Spliceosome proteins regulate timeless mRNA to control clock protein levels and daily rhythms in fruit flies

In a screen for kinases that affect circadian period length of Drosophila rest:activity rhythms in constant dark:dark (DD) conditions, we identified Pre-mRNA Processing factor 4 (PRP4), a splicing factor that also has kinase activity (Kojima et al., 2001; Schneider et al., 2010). RNA interference (RNAi)-mediated knockdown of prp4 in tim+ clock neurons with the Tim-UAS-Gal4 (TUG) Gal4 driver resulted in consistently long free-running circadian periods (Figure 1A–B). Two independent RNAi lines (GD and KK) were used to confirm these findings.

The length of the free-running circadian period is an important parameter that also affects the daily distribution of fly activity in light:dark (LD) cycles. Thus, we determined the effect of prp4 knockdown on fly diurnal activity in LD conditions (Figure 1C). Control (TUG; Dcr2/+) flies displayed characteristic bimodal activity in LD, with activity peaks that precede the onset of light (ZT0) and the onset of dark (ZT12). Compared to controls, the flies with circadian-cell-specific prp4 knockdown (TUG; Dcr2 >prp4^RNAi(GD)) exhibited a delayed evening activity peak as well as a slightly delayed morning peak. These findings are consistent with the longer endogenous period that we report for prp4 knockdown flies (Figure 1A–B).

As the small lateral ventral neurons (s-LNvs) are the most relevant clock cells for driving rest:activity rhythms under constant dark conditions, we asked if effects of PRP4 were mediated in these cells (Helfrich-Förster, 1998). To test this hypothesis, we downregulated prp4 in both s-LNvs and large lateral ventral neurons (l-LNvs) with a Gal4 transgene driven by a peptide expressed specifically in these cells, Pigment Dispersing Factor (PDF). Knockdown of prp4 with pdf-Gal4 driving the weaker RNAi line (GD) led to a modest yet significant lengthening of free-running circadian period (Figure 1F). Interestingly, the knockdown of prp4 with a stronger RNAi (KK) led to complex rhythms (Figure 1E). These complex rhythms were generally characterized by changes in the behavioral pattern, often manifest as phase shifts during day 4 or 5 of constant darkness (see ~6 hr shift in the record shown), in the midst of an otherwise rhythmic record (Figure 1D). Such a complex rhythm phenotype could reflect uncoupling of period into two components of pdf+ and pdf- cell oscillators (Yao et al., 2016). Overall, these findings suggest that PRP4 action in LNvs contributes to maintaining clock function.

Since our data indicated that PRP4 is necessary in LNvs for proper circadian rhythmicity (Figure 1), we next examined the circadian cycling of PER and TIM in s-LNvs at regular intervals around the clock (Figure 2A–C). Cycling of both PER and TIM was affected in s-LNvs upon prp4 knockdown with the clock-specific TUG driver (Figure 2A–C). Total TIM levels, as quantified by corrected total cell fluorescence (CTCF) image analysis, were decreased during the night (Figure 2B–C). PER levels were also lower in the late night (ZT20), and especially so in the early morning (ZT2), in prp4 knockdown flies (Figure 2A,C). Additionally, nuclear accumulation of PER was delayed beyond ~ZT20, the time point at which PER was already partially localized to the nucleus in control flies (Figure 2A). To further characterize the effect on nuclear expression, we profiled relative nuclear PER expression upon prp4 knockdown with each of the two RNAi lines (GD and KK) in s-LNvs around the time of nuclear accumulation of PER (ZT18-ZT22). We found that the nuclear accumulation of PER was slower in flies with decreased prp4 (both GD and KK RNAi lines) than in controls (TUG; Dcr2/+) at these time points (Figure 2—figure supplement 1). Overall, our immunohistochemistry data point to a distinct molecular clock phenotype upon prp4 downregulation. Because TIM is necessary for nuclear accumulation of PER (Vosshall et al., 1994; Saez and Young, 1996; Jang et al., 2015), reduced TIM accumulation could account for the delay and overall reduced nuclear expression of PER in s-LNvs (Figure 2, Figure 2—figure supplement 1), which together could account for the longer free-running period.

