Adult Circadian Behavior in Drosophila Requires Developmental Expression of cycle, But Not period

Circadian clocks have evolved as internal time keeping mechanisms that allow anticipation of daily environmental changes and organization of a daily program of physiological and behavioral rhythms. To better examine the mechanisms underlying circadian clocks in animals and to ask whether clock gene expression and function during development affected subsequent daily time keeping in the adult, we used the genetic tools available in Drosophila to conditionally manipulate the function of the CYCLE component of the positive regulator CLOCK/CYCLE (CLK/CYC) or its negative feedback inhibitor PERIOD (PER). Differential manipulation of clock function during development and in adulthood indicated that there is no developmental requirement for either a running clock mechanism or expression of per. However, conditional suppression of CLK/CYC activity either via per over-expression or cyc depletion during metamorphosis resulted in persistent arrhythmic behavior in the adult. Two distinct mechanisms were identified that may contribute to this developmental function of CLK/CYC and both involve the ventral lateral clock neurons (LNvs) that are crucial to circadian control of locomotor behavior: (1) selective depletion of cyc expression in the LNvs resulted in abnormal peptidergic small-LNv dorsal projections, and (2) PER expression rhythms in the adult LNvs appeared to be affected by developmental inhibition of CLK/CYC activity. Given the conservation of clock genes and circuits among animals, this study provides a rationale for investigating a possible similar developmental role of the homologous mammalian CLOCK/BMAL1 complex.


Introduction
Circadian clocks are internal daily time keeping mechanisms that allow organisms to anticipate daily environmental rhythms as well as efficiently organize behavioral and physiological functions in a daily schedule. The molecular mechanisms that form the basis for circadian rhythmicity in animals involve interlocked feedback loops controlling gene expression as well as post-translational activities [1,2]. In both insects and mammals a circadian transcription complex of two basic helix-loop-helix PAS domain transcription factors promotes the rhythmic expression of several of its negative feedback regulators. The fruit fly Drosophila melanogaster has emerged as a model system for animal circadian clocks that is both successful and representative. In the clock-bearing cells of Drosophila CLOCK/ CYCLE (CLK/CYC) acts as the central circadian transcription complex and induces peak expression of a set of transcripts including those for the negative feedback regulators period (per), timeless (tim), vrille (vri), and clock work orange (cwo) just after dusk [3][4][5][6][7][8][9]. PER and TIM proteins form a complex with the casein kinase 1e ortholog DOUBLETIME (DBT), in which TIM helps protect PER from destabilization by DBT-mediated phosphorylation [10][11][12]. PER-containing complexes enter the nucleus around midnight and trigger repression of CLK/CYC activity [5,[13][14][15][16], VRI acts as a transcriptional repressor for the Clk gene [9,17], and CWO reduces CLK/CYC activity by competitively binding CLK/CYC-regulated promoter elements [4,7,8].
The circadian clock circuits are linked to synchronizing input pathways as well as output pathways that signal time-of-day information to downstream biological functions. The extensive interconnectedness of the molecular circadian cycle complicates identification of the order of its events. We reasoned that the development of transgenic flies with conditional circadian clock function, in which the circadian cycle could be arrested or started at will, would help distinguish direct from indirect effects and determine sequential steps in circadian pathways. Moreover, transgenic flies with conditionally titratable transcription of a clock component would allow molecular, cellular, and behavioral circadian phenotypes to be determined over a range of expression levels. Finally, flies with conditionally controlled clock function would allow separation of developmental and adult functions of clock genes. Based on these arguments we created conditionally rhythmic transgenic Drosophila strains.
In the present study, we describe the generation of transgenic flies in which clock function becomes conditional on account of temperature-dependent rescue of the per 01 or cyc 01 mutations or temperature-dependent mis-expression of per. Moreover, we made use of these flies to experimentally determine the developmental requirements for a functional circadian clock as well as the individual clock components PER and CYC. We confirmed and extended previously published observations [18,19] indicating that developmental rescue of arrhythmia in per 01 mutants is not needed for restoration of circadian rhythms in adults. However, developmental mis-expression of per or failure to developmentally rescue the cyc 01 mutation led to persistent adult arrhythmia. In particular, CLK/CYC function during the pupal and pharate adult stages was associated with adult clock function. Our results suggest two distinct mechanisms underlying the developmental requirement for CLK/CYC function: (1) cyc expression contributes celltype-autonomously in the ventral lateral neurons (LN v s) to the formation of peptidergic dorsal projections containing the neuropeptide PIGMENT DISPERSING FACTOR (PDF), which are thought to be important for adult circadian behavior and (2) CLK/CYC activity during development enables normal clock gene expression rhythms in the adult LN v s.

Results
Transgenic flies with conditional rescue of per 01 in clock-bearing cells We made use of the temporal and regional gene expression targeting (TARGET) system [20] to create transgenic flies in which the essential clock components CYC and PER were expressed conditionally in relevant spatiotemporal patterns. The TARGET system combines the binary GAL4/UAS system [21] that allows transgenic expression to be directed spatiotemporally by a promoter of interest via the intermediate regulator GAL4 with a ubiquitously expressed GAL80 ts gene, which encodes a temperature sensitive inhibitor of GAL4. As a result, the TARGET system permits GAL4-mediated transgenic expression at high temperatures (e.g., 29uC), but progressively restricts it at lower temperatures. First, we generated transgenic flies that conditionally rescued the arrhythmic per 01 phenotype [22] by introducing a GAL4-driver transgene directing expression in all clock-bearing cells (tim(UAS)-Gal4) [9] along with a GAL4responsive per cDNA expression construct (UAS-per) [23] and a transgene ubiquitously expressing GAL80 ts (tubP-Gal80 ts ) [20] in a per 01 genetic background [22] (see Figure 1A). The resulting genotype is abbreviated, here, as per 01 [timP.per] ts . As expected, clock-controlled phenotypes such as behavioral rhythmicity, relative rhythmic power and period length were readily and significantly modulated by environmental temperature in these flies ( Figure 1B-1D, Figures S1 and S2). Robust circadian rhythms in locomotor activity were virtually absent at a restrictive temperature (18uC), but rescued to varying degrees over a range of higher temperatures (21-29uC) ( Figure 1B-1D, Figures S1 and  S2). The circadian period length observed at 29uC was significantly longer than those at 25uC, 27uC, and 28uC for females and those at 23uC, 25uC and 28uC for males ( Figure 1D, Figure S2A, S2C; Welch test and post-hoc Games-Howell analysis). It is noteworthy that the decrease in rhythmicity and relative rhythmic power and the increase in circadian period length found at the highest experimental temperature of transgenic induction (29uC) were also observed as a result of transgenic per over-expression in a wild-type background (see below). In comparison with wild-type controls per 01 [timP.per] ts flies were much less rhythmic at 18uC (or 29uC), but at 25uC both genotypes showed comparable percentages of rhythmic, weakly rhythmic, and arrhythmic flies ( Figure S3). At permissive temperatures the most consistent difference in the behavior of per 01 [timP.per] ts flies relative to wild-type controls was a significantly longer circadian period length increased by 2 h or more ( Figure S3B-S3D).
