantly higher ROS levels before illumination, showed high light-induced frq expression, and Dnox-1, with low ROS levels, showed lower frq expression than the Wt [10]. It has been proposed that VVD senses the cellular oxidative state under dark conditions and can change its conformation in response [38]. The ROS levels before irradiation may affect light-inducible frq expression via WCC. In any case, independent roles of cat-1 and decreasing ROS concentrations could not be elucidated for the circadian rhythm as shown in the phenotypes observed in cat-1 RIP . CAT-1 is synthesized in the growth front prior to conidial formation and may transiently reduce cellular ROS levels before conidium formation, as CAT-1 is predominantly localized to the conidium to sustain viability during germination [34]. Originally, oscillators were thought to be non-transcriptional, such as the phosphorylation cycle in cyanobacteria [44]. In Neurospora, the roles of cat-1 and loss of ROS may have substituted over the course of evolution for those of frq, TTFL genetic oscillator.
Our data indicate that, in Neurospora, the cellular redox state derived from ROS and the circadian system genetically and biochemically influence each other, leading to the stabilities of cellular redox state and conidiation banding. However, we do not show data suggesting that the cellular ROS are involved in the phase and period of circadian rhythm. Cellular redox state may not specifically control the circadian clock, but this state may contribute to an environment in which TTFL can properly regulate this process. Many studies have reported that ROS are closely linked to the circadian rhythm. For instance, in humans, the cellular redox state of NAD cofactors regulates the DNA binding activity of clock complexes such as NPAS2:BMAL1 [45]. In cyanobacteria, the circadian clock component LdpA senses the cellular redox state [46]. Recently, it was demonstrated in human red blood cells and Osteococcus tauri that oxidation of peroxiredoxins is involved in a TTFL-independent circadian rhythm. Genetic evidence reveals that the TTFL and the redox rhythm influence each other [47,48]. In Neurospora, ROS oscillation may be a TTFLindependent circadian rhythm, but further investigation is required to improve our understanding of this process. These studies indicate that the cellular redox state can be involved in the circadian system in diverse organisms.
For conidiation banding cultures, which were used as the basis for measuring ROS, for detecting transcripts and in CAT assays, an acrylic race tube (H1 6 W50 6 D1 cm; the lid is a removable) was prepared, and the growing mycelial front was harvested. The clock medium (Vogel's salt mixture, 0.1% glucose, 0.17% Larginine, 50 ng/ml biotin and 1.5% agar) was overlaid with a sterilized dialysis membrane to allow easy separation of the mycelium from the medium. The conidial suspension was inoculated onto the edge of the medium. The tube was exposed to light (20 mE m 22 s 21 ) for 12 hr and then kept in constant darkness. In the race tube assay, the conidia of each strain were harvested under a dim red light and inoculated onto the edge of the medium (clock medium) in a race tube [10].

ROS measurements
The ROS assay was performed as previously described [25]. Briefly, after culturing, a 5 mm region from the ras-1 bd and sod-1 mutants, a 9 mm region from the Wt and cat-1 RIP and a 7 mm region from the frq 10 , Dwc-1, Dwc-2 and Dnox-1 mutants were excised from the front and quickly frozen in liquid nitrogen. The size of the excised region was proportional to the growth rate of the strain. Harvesting was performed under a dim red light. The frozen samples were preserved at 280uC until analysis. Two fragments of mycelial mat were immersed in 0.5 ml of 0.2 mM lucigenin for 2 min, and the dialysis membrane was removed after vortexing for 5 sec. Luminescence, assessed as relative light units (RLUs), was measured for 20 sec in a Gene Light 55 GL-100A luminometer (MICROTEC CO., LTD, Funadashi City, Chiba, Japan). After the measurements were taken, the mycelium was dried and weighed. Luminescence values (RLU/dry weight mg) were normalized to the dry weight.