As biochemical analysis of PER and TIM oscillations requires large amounts of tissue, we conducted pan-neuronal knockdown of prp4 with the elavGal4 driver. The pan-neuronal manipulation targeted prp4 more broadly, allowing us to verify the efficiency of the knockdown through RNA analysis of whole heads. We consistently observed ~50% reduction in prp4 RNA in head lysates of flies where it was knocked down pan-neuronally (Figure 2—figure supplement 2). This broad manipulation also allowed us to test the effects of prp4 reduction on PER and TIM levels using western blotting of whole head lysates. The results obtained with this relatively crude method agreed well with our immunohistochemistry profiling of s-LNvs. First, we observed a strong effect of prp4 downregulation on total TIM levels, particularly during the initial TIM accumulation phase (ZT10-18) (Figure 2D–E). Quantification of our western blot data across multiple experiments suggested that prp4 knockdown also blunted the circadian cycling of TIM protein (Figure 2E). While the total levels of PER were not changed upon prp4 downregulation, its phosphorylation profile was dampened. In control (elavGal4; Dcr2/+) flies, as previously reported (Edery et al., 1994), PER was increasingly phosphorylated in the late night and early morning up until its degradation. In flies with reduced prp4 levels, PER migrated at a lower molecular weight on western blots (ZT18-ZT2). We speculate that the dampened phosphorylation profile of PER reflects, at least in part, the defect in nuclear accumulation of PER because a number of sites on PER get phosphorylated only when PER starts to accumulate in the nucleus (after ZT20) (Chiu et al., 2011). Alternatively, prp4 depletion could have a TIM-independent effect on PER to regulate its phosphorylation profile.

As noted above, PRP4 is associated with kinase activity, but it is also a component of the spliceosome (Kojima et al., 2001; Schneider et al., 2010). To address if the circadian role of PRP4 was spliceosome-dependent, we assayed circadian behavior following knockdown of additional tri-snRNP components. Strikingly, clock-specific downregulation of pre-mRNA processing factor 3 (prp3) and pre-mRNA processing factor 31 (prp31), which are associated primarily with U4/U6 snRNP, as well as pre-mRNA processing factor 8 (prp8) and bad response to refrigeration 2 (brr2), which are associated with U5 snRNP, caused pronounced defects in free-running circadian behavior (Table 1). The phenotypes ranged from period lengthening to complete arrhythmicity, suggesting that the overall levels and/or stability of tri-snRNP regulate circadian rhythms.

To determine if the molecular signatures of prp4 knockdown and other tri-snRNP downregulation phenotypes were similar, we performed s-LNv-specific analysis of PER and TIM levels with a few randomly selected RNAi lines. We found that the s-LNvs of flies with downregulated prp8 had lower TIM levels at night, and also showed decreased nuclear accumulation at ZT20 relative to control flies (Figure 3A). These data allow us to conclude that circadian oscillations of PER (data not shown) and TIM are sensitive to changes in the levels of multiple tri-snRNP components.

We further examined tri-snRNP function in circadian regulation by utilizing some of the previously characterized mutants of prp8 and brr2 (Coelho et al., 2005; Bivik et al., 2015). Because tri-snRNP components are essential genes with no homozygous viable knockout mutants described to date, we tested circadian behavior in the viable transheterozygous prp8/brr2 mutants. The heterozygous prp8/+ and brr2/+ mutants displayed normal circadian rhythms, readily explained by the haplosufficiency of these two genes (Figure 3B). Transheterozygous prp8/brr2 mutant flies had normal circadian period (data not shown), but displayed decreased rhythm strength compared to the prp8/+ and brr2/+ heterozygous flies, confirming a circadian function of the tri-snRNP (Figure 3B).

Next, we attempted to identify the mechanism by which PRP4 regulates circadian rhythms, starting with the broad hypothesis that downregulation of prp4 leads to the aberrant splicing of one or more core clock transcripts. Analysis of clock transcripts indicated an increase in tim levels in the late night in flies with reduced prp4 (Figure 4—figure supplement 1), but in order to apply an unbiased approach, we performed RNA-Sequencing (RNA-Seq) analysis of fly heads in which prp4 had been knocked down. As clock protein expression in the eye contributes a majority of the signal in head assays, we used an eye-specific Glass Multiple Promoter (GMR-Gal4) driver for prp4 knockdown. Overall gene expression was dramatically influenced by prp4 downregulation (433 down, 310 up at FDR < 0.05) (Supplementary file 2). Pathway enrichment analysis using DAVID identified changes in folate biosynthesis as well as in other broad pathway categories such as protein export, protein processing in endoplasmic reticulum and drug metabolism (Supplementary file 1). Interestingly, components of folate metabolism have been previously implicated in circadian clock regulation in human cells (Zhang et al., 2009). Despite the fact that PRP4 is a component of the core spliceosome required for constitutive splicing, we did not detect dramatic effects on global splicing. Using the Comprehensive AS Hunting (CASH) method, which assays for splicing events, our analysis identified 45 genes exhibiting differential splicing upon prp4 downregulation, with FDR ≤ 0.05 (Wu, W., et al., 2017) (Supplementary file 3).