Next, we examined clock-controlled molecular responses in per 01 [timP.per] ts flies released from restrictive (17 or 18uC) to permissive conditions (25uC). As expected, transgenic per expression was strongly induced in adult fly heads following this transition ( Figure S4A, S4B). In addition, the Clk, tim, vri, cwo, Par-domain Protein 1 (Pdp1), and Slow-poke binding protein (Slob) clockcontrolled transcripts showed relative expression responses that appeared consistent with their circadian phase relationships in wild-type heads. Nevertheless, the amplitude of the observed expression responses in clock-controlled genes was reduced relative to previously reported amplitudes of circadian oscillation in wild-type heads [4,[7][8][9][24][25][26][27]. Thus, upon transfer to permissive conditions, rescue of molecular circadian oscillations in adult per 01 [timP.per] ts heads, unlike behavioral rhythms, appeared to be incomplete. This discrepancy might be explained by a selective restoration of high-amplitude clock gene expression rhythms in clock neurons. We, therefore, examined circadian transcript responses in dissected adult brains of per 01 [timP.per] ts flies released under permissive conditions. However, molecular amplitudes in adult brains were comparable to those previously seen in adult heads (cf Figure S4A and S4C) suggesting incomplete restoration of molecular circadian rhythms in both peripheral clocks and the neural clock circuit. Since different time points in the Northern and Quantitative Reverse Transcriptase PCR (qRT-PCR) experiments of Figure S4 come from different samples of individual flies incomplete synchrony in the experimental population may have also contributed to the detection of relatively shallow transcript rhythms.
Adult circadian locomotor behavior does not require a functioning circadian clock or rescue of per 01 during prior development

Author Summary
The fruit fly Drosophila melanogaster is an excellent model system for studying the internal circadian clocks that animals use for daily time keeping. Since clocks exist and function in animals not only in adults, but also during prior development, the question arises if and how adult circadian rhythms depend on developmental clock circuits and components. To address this question we created transgenic flies in which the essential clock components CLOCK/CYCLE (CLK/CYC) and PERIOD (PER) can be manipulated via environmental temperature. Stopping the clock during development by depleting the negative regulator PER did not prevent restoration of circadian time keeping in the adult. However, a developmental arrest of the clock due to either depletion of the positive regulator CYC or overproduction of PER resulted in a persistent loss of clock-controlled behavior function in adults. Taken together, these observations indicate that adult clock function developmentally requires activity of the CLK/CYC transcription complex rather than a ticking clock. Based on the behavioral, molecular, and anatomical consequences of inhibiting CLK/CYC in circadian pacemaker neurons, we propose that the developmental requirement maps to these cells. It will be interesting to determine whether there is a comparable developmental requirement for the equivalent clock genes in humans. Figure 1. Conditional transgenic rescue of per 01 arrhythmic behavior. Introduction of tim(UAS)-Gal4, UAS-per, and tubP-Gal80 ts transgenes in a per 01 genetic background resulted in conditional circadian clock function, with virtually no rescue of circadian locomotor behavior at the restrictive temperature (18uC) and approximation of wild-type circadian behavior at permissive temperatures (23-29uC). (A) The lack of rescue at 18uC is flies at 17uC in constant light (LL) until adulthood and examined behavioral rhythms at restrictive (17uC) and subsequent permissive (25uC) conditions in constant darkness (DD). Consistent with the hypothesis that rescue of circadian clock function in per 01 flies can be achieved when per expression is restricted to clock-bearing cells in the adult, we observed restoration of circadian locomotor rhythms immediately following transition to permissive conditions ( Figure 2A, Figure S5). Although adult per 01 [timP.per] ts flies failed to show strongly rhythmic locomotor behavior at the restrictive temperature in DD, a subset of individual flies did exhibit residual weak rhythms under these conditions ( Figure S2). Nevertheless, we do not believe that the observed behavioral rescue in adults depends on residual clock function during the prior exposure to the restrictive temperature for two reasons: (1) The phase of the restored rhythms is determined by the phase of the prior switch from restrictive to permissive conditions rather than the phase of the light/dark transition associated with the start of the behavioral experiment ( Figure 2) and (2) our experiments included developmental exposure to LL, which is associated with both behavioral and molecular arrhythmia as well as severely reduced PER expression levels [13,28]. Therefore, we conclude that there is no developmental requirement for either a functioning clock mechanism or expression of per in the clock-bearing cells in order to allow circadian clock function in adult flies.

Over-expression of per during metamorphosis disrupts adult circadian behavior
Given the ability to restore adult clock function from a circadian cycle arrest due to PER depletion, we were wondering whether circadian cycle arrests associated with excess PER expression were equally reversible. This question was addressed experimentally with the help of transgenic flies, in which per was conditionally overexpressed to high levels in clock-bearing cells due to the introduction of the tim(UAS)-Gal4, UAS-per, and tubP-Gal80 ts transgenes in the presence of a wild-type per gene. Flies homozygous for the autosomal tim(UAS)-Gal4 and UAS-per insertions with a single X-chromosomal tubP-Gal80 ts transgene (abbreviated as [timP.per] ts ) showed conditional clock function with robust rhythms, relative rhythmic power, The phase of the offset of circadian locomotor activity at 25uC is plotted separately for median data from 6 groups of male and 6 groups of female flies as a function of the phase of the 17uC to 25uC switch along with trend lines and associated correlation coefficients (2-tailed test significance: p,0.005 for females, p,0.003 for males). doi:10.1371/journal.pgen.1002167.g002 explained by inhibition of GAL4-mediated expression of transgenic per mediated by the temperature sensitive GAL4 repressor GAL80 ts . At permissive temperatures the modulating effect of GAL80 ts is reduced allowing GAL4 expressed in clock-bearing cells to induce per and rescue circadian clock function. (B,C) Flies with a y per 01 w; tim(UAS)-Gal4/tubPGal80 ts ; UAS-per/+ genotype, abbreviated as per 01 [timP.per] ts , raised at ambient temperature (,23uC) were monitored for adult locomotor activity sequentially at restrictive and permissive conditions (see Materials and Methods). The large diagrams in (B) and (C) are double-plotted actograms representing the median locomotor activity for 8 female flies. Yellow triangles indicate the time of the temperature shifts. The small diagrams are chi-square periodograms based on the first 5 or 6 full days at the first (top) or second (bottom) experimental condition. Circadian period lengths detected at the permissive conditions are indicated in large blue type-face. Note the absence of strong circadian rhythms at 18uC. The considerably lengthened period and progressive weakening of rhythms observed at 29uC are likely attributable to excessive per expression. (D) Temperature-dependent conditional rescue of the percentage of rhythmic flies, relative rhythmic power, and period length in per 01 [timP.per] ts flies. Chi-square periodogram analysis of circadian locomotor behavior in DD was performed for 5-day intervals at the indicated temperatures. The three panels show changes as a function of environmental temperature in the percentage of rhythmic flies, relative rhythmic power (across both rhythmic and weakly rhythmic flies), and circadian period length for rhythmic flies. Error bars represent the Standard Error of the Mean (SEM). doi:10.1371/journal.pgen.1002167.g001 and only marginally increased period lengths at permissive (17uC) conditions and behavioral arrhythmia (females) or dramatically reduced rhythms (males) at the restrictive (29uC) conditions ( Figure 3, Figure 4, Figures S6 and S7). Loss of behavioral rhythms during prolonged exposure of adults to the restrictive condition could be reversed by returning the flies to the permissive (17uC) condition ( Figure 3B, Figure S7). However, comparable exposure to restrictive conditions during development resulted in irreversible adult arrhythmia ( Figure 3B, Figure 4, Figures S6 and S7) for both genders. To identify the developmental phase of sensitivity to PER over-expression flies were transferred from a permissive ambient temperature (,23uC) to 29uC or vice versa at different points during development and then analyzed for behavioral rhythmicity as adults. When exposure to restrictive conditions occurred prior to the pupal stage it did not obviously affect the percentages of flies exhibiting rhythmic, weakly rhythmic or arrhythmic adult behavior or the relative power of the detected rhythms ( Figure 4, Figure S6). However, when flies were exposed to the restrictive temperature throughout the pupal and pharate adult stages, adult locomotor rhythms were clearly inhibited ( Figure 4, Figure S6). Therefore, it appears that per over-expression in pupal/pharate adult clock cells irreversibly affects adult circadian behavior. One possible explanation for the observed effect of developmental per mis-expression on adult behavior might be the persistence of abnormally high levels of PER protein into adulthood. However, PER is known to be an unstable protein and even a 7-d exposure to 12-h light/12-h dark/ (LD) cycles at the permissive (17uC) temperature did not allow subsequent restoration of behavioral rhythms in DD. Moreover, immunofluorescence analyses of clock neurons exposed to developmental PER over-expression did not reveal a continued increase in adult PER expression. Instead, the persistent behavioral arrhythmia of [timP.per] ts flies raised at 29uC and exposed to 17uC LD for 7 d as adults appeared to be matched by blunted circadian rhythms of PER expression in the PDF-expressing ventral lateral neurons ( Figure 5). No gross morphological defects in clock neurons (including LN v , LN d , and DN cell bodies and LN v projections) were apparent in these experiments (see Figure S8). Thus, our results indicate that excess PER activity in clock cells during metamorphosis negatively affects both adult circadian locomotor activity and molecular rhythms in adult clock neurons.