Electrophoretic mobility-shift assay (EMSA)
EMSA was performed as previously described, with some modifications [5,6]. Briefly, two oligonucleotides were used as Cbox probes: c-box/for (5-gggcgtcctgatgccgctgcaagaccgatgacgctgcaaaattgagatcta-3) and c-box/rev (5-gggtagatctcaattttgcagcgtcatcgg tcttgcagcggcatcaggacg-3). The oligonucleotides were annealed and end-labeled using the BcaBEST TM Labeling kit and 32 P-labeled dCTP (TaKaRa). Excess dCTP in each 32 P-labeled probe sample was removed using a Micro Bio-Spin P-30 Tris Chromatography Column (Bio-Rad), and the probe was adjusted to 18 nM. WC-1 (0.2 mg) or WC-2 (0.1 mg) was added to 16 binding buffer, 80 mM KCl, 0.1 mg poly(dI-dC), 1 ml probe, 0.5 ml rabbit serum, and 1 mM FAD in 20 ml and incubated on ice for 30 min. The reactions were loaded into 4% polyacrylamide gels containing 0.56TBE and 2.5% (v/v) glycerol at 4uC. After electrophoresis, the radioisotopic signals were visualized using autoradiography. The binding reaction and electrophoresis were performed under a dim red light.

Catalase activity assay
Ten mycelial fragments were disrupted in 10 mM potassium phosphate buffer (pH 7.0) by homogenization and sonication. The homogenates were centrifuged at 12,0006g for 10 min at 4uC, and the supernatant (50 ml; 50 mg protein) was added to 1 ml of reaction solution (50 mM potassium phosphate buffer, pH 7.8, and 10 mM H 2 O 2 ). The decrease in absorbance (O.D. 240) after 1 min was measured using a spectrophotometer. The units are defined as 1 mmol H 2 O 2 reduced per min per mg protein. For the in-gel assay, 20 mycelial fragments were disrupted in 10 mM potassium phosphate buffer at pH 7.0. The in-gel assay was performed as previously described [34].
NADPH oxidase activity assay NOX activity was evaluated as superoxide production by lucigenin-enhanced chemiluminescence [50]. Ten mycelial fragments were disrupted in 0.2 ml of extraction buffer (20 mM sodium phosphate buffer (pH 7.0), 1 mM EDTA, 1 mM PMSF, 0.5 mM leupeptin and 0.5 mM pepstatin) by sonication on ice. The homogenates (5 ml) were added to 0.5 ml of assay buffer (50 mM sodium phosphate buffer (pH 7.0), 1 mM EDTA, 150 mM sucrose and 50 mM lucigenin). After the addition of 0.1 mM NADPH, luminescence was measured as RLUs at 1-min intervals for 20 sec in a Gene Light 55 GL-100A luminometer (MICROTEC CO., LTD, Funadashi City, Chiba, Japan). NOX activity was indicated as RLUs per minute per mg of protein.

Spectroscopic analysis
Purified recombinant WC-1 protein was exposed to light (20 mE m 22 s 21 ) for 3 min or treated with 0.5 or 1 mM H 2 O 2 for 3 min in the dark. Absorbance spectra were detected using a Beckman DU 530 Spectrophotometer (Beckman Coulter, Inc.).

Pull-down assay
Recombinant GST-WC-1 and WC-2 were purified as described above. GST-WC-1 (1.5 mg) and WC-2 (0.5 mg) were incubated in 50 ml of IP buffer (20 mM HEPES, pH 7.4, 50 mM KCl, 2 mM MgCl 2 , 0.5 mM EDTA, 0.5% NP40 and 10% glycerol) with 10 ml of Glutathione Sepharose 4B (GE Healthcare) and 1 mM DTT or 0.005 to 10 mM H 2 O 2 at 4uC for 15 min in the dark and centrifuged at 1,0006g for 5 min at 4uC. The pellets were washed five times with 1 mL of IP buffer with 1 mM DTT or 0.005 mM H 2 O 2 and then suspended in SDS sample buffer. The final pellets were analyzed using SDS-PAGE (8%), and stained with CBB. The binding reaction and electrophoresis were performed under a dim red light.

Statistical analysis
All experiments were performed at least three times independently. Data are expressed as mean plus or minus standard error of the mean (SEM). Data indicating significant differences were compared using a paired Student's t test. Differences were considered significant when P,0.05.