An intron retention event in tim that was significantly upregulated upon prp4 knockdown was of particular interest due to its potential clock function (Figure 4A–B). Our initial splicing analysis was performed with CASH, but we additionally ran the Cufflinks-2.2 pipeline to obtain psi (percent spliced in) information for differentially spliced isoforms (Supplementary file 4). Importantly, the same retention event in tim, at position chr2L:3499764–3500000 (hereafter we refer to this intron as tim-tiny for simplicity), was consistently identified as significantly upregulated with both analyses (Figure 4A). For comparison, differential alternative splicing in only four other genes (trol, pex7, Npc1a, vsg) was consistently identified with both of the algorithms (Figure 4A). Moreover, the RNA-Seq reads across the tiny-tiny junction normalized to its neighboring junction (2L:3500362–3500422), which is spliced out in all of the tim isoforms, were significantly increased in the prp4 knockdown samples (GMR >prp4^RNAi) compared to the controls (GMR/+), further pointing to the retention of tim-tiny. To estimate how common tim-tiny retention was in control flies (GMR/+), we quantified the ratio of isoforms that contain tim-tiny (tim-RM and tim-RS) to those that do not contain this intron using our Cufflinks-2.2 output (Supplementary file 4). This analysis indicated that the isoforms retaining tim-tiny were twice as abundant as other isoforms at ZT 8 (when all of the RNA-Seq samples were collected) (Supplementary file 5).

Motivated by the RNA-Seq data, we sought to verify the retention of the tim-tiny intron upon pan-neuronal prp4 knockdown. For this purpose, we designed primers that would amplify only the transcript containing the retained intron or only the spliced transcript respectively (Figure 4—figure supplement 2). The ratio of retained to spliced signal would then indicate the relative intron retention. Using this approach, increased tim-tiny retention was consistently detected upon pan-neuronal prp4 knockdown (Figure 4D). Additionally, by performing a control set of experiments, the signal from the ‘retained’ primer set targeting tim-tiny was verified not to reflect genomic DNA contamination (Figure 4—figure supplement 2A–B). First, retained intron levels were normalized to total tim mRNA. As total tim mRNA was amplified with primers that do not span junctions, but rather bind within sequences of a different exon (‘exon’), they are also expected to detect any contaminating DNA in addition to mRNA. As with the tim-tiny retained/spliced ratio, the retained/exon ratio indicated increased tim-tiny retention upon prp4 knockdown, in particular at ZT12, a time point used for our initial RNA-Seq analysis (Figure 4—figure supplement 2B). Additionally, we amplified RNA using primers that span an exon-exon junction at a different part of tim (‘mRNA’), which hence should only detect mRNA, and found that the ratio of tim ‘exon’ to ‘mRNA’ was not different between the control (elavGal4; Dcr2/+) and the prp4 knockdown flies (elavGal4; Dcr2 > prp4^RNAi(GD)), indicating that residual DNA contamination does not contribute to the increased detection of tim-tiny with prp4 knockdown (Figure 4—figure supplement 2B).

To assay the influence of prp4-dependent tim-tiny intron retention on TIM levels, we generated tim cDNA constructs that lacked the intron (‘tim-spliced‘), included the intron (‘tim-retained’) or included the intron along with a silent T > A mutation in the 5’ donor splice site at 2L:3499999 (‘tim-retained-ssM’) (Figure 5A). All constructs were transfected into S2 cells and assayed for their ability to drive expression of TIM protein. Intron inclusion had a drastic effect on the total levels of TIM, ranging from highest in the condition when no intron was present to no detectable full size TIM in the condition when splicing was blocked with the mutation (Figure 5B). Expression of the construct with the 5’ splice site mutation led to low level production of a shorter TIM isoform that we called TIM^tiny. This isoform was entirely absent when tim cDNA lacking the tim-tiny intron was expressed but was produced upon expression of the cDNA construct that included tim-tiny. Therefore, TIM^tiny is a truncated TIM isoform generated from the tim mRNA carrying an unspliced tim-tiny intron. Total mRNA levels were not different between these three experimental conditions, pointing to the post-transcriptional regulation of TIM protein abundance (data not shown).