Depletion of cyc expression during metamorphosis disrupts adult circadian behavior
Based on PER's known function as a negative regulator of CLK/CYC circadian transcription complexes the adult phenotypes associated with developmental PER over-expression are likely attributable to inhibition of CLK/CYC activity. We tested this hypothesis by determining whether adult circadian locomotor behavior required prior developmental expression of the essential clock component CYC. To this aim we generated transgenic flies that conditionally expressed cyc in postmitotic neurons by combining the elav C155 ::Gal4 driver element [29] with UAS-cyc [30], and tubP-Gal80 ts transgenes in a cyc 01 background. The resulting flies, here referred to as cyc 01 [elav.cyc] ts , showed conditional rescue of rhythmic adult locomotor activity when raised at the permissive temperature for cyc 01 rescue (29uC) ( Figure 6, Figure S9). Ambient temperature (,23uC), which acted as a mostly permissive condition for per 01 [timP.per] ts flies (see Figure 1D, Figure S2, above) represented a restrictive condition for the cyc 01 [elav.cyc] ts strain. This discrepancy is likely attributable to differences either in the amount of GAL4 protein produced in the relevant clock neurons in each of these strains or the level of GAL4-directed transgenic expression that is required to achieve behavioral rescue. Exposure of cyc 01 [elav.cyc] ts flies to the restrictive temperature during metamorphosis, severely affected adult behavioral rhythms at the permissive temperature ( Figure 6B, Figure 7, Figure S10). Therefore, depletion of cyc expression during metamorphosis, indeed, phenocopies the adult behavioral defects of per mis-expression during metamorphosis.
Selective inhibition of cyc 01 rescue in the PDF-expressing clock neurons disrupts adult locomotor rhythms as well as cell-type-specific neuro-anatomy and molecular rhythms To further explore the role of CLK/CYC expression in the PDF-positive LN v s in ensuring normal adult circadian behavior and neuro-anatomy we created flies in which rescue of cyc 01 in postmitotic neurons (by elav C155 ::Gal4 and UAS-cyc) was selectively blocked in the PDF-expressing LNvs with the help of a Pdf-Gal80 transgene [31] that expresses the GAL4 inhibitor GAL80 specifically in these cells. The behavioral phenotype of the resulting transgenic flies, indicated as cyc 01 (elav-Pdf).cyc, consists of an altered daily locomotor activity profile in the presence of light/dark cycles that includes extended activity in anticipation of lights-on, but reduced activity in anticipation of lights-off ( Figure 8A, Figure S11A) and a loss of sustained rhythmicity in constant darkness (Figure 8A-8C; Figure S11). In contrast, control cyc 01 rescue flies lacking the Pdf-Gal80 element (cyc 01 elav.cyc) showed strong behavioral rhythms in constant darkness as well as evening activity in anticipation of the lights-off transition (Figure 8A-8C; Figure S11). The cyc 01 (elav-Pdf).cyc phenotype is clearly different from that of flies with ablated PDF-expressing LN v s or defective expression of the PDF neuropeptide [32], which also lack consolidated rhythms in constant darkness but show the opposite effect on anticipation of the lights-on and lights-off transitions. The persistence of morning anticipation, which is thought to be attributable to PDF signaling from the s-LN v s [33] suggests that residual PDF expression and function persisted in s-LN v s with a cyc 01 circadian cycle arrest. Moreover, it is insightful to compare the behavior of cyc 01 (elav-Pdf).cyc flies to previously published observations for flies, in which rescue of the per 01 mutation in clock-bearing cells was blocked in the PDF-expressing clock neurons (per 01 (elav-Pdf).per) [31]. The behavior reported for per 01 (elav-Pdf).per flies resembles that of our cyc 01 (elav-Pdf).cyc flies with respect to the persistence of morning anticipation as well as reduced rhythmicity in constant darkness [31]. However, loss of evening anticipation appears to be unique to the cyc 01 -based as opposed to the per 01 -based arrest of the LN v s. Thus, cyc-depleted LN v s seemed to delay the generation of an evening activity signal by the neural clock circuits. It is possible that a slow rhythmic component in the disorganized clock circuit of cyc 01 (elav-Pdf).cyc flies is responsible for this delay in evening activity. Consistent with this notion, the few (weakly) rhythmic cyc 01 (elav-Pdf).cyc flies represented in Figure 8 and Figure S11 exhibited residual long period rhythms (t for females: 25.161.24 h, n = 5; t for males 24.460.83 h, n = 17).
