The Drosophila circadian clock gene cycle controls the development of clock neurons

Mutations in both Clk and cyc severely reduce pdf RNA and neuropeptide levels [23, 32]. In cyc null mutants, sLN_vs projections are often undetectable in larval and adult brains stained with PDF antibodies [23, 39, 40], although around half of the brains show ‘stunted’ projections [40]. We observed that cyc null mutants (cyc¹) showed a substantial reduction in PDF levels at ZT2, consistent with previous studies, but we also noticed the presence of thin, misrouted sLN_v projections in cyc¹ flies at higher magnification and intensity (S1A Fig). Upon close observation, PDF could often be detected in the sLN_v projections. However, these projections did not form the stereotypical bundle observed in control brains when extending from the anterior medulla toward the dorsal area of the brain.

Because highly defasciculated projections might contribute to the weaker PDF levels observed in cyc¹ mutants, we examined the structure of the sLN_v projections using a Pdf-RFP transgene, in which a cytosolic Red Fluorescent Protein (RFP) is controlled by the Pdf regulatory sequence [41]. Flies were raised at 28°C throughout development, and experiments were conducted at 28°C in 6–8-day old flies (Fig 1A). In control brains, the projections from the four sLN_vs remain fasciculated, forming a bundle until reaching the superior medial protocerebrum (SMP). In contrast, the sLN_vs of cyc¹ mutants often began to defasciculate in the ventral brain, near their cell bodies (Fig 1B). Some sLN_v projections were severely misrouted and did not reach the dorsal brain, extending instead toward the midline or other brain regions. As a result, it was possible to distinguish individual projections from each sLN_v even in the ventral brain in most cyc¹ mutant brains. This is almost never observed in control brains until the sLNv projections reach the main branching point in the SMP. The morphological phenotypes of cyc¹ flies are highly variable, and in some instances the projections are barely visible (S1B Fig).

To test whether the effect of cyc loss on the LN_vs is cell-autonomous, we next expressed a UAS-cyc dsRNA transgene (UAS-cyc^RNAi) using the Pdf-Gal4 driver. We quantified the length of the projections, starting at the point where the projections of the sLN_vs intersect with those of the lLN_vs (“point of origin”, POI, S1C Fig), until the first branching point (“branching point”, BP). This branching point is where the sLN_vs ramify and extend their stereotypical arborizations in the dorsal protocerebrum in control brains, and these arborizations show daily, clock-controlled rhythms in their fasciculation and outgrowth [27]. Downregulating cyc in the Pdf-expressing cells significantly decreased the distance to the branching point (Fig 1E). Using a modified Scholl’s analysis [42], we quantified the degree of branching in the ventral projections starting at the POI. We observed pronounced defasciculation in the sLN_v projections in Pdf > cyc^RNAi flies (Fig 1F). cyc¹ mutants also showed decreased distance to BP and sLN_v fasciculation (Fig 1C and 1D). The total projection length in Pdf > cyc^RNAi flies was not different from that of the controls (Fig 1G), and the defasciculation phenotype was not observed in the contralateral projections that extended from the lLN_vs (Fig 1H). Since the lLN_vs are born later in development during metamorphosis [25], this result suggests that cyc plays a role in neuronal development during an earlier developmental stage, when the sLN_vs begin to extend their projections toward the dorsal brain. These experiments were conducted at 28°C to allow subsequent comparisons with adult-specific and development-specific downregulations of cyc using Gal80^ts. Similar results were obtained with flies raised at 25°C (S1F–S1I Fig).

Expression of dominant negative forms of Clk and cyc is an effective strategy for preventing circadian molecular oscillations in specific groups of clock neurons [43–45]. In these dominant negative forms, the DNA binding ability is disrupted while the ability to heterodimerize is preserved [43]. Based on the phenotypes induced by cyc downregulation, we asked if expressing a dominant negative form of cyc in the sLN_vs also leads to aberrant projection morphology. We found that sLN_vs expressing Δ-cyc using Pdf-Gal4 had a significantly shorter distance until the branching point (S2B and S2C Fig) and a greater degree of sLNv projection defasciculation (S2B and S2D Fig), similar to the effects observed in Pdf > cyc^RNAi flies. The total projection length and the projections of the lLNvs were unaffected (S2E and S2F Fig).