Next, we overexpressed tim cDNA transgenes (Figure 5A) in flies, using a circadian cell driver (TUG). Normally, overexpression of tim using the Gal4-UAS system reduces rhythmicity and increases free-running periods, likely due to prolonged expression of the excess protein coupled with little negative feedback to downregulate production (Yang and Sehgal, 2001). Overexpression of ‘tim-spliced’ and ‘tim-retained’ cDNAs caused lengthening of free-running periods (Figure 5D). However, the circadian behavior of flies overexpressing the cDNA construct with the 5’ splice site mutation in tim-tiny (TUG; Dcr2 > tim-retained + ssM) was not different from that of the controls (TUG; Dcr2/+). This would fit with our observations from S2 cells, which suggested a loss of full-sized TIM (TIM^full) expression when tim-tiny was primed for selective retention. Thus, we hypothesized that overexpression of the construct with the 5’ splice site mutation did not produce TIM^full in flies. We assayed TIM levels in flies overexpressing different tim cDNA constructs through western blots of fly head lysates collected at ZT10 when the endogenous TIM levels are low in controls (TUG; Dcr2/+) (Figure 5C). As predicted, flies that overexpressed ‘tim-retained-ssM’ cDNA did not have increased TIM^full relative to controls, although both ‘tim-spliced’ and ‘tim-retained’ transgenes expressed abundant amounts of protein. Interestingly, we detected a band, likely corresponding to TIM^tiny, in TUG; Dcr2 > tim-retained + ssM flies, which was absent in the TUG; Dcr2/+ controls. These data further corroborate our S2 cell findings and suggest that selective tim-tiny retention acts to reduce overall TIM levels.

To further understand the effect of tim-tiny splicing on circadian behavior, we overexpressed ‘tim-spliced’, ‘tim-retained’ and ‘tim-retained-ssM’ using the TUG driver in the tim⁰ homozygous background. We hypothesized that differential splicing of tim-tiny is necessary for the maintenance of circadian rhythmicity and so the ‘tim-retained’ construct would rescue the behavioral rhythm most efficiently because it allows for both the splicing and retention of the intron. Additionally, because prp4 knockdown increases the retention of tim-tiny (Figure 4) and prolongs the rhythm (Figure 1), we speculated that the ‘tim-spliced’ construct would rescue with a shorter period than the ‘tim-retained’ construct. As expected, the expression of tim cDNA with a 5’ splice site mutation in tim-tiny (‘tim-retained + ssM’) did not restore rhythms in tim⁰ flies (Figure 5E). The other two cDNA constructs, ‘tim-retained’ and ‘tim-spliced’ rescued rhythms in 54% and 32% of tim⁰ flies, respectively. Importantly, there was a significant difference in period length between the flies that were rhythmic (Figure 5F). As discussed above, the UAS-GAL4 system typically over-expresses proteins and so rescues per/tim mutants with longer periods (Yang and Sehgal, 2001), which we observed for the ‘tim-retained’ isoform (~26 hr). Shorter periods were seen with the ‘tim-spliced’ version (Figure 5F), supporting the idea that tim-tiny retention promotes clock delays.

We next asked if the splicing of tim is under circadian regulation. For this purpose, we generated relative intron retention profiles across regular intervals under both LD and DD conditions (Figure 6A–B). In LD conditions, intron retention of tim-tiny did not display a significant cycle (by JTK analysis) but nevertheless showed a peak at ZT8. In DD conditions, interestingly, we observed robust cycling of tim-tiny with a crest at CT8. It was previously reported that the last (~850 bp) intron in tim (also known as tim-cold) is also sometimes retained (Boothroyd et al., 2007). To determine if the retention of tim-cold is rhythmic, as is the retention of tim-tiny, we profiled tim-cold intron retention across a 24 hr cycle under different conditions. We found that tim-cold retention cycled in both LD and DD with similar phases, such that peak intron retention was during the day in LD and the subjective day in DD. To further verify that the cycling we detected for tim splicing was under clock control, we assayed tim-tiny and tim-cold intron retention profiles in per¹ mutant flies. In per¹ flies under both LD and DD conditions, the cycling of both tim-tiny and tim-cold intron retention was abolished, indicating that the circadian clock drives circadian oscillations in tim splicing. We suggest that retention of the tim-tiny intron during the day serves to delay the accumulation of TIM protein, in particular when light is not available to degrade TIM.