Next, we compared molecular and neuro-anatomical phenotypes of the PDF-expressing LN v s in cyc 01 (elav-Pdf).cyc flies with those of controls lacking either the Pdf-Gal80 (cyc 01 elav.cyc) or the UAS-cyc transgenes (cyc 01 (elav-Pdf).-). Flies of these three genotypes were raised at ambient temperature and entrained as adults to LD cycles at 25uC. Brains were harvested 2 h prior to lights-on (ZT22) and stained using antibodies against PDF and the PDP1. The Pdp1 gene is a direct target gene for CLK/CYC and mutations in Clk or cyc strongly affect PDP1 protein expression in larvae and adults [26]. Indeed, cyc 01 (elav-Pdf).flies, which completely lack CLK/CYC function, exhibited greatly reduced Figure 3. Developmental over-expression of per disrupts adult circadian behavior, while the phenotype of adult per overexpression is reversible. Transgenic flies with conditional over-expression of per exhibited a reversible loss of circadian behavior when temporarily shifted to restrictive conditions (29uC) as adults, but showed long-term behavioral arrhythmia when similarly exposed to restrictive conditions during development. (A) Flies with a y tubPGal80 ts w/(FM7c or Y); tim(UAS)-Gal4; UAS-per genotype, abbreviated as [timP.per] ts , show rhythmic locomotor behavior at the permissive temperature when per over-expression is prevented by GAL80 ts , but not at the restrictive temperature when GAL80 ts is ineffective and excessive levels of PER prevent circadian clock function. (B) The top four panels are double-plotted actograms representing median locomotor activity data for groups of female [timP.per] ts flies during LD and subsequent DD conditions at the permissive temperature (17uC DD). The two actograms on the left illustrate that developmental exposure to 17uC or ,23uC has no obvious effect on adult circadian behavior, while the data in the third panel from the left reflect the loss of adult locomotor activity rhythms following development at the restrictive temperature (29uC). The right-most actogram illustrates the reversible behavioral phenotype of 17uC-raised flies that where shifted to the restrictive temperature as adults (29uC DD data included at the top) prior to the rest of the experiment. The white, light blue, and red background colors in the actograms represent 17uC light, 17uC dark, and 29uC dark conditions, respectively. The lower four panels are chi-square periodograms for the median data reflecting circadian rhythmicity in 17uC DD. Single significant (p,0.01) period lengths in the circadian range (18-35 h)   per] ts flies under permissive conditions (17uC DD) was strongly rhythmic for flies raised at ambient temperature (,23uC) or flies transferred from 29uC to ambient temperature as wandering larvae or prepupae. However, flies exposed to restrictive (29uC DD) conditions throughout development or during the pupal and pharate adult stages exhibited mostly arrhythmic adult locomotor behavior. (A) Stacked bar diagram representing the percentages of female flies with rhythmic, weakly rhythmic, or arrhythmic adult behavior at permissive conditions (17uC DD). Prior to measurement of adult locomotor activity at 17uC LD and subsequent 17uC DD (analyzed here) flies were raised at the indicated temperatures (T dev ): 23uC, 23R29uC (transferred as wandering larvae or prepupae from 23uC to 29uC), 29R23uC (transferred as wandering larvae or prepupae 29uC to 23uC), and 29uC. The numbers (n) of flies included for each condition are indicated as well as the average (6SEM) circadian period length for rhythmic flies. Chi-square analysis indicated a highly significant (p,10 223 ) association between developmental temperature and the percentages of rhythmic, weakly rhythmic, and arrhythmic adults. (B) Bar diagram of the average (6SEM) relative rhythmic power observed among the rhythmic plus weakly rhythmic flies for each developmental condition. The number of flies included in this analysis (n) is indicated for each condition. The Welch test statistic indicated a highly significant association (p,10 213 ) of relative rhythmic power with developmental condition. Significant differences found by post-hoc Games-Howell tests for pairwise comparisons of developmental treatments indicated by (*****) represent p values smaller than 10 25 . doi:10.1371/journal.pgen.1002167.g004 per] ts flies raised at either restrictive (29uC) or permissive (,23uC) conditions were transferred as adults to 17uC LD conditions and harvested 2-h before lightson (ZT22) on day 8 and 4-h after lights-on (ZT4) on day 9. Brains were dissected and subjected to immunofluorescence staining using antibodies PDP1 expression in their LN v s at ZT22. Consistent with previous studies [34], s-LN v s in cyc 01 (elav-Pdf).brains also showed a reduced PDF signal in the s-LN v s and mostly abnormal or missing PDF-positive dorsal projections (Figure 9). These phenotypes were to a large extent rescued in cyc 01 elav.cyc flies, which not only exhibited PDP1 expression in virtually all PDF-expressing LN v s (and other clock neurons) at ZT22, but also presented with normal PDF-expression levels and dorsal LN v projections ( Figure 9). The additional introduction of Pdf-Gal80 in the cyc 01 (elav-Pdf).cyc genotype, resulted in cell-type-specific phenotypes that included down-regulation of PDP1 in virtually all LN v s with detectable PDF expression as well as a reduction in the number of s-LNvs with detectable PDF expression (see Figure 9A, 9B). Moreover, PDFpositive sLN v dorsal projections were either abnormal or missing from most cyc 01 (elav-Pdf).cyc brains (see Figure 9C), although there is a formal possibility that PDF-negative sLN v projections, which would not have been detectable in these experiments, still extended to the dorsal protocerebrum. For other clock neurons no obvious differences were detected in numbers and PDP1 expression levels between the cyc 01 (elav-Pdf).cyc brains and the rescued cyc 01 elav.cyc controls.

Discussion
We created transgenic flies with conditional clock function, in which expression of the essential clock components CYC and PER was induced or repressed in relevant spatiotemporal patterns. In per 01 [timP.per] ts flies, which conditionally rescue the per 01 mutation, clock function was conditional and readily reversible. Moreover, adult circadian behavior was restored in flies raised under restrictive conditions. In earlier studies conducted by Ewer and colleagues widespread transgenic expression of per under control of a heat-shock protein 70 (hsp70) promoter was shown to partially rescue the per 01 mutation resulting in restoration of behavioral rhythms at an abnormally long period length. These long period rhythms could be generated in a conditional manner even when induction was restricted to the adult phase [18,19]. In the present study we targeted expression of transgenic per specifically to clock-bearing cells and achieved a more complete conditional rescue of the per 01 phenotype that did not require developmental per expression.
Although circadian behavior of per 01 [timP.per] ts flies at 25uC showed rhythmicity comparable to that observed for wild-type flies, period lengths were at least 2 h longer than those of wild-type flies and molecular rhythms showed a relatively low amplitude.
One key difference between the molecular clock circuits in per 01 [timP.per] ts at the permissive temperature and those of wild-type flies is the constitutively high level of per mRNA expression in the transgenically rescued flies, which could contribute to the increased circadian period length and blunted molecular rhythms in per 01 [timP.per] ts flies. Wild-type flies exhibit a trough in per transcript levels in the early morning that may facilitate subsequent down-regulation of PER protein levels and optimal induction of CLK/CYC-regulated genes [35]. The lack of a trough in per mRNA expression in the conditionally rescued flies could account for a delay in the turnover of PER protein in the morning and, therefore, a lengthened period and blunted CLK/ CYC activity. This hypothesis also explains apparent discrepancies with previous reports, in which increased per gene dosage was associated with a shortened circadian period length [36] and decreased per dosage or expression resulted in longer circadian period lengths [22,37,38]. As long as per expression shows strong circadian regulation the timing of PER nuclear entry and PERmediated transcriptional repression is predicted to be advanced by the introduction of one or two additional copies of the wild-type per gene and delayed by a reduction in per dosage, while neither manipulation is predicted to strongly affect subsequent PER turnover.
Adult circadian behavior was also conditional and reversible in [timP.per] ts flies, which exhibit temperature-dependent overexpression of per. However, developmental over-expression of per during metamorphosis was associated with irreversible behavioral arrhythmia in adults. Likewise, depletion of cyc expression during the metamorphosis in cyc 01 [elav.cyc] ts flies resulted in disruption of adult circadian locomotor behavior under permissive conditions. Both increased levels of PER and decreased levels of CYC negatively regulate CLK/CYC activity. The CLK/CYC heterodimer functions as the central transcriptional regulator in the Drosophila clock and its activity critically depends on the presence of both CLK and CYC [3,5,6]. Loss of functional cyc expression in the cyc 01 mutant results in both molecular and circadian arrhythmia and constitutively low expression levels for CLK/ CYC-regulated target genes [6], whereas PER acts as a negative regulator of CLK/CYC activity by binding and inactivating the CLK/CYC complex [5,15,39]. The arrhythmic locomotor behavior and molecular arrhythmia in the clock neurons observed as a result of per over-expression [23,40] are, therefore, interpreted to result from constitutive inhibition of CLK/CYC.