CYC activates per transcription, and thus, PER levels in the brain are significantly reduced in cyc mutants [33]. To test whether the phenotypes of cyc^RNAi expression in the Pdf-expressing neurons are consistent with what would be expected from cyc downregulation, we compared PER levels in the parental control (Pdf-Gal4, Pdf-RFP/+) with those in Pdf > cyc^RNAi flies at the end of the night (ZT23), when PER nuclear levels are highest [46]. We found that nuclear PER levels in Pdf > cyc^RNAi flies were significantly reduced in the sLN_vs (Fig 2B and 2C) and lLN_vs (Fig 2D and 2E). In contrast, PER levels were unaffected in the Dorsal Lateral Neurons (LN_ds) (Fig 2E and 2G). These results confirmed that, at least in a light-dark cycle (LD), Pdf > cyc^RNAi flies have lower PER levels in Pdf⁺ neurons.

cyc null mutant flies have pronounced behavioral phenotypes. Their activity is unimodal instead of bimodal during LD, and they are predominantly nocturnal [47]. Additionally, cyc mutants are largely arrhythmic in DD due to the key role of cyc in circadian molecular oscillations [33]. We conducted behavioral experiments to determine the extent to which downregulating cyc specifically in PDF-expressing neurons recapitulates the phenotype of the cyc mutant. We found that at 28°C, the activity pattern of Pdf > cyc^RNAi flies was still bimodal in LD (Fig 2H and 2I). However, the majority (~90%) of the experimental flies were arrhythmic in DD (Fig 2J). Pdf > Δ-cyc flies showed similar behavioral phenotypes (S2G and S2H Fig), consistent with what was reported for their free-running behavior at 25°C [43].

To knock down cyc specifically during development, we employed a temperature-sensitive Gal80 (Gal80^ts) variant with ubiquitous expression to conditionally inhibit Gal4-mediated expression of the RNAi [48]. This method enables the temporal regulation of UAS transgenes, as Gal80^ts remains active at lower temperatures but becomes inactive at higher temperatures. We raised flies at 28°C to allow cyc downregulation during development then transferred them to 18°C immediately after eclosion (Fig 3A). After 1 week at 18°C, the brains were dissected at ZT2 and stained with PDF and RFP antibodies (see methods section). As shown in Fig 3, downregulating cyc exclusively before eclosion resulted in abnormal morphology of the sLN_v axonal projections in adult flies (Fig 3B). The phenotypes resembled those observed with constitutive downregulation, with a significantly shorter distance to the branching point (Fig 3B and 3C) and a greater degree of defasciculation compared to parental controls (Fig 3B–3D). No significant differences were found in the total projection length or the degree of defasciculation of the lLN_vs (Fig 3E and 3F).

A previous study showed that panneuronal rescue of cyc expression in a cyc¹ mutant exclusively during development was sufficient to partially rescue arrhythmicity in adult flies [40]. Therefore, we asked if downregulating cyc in the Pdf+ cells specifically during development would lead to behavioral phenotypes similar to those seen in the cyc null mutants. We found that under free-running conditions at 18°C, most (~78%) of the Pdf > cyc^RNAi flies were arrhythmic (Fig 3G–3I). An analysis of PER subcellular localization in DD2 revealed clear cycling with nuclear PER higher at CT2 (Fig 3J). These results indicate that developmental downregulation of cyc specifically in the Pdf+ cells is sufficient to prevent behavioral rhythms in adults.

To determine whether adult-specific cyc knockdown in the Pdf-expressing cells would also lead to morphological phenotypes, we raised flies at 18°C and switched them to 28°C immediately after eclosion (S3A Fig, see the methods section). This manipulation did not result in morphological phenotypes either in terms of the length to the branching point or in the degree of sLN_v defasciculation (S3B–S3F Fig). Under free-running 28°C, the majority of the experimental flies were arrhythmic (S3G–S3H Fig), indicating that, as expected, cyc is required in adult clock neurons for proper circadian clock function.

Next, we asked if cyc downregulation results in clock neuron morphology phenotypes during earlier developmental stages. The four larval sLNvs, which modulate the sensitivity of larvae to light and mediate a circadian rhythm in visual sensitivity [49], appear to be identical in their anatomy and synaptic connections [50]. We expressed the cyc^RNAi transgene under the Pdf-Gal4 driver and dissected third larval instar (L3) brains (Fig 4). In brains of experimental larvae the length to the branching point did not differ from that of the controls (Fig 4B and 4C), but the degree of dorsal termini branching was significantly higher (Fig 4B–4D). This quantification is similar to that previously described when quantifying the arborization of the dorsal projections sLN_vs in adults [27], where the concentric circles are centered at the main dorsal branching point (S1B Fig; see methods section). The total sLN_v projection length was not affected by the genetic manipulation (Fig 4E).