It was previously reported that tim-cold retention is regulated by temperature and peaks at colder temperatures (Boothroyd et al., 2007). To determine if the splicing of tim-tiny was similarly sensitive to temperature, we used a temperature entrainment paradigm (12 hr:12 hr 30°C:25°C) under constant photic conditions (Figure 6C–D). In constant dark, temperature cycles were able to drive tim expression as previously reported (Glaser, F.T., and Stanewsky, R., 2005). Additionally, the profile of tim-cold retention was cyclic with the highest retention levels during the colder (25°C) temperatures. Interestingly, tim-tiny retention was robustly rhythmic under these temperature conditions, yet, unlike tim-cold, intron retention was increased by higher temperatures and decreased by lower ones. Thus, higher temperatures, which are typically associated with daytime hours, may reduce TIM levels by regulating the splicing of tim-tiny, while light does so through TIM degradation (Hunter-Ensor et al., 1996; Myers et al., 1996; Zeng et al., 1996; Naidoo et al., 1999). We observed the same relationship and even more robust cycling of tim-tiny under temperature cycles (12 hr:12 hr 30°C:25°C) in constant light conditions (Figure 6D). In these conditions, tim-cold did not cycle, further indicating different regulation of these two splicing events in tim.

Fly stocks and crosses were maintained at room temperature or at 18°C on standard cornmeal molasses medium. Stocks for RNAi overexpression were obtained from VRDC. brr2^e03171, prp8²⁰, prp8²⁰⁰ mutants were obtained from the Bloomington stock center. iso³¹, per¹ and Gal4 stocks were from the Sehgal lab stock collection. Transgenic lines for tim cDNA overexpression were generated by the site-specific PhiC31 Integration System (Rainbow Transgenics) using the attP on the 3^rd chromosome. The DNA for fly embryo injections contained tim cDNA constructs (Figure 5A) subcloned into pUASTattB vectors.

For free-running circadian analysis, male flies were entrained to 12 hr:12 hr light:dark (LD) cycles at 25°C for at least three complete cycles. Flies were loaded into TriKinetics Drosophila Activity Monitor (DAM) system (Trikinetics, Waltham, MA), released into constant darkness and recorded for at least 7 days. Circadian parameters (period and rhythm strength) were determined using Clocklab Software (Actimetrics, Wilmette, IL). Period length was determined with χ2 periodogram analysis. Rhythm strength was determined using Fast Fourier Transform (FFT) values. A fly was considered rhythmic if the FFT value was greater than 0.01.

For analysis of circadian behavior in LD, flies were stably entrained in LD at 25°C for 3 days and their behavior was recorded as described above for the next three subsequent days. The analysis of activity counts was performed using Insomniac3 Software (RP Metrix).

Fly brains were dissected in 4% paraformaldehyde (PFA) in 1x phosphate-buffered saline (PBS), fixed in PFA for 20 min at room temperature (RT) and then additionally trimmed of the air sacs and other contaminating tissues in 1xPBST. Dissected brains from each time point were stored in 1xPBST at 4°C until all of the time points were collected (never for more than 8 hr). Once all dissections and fixations were completed, brains were washed in 1xPBST for 20 min at RT, blocked with 5% Donkey Serum (DS) in 1xPBST for 20 min at RT, and incubated with the primary antibodies diluted in 5% DS overnight with gentle shaking at 4°C. The primary antibodies included rat anti-TIM (UPR42, 1:1000), guinea pig anti-PER (UP1140, 1:1000), rabbit anti-PDF (HH74, 1:500) and mouse anti-LaminC (LC28.26 from Developmental Studies Hybridoma Bank, 1:500). After a 30 min wash in 1xPBST at RT, brains were incubated with secondary antibodies diluted in 5% DS at RT. Secondary antibodies were used at 1:500 dilution and included FITC donkey-anti-guinea pig (Rockland), Alexa555 donkey-anti-rabbit (Jackson Immuno), Alexa647 donkey-anti-mouse (Jackson Immuno). Samples were washed for 30 min in 1xPBST at RT, and mounted in VectaShield. Slides were imaged with a Leica SP5 confocal microscope using a 40x oil-immersion objective and a 0.5 μM step size. The signal was adjusted to be not saturated as determined by QLUT parameters. After this original adjustment, all the settings were kept constant for a given experimental set. ImageJ software was used for analysis.