Adult behavioral arrhythmia in [timP.per] ts or cyc 01 [elav.cyc] ts flies raised under permissive conditions was reversible (see Figure 3B, Figure 6C, 6F, 6G, Figures S7, S9C, S9D, above). However, exposures to restrictive conditions of comparable duration resulted in long-term after-effects only when they occurred during development and, particularly, during the pupal and pharate adult stages. We, therefore, attribute the effects of circadian arrests during development in [timP.per] ts or cyc 01 [elav.cyc] ts flies on adult circadian behavior to a developmental requirement for CLK/CYC function beyond its immediate role in maintaining daily time keeping. The requirement for CLK/CYC activity, but not clock function per se may indicate that one or more transcriptional CLK/CYC targets play a role in enabling adult circadian locomotor behavior. Such targets would likely be expressed constitutively along with other CLK/CYC-regulated genes in conditionally arrested per 01 [timP.per] ts flies, but constitutively down-regulated in circadian arrests due to low CLK/CYC activity.
A central question that remains is what mechanism links developmental CLK/CYC activity to adult circadian behavior. Our experiments indicate that both clock neuron anatomy and the molecular oscillator itself may be involved. Previously published studies of constitutively arrhythmic alleles of the Clk and cyc genes have documented a reduction in PDF expression as well as directed against the PER and PDF proteins. Representative images illustrate the effects of both daily phase and developmental treatment on PER expression for PDF-expressing LN v cells (arrows). (B) Semi-quantitative analysis of the immunofluorescence signal for PER in adult PDF neurons. The cumulative height of the bars indicate the average number per brain hemisphere of l-LNv and s-LNv detected by PDF staining, whereas the segments with increasingly darker shades of gray represent the relative prevalence of cells with no, very faint, faint, moderate, or strong levels of PER signal, respectively. The numbers of brain hemispheres is indicated (n). Significant associations of PER signal level and daily phase found by Chi-square analyses indicated by (#), (##), and (###) represent p values smaller than 10 26 , 10 210 , and 10 217 , respectively. doi:10.1371/journal.pgen.1002167.g005 neuro-anatomical defects in the LN v s [34] that could be associated with a developmental role for the CLK/CYC transcription factor. By selectively blocking transgenic rescue of cyc 01 in the PDFexpressing clock neurons we show, here, that the reduction of PDF expression and PDF-positive dorsal projections from the s-LN v s is a cell-type specific phenotype. PDF is known to play an important role in mediating clock-controlled behavior in both LD and DD conditions. The PDF-producing s-LN v s project towards the dorsal protocerebrum as do DN1, DN2, DN3, and LN d clock neurons, suggesting that the dorsal s-LN v projections may play an important part in signaling across the neural clock circuits [41]. In this context, it may be relevant that expression of the PDF RECEPTOR in 'E' cells, a subset of clock neurons including DN1s and LN d s [31], has been associated with circadian control of locomotor activity [42]. Moreover, the axonal terminals of the dorsal s-LNv projections undergo clock-controlled rhythms in remodeling that may play a role in circadian signaling [43]. Nevertheless, the observed developmental requirement for CLK/ CYC activity also appears to involve mechanisms other than PDFmediated signaling for the following reasons. First, developmental over-expression of PER resulted in persistent adult arrhythmia, but did not lead to a loss of PDF-positive dorsal projections from the s-LN v s (see Figure S8). Second, while developmental suppression of CLK/CYC activity uniformly affected the behavior of adult flies ( Figure 4, Figure 7, Figures S6, S7, S10) constitutive depletion of CYC from the PDF-expressing neurons resulted in a variable phenotype in the s-LN v dorsal projections (see Figure 9C). Third, the light/dark activity pattern of cyc 01 (elav-Pdf).cyc flies ( Figure 8A, Figure S11A) was strikingly different from that of Pdf 01 flies or flies from which the PDF-expressing cells have been ablated [32], suggesting that PDF signaling persisted in cyc-depleted LN v s in spite of the defects in PDF-positive dorsal projections.
In principle, neuro-anatomical defects affecting intercellular connectivity rather than cell-autonomous clock function could lead to behavioral phenotypes due to asynchrony among the clock neurons or the loss of output signals. Indeed, apparent separation of molecular and behavioral phenotypes has been reported previously for genetic manipulation of CLK/CYC function [44,45]. It may be particularly relevant that rescue of the per 01 phenotype in the PDF-expressing clock neurons restores rhythmic behavior [46], while rescue of cyc 01 in the same cells restores molecular, but not behavioral rhythms [45]. However, our experimental results also provide support for developmental phenotypes at the level of the adult molecular clock circuits. Our immunofluorescence expression analyses indicated that the molecular clock circuits in the adult PDF-expressing clock neurons were affected by developmental over-expression of PER.
PDF-expressing LN v s in adults that were behaviorally arrhythmic due to developmental PER over-expression exhibited adult PER expression with an altered daily profile, but not necessarily at excessively high levels. Future studies may determine the degree to which neuro-anatomical and molecular phenotypes are linked and help determine the effect of intercellular connectivity in the neural clock circuit on the function of molecular circadian rhythms in individual clock neurons.