Since the effects of cyc knockdown via RNAi and the expression of a cyc dominant negative form in adults were similar (Figs 1 and S2), we analyzed the morphology of the sLN_v projections in L3 larvae upon Δ-cyc expression. In Pdf > Δ-cyc larval brains, the length to the branching point was significantly lower (Fig 4F and 4G) and the number of branches was significantly greater than that of controls (Fig 4H). The total projection length was not affected (Fig 4I). Taken together, these results suggest that cyc plays a role in the development of the larval sLN_v neurons.

Clk and cyc mutations produce similar effects on the expression pattern of PDF in adult brains [23]. CLK and CYC act as heterodimeric transcriptional activators, and the circadian phenotypes associated with mutations in these core circadian clock genes, both molecular and behavioral, are largely similar [43, 47, 51]. To determine if downregulating Clk in the sLN_vs leads to the same defasciculation of the sLN_vs observed with cyc manipulations, we performed similar experiments as those described above, in which we expressed Clk^RNAi in Pdf+ neurons. We found that Pdf > Clk^RNAi flies also showed neuronal morphology phenotypes (Fig 5).

In a previous study, Clk downregulation resulted in overfasciculation of the sLN_v dorsal termini when stained with anti-PDF [52]. However, RFP labeling of the sLN_v membrane indicated that these termini were actually more expanded than those of control flies, resulting in significantly higher dorsal termini branching (Figs 5B–5E and S1C). In Pdf > Clk^RNAi flies, neither the distance to the branching point (Fig 5C) nor the degree of defasciculation differed from controls (Fig 5D). Neither the sLN_v total projection length nor the lLN_v projections were affected (S4C and S4D Fig). Only ~48% of the Pdf > Clk^RNAi flies were rhythmic, and those that were rhythmic exhibited a lengthening of the free-running period (S4B Fig and Table 1). Nuclear PER levels in Pdf > Clk^RNAi flies were significantly reduced in the LN_vs (S4H–S4J Fig).

Expression of Δ-Clk in the Pdf+ cells did not result in changes in the sLN_v projection length until branching point (Fig 5F and 5G) or the total length of the projections (S4F Fig). However, the Pdf > Δ-Clk brains had increased defasciculation of the ventral projection (Fig 5H). The degree of dorsal termini branching in the Pdf > Δ-Clk flies was not significant (Fig 5I), not was the degreed of lLN_v defasciculation (S4G Fig). Under DD at 28°C, the majority of Δ-Clk expressing flies were arrhythmic (S4E Fig), consistent with what was reported at 25°C [43].

We then examined L3 larval brains to determine if the observed phenotypes were already present at this developmental stage. While expression of Clk^RNAi did not result in morphological phenotypes in larval LN_vs (S5 Fig), expression of Δ-Clk resulted in pronounced phenotypes (Fig 6A and 6B). We observed a significant increase in sLN_v dorsal termini branching (Fig 6D) and total projection length (Fig 6E) in Pdf > Δ-Clk larvae. The length to the branching point for the experimental larvae was not significantly different from that of the control lines (Fig 6C). Expressing Δ-Clk led to more pronounced phenotypes in the larval stage than Clk downregulation, possibly due to incomplete knockdown.

In addition to the main feedback loop, CLK and CYC form a secondary loop by activating vri and Pdp1ε [8, 9], which repress and activate Clk expression, respectively. The low PDF peptide in the sLN_vs projections of cyc¹ mutants can be rescued by vri overexpression [39]. To determine if vri expression also affects sLN_vs morphology we expressed a line with a CRISPR/Cas9-based gRNA targeting the vri gene [53] under the control of the Pdf-Gal4 driver. We found that in Pdf > Cas9 + vri-g flies neither the distance until branching point in the s-LN_vs (Fig 7B and 7C) nor the degree of fasciculation of the s-LN_vs (Fig 7D) was different from controls. The total projection length was significantly higher than controls (Fig 7E), but in this case due to projections extending ventrally towards the optic tract (Fig 7B). In the majority of the brains, some s-LN_v projections extended towards the ventral brain after reaching the SMP (Fig 7B) and in most cases contacted the l-LNv contralateral projections in the optic tract (Fig 7F and 7G), a phenotype that was never observed in control brains.

Flies were raised on cornmeal-sucrose yeast media in a Percival Incubator under 12:12 LD at different temperature conditions. Depending on the experiment, flies were raised under either 18°C, 25°C, or 28°C (indicated in the figure legends). The lines UAS-cyc^RNAi (BDSC #42563), UAS-Clk^RNAi (BDSC #42566), w¹¹¹⁸ (BDSC #3605) and CS (BDSC #64349) were obtained from the Bloomington Drosophila Stock Center. The lines Pdf-RFP,Pdf-Gal4;Tub-gal80^ts and w;Pdf-RFP;MKRS/TM6 were donated by Justin Blau (New York University). The cyc¹, UAS-Δ-cyc, and UAS-Δ-Clk stocks were donated by Paul Hardin (University of Texas).