Western blot assays with fly heads were performed as previously described (Garbe et al., 2013). Briefly, 7–10 fly heads per genotype/condition were lysed in 1x Passive Lysis Buffer (Promega), supplemented with protease and phosphatase inhibitors. For S2 cell extracts, 48 hr after transfection, cells were collected and lysed in the same buffer as described for fly heads. The following primary antibodies were used: anti-PER (UP1140, 1:1000), anti-TIM (UPR42, 1:1000) and anti-HSP70 (Sigma, 1:5000).

For cell culture expression, tim cDNA was subcloned into pIZ-V5 plasmids. Using standard restriction enzyme cloning technique, the tim-tiny intron was subcloned into pIZ-V5 vectors carrying intronless tim cDNA from the pBluescript-tim plasmid. Mutagenesis of the 5’ splice donor of tim-tiny intron site was performed with primers catalogued in the Key resources table using Q5 Site-Directed Mutagenesis Kit (NEB).

S2 cells were cultured in a standard Schneider medium (Invitrogen) supplemented with 10% FBS (Sigma). Transfection was done using an Effectene kit (Qiagen) according to the manufacturer’s protocol.

Flies were collected on dry ice at indicated time points and stored at −80°C until all of the time points were collected (within 24 hr). Fly heads were then collected on dry ice and homogenized with TRIzol (ThermoFischer) on ice using standard protocols. Following the phase separation, the aqueous phase was transferred into a new tube, mixed with an equal volume of 70% ethanol and loaded directly onto the RNeasy mini kit columns (Qiagen). The rest of the RNA isolation was done according to the manufacturer’s protocol. On-column DNase digestion (Qiagen) for 15 min at RT was always included. cDNA was generated with Superscript II (Invitrogen), according to the manufacturer’s protocol. Quantitative RT-PCR reaction was performed in a ViiA7 Real-Time PCR system (Applied Biosystems) using SYBR Green Master Mix (Applied Biosystems) with gene specific primers. Relative gene expression was calculated using the ΔΔCt method with rp49 as normalization control.

RNA extraction was performed as described in the ‘Quantitative RT-PCR’ section above. Tapestation (Agilent) was used to ensure that all of RNA samples were of high-quality (RIN >8). five samples per each genotype were selected and prepared with Lexogen’s SENSE mRNA-Seq library Prep Kit. Illumina Next-Generation Sequencing (NextSeq 500) with 300pb paired-end high output run (at ~50M reads per sample) was performed by the Genomics Facility at the Wistar Institute, Philadelphia, PA.

The RNA-seq reads were aligned to the Drosophila genome (dm6.BDGP6.v88) using STAR version 2.5.3a (Dobin et al., 2013). Normalization and quantification were performed with the PORT version 0.8.2a-beta pipeline (Grant, 2018) which first removes reads that map to ribosomal RNA sequences or mitochondrial DNA and then uses a read re-sampling strategy for normalization to minimize unwanted variance such as differences in sequencing depth among the samples. PORT normalization was performed before quantification, at the aligned read level. After normalization, the quantification of features (genes, exons, introns, junctions) was done with respect to the Ensemblv88 annotation. The differential expression analysis was performed using the R Bioconductor package limma-voom (Ritchie et al., 2015). General pathway enrichment analyses were performed using DAVID (https://david.ncifcrf.gov/↗). The top 743 differentially expressed genes upon prp4 downregulation, corresponding to FDR ≤ 0.05, were used for pathway enrichment analysis. Differential splicing analysis for prp4 knockdown samples (with respect to the control) was performed using CASH (Wu et al., 2018). Full transcript quantification was done using Cufflinks-2.2, and differential psi (percent spliced in) analysis was performed for various gene isoforms to identify alternate splicing events (Trapnell et al., 2013; Trapnell et al., 2010).

The statistical parameters are included in the legends of each figure. JTK_CYCLEv3.1 was run in R for circadian statistical analyses. GraphPad Prism was used for all other statistical tests.

The RNA-Seq data generated in this work are freely available at the Gene Expression Omnibus (GEO) standard repository (accession # GSE115163↗).

Sequencing data have been deposited in GEO under accession code GSE115163↗.

The following dataset was generated:

Amita Sehgal, Iryna Shakhmantsir, Soumyashant Nayak, Gregory R. Grant. 2018. RNAseq of prp4 knockdown in Drosophila. NCBI Gene Expression Omnibus. GSE115163