Drosophila stocks
Flies were raised on standard yeast cornmeal agar food either at ambient temperature (observed to range between 22uC and 24uC) or other experimental temperatures as specified. The conditional per 01 rescue flies indicated as per 01 [timP.per] ts in Figure 1, Figure 2 and Figures S1, S2, S3, S5 consisted of male and female y per 01 w; tim(UAS)-Gal4/tubPGal80 ts ; UAS-per/+ offspring from a cross between stable lines y per 01 w; tim(UAS)-Gal4 and y per 01 w; tubPGal80 ts ; UAS-per. These stocks were created by combining the previously described per 01 [22], tim(UAS)-Gal4 [9], tubPGal80 ts [20], and UAS-per [23] genetic elements. Flies with conditional overexpression of per in clock-bearing cells ([timP.per] ts in Figure 3, Figure 4, Figure 5, and Figures S6, S7, S8) were obtained from a genetically stable y tubPGal80 ts w/FM7c; tim(UAS)-Gal4; UAS-per stock as females heterozygous for FM7c and non-FM7c males. The insertion site of the tubPGal80 ts transgene in this stock appears to be associated with homozygous female lethality. An X-chromosomal period-lengthening allele present in the genetic background of the original tubPGal80 ts stocks was avoided during the creation of the [timP.per] ts stock by recombination with a control y w chromosome. Flies with conditional rescue of cyc 01 (cyc 01 [elav.cyc] ts in Figure 6, Figure 7 and Figures S9, S10) were obtained from a stable elav C155 ::Gal4; UAS-cyc/CyO; cyc 01 tubPGal80 ts stock as males and females heterozygous for CyO. The elav C155 ::Gal4 [29], UAS-cyc [30], cyc 01 [6], and tubPGal80 ts [20] elements used to create this stock had al been described previously. The cyc 01 rescue line elav C155 ::Gal4; UAS-cyc/CyO; cyc 01 (cyc 01 elav.cyc in Figure 8, Figure 9 and Figure S11) was created as a stable stock, whereas the selective cyc 01 rescue genotype elav C155 ::Gal4; UAS-cyc/Pdf-Gal80; cyc 01 and the unrescued control genotype elav C155 ::Gal4; CyO/Pdf-Gal80; cyc 01 (respectively, cyc 01 (elav-Pdf).cyc and cyc 01 (elav-Pdf).in Figure 8, Figure 9 and Figure S11) were obtained in offspring from a cross of elav C155 ::Gal4; UAS-cyc/CyO; cyc 01 flies with elav C155 ::Gal4; Pdf-Gal80; cyc 01 flies. The Pdf-Gal80 element used in the latter two genotypes has also been characterized previously [31].  Adult circadian behavior of 29uC-raised cyc 01 [elav.cyc] ts flies was strongly rhythmic under permissive conditions (29uC DD). Substantial rescue of adult locomotor rhythms was also observed for cyc 01 [elav.cyc] ts flies raised at ambient temperature (,23uC) and transferred to 29uC as wandering larvae or prepupae. However, flies exposed to restrictive (,23uC) conditions throughout development or during the pupal and pharate adult stages exhibited weakly rhythmic or arrhythmic adult locomotor behavior. (A) Stacked bar diagram representing the percentages of female flies with rhythmic, weakly rhythmic, or arrhythmic adult behavior at permissive conditions (29uC DD) following prior exposure to 7 LD days at 25uC. Prior to measurement of adult locomotor activity during 25uC LD and subsequent 29uC DD (analyzed here) flies were raised at the indicated temperatures (T dev ): 23uC, 23R29uC (transferred as wandering larvae or prepupae from 23uC to 29uC), 29R23uC (transferred as wandering larvae or prepupae 29uC to 23uC), and 29uC. The numbers (n) of flies included for each condition are indicated as well as the average (6SEM) circadian period length for rhythmic flies. Chi-square analysis indicated a significant (p,10 22 ) association between developmental temperature and the percentages of rhythmic, weakly rhythmic, and arrhythmic adults. Locomotor behavior assays Using previously described protocols [47], locomotor activity was monitored for individual adult flies of both genders in glass tubes on standard sugar agar media including 0.07% Tegosept (Genesee Scientific) using the Drosophila Activity Monitoring System (TriKinetics). Experiments were conducted in incubators kept at 70% relative humidity in 12 h L: 12 h D or DD conditions using white fluorescent light with an approximate intensity of 450 mW/cm 2 during the L condition. Due to lack of space only analyses for female flies are shown in Figure 1BC, Figure 4, Figure 6D-6G, Figure 7, and Figure 8; the corresponding analyses for male flies are found in Figures S1, S6, S9, S10, and S11, respectively.

Statistical analyses of locomotor activity data
Individual, experimental average, and experimental median activity records, as well as periodic activity profiles, and chi-square periodograms were generated using ClockLab Software (Acti-Metrics). Actograms ( Figure 1BC, Figure 2A, Figure 3B, Figure 6B and 6C, Figure 8A, Figures S1, S3A-S3C, S11A) were doubleplotted with a resolution of half-hour intervals. Each row represents a 2-day interval of Zeitgeber Time (ZT, with ZT0 as the time of lights-on; during LD) or Circadian Time (CT; during DD), of which the second day is repeated as the first day on the next row. Chi-square periodograms ( Figure 1BC, Figure 3B, Figure 6B and 6C, Figures S1 and S3A-S3C) were used to represent the experimental signal (amplitude) observed for a range of period lengths (t, x-axis) relative to threshold values associated with a p,0.01 significance (red line). Analyses of the percentages of rhythmic, weakly rhythmic, and arrhythmic flies ( Figure 1D, Figure 4A, Figure 6D and 6F, Figure 7A, Figure 8B, Figures  S2AC, S3D, S5AC, S6A, S7A, S7B, S7D, S7E, S9A, S9C, S10A, S11B) were based on chi-square periodogram statistics for locomotor activity rhythms of individual flies. For period lengths in the circadian range (,15-36 h) detected with a significance of p,0.01 relative rhythmic power was calculated by dividing the detected peak amplitude by the significance threshold value at the same period length. Flies were classified based on their values of relative rhythmic power as rhythmic (.1.5) or weakly rhythmic ([1,1.5]) and flies without significant periodicity in the circadian range were considered arrhythmic. Chi-square analyses for association of genotype or experimental protocol with the relative distribution of rhythmic, weakly rhythmic, or arrhythmic behavior were conducted using Microsoft Excel (Microsoft). Next, statistical analyses were performed using SPSS software (IBM) to detect associations between the relative rhythmic power values of rhythmic and weakly rhythmic flies with experimental conditions ( Figure 4B, Figure 6E and 6G, Figure 7B, Figures S2BD, S5BD, S6B, S7CF, S9BD, S10B) or genotypes ( Figure 8C, Figure S11C). In virtually all cases Levene's test indicated that homogeneous variances could not be assumed. Therefore, the Welch test statistic with Games-Howell post-hoc analysis was used to test for significant differences in relative rhythmic power among different genotypes and treatments. When only two conditions were compared the non-parametric Mann-Whitney rank-sum test was performed. Average (Figure 2A, Figure 6B and 6C, Figure S3A, S3B, S3C) and median ( Figure 1B and 1C, Figure 3B, Figure 8A, Figures S1, S11A) activity records that emphasize reproducible features of rhythmic locomotor activity measured in individual flies were created without prior normalization from the raw individual activity records on a point-by-point basis. For representation in illustrative double-plotted actograms we generally used median activity records, which are less susceptible to skewing by outliers and show discrete numbers of events per half-hour bin, but when a better resolution of data with relatively low activity counts was preferred average activity records were used instead. Average daily or circadian activity profiles representing records of median or average activity 6 the Standard Error of the Mean (SEM) were generated across included days under entraining ( Figure 8A, Figure S11A) or free running conditions (S3, for phase determination in Figure 2B), respectively. The phase of the offset of circadian activity ( Figure 2B) was determined from the activity profiles by interpolation as described previously [47]. Error bars throughout the manuscript represent SEM, except in cases where less than three observations were made. The parentheses surrounding individual error bars in Figure 1B (right-hand panel), Figure S7F, and Figure S9B indicate that these represent the range of two observations, instead.

Northern analysis
Extraction of total RNA from approximately 100 ml adult heads per time point using the guanidinium thiocyanate/cesium chloride method and subsequent Northern analysis were conducted according to previously published protocols [48,49]. Quantitative analysis of the radioactive signals on the blots was conducted with a Storm 840 Phosphorimager (GE healthcare) and the resulting data was graphed using Microsoft Excel (Microsoft Corporation). Five independent time course experiments were conducted addressing the transcript responses observed in the adult head upon transfer of per 01 [timP.per] ts flies from restrictive to permissive conditions. A representative example is shown in Figure S4A.