All images were acquired on an Olympus Fluoview 1000 laser-scanning confocal microscope using a 40x/1.10 NA FUMFL N objective (Olympus, Center Valley, PA) at the Advanced Science Research Center (ASRC-CUNY). For all the experiments, only one hemisphere per brain was imaged (the right hemisphere, unless it was damaged, in which case we imaged the left hemisphere).

Quantification of adult LN_vmorphology (sLN_vand lLN_v): We quantified 1) sLN_v total projection length, 2) sLN_v length from the point of origin (POI) until the branching point (BP) (‘length until BP’), 3) the degree of defasciculation of the sLN_v ventral projections, 4) the sLN_v dorsal termini branching, 5) the degree of defasciculation of the l-LNv projections, and 6) intersections between sLN_v projections and lLN_v projections along the optic tract.

1-Total projection length: The total length of the dorsal projection was determined by a line drawn from the point of intersection (POI) between the sLN_vs and the optic tract until the end of the dorsal termini. If the projection length went past the midline of the brain, the length was measured up to the midline.

2-Length until BP: The partial length of the dorsal projections was determined by a line drawn from the POI until the BP of the sLN_vs at the dorsal termini. The projection length and partial projection length of the sLN_vs were quantified using Fiji in ImageJ.

3- Defasciculation of the sLN_vventral projections (‘sLN_vDefasciculation’): A modified Scholl’s analysis [42], was used to analyze the degree of defasciculation of the ventral area of the sLN_v projections, near the cell bodies. Six concentric circles, each 25 μm apart, were placed centered in the POI (S1C Fig). Each intersection between an individual ventral projection and any of the 6 circles was counted. A value of ‘10’ denotes 10 total intersections between any of the projections and any of the circles.

4-sLNv Dorsal termini branching: A modified Scholl’s analysis was used to analyze the degree of defasciculation of the dorsal termini of the sLN_v projections. This method is similar to what was previously described to quantify sLNv dorsal termini [27]. In this study, 8 concentric circles, each 12.5 um apart, were centered in the BP (S1E Fig). Each intersection between an individual dorsal projection and any of the 8 circles was counted. A value of ‘10’ denotes 10 total intersections between any of the dorsal projections and any of the circles.

5-Defasciculation of the lLN_voptic tract projections. The degree of defasciculation of the lLNvs was determined using the same 6 concentric circles centered in the POI what were used to quantify defasciculation of the sLN_v ventral projections (2) (S1C Fig). Each intersection between an individual lLNv projection and any of the 6 circles was counted.

6- Intersections between sLN_vprojections and lLNv projections along the optic tract. We quantified the percentage of brains in which at least one sLNv dorsal projection turned ventrally and extended towards the optic tract, contacting at least one lLNv projection. This phenotype was not observed in brains of control flies but was present in more than half of the brains of flies in which vri was knocked out (shown in Fig 7F).

DAM2 Drosophila Activity Monitors (TriKinetics, Waltham, MA) were used to record the locomotor activity rhythms of adult male flies aged three- to five-days, as previously described [65]. Flies were entrained to 12:12 LD cycles for at least five days, and then transferred to constant darkness (DD) for at least eight days at a constant temperature of 28°C, unless otherwise specified. Free-running activity rhythms were analyzed with ClockLab software from Actimetrics (Wilmette, IL). We employed ClockLab’s χ-square periodogram function, which was integrated into ClockLab software, for the analysis of rhythmicity, rhythmic power, and free-running period in individual flies, using a confidence level of 0.01 [33]. For each of the tested genotypes, only significant periodicities falling within the 14 to 34-hour range were taken into consideration. In instances where an individual fly exhibited multiple periodicities with peaks surpassing the significance threshold, only the period with the highest amplitude was utilized when calculating the average periods presented in Table 1. ClockLab assigns each peak in the χ-square periodogram both a "Power" value and a "Significance" value. The "Rhythmic Power" for each designated rhythmic fly was determined by subtracting the "Significance" value from the "Power" value associated with the predominant peak. Flies that did not exhibit a periodicity peak above the threshold (10) were categorized as "arrhythmic," and their period and rhythmic power were not included in the analysis [65].

Pearson’s D’Agostino normality tests were performed for all the datasets. Depending on whether the data were normally distributed, statistical analyses were performed using either a one-way ANOVA with a Tukey’s multiple comparisons test or a Kruskal-Wallis test with a Dunn’s multiple comparisons test for 3 or more groups, or a t-test for comparisons between 2 groups. Fisher’s exact contingency tests were run to analyze the percent rhythmicity for the indicated genotypes under DD.