Supporting Information
Figure S1 Cellular ROS generation in response to growth conditions and culturing methods. (A) Wt conidia were inoculated into liquid clock medium (immersed in liquid medium) or onto sterilized dialysis membranes on solid clock medium (exposed to air) in petri dishes. The dishes were incubated at 25uC for 24 hr under constant illumination (20 mE m 22 s 21 ). (B) ROS generation in the mycelia immersed in liquid medium and exposed to air. The mycelia were harvested and cellular ROS levels were measured using a lucigenin chemiluminescence assay. All values are shown as mean 6 standard error (SEM). (C) Conidial banding in ras-1 bd mutants in an acrylic race tube (H1 6 W50 6 D1 cm).  Figure S6 Effects of entrainment on cellular ROS. Wt race tube mycelia were exposed to light at CT 6 and 18, and the temperature was shifted from 25uC to 35uC or 15uC for 1 hr. ROS levels were measured using the lucigenin chemiluminescence assay. All values are shown as mean 6 standard error (SEM). (DOC) Figure S7 The effect of light on total CAT activity. Mycelia (Wt) in the race tube growth front at CT 18 were exposed to light for 1 hr. The total CAT activity was determined using a spectrometric assay. All values are shown as mean 6 standard error (SEM). (DOC) Figure S8 The effect of the CAT inhibitor 3-AT on conidiation banding. (A) Total CAT activity in mycelia (Wt) grown in the medium containing 3-AT. Mycelia in the race tube growth front at CT 6 (DD 16.5 hr) were harvested, and the total CAT activity was determined using a spectrometric assay. (B) Cellular ROS and H 2 O 2 levels in mycelia grown on medium containing 3-AT. Mycelia in the race tube growth front were harvested at CT 18 (DD 16.5 hr). Cellular ROS and H 2 O 2 levels were determined using the lucigenin chemiluminescence assay and a hydrogen peroxide assay, respectively. To determine H 2 O 2 levels, relative values were calculated based on the values obtained for the control. Treatment with 3-AT inhibited total CAT activity and caused an increase in cellular ROS and H 2 O 2 levels. (C) Conidiation banding in Wt cells on medium containing 3-AT. All values are shown as mean 6 standard error (SEM). (DOC) Figure S9 Cellular ROS generation at CT 18 in subsequent cycles in Wt and cat-1 RIP cells. Mycelia in the race tube growth front were harvested at CT 18 of every cycle (DD 27.5, 49.5 and 71.5 hr), and cellular ROS levels were determined using the lucigenin chemiluminescence assay. All values are shown as mean 6 standard error (SEM). (DOC) Figure S10 Conidiation banding in a cat-1 loss-of-function mutant. The race tubes were exposed to light (light-resetting) for 12 hr after inoculation and then transferred to constant darkness (DD). The growth front was marked at 24-hr intervals. (DOC) Figure S11 Conidiation banding in a cat-1 loss-of-function mutant under a 12-h light/12-h dark cycle. The growth front was marked at 24-h intervals (n = 12). (DOC) Figure S12 Clock-controlled and light-induced frq expression. (A) frq mRNA accumulation in race tube growth fronts of Wt and cat-1 RIP cells at CT 4 and 16. frq transcripts were detected using RT-PCR with ß-tub as the quantitation control. (B) Light-induced frq mRNA accumulation in the Wt and cat-1 RIP . Mycelia in race tube growth fronts were exposed to light for 15 min (L15) and 60 min (L60) at CT 24/0 (DD 11 h). frq transcripts were detected using RT-PCR with ß-tub as the quantitation control. All values are shown as mean 6 standard error (SEM) (see Methods S1). (DOC) Figure S13 Control NOX activity assay. Mycelia harvested at CT 18 were disrupted and used as cell lysates in the assay. The antioxidants SOD (150 units/ml) or CAT-1 (30 mg/ml) were added to the assay buffer. Cell lysates and NADPH were added to the assay buffer at the points noted using closed and open arrows, respectively. Temporal ROS generation (RLU/mg protein) by NOX was detected using lucigenin chemiluminescence. This experiment indicates that ROS generated in this assay are superoxide radicals, but not H 2 O 2 and that NADPH-dependent superoxide generation is due to NOX activity (see Methods S1).