Quantitative reverse transcriptase PCR (qRT-PCR) analysis
Flies for the conditions of interest were harvested onto ice, and either adult heads or brains were dissected on a chilled platform and transferred to guanidinium thiocyanate buffer. DNAse Example actograms (left) and LD activity profiles (right) representing median locomotor behavior for female elav C155 ::Gal4; UAS-cyc/CyO; cyc 01 flies (abbreviated as cyc 01 elav.cyc), which are rescued for cyc expression in postmitotic neurons, versus female elav C155 ::Gal4; UAS-cyc/Pdf-Gal80; cyc 01 flies (abbreviated as cyc 01 (elav-Pdf).cyc), in which transgenic cyc rescue is selectively blocked in the PDF-expressing clock neurons. The activity profiles represent average daily activity (6SEM) indicated by the black line and red shading for the median locomotor activity of cyc 01 elav.cyc and cyc 01 (elav-Pdf).cyc flies in the presence of LD cycles. Note that cyc 01 (elav-Pdf).cyc flies exhibit loss of free running rhythms in DD as well as increased activity in anticipation of lights-on and loss of activity in anticipation of lights-off in LD. (B) Stacked bar diagram representing the percentages of female flies with rhythmic, weakly rhythmic, or arrhythmic adult behavior at 25uC DD. Along with cyc 01 elav.cyc and cyc 01 (elav-Pdf).cyc flies, the experiment also included non-rescued control flies with an elav C155 ::Gal4; CyO/Pdf-Gal80; cyc 01 genotype (abbreviated as cyc 01 (elav-Pdf).-). For each transgenic combination with cyc 01 the number of flies (n) as well as the average (6SEM) circadian period length for rhythmic flies are indicated. Chi-square analysis indicated a highly significant (p,10 29 ) association between genotype and the percentages of rhythmic, weakly rhythmic, and arrhythmic adults. (C) Bar diagram of the average (6SEM) relative rhythmic power observed among the rhythmic plus weakly rhythmic female flies for each genotype. The number of flies included in this analysis (n) is indicated for each condition. Because all cyc 01 (elav-Pdf).-flies were arrhythmic, a Mann-Whitney rank-sum test was performed to compare the effect on relative rhythmic power of the other two genotypes. As indicated, relative rhythmic power was significantly reduced in cyc 01 (elav-Pdf).cyc flies compared to cyc 01 elav.cyc flies (**; p,10 22 ). doi:10.1371/journal.pgen.1002167.g008 I-digested total RNA was obtained from the heads or brains using the RNAqueous4PCR kit (Ambion). Aliquots of the RNA samples were then analyzed with the SuperScript III Platinum SYBR Green One-Step qPCR Kit (Invitrogen) using experimental primer pairs designed to specifically amplify fragments of the circadian per, tim, vri, and cwo transcripts, the transgenic UAS-per transcript or the rp49 control transcript. Expression levels measured on a SmartCycler system (Cepheid) relative to rp49 were determined using the comparative Cycle threshold (Ct) method [50].

Immunofluorescence analysis
Adult brains were dissected, fixed, and stained for immunofluorescence analysis according to standard protocols [51]. Imaging was conducted with a spinning disk confocal microscope. Brains from [timP.per] ts flies raised under restrictive versus permissive conditions were probed with primary antibodies against PDF (mouse monoclonal; DSHB) as well as PER (rabbit polyclonal; [52]), whereas brains from cyc 01 elav.cyc and cyc 01 (elav-Pdf).cyc flies were stained with antibodies to PDF as well as PDP1 (rabbit polyclonal; [26]) and then visualized with fluorescently labeled secondary antibodies (Alexa-488 for PDF, Alexa-568 for PER or PDP1). The data for each transcript represent expression ratios relative to an internal control (rp49) that were normalized to the experimental average. The signal for per encompassed both the endogenous per 01 and transgenic per transcripts and exhibited a temperaturedependent induction. Clock-controlled transcript profiles exhibited shallow amplitudes but maintained the expected relative phase relationships, with peak expression of tim, vri, cwo, and Slob peaking Pdf).cyc but not cyc 01 elav.cyc flies at ZT22. Whole mount adult brains for both genotypes were stained with antibodies directed against PDF (green) and PDP1 (red). s-LN v , l-LN v , LN d , and 5 th s-LN v clock neurons are indicated by yellow, white, red, and gray arrows, respectively. (B) Semiquantitative analysis of the number of PDP1-positive and PDP1-negative PDF-expressing LN v s at ZT22 per brain hemisphere for cyc 01 (elav-Pdf).-, cyc 01 (elav-Pdf).cyc, and cyc 01 elav.cyc flies. The numbers (n) of brain hemispheres analyzed are indicated for each genotype. Chi-square analyses found highly significant associations between genotype and PDP1 expression in both s-LN v s (p,10 231 ) and l-LN v s (p,10 255 ). (C) Semi-quantitative analysis of neuro-anatomical defects in the dorsal PDF-expressing LN v projections. Stacked bar diagram (top panel) representing the percentages of cyc 01 (elav-Pdf).-, cyc 01 (elav-Pdf).cyc, and cyc 01 elav.cyc brain hemispheres with normal, stunted, or absent PDF-positive dorsal projections. The numbers (n) of brain hemispheres analyzed are indicated for each genotype. Chi-square analysis indicated a highly significant association between genotype and dorsal projection phenotype (p,10 25 ). The lower two panels represent examples of normal and stunted PDF-expressing dorsal projections, from cyc 01 elav.cyc and cyc 01 (elav-Pdf).cyc brains, respectively. doi:10.1371/journal.pgen.1002167.g009 ahead of Pdp1, which in turn was phase-advanced relative to Clk. Similar observations were made for several independent experiments. (B) Quantitative Reverse Transcriptase PCR (qRT-PCR) analysis was performed using primers that were (right panel) or were not (left panel) selective for transgenic per relative to native per 01 transcripts. Total RNA was extracted from the heads of per 01 [timP.per] ts flies as well as controls lacking the tim(UAS)-Gal4 driver transgene (per 01 [-.per] ts ) immediately before and at 15 and 30 h after a 18uC DD (restrictive) to 25uC DD (permissive) shift. Signals were quantified using the cycle threshold method [50] relative to rp49 transcript. Total per 01 /per transcript and transgenic per transcript were induced at 15 h in per 01 [timP.per] ts heads approximately five-fold and more than forty-fold, respectively, while per transcript levels in per 01 [-.per] ts flies remained constitutively low. (C) Total RNA was extracted from dissected adult brains of per 01 [timP.per] ts flies immediately prior to as well as at 15 and 30 h after a 18uC DD to 25uC DD shift. Signals in adult brains for clock genes (cwo, tim, vri) and total per 01 /per (relative to rp49 and normalized to experimental average) showed only modest responses after the shift to permissive conditions. (PDF) Chi-square analyses indicated a highly significant association between experimental condition and the percentages of rhythmic, weakly rhythmic, and arrhythmic females (p,10 26 ) and males (p,10 24 ). (B,D) Bar diagrams of the average (6SEM) relative rhythmic power observed among the rhythmic plus weakly rhythmic flies for each developmental condition. The number of flies included in this analysis (n) is indicated for each condition. Statistical analyses (Mann-Whitney rank-sum test) indicated a significant association between relative rhythmic power and experimental condition in both females (**; p,10 22 ) and males (***; p,10 23 ). (PDF) Figure S6 Over-expression of per in clock-bearing cells during metamorphosis disrupts locomotor activity rhythms in adult males. Circadian behavior of adult [timP.per] ts males under permissive conditions (17uC DD) was determined by the temperature of pupal and pharate adult development. While males raised at ambient temperature (,23uC) or transferred from 29uC to ambient temperature as wandering larvae or prepupae were strongly rhythmic as adults at the permissive temperature, males exposed to restrictive (29uC DD) conditions throughout development or during the pupal and pharate adult stages exhibited arrhythmic adult locomotor behavior. (A) Stacked bar diagram representing the percentages of males with rhythmic, weakly rhythmic, or arrhythmic adult behavior at permissive conditions (17uC DD). Prior to measurement of adult locomotor activity during 17uC LD and subsequent 17uC DD (analyzed here) flies were raised at the indicated temperatures (T dev ): 23uC, 23R29uC (transferred as wandering larvae or prepupae from 23uC to 29uC), 29R23uC (transferred as wandering larvae or prepupae 29uC to 23uC), and 29uC. The numbers (n) of flies included for each condition are indicated as well as the average (6SEM) circadian period length for rhythmic flies. Chi-square analysis indicated a highly significant (p,10 213 ) association between developmental temperature and the percentages of rhythmic, weakly rhythmic, and arrhythmic adults. (B) Bar diagram of the average (6SEM) relative rhythmic power observed among the rhythmic plus weakly rhythmic male flies for each developmental condition. The number of flies included in this analysis (n) is indicated for each condition. The Welch test statistic indicated a highly significant association (p,10 28 ) of relative rhythmic power with developmental condition. Significant differences found by post-hoc Games-Howell tests for pairwise comparisons of developmental treatments indicated by (****) and (*****) represent p values smaller than 10 24  In contrast, males, which have a relatively higher dosage of Gal80 ts (see Materials and Methods), showed a highly significant association between developmental temperature and rhythmicity at 29uC (Chi-square p,10 25 ) and exhibited long-period rhythms at 29uC following development under permissive conditions. (B,E) Both genders demonstrated a highly significant correlation between developmental treatment and rhythmicity at 17uC during the last step of the experiment (Chi-square females p,10 29 , males p,10 25 ) when flies exposed to permissive conditions during metamorphosis showed circadian period lengths approaching 24h. (C,F) Bar diagrams of the average(6SEM) relative rhythmic power observed among the rhythmic plus weakly rhythmic flies at 17uC DD for each developmental condition. For 29uC-raised males (n = 2) the range rather than the SEM is indicated in (F). The Welch test statistic indicated significant associations between relative rhythmic power at 17uC and developmental treatment (p,10 24 in both female and males). Significant differences found by post-hoc Games-Howell tests for pairwise comparisons of developmental treatments indicated by (**), (***) and (****) represent p values smaller than 10 22 , 10 23 and 10 24 , respectively. (PDF) Figure S8 DN and LNd subsets of clock neurons as well as PDFpositive dorsal projections persist after developmental per overexpression. DN and LNd clock neurons were detected by anti-PER immunofluorescence in adult brains of both ,23uC and 29uC-raised [timP.per] ts flies (experiment described in Figure 5). Representative images are shown. PER signal is weaker in the brain of the ,23uCraised fly because it was taken from the ZT4 time point, while the image for the 29uC-raised fly is from ZT22. LN v clock neurons are identified based on co-staining with anti-PDF antibody. Note the presence of PDF-stained dorsal projections from the LN v s in both brains (indicated by the white arrows). (PDF) Figure S9 Clock function is conditional in adult cyc 01 [elav.cyc] ts males raised at the permissive condition. Quantitative analysis of adult circadian behavior in 29uC-raised cyc 01 [elav.cyc] ts male flies at either permissive (29uC) versus two restrictive conditions (17uC, 18uC) (A,B) or permissive (29uC) versus restrictive (17uC) conditions following adult exposure to restrictive conditions ($3 days 17C) (C,D). The stacked bar diagrams (A,C) represent the percentages of 29uC-raised male cyc 01 [elav.cyc] ts flies with rhythmic, weakly rhythmic, or arrhythmic adult locomotor behavior. Rhythmicity was determined for individual flies by chi-square periodogram analysis of 7 d intervals at the indicated temperatures in constant darkness. The numbers (n) of flies included for each condition are indicated as well as the average (6SEM) circadian period length for rhythmic flies. Chi-square analyses indicated significant associations between experimental temperature and the percentages of rhythmic, weakly rhythmic, and arrhythmic adults (p,10 23 (A); p,10 213 (C)). The bar diagrams (B,D) correspond to the average (6SEM) relative rhythmic power observed among the rhythmic plus weakly rhythmic flies for each experimental condition. The number of flies included in this analysis (n) is indicated for each condition. Because there were only two observations for rhythmic/weakly rhythmic males at 18uC the range rather than the average6SEM is indicated. (B) Welch test analyses indicated a significant association of relative rhythmic power with experimental condition (p,10 22 ). A significant difference was found by post-hoc Games-Howell test for pairwise comparison of males at 29uC versus 18uC (**; p,0.02). (D) Statistical analysis (Mann-Whitney rank-sum) indicated a significant association between relative rhythmic power and experimental condition (***; p,10 23 ). (PDF) Figure S10 Association between developmental cyc expression and transgenic rescue of behavioral arrhythmia in adult males. (A) Stacked bar diagram representing the percentages of male flies with rhythmic, weakly rhythmic, or arrhythmic adult behavior at permissive conditions (29uC DD) following prior exposure to 7 LD days at 25uC. Flies were raised at the indicated temperatures (T dev ): 23uC, 23R29uC (transferred as wandering larvae or prepupae from 23uC to 29uC), 29R23uC (transferred as wandering larvae or prepupae 29uC to 23uC), and 29uC prior to analysis of adult locomotor activity during 25uC LD and subsequent 29uC DD conditions. The numbers (n) of flies included for each condition are indicated as well as the average (6SEM) circadian period length for rhythmic flies. Although the trends in rhythmicity relative to developmental conditions resemble the associations found in females chi-square analysis did not demonstrate a significant (p = 0.11) association in males. (B) Bar diagram of the average (6SEM) relative rhythmic power observed among the rhythmic plus weakly rhythmic flies for each developmental condition. The number of flies included in this analysis (n) is indicated for each condition. Again, trends of relative rhythmic power relative to developmental condition resemble the associations found in female flies, but the Welch test statistic did not demonstrate a significant association of relative rhythmic power with developmental condition (p = 0.15). (PDF) Figure S11 Male cyc 01 (elav-Pdf).cyc flies show behavioral arrhythmia. (A) Example actograms (left) and LD activity profiles (right) representing median locomotor behavior for male cyc 01 elav.cyc versus cyc 01 (elav-Pdf).cyc flies. Note that cyc 01 (elav-Pdf).cyc flies exhibit loss of free running rhythms in DD as well as increased activity in anticipation of lights-on and loss of activity in anticipation of lights-off in LD. (B) Stacked bar diagram representing the percentages of cyc 01 elav.cyc, cyc 01 (elav-Pdf).cyc, and cyc 01 (elav-Pdf).male flies with rhythmic, weakly rhythmic, or arrhythmic adult behavior at 25uC DD. The number of flies (n) as well as the average (6SEM) circadian period length for rhythmic flies are indicated. Chi-square analysis indicated a highly significant (p,10 28 ) association between genotype and the percentages of rhythmic, weakly rhythmic, and arrhythmic adults. (C) Bar diagram of the average (6SEM) relative rhythmic power observed among the rhythmic plus weakly rhythmic flies for each genotype. The number of flies included in this analysis (n) is indicated for each condition. Because all but one cyc 01 (elav-Pdf).flies were arrhythmic, a Mann-Whitney rank-sum test was performed to compare the effect on relative rhythmic power of the other two genotypes. As indicated, relative rhythmic power was significantly reduced in cyc 01 (elav-Pdf).cyc flies compared to cyc 01 elav.cyc flies (***; p,10 23 ). (PDF)